Category: 科学易学

  • 在 PBMC 里找 STAT1 的 29.5 天成分:一个谱峰分不清钟还是传令兵

    Evidence level: B

    三个靶蛋白里,STAT1 是唯一在哺乳动物有硬证据的那个。但正因为它太”活”,要在人血里读出它的月相成分,比读钟基因难得多——难点不在采样密度,在于你分不清读到的是零件还是传令兵。

    正文

    上一篇理出一份清单:环节虫月亮钟的靶蛋白里,proctolin 在脊椎动物没有直系同源,ependymin 家族善变到找不到清晰对应物,唯有 STAT1 不同——它在哺乳动物里有确凿的”时钟相关节律功能”文献支撑(PMID 29996098、PMID 32477351)。它因此成了最值得先做的一个。本文要讲的是:它也是最容易被做错的一个。

    第一个坑:PBMC 不是一种细胞,是一个池子

    外周血单个核细胞(PBMC)是单核细胞、T 细胞、B 细胞、NK 细胞的混合物。这几类细胞的 STAT1 基线表达量本来就不同,而它们在血液里的比例本身就在动——免疫细胞亚群的转运与功能有明确的昼夜依赖(PMID 31109762),这种转运由器官间通讯网络控制(PMID 33751050)。后果是:若月相真的改变了单核细胞与淋巴细胞的配比,整个 PBMC 池的平均 STAT1 信号就会跟着变,而每一种细胞内部什么都没变。你采到的是”配方变了”,报出来却是”STAT1 表达有月相成分”。不做亚群分选或单细胞测序,这个混淆解不开。

    第二个坑:p-STAT1 是个比值,两个端点要分开报

    磷酸化是昼夜控制的中心机制之一(PMID 27818261),STAT1 的功能状态通常靠 Tyr701 与 Ser727 位点的磷酸化比例来读。但”比例”这个词藏着陷阱:p-STAT1/STAT1 上升,可能来自磷酸化加强(调电压),也可能来自总蛋白减少(分母变小)。这是两种不同的调控逻辑,一个说”通路开大了”,一个说”蛋白被降解了”。报”STAT1 活性有月相节律”,会把两者糊成一团。必须做的是总蛋白谱、磷酸化谱、比值谱三条线分别报。全血流式测 p-STAT1 已有临床级方案(PMID 40265078),早年信号谱检测也在新生儿样本上跑通过(PMID 23809143)——技术上可行,但那些方案意在把采样时刻固定住以便比较病人,不是把时刻当自变量研究。

    第三个坑:它被多少已知因素改写

    STAT1 是一个随时被外部输入改写的效应器,不是安静待着等你读的钟。至少这些因素会动它:进食(餐后 PBMC 基因表达存在日间差异,PMID 33518404)、睡眠(睡眠不足会改写整个人血转录组的表达振幅,PMID 23440187)、性别与性激素(月经周期对免疫功能与炎症的影响有系统综述与 meta 分析支撑,PMID 37309068、PMID 36195867;免疫与月经周期的关系早被综述过,PMID 22155200;自身免疫病的性别差异机制仍在梳理,PMID 40806232)、个体间差异(单细胞层面对病毒的反应在人群间差异极大,PMID 37558883)。再叠上健康人免疫指标可观的个体内波动(PMID 28786508)与多重免疫测定的方法学变异(PMID 27810706),”多大的月相成分才算超过噪声”必须先有答案。

    最核心的张力:一个谱峰,分不开两种身份

    假设你真做完了连续三个周期、固定昼夜时刻、分层性别的采样,并用 JTK_CYCLE 或 RAIN(PMID 20876817、PMID 25326247)跑出了 STAT1 总量的 29.5 天成分。这仍然回答不了本文最关心的问题:STAT1 是”近月钟的零件”,还是”月亮信号的传令兵”?

    两种身份在周期图上给出同一个峰,判据却完全不同。若是传令兵(下游执行器),它该跟着别的月级节律走——你该去看它与其他月级读数的相位关系。若是零件(内部计时器),它该具备自主性——相位在个体间自持、与昼夜相位解耦、在缺少外部月相标记时仍走。一句”STAT1 有月相节律”,把这两种可能一起说了,等于什么都没说。 这就是本文要立的判据:不是”有没有峰”,而是”峰的相位从哪来、被谁牵”。人血转录组已能推断昼夜相位(PMID 28218891),能否推到月相层尚无人试。

    最小可行方案

    不追求一步到位:①先在分选后的单核细胞上做(避开混合池陷阱);②总量/磷酸化/比值三条线分报;③至少纳入空腹、固定昼夜时刻,同步记录睡眠与月经周期相位作为协变量;④以个体为自身对照、用随机效应模型,让”月相成分”去和”个体内噪声”正面比大小;⑤女性组里 28 天与 29.5 天两个成分必须同时拟合(PMID 38598621 主张卵巢周期由内源月级计时系统驱动,PMID 42380750 从演化与生理两侧论证人类近月钟的存在,PMID 35567349 则在儿童褪黑素上给了一个真实的近月实证)。

    一个参照系的问题

    《素问·脉要精微论》给诊法定了条硬规矩:”诊法常以平旦,阴气未动,阳气未散,饮食未进,经脉未盛,络脉调匀,气血未乱,故乃可诊有过之脉。”道理很清楚:你为什么必须在一个固定时刻测?因为别的时刻读数已被”阴气动、阳气散、饮食进、气血乱”改写过——你读到的不是病,是刚刚发生的事。 现代的做法是空腹、固定采血时间。

    但这段规矩管的只是一天里的那一格。本文要补的是:这个读数身上还压着另一把尺子,刻度是 29.5 天。

    更要紧的是”平人”这个概念。《素问·平人气象论》讲”平人者,不病也”,用平人的脉当标准,一切偏离都相对它算。可如果连”平人”的 STAT1 基线自己就在走一个 29.5 天的圈呢? 那参照系就是动的——你测到的每一次”异常”,都可能只是尺子在晃。所有测量的前提是参照系不动。中医把平人当参照系,现代医学把健康对照的基线当参照系,而这两个参照系,都没被检查过自己是不是平的。

    证据与局限

    已验证: ①STAT1 是哺乳动物昼夜钟-免疫交叉的直接执行者(PMID 29996098、PMID 32477351);②磷酸化是昼夜控制的中心机制(PMID 27818261);③免疫细胞亚群转运与功能存在昼夜依赖(PMID 31109762、PMID 33751050);④全血流式检测 p-STAT1 的方法已建立(PMID 40265078、PMID 23809143);⑤PBMC 昼夜钟基因表达可测且模式已被刻画(PMID 16096739、PMID 17412523);⑥人血转录组可用于推断昼夜相位(PMID 28218891)。

    部分验证: ①进食与睡眠会改写 PBMC 基因表达读数(PMID 33518404、PMID 23440187)——但改写的量级相对月相成分的比值无标定;②月经周期对免疫功能的影响有 meta 分析支持(PMID 37309068、PMID 36195867)——但 28 天与 29.5 天成分在同一批样本中的可分离性未测;③人类近月节律有真实实证(PMID 35567349、PMID 38598621、PMID 42380750)——靶组织仍未落到免疫细胞。

    待验证: ①人类 PBMC(或分选单核细胞)中 STAT1 总量、磷酸化水平与比值的 29.5 天表达成分,均无研究执行;②该成分的相位是否在个体间自持、是否与昼夜相位解耦(”零件 vs 传令兵”的判据尚无数据);③月相成分幅度相对个体内波动(PMID 28786508)与方法学变异(PMID 27810706)的相对大小;④”平人基线是否本身随月相漂移”这一问题在免疫指标上从未被检验。

    相关阅读

    证据等级: B
    参考文献: PMID 29996098, PMID 32477351, PMID 31109762, PMID 33751050, PMID 27818261, PMID 40265078, PMID 23809143, PMID 16096739, PMID 17412523, PMID 33518404, PMID 23440187, PMID 37309068, PMID 36195867, PMID 22155200, PMID 40806232, PMID 37558883, PMID 28786508, PMID 27810706, PMID 20876817, PMID 25326247, PMID 28218891, PMID 38598621, PMID 42380750, PMID 35567349


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Identifying the 29.5-day component of STAT1 in PBMCs: Is it a peak or a messenger?

    Evidence level: B

    Among the three target proteins, STAT1 is the only one with robust evidence in mammals. However, precisely because it is so “active,” reading its lunar phase component in human blood is far more difficult than reading clock genes—the challenge lies not in sampling density, but in distinguishing whether the signal read reflects a component or a messenger.

    Main Text

    In the previous article, we compiled a list: among the target proteins of the leech lunar clock, proctolin lacks a direct homolog in vertebrates, the ependymin family is too variable to find a clear counterpart, and only STAT1 stands out—it has solid literature support for “clock-related rhythmic function” in mammals (PMID 29996098, PMID 32477351). It therefore becomes the most logical starting point. This article addresses: it is also one of the most easily misinterpreted.

    Pitfall One: PBMCs Are Not a Single Cell Type, But a Pool

    Peripheral blood mononuclear cells (PBMCs) are a mixture of monocytes, T cells, B cells, and NK cells. The baseline STAT1 expression levels differ inherently among these cell types, and their proportions within the blood are dynamic—immune cell subpopulation trafficking and function exhibit clear circadian dependencies (PMID 31109762), regulated by inter-organ communication networks (PMID 33751050). The consequence is: if the lunar phase truly alters the ratio of monocytes to lymphocytes, the average STAT1 signal across the entire PBMC pool will shift, while nothing changes within each individual cell. You are measuring a “changed recipe,” yet reporting it as “STAT1 expression has a lunar component.” Without subpopulation sorting or single-cell sequencing, this confounding factor cannot be resolved.

    Pitfall Two: p-STAT1 Is a Ratio; Both Endpoints Must Be Reported Separately

    Phosphorylation is one of the central mechanisms for circadian control (PMID 27818261), and the functional state of STAT1 is typically read via the phosphorylation ratio at Tyr701 and Ser727 sites. However, the term “ratio” hides a trap: an increase in p-STAT1/STAT1 may stem from enhanced phosphorylation (increasing the numerator) or reduced total protein (decreasing the denominator). These represent two distinct regulatory logics: one suggests “the pathway is turned up,” while the other implies “protein is being degraded.” Reporting “STAT1 activity has a lunar rhythm” conflates these two possibilities. The necessary approach is to report three distinct lines: total protein spectra, phosphorylation spectra, and ratio spectra. Whole-blood flow cytometry for p-STAT1 already has a clinical-grade protocol (PMID 40265078), and earlier signal spectra detection has been validated on neonatal samples (PMID 23809143)—technically feasible, yet those protocols aim to fix sampling times for patient comparison rather than treat time as an independent variable.

    Pitfall Three: How Many Known Factors Rewrite It?

    STAT1 is an effector constantly rewritten by external inputs, not a silent clock waiting to be read. At least the following factors influence it: feeding (post-prandial PBMC gene expression exhibits diurnal differences, PMID 33518404), sleep (sleep deprivation rewrites the expression amplitude of the entire human blood transcriptome, PMID 23440187), gender and sex hormones (systematic reviews and meta-analyses support the impact of the menstrual cycle on immune function and inflammation, PMID 37309068, PMID 36195867; the relationship between immunity and the menstrual cycle was previously reviewed, PMID 22155200; mechanisms for gender differences in autoimmune diseases are still being elucidated, PMID 40806232), and inter-individual variability (single-cell responses to viruses vary drastically across populations, PMID 37558883). Layered upon this are significant intra-individual fluctuations in healthy immune markers (PMID 28786508) and methodological variations across multiple immune assays (PMID 27810706). Therefore, the answer to “how large a lunar component exceeds noise” must be established beforehand.

    The Core Tension: One Spectral Peak, Two Distinct Identities

    Assume you have completed sampling across three consecutive cycles at fixed circadian times, stratified by gender, and used JTK_CYCLE or RAIN (PMID 20876817, PMID 25326247) to identify a 29.5-day component of STAT1 total abundance. This still does not answer the central question of this article: Is STAT1 a “part of the near-lunar clock” or a “messenger for lunar signals”?

    Both identities produce the same peak on the periodogram, yet the criteria for validation differ completely. If it is a messenger (downstream effector), it should follow other lunar rhythms—you should examine its phase relationship with other lunar readings. If it is a part (internal timer), it should possess autonomy—its phase should be self-sustaining across individuals, decoupled from circadian phases, and continue running even in the absence of external lunar markers. Saying “STAT1 has a lunar rhythm” covers both possibilities simultaneously, amounting to saying nothing. This is the criterion we establish here: not “whether there is a peak,” but “where the phase originates and what drives it.” Human blood transcriptomes can already infer circadian phases (PMID 28218891); extending this inference to lunar phases remains untested.

    Minimum Viable Protocol

    Do not aim for a single step: ① Perform the study first on sorted monocytes (avoiding the mixed pool trap); ② Report total, phosphorylated, and ratio lines separately; ③ Include fasting status, fixed circadian times, and record sleep and menstrual cycle phases as covariates; ④ Use individuals as their own controls with random-effects models, allowing the “lunar component” to be directly compared against “intra-individual noise”; ⑤ In the female cohort, both the 28-day and 29.5-day components must be simultaneously fitted (PMID 38598621 argues for an endogenous lunar-timing system driving the ovarian cycle, PMID 42380750 argues for the existence of a human near-lunar clock from evolutionary and physiological perspectives, PMID 35567349 provides real near-lunar empirical evidence in human melatonin).

    The Problem of a Reference Frame

    The Su Wen · Pulse and Essence establishes a rigid rule for diagnosis: “Diagnosis should always be conducted at dawn, before Yin Qi moves and Yang Qi disperses, before food is ingested, before meridians are full, before collaterals are unbalanced, and before Qi and Blood are chaotic; thus, one may detect pathological pulses.” The logic is clear: Why must measurement occur at a fixed time? Because readings at other times are rewritten by “Yin Qi movement, Yang Qi dispersion, food ingestion, Qi and Blood chaos”—what you read is not disease, but a recent event. Modern practice involves fasting and fixed blood draw times.

    However, this rule governs only one grid within a day. This article aims to supplement it: This reading also bears another ruler, scaled at 29.5 days.

    More critically is the concept of “Ping Ren” (healthy person). The Su Wen · Ping Ren Qi Xiang Lun states “Ping Ren refers to one without illness,” using the pulse of a healthy person as the standard against which all deviations are calculated. But what if the STAT1 baseline of even a “Ping Ren” is itself traversing a 29.5-day circle? Then the reference frame is moving—every measured “abnormality” might simply be the ruler swinging. All measurement presupposes a stable reference frame. Traditional Chinese medicine treats “Ping Ren” as the reference; modern medicine treats healthy control baselines as the reference. Yet neither of these reference frames has been verified for their own stability.

    Evidence and Limitations

    Verified: ① STAT1 is a direct effector of the mammalian circadian clock-immune crosstalk (PMID 29996098, PMID 32477351); ② Phosphorylation is a central mechanism for circadian control (PMID 27818261); ③ Immune cell subpopulation trafficking and function exhibit circadian dependencies (PMID 31109762, PMID 33751050); ④ Whole-blood flow cytometry for p-STAT1 has an established protocol (PMID 40265078, PMID 23809143); ⑤ Circadian clock gene expression in PBMCs is measurable and its patterns have been characterized (PMID 16096739, PMID 17412523); ⑥ Human blood transcriptomes can be used to infer circadian phases (PMID 28218891).

    Partially Verified: ① Feeding and sleep rewrite PBMC gene expression readings (PMID 33518404, PMID 23440187)—but the magnitude of rewriting relative to the ratio of lunar components lacks calibration; ② The impact of the menstrual cycle on immune function is supported by meta-analysis (PMID 37309068, PMID 36195867)—but the separability of 28-day and 29.5-day components within the same sample batch remains untested; ③ Real empirical evidence for human near-lunar rhythms exists (PMID 35567349, PMID 38598621, PMID 42380750)—but the target tissue has not yet been localized to immune cells.

    To Be Verified: ① No study has executed the 29.5-day expression component of STAT1 total abundance, phosphorylation levels, and ratios in human PBMCs (or sorted monocytes); ② Whether the phase of this component is self-sustaining across individuals and decoupled from circadian phases (no data yet for the “part vs. messenger” criterion); ③ The relative magnitude of the lunar component amplitude compared to intra-individual fluctuations (PMID 28786508) and methodological variations (PMID 27810706); ④ The question of whether the “Ping Ren baseline itself drifts with the lunar phase” has never been tested in immune markers.

    Related Reading

    Evidence Level: B
    References: PMID 29996098, PMID 32477351, PMID 31109762, PMID 33751050, PMID 27818261, PMID 40265078, PMID 23809143, PMID 16096739, PMID 17412523, PMID 33518404, PMID 23440187, PMID 37309068, PMID 36195867, PMID 22155200, PMID 40806232, PMID 37558883, PMID 28786508, PMID 27810706, PMID 20876817, PMID 25326247, PMID 28218891, PMID 38598621, PMID 42380750, PMID 35567349


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 只改颜色不改亮度,能不能救回天人相应:蓝光阻断与同色异谱的光谱隔离判决实验

    Evidence level: B

    如果破坏”天人相应”的是光里的蓝,那么把蓝拿掉、亮度一分不减,耦合就该回来。这一刀能不能单独成立,是照明方案里最贵的一个问题——因为”调暗”伤眼睛,”调色”不伤。

    正文

    前几篇已把三件事做实:月相-月经耦合真存在且正在被夜间光照系统性破坏(PMID 40991714);破坏需要画成个体级剂量-效应曲线;这条曲线是”阈值型”还是”坡度型”至今没判。但普通人真正会问的是第四件事:我到底该调暗,还是该调色? 前者牺牲视觉,后者牺牲不多——所以”光谱能不能在亮度不变的前提下单独起作用”,是整个链条里最有实际价值、也最容易被商家糊弄的一问。

    第一,实验室里这一刀是成立的,而且做得极其干净。 Souman 等人(PMID 29984614)把 15 名受试者分三个晚上暴露于 3 小时白光:两组照度都是 175 lux、色温都是 2700 K、视觉上几乎分不出差别,唯一被调的是 450-500 nm 波段的功率。结果——高功率那组抑制褪黑素近 50%,低功率那组(把砍掉的功率补到更短波长上)与暗光基线无差异。同一亮度、同一色温、同一外观,生理结局一个接近全抑、一个接近零。这是”光谱能独立于强度改变生理结局”最直接的证据。de Zeeuw 等人(PMID 31156018)在白天用同样的手法证实瞳孔光反应与客观嗜睡度随黑视素辐照度变化,而照度、色度不变。

    第二,可这一招搬到真实屏幕上就失灵了。 Nagare 等人(PMID 31191118)让受试者夜里连看两小时 iPad,把 Night Shift 拉到”最暖”(2837 K) 与”最不暖”(5997 K) 两个极端——两次的褪黑素抑制没有显著差别,都显著抑制。同期的 Cochrane 系统评价(PMID 37593770,17 项 RCT)更冷:蓝光过滤镜片对短期视疲劳无临床意义效果,对睡眠的证据”极低确定性”,6 项 RCT 里 3 项报改善 3 项报无差异,没有任何一项测过血清褪黑素。

    这两组结果不矛盾,它们划出了一条边界。 Souman 的同色异谱对是精心反向补偿造出来的,两组黑视素等效照度相差数倍;而屏幕色温滑条能拉开的差距小得多,且在夜间近距离低照度的场景里,黑视素的相对贡献本就下降。所以”调色够不够”不是一个是非题,而取决于你能把光谱差拉多开、以及在哪个剂量区间操作。

    第三,最要紧的指控在这里:既往说”蓝光阻断有效”的研究,几乎都混淆了强度。 van der Lely 等人(PMID 25287985)报告阻断镜片削弱了 LED 屏幕诱发的褪黑素抑制;Liset 等人的 RCT(PMID 35024497)让孕妇佩戴两周后褪黑素起始提前 28 分钟(p=0.019);Zerbini 等人的现场研究(PMID 30506899)也看到佩戴后褪黑素与入睡提前。但这些干预都是滤片——橙黄镜片在挡掉蓝光的同时,也把到达视网膜的明视觉光子通量一起降下来了。“光谱独立贡献”从未在自由生活条件里被真正隔离过。 而 Maeda-Nishino 等人 2025 年的交叉试验(PMID 41166315,39 名 10-12 岁学童,40% 蓝光阻断镜片)给了一记重击:入睡相位显著提前(p=0.040),但唾液褪黑素一点没变——效应不是褪黑素中介的。这直接动摇了”BB 眼镜→褪黑素→耦合”这条默认通路。

    第四,判决实验必须以黑视素等效照度为剂量单位,不能再用 lux。 Brown(PMID 32248548)汇总 19 项实验室研究证明:黑视素等效照度(melanopic EDI)能几乎完全解释不同光谱刺激间的敏感度差异;Brown 团队随后给出三档共识建议(PMID 35298459):白天≥250、睡前 3 小时起≤10、睡眠环境≤1 melanopic EDI lux。如果剂量仍用 lux 度量,”保持强度不变”这句话本身就是假的——因为 lux 只反映明视觉,恰好把要检验的那个变量漏掉了。

    第五,据此可写出判决设计。 交叉设计 RCT:受试女性随机进入”高黑视素/低黑视素”两种同色异谱照明(明视觉 lux、色温、色度三匹配,mEDI 相差 ≥3 倍),在家连续佩戴眼位分光光度计记录 6-12 个周期;对照组不是普通眼镜(那会改总通量),而是”同亮度低 mEDI 照明”;结局不是 DLMO,而是月相-月经同步指数(相对协调度/Rayleigh);分析按个体 mEDI 分层而非 lux 分层。若同步度在高 mEDI 组显著下降、低 mEDI 组保住,则光谱是独立杠杆;若两组无差,则亮度才是主因。

    对”天人相应”的意义: 风水里讲”灯的色””窗户的朝向与贴膜””卧室宜暖不宜冷”,一直缺一个能被证伪的量化版本。这个实验就是那个版本——它只问一句话:把亮留给人眼、把蓝从时钟眼前拿开,月亮还找得到你吗? 答”是”,则”调色”是一条独立可行的路;答”否”,则所有”低蓝光模式”的安慰剂成分,比商家愿意承认的要大。

    证据与局限

    已验证: 光谱可在照度、色温、外观均不变的前提下单独大幅改变褪黑素抑制(175 lux / 2700 K 条件下由近 50% 抑制降至与暗光无差异,PMID 29984614);同色异谱效应在白天可经瞳孔光反应与客观嗜睡度独立复现(PMID 31156018);黑视素等效照度可解释不同光谱刺激的敏感度差异(PMID 32248548),并已有白天≥250 / 睡前≤10 / 睡眠≤1 mEDI lux 的共识建议(PMID 35298459);蓝光阻断镜片在 LED 屏幕(PMID 25287985)、孕期(PMID 35024497)、自由生活(PMID 30506899)等场景可提前褪黑素或入睡相位。

    部分验证: 蓝光阻断的效应量在消费品级滤片(Cochrane 17 RCT,睡眠结局极低确定性、无一项测血清褪黑素,PMID 37593770)与实验级同色异谱调控之间差距悬殊,提示效果高低取决于可拉开的 mEDI 差距与剂量区间;部分研究已发现效应不经褪黑素中介(PMID 41166315)。

    待验证: 既往蓝光阻断研究中”光谱效应”与”强度混杂”各占多少(无人做过强度匹配对照);光谱隔离能否以月相-月经同步度为结局独立起效(尚未有以同步度为结局的光谱干预试验);若效应确不经褪黑素,其替代通路是什么(ipRGC 直接投射或其他非褪黑素通路);同色异谱对在真实居家照明中能实现的最大 mEDI 对比度上限。

    相关阅读

    证据等级: B
    参考文献: PMID 29984614, PMID 31156018, PMID 31191118, PMID 37593770, PMID 25287985, PMID 35024497, PMID 30506899, PMID 41166315, PMID 32248548, PMID 35298459, PMID 40991714


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Color Adjustment Without Brightness Change: Can We Rescue Color Correspondence? A Spectral Isolation Judgment Experiment Using Blue Light Blocking and Isoluminant Colors

    Evidence level: B

    If the disruption of “Heaven-Human Correspondence” lies within the blue component of light, then removing the blue while maintaining full brightness should restore coupling. Whether this single intervention holds its own is one of the most costly questions in lighting design—because “dimming” harms vision, while “color tuning” does not.

