Systolic Blood Pressure Response in SPRINT (Systolic Blood Pressure Intervention Trial) and ACCORD (Action to Control Cardiovascular Risk in Diabetes): A Possible Explanation for Discordant Trial Results

A Novel Digital Twin Strategy to Examine the Implications of Randomized Control Trials for Real-World Populations

Randomized clinical trials (RCTs) are essential to guide medical practice; however, their generalizability to a given population is often uncertain. We developed a statistically informed Generative Adversarial Network (GAN) model, RCT-Twin-GAN, that leverages relationships between covariates and outcomes and generates a digital twin of an RCT (RCT-Twin) conditioned on covariate distributions from a second patient population. We used RCT-Twin-GAN to reproduce treatment effect outcomes of the Systolic Blood Pressure Intervention Trial (SPRINT) and the Action to Control Cardiovascular Risk in Diabetes (ACCORD) Blood Pressure Trial, which tested the same intervention but had different treatment effect results. To demonstrate treatment effect estimates of each RCT conditioned on the other RCT patient population, we evaluated the cardiovascular event-free survival of SPRINT digital twins conditioned on the ACCORD cohort and vice versa (SPRINT-conditioned ACCORD twins). The conditioned digital twins were balanced by the intervention arm (mean absolute standardized mean difference (MASMD) of covariates between treatment arms 0.019 (SD 0.018), and the conditioned covariates of the SPRINT-Twin on ACCORD were more similar to ACCORD than a sprint (MASMD 0.0082 SD 0.016 vs. 0.46 SD 0.20). Most importantly, across iterations, SPRINT conditioned ACCORD-Twin datasets reproduced the overall non-significant effect size seen in ACCORD (5-year cardiovascular outcome hazard ratio (95% confidence interval) of 0.88 (0.73-1.06) in ACCORD vs median 0.87 (0.68-1.13) in the SPRINT conditioned ACCORD-Twin), while the ACCORD conditioned SPRINT-Twins reproduced the significant effect size seen in SPRINT (0.75 (0.64-0.89) vs median 0.79 (0.72-0.86)) in ACCORD conditioned SPRINT-Twin). Finally, we describe the translation of this approach to real-world populations by conditioning the trials on an electronic health record population. Therefore, RCT-Twin-GAN simulates the direct translation of RCT-derived treatment effects across various patient populations with varying covariate distributions.

Individualising intensive systolic blood pressure reduction in hypertension using computational trial phenomaps and machine learning: a post-hoc analysis of randomised clinical trials

BACKGROUND

The cardiovascular benefits of intensive systolic blood pressure control vary across clinical populations tested in large randomised clinical trials. We aimed to evaluate the application of machine learning to clinical trials of patients without and with type 2 diabetes to define the personalised cardiovascular benefit of intensive control of systolic blood pressure.

METHODS

In SPRINT, a trial of intensive (systolic blood pressure <120 mm Hg) versus standard (systolic blood pressure <140 mm Hg) systolic blood pressure control in patients without type 2 diabetes, we defined a phenotypic representation of the study population using 59 baseline variables. We extracted personalised treatment effect estimates for the primary outcome, time-to-first major adverse cardiovascular event (MACE; cardiovascular death, myocardial infarction or acute coronary syndrome, stroke, and acute decompensated heart failure), through iterative Cox regression analyses providing average hazard ratio (HR) estimates weighted for the phenotypic distance of each participant from the index patient of each iteration. Next, we trained an extreme gradient boosting algorithm (known as XGBoost) to predict the personalised effect of intensive systolic blood pressure control using features most consistently linked to increased personalised benefit, before evaluating its performance in the ACCORD BP trial of patients with type 2 diabetes randomly assigned to receive intensive versus standard systolic blood pressure control. We stratified patients based on their predicted treatment effect, and key demographic groups (age, sex, cardiovascular disease, and smoking). We assessed the presence of heterogeneity with an interaction test, and assessed the performance of the algorithm in a simulation analysis of SPRINT in the presence or absence of an artificially introduced heterogeneous treatment effect.

