Introduction
Type 2 diabetes mellitus (T2DM) looms large over Asia. Asians, especially South Asians, are predisposed toward T2DM to a greater extent than ethnic Europeans.1 ,2 At even greater risk are ethnic Asians living in Europe or the Americas, where this predisposition is accentuated by the adoption of modern, urban lifestyles rich in processed, energy-dense foods and reduced physical activity. Examples abound: rates of T2DM are 1.3–4.8 times higher among American Asians and Pacific Islanders than in Europeans living in the Americas,3–6 2–3 times higher among American Japanese than in Japanese living in Japan,7–9 and 1.9–6 times higher among South Asians versus Europeans living in Europe.10 ,11 India (51 million) and China (43 million) already have more people with type 2 diabetes than the USA (27 million),12 but as lifestyles and diets in rapidly developing Asia become increasingly urbanized, it therefore must be expected that the burden of T2DM will continue to grow in the most populous continent.
Singapore is a microcosm of Asia. Three broad ethnicities, corresponding to the three major population centers in Asia, are represented in the city-state: East Asians, in the Chinese majority, South East Asians, via the Malay, and South Asians of mostly Indian and Sri Lankan descent. Over the past few decades, these groups have been exposed to significant changes in lifestyle, diet, and other environmental influences that are typical of a high-income society, changes that are reflected in the doubling of the prevalence of T2DM from 5% in the 1980s13 to 11% in 2010.14 Rapidly ageing, increasingly sedentary, Singapore presages the problems other Asian countries will face in the decades ahead.
Since T2DM is one of many competing public health issues that will accompany the ageing of Singapore, as in Asia, it is vital to be able to forecast the future burden of T2DM to facilitate rational planning of public health campaigns. To predict involves positing a model that encapsulates epidemiological and medical subject-area expertise on the main drivers of T2DM at the individual and population levels. Rigorous subsequent parameterization of the model ensures its relevance to the population to which it is applied. The degree of complexity of the model depends on the objective of the analysis and the data available: neither too simplistic, lest it fail in extrapolation to scenarios it was not validated for; nor more complex than is needed to meet those objectives. Methods used in other settings to forecast the evolving burden of chronic diseases include microsimulation models positing assumptions on future obesity and physical activity trends,15 extrapolating linear regressions of the prevalence of overweight and adjusting geographic distribution using deprivation indices,16 forecasting the changing demography of a country with or without increasing incidence,17–20 and modeling body mass index (BMI) and its impact on development of T2DM and related complications.21 ,22
One of the most challenging issues in developing a model for a future public health phenomenon is that the health of a population is never in a stable equilibrium. Although the observed rise in T2DM prevalence from 8.6% (95% CI 7.7% to 9.6%) in 1992 to 11.3% (95% CI 10.3% to 12.3%) in 2010 in Singapore can be attributed to ageing, as the age-specific prevalence has remained relatively static since the 1990s,14 it would be misleading to forecast the future prevalence of T2DM by applying the historic age-specific prevalence of T2DM to a projected age distribution at some future time point—for the age-specific prevalence of obesity, and overweight, another important risk factor of T2DM,23 have risen substantially in most demographic segments over this time period.24 This rise foreshadows an increase in the age-specific prevalence of T2DM, as the increasingly obese young of today become the increasingly diabetic old of tomorrow. Predictions must hence incorporate ageing and secular trends in obesity, reflecting changes in diet and physical activity, as otherwise they may severely understate the future burden. Furthermore, evidence of a genetic contribution to T2DM from familial aggregation (the risk of T2DM increases twofold to fivefold for individuals having a family history,25 ,26 while heritability of T2DM—the proportion of phenotypic variance attributed to genetic factors—has been estimated at approximately 26% in a Danish population-based twin study27) suggests that the effect of genetics should also be incorporated.
This paper describes a demographic, epidemiological model of Singapore and its use in forecasting the total prevalence of T2DM (diagnosed and undiagnosed) to 2050. The model is an individual-based model which represents each resident in the city-state, past (from 1990) and future (to 2050), thereby facilitating the incorporation of obesity trends, both secular and over an individual's life span. The model incorporates demographic processes including the mass migration Singapore has experienced over the past two decades, submodels for the evolution of each individual's yearly BMI and genetic risk of T2DM, and a T2DM onset submodel, and data from national statistics, nationally representative cross-sectional surveys, longitudinal studies, molecular epidemiological cohort studies, and the literature, analyzed using Bayesian statistical methods.