Ghana is located on the Atlantic Coast of West Africa 4 degrees north of the Equator. It occupies a land area of 238,537 km² and had a population of 23.5 million of which 56% is rural. Life expectancy at birth was 55 years for males and 60 years for females in 2000 13. The country is divided into ten administrative regions and 170 districts (Figure 1). Per capita total health expenditure as a percentage of GDP was 4.5% in 2003. Ghana ranked 135 out of 177 countries on the human development index in 2007 14. Fifty eight percent of the population lives less than 30 minutes of a public or private health facility, with geographical access being considerably better in the urban (78.5%) than in the rural populations (42.3%) 15. The two largest cities Accra, the national capital and Kumasi, the regional capital of the Ashanti Region, have the highest concentration of doctors in the country. About 60% of persons who report ill or injured consult a health practitioner while 32% purchase medicines directly from pharmacy shops for their ailments 16.

Potentially relevant papers on hypertension in Ghana were identified through a search of the PubMed and Google Scholar databases from 1970 to August 2009, supplemented by manual scan of the bibliographies of retrieved articles. Direct contact was made with some of the authors of included papers to locate some articles which were not accessible through HINARI or were unpublished. The key words used for the search were hypertension, blood pressure, cardiovascular and Ghana.

Studies included were those relating to adult non-pregnant human subjects aged 13 years and older published in English between January 1970 and August 2009. In addition, Master of Public Health student dissertations from the University of Ghana between 1996 and 2008 were reviewed and community-based prevalence studies on hypertension selected for analysis.

Published studies were excluded if they did not meet the above criteria or if they pertained to studies on patients. The abstracts of potentially relevant articles were reviewed out of which the full text articles of those deemed to be relevant were obtained for analysis. Eligible articles were assessed by the sole author of this paper. A standard data extraction sheet was used to extract information which included the study location, sampling strategy, sample size, sample demographic characteristics, prevalence, mean systolic and diastolic blood pressure, prevalence of hypertension and where available, the independent associated factors. Data on the awareness, treatment and control of those found or known to be hypertensive were obtained. Data quality issues assessed included sampling procedure, sample size estimation, response rate, selection bias, control of confounding, and quality control in blood pressure measurement. Differences in the methods of measuring blood pressure for each study were described rather than directly assessed against a gold standard. No pooled analysis was performed owing to the heterogeneity of study populations and the absence of datasets.

The primary search of the database identified 68 individual papers and the hand search provided three additional papers. Google Scholar identified two additional papers. Thus, a total of 73 papers were identified for evaluation (Figure 2). Twenty four papers estimated prevalence of hypertension but one of them was among patients attending a blood pressure (BP) clinic and so was excluded 17. Eventually, 15 unique published papers including two from Google Scholar 1819 and one from conference proceedings 20 were included in the analysis of hypertension in Ghana. One set of studies involved a follow-up study of the original cohort after five years 2021. Another study conducted among public servants was complemented by a further study on a subset of sample which involved detailed cardiovascular investigations 222324. The original and subset study were reckoned as a single study. In addition to the published studies, three out of 350 MPH student dissertations were found to be relevant 2526. Of these, one had been published 27 and so overall, 17 unique published and unpublished papers were analysed (Table 1).

The selected studies were conducted between 1973 and 2008 in four out of Ghana's ten regions - Greater Accra (10), Ashanti (4), Upper East (2) and Volta Regions (1). Nine of the studies were conducted in urban populations, five in rural populations and three in mixed populations. The study populations were varied and included general population, volunteers, public servants, factory workers and church worshippers. One study selected subjects from churches, schools and banks 28. Except for one study which involved only women 29 and another which involved only men 30, the rest of the studies involved both sexes. The proportion of women in the studies ranged from 16% to 100%. In 14 studies involving both sexes, ten enrolled more females than males and up to 77% of study subjects were females (Table 1). As expected, males were generally predominant in study populations such as civil servants, factory workers and prison officers 232531.

As with the sex composition, the age profile of study subjects was heterogeneous, ranging from 15 to 102 years. The mean age of participants was less than 40 years in six studies, between 40 and 50 years in six studies and around 55 years in one study. Although the mean ages were not provided for four studies, it is most likely that the mean age was less than 40 years in three of these studies. In one of these studies, 64% of the subjects were younger than 45 years 32 and in another 37% were aged 15 to 25 years (with only 6% aged 55-65 years) 21. In some studies, there was an attempt to oversample older subjects to compensate for their smaller size within the general population 29.

