Frontiers in medicine · 2026
Intervention-analog association between traditional Chinese medicine syndrome load and lipid/C-reactive protein biomarkers: a single-center retrospective cohort study using target trial emulation.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Guo Yu-Xuan, Han Chao, Xiang Xin, Wang Sheng, Yang Li, Li Shuang-Jian
Аннотация
BACKGROUND: Traditional Chinese Medicine (TCM) syndrome differentiation is central to individualized diagnosis, but its relationship with objective biochemical markers remains insufficiently characterized within explicit frameworks for intervention-analog causal inference.
OBJECTIVES: This study aimed to construct a syndrome load index using confirmatory factor analysis (CFA) and item response theory (IRT), assess measurement invariance, and estimate adjusted intervention-analog associations with lipid and C-reactive protein (CRP) biomarkers.
METHODS: This single-center retrospective cohort study included 1,472 adults with structured TCM assessments and repeated biochemical measurements over 90 days. Syndrome load was modeled as a continuous latent variable using CFA/IRT. A Gaussian generalized propensity score with stabilized inverse probability of treatment weighting, combined with double machine learning cross-fitting, estimated associations with low-density lipoprotein-cholesterol (LDL-C), triglycerides (TG), and CRP z-scores, as well as corresponding binary abnormality outcomes. Quartiles were used only for descriptive diagnostics. Negative controls, baseline laboratory sensitivity analyses, weight diagnostics, and E-value analyses assessed robustness.
RESULTS: CFA/IRT models demonstrated good fit (CFI = 0.96, RMSEA = 0.042) and supported metric/scalar invariance across both sex and age groups. The emulated contrast represented a 1-standard deviation (SD) difference in baseline syndrome load as an analytic intervention analog, rather than a literal randomization of a latent construct. After weighting, all absolute standardized mean differences (SMDs) were <0.10, the effective sample size was 1,289, and no material positivity violations were observed. A 1-SD higher syndrome load was associated with higher TG z-scores (+0.18; 95%CI: 0.12-0.24; approximately +0.20 mmol/L), LDL-C z-scores (+0.11; 95%CI: 0.05-0.18; approximately +0.10 mmol/L), and CRP z-scores (+0.16; 95%CI: 0.09-0.23; approximately +0.31 mg/L), corresponding to 3-5% absolute risk increases in biomarker abnormalities. Baseline laboratory sensitivity analyses yielded attenuated but directionally consistent estimates (TG + 0.16, LDL-C + 0.09, CRP + 0.13 to +0.14 z-score units). Associations were stronger among participants aged ≥65 years, those with diabetes, or those with body mass index (BMI) ≥ 28. Negative controls were null, and E-values ranged from 1.54 to 1.92.
CONCLUSION: Higher syndrome load was consistently associated with dyslipidemia and CRP elevation within an explicit intervention-analog framework. These findings support syndrome objectification and personalized cardiometabolic risk stratification; however, prospective intervention studies are required before concluding that the latent syndrome construct itself is directly modifiable.
OBJECTIVES: This study aimed to construct a syndrome load index using confirmatory factor analysis (CFA) and item response theory (IRT), assess measurement invariance, and estimate adjusted intervention-analog associations with lipid and C-reactive protein (CRP) biomarkers.
METHODS: This single-center retrospective cohort study included 1,472 adults with structured TCM assessments and repeated biochemical measurements over 90 days. Syndrome load was modeled as a continuous latent variable using CFA/IRT. A Gaussian generalized propensity score with stabilized inverse probability of treatment weighting, combined with double machine learning cross-fitting, estimated associations with low-density lipoprotein-cholesterol (LDL-C), triglycerides (TG), and CRP z-scores, as well as corresponding binary abnormality outcomes. Quartiles were used only for descriptive diagnostics. Negative controls, baseline laboratory sensitivity analyses, weight diagnostics, and E-value analyses assessed robustness.
RESULTS: CFA/IRT models demonstrated good fit (CFI = 0.96, RMSEA = 0.042) and supported metric/scalar invariance across both sex and age groups. The emulated contrast represented a 1-standard deviation (SD) difference in baseline syndrome load as an analytic intervention analog, rather than a literal randomization of a latent construct. After weighting, all absolute standardized mean differences (SMDs) were <0.10, the effective sample size was 1,289, and no material positivity violations were observed. A 1-SD higher syndrome load was associated with higher TG z-scores (+0.18; 95%CI: 0.12-0.24; approximately +0.20 mmol/L), LDL-C z-scores (+0.11; 95%CI: 0.05-0.18; approximately +0.10 mmol/L), and CRP z-scores (+0.16; 95%CI: 0.09-0.23; approximately +0.31 mg/L), corresponding to 3-5% absolute risk increases in biomarker abnormalities. Baseline laboratory sensitivity analyses yielded attenuated but directionally consistent estimates (TG + 0.16, LDL-C + 0.09, CRP + 0.13 to +0.14 z-score units). Associations were stronger among participants aged ≥65 years, those with diabetes, or those with body mass index (BMI) ≥ 28. Negative controls were null, and E-values ranged from 1.54 to 1.92.
CONCLUSION: Higher syndrome load was consistently associated with dyslipidemia and CRP elevation within an explicit intervention-analog framework. These findings support syndrome objectification and personalized cardiometabolic risk stratification; however, prospective intervention studies are required before concluding that the latent syndrome construct itself is directly modifiable.