Behavioral–Physiological Risk Discordance and Incident Cardiometabolic Multimorbidity in Middle-Aged and Ol der Chinese Adults: A Longitudinal Analysis Based on CHARLS

Technical Summary: Behavioral–Physiological Risk Discordance and Incident Cardiometabolic Multimorbidity

Problem Statement
Cardiometabolic multimorbidity (CMM), defined as the co-occurrence of multiple cardiometabolic conditions (e.g., diabetes, heart disease, stroke), represents a significant public health challenge among middle-aged and older adults. While adverse behavioral profiles (smoking, alcohol use, physical inactivity, sleep duration) and elevated physiological load (adiposity, blood pressure, glucose/lipid metabolism, inflammation, renal function) are individually established risk factors, existing research has largely examined them in isolation. There is a critical gap in understanding the "behavioral–physiological risk discordance"—the mismatch between an individual's behavioral risk profile and their accumulated physiological burden. Specifically, it remains unclear whether this discordance, or specific phenotypes arising from the interaction of these two domains, offers superior risk stratification for incident CMM compared to single-domain assessments.

Methodology
This study conducted a secondary analysis of longitudinal data from the China Health and Retirement Longitudinal Study (CHARLS).

• Cohort: The final analysis included 7,454 participants aged 45 and older who were free of CMM at baseline. Participants were followed for a median of 5 years (up to 2020).
• Exposure Metrics:
  • BPRDI (Behavioral–Physiological Risk Discordance Index): Calculated as the standardized difference between the Physiological Load Score (PLS) and the Behavioral Risk Score (BRS4). A higher BPRDI indicates a greater discordance where physiological burden exceeds behavioral risk.
  • BRS4–PLS Phenotypes: Participants were classified into four distinct groups based on their baseline BRS4 and PLS levels: Low-Low (low risk/low burden), High-Low (high risk/low burden), Low-High (low risk/high burden), and High-High (high risk/high burden).
• Outcome: Incident CMM was defined as the new diagnosis of two or more cardiometabolic conditions (diabetes, heart disease, or stroke) during the follow-up period, confirmed via self-reported physician diagnosis.
• Statistical Analysis:
  • Cox proportional hazards regression models (unadjusted, demographic-adjusted, and fully adjusted for baseline comorbidities and socioeconomic factors) were used to estimate Hazard Ratios (HR) for BPRDI and phenotypes.
  • Restricted Cubic Spline (RCS) regression assessed nonlinearity in the dose-response relationship.
  • Predictive performance was evaluated using the Area Under the Receiver Operating Characteristic Curve (AUC) for 5-year incident CMM risk, comparing base models against models incorporating BPRDI, BRS4, PLS, and phenotypes.
  • Subgroup analyses were conducted to test for interactions across age, sex, residence, and baseline disease status.

Key Results

• Incidence: Over the follow-up period, 690 participants (9.3%) developed incident CMM.
• BPRDI Association: There was a significant, approximately linear association between BPRDI and incident CMM risk. In the fully adjusted model, each 1-unit increase in BPRDI was associated with a 21% increased risk of incident CMM (HR = 1.21, 95% CI: 1.13–1.31, P < 0.001). The highest BPRDI quartile (Q4) showed a 92% higher risk compared to the lowest quartile (Q1).
• Phenotype Analysis: Compared to the Low-Low reference group:
  • Low-High (Low behavioral risk, High physiological load): HR = 2.47 (95% CI: 1.87–3.25).
  • High-High (High behavioral risk, High physiological load): HR = 2.51 (95% CI: 1.90–3.33).
  • High-Low (High behavioral risk, Low physiological load): No significant association was observed (HR = 1.16, P = 0.470).
  • This suggests that elevated physiological load is a stronger driver of CMM risk than behavioral risk alone, and that individuals with "hidden" high physiological burden despite low behavioral risk are at substantial risk.
• Predictive Value: The model combining BPRDI and BRS4–PLS phenotypes achieved the highest predictive performance (AUC = 0.839), showing a modest but statistically significant improvement over the base clinical model (AUC = 0.827, P = 0.042). Single markers (BRS4, PLS, or BPRDI alone) did not significantly improve the base model's predictive power.
• Subgroups: The associations remained consistent across subgroups stratified by age, gender, residence, and baseline comorbidities, with no significant interactions detected.

Significance and Claims
The study concludes that behavioral–physiological risk discordance is a significant predictor of incident cardiometabolic multimorbidity in middle-aged and older Chinese adults. The authors claim that:

1. Discordance Matters: Relying solely on behavioral risk or physiological load may underestimate risk. The BPRDI captures a specific risk pattern where physiological burden is high even when behavioral risk appears low.
2. Phenotype Utility: The BRS4–PLS phenotypes, particularly the "Low-High" and "High-High" groups, identify individuals at significantly elevated risk. The "High-Low" group (high behavior, low physiology) did not show a significant risk increase, suggesting that physiological accumulation is the more critical determinant for CMM onset in this context.
3. Clinical Application: BPRDI and BRS4–PLS phenotypes serve as low-cost, accessible tools for risk stratification. They may help clinicians identify individuals with elevated physiological burden who might otherwise be overlooked due to seemingly healthy behavioral profiles, thereby informing targeted prevention and early management strategies.

The authors maintain a modest tone regarding the incremental predictive value (noting the AUC improvement was small but significant) and acknowledge limitations, including reliance on self-reported diagnoses, the use of baseline assessments which may not capture dynamic changes, and the observational nature of the study which precludes causal inference. They emphasize the need for external validation in other populations.