A total of 26,649 different adults were involved in the reviewed studies (Table 1). The sample size ranged from 287 to 6,900 adults. It was more than 1,000 in 8 studies and less than 500 in five studies. Except for three studies, the sample size was either not based on any calculation or this was not reported. Thirteen studies employed rigorous sampling strategy at all levels of sampling with well defined sampling frames at the household or institutional level through the use of household census 20212629323334, electoral registers 27 or staff registers 2325283031. Three studies lacking defined sampling frames examined interested volunteers on a market day 35, Christian worshippers attending church services or persons in every other house in a rural village 28.

The response rate ranged from 25% to 98% between studies (Table 1) 2032. Only three studies reported response rates less than 60% 202936. In the study with the lowest response rate, only 25% of the original members of a cohort could be located five years later. Within studies, the response rate differed between the study groups, by sex and by location. For example, 40% to 80% of persons in 12 villages invited to take part in the study agreed to do so 36. Only five studies described reasons for non-response or characteristics of non-responders 2024273637. Non-response was due to feeling healthy, feeling ill, temporary absence, fear of premature retirement if found to be a hypertensive public servant, interference with time at work, and lack of interest.

Nearly all studies reported significant investments to ensure good quality BP readings. These investments included use of experienced nurses, periodic re-certification of personnel, confirmatory readings by supervising physician and the use of standardized measurement protocols. Reference methods guiding blood pressure measurement and analyses that were mentioned by individual studies were WHO MONICA 3438, the Joint National Committee on Prevention, Detection, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 6 or 7) 3133, the WHO and International Society of Hypertension Guidelines 313940, American Heart Association 19, other WHO guidelines 72032, video-tutored course (Shared Care, Torrance, California) 2733, other published studies 29 and unspecified standard methods 18. One study trained the blood pressure measurement team to achieve inter-observer variation of less than 4 mmHg 35.

Eleven studies (65%) employed manual mercury sphygmomanometers while six studies (35%) used electronic BP monitors, usually the Omron BP monitor (Table 2). Of the 9 studies describing the cuff size used for the BP measurement, nearly all reported using more than one cuff size to suit the small and large arms. One study used one cuff size for the arms of all subjects while another used a wrist BP monitor 26. Up to 12 trained personnel, usually nurses, measured the blood pressures in each study. In auscultatory measurements, the first and the fourth 21 or fifth 273033 Korotkoff phase sounds corresponded to the systolic (SBP) and diastolic blood pressure (DBP) respectively. Blood pressure was generally taken on the right arm of seated patients and in one case, also in supine subjects 29. Preparation commonly involved an initial rest period of at least five minutes and occasionally, prohibition of smoking 2533.

Except for four studies, all studies reported taking BP during a single visit. Studies which measured BP at two visits, repeated the measurements one day 23, 3 weeks 31 or one month 30 later for those whose BPs were greater than 140/90 mmHg during the initial visit 3031. In one study, repeated measurements were done for 16% of study subjects 2 weeks later in order to correct a regression dilution effect arising because baseline BP tends to underestimate the usual BP 34. The repeated measurements permitted a regression dilution ratio to be calculated and used to adjust the baseline casual BP to estimate the usual BP.

During a single visit, the BP was measured one to three times between different studies (Table 2). In studies with a single BP reading, measurement was only repeated in those with an initially raised reading 2021313541. Multiple BP readings were separated by intervals of 30 seconds to 27 minutes in different studies 3035. The average of two 19273342 or three 253032 readings were usually used in the analysis, with some studies ignoring the first reading and using the mean of the second and third reading 263136. The Mamprobi 1975 study analysed the initial of the two repeated measurements in those with BP higher than 160/95 mmHg 2141.

The five studies conducted up to 1995 used a threshold of 160/95 mmHg for hypertension while 13 studies conducted later used a threshold of 140/90 mmHg. Most of the latter studies also considered self-reported hypertension as hypertensives in addition to those with raised BP 18273133. However, some studies restricted their diagnosis of hypertension to only the blood pressure readings 19343543. One study specifically excluded subjects who reported being hypertensives or diabetics from the analysis 19.

