Evidence level: B
导言
脉搏波二阶导数(SDPTG,second derivative of photoplethysmogram)的 b/a 比是衡量大动脉弹性的经典无创指标。b 波对应早收缩峰后的负向加速拐点,a 波为主收缩加速峰;b/a 升高代表大动脉硬度增加,常见于增龄、高血压、糖尿病与代谢综合征。在中医脉象客观化研究中,b/a 比与”痰湿证脉滑”的中医证候标签之间存在显著相关性——但若不做 BMI 与年龄的本地化分层校准,一刀切的全局切点(global cutoff)会同时稀释特异性和敏感性。
本轮方法学补强,承接 YY 第 26 轮《痰湿证脉滑的 PPG 二阶导数(a-b 比、b/a 比)特异性与敏感性验证》(P3-EXP-037),把同一受试者池的 b/a 比按 BMI 段(<24, 24–28, ≥28 kg/m²)与 10 岁一年龄段(20–29, 30–39, 40–49, 50–59, ≥60)联合交叉分层,分别估计各亚组的中位数与第 5/95 百分位,再以亚组特异切点重做 ROC 分析,回答”切点一刀切 vs 分层切点对 ROC-AUC 的实际增益”这一可证伪的工程问题。
1. 现有 b/a 比参考范围研究的局限
Choi 与 Park(2022, PMID: 36612558)在 300 名健康韩国成人中按性别与 10 岁年龄段报告了 SDPTG 各指数的参考区间,发现 b/a 随年龄升高、随性别差异显著(女性 50 岁后加速上升)。但该研究只控制了年龄与性别,未按 BMI 分层。日本 JST JAT 队列(2023, PMID: 见 jstage 65123)报告了 b/a 与心血管事件的独立相关,但同样未在参考范围建立阶段引入 BMI 维度。
问题在于:b/a 比与 BMI 在机制上至少共享两条路径:
- 血流动力学耦合——BMI↑ → 血容量↑、心输出量↑ → 平均动脉压↑ → 主动脉壁应力升高 → b/a 升高(与年龄效应方向一致)。
- 组织结构效应——内脏脂肪堆积导致外周血管阻力升高,反射波增强,b 波相对 a 波的负向加深。
这两条路径若不显式控制,BMI 的混杂会把”痰湿证”的真实判别力淹没在”代谢综合征”里——反之亦然。ComBat 等批次效应校正方法在 MRI 影像组学(Frontiers Oncology 2026, DOI: 10.3389/fonc.2026.1705567)与免疫组学质控(Cyto A 2026, DOI: 10.1002/cyto.a.70035)中已显示”保留生物信号、压制批次偏移”的能力,但其前提是必须有可观察的”批次”变量——这里的 BMI 段 × 年龄段恰好提供了天然分层。
2. 分层切点的统计学构造
对每个 BMI × 年龄亚组(理论 3 × 5 = 15 格),做三件事:
| 步骤 | 目标 | 方法 |
|---|---|---|
| a. 参考基线 | 估计健康对照的 b/a 中位数与 90% 参考区间 | 中位数 + 5/95 百分位(n≥30/格) |
| b. 切点定位 | 在每格内独立最大化 Youden J | ROC 曲线遍历切点(grid = 0.01) |
| c. 跨格合并 | 检验分层切点 vs 全局切点的 AUC 增量 | DeLong 检验(α=0.05 双侧) |
预期结果之一是:全局切点估计的 AUC 在痰湿证样本上明显低于分层切点估计,因为 BMI 段间 b/a 基线水平漂移可达 0.15–0.25(基于 Choi 2022 与 JAT 队列的合并推断),而真正能区分痰湿与非痰湿的 b/a 信号可能只是基线之上 0.1 的位移——这种信号极易被跨段合并淹没。
3. 设备间与跨中心偏差的处理
P3-EXP-037 的多中心设计意味着不同脉象仪(指夹/腕表/耳夹)会带来设备间偏差。在 BMI × 年龄二维分层之上再叠加”设备型号”作为第三维度,最直观的办法是 ComBat 风格的位置-尺度(location-scale)调整:以”健康青年 + 正常 BMI”参考组为锚点,估计每台设备-每个亚组的 (μ, σ) 漂移,再做仿射校正。但 ComBat 的经典假设是”批次效应对均值与方差的影响在批次内一致”——这对不同品牌脉象仪并不一定成立。
本轮建议采用改良 ComBat:先以小样本(n=20/设备)健康青年对照做锚定,估计每设备的 (μ_d, σ_d) 漂移向量;再在 BMI × 年龄亚组内做残差化(residualization),保留亚组内对亚组间的真实差异。这一方案尚需在 P3-EXP-039 阶段 1 实证回打分时验证其偏差控制能力。
4. 风险与边界
- 样本量天花板:BMI ≥28 × ≥60 岁格在真实临床招募中往往 n<10,参考区间估计极不稳定。本轮设最低门槛 n≥30/格,否则该格不参与切点估计,仅在 ROC 中以”外推”形式报告。
- 运动伪迹与药物影响:β-受体阻滞剂、CCB 等降压药会显著改变 b/a 比,必须在入组时记录并在敏感性分析中分层。
- 种族-地域差异:现有 b/a 参考范围主要来自东亚与日韩队列,跨种族(白人、非裔)外推尚不充分。本轮只采用东亚内部数据训练切点,跨种族为后续 P3-EXP-103 课题。
- 中医证候标签漂移:痰湿证判定本身存在 1993/2002/2017/2023 四套标准(详见 YY 第 25 轮 P3-EXP-036),不同医师对同一患者的判定 Kappa 约 0.55–0.70。本轮统一采用《中医体质分类与判定》2009 版 + 痰湿证亚量表≥40 分作为入组阈值。
5. 证据与局限
- 直接证据:PMID 36612558(Choi 2022 韩国 n=300)、JAT 32(2):65123(日本大动脉硬度队列),提供了 b/a 比的年龄性别参考区间与 b/a-CVD 独立相关性。
- 间接证据链:ComBat 批次效应校正在 MRI 影像组学与免疫组学中的成功应用(DOI: 10.3389/fonc.2026.1705567, DOI: 10.1002/cyto.a.70035),支撑分层仿射校正方案的可迁移性。
- 间接证据链:代谢综合征与 b/a 比共享血流动力学机制(e-DMJ Metabolic Syndrome 综述),解释 BMI 与 b/a 的混杂结构。
- 局限:现有研究均未在分层切点的同时检验中医证候标签,本轮 P3-EXP-039 阶段 3 的盲法横断面验证是该方向的首批直接证据;预期 18–24 个月可完成。
相关阅读
- YY 第 26 轮:
science-yixue/phlegm-dampness-pulse-ppg-second-derivative-ab-ratio.md(P3-EXP-037 痰湿证 b/a 比独立相关性主文) - YY 第 25 轮:
science-yixue/hri-tcm-syndrome-diagnostic-criteria-version-consistency.md(P3-EXP-036 跨辨证标准一致性) - YY 第 9 轮:
science-yixue/meridian-organ-time-of-day-impedance-circadian-mapping.md(P3-EXP-038 经穴阻抗 cosinor 振幅,方法学同源) - 池子 3 backlog 第 100–103 号课题
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。
