key: cord-0976012-wfgr65di authors: Wei, Xiaolin; Zhang, Zhitong; Chong, Marc K. C.; Hicks, Joseph P.; Gong, Weiwei; Zou, Guanyang; Zhong, Jieming; Walley, John D.; Upshur, Ross E. G.; Yu, Min title: Evaluation of a package of risk-based pharmaceutical and lifestyle interventions in patients with hypertension and/or diabetes in rural China: A pragmatic cluster randomised controlled trial date: 2021-07-01 journal: PLoS Med DOI: 10.1371/journal.pmed.1003694 sha: d8ee48c8e349b5682053ea65bec54cb73a7a30b6 doc_id: 976012 cord_uid: wfgr65di BACKGROUND: Primary prevention of cardiovascular disease (CVD) requires adequate control of hypertension and diabetes. We designed and implemented pharmaceutical and healthy lifestyle interventions for patients with diabetes and/or hypertension in rural primary care, and assessed their effectiveness at reducing severe CVD events. METHODS AND FINDINGS: We used a pragmatic, parallel group, 2-arm, controlled, superiority, cluster trial design. We randomised 67 township hospitals in Zhejiang Province, China, to intervention (34) or control (33). A total of 31,326 participants were recruited, with 15,380 in the intervention arm and 15,946 in the control arm. Participants had no known CVD and were either patients with hypertension and a 10-year CVD risk of 20% or higher, or patients with type 2 diabetes regardless of their CVD risk. The intervention included prescription of a standardised package of medicines, individual advice on lifestyle change, and adherence support. Control was usual hypertension and diabetes care. In both arms, as usual in China, most outpatient drug costs were out of pocket. The primary outcome was severe CVD events, including coronary heart disease and stroke, during 36 months of follow-up, as recorded by the CVD surveillance system. The study was implemented between December 2013 and May 2017. A total of 13,385 (87%) and 14,745 (92%) participated in the intervention and control arms, respectively. Their mean age was 64 years, 51% were women, and 90% were farmers. Of all participants, 64% were diagnosed with hypertension with or without diabetes, and 36% were diagnosed with diabetes only. All township hospitals and participants completed the 36-month follow-up. At 36 months, there were 762 and 874 severe CVD events in the intervention and control arms, respectively, yielding a non-significant effect on CVD incidence rate (1.92 and 2.01 per 100 person-years, respectively; crude incidence rate ratio = 0.90 [95% CI: 0.74, 1.08; P = 0.259]). We observed significant, but small, differences in the change from baseline to follow-up for systolic blood pressure (−1.44 mm Hg [95% CI: −2.26, −0.62; P < 0.001]) and diastolic blood pressure (−1.29 mm Hg [95% CI: −1.77, −0.80; P < 0.001]) in the intervention arm compared to the control arm. Self-reported adherence to recommended medicines was significantly higher in the intervention arm compared with the control arm at 36 months. No safety concerns were identified. Main study limitations include all participants being informed about their high CVD risk at baseline, non-blinding of participants, and the relatively short follow-up period available for judging potential changes in rates of CVD events. CONCLUSIONS: The comprehensive package of pharmaceutical and healthy lifestyle interventions did not reduce severe CVD events over 36 months. Improving health system factors such as universal coverage for the cost of essential medicines is required for successful risk-based CVD prevention programmes. TRIAL REGISTRATION: ISRCTN registry ISRCTN58988083. We used a pragmatic, parallel group, 2-arm, controlled, superiority, cluster trial design. We randomised 67 township hospitals in Zhejiang Province, China, to intervention (34) or control (33) . A total of 31,326 participants were recruited, with 15,380 in the intervention arm and 15,946 in the control arm. Participants had no known CVD and were either patients with hypertension and a 10-year CVD risk of 20% or higher, or patients with type 2 diabetes regardless of their CVD risk. The intervention included prescription of a standardised package of medicines, individual advice on lifestyle change, and adherence support. Control was usual hypertension and diabetes care. In both arms, as usual in China, most outpatient drug costs were out of pocket. The primary outcome was severe CVD events, including coronary heart disease and stroke, during 36 months of follow-upAU : Ichangedafter36monthsofinterventiontoduri , as recorded by the CVD surveillance system. The study was implemented between December 2013 and May 2017. A total of 13,385 (87%) and 14,745 (92%) participated in the intervention and control arms, respectively. Their mean age was 64 years, 51% were women, and 90% were farmers. Of all participants, 64% were diagnosed with hypertension with or without