Progress in diabetes prevention or epidemiology—or both, or neither?

After decades of negative reports, in The Lancet Diabetes & Endocrinology, Dianna Magliano and colleagues¹ report encouraging signs that, among the 24 datasets they studied (of which 22 were from high-income countries or regions), comprising 5 billion person-years, the incidence of diagnosed diabetes might have declined in recent years in the majority of data sources. However, numerous questions remain. In particular, what might be driving these trends? Were these declines the result of interventions and policies designed to reduce the incidence of type 2 diabetes, or were they data artifacts? Which segments of these populations were represented in these data? And what of the global context of diabetes: given the increases in type 2 diabetes in low-income and middle-income countries, where do we go from here?

The investigators compiled an immense volume of data from national and subnational sources and used sophisticated analytics. They advanced diabetes epidemiology by sourcing data from registries, insurance and administrative claims, large health systems, and a national study, providing empirical rather than modelled estimates. To account for the heterogeneity of data types and definitions of diabetes used across countries, they reported each country’s incidence trajectories over time separately. They also did sensitivity analyses in which they stratified findings by diabetes definition and by type of data source, restricted findings to data sources reporting exclusively type 2 diabetes, and excluded data sources with a poor quality score.

The study’s findings are limited to people who had accessed health care or those who had administrative records about their health. However, if access or documentation were differentially representative of demographic, small geographical area, or socioeconomic segments of these populations, the findings might be biased. For example, more than three-quarters of these data were administrative, which might exclude lower socioeconomic and uninsured or underinsured populations. This concern is of particular importance because the epidemiological transition in high-income countries has meant a shifting of chronic disease burdens from high to low socioeconomic populations.² As an illustration of this, the countries experiencing economic and demographic transitions with representative data that included all segments of their populations such as Latvia, Lithuania, Russia, Singapore, and Ukraine all experienced upward or flat incidence trends.

Macroepidemiological findings such as these depend on the validity of micro-level data collection and documentation. In surveys, for example, the interaction between interviewers and respondents and the recall, social desirability, and documentation of responses could affect findings. With registries and administrative data, in addition to differences in diagnostic practices, we have to rely on faithful coding by clinicians of what happened in their interactions with patients. An understanding of what regulatory, clinical, or social policies or variations in practice affect the data are valuable in how we interpret the findings and also in how we use the data to assess progress.

Despite the reported declines in the rate of new cases of diagnosed diabetes, the prevalence and absolute numbers of people with diabetes globally continue to increase annually.^3,4 Notably, people with undiagnosed diabetes were not counted in this study; in the countries included, conservative estimates suggest that 38% of those with diabetes are undiagnosed.³ As such, is incidence of diagnosed disease a feasible and robust indicator to measure progress in tackling diabetes globally? For example, could these trends in incidence reflect natural or temporary fluctuations in the rate of new diabetes diagnoses? Alternatively, the observed trends might be signalling the effects of the lower diagnostic thresholds introduced in 1997, which could have resulted in increases in diabetes incidence to the point of saturation by 2005–10 and declines thereafter. Similarly, the increasing use of HbA_1c for diagnosis worldwide, which has differential sensitivity and specificity within and across populations, could have contributed to artifactual effects on incidence. To their credit, the authors attempted to address this with exploratory analyses.

Importantly, the findings are limited to mostly high-income populations and cannot speak to diabetes incidence patterns in low-income and middle-income countries, where 79% of people with diabetes worldwide reside.³ Surveillance systems to monitor diabetes incidence do not exist in low-income and middle-income countries, and empirical estimates related to diabetes are limited to scarce clinical or local epidemiological cohorts. The few estimates that do exist suggest that diabetes incidence is high.^5,6 In effect, this finding might signal stark and troubling disparities across countries whereby high-income countries are experiencing declines while diabetes burdens (including incidence) are intensifying in low-income and middle-income countries.³

In summary, the investigators demonstrate apparent downward trends in diabetes incidence from 22 million diagnoses and provocatively suggest that this finding might reflect successes in the global fight against diabetes. However, it would be a big leap of faith to attribute changes in incidence at the population level, even if real, to diabetes prevention efforts, without more rigorous examination of causality. Furthermore, the current and projected increase in overweight and obesity globally, which are major drivers of type 2 diabetes, represent a more dire prediction of future trends in metabolic disease.^7,8 Questions regarding trends in key sociodemographic groups and in low-income and middle-income countries are crucial and merit imminent attention and study. Since absolute numbers of people with diabetes are already overwhelming health systems, especially in low-income and middle-income countries, this is no time for complacency. Perhaps Magliano and colleagues’ biggest contribution will be to encourage greater investments in data systems globally—especially in low-income and middle-income countries that bear the brunt of diabetes burdens and have very fragmented health and social services—both to prompt action and to track effects with more certainty, thereby progressing the field.