Natural History of Type 2 Diabetes in Indians: Time to Progression

Indian people are at heightened risk of type 2 diabetes (1,2). However, it is unclear how Indians transition through the natural history of diabetes (e.g., from normoglycemia to prediabetes [impaired fasting glucose [IFG] or impaired glucose tolerance [IGT]), reversal to normoglycemia, and from prediabetes to diabetes), and how long people reside in each state (sojourn time).

We used data from the Chennai site of the longitudinal Centre for Cardiometabolic Risk Reduction in South Asia (CARRS) study (2010–2012) (3,4), up to the fourth follow-up (2016–2017) and implemented a continuous-time Markov model (5).

Glycemic states (normoglycemia, IGT, iIFG, diabetes) were defined according to American Diabetes Association criteria, and distributions and characteristics of each glycemic state and the rates of progression from one stage to the next were obtained from the CARRS study (3,4) (details of definitions and cohort are provided in the Supplementary Material) (6).

Among 5,961 participants with glucose measurements at baseline, 3,475 were free of diabetes. Of these, 2,714 participants had had a complete oral glucose tolerance test and at least one follow-up assessment before 2017 to estimate changes in glycemia. The characteristics of the 2,714 participants included in the final analyses (Supplementary Fig. 3) were similar to those of the overall sample (Supplementary Table 1). Because only 187 people had combined IFG and IGT, they were classified as having IGT.

We used multistate Markov models (7,8) to calculate annual transition probabilities for each state specified in Supplementary Fig. 1. The mean sojourn time was also calculated. For every participant, the outcome of interest was iIFG, IGT, or diabetes. Time was estimated from date of interview to the time of outcome diagnosis, last date of visit, or death, whichever came first. We fitted two models: 1) normoglycemia to iIFG and regression to normoglycemia or progression to diabetes (i.e., normoglycemia ↔ iIFG → diabetes); and 2) normoglycemia to IGT and regression to normoglycemia or progression to diabetes (normoglycemia ↔ IGT → diabetes). In the base case analysis, we assumed bidirectional change in states, allowing regression from prediabetes to normoglycemia (Supplementary Fig. 1). In a sensitivity analysis, we examined unidirectional progression, which assumes people cannot move back from prediabetes to normoglycemia (Supplementary Fig. 2). We performed stratified analyses by age (≤40 years vs. >40 years), sex, and BMI (<23 kg/m² vs. ≥23 kg/m²) (9) to estimate the annual transition probabilities for each set of models. The data were analyzed using the msm package in R software, version 3.2.4, and Stata 16.0/MP.

As shown in Table 1, iIFG was nearly twice as frequent as IGT, and iIFG was more frequent in women (71.5%), and IGT in men (53.1%). Those with normoglycemia, followed by iIFG, and then IGT, had the lowest mean age (normoglycemia vs. iIFG vs. IGT: 37.6 vs. 40.9 vs. 43.7 years, respectively), weight (61.6 vs. 65.5 vs. 66.3 kg, respectively), BMI (25.0 vs. 27.2 vs. 27.0 kg/m², respectively), waist circumference (81.6 vs. 85.7 vs. 88.1 cm, respectively), triglyceride levels (107.0 vs. 124.0 vs. 137.0 mg/dL, respectively), and total cholesterol level (180.4 vs. 187.4 vs. 189.9 mg/dL, respectively). The insulin levels at 0, 30, 120 min (iIFG: 7.7, 48.2, and 40.0 vs. IGT: 8.5, 53.1, and 59.6, respectively) and HOMA-β (iIFG vs IGT: 69.3 vs. 91.5 μIU/mL/mmol/L, respectively) were lower in those with iIFG compared with participants with IGT, thus indicating iIFG is a more insulin-deficient state of prediabetes.

The estimated mean annual probability of remaining in normoglycemia and in iIFG were 92.1% (95% CI 91.2, 92.9) and 68.6% (65.2, 71.8), respectively (Table 2). The annual probability of conversion from normoglycemia to iIFG was 7.5% (6.7, 8.3) and from iIFG to normoglycemia was 22.8% (19.6, 26.6). Thus, it is about three times more likely for iIFG to revert to normoglycemia than it is for normoglycemia to progress to iIFG. The annual probability of conversion from iIFG to diabetes was 8.6% (5.7, 11.9). The estimated mean sojourn times were 40.3 (34.6, 48.2) years and 9.7 (8.4, 11.4) years in normoglycemia and iIFG, respectively (Fig. 1).

The annual probability of transition for normoglycemia to iIFG was greater for those age >40 (9.9% [95% CI 8.1, 11.9]) years compared with those age ≤40 (6.4% [5.5, 7.4]) years (Table 2). The annual probability of transition from iIFG to diabetes was higher in female participants than male participants (female: 7.9% [6.8, 9.0]; male: 6.8% [5.6, 8.1]). The annual probability of transition from normoglycemia to iIFG was greater for those with BMI ≥23 kg/m² (8.9% [7.7, 10.2]) compared with BMI <23 kg/m² (5.5% [4.1, 7.1]).

The estimated annual probability of remaining in normoglycemia and IGT were 94.5% (95% CI 93.8, 95.2) and 70.1% (67.1, 73.8), respectively (Table 3). The annual probability of transition from normoglycemia to IGT was 5.1% (4.4, 5.7) and from IGT to normoglycemia was 15.4% (11.9, 19.4). Thus, it is about three times more likely for IGT to revert to normoglycemia than it is for normoglycemia to progress to IGT. The annual probability of transition from IGT to diabetes was 13.9% (12.0, 15.9). The estimated mean sojourn times were 34.5 (29.5, 40.8) years and 6.1 (5.3, 7.1) years for normoglycemia and IGT, respectively (Fig. 1).