diabetes, and 36% were diagnosed with diabetes only. All township hospitals and participants completed the 36-a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 month follow-up. At 36 months, there were 762 and 874 severe CVD events in the intervention and control arms, respectively, yielding a non-significant effect on CVD incidence rate (1.92 and 2.01 per 100 person-years, respectively; crude incidence rate ratio = 0.90 [95% CI: 0.74, 1.08; P = 0.259]). We observed significant, but small, differencesAU : HereandintheAuthorsum in the change from baseline to follow-up for systolic blood pressure (−1.44 mm Hg [95% CI: −2.26, −0.62; P < 0.001]) and diastolic blood pressure (−1.29 mm Hg [95% CI: −1.77, −0.80; P < 0.001]) in the intervention arm compared to the control arm. Self-reported adherence to recommended medicines was significantly higher in the intervention arm compared with the control arm at 36 months. No safety concerns were identified. Main study limitations include all participants being informed about their high CVD risk at baseline, non-blinding of participants, and the relatively short follow-up period available for judging potential changes in rates of CVD events. The comprehensive package of pharmaceutical and healthy lifestyle interventions did not reduce severe CVD events over 36 months. Improving health system factors such as universal coverage for the cost of essential medicines is required for successful risk-based CVD prevention programmes. ISRCTN registry ISRCTN58988083. Why was this study done? • Pharmaceutical interventions AU : Pleasecheckthattheeditstothesentence}Pharmaceuticalintervent have been shown to reduce cardiovascular disease (CVD) events, but the evidence has limited policy implications as medicines in trials have been typically been provided free of charge. • Four previous trials examining CVD-risk-based management treatments in highincome countries were focused on healthy lifestyle interventions and did not identify any clear effects on CVD events. Two previous trials in low-and middle-income countries implemented both pharmaceutical and lifestyle interventions, but they were not designed to examine impacts on CVD events. • We conducted a cluster randomised controlled trial in rural China among 31,326 patients with (1) hypertension who had a CVD risk of 20% or higher or (2) diabetes, and followed participants up for 36 months. Interventions included prescription of a standardised package of medicines, individual advice on lifestyle change, and adherence support provided by a team led by family doctors. We compared this to usual existing care. • The study found that our comprehensive pharmaceutical and lifestyle interventions did not reduce severe CVD events. • We observed significant, but small, differences in the change from baseline to follow-up of systolic blood pressure (−1.44 mm Hg) and diastolic blood pressure (−1.29 mm Hg) in the intervention arm compared to the control arm. Self-reported adherence to recommended medicines was significantly higher in the intervention arm compared with the control arm at 36 months. Cardiovascular disease (CVD) is the world's leading cause of mortality, representing 31% of all global deaths in 2016 [1] . The United Nations Sustainable Development Goal (SDG) 3 calls for a one-third reduction in CVD deaths by 2030. However, this goal will not be achieved without progressive prevention activities that include adequate control of blood pressure, glucose, and lipid levels, as well as smoking cessation, and reducing salt, alcohol, sugar, and trans fat intake [2] . Primary care provides the best setting for these solutions because hypertension and diabetes, the 2 most common risk factors for CVD, are managed by family doctors, who can provide continuous, coordinated, and comprehensive care. However, hypertension and diabetes are often treated and managed separately, as reflected in the frequent existence of separate guidelines and programmes. In addition, the 2 diseases are both typically inadequately controlled in low-and middle-income countries (LMICs) compared to high-income countries (HICs). Recent multi-country studies reported that only 10% of patients with hypertension had controlled hypertension in LMICs, compared to 50% in HICs [3, 4] . To achieve SDG 3, a comprehensive treatment approach consisting of both pharmaceutical and lifestyle strategies is urgently needed in LMICs. The ability of recommended antihypertensive medicines, aspirin, and statins to prevent CVD events has been well documented in trials [1] . Recommended AU : Pleasecheckthattheeditstothesentence}Recommendedpharmaceuticaltherapy::: pharmaceutical therapy, often in the form of a combined polypill, can reduce CVD risk by 50%-60% in the long term [5] , but the effects are often more modest, as shown in the PolyIran trial [6] (34% reduction