Scholarlink

Background <p>Hypertension and type 2 diabetes (T2D) frequently coexist, but whether hypertension independently contributes to earlier diabetes onset beyond shared metabolic and genetic susceptibility remains incompletely understood. This study investigates the longitudinal relationship between hypertension and subsequent T2D risk using survival analysis and survival machine learning approaches. </p> Methods <p>A matched cohort analysis was conducted using UK Biobank participants free of diabetes at baseline, and a 6-month lag period was applied to reduce reverse causation. Time-to-event analyses were performed using Cox proportional hazards models. Genetic susceptibility was evaluated using both a genome-wide polygenic risk score (PRS) and a targeted multi-variant score (TMS). Survival machine learning models including XGBoost, Random Survival Forests, and ridge-penalized Cox regression were evalueted under identical train/test splits. Additional robustness analyses incorporated antihypertensive medication classes, genetic principal components, broader hypertension phenotype definitions integrating blood pressure measurements and medication use, stricter baseline diabetes exclusion criteria, extended lag-period analyses, and time-stratified sensitivity analyses.</p> Results <p>The final multivariable matched analytic cohort included 76,439 participants (37,951 individuals with HTN and 38,488 matched controls). During follow-up, 4,315 incident T2D events occurred, including 3,571 among individuals with HTN and 744 among controls. Hypertension remained independently associated with T2D after adjustment for adiposity, lipid traits, statin use, antihypertensive medication classes, genetic principal components, and genetic susceptibility (HR 1.30, p &lt; 0.001). Individuals with hypertension developed T2D earlier, with reduced diabetes-free survival over follow-up. Both PRS and TMS independently predicted T2D risk and demonstrated significant interactions with hypertension. Survival machine learning analyses showed progressive improvement in discrimination following incorporation of hypertension and genetic susceptibility, with XGBoost demonstrating the highest performance (C-index 0.823, 95% CI 0.812–0.833). Importantly, the observed association remained robust across multiple sensitivity analyses, demonstrating persistence of the association beyond the early post-index period.</p> Conclusions <p>Hypertension is associated with an increased and earlier T2D risk independent of metabolic risk factors and genetic susceptibility. Polygenic risk further stratifies diabetes risk but does not explain the relationship between hypertension and earlier T2D onset. The consistency of the findings across sensitivity analyses and survival-learning frameworks supports hypertension as an important component of the cardiometabolic trajectory preceding T2D development.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Patients with hypertension consistently showed earlier onset of diabetes: A UK Biobank survival analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月5日

查看全文
  • Cynthia AL Hageh,
  • Haralampous Hatzikirou,
  • Lithe Basbous,
  • Siobhán O’Sullivan,
  • Nantia Iakovidou,
  • Andreas Henschel,
  • Jorge Zubelli,
  • Hao Zhou,
  • Basem Al Omari,
  • Laurent Alain Najman,
  • Antoine Abchee,
  • Pierre A Zalloua

摘要

Background

Hypertension and type 2 diabetes (T2D) frequently coexist, but whether hypertension independently contributes to earlier diabetes onset beyond shared metabolic and genetic susceptibility remains incompletely understood. This study investigates the longitudinal relationship between hypertension and subsequent T2D risk using survival analysis and survival machine learning approaches.

Methods

A matched cohort analysis was conducted using UK Biobank participants free of diabetes at baseline, and a 6-month lag period was applied to reduce reverse causation. Time-to-event analyses were performed using Cox proportional hazards models. Genetic susceptibility was evaluated using both a genome-wide polygenic risk score (PRS) and a targeted multi-variant score (TMS). Survival machine learning models including XGBoost, Random Survival Forests, and ridge-penalized Cox regression were evalueted under identical train/test splits. Additional robustness analyses incorporated antihypertensive medication classes, genetic principal components, broader hypertension phenotype definitions integrating blood pressure measurements and medication use, stricter baseline diabetes exclusion criteria, extended lag-period analyses, and time-stratified sensitivity analyses.

Results

The final multivariable matched analytic cohort included 76,439 participants (37,951 individuals with HTN and 38,488 matched controls). During follow-up, 4,315 incident T2D events occurred, including 3,571 among individuals with HTN and 744 among controls. Hypertension remained independently associated with T2D after adjustment for adiposity, lipid traits, statin use, antihypertensive medication classes, genetic principal components, and genetic susceptibility (HR 1.30, p < 0.001). Individuals with hypertension developed T2D earlier, with reduced diabetes-free survival over follow-up. Both PRS and TMS independently predicted T2D risk and demonstrated significant interactions with hypertension. Survival machine learning analyses showed progressive improvement in discrimination following incorporation of hypertension and genetic susceptibility, with XGBoost demonstrating the highest performance (C-index 0.823, 95% CI 0.812–0.833). Importantly, the observed association remained robust across multiple sensitivity analyses, demonstrating persistence of the association beyond the early post-index period.

Conclusions

Hypertension is associated with an increased and earlier T2D risk independent of metabolic risk factors and genetic susceptibility. Polygenic risk further stratifies diabetes risk but does not explain the relationship between hypertension and earlier T2D onset. The consistency of the findings across sensitivity analyses and survival-learning frameworks supports hypertension as an important component of the cardiometabolic trajectory preceding T2D development.