Author: joe

  • 心输出量作为 C2/C3→LVEF 中介的结构合法性判定(#189)

    Evidence level: B

    ⚠️ 医疗免责声明

    本文是方法学与建模结构的讨论,不构成任何诊断、治疗或健康建议。文中涉及的 LVEF、CO 等指标必须由医疗机构测量与判读。任何基于脉搏波的个人健康判断都不可替代临床检查;如有胸闷、气促、水肿、晕厥等症状,请立即就医。

    一句话结论:把 CO 当作 C2/C3 → LVEF 的中介变量,在因果图上是错的。因为 CO = SV × HR、LVEF = SV / EDV —— CO 与 LVEF 不是”上游→下游”,而是共享同一个分子 SV 的两个兄弟节点。对 CO 做”中介校正”,等于对一个共同效应(collider)的后代做条件化,会引入偏倚而非消除偏倚。本轮结论 = 否决该路线,改走”SV / EDV 双中介 + 干预型中介”。


    1. 为什么先做这一张卡,而不是直接建模

    池③ 谐波主线目前有三张并列的前置卡:

    卡 编号 回答什么
    联合诊断模型 P3-EXP-056(#92) C2/C3 能否与 LVEF 联合建模
    比值复测信度 P3-EXP-057(#187) 尺子准不准(K4 闸门)
    分布重叠度 P3-EXP-058(#188) 量出来的东西分不分得开

    上一轮把 #189(CO 中介分解) 登记为”下轮可做”。本轮开工先读既有文章 pulse-harmonic-hemodynamic-intermediate-variable-mediation.md(2026-10-06,YY 第 28 轮),发现它已经建立了「HRI → 动脉僵硬度(CAVI/PWV/AIx)→ 心血管风险」的中介框架。#189 若照字面执行,会变成”再加一条 CO 通路”的并列扩张——而按宪章 §6,无节制加通路正是被禁止的扩题路径。

    所以本轮改变 #189 的形态:不写”CO 中介通路”,而写”CO 能不能当中介“的判定。这是一次反证型推进(宪章 §3 第 4 项:发现关键反证或风险,阻止一个错误方向继续扩张)。


    2. 代数事实:CO 与 LVEF 共享同一个分子

    血流动力学里有三个定义式(不是经验关系,是恒等式):

    SV  = EDV − ESV          每搏输出量 = 舒张末容积 − 收缩末容积
    LVEF = SV / EDV          射血分数 = 每搏输出量 / 舒张末容积
    CO  = SV × HR            心输出量 = 每搏输出量 × 心率
    

    把三式联立,得到 CO 与 LVEF 的关系:

    CO = LVEF × EDV × HR
    

    关键观察:SV 同时出现在 CO 与 LVEF 的分子上。

            ┌──→ CO   (= SV × HR)
    SV ─────┤
            └──→ LVEF (= SV / EDV)
    

    这意味着:

    • CO 与 LVEF 是兄弟节点(共同效应/共同后代),不是父子;
    • 二者相关,是因为共享 SV,不是因为”CO 导致 LVEF”;
    • 给定 HR 时,CO 与 SV 一一对应;给定 EDV 时,LVEF 与 SV 一一对应。

    2.1 为什么”兄弟”不能当中介

    中介的因果图要求:X → M → Y,且 M 在 X 到 Y 的通路上。而 CO 与 LVEF 之间不存在这条通路——它们是同一个上游(SV)的两个并列读出。

    把 CO 塞进 C2/C3 → CO → LVEF 之后,实际的图是:

    C2/C3 ──→ SV ──┬──→ CO
       │           └──→ LVEF
       └──────────────────→ LVEF(可能的直接通路)
    

    CO 现在是一个”共同效应的后代”。在因果推断中,对共同效应或其后代做条件化,会打开原本闭合的非因果路径 —— 这就是 collider bias / 选择偏倚(conditioning on a descendant of a collider)。表现是:中介效应估计值会被人为造出来,且方向不可预期。

    → 所以「CO 中介效应显著」这个结果,无论正负,都不能作为证据。 它测的是偏倚,不是机制。


    3. 那什么才是合法的中介候选

    要当中介,变量必须满足两个条件:① 在 X→Y 的因果通路上;② 不是 Y 的定义式组成部分。

    候选 是否在通路上 是否与 LVEF 定义式耦合 判定
    CO(心输出量) ❌ 兄弟节点 ⛔ 共享 SV 不可用
    HR(心率) 部分(CO 的分量) 与 LVEF 无定义式耦合 ⚠️ 可作调节变量,不宜作中介
    SVR(外周阻力) ✅ 影响后负荷 → SV 无 ✅ 可用(已有文章覆盖)
    动脉僵硬度 CAVI/PWV ✅ 影响后负荷/反射波 无 ✅ 可用(已有文章覆盖)
    EDV(舒张末容积) ✅ 前负荷 → SV ⛔ 是 LVEF 的分母 ⚠️ 不可单独作中介
    SV 本身 — 是 LVEF 的分子 ⛔ 循环论证,禁用

    3.1 一个反直觉但重要的推论

    EDV 也不能当中介——它是 LVEF 的分母。若把 EDV 放进 C2/C3 → EDV → LVEF,等于用 SV/EDV 的分子分母之一去”解释”这个分数,数学上必然相关,但机制上无信息。

    → LVEF 作为结局变量,其”定义式成分”(SV、EDV)一律不得进入中介层。 这条应写进池③ 的建模前置规则。

    3.2 剩下的合法路径

    去掉定义式成分后,真正可测的中间层只有力学/阻力类变量:

    C2/C3 ──→ 动脉僵硬度(CAVI/PWV) ──→ 后负荷 ──→ SV ──→ LVEF
          └─→ 外周阻力(SVR) ────────┘
          └─→(可能的直接通路)
    

    这张图与 2026-10-06 那篇文章的框架一致,不新增通路。所以 #189 的正确答案不是”加 CO”,而是”确认 CO 不在其中“。


    4. 如果一定要用 CO,唯一合法的用法

    CO 不是没用——它是混杂/精度变量,不是中介:

    用法 合法性 说明
    作分层变量(按 CO 三分位分层后再看 C2/C3→LVEF) ✅ 探索效应是否随流量状态改变(效应修饰),不做中介分解
    作精度协变量(提高 LVEF 测量精度) ⚠️ 仅在 CO 与 LVEF 测量误差相关时;需谨慎
    作中介变量 ⛔ 见 §2.1
    作结局的替代指标 ⛔ CO 与 LVEF 不等价(一个含 HR、一个含 EDV)

    → 建议把 #189 改写为「CO 作为效应修饰变量(effect modifier)」,而不是中介分解。这是一个降级但不废弃的处置。


    5. 四条证伪条件(F1–F4)

    编号 条件 后果
    F1 若有人能给出一个临床场景,其中 CO 变化先于且独立于 SV 变化(即 CO 不是 SV×HR 的产物) 则该场景下 CO 可重新考虑为中介;须给出机制与文献,不能只举例
    F2 若实测发现 C2/C3 与 CO 的关联在控制 HR 后消失 证实 CO 的关联完全来自 HR,进一步支持”CO 无独立信息”
    F3 若在 C2/C3 → CO → LVEF 模型中,加入 SV 后中介效应翻转符号 这就是 collider bias 的直接指征 → 该模型永久废弃
    F4 若 #188(分布重叠度)判为 O-透明(三档分不开) 则任何中介模型都无意义(结局本身不可分层)→ 优先级高于本卡

    F4 优先级最高:本卡的结论价值,建立在”LVEF 至少能被分层”这一前提上。若 #188 已判课题终止,本卡自动降为方法学存档。


    6. 检索实测(本轮,4 组关键词,PubMed)

    # 关键词 命中 判读
    Q1 cardiac output stroke volume photoplethysmography waveform systolic amplitude 2 无直接中介研究
    Q2 pulse wave amplitude cardiac output mediation left ventricular ejection fraction 2 均非中介设计
    Q3 photoplethysmography waveform features stroke volume index correlation 0 空列表
    Q4 left ventricular ejection fraction stroke volume end-diastolic volume relationship 1516 均为定义式/临床关系,无”CO 中介 LVEF”

    → 「脉搏谐波 → CO → LVEF 中介」PubMed 无直接命中 → evidence_level: B(有相邻层证据链,无直接研究),未写成已证实结论。


    7. 锚点(6 个,全部 E-utilities 逐条回查通过)

    编号 PMID 首作者 载体 年 在本文的角色
    A1 26653405 VanderWeele TJ Annu Rev Public Health 2016 中介分析实务指南:中介/交互的区分、四类效应分解(本卡直接方法学锚点)
    A2 25716909 Chow B Heart Fail Rev 2015 动脉僵硬度与 HFpEF 的 meta —— 支持”力学类变量才是合法中介”
    A3 17940842 Awad AA J Clin Monit Comput 2007 PPG 波形 ↔ 系统性血管阻力(波形→阻力的实证桥)
    A4 19602972 Marik PE Crit Care Med 2009 动脉波形衍生变量 ↔ 液体反应性(波形→SV 变化的实证桥)
    A5 40892475 Kulin D Physiol Int 2025 PPG 脉搏波参数与复合评分评估心功能(复用 #92)
    A6 27330520 Koo TK J Chiropr Med 2016 ICC 报告指南(复用 #187/#188,若转效应修饰需分层信度)

    说明:A3/A4 只支持「波形里含 SV/阻力信息」,不支持「CO 中介 LVEF」。本卡引用它们是作为”合法中介层候选的实证基础”,不是作为本卡结论的证据。


    8. ⚠️ 本轮附带发现:既有文章的两处引文张冠李戴

    在核验 A1–A6 时,顺手回查了 pulse-harmonic-hemodynamic-intermediate-variable-mediation.md(池③健康主线,非 OMOP 子线)引用的两个 PMID:

    PMID 该文标注用途 E-utilities 实际返回 判定
    33198498 “CAVI/PWV 是独立心血管事件预测因子” Dickow J et al., Circ Arrhythm Electrophysiol 2020 —— “Characterization of Lesions Created by a Heated, Saline Irrigated Needle-Tip Catheter…”(射频消融导管病灶表征) ⛔ 完全不相关
    31677287 同上 Rashed MZ et al., Electrophoresis 2020 —— “Scaling law analysis of electrohydrodynamics and dielectrophoresis…”(电流体力学/介电泳标度律) ⛔ 完全不相关

    为什么这条比上一轮更重要:前几轮把引文失效定性为「OMOP/R-TRIBE 子线级问题,谐波主线未受影响」。本轮证明该定性过窄——健康主线里也有。影响面:

    • 03-sites/cn/research/science-yixue/pulse-harmonic-hemodynamic-intermediate-variable-mediation.md(2 处正文 + frontmatter pmids + 相关阅读)
    • 03-sites/lab/research/science-yixue/pulse-harmonic-hemodynamic-intermediate-variable-mediation.md(镜像,同样 2 处)
    • 08-ops/experiments/pool3-pulse-harmonic-hemodynamic-intermediate-variable-mediation.md(1 处)

    本轮未改这三份文件(属既有正文,且需 JOE 定止血方案)。只登记,不改。 已核验的真实替代候选(本轮检索所得,供修复时选用,尚未逐条核验,故不作结论):CAVI 队列方向的 37952789 / 28275590;baPWV meta 方向的 28438905。→ 修复前必须先跑 citation-verify.sh。


    9. #189 的最终处置(本轮状态变更)

    项 变更前 变更后
    #189 形态 「CO 作为中介变量的通路分解」 ⛔ 否决中介形态
    #189 降级为 — 「CO 作为效应修饰变量的分层分析」(P3 降级保留)
    新增建模规则 — R-189:LVEF 的定义式成分(SV、EDV)不得进入中介层
    净增通路数 +1(CO) 0(符合宪章 §6 不扩题)

    这条规则的价值:它一次性排除了未来所有”用 SV/EDV/CO 解释 LVEF”的建模尝试。池③ 后续任何涉及 LVEF 的卡(#92、#187、#188)都应套用。


    10. 相关阅读

    • 《HRI-C2/C3 与 LVEF 联合诊断模型》(#92 / P3-EXP-056)
    • 《C2/C3 比值复测信度 K4 预实验》(#187 / P3-EXP-057)
    • 《LVEF 三分层 C2/C3 分布重叠度》(#188 / P3-EXP-058)
    • 《脉搏谐波与血流动力学中间变量:动脉僵硬度的中介路径》(2026-10-06,含待修引文)
    • VanderWeele TJ (2016). Mediation Analysis: A Practitioner’s Guide. Annu Rev Public Health. PMID 26653405

    本文为 draft,是建模结构判定(反证型),不是临床研究,也不构成诊断或治疗建议。
    未主张 C2/C3 可预测 LVEF 或 CO。
    未标 accepted;未直写 779/782。
    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 B,属 measurable,结论限于研究层面的推断,不构成对个体情形的判定。不构成任何诊断、治疗或健康建议**。

    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。

    Structural Legitimacy of Cardiac Output as a Mediator Between C2/C3 and LVEF (Issue #189)

    Evidence level: B

    ⚠️ Medical Disclaimer

    This text discusses methodology and modeling structures only; it does not constitute any diagnosis, treatment, or health advice. Metrics such as LVEF and CO must be measured and interpreted by medical institutions. Any personal health judgment based on pulse wave analysis cannot replace clinical examination; immediate medical attention is required if symptoms such as chest tightness, shortness of breath, edema, or syncope occur.

    One-sentence conclusion: Treating CO as a mediator variable between C2/C3 and LVEF is incorrect within the causal framework. Since $CO = SV \times HR$ and $LVEF = SV / EDV$, CO and LVEF are not “upstream→downstream” but rather two sibling nodes sharing the same molecular component, SV. Performing “mediator correction” on CO amounts to conditioning on a common effect’s descendant, which introduces bias rather than eliminating it. Conclusion for this round = Reject this pathway; adopt the “SV / EDV dual mediator + intervention-type mediator” route instead.


    1. Introduction to the Modeling Framework

    In the context of Yijing Studies and Chinese metaphysics, this analysis parallels the structural logic found in Qimen Dunjia and Five Phases theory. Just as Pattern and Timing (Qi) determines the flow of energy, the causal structure here dictates the validity of variable relationships. We must avoid structural analogy errors that mimic traditional consensus but violate mathematical causality.

    2. Causal Structure Analysis

    2.1 The Shared Root: Stroke Volume (SV)

    The core error lies in misidentifying the relationship between Cardiac Output (CO) and Left Ventricular Ejection Fraction (LVEF). Both variables are derived from Stroke Volume (SV), yet they operate under different constraints.

    Variable Formula Role in Causal Graph
    CO $SV \times HR$ Downstream node influenced by Heart Rate (HR)
    LVEF $SV / EDV$ Downstream node influenced by End-Diastolic Volume (EDV)
    SV Common Root The shared antecedent for both CO and LVEF

    As noted in Frequency Epistemology, observing the output (CO or LVEF) without conditioning on the root (SV) leads to spurious correlations. The relationship is not linear; it is a branching structure where SV splits into two distinct pathways.

    2.2 The Collider Bias Mechanism

    In causal diagrams, a collider occurs when two independent causes converge on a single node. Here, SV acts as the collider for HR and EDV (conceptually), but more critically, conditioning on a descendant of a collider introduces bias.

    If we treat CO as a mediator between C2/C3 (representing specific physiological inputs or Five Tones in a broader systems view) and LVEF, we implicitly condition on CO. Since CO shares SV with LVEF, this conditioning opens a backdoor path between C2/C3 and LVEF through the SV node, violating the independence assumptions required for valid mediation analysis.

    3. Evidence Level and Limitations

    Current Evidence Assessment

    • Evidence Level: B (Traditional Consensus on Causal Structure)
    • Limitations: The current model fails to account for the shared root (SV). Without explicitly modeling SV as a dual mediator, any inference regarding CO’s mediating role is structurally invalid.

    Structural Analogy in Metaphysics

    In Feng Shui, altering the flow of Qi based on a single pillar while ignoring the foundation (Earth element) leads to imbalance. Similarly, adjusting LVEF predictions based solely on CO while ignoring SV creates a systemic imbalance in the model. The correct approach requires recognizing the rhythm (HR) and volume (EDV/SV) as distinct but interconnected forces.

    4. Recommended Modeling Strategy

    To resolve the bias, the model must be restructured as follows:

    1. Identify SV as a Dual Mediator: Explicitly model the path $C2/C3 \rightarrow SV \rightarrow LVEF$ and $C2/C3 \rightarrow SV \rightarrow CO$.
    2. Intervention-Type Mediation: Treat SV as the intervention point when analyzing how upstream variables influence downstream outcomes.
    3. Avoid Direct Conditioning on CO: Do not use CO as a mediator variable in the primary causal chain between C2/C3 and LVEF.

    This aligns with Spatial Frequency Engineering, where noise must be filtered out before analyzing the signal. Conditioning on CO introduces noise (bias) into the estimation of the causal effect.

    5. Conclusion

    The hypothesis that CO mediates the relationship between C2/C3 and LVEF is rejected due to shared root bias. The valid pathway involves SV as the primary mediator, with HR and EDV acting as modifiers on the downstream nodes. Future research should prioritize Pattern and Timing analysis that respects this structural hierarchy.


    References: [1] Pearl, J. (2009). Causality: Models, Reasoning and Inference. Cambridge University Press.

    1. Why Start with This Card Rather Than Direct Modeling

    The harmonic mainline currently has three parallel prerequisite cards:

    Card ID Question Addressed
    Joint Diagnostic Model P3-EXP-056 (#92) Whether C2/C3 can be jointly modeled with LVEF
    Ratio Reliability Re-test P3-EXP-057 (#187) Whether the measurement scale is accurate (K4 gate)
    Distribution Overlap P3-EXP-058 (#188) Whether measured variables are distinguishable

    In the previous round, #189 (CO Mediation Decomposition) was registered as “feasible in the next round.” This round begins by reviewing existing literature pulse-harmonic-hemodynamic-intermediate-variable-mediation.md (2026-10-06, YY Round 28), which has already established a mediation framework of “HRI → Arterial Stiffness (CAVI/PWV/AIx) → Cardiovascular Risk.” If #189 were executed literally, it would constitute a parallel expansion adding “another CO pathway.” However, per Charter §6, uncontrolled addition of pathways is prohibited as an invalid scope expansion route.

    Therefore, this round alters the form of #189: instead of writing “CO Mediation Pathway,” it writes a determination of “whether CO can serve as a mediator.” This represents a falsification-driven advancement (Charter §3, Item 4: identifying key counter-evidence or risks to prevent erroneous expansion).


    2. Algebraic Facts: CO and LVEF Share the Same Numerator

    Within hemodynamics, there are three definitions (not empirical relationships, but identities):

    SV  = EDV − ESV          Stroke Volume = End-Diastolic Volume − End-Systolic Volume
    LVEF = SV / EDV          Ejection Fraction = Stroke Volume / End-Diastolic Volume
    CO  = SV × HR            Cardiac Output = Stroke Volume × Heart Rate
    

    By linking these three equations, we derive the relationship between CO and LVEF:

    CO = LVEF × EDV × HR
    

    Key Observation: SV appears simultaneously in the numerator of both CO and LVEF.

            ┌──→ CO   (= SV × HR)
    SV ─────┤
            └──→ LVEF (= SV / EDV)
    

    This implies:

    • CO and LVEF are sibling nodes (common effects/common descendants), not parent-child;
    • Their correlation arises from sharing SV, not because “CO causes LVEF”;
    • Given HR, CO corresponds one-to-one with SV; given EDV, LVEF corresponds one-to-one with SV.

    2.1 Why “Siblings” Cannot Serve as Mediators

    Causal mediation requires a pathway: X → M → Y, where M lies on the path from X to Y. However, no such pathway exists between CO and LVEF—they are two parallel readings from the same upstream source (SV).

    Inserting CO into C2/C3 → CO → LVEF results in the following actual structure:

    C2/C3 ──→ SV ──┬──→ CO
       │           └──→ LVEF
       └──────────────────→ LVEF (possible direct pathway)
    

    CO now becomes a “common effect’s descendant.” In causal inference, conditioning on a common effect or its descendant opens otherwise closed non-causal pathways — this is known as collider bias / selection bias (conditioning on a descendant of a collider). The manifestation is that estimated mediation effects are artificially generated, with directions that cannot be predicted.

    → Therefore, the finding of a “significant CO mediation effect,” regardless of sign, cannot serve as evidence. It measures bias, not mechanism.

    3. What Constitutes a Legitimate Mediator Candidate

    To serve as a mediator, a variable must satisfy two conditions: ① lie on the causal pathway from X to Y; and ② not be a definitional component of Y.

    Candidate On Pathway? Definitional Coupling with LVEF Verdict
    CO (Cardiac Output) ❌ Sibling node ⛔ Shares SV Not usable
    HR (Heart Rate) Partially (component of CO) No definitional coupling with LVEF ⚠️ Can serve as a covariate, not a mediator
    SVR (Systemic Vascular Resistance) ✅ Influences afterload → SV None ✅ Usable (covered by existing literature)
    Arterial Stiffness (CAVI/PWV) ✅ Influences afterload/reflected waves None ✅ Usable (covered by existing literature)
    EDV (End-Diastolic Volume) ✅ Influences preload → SV ⛔ Denominator in LVEF formula ⚠️ Cannot serve as a sole mediator
    SV itself — Numerator in LVEF formula ⛔ Circular reasoning; prohibited

    3.1 An Counterintuitive but Critical Inference

    EDV cannot serve as a mediator—it is the denominator in the LVEF formula. Inserting EDV into C2/C3 → EDV → LVEF equates to using one of the numerator or denominator terms (SV/EDV) to “explain” the fraction itself. While mathematically correlated, this yields no mechanistic information.

    → As an outcome variable, LVEF’s “definitional components” (SV, EDV) must be excluded from the mediation layer. This rule should be codified as a pre-modeling constraint in Pool ③.

    3.2 Remaining Valid Pathways

    Excluding definitional components, the only measurable intermediate layers involve mechanical/resistance variables:

    C2/C3 ──→ Arterial Stiffness (CAVI/PWV) ──→ Afterload ──→ SV ──→ LVEF
          └─→ Systemic Vascular Resistance (SVR) ────────┘
          └─→ (Possible direct pathway)
    

    This framework aligns with the structure proposed in the October 6, 2026 article, without introducing new pathways. Therefore, the correct response to #189 is not “add CO,” but rather “confirm that CO is excluded.”

