A meta-analysis of the long-term effects of antihypertensive therapy on the risk of major cardiovascular disease across 51 randomized trials

Abstract Blood pressure (BP)-lowering therapy reduces cardiovascular risk, but whether its proportional benefits increase with longer treatment duration remains unclear. We conducted an individual participant-level data meta-analysis of 51 randomized trials from the Blood Pressure Lowering Treatment Trialistsʼ Collaboration (358,642 participants; median follow-up: 4.2 years). Using Cox proportional hazards models, we estimated time-stratified hazard ratios (HRs) for major cardiovascular events (MACE; fatal or non-fatal stroke, ischemic heart disease or heart failure) across annual follow-up intervals up to more than 5 years, standardized to a 5-mmHg systolic BP reduction. Network meta-analysis examined whether temporal patterns differed across antihypertensive drug classes. Annual MACE incidence was highest during year 1 (3.0% treatment versus 3.6% control), declined during years 1−5 and then rose at more than 5 years (3.1% versus 3.4%). BP lowering reduced MACE risk, with benefits established early and not progressively increasing over time. A 5-mmHg systolic BP reduction was associated with a 12% lower MACE risk in year 1 (HR = 0.88, 95% confidence interval (CI): 0.84−0.91), with modest attenuation thereafter: HRs were 0.88 (0.85−0.92) in years 1−2, 0.94 (0.90−0.98) in years 2−3, 0.87 (0.83−0.92) in years 3−4, 0.97 (0.91–1.03) in years 4−5 and 0.94 (0.87−1.01) at more than 5 years (P for trend = 0.006). Similar patterns occurred across five drug classes. These findings indicate that the relative cardiovascular benefits of BP lowering emerge within months and do not increase over time, suggesting that prioritizing higher-risk individuals for treatment yields greater clinical utility than prolonged treatment in low-risk individuals. Main With growing emphasis on lifecourse cardiovascular risk management1, BP lowering has become central to discussions of earlier intervention2,3. Observational studies have shown that cumulative BP burden, thought to reflect progressive vascular and cardiac damage4, is strongly associated with cardiovascular disease5,6. In parallel, long-term follow-up of randomized trials has reported so-called ‘legacy effects’7, whereby early BP-lowering treatment is associated with persistent cardiovascular protection beyond the trial period. Mendelian randomization studies, often considered proxies for lifelong BP exposure, have further reinforced these observations, showing substantially larger relative reductions in cardiovascular risk associated with genetically lower BP than those observed in short-term pharmacological trials8,9,10. Together, this evidence has motivated the hypothesis that longer BP-lowering treatment duration could yield progressively greater relative cardiovascular benefit10. Similar time-dependent patterns have also been observed in lipid-lowering treatment11,12, where statin trials demonstrate modest early effects followed by larger proportional risk reductions with longer follow-up. By analogy, these findings have informed expectations of progressively increasing relative benefits with longer BP-lowering treatment. On this basis, earlier and more proactive BP-lowering strategies have been proposed, including age-specific treatment thresholds13 and broader lifecourse prevention paradigms. Such approaches have been framed metaphorically as ‘vaccine-like’14, in the sense that BP-lowering and lipid-lowering interventions initiated earlier in life may confer long-term cardiovascular protection, on the premise that earlier intervention may yield disproportionately larger relative risk reductions over the lifecourse15. However, no randomized evidence has directly tested whether BP-lowering therapy itself exhibits a similar time-dependent amplification of relative benefit. Most trials are neither designed nor sufficiently powered to assess whether the magnitude of cardiovascular risk reduction varies by treatment duration. Consequently, it remains unclear whether BP-lowering therapy follows the compounding efficacy profile observed in cholesterol management or whether treatment effects simply plateau after an initial period. Utilizing an individual participant-level data (IPD) meta-analysis of randomized controlled trials, we investigated whether antihypertensive therapy confers a cumulative temporal benefit analogous to lipid-lowering interventions, aiming to optimize BP management strategies across the lifecourse. Results Summary of included studies Of the 52 randomized trials included in the Blood Pressure Lowering Treatment Trialists’ Collaboration (BPLTTC) resource, one trial was excluded owing to a lack of time-to-event data16, leaving 51 trials with a total of 358,642 participants. A total of 1,328 participants were excluded because follow-up time for the primary outcome was missing, leaving 357,314 participants in the analytic cohort. The median follow-up duration across these trials was 4.2 years (interquartile interval: 3.0−5.0 years), ranging from 1.6 years in the Comparison of Amlodipine vs Enalapril to Limit Occurrences of Thrombosis (CAMELOT) trial17 to 7.9 years in the United Kingdom Prospective Diabetes Study (UKPDS)18. The number of participants contributing data to the primary outcome declined progressively across annual follow-up intervals (0−1, 1−2, 2−3, 3−4, 4−5 and >5 years), due to censoring and event occurrence, from 357,314 participants in the first year to 82,678 participants at more than 5 years (Table 1). Baseline characteristics Baseline characteristics at randomization are presented for the subsets of participants contributing data to each follow-up interval. Within each interval, treatment and control groups remained well balanced. However, participants contributing data to later follow-up intervals were, on average, younger at baseline and had higher baseline BP levels and fewer cardiovascular comorbidities. In the treatment group, the mean age at randomization was 64.8 years in the years 0−1 dataset and 61.7 years in the >5-year group. Mean systolic BP was 152.6 mmHg in the years 0−1 cohort and 155.8 mmHg in the >5-year group, and mean diastolic BP was 87.5 mmHg and 91.4 mmHg, respectively. The proportion of treatment participants with a history of cardiovascular disease was 44.6% in years 0−1 and 26.6% at more than 5 years. By contrast, the prevalence of chronic kidney disease and the history of angiotensin-converting enzyme (ACE) inhibitor and β-blocker use were higher in the groups with longer follow-up durations. Detailed baseline characteristics for each interval are provided in Table 1. Effects of BP-lowering treatment on MACE and death across successive follow-up intervals Using prespecified 1-year follow-up intervals, we estimated interval-specific HRs with trial-stratified Cox models, standardized treatment effects to a 5-mmHg reduction in systolic BP and formally tested for linear trend across follow-up periods. The annual incidence of MACE was highest during the first year of follow-up (3.0% in the treatment group versus 3.6% in controls) and then declined during years 1−5, followed by a subsequent rise at more than 5 years (3.1% versus 3.4%). The composition of MACE varied over time: the incidence of stroke and heart failure remained