Statistical and Methodological Reviews for Cardiologists

Statistical and Methodological Review of the SPRINT Trial

Reviewer: Robert Long
Position: Principal Consultant
Company: The Data Guru Ltd
Date: 10 May 2025
Document Version: 3.3

I. Introduction & Background

The Systolic Blood Pressure Intervention Trial (SPRINT), published in 2015, was a landmark randomised controlled trial (RCT) evaluating whether intensive lowering of systolic blood pressure (SBP) to a target of <120 mm Hg provided greater cardiovascular protection than the conventional <140 mm Hg target in high-risk, non-diabetic patients.

Prior to SPRINT, guidance on blood pressure targets in high-risk individuals remained inconsistent, partly due to inconclusive or population-specific findings from previous trials. For example, the ACCORD-BP trial showed no significant benefit from intensive control in patients with type 2 diabetes, while the HYVET study demonstrated positive outcomes in older hypertensive adults. In this context, SPRINT aimed to clarify the balance of benefit and risk in a large, high-risk but non-diabetic cohort.

With over 9,300 participants and a primary endpoint focused on major cardiovascular events, SPRINT generated substantial impact on clinical guidelines — particularly in the United States, where it helped lower treatment thresholds. However, as this review will argue, the trial’s design, statistical analysis, population characteristics, and long-term implications warrant critical evaluation before its findings can be applied broadly to everyday cardiology practice.

This review re-examines the trial's methodology, evaluates the internal and external validity of its findings, and integrates recent follow-up studies and meta-analyses to place the results in broader clinical perspective.

II. Trial Design and Methodology

SPRINT was a multicentre, open-label, randomised controlled trial funded by the U.S. National Institutes of Health. It enrolled 9,361 participants across 102 clinical sites in the United States and Puerto Rico between 2010 and 2013. Participants were randomised 1:1 to either an intensive SBP target (<120 mm Hg) or a standard target (<140 mm Hg), with the primary outcome being a composite of myocardial infarction (MI), acute coronary syndrome (ACS), stroke, heart failure (HF), or cardiovascular death.

Inclusion and Exclusion Criteria

Eligible participants were aged ≥50 years with systolic BP between 130 and 180 mm Hg, and at least one additional cardiovascular risk factor: established clinical or subclinical CVD, chronic kidney disease (eGFR 20–60 mL/min/1.73m²), age ≥75 years, or a 10-year Framingham CVD risk score ≥15%.

Key high-risk groups were explicitly excluded — including patients with diabetes mellitus, prior stroke, proteinuria >1 g/day, symptomatic heart failure, and LVEF <35%. These exclusions aimed to isolate a non-diabetic high-risk cohort distinct from ACCORD-BP and other stroke-focused trials. However, they also limit generalisability to real-world cardiovascular practice, where such comorbidities are common.

Randomisation and Stratification

Randomisation was performed centrally with stratification by clinical site, age group, presence of cardiovascular disease, and CKD status. This strategy helped balance key risk factors across arms and allowed for prespecified subgroup analyses. While this design improved internal validity, it also diluted some of the real-world variation present in routine care.

BP Measurement Protocol

Blood pressure was measured using the Omron HEM-907XL device following a strict protocol: participants rested for five minutes in a seated position, after which three BP readings were taken and averaged. Crucially, measurements were intended to be unattended — a deliberate step to reduce the white-coat effect and inter-operator variability.

Subsequent analysis by Wright et al. (2021) confirmed that staff presence had little effect on readings. Nevertheless, this protocol differs sharply from typical outpatient practice, where readings may be taken under time pressure, with active staff interaction and without prolonged rest. As such, the SBP targets defined in SPRINT are not directly interchangeable with clinic-based values recorded under less controlled conditions.

Intervention and Follow-up

Antihypertensive treatment was adjusted monthly to meet BP targets, using a stepped-care algorithm consistent with contemporary U.S. guidelines. The mean number of medications was higher in the intensive group, as expected. Patients were followed for a median of 3.26 years until the trial was stopped early due to apparent benefit in the primary outcome.

III. Statistical Methods and Analysis

The SPRINT investigators used a time-to-event framework, with the primary outcome being the time to first occurrence of a major cardiovascular event from a pre-specified composite. Kaplan–Meier estimates were used to visualise event-free survival, and comparisons were made using the log-rank test. Cox proportional hazards models, stratified by clinical site and adjusted for pre-specified covariates, were used to estimate hazard ratios (HRs) with 95% confidence intervals.

Proportional Hazards Assumption

While the assumption of proportional hazards was stated, formal testing was limited. Schoenfeld residual diagnostics were performed for the primary outcome, but not systematically reported across secondary endpoints or key subgroups. Visual inspection of survival curves, particularly in CKD strata, reveals delayed separation suggestive of time-varying effects. For a trial that stopped early, such non-proportionality could distort the perceived benefit and should have been explicitly addressed.

Interim Monitoring and Early Stopping

Interim analyses were guided by O’Brien–Fleming boundaries to preserve type I error. The trial was stopped at a median of 3.26 years due to significant benefit in the primary endpoint. While this is procedurally defensible, early stopping often leads to overestimation of effect sizes and uncertainty in secondary endpoints, as discussed by Pocock & Hughes (1989).