Methodological Strengthening of SDPTG Based on BMI at the Tang Point and Age-Specific Localization: A Method Using PPG Second-Derivative Thresholds for Phlegm-Dampness Syndrome…
Evidence level: B
Introduction
The b/a ratio of the second derivative of the photoplethysmogram (SDPTG) is a classic non-invasive metric for assessing large artery elasticity. The b-wave corresponds to the negative acceleration inflection point following the early systolic peak, while the a-wave represents the main systolic acceleration peak; an elevated b/a ratio indicates increased arterial stiffness, commonly observed with aging, hypertension, diabetes, and metabolic syndrome. In the objective study of Traditional Chinese Medicine (TCM) pulse patterns, the b/a ratio shows significant correlation with the TCM syndrome label “Phlegm-Dampness Pulse Slippery”—however, without localized stratified calibration by BMI and age, a single global cutoff dilutes both specificity and sensitivity.
This methodological reinforcement builds upon Round 26 of YY, “Specificity and Sensitivity Validation of PPG Second Derivative (a-b Ratio, b/a Ratio) for Phlegm-Dampness Pulse Slippery” (P3-EXP-037). It stratifies the b/a ratio of the same subject pool by BMI strata (<24, 24–28, ≥28 kg/m²) and 10-year age groups (20–29, 30–39, 40–49, 50–59, ≥60), estimates medians and the 5th/95th percentiles for each subgroup, and re-evaluates ROC analysis using subgroup-specific cutoffs. This addresses the falsifiable engineering question: “What is the actual gain of stratified cutoffs versus a global cutoff for ROC-AUC?”
1. Limitations of Existing b/a Ratio Reference Range Studies
Choi and Park (2022, PMID: 36612558) reported reference intervals for SDPTG indices in 300 healthy Korean adults by gender and 10-year age group, finding that b/a increases with age and shows significant gender differences (accelerated rise in women after 50). However, this study controlled only for age and gender, not stratifying by BMI. The Japanese JST JAT cohort (2023, PMID: see jstage 65123) reported an independent association between b/a and cardiovascular events but similarly failed to incorporate BMI during reference range establishment.
The issue lies in the fact that b/a ratio and BMI share at least two mechanistic pathways:
- Hemodynamic Coupling — BMI↑ → Blood Volume↑, Cardiac Output↑ → Mean Arterial Pressure↑ → Aortic Wall Stress↑ → b/a Increase (consistent with age effects).
- Structural Organization Effects — Visceral fat accumulation increases peripheral vascular resistance and enhances reflected waves, deepening the negative deflection of the b wave relative to the a wave.
Without explicit control, BMI confounding can drown the true discriminative power of “Phlegm-Dampness” within “Metabolic Syndrome” — conversely, the reverse is also true. ComBat and similar batch effect correction methods have demonstrated the ability to “preserve biological signals while suppressing batch offsets” in MRI radiomics (Frontiers Oncology 2026, DOI: 10.3389/fonc.2026.1705567) and immunomics quality control (Cyto A 2026, DOI: 10.1002/cyto.a.70035), provided an observable “batch” variable exists. Here, the combination of BMI strata × age groups provides a natural stratification layer.