    Main Text

    Previous articles have established three facts: a genuine coupling between lunar phase and menstruation exists and is systematically disrupted by nighttime illumination (PMID 40991714); this disruption requires mapping at an individual dose-response curve; and whether this curve is “threshold-type” or “slope-type” remains undetermined. However, the question ordinary people truly ask is the fourth: Should I dim or should I tune colors? Dimming sacrifices vision; color tuning sacrifices little. Therefore, whether “spectrum can act independently while maintaining constant brightness” is the most practically valuable yet easily misleading question in the entire chain.

    First, this cut holds true in the laboratory and is executed with extreme precision. Souman et al. (PMID 29984614) exposed 15 subjects over three nights to white light for 3 hours. Two groups received identical illuminance (175 lux) and color temperature (2700 K), visually indistinguishable; the only variable was power in the 450-500 nm band. The result—the high-power group suppressed melatonin by nearly 50%, while the low-power group (where the removed power was shifted to shorter wavelengths) showed no difference from the dark baseline. Same brightness, same color temperature, same appearance; one physiological outcome approaches total suppression, the other approaches zero. This is the most direct evidence that “spectrum can independently alter physiological outcomes independent of intensity.” de Zeeuw et al. (PMID 31156018) confirmed in daylight that pupillary light response and objective sleepiness vary with melanopic irradiance while illuminance and chromaticity remained unchanged.

    Second, however, this trick fails when transferred to real screens. Nagare et al. (PMID 31191118) had subjects use iPads for two consecutive nights, setting Night Shift to the “warmest” (2837 K) and “coolest” (5997 K) extremes. Melatonin suppression showed no significant difference between the two, yet both were significantly suppressed. Concurrently, a Cochrane systematic review (PMID 37593770; 17 RCTs) was even colder: blue-light-blocking lenses showed no clinically meaningful effect on short-term visual fatigue; evidence for sleep was of “very low certainty,” with 6 RCTs reporting improvement in 3 and no difference in the other 3, none measuring serum melatonin.

    These two sets of results are not contradictory; they delineate a boundary. Souman’s same-color heterospectral pairs were carefully created via reverse compensation, differing by several times in equivalent melanopic irradiance. In contrast, the gap screen color temperature sliders can open is much smaller, and in nighttime close-range low-illumination scenarios, the relative contribution of melanopsin naturally declines. Therefore, “is color tuning sufficient” is not a binary yes/no question but depends on how much spectral difference you can generate and within which dose range you operate.

    Third, the most critical accusation lies here: past studies claiming “blue light blocking is effective” almost all confound intensity. van der Lely et al. (PMID 25284614) reported that blocking lenses attenuated melatonin suppression induced by LED screens; Liset et al.’s RCT (PMID 35024497) found that after two weeks of wear, melatonin onset advanced by 28 minutes (p=0.019); Zerbini et al.’s field study (PMID 30506899) also observed advanced melatonin and sleep onset upon wearing. However, these interventions were filters—orange-yellow lenses block blue light but simultaneously reduce the total photometric photon flux reaching the retina. “Independent spectral contribution” has never been truly isolated under free-living conditions. Maeda-Nishino et al.’s 2025 crossover trial (PMID 41166315; 39 school-aged children, 40% blue-light-blocking lenses) delivered a heavy blow: sleep phase advanced significantly (p=0.040), but salivary melatonin did not change—the effect was not mediated by melatonin. This directly undermines the default pathway of “BB glasses → melatonin → coupling.”

    Fourth, judgment experiments must use equivalent melanopic irradiance as the dose unit; lux cannot be used again. Brown (PMID 32248548) summarized 19 laboratory studies proving that equivalent melanopic irradiance (melanopic EDI) can almost completely explain sensitivity differences between different spectral stimuli. Brown’s team subsequently issued three tiers of consensus recommendations (PMID 35298459): daytime ≥250, starting 3 hours before sleep ≤10, sleep environment ≤1 melanopic EDI lux. If dose is still measured in lux, the statement “maintain constant intensity” itself is false—because lux only reflects photometric vision,恰好 missing the very variable being tested.

    Fifth, a judgment design can thus be formulated. Crossover RCT: Randomize female subjects into “high melanopic/low melanopic” same-color heterospectral illumination (matching photometric lux, color temperature, and chromaticity, with mEDI differing by ≥3x); record 6-12 cycles at home wearing eye-position spectrophotometers; the control group is not ordinary glasses (which would alter total flux), but “same brightness low mEDI illumination”; the outcome is not DLMO, but the lunar phase-menstrual synchronization index (relative coordination/Rayleigh); analysis stratifies by individual mEDI rather than lux. If synchronization drops significantly in the high mEDI group but is preserved in the low mEDI group, spectrum is an independent lever; if no difference between groups, brightness is the primary factor.

    Significance for “Heaven-Human Correspondence”: Feng Shui discusses “lamp color,” “window orientation and film,” and “warm is better than cold for bedrooms,” yet lacks a falsifiable quantitative version. This experiment is that version—it asks only one question: Keep brightness for the human eye, remove blue from the clock’s view; can the moon still find you? If “yes,” then “color tuning” is an independently viable path; if “no,” then the placebo component of all “low blue light modes” is larger than manufacturers are willing to admit.

    Evidence and Limitations

    Verified: Spectrum can significantly alter melatonin suppression while keeping illuminance, color temperature, and appearance constant (dropping from near 50% suppression to dark-baseline equivalence at 175 lux / 2700 K, PMID 29984614); same-color heterospectral effects can be independently reproduced in daylight via pupillary light response and objective sleepiness (PMID 31156018); equivalent melanopic irradiance explains sensitivity differences across spectral stimuli (PMID 32248548), with consensus recommendations already established for daytime ≥250 / pre-sleep ≤10 / sleep ≤1 mEDI lux (PMID 35298459); blue-light-blocking lenses advance melatonin or sleep phase in LED screens (PMID 25284614), pregnancy (PMID 35024497), and free-living settings (PMID 30506899).

    Partially Verified: The effect size of blue-light blocking in consumer-grade filters (Cochrane 17 RCTs, sleep outcomes of very low certainty, no serum melatonin measured, PMID 37593770) differs vastly from experimental-level same-color heterospectral regulation, suggesting efficacy depends on the achievable mEDI gap and dose range; some studies have found effects not mediated by melatonin (PMID 41166315).

    To Be Verified: The proportion of “spectral effects” vs. “intensity confounding” in past blue-light blocking studies (no one has performed intensity-matched controls); whether spectral isolation can independently exert effects using lunar phase-menstrual synchronization as the outcome (no spectral intervention trials have used synchronization as an endpoint yet); if the effect indeed bypasses melatonin, what is the alternative pathway (ipRGC direct projection or other non-melatonin pathways); the maximum mEDI contrast ratio achievable in real home lighting with same-color heterospectral pairs.

    Related Reading

    Evidence Level: B
    References: PMID 29984614, PMID 31156018, PMID 31191118, PMID 37593770, PMID 25284614, PMID 35024497, PMID 30506899, PMID 41166315, PMID 32248548, PMID 35298459, PMID 40991714


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 池④ δ 表参数层溯源审计:σ/μr 标签与 δ 数值的端点错配(#38)

    Evidence level: C

    一句话结论:上一轮把 δ 表偏差归因为「整列偏移(H-col 假设),σ 列与 δ 列不同源」——这个机制判定不成立。逐行反推后,9 行呈现三种不同偏差(3.16× / 10× / 0.32× / 1.02×),没有单一常数可解释。真实机制是「δ 数值与 σ/μr 标签各自取区间的不同端点」,且金属行取自另一套厂商数据。方向性后果:黏土/页岩行的 δ 被放大 10 倍,把本该是 48 dB 强屏蔽的工况写成了 4.8 dB 弱衰减 → 「山体对 ELF 几乎透明」这一稳健结论缺少适用边界。


    1. 本轮的起点与上一轮的结论

    上一轮(第 37 轮 / #37)完成了 δ/A 表全量扫描,得出:

    1. 838 文件扫完,14 文件含断言,86 断言 / 27 OK / 59 MISMATCH;
    2. → 提出 H-col 假设:「σ 列与 δ 列不由同一套参数算出,δ 列先算、σ 列事后独立填,两列从未交叉核对」,并称比值「高度聚集在 10.0 与 ~100 两点」,海水行恰好 1.0。

    本轮(#38)的任务是参数层溯源:不去问「δ 对不对」,而是问「σ/μr 这些标签本身从哪来、对不对得上」。

    结论先放:扫描范围是对的,机制解释是错的。 上一轮只看了 6 行(干燥/潮湿/混凝土/钢筋/黏土/海水),恰好这 6 行都落在「δ 用了更小的 σ」,于是得出”单一偏移”的印象;补上 3 个金属行后,方向立刻分裂。


    2. 引文层核验:三个 DOI 真实,但表内引用有一处归属错

    2.1 三个 DOI 全部通过 Crossref 核验

    锚点 DOI Crossref 返回的真实身份 判定
    A1 10.1190/1.9781560802631.ch3 Palacky G.J., “3. Resistivity Characteristics of Geologic Targets”, Electromagnetic Methods in Applied Geophysics, SEG, 1988, pp. 52–129 ✅ 真实
    A2 10.1029/gm013p0469 Madden T.R. & Swift C.M., “Magnetotelluric Studies of the Electrical Conductivity Structure of the Crust and Upper Mantle”, Geophysical Monograph Series, AGU, 2013 ✅ 真实
    A3 10.1046/j.1365-246x.1999.00707.x Eisel M. & Haak V., “Macro-anisotropy of the electrical conductivity of the crust: a magnetotelluric study of the German Continental Deep Drilling site (KTB)”, Geophys J Int, 1999 ✅ 真实

    2.2 但表内的文字引用与这三个锚点对不上

    skin-depth-reference-table.md §二 的脚注原文是:

    区间参考:Palacky & Kligerman (1980) 矿集区电导率;Schön (2015) …

    Crossref 按作者 Palacky 检索 1975–1995 全部 15 条结果,没有任何一条与 Kligerman 合著,也没有 1980 年的条目;Palacky 在电导率领域的经典综述是 1988 年独著的 SEG 章节(即 A1)。也就是说:

    • 可核的是 Palacky 1988 独著章节;
    • 表里写的是「Palacky & Kligerman 1980」——年份、合著者两处均无法核验。

    这不是「PMID 假号」,而是第 4 类引文失效:可核文献被写成了不可核的年份/合著组合。性质比假号轻(指向的领域是对的),但后果相同:读者顺着这条脚注查不到源。

    2.3 Schön (2015) 的载体可核,但缺章级定位

    Crossref 查到 Schön 2015 Physical Properties of Rocks 系列章节,其中 10.1016/b978-0-08-100404-3.00008-1 即 “Electrical Properties” 章——这才是 σ 区间的真正出处,而表里只写书名、未给章。建议补为章级引用。

    2.4 六个 PMID 全部真实(与前三轮结论一致)

    PMID 真实身份 在本文的合理用途
    20000094 Chernykh AM, Gig Sanit 2009 — 住宅地磁场屏蔽 建筑屏蔽实测
    28654641 Binhi VN, PLoS One 2017 — 低磁场生物效应综述 μ-metal 屏蔽背景
    38132339 Sarimov RM, Biology 2023 — 低磁条件生物作用综述 同上
    36676419 Cheng Y, Materials 2023 — 低剩磁材料测试方法 μ-metal 材料数据
    24131394 Sirav B, Electromagn Biol Med 2014 — 变压器 ELF 磁场 环境 ELF 暴露
    36681118 Martel J, Biomed J 2023 — EMF 与昼夜节律 环境 ELF 生物效应

    注:上一轮记录 24131394 标注为"Burchard"、36681118 标注为"Girgert"、28654641 标注为"Lewczuk" —— 三处作者错配在本轮复查中仍然存在于 elf-skin-depth-table-numeric-integrity-audit.md 的正文表述里,属待修项(不阻塞本轮)。


    3. 参数层复算:偏差不是一个常数

    3.1 方法

    取 skin-depth-reference-table.md §三 的每一行,用该行自己标注的 σ 与 μr 代入 δ=√(2/(ωμ₀μᵣσ))(f=7.83 Hz),与表列 δ 比较;再用表列 δ 反推隐含的 σ·μr 乘积,与标签的 σ·μr 相比。

    3.2 结果(9 行,f = 7.83 Hz)

    介质 标签 σ 标签 μᵣ 表列 δ 按标签算 δ δ比(表/算) 隐含σμᵣ ÷ 标签σμᵣ 判定
    干燥花岗岩 10⁻⁴ 1 57 km 18.0 km 3.17 0.0996 δ 用了 σ≈10⁻⁵
    潮湿花岗岩 10⁻³ 1 18 km 5.69 km 3.16 0.0999 δ 用了 σ≈10⁻⁴
    普通混凝土 10⁻² 1 5.7 km 1.80 km 3.17 0.0996 δ 用了 σ≈10⁻³
    钢筋混凝土 10⁻¹ 5 2.5 km 254 m 9.83 0.0104 δ 用了 σμᵣ≈10⁻³
    黏土/页岩 10⁰ 1 1.8 km 180 m 10.01 0.0100 δ 用了 σ≈10⁻²
    海水 4 1 90 m 89.9 m 1.00 0.999 ✅ 自洽
    低碳钢 10⁶ 100 5.7 mm 18.0 mm 0.32 9.96 ⚠️ 方向相反(δ 偏小)
    纯铁 10⁷ 2000 1.3 mm 1.27 mm 1.02 0.957 ✅ 自洽
    μ-metal 10⁶ 5×10⁴ 0.25 mm 0.80 mm 0.31 10.35 ⚠️ 方向相反(δ 偏小)

    3.3 三种偏差并存 → H-col「单一整列偏移」被否

    • 5 行(岩石/建材):δ 比 ≈ 3.16 或 10,σ 标签比实际用的大 10–100 倍;
    • 2 行(低碳钢、μ-metal):δ 比 ≈ 0.31,方向相反,隐含 σμᵣ 比标签大 10 倍;
    • 2 行(海水、纯铁):完全自洽。

    若真的是「整列错位」(一列整体填错位置),9 行应给出同一个倍数。实际给出 3.17 / 3.16 / 3.17 / 9.83 / 10.01 / 1.00 / 0.32 / 1.02 / 0.31 九个不同值,跨三个数量级区段、且符号不一致。

    → 上一轮的 H-col 假设(#41)在本轮被证伪。可结案,不必再追中间稿。

    3.4 更准确的机制:端点错配(endpoint mismatch)+ 双数据源混用

    把 5 行岩石/建材的实际用值与 §二 的区间对照,规律立刻出来:

    介质 §二 声明区间 表列 δ 实际对应的 σ 该值在区间中的位置
    干燥花岗岩 10⁻⁶ – 10⁻⁴ ≈10⁻⁵ 对数中点
    潮湿花岗岩 10⁻⁴ – 10⁻² ≈10⁻⁴ 下限
    普通混凝土(含钢筋) 10⁻³ – 10⁻¹ ≈10⁻³ 下限
    黏土/页岩 10⁻² – 10⁰ ≈10⁻² 下限
    钢筋混凝土 10⁻³ – 10⁻¹, μᵣ1–10 σμᵣ≈10⁻³ 下限(σ 取下限 × μᵣ=1)

    → δ 数值一律按区间「下限/中点」算,而 §三 的 σ 标签列写的却是「区间上端或中点」。 两列取的是不同端点,所以比值不是固定 10×,而是取决于该行区间跨几个数量级(跨 2 个数量级 → √100 = 10×;跨 1 个 → √10 ≈ 3.16×)。这解释了为何 5 行分散在 3.16 与 10 两档。

    金属行的方向相反,则是另一套来源:低碳钢标签 μᵣ=100、μ-metal 标签 μᵣ=5×10⁴ 均取区间上端,而 δ 值对应 μᵣ≈10 与 μᵣ≈5.2×10⁵ —— 前者低于文献区间下限(50–100 之外),后者高于上限(10⁴–10⁵ 之外)。这说明金属行 δ 不是由表内 μᵣ 算出的,而是从厂商/材料手册的另一套数抄来的。

    三条独立证据合起来:这不是”一列填错位置”,而是「同一张表里三套数据源拼接,且未统一端点口径」。


    4. 表内自洽性检验:A(dB) 表与 δ 表是一致的

    把 §四 的 A(dB) 用同一张表的 δ 复算:

    场景 介质 z 表列 δ 表列 A(dB) 按表列 δ 复算 判定
    千米大山 花岗岩 1000 m 57 km 0.15 dB 0.152 ✅
    百米山包 潮湿花岗岩 100 m 18 km 0.05 dB 0.048 ✅
    千米山体 黏土页岩 1000 m 1.8 km 4.8 dB 4.83 ✅
    20 层楼 钢筋混凝土 60 m 2.5 km 0.21 dB 0.208 ✅
    1 cm 钢板 低碳钢 0.01 m 5.7 mm 15.2 dB 15.2 ✅
    1 mm μ-metal μ-metal 0.001 m 0.25 mm 34.7 dB 34.7 ✅

    → A(dB) 表完全忠实于 δ 表。所以错误不在换算,只在 §三 的 δ ↔ σ/μr 配对。这个定位很重要:修一处(δ 列或 σ 列),A 表自动跟着对;不需要逐格改 A 值。


    5. ⭐ 方向性后果:黏土/页岩行把「强屏蔽」写成了「弱衰减」

    这是本轮唯一会改变实质结论的发现。

    5.1 两种算法的差距

    介质 表列 δ@7.83 表列 A(1 km) 按标签 σ 的 δ 按标签 σ 的 A(1 km) 后果
    干燥花岗岩 57 km 0.15 dB 18.0 km 0.48 dB 同向,量级不变
    潮湿花岗岩 18 km 0.27 dB(估) 5.69 km 1.53 dB 同向,量级不变
    钢筋混凝土 2.5 km 0.21 dB(60 m) 254 m 2.05 dB(60 m) 同向,仍弱
    黏土/页岩 1.8 km 4.8 dB 180 m 48.3 dB ⛔ 方向性反转

    5.2 48 dB 意味着什么

    δ=180 m 时,1 km 厚度 = 5.56 个趋肤深度,振幅衰减到 e⁻⁵·⁵⁶ ≈ 0.38%。这不是「衰减 37%」,这是衰减 99.6%,等效约 48 dB,已是工程意义上的强屏蔽。

    而表 §四 把这一行写成「千米大山 / 黏土页岩 / 4.8 dB / 衰减 37%」,§五 又据此在「靠山=屏蔽」审查表里给出「山体 ≈ 0.05–0.15 dB,基本无屏蔽作用」的总结。

    5.3 稳健结论必须加边界

    当前表述(表 §五、下游多处引用):「靠山屏蔽地磁 ELF 在 1–30 Hz 段没有物理基础」。

    应修正为:

    「靠山屏蔽 ELF」的否定性结论,只对低电导结晶岩(σ ≤ 10⁻³ S/m,如干燥/一般花岗岩、玄武岩)成立。对高含水黏土、页岩、湿沉积层(σ ≈ 10⁻²–10⁰ S/m),百米级即可产生 dB 级衰减、千米级可达数十 dB——这条路径不能一概否定,反而值得作为”地形-岩性”分层假说单独立题。

    5.4 对下游的传导

    • #22 A 档门槛(σ≥10⁻² → 10 dB):用表列 δ 会得到「10⁻² 只衰减 4.83 dB,不达门槛」;按标签 σ 算是 48 dB,远超门槛。→ 门槛判定方向相反,A 档样本筛选逻辑需重定。
    • #23 铁律(A(dB) < 本底−10 dB 则禁入 E1):铁律本身没问题(它不依赖 δ 数值),但输入 δ 若取自该表,会把高 σ 岩体误判为”透明”,从而误放行。
    • #16 表的”千米山体几乎透明”:结论对花岗岩成立,对黏土质山体不成立。
    • #18 T4 / #24 阈值:不受影响(不依赖 δ 绝对值)。

    6. 更正建议(待授权稿,本轮不改正文)

    6.1 引文层(可直接改,属纠错)

    位置 现文 建议
    §二 脚注 Palacky & Kligerman (1980) Palacky (1988), DOI 10.1190/1.9781560802631.ch3(§3 Resistivity Characteristics of Geologic Targets, pp. 52–129)
    §二 脚注 Schön (2015) Physical Properties of Rocks 补章级:“Electrical Properties”, DOI 10.1016/b978-0-08-100404-3.00008-1
    §一/§七 未提 Madden / Eisel 建议补 10.1029/gm013p0469(地壳上地幔 σ 结构)、10.1046/j.1365-246x.1999.00707.x(KTB 宏各向异性),使 σ 溯源可查三层

    6.2 数值层(需 JOE 定 canonical,属 #40)

    两条互斥路线,不要混用:

    • 路线 A(保标签,重算 δ):σ/μr 保持现表(它们是文献区间取值,可与 §二 对齐),整列重算 δ 与 A。优点:物理依据明确、可追溯;缺点:A 档门槛、下游 12 文件全部要跟着改。
    • 路线 B(保 δ,改标签):δ 保持现值(下游自洽),把 σ 列改成”计算所用值”并注明取自区间哪一端。优点:改动面小;缺点:现表 δ 对应的某些 σ(如低碳钢 μᵣ≈10、μ-metal μᵣ≈5.2×10⁵)落在文献区间之外,改标签等于写入不实参数 → 不推荐。

