Received: 26 December 2019  |  Revised: 12 February 2020  |  Accepted: 4 March 2020 DOI: 10.1111/jch.13873 O R I G I N A L P A P E R Longitudinal control of blood pressure among a cohort of Ghanaians with hypertension: A multicenter, hospital-based study Fred S. Sarfo MD, PhD, PhD1,2  | Linda Mobula MD, MPH3,4 | Jacob Plange-Rhule MD, PhD5 | Mulugeta Gebregziabher PhD6 | Daniel Ansong MD, MSc1,2 | Osei Sarfo-Kantanka MD2 | Lynda Arthur BSc7 | Jasper Sablah MSc7 | Edith Gavor BSc8 | Gilbert Burnham PhD3,4 | David Ofori-Adjei MD9 1Department of Medicine, Kwame Nkrumah University of Science & Technology, Kumasi, Ghana 2Department of Medicine, Komfo Anokye Teaching Hospital, Kumasi, Ghana 3Johns Hopkins University School of Medicine, Baltimore, MD, USA 4Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD, USA 5Ghana College of Physicians and Surgeons, Accra, Ghana 6Department of Biostatistics, Medical University of South Carolina, Charleston, SC, USA 7Ghana Access and Affordability Program, Accra, Ghana 8Ghana National Drugs Programme, Ministry of Health, Accra, Ghana 9Department of Medicine & Therapeutics, University of Ghana School of Medicine and Dentistry, Accra, Ghana Correspondence Fred S. Sarfo, Department of Medicine, Abstract Division of Neurology, Kwame Nkrumah There are limited data on factors associated with longitudinal control of blood pres- University of Science and Technology, Private Mail Bag, Kumasi, Ghana. sure (BP) among Ghanaians on antihypertensive treatment. We sought to evalu- Email: stephensarfo78@gmail.com ate associations between prospective BP control and 24 putative factors within Funding information socio-demographic, biological, and organizational domains. This is a cohort study Funding for this study was provided by involving 1867 (65%) adults with hypertension and 1006 (35%) with both hyperten- MSD, Novartis, Pfizer, Sanofi and the Bill and Melinda Gates Foundation (collectively, sion and diabetes mellitus at five public hospitals. Clinic BP was measured every the Funders) through the New Venture Fund 2 months for 18 months of follow-up. A multivariate logistic regression analysis was (NVF). The funders had no role in study design, data collection, data analysis or in fitted via generalized linear mixed models to identify factors associated with clinic study report writing. BP ≥ 140/90 mm Hg at each clinic visit during follow-up. Mean age of study partici- pants was 58.9 ± 16.6 years and 76.8% were females. Proportions with controlled BP increased from 46.3% at baseline to 59.8% at month 18, P < .0001. Eight factors with adjusted OR (95% CI) associated prospectively with uncontrolled BP were male gen- der: 1.37 (1.09-1.72), secondary education: 1.32 (1.00-1.74), non-adherence to anti- hypertensive treatment: 1.03 (1.00-1.06), fruit intake: 0.94 (0.89-1.00), duration of hypertension diagnosis: 1.01 (1.00-1.02), hypertension with diabetes mellitus: 2.05 J Clin Hypertens. 2020;22:949–958. wileyonlinelibrary.com/journal/jch © 2020 Wiley Periodicals, Inc.  |  949 950  |     SARFO et Al. (1.72-2.46), number of antihypertensive medications: 1.63 (1.49-1.79), and estimated glomerular filtration rate (mL/min rise): 0.82 (0.76-0.89). Interventions aimed at ad- dressing modifiable factors associated with poorly controlled BP would be critical in prevention of cardiovascular diseases among Ghanaians. 1  | INTRODUC TION putative factors within socio-demographic, biological, and orga- nizational domains which might contribute to poor BP control in Hypertension is a leading risk factor responsible for significant the context of a large prospective cohort of Ghanaians with hyper- morbidity and mortality from cardiovascular diseases (CVDs).1,2 tension recruited into the Ghana Access and Affordability Program Although effective treatments are available for the management of (GAAP). The GAAP is a public-private partnership with an overar- hypertension, blood pressure (BP) control is sub-optimal worldwide ching objective to improve the management of hypertension and and the greatest burden of uncontrolled BP is reported in low- and type II diabetes through improved access to medicines and systems middle-income countries (LMICs) where CVD rates are rapidly ris- strengthening activities. As part of this initiative, a prospective co- ing.3-5 A multi-national, community-based study has shown that hort of participants with hypertension was enrolled and followed 46.5% of participants were aware they had hypertension and BP up for 18 months to assess factors associated with poor control of control was 32.5% among those being treated. The need for im- hypertension in Ghanaian public health institutions. provement in hypertension control is particularly urgent in LMICs with two-thirds of global burden of hypertension.6 The principal factors associated with BP control among those 2  | METHODS receiving treatment include poor adherence, clinical inertia, and or- ganizational failure.7-9 The influence of the aforementioned factors 2.1 | Study design and participants on the pervasively high prevalence of uncontrolled hypertension among individuals seeking health care in resource-limited settings This is a prospective cohort study involving adults with hypertension, have seldom been characterized in prospective studies. It has been hypertension with diabetes mellitus and diabetes mellitus at public suggested that single BP measurements may underestimate associ- hospitals in Ghana. Ethical approval was obtained from the Ghana ations between BP and cardiovascular events due to the inherent Health Services Ethical Review