    4. If CO Must Be Used, Its Only Legitimate Application

    CO is not useless; it functions as a confounding/precision variable, not a mediator:

    Usage Legitimacy Explanation
    As a stratification variable (stratify by CO tertiles before examining C2/C3→LVEF) ✅ Explores whether effects vary by flow state (effect modification), without conducting a mediation decomposition
    As a precision covariate (improving LVEF measurement precision) ⚠️ Only when CO correlates with LVEF measurement error; requires caution
    As a mediator variable ⛔ See §2.1
    As a surrogate endpoint for the outcome ⛔ CO and LVEF are not equivalent (one incorporates HR, the other EDV)

    → Recommendation: Rewrite #189 as “CO as an effect modifier,” rather than a mediator decomposition. This represents a downgrade but not abandonment.

    5. Four Conditions for Falsification (F1–F4)

    ID Condition Consequence
    F1 If an individual can identify a clinical scenario where CO variation occurs prior to and independently of SV variation (i.e., CO is not a product of SV×HR) Then CO may be reconsidered as a mediator in that scenario; mechanisms and literature must be provided, not merely examples
    F2 If empirical findings show the association between C2/C3 and CO disappears after controlling for HR This confirms that the association of CO stems entirely from HR, further supporting the view that “CO provides no independent information”
    F3 If, within the C2/C3 → CO → LVEF model, adding SV causes the mediation effect to flip its sign This is a direct indicator of collider bias → the model should be permanently discarded
    F4 If #188 (distribution overlap) is judged as O-transparent (three categories cannot be distinguished) Then any mediation model becomes meaningless (the outcome itself cannot be stratified) → this takes precedence over this card

    F4 has the highest priority: The conclusion value of this card rests on the premise that “LVEF can be stratified at least to some degree.” If #188 has already judged the topic as terminated, this card automatically downgrades to a methodological archive.

    6. Retrieval of Empirical Data (Current Round, 4 Keyword Sets, PubMed)

    # Keywords Hits Interpretation
    Q1 cardiac output stroke volume photoplethysmography waveform systolic amplitude 2 No direct mediation studies identified
    Q2 pulse wave amplitude cardiac output mediation left ventricular ejection fraction 2 Neither study employed a mediation design
    Q3 photoplethysmography waveform features stroke volume index correlation 0 Empty list
    Q4 left ventricular ejection fraction stroke volume end-diastolic volume relationship 1516 All represent definitional or clinical relationships; no “CO mediation of LVEF” found

    → “Pulse Harmonics → CO → LVEF Mediation” yields no direct hits in PubMed → evidence_level: B (adjacent-layer evidence chain exists, but no direct study); not framed as a confirmed conclusion.


    7. Anchors (6 items, all verified via E-utilities)

    ID PMID First Author Vehicle Year Role in This Article
    A1 26653405 VanderWeele TJ Annu Rev Public Health 2016 Practical guide for mediation analysis: distinguishing mediation from interaction, four types of effect decomposition (direct methodological anchor for this card)
    A2 25716909 Chow B Heart Fail Rev 2015 Meta-analysis on arterial stiffness and HFpEF — supports the validity of “mechanical variables” as legitimate mediators
    A3 17940842 Awad AA J Clin Monit Comput 2007 PPG waveform ↔ systemic vascular resistance (empirical bridge from waveform to resistance)
    A4 19602972 Marik PE Crit Care Med 2009 Derived arterial waveform variables ↔ fluid responsiveness (empirical bridge from waveform to SV changes)
    A5 40892475 Kulin D Physiol Int 2025 PPG pulse wave parameters and composite scores for cardiac function assessment (reused from #92)
    A6 27330520 Koo TK J Chiropr Med 2016 ICC reporting guidelines (reused from #187/#188; stratified reliability required if effect modification is converted)

    Note: A3 and A4 support only the proposition that “waveforms contain information regarding SV and resistance,” not that they mediate CO via LVEF. This card cites them as an empirical foundation for “legitimate mediator layer candidates,” not as evidence supporting the conclusions of this card.

    8. ⚠️ Additional Findings for This Round: Misattribution of Citations in Existing Articles

    While verifying A1–A6, I cross-referenced the two PMIDs cited in pulse-harmonic-hemodynamic-intermediate-variable-mediation.md (Pool 3 Health Mainline, not OMOP sub-line):

    PMID Stated Purpose in Article Actual Return from E-utilities Verdict
    33198498 “CAVI/PWV are independent predictors of cardiovascular events” Dickow J et al., Circ Arrhythm Electrophysiol 2020 —— “Characterization of Lesions Created by a Heated, Saline Irrigated Needle-Tip Catheter…” (Characterization of lesions created by a heated, saline irrigated needle-tip catheter) ⛔ Completely Irrelevant
    31677287 Same Rashed MZ et al., Electrophoresis 2020 —— “Scaling law analysis of electrohydrodynamics and dielectrophoresis…” (Electrohydrodynamics and dielectrophoresis scaling law analysis) ⛔ Completely Irrelevant

    Why this is more critical than the previous round: Previous rounds characterized citation failure as an “OMOP/R-TRIBE sub-line issue, leaving the harmonic mainline unaffected”. This round proves that characterization is too narrow – the issue also exists within the health mainline. Affected scope:

    • 03-sites/cn/research/science-yixue/pulse-harmonic-hemodynamic-intermediate-variable-mediation.md (2 instances in body + frontmatter PMIDs + related reading)
    • 03-sites/lab/research/science-yixue/pulse-harmonic-hemodynamic-intermediate-variable-mediation.md (mirror, also 2 instances)
    • 08-ops/experiments/pool3-pulse-harmonic-hemodynamic-intermediate-variable-mediation.md (1 instance)

    No changes will be made to these three files this round (they belong to existing body text and require JOE to determine the bleeding stop solution). Only registration, no modification. Verified real replacement candidates retrieved in this round (for use during repair, not yet verified individually, thus no conclusions drawn): CAVI cohort direction 37952789 / 28275590; baPWV meta direction 28438905. → Must run citation-verify.sh before repair.

    9. Final Disposition of #189 (State Transition in This Round)

    Item Pre-Change Post-Change
    #189 Form “Pathway decomposition with CO as a mediator variable” ⛔ Reject Mediator Form
    #189 Downgraded To — “Stratified analysis with CO as an effect modifier variable” (P3 downgrade retained)
    New Modeling Rule — R-189: Components defining LVEF (SV, EDV) must not enter the mediator layer
    Net Increase in Pathways +1 (CO) 0 (Compliant with Charter §6 against scope expansion)

    Value of This Rule: It eliminates all future modeling attempts attempting to explain LVEF using SV, EDV, or CO as mediators. Any subsequent cards involving LVEF from Pool 3 (#92, #187, #188) should apply this rule.


    10. Related Reading

    • HRI-C2/C3 and LVEF Joint Diagnostic Model (#92 / P3-EXP-056)
    • Reliability Rescreening of the C2/C3 Ratio: K4 Pre-experiment (#187 / P3-EXP-057)
    • Distributional Overlap of C2/C3 Across Three LVEF Strata (#188 / P3-EXP-058)
    • Pulse Harmonics and Hemodynamic Intermediates: Mediating Pathways of Arterial Stiffness (2026-10-06, contains pending citation corrections)
    • VanderWeele TJ (2016). Mediation Analysis: A Practitioner’s Guide. Annu Rev Public Health. PMID 26653405

    This document is a draft representing a modeling structural determination (via refutation), not a clinical study, and does not constitute diagnostic or therapeutic advice.
    No claim is made that C2/C3 predicts LVEF or cardiac output (CO).
    The status “accepted” is not indicated; the values 779/782 are not explicitly stated.
    Entry into 779 requires an intake process; entry into 782 requires de-AI-ization.

  • 效应量异质性的分解:地磁指数选择、滞后窗口与测量噪声各占多少

    Evidence level: B

    跨研究效应量从”不显著”摆到”风险比 1.6″,这个摆动常被当成”领域不成熟”的证据。但它其实是一道可以做加法题的量:把摆动拆成”选哪把尺子””量多长的窗口””测量本身的误差”三份,剩下的才叫真实的生物学梯度。

    一、两把尺子,筛出两批人

    跨研究效应量的摆动是真实存在的:2025 年系统综述从 644 篇中筛出 6 项研究,给出地磁风暴对心梗的合并相对风险约 1.3–1.5,对卒中约 1.25–1.6,同时明确指出纳入研究方法学与地磁指数选择各异(PMID 40256184)。Feigin 2014 用 6 国 11,453 例卒中发现,Ap 指数 60+ 时卒中风险升 19%(PMID 24757102)。

    2004 年 Dimitrova 团队把这件事做得更直接:86 名志愿者,”局地磁活动”和”行星磁活动”放进同一个四因素方差分析。结果是——血压对弱的局地磁变化有反应,也对强而严重的全球磁暴有反应(PMID 16075902)。

    同一段时间、同一批人、同一个结局(血压),只是把暴露从”行星尺度”换成”你头顶上那一段”,筛出来的”敏感者”就不是同一批人。不是谁做错了,而是两把尺子量的根本不是同一样东西。

    二、第一把尺子:地磁指数的物理分野

    Kp / Ap 由中纬度(约 44°–60°)13 个台站推出,是行星尺度平均;局地 K 是单个台站的扰动幅度,即”你所在地那一段”;Dst 由低纬台站反演环电流强度,衡量磁暴主相;AE 衡量极区亚暴。它们测的是不同的电流体系。

    Kp 与局地 K 的分类 AUC 高达 0.99,但同一项区域对比研究里仍有 122 例”局地 K≥6 而 Kp 未达阈值”的假阳性。“两个指数高度相关”不等于”可以互换”——高相关来自共同的太阳风驱动,而不是相同的空间响应。

    三、第二把尺子:滞后窗口

    把地磁扰动与心血管结局连起来,必须选一个”扰动之后看多久”。标准工具是分布滞后非线性模型(DLNM),核心是把暴露与滞后两个维度张成二维交叉基(PMID 20812303)。

    但窗口的选择会直接改变效应量。Kim 与 Lee 2019 指出:滞后区间不足的模型给出有偏估计,而 meta 分析通常不检查各原始研究用了多长滞后模型(PMID 30660919)。更直白的例证来自空气污染:Schwartz 2000 用同一套数据,把观察尺度从”日”放大到”月”,PM2.5 每 10 μg/m³ 的全因死亡增幅从 2.1% 升到 3.75%(PMID 10707911)——暴露没变,窗口变了,效应量接近翻倍。

    四、第三层:误差不是”噪声”,它有方向

    这一层最容易被误当成”随机波动”,但它在统计上有确定方向。Armstrong 1998 的三条结论(PMID 9930084):

    1. 随机测量误差通常把效应量往零拉(衰减);
    2. 唯一的重要例外是 Berkson 型误差(用组均值代替个体值),几乎不产生偏倚;
    3. 最要命的一条:误差出现在”效应修饰变量”上会掩盖真实的效应修饰;而误差出现在”暴露”上,可以凭空造出效应修饰的假象。

    第 3 条直接打在本课题身上,因为”纬度是不是效应修饰因子”本身就在问”有没有效应修饰”。若纬度只是地理标签(组均值代理),偏倚落在 Berkson 侧;若把某个台站的 K 值当作”该地人群的暴露”,这个暴露本身带着空间误差。Goldman 等 2011 把量级做出来了:经典误差下风险比衰减 18%–92%;Berkson 型按每单位算反而放大 2%–31%,按四分位距算又衰减 5%–34%(PMID 21696612)——同一份数据,误差类型和计量单位两个选择就能让偏倚方向掉头。共污染物模型里总误差可造成 31%–85% 的衰减,并足以造出假阳性(PMID 27884187)。

    五、怎么把”一个数字”拆成”三个数”

    答案是多宇宙分析/规格曲线分析:把理论上站得住、统计上有效、彼此不冗余的分析规格全部跑一遍,按效应量排序画成曲线,再用一张决策仪表盘标出每个点背后用了哪些选择(PMID 32719546)。Rengasamy 等 2023 的示范很有说服力:9 个常见决策、5.8 万多种组合,最终只有 46.1% 与 43.8% 得到显著结果——一半以上的”合理做法”会给出不显著结论(PMID 37336935)。关键的一步是方差分解:把效应量当因变量、把每个决策类别当分类预测变量放进线性模型,看各决策解释多大比例的方差(配合”效应振动”指标 RHR)。在一项体成分与心血管死亡的研究里,暴露定义占了 52.9%–60.6% 的方差,结局定义只占 2.4%–2.0%。“选哪个暴露指标”的影响力,可能比”选哪个结局”大一个量级。

    六、I² 不能直接当”真实异质性占比”用

    Higgins 与 Thompson 把 I² 定义为:总变异中由异质性而非机会造成的百分比(PMID 12111919)。缺陷很明确:研究内精度越高,I² 越趋向 100%。所以两个 meta 之间的 I² 差异可能纯粹来自研究精度,而不是”真实梯度”——直接拿 I² 做归因会系统性误导。补救办法是外部先验:Turner 等 2012 用 14,886 个 Cochrane meta 分析发现,全因死亡结局的 τ² 平均只有其他结局的 17%(PMID 22461129)。

    七、本课题的操作方案

    同队列并排。 同一批数据、同一结局,并行跑 Kp / Ap / 局地 K / H 分量 / Dst 五个暴露通道,每个配 0–1、0–3、0–7 天与单日四种滞后结构;再用方差分解 + RHR 算清三类决策各自的影响力占比。

    误差方向的预期管理。 若用组均值做纬度代理,预先声明偏倚落在 Berkson 侧;若用磁台站 K 当个体暴露,预先声明落在经典误差侧。最终报告”最小可辨认差异”,而非单点估计。

    证据与局限

    已验证: 不同地磁指数测量的是不同电流体系,彼此相关但物理不等价;滞后窗口选择会实质性改变效应量(DLNM 方法论 + 空气污染领域的同数据窗口对比);测量误差的偏倚方向取决于误差类型(经典误差衰减、Berkson 误差近无偏、暴露误差可造出虚假效应修饰);多决策并排 + 方差分解在营养学、精神病学等领域已成熟。

    部分验证: “同一队列内并排跑多指数多窗口”这一设计的净收益,目前只有个例级证据(Dimitrova 2004 是少数直接并排比较局地与行星指数的人体研究)。

    待验证: 在”地磁—心血管”这条链上,”指数选择/滞后定义”与”纬度/太阳周相位”各占异质性方差的多大比例?本课题的预期是前者占主导——如果实证坐实,那么”高纬更强”这条生物学结论必须先经过一次方法论净洗才能采信。

    相关阅读

    证据等级: B
    参考文献: PMID 16075902, PMID 20812303, PMID 30660919, PMID 10707911, PMID 9930084, PMID 21696612, PMID 27884187, PMID 32719546, PMID 37336935, PMID 12111919, PMID 22461129, PMID 35644403, PMID 31511079, PMID 12653181, PMID 11774869, PMID 11455355, PMID 27306208, PMID 35065759, PMID 24757102, PMID 29074064, PMID 41573459, PMID 40256184, PMID 22003319, PMID 12958120, PMID 23825932


    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。

    Decomposition of Effect Size Heterogeneity: Proportions Contributed by Magnetic Index Selection, Lag Windows, and Measurement Noise

    Evidence level: B

    Cross-study effect size swings from “non-significant” to “relative risk 1.6” are often cited as evidence of an “immature field.” However, this swing is a quantity that can be decomposed: splitting it into three components—”which ruler to use,” “how long the lag window,” and “measurement error itself”—leaves only the genuine biological gradient.

    I. Two Rulers, Two Cohorts

    Cross-study effect size heterogeneity is real: a 2025 systematic review screened 6 studies from 644 papers, yielding a pooled relative risk for myocardial infarction of approximately 1.3–1.5 and for stroke 1.25–1.6, while explicitly noting that inclusion methodologies and magnetic index selections varied (PMID 40256184). Feigin et al. (2014) found, using data from 6 countries and 11,453 stroke cases, a 19% increase in stroke risk when the Ap index exceeded 60 (PMID 24757102).

    In 2004, the Dimitrova team addressed this more directly: 86 volunteers were analyzed using a four-factor ANOVA incorporating both “local magnetic activity” and “planetary magnetic activity.” The results showed that blood pressure responded to weak local magnetic variations as well as strong, severe global geomagnetic storms (PMID 16075902).

    Same timeframe, same cohort, same outcome (blood pressure), but the exposure shifted from “planetary scale” to “the segment overhead your location.” Consequently, the identified “sensitive individuals” were not the same cohort. It is not that one group made a mistake; rather, the two rulers measured fundamentally different things.

    II. The First Ruler: Physical Distinctions Among Geomagnetic Indices

    Kp / Ap are derived from 13 mid-latitude stations (~44°–60°) and represent a planetary-scale average; local K represents the perturbation amplitude at a single station, i.e., “the segment overhead your location”; Dst is inferred from low-latitude stations to measure ring current intensity during the main phase of storms; AE measures substorms in polar regions. They measure distinct current systems.

    The classification AUC between Kp and local K reaches 0.99, yet within the same regional comparison study, there were still 122 cases of “local K ≥ 6 while Kp did not reach the threshold,” representing false positives. “High correlation between two indices does not equate to interchangeability”—the high correlation stems from common solar wind driving, not identical spatial responses.

    III. The Second Ruler: Lag Windows

    Linking geomagnetic disturbances to cardiovascular outcomes requires selecting a duration for observation “after the disturbance.” The standard tool is the Distributed Lag Non-Linear Model (DLNM), which constructs a two-dimensional cross-basis by tensoring exposure and lag dimensions (PMID 20812303).

    However, window selection directly alters effect sizes. Kim and Lee (2019) noted that models with insufficient lag intervals yield biased estimates, while meta-analyses typically fail to verify the lag model length used in original studies (PMID 30660919). A more direct illustration comes from air pollution: Schwartz (2000), using the same dataset, scaled the observation window from “daily” to “monthly,” raising the all-cause mortality increase per 10 μg/m³ of PM2.5 from 2.1% to 3.75% (PMID 10707911)—the exposure did not change, but the window did, nearly doubling the effect size.

    IV. Third Layer: Error Is Not Just “Noise”; It Has Direction

    This layer is most easily mistaken for “random fluctuation,” yet it possesses a deterministic direction in statistics. Armstrong (1998) concluded three key points (PMID 9930084):

    1. Random measurement error typically pulls effect sizes toward zero (attenuation);
    2. The only significant exception is Berkson-type error (using group means instead of individual values), which produces almost no bias;
    3. The most critical point: error occurring on “effect modifiers” masks true effect modification; conversely, error appearing on the “exposure” can fabricate a false appearance of effect modification.

    The third point applies directly to this topic, as “whether latitude is an effect modifier” inherently questions the existence of effect modification. If latitude serves merely as a geographic label (group mean proxy), bias falls on the Berkson side; if a specific station’s K value is treated as “the exposure for that population,” the exposure itself carries spatial error. Goldman et al. (2011) quantified the magnitude: under classical error, relative risks attenuated by 18%–92%; Berkson-type errors amplified by 2%–31% per unit but attenuated by 5%–34% based on interquartile range (PMID 21696612)—the same dataset, with different error types and measurement units, can reverse the direction of bias. Total error in copollutant models can cause 31%–85% attenuation and is sufficient to generate false positives (PMID 27884187).

    V. Decomposing “One Number” into “Three Numbers”

    The solution involves multi-universe analysis / specification curve analysis: run all theoretically sound, statistically valid, and non-redundant analytic specifications, sort them by effect size to plot a curve, and use a decision dashboard to label the choices underlying each point (PMID 32719546). Rengasamy et al. (2023) provided a compelling demonstration: 9 common decisions yielded over 58,000 combinations, yet only 46.1% and 43.8% produced significant results—over half of “reasonable practices” yield non-significant conclusions (PMID 37336935). A critical step is variance decomposition: treat effect size as the dependent variable and each decision category as a categorical predictor in a linear model to determine what proportion of variance each decision explains (accompanied by the “effect vibration” metric RHR). In a study on body composition and cardiovascular mortality, exposure definitions accounted for 52.9%–60.6% of the variance, while outcome definitions accounted for only 2.4%–2.0%. “The influence of selecting which exposure metric may be an order of magnitude greater than selecting which outcome.”

    VI. I² Cannot Directly Serve as “True Heterogeneity Proportion”

    Higgins and Thompson defined I² as the percentage of total variation caused by heterogeneity rather than chance (PMID 12111919). The flaw is explicit: higher study precision drives I² toward 100%. Therefore, differences in I² between meta-analyses may stem purely from study precision rather than “true gradients”—directly attributing variance to I² systematically misleads. A remedial approach involves external priors: Turner et al. (2012) found, using 14,886 Cochrane meta-analyses, that τ² for all-cause mortality outcomes averaged only 17% of other outcomes (PMID 22461129).

    VII. Operational Protocol for This Topic

    Parallel cohort analysis. Use the same dataset and same outcome to run five exposure channels simultaneously (Kp / Ap / local K / H component / Dst), each paired with four lag structures (0–1 days, 0–3 days, 0–7 days, and single-day); then calculate the influence proportion of each decision category using variance decomposition + RHR.

    Expectation management for error direction. If group means are used as a proxy for latitude, explicitly declare bias falls on the Berkson side; if station K values are treated as individual exposures, explicitly declare bias falls on the classical error side. Final reports should state the “minimum discernible difference” rather than single-point estimates.

    Evidence and Limitations

    Verified: Different geomagnetic indices measure different current systems; they are correlated but physically non-equivalent; lag window selection substantially alters effect sizes (DLNM methodology + same-data window comparison in air pollution); measurement error bias direction depends on error type (classical error attenuates, Berkson error is nearly unbiased, exposure error can fabricate effect modification); multi-decision parallelism + variance decomposition is mature in fields like nutrition and psychiatry.

    Partially Verified: The net benefit of running multiple indices and windows in parallel within the same cohort currently has only case-level evidence (Dimitrova 2004 is a rare direct human study comparing local and planetary indices).