    我的建议:走路线 A,并在每行显式声明端点口径(例如「取区间下限,代表最保守屏蔽估计」或「取上限,代表最强屏蔽场景」),从制度上消除”标签与数值各取一端”的空间。理由见 §6.3。

    6.3 为什么必须先定端点口径,而不是先定数值

    本轮最根本的教训是:这张表的 9 行里有 3 套数据源、2 种端点习惯、2 个方向的偏差。只要「端点口径」这条规则缺位,无论把数字改对多少次,下一版仍会漂。

    → 建议把「每行必须声明 σ 取区间哪一端」写进 delta-table-scan.py 的断言格式,作为 #37 的复发防线。


    7. 本轮没有做的事(守边界)

    • 未改任何既有正文:skin-depth-reference-table.md、elf-skin-depth-table-numeric-integrity-audit.md、以及受污染的 12 个文件,一字未动;
    • 未做 σ 实测:只核”标签与数值是否自洽”,未核”σ 的物理真值”。Palacky/Schön 的区间本身跨 2 个数量级,这才是最大的不确定性来源,比 δ 列的错配更根本;
    • 未替 JOE 定 canonical:路线 A/B 的取舍写在 §6.2,属红灯。

    8. 与上一轮的关系(supersedes / extends)

    维度 第 37 轮(#37) 本轮(#38)
    扫描范围 ✅ 838 文件 / 86 断言 / 59 MISMATCH 沿用,不复扫
    偏差机制 H-col「单一整列偏移,比值聚在 10 与 100」 ❌ 被否:9 行三个区段、两个方向
    覆盖行数 6 行(岩石/建材/海水) 9 行(补入低碳钢/纯铁/μ-metal)
    引文层 3 处作者错配 +1 处归属错(Palacky & Kligerman 1980 不可核)
    稳健结论 「山体透明」不变 ⚠️ 加边界:对高含水黏土/页岩不成立(48 dB vs 4.8 dB)
    交付物 扫描器 delta-table-scan.py 本审计 + 端点口径 SOP 提案

    #41(H-col 中间稿追查):本轮证据已足以否证 H-col,建议关闭,不再耗成本找中间稿。
    #38 状态:从「P3 登记未启动」→ ✅ 本轮完成参数层溯源。


    9. 相关阅读

    • 《趋肤深度对照表:基岩/建材/μ-metal 在 1–30 Hz ELF 段的 δ(ω)》(#16,被审对象)
    • 《池④ 趋肤深度表数值完整性审计》(第 36 轮,复算层)
    • 《δ/A 数值失真全量扫描与整列偏移机制》(第 37 轮,机制判定本轮修正)
    • 《δ 表量级过滤 SOP 推广》(#23,铁律上游)
    • 《靠山 ELF 屏蔽山体样本优先级重排》(#22,A 档门槛受本轮影响)

    本文为 draft,是参数层溯源审计 + 复算,不引入新的地球物理结论。
    不构成对具体山体/建材屏蔽性能的实测声明。
    更正建议需 JOE 授权后方可落地(对外可见材料)。
    进入 779 前需走 intake;进入 782 前需去 AI 化。

    证据与局限

    本文证据等级为 C,属 measurable,结论限于研究层面的推断,不构成对个体情形的判定。不同源」——这个机制判定不成立**。

    局限:本文未覆盖全部相关文献,效应量与结论随样本与方法变化;传统体系与现代研究的对照属结构类比而非等价证明,请结合专业判断参考。


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Pool 4 δ Table Parameter Layer Traceback Audit: σ/μr Tag and δ Value Endpoint Mismatch (#38)

    Evidence level: C

    One-Sentence Conclusion: The previous iteration attributed the δ table deviation to the “H-column shift hypothesis” (σ column and δ column originate from different sources) — this mechanism is invalid. A row-by-row backtracking reveals three distinct deviations (3.16× / 10× / 0.32× / 1.02×), none of which can be explained by a single constant. The true mechanism is that “δ values and σ/μr labels draw from different interval endpoints,” with the metal rows sourced from a separate manufacturer dataset. Directional Consequence: The δ for clay/shale rows is amplified 10-fold, rendering a scenario that should be 48 dB strong shielding as 4.8 dB weak attenuation → The robust conclusion that “mountains are nearly transparent to ELF” lacks defined applicability boundaries.


    1. Rejection of the H-Column Shift Hypothesis

    The hypothesis that a systematic column shift explains the discrepancies across all rows is refuted by empirical verification. If such a global offset existed, the ratio between calculated and observed values would remain constant across all material types. Instead, the variance indicates a structural inconsistency in data mapping rather than a simple translational error.

    2. Empirical Evidence of Multi-Modal Deviation

    Analysis of the nine rows in the dataset reveals three distinct deviation factors, confirming the absence of a unifying constant:

    Row Type Observed Deviation Factor Interpretation
    Clay/Shale 10× Significant overestimation of attenuation
    Metal 3.16× Moderate deviation linked to data source
    Others 0.32× / 1.02× Underestimation or near-match

    This pattern confirms that a single linear transformation cannot account for the data discrepancies.

    3. Root Cause: Endpoint Mismatch and Data Source Discrepancy

    The core issue lies in the Pattern and Timing (Qi) of data extraction. Specifically:
    * δ Values: Derived from one set of interval endpoints.
    * σ/μr Labels: Derived from a different set of interval endpoints.

    Furthermore, the metal rows utilize data from a distinct manufacturer database compared to the clay/shale rows. This structural analogy failure results in a misalignment of physical parameters, leading to the observed directional errors in shielding calculations.

    4. Implications for ELF Shielding Models

    The error propagates directly into the interpretation of electromagnetic shielding efficacy:
    * Correct Physics: Clay and shale materials provide approximately 48 dB of attenuation for ELF frequencies.
    * Current Model Error: Due to the 10× amplification error, the model predicts only 4.8 dB.
    * Consequence: This invalidates the generalization that “mountains are nearly transparent to ELF” without specifying material composition and frequency bands. The conclusion holds only within a specific, unverified boundary condition.

    5. Evidence Level and Limitations

    • Evidence Level: B (Pattern recognition from tabular data; direct calculation verification).
    • Limitations: Relies on the availability of raw manufacturer datasets for cross-validation. Without access to the specific source files, the exact endpoint definitions remain inferred rather than explicitly verified.

    1. Starting Point of This Round and Conclusion of the Previous Round

    The previous round (Round 37 / #37) completed a full scan of the δ/A table, yielding:

    1. 838 files scanned; 14 files contained assertions; 86 assertions / 27 OK / 59 MISMATCH;
    2. → Proposed the H-col hypothesis: “The σ column and δ column are not calculated by the same set of parameters; the δ column is calculated first, while the σ column is filled in independently afterward, with no cross-verification between the two columns,” and noted that the ratios are “highly clustered at 10.0 and ~100,” with the sea water row exactly at 1.0.

    The task of this round (#38) is parameter-level tracing: not asking “is δ correct,” but rather asking “where do these labels σ/μr originate, and do they align correctly?“

    Conclusion placed first: The scan scope is correct; the mechanistic explanation is wrong. The previous round examined only 6 rows (dry / moist / concrete / rebar / clay / sea water), all of which fell under the pattern where “δ used a smaller σ,” leading to an impression of a “single offset.” After adding the three metal rows, the direction immediately split.


    2. Citation Layer Verification: Three DOIs are authentic, but one citation in the table has a misattribution

    2.1 All three DOIs verified successfully via Crossref

    Anchor DOI Verified Identity via Crossref Verification Status
    A1 10.1190/1.9781560802631.ch3 Palacky G.J., “3. Resistivity Characteristics of Geologic Targets”, Electromagnetic Methods in Applied Geophysics, SEG, 1988, pp. 52–129 ✅ Authentic
    A2 10.1029/gm013p0469 Madden T.R. & Swift C.M., “Magnetotelluric Studies of the Electrical Conductivity Structure of the Crust and Upper Mantle”, Geophysical Monograph Series, AGU, 2013 ✅ Authentic
    A3 10.1046/j.1365-246x.1999.00707.x Eisel M. & Haak V., “Macro-anisotropy of the electrical conductivity of the crust: a magnetotelluric study of the German Continental Deep Drilling site (KTB)”, Geophys J Int, 1999 ✅ Authentic

    2.2 However, the textual citations in the table do not align with these anchors

    The footnote text in §2 of skin-depth-reference-table.md reads:

    Interval reference: Palacky & Kligerman (1980) for mineralization conductivity; Schön (2015) …

    Crossref searches for author Palacky between 1975–1995 return 15 results, but none are co-authored with Kligerman, nor is there any entry from 1980. Palacky’s seminal review in the field of conductivity is the 1988 solo-authored SEG chapter (i.e., A1). Therefore:

    • The verifiable record is Palacky’s 1988 solo-authored chapter;
    • The table states “Palacky & Kligerman 1980” — both the year and co-author are unverifiable.

    This is not a “fake PMID,” but rather Category 4 citation failure: verifiable literature presented as unverifiable year/co-author combination. The severity is lighter than a fake identifier (the domain remains correct), but the consequence is identical: readers cannot trace the source via this footnote.

    2.3 Schön (2015) is verifiable at the book level, but lacks chapter-level specificity

    Crossref identifies Schön’s 2015 Physical Properties of Rocks series chapters, where 10.1016/b978-0-08-100404-3.00008-1 corresponds to the “Electrical Properties” chapter — this is the true source for the $\sigma$ interval. The table currently lists only the book title without specifying the chapter. It is recommended to supplement with chapter-level citation details.

    2.4 All six PMIDs are authentic (consistent with previous rounds)

    PMID Verified Identity Appropriate Use in This Text
    20000094 Chernykh AM, Gig Sanit 2009 — Residential geomagnetic shielding Building shielding measurements
    28654641 Binhi VN, PLoS One 2017 — Review of low-field biological effects Background for $\mu$-metal shielding
    38132339 Sarimov RM, Biology 2023 — Review of biological effects under low-magnetic conditions Same as above
    36676419 Cheng Y, Materials 2023 — Testing methods for low-remanence materials $\mu$-metal material data
    24131394 Sirav B, Electromagn Biol Med 2014 — Transformer ELF magnetic fields Environmental ELF exposure
    36681118 Martel J, Biomed J 2023 — EMF and circadian rhythms Environmental ELF biological effects

    Note: In the previous round, PMID 24131394 was mislabeled as “Burchard,” 36681118 as “Girgert,” and 28654641 as “Lewczuk.” These author mismatches persist in the main text of elf-skin-depth-table-numeric-integrity-audit.md during this re-examination and constitute items requiring correction (not blocking this round).

    3. Parameter-Level Recalculation: Deviation Is Not Constant

    3.1 Methodology

    For each row in skin-depth-reference-table.md §3, substitute the σ and μᵣ values explicitly labeled in that row into the formula $\delta = \sqrt{2/(\omega\mu_0\mu_r\sigma)}$ (with $f = 7.83 \text{ Hz}$) and compare the calculated result against the listed $\delta$. Subsequently, reverse-calculate the implied $\sigma \cdot \mu_r$ product from the listed $\delta$ and compare it with the labeled $\sigma \cdot \mu_r$.

    3.2 Results (9 Rows, f = 7.83 Hz)

    Medium Labeled $\sigma$ Labeled $\mu_r$ Listed $\delta$ Calculated $\delta$ Ratio (Listed/Calc) Implied $\sigma\mu_r$ / Labeled $\sigma\mu_r$ Verdict
    Dry Granite $10^{-4}$ 1 57 km 18.0 km 3.17 0.0996 $\delta$ used $\sigma \approx 10^{-5}$
    Wet Granite $10^{-3}$ 1 18 km 5.69 km 3.16 0.0999 $\delta$ used $\sigma \approx 10^{-4}$
    Ordinary Concrete $10^{-2}$ 1 5.7 km 1.80 km 3.17 0.0996 $\delta$ used $\sigma \approx 10^{-3}$
    Reinforced Concrete $10^{-1}$ 5 2.5 km 254 m 9.83 0.0104 $\delta$ used $\sigma\mu_r \approx 10^{-3}$
    Clay/Shale $10^{0}$ 1 1.8 km 180 m 10.01 0.0100 $\delta$ used $\sigma \approx 10^{-2}$
    Seawater 4 1 90 m 89.9 m 1.00 0.999 ✅ Consistent
    Low-Carbon Steel $10^{6}$ 100 5.7 mm 18.0 mm 0.32 9.96 ⚠️ Inverse Direction ($\delta$ underestimated)
    Pure Iron $10^{7}$ 2000 1.3 mm 1.27 mm 1.02 0.957 ✅ Consistent
    $\mu$-metal $10^{6}$ $5 \times 10^{4}$ 0.25 mm 0.80 mm 0.31 10.35 ⚠️ Inverse Direction ($\delta$ underestimated)

    3.3 Three Concurrent Deviations → “Single Column Shift” Hypothesis Refuted

    • 5 Rows (Rock/Construction Materials): Ratio $\approx 3.16$ or $10$; labeled $\sigma$ values are 10–100 times larger than those actually used in the calculation.
    • 2 Rows (Low-Carbon Steel, $\mu$-metal): Ratio $\approx 0.31$, exhibiting an inverse direction; implied $\sigma\mu_r$ is 10 times larger than labeled values.
    • 2 Rows (Seawater, Pure Iron): Fully consistent.

    If a “single column shift” were true (an entire column filled with values from the wrong position), all 9 rows should yield the same multiplier. Instead, the values are 3.17 / 3.16 / 3.17 / 9.83 / 10.01 / 1.00 / 0.32 / 1.02 / 0.31, spanning three orders of magnitude with inconsistent signs.

    → The H-col hypothesis from the previous round (#41) is falsified in this iteration. The case can be closed without further chasing intermediate drafts.

    3.4 More Accurate Mechanism: Endpoint Mismatch + Mixed Data Sources

    Comparing the actual values used for the 5 rock/construction rows against the intervals declared in §2 reveals an immediate pattern:

    Medium §2 Declared Interval Actual $\sigma$ Corresponding to Listed $\delta$ Position Within Interval
    Dry Granite $10^{-6} – 10^{-4}$ $\approx 10^{-5}$ Logarithmic Midpoint
    Wet Granite $10^{-4} – 10^{-2}$ $\approx 10^{-4}$ Lower Bound
    Ordinary Concrete (incl. rebar) $10^{-3} – 10^{-1}$ $\approx 10^{-3}$ Lower Bound
    Clay/Shale $10^{-2} – 10^{0}$ $\approx 10^{-2}$ Lower Bound
    Reinforced Concrete $10^{-3} – 10^{-1}, \mu_r=1–10$ $\sigma\mu_r \approx 10^{-3}$ Lower Bound ($\sigma$ takes lower limit $\times \mu_r=1$)

    → The $\delta$ values are calculated using the interval’s lower bound or midpoint, whereas the $\sigma$ labels in §3 correspond to the upper bound or midpoint. Since the two columns utilize different endpoints, the ratio is not a fixed $10\times$ but depends on how many orders of magnitude the interval spans (spanning 2 orders $\rightarrow \sqrt{100} = 10\times$; spanning 1 order $\rightarrow \sqrt{10} \approx 3.16\times$). This explains why the 5 rows cluster around either 3.16 or 10.

    The inverse direction in the metal rows stems from a different source: both the low-carbon steel label ($\mu_r=100$) and $\mu$-metal label ($\mu_r=5 \times 10^{4}$) take the upper interval bound, while the $\delta$ values correspond to $\mu_r \approx 10$ and $\mu_r \approx 5.2 \times 10^{5}$ respectively. The former falls below the literature lower limit (outside 50–100), and the latter exceeds the upper limit (outside $10^{4}–10^{5}$). This indicates that the metal row $\delta$ values were not derived from the table’s internal $\mu_r$ but copied from a separate manufacturer/material handbook dataset.

    Synthesizing these three independent lines of evidence: This is not a “single column filled with misplaced values,” but rather “three distinct data sources spliced within the same table without a unified endpoint convention.”

    4. Internal Consistency Check of Tabular Data: Alignment Between the A(dB) Table and the δ Table

    Recalculate the values in §4’s A(dB) table using the same table’s δ values:

    Scenario Medium z Tabulated δ Tabulated A(dB) Recalculated via Tabulated δ Verdict
    Kilometer-scale mountain Granite 1000 m 57 km 0.15 dB 0.152 ✅
    Hundred-meter hill Moist granite 100 m 18 km 0.05 dB 0.048 ✅
    Kilometer-scale mountain body Clay shale 1000 m 1.8 km 4.8 dB 4.83 ✅
    20-story building Reinforced concrete 60 m 2.5 km 0.21 dB 0.208 ✅
    1 cm steel plate Low-carbon steel 0.01 m 5.7 mm 15.2 dB 15.2 ✅
    1 mm μ-metal μ-metal 0.001 m 0.25 mm 34.7 dB 34.7 ✅

    → The A(dB) table is fully consistent with the δ table. Therefore, the error lies not in the conversion but in the δ ↔ σ/μr pairing in §3. This localization is critical: correcting one column (either δ or σ) automatically aligns the A table; individual adjustments to A values are unnecessary.


    Glossary Note:
    * Yijing Studies: The academic discipline concerned with the I Ching, focusing on its philosophical, textual, and metaphysical frameworks rather than fortune-telling.

    5. ⭐ Directional Consequences: Clay/Shale Misrepresenting “Strong Shielding” as “Weak Attenuation”

    This is the only finding in this round that alters substantive conclusions.

    5.1 Discrepancy Between Two Algorithms

    Medium Listed $\delta@7.83$ Listed $A(1 \text{ km})$ $\delta$ by Tagged $\sigma$ $A(1 \text{ km})$ by Tagged $\sigma$ Consequence
    Dry Granite 57 km 0.15 dB 18.0 km 0.48 dB Same direction, magnitude unchanged
    Wet Granite 18 km 0.27 dB (est.) 5.69 km 1.53 dB Same direction, magnitude unchanged
    Reinforced Concrete 2.5 km 0.21 dB (60 m) 254 m 2.05 dB (60 m) Same direction, still weak
    Clay/Shale 1.8 km 4.8 dB 180 m 48.3 dB ⛔ Directional Reversal

    5.2 Implications of 48 dB

    When $\delta = 180 \text{ m}$, a thickness of $1 \text{ km}$ equals 5.56 skin depths, reducing amplitude to $e^{-5.56} \approx 0.38\%$. This is not “attenuation of 37%”; this is attenuation of 99.6%, equivalent to approximately 48 dB, constituting strong shielding in an engineering sense.

    However, Table §IV records this row as “kilometer-scale mountain / clay/shale / 4.8 dB / attenuation of 37%”. Consequently, §V summarizes this to conclude that “mountains ≈ 0.05–0.15 dB, essentially no shielding effect” within the “Mountain-as-Shield” review table.

    5.3 Robust Conclusions Require Boundary Conditions

    Current Statement (Table §V and downstream citations): “There is no physical basis for ‘mountain shielding’ of ELF geomagnetic fields in the 1–30 Hz band.”

    Should be corrected to:

    The negative conclusion regarding “mountain shielding of ELF” holds only for low-conductivity crystalline rocks ($\sigma \leq 10^{-3} \text{ S/m}$, such as dry or general granite, basalt). For high-moisture clays, shales, and wet sedimentary layers ($\sigma \approx 10^{-2}–10^{0} \text{ S/m}$), attenuation reaches the dB range at the百米 scale and tens of dB at kilometer scales—this pathway cannot be universally negated; rather, it merits independent thematic treatment as a “topography-lithology” stratified hypothesis.

    5.4 Transmission to Downstream Applications

    • #22 A-Class Threshold ($\sigma \geq 10^{-2} \rightarrow 10 \text{ dB}$): Using the listed $\delta$ yields “only 4.83 dB attenuation at $10^{-2}$, failing the threshold”; however, based on tagged $\sigma$, it calculates to 48 dB, far exceeding the threshold. → The threshold determination is reversed in direction, requiring a redefinition of the A-class sample screening logic.
    • #23 Iron Rule ($A(\text{dB}) < \text{Background} – 10 \text{ dB} \rightarrow$ Prohibit Entry E1): The rule itself remains valid (it does not depend on $\delta$ values), but if $\delta$ is sourced from this table, high-$\sigma$ rock masses will be erroneously classified as “transparent,” leading to false clearance.
    • #16 Table’s “Kilometer-scale Mountains Are Almost Transparent”: This conclusion holds for granite but does not hold for clay-rich mountains.
    • #18 T4 / #24 Thresholds: Unaffected (do not depend on absolute $\delta$ values).

    6. Corrective Recommendations (Draft pending authorization; do not modify main text in this round)

    6.1 Citation Layer (Directly editable; constitutes error correction)

    Location Current Text Suggestion
    Footnote §2 Palacky & Kligerman (1980) Palacky (1988), DOI 10.1190/1.9781560802631.ch3 (§3 Resistivity Characteristics of Geologic Targets, pp. 52–129)
    Footnote §2 Schön (2015) Physical Properties of Rocks Add chapter-level: “Electrical Properties”, DOI 10.1016/b978-0-08-100404-3.00008-1
    §1 / §7 Madden / Eisel not mentioned Suggest adding 10.1029/gm013p0469 (σ structure in crust-mantle transition) and 10.1046/j.1365-246x.1999.00707.x (KTB macro-anisotropy) to ensure σ attribution covers three layers

    6.2 Numerical Layer (Requires JOE to define canonical; classified as #40)

    Two mutually exclusive paths; do not mix them:

    • Path A (Preserve labels, recalculate δ): Maintain current tabulated values for σ/μr (they represent literature interval ranges and can align with §2); recalculate the entire column for δ and A. Pros: Clear physical basis, traceable. Cons: Raises the threshold for Path A and requires updating all downstream 12 files.
    • Path B (Preserve δ, modify labels): Maintain current δ values (ensuring downstream consistency); change the σ column to “values used in calculation” and specify which interval endpoint was sourced. Pros: Smaller scope of change. Cons: Certain σ values corresponding to current δ (e.g., low-carbon steel μᵣ≈10, μ-metal μᵣ≈5.2×10⁵) fall outside the literature intervals; changing labels would imply writing in false parameters → Not recommended.

    My recommendation: Proceed with Path A, and explicitly declare the endpoint convention for each row (e.g., “take lower interval bound, representing the most conservative shielding estimate” or “take upper bound, representing the strongest shielding scenario”) to eliminate institutionally the possibility of “taking one endpoint for labels and another for values.” Rationale is detailed in §6.3.

    6.3 Why Endpoint Convention Must Be Defined Before Numerical Values

    The fundamental lesson from this round is that among the 9 rows in this table, there are 3 distinct data sources, 2 different endpoint conventions, and 2 directional biases. As long as the rule regarding “endpoint convention” is absent, no matter how many times the numbers are corrected, the next version will still drift.

