Adolescent BMI and early-onset type 2 diabetes among Ethiopian immigrants and their descendants: a nationwide study

This was a retrospective cohort study. At age 17 years, Israeli adolescents are approached to undergo a medical evaluation prior to mandatory military service. The study sample comprised all Israeli male and female adolescents who were medically evaluated for military service between ages 16–20 years from January 1, 1996 for men, and January 1, 1997 for women, until December 31, 2011. We excluded examinees with missing BMI data (n = 3,291, 2.6%) and those with a history of diabetes or dysglycemia at study entry (n = 325, 0.3%) that was solely based on reports by primary care physicians. We also excluded those who died before the establishment of the Israel National Diabetes Registry (INDR) in 2012 (n = 378; 0.3%) as these individuals could not be reported to the registry (See Additional file 1: Figure S1). The study sample included 121,490 examinees for whom we had continuous follow-up from adolescence until diabetes onset, death, or December 31st, 2016, whichever came first. Data of an additional 121,997 male adolescents who immigrated to Israel from former countries of the Union of Soviet Socialist Republics (USSR) during the same years as the immigration from Ethiopia were analyzed as part of sensitivity analyses. The Israel Defense Forces Medical Corps Institutional Review Board approved this study.

The primary outcome of the study was incident type 2 diabetes that was diagnosed either during military service, or later in life as recorded by the INDR. All health medical organizations in Israel have been requested by law to annually report prevalent cases of diabetes to the INDR since 2012. These organizations provide medical care to nearly 100% of permanent residents in Israel. The INDR retrospectively recorded the date of diabetes diagnosis for all individuals diagnosed in Israel between January 1, 2000 and December 31, 2011, and prospectively recorded all new incidences of diabetes from January 1, 2012 onwards. Data of the INDR were linked to the IDF database using the participants’ national ID number, to enable the linkage of body weight and height measured at adolescence with diabetes incidence recorded later in life. Diabetes was reported to the INDR when one or more of the following criteria were met in the previous year of the report to the registry: (i) glycated hemoglobin (HbA1c) ≥ 6.5%; (ii) serum glucose concentrations of ≥ 200 mg/dL in two tests performed at an interval of at least one month; (iii) ≥ 3 purchases of glucose lowering medications in different months. Data regarding prescribed diabetes medications were used to classify diabetes as type 1, type 2, or uncertain type, as reported previously [13] (for details see the Additional file 1: Appendix). The sensitivity of the INDR for the detection of type 2 diabetes is 95% [13] The INDR also includes weight and height measurements that were documented at routine clinic visits during the year of diabetes diagnosis. Measurements regarding weight and height at adulthood among those who did not develop diabetes were unavailable. Individuals who were reported to the INDR as having type 2 diabetes, but without a formal date of diagnosis were considered missing in the time-to-event analysis if they were first reported in the first year of the registry, 2012; and were included only in logistic regression models. For individuals without a formal date, who were first reported in 2013 or later, we assigned the date of diagnosis as July 1^st of the year before the report year to the INDR, as described previously [14].

Age at examination and year of birth were treated as continuous variables. The health examination was performed by military physicians who reviewed the participants’ medical records and provided diagnostic codes when applicable. The physical examination included measurements of weight and height (barefoot and in underwear) by trained medics using a beam balance and stadiometer. Measurements were recorded and rounded to the nearest 0.5 kg for weight and one cm for height. BMI was calculated (weight in kilograms divided by height squared in meters) and categorized into 3 subgroups: < 18.5 kg/m² (underweight), 18.5–24.9 kg/m² (normal BMI), and ≥ 25 kg/m² (high BMI). Because at age 17 adolescents have completed > 98% of their growth potential, we used the world health organization (WHO) BMI categories intended for adults [15]. Individuals whose medical assessment revealed no history of major operation or comorbidity and who had no illness that requires a chronic medical treatment were defined as having unimpaired health. Data regarding education and residential socioeconomic status (SES) were obtained from governmental ministries, and were used as previously reported [16]; details are provided in the Additional file 1: Appendix. Cognitive performance at adolescence was assessed by a general intelligence test that was delivered to all individuals at study entry, and was categorized (low/medium/high) as reported previously [17]. Notably, this variable was identified as an independent risk factor for early-onset type 2 diabetes in a subpopulation of this cohort [18]. Examinees were classified as of Ethiopian origin if they or their fathers were born in Ethiopia, whereas Israeli-born individuals whose fathers and grandfathers were born in Israel were classified as native Israelis [19]. All individuals of Ethiopian origin were either immigrants from Ethiopia or were born in Israel to immigrants from Ethiopia (second-generation). In a similar manner, we defined in a sensitivity analysis the region of origin of immigrants or offspring of immigrants from former countries of the USSR [20].

