Ethnic differences in prediabetes incidence among immigrants to Canada: a population-based cohort study

The recognition of prediabetes in high-risk populations offers a potential window to intervene in the natural course of type 2 diabetes development. Clinical trials have demonstrated that structured lifestyle interventions leading to modest weight loss can prevent or delay the progression from prediabetes to diabetes [1]. Applying this evidence on a population scale poses significant challenges due to the growing number of individuals with prediabetes. According to the International Diabetes Federation, more than 318 million persons are estimated to have prediabetes worldwide and this figure is expected to rise to 482 million by 2040 in parallel with global trends in obesity [2]. Low- and middle-income countries are projected to have particularly sharp increases in prediabetes prevalence because of rapid changes in urbanization and lifestyle.

While several cohort studies have provided insight into the population prevalence of prediabetes, estimates have varied depending on the thresholds used to diagnose prediabetes and the underlying characteristics of the population [3‐10]. Recent studies from the United States (US) and United Kingdom (UK) place the prevalence of prediabetes among adults at ~ 38% and 35%, respectively, although higher estimates have been noted in older populations [6, 8]. Moreover, the prevalence of prediabetes appeared to increase nearly threefold between 2003 and 2011 across multiple ethnic groups in the UK and from 29 to 36% between earlier (1999–2002) and later (2007–2010) waves of the US National Health and Nutrition Examination Survey, affecting lower- and higher-risk ethnic groups alike [8, 9]. Similar increases were noted in Southeast and East Asia, including in Japan [11] and China [4, 12].

Few studies have examined the risk of progression from normoglycemia to prediabetes and those that have focused largely on older adults or a single ethnic group [13‐17]. To our knowledge, no studies have compared the incidence of prediabetes in higher- vs lower-risk ethnic groups or have examined the age at which ethnic differences emerge. The purpose of our study was to examine the incidence of prediabetes among immigrants to Canada of different ethnic origins and in relation to the general Canadian population. Furthermore, we investigated the extent to which the progression from normoglycemia to prediabetes is influenced by sociodemographic and immigration factors. We hypothesized that non-Europeans have a higher risk of progression compared to those of European descent, especially adults of South Asian and East Asian ancestries.

Overall, there were 1,772,180 adults with normoglycemia in our cohort—of whom 334,678 were immigrants who had arrived in Canada since 1985. Baseline characteristics by immigration status and ethnicity are outlined in Tables 1 and 2, respectively, with more detailed data provided in Additional file 1: Table S2. Overall, immigrants were younger at baseline (mean age 41.3 ± 12.8) compared to long-term Canadian residents (mean age 47.4 ± 15.4) and more likely to live in low-income neighborhoods. Among immigrants, South Asian and East Asians accounted for the largest number of individuals in our study. More than half of the immigrants in our cohort were married, had high school education or less, and had arrived more than 10 years prior to baseline. Additionally, about three quarters of immigrants arrived under the economic visa category or for the purpose of reuniting with family members who had previously migrated to Canada.

Over a median follow-up of 8.0 years, 66,085 immigrants and 271,523 long-term Canadian residents developed prediabetes. Over the period of follow-up, 129,710 (5.63%) died and 115,068 (4.99%) developed diabetes. Loss to follow-up during the observation period due to lack of eligibility for health care coverage in Ontario was 3.93% for the entire cohort, and 0.43% for immigrants and 3.5% for long-term Canadian residents.

Overall, prediabetes incidence rates were 40% higher among immigrants compared to long-term Canadian residents after adjusting for age, sex, and area income (Table 3). Over the entire follow-up period, the adjusted cumulative incidence of prediabetes was 21.2% among immigrants and 16.0% among long-term Canadian residents (Additional file 1: Figure S1).

