Usual Blood Pressure and Risk of New-Onset Diabetes: Evidence From 4.1 Million Adults and a Meta-Analysis of Prospective Studies

Abstract

Background: Reliable quantification of the association between blood pressure (BP) and risk of type 2 diabetes is lacking.

Objectives: This study sought to determine the association between usual BP and risk of diabetes, overall and by participant characteristics.

Methods: A cohort of 4.1 million adults, free of diabetes and cardiovascular disease, was identified using validated linked electronic health records. Analyses were complemented by a meta-analysis of prospective studies that reported relative risks of new-onset diabetes per unit of systolic blood pressure (SBP).

Results: Among the overall cohort, 20 mm Hg higher SBP and 10 mm Hg higher diastolic BP were associated with a 58% and a 52% higher risk of new-onset diabetes (hazard ratio: 1.58; 95% confidence interval [CI]: 1.56 to 1.59; and hazard ratio: 1.52; 95% confidence interval: 1.51 to 1.54), respectively. There was no evidence of a nadir to a baseline BP of 110/70 mm Hg. The strength of the association per 20 mm Hg higher SBP declined with age and with increasing body mass index. Estimates were similar even after excluding individuals prescribed antihypertensive or lipid-lowering therapies. Systematic review identified 30 studies with 285,664 participants and 17,388 incident diabetes events. The pooled relative risk of diabetes for a 20 mm Hg higher usual SBP across these studies was 1.77 (1.53 to 2.05).

Conclusions: People with elevated BP are at increased risk of diabetes. The strength of the association declined with increasing body mass index and age. Further research should determine if the observed risk is modifiable.

Keywords: body mass index; meta-analysis; regression dilution bias.

Central Illustration

Elevated Blood Pressure and Risk of New-Onset Diabetes (Top) Kaplan-Meier curves illustrating time to incident diabetes by (A) usual systolic blood pressure categories and (B) usual diastolic blood pressure categories. No adjustments were applied. (Bottom) Hazard ratios per 20 mm Hg higher systolic blood pressure and 10 mm Hg higher diastolic blood pressure for new-onset diabetes, with progressive adjustment for age, sex, BMI, smoking and baseline antihypertensive use and baseline lipid-lowering drug use. BMI = body mass index; CI = confidence interval; DBP = diastolic blood pressure; HR = hazard ratio; SBP = systolic blood pressure.

Figure 1

Adjusted HR for Diabetes by Systolic Blood Pressure and Diastolic Blood Pressure Hazard ratios for diabetes rose with increasing blood pressure. Adjustments were made for age, body mass index, smoking status, sex, and blood pressure category (plotted). CIs are displayed as floating absolute risks. Area of each square is proportional to the inverse variance of the estimate. Usual SBP categories were defined by the measured SBP categories: <95 mm Hg, >195 mm Hg, and increments of 10 mm Hg for everything in between (e.g., 95 to 105 mm Hg, 106 to 115 mm Hg, and so on); usual DBP categories were defined by the measured DBP categories: <65 mm Hg, >115 mm Hg, and intervening increments of 10 mm Hg (66 to 75 mm Hg, and so on). Blood pressure categories were entered simultaneously into the Cox model (separate models for SBP and DBP) and estimated simultaneously. Floating absolute risks were used to display all hazard ratios . The variance of each estimate approximates the variance in the underlying category. CI = confidence interval.

Figure 2

Adjusted Hazard Ratios for Diabetes by SBP and BMI or Age Adjustments were for smoking status, sex, and the interaction between SBP as a categorical variable and (A) BMI category (plotted) and (B) age category (plotted). BMI = body mass index; other abbreviations as in Figure 1.

Figure 3

Association Between Blood Pressure and Diabetes per Baseline Variables Adjustments were for age, sex, BMI, baseline antihypertensive use, and baseline lipid-lowering therapy use. For subgroups of age, adjustment was also for age category and the interaction between SBP and age category (plotted). For subgroups of sex, adjustment was also for the interaction between sex and SBP (plotted). For subgroups of BMI, adjustments were also for BMI category and the interaction between SBP and BMI category (plotted). Area of each square is proportional to the inverse variance of the estimate. In various subgroups (usual SBP or DBP, sex, BMI, and age), proportional associations were seen between newly diagnosed diabetes and a 20 mm Hg higher SBP or 10 mm Hg higher DBP. A greater proportional risk was seen with the lowest BMI category versus the highest in both blood pressure groups; similarly, increasing age was associated with decreasing risk. Abbreviations as in Figures 1 and 2.

Figure 4

Association Between Higher Usual SBP and Diabetes Risk A meta-analysis showed that each 20 mm Hg higher usual SBP from previous cohort studies was associated with a 77% higher risk of new diabetes. Study refers to first author of study. Abbreviations as in Figures 1 and 2.