Committee (GHS-ERC: 12/07/14). biological variability of BP. Serial measures of BPs over time have This research study was performed in accordance with the relevant been adversely linked with CVD outcomes.10-12 A plethora of studies regulations. The study protocol is published elsewhere.34 Briefly, the conducted in sub-Saharan Africa using cross-sectional designs have study was conducted at five hospitals in Ghana with hypertension consistently reported poor control rates of hypertension with an and diabetes specialty and general clinics. The five study sites in- average of <45% of adult study samples with blood pressure on tar- cluded the Agogo Presbyterian Hospital, (APH), Atua Government get.13-31 There is, however, a paucity of studies on the determinants Hospital, (AGH), Komfo Anokye Teaching Hospital, (KATH), Kings of poor BP control from prospective cohort studies in sub-Saharan Medical center, (KMC), and the Tamale Teaching Hospital, (TTH). Africa (SSA). The WHO has estimated that 46% of adults aged 25 years and older in SSA have hypertension with a projected rise in com- 2.2 | Recruitment of study participants ing decades.32,33 With awareness rates of <30%, treatment rates of <20%, and control rates of <10%, there is substantial room for Participants were eligible if they were 18 years or older with known improvement in control of hypertension in Africa.33 Characterizing diagnosis of hypertension and/or type II diabetes presenting for the factors associated with poor BP control in SSA is an import- routine care at either a general polyclinic (AGH, KMC, TTH) or a ant first step in helping design evidence-based interventions to- dedicated diabetes or hypertension clinic, (KATH, APH). Participants ward mitigating the devastating consequences of uncontrolled BP. were excluded if they had hypertensive urgency or emergency or had Previous studies assessing predictors of BP control in SSA have glycemic complications at initial contact for enrollment. Informed been cross-sectional and mostly single-center studies.32,33 These consent was obtained from all consecutively enrolled participants. previous studies have not comprehensively evaluated the contri- butions of organizational factors such as level of healthcare insti- tution, care provider-related factors such as therapeutic inertia 2.3 | Evaluation of study participants and patient-level factors such as socio-demographic and patho-bi- ological variables all of which may differentially contribute to BP Trained Research Assistants interviewed study participants control. We sought to characterize the associations between 24 and collected demographic information such as age, gender, SARFO et Al.      |  951 educational attainment, employment status, number of depend- Each clinic BP recorded for study participants during follow-up was ents on monthly income and health expenditures. Information on scored 1 if ≥140/90 mm Hg or scored 0 if <140/90 mm Hg. Secondary lifestyle behaviors such as alcohol use, cigarette smoking, level outcomes were systolic and diastolic BPs as continuous variables. of physical activities, frequency and daily quantities of fruits and vegetable consumption and table added salt was also recorded. The duration of hypertension or diabetes diagnosis was noted and 2.8 | Statistical analysis compliance with hypertension treatment was assessed using the 14-item version of Hill-Bone compliance to high blood pressure We investigated 24 potential factors for associations with poor BP therapy scale.35 Stroke was self-reported if participant had ever control based on literature search, our understanding of the epi- experienced sudden onset of weakness or sensory loss on one demiology of BP control and empirical evidence from our data. For side of the body, sudden loss of vision, or sudden loss of speech. baseline characteristics, means were compared using the Student's Heart failure was self-reported if participant had ever experienced t test and proportions were compared using the chi-squared test or shortness of breath on exertion, on lying down and swelling of Fisher's exact test. A multivariate logistic regression analysis was both feet. BP measurements were performed following a stand- fitted via generalized linear mixed models (GLMM) with a random ardized operating procedure implemented across study sites. intercept (to account for the repeated nature of the data and clus- Anthropometric assessments performed by study nurses include tering by hospitals) to identify factors independently associated measurement of weight and height for body mass index (BMI) deri- with the risk of clinic BP reading of ≥140/90 mm Hg. Independent vation as well as waist circumference. variables evaluated included the following socio-demographic fac- tors: age, gender, educational attainment, marital status, monthly income, employment status, dependents on household income 2.4 | Laboratory measurements and location of residence; lifestyle/behavioral factors: cigarette smoking, current alcohol use, physical activity, table added salt, An International Organization for Standardization (ISO)-certified fruit and vegetable intake, antihypertensive therapy adherence; laboratory was contracted to analyze serum creatinine, lipid profile, patho-biologic factors: duration of hypertension diagnosis, co-mor- and hemoglobin A1C for study participants. bid diabetes, number of antihypertensive medications, estimated glomerular filtration rate, waist circumference and diagnoses of stroke or cardiac failure; and finally health system factors: level 2.5 | Systems strengthening activities at study