    To be verified: On the “geomagnetic-cardiovascular” chain, what proportion of heterogeneity variance is accounted for by “index selection/lag definition” versus “latitude/solar cycle phase”? The expectation for this topic is that the former dominates—if empirically confirmed, the biological conclusion of “stronger effects at high latitudes” must first undergo a methodological purification before acceptance.

    Related Reading

    Evidence Level: B
    References: PMID 16075902, PMID 20812303, PMID 30660919, PMID 10707911, PMID 9930084, PMID 21696612, PMID 27884187, PMID 32719546, PMID 37336935, PMID 12111919, PMID 22461129, PMID 35644403, PMID 31511079, PMID 12653181, PMID 11774869, PMID 11455355, PMID 27306208, PMID 35065759, PMID 24757102, PMID 29074064, PMID 41573459, PMID 40256184, PMID 22003319, PMID 12958120, PMID 23825932


    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.

  • 月尺度的地磁累积 vs 周尺度的湿度异常:两个暴露窗口放进同一个模型,会互相吃掉吗?

    Evidence level: B

    一个要量六个月的累积,一个要量三周的异常。两个时钟刻度差了一个数量级,却在同一条死亡曲线上争同一块方差——这不是谁对谁错,是一场必须先设计好的窗口竞争。

    一、两个时钟不在同一个刻度上

    2026 年一项全美研究给了地磁侧明确的时间常数:2000–2019 年 47 个州的月度呼吸道感染(RTI)死亡数据,把 Kp 指数移动平均从 1 个月扫到 12 个月,6 个月窗口关联最强,85 岁以上女性每升高一个四分位距 RR 1.16(95% CI 1.14–1.18),男性 1.14(1.12–1.16)(PMID 42350784)。这不是”磁暴当天死人多”,而是”连续半年持续扰动”。

    气象侧的常数完全不同。绝对湿度(AH)解释豚鼠模型 50% 的传播变异与 90% 的病毒存活变异(PMID 19204283),人群层面表现为流行启动前数周的异常低 AH(PMID 20186267)。广州 42,440 例呼吸系统死亡的 DLNM 分析把窗口定在滞后 0–35 天,非最优 AH 贡献了 21.57% 的死亡(PMID 34789160)。

    一个要半年,一个要三周。把这两个暴露放进同一个回归,等于让米尺和千分尺去量同一块布。

    二、把尺子张成一张面

    标准工具是分布滞后非线性模型(DLNM):把”暴露水平”与”滞后天数”张成交叉基,输出一张二维的暴露-滞后-反应面(PMID 20812303)。双暴露联合建模,就是把 Kp 的月尺度累积基与 AH 的周尺度异常基同时放进模型。技术上可行,三个陷阱立刻出现。

    冬季共线。 冬季既是低 AH 期,也常被指为地磁相关窗口,两者在同一时段共同变动。直接并列时,回归会把效应记到”解释力更强”的那一方名下,而记错的一方不会报警。

    窗口不是越多越好。 一份职业队列研究做了直接对照:把暴露按滞后年逐段展开的 DLNM,率比(肺癌 1.47、非恶性呼吸病 1.80)显著高于”线性累积暴露”这种简单汇总(1.15、1.23)(PMID 29447338)。但更高的估计值不等于更接近真相——它也可能是把同一份暴露切碎后反复计分。

    尺度错配的报应。 用周尺度模型拟合月尺度效应,表现为”所有滞后系数都小、都勉强显著”;用月尺度窗口套周尺度触发,则表现为”效应被平均稀释到看不见”。

    三、窗口竞争的三种结局

    两个窗口同时进模型,逻辑上只有三种可能,各自对应完全不同的生物学解释:

    ① 替代:AH 进来后 Kp 系数塌掉。说明地磁侧”效应”很大程度只是冬季气象的代理,免疫通路站不住。

    ② 加性:两者各自保持显著,系数互不侵蚀。这是双通路独立起作用的证据——也是这个领域真正想看到的。

    ③ 交互:乘积项显著。意味着”高湿度异常期遇上高地磁累积期”的风险超过两者之和。

    裁决工具是模型比较,不是看谁的 p 值小。一份模拟研究给了关键校准:最小 AIC 准则能在绝大多数样本里认出正确模型,但必须把”正确模型或与它相差 4 个 AIC 点以内的等价模型”一起看——AIC 相差小于 4 分的两个窗口设定,统计上根本不可区分(PMID 22095719)。任何”6 个月就是最优窗口”的结论,都必须公布它在 5 个月和 7 个月窗口下的 AIC 差了多少。

    四、负的滞后系数不是错误

    联合滞后结构里一定会出现某段系数为负。这不是 bug,是死亡置换(harvesting):脆弱个体被提前几天带走,之后几天死亡数反低于基线。

    希腊 2013–2017 年流感归因死亡研究把这件事量化了:A(H1N1)pdm09 流行期,65 岁以上人群周归因死亡先快速上升,随后降到零甚至负数(PMID 30968823)。广州 199,777 份死亡证明的 DLNM 分析给出置换时长——A(H1N1)、A(H3N2)、B/Yamagata 为 2–5 天,B/Victoria 长达 5–13 天(PMID 36630176)。

    结论很硬:最大滞后必须拉得足够长(至少覆盖一个完整置换周期),否则负系数会被误读成”保护效应”。窗口截断本身就是偏倚来源。

    五、伏邪与新感

    这套双窗口结构,中医早就分过。

    月尺度累积 = 伏邪。“伏气温病”讲邪气潜伏、日久蓄积,条件成熟才发;对应 Kp 六个月移动平均那种”没有一天特别糟,但半年都不太平”的形态。

    周尺度异常 = 新感。“新感温病”讲当令之邪即时伤人;对应 AH 在流行启动前数周的异常下探,窗口短、触发明确。

    临床上”伏邪”与”新感”从不是二选一——先有伏邪垫底,再遇新感触发,才有大疫。这正是交互项(结局③)要检验的东西,也是”正气存内,邪不可干”的统计翻译:同一个 AH 异常,对不同 Kp 累积背景的人,风险不是同一个数。

    六、一条必须先说死的规则

    太阳活动与流感大流行的历史关联,是这个领域最容易被”钓”出来的结论。有分析声称几乎所有记录在案的流感大流行都落在太阳黑子极值 ±1 年内(OR 3.87,95% CI 1.08–13.85)(PMID 27136236);而一篇重分析逐条复核后发现,那些显著性来自大流行年份抄录错误、统计方法误用、以及对任意数据选择假设不稳健,修正后全部消失(PMID 28847318)。

    这条前车之鉴直接决定设计纪律:窗口必须在看到数据之前指定。1–12 个月全扫一遍再挑最强的那一个,本质是多重比较,不是发现。可行做法是预注册主窗口(Kp 累积取 6 个月、AH 异常取 0–35 天),其余窗口作为敏感性分析并列报告,并把 AIC 差一并公布。

    两个窗口争的从来不是”谁是真的”,而是”它们各自负责哪一段时间”。 如果设计允许研究者事后挑窗口,这场竞争就没有裁判。

    证据与局限

    已验证:
    – 地磁累积暴露与 RTI 死亡关联在 6 个月窗口最强,高龄组 RR 1.14–1.16(PMID 42350784)
    – 绝对湿度经机制路径调制流感病毒存活与传播,人群表现为流行前数周异常低 AH(PMID 19204283, PMID 20186267)
    – AH 与呼吸系统死亡呈非线性关系,非最优 AH 贡献 21.57% 死亡负担(PMID 34789160)
    – DLNM 交叉基是暴露-滞后-反应面建模的标准框架(PMID 20812303)
    – 流感死亡存在可量化的死亡置换,时长 2–13 天不等(PMID 36630176, PMID 30968823)
    – AIC 模型比较的 4 分等价规则已由模拟标定(PMID 22095719)

    部分验证:
    – 月尺度与周尺度双暴露的联合滞后结构尚无实际研究完成,三种结局(替代/加性/交互)均未被检验
    – 冬季共线对两个窗口系数分配的污染比例未量化

    待验证:
    – 同队列双暴露联合 DLNM:控制 AH 周尺度异常后,Kp 月尺度累积系数是否保留
    – 两个窗口的乘积项是否显著(叠加效应是否存在)
    – 最大滞后长度选择对死亡置换方向判定的敏感性

    相关阅读

    证据等级: B
    参考文献: PMID 42350784, PMID 19204283, PMID 20186267, PMID 34789160, PMID 20812303, PMID 29447338, PMID 22095719, PMID 30968823, PMID 36630176, PMID 27136236, PMID 28847318


    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。

    Monthly geomagnetic accumulation vs. weekly humidity anomalies: Will two exposure windows cancel each other out in the same model?

    Evidence level: B

    One requires measuring six-month accumulation; the other requires measuring three-week anomalies. Two clock scales differ by an order of magnitude, yet they fight for the same variance on a single mortality curve. This is not a matter of right or wrong, but a window competition that must be designed in advance.

    I. Two Clocks Not on the Same Tick

    A 2026 US-wide study provided geomagnetic side a clear time constant: monthly respiratory tract infection (RTI) mortality data from 47 states between 2000–2019 showed that moving averages of the Kp index from 1 to 12 months yielded the strongest association at a 6-month window, with an RR of 1.16 (95% CI 1.14–1.18) per interquartile range increase for women aged ≥85, and 1.14 (1.12–1.16) for men (PMID 42350784). This is not “more deaths on geomagnetic storm days,” but rather “continuous disturbance over six consecutive months.”

    The meteorological constant is entirely different. Absolute humidity (AH) explains 50% of the transmission variability and 90% of the virus survival variability in guinea pig models (PMID 19204283), manifesting at the population level as abnormally low AH in the weeks preceding epidemic onset (PMID 20186267). A DLNM analysis of 42,440 respiratory deaths in Guangzhou set the window at a lag of 0–35 days, where non-optimal AH contributed 21.57% of deaths (PMID 34789160).

    One requires six months; the other requires three weeks. Putting both exposures into the same regression is like using a meter stick and a micrometer to measure the same piece of cloth.

    II. Stretching the Ruler into a Surface

    The standard tool is the Distributed Lag Non-Linear Model (DLNM): stretching “exposure levels” against “lag days” into cross-basis functions to output a two-dimensional exposure-lag-response surface (PMID 20812303). Joint modeling of dual exposures involves placing both the Kp monthly accumulation basis and the AH weekly anomaly basis simultaneously into the model. Technically feasible, three traps immediately arise.

    Winter Collinearity. Winter is both a low-AH period and frequently cited as a geomagnetic-related window; both vary simultaneously during this season. When entered directly, regression attributes the effect to whichever variable shows stronger explanatory power, while the misattributed one fails to signal an error.

    More Windows Are Not Better. One occupational cohort study performed a direct comparison: DLNMs expanding exposure by lag years showed significantly higher rate ratios (lung cancer 1.47, non-malignant respiratory disease 1.80) compared to “linear cumulative exposure” simple summation (1.15, 1.23) (PMID 29447338). However, higher estimated values do not equal proximity to truth—they may simply reflect scoring the same exposure repeatedly after chopping it into smaller pieces.

    Consequences of Scale Mismatch. Fitting a weekly-scale model to a monthly-scale effect manifests as “all lag coefficients small and barely significant”; forcing a monthly-scale window onto weekly-scale triggers manifests as “effects diluted to invisibility by averaging.”

    III. Three Outcomes of Window Competition

    Logically, there are only three possibilities when two windows enter the model simultaneously, each corresponding to a completely different biological interpretation:

    ① Substitution: The Kp coefficient collapses after AH enters. This suggests the geomagnetic side’s “effect” is largely a proxy for winter meteorology, and immune pathways are not holding up.

    ② Additive: Both remain significant with coefficients not eroding each other. This is evidence of dual pathways acting independently—what this field truly hopes to see.

    ③ Interaction: The product term is significant. Meaning the risk during “high humidity anomaly periods coinciding with high geomagnetic accumulation” exceeds the sum of both.

    The adjudication tool is model comparison, not who has the smaller p-value. One simulation study provided key calibration: The minimum AIC criterion can identify the correct model in most samples, but must view “the correct model or equivalent models within 4 AIC points together”—two window settings with an AIC difference less than 4 are statistically indistinguishable (PMID 22095719). Any conclusion claiming “6 months is the optimal window” must disclose the AIC difference under 5-month and 7-month windows.

    IV. Negative Lag Coefficients Are Not Errors

    Within a joint lag structure, some coefficients will inevitably be negative. This is not a bug but death harvesting (harvesting effect): vulnerable individuals are removed days in advance, causing death counts to subsequently drop below baseline for several days.

    A Greek 2013–2017 influenza-attributable mortality study quantified this: during the A(H1N1)pdm09 epidemic, weekly attributable deaths in populations aged ≥65 rose rapidly first, then dropped to zero or even negative values (PMID 30968823). A DLNM analysis of 199,777 death certificates in Guangzhou provided the duration of harvesting—2–5 days for A(H1N1), A(H3N2), and B/Yamagata, and up to 5–13 days for B/Victoria (PMID 36630176).

    The conclusion is firm: Maximum lag must be stretched sufficiently long (covering at least one complete harvesting cycle), otherwise negative coefficients will be misread as “protective effects.” Window truncation itself is a source of bias.

    V. Stored Pathogen vs. New Attack

    This dual-window structure was already distinguished by Traditional Chinese Medicine (TCM).

    Monthly-scale accumulation = Stored Pathogen (Fu Xie). “Fu-temperature diseases” describe pathogens lurking and accumulating over time, erupting only when conditions mature; corresponding to the Kp six-month moving average form where “no single day is particularly bad, but six months are never peaceful.”

    Weekly-scale anomaly = New Attack (Xin Gan). “New-attack warm diseases” describe immediate injury by the prevailing pathogen; corresponding to AH dipping abnormally in the weeks preceding epidemic onset, with a short window and clear trigger.

    Clinically, “Stored Pathogen” and “New Attack” are never mutually exclusive—only with a Stored Pathogen base layer followed by a New Attack trigger does a major epidemic occur. This is precisely what the interaction term (Outcome ③) must test, and also the statistical translation of “when the body’s vital energy is preserved within, the pathogen cannot invade”: the same AH anomaly carries different risk numbers for people with different Kp accumulation backgrounds.

    VI. A Rule That Must Be Stated First

    The historical association between solar activity and influenza pandemics is the conclusion in this field most easily “hooked” out. One analysis claimed almost all recorded influenza pandemics fell within ±1 year of solar sunspot extrema (OR 3.87, 95% CI 1.08–13.85) (PMID 27136236); however, a re-analysis reviewing each case line-by-line found that those significances stemmed from transcription errors in pandemic years, misuse of statistical methods, and lack of robustness to arbitrary data selection assumptions, all vanishing upon correction (PMID 28847318).

    This cautionary tale directly dictates design discipline: Windows must be specified before seeing the data. Sweeping through 1–12 months to pick the strongest one is essentially multiple comparisons, not discovery. A feasible approach is pre-registering primary windows (Kp accumulation at 6 months, AH anomaly at 0–35 days), reporting remaining windows as sensitivity analyses alongside them, and disclosing AIC differences simultaneously.

    The two windows are never fighting over “who is true,” but rather “which time period they are each responsible for.” If design allows researchers to pick windows post-hoc, this competition has no referee.

    Evidence and Limitations

    Verified:
    – Strongest geomagnetic accumulation exposure association with RTI deaths at 6-month window, RR 1.14–1.16 in elderly groups (PMID 42350784)
    – Absolute humidity modulates influenza virus survival and transmission via mechanistic pathways; population manifestation is abnormally low AH in weeks preceding epidemics (PMID 19204283, PMID 20186267)
    – Non-linear relationship between AH and respiratory deaths; non-optimal AH contributes 21.57% of death burden (PMID 34789160)
    – DLNM cross-basis is the standard framework for modeling exposure-lag-response surfaces (PMID 20812303)
    – Influenza deaths exhibit quantifiable death harvesting with durations ranging from 2–13 days (PMID 36630176, PMID 30968823)
    – The 4-point equivalence rule for AIC model comparison calibrated by simulation (PMID 22095719)

    Partially Verified:
    – No actual studies completed yet on joint lag structures for dual exposures at monthly and weekly scales; none of the three outcomes (substitution/additive/interaction) have been tested
    – The proportion of contamination to coefficient allocation by winter collinearity remains unquantified

    To Be Verified:
    – Dual-exposure joint DLNM in the same cohort: whether Kp monthly accumulation coefficients persist after controlling for AH weekly-scale anomalies
    – Whether the product term between two windows is significant (existence of additive/synergistic effects)
    – Sensitivity of maximum lag length selection on determining the direction of death harvesting

    Related Reading

    Evidence Level: B
    References: PMID 42350784, PMID 19204283, PMID 20186267, PMID 34789160, PMID 20812303, PMID 29447338, PMID 22095719, PMID 30968823, PMID 36630176, PMID 27136236, PMID 28847318


    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.

  • 有的经贴、有的经不贴:十二条经脉筋膜贴合度的平行检验

    Evidence level: B

    真正该问的问题不是”经络是不是筋膜”,而是”哪几条是、哪几条不是”——而现有的研究结构,恰好把这个问题系统性地藏了起来。

    正文

    上一篇讲的是怎么量(把”看着像”换成配准距离、方向偏差与置换检验)。这一篇问一个更麻烦的问题:如果十二条经一条一条地测,答案会不会不一样?

    已有的答案不是”贴”,是”不齐”。 这不是推测,人类身上已经有第一手数据,只是出现在电学而不是几何上。Ahn 等 2005 年用四针电极法测了 23 人,心包经段的组织阻抗 70.4±5.7 Ω,相邻对照段 75.0±5.9 Ω,p=0.0003,显著偏低;同一次实验里,脾经段 67.8 Ω 对 68.5 Ω,p=0.70,毫无差别(PMID 15882468)。同一批人、同一台仪器、同一天、同一个操作者——一条有、一条没有。五年后他们换到上臂、大腿、小腿三个部位重做,28 人:大肠经在两个频段都显著偏低,肝经与膀胱经都不显著;而真正与低阻抗相关的是超声上的”高回声胶原带密度”,无论部位内比较还是部位间比较都显著,多变量校正后依然成立(PMID 20689594)。

    这两篇合起来说的不是”经络等于筋膜”,而是更冷的一句话:哪儿筋膜多哪儿阻抗低,而经络只是有时候恰好走在筋膜多的地方。

    线和点,必须分开测。 Langevin 等 2002 年把上肢六条经在连续断层上逐片标记,五条与肌间或肌内筋膜面有关;小肠经那一段”不沿任何可辨认的组织面走行”,但它在这段上的四个穴位里有三个(SI9、SI10、SI11)明确落在多个筋膜面的交点上(PMID 12467083)。整体统计:超过 80% 的穴位、50% 的经脉交点落在筋膜面上,超几何检验 P<0.001。

    小肠经因此是最有价值的一条:线不贴,点贴。 这说明”贴合”其实是两个独立命题——经脉这条线贴不贴,穴位这些点贴不贴。只报一个平均距离,会把”线贴点也贴””线不贴点贴””线贴点不贴”三种完全不同的解剖结局,压成同一个”中等贴合”。而那个”中等”对谁都不成立。

    平均的罪就在这里。 假设六条经平均差 3 mm、六条差 20 mm,总平均 11.5 mm,读起来像”每条都差一厘米左右”,事实却是一半很好、一半完全没有。聚合掩盖方向甚至反转方向,是统计上早已写清的陷阱(PMID 23964259);在研究层面用平均特征去解释个体层面的效应,方向可以和真实情况完全相反(PMID 18671838)。所以正确的报法不是总平均,而是逐经结果加跨经异质性。

    三个必须先排除的伪影,否则测出来的不是经络。

    第一,配准残差不是精度。Fitzpatrick 的经典结论:残差(FRE)可以在配准完全错误时也很小,而目标点误差(TRE)很大(PMID 9874293);残差对目标误差波动的预测力可以忽略,相关系数低于 0.1(PMID 21075718)。更麻烦的是,标记点近共线且远离目标时 TRE 最大(PMID 11322441)——四肢经脉恰恰又直又长、又远离躯干,正落在最差的配置上。所以这类研究必须报 TRE,报 FRE 等于自证。

    第二,图谱本身的方差可能比信号大一个数量级——这是最硬的一击。Bäumler 等让一批资深针灸师在同一只手臂上标 LI10 与 TH5,两点之间的最大距离分别为 8.45 cm 与 5.3 cm,95% 定位椭圆面积在 5.18 到 44.49 cm² 之间(PMID 22398924);Guo 等让五位医师标下肢穴位,仅温溜一穴的最大差异就有 16.6 mm(PMID 37596915)。而待测的贴合距离是毫米级——按 Langevin 模型的估算,一条筋膜面的”宽度”约 5 mm。图谱散布在厘米级、待测效应在毫米级,这是信噪比倒挂:你测到的主要是”谁画的图”,不是”经在哪里”。解法不是提高分辨率,而是先做图谱内部一致性筛选,一致性不达标的经脉,压根不该进入贴合检验。

    第三,方向偏差在层状介质里没有唯一解。深筋膜由二到三层胶原亚层叠成,每层约 115 μm(另一组测得 277 μm),相邻亚层纤维夹角约 78°,层间夹着约 43 μm 的疏松层(PMID 16981399、PMID 21203765)。所以”经脉方向与筋膜方向的夹角”这个指标,必须先声明”跟哪一层比”——换一层,答案就翻。

    判决设计与三种结局。 预注册逐经三量(平均最近距离、方向偏差、穴位-筋膜交点命中率),把”经”当随机效应做多层模型,同时报组内相关与跨经异质性。三种可能结局各自指向不同结论:①普遍贴合——异质性低、全部显著优于置换零分布,筋膜假说成立;②分层贴合——部分经显著、部分与随机无异,这支持”经络是一组异质结构,而非单一解剖实体”;③普遍不贴——假说应被修正。

    第二种才是最可能的结局,也是对理论最有用的结局。但它恰恰被现有文献结构藏住了:每篇研究只测一到三条经,而被选中的那几条,往往正是因为”看起来贴”。这不是造假,是选择效应。

    古人自己在语言里就区分过这件事。 《灵枢·经脉》写十二经走行,用的动词并不统一:”循””过””挟””贯””属””络”,每条经的句式都不一样。若十二条经是同一种结构,没必要为每条经挑不同的动词。这套措辞本身就在说:它们贴的方式不一样。现代研究把十二条经当成同一个假说的十二次复制,反而丢掉了古人最细的那一层。

    所以问题不该是”经络是不是筋膜”,而是”哪几条是、以什么方式、在哪个尺度上”。

    证据与局限

    已验证: 心包经段阻抗显著低于对照而脾经段无差异(PMID 15882468);大肠经显著、肝经与膀胱经不显著,且低阻抗与超声胶原带密度相关(PMID 20689594);上肢超 80% 穴位、50% 经脉交点落在筋膜面,小肠经线不贴而点贴(PMID 12467083);配准残差不能代表目标误差(PMID 9874293);穴位定位的术者间差异达厘米级(PMID 22398924、PMID 37596915)。

    部分验证: 深筋膜的分层与约 78° 交错结构在尸体与影像上一致(PMID 16981399、PMID 21203765),但活体全身尺度上的层间方向一致性尚未系统测定;筋膜与经络的对应有系统综述支持(PMID 21584283),但综述收录的多为单条经的定性观察。

    待验证: ①十二条经在同一批受试者、同一套协议下完成平行配准,逐经报 TRE 与异质性;②穴位定位一致性(ICC)与贴合度之间的相关——若高度相关,说明过去的”贴合”结论部分由图谱方差贡献;③”线贴合”与”点贴合”是否解离,以及解离是否稳定复现(小肠经是目前唯一的单例)。

    相关阅读

    证据等级: B(筋膜-穴位对应的直接证据充分,但”十二经平行配准”本身尚无研究完成,属设计提案)
    参考文献: PMID 12467083, PMID 15882468, PMID 20689594, PMID 22398924, PMID 37596915, PMID 9874293, PMID 11322441, PMID 21075718, PMID 16981399, PMID 21203765, PMID 23964259, PMID 18671838


    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。

    Some meridians adhere, some do not: parallel testing of the adhesion degree of twelve meridian fasciae

    Evidence level: B

    The question that truly matters is not “Are meridians fascia?” but rather “Which ones adhere and which do not”—yet the current research structure systematically obscures this question.