Most studies reported a crude prevalence of hypertension between 25% and 48%, using the newer threshold of 140/90 mmHg (Table 1). The Women's Health Study of Accra (WHSA) reported a crude prevalence of 54.6% among 1,303 women in upright position 44. Only four studies reported a prevalence of less than 20%. Regardless of the cut-off used, most studies reported a higher prevalence among men than among women though differences were frequently small (< 4 percentage points) and where assessed, not statistically significant 36. Only four of 15 studies reported a higher prevalence in women 19273233. The widest sex disparities in prevalence were observed in the rural Kpando district 35 and in the early urban studies of Accra 212324. In one study, the prevalence in men was more than twice that in women 23. The pattern of sex differences in the prevalence of hypertension remained after adjusting for age 273133.

In mixed populations, the prevalence of hypertension was higher in urban than in rural populations (Table 3). In four of six rural populations, prevalence of hypertension (BP ≥ 140/90 mmHg) was 24% or higher 27283536. It is striking that in the other two studies with lower prevalence, both conducted in the same rural district in the Upper East Region, there was a three-fold disparity in the prevalence of hypertension 2634.

Within studies, the prevalence of hypertension increased with increasing age although the gradient was not always monotonic 2337. The age-standardized prevalence 3145 or the weighted prevalence 29 was similar or lower than the crude prevalence where these were provided. For instance, the 48% crude prevalence of hypertension in the Women's Health Study of Accra compares with a weighted prevalence of 40% 29. Hypertension also increased with increasing body mass index.

Between studies with available data, the mean systolic BP was in the range 113.0-140.9 mmHg and the diastolic BP in the range 69.0-86.4 mmHg (Table 3). There was weak correlation between the mean age and the mean SBP (R = 0.20, p = 0.29) or the mean DBP (R = 0.05, p = 0.45). The mean systolic and diastolic blood pressures were higher in urban than in rural areas.

The age-specific mean blood pressures varied between studies and were higher in the older age groups (Table 4). The mean age-specific blood pressures were generally higher in men in the younger age groups and higher in women from 55 years. The widest disparities in mean age-specific systolic and diastolic blood pressures were observed in the older age groups. The lowest age-specific diastolic pressures were recorded by the rural Danfa study 32 while the highest DBPs were recorded by the urban Mamprobi cardiovascular study 20, both studies conducted in the 1970s. The latter study 20 also recorded the highest age-specific systolic BPs whereas the lowest age-specific BPs varied inconsistently between studies.

In bivariate analyses in which this was assessed, the mean SBP, mean DBP or prevalence of hypertension increased with increasing age 71920272829313233343637. Similarly, these same parameters tended to increase with increasing BMI 1819212731344244.

Multivariate analyses to adjust for the effects of potential confounders were performed by six studies 192728303134. Factors independently associated with SBP or DBP in various combinations in the different studies were older age, menopause before age 50 years, no formal education, overweight or obesity, urban residence, land ownership, and male sex 272829303134. On the other hand, having completed six years of education 27, alcohol consumption 28 and time of blood pressure measurement 34 were inversely associated with blood pressure. In the study reporting the highest association with BMI, obese persons had 6.9 (95% CI 1.7, 28.2) times the odds of having hypertension as those with normal BMI.

In multiple linear regression analysis, both age and BMI were positively and independently associated with SBP and DBP in females, while age was positively and independently associated with DBP in males in the rural Kassena Nankana district 34. Findings were not always consistent between studies. In Kumasi, abnormal waist circumference was associated with higher odds for SBP or DBP compared to those with normal waist circumference, with the difference barely reaching statistical significance. However, the association between abnormal WHR, BMI and waist to height ratio and systolic or diastolic hypertension was not statistically significant 19.

Detailed analysis of the awareness, treatment and control of hypertension was available for seven studies (Table 5). Of the hypertensives identified in these studies, only 22%-54% were aware of their condition, 7%-31% were on treatment and 0%-13% had their blood pressures controlled. Between studies, the proportion of hypertensives aware, on treatment and who had controlled blood pressure in urban populations was not markedly different from that in rural populations. The most favourable awareness, treatment and response was observed among civil servants in Accra 31 and the worst among public servants in Accra in 1973 22. Although the differences in study populations hinder direct comparison, there appeared to be a trend towards improved awareness, treatment and control (Table 5) from the mid-1970s to the late 2000s.

In one study, age group older than 35 years relative to the 16-34 years was independently associated with being aware of hypertension after adjusting for age group, sex, locality, educational level, occupation, BMI, land ownership, smoking and alcohol consumption 28. Older age from 50 years, trading and overweight were independently associated with being on drug treatment while trading was independently was associated with adequate blood pressure control.