Commentary: The salience of socioeconomic status in assessing cardiovascular disease and risk in low- and middle-income countries

Emerging data suggest that cardiovascular disease and risk factor (CVD/RF) prevalence trends are rising in Bangladesh and other low- and middle-income countries (LMICs). 1 It is clear, however, that average trends and patterns mask the fact that the distribution of CVD/RF is uneven across different socioeconomic groups, 2,3 with the prevalence being substantially higher among high socioeconomic groups. Harshfield and colleagues report that the prevalences of hypertension/hyperglycaemia are substantially higher among the high-socioeconomic status (SES) individuals in Bangladesh. 4 Whereas early prevention efforts may be ideal in all population groups, as is seen in the global effort to implement universal health coverage, 5 this remains a challenge when allocating scant resources, particularly as the threat of communicable diseases still remains high in many of the LMICs. 2,6–8 In this commentary, we start by pointing to a methodological concern related to the analysis by Harshfield and colleagues, resulting in the distortion of CVD/RF prevalence estimates. We then take this opportunity to present a systematic review of the existing literature of SES and CVD/RF in Bangladesh and underscore the importance of collecting systematic national surveillance data. With an increasing number of studies considering SES in assessments of CVD/RF, we present a rationale to consider both the socioeconomic gradient of CVD/RF and the percentage of the total CVD/RF burden attributable to each SES group. In the final section of this commentary, we highlight the explicit need to consider gender in assessing SES and CVD/RF in LMICs, and especially in countries characterized by substantial gender inequalities. 9 The analysis conducted by Harshfield and colleagues contains some notable concerns related to the sample sizes along with incorrect adjustments for survey design and sampling weights. Together, these errors have distorted the CVD/RF prevalence estimates reported by the authors, which we clarify below. The study is based on the 2011 Bangladesh Demographic and Health Survey (BDHS), which consists of a representative sample of households in Bangladesh. In one-third of households, ever-married men, ages 15–54 years, were invited to participate in a men’s health survey; however, all men and women aged 35 years and older were invited to participate in a biomarker survey that required participants to have blood pressure and blood glucose measured. There is a discrepancy in how Harshfield and colleagues report the number of men and women eligible for blood pressure and glucose measurement compared with the number of individuals who participated and on whom valid and complete data were obtained. The final BDHS survey report indicates a total sample size of 4311 women and 4524 men aged 35years and older, who were eligible for the biomarker component, 10 of which 92% of women and 86% of men participated in blood pressure measurement, and 89% of women and 83% of men participated in blood glucose measurement. The sample sizes of 4311 women and 4523 men reported (and used incorrectly as denominators) by Harshfield and colleagues represent the eligible population and not the participating population (note a further discrepancy of n = 1 in men’s sample size between Harshfield et al . and the BDHS report). Consequently, the incorrect sample sizes comprising the denominators of the authors’ calculations result in an underestimation of the reported prevalences of hypertension and hyperglycemia. We illustrate the magnitude of this error in Table 1 , which reproduces part of Harshfield’s and colleagues Table 3 . For instance, the prevalence of medical hypertension is reported as 19.6% in men and 33.6% in women, whereas we calculated the prevalence as 21.6% in men and 35.2% in women based on the correct sample size and with adjustments for survey sampling weights and survey design characteristics. We draw the attention of the reader to this point mainly as clarification, as it appears that these errors affect the prevalence estimates and do not, as far as we can determine, influence the SES risk factor associations or inferences drawn by the study authors. Corrected prevalence estimates of selected blood pressure and blood glucose outcomes by sex, Bangladesh Demographic Health Survey 2011 Unweighted counts, frequencies adjusted for survey weights and survey design characteristics. Definitions of measured/medical hypertension and hyperglycaemia given by Harshfield et al. Corrected prevalence estimates of selected blood pressure and blood glucose outcomes by sex, Bangladesh Demographic Health Survey 2011 Unweighted counts, frequencies adjusted for survey weights and survey design characteristics. Definitions of measured/medical hypertension and hyperglycaemia given by Harshfield et al. Following a previous critical review of studies investigating the association between SES and CVD/RF in India, 2 we conducted a similar systematic review for Bangladesh. Using MEDLINE