sites of healthcare institution and availability of all prescribed antihy- pertensives on NHIS. The scores on Hill-Bone compliance to high Study nurses at each site led educational programs on hyperten- blood pressure therapy scale assessed every 6 months were av- sion management by presenting materials on the need for adher- eraged. We constructed sequential models first to assess inde- ence to therapeutic lifestyle and medications aimed at controlling pendent socio-demographic variables followed by patho-biologic hypertension. Furthermore, treatment guidelines were developed and behavioral variables associated with primary and secondary by local experts and used to train physicians on hypertension outcome measures. In all analyses, two-tailed P-values <.05 were management. Supply chain systems were also strengthened to considered statistically significant. Secondary analysis considering enhance availability of antihypertensive medications at hospital SBP and DBP as continuous outcomes was also performed using pharmacies.34 linear mixed modeling to account for within-subject correlation of SBP and DBP and missing at random. Model diagnosis and fit were assessed using residual plots analysis. Statistical analysis was per- 2.6 | Prospective evaluations formed using SAS 9.4. Study participants visited study sites every two months to have clinic BP measured for 18 months. At every clinic visit, physician decision to 3  | RESULTS alter antihypertensive medications was recorded. Failure to intensify treatment when BP was uncontrolled was classified as therapeutic in- 3.1 | Demographic, lifestyle, and clinical ertia and this was assessed post hoc. The Hill-Bone questionnaire was characteristics of cohort at enrollment administered six monthly to assess adherence to therapy. Between July 1, 2015 and April 30, 2016, we enrolled 1867 (65%) participants with hypertension (HTN) and 1006 (35%) with both 2.7 | Study outcome measures hypertension and diabetes (HTN + DM). Follow-up ended on July 31, 2017. There were significant differences in demographic char- The main outcome measure was serially measured clinic acteristics between the HTN and HTN + DM groups as shown in BP ≥ 140/90 mm Hg during follow-up as a dichotomized variable. Table 1. The mean ± SD duration of hypertension diagnosis among 952  |     SARFO et Al. the HTN + DM group of 9.1 ± 7.2 years was significantly longer than 6 (n = 2,329), 16.4 ± 2.8 at month 12 (n = 2,092), and 16.4 ± 2.8 7.2 ± 7.2 years among the HTN group, P < .0001. The HTN group (n = 1,570) at month 18, (lowest score of 14 indicates excellent were on an average of 2.2 ± 0.9 antihypertensive medications com- adherence, highest score 56 indicate poor adherence). Physicians pared with 1.8 ± 0.9 among the DM + HTN group, P < .0001 with documented decisions whether or not they altered antihypertensive angiotensin-converting enzyme inhibitors and angiotensin recep- medication in the face of uncontrolled BP during clinic visit, as an in- tor blockers being used more commonly among DM + HTN group dicator of therapeutic inertia. Among participants with uncontrolled than HTN group while calcium channel blockers and beta blockers BP, physicians changed antihypertensive medications at rates that were more commonly prescribed among HTN group compared with varied between 8.6% and 19.8% of clinic visits with therapeutic iner- DM + HTN group. Among the HTN group, 57.1% reported having tia worsening over the course of 18 months (Shown in Figure S2A). all medications for disease control covered by the national health insurance and 47.9% in the DM + HTN group. Monthly expenditures on medications were higher in the DM + HTN group than in the HTN 3.4 | Determinants of blood pressure control group (Table 1). Factors associated with poor BP control were assessed among a pro- spective cohort of 2632 participants with at least one BP measure- 3.2 | Blood pressure control during follow-up ment after baseline visit. Eight factors with adjusted OR (95% CI) independently associated with uncontrolled BP were male gender: During follow-up, there was a mean ± SD of 6.5 ± 3.0 BP recordings per 1.37 (1.09-1.72), secondary education: 1.32 (1.00-1.74), non-adher- participant. Out of 9 total study visits: 1084 (41.2%) attended all visits, ence to antihypertensive treatment: 1.03 (1.00-1.06), fruit intake: 428 (16.3%) attended eight visits, 302 (11.5%) 7 visits, 217 (8.2%) at- 0.94 (0.89-1.00), duration of hypertension diagnosis: 1.01 (1.00- tended 6 visits, 155 (5.9%) had 5 visits, and 447 (17.0%) attended 4 or less 1.02), hypertension with diabetes mellitus: 2.05 (1.72-2.46), number visits. Differences in baseline demographic characteristics of those who of antihypertensive medications prescribed: 1.63 (1.49-1.79) and in- completed all study visits and those who missed some visits are shown creasing estimated glomerular filtration rate (each mL/min increase): in Table S1. The unadjusted mean ± SD systolic BP at enrollment of 0.82 (0.76-0.89), Table 2. Furthermore, we identified 8 factors 143.5 ± 21.9 mm Hg significantly declined to 139.2 ± 22.7 mm Hg at month namely increasing age, male gender, primary level education, dura- 6, 137.1 ± 21.9 mm Hg at month 12, and 136.6 ± 22.0 mm Hg at month tion of hypertension, comorbid diabetes, increasing number of anti- 18. Adjusted mean ± SD reductions in SBP were −3.81 ± 0.52 mm Hg at hypertensive medications, eGFR, and heart failure were associated month 6, −5.21 ± 0.55 mm Hg at month 12 and −5.53 ± 0.59 mm Hg at with systolic BP as a continuous variable while eight factors were month 18. Similarly, DBP declined from 83.0 ± 18.3 mm Hg at month 0, associated with poor diastolic BP control, Table 2. We also assessed to 80.7 ± 13.9 mm Hg at month 6, 79.5 ± 12.7 mm Hg at month 12 and BP control per participant over the course of follow-up defined as 79.5 ± 12.6 mm Hg at month 18. Adjusted mean ± SD reductions in DBP number of clinic visits with BP ≥140/90 mm Hg divided by number were −1.74 ± 0.32 mm Hg at month 6, −2.59 ± 0.33 mm Hg at month 12 of clinic visits during follow-up. Only 287 (10.9%) patients had BP and −2.63 ± 0.35 mm Hg at month 18. controlled at all clinic visits (Figure S2B). Overall, the proportion of study participants with target BP <140/90 mm Hg increased from baseline value of 46.3% to 54.5% at month 6, 58.6% at month 12 and 59.8% at month 18, P < .0001. 