largely constant, whereas ischemic heart disease and cardiovascular death were more frequent in later follow-up periods (Fig. 1). In the primary standardised analysis, there was no evidence of compounding treatment effects with longer follow-up. Relative risk reductions were established early after treatment initiation and then modestly attenuated over time (P for trend = 0.006). Specifically, a pharmacological reduction of 5-mm Hg in systolic BP reduced MACE by 12% in the first year (HR 0.88, 95% CI 0.84−0.91), with evidence of effect attenuation in subsequent years: HR 0.88 (0.85−0.92) in year 1−2, HR 0.94 (0.90−0.98) in year 2−3, HR 0.87 (0.83−0.92) in year 3−4, HR 0.97 (0.91−1.03) in year 4−5, and HR 0.94 (0.87−1.01) at >5 years. Secondary outcomes showed similar patterns of relative risk reduction, without evidence of compounding effects, although statistical precision was reduced (Fig. 1). The most pronounced attenuation was observed for heart failure, where the relative effects were largest in the first year (27% reduction, HR = 0.73, 0.65−0.80), decreased significantly over time (P for trend < 0.001) and were compatible with no effect at more than 5 years (1.02 (0.85−1.23)). All-cause death showed suggestive evidence of diminishing effect (P for trend = 0.042). For stroke, relative effects appeared smaller in later years (HR = 0.99, 0.87−1.13 at >5 years), but there was no statistically significant trend (P for trend = 0.115). Relative effects for ischemic heart disease and cardiovascular death were broadly consistent across follow-up intervals (P for trend = 0.569 and 0.373, respectively). Continuous effects of BP-lowering treatment on major cardiovascular disease and death over time To complement the interval-specific analyses, we used flexible parametric survival models to visualize treatment effects continuously over follow-up (Fig. 2). For MACE, significant treatment effects emerged rapidly after randomization and were established early after initiation, with relative risk reductions persisting throughout follow-up without evidence of progressive amplification over time. Secondary outcomes showed a similar rapid onset of treatment effects, although temporal patterns varied across outcomes, and estimates became less precise at longer follow-up. For ischemic heart disease, relative risk reductions were apparent early and appeared broadly stable over time. By contrast, heart failure exhibited a pronounced early treatment effect that attenuated over follow-up and then plateaued. For stroke, early treatment effects were observed, followed by gradual attenuation with increasing uncertainty at later timepoints. For cardiovascular death and all-cause death, early relative risk reductions were evident, but estimates became imprecise and approached the null with longer follow-up (Fig. 2). Sensitivity analyses Sensitivity analyses were performed to assess the robustness of the primary findings to alternative modeling of follow-up time, informative censoring, treatment adherence, trial selection and BP standardization, yielding results consistent with the primary analysis. First, to assess whether the results were sensitive to the discretization of follow-up time into ordered 1-year intervals, we fitted a continuous-time Cox model using a counting process framework. The results were consistent with those of the primary analysis. There was no evidence of time-dependent modification of treatment effects for MACE (interaction HR = 1.014 per unit increase in log-year, 95% CI: 0.998−1.030; P = 0.095) or for secondary outcomes including ischemic heart disease (P = 0.704), stroke (P = 0.524), cardiovascular death (P = 0.992) and all-cause death (P = 0.236). The only exception was heart failure, for which treatment effects attenuated significantly over time (interaction HR = 1.108, 95% CI: 1.064−1.153; P < 0.001) (Extended Data Table 1). Second, to help address potential selection bias arising from informative censoring across follow-up intervals, we applied inverse probability of censoring weighting (IPCW) based on measured baseline covariates. After weighting, key covariates were well balanced across follow-up cohorts compared to the unweighted analyses (Extended Data Table 2). IPCW-adjusted estimates aligned with the primary analysis and again showed no evidence of compounding treatment effects (Extended Data Fig. 1). For MACE, the HR per 5-mmHg systolic BP reduction was 0.88 (0.84−0.91) in years 0−1 and tended to diminish but remained effective after 5 years with HR = 0.91 (0.84−1.00) (P for trend = 0.022). Attenuation remained most evident for heart failure, whereas ischemic heart disease and cardiovascular death showed no clear time-varying effects. To further account for time-varying factors related to treatment persistence and evolving clinical status, we extended the IPCW model in the subset of nine trials with repeated adherence assessments by incorporating time-updated adherence. Results from this expanded model were again consistent with the primary findings (Extended Data Fig. 2). Third, to examine the potential influence of declining adherence over time, we restricted the analyses to participants with high adherence (defined as taking ≥80% of assigned medication) in the 12 trials with available overall adherence data. In this high-adherence subgroup, there was no evidence of an increase in treatment benefit with longer follow-up for MACE (P for trend = 0.111). Notably, the attenuation observed in the primary analysis was not evident: the HR at more than 5 years was 0.84 (95% CI: 0.71−0.99), similar to the first-year effect (HR = 0.87, 95% CI: 0.82−0.93). Temporal patterns for secondary outcomes were broadly similar to the primary analysis, with heart failure again showing the largest early benefit, followed by attenuation (P for trend = 0.002) (Extended Data Fig. 3). Finally, results were also consistent after excluding trials terminated early for benefit (Supplementary Fig. 1), excluding trials comparing different antihypertensive drug classes (Supplementary Fig. 2), analyzing treatment effects without BP reduction standardization (Supplementary Fig. 3), standardization using within-interval BP differences (Supplementary Fig. 4) and standardization to a 3-mmHg reduction in diastolic BP (Supplementary Fig. 5). Across these analyses, there was no evidence of compounding treatment effects over time across outcomes. Subgroup analyses Subgroup analyses by age category (<55, 55−64, 65−74, 75−84 and ≥85 years) showed broadly similar patterns across all groups. Relative treatment effects for MACE remained stable or tended to decline across follow-up intervals, with no evidence of compounding effects (Fig. 3). Broadly similar patterns were observed for secondary outcomes (Supplementary Figs. 6−10) as well as in analyses stratified by sex (Supplementary Figs. 11 and 12) and baseline diabetes status (Supplementary Figs. 13 and 14). Effects of major antihypertensive drug classes on major cardiovascular disease and death across successive follow-up intervals To examine whether the temporal pattern of treatment effects differed by antihypertensive drug class, we conducted an IPD network meta-analysis across the drug classes included in the eligible trials. The network meta-analysis included 34 trials comparing five major antihypertensive drug classes. Across all