and EMBASE databases, we reviewed 260 articles that specifically matched our search terms ( Box 1 ) and identified a total of 67 published articles that assessed the association of seven CVD/RF (hypertension, cholesterol, diabetes, obesity, body mass index, physical inactivity, diet and smoking) with SES factors (including education and social class) in Bangladesh for the period January 1980 to June 2015. We excluded studies whose primary outcome was mortality and Bangladeshi populations not residing in Bangladesh. Individual study details extracted from the studies included in this systematic review are outlined in Table 2 . Box 1Systematic review search terms We conducted a cursory search on MEDLINE of the following MeSH terms: [‘Neoplasms’ OR ‘Cardiovascular Diseases’ OR ‘Wounds and Injuries’ OR ‘Hypertension’ OR ‘Diabetes Mellitus’ OR ‘Obesity’ OR ‘Chronic Obstructive Pulmonary Disease’ OR ‘Respiratory Tract Diseases’ OR ‘Schizophrenia’ OR ‘Psychotic Disorders’ OR ‘Mental Disorders’ OR ‘Depressive Disorder’ AND ‘Bangladesh’] which yielded 1087 articles, of which 209 (19%) were indexed as ‘cardiovascular disease’. We systematically searched the following MeSH terms in the MEDLINE and EMBASE databases: (‘Socioeconomic factors’ OR ‘education’ OR ‘social class’) AND (‘Cardiovascular diseases’ OR ‘Obesity’ OR ‘Hypertension’ OR ‘Diabetes mellitus’ OR ‘Cholesterol’ OR ‘Physical inactivity’ OR ‘Smoking’ OR ‘Motor activity’ OR ‘Physical activity’ OR ‘Diet’) AND ‘Bangladesh’. In addition, we supplemented relevant articles from the reference lists of identified articles in our search. Studies in systematic review of cardiovascular risk factors and socioeconomic factors in Bangladesh Studies in systematic review of cardiovascular risk factors and socioeconomic factors in Bangladesh Prevalence, odds ratios (OR) and percentage of total burden of hypertension, hyperglycaemia and overweight among the poorest and richest 20% of men and women in Bangladesh, 2011 Prevalence, odds ratios (OR) and percentage of total burden of hypertension, hyperglycaemia and overweight among the poorest and richest 20% of men and women in Bangladesh, 2011 Of the 67 studies, 84% ( n = 56) were cross-sectional studies, 3% ( n = 2) were longitudinal studies and 12% ( n = 8) were non-systematic reviews or commentaries. The sample sizes of the studies varied widely from n = 32 to n = 3 651 921. In all, 15 studies included only urban populations, 11 studies included only rural populations and 36 studies included both urban and rural Bangladeshi populations. Among the published studies, the most commonly studied CVD/RF included measures of obesity ( n = 37 studies, 55%) and hypertension ( n = 36 studies, 54%), followed by smoking ( n = 28 studies, 42%) and diabetes ( n = 27 studies, 40%). A quarter of the published studies reported relationships with dietary factors ( n = 18 studies, 27%) and 11 studies (16%) reported results on cholesterol and physical activity. Of the 58 studies that reported on sex, 4/58 (7%) included only males, 10/58 (17%) studies included only females and, although 44/58 (76%) studies reported results on both males and females, only 16/44 studies (36%) stratified any results by sex. The quality of measurement varied between the studies, where 25/59 (42%) studies relied strictly on self-report of risk factors, 9/59 (15%) studies relied on trained personnel and/or laboratory tests and 25/59 (42%) studies used a combination of self-report and some measured outcomes. Important to note, whereas five studies looked at access to health care/ knowledge or some type of adherence to therapy, none of the studies differentiated between medically controlled versus medically treated risk factors and this important nuance is often missed, not only in epidemiological studies but also in clinical research. The proportion of individuals who are receiving medical treatment but are poorly controlled (their biomarker values are still not within normal ranges) are not differentiated in the outcomes of the studies reviewed. This discrepancy may be an issue of adherence, in combination with physician/patient perceptions, social beliefs, education and income, all important factors in LMICs. 11 Of the 67 studies analysed, although all studies mentioned SES, 14 (21%) did not report their results by SES. Of the remaining 53 studies, 30 (57%) found a positive association between higher prevalence of CVD/RF in higher SES groups. Among the studies that reported results by SES, positive SES-CVD/RF associations were found for obesity (94%, n = 29), hypertension (96%, n = 8), diabetes (95%, n = 21) and dyslipidaemia (100%, n = 3). Poor diet as a risk factor for CVD was seen consistently among lower SES groups (100%, n = 5). Negative CVD-SES associations, where higher SES groups report a lower prevalence of CVD/RF, were consistently reported with smoking/ tobacco use (95%, n = 19), although cigarette smoking specifically was greater among higher-SES groups. 