Design and study population
The rationale and design has been described in detail elsewhere (12). In short, a cluster randomized controlled trial was conducted in Dutch general practices (10 intervention and 12 usual care). Eligible were people enlisted with the GP, 18 to 80 years of age, with an office BP >140/90 mmHg, while managed with three or more antihypertensive drugs from different therapeutic classes simultaneously for three or more months in an adequate dose based on the current Dutch guideline on cardiovascular risk management (13). Detailed inclusion and exclusion criteria are presented in table 1.
The study protocol was approved by the Research Ethics Committee of the University Medical Center Utrecht in the Netherlands. All patients provided written informed consent. This study was conducted between October 2018 and June 2021 partly during the full COVID-19 epidemic.
Randomisation and blinding
Randomisation was done at GP practice level, enabling uniform care of GPs working in one practice (the so-called cluster). The randomization was 1:1, stratified by region (Utrecht/Nijmegen) and practice size (1 norm practice, 2 norm practices 3 ≥ norm practices, 2350 patients are the norm for one “standard GP practice”), using a web-based randomization program (Julius Center, UMC Utrecht, Utrecht, the Netherlands)(14,15). Due to the nature of the intervention, blinding was not possible. A cluster randomized design was adopted to avoid contamination between the intervention and usual care group. Patients in the control group were not informed about the intervention, and were asked to participate in a study to monitor the treatment of patients with uncontrolled hypertension.(16)
Interventions
The stepwise treatment of patients in the intervention group (figure 1) consisted of 5 steps: (i) excluding a white coat effect using 24-hour blood pressure measurement, (ii) re-evaluation of lifestyle, (iii) re-evaluation of adherence, (iv) optimalisation of medication, and (v) referral to a hospital specialist to rule out secondary hypertension. A more detailed description of these five steps of the intervention group has been described elsewhere (12). When participants achieved a controlled BP level at a certain step, the next steps were not performed and only a 8 months end of study assessment was performed. In the mean time, patients received care as usual. Participants in the usual care group were in principle seen at least once a year for cardiovascular risk management consultation in the general practice and were managed according to the Dutch guideline for cardiovascular risk management (13).
For the usual care, follow-up information was obtained at the end of study visit (at 8 months) and/or through the linkage of the GP information in Electronic Health Care records with the study database. The latter holds for information on co-morbidity, medication prescription and office SBP and BMI measurements and laboratory values.
Outcomes
The primary outcome was the difference in the mean 24–hour systolic BP between the intervention and usual care group assessed 8 months after the enrolment. Secondary outcome measures after 8 months were: (1) The percentage of patients achieving a controlled BP, that is an office BP ≤140/90 mmHg or a mean 24-hour BP ≤130/80 mmHg, measured at the end of study (8 months); (2) The time window (time from baseline) to reach a controlled BP during the study, defined as twice an office BP of ≤140/90 mmHg, or a mean 24-hour BP of ≤130/80 mmHg; (3) The number and/or dosage prescribed of BP lowering drugs, measured during the study time; (4) Referral to the medical specialist during the study time; (5) Health care use during the study period; (6) The difference in health-related quality of life as measured with the EQ5D and EQ-VAS at baseline and after 8 months between both arms (7) Self-reported weight/body mass index.
Sample size
The sample size calculation for the primary endpoint was based on an estimated standard deviation (SD) of 16 mmHg around a group mean systolic blood pressure based on 24-hour systolic blood pressure measurements as described elsewhere (12). We needed 233 participants to detect a clinically relevant difference of 7.5 mmHg in the mean 24 hour systolic BP between both groups, with an intracluster correlation coefficient of 0.05, a 2-sided alpha of 5% and 90% power.
We estimated the prevalence of eligible patients to be 15 per GP practice. Anticipating a conservative 50% response for participation, and a drop-out rate of around 20%, we needed to enrol 39 GP practices to be sure to end up with 583 participants to be invited to participate (i.e., 233 * 2 * 1.25). Rounded up, we planned to include 40 general GP practices (20 in the intervention and 20 in the care as usual arm) and 240 patients (40 * 15 * 0.50 * 0.80).
Statistical analyses
The baseline characteristics were provided as percentages or means with corresponding standard deviation. No statistical testing for differences at baseline was conducted since all potential difference are considered to be due to chance since a central randomization procedure was applied. The primary outcome was the difference in 24-hour systolic BP between groups measured at 8 months follow-up. The end-of-study 24-hour systolic BP values were presented for each treatment arm using a modified intention to treat approach, i.e., taking those with end of study measurements available. The difference between treatment arms was assessed using a linear regression model and presented as mean difference with standard errors (SE). We did not collect baseline 24-hour blood pressure measurements in the usual care group, since these measurements were considered the first step in the intervention. Therefore we could not adjust the end of study measurements for the baseline level as recommend to do so (17). Yet, based on the assumption that baseline office systolic pressure is a major determinant to end of study 24-hour blood pressure, we ran an analysis for end of 24-hour blood pressure with baseline adjustment for office blood pressure.
For the secondary outcomes, we applied regression models with end of study value adjusted for the baseline value (for blood pressure and body weight). The difference in EQ-5D score parameters between intervention groups at 8 months was evaluated using a Spearman correlation coefficient. A similar approach was taken for health care use. SPSS 25.0 was used as for the above mentioned analyses. Statistical significance was based on a two-sided p value of 0.05.
As this is a cluster randomized trial we additionally performed analyses taking clustering by GP practice into account in a linear mixed model approach with a random intercept using SAS statistical package. These results did not differ in direction, magnitude or significance from the results presented.