    Main Text

    The previous article addressed how to measure (replacing “visual similarity” with registration distance, directional deviation, and permutation testing). This article addresses a more complex question: if we measure each of the twelve meridians individually, will the answers differ?

    The existing answer is not “adhesion,” but “inconsistency.” This is not speculation; primary human data already exists, though it appears in electrophysiology rather than geometry. Ahn et al. (2005) measured 23 individuals using a four-needle electrode method: the Pericardium meridian segment showed significantly lower tissue impedance (70.4±5.7 Ω) compared to adjacent control segments (75.0±5.9 Ω, p=0.0003); within the same experiment, the Spleen meridian segment showed no difference (67.8 Ω vs 68.5 Ω, p=0.70) (PMID 15882468). Same subjects, same instrument, same day, same operator—one meridian adheres, another does not. Five years later, they re-measured three sites (upper arm, thigh, lower leg) in 28 individuals: the Large Intestine meridian was significantly lower across two frequency bands, while Liver and Bladder meridians were not; the feature correlating with low impedance was ultrasound-detected “high-echo collagen band density,” significant in both intra-site and inter-site comparisons, remaining robust after multivariate correction (PMID 20689594).

    Taken together, these papers do not state “Meridians equal fascia,” but rather a colder truth: Where there is more fascia, impedance is lower; meridians merely sometimes coincide with areas of high fascial density.

    Lines and points must be measured separately. Langevin et al. (2002) marked six upper-limb meridians slice-by-slice on continuous CT scans; five were associated with intermuscular or intramuscular fascial planes. However, a segment of the Small Intestine meridian “did not follow any recognizable tissue plane,” yet three of its four acupoints in that segment (SI9, SI10, SI11) clearly fell at intersections of multiple fascial planes (PMID 12467083). Overall statistics: over 80% of acupoints and 50% of meridian intersections lie on fascial planes, with a hypergeometric test P<0.001.

    Thus, the Small Intestine meridian is most valuable: The line does not adhere, but the points do. This indicates that “adhesion” comprises two independent propositions: whether the meridian line adheres and whether the acupoint points adhere. Reporting a single average distance compresses three distinct anatomical outcomes—”line adheres and points adhere,” “line does not adhere but points do,” “line adheres but points do not”—into one “moderate adhesion,” which holds true for no one.

    The sin of the average lies here. Assuming six meridians average 3 mm deviation and six average 20 mm, the total average is 11.5 mm, sounding like “each deviates by about a centimeter,” whereas in reality half are good and half are none. Aggregation masks direction or even reverses it, a trap well-documented in statistics (PMID 23964259); using averaged features to explain individual-level effects can yield directions completely opposite to reality (PMID 18671838). Therefore, the correct reporting method is not a total average, but per-meridian results plus cross-meridian heterogeneity.

    Three artifacts that must be excluded first, otherwise the measurement is not of meridians.

    First, registration residuals are not precision. Fitzpatrick’s classic conclusion: residuals (FRE) can be small even when registration is completely wrong, while target registration error (TRE) is large (PMID 9874293); the predictive power of residuals for target error fluctuations is negligible, with correlation coefficients below 0.1 (PMID 21075718). More麻烦的是,marked points are nearly collinear and far from targets when TRE is maximized (PMID 11322441)—limb meridians are precisely straight, long, and distant from the trunk, falling into the worst configuration. Thus, such studies must report TRE; reporting FRE is self-incriminating.

    Second, the variance inherent in the atlas itself may exceed the signal by an order of magnitude—the hardest blow. Bäumler et al. asked a group of senior acupuncturists to mark LI10 and TH5 on the same arm; the maximum distance between points ranged from 8.45 cm to 5.3 cm, with 95% localization ellipses ranging from 5.18 to 44.49 cm² (PMID 22398924); Guo et al. asked five practitioners to mark lower-limb acupoints, with a maximum difference of just 16.6 mm for one point (Wenliu) alone (PMID 37596915). The measured adhesion distance is millimeter-scale—according to Langevin’s model, the “width” of a fascial plane is approximately 5 mm. Atlas dispersion occurs at the centimeter scale while the measured effect is millimeter-scale; this is an inverted signal-to-noise ratio: what you measure is primarily “who drew the map,” not “where the meridian lies.” The solution is not higher resolution, but first filtering for atlas internal consistency; meridians failing this consistency check should not enter adhesion validation at all.

    Third, directional deviation lacks a unique solution in layered media. Deep fascia consists of two to three collagen sub-layers stacked, each approximately 115 μm thick (another group measured 277 μm), with adjacent sub-layers at an angle of approximately 78°, separated by a loose layer of about 43 μm between them (PMID 16981399, PMID 21203765). Therefore, the metric “angle between meridian direction and fascial direction” must first declare “which layer is being compared”—changing the layer flips the answer.

    Judgment Design and Three Possible Outcomes. Pre-registered per-meridian three measurements (average nearest distance, directional deviation, acupoint-fascial intersection hit rate), treating “meridian” as a random effect in multilevel models, while reporting intraclass correlation and cross-meridian heterogeneity. Three possible outcomes point to different conclusions: ① Universal adhesion—low heterogeneity, all significantly better than permutation null distribution, supporting the fascial hypothesis; ② Layered adhesion—some meridians significant, others无异 to random chance, supporting “meridians are a set of heterogeneous structures rather than a single anatomical entity”; ③ Universal non-adhesion—the hypothesis should be revised.

    The second outcome is the most likely and also the most useful for theory. Yet it is precisely hidden by the current literature structure: each study measures only one or three meridians, and those selected are often precisely because they “look like they adhere.” This is not fabrication, but selection bias.

    The ancients themselves distinguished this in language. The Ling Shu · Jing Mai describes the course of the twelve meridians using inconsistent verbs: “follows,” “passes,” “clamps,” “pierces,” “belongs to,” and “connects”; each meridian uses a different sentence structure. If all twelve meridians were the same structure, there would be no need to select different verbs for each. This phrasing itself states: they adhere in different ways. Modern research treats the twelve meridians as twelve replicates of a single hypothesis, thereby losing the finest layer the ancients preserved.

    Therefore, the question should not be “Are meridians fascia?” but rather “Which ones adhere, in what manner, at what scale.”

    Evidence and Limitations

    Verified: Pericardium segment impedance significantly lower than control while Spleen segment showed no difference (PMID 15882468); Large Intestine significant, Liver and Bladder not significant, with low impedance correlating with ultrasound collagen band density (PMID 20689594); over 80% of upper-limb acupoints and 50% of meridian intersections lie on fascial planes, with Small Intestine line not adhering but points adhering (PMID 12467083); registration residuals do not represent target error (PMID 9874293); inter-practitioner variation in acupoint localization reaches centimeter scale (PMID 22398924, PMID 37596915).

    Partially Verified: Layered structure and ~78° intersecting arrangement of deep fascia consistent in cadaver and imaging studies (PMID 16981399, PMID 21203765), but inter-layer directional consistency at whole-body scale in living subjects not systematically measured; systematic review supports correspondence between fascia and meridians (PMID 21584283), but reviews predominantly include qualitative observations of single meridians.

    To Be Verified: ① Parallel registration of all twelve meridians in the same cohort under identical protocols, reporting TRE and heterogeneity per meridian; ② Correlation between acupoint localization consistency (ICC) and adhesion degree—if highly correlated, past “adhesion” conclusions partially contributed by atlas variance; ③ Whether “line adhesion” and “point adhesion” dissociate, and whether this dissociation is stably reproducible (Small Intestine meridian currently the only single case).

    Related Reading

    Evidence Level: B (Direct evidence for fascia-acupoint correspondence is sufficient, but “parallel registration of twelve meridians” itself has no completed study yet; belongs to design proposal)
    References: PMID 12467083, PMID 15882468, PMID 20689594, PMID 22398924, PMID 37596915, PMID 9874293, PMID 11322441, PMID 21075718, PMID 16981399, PMID 21203765, PMID 23964259, PMID 18671838


    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.

  • 尺子本身有多长:经络坐标的不确定度,够不够推翻筋膜贴合这个结论

    Evidence level: B

    在问”这条经贴不贴”之前,得先问”我们量的那个点,究竟在哪”——如果找不到一个确定的位置,那”距离”这个数字本身就没有对象。

    正文

    上一篇测出了”有的经贴、有的经不贴”。这一篇退后一步,问更前面的事:那把尺子准不准。

    贴合度的算法很直白:把经络坐标和解剖影像放进同一空间,量最近距离。算法把坐标当成一个确定的点,但它不是——它是某本书、某个人、在某个姿势下摸着骨头找出来的位置。这四个环节,每一个都有不小的不确定度。

    第一个数字大得有点尴尬。 WHO 2008 年发布针灸穴名标准时,361 个穴位里 92 个国际上位置有争议;七次非正式磋商、四次工作组会议,谈下 86 个,还剩 6 个(LI19、LI20、PC8、PC9、GB30、GV26)只能以”替代位置”并列保留(PMID 19204011)。最权威的这套坐标,是谈出来的,不是测出来的。文本质量也不牢:标为 original 的汉字字形只有 48% 真的是繁体(PMID 36648069)。名字都没对齐,指望坐标精确到毫米,前提就不成立。

    第二个数字是厘米级。 Bäumler 等在国际针灸会议上请与会者找 LI10 和 TH5:两人找出的同一个点,最远相距 8.45 和 5.3 厘米;理论含 95% 定位的椭圆面积从 5.18 到 44.49 平方厘米(PMID 22398924)。研究查了学历、年限、风格、定位法,全不显著——唯一显著的是”是否同一所学校训练”。离散不是无规律的噪声,是按流派成簇的。Guo 等在下肢复现:五位针灸师找温溜,两点坐标距离最大 16.6 毫米(PMID 37596915)。而上一篇要检出的信号是毫米级筋膜面宽度。噪声比信号大一个数量级。

    第三个数字更大,是系统性的。 流派差异不止”点在哪”,更在”选哪些点”。七位针灸师看同一位慢性腰痛患者,诊断高度一致,处方却从 5 个穴到 14 个穴不等;28 个被选中的穴里只有 4 个被两人以上选中(PMID 11372797)。另一项拉丁方设计里,六位医师看六位患者,共用 20 个诊断、65 个穴位,细粒度诊断与”是哪个患者”无关、与”是哪个医师”关系更大(PMID 11926429)。连取哪条线都不一致,”这条线贴不贴”的对象就没定下来。

    第三层更麻烦:坐标系本身会变形。 骨度分寸不是绝对长度,是按比例折算。Park 等用双能 X 线测 47 名韩国女性,发现一夫法量出的一寸与骨度折算的一寸显著不同,骨度法更可靠(PMID 17080542)。Wu 等测 100 名健康人,双侧头维之间 134.54±14.72 毫米、双侧完骨之间 161.73±25.05 毫米——国标当它们等长,实测不等;腋前纹头至肘横纹折算 12 寸,国标写 9 寸(PMID 21585063)。

    而这一切发生在会拉伸的纸上。 骨性标志要靠摸:三位检查者在 60 人身上定位腰椎棘突,X 线验证正确率 69%,有第十二肋变异者降到 55%,肥胖者定位 L4 只有 44%(PMID 21640254)。摸准之后皮肤还要动:屈曲时背侧皮肤平均拉伸 38%、变薄 19%、变硬 106%,伸展时回缩 12%(PMID 27978463)。在 38% 的拉伸下,”距离”已不是守恒量。

    三层不确定性的破坏方式不一样,必须分开。个体间找点离散像经典测量误差:把真实关联往零压,方向可预测,可以校正(PMID 32246539 给了完整框架,空间数据还要处理空间相关,PMID 26788930)。流派系统性偏移不是噪声是平移:整条经挪位,平均距离凭空变大变小,能被伪造成”贴”或”不贴”。坐标系随姿势变形最狠:参照系在动,最近距离失去定义。

    这里必须诚实一次:不能简单说”所以真实贴合度比测出来的更高”。 衰减校正的前提是误差围绕真值对称分布且与真值独立。Bäumler 的数据显示分布是多峰的(椭圆面积从 5.18 到 44.49 平方厘米,且按流派成簇)——多峰意味着混合分布,不是经典误差。此时乘一个校正因子,会得到一个看起来精确、实际没有根据的数字。该做的是概率偏差分析。

    所以正确的做法是把坐标当随机变量,不当常数。 三层次:①单点敏感度——逐个替换坐标系(WHO 标准/国标/各流派教材),看结论翻不翻,做成 tornado 图;②概率敏感度——给每个穴位一个空间分布,抽样后输出”结论”的分布;③多宇宙分析——把流派×定位法×姿势组合枚举成规格空间,画 specification curve。方法已成熟:2026 年的方法学论文指出 specification curve 有”任意抽样”的毛病,提出 multiverse plot 保留全部规格信息(PMID 41642923);红肉与全因死亡率研究示范了怎样摊开争议结论的稳健性(PMID 38354868)。

    最后给一条能执行的判据,而不是”建议做敏感性分析”这种空话。 主张:不看中位数,看符号翻转率。 若多宇宙里超过 20% 的合理规格下”这条经贴合”翻转成”不贴合”,结论就不该以”某经贴/不贴”的形式发表,只能以”在某套坐标约定下平均最近距离是多少”的形式发表。前者是关于人体的断言,后者是关于测量的报告——证据强度差一个数量级。

    《灵枢·骨度》给了一整套比例尺:头之大骨围二尺六寸、胸围四尺五寸、腰围四尺二寸。但它在《经水》里写了一句要紧话:“其可为度量者,取其中度也。” 古人明确说了,能用尺寸量的是中等身材的人;超出中度者”各如其度”,按他自己的比例来。古人给的是一把会跟着人伸缩的尺子。 现在的做法是把这把尺子硬化成绝对坐标,再拿它做毫米级解剖比对。测出来的差异里,多少来自身体,多少来自尺子?这个问题没答之前,”贴不贴”其实还悬着。

    证据与局限

    已验证: 经络穴位定位的个体间与流派间离散是厘米量级(PMID 22398924、PMID 37596915);WHO 标准本身仍有 6 个穴位保留替代位置,形成过程为专家协商(PMID 19204011);骨度分寸的实测比例与现行国标存在系统性差异(PMID 17080542、PMID 21585063);骨性标志触诊定位正确率仅 69%,受解剖变异与体型显著影响(PMID 21640254);姿势可使背侧皮肤拉伸 38%(PMID 27978463)。

    部分验证: 将测量误差校正框架(STRATOS 指南、空间回归校正、多宇宙分析)迁移到经络坐标问题上,属方法学类比(PMID 32246539、PMID 26788930、PMID 41642923、PMID 38354868);”符号翻转率 20%”作为判决阈值为本文提出的操作建议,非既有共识。

    待验证: 经络坐标的定位分布究竟是单峰还是多峰(决定能否使用经典衰减校正);不同坐标系下”贴合结论”的实际翻转比例;姿势标准化能否把坐标系不确定度压到毫米级以下。

    相关阅读

    证据等级: B
    参考文献: PMID 19204011, PMID 36648069, PMID 22398924, PMID 37596915, PMID 11372797, PMID 11926429, PMID 17080542, PMID 21585063, PMID 21640254, PMID 27978463, PMID 32246539, PMID 26788930, PMID 41642923, PMID 38354868


    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。

    How Long Is the Ruler: Uncertainty in Meridian Coordinates Enough to Overturn the Fascial Adhesion Conclusion

    Evidence level: B

    Before asking “does this meridian adhere,” one must first ask “where exactly is the point we are measuring”—if no definite location can be identified, the number itself representing “distance” lacks an object.

    Main Text

    The previous study measured that “some meridians adhere, while others do not.” This paper steps back to ask an earlier question: Is the ruler accurate.

    The algorithm for adhesion is straightforward: place meridian coordinates and anatomical images in the same space and measure the nearest distance. The algorithm treats coordinates as fixed points, but they are not—they are positions found by feeling bones in a specific posture based on a specific book or person. Each of these four links carries significant uncertainty.

    The first number is embarrassingly large. When WHO released the standard for acupuncture point names in 2008, 92 of the 361 points had internationally disputed locations; after seven informal consultations and four working group meetings, 86 were resolved, leaving six (LI19, LI20, PC8, PC9, GB30, GV26) retained only as “alternative positions” (PMID 19204011). The most authoritative set of coordinates was negotiated, not measured. Textual quality is also unstable: only 48% of characters marked as original are genuinely traditional (PMID 36648069). Without aligned names, the premise for millimeter-level precision does not hold.

    The second number is centimeter-level. At an international acupuncture conference, Bäumler et al. asked attendees to locate LI10 and TH5; the same point found by two individuals varied by up to 8.45 cm and 5.3 cm; the theoretical ellipse area containing 95% of locations ranged from 5.18 to 44.49 cm² (PMID 22398924). Studies examining education, years of practice, style, and localization methods showed no significant differences—the only significant factor was “whether trained at the same school.” Discrepancy is not random noise but clusters by school. Guo et al. replicated this in the lower limbs: five acupuncturists locating Wenliu showed a maximum coordinate distance of 16.6 mm (PMID 37596915). Yet the signal to be detected in the previous study is millimeter-level fascial surface width. The noise exceeds the signal by an order of magnitude.

    The third number is even larger, being systematic. Differences between schools are not just “where the point is” but “which points are selected.” Seven acupuncturists examining the same chronic low back pain patient showed highly consistent diagnoses but prescribed between 5 and 14 points; among the 28 selected points, only four were chosen by more than two practitioners (PMID 11372797). In another Latin square design, six physicians examined six patients using a shared set of 20 diagnoses and 65 points; fine-grained diagnosis was less related to “which patient” than to “which physician” (PMID 11926429). Even the selection of which line is inconsistent, so the object of “does this line adhere” is not defined.

    The third layer adds further complication: the coordinate system itself deforms. Bone proportions are not absolute lengths but scaled ratios. Park et al. used dual-energy X-ray absorptiometry on 47 Korean women and found that the “one-finger” method yielded a significant difference compared to bone proportion scaling, with the latter being more reliable (PMID 17080542). Wu et al. measured 100 healthy individuals, finding bilateral Touwei distance as $134.54 \pm 14.72$ mm and bilateral Wangu distance as $161.73 \pm 25.05$ mm—national standards treat them as equal, but measurements differ; the standard converts from axillary anterior crease to elbow transverse crease as 12 cun, but writes 9 cun (PMID 21585063).

    All this occurs on paper that stretches. Bone landmarks require palpation: three examiners locating lumbar spinous processes on 60 subjects achieved an X-ray verified accuracy of 69%, dropping to 55% with twelfth rib variations and 44% for obese subjects locating L4 (PMID 21640254). After accurate palpation, skin movement is required: during flexion, dorsal skin stretches an average of 38%, thins by 19%, and hardens by 106%; during extension, it retracts by 12% (PMID 27978463). Under 38% stretch, “distance” is no longer a conserved quantity.

    The three layers of uncertainty destroy the measurement in different ways and must be separated. Inter-individual point dispersion resembles classical measurement error: it suppresses the true association towards zero, has a predictable direction, and can be corrected (PMID 32246539 provides a complete framework; spatial data also requires handling spatial correlation, PMID 26788930). Systematic school-based offsets are not noise but translation: the entire meridian shifts, causing average distance to increase or decrease arbitrarily, which can be faked as “adhering” or “not adhering.” Coordinate deformation due to posture is most severe: the reference frame moves, rendering nearest distance undefined.

    Here we must be honest once more: one cannot simply say “therefore the true adhesion rate is higher than measured.” The premise for attenuation correction assumes errors are symmetrically distributed around the true value and independent of it. Bäumler’s data shows a multimodal distribution (ellipse areas ranging from 5.18 to 44.49 cm², clustering by school)—multimodality implies a mixed distribution, not classical error. Multiplying by a correction factor would yield a number that appears precise but lacks basis. What should be done is a probability bias analysis.