4  | DISCUSSION Proportion of participants with optimal BP <120/80 mm Hg in- creased to 17.6% at month 6, 20.0% at month 12 and 21.0% at month We have identified 8 independent factors associated with poor 18 compared with baseline of 12.5% while those with stage II hyper- control of prospectively measured BP among Ghanaians seek- tension of 160/100 mm Hg declined from 22.7% at baseline to 17.7% ing health care at public hospitals. These include male gender at month 6, 14.1% at month 12 and 14.3% at month 18. Mean SBP and educational status (demographic factors), poor adherence to and DBP at baseline and during follow-up declined for both those treatment and regularity of fruit intake (lifestyle/behavioral fac- with hypertension only and hypertension plus diabetes as were the tors) and four patho-biologic factors namely, longer duration of proportions with blood pressure readings at cut-offs of optimal, ade- hypertension, comorbid diabetes, higher number of antihyperten- quate, stages 1 and 2 for both groups as shown in Figure S1A-D. sive medications and renal impairment. Several previous studies assessing factors associated with poor BP control have been based on BP measures obtained in a cross-sectional fashion and may be 3.3 | Adherence to treatment and failure to subject to the biases due to the inherent variability of blood pres- intensify antihypertensive treatment during follow-up sure. For instance, we found receiving health care at a tertiary level, longer duration of hypertension, poor adherence to therapy, Adherence to hypertension therapy assessed using the Hill Bone reported difficulties in obtaining antihypertensive medications antihypertensive adherence scale improved significantly from a and increasing number of antihypertensive medications to be the score of 18.2 ± 4.0 at baseline (n = 2,783) to 16.8 ± 3.1 at month five factors associated with poor BP control in the same cohort SARFO et Al.      |  953 TA B L E 1   Comparison of baseline demographic and clinical characteristics of study participants according to disease status Hypertension (HPT) Hypertension and diabetes mellitus Characteristic N = 1867 N = 1006 P-value Age, mean ± SD 58.0 ± 13.0 60.0 ± 10.8 <.0001 Female, n (%) 1434 (76.9) 778 (77.4) .77 Location of residence Urban 635 (34.1) 564 (56.1) <.0001 Semi-urban 384 (20.6) 247 (24.6) Rural 843 (45.3) 195 (19.4) Highest educational status No formal education 743 (39.8) 361 (36.0) .19 Primary level 294 (15.7) 172 (17.1) Secondary level 627 (33.6) 347 (34.6) Tertiary level or more 203 (10.9) 124 (12.4) Monthly household income >1000 Ghana cedis (GHc) 127 (9.2) 87 (12.6) .0004 500-1000 GHc 153 (11.1) 109 (15.8) 300-500 GHc 203 (14.8) 111 (16.1) 210-300 GHc 120 (8.7) 57 (8.3) <210 GHc 773 (56.2) 325 (47.2) Level of health institution Tertiary referral level 765 (41.0) 781 (77.6) .0001 Secondary/district level 936 (50.1) 217 (21.6) Primary level 166 (8.9) 8 (0.8) Vascular risk factors Duration of hypertension, (y) 7.2 ± 7.2 9.1 ± 7.2 <.0001 Duration of diabetes mellitus, mean ± SD (y) NA 9.8 ± 7.2 NA Systolic blood pressure at enrollment (mm Hg), mean ± SD 142.5 ± 21.3 145.3 ± 22.7 .001 Diastolic blood pressure at enrollment (mm Hg), mean ± SD 83.2 ± 13.1 81.8 ± 12.5 .005 Medical co-morbidities Previous stroke diagnosis 84 (4.5) 70 (7.0) .005 Previous heart failure diagnosis 121 (6.5) 51(5.1) .13 Lifestyle/Behavioral factors Current alcohol use 160 (8.6) 60 (6.0) .01 Current cigarette smoking 9 (0.48) 5 (0.50) .96 Previous cigarette smoking 111 (5.9) 86 (8.5) .008 Fruit consumption Daily intake of fruits in a week, mean ± SD 2.54 ± 2.01 2.61 ± 2.01 .41 Fruit servings per day, mean ± SD 1.64 ± 1.68 1.69 ± 1.33 .42 Vegetable consumption Daily intake of vegetables in a week, mean ± SD 5.00 ± 2.20 4.86 ± 2.11 .10 Vegetable servings per day, mean ± SD 2.19 ± 1.60 2.28 ± 1.50 .15 Added salt at table Never 1491 (79.9) 866 (86.1) <.0001 Rarely 157 (8.4) 42 (4.2) Occasionally 117 (6.3) 54 (5.3) Very often 101 (5.4) 44 (4.4) (Continues) 954  |     SARFO et Al. TA B L E 1   (Continued) Hypertension (HPT) Hypertension and diabetes mellitus Characteristic N = 1867 N = 1006 P-value Regular physical activity Participants engaged in physical activities daily 1,092 (58.5) 640 (61.5) .008 Duration of time spent on physical activities in minutes, 18.2 ± 23.0 20.5 ± 23.9 .01 mean ± SD >60 min 141 (7.6) 88 (8.7) .07 20-59 min 723 (38.7) 422 (41.9) <20 min 1,003 (53.7) 496 (49.3) Anthropometric indicators Waist circumference, mean ± SD 95.0 ± 13.2 98.2 ± 12.6 <.0001 Waist circumference elevated, n (%) 1,075 (57.9) 704 (70.1) <.0001 Body mass index, mean ± SD 26.9 ± 5.9 26.7 ± 5.2 .42 Health expenditure indicators Monthly expenditure on medicines, mean ± SD (cedis) 34.1 ± 46.7 67.1 ± 81.9 <.0001 Travel cost to hospital, mean ± SD (cedis) 7.2 ± 12.7 