Evidence level: B
工业化不是一个开关,而是三个:电气化切掉光周期、供暖空调切掉温度、冷藏与食品工业切掉营养的季节性。三根线被剪断的年代不同,所以「印记衰减」这条曲线,其实是三条被遮挡的曲线叠在一起。
正文
上一篇《生育节律与健康印记的衰减斜率》把两条衰减曲线对齐,试图从断点时间差里读出代际时滞。但那个设计留下一个没被检验的假设:把「工业化」当成一个单一开关。真实的工业化是三个开关,而且合上的年代各不相同。于是问题变成:这三根线各自的贡献,能不能分开算?
为什么难分,难在共线。 在任何传统农业社会,同一个季节里光照变短、气温变冷、食物变少是同时到来的——冬天就是冬天。回归里放进三个变量,模型分不清是谁在起作用,甚至可能三个都”显著”、加起来却超过 100%(方差被重复计算)。这和中医里”风寒湿同时来、分不清主次”是同一类困难。要真分开,不能靠统计技巧硬拆,只能靠历史自然实验:找到某一根线被单独切断的时间点,看印记衰减是否跟着它走。
光周期这根线:有最干净的历史断点。 英格兰与威尔士 1955–2015 年的月度活产记录,与全国光周期、地表温度序列做互相关后发现:1974 年之前光周期领先出生计数约 11 个月(即”受孕前一个月的日照长度”预测出生数),温度只起次要作用;1974–1976 年出现极陡的相位断点,此后光周期相关彻底消失,只剩一个更弱、方向相反的温度关联(PMID 41299814)。该断点与全国免费激素避孕药普及及社会变迁同期。这是一根近乎纯”光”的线被剪断的现场记录。
个体层面,光周期可以做成连续暴露而不是季节标签。英国生物样本库中,用第三孕季平均日光时长与相对光周期幅度(相对均值的波动范围)建模,抑郁的比值比达 2.85(单次发作)、2.20(复发)、1.48(确诊),双相障碍则无一致模式(PMID 38581863)。同期另一项 UK Biobank 研究显示,夏季出生与颞叶、枕叶皮层厚度略高相关,冬季出生与白质完整性略高相关(PMID 38776272)——效应量都很小,这本身就是”印记在现代队列里已弱化到接近噪声”的量化体现。
温度这根线:需要一个”没有供暖的社会”才看得清。 瑞典乌普萨拉 1915–1929 年近 1.4 万例分娩提供了这种社会——室内供暖简陋、缺避寒设施。结果显示孕期环境温度每下降 1°C,死产风险比升高 1.08;寒冷极端影响早产与出生身长,炎热极端同样升高早产(PMID 24332864)。现代人群里效应仍存在但更小:覆盖 70 项研究、27 个国家的系统综述与随机效应 meta 分析给出”温度每升 1°C 早产比值比 1.05、热浪期间 1.16″(PMID 33148618);更早的综述已明确季节与气温对出生结局的共同作用(PMID 21333980)。注意时代差:1915 年的瑞典温度效应比 2019 年的全球 meta 更清晰,因为现代人绝大部分时间在恒温室内——温度这根线被建筑切掉了。
营养这根线:在还有”青黄不接”的地方最清楚。 坦桑尼亚农村队列中,出生在收获前(9–10 月)的儿童 18 月龄发育评估总分低 1.05 个标准差,语言与精细动作各低 1.67(PMID 39032770);同一地区的 MAL-ED 研究显示食物不安全高度季节化,12 月–2 月出生儿童入组体重低 0.35 个 z 分,差距持续到 2 岁(PMID 30608052)。南非 VHEMBE 队列证实孕期饮食构成、产检利用与胎儿大小均有季节性(PMID 31553767);马拉维研究用分位数回归发现收获后出生者在第 90 百分位的出生体重高出 281.4 克(PMID 35897009)。营养效应集中在低分位人群——它不是均匀平移,而是”弱者更弱”。
怎么把这个分解算出来。 三根线各自的证据都有,但”相对贡献”要给数,需要专门工具。环境健康领域成熟的做法是相对重要性分析(dominance analysis / Johnson 相对权重):把每个变量在全部子模型集合中的边际 R² 增量求平均,得到不依赖变量进入顺序的贡献份额,并用自助法给出置信区间(PMID 18569519)。它比”逐步回归看 ΔR²”稳健,也比单看系数可靠——共线下系数会被彼此挤压,相对权重不会。操作设计是:跨断代出生队列 + 围产期三暴露重建(按天文算日照时长、按气象站建气温、按农业统计或食物平衡表建营养可得性)+ 三暴露×出生队列交互回归 + 相对权重分解。把队列按”三根线被切断的程度”分层(1940 前 / 1940–1970 / 1970 后),看每一层的相对权重排序如何变化。
对易学这边的落点。 这一轮拆解最耐人寻味处,是它回到了传统术数本来的分类。八字月令把光、温、营养混在一个格子里,但古人用它们时,词是分开的:调候看的是”寒暖燥湿”——这是温度;择时看日躔与节气——这是光照;节令进补、食忌、物候出产——这是营养。三套参数、三套判断,古人从未把它们当成一个”出生季节”处理。所以”把三根线分开算”不是现代人对传统的修正,而是把被现代人简化掉的传统结构重新装回去。它给出的可检验推论也很清楚:若”天时印记”确由这三条独立通路承载,那么在现代城市里,光与温这两条已被建筑与电气化屏蔽得更彻底、营养那条被食品工业削平得最晚,它们对印记的贡献份额应呈现可预测的先后次序,而不是同步归零。