Among immigrants to Canada, the crude and adjusted annual incidence rates of prediabetes were significantly higher for all non-European groups relative to European populations, after adjusting for sociodemographic and immigration-related covariates (Table 3), both overall and when analyzed by sex (Additional file 1: Figure S2). Over the 12-year follow-up period, the highest adjusted cumulative incidence of prediabetes was seen among people of South Asian, Sub-Saharan African and Caribbean, and Southeast Asian descent, with rates of conversion from normoglycemia to prediabetes of 23.6%, 20.3%, and 19.8%, respectively, based on WHO/DC criteria for defining prediabetes (Fig. 1). These figures were substantially higher relative to the Western European comparison group (South Asian: adjusted HR 1.95 (95% CI 1.87–2.03); Southeast Asian: HR 1.52 (95% CI 1.46–1.59)), among whom the adjusted 12-year cumulative incidence of prediabetes was only 13.1%, using the same criteria. The adjusted cumulative incidence of prediabetes was two- to threefold higher when using ADA thresholds, at 12 years ranging from 32.3% among Western Europeans (referent) to 47.1% (HR 1.64, CI 1.59–1.68) among South Asian immigrants (Additional file 1: Figure S3; Table 4).

Ethnic differences in prediabetes incidence emerged at young ages with similar patterns as observed in the overall population (Fig. 2). Incidence rates increased sharply with age but peaked by age 50–64 in high-risk populations with a little further rise in incidence with ages above 65. Among the highest risk group (South Asians), the 12-year cumulative incidence of prediabetes was 14.3% at 20–34 years, 31.1% at 35–49, 47.9% at 50–64 years, and 48.5% at 65 years and older (Fig. 2). Over this follow-up period, Western Europeans had consistently lower adjusted cumulative incidence rates, ranging from 3.3% among the 20–34 year age group to 38.1% among those aged 65 years and older. In comparison with the WHO thresholds, the risk of conversion to prediabetes according to the ADA thresholds was two- to nearly threefold higher (Fig. 3), ranging from 17.1 to 38.0% and 54.6 to 61.6% among the youngest and oldest groups, respectively. Furthermore, over the 12-year follow-up period, the cumulative incidence curves for prediabetes among South Asians immigrants aged 20–34 (14.3%) were equivalent to that of Western Europeans who were 15 years older (Figs. 2 and 3), regardless of which thresholds were used to define prediabetes (WHO/DC 14.4% vs 15.7%; ADA 38.0% vs 33.0%).

The above patterns were apparent for both men and women (Additional file 1: Figure S2); however, ethnic differences in prediabetes incidence were consistently greater among women. For example, prediabetes incidence was more than twofold higher among South Asian women compared to Western European women (HR 2.38, 95% CI 2.25–2.52), whereas the same comparison among men showed a lesser magnitude of difference across the groups (HR for South Asian vs Western European men 1.62, 95% CI 1.53–1.71).

Overall, 106,920 immigrants (31.9%) and 426,409 long-term residents (29.7%) did not have a second laboratory test for glucose screening during the observation window. The majority of those without follow-up test records were under age 50. Censoring on the date of the last laboratory test resulted in even higher cumulative prediabetes incidence probabilities (26.4% among immigrants and 24.7% among long-term residents at 10 years). This was particularly so among the younger age groups. For instance, the adjusted cumulative incidence for prediabetes among South Asians aged 20–34 years was considerably higher (24.9% at 10 years) when censoring on the last test than that generated when censoring at the end of the study period (Fig. 2).

This study demonstrates a strong association between ethnicity and the conversion from normoglycemia to prediabetes. Among immigrants to Canada, individuals of South Asian, Sub-Saharan African and Caribbean, and Southeast Asian descent experienced a much higher incidence of prediabetes in young to middle adulthood, relative to Western Europeans. Nearly one in three South Asians aged 35–49 and one in two aged 50 years and older developed prediabetes over a median follow-up of 8 years. When the ADA thresholds were applied, conversion rates to prediabetes were on average two to nearly threefold higher than when WHO thresholds were used. Furthermore, high-risk ethnic groups experienced a heightened risk of developing prediabetes at early ages. For instance, South Asians in their 20s and 30s had a similar incidence of prediabetes as Western Europeans that were 15 years older. These findings support a potential role for addressing upstream risk factors in high-risk ethnic groups prior to the development of prediabetes. In the absence of intensive lifestyle changes, a vast majority of individuals with prediabetes are expected to develop type 2 diabetes [15, 25, 26]. Thus, the early identification of prediabetes may help to reduce the burden of diabetes in high-risk populations by providing them with opportunities to adopt intensive lifestyle changes.