9.2 ± 20.2 .001 Dependents on monthly household income, mean ± SD (cedis) 5.7 ± 4.3 5.5 ± 4.1 .14 Health insurance coverage for all medications, n (%) 1059 (57.1) 481 (47.9) <.0001 Laboratory Indicators eGFR, mean ± SD 76.6 ± 15.4 74.4 ± 18.4 .003 Proportion with dyslipidemia, n (%) 220 (82.7) 212 (80.0) .42 Antihypertensive Medications, mean ± SD 2.2 ± 0.9 1.8 ± 0.9 <.0001 Classes of Antihypertensive Medications Angiotensin-converting enzyme-inhibitor 770 (41.2) 471 (46.8) .004 Angiotensin receptor blocker 427 (22.9) 363 (36.1) <.0001 Beta blockers 218 (11.7) 55 (5.5) <.0001 Calcium channel blocker 1,570 (84.1) 592 (58.8) <.0001 Diuretics 732 (39.2) 159 (15.8) <.0001 Methyldopa 282 (15.1) 163 (16.2) .44 Hydralazine 22 (1.2) 21 (2.1) .06 using only enrollment BP measures.36 By using approximately hypertensive patients presenting regularly to hospitals for routine seven BP measurements per participant with 18 months of follow- follow-up had uncontrolled BPs, yet therapeutic modifications by up, we have been able to characterize factors which may possibly physicians were very infrequent in this sub-group of patients. The be consistent predictors of poor hypertension control among pa- contribution of therapeutic inertia to poor control of hypertension tients under routine care settings and could be potential targets among Ghanaians is quite substantial and estimated to be between for remediation. 80.2% and 91.4% of clinic visits. This observation is higher than In addition to more frequent clinic visits during the study period 75% reported in previous studies conducted in Spain and China.37- for BP monitoring, participants also received educational sessions 39 Clinical inertia and under-treatment of high-risk hypertensive pa- on hypertension as part of systems strengthening activities for the tients has been associated with physician beliefs such as awaiting for study. This may have accounted for the modest improvements in BP full drug effect, patients almost near target, poor compliance, fear of control during follow-up. Nearly, 46% of all hypertensive patients side effects, poor BP measurement techniques or white-coat hyper- had controlled BP at enrollment36 with salutary increment of up to tension and lack of awareness of treatment guidelines occasionally.40 60% at month 18. Of note, adherence to antihypertensive treatment Although the high burden of undiagnosed and untreated hy- improved significantly during follow-up. In spite of these positive pertension in SSA remains a major challenge, treating patients with trends, there is still considerable room for improvement. Only 11% known hypertension to therapeutic goals represents an additional of study participants had BP under control at all clinic visits and a level of challenge in the fight against hypertension. The challenge substantial majority had gradations of uncontrolled BP during fol- of poorly controlled hypertension can be addressed by design- low-up. A notable observation was that between 30% and 45% of ing evidence-based interventions that are culturally attuned and SARFO et Al.      |  955 TA B L E 2   Determinants of blood pressure control in a prospective Ghanaian cohort (Full models) Adjusted OR (95%CI) Coefficient (95% CI) Coefficient (95% CI) Determinants BP control (≥140/90 mm Hg) Mean systolic BP Mean diastolic BP Socio-demographic Age 1.03 (0.95-1.12) 0.81 (0.10-1.52) −2.62 (−3.04 to −2.19) Male gender 1.37 (1.09-1.72) 3.80 (1.83-5.77) 0.42 (−0.75 to 1.59) Education Tertiary 1.03 (0.79-1.33) 1.44 (−0.84 to 3.72) 1.26 (−0.10 to 2.61) Secondary 1.32 (1.00-1.74) 0.68 (−1.91 to 3.27) 1.07 (−0.47 to 2.61) Primary 1.06 (0.79-1.42) 3.03 (0.61-5.46) 2.05 (0.61-3.49) None 1.00 0.00 0.00 Marital status Married 1.07 (0.85-1.35) 0.61 (−1.43 to 2.66) 0.68 (−0.53 to 1.89) Divorced 1.02 (0.83 - 1.25) 0.42 (−1.37 to 2.20) 0.21 (−0.85 to 1.27) Widow 0.67 (0.41-1.09) −3.07 (−7.37 to 1.23) −1.10 (−3.65 to 1.46) Single 1.00 0.00 0.00 Income (per each 100 GHc rise) 1.02 (0.94-1.12) 0.73 (−0.04 to 1.51) −0.12 (−0.59 to 0.34) Unemployed 0.88 (0.73-1.06) −0.97 (−2.59 to 0.65) −0.90 (−1.86 to 0.06) Dependents on household income 0.99 (0.97-1.01) −0.08 (−0.26 to 0.10) −0.06 (−0.16 to 0.05) Location of residence Urban 1.15 (0.89-1.49) 0.86 (−1.16 to 2.89) 0.44 (−0.76 to 1.65) Semi-urban 1.12 (0.88-1.41) 0.41 (−1.84 to 2.65) 1.17 (−0.16 to 2.50) Rural 1.00 0.00 0.00 Lifestyle/Behavioral Any cigarette use 1.19 (0.28-5.14) −0.15(−13.00 to 12.70) −1.83 (−9.49 to 5.83) Current alcohol use 0.86 (0.65-1.13) −1.52 (−3.92 to 0.89) 0.54 (−0.89 to 1.97) Physical activity 0.95 (0.81-1.12) −0.64 (−2.02 to 0.75) 0.16 (−0.66 to 0.99) Fruit intake (servings/d) 0.94 (0.89-1.00) −0.41 (−0.91 to 0.09) −0.24 (−0.54 to 0.05) Vegetable intake (servings/d) 0.97 (0.91-1.03) −0.15 (−0.66 to 0.37) −0.10 (−0.41 to 0.21) Salt intake 1.03 (0.82-1.30) −0.05 (−2.06 to 1.97) −0.48 (−1.68 to 0.72) Adherence to Hypertension treatment 1.03 (1.00-1.06) 0.18 (−0.08 to 0.44) 0.23 (0.08-0.39) Patho-biologic factors Duration of hypertension 1.01 (1.00-1.02) 0.14 (0.03-0.24) 0.05 (−0.01 to 0.11) Hypertension with DM 2.05 (1.72-2.46) 6.83 (5.26-8.39) 1.16 (0.23-2.09) Number of antihypertensives 1.63 (1.49-1.79) 4.88 (4.07-5.68) 1.62 (1.14-2.09) Estimated GFR (mL/min rise) 0.82 (0.76-0.89) −2.02 (−2.65 to −1.38) −0.50 (−0.88 to −0.12) Waist circumference 0.97 (0.91-1.03) −0.49 (−1.00 to 0.03) 0.82 (0.52-1.13) Heart failure 0.76 (0.56-1.04) −3.13 (−5.79 to −0.47) 0.70 (−0.88 to 2.28) Stroke 1.03 (0.74-1.43) 0.58 (−2.31 to 3.47) 1.01 (−0.71 to 2.72) Organizational Antihypertensives covered by National 1.07 (0.92-1.25) −0.25 (−1.60 to 1.09) 0.24 (−0.56 to 1.04) Insurance Level of Health Institution Tertiary level 1.38 (0.85-2.23) 5.07 (0.70-9.44) −3.43 (−6.03 to −0.83) Secondary/district level 1.50 (0.90-2.48) 5.04 (0.89-9.18) −4.36 (−6.82 to −1.89) Primary 1.00 0.00 0.00 Bold indicate variables which attained significant independent associations with dependent variables. 