Indian Phenotype Characteristics Among Patients with Type 2 Diabetes Mellitus: Insights from a Non-interventional Nationwide Registry in India

Background: Indian patients with type 2 diabetes mellitus (T2D) constitute one-sixth of affected adults globally. Here, we evaluate the association of body mass index (BMI) with body fat percentage (BF%) and glycated haemoglobin (HbA1c) levels among patients with T2D in India. Method: This was a cross-sectional Indian registry study across 845 geographically diverse zones between December 2017 and August 2019. Results: Of 37,927 patients, 55.6% were men, with a mean ± standard deviation age of 54.2 ± 11.5 years and HbA1c of 8.3 ± 1.71%. Mean ± standard deviation BMI and BF% were 27.0 ± 4.6 kg/m2 and 32.0 ± 8.0%, respectively. Overall, 15.4% of patients were overweight, and 25.0% were obese. Despite fewer males (20.7%) having BMI-based obesity than females (31.2%), around three-quarters of both sexes had BF%-defined obesity (males 77.2%; females 71.2%). One-third of males (34.6%) and 41.9% of females had BF%-defined obesity despite normal BMI. The association was substantiated by a moderately significant correlation (r=0.51) between BMI and BF% in the overall population (p<0.0001). Conclusion: This pan-India registry presents a real-world reflection of the Asian Indian phenotype: high BF% despite lower BMI in people with T2D. This highlights the importance of primordial and primary prevention, and may guide decisions on the choice of agents for glycaemic control, with a preference for drugs that promote weight loss or are weight neutral.

1 in 6 adults with T2D globally, with marked differences in prevalence across the states. 2,3 The younger age of onset and faster progression from prediabetes to diabetes among Indians increases the disease burden. 4 With a 10.4% age-adjusted comparative prevalence of T2D, India accounts for the highest mortality in the Southeast Asian region, with 1,010,262 deaths due to T2D in 2019. 2,5 The age-standardized disability-adjusted life year rate for T2D increased in India by 39.6% (95% uncertainty interval [UI] 32.1-46.7%) from 1990 to 2016. 5 Notably, of patients who died due to T2D in India in 2016, 42.6% (95% UI 41.6-43.9%) were younger than 70 years. 5 Nearly half (47.3%) of the patients diagnosed with diabetes had not been diagnosed previously. 6 Although the prevalence of T2D remains higher in the economically advanced states in India, it has surged rapidly in the less-developed states. 5 Rapid epidemiological transition with an ageing population, compounded by modifiable risk factors such as an unhealthy diet, sedentary lifestyle, tobacco use and obesity, is an important driver of the T2D epidemic in India. 5 Among these, obesity is one of the most pivotal and dominant risk factors; prevalence of overweight in India markedly increased from 9.0% in 1990 to 20.4% in 2016. 5 Anthropometric analysis from the National Family Health Survey III and IV highlighted a rising prevalence of overweight/obesity across urban and rural locations -the prevalence among men and women was observed to be 38.4% and 36.2%, respectively. 7 A systematic review reported that more than 135 million individuals are affected by obesity in India, with variations in prevalence rates of obesity and central obesity (11.8-31.3% and 16.9-36.3%, respectively). 8 It is estimated that the percentage of touchREVIEWS in Endocrinology overweight people will more than double and obesity will triple among Indian adults between 2010 and 2040. 9 Excessive accumulation of visceral fat causes an imbalance in endocrine function and release of proinflammatory factors, which results in the development of insulin resistance, T2D and other poor cardiometabolic outcomes. 10,11 The Asian Indian phenotype (IP) is characterized by unique clinical and biochemical abnormalities, including increased insulin resistance and greater abdominal adiposity (i.e. higher waist circumference and waist-to-hip ratio), despite lower body mass index (BMI), lower adiponectin and higher high-sensitivity C-reactive protein levels. These, together with the dyslipidaemia triad -low high-density lipoprotein (HDL), high low-density lipoprotein (LDL) and high triglycerides -make Indians more prone to developing T2D. 12,13 South Asians tend to have a higher body fat percentage (BF%) compared with other ethnicities, despite lower BMI values -commonly referred to as the Yajnik and Yudnik (Y-Y) paradox. 14,15 Evaluating the body composition in terms of BF% in patients with T2D can identify risk factors, facilitating early prevention and reducing