Although, previous studies have demonstrated clear ethnic differences in the predisposition towards metabolic diseases, [27, 28] few studies have looked at the early stages of type 2 diabetes development. Prospective studies that reported risk estimates for the development of prediabetes were marred by small sample sizes or focused on a single ethnic group [3, 4, 7‐17]. Our study adds to this literature by comparing prediabetes incidence in a large multi-ethnic population and revealed that adults who were of non-European descent, particularly South Asians, experienced a more rapid and earlier onset of prediabetes relative to Western Europeans. Previous studies have also demonstrated that South Asian migrants living in England [8], the USA [3, 12], and Canada [27] had a higher likelihood of having either prediabetes or diabetes, relative to other ethnicities. This finding could be explained by the differences in the underlying genetic predisposition towards type 2 diabetes. Furthermore, South Asian and Chinese populations develop insulin resistance and subsequent diabetes at a lower BMI [29]. Thus, these differences would likely translate into a higher incidence of prediabetes in these ethnic groups. Among many non-European countries, women have higher rates of obesity, which may explain why we found greater ethnic differences in prediabetes incidence among women. [30, 31] Gender differences in physical activity and diet may compound these effects. Among immigrants, women are also less likely to be employed which may lead to fewer opportunities for physical activity [32, 33].

The risk of prediabetes also rose with increasing time spent in Canada, possibly because of acculturation to a more Western lifestyle. In low- and middle-income countries, trends in urbanization and economic development have led to the adoption of Western diets, and sedentary behaviors [34], leading to higher rates of obesity and related diseases. This may contribute to the heightened susceptibility of immigrants to prediabetes prior to immigration. In addition to macroeconomic factors in the home countries of immigrants, post-migration factors likely also play an important role [34]. Psychosocial stressors, cultural barriers, unemployment, and economic instability may collectively contribute to ethnic variations in prediabetes incidence among immigrant populations through differential opportunities for physical activity and healthy eating. In particular, immigrants from lower-income groups are more likely to experience food insecurity, due to the high costs of healthy foods [35].

Our study has a few limitations to be noted. First, some misclassification of ethnicity may have occurred, since ethnic groups within a given country or world region may be quite heterogeneous. We supplemented information on country of origin using mother tongue and a validated surname algorithm; however, the latter could only identify individuals of South Asian and Chinese descent. Second, although validated algorithm based on diagnostic codes and prior laboratory tests were used to exclude potential diabetes cases, our data sources lacked information on medication use for those under 65; hence, some misclassification may have occurred in excluding younger members of the cohort with diabetes. Third, because we captured individuals who had glucose tests performed in a single commercial laboratory in Ontario, we may have underestimated the incidence of prediabetes since people without subsequent tests were assumed to be free of prediabetes. However, people who underwent blood glucose testing were likely at higher risk of prediabetes than those in the general population simply because a doctor screened them for the condition. Fourth, while we were able to account for numerous sociodemographic and immigration factors, we did not have information on BMI, family history, and other lifestyle measures. Therefore, we cannot fully assess the extent to which genetic versus non-genetic factors contributed to ethnic differences in the susceptibility towards prediabetes development.

To our knowledge, this was the first prospective, population-based analysis of prediabetes incidence in a large multi-ethnic population. Prediabetes has been shown to be associated with an increased risk of diabetes incidence, cardiovascular diseases, and all-cause mortality [36]. Given that high-risk populations are more susceptible to developing diabetes [27], the accentuated risk of prediabetes observed among high-risk ethnic groups at younger ages has significant implications for health care systems and societies as a whole. Identifying high-risk groups provides the opportunity to intervene earlier in the pathway to diabetes development through implementation of intensive lifestyle interventions. However, the lifestyle interventions tested in clinical trials have been poorly implemented in primary care settings due to the growing numbers of people living with prediabetes and limited infrastructure, resources, and coordination of diabetes prevention efforts [37]. As a result, further research on the upstream socioeconomic, environmental, political determinants of prediabetes is necessary [38]. Moreover, ethnic- and age-specific thresholds that predict the future risk of prediabetes are needed to help guide the decisions regarding which patients would benefit from early screening. As such, the integration of culturally competent and accessible screening and prevention programs are a critical step to reduce the burden of diabetes in immigrants from high-risk ethnic populations [39].