in hazard ratio for major CVD events [95% CI: 20%, 45%] over 60 months). Meta-analyses of healthy lifestyle interventions, including those targeting weight loss, alcohol reduction, and smoking cessation, have shown moderate improvements in biomarkers of CVD risk factors, such as blood pressure, body mass index (BMI), and total serum cholesterol [7] . Thus, combined pharmaceutical and lifestyle interventions could be more beneficial to patients at high risk of CVD. Several trials have reported positive effects of combined pharmaceutical and lifestyle interventions, including reduced patient CVD risks, blood pressure, and lipid profiles, but none demonstrated reductions in CVD events [8] [9] [10] [11] . The strategy of risk-based pharmaceutical and lifestyle interventions has been suggested in national clinical guidelines; however, the policy question of whether this strategy has any impact to achieve SDG 3 remains unanswered due to insufficient evidence [12] . Our previous study identified health system barriers to sustainable CVD risk management, such as insufficient knowledge amongst family doctors, use of ineffective medicines, and lack of treatment support [13] . China's recent health reforms provided an opportunity for better control of non-communicable diseases as they strengthened primary care capacities and improved hypertension and diabetes management [14] . Based on this, we designed and implemented a comprehensive package of pharmaceutical and healthy lifestyle interventions embedded into routine primary care practice for primary prevention of CVD among patients with diabetes and hypertension. We have previously reported feasibility results from our pilot study [15] , the trial protocol [16] , and the effect on pharmaceutical management [17] . Here we report the evaluation of relevant interventions against the trial's primary and secondary outcomes during 36 months of follow-up. Our study is a pragmatic, parallel group, 2-arm, cluster randomised, controlled, superiority trial. It aims to assess the effectiveness of a package of comprehensive pharmaceutical and healthy lifestyle interventions targeting primary prevention of CVD among patients with high CVD risk in rural China. We conducted the research in township hospitals, which are rural primary care facilities. Township hospitals and their catchment populations are the clusters of the trial. Details of the trial design and analysis plan have been published elsewhere [16] . In brief, we selected all township hospitals, except 1 that served as the pilot site, in 3 counties in Zhejiang Province. Cluster eligibility criteria included having electronic medical records covering the last 2 years for residents in these townships. This study was approved by the Ethics Review Board of the University of Leeds, UK (reference HSLTLM/12/010) and the Ethics Committee of Zhejiang Provincial Centre for Disease Control and Prevention, China (reference 18/06/2012). Written informed consent was obtained from all participating individuals. We used the CONSORT cluster trials checklist (S1 CONSORT Checklist) to help in reporting the study [18] . We recruited participants from township hospitals based on collected health records from all residents in their catchment areas. Patients were considered eligible if they had the following characteristics: (1) aged 50 to 74 years; (2) permanent residents in the township; (3) having either hypertension with a 10-year CVD risk � 20% calculated using the Asian equation [19] , or type 2 diabetes (with or without hypertension); (4) free from any diagnosed mental diseases or physical disabilities, any history of severe CVD events, or other diseases as defined in the protocol [16] ; (5) not currently hospitalised or living in a long-term care facility; (6) no serious adverse effects to the recommended medicines, and (7) having a diastolic blood pressure � 60 mm Hg. After cluster randomisation, all potentially eligible participants were invited to visit the township hospital, where family doctors explained the study, recruited patients, and obtained their consent. In December 2013, we randomised a total of 67 township hospitals, without stratification, to the intervention and control arms in a 34:33 ratio, via an independent statistician using computer-generated random numbers. All eligible consenting participants received either the intervention or control treatment based on the treatment allocated to their township hospital. Due to the nature of the interventions, health providers and patients could not be blinded, but the analysis of the trial was blinded. In both arms patients had recurring booked consultations with their doctors at least quarterly. All recommended medicines were available in township hospitals or pharmacies in both arms. Patients purchased their medications from township hospitals or