    → Recommendation: Incorporate a requirement into the assertion format of delta-table-scan.py stating that “each row must declare which interval endpoint σ takes”, serving as a recurrence defense for #37.

    7. Actions Omitted in This Round (Boundary Enforcement)

    • No modifications to existing core text: The files skin-depth-reference-table.md, elf-skin-depth-table-numeric-integrity-audit.md, and the 12 affected documents remain untouched, word for word;
    • No direct σ validation: Only verified the consistency between “labels and numerical values,” not the physical truth of σ itself. The Palacky/Schön intervals span two orders of magnitude; this represents the primary source of uncertainty, outweighing mismatches in the δ column;
    • No canonical assignment for JOE: The decision between Route A and Route B is documented in §6.2 and currently stands as a red light.

    8. Relationship with the Previous Iteration (Supersedes / Extends)

    Dimension Round #37 Current Round (#38)
    Scan Scope ✅ 838 files / 86 assertions / 59 MISMATCH Retained; no re-scanning
    Deviation Mechanism H-col “Single column shift, ratios clustered at 10 and 100” ❌ Rejected: 9 rows across three sections and two directions
    Covered Rows 6 rows (rock/construction materials/seawater) 9 rows (added low-carbon steel/pure iron/μ-metal)
    Citation Layer 3 author mismatches +1 attribution error (Palacky & Kligerman 1980 unverifiable)
    Robust Conclusion “Transparency of mountain” remains unchanged ⚠️ Added boundary: Not valid for high-moisture clays/shale (48 dB vs 4.8 dB)
    Deliverable Scanner delta-table-scan.py This audit + endpoint scope SOP proposal

    #41 (H-col intermediate draft tracking): Current evidence is sufficient to refute H-col; recommend closure to avoid further expenditure on locating intermediate drafts.
    #38 Status: Transitioned from “P3 registration not initiated” → ✅ Parameter-level tracing completed in this round.


    Note: The term “H-col” refers to a specific column identifier within the data structure used for tracking hypothesis deviations.

    9. Related Reading

    • Skin Depth Reference Table: Basalt/Construction Materials/μ-metal in the 1–30 Hz ELF Band δ(ω) (#16, Under Review)
    • Pool ④ Skin Depth Table Numerical Integrity Audit (Round 36, Recalculation Layer)
    • Full-Scale Scan of δ/A Numerical Distortion and Column Offset Mechanisms (Round 37, Mechanism Determination Round Correction)
    • δ Table Magnitude Filtering SOP Dissemination (#23, Upstream Iron Rule)
    • Priority Reordering of Backing ELF Shielding Mountain Samples (#22, A-tier Threshold Affected by This Round)

    This document is a draft serving as a parameter-layer traceability audit and recalculation; it does not introduce new geophysical conclusions.
    It does not constitute a measured declaration regarding the specific shielding performance of certain mountains or construction materials.
    Correction suggestions may only be implemented upon authorization by JOE (for externally visible materials).
    Entry into 779 requires intake processing; entry into 782 requires de-AI-ization.


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 去趋势的三条岔路:同一个中子通量与病毒突变率,三种洗法能洗出三个结论

    Evidence level: B

    上一篇已经把”天时改疫”翻译成一道可算的统计题。这一篇要说的是:这道题在动手计算之前,还横着一个不起眼的岔路口——你怎么把趋势洗掉,决定你会拿到什么答案。

    正文

    上一篇(neutron-flux-virus-mutation-lag-cross-correlation)把”岁运影响疫变”从”拿疫情年份去对太阳极小期”改写成了”时序对时序”:中子监测站的通量序列,对病毒系统发育的 root-to-tip(RTT)突变速率序列,做滞后互相关。设计是对的,但它内部埋了一个动词——去趋势。而这一个动词,是整条链路上最容易让结论掉头的地方。

    为什么非洗不可。 两条真实的地球/生物序列,几乎必然都带趋势。中子通量侧随太阳活动周期涨落,突变率侧则有”时间依赖速率现象”(TDRP):速率估计随观测时间尺度系统性衰减,机制被建模为”最快进化位点饱和 + 纯化选择”的组合(PMID 34478645)。再叠加 RNA 病毒本就存在的谱系间速率变异(PMID 11821909 对 50 种 RNA 病毒的定量系统发育分析显示严格分子钟常被拒绝),”突变率序列”本身就是一个带强趋势、带低频漂移的量。两条各自带趋势的序列做相关,即使物理上毫无关系,也能因共享低频漂移获得虚假的正相关——这在气候检测与归因里已经被系统论证过(PMID 35345504)。更麻烦的是”趋势”这个词本身不唯一:对非线性非平稳序列,趋势的定义取决于你用哪种自适应分解(PMID 17846430),而温度类序列的长程持续性又会让”去掉趋势”这个动作本身产生方法依赖(PMID 22181223)。所以去趋势不是清洁步骤,而是建模选择。

    三条岔路,是三台不同的机器。 一阶差分(Δxₜ = xₜ − xₜ₋₁)是最常见的做法,它的传递函数是 1 − z⁻¹——一台在低频增益趋零、在奈奎斯特频率增益为 2 的高通滤波器。它的副作用是双重的:既把低频真效应一起清零,又会在白噪声上引入约 −0.5 的负自相关。低通滤波减去低通成分(或滑动平均去趋势)表面上收窄了频段,但有限冲激响应带来的边缘效应和相位失真会跟着进残差。谐波回归走的是第三条路:把趋势项和季节谐波一起放进设计矩阵做最小二乘,再从残差里取信号——它相当于在频域定点扣除若干频率,既不做全频段高通,也不引入差分式负自相关,但代价是”扣几个谐波、趋势用几次多项式”变成一串人为的自由度选择。三种做法的共同反讽是:它们都声称”去掉了趋势”,却给出了三份互不相同的残差序列,而残差的自相关结构,正是决定下一步互相关显著性的唯一输入(PMID 39815636 对随机/确定性趋势处理的系统讨论)。

    洗完之后,显著性还得重算。 互相关的 p 值绝不能按样本量 n 直接查表。序列的残余自相关会把独立观测的”有效数量”压到远低于 n,此时名义 5% 的检验可能实际有 30% 以上的假阳性——这正是单例研究与过程控制数据里反复被证实的推断陷阱(PMID 26463724、PMID 23705957)。正确做法有三条并行的路:一是有效自由度校正,用序列自身的谱密度换算出等价的独立观测数,再据此构造相关系数的置信区间(PMID 39846001 给出了适用于生理信号的参数化方案);二是预白化(prewhitening),先按 ARIMA 结构把序列白化再算互相关——这个方法已经在真实流行病学问题上落地,比如韩国自杀相关网络搜索与自杀率的时间关系检验就明确采用了预白化互相关(PMID 41662218);三是重采样/替代数据,在彩色噪声的框架下用时间域与波域的重采样构造零分布(PMID 11169871),对功率律互相关过程也有专门的检验(PMID 22304166)。此外,当序列存在长程记忆时,”趋势是否显著”本身就要用考虑记忆的显著性框架来判(PMID 37271384)。

    最难的是把话说对。 三种洗法跑完,如果结论一致,说明”通道”的信号强到能扛过方法选择——这是最理想的结局。如果不一致,则眼下的结论其实是方法的函数,而不是自然的函数,此时任何单点显著都不该被当成证据。还有两个方向相反的表述陷阱必须避免:去趋势后不显著,不等于通道不存在(可能是真阴性,也可能是方法把真效应一起切掉了);去趋势后显著,也不等于通道存在(可能只是残余自相关下的假自由度)。上一篇引过的反证在这里尤其重要:2008–2009 与 2019–2020 两次太阳深极小期中子通量同样偏高,却并未对应同等的病毒登场节奏(PMID 35194566)。

    对易学的意义。 这一篇没有一个新假说,它只做一件事:把”天时改疫”能不能被检验这件事,与”检验者怎么选方法”彻底分开。做法就是把三种去趋势算子在同一份中子通量序列、同一份突变率序列上并行跑,把三套相关系数、三套置信区间、三套显著性并排列出来,并预先登记方法选择与判定规则。若三套一致指向零,那”岁运直接改写病毒基因组”这条通道基本可以放下;若三套都指向非零且方向一致,它才值得被认真追查。五运六气的科学价值从来不在于”某年对上了”,而在于”它敢被放在一套会失败的检验里”——去趋势方法学正是这套检验里最容易被糊过去、也最必须写清楚的一环。

    证据与局限

    已验证: 病毒替换速率存在时间依赖衰减(TDRP),机制为位点饱和与纯化选择的组合(PMID 34478645);RNA 病毒进化速率存在显著谱系间变异,严格分子钟常被拒绝(PMID 11821909);带趋势的非平稳序列做相关/回归可产生伪结果(PMID 35345504);趋势与去趋势的定义依赖方法选择(PMID 17846430);自相关会压缩有效自由度、膨胀 I 类错误(PMID 26463724, PMID 23705957);有效自由度校正(PMID 39846001)、预白化互相关(PMID 41662218)、时间/波域重采样(PMID 11169871)、功率律互相关检验(PMID 22304166)、多尺度去趋势互相关(PMID 39605794)均为成熟可用工具。

    部分验证: 预白化互相关在流行病学与公共卫生监测中已有实际应用案例(PMID 41662218),但尚未成为病毒进化-天体指数类研究的常规步骤;长程记忆下的趋势显著性框架(PMID 37271384)验证于大气污染物趋势,尚未在病毒速率序列上标定。

    待验证: 同一中子通量—突变率序列在一阶差分、低通滤波去趋势、谐波回归三种算子下的相关系数与显著性漂移幅度;去趋势后残差自相关结构(及其对应的有效自由度)对互相关显著性阈值的影响量级;三种算子下结论一致性是否构成”通道存在”的充分判据。

    相关阅读

    证据等级: B(TDRP 机制、RNA 病毒速率变异、伪相关与去趋势方法学、有效自由度/预白化/替代数据工具均有 PubMed 直接支撑;”三种去趋势算子导致互相关显著性漂移”这一具体命题尚无直接实证,属方法学类比与间接证据链)
    参考文献: PMID 34478645, PMID 11821909, PMID 35345504, PMID 17846430, PMID 22181223, PMID 39815636, PMID 39846001, PMID 41662218, PMID 11169871, PMID 22304166, PMID 39605794, PMID 37271384, PMID 26463724, PMID 23705957, PMID 35194566


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Three Paths to De-trending: Same Neutron Flux and Virus Mutation Rate Yield Three Different Conclusions via Three Washing Methods

    Evidence level: B

    The previous article translated “heavenly timing alters epidemics” into a calculable statistical problem. This piece addresses the fork encountered before calculation begins: how you choose to de-trend determines which answer you receive.

    Main Text

    The previous article (neutron-flux-virus-mutation-lag-cross-correlation) reframed “solar cycle influences epidemic shifts” from matching epidemic years to solar minima into a time-series-to-time-series analysis: cross-correlating neutron monitor flux sequences with root-to-tip (RTT) mutation rate sequences from viral phylogenies. The design is sound, but it contains an embedded verb—de-trending. This single verb is the most critical point where conclusions can diverge along the entire chain.

    Why de-trending is non-negotiable. Two real Earth/biological sequences almost certainly carry trends. On the neutron flux side, fluctuations follow solar activity cycles. On the mutation rate side, there is a “Time-Dependent Rate Phenomenon” (TDRP): rate estimates systematically decay with observation time scales, modeled as a combination of “fastest-evolving site saturation + purifying selection” (PMID 34478645). Adding intrinsic rate variation across lineages in RNA viruses (PMID 11821909 shows strict molecular clocks are frequently rejected in quantitative phylogenetic analyses of 50 RNA viruses), the “mutation rate sequence” itself is a quantity with strong trends and low-frequency drift. Correlating two trend-bearing sequences can yield spurious positive correlation due to shared low-frequency drift even if they are physically unrelated—a phenomenon systematically demonstrated in climate detection and attribution (PMID 35345504). Complicating matters further, the definition of “trend” is not unique for nonlinear non-stationary sequences; it depends on the adaptive decomposition method chosen (PMID 17846430), and long-range persistence in temperature-like sequences means the act of “removing trend” itself introduces method dependence (PMID 22181223). Thus, de-trending is not a cleaning step but a modeling choice.

    Three paths correspond to three different machines. First-order differencing (Δxₜ = xₜ − xₜ₋₁) is the most common approach; its transfer function is 1 − z⁻¹—a high-pass filter with zero gain at low frequencies and a gain of 2 at the Nyquist frequency. Its side effects are dual: it zeroes out low-frequency true effects while introducing approximately −0.5 negative autocorrelation on white noise. Subtracting low-pass components (or using sliding averages) narrows the band superficially, but finite impulse response introduces edge effects and phase distortion into the residuals. Harmonic regression takes a third path: including trend terms and seasonal harmonics in the design matrix for least squares, then extracting signals from residuals. This effectively subtracts specific frequencies in the frequency domain without full-band high-pass filtering or introducing differential-based negative autocorrelation, but the cost is that “which harmonics to subtract and how many polynomial terms for trend” becomes a series of arbitrary degree-of-freedom choices. The common irony is that all three claim to “remove trend” yet produce three distinct residual sequences, and the autocorrelation structure of these residuals is the sole input determining the significance of the subsequent cross-correlation (PMID 39815636 offers a systematic discussion on stochastic vs. deterministic trend handling).

    Recalculate significance after cleaning. Cross-correlation p-values cannot be looked up directly using sample size $n$. Residual autocorrelation compresses the “effective number of independent observations” far below $n$; at this point, a nominal 5% test may yield over 30% false positives—a trap repeatedly confirmed in single-case studies and process control data (PMID 26463724, PMID 23705957). Correct approaches include three parallel paths: First, effective degrees of freedom correction, converting the sequence’s spectral density into an equivalent number of independent observations to construct confidence intervals for the correlation coefficient (PMID 39846001 provides a parameterized scheme suitable for physiological signals). Second, pre-whitening, where sequences are whitened via ARIMA structure before calculating cross-correlation—this method has already been applied to real epidemiological issues, such as the time relationship test between suicide-related network searches and suicide rates in Korea explicitly using pre-whitened cross-correlation (PMID 41662218). Third, resampling/alternative data, constructing null distributions via time-domain and wavelet-domain resampling under a colored noise framework (PMID 11169871), with specific tests available for power-law cross-correlation processes (PMID 22304166). Furthermore, when sequences exhibit long-range memory, the significance of “trend” itself must be judged using a significance framework accounting for memory (PMID 37271384).

    The hardest part is speaking correctly. After running all three cleaning methods, if conclusions align, it indicates the “channel” signal is strong enough to withstand methodological choices—the ideal outcome. If they do not align, the current conclusion is actually a function of the method, not nature; thus, no single-point significance should be treated as evidence. Two opposing rhetorical traps must also be avoided: non-significance after de-trending does not mean the channel is absent (could be true negative or method cutting out true effects); significance after de-trending does not mean the channel exists (could be false degrees of freedom under residual autocorrelation). The counter-evidence cited previously is particularly important here: neutron flux was similarly elevated during the 2008–2009 and 2019–2020 solar grand minima, yet did not correspond to equivalent viral emergence rhythms (PMID 35194566).

    Significance for Yijing Studies. This article proposes no new hypothesis; it performs one task: completely separates whether “heavenly timing alters epidemics” can be tested from how the “tester selects a method.” The procedure involves running all three de-trending operators in parallel on the same neutron flux sequence and the same mutation rate sequence, listing three sets of correlation coefficients, three sets of confidence intervals, and three sets of significance levels in advance, while registering method selection and decision rules. If all three point to zero, the channel of “solar cycles directly rewriting viral genomes” can be largely set aside. If all three point to non-zero with consistent direction, it warrants serious investigation. The scientific value of the Five Movements and Six Qi (Wu Yun Liu Qi) never lies in “matching a specific year,” but in its willingness to be placed within a suite of tests prone to failure—de-trending methodology is precisely the most easily overlooked yet must-be-clarified link in this testing suite.

    Evidence and Limitations

    Verified: Time-dependent decay exists in virus substitution rates (TDRP), mechanism being a combination of site saturation and purifying selection (PMID 34478645); RNA virus evolutionary rates exhibit significant lineage-to-lineage variation, strict molecular clocks frequently rejected (PMID 11821909); correlating/regressing non-stationary sequences with trends produces spurious results (PMID 35345504); definitions of trend and de-trending depend on methodological choices (PMID 17846430); autocorrelation compresses effective degrees of freedom and inflates Type I errors (PMID 26463724, PMID 23705957); effective degrees of freedom correction (PMID 39846001), pre-whitened cross-correlation (PMID 41662218), time/wavelet domain resampling (PMID 11169871), power-law cross-correlation tests (PMID 22304166), and multi-scale de-trending cross-correlation (PMID 39605794) are mature, available tools.

    Partially Verified: Pre-whitened cross-correlation has practical application cases in epidemiology and public health monitoring (PMID 41662218), but has not yet become a routine step in virus evolution-astronomical index studies; trend significance frameworks under long-range memory (PMID 37271384) are validated for atmospheric pollutant trends but not yet calibrated on virus rate sequences.

    To be Verified: Magnitude of correlation coefficient and significance drift across first-order differencing, low-pass filtering de-trending, and harmonic regression operators on the same neutron flux-mutation rate sequence; magnitude of impact of residual autocorrelation structure (and corresponding effective degrees of freedom) on cross-correlation significance thresholds after de-trending; whether consistency across three operators constitutes a sufficient criterion for “channel existence.”

    Related Reading

    Evidence Level: B (TDRP mechanisms, RNA virus rate variation, spurious correlation and de-trending methodology, effective degrees of freedom/pre-whitening/alternative data tools all have direct PubMed support; the specific proposition “three de-trending operators cause cross-correlation significance drift” lacks direct empirical evidence, belonging to methodological analogy and indirect evidence chains)
    References: PMID 34478645, PMID 11821909, PMID 35345504, PMID 17846430, PMID 22181223, PMID 39815636, PMID 39846001, PMID 41662218, PMID 11169871, PMID 22304166, PMID 39605794, PMID 37271384, PMID 26463724, PMID 23705957, PMID 35194566


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 同一个风险基因,为什么在夜班下是保护、在月光下未必

    Evidence level: B

    MTNR1B 的风险等位在人造光照(夜班)下是缓冲器,但人造光和月光是两片”不相似的环境”——按基因×环境交互的普遍规律,方向未必跟过去。

    正文

    前一篇问的是”基因型×月相”这个交互有没有。这一篇问更难的一层:如果真有,方向是哪一边。

    一、G 等位到底是好是坏,已经开始分裂

    MTNR1B 的 rs10830963(G 等位)编码褪黑素 MT2 受体上一个常见变异,生理后果很具体:G 携带者的褪黑素升高时长比非携带者多 41 分钟,褪黑素终止相位平均延后 1.37 小时(PMID 26868293)——夜里褪黑素信号更长、更晚才关掉。

    那它是好是坏?在代谢线上是明确的”坏”:337,083 人的 UK Biobank 分析里,CG 携带者 2 型糖尿病风险 1.10 倍、GG 1.21 倍(PMID 31623012)。但在光照扰动这条线上,它突然变成了”好”的。

    二、夜班证据:G 是缓冲器

    这是本文最重要的对照。UK Biobank 242,194 人、12 年随访的卒中分析发现:夜班只在 CC 纯合子里把卒中风险推到 HR 1.23,GC/GG 携带者完全没有这个风险,交互 P=0.010(PMID 35411403)。同一人群另一结局——前列腺癌,133,416 名男性,交互 P=0.009,夜班者里 GC 对 CC 是 HR 0.69、GG 对 CC 是 HR 0.33(PMID 35064058)。两个独立结局、同一方向:G 等位在人造光照扰动下是缓冲器,不是放大器。作者的解释也顺:夜班把外在明暗循环整个翻了过来,而 G 携带者的内源时钟本来就偏晚,外部反转反而与内部节奏对上了。

    三、可方向并不一致

    换个分析、换个扰动量,方向会翻。216,702 人的 UK Biobank 报告(PMID 35467761)显示 rs10830963 与乳腺癌、前列腺癌均不独立相关(Ptrend = 0.103 / 0.281),但一分层,晚型女性的乳腺癌风险升高只出现在 G 等位携带者里(Ptrend = 0.015)——这里 G 是放大器。另一个 337,083 人分析(PMID 31623012)里,”晚型→风险升高”只出现在 CC 与 CG 组,GG 组消失;而最初那篇生理学论文(PMID 26868293)发现糖尿病风险升高在早起的人里更明显。更关键的是有一份大样本直接说”没有交互”:189,488 人(PMID 31757795)报告 rs10830963 与 2 型糖尿病的关联完全没被轮班工作修改,晨昏偏好那项只有边际证据,且扩展到全体受试者、改用加速度计客观睡眠中点后就不成立。

    于是现实是:同一位点、同一人群、同一框架,方向随结局、随扰动量定义、随分层方式而变。

    四、为什么方向可能反转:酵母给的骨架

    这不是人类遗传学独有的混乱。Wei 与 Zhang 用 1005 个酵母分离体、47 种环境做全基因组 g×eQTL 扫描(PMID 27903611),得到两条规律:① 多数交互位点在两种环境里方向一致,但某环境中效应越大,另一环境出现”拮抗”(方向相反)的概率越高;② 拮抗位点显著富集在性质不相似的环境中。更早的酵母表达研究(PMID 18416601)补上机制:顺式变异跨环境方向稳定,反式变异高度依赖环境、经常在两种条件下朝向相反;实验若不控制条件,这些相反效应会互相抵消,最后显示”没关系”。

    搬到人和月亮上就清楚了:月相(0.1-0.3 lux,天然、微弱、慢性)与夜班(人造、强烈、急性)是两片不相似的环境。G 等位在夜班下的缓冲,没有理由被直接搬给月相——按酵母的规律,环境越不相似,方向反转概率越高,”翻转”和”没交互”都是合理结局。

    五、难点不是假设,是功效

    Duncan 与 Keller 审查精神病遗传学基因×环境交互研究的最初十年、103 项研究(PMID 21890791):新发现 96% 阳性,复制尝试只有 27% 阳性;按观测样本量算功效,即便假设交互效应量大于已知的任何主效应,多数研究仍功效不足;按功效与先验估计的假发现率超过 97%。根子在数学上很朴素——交互项是”基因×环境”的乘积,在非实验人群里方差天然很小,方差小就检不出。作者还纠正了一个流行误解:乘积项与主效应项的相关性不影响交互检验的功效。

    这对”月相×MTNR1B”是硬约束。月相主效应本就只剩十几分钟量级,交互项只会更小,基因型又无法按极端分组抽样。要检验方向,样本量与记录时长必须按”方向可检出”而非按主效应设计,且方向假设必须预注册——否则看到什么都能解释成翻转。同时个体光敏感度差异极大:55 人的个体化剂量-反应显示 50% 抑制有效剂量从 6 lux 到 350 lux、横跨一个数量级,变异系数 26%(PMID 31138694);褪黑素抑制起效阈值约 1.5 melanopic lux、饱和在 305(PMID 30697806)。月照确实落在低剂量敏感段——这既是”可能有作用”的理由,也是”极难稳定检出”的理由。

    六、对易学的意义

    命理里有个基本规矩:同一个字,在不同的运里角色可以完全反过来——”用神”随大运流年而变,不存在天生就好的字。G 等位在这里扮演的正是这个角色:在人造光照下是缓冲器,在天然光照下该是缓冲还是放大,没有人测过。而酵母的规律说明这不是含糊其辞,而是有具体预测的:扰动源越不相似,方向反转概率越高。

    这也正是”天人相应”最容易被讲错的地方。它讲的不是”天好你就好”,而是同一个天、不同的人、接法不同。月亮对所有人一样圆,但”顺”与”逆”是相对于各自的节奏说的——所以正确的问法从来不是”月亮有害吗”,而是”对我这条作息线,月亮是推力还是阻力”。

    证据与局限

    已验证: MTNR1B rs10830963 G 等位独立调节褪黑素分泌动力学(+41 分钟、终止相位延迟 1.37 小时,PMID 26868293)并升高 2 型糖尿病风险(PMID 31623012);夜班×rs10830963 交互在两个独立结局中方向一致地显示 G 等位缓冲(卒中 P=0.010、PMID 35411403;前列腺癌 P=0.009、PMID 35064058);分层分析中出现方向相反的证据(乳腺癌晚型 Ptrend=0.015、PMID 35467761);同一交互在一份 189,488 人样本中未复现(PMID 31757795);酵母 g×eQTL 的拮抗方向富集于不相似环境(PMID 27903611)、反式变异跨条件方向常相反且可相互抵消(PMID 18416601);cG×E 首十年 96% 新发现阳性对 27% 复制阳性、假发现率 >97%(PMID 21890791);个体光敏感度 ED50 跨 6-350 lux(PMID 31138694);褪黑素抑制起效约 1.5 melanopic lux(PMID 30697806)。

    部分验证: “月相→ipRGC→SCN→松果体→褪黑素→MT2→睡眠”链条逐环有证据,但整链在真实月照(0.1-0.3 lux)下能否驱动可测的褪黑素变化缺少直接测量;月照落于低剂量敏感段属换算推断,非实测抑制量。

    待验证: 月相×MTNR1B 交互的方向——G 等位在天然光照下是放大还是缓冲,无任何研究直接检验;该方向是否与夜班下的缓冲方向相反;交互方向是否随扰动源强度与光谱比例系统性变化;在按方向检出设计的样本量下,交互能效是否足够。

    相关阅读

    证据等级: B
    参考文献: PMID 26868293, PMID 31623012, PMID 35411403, PMID 35064058, PMID 35467761, PMID 31757795, PMID 27903611, PMID 18416601, PMID 21890791, PMID 31138694, PMID 30697806


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Why Does the Same Risk Gene Act as a Protector Under Night Shifts but Not Under Moonlight?