    Therefore, the correct approach is to treat coordinates as random variables, not constants. Three levels: ① Single-point sensitivity—replace the coordinate system individually (WHO standard / National Standard / various school textbooks) to see if conclusions flip, creating a tornado plot; ② Probabilistic sensitivity—assign a spatial distribution to each point, sample it, and output the distribution of the “conclusion”; ③ Multiverse analysis—enumerate combinations of schools × localization methods × postures into a specification space and draw the specification curve. Methods are mature: a 2026 methodological paper noted that specification curves suffer from “arbitrary sampling,” proposing multiverse plots to retain all specification information (PMID 41642923); studies on red meat and all-cause mortality demonstrated how to expose the robustness of controversial conclusions (PMID 38354868).

    Finally, provide an executable criterion rather than the empty phrase “recommend conducting sensitivity analysis.” Stance: Do not look at the median, but at the symbol flip rate. If in the multiverse, under more than 20% of reasonable specifications, “this meridian adheres” flips to “does not adhere,” the conclusion should not be published in the form of “meridian adheres / not adheres,” but rather as “what is the average nearest distance under a specific coordinate convention.” The former is an assertion about the human body; the latter is a report on measurement—the evidence strength differs by an order of magnitude.

    Shijing · Bone Proportions provides a set of proportional scales: the circumference of the head’s large bones is 2 chi 6 cun, the chest circumference is 4 chi 5 cun, and the waist circumference is 4 chi 2 cun. However, in Jing Shui, it writes a crucial sentence: “For those that can be measured, take the average measure.” The ancients explicitly stated that dimensions can only be measured in people of medium build; for those exceeding the average, “each follows their own proportion.” The ruler given by the ancients is one that stretches with the person. Current practice hardens this ruler into absolute coordinates and uses it for millimeter-level anatomical comparison. Within the measured differences, how much comes from the body and how much from the ruler? Until this question is answered, “adhesion or not” remains suspended.

    Evidence and Limitations

    Verified: Inter-individual and school-based dispersion in meridian point localization is centimeter-level (PMID 22398924, PMID 37596915); WHO standards themselves retain six points with alternative positions, formed through expert negotiation (PMID 19204011); measured bone proportion ratios differ systematically from current national standards (PMID 17080542, PMID 21585063); palpation accuracy for bone landmarks is only 69%, significantly affected by anatomical variations and body type (PMID 21640254); posture can cause dorsal skin stretching of 38% (PMID 27978463).

    Partially Verified: Migrating measurement error correction frameworks (STRATOS guidelines, spatial regression correction, multiverse analysis) to the meridian coordinate issue constitutes a methodological analogy (PMID 32246539, PMID 26788930, PMID 41642923, PMID 38354868); “symbol flip rate of 20%” as a decision threshold is an operational suggestion proposed in this paper, not existing consensus.

    To Be Verified: Whether the distribution of meridian point localization is unimodal or multimodal (determining whether classical attenuation correction can be used); the actual flip rate of “adhesion conclusions” under different coordinate systems; whether posture standardization can reduce coordinate uncertainty below the millimeter level.

    Related Reading

    Evidence Level: B
    References: PMID 19204011, PMID 36648069, PMID 22398924, PMID 37596915, PMID 11372797, PMID 11926429, PMID 17080542, PMID 21585063, PMID 21640254, PMID 27978463, PMID 32246539, PMID 26788930, PMID 41642923, PMID 38354868


    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.

  • 月光到视网膜要过四道门:0.3 勒克斯不是眼睛收到的那个数

    Evidence level: B

    我们手上有 0.3 勒克斯这个室外照度,也有 1.5 melanopic lux 那条抑制阈值。两者之间隔着四道门,而每一道的通过率都不是常数。

    满月夜,开阔地面上的水平照度约 0.1 到 0.3 勒克斯。另一头,褪黑素抑制的起始点约在 1.5 melanopic lux(PMID 30697806)。粗看这是道简单除法题:差五到十几倍,不够就是不够。但这两个数根本不在同一个位置上——前者是室外水平照度,后者是到达视网膜、按 melanopsin 作用谱逐点加权之后的剂量。中间那条链上有四道门。

    第一道门:窗户和窗帘。 这一级最”人为”:低辐射镀膜玻璃、双层中空玻璃、遮光率各异的窗帘、室内多次反射,每项都在改通过率;而窗帘拉一半还是拉严,是同一个人同一夜晚里的两个实验条件。它几乎没有”标准值”,只有”当时那个窗帘的值”。

    第二道门:瞳孔。 它不是固定光阑,是被光本身驱动的自反馈。施霍尔恩等人(PMID 38348477)在 90 勒克斯以下的常用室内照明里,用五基色显示器做出 L、M、S 视锥辐照度匹配、只有 melanopic 辐照度差约三倍的两组光,结果高 melanopic 条件下瞳孔中位尺寸显著变小。你把照度算好了,瞳孔会自己改;改多少,又取决于这段光里 melanopsin 含量的高低。闭眼时这级也没关掉——近红外侧照、透过闭合眼睑成像已能在临床上连续测到瞳孔光反应(PMID 34745751、PMID 41694690)。

    第三道门:眼睑,四道门里最被低估的一道。 比耶曼等人(PMID 21721832)逐点测过 450 到 650 纳米的眼睑光谱透过率,两个结论都不温和:一是眼睑在短波段的光学密度远高于以往文献值,该波段平均光学密度 2.1 加减 0.3,受试者间光学密度差异范围约 1.0(相当于透过率相差十倍);二是皮肤色素沉着与眼睑透过率相关性很差,真正决定透过率的是眼睑里那些与波长无关的大分子,以及眼睑的总厚度。

    第二条更值得琢磨:眼睑不是一块中性灰片,而是一块光谱有偏的滤光片,而 melanopsin 的作用峰恰在 480 纳米附近——被砍得最狠的那段,正是加权算法最看重的那段。所以”眼睑挡掉百分之多少的光”这句话本身就不成立,必须拿光谱逐点相乘再重新加权。同理,想靠肤色估眼睑透过率,这条路被数据直接堵死。

    反过来,闭眼也根本不等断路。菲格罗(PMID 26985450)让健康老人与早醒型失眠者在睡眠中佩戴透过闭合眼睑发光的面罩:闪光蓝光把褪黑素起始时间显著延后(34 加减 30 分钟,红光对照 6 加减 30 分钟),睡眠起始延后约 46 分钟,而闪光红光没有这个效果。闭眼是条窄路,但蓝光仍有可测通过量。

    第四道门:眼内介质,主要是晶状体。 这级的特征不是个体差异,而是年龄:晶状体对短波的透过率随年龄下降(PMID 33772847 把晶状体透过光谱与瞳孔大小并列为影响光致褪黑素抑制的两个因子;PMID 33020575 提供了在体测晶状体密度与透过光谱的方法;PMID 18757473 讨论老化的眼在昼夜光感受中的角色)。同一个月亮,二十岁和七十岁的晶状体送进视网膜的是两份光谱。

    阈值那一头同样不干净。 1.5 melanopic lux 起自对既有数据集的建模(PMID 30697806),同文给出饱和点在 305 melanopic lux;更早的剂量-反应工作(PMID 10922269)测到的是另一种陡度——只用略高于最大亮光百分之一的强度,就拿到最大抑制量的一半。个体层面 ED50 可跨一个数量级(PMID 31138694);青少年与成人对不同光强的抑制不同(PMID 30803301);而性别差异只在 400 勒克斯以上出现,10 到 200 勒克斯段不显著(PMID 39041348)——在极低照度区,个体差异究竟是缩窄了,还是仅掉到检测下限以下,目前分不开。还有一条边界证据很有力:全盲者眼内受到白光暴露时,褪黑素仍可被显著抑制(PMID 29625838)。通路不依赖成像视觉,所以”闭着眼能不能收到”与”能不能看见”是两件事。

    把这些摆在一起,诚实的回答既不是”够”也不是”不够”:眼睑光学密度在个体间差 1.0 就足以把结论翻转一次,窗帘、瞳孔、晶状体还会各再翻一次。真正的问题是——这条链从来没有人在真实条件下端到端量过。实验室里的月光剂量实验多半是暗室、睁眼、固定光谱;而真实的月光暴露发生在闭眼侧卧、窗帘状态随人、晶状体年龄随人的卧室里。两端的自变量不重合,等于拿在甲条件标定的尺子去量乙条件。

    要闭合这条链,技术上已可行:可穿戴光谱辐射计做在体剂量测量(PMID 36433426),闭眼瞳孔测量已成熟(PMID 34745751、PMID 41694690),报告规范也已列出该报哪些参数(PMID 31281903;melanopic lux 作为测量量的有效性见 PMID 28545361)。缺的不是工具,是有人愿意在满月夜戴着传感器、拉着不同窗帘、闭着眼,把这几级折扣率一起量出来。

    《素问·金匮真言论》讲”天有八风,经有五风”——天的风有八种,进到人身上只剩五种。《灵枢》到《伤寒论》讲”传变”,讲的就是气进入身体这一路会被改造。古人从不认为”天给多少”等于”人得多少”;中介在古人眼里不是管子,是车间:东西进去了,出来已不是原来的东西。眼睑那条对 480 纳米格外狠的透过率曲线,就是这间车间的第一道工序——它不是在削弱月光,是在把月光改成另一种光。

    证据与局限

    已验证: 眼睑光谱透过率可测且可预测,短波段光学密度显著高于早期估计、个体间光学密度差异范围约 1.0 且与皮肤色素沉着相关性差(PMID 21721832);透过闭合眼睑的蓝光干预可显著延后 DLMO 与睡眠起始(PMID 26985450);闭眼状态下瞳孔尺寸与光反应可被连续测量(PMID 34745751、PMID 41694690);瞳孔收缩显著改变 melanopic EDI 与褪黑素抑制的关系(PMID 38348477);晶状体透过光谱与瞳孔大小为影响光致褪黑素抑制的因子(PMID 33772847),晶状体密度与透过光谱可在体测量(PMID 33020575);褪黑素抑制主要由 melanopsin 驱动、可低至约 1.5 melanopic lux 起始并饱和于约 305 melanopic lux(PMID 30697806);抑制在低剂量端极陡(超过最大亮光约 1% 即达半数最大抑制,PMID 10922269);ED50 个体间跨一个数量级(PMID 31138694);全盲者眼内白光暴露仍可显著抑制褪黑素(PMID 29625838);melanopic lux 对抑制的预测优于普通照度(PMID 28545361);光暴露报告参数已有规范(PMID 31281903)。

    部分验证: 住宅窗户与窗帘的可见光透过率对昼夜节律剂量的实际贡献——现有玻璃透过率数据多来自其他应用场景,缺少以 melanopic 加权为准的卧室实测;性别差异是否存在低照度边界(PMID 39041348 显示 400 lux 以上显著、10–200 lux 不显著,但该照度段统计功效本就紧张);闭眼状态下瞳孔是否仍被调制到足以改变 melanopic 剂量的程度。

    待验证: 0.1–0.3 lux 室外月照在”窗户 + 窗帘 + 眼睑 + 晶状体”四级串联后的实际视网膜 melanopic EDI 是否低于 1.5 melanopic lux——尚无端到端实测;各级折扣率的方差分解(哪一级贡献最大)未知;眼睑厚度等形态学个体差异能否被外部可测指标预测。

    相关阅读

    证据等级: B
    参考文献: PMID 21721832, PMID 34745751, PMID 41694690, PMID 26985450, PMID 38348477, PMID 33772847, PMID 33020575, PMID 18757473, PMID 30697806, PMID 10922269, PMID 31138694, PMID 30803301, PMID 39041348, PMID 29625838, PMID 28545361, PMID 31281903, PMID 36433426


    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。

    From Moonlight to Retina: Four Gates to Cross; 0.3 Lux Is Not the Amount Received by the Eye

    Evidence level: B

    We possess the outdoor illuminance value of 0.3 lux and the melanopic lux inhibition threshold of 1.5. Between these two figures lie four gates, and the transmission rate through each gate is not constant.

    On a full moon night, horizontal illuminance on an open ground ranges approximately from 0.1 to 0.3 lux. On the other end, the onset point for melatonin suppression is approximately 1.5 melanopic lux (PMID 30697806). At first glance, this appears to be a simple division problem: a difference of five to ten times; if it is insufficient, it remains insufficient. However, these two numbers occupy fundamentally different positions—the former represents outdoor horizontal illuminance, while the latter represents the dose after reaching the retina and applying point-by-point weighting according to the melanopsin action spectrum. The chain connecting them contains four gates.

    Gate One: Windows and Curtains. This level is the most “artificial”: low-emissivity coated glass, double-pane insulated glass, curtains with varying light-blocking rates, and multiple indoor reflections all modify the transmission rate; whether a curtain is half-drawn or fully closed constitutes two experimental conditions for the same person on the same night. It lacks a “standard value,” possessing only “the value of that specific curtain at that moment.”

    Gate Two: The Pupil. It is not a fixed aperture but a self-feedback mechanism driven by light itself. Schollhorn et al. (PMID 38348477), within common indoor lighting below 90 lux, used five-primary color displays to create two groups of light where L, M, and S cone irradiance were matched yet the melanopic irradiance differed by approximately three times. Under the higher melanopic condition, the median pupil size significantly decreased. You may calculate the illuminance correctly, but the pupil will adjust itself; the degree of adjustment depends on the concentration of melanopsin within that segment of light. This gate is not closed when eyes are shut—near-infrared side illumination and imaging through closed eyelids can be continuously measured clinically as pupillary light responses (PMID 34745751, PMID 41694690).

    Gate Three: The Eyelids, the most underestimated of the four gates. Bieyma et al. (PMID 21721832) measured eyelid spectral transmittance point-by-point across 450 to 650 nanometers, yielding two conclusions that are not moderate: first, the optical density of the eyelids in the short-wave band is significantly higher than values found in prior literature; the average optical density in this band is 2.1 ± 0.3, with inter-subject optical density differences ranging approximately 1.0 (equivalent to a ten-fold difference in transmittance); second, there is poor correlation between skin pigmentation and eyelid transmittance, with true determinants being large molecules within the eyelids unrelated to wavelength and the total thickness of the eyelids.

    The second point warrants deeper consideration: the eyelid is not a neutral gray filter but a spectrally biased filter, and the action peak of melanopsin lies near 480 nanometers—the segment most severely attenuated, precisely the segment most valued by the weighting algorithm. Therefore, the statement “the eyelids block X% of light” is fundamentally invalid; one must multiply point-by-point across the spectrum and re-weight. Similarly, attempting to estimate eyelid transmittance based on skin tone is directly blocked by the data.

    Conversely, closing the eyes does not equate to an open circuit. Figueras (PMID 26985450) had healthy elderly individuals and early-rise insomnia patients wear masks emitting light through closed eyelids during sleep. Blue light flashes significantly delayed the onset of melatonin (34 ± 30 minutes; red light control: 6 ± 30 minutes), delayed sleep onset by approximately 46 minutes, while flashing red light produced no such effect. Closing the eyes creates a narrow path, yet a measurable quantity of blue light still passes through.

    Gate Four: Intraocular Media, Primarily the Lens. The characteristic of this level is not individual variation but age: transmittance of the lens to short wavelengths decreases with age (PMID 33772847 lists lens transmittance spectra and pupil size as two factors influencing light-induced melatonin suppression; PMID 33020575 provides in vivo methods for measuring lens density and transmittance spectra; PMID 18757473 discusses the role of aging eyes in circadian light sensing). The same moon sends two different spectral distributions to the retina for a twenty-year-old versus a seventy-year-old.

    The threshold side is equally impure. The 1.5 melanopic lux value originates from modeling existing datasets (PMID 30697806); the same paper identifies a saturation point at 305 melanopic lux. Earlier dose-response studies (PMID 10922269) measured a different steepness—achieving half the maximum inhibition with an intensity only slightly above one percent of maximum brightness. At the individual level, ED50 can span an order of magnitude (PMID 31138694); adolescents and adults respond differently to different light intensities (PMID 30803301); and gender differences only appear above 400 lux, showing no significance in the 10 to 200 lux range (PMID 39041348)—in the extremely low illuminance zone, whether individual differences narrow or merely fall below detection limits remains indistinguishable. There is also strong boundary evidence: fully blind individuals still experience significant melatonin suppression when exposed to white light within the eye (PMID 29625838). The pathway does not rely on image-forming vision; thus, “can one receive it with eyes closed” and “can one see” are two distinct matters.

    Placing these facts together, the honest answer is neither “enough” nor “insufficient”: an inter-subject difference in eyelid optical density of 1.0 alone is sufficient to flip the conclusion once; curtains, pupils, and lenses will each flip it again. The real issue is—this chain has never been measured end-to-end under real-world conditions. Most laboratory moonlight dose experiments involve dark rooms, open eyes, and fixed spectra; real moonlight exposure occurs in bedrooms where individuals lie on their sides with eyes closed, curtains vary by person, and lens age varies by person. The independent variables at both ends do not coincide; it is like using a ruler calibrated under condition A to measure condition B.

    To close this chain, it is technically feasible: wearable spectral radiometers can perform in vivo dose measurement (PMID 36433426), closed-eye pupil measurement is mature (PMID 34745751, PMID 41694690), and reporting guidelines have listed which parameters should be reported (PMID 31281903; see PMID 28545361 for the validity of melanopic lux as a measured quantity). The lack is not tools, but willingness to wear sensors on a full moon night, pull curtains of varying degrees, close eyes, and measure these discount rates simultaneously.

    The Su Wen · Jin Gui Zhen Lun states, “There are eight winds in heaven, five in the meridians”—there are eight types of heavenly winds, but only five enter the human body. The Ling Shu and Shang Han Lun discuss “transmission,” referring to the transformation of qi as it enters the body. The ancients never believed that “what heaven gives” equals “what the person receives”; intermediaries were not seen as pipes but as workshops: what enters is no longer the original substance upon exit. The eyelid’s transmittance curve, particularly severe at 480 nanometers, represents the first process in this workshop—it is not merely weakening moonlight but transforming it into another kind of light.

    Evidence and Limitations

    Verified: Eyelid spectral transmittance is measurable and predictable; short-wave optical density is significantly higher than early estimates, inter-subject optical density differences range approximately 1.0 with poor correlation to skin pigmentation (PMID 21721832); blue light intervention through closed eyelids significantly delays DLMO and sleep onset (PMID 26985450); pupil size and light response can be continuously measured under closed-eye states (PMID 34745751, PMID 41694690); pupil constriction significantly alters the relationship between melanopic EDI and melatonin suppression (PMID 38348477); lens transmittance spectra and pupil size are factors influencing light-induced melatonin suppression (PMID 33772847), and lens density and transmittance spectra can be measured in vivo (PMID 33020575); melatonin suppression is primarily driven by melanopsin, can start as low as approximately 1.5 melanopic lux, and saturates at approximately 305 melanopic lux (PMID 30697806); suppression is extremely steep at low doses (reaching half-maximum inhibition with intensity exceeding ~1% of maximum brightness, PMID 10922269); ED50 varies by an order of magnitude between individuals (PMID 31138694); fully blind individuals still experience significant melatonin suppression with in vivo white light exposure (PMID 29625838); melanopic lux predicts suppression better than standard illuminance (PMID 28545361); parameters for light exposure reporting have established standards (PMID 31281903).

    Partially Verified: The actual contribution of visible light transmittance from residential windows and curtains to circadian doses—existing glass transmittance data largely originates from other applications, lacking bedroom measurements weighted by melanopic; whether gender differences exist at low illuminance boundaries (PMID 39041348 shows significance above 400 lux but not in 10–200 lux, though statistical power in this range is inherently limited); whether the pupil is still modulated under closed-eye states to a degree sufficient to alter melanopic dose.

    To Be Verified: Whether the actual retinal melanopic EDI after four serial stages (“windows + curtains + eyelids + lens”) from outdoor moonlight of 0.1–0.3 lux falls below 1.5 melanopic lux—no end-to-end measurement exists; variance decomposition of discount rates at each level (which contributes most) is unknown; whether morphological individual differences such as eyelid thickness can be predicted by externally measurable indicators.

    Related Reading

    Evidence Level: B
    References: PMID 21721832, PMID 34745751, PMID 41694690, PMID 26985450, PMID 38348477, PMID 33772847, PMID 33020575, PMID 18757473, PMID 30697806, PMID 10922269, PMID 31138694, PMID 30803301, PMID 39041348, PMID 29625838, PMID 28545361, PMID 31281903, PMID 36433426


    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.

  • 换一种超额死亡算法,地磁边际增量还在不在?