drug classes, there was no evidence of a compounding benefit for MACE across follow-up intervals (all P for trend > 0.05; Fig. 4), and a similar pattern was observed for most individual cardiovascular outcomes and all-cause death (Supplementary Figs. 15−19). However, angiotensin II receptor blockers (ARBs) demonstrated a non-significant effect in preventing stroke during the first year (HR = 0.92, 95% CI: 0.84−1.02), with a greater reduction emerging over time (P for trend = 0.046; Supplementary Fig. 16). By contrast, the early protective effects of thiazide diuretics on heart failure peaked in the first year (HR = 0.40, 95% CI: 0.30−0.54) and declined soon thereafter (P for trend = 0.034; Supplementary Fig. 17), which may explain the attenuation of heart failure risk reduction observed in the primary analysis. Additionally, diminishing effects were observed for heart failure with calcium channel blockers (CCBs) (P for trend = 0.010) and for all-cause death with thiazide diuretics (P for trend = 0.012). Cumulative absolute risk reduction analyses To complement the primary analyses on the relative scale, we estimated fixed-horizon cumulative absolute risk reductions from randomization. In these analyses, which were anchored to the original randomized population, cumulative absolute risk reductions standardized to a 5-mmHg lower systolic BP increased over time for MACE and several secondary outcomes (Extended Data Table 3). For MACE, the absolute risk reduction increased from 0.30% (95% CI: 0.20−0.40) at 1 year to 1.25% (0.94−1.57) at 5 years, corresponding to numbers needed to treat of 337 and 80, respectively. For stroke, the corresponding estimates were 0.07% (0.02−0.12) at 1 year and 0.52% (0.36−0.68) at 5 years, corresponding to numbers needed to treat of 1,535 and 192, respectively. Absolute benefits also accumulated for death outcomes: for cardiovascular death, from 0.04% (−0.00 to 0.08) at 1 year to 0.29% (0.12−0.47) at 5 years; and for all-cause death, from 0.10% (0.04−0.17) to 0.36% (0.11−0.62). For ischemic heart disease, the benefit became more apparent after 1 year, whereas estimates for heart failure were smaller and less consistent. These findings indicate that cumulative absolute benefit accrues over time even though interval-specific proportional effects do not progressively amplify. Discussion This IPD meta-analysis of 51 randomized trials provides robust evidence of rapid-onset cardiovascular benefits after initiation of BP-lowering treatment, with benefits sustained but without evidence of compounding. Across most outcomes, relative risk reductions were largely established soon after treatment initiation and then remained stable or modestly attenuated over time, with later interval-specific estimates conditional on the evolving risk set. For heart failure prevention, relative risk reductions were most pronounced in the first year and decreased thereafter, a pattern consistent with the rapid hemodynamic effects of antihypertensive therapy. Our network meta-analysis further confirmed that this pattern was broadly consistent across the major antihypertensive drug classes, including ACE inhibitors, ARBs, β-blockers, CCBs and thiazide diuretics. Complementary fixed-horizon analyses on the absolute scale showed that cumulative absolute benefit continued to accrue with longer treatment exposure. Studies of BP lowering typically report average treatment effects across the entire follow-up period. Owing to limited sample sizes, few studies have evaluated how relative risk reductions may vary with treatment duration. Thus, it has remained uncertain whether cardiovascular benefits from BP-lowering therapy continue to accrue or plateau after initial gains. Mendelian randomization studies, often interpreted as proxies for lifelong BP reduction, tend to show lower HRs than short-term trials8,9,15, fueling the hypothesis of compounding benefits over time and motivating recent calls for early cardiovascular prevention strategies19; however, differences in study design and assumptions limit direct comparability20. Analysis of long-term treatment effects from individual clinical trials after the end of randomized therapy can provide another perspective on potential compounding effects. So far, some studies have provided supporting evidence for the existence of ‘legacy effects’7,21, implying that longer duration of antihypertensive therapy may lead to cumulative vascular protection that persists even after treatment cessation. For example, in the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT) Legacy Study, participants originally assigned to the amlodipine-based regimen had lower long-term cardiovascular death21. However, these findings have not been confirmed in long-term follow-up studies of large trials, such as the Systolic Blood Pressure Intervention Trial (SPRINT) and the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT)22,23. By leveraging the largest IPD resource available to date, our study directly addresses this evidence gap by estimating treatment effects across successive annual follow-up intervals. Notably, each interval included large numbers of participants, ranging from over 350,000 in the first year to over 80,000 at more than 5 years. Across these intervals, we observed no evidence of compounding relative benefits over time for MACE or all-cause death. Treatment benefits were evident within the first months of follow-up and did not increase with longer treatment duration; instead, they tended to plateau or attenuate modestly over time. When treatment effects were modeled as continuous functions of follow-up time, estimates derived at finer temporal resolution demonstrated patterns consistent with the primary time-stratified analyses, providing no evidence of progressive amplification with longer treatment duration. For individual components of MACE, broadly similar results were observed for ischemic heart disease and cardiovascular death, whereas variations have been found in stroke and heart failure. Regarding stroke prevention, ARBs demonstrated a gradual increase in relative risk reduction over time, consistent with previous observations of delayed stroke benefit in trials such as the PRoFESS trial24,25. By contrast, for heart failure, we observed a pronounced initial reduction in incidence during the first year, which diminished significantly in subsequent years. Our network meta-analysis indicates that this pattern is largely driven by diuretics, which exhibit a similar early benefit in year 1 that declines soon afterwards. This may be attributed to the fact that diuretics rapidly relieve heart failure symptoms associated with fluid overload, thereby preventing the clinical manifestation of heart failure events26. A directionally similar temporal pattern was also observed for CCBs; however, none of the estimates was statistically significant and should be interpreted cautiously. Notably, the absence of progressively increasing proportional effects should not be interpreted as evidence of no cumulative absolute treatment benefit. In complementary fixed-horizon analyses anchored to the original randomized population, cumulative absolute risk reductions increased steadily over time for MACE and several secondary outcomes. These analyses provide an important complement to the interval-specific HRs because they are estimated