    Evidence level: B

    The risk allele of MTNR1B acts as a buffer under artificial light (night shifts), yet artificial light and moonlight represent two “dissimilar environments.” According to the general law of gene-by-environment interactions, the direction need not mirror the past.

    Main Text

    The previous inquiry asked whether a “genotype × lunar phase” interaction exists. This piece addresses the more difficult layer: if it exists, which direction does it take?

    I. The “Good” or “Bad” Nature of the G Allele Is Already Splitting

    The rs10830963 (G allele) of MTNR1B encodes a common variant at the melatonin MT2 receptor. The physiological consequence is specific: G carriers experience a 41-minute longer duration of elevated melatonin and an average delay of 1.37 hours in the termination phase compared to non-carriers (PMID 26868293)—meaning the nocturnal melatonin signal is longer and turns off later.

    So, is it good or bad? On the metabolic line, it is clearly “bad”: in a UK Biobank analysis of 337,083 individuals, CG carriers had a 1.10-fold risk and GG carriers a 1.21-fold risk for type 2 diabetes (PMID 31623012). However, on the line of light perturbation, it suddenly becomes “good.”

    II. Night Shift Evidence: G Acts as a Buffer

    This is the most critical comparison in this text. A stroke analysis of 242,194 individuals from the UK Biobank with 12 years of follow-up found that night shifts only elevated stroke risk to HR 1.23 in CC homozygotes, while GC/GG carriers showed no such risk, with an interaction P=0.010 (PMID 35411403). In the same population for a different outcome—prostate cancer, in 133,416 males—the interaction P=0.009; among night shift workers, GC showed HR 0.69 relative to CC and GG showed HR 0.33 relative to CC (PMID 35064058). Two independent outcomes, same direction: the G allele acts as a buffer under artificial light perturbation, not an amplifier. The authors’ explanation aligns: night shifts flip the external light-dark cycle entirely, while G carriers already have an endogenous clock biased late; the external reversal thus aligns with their internal rhythm.

    III. Direction Is Not Consistent Across Analyses

    Change the analysis, change the perturbation magnitude, and the direction flips. A UK Biobank report of 216,702 individuals (PMID 35467761) showed that rs10830963 was not independently associated with breast cancer or prostate cancer (Ptrend = 0.103 / 0.281), but upon stratification, breast cancer risk elevation in late-type women appeared only in G allele carriers (Ptrend = 0.015)—here, G acts as an amplifier. In another analysis of 337,083 individuals (PMID 31623012), “late-type → risk elevation” appeared only in CC and CG groups, disappearing in the GG group; conversely, the initial physiological paper (PMID 26868293) found that diabetes risk elevation was more pronounced in early risers. More critically, one large sample directly states “no interaction”: an analysis of 189,488 individuals (PMID 31757795) reported that the association between rs10830963 and type 2 diabetes was not modified by shift work, with circadian preference showing only marginal evidence, and the finding failed when extended to all subjects or when using accelerometers for objective sleep timing.

    Thus, reality is: at the same locus, same population, same framework, the direction varies by outcome, by definition of perturbation magnitude, and by stratification method.

    IV. Why Direction Might Reverse: A Skeleton from Yeast

    This is not chaos unique to human genetics. Wei and Zhang used 1,005 yeast isolates and 47 environmental conditions for a whole-genome g×eQTL scan (PMID 27903611), deriving two rules: ① Most interaction loci show consistent direction across environments, but the probability of “antagonism” (opposite direction) increases in the other environment where the effect is larger; ② Antagonistic loci are significantly enriched in dissimilar environments. Earlier yeast expression studies (PMID 18416601) supplemented the mechanism: cis-variants show stable direction across environments, while trans-variants are highly dependent on the environment and often orient oppositely under two conditions; if experimental conditions are not controlled, these opposing effects cancel each other out, finally displaying “no relationship.”

    Applying this to humans and the moon clarifies the situation: lunar phase (0.1–0.3 lux, natural, weak, chronic) and night shifts (artificial, strong, acute) are two dissimilar environments. The buffering effect of the G allele under night shifts has no reason to be directly transposed to lunar phase—according to yeast rules, the less similar the environments, the higher the probability of direction reversal; both “flip” and “no interaction” are reasonable outcomes.

    V. The Difficulty Is Not Assumption, But Power

    Duncan and Keller reviewed the first decade of gene-by-environment interaction studies in psychiatric genetics, covering 103 studies (PMID 21890791): new discoveries showed 96% positivity, while replication attempts yielded only 27% positivity; by observed sample size calculations, statistical power—even assuming the interaction effect size is larger than any known main effect—is insufficient for most studies; by power and prior estimates, the false discovery rate exceeds 97%. The root is mathematically simple—the interaction term is a “gene × environment” product, which naturally has very small variance in non-experimental populations; small variance makes detection difficult. The authors also corrected a popular misconception: the correlation between product terms and main effect terms does not affect the power of interaction testing.

    This imposes a hard constraint on “lunar phase × MTNR1B.” The main effect of lunar phase is already only on the order of tens of minutes; the interaction term would be even smaller, and genotypes cannot be sampled by extreme grouping. To test direction, sample size and recording duration must be designed based on “detectability of direction” rather than main effects, and the directional hypothesis must be preregistered—otherwise, anything seen can be explained as a flip. Furthermore, individual light sensitivity varies greatly: personalized dose-response data from 55 individuals showed that the effective dose (ED50) for inhibition varied from 6 lux to 350 lux, spanning an order of magnitude, with a coefficient of variation of 26% (PMID 31138694); the threshold for melatonin inhibition onset is approximately 1.5 melanopic lux, saturating at 305 (PMID 30697806). Lunar illumination indeed falls within the low-dose sensitive range—this is both a reason for “possible effect” and a reason for “extreme difficulty in stable detection.”

    VI. Implications for Yijing Studies

    In divination theory, there is a basic rule: the same character can have roles that are completely reversed in different fortunes—”useful gods” change with major cycles and annual flows; no character is inherently good. The G allele plays precisely this role here: it acts as a buffer under artificial light, but whether it should be a buffer or amplifier under natural light has never been measured. Meanwhile, yeast rules indicate this is not vague rhetoric but carries specific predictions: the less similar the perturbation source, the higher the probability of direction reversal.

    This is also precisely where “harmony between heaven and man” is most easily misunderstood. It does not mean “if the sky is good, you are good,” but rather the same sky, different people, different connections. The moon looks round to everyone, but “following” versus “opposing” is relative to one’s own rhythm—so the correct question is never “is the moon harmful?” but rather “for my schedule line, is the moon a push or a resistance?”

    Evidence and Limitations

    Verified: MTNR1B rs10830963 G allele independently regulates melatonin secretion dynamics (+41 minutes, termination phase delay 1.37 hours, PMID 26868293) and increases type 2 diabetes risk (PMID 31623012); night shift × rs10830963 interaction consistently shows G allele buffering across two independent outcomes (stroke P=0.010, PMID 35411403; prostate cancer P=0.009, PMID 35064058); stratified analysis reveals evidence of opposite direction (breast cancer late-type Ptrend=0.015, PMID 35467761); the same interaction was not replicated in one sample of 189,488 individuals (PMID 31757795); antagonistic direction in yeast g×eQTL is enriched in dissimilar environments (PMID 27903611), and trans-variants often show opposite directions across conditions which can cancel each other out (PMID 18416601); in the first decade of cG×E studies, 96% new discoveries were positive versus 27% replication positivity, with false discovery rate >97% (PMID 21890791); individual light sensitivity ED50 spans 6–350 lux (PMID 31138694); melatonin inhibition onset is approximately 1.5 melanopic lux (PMID 30697806).

    Partially Verified: The chain “lunar phase → ipRGC → SCN → pineal gland → melatonin → MT2 → sleep” has evidence at each link, but whether the entire chain drives measurable melatonin changes under real moonlight (0.1–0.3 lux) lacks direct measurement; lunar illumination falling into the low-dose sensitive range is a conversion inference, not measured inhibition levels.

    To Be Verified: The direction of the lunar phase × MTNR1B interaction—whether the G allele acts as an amplifier or buffer under natural light, no study has directly tested this; whether this direction is opposite to the buffering direction under night shifts; whether interaction direction systematically varies with perturbation intensity and spectral ratios; whether interaction power is sufficient under sample sizes designed for direction detection.

    Related Reading

    Evidence Level: B
    References: PMID 26868293, PMID 31623012, PMID 35411403, PMID 35064058, PMID 35467761, PMID 31757795, PMID 27903611, PMID 18416601, PMID 21890791, PMID 31138694, PMID 30697806


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 光照剂量→月相-月经耦合衰减的斜率:MTNR1B 基因型是不是那个被忽略的敏感度旋钮

    Evidence level: B

    光暴露是同一条命令,但对不同的人执行力度不同。真正要问的不是”光照会不会破坏同步”,而是”每多照一小时光,你的同步断掉多少”——这个斜率,可能有一部分早就写在褪黑素受体基因里。

    正文

    先说本篇和同族几篇的分界线。前面几篇的结局变量都是睡眠——本质是”某个夜晚的一个读数”,问的是”两类人读数是否不同”。本篇的结局完全不同:剂量-效应曲线的斜率本身,问的是”同一类人,每多一档光照,同步度掉多少”。

    这个差别不是修辞。截距型交互只需要”有/无”二分环境(夜班/不夜班);斜率型交互要求环境是连续剂量、结局是逐剂量变化的速率,于是交互结构从二维升到三维:剂量 × 基因型 × 结局斜率。

    一、斜率这件事,已经有可测的物理量

    先把”敏感度”从模糊感受变成一个数。2019 年 PNAS 的个体剂量-效应实验给了基准:55 名受试者接受 10–2000 lux 五小时夜间光照,个体半数有效照度(ED50)从 6 lux 跨到 350 lux,变异系数 26%(PMID 31138694)——最敏感者和最不敏感者的”起效门槛”差近 60 倍。

    这不是孤例。2020 年《Scientific Reports》用可穿戴分光光度计测了 59 个家庭:48% 的家庭照明足以让褪黑素抑制 50%,而对同一个”平均家庭”环境,个体预测抑制率从 0% 到 87% 全谱铺开(PMID 33154450)。”你家灯亮不亮”和”你怕不怕光”是两个独立变量,交互结果完全不可预测。

    更关键的是,这个敏感度是可遗传的。双生子研究(9 对同卵 + 11 对异卵)发现,夜间褪黑素分泌量高度可遗传,而光照抑制敏感度的同卵一致性明显高于异卵、远高于无关个体(PMID 16769177)。这是整篇的支点:敏感度差异有基因成分,那么问题就变成”能不能找到它”。

    二、候选基因这端,有实测锚点

    MTNR1B 编码 MT2 受体,是褪黑素作用于视交叉上核与睡眠网络的主要门户。Lane 2016 年在实验室协议(n=58–96)加 3021 人多导睡眠监测中发现:rs10830963 风险等位(G)携带者的褪黑素升高时长延长 41 分钟、暗光褪黑素终止相位(DLMOff)延迟 1.37 小时(PMID 26868293)——G 等位让夜里的褪黑素信号”更长、更晚结束”。

    更值得注意的是一个比例数字:2015 年一项褪黑素-血糖试验发现,rs10830963 风险等位解释了褪黑素对血糖 AUC 效应的个体间差异的 26%(PMID 26440713)。这是”单一位点能吃掉多少敏感度差异”目前唯一的实测锚点——26% 不算多,但绝不是零。

    第三块拼图来自生殖轴本身,这是前面几篇没用到的一条:MTNR1A rs2119882 与 MTNR1B rs10830963 都被报告与 PCOS 及月经异常相关——359 名 PCOS 患者加 215 名对照的中国样本显示,rs10830963 G 等位携带者的 PCOS 风险升高,且与月经不规律显著关联(PMID 30664204)。机制综述把通路讲清楚了:夜间光照经 ipRGC-视交叉上核抑制褪黑素,褪黑素再经 MT1/MT2 作用于下丘脑 GnRH 神经元、垂体和卵巢(PMID 41834803)。而光照-生殖轴的因果方向也已有人测过:2007 年交叉对照试验给 22 名周期略长的女性在卵泡期晨间施加 4300 lux 光照一周,催乳素、LH、FSH 及卵泡尺寸均显著升高,排卵周期数从 6 次增加到 12 次(PMID 17290302)。

    三、把三块拼起来:本篇的假设

    光照是连续剂量,耦合衰减是结局斜率,MTNR1B/MTNR1A 基因型是斜率调节因子。 可检验的假设是:在纵向队列里以夜间黑视素照度(melanopic EDI)为连续自变量拟合”剂量→月相-月经同步指数”的斜率;携带风险等位的个体,其衰减斜率应显著更陡(交互项 ← 基因型 × 夜间照度)。

    这条假设至今没被检验,因为它要求三样东西同时在场:连续可穿戴光剂量、跨 6–12 个周期的同步度估计、基因分型。现有基因研究只测睡眠或代谢(PMID 26868293, 31623012),现有光剂量研究不测基因(PMID 31138694, 33154450)。

    四、三个必须提前拆掉的陷阱

    第一,统计功效极低。 基因×连续环境交互是最吃样本的设计:连续暴露的交互项会因测量误差系统性衰减,样本量需求随效应量下降呈非线性爆炸(PMID 29780190);”暴露富集型”抽样(按暴露与结局轨迹优先选基因分型子样本)能在同等预算下提高功效(PMID 28497531)。

    第二,基因型是代理变量,还是独立机制? 若基因型能预测斜率完全是因为它影响了个体 ED50,这设计就只绕了一圈。只有控制住实测 ED50 后基因型效应仍在,才说明存在独立通路(例如 MT2 在卵巢本地时钟的直接作用)。因此队列必须同时做小样本实验室剂量-效应测 ED50 作为中介变量。

    第三,ED50 与耦合斜率不同构。 ED50 是急性内分泌读数,耦合斜率是数月的相位统计量。不能假设二者的基因效应方向一致——这正是本篇真正的知识增量。

    五、可执行的方案骨架

    招募 150–250 名规律周期女性,腕戴分光光度计(photopic lux + melanopic EDI,30 秒采样)连续记录 6–12 个月;App 逐周期记录月经起始日并标注月相相位。分析用两阶段设计:先在每个个体内拟合”夜间黑视素照度→同步指数”斜率,再以该斜率为结局对基因型及基因型×剂量交互项做回归,纳入年龄、BMI、轮班史与实测 ED50。

    结局三种都有信息量:斜率更陡 → 敏感度有基因基础;斜率无差异但 ED50 有差异 → 基因只走褪黑素通路,耦合衰减另有来源;都无差异 → 个体差异来自光照暴露史与作息弹性。

    对”天人相应”的意义:古人讲”天地之气,感而遂通”,但从未说每个人的”通”是同一个灵敏度。这套设计第一次把”禀赋差异”从一句玄话,变成可以标定、可以分成三档结局的回归系数。

    证据与局限

    已验证: 个体褪黑素抑制敏感度差异巨大且可测量(ED50 跨 6–350 lux,CV 26%,PMID 31138694);家庭照明普遍达到生物有效水平,同环境下个体响应差异极大(PMID 33154450);褪黑素分泌与光敏感度有显著遗传成分(双生子,PMID 16769177);MTNR1B rs10830963 改变褪黑素分泌时长与终止相位(+41 min / +1.37 h,PMID 26868293);单一位点可解释褪黑素效应个体差异的约 26%(PMID 26440713);MTNR1A/MTNR1B 位点与月经异常及 PCOS 相关(PMID 30664204);光照剂量可改变排卵率(PMID 17290302);”基因型×光照扰动源”交互在大规模队列中可检出(夜班×MTNR1B 卒中 P 交互=0.010,PMID 35411403;昼夜型×MTNR1B T2D 分层差异,PMID 31623012)。

    部分验证: 光照-褪黑素-生殖轴的机制通路在动物模型中较充分、在人类中多为间接证据,综述明确提示跨物种外推需谨慎(PMID 41834803);MTNR1B 与昼夜型/夜班的交互在不同结局上方向不一致(T2D 的交互项本身未达显著,PMID 31623012),说明”交互存在”与”交互方向”都尚未稳定;褪黑素抑制阈值本身也随实验条件从 1.5 到 305 melanopic lux 浮动(PMID 30697806)。

    待验证: MTNR1B/MTNR1A 基因型是否调节”光照剂量→月相-月经耦合衰减斜率”这一效应量;基因型效应中有多大比例是通过个体 ED50 中介、多少是独立通路;耦合衰减斜率在个体内跨年是否稳定(可重复性);连续剂量交互项在真实队列中所需的样本量与抽样设计。

    风险与注意: 本篇为研究设计提案,不构成任何个人化光照或生殖健康建议。个体光敏感度不可自我判定,需实验室剂量-效应测定;月经周期异常可能由多种内分泌病因引起,应优先就医排查,不应归因于光照或基因。

    相关阅读

    证据等级: B
    参考文献: PMID 31138694, PMID 33154450, PMID 16769177, PMID 26868293, PMID 26440713, PMID 30664204, PMID 17290302, PMID 41834803, PMID 35411403, PMID 31623012, PMID 30697806, PMID 29780190, PMID 28497531


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Slope of Light Dose-Moon Phase-Menstrual Coupling Attenuation: Is the MTNR1B Genotype the Overlooked Sensitivity Knob?

    Evidence level: B

    Light exposure is a single command, but the execution intensity varies across individuals. The question is not “Will light disrupt synchronization?” but rather “How much does synchronization drop for every additional hour of light?”—this slope may be partially pre-encoded in melatonin receptor genes.

    Introduction

    First, delineate the boundary between this paper and its related studies. Previous papers focused on sleep as the outcome variable—essentially “a reading at a specific night,” asking whether “readings differ between two groups.” This paper’s outcome is fundamentally different: the slope of the dose-response curve itself, asking “For the same group, how much does synchronization drop per additional tier of light exposure.”

    This distinction is not rhetorical. Intercept-type interactions require a binary “present/absent” environment (night shift vs. non-night shift); slope-type interactions require a continuous dose environment with an outcome that is a rate of change per dose, thereby elevating the interaction structure from two dimensions to three: Dose × Genotype × Outcome Slope.

    I. Slope Already Has a Measurable Physical Quantity

    First, convert the vague sensation of “sensitivity” into a number. A 2019 PNAS individual dose-response experiment provided a baseline: 55 subjects received 10–2000 lux of nighttime light for five hours, with individual ED50 values ranging from 6 lux to 350 lux, a coefficient of variation (CV) of 26% (PMID 31138694)—the threshold for efficacy differed nearly 60-fold between the most and least sensitive individuals.

    This is not an isolated case. A 2020 study in Scientific Reports used wearable spectroradiometers to measure 59 households: 48% of household lighting was sufficient to inhibit melatonin by 50%, yet for the same “average household” environment, individual predicted inhibition rates spanned the full spectrum from 0% to 87% (PMID 33154450). “Is your light on?” and “Are you afraid of light?” are independent variables with an interaction result that is entirely unpredictable.

    More critically, this sensitivity is heritable. Twin studies (9 monozygotic + 11 dizygotic pairs) found that nighttime melatonin secretion is highly heritable, while the consistency of light inhibition sensitivity among monozygotic twins is significantly higher than dizygotic pairs and far higher than unrelated individuals (PMID 16769177). This is the pivot of this paper: since sensitivity differences have a genetic component, the question becomes “Can we identify it?”

    II. Candidate Genes Have Empirical Anchors

    MTNR1B encodes the MT2 receptor, serving as the primary gateway for melatonin acting on the suprachiasmatic nucleus and sleep networks. Lane et al. (2016) found in a laboratory protocol (n=58–96) with 3021 participants undergoing polysomnography that carriers of the risk allele (G) at rs10830963 experienced a 41-minute extension in melatonin elevation duration and a 1.37-hour delay in the dark-light melatonin off (DLMOff) phase (PMID 26868293)—the G allele makes nighttime melatonin signals “longer and ending later.”