pharmacies using prescriptions from their family doctors, and then reported their adherence to the prescriptions in the next consultation. In the control arm, usual care of hypertension and diabetes continued per existing practice. No additional recommendations for pharmaceutical or healthy lifestyle modifications were provided. The 2010 Chinese national hypertension [20] and diabetes [21] guidelines were available, but they were often not referred to in primary care consultations because the guidelines contain large amounts of information for tertiary care [22] . Family doctors treated hypertension and diabetes according to their own discretion and group practice. No specific training regarding CVD risk management was provided. Healthy lifestyle change education was given based on controlling either hypertension or diabetes, but compared to the intervention arm this was not using a holistic approach to CVD risk reduction. Treatment adherence support was not provided. In the intervention arm, patients received prescriptions and healthy lifestyle education for treatment of their existing medical conditions and CVD prevention. All patients with hypertension, or hypertension and diabetes, were prescribed a standard combination [23] of (1) 2 antihypertensives (2 different kinds selected from thiazide diuretics, calcium channel blockers, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, and beta-blockers) [24] , (2) a statin, and (3) a low dose aspirin, unless contraindicated. Patients with only diabetes received the same pharmaceutical package but with only 1 antihypertensive, plus their anti-diabetic medicines, if any. Patients already on other older antihypertensive medicines were advised to switch to the recommended medicines in standardised packages. Patients who had a history of, or showed signs of, gastrointestinal bleeding were not prescribed aspirin. During consultation, family doctors provided individualised health education focusing on smoking cessation, salt reduction, and reduction of alcohol consumption depending on the patient's situation. A family treatment supporter was selected at the patient's home to support him/her taking medicines and adhering to lifestyle changes. At the facility level, we provided annual training to all family doctors using our CVD risk management guidelines, which covered (1) prescribing for primary prevention of CVD, (2) personalised advice for healthy lifestyle changes, and (3) advice to improve adherence [16] . Township hospitals held monthly meetings to discuss doctors' experience of using the guidelines and chronic disease management (Table 1 ). On enrolment, family doctors measured participants' blood pressure using a standardised mercury sphygmomanometer after 5 minutes of seated rest. They also measured their body weight and height, and asked for details on their medication history. These measures were repeated on a quarterly basis when participants were followed up in township hospitals. All information was recorded in an internet-based public health management information system. The trial stopped 36 months (May 2017) after enrolment of the last participant based on our protocol. All acute severe CVD events were collected through Zhejiang CDC's surveillance system [25] based on the adapted WHO MONICA definitions [26] compatible with ICD-10 codes. These included coronary heart disease (CHD) and stroke, where CHD includes acute myocardial infarction, ischaemic cardiac arrest, and unclassifiable CHD deathsAU : unclassifiabledeathsðfroma , and stroke includes haemorrhagic stroke (e.g., subarachnoid and intracerebral), ischaemic stroke/infarction (e.g., thrombosis and embolism), and unclassifiable stroke [27] . We did not include minor CVD events such as angina and transient ischaemic attacks because they are often unrecorded. To be eligible as endpoints, all CVD events had to be verified from hospital records, and were reviewed and verified by centres for disease control at the county and provincial levels. All reported CVD events were then also been verified by primary care facilities through home visits. The process was supervised under the Validation Committee in the Zhejiang CDC. We previously reported on the data collection, reporting, and validation of the Zhejiang CDC's surveillance system [25] . In addition, family doctors asked participants if they had experienced Table 1 . Intervention strategies to reduce the risk of cardiovascular disease (CVD). Family doctors Operational guidelines: The guidelines contained a definition of high risk of CVD, workflow chart for patient management, how to diagnose hypertension and type 2 diabetes, suggested prescriptions and CVD risk management practices, lifestyle evaluation and recommendations, treatment support, and managing the side effects of medicines. In addition, we also incorporated elements