Burden and Socio-Behavioral Correlates of Uncontrolled Abnormal Glucose Metabolism in an Urban Population of India

Background Progressive burden of diabetes mellitus is a major concern in India. Data on the predictors of poor glycemic control among diabetics are scanty. A population-based cross-sectional study nested in an urban cohort was thus conducted in West Bengal, India to determine the burden and correlates of total and uncontrolled abnormalities in glucose metabolism (AGM) in a representative population. Methods From 9046 adult cohort-members, 269 randomly selected consenting subjects (non-response = 7.24%) were interviewed, examined [blood pressure (BP), anthropometry], tested for fasting plasma glucose (FPG) and glycosylated hemoglobin (HbA1C). Those having pre-diagnosed diabetes or FPG ≥126 or HbA1c≥6.5 were defined as diabetic. Among non-diabetics, subjects with FPG (mg/dl) = 100–125 or HbA1C(%) = 5.7–6.4 were defined as pre-diabetic. Pre-diagnosed cases with current FPG ≥126 were defined as uncontrolled AGM. Descriptive and regression analyses were conducted using SAS-9.3.2. Results Among participants, 28.62% [95% Confidence Interval (95%CI) = 23.19–34.06)] were overweight [body mass index(BMI) = (25–29.99)kg/meter2], 7.81% (4.58–11.03) were obese(BMI≥30kg/meter2), 20.82% (15.93–25.70) were current smokers, 12.64% (8.64–16.64) were current alcohol-drinkers and 46.32% of responders (39.16–53.47) had family history of diabetes. 17.84% (13.24–22.45) had stage-I [140≤average systolic BP (AvSBP in mm of mercury)<160 or 90≤average diastolic BP (AvDBP)<100] and 12.64% (8.64–16.64) had stage-II (AvSBP≥160 or AvDBP≥160) hypertension. Based on FPG and HbA1c, 10.41% (6.74–14.08) were diabetic and 27.88% (22.49–33.27) were pre-diabetic. Overall prevalence of diabetes was 15.61% (11.25–19.98). Among pre-diagnosed cases, 46.43% (26.74–66.12) had uncontrolled AGM. With one year increase in age [Odds Ratio(OR) = 1.05(1.03–1.07)], retired subjects [OR = 9.14(1.72–48.66)], overweight[OR = 2.78(1.37–5.64)], ex-drinkers [OR = 4.66(1.35–16.12)] and hypertensives [ORStage I = 3.75(1.42–9.94); ORStage II = 4.69(1.67–13.17)] had higher odds of diabetes. Relatively older subjects [OR = 1.06(1.02–1.10)], unemployed [OR = 19.68(18.64–20.78)], business-owners [OR = 25.53(24.91–16.18)], retired [OR = 46.53(45.38–47.72)], ex-smokers [OR = 4.75(1.09–20.78)], ex-drinkers [OR = 22.43(4.62–108.81)] and hypertensives [ORStage II = 13.17(1.29–134.03)] were more likely to have uncontrolled AGM. Conclusions Burden of uncontrolled AGM was high among participants. Efforts to curb the diabetes epidemic in urban India should include interventions targeting appropriate diabetic control among relatively older persons, unemployed, business-owners, retired, ex-smokers, ex-drinkers and hypertensives.