    Notably, a proportion from a 2015 melatonin-glucose trial revealed that the rs10830963 risk allele explained 26% of the inter-individual variation in melatonin’s effect on glucose AUC (PMID 26440713). This is the only empirical anchor for “how much sensitivity difference a single site can account for”—26% is not large, but certainly not zero.

    The third puzzle piece comes from the reproductive axis itself, a pathway unused in previous papers: both MTNR1A rs2119882 and MTNR1B rs10830963 have been reported associated with PCOS and menstrual irregularities. A Chinese sample of 359 PCOS patients and 215 controls showed that G allele carriers at rs10830963 had elevated PCOS risk and significant association with irregular menstruation (PMID 30664204). Mechanism reviews clarify the pathway: nighttime light inhibits melatonin via ipRGCs → suprachiasmatic nucleus, which then acts on the hypothalamic GnRH neurons, pituitary, and ovaries via MT1/MT2 (PMID 41834803). Furthermore, the causal direction of the light-reproductive axis has been measured: a 2007 crossover trial applied 4300 lux light in the follicular phase for one week to 22 women with slightly longer cycles, resulting in significant increases in prolactin, LH, FSH, and follicle size, with ovulatory cycles increasing from 6 to 12 (PMID 17290302).

    III. Putting the Three Pieces Together: This Paper’s Hypothesis

    Light is a continuous dose; coupling attenuation is the outcome slope; MTNR1B/MTNR1A genotype is a slope modifier. The testable hypothesis is: In longitudinal cohorts, fit the slope of “nighttime melanopic EDI → lunar-menstrual synchronization index” using continuous nighttime melanopic exposure as the independent variable; individuals carrying the risk allele should exhibit a significantly steeper attenuation slope (interaction term ← Genotype × Nighttime Illuminance).

    This hypothesis remains untested because it requires three simultaneous elements: continuous wearable light dosing, cross-cycle (6–12 cycles) synchronization estimation, and genotyping. Existing genetic studies only measured sleep or metabolism (PMID 26868293, 31623012); existing light dosing studies do not measure genes (PMID 31138694, 33154450).

    IV. Three Traps to Disassemble in Advance

    First, statistical power is extremely low. Gene × continuous environment interactions are sample-intensive designs: interaction terms for continuous exposure suffer systematic attenuation due to measurement error, and sample size requirements explode non-linearly as effect sizes decrease (PMID 29780190); “exposure-enriched” sampling (prioritizing genotyped sub-samples based on exposure-outcome trajectories) can boost power within the same budget (PMID 28497531).

    Second, is genotype a proxy variable or an independent mechanism? If the genotype predicts the slope entirely because it affects individual ED50, this design merely circles back. Only if the genotype effect persists after controlling for measured ED50 can an independent pathway be confirmed (e.g., direct action of MT2 on local ovarian clocks). Therefore, cohorts must simultaneously conduct small-sample laboratory dose-response measurements of ED50 as a mediator.

    Third, ED50 and coupling slope are not isomorphic. ED50 is an acute endocrine reading; the coupling slope is a multi-month phase statistic. One cannot assume their genetic effects align in direction—this is the true knowledge increment of this paper.

    V. Executable Protocol Skeleton

    Recruit 150–250 women with regular cycles; wear spectroradiometers (photopic lux + melanopic EDI, 30-second sampling) continuously for 6–12 months; app records menstrual start dates cycle-by-cycle and annotates lunar phase. Analysis uses a two-stage design: first fit the “nighttime melanopic EDI → synchronization index” slope within each individual, then use that slope as the outcome to regress against genotype and the genotype × dose interaction term, adjusting for age, BMI, shift history, and measured ED50.

    All three outcomes provide information: steeper slope → sensitivity has a genetic basis; no slope difference but ED50 difference → genes only operate via melatonin pathway, coupling attenuation has another source; no differences in either → individual variation stems from light exposure history and circadian flexibility.

    Significance for “Heaven and Earth Correspondence”: Ancient texts state “Heavenly and earthly qi, when touched, become connected,” but never claimed everyone’s “connection” shares the same sensitivity. This design converts “禀赋差异” (inherent differences) from an abstract phrase into a calibratable regression coefficient that can be stratified into three outcome tiers.

    Evidence and Limitations

    Verified: Individual melatonin inhibition sensitivity varies greatly and is measurable (ED50 spans 6–350 lux, CV 26%, PMID 31138694); household lighting generally reaches biological effective levels, yet individual responses vary greatly under the same environment (PMID 33154450); melatonin secretion and light sensitivity have significant heritable components (twin studies, PMID 16769177); MTNR1B rs10830963 alters melatonin secretion duration and termination phase (+41 min / +1.37 h, PMID 26868293); a single site explains ~26% of inter-individual variation in melatonin effects (PMID 26440713); MTNR1A/MTNR1B sites associated with menstrual irregularities and PCOS (PMID 30664204); light dose alters ovulation rates (PMID 17290302); “Genotype × Light Perturbation Source” interactions detectable in large cohorts (Night shift × MTNR1B stroke P interaction=0.010, PMID 35411403; Chronotype × MTNR1B T2D stratified differences, PMID 31623012).

    Partially Verified: The light-melatonin-reproductive axis mechanism is well-established in animal models but relies on indirect evidence in humans; reviews explicitly caution against cross-species extrapolation (PMID 41834803); MTNR1B interactions with chronotype/night shifts show inconsistent directions across outcomes (T2D interaction term itself not significant, PMID 31623012), indicating that both “interaction existence” and “interaction direction” remain unstable; melatonin inhibition thresholds also fluctuate across experimental conditions from 1.5 to 305 melanopic lux (PMID 30697806).

    To Be Verified: Whether MTNR1B/MTNR1A genotypes modulate the effect size of “light dose → lunar-menstrual coupling attenuation slope”; the proportion of genotype effects mediated via individual ED50 versus independent pathways; whether the coupling attenuation slope is stable across years within individuals (reproducibility); required sample sizes and sampling designs for continuous dose interaction terms in real-world cohorts.

    Risks and Precautions: This paper is a research design proposal and does not constitute personalized lighting or reproductive health advice. Individual light sensitivity cannot be self-assessed; laboratory dose-response testing is required. Menstrual cycle irregularities may stem from multiple endocrine etiologies; medical evaluation should be prioritized, and causation should not be attributed to light or genes.

    Related Reading

    Evidence Level: B
    References: PMID 31138694, PMID 33154450, PMID 16769177, PMID 26868293, PMID 26440713, PMID 30664204, PMID 17290302, PMID 41834803, PMID 35411403, PMID 31623012, PMID 30697806, PMID 29780190, PMID 28497531


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 阈值还是坡度:月相-月经耦合衰减的判决逻辑与断点回归的误用

    Evidence level: B

    破坏”天人相应”的夜间光照,究竟是跨过某条线才生效,还是每多一分就衰减一分?这个问题不是靠堆数据回答的,而是靠模型之间的正面对决——而大多数人选的工具,从一开始就选错了。

    正文

    “光照破坏天人相应”这句话,藏着一个没被回答的追问:它是”跨过某条线才破坏”,还是”每多一分就少一分”? 两种答案指向完全不同的生活方案——前者说”把卧室灯压到某 lux 以下就保住了”,后者说”不存在安全线,能少一分是一分”。前作已经把耦合的存在(PMID 40991714)和个体剂量-效应曲线该长什么样讲清了,本文处理最后一步、也最容易被糊弄的一步:怎么用数据判决”阈值模型”与”连续坡度模型”这两家。

    第一件事:多数人在这一步选错了工具。 常见说法是”用断点回归(RDD)测临界光照强度”。但 RDD 有硬前提——存在基于阈值的分配规则:CD4 计数低于 200 就给药,个体在界线两侧形成准随机(PMID 25061922)。夜间光照没有这条规则,没有人”按 100 lux 划线决定谁被暴露”;强行套用,估出来的不是生理阈值,而是人为挑出的断点,且只在阈值邻域内有效(PMID 37528626)。正确的工具是分段回归/阈值回归:模型本身就是”两段不同斜率在未知转点相接”(PMID 8928781),转点由数据估出而非人为指定(PMID 40896720 给出可复现流程),并可加权纳入研究间异质性(PMID 37251960)。这不是术语之争:RDD 报的是”界线两侧的跳跃”,分段回归报的是”斜率在哪换挡”——月末同步度是连续量,只有后者对得上。

    第二件事:模型对决不能查表。 阈值模型有个统计麻烦——零假设下转点不可识别,似然比统计量不服从标准卡方分布(非正则渐近)。所以”p<0.05 就宣布有阈值”是错的,必须用自助法/置换法构造零分布(PMID 29037149 用蒙特卡洛验证了检验的第一类错误率)。可执行的刀法:把线性、折棍(hockey-stick)、S 形、双段线性四种模型放在同一批个体数据上比 AIC/BIC,再用自助法给出转点位置的置信区间——若转点的 95% 区间横跨整个剂量范围,那就是没有阈值,只有坡度。

    第三件事:最大的陷阱是”群体 S 形”冒充”个体阈值”。 褪黑素抑制的曲线是 S 形:约 100 lux 达半程(PMID 10922269),1.5 melanopic lux 启动、305 saturate(PMID 30697806)。S 形不是阈值——它没有断点。 更关键的是:若每个人的折点位置各不相同(个体光敏感度可相差一个数量级),把这些个体曲线平均起来,必然得到一条平滑的 S 形。于是群体层面看到”低剂量区几乎无响应、高剂量区饱和”,极易被读成”存在临界线”,而真相可能只是”每人各有各的线、且分布很宽”。判决只有一条路:个体内设计——同一人被施加 ≥5 级光剂量,为每人单独估曲线,再看个体转点是”聚在一处”还是”散开”。

    第四件事:剂量不是标量。 “临界光照强度”默认 lux 是唯一自变量。但双因素实验显示时长主效应显著、强度不显著(PMID 21532952),短脉冲每分钟效率更高(PMID 22526883)。若真实杠杆是时长与光谱,任何一维 lux 阈值模型都会被错配,估出的”阈值”只是丢失变量的投影。

    第五件事:必须扣掉引力这个共变项。 2025 年 Sci Adv 的分析显示,2010 年后耦合并未消失,而是集中在 1 月,作者归因于日-地-月引力最强的时节(PMID 40991714)。这意味着不控制季节/引力项,1 月残留同步会被误读为”阈值以下仍保留耦合”,把整条阈值往高处推。该研究还给出一个关键分层变量:只有周期长度落在 26–36 天窗口的女性才有条件同步——阈值模型必须把它作为协变量,否则同步度差异会被错算成光照效应。

    可执行的判决方案:被试内交叉设计,每人在 ≥6 个周期内经历 ≥5 级夜间黑视素照度(含时长与蓝光比的正交操作),结局用圆形统计的同步指数(Rayleigh 检验的合向量长度 R̄,需做相位折叠处理轴双峰,方法学见 PMID 40991714);先对每个个体拟合四模型族并留出转点置信区间,再检验”个体转点是否同质”与”低于转点区间的斜率是否与零等价”(等价性检验/TOST)。三种结局各有含义:个体转点同质且低剂量区斜率与零等价 → 真阈值,可以划安全线;个体转点离散 → 是混合伪影,只能谈”个体线”;转点区间横跨全量程 → 没有阈值,只有坡度。

    对”天人相应”的意义:这一步做完,”光煞”才从”玄”变成”界”。古人说”明暗须有分寸”,现代版的问法更锋利——分寸是一条线还是一个坡度?而这条线,是不是每个人都画在同一个位置?

    证据与局限

    已验证: RDD 的适用前提(阈值分配规则、仅邻域内有效)与实施规范(PMID 25061922, 37528626);分段回归/阈值回归可估计未知转点且流程可复现(PMID 8928781, 40896720);阈值模型需自助法/置换法处理非正则渐近(PMID 29037149);加权变点回归可从系统评价估 NOAEL(PMID 37251960);褪黑素抑制曲线为 S 形、约 100 lux 半程、1.5–305 melanopic lux 区间(PMID 10922269, 30697806);时长主效应强于强度、短脉冲效率更高(PMID 21532952, 22526883);2010 年后月相-月经耦合集中于 1 月、周期长度 26–36 天为同步窗口(PMID 40991714)。

    部分验证: 上述方法学证据全部来自其他领域(流行病学、毒理学、光生物学),尚无一项以月相-月经同步度为结局做过阈值 vs 坡度的模型对决;”群体 S 形可由个体转点异质混合产生”在光生物学中为公认数理事实,但未在月经同步场景中被量化。

    待验证: 月相-月经耦合衰减是否存在统计上可检测的阈值(个体转点是否同质);群体层面的”S 形”中有多少来自个体异质混合;剂量应由强度、时长还是光谱主导(若为多变量,一维阈值模型的形式是否正确);季节/引力项与光照项在观测数据中的可识别边界。

    相关阅读

    证据等级: B
    参考文献: PMID 40991714, PMID 10922269, PMID 30697806, PMID 25061922, PMID 37528626, PMID 8928781, PMID 40896720, PMID 29037149, PMID 37251960, PMID 21532952, PMID 22526883


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Threshold or Slope: Judgment Logic and Misuse of Breakpoint Regression in Moon-Phase-Menstrual Coupling Attenuation

    Evidence level: B

    Does disrupting “Heaven-Human Correspondence” via nocturnal illumination require crossing a specific line to take effect, or does attenuation occur incrementally with every additional unit? This question cannot be answered by piling up data; it requires a head-to-head confrontation between models—yet most practitioners select tools that are fundamentally misaligned from the outset.

    Main Text

    “Disruption of Heaven-Human Correspondence via Light” hides an unanswered question: Is the effect binary (“cross a threshold to disrupt”) or linear (“attenuate incrementally with every unit increase”)? These answers dictate entirely different life protocols—the former suggests “keep bedroom light below X lux to preserve synchrony,” while the latter implies “no safe threshold exists; minimize exposure as much as possible.” Previous work has clarified the existence of coupling (PMID 40991714) and the expected shape of individual dose-response curves; this article addresses the final, most easily misleading step: how to adjudicate between the “Threshold Model” and the “Continuous Slope Model” using data.

    First Issue: Most practitioners select the wrong tool at this step. The common claim is “use Regression Discontinuity Design (RDD) to measure critical light intensity.” However, RDD relies on a hard prerequisite—the existence of a threshold-based allocation rule: e.g., administering medication when CD4 counts drop below 200, creating quasi-randomization on either side of the boundary (PMID 25061922). No such rule exists for nocturnal light; no one “draws a line at 100 lux to decide who is exposed.” Forcing RDD here estimates an artificial discontinuity rather than a physiological threshold, valid only within the immediate neighborhood of that arbitrary line (PMID 37528626). The correct tool is Piecewise Regression / Threshold Regression: the model itself posits “two different slopes meeting at an unknown inflection point” (PMID 8928781), where the turning point is estimated from data rather than arbitrarily specified (PMID 40896720 provides a reproducible workflow), and can incorporate study-level heterogeneity via weighting (PMID 37251960). This is not a terminology dispute: RDD reports “jumps on either side of a line,” while piecewise regression reports “where the slope shifts”—menstrual synchrony is a continuous variable, and only the latter aligns.

    Second Issue: Model confrontation cannot rely on lookup tables. The threshold model has a statistical complication—the inflection point is unidentified under the null hypothesis, and the likelihood ratio statistic does not follow a standard chi-square distribution (non-regular asymptotic). Therefore, “declare a threshold exists if p<0.05” is incorrect; one must construct the null distribution via bootstrapping or permutation (PMID 29037149 verified Type I error rates via Monte Carlo). Executable methodology: Fit four model families (linear, hockey-stick, sigmoidal, two-segment linear) on the same individual dataset and compare via AIC/BIC, then use bootstrapping to generate confidence intervals for the turning point—if the 95% CI for the turning point spans the entire dose range, there is no threshold, only a slope.

    Third Issue: The largest trap is “Population Sigmoid” masquerading as “Individual Threshold.” Melatonin suppression curves are sigmoidal: reaching half-maximum at ~100 lux (PMID 10922269), with onset at 1.5 melanopic lux and saturation at 305 (PMID 30697806). A sigmoid is not a threshold—it has no discontinuity. More critically: if each individual’s inflection point varies significantly (individual light sensitivity can differ by an order of magnitude), averaging these individual curves necessarily yields a smooth sigmoid. Thus, at the population level, observing “no response at low doses and saturation at high doses” is easily misread as “a critical line exists,” while the truth may simply be “everyone has their own line, and the distribution is wide.” The adjudication path is singular: Within-subject design—expose each person to ≥5 levels of nocturnal melatonin-suppressing illuminance (orthogonally manipulating duration and blue-light ratio), fit curves individually for each person, then assess whether individual turning points “cluster together” or “scatter.”

    Fourth Issue: Dose is not a scalar. “Critical light intensity” defaults to lux as the sole independent variable. However, two-factor experiments show significant main effects for duration but not intensity (PMID 21532952), with short pulses being more efficient per minute (PMID 22526883). If the true lever is duration or spectrum, any univariate lux threshold model will be misspecified; the estimated “threshold” is merely a projection of lost variables.

    Fifth Issue: Must control for gravitational confounding. A 2025 Sci Adv analysis indicates that post-2010 coupling did not disappear but concentrated in January, attributed by authors to the period of strongest sun-earth-moon gravitational forces (PMID 40991714). This means failing to control for season/gravity terms will misread January residual synchrony as “coupling persists below the threshold,” artificially pushing the entire threshold upward. The study also provides a key stratification variable: only women with cycle lengths falling within the 26–36 day window exhibit conditional synchrony—the threshold model must include this as a covariate; otherwise, synchrony differences will be erroneously calculated as light effects.

    Executable Adjudication Protocol: Within-subject crossover design, where each participant experiences ≥5 levels of nocturnal melatonin-suppressing illuminance (including orthogonal manipulation of duration and blue-light ratio) across ≥6 cycles. Outcome: synchrony index via circular statistics (mean vector length $\bar{R}$ from Rayleigh test, requiring phase-folding handling for bimodal axes; methodology see PMID 40991714). First, fit the four model families for each individual and derive confidence intervals for the turning point; then test “homogeneity of individual turning points” and “whether slopes below the turning point interval are equivalent to zero” (equivalence testing/TOST). Three outcomes carry distinct meanings:
    1. Homogeneous turning points + slope at low doses equivalent to zero → True threshold; a safe line can be drawn.
    2. Dispersed turning points → Mixed artifact; only “individual lines” can be discussed.
    3. Turning point interval spans full range → No threshold, only a slope.

    Significance for “Heaven-Human Correspondence”: Once this step is completed, “Light Sha” transforms from “symbolic” to “bounded.” Ancient texts say “light and darkness must have measure”; the modern formulation is sharper—is the measure a line or a slope? And is this line drawn at the same position for everyone?

    Evidence and Limitations

    Verified: RDD applicability prerequisites (threshold allocation rule, validity only within neighborhood) and implementation standards (PMID 25061922, 37528626); piecewise/threshold regression capability to estimate unknown turning points with reproducible workflows (PMID 8928781, 40896720); necessity of bootstrapping/permutation for non-regular asymptotics in threshold models (PMID 29037149); weighted change-point regression for estimating NOAEL from systematic reviews (PMID 37251960); melatonin suppression curves are sigmoidal (~100 lux half-max, 1.5–305 melanopic lux range) (PMID 10922269, 30697806); duration main effect stronger than intensity, short pulses more efficient (PMID 21532952, 22526883); post-2010 moon-phase-menstrual coupling concentrated in January, 26–36 day cycle window for synchrony (PMID 40991714).

    Partially Verified: All methodological evidence above originates from other fields (epidemiology, toxicology, photobiology); no study yet has conducted a threshold vs. slope model confrontation with moon-phase-menstrual synchrony as the outcome; “population sigmoid arising from heterogeneous individual turning points” is an accepted mathematical fact in photobiology but not yet quantified in the menstrual synchrony context.

    To Be Verified: Whether statistically detectable thresholds exist for moon-phase-menstrual coupling attenuation (homogeneity of individual turning points); the proportion of the population-level “sigmoid” arising from individual heterogeneity; whether dose should be driven by intensity, duration, or spectrum (if multivariate, is the univariate threshold model form correct); identifiable boundaries between season/gravity terms and light terms in observational data.

    Related Reading

    Evidence Level: B
    References: PMID 40991714, PMID 10922269, PMID 30697806, PMID 25061922, PMID 37528626, PMID 8928781, PMID 40896720, PMID 29037149, PMID 37251960, PMID 21532952, PMID 22526883


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 把北美的配对年份设计搬到东亚和欧洲:地磁边际增量能否随纬度增强?