Study design and setting
We conducted this non-interventional, multicentre, cross-sectional study

Statistical analysis
The patient characteristics and variables were described using frequency distributions and proportions for categorical variables. Continuous variables were described using mean ± standard deviation (SD). To understand the effect of sex, subgroup analysis for males and females was conducted for anthropometric variables. Correlation between BF% and BMI was evaluated using the Pearson correlation coefficient (r).
Statistical analyses were performed with statistical software, SAS ® 9.4 (SAS Institute Inc., Cary, NC, USA), and a p value of <0.05 was considered statistically significant.  Table 3 shows the anthropometric and clinical characteristics of the study population. 19 The mean body weight and height of enrolled subjects were 72.0 ± 13.0 kg and 161.5 ± 9.1 cm, respectively. The

Sex-stratified subgroup analysis
The sex-stratified analysis revealed cases of NWO (BMI within the normal range and a high BF%   (Table 4B). 19 Figure   1 characterizes IP among individuals with a normal, overweight or pre-obese BMI, with a BF% in 'acceptable' or 'obese' category. Conversely, 27.6% of females (n=1,952) with a pre-obese BMI and 9.8% (n=514) with an obese BMI had a normal BF% (range 10-31%).

Correlation between body fat percentage and body mass index
A statistically significant, moderate positive correlation (r=0.51; p<0.0001) between BF% and BMI was seen in the overall population. Similar findings were reflected for both males and females, with a significant positive relationship between BF% and BMI in both groups ( Figure 2). The scatter plot illustrates that even patients at the lower end of the spectrum of BMI tend to have a high BF%.

Correlation between glycated haemoglobin level and body mass index categories
Among the patients with HbA1c <7.0%, nearly one quarter (24.9%, n=2,144) were obese, while 44.6% (n=3,835) were pre-obese ( Table 5). 19 Similar trends were observed for higher HbA1c levels; among patients with high HbA1c levels (≥7.0), the proportion of obese patients ranged from 25.0% to 26.1%, while that for pre-obese patients ranged from 43.0% to 45.4% (Table 5). However, the correlation analysis did not demonstrate any relationship between HbA1c level and BMI (Supplementary Figure 1).

Body fat content with increasing age and diabetes duration
The sex-stratified subgroup analysis among males and females showed a significant association between BF% and age (Supplementary Table   1). Most males with an obese BF% were aged >60 years; most female subjects with an obese BF% were aged 50-59 years. Similarly, a strong significant association was found between BF% and duration of diabetes (p<0.05) in both sexes. Most males and females with an obese BF% had a duration of diabetes between 10 and 20 years (Supplementary Table 2).