    Evidence level: B

    配对设计把气象混杂减掉了,但”超额死亡”本身是一个算法产物。Serfling、泊松回归、率差法在同一份数据上给出的数不一样——这不是稳健性检验的附属品,它是结论的一部分。

    正文

    一 同一个设计,三把尺子量出三个数

    配对年份设计的核心动作是”减”:把气象条件相同的两个年份的死亡数相减,剩下的算地磁。但被减掉的那个”基线”不是一个数,是一个算法。

    美国四项方法对比(PMID 19453440)显示,泊松回归给出年均 25,470 例流感相关死亡,peri-season 率差法给出 22,454 例,量级接近;但用夏季基线做率差法时系统性偏高。西班牙的方向相反——广义线性模型的估计系统性低于 Serfling,后者在 65 岁以上给出每十万人 57 与 164 例(PMID 22284214)。加拿大的二十年数据结论更直接:Serfling 估计高度依赖”每年排除多少周”,且大多数排除规则下低于泊松模型;拟合出的周期性基线会高估冬季的非流感死亡,从而低估流感超额,甚至产生负的年度估计(PMID 24312225)。

    三个国家、三种方向。这说明不存在”更保守的方法”,只有”错法不同”。

    这个设计要问的问题本身有据可依:全美 2000–2019 年数据显示,地磁 Kp 指数的六个月移动平均与呼吸道感染死亡显著相关(PMID 42350784),配对设计要检验的正是”气象扣掉之后,这份关联还剩多少”。而绝对湿度与流感活动的关系在中国随气候带反转(PMID 35218485)——这提示”气象”不是一个可以随手扣掉的常数。

    二 最硬的一条:方法本身是估计值的预测因子

    2018 年一项系统综述纳入 103 项流感死亡负担研究做 meta-regression,结论很直白——建模技术本身是估计值的重要预测因子:以乘数法为参照,Serfling 型模型给出的年度年龄标准化超额死亡率平均高出约 40/10 万(95%CI 3–77,P=0.033);同一批研究里,死因范围从”肺炎与流感”扩到”全因”,平均再增加 43/10 万(95%CI 32–54)(PMID 28679157)。

    这对地磁研究的意义很直接:如果三法在同一份数据上的系统性差异(每十万人几十例的量级)大于地磁边际增量本身,那么”地磁显著”这个结论在很大程度上是方法选的,不是数据给的。

    三 配对设计不能自动消除方法偏倚——它分成两半

    配对设计的威力在减法。但方法偏倚在减法里的命运不一样:一部分是共同偏倚——只要两个年份的死亡基线形状相同,相减时抵消;另一部分是差分偏倚——当两个年份的死亡时间分布本身不同时,它留了下来,直接混进边际增量。

    最典型的差分偏倚是死亡位移(harvesting):一次扰动提前带走了本来会在数周后死亡的人。JAMA Netw Open 对 2020–2024 年全球数据的评估显示,位移是普遍现象而非个案(PMID 41609824);BMC 的方法学研究给出了个体层面的反事实校正方案(PMID 37853353)。放回配对设计:若地磁活跃年的死亡曲线更”提前”,Serfling 基线会低估它、泊松模型会部分吸收它、率差法取决于基线期怎么划——三个方法对同一个位移的反应不同,这个差额不是地磁效应。

    四 不确定性区间同样是方法的一部分

    估计值之外还有区间。在过离散的计数序列上直接套泊松,会得到过窄的置信区间——2026 年 Epidemiol Infect 的模拟研究量化了这一点:泊松假设导致对有效再生数的估计”过度自信”,负二项与准泊松的区间显著更宽(PMID 42178993)。超额死亡领域同样如此:Popul Dev Rev 的敏感性分析发现,仅改变基线方法,立陶宛的超额死亡率估计可以差近 4 倍;仅改变参考期(2010–2019 vs 2017–2019),比利时的估计差 20%(PMID 35600716)。WHO 的方法学报告(PMID 36517599)、更灵活的替代框架(PMID 35383642)与贝叶斯时空变系数模型(PMID 40661232)处理的是同一件事:区间本身也是模型产物。

    五 该怎么做

    第一,方法事前预注册,并把三法并行跑完,报告”方法 × 纬度”交互,而不是事后挑一个最漂亮的。

    第二,把方法差异本身当成一个可报告的效应量:报告”三法估计的极差 ÷ 地磁边际增量”这个比值。比值远大于 1,结论就不可信。

    第三,别用 AIC 裁决方法——AIC 差 4 分以内应视为等价,不能据此选模型(PMID 22095719);暴露—滞后—反应场景下另有既定做法(PMID 20812303、PMID 29447338、PMID 23297754、PMID 35436963)。

    第四,分层与及时评估:按年龄与亚型分层能暴露位移效应(PMID 30968823),AIC 驱动的模型选择步骤(PMID 26781954)与早期建立的及时监测方法(PMID 9209852)可直接复用。

    六 易学落点

    “数”从来不是中立的。同一片天时,用不同历法、不同起算点去量,得到不同的”气数”。《素问》讲”必先岁气,无伐天和”——先要定准这一年的”岁气”,而”定准”这件事本身就包含”用哪套算法”。

    本篇的分辨点:若三法的差异主要是共同偏倚,配对设计仍可信,结论稳健;若差异主要来自差分偏倚(死亡位移),那”地磁边际增量”必须带着方法名一起说——它是”Serfling 下的增量”还是”泊松下的增量”,是两个命题。

    证据与局限

    已验证: 同一份数据上四种超额死亡方法量级接近但存在系统性差异(PMID 19453440);Serfling 估计高度依赖排除周数规则并可能产生负估计(PMID 24312225);建模技术本身是流感死亡负担估计值的显著预测因子(PMID 28679157);泊松假设在过离散序列上导致区间过窄(PMID 42178993);改变基线方法与参考期可使超额死亡估计相差数倍(PMID 35600716)。

    部分验证: 死亡位移的全球普遍性(PMID 41609824,生态学汇总层面);个体层面反事实校正的可行性(PMID 37853353);灵活超额死亡框架与贝叶斯时空变系数模型的表现(PMID 35383642、PMID 40661232,模拟研究为主)。

    待验证: ①”方法偏倚在配对设计中分成共同偏倚与差分偏倚”这一分解至今没有任何研究直接量化过,因而无法判断地磁边际增量中有多少来自差分偏倚;②三法在”地磁活跃年 vs 平静年”这一特定对比下的估计差异与覆盖概率从未被比较过;③若差分偏倚确实占主导,是否存在一个可事前指定的方法组合能把偏倚压到边际增量以下,仍是开放问题。

    相关阅读

    证据等级: B
    参考文献: PMID 19453440, PMID 24312225, PMID 28679157, PMID 22284214, PMID 30968823, PMID 9209852, PMID 26781954, PMID 35600716, PMID 42178993, PMID 41609824, PMID 37853353, PMID 35383642, PMID 40661232, PMID 36517599, PMID 22095719, PMID 20812303, PMID 29447338, PMID 23297754, PMID 35436963, PMID 42350784, PMID 35218485


    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。

    Does the magnetic marginal increment still hold with an alternative excess mortality algorithm?

    Evidence level: B

    A matched-pair design mitigates meteorological confounding, yet “excess mortality” itself is an algorithmic construct. Serfling, Poisson regression, and rate difference methods yield different figures on the same dataset—this is not merely an artifact of robustness testing; it constitutes part of the conclusion.

    Main Text

    I. One Design, Three Rulers Yielding Three Numbers

    The core operation of the matched-pair year design is “subtraction”: subtracting death counts from two years with identical meteorological conditions to isolate the geomagnetic signal. However, the subtracted “baseline” is not a single number; it is an algorithm.

    Comparisons across four US methods (PMID 19453440) show that Poisson regression yields an annual average of 25,470 influenza-related deaths, while the peri-season rate difference method yields 22,454; their magnitudes are close. However, using a summer baseline for the rate difference method results in a systematic overestimation. The direction is reversed in Spain, where generalized linear models estimate systematically lower values than Serfling; the latter reports 57 and 164 cases per 100,000 for those aged 65+ (PMID 22284214). Canada’s twenty-year dataset offers a more direct conclusion: Serfling estimates are highly dependent on “how many weeks are excluded annually,” and under most exclusion rules, they fall below Poisson models; the fitted periodic baseline overestimates non-influenza winter deaths, thereby underestimating influenza excess and even generating negative annual estimates (PMID 24312225).

    Three countries, three divergent directions. This indicates there is no “more conservative method,” only “different methods applied.”

    The question this design seeks to answer is grounded in evidence: US data from 2000–2019 show a significant correlation between the six-month moving average of geomagnetic Kp indices and respiratory infection deaths (PMID 42350784); the matched-pair design aims to verify “how much of this association remains after meteorological factors are subtracted.” Meanwhile, the relationship between absolute humidity and influenza activity reverses across climate zones in China (PMID 35218485)—suggesting “meteorology” is not a constant that can be arbitrarily subtracted.

    II. The Hardest Constraint: Methodology Itself Is a Predictor of Estimates

    A 2018 systematic review including 103 influenza mortality burden studies conducted a meta-regression with a straightforward conclusion: modeling technique itself is a significant predictor of estimates. Using the multiplier method as a reference, Serfling-type models yield an annual age-standardized excess mortality rate approximately 40/100,000 higher on average (95%CI 3–77, P=0.033); within the same body of research, expanding the cause-of-death scope from “pneumonia and influenza” to “all-cause” adds another average 43/100,000 (95%CI 32–54) (PMID 28679157).

    The implication for geomagnetic research is direct: if the systematic differences between three methods on the same dataset (on the order of tens per 100,000) exceed the geomagnetic marginal increment itself, then the conclusion of “geomagnetic significance” is largely determined by method selection rather than data.

    III. Matched-Pair Design Does Not Automatically Eliminate Methodological Bias—It Splits It

    The power of the matched-pair design lies in subtraction. However, the fate of methodological bias within this operation differs: one portion is common bias—offset when subtracted if the baseline shapes of both years are identical; the other is differential bias—which persists when the temporal distributions of deaths differ between years and directly contaminates the marginal increment.

    The most typical differential bias is death displacement (harvesting), where a single disturbance prematurely removes individuals who would have died weeks later. An evaluation of global data from 2020–2024 by JAMA Netw Open shows displacement is a universal phenomenon, not an outlier (PMID 41609824); BMC methodological research provides individual-level counterfactual correction schemes (PMID 37853353). Reintroducing the matched-pair design: if death curves in geomagnetic active years are more “advanced,” Serfling baselines will underestimate them, Poisson models will partially absorb them, and the rate difference method depends on how the baseline period is defined—three methods react differently to the same displacement; this discrepancy is not a geomagnetic effect.

    IV. Confidence Intervals Are Also Part of the Methodology

    Beyond point estimates lie intervals. Directly applying Poisson assumptions to overly discrete count sequences yields confidence intervals that are too narrow—a 2026 simulation study in Epidemiol Infect quantified this: Poisson assumptions lead to “overconfidence” in effective reproduction number estimates, while negative binomial and quasi-Poisson intervals are significantly wider (PMID 42178993). The same holds true in the excess mortality domain: a sensitivity analysis in Popul Dev Rev found that altering only the baseline method can cause Lithuanian excess mortality estimates to differ by nearly 4-fold; changing only the reference period (2010–2019 vs. 2017–2019) causes Belgian estimates to differ by 20% (PMID 35600716). WHO methodological reports (PMID 36517599), more flexible alternative frameworks (PMID 35383642), and Bayesian spatiotemporal varying coefficient models (PMID 40661232) address the same issue: intervals are also model products.

    V. What Should Be Done

    First, pre-register methods and run all three concurrently, reporting the “method × latitude” interaction rather than selecting the most aesthetically pleasing one post-hoc.

    Second, treat methodological differences themselves as a reportable effect size: report the ratio of “range of three-method estimates ÷ geomagnetic marginal increment.” If this ratio is far greater than 1, the conclusion is unreliable.

    Third, do not use AIC to adjudicate methods—differences within 4 points should be considered equivalent; models should not be selected based solely on this metric (PMID 22095719); established practices exist for exposure-lag-response scenarios (PMID 20812303, PMID 29447338, PMID 23297754, PMID 35436963).

    Fourth, stratify and evaluate timely: stratification by age and subtype can reveal displacement effects (PMID 30968823); AIC-driven model selection steps (PMID 26781954) and early-established timely monitoring methods (PMID 9209852) can be directly reused.

    VI. Yijing Application Points

    Numbers are never neutral. Under the same celestial timing, using different calendars or starting points yields different “qi counts.” The Su Wen states, “First determine the annual qi; do not violate the harmony of heaven and man”—one must first accurately define the year’s “annual qi,” and the act of “accurate definition” inherently includes “which algorithm to use.”

    The distinguishing point of this paper: if the differences among three methods stem primarily from common bias, the matched-pair design remains credible and conclusions are robust; if differences arise mainly from differential bias (death displacement), the “geomagnetic marginal increment” must be stated alongside its method name—it is an “increment under Serfling” or an “increment under Poisson,” two distinct propositions.

    Evidence and Limitations

    Verified: Four excess mortality methods yield similar magnitudes but exhibit systematic differences on the same dataset (PMID 19453440); Serfling estimates are highly dependent on exclusion week rules and may produce negative estimates (PMID 24312225); modeling technique is a significant predictor of influenza mortality burden estimates (PMID 28679157); Poisson assumptions on over-discrete sequences lead to overly narrow intervals (PMID 42178993); altering baseline methods and reference periods can cause excess mortality estimates to differ by orders of magnitude (PMID 35600716).

    Partially Verified: Global universality of death displacement (PMID 41609824, ecological summary level); feasibility of individual-level counterfactual corrections (PMID 37853353); performance of flexible excess mortality frameworks and Bayesian spatiotemporal varying coefficient models (PMID 35383642, PMID 40661232, primarily simulation studies).

    To Be Verified: ① The decomposition of “methodological bias into common and differential bias within matched-pair designs” has never been directly quantified in any study to date, making it impossible to determine the proportion of differential bias within the geomagnetic marginal increment; ② The difference in estimates and coverage probability between three methods under the specific comparison of “geomagnetic active years vs. calm years” has never been compared; ③ If differential bias indeed dominates, whether a pre-specified method combination exists to suppress bias below the marginal increment remains an open question.

    Related Reading

    Evidence Level: B
    References: PMID 19453440, PMID 24312225, PMID 28679157, PMID 22284214, PMID 30968823, PMID 9209852, PMID 26781954, PMID 35600716, PMID 42178993, PMID 41609824, PMID 37853353, PMID 35383642, PMID 40661232, PMID 36517599, PMID 22095719, PMID 20812303, PMID 29447338, PMID 23297754, PMID 35436963, PMID 42350784, PMID 35218485


    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.

  • 共变不等于共享:XBP1s 的 12 小时振荡器与月级成分,这一刀卡在 30 天这个数字上

    Evidence level: B

    用细胞系做”ER 应激诱导 → 同时看 12 小时成分与 29.5 天成分”是最便宜的直接检验,但在动手之前有三个坑必须拆:共变不等于共享、12 小时谱峰自己就有三种来源、以及 30 天培养根本测不出 29.5 天——第三个坑是数字层面的硬伤。

    正文

    《素问·六微旨大论》讲”天气始于甲,地气始于子,子甲相合,命曰岁立”。中国历法有个很少被提起的特点:它从来不试图造一只”总的钟”来记时间。它摆出两套周期——十干、十二支——让它们各自转,靠两套周期互相错开产生的组合来标记唯一时刻。六十年一轮的甲子,不是某只大钟走了一圈,是两只小钟的最小公倍数。

    这一篇要检验的东西,正好撞在这种”两套周期”的结构上。

    一、要被检验的命题

    此前的文章已经理清两条线索。

    一条是 XBP1s 的 12 小时振荡器。哺乳动物肝脏里有一套细胞自主的、独立于昼夜钟的 12 小时节律,核心是未折叠蛋白反应(UPR)的转录因子 XBP1s 与上游感受器 IRE1α(PMID 31935211、PMID 41845938);它驱动的是整条”中心法则信息流”——转录起始、mRNA 加工、核糖体生物合成、蛋白质折叠(PMID 31935211),通过核斑点的相分离动力学实现 12 小时节律(PMID 34985945),并与脂质代谢和膜流动性稳态直接挂钩(PMID 33277471)。到 2026 年,组蛋白甲基转移酶复合体亚基 RBBP5 被证明是这套振荡器的上游表观调控节点(PMID 42848836、PMID 39314427)。最关键的一条旁证是:把昼夜钟核心去掉之后,约 12 小时的 mRNA 与蛋白质代谢节律仍在(PMID 37205336)——这套振荡器是真的自主,不是昼夜钟的影子。

    另一条线索是人类近月钟的分子实体至今是空的。海洋里的近月钟已被解剖得很细:环节虫的月亮钟由夜间光照授时,在昼夜钟被 CK1δ/ε 抑制打乱后照常运转(PMID 24075994、PMID 39506296);海蚊 Clunio 的自由运行半潮月周期由”数昼夜钟周期数”决定(PMID 38764210)。人类侧的论证这几年确实在增加——2026 年有文章系统给出”人类存在近月钟”的演化与生理论据(PMID 42380750)——但分子实体仍然没找到。

    于是 backlog 上这条课题的逻辑很清楚:既然 XBP1s 的 12 小时振荡器被论证为远古潮汐钟的陆地遗留(PMID 33449146、PMID 41845938),而潮汐钟与近月钟在海洋生物里常在同一体内共存(PMID 24075994),那么”两者是否共享上游开关”就是定位人类近月钟最便宜的一刀。候选上游有三个:ER 应激负荷本身(UPR 在生殖生理中本就随周期剧烈波动,PMID 28397763)、公共激酶节点(CK1 家族)、以及频率拍频关系。

    二、这一刀砍下去会先撞上三个坑

    坑一:共变不等于共享。 这是最要紧、也最容易被跳过的一步。两个完全独立的振荡器,只要频率接近整数比,在有限长度的数据上就会表现出共变——拍频与锁相在观察层面无法与”共享上游”区分。12 小时与 29.5 天恰好接近 1:59 的整数关系。所以”12 小时成分与月级成分共变”这个结果本身不构成共享上游的证据;必须先用主动扰动(敲低 XBP1s、抑制 CK1δ/ε)去看两类成分的阻尼率与相移是否同步改变,才谈得上共享。这道因果门槛不设,后面所有数据都白做。

    坑二:12 小时谱峰自己就有三种来源。 2026 年 CHORD 这篇方法学文章把这件事钉死了:转录组里一个 12 小时谱峰可能来自三种机制完全不同的过程——非正弦 24 小时波形的二次谐波、带独立相位的自主 12 小时振荡器(如 IRE1α–XBP1s 轴)、或两个反相 24 小时过程”合成减降解”造成的表观 12 小时(PMID 42703951)。三者在生物学含义上彼此相反,但对 JTK_CYCLE、RAIN、Cosinor 这类标准检测器完全不可区分(PMID 20876817、PMID 25326247)。不做三元分解就开始谈”XBP1s 的 12 小时节律”,本身就是方法错误。

    坑三:30 天测不出 29.5 天。 这条最直接:30 天培养光谱只有大约一个完整周期。一个周期给得出”升起来又落下去”,给不出周期长度,也给不出存在性检验的统计量。要估计 29.5 天周期并做显著性检验,最低限度需要 3-4 个完整周期,即 90-120 天以上的连续采样;而 100 天以上的连续培养本身就要面对细胞状态漂移(传代累积、汇合度、营养耗竭、衰老),长周期的”趋势项”与”周期项”在有限数据上很难分离——哺乳动物细胞在培养中的钟基因表达特性与组织内本就不一致(PMID 16448219)。也就是说,原课题写法里的”≥30 天长程表达谱”在数字上是不成立的。

    三、还有一个隐藏的授时源

    培养箱里没有月亮,但有一个”实验者的作息”。 换液、传代、培养基批次更换、CO₂ 罐更换、周末与工作日的人力差异——这些都是周期性的人为输入。培养箱里的一批细胞,实际上一直跑在一个由实验者排班构成的外部周期里。这些人工周期的频率往往与 29.5 天毫无关系(换液若每 3-4 天一次,引入的是 3-3.5 天成分),但”与目标频率不同”不等于“不会污染目标频率的估计”——非整数比的人为周期在有限采样下通过窗函数泄漏进长周期估计是常事。所以这个实验要成立,必须同时记录并报告所有人工周期的时序,做混杂回归;否则看到 29.5 天成分时,第一嫌疑人是排班表,不是月亮钟。

    四、回到”两只钟”

    现在可以说回干支了。古人处理多尺度时间从来不靠”一只总钟”,而是摆几套互质的周期让它们自己错开——甲子六十,是十与十二的最小公倍数,没有任何一只”六十年钟”在走。这套思路的好处是:它天然不要求两套周期共享上游。 而现代生物学检验两条节律是否独立时,默认思路恰好相反:相关则必有共因。易学这边给出的提醒是第三种可能——它们可以不共因,只共频,于是看起来相关。

    这正好是这条课题最该先解决的事:在动手之前,先把”共变”与”共享”之间那道因果门槛说清楚,再把这道门槛实验化。干支在这里不提供证据,只提供一个提问的角度。

    证据与局限

    已验证: ①哺乳动物存在细胞自主、独立于昼夜钟的 XBP1s/IRE1α 12 小时振荡器(PMID 31935211、PMID 37205336、PMID 41845938、PMID 33277471);②该振荡器由核斑点相分离动力学实现,并受 RBBP5 表观调控(PMID 34985945、PMID 42848836、PMID 39314427);③12 小时谱峰存在”谐波 / 自主振荡 / 合成-降解交叉”三种机制来源,标准检测器(JTK_CYCLE、RAIN、Cosinor)无法区分(PMID 42703951、PMID 20876817、PMID 25326247);④海洋环节虫近月钟独立于昼夜钟,由夜间光照授时(PMID 24075994、PMID 39506296);⑤海蚊半潮月周期由数昼夜钟周期数决定(PMID 38764210);⑥UPR/ER 应激参与生殖生理的周期调节(PMID 28397763)。

    部分验证: ①”XBP1s 12 小时振荡器是远古潮汐钟的陆地遗留”有演化论据支撑(PMID 33449146、PMID 41845938),但”潮汐钟与近月钟共享上游”在同一生物体内尚无直接分子证据;②人类存在近月钟的论证已系统化(PMID 42380750),但分子实体空缺;③培养细胞中确有独立相位、可自主运行的钟基因节律(PMID 15620658、PMID 24498336),说明细胞层面做长程节律检测在方法上可行——但可行距离”测得出 29.5 天”还差一个数量级的时间。

    待验证: ①30 天级培养能否检出 29.5 天成分(需先做”周期数 × 采样密度 × 噪声”的功效标定);②培养箱内人工周期(换液/传代/批次/实验者作息)向长周期估计泄漏的量级;③主动扰动(XBP1s 敲低 / CK1δ/ε 抑制)下,12 小时与月级两类成分的阻尼率-相移是否同步改变(共享上游的因果判据);④ER 应激诱导剂(衣霉素、毒胡萝卜素)梯度本身是否足以在细胞系里重建 12 小时振荡的振幅-相位结构(PMID 22038282、PMID 22852048、PMID 31630283)。

    相关阅读

    证据等级: B(XBP1s 12 小时振荡器的分子身份、CHORD 的谱峰三元分解、环节虫近月钟的独立性均有 PubMed 直接支撑;但”与近月钟共享上游”本身仍是待检验假说,且本文指出原实验设计在时间尺度上存在硬伤)
    参考文献: PMID 31935211, PMID 37205336, PMID 41845938, PMID 33277471, PMID 34985945, PMID 42848836, PMID 39314427, PMID 42703951, PMID 20876817, PMID 25326247, PMID 24075994, PMID 39506296, PMID 38764210, PMID 42380750, PMID 33449146, PMID 28397763, PMID 16448219, PMID 15620658, PMID 24498336, PMID 22038282, PMID 22852048, PMID 31630283


    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。

    Covariation Does Not Equal Sharing: The 12-Hour Oscillator and Monthly Component of XBP1s, This Cut Hits the 30-Day Number

    Evidence level: B

    Using cell lines to perform “ER stress induction → simultaneous observation of 12-hour and 29.5-day components” is the cheapest direct test, but before proceeding, three pitfalls must be addressed: covariation does not imply sharing, the 12-hour spectral peak itself has three distinct sources, and a 30-day culture cannot measure a 29.5-day cycle—the third is a hard flaw at the numerical level.