FINDINGS

From SPRINT, we included all 9361 study participants (mean age 67·9 years [SD 9·4], 3332 [35·6%] female) who underwent randomisation to either intensive (n=4678) or standard (n=4683) treatment. The median individualised HR for MACE was 0·63 (IQR 0·53-0·78). An eight-feature tool built for this analysis to predict personalised benefit in SPRINT was externally tested in ACCORD BP (4733 participants (mean age 62·7 years [SD 6·7], 2258 [47·7%] female), wherein it successfully identified individuals with differential benefit from intensive versus standard systolic blood pressure control (adjusted HR for MACE of 0·70 [95% CI 0·55-0·90] in individuals with above-median MACE benefit versus 1·05 [95% CI 0·84-1·32] for below-median predicted benefit; p_interaction=0·0184). Subgroup analysis based on age (<65 years: HR 0·89 [95% CI 0·71-1·12]; ≥65 years: 0·85 [0·67-1·09]), sex (male: 0·89 [0·72-1·10]; female: 0·85 [0·65-1·10]), established cardiovascular disease (no: 0·89 [0·70-1·14]; yes: 0·84 [0·67-1·06]), or active smoking (no: 0·85 [0·71-1·02]; yes: 1·01 [0·64-1·60]) did not identify groups with heterogeneity of treatment effect. In a simulation analysis of SPRINT, the proposed algorithm detected groups with heterogeneous treatment effects in the presence, but not absence, of simulated subgroup differences.

INTERPRETATION

By use of machine learning to define an individual's personalised benefit through phenotypic representations of clinical trials, we created a practical tool for individualising the selection of intensive versus standard systolic blood pressure control in patients without and with type 2 diabetes.

FUNDING

National Heart, Lung, and Blood Institute of the US National Institutes of Health.

Artificial Intelligence (AI) approach to identifying factors that determine systolic blood pressure in type 2 diabetes (study from the LOOK AHEAD cohort)

BACKGROUND AND AIMS

Artificial Intelligence (AI) methods have recently become critical for research in diabetes in the era of big-data science.

METHODS

In this study, we used the data from the LOOK AHEAD and applied Random Forest to examine the factors determining SBP in persons with diabetes using the software R (version 4.0.3).

RESULTS

Our analysis (that included 4723 participants) showed that maximal exercise capacity, age, albumin to creatinine ratio, and serum creatinine were the key variables that determined systolic blood pressure.

CONCLUSIONS

Maximum exercise capacity is an important predictor of systolic blood pressure in patients with type 2 diabetes.

Prognostic importance of long-term SBP variability in high-risk hypertension

OBJECTIVE

In addition to high blood pressure variability (BPV), low BPV was associated with adverse cardiovascular prognosis in selected high-risk patients. We explored this issue in the Systolic Blood Pressure Intervention Trial (SPRINT) using a nonlinear approach with BPV as a continuous variable.

METHODS

Long-term systolic BPV (SBPV) (coefficient of variation, CoV %) was calculated on quarterly visits until a fatal/nonfatal cardiovascular event or all-cause mortality, excluding titration period and patients with missing visits. We used Cox proportional hazard models with penalized smoothing splines to shape the risk of outcomes against the continuum of SBPV (independent variable). Adjusted hazard ratios (aHR, 95% CI) were calculated using the reference range derived from the nonlinear model. Sensitivity analysis based on propensity score matching (PSM) was performed.

RESULTS

The association of SBPV with fatal/nonfatal cardiovascular events was J-shaped, whereas that with all-cause mortality was linear. After multivariate adjustment, however, the only significant associations remained that of low SBPV (CoV <5%) with cardiovascular events (hazard ratio 1.85, 95% CI 1.24-2.75, P = 0.003), and of high SBPV (CoV >10%) with the composite of cardiovascular events and all-cause mortality (hazard ratio 1.35, 95% CI 1.02-1.80; P = 0.037). Low SBPV was associated with ischemic heart disease (hazard ratio 2.76, 95% CI 1.55-4.91; P < 0.001). There was a significant U-shaped association of SBPV with cardiovascular events in the PSM cohort.