Trials stopped early often capture peak statistical signals by chance. This inflates perceived benefit, particularly if long-term trends would have regressed toward null.

Multiplicity and Subgroup Analyses

Although prespecified subgroup analyses were conducted — including age, CKD status, baseline SBP, and prior CVD — adjustment for multiple testing was limited. Notably, the interaction p-value for cognitive outcomes in CKD subgroups was reported as marginally significant (p = 0.04). However, without correction for multiplicity, this result may reflect chance. More importantly, the interaction is uninterpretable without context on effect size, absolute risk difference, and clinical relevance — particularly in subgroup populations with fewer events.

Effect Size Interpretation

The primary result — a 25% relative reduction in the composite outcome (HR 0.75) — must be interpreted alongside the absolute risk reduction of 1.6% over ~3 years. This translates to a number needed to treat (NNT) of 62. While clinically meaningful, cardiologists should balance this against the increased risk of adverse events, particularly in older and multimorbid populations. The trial was not powered for mortality benefit, and observed reductions in CV death should be interpreted cautiously in light of low event counts and early termination.

Handling of Missing Data

SPRINT reported low rates of missing data and used censoring at last contact for participants lost to follow-up. No multiple imputation or sensitivity analysis for missingness was provided, although the overall impact was likely modest. Still, even small degrees of informative censoring could introduce bias, particularly in subgroup analyses where events were sparse.

V. Critical Appraisal

Internal Validity

Randomisation and allocation concealment were robust. Outcomes were adjudicated blind to treatment. BP measurement was standardised, mostly unattended, and consistent across sites. The main threats to internal validity were early stopping and possible unmeasured site-level effects. Protocol adherence was high, but the open-label design means performance bias cannot be ruled out entirely, especially in subjective outcomes such as syncope.

External Validity

SPRINT excluded major patient groups, including those with diabetes, prior stroke, or significant proteinuria. Women, the frail elderly, and racially diverse patients were also underrepresented relative to their burden of hypertension in practice. These exclusions limit generalisability, particularly for primary care physicians managing multimorbid patients with less intensive follow-up resources.

Composite Outcomes

The primary endpoint was a composite of five major events. Most of the benefit was driven by reductions in heart failure and cardiovascular death. Stroke and MI contributed less. The lack of weighting and minimal justification for the component structure leaves open the possibility that the composite overstates clinical relevance. Disaggregated effect sizes should be given greater prominence in future publications.

Adverse Events

Adverse events such as hypotension, syncope, electrolyte abnormalities, and acute kidney injury were significantly more common in the intensive group. Although most were transient, they have serious implications for older or frailer patients. Wright et al. (2021) acknowledged that for some subgroups — particularly those with low baseline CV risk — the net benefit may be negative.

BP Measurement Context

SPRINT used a protocolised, automated measurement system that is not representative of routine care. In-office readings are typically attended, rushed, and less reproducible. Therefore, the “<120 mm Hg” target in SPRINT cannot be interpreted as equivalent to an outpatient clinic target of <120 mm Hg. This undermines direct application of SPRINT thresholds in settings without the same measurement rigour.

VI. Clinical and Policy Implications

SPRINT helped catalyse a shift toward more intensive SBP targets in clinical guidelines, particularly in the United States. The 2017 ACC/AHA guidelines adopted a general target of <130 mm Hg for high-risk patients, citing SPRINT as key evidence. The European Society of Hypertension and ESC 2023 guidelines also endorse lower targets — including <130 mm Hg for most patients, and potentially <120 mm Hg for those under 70 if tolerated — though with more emphasis on individualised care.

However, the trial's strict measurement protocol, short follow-up, and selective eligibility criteria mean that applying its findings in typical outpatient care requires careful interpretation. For frail older adults, or those with multiple comorbidities, the risks of adverse events may outweigh modest cardiovascular benefits. In primary care, with less frequent monitoring and variable adherence, the same intensive strategy may lead to overtreatment or harm.

Therefore, SPRINT should inform clinical decision-making — not dictate it. In well-resourced settings with capacity for close monitoring, SPRINT-style intensive control may improve outcomes in a subset of high-risk patients. Elsewhere, especially among elderly or medically complex patients, a more conservative, personalised approach remains prudent.

VII. Conclusions

SPRINT represents a landmark contribution to the evidence base for blood pressure management in high-risk, non-diabetic populations. It demonstrated that targeting an SBP of <120 mm Hg — under tightly controlled conditions — reduces cardiovascular events, particularly heart failure and CV death.

However, the trial's applicability to broader cardiology practice is limited by exclusion criteria, protocol-specific BP measurement techniques, early stopping, and modest absolute risk reductions. Moreover, the increased incidence of hypotension, syncope, and AKI — especially in older adults — warrants caution.

Subsequent meta-analyses confirm that while MACEs and stroke are reduced, overall mortality is not clearly affected, and adverse events remain a limiting factor. Clinicians should view the SPRINT findings not as a mandate for universal intensive control, but as a guide for informed, individualised therapy in well-selected patients with capacity for close monitoring and titration.

In summary, SPRINT reshaped guidelines — but its translation into cardiology practice must be tempered by a nuanced appreciation of risk, context, and generalisability.

References