Article Text
Statistics from Altmetric.com
The concept of the ‘social cross-over’ in cardiovascular disease (CVD) mortality was first described in 1978 by Marmot et al.1 By analysing mortality trends between 1931 and 1971 in England and Wales, Marmot et al found that the pattern of mortality from coronary heart disease changed over the years, from being more common among middle-class and upper-class men and women to being more common among working-class men and women.1 Underlying this change was a steep increase in CVD mortality among the working-class group starting in the 1950s for men and in the 1960s for women, concurrent with increases in smoking rates and dietary changes.1
In the study by Mallinson and colleagues, the authors describe the absence of a social cross-over in CVD mortality in Brazil.2 The study used cross-sectional data for 2010 from states at various degrees of economic development to proxy the changes in economic development experienced by high-income countries over time. Mallinson et al found no support for the social cross-over hypothesis. Instead, they found that education-based inequalities in CVD mortality were actually widest in the least developed states, meaning that CVD mortality was highest among the population with the lowest levels of educational attainment.2
Previous studies on the social cross-over have relied on longitudinal samples to examine how inequalities in CVD change over time. Mallinson et al instead leverage intracountry heterogeneity in social development. The use of cross-sectional data from one country has two important advantages: (1) uniformity of the vital statistics system across the country, which reduces inconsistencies in the quality of the data often present in cross-country comparisons, and (2) consistency of coding, one of the biggest issues in studies using historical data.2 The authors also addressed issues of variation in the quality of data across states and examined the impact of potential measurement biases in the results. However, using cross-sectional data comes with its own problems. In this piece, we will discuss the results from Mallinson et al in the context of changing patterns of CVD mortality in low-income and middle-income countries (LMICs) and their implications on clinical care and population health in LMICs.
The original social cross-over paper by Marmot et al 1 used data for a historical period marked by consistent increases in CVD mortality, following a previous transition from high mortality due to infectious diseases to high mortality due to non-communicable diseases. This transition, also experienced by other high-income countries, was followed by rapid declines in CVD mortality, leading to a subsequent continued (although slowed, compared with previous decades) growth in life expectancy.3 In addition to the overall decline in CVD mortality, high-income countries also presented a progressively strong social gradient in the distribution of CVD mortality in the 1970s and 1980s, with higher rates of CVD mortality among socially disadvantaged groups compared with advantaged groups.4 In summary, the phenomenon described as social cross-over of CVD mortality occurred in a specific historical context in high-income countries.
However, the patterns of mortality observed in high-income countries do not seem to be a good proxy to explain present patterns of CVD mortality and CVD inequalities in LMICs, which currently account for three in four CVD deaths worldwide.5 The idea of a clear transition from high burden of communicable diseases to high burden of non-communicable diseases is unlikely to be seen in LMICs.5 Many LMICs, including Brazil, have experienced demographic changes such as rapid urbanisation and population ageing, which contributed to a higher burden of non-communicable diseases.5 This higher burden has occurred in (1) populations in a different baseline status, as compared with high-income countries, including high levels of malnutrition and stunting, creating a double burden of disease, and (2) a historical context marked by the globalisation of economic activity and a consistent increase in the availability of cheap, poorly nutritious calories.5 In Brazil specifically, ischaemic heart disease and stroke have been the leading cause of death for the past 50 years. At the same time, income inequality and poor infrastructure in large cities have allowed the persistence of communicable, maternal, neonatal and nutritional disease despite expansion of primary care, improvements in vaccine coverage and reductions in infant mortality thorough the country.6 These improvements have occurred concurrently with the increased availability of ultraprocessed foods through multinational food corporations.5
In this regard, the study by Mallinson et al cannot account for these broad changes that happen over time as they used cross-sectional data to mimic a historical design: the authors compared CVD mortality across Brazilian states with different levels of economic development based on the Human Development Index (HDI), relying on the assumption that a cross-sectional comparison of states at different levels of development provides information on longitudinal changes in development. To give a concrete example, the analysis uses CVD mortality in Alagoas, a low-HDI state, as a counterfactual for CVD mortality in São Paulo, a high-HDI state, as if Alagoas in 2010 was a good proxy for São Paulo around 1995, when the HDI in São Paulo matched that of Alagoas in 2010 (figure 1). One challenge of this strategy is that a number of factors affecting CVD mortality occur in all states concurrently, regardless of the economic development, including issues linked to globalisation (eg, the role of multinational tobacco or food/agriculture corporations), and public health policies implemented at the national level, such as a robust set of tobacco control policies between 1989 and 2010.7 In this context, Alagoas in 2010 may not be a good counterfactual for São Paulo in 1995.