on the absolute scale and do not condition on event-free survival within successive follow-up intervals. Together, the relative-scale and absolute-scale results suggest that BP-lowering treatment confers protection early and that continued exposure allows this benefit to accumulate over time in absolute terms but without progressive amplification of proportional effects. These findings have potentially major implications for cardiovascular prevention and clinical decision-making. Our results do not support the assumption that the relative effects of BP lowering necessarily amplify with longer treatment duration during the period of available randomized follow-up. Rather, the rationale for early and sustained treatment lies in the early establishment of relative benefit and the continued accrual of absolute risk reduction with ongoing exposure. As a result, treatment decisions should remain guided primarily by baseline cardiovascular risk and expected absolute benefit13 rather than by an expectation of progressively larger proportional effects over time. Combined with previous BPLTTC findings demonstrating consistent relative risk reductions across age groups2, and age-stratified analyses in this study, our findings do not support changing treatment thresholds solely on the assumption that earlier exposure yields progressively larger proportional benefits. The non-compounding benefit observed with BP-lowering treatment contrasts markedly with that of cholesterol-lowering therapies. Meta-analyses from the Cholesterol Treatment Trialists’ (CTT) Collaboration found that reducing low-density lipoprotein cholesterol (LDL-C) lowers cardiovascular risk by about 14% in the first year but by 20−30% in each subsequent year up to more than 5 years (P for interaction < 0.001)12. Similarly, Wang et al.11 demonstrated a compounding effect of LDL-C lowering, with meta-regression analyses showing that cholesterol lowering leads to progressively larger reductions in cardiovascular risk with longer follow-up. Although similar compounding effects might have been anticipated for antihypertensive treatment, our results do not support this expectation. These contrasting efficacy profiles may reflect differences in underlying mechanisms: continued exposure to LDL-C leads to the development of atherosclerotic plaque27,28, and cholesterol-lowering therapies (such as statins) act on atherosclerotic plaques, promoting gradual stabilization and regression and resulting in accumulating benefit over time29,30, whereas antihypertensive therapies primarily alleviate mechanical arterial stress, which may explain their rapid and stable risk reductions31. Although BP reduction may also favorably modify arterial structure32, it may not contribute to a compounding effect. Key strengths of our study include access to the largest IPD resource from randomized trials to date, facilitating robust estimates for each year of treatment duration. Notably, consistent findings across multiple complementary analytical approaches strengthened confidence in the robustness of the results. Additionally, the inclusion of multiple drug classes and comparison types (placebo and active-controlled trials) enabled, to our knowledge, the first comprehensive network meta-analyses investigating drug-class-specific effects over time. This study also has some limitations. First, trials included in this study had a median follow-up of only 4.2 years, which is potentially insufficient to detect longer-term benefits. Considering that BP-lowering therapy may prevent or delay hypertension-induced pathological changes such as atherosclerosis33, left ventricular hypertrophy34 and carotid intima-media thickening35, longer-term evidence is needed to fully elucidate any potential compounding effects. Because trials typically end early once benefit is proven, post-trial extensions are necessary to assess long-term outcomes. Second, the interpretation of the time-stratified estimates is limited because participants contributing to later intervals had survived and remained event free and uncensored in earlier intervals. Interval-specific HRs, therefore, reflect evolving risk sets rather than a fixed baseline population, and IPCW cannot fully account for unmeasured factors related to censoring. However, the overall pattern was similar across sensitivity analyses and was not materially altered by alternative modeling of follow-up time, changes in cohort composition, attenuation of BP separation or declining adherence. These findings make it less likely that the absence of progressively stronger effects was explained solely by informative censoring. Third, the present analysis did not evaluate the effects of treatment on other outcomes, such as cognitive function and chronic kidney disease, over time, and further studies will be essential to clarify these patterns. Fourth, because the mean age of the trial population was over 60 years, and even the youngest subgroup (<55 years) represented midlife rather than early adulthood, these findings should not be extrapolated directly to treatment initiated in early adulthood. However, to our knowledge, trials conducted in low-risk younger individuals are lacking. In conclusion, this study challenges the widely held assumption that the benefits from antihypertensive treatment compound over a longer duration of therapy. Instead, our findings show that antihypertensive therapy provides substantial cardiovascular benefits early, which are maintained over time without progressive amplification, although later interval estimates should be interpreted in the context of changing risk sets over time. Sustaining these benefits likely requires continued adherence over the long term. The early emergence of relative treatment effects suggests that clinically meaningful benefit can be achieved soon after antihypertensive treatment is initiated. On the other hand, the absence of compounding effects suggests that targeting individuals at higher cardiovascular risk may be more efficient than expanding treatment to lower-risk individuals on the assumption that longer treatment duration yields increasingly larger proportional benefits. Methods This IPD meta-analysis follows the current recommendations of the PRISMA-IPD statement36. The PRISMA-IPD checklist was completed (Supplementary Table 1). Study design We conducted a time-stratified IPD meta-analysis of eligible BP-lowering trials using the resource provided by the BPLTTC37,38. A study protocol was developed before releasing a dataset for statistical analysis and was finalized with extensive feedback from international collaborators and the BPLTTC steering committee. To date, this resource provides IPD from 52 randomized trials and 363,684 participants2,3,37. Participants in each study were randomly assigned to intervention or control groups, and analyses followed the intention-to-treat principle. Characteristics at trial baseline, including age, sex, systolic and diastolic BP, body mass index, history of cardiovascular disease, diabetes, prior use of antihypertensive medication and other relevant clinical variables, were summarized according to different follow-up durations. In addition, we reported the median follow-up time of participants included in the time-stratified analyses for each follow-up interval. Eligibility criteria We used the most current