956  |     SARFO et Al. sensitive to the local settings where patients are receiving care deployment of Community Health Nurses would be instrumental in based on factors identified in the local population. For instance, overcoming some of the system-level barriers to hypertension man- patient-level factors associated with poor BP control such as male agement due critical shortage of skilled personnel.49,50 gender and educational level require further studies to unravel the unique contributors to poor control in these specific demographic sub-groups so interventions could be targeted to improve their BP 4.1 | Strengths and limitations control. We observed significant improvements in adherence to antihypertensive treatment during follow-up for the entire cohort This is one of the few studies emanating from SSA to identify the de- for which we speculate could have due to the regular education terminants of poor BP control in a prospective cohort of hypertensive given to patients at each visit. However, the fact that adherence patients. There are some limitations worth mentioning. Although we scores measured every 6 months were better among those with observed significant improvements in clinic BP control over the course BP on target than those not on target suggests that individual level of follow-up, many participants found the two monthly visits difficult rather than group educational intervention for the poorly adher- to comply with leading to attrition in follow-up. Hence, the improved ent might improve adherence further. There is sufficient evidence outcomes might reflect healthy survivorship bias, however, our statisti- of beneficial associations between high adherence to antihyper- cal approach accounted for the missing data. Many of the independent tensive treatment and reduced risk of cardiovascular events.41,42 variables such as fruit intake and income levels were assessed only at Furthermore, the Dietary Approaches to Stopping Hypertension baseline but could have changed during follow-up. The dynamics of (DASH) diet has emphasized the importance of increasing fruit and such time-dependent variables could influence prospective BP control vegetable intake while reducing the consumption of red meat.43 but these were not assessed in the present study. In our cohort, a unit increase in fruit serving per week was asso- ciated with improved BP control over follow-up with an adjusted OR of 0.94 (95% CI: 0.89-1.00) and that for vegetable intake was 5  | CONCLUSION 0.97 (0.91-1.03). Indeed, increased long-term fruit but not vegeta- ble consumption was found to reduce the risk of developing hy- Modest improvements in BPs measured prospectively were ob- pertension from an analysis of 3 large prospective US cohorts.44 served in this Ghanaian cohort due probably to more frequent A recent study has reported on a protective association between clinic visits, and patient education. Our study, conducted in a green leafy vegetable consumption and stroke occurrence among resource-limited setting has uncovered potential targets such Ghanaians and Nigerians.45,46 as therapeutic inertia and specific demographic subgroups for The findings from our study have important public health ram- which interventions aimed at improving BP control further may ifications. The WHO has identified hypertension in Africa as an ep- be directed. idemic and in 2015, the World Heart Federation set a key target of achieving a 25% relative reduction in the prevalence of raised blood CONFLIC T OF INTERE S T pressure globally via enhanced awareness, detection, and control.47 The authors have the following competing interests: Funding for However, geopolitical, health-system, healthcare professional, and this study was provided by MSD, Pfizer, Sanofi (each a Participant patient-related factors that hamper hypertension control require care- company) and Bill and Melinda Gates Foundation (collectively, the fully crafted solutions to achieve these laudable goals. Universal health funders) through the New Venture Fund (NVF). FSS, LMM, GB, DA, coverage across Africa to support patients with hypertension is a major JPR, DOA, LA, JS received honoraria from NVF for participation in challenge. In this vein, the widespread availability of national health the present study. There are no patents, products in development or insurance coverage policy for hypertension care in Ghana is a signif- marketed products to declare. icant step forward and may have been contributory to the measure of success in controlling BP at public hospitals. Hence in our analy- AUTHOR CONTRIBUTIONS ses, insurance coverage, health expenditures, household income, and Fred Stephen Sarfo, Linda M. Mobula, Gilbert Burnham, and David other socio-economic indicators were not independent predictors of Ofori-Adjei involved in conceptualization. Daniel Ansong involved BP control. Having insurance coverage has been shown in the United in data curation. Mulugeta Gebregziabher and Fred Stephen Sarfo States to be associated with improved BP control among hypertensive involved in formal analysis. Fred Stephen Sarfo and Osei Sarfo- patients.48 Furthermore, health system factors such as the level of Kantanka involved in investigation. Fred Stephen Sarfo, Daniel