CONCLUSION

Nonlinear modeling indicates that low and high long-term SBPV have prognostic relevance in high-risk hypertensive individuals from SPRINT. Randomized trials are needed to test these findings and their potential therapeutic implications.

Cardiovascular outcomes and achieved blood pressure in patients with and without diabetes at high cardiovascular risk

AIMS

Studies have shown a non-linear relationship between systolic blood pressure (SBP) and diastolic blood pressure (DBP) and outcomes, with increased risk observed at both low and high blood pressure (BP) levels. We hypothesized that the BP-risk association is different in individuals with and without diabetes at high cardiovascular risk.

METHODS AND RESULTS

We identified patients with (N = 11 487) or without diabetes (N = 19 450), from 30 937 patients, from 133 centres in 44 countries with a median follow-up of 56 months in the ONTARGET/TRANSCEND studies. Patients had a prior history of stroke, myocardial infarction (MI), peripheral artery disease, or were high-risk diabetics. Patients in ONTARGET had been randomized to ramipril 10 mg daily, telmisartan 80 mg daily, or the combination of both. Patients in TRANSCEND were ACE intolerant and randomized to telmisartan 80 mg daily or matching placebo. We analysed the association of mean achieved in-trial SBP and DBP with the composite outcome of cardiovascular death, MI, stroke and hospitalization for congestive heart failure (CHF), the components of the composite, and all-cause death. Data were analysed by Cox regression and restricted cubic splines, adjusting for risk markers including treatment allocation and accompanying cardiovascular treatments. In patients with diabetes, event rates were higher across the whole spectrum of SBP and DBP compared with those without diabetes (P < 0.0001 for the primary composite outcome, P < 0.01 for all other endpoints). Mean achieved in-trial SBP ≥160 mmHg was associated with increased risk for the primary outcome [diabetes/no diabetes: adjusted hazard ratio (HR) 2.31 (1.93-2.76)/1.66 (1.36-2.02) compared with non-diabetics with SBP 120 to <140 mmHg], with similar findings for all other endpoints in patients with diabetes, and for MI and stroke in patients without diabetes. In-trial SBP <120 mmHg was associated with increased risk for the combined outcome in patients with diabetes [HR 1.53 (1.27-1.85)], and for cardiovascular death and all-cause death in all patients. In-trial DBP ≥90 mmHg was associated with increased risk for the primary outcome [diabetes/no diabetes: HR 2.32 (1.91-2.82)/1.61 (1.35-1.93) compared with non-diabetics with DBP 70 to <80 mmHg], with similar findings for all other endpoints, but not for CHF hospitalizations in patients without diabetes. In-trial DBP <70 mmHg was associated with increased risk for the combined outcome in all patients [diabetes/no diabetes: HR 1.77 (1.51-2.06)/1.30 (1.16-1.46)], and also for all other endpoints except stroke.

CONCLUSION

High on treatment BP levels (≥160 or ≥90 mmHg) are associated with increased risk of cardiovascular outcomes and death. Also low levels (<120 or <70 mmHg) are associated with increased cardiovascular outcomes (except stroke) and death. Patients with diabetes have consistently higher risks over the whole BP range, indicating that achieving optimal BP goals is most impactful in this group. These data favour guidelines taking lower BP boundaries into consideration, in particular in diabetes.

CLINICAL TRIAL REGISTRATION

http://clinicaltrials.gov.Unique identifier: NCT00153101.

Effects of Intensive Systolic Blood Pressure Lowering on Cardiovascular Events and Mortality in Patients With Type 2 Diabetes Mellitus on Standard Glycemic Control and in Those Without Diabetes Mellitus: Reconciling Results From ACCORD BP and SPRINT

Background

Intensive systolic blood pressure (SBP) lowering significantly reduced cardiovascular disease (CVD) events in SPRINT (Systolic Blood Pressure Intervention Trial) but not in ACCORD BP (Action to Control Cardiovascular Risk in Diabetes Blood Pressure).