BPLTTC dataset, which comprised data contributed by participating trials, including BP values at baseline and during follow-up, primary and secondary outcomes with event timing and baseline characteristics of interest. Participants with a known diagnosis of heart failure at trial baseline were excluded from this analysis, because established heart failure defines a clinically distinct population, with different treatment goals, risk profiles and outcome definitions from the broader hypertensive population, which is commonly excluded from individual antihypertension trials2,37. Outcomes The primary outcome was MACE, defined as a composite of fatal or non-fatal stroke, myocardial infarction or ischemic heart disease or heart failure causing death or requiring admission to hospital. Secondary outcomes were all-cause death and each component of the primary outcome. Comparison groups Individual trials were categorized into two groups: intervention and comparator. For placebo-controlled trials, the placebo group was regarded as the comparator, and the active drug group served as the intervention. In trials with two or more active groups, including those comparing different drug classes, the group with the greatest BP reduction was designated the intervention, and the other treatment group(s) were the comparators. Trials that compared more intense versus less intense treatments were classified into intervention and comparator groups, respectively. Detailed information concerning the comparison groups, trial design, participant characteristics and the extent of BP reduction in each trial was reported previously2,3,37. Statistical analyses We analyzed the data on an intention-to-treat basis, according to the groups to which participants were originally randomized (intervention versus comparator). IPD from each trial were harmonized before statistical analysis and merged into a single dataset, enabling a one-stage IPD meta-analysis. Participants contributed data from randomization until the earliest occurrence of the outcome of interest or censoring. Primary time-stratified analysis To evaluate how treatment effects evolved over time, the median intervention duration of 4.2 years (interquartile interval: 3.0−5.0 years) was divided into discrete 1-year intervals: 0−1, 1−2, 2−3, 3−4, 4−5 and >5 years39,40. Participants contributed follow-up time to each interval until they developed the outcome, died or were otherwise censored. All participants were included in the 0−1-year interval, but only event-free participants continued to contribute in subsequent intervals. For each interval, follow-up was truncated at the end of the interval, so that only events occurring within that interval were counted in its analyses. Separate datasets were created for each interval using a consistent censoring and truncation rule. Within each interval’s dataset, a Cox proportional hazards model was employed, stratified by trial, to estimate HRs for primary and secondary outcomes over time. Effect sizes of each interval were standardized to a 5-mmHg reduction in systolic BP using trial-specific mean between-group BP differences estimated from follow-up measurements, excluding the first year2,3. These differences were estimated using a linear mixed-effects model and incorporated into the Cox model as a continuous trial-level variable interacting with treatment allocation, allowing treatment effects to be compared on a common, clinically interpretable scale while accounting for between-trial differences in treatment intensity40. Formal assessment of temporal trend in treatment effects To formally assess whether treatment effects changed across follow-up periods, follow-up time was modeled using a time-split approach, with successive 1-year intervals represented by a numeric time period index. In the survival analysis, a stratified Cox proportional hazards model including an interaction term between treatment allocation and this time variable was fitted, and the interaction term was assessed using a Wald test to evaluate a linear trend in treatment effects with increasing follow-up duration38,41. Continuous-time modeling of treatment effects To complement the primary time-split analyses and provide a smooth graphical description of how treatment effects evolved over time without relying on arbitrary interval cutpoints, we additionally modeled treatment effects using a complementary continuous-time approach based on flexible parametric survival models. Specifically, for each outcome, we fitted Royston−Parmar spline-based survival models with three degrees of freedom for the baseline log cumulative hazard42, allowing the treatment effect to vary smoothly over time by including a time-varying coefficient treatment allocation. Models were stratified by trial, and effect estimates were standardized to a 5-mmHg difference in systolic BP. Time-varying HRs were estimated across follow-up by predicting model-based treatment effects at 30-day intervals from 30 days after randomization up to 6 years of follow-up. Sensitivity analyses and subgroup assessment Several sensitivity analyses were conducted to assess the robustness of the primary findings. First, to evaluate whether the results were sensitive to the modeling of follow-up time, we fitted an alternative time-dependent Cox proportional hazards model using a counting process framework, in which follow-up time was treated as a continuous variable and the treatment effect was allowed to vary with time43, complementary to the primary time-split analysis. Second, to account for potential selection bias due to informative censoring across follow-up intervals, estimates were adjusted using IPCW44. Stabilized weights were derived from logistic regression models estimating the probability of remaining uncensored beyond each follow-up interval, conditioned on randomized treatment allocation, age, sex, baseline systolic BP, body mass index, history of cardiovascular disease, history of diabetes and prior use of antihypertensive drugs. To assess covariate balance, the distributions of these variables were examined before and after weighting across follow-up cohorts. The IPCW-adjusted HRs were estimated using weighted Cox proportional hazards models stratified by trial and were standardized to a 5-mmHg reduction in systolic BP. In addition, to account for potential informative censoring related to time-varying treatment adherence, we repeated the IPCW analyses in the subset of nine trials with repeated adherence assessments. Stabilized weights were estimated using the same baseline covariates as above, plus the most recently observed adherence status at each landmark, and weighted Cox models additionally adjusted for interval-specific adherence status. Third, to examine the potential influence of declining treatment adherence over time, analyses were restricted to participants with high adherence (defined as taking at least 80% of the prescribed medication) across 12 trials with available adherence data. Fourth, trials comparing two antihypertensive drug classes were excluded, as they typically achieve similar BP control across treatment groups. Fifth, analyses were repeated after excluding trials that were terminated early because of clear evidence of benefit in the treatment groups. Finally, additional sensitivity analyses were performed to evaluate the