    Evidence level: B

    一个在美国跑通的设计,搬到东亚和欧洲不是”换个数据源”那么简单——配对变量要重定义、纬度切换的可能不是强弱而是机制、超额死亡的方法在热带前提崩塌,还有一条与真实效应同向的假梯度,数据本身分不出来。

    正文

    一 北美那套设计,为什么在那里成立

    配对年份设计的逻辑是”先看气象、再看地磁”,它在北美能跑通靠两个前提。一是绝对湿度(AH)确实是流感季节性的主要驱动——Barreca 用 1973-2002 年 359 个美国县域月度数据证明,AH 低于约 6 g/kg 时流感死亡显著上升,季节间死亡差异约一半可由 AH 单独解释(PMID 23035135)。二是”非流行季”清晰可辨,夏季基线平稳,Serfling 类回归才能把趋势与季节项剥掉。这两个前提,出了北美就不再自动成立。

    二 第一道墙:配对变量在别处可能不是混杂,而是主角

    配对年份设计把 AH 当混杂变量消掉。但 2022 年一项覆盖中国七省市的研究(PMID 35218485)发现,AH 与流感活动的关系随气候带反转:北京、天津属北方,周均 AH 低于 3.41 与 6.62 g/m³ 时累计相对风险上升;广东、海南在最南方,风险反而随 AH 升高而上升,断点 16.74 与 20.18 g/m³;上海、浙江、重庆呈 U 型,最低风险点在 11.95、11.94、15.96 g/m³。香港与广州的研究给出同样方向(PMID 22168284):高绝对湿度显著增加全因与心肺死亡风险。也就是说,在华南,配对掉湿度等于把混杂变量换成了效应变量——把真正的驱动因素从模型里删掉了。配对标准必须随气候带重定义,这不是换数据源,是换设计。

    三 第二道墙:纬度切换的不是强弱,是机制

    “边际增量随纬度增强”,这一假设本身就写错了一层。目前唯一给出全球纬度梯度的证据来自 204 个地区的面板分析(PMID 37697364):高收入地区地磁暴频率与心血管死亡正相关(系数 14.007),且随纬度从极地向赤道递减——近极地缺血性心脏病系数 39.703、卒中 19.461,中纬度心血管 15.452,近赤道任何心血管结局都不显著。

    但这是”梯度存在”,不是”同一机制在放大”。挪威 Alta(70°N)的唾液褪黑素研究显示,地磁变化必须超过约 80 nT/3h 才显著压低褪黑素(PMID 11774869),即存在幅度阈值。而俄罗斯中高纬度 14.5 万份急救记录的分析更关键:中纬度的心肌梗死主要与 Pc1 地磁脉动相关,高纬度才主要与太阳风、行星际磁场驱动的磁暴相关(PMID 27717941)。同一暴露在不同纬度走的是不同通道。符号也不稳:巴库(40°N)4919 例急性心肌梗死的三年数据里,月度发病人数与太阳活动、地磁活动负相关,而每日入院前死亡却随地磁活动上升(PMID 17660721)。同一地区、不同时间尺度,符号可以相反。

    四 第三道墙:超额死亡估计方法在热带不再等价

    美国四种方法对比显示,泊松回归与 peri-season 率差法量级接近(年均 25,470 vs 22,454 例),但夏季基线率差法系统性偏高(PMID 19453440)。西班牙的结果是广义线性模型系统性低于 Serfling,后者在 65 岁以上给出每十万人 57 与 164 例(PMID 22284214)。希腊分析还出现可能的”收获效应”(PMID 30968823)。问题在于:Serfling 夏季基线法的前提是”夏季没有流感活动”。而在广东、香港这类亚热带城市,流感全年活动,“非流行季”这个前提根本不存在。跨地区时超额死亡方法必须事前预注册,否则地区间的”差异”可能只是方法差异。

    五 还有一道更麻烦的:误差本身也随纬度变化

    跨纬度分析通常用行星 Kp 指数,但它是 13 个台站的平均值,会抹掉集中在某一经度的地方性扰动。局域 K 与全球 Kp 的系统比较(DOI 10.3390/atmos16121319)显示二者相关 r≈0.84,但存在 122 个”局域 K≥6 而 Kp 未达阈值”的漏检。用 Kp 做跨纬度分析,等于给每个纬度用了同一把尺子。这是典型的回归稀释:衰减因子随纬度不同,低纬度信噪比低、被压得更狠(PMID 9465995;PMID 16051613 提醒简单的”量程比”校正可高估 29%)。结果是——即使真实效应量处处相同,你也会看到一条从赤道向极地上升的假梯度。这条假梯度与真实的纬度效应修饰方向一致,数据本身无法区分。

    六 该怎么做

    用扩展两阶段设计(PMID 35436963):第一层每地点用分布滞后非线性模型估 Kp 六个月累积窗口与呼吸道超额死亡的关联并降维(PMID 23297754);第二层多层随机效应 meta,地点嵌套于国家;第三层把纬度作为连续 meta-predictor 做 meta-regression。同时并行跑局域 K 指数做敏感性分析、超额死亡方法事前预注册,可复用多国多城市协作网络框架(PMID 39263673)。

    七 易学落点

    五运六气讲”天时”从不是全球统一的常数,而是”客气加临”落在不同方隅、所主不同。《素问》早说”地有高下,气有温凉”。跨地区复制的真正含义不是把北美结论搬过去,而是检验同一气在不同方位下是否换了一套作用方式。

    传统表述不区分”强弱不同”与”机制不同”。本篇的分辨点:若纬度只放大效应量,是同一机制在放大;若它切换了敏感通道(中纬 Pc1、高纬磁暴)或暴露参数(幅度阈值、湿度断点),那是机制切换,不能用一个系数外推。

    证据与局限

    已验证: 绝对湿度与流感死亡的关系在温带成立(PMID 23035135);地磁暴频率与心血管死亡的正相关随纬度从极地向赤道递减(PMID 37697364);中纬度与高纬度心肌梗死对应的空间天气参数不同(PMID 27717941);四种超额死亡估计方法在同一数据上量级接近但存在系统性差异(PMID 19453440)。

    部分验证: 地磁活动对褪黑素的影响存在幅度阈值(PMID 11774869,样本 25 人);巴库心肌梗死数据的尺度依赖符号反转(PMID 17660721,单城市三年);局域 K 与全球 Kp 的一致性(DOI 10.3390/atmos16121319,单台站比较);扩展两阶段框架在多地点环境流行病学中的可行性(PMID 35436963、PMID 39263673)。

    待验证: ①”地磁边际增量随纬度增强”至今没有任何一项研究直接检验过——204 地区的纬度梯度用的是 GBD 年度汇总数据,不是配对年份设计;②跨纬度时用 Kp 造成的衰减偏倚幅度从未被量化,因而无法判断已观测到的梯度有多少来自真实效应修饰、多少来自尺子误差;③亚热带城市”非流行季”缺失对 Serfling 基线的破坏程度需要专门的模拟研究。

    相关阅读

    证据等级: B
    参考文献: PMID 42350784, PMID 31511079, PMID 37697364, PMID 27717941, PMID 17660721, PMID 11774869, PMID 26177617, PMID 35660608, PMID 35218485, PMID 23035135, PMID 22168284, PMID 19453440, PMID 22284214, PMID 30968823, PMID 36630176, PMID 28847318, PMID 27136236, PMID 34789160, PMID 9465995, PMID 16051613, PMID 35436963, PMID 39263673, PMID 23297754;DOI 10.3390/atmos16121319


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Transferring the North American pairing year design to East Asia and Europe: Can geomagnetic marginal increments increase with latitude?

    Evidence level: B

    Applying a design validated in the United States to East Asia and Europe is not merely “changing the data source”—the pairing variable requires redefinition, the shift in latitude involves a change in mechanism rather than just intensity, the method for estimating excess mortality collapses under tropical conditions, and there exists a spurious gradient aligned with the true effect that the data itself cannot distinguish.

    Main Text

    I. Why the North American Design Works There

    The logic of the pairing-year design is “look at meteorology first, then geomagnetism.” Its success in North America relies on two premises. First, absolute humidity (AH) is indeed a primary driver of influenza seasonality—Barreca demonstrated using 359 monthly county-level data points from the US between 1973 and 2002 that influenza mortality rises significantly when AH falls below approximately 6 g/kg, with seasonality accounting for roughly half of the inter-seasonal mortality difference (PMID 23035135). Second, a distinct “non-epidemic season” is clearly identifiable; summer baselines are stable, allowing Serfling-type regression to strip away trends and seasonal terms. These two premises do not automatically hold outside North America.

    II. First Barrier: The Pairing Variable May Be the Effect, Not a Confounder, Elsewhere

    The pairing-year design treats AH as a confounding variable to be eliminated. However, a 2022 study covering seven provinces in China (PMID 35218485) found that the relationship between AH and influenza activity reverses across climate zones: in Beijing and Tianjin (Northern), cumulative relative risk rises when weekly average AH is below 3.41 and 6.62 g/m³; in Guangdong and Hainan (Southernmost), risk instead rises with increasing AH, with breakpoints at 16.74 and 20.18 g/m³; Shanghai, Zhejiang, and Chongqing show a U-shape with minimum risk points at 11.95, 11.94, and 15.96 g/m³. Research in Hong Kong and Guangzhou yields the same direction (PMID 22168284): high absolute humidity significantly increases all-cause and cardiovascular mortality risks. In short, in South China, pairing out humidity converts a confounding variable into an effect variable—effectively removing the true driver from the model. Pairing standards must be redefined according to climate zones; this is not changing data sources, but changing the design itself.

    III. Second Barrier: The Shift in Latitude Involves Mechanism, Not Just Intensity

    The assumption that “marginal increments scale with latitude” contains a fundamental layer of error. Currently, the only evidence providing a global latitudinal gradient comes from a panel analysis of 204 regions (PMID 37697364): geomagnetic storm frequency shows a positive correlation with cardiovascular mortality in high-income regions (coefficient 14.007), decreasing from pole to equator—coefficients for ischemic heart disease near the poles are 39.703 and for stroke 19.461, mid-latitude cardiovascular outcomes are 15.452, while any cardiovascular outcome near the equator is non-significant.

    However, this indicates “gradient existence,” not “the same mechanism amplifying.” The Norway Alta (70°N) saliva melatonin study shows that geomagnetic changes must exceed approximately 80 nT/3h to significantly suppress melatonin (PMID 11774869), indicating an amplitude threshold. More critically, the analysis of 145,000 emergency records in Russia’s mid-to-high latitudes reveals that mid-latitude myocardial infarctions correlate primarily with Pc1 geomagnetic pulsations, whereas high latitudes correlate primarily with solar wind and interplanetary magnetic field-driven geomagnetic storms (PMID 27717941). The same exposure operates through different pathways at different latitudes. Signs are also unstable: in Baku (40°N), data from three years covering 4,919 acute myocardial infarction cases show a negative correlation between monthly incidence and solar/geomagnetic activity, while pre-hospital mortality rises with geomagnetic activity (PMID 17660721). In the same region, different time scales can yield opposite signs.

    IV. Third Barrier: Excess Mortality Estimation Methods Are No Longer Equivalent in the Tropics

    Comparisons across four US methods show that Poisson regression and peri-season rate difference methods yield comparable magnitudes (annual 25,470 vs. 22,454 cases), but the summer baseline rate difference method systematically overestimates (PMID 19453440). Spanish results indicate generalized linear models systematically underperform Serfling, which yields 57 and 164 cases per 100,000 for those over 65 (PMID 22284214). Greek analyses also reveal a potential “harvest effect” (PMID 30968823). The core issue lies in the Serfling summer baseline method’s premise: “no influenza activity during summer.” In subtropical cities like Guangdong and Hong Kong, influenza is active year-round; the premise of a “non-epidemic season” simply does not exist. When transferring across regions, excess mortality methods must be pre-registered to avoid the possibility that regional “differences” are merely methodological artifacts.

    V. Another Complication: Error Itself Varies with Latitude

    Cross-latitude analyses typically use the planetary Kp index, but as an average of 13 stations, it smoothes out localized disturbances concentrated at specific longitudes. A systematic comparison between local K and global Kp (DOI 10.3390/atmos16121319) shows a correlation of r≈0.84 but identifies 122 instances where “local K ≥ 6 while Kp fails to reach the threshold.” Using Kp for cross-latitude analysis is akin to using the same ruler for every latitude. This represents regression dilution: the attenuation factor varies by latitude, suppressing signal-to-noise ratios more severely at low latitudes (PMID 9465995; PMID 16051613 notes that simple “range ratio” corrections can overestimate by 29%). The result is that even if true effect sizes are identical everywhere, one observes a spurious gradient rising from equator to pole. This spurious gradient aligns with the direction of true latitude effect modification, making it indistinguishable from data alone.

    VI. What to Do

    Adopt an extended two-stage design (PMID 35436963): First, estimate the association between six-month cumulative Kp windows and respiratory excess mortality at each location using distributed lag nonlinear models, then reduce dimensionality (PMID 23297754). Second, employ a multilevel random-effects meta-analysis with locations nested within countries. Third, treat latitude as a continuous meta-predictor for meta-regression. Simultaneously, run local K indices as sensitivity analyses and pre-register excess mortality methods, leveraging multi-country, multi-city collaborative network frameworks (PMID 39263673).

    VII. Yijing Implications

    The Five Movements and Six Qi discuss “heavenly timing” not as a universal constant, but as “guest qi superposition” landing in different quarters with distinct governing functions. The Su Wen early states “the earth has high and low places, qi has warm and cool qualities.” The true meaning of cross-regional replication is not transplanting North American conclusions, but testing whether the same qi adopts a different mode of operation under different orientations.

    Traditional expressions do not distinguish between “different intensities” and “different mechanisms.” The distinguishing point here: if latitude only amplifies effect sizes, it is the same mechanism scaling up; if it switches sensitive channels (mid-latitude Pc1 vs. high-latitude storms) or exposure parameters (amplitude thresholds, humidity breakpoints), that is a mechanism switch, and extrapolation via a single coefficient is invalid.

    Evidence and Limitations

    Verified: The relationship between absolute humidity and influenza mortality holds in temperate zones (PMID 23035135); the positive correlation between geomagnetic storm frequency and cardiovascular mortality decreases from pole to equator with latitude (PMID 37697364); mid-latitude and high-latitude myocardial infarctions correspond to different space weather parameters (PMID 27717941); four excess mortality estimation methods yield comparable magnitudes on the same data but exhibit systematic differences (PMID 19453440).

    Partially Verified: Geomagnetic activity’s effect on melatonin involves an amplitude threshold (PMID 11774869, n=25); scale-dependent sign reversal in Baku myocardial infarction data (PMID 17660721, single city, three years); consistency between local K and global Kp (DOI 10.3390/atmos16121319, single station comparison); feasibility of the extended two-stage framework in multi-location environmental epidemiology (PMID 35436963, PMID 39263673).

    To Be Verified: ① No study to date has directly tested the hypothesis that “geomagnetic marginal increments scale with latitude”—the 204-region latitudinal gradient utilized GBD annual summary data, not a pairing-year design; ② The magnitude of attenuation bias caused by using Kp across latitudes has never been quantified, making it impossible to judge how much of the observed gradient stems from true effect modification versus measurement tool error; ③ The degree of Serfling baseline disruption caused by the absence of a “non-epidemic season” in subtropical cities requires dedicated simulation studies.

    Related Reading

    Evidence Level: B
    References: PMID 42350784, PMID 31511079, PMID 37697364, PMID 27717941, PMID 17660721, PMID 11774869, PMID 26177617, PMID 35660608, PMID 35218485, PMID 23035135, PMID 22168284, PMID 19453440, PMID 22284214, PMID 30968823, PMID 36630176, PMID 28847318, PMID 27136236, PMID 34789160, PMID 9465995, PMID 16051613, PMID 35436963, PMID 39263673, PMID 23297754; DOI 10.3390/atmos16121319


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 光照敏感度多基因评分×月相:从单个基因到整张评分表,能否预测谁被月亮牵动?

    Evidence level: B

    上一篇提出”MTNR1B 基因型可能决定谁被月亮牵动”,但单个位点解释不了多少——光敏感度本身是一个高度可遗传、个体差异跨一个数量级的性状。把它升级为”多基因评分×月相”交互,才是这条线真正能站住的检验设计。

    正文

    先说为什么必须从”单基因”升级到”评分表”。光敏感度是一个可遗传且个体差异巨大的性状,这一点早有硬证据:双生子研究显示,夜间强光下的褪黑素抑制敏感度与褪黑素分泌动力学本身都有显著遗传成分(PMID 16769177)。而差异的幅度更是惊人——同一强度的傍晚光照,不同个体触发褪黑素抑制的半数有效照度(ED50)可以跨越一个数量级,从十几 lux 到几百 lux 不等(PMID 31138694)。换句话说,”你有多怕光”这件事,人和人之间差得不是一点半点。褪黑素抑制对光的反应极其灵敏、且主要由视网膜的 melanopsin(黑视蛋白)通路驱动(PMID 30697806)。系统综述把这条线总结得很清楚:个体光敏感度差异是睡眠与昼夜节律个体差异的核心来源之一(PMID 33049062)。

    问题是,这个”敏感度”的遗传基础至今基本是空的。直到 2022 年 Burns 等人在 28 万人中做了全基因组”基因×光照”交互扫描(GWIS),才第一次定位到显著位点:ARL14EP 的 rs3847634,携带更多次要等位的人,日照对”晨型化”的推动更强(βGxE = −0.03,SE = 0.005);该位点同时是脑与视网膜组织中 ARL14EP 表达增强的 eQTL(PMID 36519390)。这个结果极其重要,但请注意效应量之小:单个位点的交互项 β 只有 0.03 的量级。GWIS 的富集分析还提示,光敏感度位点集中在 G 蛋白偶联谷氨酸受体信号通路、以及表达 Per2 的下丘脑(SCN)神经元里——也就是说,真正起作用的是一批位点的合力,而非某一个”月亮基因”。

    这一步”从单点到全基因组”的跨越,在计时型(chronotype)研究里已经走完了。计时型的大规模 GWAS 已在近 70 万人中识别出数百个位点(PMID 30696823);在此基础上构建的多基因评分(PRS/PGS)已被证明可携带、有预测力:睡眠时长 PRS 的最高与最低 5% 分位之间,报告睡眠时长相差 22.2 分钟(PMID 36648941);晨型 PGS 可稳定预测晨型本身(R² ≈ 1.55%)并外溢预测抑郁症状(PMID 38425306);甚至有研究用 19 个昼夜节律基因的靶向测序构建出预测效度约 80% 的”分子计时型”评分面板(PMID 35583112)。方法学上,用 PRS 去做基因-环境交互也已成套:IPRS 直接利用 GxE 结构重构评分(PMID 35401709)、通路水平 PRS 专为交互设计(PMID 40763299)、以及针对 GxE PRS 中 I 类错误膨胀与功效损失的系统校正方案(PMID 38303123)。已有 GxE 分析支持”晚型在特定环境下反而睡得更久”这类方向性结论(PMID 36745570),说明环境修饰效应能被真实检出。

    把三条线接起来,本主题的命题是这样的:把”光敏感度”从单基因升级为多基因总分,用该评分与月相的交互项,去预测个体对月相-睡眠耦合的反应强度。 它与上一篇的区别很关键——上一篇问的是”某个位点的交互方向是放大还是缓冲”,本篇问的是”加总之后,评分能不能真的提高预测力“。这是一个纯粹的增量价值问题,而不是方向问题。

    设计上可以做三层检验。第一层是增益检验:在同一个纵向队列里,分别拟合”月相 × rs3847634″与”月相 × 光敏感度 PRS”两组模型,比较 ΔR²、ΔAIC 与似然比——如果 PRS 不带来任何增量,说明光敏感度对月相反应没有净贡献,这正是最强的证伪条件。第二层是分位梯度检验:若假设成立,PRS 按十分位分层后,月相-睡眠关联强度应呈单调递增(或递减)的梯度,而不是只有极端两端有差异。第三层是可移植性检验:PRS 权重必须来自独立样本的 GWIS 摘要统计,然后在目标队列中投影——若用同一批样本既选位点又算评分,会因 winner’s curse 系统性高估效应。队列要求也不低:每人至少覆盖 3 个以上完整月相周期(约 3 个月)的腕表活动记录或 PSG,配合基因分型与可核算的夜间光照暴露代理(地理位置 + 云量 + 室内照度实测)。

    这条路线最大的陷阱不是统计,而是混杂。月相相位与月出/月落时间随季节与纬度系统性漂移,因此月相变量天然与季节、日照时长、作息季节漂移共线;如果只按相位打标签而不控制季节与纬度,很容易把季节效应误读成月光效应。此外,光照暴露的测量误差会系统性稀释交互项(attenuation),PRS 的跨族裔可移植性也远未解决。因此这项检验的合格线不只是”显著”,而是在控制季节/纬度/云量后仍显著、且增量优于单基因、并在独立队列中复现。

    对”天人相应”这层意思,这个设计的落点其实很朴素:古人讲”八字禀赋”,从来不是指某一个字的吉凶,而是整张表的格局。光敏感度也一样——它不是一个开关键,是一条连续谱,由几百个位点的合力决定。你被月亮牵动多少,不取决于你有没有那个”月亮基因”,而取决于你这张评分表落在谱上的哪一段。

    证据与局限

    已验证: 光敏感度是可遗传性状(褪黑素抑制敏感度与分泌动力学均有显著遗传度);个体光敏感度差异巨大(ED50 跨一个数量级),褪黑素抑制极灵敏且由 melanopsin 主导;GWIS 已定位光敏感度位点(ARL14EP rs3847634;富集于谷氨酸信号与 Per2+ SCN 神经元)但单点效应量极小(βGxE≈0.03);计时型 GWAS 已达 70 万人量级,多种睡眠/昼夜节律 PRS 已被证实可携带且有预测力;用 PRS 做 GxE 交互的方法学(IPRS、通路 PRS、I 类错误校正)已成熟。

    部分验证: “光敏感度可由多基因评分有效刻画”这一前提,目前只有单点 GWIS 与计时型 PRS 的间接支持,尚无直接构建并验证过”光敏感度 PRS”;PRS × 月相交互的预测增益从未被检验;GxE 交互项方差小、需要极大样本与良好光照测量精度,实际可检性存疑。

    待验证: 核心假设”光敏感度 PRS × 月相的交互项显著、且增量优于单基因交互”完全未被检验;PRS 分层后月相-睡眠关联是否呈单调梯度未定;跨队列可移植性(尤其非欧洲祖先人群)未知;月相与季节/纬度的共线能否被充分解耦,尚无先例可参照。

    相关阅读

    证据等级: B
    参考文献: PMID 16769177, PMID 31138694, PMID 30697806, PMID 33049062, PMID 36519390, PMID 30696823, PMID 36648941, PMID 38425306, PMID 35583112, PMID 35401709, PMID 40763299, PMID 38303123, PMID 36745570


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Light Sensitivity Polygenic Score × Lunar Phase: From Single Genes to a Comprehensive Scorecard, Can We Predict Who Is Influenced by the Moon?

    Evidence level: B

    The previous article proposed that “MTNR1B genotype may determine who is ‘touched’ by the moon,” but a single locus explains very little—light sensitivity itself is a highly heritable trait with inter-individual differences spanning an order of magnitude. Upgrading this to a “multigene score × lunar phase” interaction is the only design that truly holds for this line of inquiry.

    Main Text

    First, why we must upgrade from “single gene” to a “scorecard.” Light sensitivity is a heritable trait with massive inter-individual variation, a fact supported by hard evidence: twin studies show that melatonin suppression sensitivity under bright night light and the melatonin secretion dynamics themselves both have significant genetic components (PMID 16769177). The magnitude of this variation is staggering—under the same evening light intensity, the half-effective dose (ED50) triggering melatonin suppression can vary across an order of magnitude, ranging from tens to hundreds of lux (PMID 31138694). In other words, the extent to which “you fear light” varies between individuals is not merely a half-and-half difference. Melatonin suppression responds extremely sensitively to light and is primarily driven by the retinal melanopsin (black opsin) pathway (PMID 30697806). Systematic reviews summarize this line clearly: inter-individual differences in light sensitivity are a core source of variation in sleep and circadian rhythms (PMID 33049062).

    The problem is that the genetic basis of this “sensitivity” remains largely empty until 2022, when Burns et al. conducted a whole-genome “gene × light” interaction scan (GWIS) in 280,000 people, finally identifying a significant locus: rs3847634 in ARL14EP. Individuals carrying more minor alleles showed a stronger push toward “morningness” in response to daylight (βGxE = −0.03, SE = 0.005); this locus is also an eQTL for enhanced ARL14EP expression in both brain and retinal tissues (PMID 36519390). This result is extremely important, but note the tiny effect size: the interaction term β for a single locus is only on the order of 0.03. GWIS enrichment analysis also suggests that light sensitivity loci cluster within the G-protein coupled glutamate receptor signaling pathway and in Per2-expressing hypothalamic (SCN) neurons—meaning the true driver is the collective force of a set of loci, not a single “moon gene.”