证据与局限
已验证: 英格兰与威尔士 1955–2015 年出生序列存在 1974–1976 年相位断点,断点前光周期领先出生数约 11 个月、断点后光周期相关消失而弱的反向温度关联保留(PMID 41299814);瑞典乌普萨拉 1915–1929 队列中孕期温度每降 1°C 死产风险比 1.08(PMID 24332864);高温与早产/低出生体重的系统综述与 meta 分析(PMID 33148618、PMID 21333980);坦桑尼亚、南非、马拉维三个低资源队列中”出生时点相对食物季节”与生长/认知/出生体重的关联(PMID 39032770、PMID 30608052、PMID 31553767、PMID 35897009);UK Biobank 中围产期光周期连续指标与抑郁的关联(PMID 38581863、PMID 38776272)。
部分验证: 三根线各自的可分离性有间接支持——光周期有历史断点、温度有”前供暖社会”队列、营养有收获/lean season 对照,说明三条通路可被分别观测;但同一队列内三暴露的贡献份额从未被直接分解,目前只是”各有独立证据”,不等于”相对权重可估”。相对重要性分析方法本身在公共卫生领域已有成熟验证(PMID 18569519),但尚未用于围产期三暴露分解。
待验证: ①三暴露在季节内高度共线,相对权重分解的可识别性下界(共线到什么程度就分不开)未标定;②三暴露的断代时间点是否真能作为独立”自然实验臂”,需要把电气化/供暖/冷藏的地区先后的制度时间线与暴露序列对齐验证;③贡献份额是否随出生队列分层单调变化,需要跨断代队列的实际数据;④本分解假设三通路可加,若存在交互(如低温+营养不足的协同),加性方差分解会低估个体通路贡献。
相关阅读
证据等级: B
参考文献: PMID 41299814, PMID 24332864, PMID 33148618, PMID 21333980, PMID 39032770, PMID 30608052, PMID 31553767, PMID 35897009, PMID 38581863, PMID 38776272, PMID 18569519
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。
Can the three lines of \”Heavenly Time Mark\” attenuation be calculated separately: decomposing the contributions of photoperiod, temperature, and nutrition
Evidence level: B
Industrialization is not a single switch but three: electrification cuts photoperiod, HVAC cuts temperature, and refrigeration plus the food industry cut seasonal nutrition. These three lines were severed at different times; therefore, the curve of “marking attenuation” is actually a superposition of three obscured curves.
Main Text
The previous article, “Birth Rhythms and Health Marking Attenuation Slopes,” aligned two attenuation curves attempting to read intergenerational lag from the difference in breakpoint timing. However, that design left an untested assumption: treating “industrialization” as a single switch. Real industrialization comprises three switches, each closed at different eras. Thus the question becomes: Can the contributions of these three lines be calculated separately?