of China's national hypertension and type 2 diabetes guidelines for township hospitals. Training workshops: Training workshops were provided on an annual basis by local senior doctors based on the operational guidelines, including hypertension/diabetes management, prevention, and communication skills, using lectures, health education videos, case discussions, role plays, and question and answer sessions. Monthly meetings in township hospitals for performance monitoring: A senior doctor led the review of the implementation of intervention strategies, and provided comments to related questions on CVD risk management and treatment of patients by doctors. Guidelines and training: National hypertension and type 2 diabetes guidelines were available but were not referred to in practice because the guidelines focus on tertiary care. No specific training/guidelines were given on reducing CVD risks based on a holistic approach. Training workshops: Routine annual training workshops on hypertension and type 2 diabetes management were provided, but there was no research-driven systematic training provided on CVD risk management. Monthly meetings: Routine monthly internal meetings were held, but there were no specific discussions dedicated to CVD risk management. Recommended medicines: A combination of 2 antihypertensives, a statin, and a low dose of aspirin were recommended to patients with hypertension. One antihypertensive, a statin, and a low dose of aspirin were recommended for patients with type 2 diabetes but not hypertension. Anti-diabetic medicine or insulin was given for diabetic patients if necessary. In their consultations, family doctors gave targeted health education messages to participants based on their health conditions, e.g., smoking cessation messages to smokers, and messages regarding the benefits of the combined medicines to those who did not adhere to prescriptions. Enhanced follow-up appointment: Participants attended follow-up consultations in township hospitals at least quarterly with their family doctors. Nurses communicated with participants monthly through phone calls/SMS messages. Treatment supporter: Family doctors or nurses helped to identify and train a treatment supporter (normally a family member) who reminded patients to take medicines, maintain a healthy lifestyle, and adhere to their follow-up consultations. Recommend medications and health education: Treatment and lifestyle changes were offered according to existing knowledge and at the individual clinician's discretion, but based on either hypertension or type 2 diabetes instead of a holistic approach for CVD prevention. In practice, most doctors prescribed only 1 antihypertensive for hypertensive patients, and no antihypertensive for diabetic patients. Very few patients were given a statin or aspirin. Anti-diabetic medicine or insulin was given for diabetic patients if necessary. Existing general health education messages were provided. Follow-up appointment: Patients with hypertension or type 2 diabetes were followed up at least quarterly by family doctors according to national guidelines. Treatment supporter: The option of a treatment supporter was not provided. Modern medicines were available in both the intervention and control arms, and their prices were the same in the 2 arms. https://doi.org/10.1371/journal.pmed.1003694.t001 any stroke or heart attack events that were diagnosed by hospitals at the 12th, 24th, and 36th month following randomisation; if any, the CVD events were reported to the Validation Committee and then verified and included in the records. Participants were classified as lost to follow-up if they could not be contacted after 3 attempts (via telephone, message, or home visit) and their CVD status was unknown by the 36th month. All outcomes were measured at the patient level. The primary outcome was the number of severe CVD events recorded by Zhejiang's CVD surveillance system [25] . We also collected the following secondary outcomes: (1) AU : Ichangedthenumberingofthislistsothatthenumbereditemsallh mortality due to severe CVD event, (2) number of CHD events, (3) mortality due to CHD event, (4) number of stroke events, (5) mortality due to stroke, (6) time to the first reported severe CVD event, (7) time to mortality due to CVD, (8) change in diastolic and systolic blood pressure (mm Hg) between baseline and 36 months, (9) adherence to the final quarterly follow-up consultation at 36 months, and (10) self-reported adherence at 36 months to 2 antihypertensive drugs for patients with hypertension or 1 antihypertensive drug for patients with only diabetes, aspirin, and statin. We also recorded any minor or serious adverse events. We added secondary outcomes 2-5 to identify any changes in CVD subgroups. In subsequent papers, we will report our other protocol-defined secondary outcomes