Prior to undertaking analysis, due to clinical plausibility, we decided to stratify the cohort by sex and adolescent BMI (underweight, normal, and high BMI). Analysis of the entire cohort demonstrated a significant interaction term between sex and origin (p = 0.038), and a borderline significant interaction term between BMI and origin (p = 0.086).

The incident rate of type 2 diabetes was calculated per person-year of follow-up for native Israelis and Israelis of Ethiopian origin per a given BMI group. Kaplan–Meier survival curves were computed for the BMI percentile categories, with 95% confidence intervals. Cox proportional hazard models were used to estimate the hazard ratios (HRs) and 95%-confidence intervals (CIs) for incident diabetes, considering native Israelis with normal BMI as the reference group. Covariates were added in a stepwise manner to a model adjusted for age and birth year. Variables that were significant (p < 0.05) in the minimally adjusted model were included in the final multivariable analysis. The assumption of proportionality of hazards was visually confirmed for all variables.

We conducted several sensitivity analyses that were limited to men, given the smaller sample of women in this cohort. (i) The analysis was restricted to those with unimpaired health status at enrollment in order to minimize residual confounding by co-existing morbidities. (ii) The analysis was restricted to second-generation, Israeli-born men of Ethiopian origin, using the same reference group as the main analysis. (iii) We assessed the risk for type 2 diabetes that was related to age of immigration among immigrants from Ethiopia with normal adolescent BMI. (iv) Diabetes risk of men from former countries of the USSR were compared to the main study groups, since immigration to Israel from this region occurred in parallel to immigration from Ethiopia to Israel; 1990–2000 vs. 1985–2000. To better assess the contribution of immigration status to diabetes risk, analysis was stratified to immigrants (USSR immigrants were the reference) and second-generation Israeli-born (native Israelis were the reference). (v) Logistic regression models were applied to account for cases with missing date of type 2 diabetes diagnosis. (vi) We assessed potential misclassification of type 2 diabetes by accounting for cases of diabetes of uncertain type, or by using a stricter definition of type 2 diabetes with a positive predicted value of over 98% [13], in which a case of type 2 diabetes is defined only if reported to the INDR in two separate years.

During 992,980 person-years there were 226 incident cases of type 2 diabetes in men: 89 (0.52%) in those of Ethiopian origin and 137 (0.22%) in native Israelis. The mean ages at diabetes diagnosis were 30.3 and 30.4 years, respectively (p = 0.9). The crude incidence rates for type 2 diabetes (per 10⁵ person-years) in Ethiopian vs. native Israeli men were 17.3 vs. 5.1 in the underweight group, 46.9 vs. 9.5 in the normal BMI group, and 112.3 vs. 62 in the high BMI group (Table 2). The mean BMI values that were measured at the time when diabetes diagnoses were first reported to the INDR were lower in Ethiopian than native Israeli men for the normal and high adolescent BMI categories (Additional file 1: Figure S2A). Most native (69.4%), but not Ethiopian (49.0%), Israelis were with overweight or obesity at diabetes diagnosis (p = 5.2*10^–7).

Kaplan–Meier survival analysis curves for men are shown in Fig. 1a, c, e. The unadjusted HRs for type 2 diabetes were consistently higher in men of Ethiopian origin than in native Israeli men (Table 2). Further adjustment for sociodemographic data (education level, residential SES status, and cognitive performance) yielded the following HRs in Ethiopian vs. native Israelis (reference group, native Israelis with normal BMI): 1.2 (0.6–2.3) vs. 0.5 (0.2–1.3) in underweight, 3.4 (2.3–5.1) vs. 1 (reference) in normal BMI, and 15.8 (8.3–30.3) vs. 6.9 (4.9–9.9) in the high BMI group. The results persisted when the study sample was limited to persons with unimpaired health at baseline to minimize confounding by pre- or coexisting morbidities (Table 2).