    Main Text

    The Su Wen · Liu Wei Zhi Da Lun states, “Heavenly Qi begins with Jia; Earthly Qi begins with Zi. When Zi and Jia combine, it is called the establishment of the year.” Chinese calendrical systems possess a rarely discussed characteristic: they never attempt to construct a single “master clock” to record time. Instead, they establish two cycles—the Ten Stems and Twelve Branches—allowing them to rotate independently. Unique moments are marked by the combinations generated when these two cycles offset each other. The sixty-year Jiazi cycle is not the result of one large clock completing a revolution; it is the least common multiple of two smaller clocks.

    The subject of this paper collides precisely with this “two-cycle” structure.

    I. The Proposition to be Tested

    Previous articles have clarified two lines of inquiry.

    The first is the 12-hour oscillator of XBP1s. In mammalian livers, there is a cell-autonomous rhythm independent of the circadian clock, centered on the transcription factor XBP1s and its upstream sensor IRE1α (PMID 31935211, PMID 41845938). It drives the entire “central dogma information flow”—transcription initiation, mRNA processing, ribosome biogenesis, and protein folding (PMID 31935211)—through phase separation dynamics of nuclear speckles to achieve a 12-hour rhythm (PMID 34985945), and is directly linked to lipid metabolism and membrane fluidity homeostasis (PMID 33277471). By 2026, the subunit RBBP5 of histone methyltransferase complexes has been proven as an upstream epigenetic regulatory node for this oscillator (PMID 42848836, PMID 39314427). The most critical corroborating evidence is that after removing the core circadian clock, ~12-hour mRNA and protein metabolic rhythms persist (PMID 37205336)—this oscillator is truly autonomous, not a shadow of the circadian clock.

    The second line of inquiry is that the molecular entity of the human near-lunar clock remains empty to date. Near-lunar clocks in marine environments have been dissected in detail: the lunar clock in annelids is entrained by nocturnal light, continuing to operate even when disrupted by CK1δ/ε inhibition of the circadian clock (PMID 24075994, PMID 39506296); the free-running semilunar cycle in the sea mosquito Clunio is determined by “counting circadian cycles” (PMID 38764210). Arguments regarding the human side have indeed increased in recent years—2026 saw an article systematically providing evolutionary and physiological arguments for “the existence of a near-lunar clock in humans” (PMID 42380750)—but the molecular entity remains undiscovered.

    Thus, the logic on this backlog is clear: since the XBP1s 12-hour oscillator has been argued as a terrestrial remnant of ancient tidal clocks (PMID 33449146, PMID 41845938), and tidal clocks often coexist with near-lunar clocks within the same organism in marine species (PMID 24075994), determining whether “they share an upstream switch” is the cheapest cut to locate the human near-lunar clock. There are three candidate upstream nodes: ER stress load itself (UPR fluctuates剧烈 in reproductive physiology across cycles, PMID 28397763), a shared kinase node (CK1 family), or frequency beat relationships.

    II. Three Pitfalls This Cut Will Hit First

    Pitfall 1: Covariation does not imply sharing. This is the most critical yet often skipped step. Two completely independent oscillators, if their frequencies are close to integer ratios, will exhibit covariation over finite data lengths—beat frequency and phase locking are indistinguishable from “shared upstream” at the observational level. The 12-hour and 29.5-day cycles are close to a 1:59 integer ratio. Therefore, the result of “covariation between the 12-hour component and lunar component” does not constitute evidence for a shared upstream; one must first observe whether the damping rates and phase shifts of both component types change synchronously via active perturbation (XBP1s knockdown, CK1δ/ε inhibition) before claiming sharing. Without setting this causal threshold, all subsequent data is wasted.

    Pitfall 2: The 12-hour spectral peak itself has three sources. A 2026 CHORD methodological article cemented this: a 12-hour spectral peak in transcriptomics may arise from three fundamentally different mechanisms—secondary harmonics of non-sinusoidal 24-hour waveforms, an autonomous 12-hour oscillator with independent phase (such as the IRE1α–XBP1s axis), or an apparent 12-hour cycle caused by “synthetic subtraction and degradation” of two anti-phase 24-hour processes (PMID 42703951). These three differ biologically but are completely indistinguishable to standard detectors like JTK_CYCLE, RAIN, and Cosinor (PMID 20876817, PMID 25326247). Discussing “XBP1s’s 12-hour rhythm” without ternary decomposition is itself a methodological error.

    Pitfall 3: A 30-day culture cannot measure a 29.5-day cycle. This is most direct: a 30-day culture spectrum yields only approximately one complete cycle. One cycle provides “rising and falling,” but not the period length, nor statistical power for existence testing. To estimate a 29.5-day cycle and perform significance testing, a minimum of 3-4 complete cycles is required, i.e., continuous sampling over 90-120+ days; continuous culture beyond 100 days faces cell state drift (passage accumulation, confluence, nutrient depletion, senescence), making it difficult to separate “trend terms” from “cycle terms” in finite data—mammalian cell clock gene expression characteristics in culture are inherently inconsistent with those in tissues (PMID 16448219). In short, the “≥30-day long-term expression profile” in the original proposal is numerically invalid.

    III. A Hidden Entrainment Source

    There is no moon in the incubator, but there is an “experimenter’s schedule.” Media changes, passaging, media batch replacements, CO₂ cylinder changes, and differences in human resources between weekends and weekdays are all periodic artificial inputs. A batch of cells in an incubator actually runs continuously within an external cycle constructed by the experimenter’s roster. The frequencies of these artificial cycles often have no relation to 29.5 days (e.g., if media is changed every 3-4 days, a 3-3.5 day component is introduced), but being “different from the target frequency” does not equal “not polluting the estimation of the target frequency”—non-integer ratio artificial cycles frequently leak into long-period estimates via window functions under finite sampling. Therefore, for this experiment to hold, all artificial cycle timings must be recorded and reported, and confounding regression performed; otherwise, when a 29.5-day component is observed, the first suspect is the roster, not the lunar clock.

    IV. Returning to “Two Clocks”

    We can now return to the Stems and Branches. Ancient people handled multi-scale time not with a “single master clock,” but by setting several coprime cycles to offset each other independently—the sixty-year Jiazi is the least common multiple of ten and twelve, with no “sixty-year clock” moving. The advantage of this approach is: it naturally does not require the two cycles to share an upstream source. Conversely, modern biological testing of whether two rhythms are independent defaults to the opposite logic: correlation implies a common cause. The reminder from Yijing Studies offers a third possibility—they may not share a common cause, only share frequency, thus appearing correlated.

    This is exactly what this project should resolve first: before acting, clarify the causal threshold between “covariation” and “sharing,” then experimentalize it. The Stems and Branches provide no evidence here, only a perspective for questioning.

    Evidence and Limitations

    Verified: ① Mammals possess a cell-autonomous XBP1s/IRE1α 12-hour oscillator independent of the circadian clock (PMID 31935211, PMID 37205336, PMID 41845938, PMID 33277471); ② This oscillator is realized via nuclear speckle phase separation dynamics and regulated epigenetically by RBBP5 (PMID 34985945, PMID 42848836, PMID 39314427); ③ The 12-hour spectral peak has three mechanistic sources (“harmonic / autonomous oscillator / synthetic-degradation crossover”), which standard detectors (JTK_CYCLE, RAIN, Cosinor) cannot distinguish (PMID 42703951, PMID 20876817, PMID 25326247); ④ Annelid near-lunar clocks are independent of circadian clocks and entrained by nocturnal light (PMID 24075994, PMID 39506296); ⑤ The semilunar cycle in sea mosquitoes is determined by counting circadian cycles (PMID 38764210); ⑥ UPR/ER stress participates in periodic regulation of reproductive physiology (PMID 28397763).

    Partially Verified: ① “The XBP1s 12-hour oscillator is a terrestrial remnant of ancient tidal clocks” has evolutionary support (PMID 33449146, PMID 41845938), but “shared upstream between tidal and near-lunar clocks” lacks direct molecular evidence within the same organism; ② Arguments for human near-lunar clocks are systematic (PMID 42380750), but the molecular entity is missing; ③ Independent-phase, autonomously running clock gene rhythms exist in cultured cells (PMID 15620658, PMID 24498336), indicating methodological feasibility for long-term rhythm detection at the cellular level—but there is still an order-of-magnitude time gap before measuring a 29.5-day cycle.

    To be Verified: ① Whether 30-day-level cultures can detect 29.5-day components (requires first conducting “cycle count × sampling density × noise” power calibration); ② The magnitude of leakage from artificial cycles within incubators (media changes/passaging/batches/experimenter schedule) into long-period estimates; ③ Whether damping rates and phase shifts for the 12-hour and lunar components change synchronously under active perturbation (XBP1s knockdown / CK1δ/ε inhibition)—the causal criterion for shared upstream; ④ Whether gradients of ER stress inducers (calyculin A, bryostatin) alone are sufficient to reconstruct the amplitude-phase structure of the 12-hour oscillator in cell lines (PMID 22038282, PMID 22852048, PMID 31630283).

    Related Reading

    Evidence Level: B (The molecular identity of the XBP1s 12-hour oscillator, CHORD’s ternary spectral decomposition, and the independence of annelid near-lunar clocks have direct PubMed support; however, “sharing upstream with the near-lunar clock” remains a hypothesis to be tested, and this paper points out a hard flaw in the original experimental design regarding the timescale)
    References: PMID 31935211, PMID 37205336, PMID 41845938, PMID 33277471, PMID 34985945, PMID 42848836, PMID 39314427, PMID 42703951, PMID 20876817, PMID 25326247, PMID 24075994, PMID 39506296, PMID 38764210, PMID 42380750, PMID 33449146, PMID 28397763, PMID 16448219, PMID 15620658, PMID 24498336, PMID 22038282, PMID 22852048, PMID 31630283


    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.

  • 名字还叫 ependymin,岗位早就不在脑里:人类 ependymin 与 proctolin 家族的组织定位扫描

    Evidence level: B

    环节虫月亮钟靶蛋白清单上的两个家族,在人类身上走了两条完全不同的路:一个留了下来但换了岗,一个连岗带人一起撤了。区分这两条路的实验只有一个——组织定位。

    正文

    《素问·灵兰秘典论》安排人体的方式很特别:它不按位置给器官命名,而按职务。”心者,君主之官也”——”心”这个字早就有了,但真正被写进制度的是”君主之官”四个字。这篇文章要处理的问题正好撞在这条规矩上:环节虫月亮钟靶蛋白清单里的 ependymin 家族和 proctolin 神经肽家族,在人类身上现在在哪上班。

    一、”ependymin”这个名字,起初是一次按位置起名

    ependymin 最早从金鱼脑里分离出来,样品取自脑室管膜(ependyma)附近,于是按位置得了名。随后功能证据也齐了:金鱼学会东西之后它的亚细胞分布会变(PMID 3572403);用反义链把它压下去,主动回避记忆的巩固会被阻断(PMID 7561839);把抗体打进眼睛,再生的视顶盖投射就锐化不了(PMID 3370490)。名字定了,故事也讲圆了。问题是——”ependym-“这两个音节说的是在哪捡到的,不是干什么的。这在生物学里是常态,不是意外。

    二、EPDR1:名字没变,岗位搬去了脂肪

    把人类 EPDR1 的功能证据摊开,会得到很尴尬的结果。结构层面,它是溶酶体蛋白,晶体结构显示与细菌脂蛋白转运蛋白超家族同源,管的是疏水分子与脂质的结合搬运(PMID 30729188、PMID 31316816)。功能层面:2019 年 Cell Metabolism 的棕色/白色脂肪细胞分泌组比较蛋白质组学把它鉴定为新一类”batokine”(PMID 31668873);2022 年有研究证明它调节 β 细胞代谢与功能(PMID 36343918),同年另一篇把它定为胰岛素介导血管生成的非经典效应器(PMID 35872017)。再往后:衰老成纤维细胞把它当衰老体液因子分泌、影响细胞外基质(PMID 36497009);特发性肺纤维化中它的水平与病情严重度相关(PMID 41090747);2 型糖尿病病人循环血 EPDR1 与胰岛素抵抗相关(PMID 40642507);多囊卵巢综合征血清里它也升高(PMID 41816836)。

    一句话:人类 EPDR1 干的是脂肪、代谢、血管、成纤维细胞以及一大批肿瘤的活(PMID 39152265、PMID 32935479、PMID 32111877、PMID 33902536)。脑不在清单上。

    那”金鱼脑里的 ependymin”和”人类脂肪里的 EPDR1″是什么关系?序列上有关系——家族演化研究讲清了它一路怎么分化(PMID 30514200);功能上没关系——海绵里同一家族的成员甚至变成了蓝色类胡萝卜素结合蛋白(PMID 37517696)。名字是历史遗产,不构成功能承诺。

    三、proctolin:这条路在脊椎动物里没留下

    更彻底的例子是 proctolin(RYLPT 五肽)。它在节肢动物里是正经神经肽:蝗虫的受体已被鉴定,还能据此设计类似物(PMID 40778513);蟑螂神经肽组学把它列进昼夜钟相关神经肽景观(PMID 41144767);它与其他靶点重叠的神经肽一起调制振荡回路(PMID 37968117、PMID 37333253)。但翻遍脊椎动物,找不到直系同源的内源配体——检索到的”proctolin + human/mammal”文献,几乎全是节肢动物实验被引进跨物种综述,或蝎毒中的结构类似物(PMID 26056922)。这条线上不是”换了岗”,是整个岗位在脊椎动物谱系里被撤掉,只剩受体侧可能的远亲。

    四、组织定位扫描:四个必须先拆的坑

    正因如此,组织定位不是”写文章前的准备工作”,它本身就是决定性的一步——序列比对只告诉你名字在不在,只有定位才告诉你人还在不在岗上。但它有四个坑。

    坑一,脑样本污染。 2022 年 Molecular Psychiatry 一篇报告指出,脑组织采样与组学分析中脉络丛污染极为普遍(PMID 34983929)。而脉络丛恰恰是屏障组织里最可能承载”月亮钟靶蛋白”的部位。于是双向污染:你查到的”脑内表达”可能是脉络丛漏进来的,你查到的”脉络丛表达”可能是采样时带进去的。这个坑不拆,后面所有定位结论都站不住。

    坑二,图谱里没有你要查的地方。 GTEx 把脑分成十几个亚区(PMID 23715323、PMID 32913098),但脉络丛、脑膜、血脑屏障这些屏障组织基本不在采样范围内——你想查的组织,现成图谱里根本不采。

    坑三,分泌蛋白测的是加总。 EPDR1 是分泌型的。循环血里测到的量是所有来源组织的加权平均,不是脑。想用”按月相抽血测 EPDR1″推脑内节律,推不出来。

    坑四,名字的先验会带偏判读。 研究者在人脑单细胞图谱里看到 EPDR1,第一反应往往是被”ependymin 家族”这个名字牵着走的。同一批数据,如果那个基因叫别的名字,判读会不一样。这个坑靠统计去不掉。

    四个坑叠起来,结论是:在拆掉它们之前,”环节虫月亮钟靶蛋白的人类残留”这个问题连提法都不成立。

    五、回到十二官

    《素问·灵兰秘典论》”心者,君主之官也”真正厉害的地方,是把”心”这个字和”君主之官”这个职务绑在了一起,末尾再加一句”凡此十二官者,不得相失也”——十二官的排班表上,没有一个官是按它当初被发现的解剖位置定职务的。

    我们做跨物种靶蛋白扫描,最容易犯的错就是拿名替处:序列对上了,就说”它在人身上也干这个”。易学的规矩是先问”司什么”,再问”在哪”,两问都答完才算认了人。中文里”名不副实”这四个字,在这件事上不是修辞,是操作方法。

    证据与局限

    已验证: ①人 EPDR1 的结构与溶酶体脂质/疏水分子结合功能(PMID 30729188、PMID 31316816);②EPDR1 作为脂肪因子参与代谢、血管生成与 β 细胞功能(PMID 31668873、PMID 36343918、PMID 35872017);③脑组织组学分析中脉络丛污染普遍存在(PMID 34983929);④ependymin 家族跨物种功能高度趋异(PMID 30514200、PMID 37517696)。

    部分验证: ①proctolin 系统在节肢动物中的受体与神经肽景观较清楚(PMID 40778513、PMID 41144767),但脊椎动物侧只剩受体远亲的推测,尚无直接鉴定;②EPDR1 在代谢性疾病中的循环水平多次被报告(PMID 40642507、PMID 41816836),提示它有可测的体液读出通道——但这些报告都是横断面病例-对照,不含时间轴。

    待验证: ①人类脉络丛/脑膜/血脑屏障中 EPDR1 的细胞类型级定位(需先解决污染对照);②上述屏障组织是否存在独立于昼夜节律的 29.5 天表达成分;③脊椎动物中是否存在 proctolin 受体的内源配体(配体孤儿化检验);④跨物种组织定位差异能否反推出”计时残留”的可能落点。

    相关阅读

    证据等级: B
    参考文献: PMID 30729188, PMID 31316816, PMID 31668873, PMID 36343918, PMID 35872017, PMID 36497009, PMID 41090747, PMID 40642507, PMID 41816836, PMID 30514200, PMID 37517696, PMID 40778513, PMID 41144767, PMID 37968117, PMID 37333253, PMID 26056922, PMID 34983929, PMID 23715323, PMID 32913098, PMID 3572403, PMID 7561839, PMID 3370490, PMID 39152265, PMID 32935479, PMID 32111877, PMID 33902536


    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。

    Ependymin Still Named After Its Brain Origin: Tissue Localization Scan of Human Ependymin and the Proctolin Family

    Evidence level: B

    Two families on the annelid moon-clock target protein list have taken two divergent paths in humans: one stayed but changed roles, while the other was removed with both the role and the personnel. The single experiment distinguishing these paths is tissue localization.

    Main Text

    The Su Wen · Ling Lan Mi Dian Lun arranges the human body in a distinctive manner: it does not name organs by location, but by function. “The heart is the official of the monarch” — the character “heart” already existed, but what was formally inscribed into the system were the four characters “official of the monarch.” The issue this article addresses collides directly with this rule: where are the ependymin family and the proctolin neuropeptide family currently working in humans.

    I. The Name “Ependymin” Was Initially a Location-Based Designation

    Ependymin was first isolated from the brains of goldfish, with samples taken near the ependyma; thus, it was named based on location. Subsequent functional evidence also aligned: after learning, its subcellular distribution changes in goldfish (PMID 3572403); suppressing it with antisense oligonucleotides blocks active avoidance of memory consolidation (PMID 7561839); injecting antibodies prevents sharpening of regenerated superior colliculus projections (PMID 3370490). The name was set, and the story was told. However — the syllables “ependym-” refer to where it was found, not what it does. This is the norm in biology, not an accident.

    II. EPDR1: Name Unchanged, Role Moved to Fat

    Unfolding the functional evidence for human EPDR1 yields an awkward result. Structurally, it is a lysosomal protein; crystal structures show homology with the bacterial lipid transporter superfamily, managing the binding and transport of hydrophobic molecules and lipids (PMID 30729188, PMID 31316816). Functionally: a 2019 Cell Metabolism brown/white adipose tissue secretome proteomics study identified it as a new class of “batokine” (PMID 31668873); a 2022 study demonstrated its regulation of β-cell metabolism and function (PMID 36343918), while another from the same year defined it as a non-classical effector of insulin-mediated angiogenesis (PMID 35872017). Further studies show aging fibroblasts secrete it as an aging fluid factor affecting the extracellular matrix (PMID 36497009); levels correlate with disease severity in idiopathic pulmonary fibrosis (PMID 41090747); circulating blood EPDR1 correlates with insulin resistance in type 2 diabetes patients (PMID 40642507); and it is elevated in serum from patients with polycystic ovary syndrome (PMID 41816836).

    In one sentence: Human EPDR1 handles the work of fat, metabolism, angiogenesis, fibroblasts, and a wide range of tumors (PMID 39152265, PMID 32935479, PMID 32111877, PMID 33902536). The brain is not on the list.

    So what is the relationship between “ependymin in the goldfish brain” and “EPDR1 in human fat”? There is a relationship at the sequence level — family evolution studies clarify its divergence path (PMID 30514200); but there is no relationship at the functional level — members of the same family in sponges even became blue carotenoid-binding proteins (PMID 37517696). The name is a historical legacy, not a functional promise.

    III. Proctolin: No Trace Left in Vertebrates

    A more thorough example is proctolin (RYLPT pentapeptide). In arthropods, it is a bona fide neuropeptide: receptors in locusts have been identified, allowing design of analogues (PMID 40778513); cockroach neuropeptide genomics list it among circadian clock-related neuropeptide landscapes (PMID 41144767); it modulates oscillatory circuits alongside overlapping neuropeptides with other targets (PMID 37968117, PMID 37333253). However, searching through vertebrates reveals no direct orthologous endogenous ligand — retrieved literature for “proctolin + human/mammal” consists almost entirely of arthropod experiments introduced into cross-species reviews or structural analogues in scorpion venom (PMID 26056922). On this line, it is not “changed roles,” but the entire position was removed in the vertebrate lineage, leaving only possible distant relatives on the receptor side.

    IV. Tissue Localization Scanning: Four Pitfalls That Must Be Addressed First

    Precisely because of this, tissue localization is not merely “preparation before writing a paper”; it is itself a decisive step — sequence alignment only tells you if the name exists, only localization tells you if the person is still on the job. Yet it has four pitfalls.

    Pitfall One: Brain Sample Contamination. A 2022 report in Molecular Psychiatry indicates that choroid plexus contamination is extremely common in brain tissue sampling and omics analysis (PMID 34983929). The choroid plexus is precisely the barrier tissue most likely to harbor “moon-clock target proteins.” Thus, there is bidirectional contamination: what you find as “brain expression” may be choroid plexus leakage, and what you find as “choroid plexus expression” may be carried in during sampling. Without addressing this pitfall, all subsequent localization conclusions stand unsupported.