Why it is hard to separate lies in collinearity. In any traditional agrarian society, shorter daylight, colder temperatures, and reduced food availability arrive simultaneously in the same season—winter is simply winter. If three variables are entered into a regression, the model cannot distinguish which is acting, and all three might appear “significant” while summing to over 100% (variance double-counted). This mirrors the difficulty in Traditional Chinese Medicine where “wind, cold, and dampness arrive together with unclear primary roles.” True separation cannot rely on statistical tricks alone but requires historical natural experiments: identifying a point in time where one line is severed independently to observe if marking attenuation follows it.
The Photoperiod Line: Has the cleanest historical breakpoints. Monthly live birth records in England and Wales from 1955–2015, cross-correlated with national photoperiod and surface temperature sequences, revealed that prior to 1974, photoperiod led birth counts by approximately 11 months (i.e., “daylight length one month prior to conception” predicts birth counts), while temperature played a secondary role; between 1974–1976, an extremely sharp phase breakpoint appeared, after which photoperiod correlation disappeared entirely, leaving only a weaker, directionally opposite temperature association (PMID 41299814). This breakpoint coincided with the nationwide rollout of free hormonal contraceptives and social shifts. This is a field record of a line nearly pure in “light” being cut.
At the individual level, photoperiod can be modeled as continuous exposure rather than seasonal labels. In the UK Biobank, modeling average daylight duration in the third trimester against relative photoperiod amplitude (range of fluctuation relative to the mean) yielded odds ratios for depression of 2.85 (single episode), 2.20 (recurrence), and 1.48 (diagnosis), with no consistent pattern for bipolar disorder (PMID 38581863). A concurrent UK Biobank study showed that summer births correlated slightly higher with temporal and occipital cortex thickness, while winter births correlated slightly higher with white matter integrity (PMID 38776272)—effect sizes are very small, which itself is a quantitative manifestation of “markings weakened to near noise in modern cohorts.”
The Temperature Line: Requires a “non-heated society” to be visible. The Uppsala, Sweden cohort from 1915–1929 with nearly 14,000 deliveries provided such a society—indoor heating was rudimentary and shelter from cold was lacking. Results showed that for every 1°C decrease in prenatal ambient temperature, stillbirth risk ratio increased by 1.08; cold extremes affected preterm birth and birth length, while heat extremes also increased preterm birth (PMID 24332864). Effects persist but are smaller in modern populations: a systematic review and random-effects meta-analysis covering 70 studies across 27 countries found “preterm birth risk ratio increases by 1.05 per 1°C rise and 1.16 during heatwaves” (PMID 33148618); earlier reviews already clarified the joint role of season and air temperature on birth outcomes (PMID 21333980). Note the temporal gap: The temperature effect in 1915 Sweden was clearer than the global meta-analysis in 2019 because modern humans spend most time in temperature-controlled buildings—the temperature line was cut by architecture.
The Nutrition Line: Clearest where “lean seasons” still exist. In a Tanzanian rural cohort, children born before harvest (September–October) scored 1.05 standard deviations lower on the 18-month developmental assessment total, with language and fine motor skills each 1.67 lower (PMID 39032770); the same region’s MAL-ED study showed food insecurity is highly seasonal, with children born December–February having lower enrollment weight by 0.35 z-scores, a gap persisting to age 2 (PMID 30608052). The South African VHEMBE cohort confirmed seasonal patterns in prenatal diet composition, utilization of antenatal care, and fetal size (PMID 31553767); a Malawian study using quantile regression found post-harvest births had birth weights 281.4 grams higher at the 90th percentile (PMID 35897009). Nutrition effects concentrate in low-percentile populations—it is not a uniform shift but “the weak are weaker.”