12,13 Following Harshfield and colleagues, we examined the strength of SES-CVD associations according to urban/rural locality. Wide variability was found in the strength of the associations reported for CVD/RF and SES in urban compared to rural populations even of similar education and social class; however, this may be related to the quality of the sampling and design of the individual studies. For example, many of the prevalence estimates of CVD/RF were obtained from hospital/clinic populations, which inherently contain selection bias. Regardless of individual study nuances, with the exception of smoking and poor diet, this systematic review indicates that there is a robust association between greater levels of CVD/RF among higher-SES groups in Bangladesh, demonstrating that CVD/RF conditions disproportionally affect the socially advantaged. The findings of this review strongly support a previous study in India, which reported greater prevalence of CVD/RF and CVD among high-SES groups in India. 2 With this in mind, it is important to note that high SES groups represent overall a small proportion of the total population of Bangladesh. By contextualizing these findings we caution policy makers to resist ‘jumping the gun’ and that they do not shift focus from the public health issues that affect the broader population. 2 Concern has been raised over the anticipated increase in the prevalence of CVD/RF and NCDs in Bangladesh and other LMIC. 1 Evidence on the secular increases in such conditions, however, have typically focused on the mean rates at the country level rather than how CVD/RF are distributed within in the population, particularly along the SES dimension. There is a need to understand variations in the burden of disease within poor and rich segments of the population in order to inform health policies and target resource allocation to the proportional burden of disease on different population groups, to improve health equity. 7 We take this opportunity to expand upon findings related to the positive association between SESCVD/RF in Bangladesh. The study by Harshfield et al. clearly demonstrates a greater prevalence of hypertension and hyperglycaemia among the wealthiest groups in Bangladesh compared with the poorest, with the likelihood of having these risk factors two to four times greater among individuals in the highest wealth quintile compared with the lowest. Although this metric provides one extremely useful starting point to assess which groups are more likely to be affected by CVD/RF, it does not fully capture the extent to which the total burden of CVD/RF may be concentrated within certain population groups. 7 Using data from the BDHS, we illustrate a more complete metric which captures the extent to which the burden of CVD/RF falls among the richest or poorest 20% of the population. Among the poorest and richest 20% of men and women in Bangladesh, we present the prevalence, a measure of the number of existing cases at a given time, and the odds ratio (OR) indicating strength of association, comparing the proportion of cases among the poorest and richest for hypertension, hyperglycaemia and overweight in men and women ( Table 3 ). It is surprising that Harshfield and colleagues do not consider overweight/obesity in their analysis (although these data are also available in the BDHS) as overweight/obesity are well-established as both a ‘disease’ and a major CVD/RF. 79 Our analyses demonstrate that both the prevalence and the proportion of the total burden of the three CVD/RF were greater among the richest population groups in Bangladesh. The latter metric in particular is important for equity analyses and disease priority setting in this context, demonstrating, for example, that nearly half of the burden of hyperglycaemia and overweight is concentrated among the richest 20% of population, compared with less than 10% among the poorest 20% of the population. Priorities to address CVD/RF in settings such as Bangladesh would only benefit a minority of the poor, despite advocates suggesting that such disease and risk factors are increasingly concentrated among these groups. 2,7,8 Studies often report results by sex, yielding important information, but the richer SES information lies in accounting for gender and not just sex. While sex and gender are two terms often used interchangeably, the concepts they represent are not; sex is used to capture the biological differences between males and females, whereas gender is a construct that embodies the social roles, relationships, beliefs, norms, attitudes, behaviours, values and relative power that society ascribes to the two sexes on a differential basis. 80 Women, particularly in LMICs, have limited access to health care because of many of the components that constitute gender, such as cultural beliefs, socioeconomic circumstances, illiteracy, disempowerment, accessibility and negative cultural and religious customs, to name a few. 81 Women’s health has been primarily focused on reproduction and issues related to maternal health, 82 and considerably less attention has been focused on viewing their health and well-being from a more intrinsic perspective. 