Phenomapping of subgroups in hypertensive patients using unsupervised data-driven cluster analysis: An exploratory study of the SPRINT trial

Abstract

Background Hypertensive patients are highly heterogeneous in cardiovascular prognosis and treatment responses. A better classification system with phenomapping of clinical features would be of greater value to identify patients at higher risk of developing cardiovascular outcomes and direct individual decision-making for antihypertensive treatment. Methods An unsupervised, data-driven cluster analysis was performed for all baseline variables related to cardiovascular outcomes and treatment responses in subjects from the Systolic Blood Pressure Intervention Trial (SPRINT), in order to identify distinct subgroups with maximal within-group similarities and between-group differences. Cox regression was used to calculate hazard ratios (HRs) with 95% confidence intervals (CIs) for cardiovascular outcomes and compare the effect of intensive antihypertensive treatment in different clusters. Results Four replicable clusters of patients were identified: cluster 1 (index hypertensives); cluster 2 (chronic kidney disease hypertensives); cluster 3 (obese hypertensives) and cluster 4 (extra risky hypertensives). In terms of prognosis, individuals in cluster 4 had the highest risk of developing primary outcomes. In terms of treatment responses, intensive antihypertensive treatment was shown to be beneficial only in cluster 4 (HR 0.73, 95% CI 0.55–0.98) and cluster 1 (HR 0.54, 95% CI 0.37–0.79) and was associated with an increased risk of severe adverse effects in cluster 2 (HR 1.18, 95% CI 1.05–1.32). Conclusion Using a data-driven approach, SPRINT subjects can be stratified into four phenotypically distinct subgroups with different profiles on cardiovascular prognoses and responses to intensive antihypertensive treatment. Of note, these results should be taken as hypothesis generating that warrant further validation in future prospective studies.

SPRINT trial phenotypes and cardiovascular risk: The ‘Marathon’ of data analysis is still on

“…In the present issue of the Journal of Preventive Cardiology, Yang et al 1 investigated, using a data driven approach of the SPRINT trial, the influence of phenotypically distinct subgroups with different profiles on cardiovascular prognosis and responses to intensive anti-hypertensive treatment. This paper is an interesting analysis of the guideline-changing SPRINT study [2][3][4][5] and provides novel insights into the results, referring almost to the totality of the SPRINT participants.…”

“…This paper is an interesting analysis of the guideline-changing SPRINT study [2][3][4][5] and provides novel insights into the results, referring almost to the totality of the SPRINT participants. 1 Addressing the clinically relevant reduction of overall cardiovascular risk in hypertension remains an everlasting need. [5][6][7] The reasons for the above are still to be fully determined but it is mainly the diversity of phenotypes of hypertensives involved in the trials, the complexity of the effect of different anti-hypertensive regimens, the type and intensity of follow-up, the degree of adherence/compliance to therapy and, of course, the methodology for end-point verification.…”

“…The authors divided the SPRINT population into Obese, CKD, Extra Risky hypertensives and Index hypertensives using appropriate statistical models. 1 In this sense an enhanced understanding of the trial findings can be provided.…”

“…Starting from the Obese population, there is the interesting finding that these hypertensives compared with the index cluster do not respond better to the intensive therapy. 1 This can be justified by the fact that in patients with similar risk profile, apart from the obesity-related augmentation of probability of events, the lower BP targets cannot show, at least during the follow-up of the SPRINT trial, any additional benefit. It is important to notice that no diabetics are included in the trial, and thus the results are not applicable to obese patients with hypertension and diabetes.…”

“…2 Defining obesity based strictly on body mass index is a limitation and other measures like waist circumference and waist to hip ratio can be useful 'tools' for determining cardiovascular prognosis in the SPRINT trial. 1,8 If one focuses on the body mass index values of the index cluster, as well as of the other two clusters (CKD and Extra Risky) the values are similar and in the ranges of overweight. 1 Moreover, the Obese cluster is characterized by high probability adverse events in the intensive therapy group such as syncope and this should be acknowledged and further examined.…”