Methods and Results

SPRINT tested the effects of intensive (<120 mm Hg) versus standard (<140 mm Hg) SBP goals on CVD events and all‐cause mortality. Using 2×2 factorial design, ACCORD BP tested the same SBP intervention in addition to an intensive versus standard glycemia intervention. We compared the effects of intensive SBP lowering on the composite CVD end point and all‐cause mortality in SPRINT with its effects within each of the glycemia arms in ACCORD BP. Intensive SBP lowering decreased the hazard of the composite CVD end point similarly in SPRINT (hazard ratio: 0.75; 95% confidence interval, 0.64–0.89) and in the ACCORD BP standard glycemia arm (hazard ratio: 0.77; 95% confidence interval, 0.63–0.95; interaction P=0.87). However, the effect of intensive SBP lowering on the composite CVD end point in the ACCORD BP intensive glycemia arm (hazard ratio: 1.04; 95% confidence interval, 0.83–1.29) was significantly different from SPRINT (interaction P=0.023). Patterns were similar for all‐cause mortality.

Conclusions

The effects of intensive SBP control on CVD events and all‐cause mortality were similar in patients without diabetes mellitus and in those with diabetes mellitus on standard glycemic control. An interaction between intensive SBP lowering and intensive glycemic control may have masked beneficial effects of intensive SBP lowering in ACCORD BP.

Clinical Trial Registration

URL: http://www.clinicaltrials.gov. Unique identifiers: NCT01206062, NCT00000620.

Personalized prediction of adverse heart and kidney events using baseline and longitudinal data from SPRINT and ACCORD

BACKGROUND

The 2017 guidelines of the American College of Cardiology and the American Heart Association propose substantial changes to hypertension management. The guidelines lower the blood pressure threshold defining hypertension and promote more aggressive treatments. Thus, more individuals are now classified as hypertensive and as a result, medication usage may become more extensive. An inevitable byproduct of greater medication use is higher incidence of adverse effects. Here, we examined these issues by developing models that predict both cardiovascular events and other adverse events based on the treatment chosen and other patient's data.

METHODS AND RESULTS

We used data from the SPRINT trial to produce patient-specific predictions of the risks for adverse cardiovascular or kidney outcomes. Unlike prior models, we used both the baseline characteristics collected upon recruitment and the longitudinal data during the follow-up. Importantly, our cardiovascular predictor outperformed extant models on SPRINT participants, achieving AUC = 0.765, and was validated with good performance in an independent cohort of the ACCORD trial.

CONCLUSIONS

Our study illustrates the importance of including longitudinal data for assessing personalized risk and provides means for recommending personalized treatment decisions.

KDOQI US Commentary on the 2017 ACC/AHA Hypertension Guideline

Hypertension is a modifiable risk factor for cardiovascular morbidity and mortality and reduction of elevated blood pressure (BP) remains an important intervention for slowing kidney disease progression. Over the past decade, the most appropriate BP target for initiation and titration of BP-lowering medications has been an area of intense research and debate within the clinical community. In 2017, the American College of Cardiology and the American Heart Association (ACC/AHA) in conjunction with several other professional societies released new hypertension guidelines based on data from a systematic review of clinical trials and observational data. While many of the recommendations in the ACC/AHA hypertension guideline are relevant to nephrology practice, BP targets and management strategies for patients receiving dialysis are not discussed. This Kidney Disease Outcomes Quality Initiative (KDOQI) commentary focuses largely on recommendations from the ACC/AHA hypertension guidelines that are pertinent to individuals at risk of chronic kidney disease or with non-dialysis-dependent chronic kidney disease. This KDOQI commentary also includes a brief discussion of the consensus statement regarding hypertension diagnosis and management for adults receiving maintenance dialysis published by the European Renal and Cardiovascular Medicine Working Group of the European Renal Association-European Dialysis and Transplant Association (ERA-EDTA) and the Hypertension and the Kidney working group of the European Society of Hypertension. Overall, we support the vast majority of the ACC/AHA recommendations and highlight select areas in which best diagnosis and treatment options remain controversial.

How Low Do We Go (in the Post-SPRINT Era)?