Discussion
This pan-India registry presents a comprehensive real-world reflection of the IP in patients with T2D. The study validates that Indians have a high BF%, despite having relatively lower BMI. Among the patients with normal BMI, many had an obese BF% despite using the Asia-specific BMI cut off, which is lower than that used for Caucasians. 22 Of these patients with NWO, more females had an obese BF% than males. The correlation coefficient revealed a moderate positive relation between BMI and BF% in both males and females. Overall, the most prevalent comorbidities were hypertension and dyslipidaemia. Although most were taking on-going anti-diabetic medications, the mean HbA1c levels were higher than those recommended by the American Diabetes Association guidelines. 21 There was a similar distribution of individuals with an obese and pre-obese BMI across HbA1c categories.   (2) in childhood and early adulthood, excess calorie intake deposits preferentially in the upper body and ectopic fat stores (rather than lower body or as superficial subcutaneous fat); (3) a vicious cycle of high levels of plasma insulin, triglycerides and glucose, and a fatty liver appears, exacerbated by low physical activity and excess calories; (4) pancreatic β cells fail due to fewer β cells at birth, exposure to apoptotic triggers such as fat in the pancreas, and high demand from insulin resistance. 26 A study from Sri Lanka demonstrated a significant positive correlation between BMI and BF% in males (r=0.75, p<0.01) and females (r=0.82, p<0.01) of all ages. 27 The paradox of low BMI and high BF% was starkly reported for Indians in Singapore, with Indians having the highest BF% among a mixed population of Indian, Chinese and Malayan people. 28 NWO is an under-recognized arena; however, evidence on its pathophysiology and its association with metabolic diseases such as T2D, hypertension and dyslipidaemia is evolving. 29 Results from the Kerala Diabetes Prevention Program demonstrated that about one-third of the study subjects had NWO. 30 The study also reported a significantly higher proportion of individuals with T2D, hypertension and dyslipidaemia in the NWO group compared with the non-obese group. 30 NWO was also identified as an independent strong predictor of cardiovascular mortality, and a widely prevalent problem in individuals of Asian descent. 29 A study among males in Lucknow reported that 44.0% of subjects showed a high BF% (>25%) with a BMI of 24.0-24.9 kg/m 2 , and 4.7% at a lower BMI (<20 kg/m 2 ). Rates of high BF% in the BMI range 20-21.9 kg/m 2 and 22-23.9 kg/m 2 were 9.5% and 18.4%, respectively. In addition, BMI was highly correlated with BF% (r=0.73, p<0.001). 17 The study results demonstrate that females have a proportionally higher BF% than males, despite having similar BMI. A real-world study including data from the Korea National Health and Nutrition Examination Survey (2007-2010) revealed that whole BF% content was higher in women than in men. 31 This study also demonstrated that the average BMI and whole BF% content of women were higher than those of men; however; the average waist circumference at the time of diabetes diagnosis was similar (approximately 88 cm) in both sexes. 31   in patients with T2D. 16 In addition, a prospective study from India also reported that centrally and peripherally obese subjects with dyslipidaemia had a significant association with HbA1c in T2D. 39 A community-based Korean cohort study demonstrated that, compared with people with a lower BF% (quintile 1), the risk for T2D significantly increased among those with a higher BF% (22.8% in men and 32.9% in women; ≥ quintile 4). 40 However, our study showed a similar distribution for obese and pre-obese individuals with respect to BMI across the different HbA1c levels.
There is a significant loss of skeletal muscle mass and an increase in BF% with increasing age; the term 'sarcopenia' relates to age-related decreases in muscle mass and strength. Low muscle mass and increased BF% are associated with a risk of developing metabolic disorders, including T2D. 41 Because of lifestyle changes and longer life expectancy, the burden of T2D and sarcopenic obesity is projected to increase globally; both share common risk factors, such as ageing and general obesity. 42 Individuals with T2D tend to develop sarcopenic obesity, which is likely to increase with age. 43  and GLP-1 agonists minimize weight gain when added to metformin and/or sulfonylurea, and the clinically meaningful body weight reductions can further contribute reduced HbA1c and systolic blood pressure. 49,50 The co-administration of these novel oral anti-diabetic drugs that target complementary mechanisms represents an effective strategy for weight loss, with additional cardiorenal benefits among South Asian people. 51 The Indian Phenotype Registry is a real-world registry based on data collected from routine clinical practice, with no follow-up visits.
Hence, issues related to an observational registry, such as loss to follow-up and missing data, as well as the unavailability of zone-specific data, form some important limitations. Data on dyslipidaemia, such as HDL, LDL, total cholesterol and triglycerides, were not collected in this study. Although we assessed the correlation between HbA1c and BMI, the correlation between HbA1c and BF% was not investigated.
Additionally, being a cross-sectional analysis, the study cannot affirm a causal association between obesity and other variables. However, this is one of the largest registries worldwide exploring IP characteristics. In India, the relationship between BMI and BF% has been investigated in region-specific prevalence studies, but with smaller sample sizes. Results from the IP registry can augment and substantiate the current evidence pool describing the Asian phenotype. The large sample size, with a representative population from diverse geographies and healthcare tiers of India, strengthen the results of the study.

Conclusion
The Indian Phenotype Registry is a pan-India cross-sectional registry that aims to generate nationwide data and provide clear insights about the phenotypic characteristics specific to Indian patients with T2D. Results from this study affirm the key characteristics of the IP of a low BMI with a high BF%. Additionally, the mean HbA1c levels were high, despite the majority of patients receiving anti-diabetic medications. Insights on the high BF distribution in Indian patients with T2D highlight the importance of effectively identifying risk factors (primordial prevention), diagnosing early (primary prevention) and aggressively managing obesity with intensive diet, exercise and therapy interventions to reduce complications and comorbidities (secondary prevention). These findings will guide therapeutic decisions on the choice of agents for glycaemic control, with preference for drugs that promote weight loss, such as SGLT2is and GLP-1 agonists, or are weight neutral, such as metformin, α-glucosidase inhibitors and DPP4is. 47-50 ❑