health institution were not independent predictors of poor BP control, Ansong, and Jacob Plange-Rhule involved in methodology. Lynda perhaps due to clustering effect of having standard study procedures Arthur and Jasper Sablah involved in project administration. across sites. Health professional level factors, in particular, therapeutic David Ofori-Adjei, Gilbert Burnham, Jacob Plange-Rhule, Daniel inertia is a major challenge and could be addressed by the implementa- Ansong, and Edith Gavor involved in supervision. Fred Stephen tion of clinical guidelines as only 16 (26%) of 62 African countries have Sarfo involved in writing-original draft. Linda M. Mobula, David clinical practice guidelines on hypertension.31 Certainly, the paucity of Ofori-Adjei, and Jacob Plange-Rhule involved in writing-review physicians across SSA means that task shifting strategies such as the and editing. SARFO et Al.      |  957 ORCID 18. Tesfaye F, Byass P, Wall S. Population based prevalence of high Fred S. Sarfo https://orcid.org/0000-0001-7274-5093 blood pressure among adults in Addis Ababa: uncovering a silent epidemic. BMC Cardiovasc Disord. 2009;9(39):1471-2261. 19. van de Vijver SJ, Oti SO, Agyemang C, Gomez GB, Kyobutungi C. R E FE R E N C E S Prevalence, awareness, treatment and control of hypertension 1. Lim SS, Vos T, Flaxman AD, et al. A comparative risk assessment among slum dwellers in Nairobi, Kenya. J Hypertens. 2013;19:19. of burden of disease and injury attributable to 67 risk factors 20. Jenson A, Omar AL, Omar MA, Rishad AS, Khoshnood K. Assessment and risk factor clusters in 21 regions, 1990–2010: a systematic of hypertension control in a district of Mombasa Kenya. Glob Public analysis for the Global Burden of Disease Study 2010. Lancet. Health. 2011;6(3):293-306. 2012;380:2224-2260. 21. Hendriks ME, Wit FWNM, Roos MTL, et al. Hypertension in Sub- 2. Rodgers A, Ezzati M, Vander Hoorn S, et al. Distribution of major Saharan Africa: cross-sectional surveys in four rural and urban com- health risks: findings from the Global Burden of Disease Study. PLoS munities. PLoS ONE. 2012;7(3):12. Med. 2004;1:44-55. 22. Damasceno A, Azevedo A, Silva-Matos C, Prista A, Diogo D, Lunet 3. Ikeda N, Sapienza D, Guerrero R, et al. Control of hypertension with N. Hypertension prevalence, awareness, treatment, and control in medication: a comparative analysis of national surveys in 20 coun- mozambique: urban/rural gap during epidemiological transition. tries. Bull World Health Organ. 2014;92:10-19. Hypertension. 2009;54(1):77-83. 4. NCD Risk factor. Collaboration (NCD-RisC). Worldwide trends in 23. Bovet P. The epidemiologic transition to chronic diseases in devel- blood pressure from 1975 to 2015: a pooled analysis of 1479 pop- oping countries: cardiovascular mortality, morbidity, and risk fac- ulation-based measurement studies with 19.1 million participants. tors in Seychelles (Indian Ocean): investigators of the Seychelles Lancet. 2017;389(10064):37-55. heart study. Soz Praventivmed. 1995;40(1):35-43. 5. Sarfo FS, Akassi J, Awuah D, et al. Trends in stroke admission and 24. Bovet P, Ross AG, Gervasoni J-P, et al. Distribution of blood pres- mortality rates from 1983 to 2013 in central Ghana. J Neurol Sci sure, body mass index and smoking habits in the urban population 2015;357(1):240-245. of Dar es Salaam, Tanzania, and associations with socioeconomic 6. Chow CK, Teo KK, Rangarajan C, et al. Prevalence, awareness, status. Int J Epidemiol. 2002;31(1):240-247. treatment, and control of hypertension in rural and urban com- 25. Dewhurst MJ, Dewhurst F, Gray WK, Chaote P, Orega GP, Walker munities in high-, middle-, and low-income countries. JAMA. RW. The high prevalence of hypertension in rural-dwelling 2013;310(9):959-968. Tanzanian older adults and the disparity between detection, treat- 7. Phillips LS, Branch WT, Cook CB, et al. Clinical inertia. Ann Intern ment and control: a rule of sixths? J Hum Hypertens. 2012;13(10):59. Med. 2001;135:825-834. 26. Edwards R, Unwin N, Mugusi F, et al. Hypertension prevalence 8. Chowdhury R, Khan H, Heydon E, et al. Adherence to cardiovas- and care in an urban and rural area of Tanzania. J Hypertens. cular therapy: a meta-analysis of prevalence and clinical conse- 2000;18(2):145-152. quences. Eur Heart J. 2013;34:2940-2948. 27. Musinguzi G, Nuwaha F. Prevalence awareness and control of hy- 9. Glynn LG, Murphy AW, Smith SM, et al. Interventions used to pertension in Uganda. PLoS ONE. 2013;4:e62236. improve control of blood pressure in patients with hypertension. 28. Pires JE, Sebastiao YV, Langa AJ, Nery SV. Hypertension in Cochrane Database Syst Rev. 2010;3:CD005182. Northern Angola: prevalence, associated factors, awareness, treat- 10. Sesso HD, Stampfer MJ, Rosner B, et al. Systolic and diastolic blood ment and control. BMC Publ Health. 2013;13(1):90. pressure, pulse pressure, and mean arterial pressure as predictors of 29. Kamadjeu RM, Edwards R, Atanga JS, Unwin N, Kiawi EC, Mbanya cardiovascular disease risk in men. Hypertension. 2000;36:801-807. JC. Prevalence, awareness and management of hypertension in 11. Van Trijp MJ, Grobbee DE, Peeters PH, van Der Schouw YT, Bots Cameroon: findings of the 2003 Cameroon burden of diabetes ML. Average blood pressure and cardiovascular disease-related baseline survey. J Hum Hypertens. 2006;20(1):91-92. mortality in middle-aged women. Am J Hypertens. 2005;18:197-201. 30. Hammami S, Mehri S, Hajem S, et al. Awareness, treatment and 12. Lewington S, Clarke R, Qizibash N, Peto R, Collins R. Prospective control of hypertension among the elderly living in their home in Studies Collaboration. Age-specific relevance of usual blood Tunisia. BMC Cardiovasc Disord. 