Conclusions
In this study of middle-aged and older adults followed between 2010 and 2020 from population-based cohorts in five economically and geographically diverse countries (three of which were nationally representative), we found that people with diabetes consistently had higher all-cause mortality than people without diabetes. Relative mortality differences were similar across cohorts, ranging from mortality rate ratios of 1.53 (95% CI: 1.39 to 1.68) in the USA to 2.02 (95% CI: 1.34 to 3.06) in Mexico. Absolute mortality differences had more variation across cohorts, ranging from mortality rate differences (per 1000 person-years) of 11.9 (95% CI: 4.8 to 18.9) in England to 24.6 (95% CI: 12.2 to 37.0) in South Africa. These findings using recent and comparable data highlight the immense burden of diabetes around the world, particularly in low-income and middle-income countries (represented in our study by China, South Africa, and Mexico), where the absolute mortality impact of diabetes appears greatest. These are also settings where diabetes care is thought to be least robust.3 27 28
Many prior studies assessing the association between diabetes and all-cause mortality have been conducted in high-income countries and among younger age groups.3 7 8 Large-scale meta-analyses in the last two decades have reported relative mortality differences among people with diabetes, as compared with those without diabetes, that are generally similar to findings in our study.4 6 29 30 However, these meta-analyses primarily included non-representative cohorts from high-income countries, limiting population inferences globally. A multicountry analysis from 1995 to 2016 in 16 countries provides updated evidence of a reduction in all-cause mortality among people with diagnosed diabetes, but data were only available from high-income countries.8 The Prospective Urban Rural Epidemiology study reported greater absolute mortality among people with diabetes in middle-income and low-income countries, as compared with people with diabetes in high-income countries.5 While studies on diabetes-related mortality previously have been performed in each of the countries included in our analysis, including at times using the same underlying cohorts,31–37 our study uniquely assesses diabetes-related mortality in multiple countries using similar methods across the entire continuum of middle-aged and older adults. Individuals in this age range are sometimes excluded from population-based studies worldwide. However, they have the highest diabetes prevalence and require comprehensive clinical management to prevent diabetes complications.
An important secondary finding in our study was the tendency of higher mortality among people with diagnosed diabetes compared with undiagnosed diabetes. This finding was most marked in Mexico. Many—though not all—prior high-quality population-based studies have reported similar findings.5 31–33 38 We hypothesize that the greater mortality among people with previously diagnosed compared with undiagnosed diabetes likely reflects a selection effect related to diabetes severity and/or diabetes duration. Patients with diabetes with the highest disease severity or progression are most likely to experience symptoms, to seek a diagnosis in the healthcare system, and, despite obtaining a diagnosis, to die. This selection effect may be most salient in countries at lower income levels, where the proportion of adults with diabetes who are diagnosed is as low as 20%, compared with 80% or greater in some high-income countries such as the USA.3
What are the policy implications emerging from this work? We speculate that the higher absolute mortality rates in South Africa and Mexico are a result of people with diabetes in these countries experiencing challenges accessing quality diabetes care34 39 and being impacted by the broader social determinants of health and diabetes.40 There is an urgent need to scale up evidence-based interventions to manage diabetes, particularly in low-income and middle-income countries where societies are aging, absolute diabetes mortality is highest, and the population with diabetes is rapidly growing.2 Evidence from Sweden shows that people with diabetes who are appropriately managed and achieve risk factor control have little or no excess mortality compared with those without diabetes.41 Yet only 10% of people with diabetes in low-income and middle-income countries receive comprehensive diabetes management aligned with guidelines.27 In the coming decades, diabetes will cause a staggering degree of premature mortality unless health systems are strengthened to improve diabetes care.1 The WHO Global Diabetes Compact is a crucial international effort to stimulate improvements in equitable, affordable, and quality care for people with diabetes.1 3 A key pillar of these efforts is the inclusion of stakeholders from the public and private sectors, as well as individuals with lived experiences of diabetes.
Our study has several limitations. First, our analysis did not include people aged 50 years or younger. The younger population with diabetes tends to have a greater hazard of diabetes mortality than the older population without diabetes.31–33 Our results should not be generalized to the entire population or young population. Still, they can be generalized to the population aged 51 years or older, which represents approximately two-thirds of people with diabetes worldwide.2 Second, our use of Poisson models in the main analysis assumes that an individual’s hazard of dying remains constant throughout the observation period, which ranged from a median of 4.6–8.3 years across the five cohorts. We chose this approach because Poisson models allow us to estimate and compare absolute mortality rates directly, expressed as events per person-time, while adjusting for covariates. Furthermore, sensitivity analyses that relax the constant mortality assumption by using Cox and Gompertz models with age as the time scale yielded results consistent with our primary analysis. Third, differences in the blood-based diabetes biomarkers collected in each cohort (eg, glucose vs HbA1c) may contribute to slightly different phenotypes of individuals classified as having undiagnosed diabetes.42 43 This limitation could decrease the comparability of estimates across cohorts. As an example of this dynamic, studies in Asian Indians suggest that HbA1c-based diabetes diagnoses may identify individuals with milder glucose intolerance, potentially reflecting less severe disease and lower associated mortality.44 Fourth, our study lacks data on cause-specific mortality, preventing us from distinguishing between microvascular and macrovascular patterns of death among individuals with diabetes. Fifth, the Mexican and South African cohorts were not nationally representative, though they were representative of four states in Mexico and a rural community in South Africa like many others in Southern Africa, respectively. Sixth, available cohort data do not allow us to distinguish between type 1 versus type 2 diabetes. Given the age profile of the cohorts, it can be assumed that the vast majority of individuals have type 2 diabetes.1 Finally, while this analysis used data from a geographically and economically diverse set of countries, the included cohorts may not fully represent population with diabetes worldwide. In particular, none of the cohorts were drawn from low-income or lower-middle-income countries. Estimating diabetes mortality in these settings is an important area of future research.
In summary, we observed that diabetes was consistently associated with increased all-cause mortality across five diverse settings, and absolute diabetes mortality was particularly high in low-income and middle-income countries, where systems of care for diabetes are known to be weaker. The findings reinforce the need to implement clinical and public health interventions to improve diabetes outcomes in countries worldwide.
Leave a Reply