Introduction
Diabetes Mellitus is a heterogeneous syndrome of abnormalities in carbohydrate and fat metabolism, characterized by multi-factorial interplay between genetic and environmental factors culminating into beta-cell dysfunction and reduced tissue insulin sensitivity. [1] Most adult type 2 diabetics are found to be overweight and centrally obese. The three explanatory paradigms are: portal/visceral hypothesis (delineating the key role of elevated non-esterified fatty acids), ectopic fat storage (triglyceride deposits in muscle/liver/pancreas) and role of adipose tissue endocrine organs (secretions of various adipocytokines implicated in insulin resistance and probably beta-cell dysfunction). [2][3][4] Thus an obesogenic environment comprising of lifestyle, eating habits, addictions (smoking and alcoholism), stress and physical inactivity, are considered to be the predominant yet modifiable determinants of diabetes. [5] On the other hand in case of Type 1 diabetes (immune-mediated, non-autoimmune or Idiopathic leading to absolute insulin deficiency) of younger age group,the hyperglycemia is due to Beta cell destruction mostly with immunologic marker from insulin deficiency.
Diabetes has become a major public health challenge due to its increasing prevalence in most countries. [6,7] The International Diabetes Federation (IDF) estimated that in 2012, globally 4.8 million persons died of diabetes while 347 million were suffering from it (adult prevalence of 8.3%). [6] More than 80% of these cases and deaths were reported from low and middle income countries, while India and China contributed the most. [6,8] The burden of diabetes was reported to have increased rapidly in India. [9,10] The IDF estimated the prevalence of diabetes in India to be 9.01% among persons aged !20 years in 2012. [6,11] According to the India Diabetes (ICMR-INDIAB) study by ICMR in 2011, approximately 62.4 million people aged !20 years were suffering from diabetes and the finding was consistent with that of IDF. [6,11] Reviews on abnormalities in glucose metabolism (AGM) in India showed that the majority of the earlier studies were conducted in southern and northern parts of the country [12][13][14] and a few in the northeastern states. Data from the eastern region, particularly from the populous state of West Bengal, are scanty [12][13][14][15]. Related studies were mostly concerned with prevalence rather than risk factors, the predictors of uncontrolled disease, or the population subgroups to target preventive and treatment efforts. [16,17] Moreover, methodological shortcomings related to study design, sampling, geographical diversities, non-standardized diagnostic criteria, incomplete data, high non-response and non-representative sampling precluding extrapolation of observations beyond the study samples. [12] A dearth of information on socio-behavioral determinants of uncontrolled AGM thus called for a detailed investigation in a representative, population-based sample. A cross-sectional study nested in a cohort of an urban population in West Bengal was thus conducted with following objectives: 1. To estimate the prevalence of overall and uncontrolled cases of AGM among adults and 2. To measure the strength of association between these abnormalities and their socio-demographic, physical and behavioral correlates. where,all households were identified and enumerated. Using empirical most conservative parameter value of 50%, assuming α = 0.05, as per the sample size calculation formula: s = z 2 Ã p(1-p)/d 2 , (d = desired precision = 5% of 50 = 2.5) and finite population correction formula: n = s/(1 + (s/N)) where N = population size = 31715 households altogether, 489 (using Epi Info) households from each of the six randomly selected wards were required to be recruited assuming equal cluster (ward) population size and empirically assumed designed effect = 2.

Study population
Hence from the list of all houses in each of the six selected wards approximately 500 households were selected randomly by using random number generation and selection using computer programming. The number of selected household in each selected ward did not remain exactly 500 because some of the randomly selected identification numbers for one household actually represented more than one because there was more than one family residing (and cooking separately) at the same address. Thus 3030 households from six selected wards with 12734 residents were included in the cohort at baseline in 2001 and are being continuously followed. People were removed from the cohort by death or migration-out and were included by birth or migration-in.

Sample size for current study and recruitment
Due to non-availability of appropriate parameter values, the previously observed burden of abnormal FPG levels within this cohort (in 2006 among 1194 randomly selected representative individuals 22.65% had FPG!100mg/dl) was used for the sample size calculation. [18,19] Based on this parameter estimate in the source population, using the population-survey tool of Epi-Info, [20] to determine the prevalence of abnormal FPG level among adults in the study population with a relative precision of 10% and 1 = 0.05, 261 subjects were required for a cross-sectional study. Exclusion criteria included: migration out, refusal to provide informed consent and non-availability in three consecutive follow-ups. Assuming a non-response rate of 10%, 290 adults (irrespective of their diabetic and treatment status) were randomly selected from the list of all 9046 adult cohort members in 2009 and invited to participate between November 2009 and March 2010 (using age and gender stratification to preserve the distribution). Altogether 269 of these 290 invited eligible subjects provided consent, were interviewed, tested and included in the analysis. (Fig 1)