Differences across countries and across geographies and socioeconomic groups within countries can be the result of variations in underlying CVD risk factors and differences in access to healthcare for primary and secondary CVD prevention.3 5 In addition to differential exposure to CVD risk factors, socioeconomic disadvantaged groups may also be more vulnerable to rapid changes in society, such as rapid increase in the availability of highly processed foods.5 The social cross-over hypothesis assumes that low-socioeconomic status (SES) populations in early stages of development are less exposed to CVD risk factors and that, over time, as economic development advances, low-SES populations become more exposed to CVD risk factors. However, data from Brazilian state capitals in 2010 show that low-SES populations in less developed cities are not less likely to be exposed to CVD risk factors, compared with those in more developed cities (table 1).
Table 1 shows that women and men with less than 9 years of education had higher prevalence of several risk factors, including obesity, smoking, and unhealthy dietary behaviours, compared with those with 12 or more years of education regardless of whether they lived in São Paulo City (capital of São Paulo, high HDI) or Maceió (capital of Alagoas, low HDI). Large education inequalities in prevalence of risk factors are present, particularly for current smoking among women (11.7% among women with less than 9 years of education vs 3.9% among those with 12 or more years of education), and past smoking among men (36.4% vs 13.2%) and women (19.9% vs 9.9%) in Maceió. High prevalence of CVD risk factors in the group with less than 9 years of education in a state with low HDI demonstrates that this group is not protected from CVD risk factors as the social cross-over hypothesis would suggest.
The clinical and policy implications of understanding social gradients in CVD mortality and underlying risk factors are many. First, the increasing evidence of a social gradient in CVD mortality strengthens the argument that CVD prevention interventions should not be socially neutral and should rather focus on low SES populations. Second, the strong social patterning of CDV mortality observed in Brazil and potentially present in other LMICs makes the case for population-level policies as a preferred strategy to reduce inequalities. Previous evidence in the UK has shown that individual clinical interventions tend to widen inequities, while population-level policies tend to narrow them.8 Third, understanding social gradients in CVD is important to set a clear narrative, particularly around differential vulnerability to diseases across SES groups and which groups benefit from advances in diagnostic and treatment. We have recently witnessed the consequences of differential vulnerability across SES groups, with the COVID-19 pandemic being initially thought to transcend SES groups (as the wealthy had improved access to testing, in yet another instance of measurement error), and with a delay in understanding how low SES populations may be differentially harmed.
In summary, the study by Mallinson et al adds to the literature that challenges the validity and the generalisability of the social cross-over of CVD mortality, which was first described in a specific historical context in high-income countries. Inequalities in CVD mortality and CVD risk factors in Brazil demonstrate that low SES groups are indeed at a greater risk compared with high SES groups. CVD prevention efforts should focus on low-SES populations, that continue to bear the burden of higher CVD rates.
Footnotes
Twitter @usama_bilal
Contributors PHM conceived the commentary with support from UB. PHM and UB conducted the statistical analyses. PHM drafted the first version of the manuscript with support from UB. Both authors participated in the interpretation of results and approved the final version of the manuscript.
Funding PHM and UB were funded with support from the National Institutes of Health (under grant DP5OD26429). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.
Provenance and peer review Commissioned; externally peer reviewed.
Linked Articles
- Healthcare delivery, economics and global health