2011;11(65):1471-2261. pressure to vascular mortality: a meta-analysis of individual 31. Dzudie A, Rayner B, Ojji D, et al. Roadmap to achieve 25% hyper- data for one million adults in 61 prospective studies. Lancet. tension control in Africa by 2025. Global Heart. 2018;13:45-49. 2002;360(9349):1903-1913. 32. Adeloye A, Basquill C. Estimating the prevalence and awareness 13. Maginga J, Guerrero M, Koh E, et al. Hypertension control and its rates of hypertension in Africa: a systematic analysis. PLoS ONE. correlates among adults attending a hypertension clinic in Tanzania. 2014;e104300. J Clin Hypertens (Greenwich). 2016;18(3):207-216. 33. Ataklte F, Erqou S, Kaptoge S, et al. Burden of undiagnosed hyper- 14. Anbaw A, Alemie G, W/Yohannes S, Mengesha Z. Adherence to tension in sub-Saharan Africa: a systematic review and meta-analy- antihypertensive treatment and associated factors among patients sis. Hypertension. 2015;65:291-298. on follow up at University of Gondat Hospital, Northwest Ethiopia. 34. Mobula LM, Sarfo S, Arthur L, et al. A multicenter prospective co- BMC Public Health. 2012;12:282. hort study to evaluate the effect of differential pricing and health 15. Solet J-L, Baroux N, Pochet M, et al. Prevalence of type 2 diabe- systems strengthening on access to medicines and management of tes and other cardiovascular risk factors in Mayotte in 2008: the hypertension and diabetes in Ghana: A study protocol. Gates Open MAYDIA study. Diabetes Metab. 2011;37(3):201-207. Res. 2018;2:6. 16. Muluneh AT, Haileamlak A, Tessema F, et al. Population based 35. Kim MT, Hill MN, Bone LR, Levine DM. Development and testing of survey of chronic non-communicable diseases at gilgel gibe the Hill-Bone compliance to high blood pressure therapy scale. Prog field research center, southwest Ethiopia. Ethiop J Health Sci. Cardiovasc Nurs. 2000;15(3):90-96. 2012;22:7-18. 36. Sarfo FS, Mobula LM, Burnham G, et al. Factors associated with 17. Awoke A, Awoke T, Alemu S, Megabiaw B. Prevalence and associ- uncontrolled blood pressure among Ghanaians: Evidence from a ated factors of hypertension among adults in Gondar, Northwest multicenter hospital-based study. PLoS ONE. 2018;13(3):e193494. Ethiopia: a community based cross-sectional study. BMC Cardiovasc 37. Redón J, Coca A, Lázaro P, et al. Factors associated with thera- Disord. 2012;12(113):1471-2261. peutic inertia in hypertension: validation of a predictive model. J Hypertens. 2010;28(8):1770-1777. 958  |     SARFO et Al. 38. Domenech CS, Caro JL, Sanz VP, et al. Objective KONTROL study: 47. Adler AJ, Prabhakaran D, Bovet P, et al. Reducing cardiovas- therapeutic inertia in hypertensive patients attended in primary cular ortality through prevention and management of raised care of Spain. Aten Primaria. 2011;43(12):638-647. blood pressure: a World Heart Federation roadmap. Glob Heart. 39. Feldman RD, Liu L, Wu Z, Zhang Y, Yu X, Zhang XH. Hypertension 2015;10:111-122. attitude Perspectives and Needs (HAPPEN): A real-world survey of 48. Duru O, Vargas R, Kermah D, et al. Health insurance status and physicians and patients with hypertension in China. J Clin Hypertens hypertension monitoring and control in the United States. Am J (Greenwich). 2017;19(3):256-264. Hypertens. 2007;20:348-353. 40. Ferrari P. National Coordinators for the Reasons for not Intensifying 49. Ogedegbe G, Plange-Rhule J, Gyamfi J, et al. Health insurance cov- Antihypertensive Treatment (RIAT) trial12. Reasons for thera- erage with or without a nurse-led task shifting strategy for hyper- peutic inertia when managing hypertension in clinical practice in tension control: a pragmatic cluster randomized trial in Ghana. PLoS non-Western countries. J Hum Hypertens. 2009;23(3):151-159. Med. 2018;15(5):e1002561. 41. Blood Pressure Lowering Treatment Trialists’ Collaboration, 50. Sarfo F, Treiber F, Gebregziabher M, et al. PINGS (Phone-Based Turnbull F, Neal B, et al. Effects of different regimens to lower Intervention under Nurse Guidance after Stroke): interim results of blood pressure on major cardiovascular events in older and a pilot randomized controlled trial. Stroke. 2018;49(1):236-239. younger adults: meta-analysis of randomized trials. BMJ. 2008;336:1121-1123. 42. Sokol MC, McGuigan KA, Verbrugge RR, et al. Impact of medication SUPPORTING INFORMATION adherence on hospitalizations risk and healthcare cost. Med Care. Additional supporting information may be found online in the 2005;43:521-530. 43. Sacks FM, Moore TJ, Appel LJ, et al. A dietary approach to pre- Supporting Information section. vent hypertension: a review of the Dietary Approaches to Stop Hypertension (DASH) Study. Clin Cardiol. 1999;54:381-387. 44. Borgi L, Muraki I, Satija A, Willett WC, Rimm EB, Forman JP. Fruit How to cite this article: Sarfo FS, Mobula L, Plange-Rhule J, et and vegetable consumption in three prospective cohort studies. al. Longitudinal control of blood pressure among a cohort of Hypertension. 2016;67:288-293. 45. Owolabi M, Sarfo F, Akinyemi R, et al. Dominant modifiable risk fac- Ghanaians with hypertension: A multicenter, hospital-based tors for stroke in Ghana and Nigeria (SIREN): a case-control study. study. J Clin Hypertens. 2020;22:949–958. https://doi. Lancet Glob Health. 2018;6(4):e436-e446. org/10.1111/jch.13873 46. Sarfo FS, Ovbiagele B, Gebregziabher M, et al. Stroke among young west Africans: evidence from the SIREN (Stroke Investigative Research and Educational Network) Large Multisite case-control study. Stroke. 2018;49(5):1116-1122.