The annual probability of transition from normoglycemia to IGT was higher among those aged >40 years (7.6% [95% CI 6.2, 9.1]) as compared with those aged ≤40 years (3.9% [3.3, 4.7]) (Table 3). The annual transition probability from normoglycemia to IGT was higher in men (6.8% [5.7, 8.0]) than women (3.9% [3.1, 4.6]). The annual probability of transition from normoglycemia to IGT was higher among those with BMI ≥23 kg/m² (6.2% [5.2, 7.1]) as compared with those with BMI <23 kg/m² (3.5% [2.6, 4.5]).

In a sensitivity analysis, the probabilities of transition based on unidirectional models (i.e., not allowing regression from prediabetes to normoglycemia) were similar to those of base case bidirectional models (Supplementary Tables 1–4).

In an urban Indian population aged ≥20 years, progression to diabetes is rapid once an individual has prediabetes. On average, people reside 35–40 years in normoglycemic states, and only 9.7 years in iIFG or 6.1 years in IGT before advancing to diabetes (assuming bidirectional transition from normoglycemia to prediabetes). Prediabetes represents a fragile state, with a nearly three times likelihood of either iIFG or IGT reverting to normoglycemia than normoglycemia progressing to prediabetes. However, at the onset of prediabetes, and after accounting for reversibility, the rate of progression from prediabetes to diabetes was rapid, and those with IGT progressed to diabetes faster (13.9% per annum) than those with iIFG (8.6% per annum).

Similar to our findings, several studies have reported a high incidence of diabetes and prediabetes in Indians (1,10–12). In addition, we found a higher rate of conversion from normoglycemia to iIFG than to IGT, suggesting reduced insulin secretion (lower HOMA of β-cell function [HOMA-β]) as an early defect (13). However, among those with prediabetes, those with IGT had a more rapid conversion to diabetes, suggesting poorer insulin sensitivity as a key factor at later stages in those already susceptible, and iIFG and IGT being potentially different phenotypes with differences in pathophysiology (13–15). Similar to previous reports (16), we also found that iIFG is the more frequent (almost two-thirds) prediabetes manifestation in Indians, iIFG is the common phenotype in women, and IGT the more common phenotype in men. Although lifestyle interventions are effective in reducing the incidence of diabetes among adults with IGT (17,18), these interventions seem not effective in individuals with iIFG (19).

The strengths of our study include data from a representative sample, high response and retention rates, multiple time points of follow-up, and objective measures of glycemia derived from three-step oral glucose tolerance tests. To our knowledge, no previous study of diabetes in Indians has estimated time to progression or time spent in each glycemic state. Given the fragility of the prediabetes state, we conservatively assumed bidirectional transition from normoglycemia to iIFG or IGT (in separate models), but we also performed sensitivity analyses to explore the effect of unidirectional transitions and other assumptions. The results for transition probabilities across states were robust regardless of assumption of bidirectional or unidirectional progression, but the estimates of time in each state were substantially longer under assumption of bidirectionality. Last, we performed stratified analyses by age, sex, and BMI. Our study has some limitations, including that data are from one city; however, diabetes incidence across urban India is quite similar, and the sample reflects the age and sociodemographic distribution of populations of cities in India.

In conclusion, we found a high rate of conversion from normoglycemia to IFG or IGT in Indians, and once an individual has prediabetes, the conversion to diabetes is rapid. On the hopeful side, people at risk for diabetes reside in normoglycemic states for an average of 35–40 years, and those transitioning through the more frequent iIFG stage reside there for an average of 9.7 years (as opposed to 6.1 years in IGT). These findings suggest the need to test interventions to prevent the occurrence of prediabetes.

Acknowledgments. The authors thank the staff and participants of the CARRS for their important contributions.

Funding. The CARRS Study was funded in part by the National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health (NIH), Department of Health and Human Services (contract HHSN268200900026C and grant PO1HL154996); and the United Health Group (Minneapolis, MN). K.M.V.N., M.K.A., U.P.G., and S.A.P. were funded in part by the National Institute of Diabetes and Digestive and Kidney Diseases, NIH (grant P30DK111024). K.M.V.N. was funded in part by the Worksite Lifestyle Program for Reducing Diabetes and Cardiovascular Risk in India project funded by the NHLBI (grant R01HL125442). S.A.P. was funded in part by the NHLBI (grant 5U01HL138635-02). S.A.P., K.M.V.N., M.K.A., N.T., and D.P. were supported in part by the NIH (grant 5U01HL138635) under the Hypertension Outcomes for T4 Research Within Lower Middle-Income Countries program. D.K. has been supported by the NIH’s Fogarty International Center for the Public Health Leader course (grant D43TW009135).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.M.V.N. wrote the first draft of the manuscript and reviewed and edited the manuscript. D.K. and H.H.C. researched the data and performed the formal analysis for the article. D.M., R.M.A., and M.K.A. reviewed and edited the manuscript. U.P.G., L.R.S., and S.A.P. contributed to methods and reviewed and edited the manuscript. D.P., N.T., and V.M. contributed to discussion and reviewed and edited the manuscript. All authors have read and agreed to the published version of the manuscript. K.M.V.N. and D.P. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Namratha R. Kandula.