robustness of standardization methods. These included estimating unstandardized treatment effects without accounting for BP reduction, as well as standardizing treatment effects using within-interval BP differences, by calculating the mean BP difference between treatment groups separately for each follow-up interval rather than using a single overall BP difference as in the primary analysis, and standardized to a 3-mmHg reduction in diastolic BP, which was the mean diastolic BP reduction between randomized groups, excluding the first 12 months across all trials2. We examined the time-stratified effects of BP-lowering treatment on cardiovascular disease risk by baseline age groups (<55, 55−64, 65−74, 75−84 and ≥85 years), sex and diabetes status. Potential variations in treatment effects across follow-up periods were then evaluated within each subgroup. Network meta-analysis We used IPD network meta-analysis to compare the effects of five major antihypertensive drug classes across different treatment intervals: ACE inhibitors, ARBs, β-blockers, CCBs and thiazide diuretics. Unlike the primary analyses, this network meta-analysis estimated total class effects, capturing both BP-mediated and non-BP-mediated effects, rather than standardizing effects to a fixed BP reduction magnitude. Because BP-lowering intensity trials did not compare specific antihypertensive drug classes, only placebo-controlled and drug class comparison trials were included in this analysis. The network meta-analysis was conducted at the level of randomized drug class allocation, following the intention-to-treat principle. Estimated effects, therefore, reflected assignment to a given drug class strategy, including any protocol-permitted background or add-on therapy. Trial arms in which the randomized intervention consisted of fixed combination therapies were excluded to ensure that each comparison could be attributed to a single drug class. A Cox proportional hazards model was used to estimate the HRs for each pairwise comparison using IPD from each trial, with censoring applied at the upper boundary of each follow-up interval. To run the network meta-analysis, we used the Markov chain Monte Carlo simulation approach with four chains and 100,000 iterations after an initial burn-in of 10,000 (ref. 45). To assess whether treatment effects changed across follow-up, we performed a meta-regression of the interval-specific treatment estimates derived from the IPD network meta-analysis, with the Knapp−Hartung adjustment applied46. Complementary analysis As a complementary analysis, we estimated cumulative absolute risk reductions from randomization at 1, 2, 3, 4 and 5 years. Within each trial, fixed-horizon risks were estimated separately by randomized group using Kaplan−Meier methods for all-cause death and cumulative incidence functions for non-fatal outcomes, with death treated as a competing event where applicable. Trial-specific absolute risk reductions were defined as the difference in cumulative risk between control and treatment groups at each horizon and were pooled using two-stage fixed-effect meta-analysis. To standardize estimates across trials, absolute risk reductions were additionally rescaled to a 5-mmHg lower systolic BP on the basis of the achieved trial-level between-group systolic BP difference. Numbers needed to treat were derived as the inverse of the pooled absolute risk reduction when positive. Statistical analyses were performed using R (v.4.2.2). Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Data availability The governance of the BPLTTC was reported previously37. Individual participant data used in this study were contributed to the BPLTTC by the original trial custodians under data-sharing agreements and cannot be deposited in a public repository or redistributed by the BPLTTC because they are subject to confidentiality and data governance restrictions. Scientific activities based on the BPLTTC dataset are overseen by the BPLTTC steering committee. Requests for data should be made directly to the data custodians for each trial. Information about individual projects is posted at https://www.wrh.ox.ac.uk/research/Blood_Pressure_Lowering_Treatment_Trialists_Collaboration_BPLTTC. Code availability Code used for the analyses is available upon reasonable request. References Wagner, C. et al. Life course epidemiology and public health. Lancet Public Health 9, e261–e269 (2024). Blood Pressure Lowering Treatment Trialistsʼ Collaboration. Age-stratified and blood-pressure-stratified effects of blood-pressure-lowering pharmacotherapy for the prevention of cardiovascular disease and death: an individual participant-level data meta-analysis. Lancet 398, 1053–1064 (2021). Rahimi, K. et al. Pharmacological blood pressure lowering for primary and secondary prevention of cardiovascular disease across different levels of blood pressure: an individual participant-level data meta-analysis. Lancet 397, 1625–1636 (2021). Masrouri, S. et al. Cumulative blood pressure exposure and global and regional cardiac structure and function: the MESA study. Eur. J. Prev. Cardiol. 32, 1296–1309 (2025). Reges, O. et al. Association of cumulative systolic blood pressure with long-term risk of cardiovascular disease and healthy longevity: findings from the lifetime risk pooling project cohorts. Hypertension 77, 347–356 (2021). Wang, C. et al. Association of age of onset of hypertension with cardiovascular diseases and mortality. J. Am. Coll. Cardiol. 75, 2921–2930 (2020). Kostis, W. J., Thijs, L., Richart, T., Kostis, J. B. & Staessen, J. A. Persistence of mortality reduction after the end of randomized therapy in clinical trials of blood pressure-lowering medications. Hypertension 56, 1060–1068 (2010). Higgins, H. et al. Estimating the population benefits of blood pressure lowering: a wide-angled Mendelian randomization study in UK Biobank. J. Am. Heart Assoc. 10, e021098 (2021). Malik, R. et al. Relationship between blood pressure and incident cardiovascular disease: linear and nonlinear Mendelian randomization analyses. Hypertension 77, 2004–2013 (2021). Ference, B. A. et al. Clinical effect of naturally random allocation to lower systolic blood pressure beginning before the development of hypertension. Hypertension 63, 1182–1188 (2014). Wang, N., Woodward, M., Huffman, M. D. & Rodgers, A. Compounding benefits of cholesterol-lowering therapy for the reduction of major cardiovascular events: systematic review and meta-analysis. Circ. Cardiovasc. Qual. Outcomes 15, e008552 (2022). Baigent, C. et al. Efficacy and safety of cholesterol-lowering treatment: prospective meta-analysis of data from 90,056 participants in 14 randomised trials of statins. Lancet 366, 1267–1278 (2005). McCarthy, C. P., Rahimi, K. & McEvoy, J. W. Age-stratified risk categories for cardiovascular disease prevention therapies. JAMA Cardiol. 