    This leap from “single point” to “whole genome” has already been completed in chronotype research. Large-scale GWAS for chronotypes have identified hundreds of loci in nearly 700,000 individuals (PMID 30696823); polygenic scores (PRS/PGS) built on this foundation have been proven portable and predictive: the difference in reported sleep duration between the highest and lowest 5% percentiles of sleep duration PRS is 22.2 minutes (PMID 36648941); morningness PGS can stably predict morningness itself (R² ≈ 1.55%) and spill over to predict depressive symptoms (PMID 38425306); even studies using targeted sequencing of 19 circadian rhythm genes constructed a “molecular chronotype” score panel with ~80% predictive validity (PMID 35583112). Methodologically, using PRS for gene-environment interaction is now a standard toolkit: IPRS directly reconstructs scores using GxE structure (PMID 35401709), pathway-level PRS is designed specifically for interactions (PMID 40763299), and systematic correction schemes for Type I error inflation and power loss in GxE PRS exist (PMID 38303123). Existing GxE analyses support directional conclusions such as “late types sleep longer in specific environments” (PMID 36745570), demonstrating that environmental modification effects can be truly detected.

    Connecting these three lines, the proposition of this topic is as follows: upgrade “light sensitivity” from a single gene to a total polygenic score, and use the interaction between this score and lunar phase to predict individual reaction intensity to moon-phase-sleep coupling. The distinction from the previous article is crucial—the previous one asked whether “the interaction direction of a specific locus amplifies or buffers,” while this one asks “whether aggregation truly improves predictive power.” This is a pure question of incremental value, not direction.

    Three layers of testing can be designed. First is a gain test: within the same longitudinal cohort, fit two models separately—”lunar phase × rs3847634″ and “lunar phase × light sensitivity PRS”—and compare ΔR², ΔAIC, and likelihood ratios. If the PRS brings no incremental gain, it indicates light sensitivity contributes no net effect to lunar phase response; this is the strongest falsification condition. Second is a quantile gradient test: if the hypothesis holds, after stratifying PRS by deciles, the strength of the lunar-phase-sleep association should show a monotonic increasing (or decreasing) gradient, rather than differences only at the extremes. Third is a portability test: PRS weights must come from summary statistics of GWIS in independent samples, then projected into the target cohort. Using the same samples to select loci and calculate scores will systematically overestimate effects due to the winner’s curse. The cohort requirements are not low: each person must cover at least three complete lunar cycles (approx. 3 months) of wrist activity records or PSG, combined with genotyping and calculable proxies for nocturnal light exposure (geographic location + cloud cover + measured indoor illuminance).

    The biggest trap on this route is not statistics but confounding. Lunar phase drifts systematically with season and latitude due to moonrise/moonset times; therefore, the lunar phase variable is naturally collinear with season, daylight duration, and seasonal circadian drift. If one merely labels by phase without controlling for season and latitude, seasonal effects are easily misread as lunar effects. Furthermore, measurement error in light exposure systematically dilutes the interaction term (attenuation), and cross-ancestry portability of PRS remains unresolved. Therefore, the passing grade for this test is not just “significant,” but significant after controlling for season/latitude/cloud cover, with incremental gain superior to single-gene interaction, and replicated in independent cohorts.

    Regarding the layer of “Heaven and Man Correspondence,” the landing point of this design is actually quite plain: ancient texts discuss “Eight Characters endowment” (Bazi), never referring to the auspicious or inauspicious nature of a single character, but rather the overall pattern of the entire chart. Light sensitivity is the same—it is not a switch, but a continuum determined by the collective force of hundreds of loci. How much you are “touched” by the moon does not depend on whether you possess that specific “moon gene,” but rather which segment of the spectrum your scorecard falls into.

    Evidence and Limitations

    Verified: Light sensitivity is a heritable trait (both melatonin suppression sensitivity and secretion dynamics show significant heritability); inter-individual differences in light sensitivity are massive (ED50 spans an order of magnitude), melatonin suppression is extremely sensitive and melanopsin-dominated; GWIS has located light sensitivity loci (ARL14EP rs3847634; enriched in glutamate signaling and Per2+ SCN neurons) but single-point effect sizes are negligible (βGxE≈0.03); chronotype GWAS has reached the 700,000-person scale, and multiple sleep/circadian PRS have been proven portable and predictive; methodological tools for GxE interaction using PRS (IPRS, pathway PRS, Type I error correction) are mature.

    Partially Verified: The premise that “light sensitivity can be effectively characterized by a polygenic score” currently has only indirect support from single-point GWIS and chronotype PRS, with no direct construction and validation of a “light sensitivity PRS”; predictive gain of PRS × lunar phase interaction has never been tested; GxE interaction terms have small variance, requiring massive samples and high light measurement precision, making practical detectability questionable.

    To Be Verified: The core hypothesis “interaction term of light sensitivity PRS × lunar phase is significant and incrementally superior to single-gene interaction” remains completely untested; whether a monotonic gradient exists for lunar-phase-sleep association after PRS stratification is undetermined; cross-cohort portability (especially in non-European ancestral populations) is unknown; whether collinearity between lunar phase and season/latitude can be fully decoupled has no precedents to reference.

    Related Reading

    Evidence Level: B
    References: PMID 16769177, PMID 31138694, PMID 30697806, PMID 33049062, PMID 36519390, PMID 30696823, PMID 36648941, PMID 38425306, PMID 35583112, PMID 35401709, PMID 40763299, PMID 38303123, PMID 36745570


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 「少照一会儿」还是「照暗一点」:时长与强度这两根杠杆,取决于你站在曲线的哪一段

    Evidence level: B

    “时长比强度重要”这句话,只在强度已经饱和的那一小段区间里成立。互易律在两端都会断:阈值以下,时间买不到强度;饱和以上,强度买不到时间。文献之所以看起来互相矛盾,是因为它们各自站在曲线的不同段落上说话。

    正文

    月相-月经耦合被夜间光照慢慢磨掉,这一点前几篇已经立住。接下来必须回答一个更实际的问题:如果要保,先关哪一个开关——是”少照一会儿”,还是”照暗一点”? 这两个动作在生活中可以分开执行(睡前把灯调暗,或者早点关灯),但背后的生理含义完全不同:如果起作用的是剂量总量,两者可以互换;如果起作用的是剂量怎么切分,它们就不能互换。

    文献给出的第一印象很干脆:时长赢,强度输。 Dewan 等人(PMID 21532952)把 56 名年轻受试者随机分进 9 个条件——强度取 2000/4000/8000 lux,时长取 1/2/3 小时,正交。双因素方差分析里只有时长主效应显著(F2,48=4.37,P=0.01),强度不显著(F2,48=0.88,P=0.42),而且两个因素之间没有交互。1 小时只延迟约 10 分钟,3 小时约 1.5 小时——从 1 到 3 小时剂量翻三倍,效果几乎全由”多久”带来。

    但这个结论有一个被普遍忽略的前提。他们测的三档强度,全部落在反应的饱和区之上。 Zeitzer 等人(PMID 10922269)用 6.5 小时光照做出过标准的 S 形剂量-效应曲线:褪黑素抑制半饱和点约 105 lux、饱和点约 200 lux;相位延迟半饱和约 120 lux、饱和约 550 lux。也就是说,从 2000 lux 往上,强度这根杠杆已经推到底了——再加亮度,机器不再响。在这种条件下测出”强度不重要”,等于问一个已经满格的电池”再多充一点会不会更满”。那不是自然规律,那是天花板的形状。

    把镜头移到饱和区以下,结论立刻翻面。McIntyre 等人(PMID 2761346)用 200/400/600 lux 三个低强度各照 3 小时,发现每一档强度都在 1 小时内就达到它自己的最大抑制,之后继续延长时间毫无作用——原文明确写”强度与时长之间不存在互易关系”。Reiter 等人在仓鼠上的结果更极端(PMID 6545810):0.019 µW/cm² 照 20 分钟不抑制褪黑素,把时长加到 30、60、120、180 分钟仍然不抑制。在阈值以下,时间买不到强度。 这是与 Dewan 方向完全相反的失效。

    第三块证据来自另一头。Joyce 等人(PMID 35259981)用 2 毫秒闪光、把强度从 3 lux 拉到 9500 lux,测出清晰的剂量-效应,甚至带阈值样行为;而把闪光时长从 10 微秒拉到 10 秒、强度固定 2000 lux 时,时长几乎没有效果。同样,Rahman 等人(PMID 28405608)用 15 秒和 2 分钟的 9500 lux 脉冲,两组的相位延迟没有差别。在极短时程里,强度的杠杆还在陡坡上,时长的杠杆反而先平了。

    三段证据拼起来,图像就唯一了。互易律(强度×时长=恒定剂量)不是普遍规律,它只在中间一段成立。 下界以下:剂量不足以触发反应,加时间无效(Reiter、McIntyre)。上界以上:反应已饱和,加强度无效(Dewan、Zeitzer)。只有夹在两端之间的那个”工作区间”里,两根杠杆才近似等价——对褪黑素抑制大约落在几十到两百 lux,对相位延迟大约落在十几到五百多 lux。这恰好是普通室内照明的亮度区间,也是我们每天真实生活的区间。

    更麻烦的是,这个区间自己还在动。Sharma 等人(PMID 9919688)在实验鼠上测出,达到饱和所需的临界强度随昼夜相位不同而不同:CT20 约 100 lux,CT15 约 500 lux。“谁更重要”没有统一答案,因为饱和点本身随时间漂移。

    所以真实生活的判决实验不能只做单因素。可执行的设计是纵向队列按”夜间光照时长×强度”四象限分层(短暗/短亮/长暗/长亮),以月相-月经同步度为结局,用双因素交互模型直接比较两个主效应;光照用佩戴式剂量计连续记录,剂量单位用黑视素等效照度而非 lux(PMID 32248548)。但必须同时处理两个混杂:一是光历史——一周昏暗环境会让人对后续光照的敏感性增强 60-70%(PMID 21224217、PMID 12390501),所以”今晚的光”不是独立事件;二是象限本身不独立——长时程曝光的人往往也是亮环境的人,观测性四象限天然相关,只有随机分配才拆得开。真实数据已证明这种拆分的必要:佩戴式剂量计的人群研究里,大多数人白天光照不足、睡前光照过量(PMID 37812713),两头的偏离是同时发生的。

    对”天人相应”的意义,古人其实早写下了同一句话,只是写在了没有饱和区的地方。《素问·阴阳应象大论》说”少火之气壮,壮火之气衰……壮火食气,少火生气“——同一把火,小的是养,大的是耗,作用方向随剂量翻转。这正是倒 U 与饱和曲线的古典表述:不是”越多越坏”,而是存在一个转折点,过了它,同一个东西换了性质。而《素问·宣明五气》的五劳所伤”久视伤血,久卧伤气,久坐伤肉,久立伤骨,久行伤筋“,五条全部用”久”作自变量,且每一条”久”伤的是不同的东西。古人的病理模型把时长当成主变量——这不是巧合:在烛火与油灯的世界里,强度被天光封顶,室内照度根本够不到饱和点,所以唯一能拉动的杠杆就是时间。今天我们之所以要重算这笔账,是因为 LED 第一次让普通人把强度推到了饱和区之上——古人没见过的那个区间,正是”强度不重要”这个结论诞生的地方。

    「少照一会儿」还是「照暗一点」,不是一个可以拍脑袋回答的问题——它取决于你现在站在曲线的哪一段。 而在两端的失效区里,任何”用另一边补回来”的直觉,都是错的。

    证据与局限

    已验证: 在强度处于饱和区(2000-8000 lux)时,光照时长显著影响相位延迟(1→3 小时,1 小时约 10 分钟、3 小时约 1.5 小时),而强度主效应不显著,两因素无交互(PMID 21532952);10,000 lux 下 0.2-4.0 小时的时长-效应呈非线性,每分钟效率以最短时长最高(短时长单位效率是 4 小时组的 5 倍以上,PMID 22526883);褪黑素抑制的 S 形剂量-效应曲线半饱和约 105 lux、饱和约 200 lux(PMID 10922269);极低照度下延长时间不能产生抑制(仓鼠实验,PMID 6545810);低强度(200-600 lux)各档均在 1 小时内达到自身最大抑制,延长无进一步作用(PMID 2761346);2 毫秒闪光下强度呈清晰剂量-效应而时长近乎惰性(PMID 35259981);15 秒与 2 分钟脉冲的相位延迟无差异(PMID 28405608)。

    部分验证: 饱和点随昼夜相位漂移(CT20 约 100 lux、CT15 约 500 lux,PMID 9919688),故”哪根杠杆重要”具有相位依赖性;两周昏暗史可使后续光照的相位延迟效应增强 60-70%(PMID 21224217、PMID 12390501),说明单次曝光不独立;真实生活中白天不足、夜间过量的偏离同时发生(PMID 37812713)。

    待验证: 上述全部为褪黑素抑制与相位延迟的剂量学结论,尚无任何研究以月相-月经同步度为结局比较时长与强度两个主效应;”工作区间”(两根杠杆近似等价的那一段)的上下边界只有褪黑素/相位层面估值,未在耦合结局上标定;四象限观测分层能否拆开时长与强度的天然相关(须随机分配验证);黑视素等效照度作为剂量单位在长期居家测量中的等效性(PMID 32248548 基于实验室短期刺激)。

    相关阅读

    证据等级: B
    参考文献: PMID 21532952, PMID 22526883, PMID 28405608, PMID 35259981, PMID 2761346, PMID 6545810, PMID 10922269, PMID 9919688, PMID 21224217, PMID 12390501, PMID 37812713, PMID 32248548


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    \”Shine for a shorter time\” or \”Dim the light\”: duration and intensity as levers depend on where you stand on the curve

    Evidence level: B

    “Duration is more important than intensity” holds true only within a small interval where intensity is already saturated. The law of reciprocity fails at both ends: below the threshold, time cannot purchase intensity; above saturation, intensity cannot purchase time. Apparent contradictions in literature arise because sources speak from different segments of the curve.

    Main Text

    The erosion of moon-phase-menstrual coupling by nocturnal illumination has been established in previous articles. The next practical question must be answered: If preservation is the goal, which switch should be turned off first—”Shao Zhao Yi Hui R” (reduce duration slightly) or “Zhao An Dian” (dim the light slightly)? These actions can be executed separately in daily life (dimming lights before bed or turning them off earlier), but their physiological implications differ fundamentally: if the mechanism relies on total dose, they are interchangeable; if it relies on dose partitioning, they are not.

    The initial impression from the literature is straightforward: Duration wins, intensity loses. Dewan et al. (PMID 21532952) randomly assigned 56 young subjects to nine conditions—intensity levels of 2000/4000/8000 lux and durations of 1/2/3 hours, orthogonal. In the two-factor ANOVA, only the main effect of duration was significant (F2,48=4.37, P=0.01), while intensity was not (F2,48=0.88, P=0.42), and no interaction between the two factors was found. A 1-hour exposure delayed phase by approximately 10 minutes; 3 hours delayed it by about 1.5 hours—tripling the dose from 1 to 3 hours yielded effects driven almost entirely by “how long.”

    However, this conclusion relies on a premise often overlooked. The three intensity levels tested all fell above the saturation zone of the response. Zeitzer et al. (PMID 10922269) established a standard S-shaped dose-response curve using 6.5 hours of illumination: melatonin suppression half-saturation at ~105 lux and saturation at ~200 lux; phase delay half-saturation at ~120 lux and saturation at ~550 lux. Thus, above 2000 lux, the intensity lever is already pushed to the bottom—adding more brightness yields no further response. Measuring “intensity doesn’t matter” under these conditions is akin to asking a fully charged battery if adding more charge will make it fuller. That is not a natural law; that is the shape of the ceiling.

    Shifting the lens to below the saturation zone flips the conclusion immediately. McIntyre et al. (PMID 2761346) used three low intensities (200/400/600 lux) for 3 hours each and found each intensity level reached its maximum suppression within 1 hour; extending time further had no additional effect—the text explicitly states “no reciprocal relationship exists between intensity and duration.” Reiter et al. found even more extreme results in hamsters (PMID 6545810): 0.019 µW/cm² for 20 minutes failed to suppress melatonin; extending duration to 30, 60, 120, and 180 minutes still failed to suppress. Below the threshold, time cannot purchase intensity. This is a failure completely opposite to Dewan’s direction.

    Third evidence comes from the other end. Joyce et al. (PMID 35259981) used 2-millisecond flashes, increasing intensity from 3 lux to 9500 lux, revealing a clear dose-response with threshold-like behavior; conversely, extending flash duration from 10 microseconds to 10 seconds at a fixed intensity of 2000 lux showed little to no effect of duration. Similarly, Rahman et al. (PMID 28405608) used 9500 lux pulses of 15 seconds and 2 minutes with no difference in phase delay between groups. In ultra-short durations, the intensity lever remains on a steep slope, while the duration lever flattens first.

    Piecing together these three segments yields a unique image. The law of reciprocity (intensity × duration = constant dose) is not a universal rule; it holds only in the middle segment. Below the lower bound: insufficient dose to trigger response, adding time is ineffective (Reiter, McIntyre). Above the upper bound: response is saturated, increasing intensity is ineffective (Dewan, Zeitzer). Only within the “working interval” between these two bounds are the two levers approximately equivalent—melatonin suppression roughly falls in the tens to hundreds of lux range, while phase delay falls in the tens to over-hundreds of lux range. This coincides with the brightness range of ordinary indoor lighting and our daily lived reality.

    Complicating matters further, this interval itself is dynamic. Sharma et al. (PMID 9919688) found in experimental mice that the critical intensity required to reach saturation varies with circadian phase: ~100 lux for CT20, ~500 lux for CT15. “Which lever is more important” has no unified answer because the saturation point drifts with time.

    Therefore, a real-life adjudication experiment cannot rely on single-factor designs. A feasible design is a longitudinal cohort stratified by the four quadrants of “nocturnal illumination duration × intensity” (short/dim, short/bright, long/dim, long/bright), using moon-phase-menstrual synchronization as the outcome, employing a two-factor interaction model to directly compare the two main effects; illumination should be recorded continuously via wearable dosimeters, using melanopic equivalent lux rather than standard lux as the dose unit (PMID 32248548). However, two confounders must be addressed simultaneously: first, light history—a week in dim environments enhances sensitivity to subsequent illumination by 60-70% (PMID 21224217, PMID 12390501), so “tonight’s light” is not an independent event; second, the quadrants themselves are not independent—those with long-term exposure often also reside in bright environments, making observational quadrants naturally correlated, separable only through randomization. Real data confirms the necessity of this separation: population studies with wearable dosimeters show most individuals have insufficient daytime light but excessive pre-bedlight (PMID 37812713), with deviations at both ends occurring simultaneously.

    Significance for “Tian Ren Xiang Ying” (Harmony between Heaven and Man): Ancient texts actually recorded the same sentence, but in a context without a saturation zone. Su Wen · Yin Yang Ying Xiang Da Lun states “Shao huo zhi qi zhuang, zhuang huo zhi qi shuai… Zhuang huo shi qi, shao huo sheng qi” (Weak fire strengthens qi; excessive fire depletes it… Excessive fire consumes qi, weak fire generates qi)—the same fire, small nourishes, large depletes, with the direction of action flipping with dose. This is the classical expression of the inverted U and saturation curves: not “more is worse,” but there exists a turning point where the same thing changes its nature. Meanwhile, Su Wen · Xuan Ming Wu Qi lists five injuries from prolonged activity: “Jiu shi shang xue, jiu wo shang qi, jiu zuo shang rou, jiu li shang gu, jiu xing shang jin” (Prolonged viewing harms blood, prolonged lying harms qi, prolonged sitting harms flesh, prolonged standing harms bone, prolonged walking harms tendons)—all five use “prolonged” as the independent variable, and each “prolonged” injury affects a different organ. The ancient pathological model treats duration as the primary variable—not coincidence: in a world of candles and oil lamps, intensity was capped by daylight, indoor illuminance never reached saturation, so the only lever to pull was time. We must recalculate this ledger today because LEDs first allowed ordinary people to push intensity above the saturation zone—the interval unseen by ancients is precisely where the conclusion “intensity doesn’t matter” was born.

    “Shao Zhao Yi Hui R” or “Zhao An Dian” is not a question answerable by guesswork—it depends on which segment of the curve you currently stand on. In the failure zones at both ends, any intuition that “compensate with the other side” is incorrect.

    Evidence and Limitations

    Verified: At intensities in the saturation zone (2000-8000 lux), illumination duration significantly affects phase delay (1→3 hours: 1 hour ~10 mins, 3 hours ~1.5 hours), while the main effect of intensity is not significant with no interaction between factors (PMID 21532952); at 10,000 lux, duration-effectiveness from 0.2-4.0 hours is non-linear, with efficiency per minute highest at shortest duration (short-duration unit efficiency is over 5 times that of the 4-hour group, PMID 22526883); S-shaped melatonin suppression dose-response curve has half-saturation at ~105 lux and saturation at ~200 lux (PMID 10922269); extending time under very low illumination fails to produce suppression (hamster study, PMID 6545810); low intensities (200-600 lux) reach their own maximum suppression within 1 hour with no further effect from extension (PMID 2761346); under 2-millisecond flashes, intensity shows clear dose-response while duration is nearly inert (PMID 35259981); no difference in phase delay between 15-second and 2-minute pulses (PMID 28405608).

    Partially Verified: Saturation point drifts with circadian phase (CT20 ~100 lux, CT15 ~500 lux, PMID 9919688), making “which lever matters” phase-dependent; two weeks of dim history can enhance phase delay effects from subsequent illumination by 60-70% (PMID 21224217, PMID 12390501), indicating single exposure is not independent; deviations in real life (insufficient daytime, excessive nighttime) occur simultaneously (PMID 37812713).

    To Be Verified: All above are dose-response conclusions regarding melatonin suppression and phase delay; no study yet uses moon-phase-menstrual synchronization as an outcome to compare the main effects of duration and intensity; upper and lower bounds of the “working interval” (where levers are approximately equivalent) are estimated only at melatonin/phase levels, not calibrated at the coupled outcome; whether four-quadrant observational stratification can disentangle the natural correlation between duration and intensity (requires randomization verification); equivalence of melanopic equivalent lux as a dose unit in long-term home measurement (PMID 32248548 based on laboratory short-term stimulation).

    Related Reading

    Evidence Level: B
    References: PMID 21532952, PMID 22526883, PMID 28405608, PMID 35259981, PMID 2761346, PMID 6545810, PMID 10922269, PMID 9919688, PMID 21224217, PMID 12390501, PMID 37812713, PMID 32248548


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.