The rationale and design of the STEP trial, which was a prospective, multicenter, randomized, controlled trial performed at 42 clinical centers throughout China, have been published previously [16, 17]. Briefly, the STEP trial compared the effects of an intensive SBP target of 110–130 mmHg versus a standard target of 130–150 mmHg on cardiovascular outcomes. The STEP trial included patients aged 60–80 years who had an SBP of 140–190 mmHg or were receiving antihypertensive medication. Participants with prior ischemic or hemorrhagic stroke were excluded. From 10th January 2017 to 31st December 2017, 8511 patients were screened, recruited, and randomly divided into an intensive treatment group of 4243 participants and a standard treatment group of 4268 participants. The STEP trial was approved by the Ethics Committee of Fuwai Hospital and collaborating centers. All participants provided written informed consent.

A central computer program accessed via a website was used to randomly divide all eligible participants into the intensive treatment group or the standard treatment group in a 1:1 ratio. The participants were stratified by clinical center. All participants were followed up at specific timepoints after randomization. Participants were screened monthly for the first 3 months, and then every 3 months until the end of the trial or the death of the participant.

We used the information of antihypertensive medications prescribed for participants at randomization to denote overall antihypertensive medication burden, since the prescription was almost the same during the whole follow-up visits for most participants. A high antihypertensive medication burden was defined as treating with three or more different classes of antihypertensive medications, because this number of medications has frequently been used to identify resistant hypertension.

Trained physicians evaluated all participants using a standardized questionnaire during face-to-face visits. The questionnaire collected baseline information, including demographic characteristics (e.g., age, sex, birth date, residence, weight, height), lifestyle behaviors (e.g., physical activity, drinking status, smoking status), medical history (e.g., diabetes mellitus, hyperlipidemia, CVD, other chronic diseases), and medication use (e.g., antihypertensive agents, statins, aspirins). At each visit, the physicians measured the office blood pressure and heart rate, and collected information on concomitant medication use, antihypertensive drug adherence (assessed using the Morisky Medication Adherence Scale-8), and trial outcomes. Patients were required to rest for at least 5 min in a seated position, then a well-trained physician measured the blood pressure three times with an interval of 1 min, and the average value was recorded as the final datum. This process was standardized and all of the office blood-pressure monitors (Omron Healthcare) were same at all participating centers and validated during all clinic visits.

All participants underwent laboratory blood testing at baseline and each year thereafter, and biological examinations were conducted to acquire the total cholesterol, triglyceride, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol, fasting plasma glucose, creatinine, and uric acid concentrations. Body mass index (BMI) was calculated as the weight (kilograms) divided by the squared height (meters). The Chronic Kidney Disease Epidemiology Collaboration formula was used to estimate the glomerular filtration rate. The Framingham risk score was calculated to evaluate the 10-year CVD risk of all participants [18].

The primary outcome was a composite of stroke (ischemic or hemorrhagic), acute decompensated heart failure, acute coronary syndrome (acute myocardial infarction and hospitalization for unstable angina), atrial fibrillation, coronary revascularization, or death from cardiovascular causes. The secondary outcomes were SBP level and all-cause death within each treatment group. The detailed definitions and ascertainment criteria for the study outcomes have been published [16].

Continuous variables are presented as mean ± standard deviation, while categorical variables are presented as n (%). Baseline characteristics were compared across baseline medication burden groups within the intensive and standard treatment groups using one-way analysis of variance for continuous variables and the chi-squared test for categorical variables.

The relationship between the antihypertensive medication burden and cardiovascular outcomes was analyzed in the total cohort including all participants, and in the intensive and standard treatment groups separately. The ordinary least squares regression model was applied to calculate the mean SBP and changes in SBP from baseline to 12, 24, and 36 months. Modified Poisson regression was used to calculate the risk ratios (RRs) for achieving the target SBP. The Fine-Gray model was used to calculate hazard ratios (HRs) for the primary outcome, and Cox proportional hazard regression was used to calculate HRs for all-cause mortality. Models were adjusted for potential confounders including age, sex, BMI, baseline SBP, estimated glomerular filtration rate, fasting plasma blood glucose concentration, LDL-C concentration, smoking frequency, drinking frequency, and physical activity frequency. In addition, we calculated each outcome by including the product term (treatment group × medication burden group) in regression models to detect interactions between the treatment group and medication burden group.

To determine whether the antihypertensive medication burden could modify the effect of intensive versus standard treatment, the Fine-Gray model and Cox proportional hazard regression were used to calculate HRs for the primary outcome and all-cause mortality, respectively, associated with intensive versus standard treatment within the high and low antihypertensive medication burden groups. Furthermore, to investigate whether the results were sensitive to our definition of medication burden, we repeated all analyses with the medication burden groups classified as less than two medications and two or more medications.

All analyses were performed using R version 4.1.2. A two-sided P < 0.05 was considered statistically significant.