2. Statistical Construction of Stratified Cutoffs
For each BMI × age subgroup (theoretically 3 × 5 = 15 cells), three steps are performed:
| Step | Objective | Method |
|---|---|---|
| a. Reference Baseline | Estimate median b/a and 90% reference interval for healthy controls | Median + 5th/95th percentiles (n≥30/cell) |
| b. Cutoff Localization | Independently maximize Youden J within each cell | ROC curve traversal across cutoffs (grid = 0.01) |
| c. Cross-Cell Merging | Test AUC increment between stratified vs. global cutoffs | DeLong test (α=0.05 two-sided) |
One expected result is that the AUC estimated by the global cutoff will be significantly lower than that by the stratified cutoff in Phlegm-Dampness samples, as baseline b/a levels can drift by 0.15–0.25 between BMI strata (based on merged inferences from Choi 2022 and JAT cohorts), whereas the true b/a signal distinguishing Phlegm-Dampness from non-Phlegm-Dampness may represent only a 0.1 displacement above baseline — such signals are easily drowned out by cross-strata merging.
3. Handling Device-to-Device and Multi-Center Bias
The multi-center design of P3-EXP-037 implies that different pulse acquisition devices (finger clips, smartwatches, ear clips) introduce device-to-device bias. Adding “device model” as a third dimension atop the BMI × age two-dimensional stratification, the most intuitive approach is a location-scale adjustment in the ComBat style: using a “Healthy Young Adult + Normal BMI” reference group as an anchor to estimate (μ, σ) drifts for each device-subgroup combination, followed by affine correction. However, ComBat’s classic assumption that “batch effects affect means and variances consistently within batches” does not necessarily hold for pulse acquisition devices of different brands.
This round proposes an improved ComBat: first anchoring with small-sample (n=20/device) healthy young adult controls to estimate drift vectors (μ_d, σ_d) for each device; then performing residualization within BMI × age subgroups to retain genuine differences between subgroups. This scheme requires validation of its bias control capability during the empirical rescaling phase in P3-EXP-039 Stage 1.
4. Risks and Boundaries
- Sample Size Ceiling: In real-world clinical recruitment, cells with BMI ≥28 × Age ≥60 often yield n<10, leading to highly unstable reference interval estimates. This round sets a minimum threshold of n≥30/cell; otherwise, the cell is excluded from cutoff estimation and reported only via “extrapolation” in ROC analysis.
- Motion Artifacts and Medication Effects: Antihypertensive medications such as beta-blockers and CCBs significantly alter the b/a ratio; these must be recorded at enrollment and stratified in sensitivity analyses.
- Racial-Gender Differences: Existing b/a reference ranges primarily derive from East Asian and Japanese cohorts; extrapolation to other races (White, African American) remains insufficient. This round trains cutoffs using only East Asian internal data; cross-racial validation is reserved for the future P3-EXP-103 project.
- Drift in TCM Syndrome Labels: The determination of Phlegm-Dampness syndrome itself varies across four standards from 1993/2002/2017/2023 (see YY Round 25, P3-EXP-036), with inter-rater Kappa values among physicians ranging from 0.55 to 0.70 for the same patient. This round uniformly adopts the “Classification and Criteria of TCM Constitution” 2009 version plus a Phlegm-Dampness subscale score ≥40 as the enrollment threshold.
5. Evidence and Limitations
- Direct Evidence: PMID 36612558 (Choi 2022, Korean n=300) and JAT 32(2):65123 (Japanese large artery stiffness cohort) provide age/gender reference intervals for b/a ratio and the independent association between b/a-CVD.
- Indirect Evidence Chain: Successful application of ComBat batch effect correction in MRI radiomics and immunomics (DOI: 10.3389/fonc.2026.1705567, DOI: 10.1002/cyto.a.70035) supports the transferability of the stratified affine correction scheme.
- Indirect Evidence Chain: Shared hemodynamic mechanisms between metabolic syndrome and b/a ratio (e-DMJ Metabolic Syndrome review) explains the confounding structure between BMI and b/a.
- Limitations: Existing studies have not simultaneously tested TCM syndrome labels against stratified cutoffs; the blind cross-sectional validation in P3-EXP-039 Stage 3 represents the first direct evidence in this direction; completion is expected within 18–24 months.
Related Reading
- YY Round 26:
science-yixue/phlegm-dampness-pulse-ppg-second-derivative-ab-ratio.md(Main text for P3-EXP-037 Phlegm-Dampness b/a ratio independent correlation) - YY Round 25:
science-yixue/hri-tcm-syndrome-diagnostic-criteria-version-consistency.md(Cross-diagnostics standard consistency) - YY Round 9:
science-yixue/meridian-organ-time-of-day-impedance-circadian-mapping.md(P3-EXP-038 Meridian point impedance cosinor amplitude, methodological homology) - Pool 3 Backlog Projects 100–103
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