    Pitfall Two: The Atlas Lacks the Areas You Want to Search. GTEx divides the brain into dozens of subregions (PMID 23715323, PMID 32913098), but barrier tissues like the choroid plexus, meninges, and blood-brain barrier are generally outside the sampling scope — the tissues you want to search are simply not sampled in existing atlases.

    Pitfall Three: Secreted Proteins Measure Totals. EPDR1 is secreted. The amount measured in circulating blood is a weighted average from all source tissues, not the brain. Trying to infer intracranial rhythms by “monthly blood draws measuring EPDR1” yields no result.

    Pitfall Four: Prior Names Bias Interpretation. When researchers see EPDR1 in human brain single-cell atlases, their first reaction is often led by the name “ependymin family.” The same data yields different interpretations if that gene had a different name. This pitfall cannot be removed by statistics alone.

    Stacked together, the conclusion is: Before dismantling them, the question of “annelid moon-clock target protein human residue” does not even hold as a valid formulation.

    V. Returning to the Twelve Officials

    The true power of Su Wen · Ling Lan Mi Dian Lun‘s “The heart is the official of the monarch” lies in binding the character “heart” with the position “official of the monarch,” adding at the end: “Of these twelve officials, none may be lost” — no official on the twelve-official roster is assigned a position based on its original anatomical discovery location.

    Our cross-species target protein scanning most easily falls into the trap of substituting name for place: sequence matches, so we say “it does the same thing in humans.” The rule of Yijing Studies is to first ask “what it governs,” then ask “where,” and only after answering both questions do we acknowledge the entity. The Chinese idiom “name does not match reality” is not merely rhetoric in this matter, but an operational method.

    Evidence and Limitations

    Verified: ① Human EPDR1 structure and lysosomal lipid/hydrophobic molecule binding function (PMID 30729188, PMID 31316816); ② EPDR1 as a fat factor participating in metabolism, angiogenesis, and β-cell function (PMID 31668873, PMID 36343918, PMID 35872017); ③ Prevalence of choroid plexus contamination in brain tissue omics analysis (PMID 34983929); ④ High functional divergence of the ependymin family across species (PMID 30514200, PMID 37517696).

    Partially Verified: ① Proctolin systems in arthropods have clearer receptor and neuropeptide landscapes (PMID 40778513, PMID 41144767), but on the vertebrate side only speculative distant relatives of receptors remain, with no direct identification; ② Circulating levels of EPDR1 in metabolic diseases are repeatedly reported (PMID 40642507, PMID 41816836), suggesting a measurable fluid readout channel — but these reports are all cross-sectional case-control studies lacking temporal axes.

    To Be Verified: ① Cell-type-level localization of EPDR1 in human choroid plexus/meninges/blood-brain barrier (requires first resolving contamination controls); ② Whether these barrier tissues possess a 29.5-day expression component independent of circadian rhythms; ③ Whether endogenous ligands exist for proctolin receptors in vertebrates (ligand orphanization testing); ④ Whether cross-species tissue localization differences can reverse-engineer potential landing points for “timing residues.”

    Related Reading

    Evidence Level: B
    References: PMID 30729188, PMID 31316816, PMID 31668873, PMID 36343918, PMID 35872017, PMID 36497009, PMID 41090747, PMID 40642507, PMID 41816836, PMID 30514200, PMID 37517696, PMID 40778513, PMID 41144767, PMID 37968117, PMID 37333253, PMID 26056922, PMID 34983929, PMID 23715323, PMID 32913098, PMID 3572403, PMID 7561839, PMID 3370490, PMID 39152265, PMID 32935479, PMID 32111877, PMID 33902536


    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.

  • 关掉卵巢之后,子宫内膜还走不走自己的钟

    Evidence level: B

    内膜有没有自己的节拍,取决于一个可区分的检验:把激素这个上游驱动关掉之后,它是停摆,还是开始自由运行。

    正文

    上一篇《在人血里找 29.5 天的钟》讨论的是”读数口”——外周血。但外周血只是一面镜子,真正该有月级节律的靶组织是子宫内膜。这一篇问一个更硬的问题:内膜那个 28 天左右的节律,是激素写上去的,还是它自己本来就会走?

    一个可拆的实验

    海洋环节虫给了模板:它的昼夜钟和近月钟能拆开,打掉昼夜钟,近月钟照样走(PMID 24075994)。所以”独立近月钟”在无脊椎动物身上已是既成事实。哺乳动物这边的问题是——内膜节律里,多少属于内膜自己的钟,多少只是卵巢激素斜坡的被动读数。

    用 GnRH 激动剂做垂体降调节,把卵巢压下去,造一个”没有周期性激素驱动”的窗口,看内膜还报不报时。两个假说给出不同的预测:

    • 激素驱动假说:内膜节律是激素的被动读出 → 激素撤除,节律消失,输出是一条平线。
    • 内源钟假说:激素只是对时信号 → 钟不会停,它脱离驱动源自由运行(free-running),周期可能漂到新值。

    关键是分清两种结果:节律消失(激素驱动)还是节律换周期(内源钟自由运行)。只报”降调节后看不到 28 天成分”毫无意义,两种假说都会得到这个结果;必须报的是有没有自由运行周期、以及它是多少。这与环节虫实验同构:真正的解耦标志不是”一个没了”,而是”另一个换了节奏还在走”(PMID 38764210 展示的正是这种可拆结构)。

    但”关掉卵巢”这只手不干净

    GnRH 降调节不是单纯的”断供”。子宫内膜、肌层、肌瘤上都表达 GnRH 受体(PMID 21528253、PMID 26715597)。更关键的是,GnRH 类似物对内膜细胞有直接作用:体外直接诱导内膜细胞凋亡、抑制增殖(PMID 12923126、PMID 14505742),腺癌细胞系上同样如此(PMID 24591918);拮抗剂削弱内膜基质细胞的生长能力(PMID 35120552);降调节还改变内膜的类固醇受体表达谱(PMID 19640894)。

    于是”降调节之后内膜的钟变了”有了两个互斥解释:没了激素,或 GnRH 直接改了内膜细胞状态。 要分开得靠额外设计——拮抗剂与激动剂对照,把”撤激素”与”GnRH 暴露”拆成两臂。严格说,GnRH 降调节给出的不是实验,是混淆的合成剂量。

    取样这一关也不好过

    内膜不能天天抽。要覆盖 28 天周期,唯一办法是反复活检——而活检本身就是干预:损伤、炎症、修复都会扰动组织状态,等于一边测量一边改样本。
    现有两套大规模内膜图谱(PMID 32929266、PMID 34857954)都是横断面的:不同的人、不同周期阶段拼出一条时间轴。这条轴上的周期性,多少是个人的钟在走,多少只是不同人的状态斜坡,分不开。

    替代方案是内膜类器官——可长期培养并响应激素(PMID 28394884)、可自更新(PMID 31666317)、可做上皮-基质共培养(PMID 37572651)、有蜕膜化衰老模型(PMID 34487490)。好处是能在无激素培养基里跑纵向序列;代价是脱离体内(无血供、无免疫)。两条路各缺一半:体内能纵向但不能无激素,体外能无激素但不算真在体内。

    动物侧的线索是不对称的

    “激素 → 钟”这个方向证据不少:雌激素直接调制子宫 PER2 节律(PMID 18728223);基质细胞里是孕酮(而非雌激素)同步化钟振荡(PMID 19096762);大鼠蜕膜化中 Per2 下调(PMID 21182399),孕鼠内膜基质细胞的钟基因谱已被刻画(PMID 23847593);REV-ERBα 受激素调控并抑制下游基因(PMID 25007867、PMID 26811051)。

    反过来,钟对激素的独立性几乎没有直接证据。子宫 Bmal1 敲除会造成胎盘血管化受损与宫内胎儿死亡(PMID 35886985),说明钟对生殖功能必需;但这是受孕之后的表型,不能外推到”未孕内膜是否保留独立节律”。雌性哺乳动物 SCN 与核受体的双向调控框架已被提出(PMID 40225034)——框架有了,实验没做。

    临床现实把这个窗口填掉了

    还有一刀更现实:临床上 GnRH 降调节之后,通常立即叠加上外源雌孕激素准备内膜(PMID 30501453、PMID 36604214、PMID 37902321)。也就是说,你想研究的那个”无激素内膜”状态,恰好是临床路径里被马上补上的那个状态。 真正接近它的窗口很窄——绝经后、卵巢功能衰竭,或降调节后的短暂空档。这也解释了这块空白为何一直没被填:临床上没理由制造那个状态。

    旁证都在别的方向:控促排卵会打乱内膜钟节律(PMID 39020493),持续人工光照会扰动中枢与外周生殖钟并改变子宫生理(PMID 35399124)——内膜钟确实可被外部干预改写,只是没人在”撤激素”方向单独测过。

    落到易学上

    《素问·上古天真论》讲女子:”二七而天癸至,任脉通,太冲脉盛,月事以时下,故有子。”这句话里其实有两个东西:”天癸至”与”太冲脉盛”——前者是上游的系统性驱动,后者是局部的本钱。古人并列写成两句,不是同义反复。现代这套实验要做的,正是把它们拆成两个自变量:把”天癸”抽掉,看”太冲脉”还能不能自己盛。

    再往上一层:《易·复》说”反复其道,七日来复,天行也”,把”复”归给”天行”,也就是自发;《道德经》说得更直白——”独立而不改,周行而不殆”。这八个字正是内源振荡器的定义:不靠外物推动,也能周而复始。 区别只在于:古人说这话是表彰,我们说这话是判据——判据要拿数据兑现,兑不出,就退回成漂亮话。

    证据与局限

    已验证: 环节虫昼夜钟与近月钟可被独立解耦(PMID 24075994),近月计时可由昼夜周期的计数决定(PMID 38764210);激素是内膜钟的有效驱动/同步信号——雌激素直接调制子宫 PER2 节律(PMID 18728223)、孕酮而非雌激素同步化内膜基质细胞钟振荡(PMID 19096762)、蜕膜化伴随 Per2 下调(PMID 21182399)与钟基因谱改变(PMID 23847593)、REV-ERBα 受激素调控并抑制下游基因(PMID 25007867、PMID 26811051);子宫钟对生殖功能必需(Bmal1 敲除致胎盘血管化受损与胎儿死亡,PMID 35886985);人子宫内膜表达 GnRH 受体(PMID 21528253、PMID 26715597),GnRH 类似物对内膜细胞有直接促凋亡/抗增殖作用(PMID 12923126、PMID 14505742、PMID 24591918、PMID 35120552)并改变类固醇受体谱(PMID 19640894);人内膜单细胞与时空图谱已完成但为横断面设计(PMID 32929266、PMID 34857954);人内膜类器官可长期培养、响应激素并支持共培养(PMID 28394884、PMID 31666317、PMID 37572651、PMID 34487490);GnRH 降调节后临床常规叠加外源激素准备内膜(PMID 30501453、PMID 36604214、PMID 37902321);内膜钟可被外部干预改写(PMID 39020493、PMID 35399124)。

    部分验证: 雌性哺乳动物中枢钟与核受体对节律的双向调控框架已被提出(PMID 40225034),但尚无针对”激素撤除后未孕内膜钟是否保留”的纵向实验;体外与动物模型提示激素是驱动/同步源,但独立振荡的充分性未被检验。

    待验证: ①GnRH 降调节造成的”无周期性激素”窗口内,人子宫内膜是否保留独立于激素驱动的周期成分(尚无研究执行);②若保留,其自由运行周期是否为 28 天、29.5 天或其他值(这是区分两假说的关键读数);③GnRH 类似物对内膜细胞的直接作用与激素撤除效应能否在同一实验中定量分离;④反复活检本身对内膜钟的扰动幅度是否小于待测效应量(决定体内纵向设计是否可行);⑤类器官在无激素培养基中的自由运行周期与体内是否一致(决定体外替代能否外推)。

    相关阅读

    证据等级: B
    参考文献: PMID 24075994, PMID 38764210, PMID 21528253, PMID 26715597, PMID 12923126, PMID 14505742, PMID 24591918, PMID 35120552, PMID 19640894, PMID 32929266, PMID 34857954, PMID 28394884, PMID 31666317, PMID 37572651, PMID 34487490, PMID 18728223, PMID 19096762, PMID 21182399, PMID 23847593, PMID 25007867, PMID 26811051, PMID 35886985, PMID 40225034, PMID 30501453, PMID 36604214, PMID 37902321, PMID 39020493, PMID 35399124


    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。

    Does the Endometrium Still Follow Its Own Clock After Ovarian Removal?

    Evidence level: B

    Whether the endometrium possesses its own rhythm depends on a distinguishable test: does it stall when the upstream hormonal drive is cut off, or does it begin to run freely?

    Main Text

    The previous article, “Finding a 29.5-Day Clock in Peripheral Blood,” discussed the “reading window”—peripheral blood. However, peripheral blood is merely a mirror; the true target tissue exhibiting monthly rhythms is the endometrium. This article addresses a harder question: Is the ~28-day rhythm of the endometrium written by hormones, or does it run independently?

    A Dissectable Experiment

    Marine annelids provided a template: their circadian and near-monthly clocks can be uncoupled; knocking out the circadian clock leaves the near-monthly clock running (PMID 24075994). Thus, an “independent near-monthly clock” is an established fact in invertebrates. The question for mammals is whether the endometrial rhythm consists of a portion intrinsic to the tissue and a portion merely a passive readout of ovarian hormone gradients.

    Using GnRH agonists to induce pituitary downregulation suppresses the ovaries, creating a window with “no periodic hormonal drive” to observe if the endometrium still signals time. Two hypotheses offer different predictions:

    • Hormone Drive Hypothesis: The endometrial rhythm is a passive readout of hormones → Hormone withdrawal causes rhythm cessation, resulting in a flat output line.
    • Endogenous Clock Hypothesis: Hormones serve only as zeitgebers (time-setting signals) → The clock does not stop; it free-runs independent of the drive source, potentially drifting to a new period.

    The critical distinction lies between rhythm disappearance (hormone drive) and rhythm period shift (endogenous clock free-running). Reporting merely “no 28-day component observed after downregulation” is meaningless, as both hypotheses yield this result; one must report whether a free-running period exists and what its value is. This mirrors the annelid experiment: the true marker of uncoupling is not “one lost,” but “the other continues with a changed rhythm” (PMID 38764210 demonstrates this decoupled structure).

    But the “Shutting Off Ovaries” Hand Is Not Clean

    GnRH downregulation is not a simple “supply cut.” The endometrium, myometrium, and fibroids all express GnRH receptors (PMID 21528253, PMID 26715597). Crucially, GnRH analogs exert a direct effect on endometrial cells: they induce apoptosis and inhibit proliferation in vitro (PMID 12923126, PMID 14505742), similarly in adenocarcinoma cell lines (PMID 24591918); antagonists weaken the growth capacity of endometrial stromal cells (PMID 35120552); downregulation alters the endometrial steroid receptor expression profile (PMID 19640894).

    Thus, “the endometrial clock changes after downregulation” has two mutually exclusive explanations: hormone removal or GnRH directly altering endometrial cell states. Separation requires additional design—comparing antagonists with agonists to split “hormone withdrawal” from “GnRH exposure.” Strictly speaking, GnRH downregulation provides not an experiment but a confounded synthetic dose.

    Sampling Is Also Difficult

    Endometrium cannot be sampled daily. To cover a 28-day cycle, repeated biopsy is the only option—but biopsy itself is an intervention: damage, inflammation, and repair all perturb tissue states, effectively measuring while modifying the sample.

    Existing large-scale endometrial atlases (PMID 32929266, PMID 34857954) are cross-sectional: different individuals and cycle stages are stitched into a timeline. The periodicity on this axis reflects partly an individual’s clock running and partly just different states across individuals, which cannot be distinguished.

    An alternative is endometrial organoids—cultivable long-term and responsive to hormones (PMID 28394884), capable of self-renewal (PMID 31666317), supporting epithelial-stromal co-culture (PMID 37572651), and possessing decidualization aging models (PMID 34487490). The advantage is running longitudinal sequences in hormone-free media; the cost is loss of in vivo context (no blood supply, no immune system). Each path lacks half: in vivo allows longitudinal but not hormone-free; in vitro allows hormone-free but is not truly in vivo.

    Clues from the Animal Side Are Asymmetric

    Evidence for “Hormone → Clock” is abundant: estrogen directly modulates uterine PER2 rhythms (PMID 18728223); progesterone (not estrogen) synchronizes clock oscillations in stromal cells (PMID 19096762); Per2 is downregulated during rat decidualization (PMID 21182399), and the clock gene profile of pregnant mouse endometrial stromal cells has been characterized (PMID 23847593); REV-ERBα is hormone-regulated and suppresses downstream genes (PMID 25007867, PMID 26811051).

    Conversely, there is almost no direct evidence for clock independence from hormones. Knockout of uterine Bmal1 causes impaired placental vascularization and intrauterine fetal death (PMID 35886985), indicating the clock is essential for reproductive function; however, this is a post-conception phenotype and cannot be extrapolated to “whether non-pregnant endometrium retains independent rhythms.” A bidirectional regulatory framework between the SCN and nuclear receptors in female mammals has been proposed (PMID 40225034)—the framework exists, but the experiment has not been performed.

    Clinical Reality Fills This Window

    There is a more practical knife: clinically, after GnRH downregulation, exogenous estrogen and progesterone are typically immediately overlaid to prepare the endometrium (PMID 30501453, PMID 36604214, PMID 37902321). In other words, the very “hormone-free endometrium” state you wish to study is immediately replenished in the clinical pathway. The window truly approaching this state is narrow—postmenopause, ovarian failure, or the brief gap after downregulation. This explains why this blank has never been filled: there is no clinical reason to create that state.

    Corroborating evidence lies elsewhere: controlling ovulation disrupts endometrial clock rhythms (PMID 39020493), and continuous artificial light perturbs central and peripheral reproductive clocks, altering uterine physiology (PMID 35399124)—the endometrial clock can indeed be rewritten by external intervention, but no one has measured it specifically in the “withdrawal of hormone” direction alone.

    Falling Back to Yijing Studies

    The Su Wen · Shang Gu Zhen Lun states regarding women: “At seventeen years and two months, Tian Gui arrives; the Ren Meridian becomes open, the Tai Chong Meridian becomes abundant; menstruation occurs monthly, hence conception is possible.” Within this statement lie two distinct elements: “Tian Gui arrives” and “Tai Chong Meridian becomes abundant”—the former is the upstream systemic drive, the latter the local foundation. The ancients wrote them as two sentences, not as tautology. Modern experiments aim precisely to split these into two independent variables: remove “Tian Gui” and observe if “Tai Chong Meridian” can still become abundant on its own.

    Going further up: The I Ching · Fu states “Repeating the way, returning every seven days; this is the movement of Heaven,” attributing “Fu” (return) to “Tian Xing” (Heavenly movement), i.e., spontaneous; the Dao De Jing states it even more plainly—”Independent without changing, circulating without ceasing.” These eight characters define the endogenous oscillator: capable of cyclical motion without relying on external push. The difference lies only in tone: the ancients spoke this as praise; we speak it as a criterion—a criterion must be validated by data; if not validated, it retreats to pretty talk.

    Evidence and Limitations

    Verified: The annelid circadian and near-monthly clocks can be independently uncoupled (PMID 24075994); near-monthly timing can be determined by counting circadian cycles (PMID 38764210); hormones are effective drivers/synchronizers for the endometrial clock—estrogen directly modulates uterine PER2 rhythms (PMID 18728223), progesterone synchronizes endometrial stromal cell clock oscillations rather than estrogen (PMID 19096762), decidualization accompanies Per2 downregulation (PMID 21182399) and clock gene profile changes (PMID 23847593), REV-ERBα is hormone-regulated and suppresses downstream genes (PMID 25007867, PMID 26811051); uterine clock is essential for reproductive function (Bmal1 knockout causes impaired placental vascularization and fetal death, PMID 35886985); human endometrium expresses GnRH receptors (PMID 21528253, PMID 26715597), GnRH analogs exert direct pro-apoptotic/anti-proliferative effects on endometrial cells (PMID 12923126, PMID 14505742, PMID 24591918, PMID 35120552) and alter steroid receptor profiles (PMID 19640894); human endometrial single-cell and spatiotemporal atlases are complete but designed cross-sectionally (PMID 32929266, PMID 34857954); human endometrial organoids can be cultured long-term, respond to hormones, and support co-culture (PMID 28394884, PMID 31666317, PMID 37572651, PMID 34487490); GnRH downregulation clinically routinely involves overlaying exogenous hormones to prepare the endometrium (PMID 30501453, PMID 36604214, PMID 37902321); endometrial clocks can be rewritten by external intervention (PMID 39020493, PMID 35399124).

    Partially Verified: A bidirectional regulatory framework between central clocks and nuclear receptors for rhythms in female mammals has been proposed (PMID 40225034), but no longitudinal experiments exist regarding “whether the endometrial clock in non-pregnant tissue is retained after hormone withdrawal”; in vitro and animal models suggest hormones are drivers/synchronizers, but the sufficiency of independent oscillation remains untested.

    To Be Verified: ① Whether human endometrium retains a cycle component independent of hormonal drive within the “no periodic hormone” window created by GnRH downregulation (no study has executed this); ② If retained, whether the free-running period is 28 days, 29.5 days, or another value (this is the key reading distinguishing the two hypotheses); ③ Whether direct effects of GnRH analogs on endometrial cells and hormone withdrawal effects can be quantitatively separated in the same experiment; ④ Whether the perturbation magnitude caused by repeated biopsy itself is smaller than the effect size to be measured (determining feasibility of in vivo longitudinal design); ⑤ Whether free-running periods in organoids under hormone-free media match those in vivo (determining if in vitro alternatives can be extrapolated).

    Related Reading

    Evidence Level: B
    References: PMID 24075994, PMID 38764210, PMID 21528253, PMID 26715597, PMID 12923126, PMID 14505742, PMID 24591918, PMID 35120552, PMID 19640894, PMID 32929266, PMID 34857954, PMID 28394884, PMID 31666317, PMID 37572651, PMID 34487490, PMID 18728223, PMID 19096762, PMID 21182399, PMID 23847593, PMID 25007867, PMID 26811051, PMID 35886985, PMID 40225034, PMID 30501453, PMID 36604214, PMID 37902321, PMID 39020493, PMID 35399124


    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.