How to calculate this decomposition. Each line has its own evidence, but assigning numbers to “relative contributions” requires specialized tools. A mature approach in environmental health is Relative Importance Analysis (dominance analysis / Johnson relative weights): averaging the marginal R² increment of each variable across all sub-model sets to obtain contribution shares independent of variable entry order, with bootstrap confidence intervals (PMID 18569519). It is more robust than “stepwise regression looking at ΔR²” and more reliable than single coefficients alone—coefficients are squeezed under collinearity, but relative weights are not. The operational design involves: cross-generational birth cohorts + reconstruction of three prenatal exposures (calculated photoperiod from astronomy, temperature from weather stations, nutritional availability from agricultural statistics or food balance sheets) + three-exposure × birth cohort interaction regression + relative weight decomposition. Stratify cohorts by “degree of severance of the three lines” (pre-1940 / 1940–1970 / post-1970) and observe how the ranking of relative weights changes across layers.
The Landing Point for Yijing Studies. The most intriguing aspect of this dissection is that it returns to the original classification of traditional divination systems. The Bazi monthly pillar mixes light, temperature, and nutrition into one grid, but ancients used them separately: Tiao Hou (adjustment of climate) observes “cold, warm, dry, damp”—this is temperature; Zhe Shi (timing selection) observes solar position and solar terms—this is light; seasonal tonification, dietary taboos, and phenological harvests—this is nutrition. Three sets of parameters, three sets of judgments; ancients never treated them as a single “birth season.” Therefore, “calculating the three lines separately” is not a modern correction of tradition but reinstalling the traditional structure simplified by modernity. It also yields clear testable inferences: If “celestial markings” are indeed carried by these three independent pathways, then in modern cities, where light and temperature are more thoroughly shielded by buildings and electrification, and nutrition flattened last by the food industry, their contribution shares should show a predictable sequence rather than simultaneous zeroing.
Evidence and Limitations
Verified: England and Wales birth sequences from 1955–2015 exhibit a phase breakpoint between 1974–1976, with photoperiod leading birth counts by ~11 months pre-breakpoint and disappearing post-breakpoint while a weak reverse temperature association persists (PMID 41299814); Uppsala, Sweden cohort from 1915–1929 shows stillbirth risk ratio increases by 1.08 per 1°C prenatal temperature drop (PMID 24332864); systematic reviews and meta-analyses on heat and preterm birth/low birth weight (PMID 33148618, PMID 21333980); associations between “birth timing relative to food seasonality” and growth/cognition/birth weight in low-resource cohorts in Tanzania, South Africa, and Malawi (PMID 39032770, PMID 30608052, PMID 31553767, PMID 35897009); associations between continuous prenatal photoperiod metrics and depression in UK Biobank (PMID 38581863, PMID 38776272).
Partially Verified: Separability of the three lines has indirect support—photoperiod has historical breakpoints, temperature has pre-heating society cohorts, and nutrition has harvest/lean season controls, indicating three pathways can be observed separately; however, contribution shares within the same cohort have never been directly decomposed, currently only “each has independent evidence,” not equivalent to “relative weights estimable.” Relative importance analysis itself has mature verification in public health (PMID 18569519), but has not yet been applied to three-exposure decomposition in the perinatal context.
To Be Verified: ① High collinearity of three exposures within seasons means the identifiability lower bound (to what degree of collinearity they cannot be separated) is not calibrated; ② Whether generational breakpoints truly function as independent “natural experiment arms” requires aligning exposure sequences with the regional sequential institutional timelines of electrification/heating/refrigeration; ③ Whether contribution shares change monotonically across birth cohort stratification requires actual data from cross-generational cohorts; ④ This decomposition assumes additivity of three pathways; if interactions exist (e.g., synergy between low temperature and malnutrition), additive variance decomposition will underestimate individual pathway contributions.
Related Reading
- Birth Seasonality and Health Marking Attenuation Contribution Split
- Quantification of Lag Between Two Breakpoints: Aligning Rhythmic and Marking Attenuation
- Latitudinal Gradient of Birth Season Effects
- Mechanism Switch in Monsoon Regions: Light Markings vs. Infection-Nutrition Markings
- Research Pool 2: Spatiotemporal Frequency Topology
Evidence Level: B
References: PMID 41299814, PMID 24332864, PMID 33148618, PMID 21333980, PMID 39032770, PMID 30608052, PMID 31553767, PMID 35897009, PMID 38581863, PMID 38776272, PMID 18569519
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.
Leave a Reply