81 In many of the studies we reviewed, including the study by Harshfield et al ., risk factor patterns are relatively consistent across the socioeconomic measures of wealth and education in men, but these trends are not consistent in women, particularly along the education metric. For example, in the study by Harshfield et al ., there are no differences in household wealth between men and women in the study ( P = 0.548). However, over 55% of women and just over one-third of men (35%) have no education, compared with only 4% of women and 14% of men who have higher education ( P < 0.001). This discrepancy highlights an important and all too common gender difference, particularly among LMICs. The greater variation commonly seen among women in higher SES groups could be attributed to: artefact; an issue of power (often there are low numbers of women in higher SES groups); or bias within the development of the wealth metric, which is based on household assets and inherently incorporates aspects of men’s wealth, underestimating wealth in women, with no real measures to adequately capture household circumstances, social capital or area gender equity, factors important when measuring SES in women. 83 The measurement of gender factors in Bangladesh are particularly salient, as recent health reforms aimed to reduce gender inequity have been implemented as an impetus for health improvement. 84 Whereas important aspects of gender remain underexplored in health research, particularly in LMICs where social roles between men and women can vary greatly, there is often important information overlooked in the sex analyses as well. Study authors often report their findings adjusted according to age and sex but, beyond reporting sex differences, rarely are any insights explored. In the publicly available BDHS dataset used by Harshfield et al . and others, the average age of the population is relatively older (mean age ∼ 51 years) and it is difficult to interpret CVD risk factors findings among women without accounting for menopausal status which, although is possible to determine using data from the BDHS, was not considered by the authors. It has been well established from large epidemiological studies (mostly of HIC) that the prevalence of CVD risk factors such as hypertension and hyperglycaemia increase with age. However, CVD risk factors in women that manifest in the premenopausal stage are associated with a higher risk for adverse CVD events compared with risk factors that develop postmenopausally. 85 Although study authors may have adjusted the analysis by age, not stratifying or adjusting by menopausal status may mask the magnitude of the effect size of CVD/RF among younger women. Generally speaking, information about cardiovascular health in women in the developing world is lacking or neglected, 9,86 and among the few studies of women in LMICs they are reported to receive less appropriate management than men, with the greatest disparities seen among the lower-SES groups. 86 In addition to sex and gender differences in the manifestation of disease, there is a worldwide perception that cardiovascular disease is ‘a man’s disease’ and it is difficult to gauge how much this perception affects the health literacy of patients and healthcare providers alike. Health improvement efforts in Bangladesh are almost exclusively centred around issues of maternal health and reproduction, with great strides already achieved, 87,88 but the incidence of pre-eclampsia, a serious condition in pregnancy that characterized by hypertension, remains a concern (∼ 9%). 88 An association between women with pre-eclampsia and an increased risk of major adverse cardiovascular events later in life has been well established. 89 The example of pre-eclampsia also underscores the life-course aspects of health and disease, where the benefits of controlling hypertension during pregnancy may be important for securing maternal and child health in the short term and preventing CVD later in life. In summary, the findings by Harshfield and colleagues that the prevalences of hypertension and hyperglycaemia are substantially higher among high-SES groups in Bangladesh (a resource-poor country) are critical and need to be emphasized, as the influential research community and policy makers increase their support to focus on non-communicable diseases (NCDs). 90–92 Indeed, a focus on CVD/RF need not be pitched against the continuing agenda of communicable diseases in such settings; resource constraints are likely to force policy makers to prioritize and channel resources towards addressing disease burden that afflicts the majority of the population, and especially the poor. As was observed in India, 8 a public policy focus on addressing CVD/RF in Bangladesh will invariably benefit the high-SES groups at the expense of the low-SES groups and the poor. With 43.3% and 76.5% of population living on less than $1.25/day and $2/day, respectively, and with such burden perhaps falling more on women, Bangladesh needs to focus on disease burdens that are more common among this group. Conflict of interest : None of the authors have any conflicts of interest to report.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.