Investigators examined whether visit-to-visit blood pressure variability (BPV) modifies the known gradient of risk that runs from normoglycemia through prediabetes to type 2 diabetes for vascular and neurological outcomes. The analysis used UK Biobank data to test associations with incident dementia, stroke, and all-cause mortality. The cohort included 57,611 participants with a median follow-up of 5.1 years (IQR 3.3–9.5). The study is reported as a medRxiv preprint and has not been peer reviewed.
The authors quantified visit-to-visit variability for systolic (SBP), diastolic (DBP), mean arterial (MAP), and pulse pressures (PP) using absolute and directional measures, as stated in the abstract. Time-updated Cox proportional-hazards models were applied, and models adjusted for potential confounders plus mean blood pressure. The team evaluated both joint and independent associations of BPV and glycemic status with outcomes, probed effect modification, and assessed whether adjusting for BPV attenuated glycemia-associated risk. The abstract did not supply additional methodological specifics beyond these model-level descriptions.
When the investigators compared participants with normoglycemia and low SBP variability to those with type 2 diabetes plus high SBP variability, they observed substantially elevated hazards for mortality and a signal for dementia. Specifically, T2D with high SBP variability was associated with increased risk of all-cause mortality (hazard ratio 2.80, 95% confidence interval 2.12–3.69) and a nominally higher risk of all-cause dementia (HR 2.46, 95% CI 1.37–4.42), as reported in the abstract. Those effect estimates derive from fully adjusted, time-updated models.
Absolute measures of BPV showed differential associations by outcome and pressure component. All absolute variability metrics—across SBP, DBP, MAP and PP—were independently associated with mortality in the adjusted models. Variability in diastolic blood pressure and mean arterial pressure bore associations with overall stroke that remained significant after controlling the false-discovery rate. The authors therefore present a pattern in which different pressure metrics signal distinct outcome risks rather than a uniform BPV effect across all endpoints.
The team examined whether including BPV in models materially changed the estimated risks linked to glycemic status. Additional adjustment for BPV produced only minimal attenuation of the association between glycemia and outcomes, according to the abstract. The investigators also tested multiplicative interaction between BPV and glycemic status; no interaction term remained statistically significant following correction for multiple testing. Those findings indicate that BPV and glycemic status each contribute risk information, but BPV did not explain away the elevated risk seen with worsening glycemia in this cohort.
The report uses UK Biobank data and notes ethical oversight: the North West Multi-centre Research Ethics Committee provided approval for UK Biobank, and analyses were conducted under an approved application. The authors declared no competing interests and stated that data are available on reasonable request. The manuscript is a preprint on medRxiv and has not been certified by peer review; that status limits immediate clinical translation. The abstract does not provide all granular methodological details—for example, the exact computational formulas for each BPV metric and the complete covariate list in each model are not laid out in the abstract text—and those items would require consultation of the full manuscript or supplementary files.
Limitations reported or implicit from the abstract include reliance on observational data, which precludes causal inference, and the preprint status that leaves findings pending peer review. The abstract did not report subgroup demographic breakdowns, nor did it provide the number of events for each outcome in the cohort summary; the available results focus on hazard ratios and which associations survived false-discovery-rate correction. Future peer-reviewed reporting should clarify these details and provide full model specifications and event counts to support reproducibility and clinical interpretation.