9, 1069–1070 (2024). Holmes, M. V. & Bhala, N. The physiological paradox: reframing the polypill as a vaccine for cardiovascular disease. J. Epidemiol. Community Health 67, 897–902 (2013). Ference, B. A. et al. Association of genetic variants related to combined exposure to lower low-density lipoproteins and lower systolic blood pressure with lifetime risk of cardiovascular disease. JAMA 322, 1381–1391 (2019). Suzuki, H. & Kanno, Y. Effects of candesartan on cardiovascular outcomes in Japanese hypertensive patients. Hypertens. Res. 28, 307–314 (2005). Nissen, S. E. et al. Effect of antihypertensive agents on cardiovascular events in patients with coronary disease and normal blood pressure: the CAMELOT study: a randomized controlled trial. JAMA 292, 2217–2225 (2004). UK Prospective Diabetes Study Group. Tight blood pressure control and risk of macrovascular and microvascular complications in type 2 diabetes: UKPDS 38. BMJ 317, 703–713 (1998). Masson, G. Aiming to disrupt cardiovascular space: 2 industry vets unveil Boston biotech preventative RNAi. Fierce Biotech https://www.fiercebiotech.com/biotech/aiming-disrupt-cardiovascular-space-2-industry-vets-unveil-boston-biotech-preventative-rnai (2025). Ference, B. A., Holmes, M. V. & Smith, G. D. Using Mendelian randomization to improve the design of randomized trials. Cold Spring Harb. Perspect. Med. 11, a040980 (2021). Gupta, A. et al. Long-term mortality after blood pressure-lowering and lipid-lowering treatment in patients with hypertension in the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT) Legacy study: 16-year follow-up results of a randomised factorial trial. Lancet 392, 1127–1137 (2018). Ho, C. L. B. et al. Lack of a significant legacy effect of baseline blood pressure ‘treatment naivety’ on all-cause and cardiovascular mortality in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial. J. Hypertens. 38, 519–526 (2020). Jaeger, B. C. et al. Longer-term all-cause and cardiovascular mortality with intensive blood pressure control: a secondary analysis of a randomized clinical Trial. JAMA Cardiol. 7, 1138–1146 (2022). Yusuf, S. et al. Telmisartan to prevent recurrent stroke and cardiovascular events. N. Engl. J. Med. 359, 1225–1237 (2008). Dahlöf, B. Prevention of stroke: new evidence. Eur. Heart J. Suppl. 11, F33–F38 (2009). Kjeldsen, S. E., Krzysztof, N., Thomas, H. & Mancia, G. The SPRINT study: outcome may be driven by difference in diuretic treatment demasking heart failure and study design may support systolic blood pressure target below 140 mmHg rather than below 120 mmHg. Blood Press. 25, 63–66 (2016). Ference, B. A., Graham, I., Tokgozoglu, L. & Catapano, A. L. Impact of lipids on cardiovascular health: JACCHealth Promotion Series. J. Am. Coll. Cardiol. 72, 1141–1156 (2018). Goldstein, J. L. & Brown, M. S. A century of cholesterol and coronaries: from plaques to genes to statins. Cell 161, 161–172 (2015). Rosenson, R. S., Hegele, R. A. & Koenig, W. Cholesterol-lowering agents. Circ. Res. 124, 364–385 (2019). Cesaro, A., Acerbo, V., Indolfi, C., Filardi, P. P. & Calabrò, P. The clinical relevance of the reversal of coronary atherosclerotic plaque. Eur. J. Intern. Med. 129, 16–24 (2024). Safar, M. E. Mechanism(s) of systolic blood pressure reduction and drug therapy in hypertension. Hypertension 50, 167–171 (2007). Laurent, S. & Boutouyrie, P. The structural factor of hypertension. Circ. Res. 116, 1007–1021 (2015). Ruddy, T. D., Kadoya, Y. & Small, G. R. Targeting atherosclerosis with antihypertensive therapy. J. Nucl. Cardiol. 30, 1627–1629 (2023). Deng, Y. et al. Intensive blood pressure lowering improves left ventricular hypertrophy in older patients with hypertension: the STEP trial. Hypertension 80, 1834–1842 (2023). Wang, J. G. et al. Carotid intima-media thickness and antihypertensive treatment: a meta-analysis of randomized controlled trials. Stroke 37, 1933–1940 (2006). Stewart, L. A. et al. Preferred Reporting Items for Systematic Review and Meta-Analyses of individual participant data: the PRISMA-IPD statement. JAMA 313, 1657–1665 (2015). Rahimi, K. et al. Investigating the stratified efficacy and safety of pharmacological blood pressure-lowering: an overall protocol for individual patient-level data meta-analyses of over 300 000 randomised participants in the new phase of the Blood Pressure Lowering Treatment Trialistsʼ Collaboration (BPLTTC). BMJ Open 9, e028698 (2019). Copland, E. et al. Antihypertensive treatment and risk of cancer: an individual participant data meta-analysis. Lancet Oncol. 22, 558–570 (2021). Reith, C. et al. Effect of statin therapy on muscle symptoms: an individual participant data meta-analysis of large-scale, randomised, double-blind trials. Lancet 400, 832–845 (2022). Canoy, D. et al. Antihypertensive drug effects on long-term blood pressure: an individual-level data meta-analysis of randomised clinical trials. Heart 108, 1281–1289 (2022). Bellera, C. A. et al. Variables with time-varying effects and the Cox model: Some statistical concepts illustrated with a prognostic factor study in breast cancer. BMC Med. Res. Method. 10, 20 (2010). Royston, P. & Parmar, M. K. Flexible parametric proportional-hazards and proportional-odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects. Stat. Med. 21, 2175–2197 (2002). Andersen, P. K. & Gill, R. D. Coxʼs regression model for counting processes: a large sample study. Ann. Stat. 10, 1100−1120 (1982). Mansournia, M. A. & Altman, D. G. Inverse probability weighting. BMJ 352, i189 (2016). Nazarzadeh, M. et al. Blood pressure lowering and risk of new-onset type 2 diabetes: an individual participant data meta-analysis. Lancet 398, 1803–1810 (2021). Hartung, J. & Knapp, G. On tests of the overall treatment effect in meta-analysis with normally distributed responses. Stat. Med. 20, 1771–1782 (2001). Acknowledgements This paper was prepared using Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) and Prevention of Events with Angiotensin-Converting Enzyme Inhibition (PEACE) research materials obtained from the National Heart, Lung, and Blood Institute (NHLBI) Biologic Specimen and Data Repository Information Coordinating Center and does not necessarily reflect the opinions or views of ALLHAT, PEACE or NHLBI. We acknowledge original depositors of the Australian National Blood Pressure Study data and the Australian Data Archive and declare that those who did the original analysis and collection of the data bear no responsibility for the further analysis or interpretation of the data. This study used data from a trial supported by Boehringer Ingelheim. Boehringer Ingelheim had no role in the design, analysis or interpretation of the results; however, they were allowed to review the paper for medical and scientific accuracy regarding Boehringer Ingelheim substances and for intellectual property considerations. Funding This research was funded by the British Heart Foundation (FS/19/36/34346). The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Author information Authors and Affiliations Contributions Q.Y., M.N. and K.R. have full access to the study data and are responsible for the integrity and accuracy of the data analysis. K.R., M.N. and Q.Y. conceived the study. Q.Y., Z.B., M.N. and K.R. were responsible for data curation. All authors were responsible for protocol writing and investigation. Q.Y. conducted the data analyses. All authors interpreted the data. Q.Y. drafted the original paper, which was reviewed and edited by K.R., M.N., M.W., A.R., J.S., W.C.C., Z.B., D.C., G.Z., K.T., B.R.D., J.C. and C.J.P. Q.Y. was responsible for data visualization. K.R. and M.N. supervised the project. K.R. was responsible for the decision to submit the paper for publication. All authors gave final approval of the version to be published. Corresponding author Ethics declarations Competing interests All authors