from our panel data and cost-effectiveness analysis, which are based on different datasets as per our protocol [16] . We also previously reported feasibility measures in our process evaluation paper [28] . We calculated AU : Pleasecheckthattheeditstothesentence}Wecalculated:::}captureyourmeaning; partic the necessary sample size as 32 clusters per arm and 450 patients per cluster, to have 90% power to detect a 20% reduction in the severe CVD event incidence rate after 36 months, based on an assumed CVD event rate of 5% in the control arm and a coefficient of variation of 0.15, with hypothesis testing using a 2-sided P value with a 5% threshold for statistical significance. We calculated the exact follow-up time for each participant until the end of the trial or death. We estimated crude and covariate-adjusted ("adjusted") estimates of the treatment effect on all outcomes using generalised linear mixed models (GLMMs). To estimate the crude treatment effect (as an incidence rate ratio [IRR]) on our primary outcome, we used a GLMM with a fixed effect for treatment arm, a random intercept for cluster, and an "offset" variable for the log of person follow-up time, which used a log-link with Poisson errors. To estimate the adjusted treatment effect, we repeated the same model but included patient-level variables as per table footnotes. We also estimated crude and adjusted treatment effects for all our secondary outcomes using a variety of mixed-effects models, with the same random and fixed effects (unless stated: see relevant table footnotes) included. We employed GLMMs with Poisson errors and log-links (if no evidence of overdispersion) or GLMMs with negative binomial distributions and log-links (with treatment effects estimated as IRRs) to analyse event and mortality outcomes, Cox proportional hazard mixed-effects models (with treatment effects estimated as HRs) for time to event outcomes, GLMMs with normal errors and identity links (with treatment effects estimated as mean differences) for blood pressure outcomes, and GLMMs with binomial errors and logistic links (with treatment effects estimated as odds ratios [ORs]) for adherence and adverse event outcomes. The intra-cluster correlation coefficient for the primary outcome was calculated by dividing the between-cluster variance by the total variance. The variance terms were estimated by fitting the relevant (Poisson/negative binomial) unconditional mixed-effects model with a random intercept for cluster [29, 30] . All 95% confidence intervals and P values for GLMMs were based on either the Wald statistic (nonnormal errors) or the t statistic (normal errors). We analysed all data using SAS statistical software (PROC MIXED and PROC GLIMMIX functions). Where any patients were missing outcomes or covariates in adjusted analyses, we excluded these patients and performed complete case analyses. We also ran cluster-level analyses for all our outcomes, as reported in S1 Text. Our primary AU : Pleasecheckthattheeditstothesentence}Ourprimaryanalysis:::}captureyourmeanin analysis was of the intention-to-treat (ITT) population and included all participants within all clusters as originally recruited and randomised, including all participants who left the trial after their first consultation or moved out of their cluster catchment area, and those who were wrongly classified as eligible: They all remained in the province so their CVD status was still tracked by the surveillance system. We also defined a "modified ITT" analysis population, which included all clusters and patients as originally recruited and randomised but excluding any patients with missing outcome or covariate data. We also repeated our analyses for all outcomes reported here in a per-protocol analysis population using the same methods described above. This per-protocol analysis population included all clusters as originally recruited and randomised, and across both arms it included all eligible patients except those patients who left the trial within the first quarter or those who were wrongly recruited. In addition, in the intervention arm this population excluded patients who did not have at least 50% adherence to their prescriptions across all their follow-up consultations. This included taking recommended antihypertensives (specifically, patients with hypertension taking a standard combination of 2 antihypertensives, and patients with diabetes only taking 1 antihypertensive) and statins or aspirin. These per-protocol analyses were designed after the trial because we observed a low rate of adherence to prescriptions, beyond what we expected. We did a range of exploratory subgroup analyses of the primary outcome based on the prerandomisation baseline characteristics of diagnosis (hypertension or hypertension and diabetes versus diabetes only), patient income level (<10,000 RMB [