Hypertension is frequently both a cause and complication of CKD. The optimal blood pressure target in CKD has been of hot debate over the decades with little data to inform the goals. Here, we review the data from the Systolic Blood Pressure Intervention Trial (SPRINT), Modification of Diet in Renal Disease (MDRD), African American Study of Kidney Disease and Hypertension (AASK), Ramipril Efficacy in Nephropathy-2 (REIN-2), and Action to Control Cardiovascular Risk in Diabetes (ACCORD) trials to use the available evidence to better inform what blood pressure goal should be recommended in patients with CKD and to answer the question "How low should we go?".

A Comparison of the 2017 American College of Cardiology/American Heart Association Blood Pressure Guideline and the 2017 American Diabetes Association Diabetes and Hypertension Position Statement for U.S. Adults With Diabetes

OBJECTIVE

To determine the concordance in the prevalence of hypertension and pharmacological antihypertensive treatment recommendations for U.S. adults with diabetes using definitions from the 2017 American College of Cardiology/American Heart Association (ACC/AHA) blood pressure (BP) guideline and the 2017 American Diabetes Association (ADA) diabetes and hypertension position statement.

RESEARCH DESIGN AND METHODS

We analyzed data for U.S. adults with diabetes in the U.S. National Health and Nutrition Examination Survey (NHANES), 2011-2016 (n = 2,266). Diabetes was defined by treatment with glucose-lowering medication, glycosylated hemoglobin ≥6.5%, fasting serum glucose ≥126 mg/dL, or nonfasting serum glucose ≥200 mg/dL. BP was measured three times and antihypertensive medication use was self-reported.

RESULTS

The prevalence (95% CI) of hypertension among U.S. adults with diabetes was 77.1% (73.9, 80.0) and 66.3% (63.4, 69.1) according to the ACC/AHA and ADA definitions, respectively. Also, 22.9% (20.0, 26.1) did not have hypertension according to either definition, and the concordance in hypertension status was 89.2% (87.2, 91.0). Among U.S. adults with diabetes not taking antihypertensive medication, 52.8% (47.7, 57.8) were not recommended to initiate antihypertensive medication by either the ACC/AHA or the ADA document and 22.4% (19.2, 25.9) were recommended to initiate it by both documents (overall concordance 75.2% [70.4, 79.4]). Among those taking antihypertensive medication, 45.3% (41.3, 49.4) and 50.4% (46.5, 54.2) had BP above the goal in neither and both documents, respectively (overall concordance 95.7% [93.4, 97.2]).

CONCLUSIONS

A high percentage of U.S. adults with diabetes are provided identical antihypertensive treatment recommendations by the ACC/AHA BP guideline and the ADA diabetes and hypertension position statement.

Effect of intensive blood pressure control in patients with type 2 diabetes mellitus over 9 years of follow-up: A subgroup analysis of high-risk ACCORDION trial participants

Although guidelines recommend strict blood pressure (BP) control in patients with type 2 diabetes mellitus (T2DM) and elevated cardiovascular risk, the long-term effects of this approach are unknown. We investigated the effect of intensive BP control on clinical outcomes in patients with T2DM over 9 years of follow-up. We included Action to Control Cardiovascular Risk in Diabetes - Blood Pressure participants in the standard glucose control arm who had established cardiovascular disease, chronic kidney disease, were ≥75 years of age or who had a 10-year coronary heart risk ≥15%. Participants were randomized to either intensive (systolic BP < 120 mm Hg) or standard (systolic BP < 140 mm Hg) BP control for an average of 5 years. Observational follow-up occurred for an average of 4 years thereafter. After an average total follow-up of 9 years, intensive BP control reduced the composite of cardiovascular death, nonfatal myocardial infarction and nonfatal stroke by 25% (hazard ratio, 0.75; 95% confidence interval, 0.60-0.95; P = .02). The overall benefit was driven by a reduction in nonfatal myocardial infarction (P = .01). In this post-hoc analysis, the benefits of a fixed-duration intensive BP control intervention in patients with T2DM persisted throughout 9 years of follow-up.