have completed the International Committee of Medical Journal Editors (ICMJE) uniform disclosure form at https://www.icmje.org/disclosure-of-interest/ and declare the following competing interests. K.R. has received grants outside the submitted work from the British Heart Foundation, the Horizon Europe AI4HF consortium (grant R79992/CN001), the Novo Nordisk Oxford Big Data Partnership, the University of Oxford and the UK Research and Innovation Global Challenge Research Fund (grant ES/P011055/1); has received consulting fees from Medtronic CRDN; has received honoraria or fees from Heart, PLoS Medicine, AstraZeneca MEA Region, Medscape and WebMD Medscape UK; and is the editor-in-chief of Heart. Q.Y. has received scholarship support from the China Scholarship Council−University of Oxford Scholarships and the China Oxford Scholarship Fund. M.N. is supported by an individual research fellowship from the British Heart Foundation (grant FS/IPBSRF/22/27060) and has received reimbursement and honoraria from AstraZeneca, Nemysis, Albus Health and BMJ Publishing Group (as statistical adviser of Heart) outside the submitted work. Z.B. has received a PhD fellowship from the British Heart Foundation (FS/PhD/25/29632). D.C. has received support from the UK Research and Innovation Medical Research Council (MR/Y010825/1), the Vivensa Foundation (formerly Dunhill Medical Trust) (ARVHF2402/7) and the National Institute for Health and Care Research (NIR203982) outside the submitted work; the views expressed are not necessarily those of these funders. D.C. has also received an honorarium as Specialty Chief Editor of Frontiers in Cardiovascular Medicine (Cardiovascular Epidemiology and Prevention). J.C. has received grants from the National Health and Medical Research Council of Australia for work unrelated to this submission. M.W. has received personal fees from Amgen, Kyowa Kirin and Freeline. C.J.P. received grants outside the submitted work from the NHLBI (R01-MPI; Angiotensin, resistant hypertension and microbiota), the US Department of Defence (WARRIOR and QUIET WARRIOR W81XWH-17-2-0030), University of Florida Health, the McJunkin Family Charitable Foundation Trust and the Gatorade Trust. All payments were to the University of Florida. C.J.P. has also received honoraria from Elsevier as the Editor-in-Chief of American Heart Journal Plus. W.C.C. has received consulting fees from Alnylam, Idorsia and Azurity Pharmaceuticals and a grant from George Medicines. The other authors declare no competing interests. Peer review Peer review information Nature Medicine thanks Christian Delles, John Sim and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available. Primary Handling Editor: Michael Basson, in collaboration with the Nature Medicine team. Additional information Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Extended data Extended Data Fig. 1 Standardised effects of systolic blood pressure-lowering treatment on outcomes, stratified by follow-up duration, adjusted for selection bias using inverse probability of censoring weighting. Squares indicate hazard ratio (HR) point estimates and horizontal lines indicate 95% confidence intervals. The vertical line indicates HR = 1. n/N indicates the number of participants with events divided by the number of participants included in the corresponding analysis interval. For each follow-up interval, inverse probability of censoring weights (IPCW) were estimated from logistic models for the probability of remaining event-free beyond the corresponding landmark. Stabilised weights used a numerator model including randomised treatment allocation only and a denominator model additionally adjusted for baseline covariates. Weights were estimated among participants surviving beyond each landmark and truncated at the 1st and 99th percentiles. Interval-specific HRs were estimated using trial-stratified weighted Cox models with robust variance estimation and standardised to a 5 mm Hg reduction in systolic blood pressure. P values for trend were calculated using two-sided Wald tests from weighted trial-stratified Cox proportional hazards models with robust variance estimation to assess linear changes in the treatment log-hazard ratio across ordered follow-up intervals. Extended Data Fig. 2 Standardised effects of systolic blood pressure-lowering treatment on outcomes, stratified by follow-up duration, using inverse probability of censoring weighting, incorporating time-varying adherence. Nine trials (AASK, ACCORD, ALLHAT, ASCOT, HIJ-CREATE, IDNT, PROGRESS, SHEP and SPRINT) with longitudinal adherence data were included. Squares indicate hazard ratio (HR) point estimates and horizontal lines indicate 95% confidence intervals. The vertical line indicates HR = 1. n/N indicates the number of participants with events divided by the number of participants included in the corresponding analysis interval. Treatment effects were estimated using trial-stratified weighted Cox models, standardised to a 5 mm Hg reduction in systolic blood pressure. Stabilised inverse probability of censoring weights incorporated baseline covariates and the most recently observed adherence status (high adherence was defined as taking ≥80% of prescribed medication) at each landmark. P values for trend were calculated using two-sided Wald tests from weighted trial-stratified Cox proportional hazards models with robust variance estimation to assess linear changes in the treatment log-hazard ratio across ordered follow-up intervals; for heart failure, P = 1.18 × 10−5. No adjustment was made for multiple comparisons. Extended Data Fig. 3 Standardised effects of systolic blood pressure-lowering treatment on outcomes among participants with high adherence, stratified by follow-up duration. Twelve trials collected and shared adherence data have been included in this analysis. For nine trials (AASK, ACCORD, ALLHAT, ASCOT, HIJ-CREATE, IDNT, PROGRESS, SHEP, and SPRINT) with longitudinal adherence data, a counting process model was employed. Only follow-up periods where participants maintained high adherence status (defined as taking ≥80% of prescribed medication) were included in the analysis. For three trials (ANBP, PREVENT, and PROFESS) lacking longitudinal data, the entire follow-up period of participants who demonstrated overall high adherence (taking ≥80% of prescribed medication) was included. Squares indicate hazard ratio (HR) point estimates and horizontal lines indicate 95% confidence intervals. The vertical line indicates HR = 1. n/N indicates the number of participants with events divided by the number of participants included in the corresponding analysis interval. Treatment effects were estimated using trial-stratified Cox models and standardised to a 5 mm Hg reduction in systolic blood pressure. P values for trend were calculated using two-sided Wald tests of the treatment-by-time interaction term. Supplementary information Rights and permissions Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. About this article Cite this article Yang, Q., Bidel, Z., Canoy, D. et al. A meta-analysis of the long-term effects of antihypertensive therapy on the risk of major cardiovascular disease across 51 randomized trials. Nat Med (2026). https://doi.org/10.1038/s41591-026-04514-3 Received: Accepted: Published: Version of record: DOI: https://doi.org/10.1038/s41591-026-04514-3