Point estimates for incident diabetes were directly related to the time interval from immigration. Point estimates were higher among those who immigrated to Israel before age 6 years (HR = 4.6, 95%CI = 2.6–8.2) than for those who immigrated after age 12 years (HR = 3.1, 95%CI = 1.6–5.9) (Additional file 1: Table S3). Also, the HR for diabetes among Israelis who were born in Ethiopia was lower than among second-generation Israelis of Ethiopian origin – 2.8 (2.0–3.9) vs. 4.0 (2.2–7.3) (native Israelis were the reference group; the model was adjusted for birth year, age at study entry, education level, cognitive score, and BMI (continuous); Additional file 1: Table S3). When the analysis was limited to men of Ethiopian origin who were born in Israel, the HRs were 4.4 (1.7–11.4) in the normal BMI group and 29.1 (12.0–70.6) in the high BMI group compared to native Israelis with normal BMI (Table 2). In contrast, risk for diabetes was comparable among Israeli-born men of USSR origin to that of native Israelis at any BMI stratum (Fig. 2a). For immigrants from Ethiopia, the adjusted HR to develop type 2 diabetes was 3.9 (2.8–5.5) compared to immigrants from the USSR (Fig. 2b). This risk was more pronounced among immigrants from Ethiopia with normal BMI at study entry [HR = 5.1 (3.4–7.8)] when immigrants from the USSR with normal BMI served as the reference (Fig. 2b).

The results were accentuated when logistic regression models were applied, and with the inclusion of incident cases of type 2 diabetes for whom the date of diagnosis was missing (n = 64; Additional file 1: Table S1). Odds ratios (ORs) were calculated for the development of type 2 diabetes, using native Israelis with normal BMI as the reference group. The adjusted odds ratio (OR) for Israelis of Ethiopian origin with normal BMI was 4.1 (2.9–5.9). Among individuals with high BMI, the OR for those of Ethiopian origin was 18.8 (10.5–33.6), and for native Israelis 7.4 (5.4–10.2); this difference between the culture groups was statistically significant (p = 0.028). Accounting for misclassification of diabetes diagnosis, either by including cases of uncertain diabetes type or by applying a stricter definition of type 2 diabetes with over 98% positive predicted value, yielded similar results (Additional file 1: Table S2).

During 530,814 person-years there were 79 cases of incident type 2 diabetes among women: 29 (0.27%) in those of Ethiopian origin and 50 (0.15%) in native Israelis. The mean ages at diabetes diagnosis were 28.3 and 27.3 years, respectively. Notably, high BMI among female adolescents of Ethiopian origin was rare, only 6 persons, in the first decade of the study. The crude incident rate for type 2 diabetes (per 10⁵ person-years) in Ethiopian vs. native Israeli women were 10.5 vs. 0 in the underweight group, 23.7 vs. 5.7 in the normal BMI group, and 54.0 vs. 55.8 in the high BMI group (Table 2). These differences were well apparent in Kaplan–Meier survival analysis (Fig. 1b, d, f). The BMI recorded at the time of entrance to the INDR had no consistent pattern. Women of Ethiopian origin who developed type 2 diabetes in the normal BMI group gained more weight than did their native counterparts, whereas the opposite was observed for the high BMI group (Additional file 1: Figure S2B). Among women with normal BMI in adolescence, the HR for type 2 diabetes was 4.2 (2.2–8.1) for those of Ethiopian origin compared to their native peers; and 2.4 (1.1–5.1) in the adjusted model. The latter was attenuated to borderline significance when adjusted logistic regression models were applied to include incident cases without the date of diabetes diagnosis (n = 47; OR = 1.8, 95%CI 0.98–3.2, p = 0.058; Additional file 1: Table S1). HRs of Ethiopian and native Israeli women with high adolescent BMI were comparable.