Ethics statement
Prior to the interview and blood sample collection, written informed consent was collected. Data were securely preserved with confidentiality. The study protocol was reviewed and approved by the Institutional Ethics Committee of National Institute of Cholera and Enteric
Personal behavior/habits included physical exercise (hour/wk), smoking and alcohol-drinking. For the quantification of smoking, number of cigarettes smoked/day (converted to packs smoked/day assuming Indian standard pack of 10 cigarettes), year of starting smoking and duration of smoking were determined. Pack-years (py) were calculated by combining the frequency and duration. Type of smoking was then categorized into non-smoker/ex-smoker/current light smoker (0.1-20.0py)/current moderate smoker (20.1-40.0py)/current heavy smoker (>40py). [27] Based on similar questions, type of alcohol-drinking was classified as: nondrinker/ex-drinker/infrequent drinker (drinking rarely or 1-3 times/month)/frequent drinker (once wkly drinking <daily)/daily drinker. Duration of alcohol intake in completed years, family history of diabetes (yes/no/don't know) and whether the subject had already been diagnosed as a diabetic (yes/no) were enquired additionally. [28] Three arterial blood pressure readings were taken at 10 minutes interval with a bound of ±3 mm of Hg and the averages (mm of Mercury) were used to define hypertension status as: normotensive [average systolic BP (AvSBP)<120 and average diastolic BP (AvDBP)<80), prehypertensive [(120 AvSBP<140 and AvDBP<90) or (AvSBP<140 and 80 AvDBP<90)], stage-I hypertension [(140 AvSBP<160 and AvDBP<100) or (AvSBP<160 and 90 AvDBP<100)] and stage-II hypertension (AvSBP!160 or AvDBP !100). [29] Laboratory testing and parameters To determine AGM through several parameters (observed level of FPG, glycemic control, current diagnosis of diabetes and overall diagnosis) and then to define AGM status based on demonstrated control over the aforementioned parameters we needed to measure FPG and glycosylated hemoglobin (HbA1c). For this purpose, approximately 3ml of venous blood was collected in double oxalate and EDTA vials on the day of interview, from each participant after overnight fasting. Plasma separation from the double-oxalated blood was done in the study laboratory in Barrackpore and then carried in a cold chamber to the central laboratory at National Institute of Cholera and Enteric Diseases, Kolkata where preparation for assays and storage were performed. Samples were next sent to the processing Laboratory at Mediland, Diagnostic Institute, Kolkata in duly labeled Eppendorf tube maintaining a temperature of 2-8°C. For HbA1cwhole blood in EDTA vials was carried to the same laboratory in a cool box maintaining a temperature of 2-8°C. Samples were stored for the next six months in a refrigerator at -20 degree centigrade temperature.
FPG was estimated from plasma separated from the double oxalated blood using GOD/ POD enzymatic method on fully automated Biochemistry analyzer [A-25 Biosystems (Spain)] and the assay took 5 minutes time/sample. A. HbA1c was estimated from whole blood collected in EDTA vials, using Column Method [Biosystems (Spain)] taking 30 minutes /sample.
Next, among previously diagnosed diabetics, AGM status (based on prior diagnosis of diabetes: yes/no and current FPG results) was categorized as: 1. Never diagnosed with AGM (nondiabetic FPG with no prior diagnosis of diabetes), 2. Well-controlled AGM (non-diabetic FPG with prior diagnosis of diabetes) 3. Moderately-controlled AGM (pre-diabetic FPG with prior diagnosis of diabetes) or 4. Un-controlled AGM (diabetic FPG with prior diagnosis of diabetes). [30] Data analysis Following descriptive analyses [means with standard deviation (SD) and proportions with 95% Confidence Intervals (95%CI)], measurement of the strengths of associations [Odds Ratios (OR) and corresponding 95%CIs] of potential socio-demographic, physical and behavioral correlates with binary (prior diagnosis of diabetes, type of diagnosis of abnormal FPG and glycemic control) and multi-category (FPG level, glycemic control and AGM status) outcome variables were conducted using simple and multinomial bivariate logistic regressions respectively, [31][32][33] using SAS version 9.3.2. [34] We modeled fasting plasma glucose level (ref: normal), glycemic control (reference: normal), current diagnosis of diabetes (reference: non-diabetic), overall diagnosis of diabetes (reference: non-diabetic) and AGM status as our five outcome variables and regressed each of them on age, gender (reference = female), occupation (reference = laborer), physical exercise (hrs/wk), body mass index (reference = normal), waist circumference (cm), abdominal obesity (reference = no), smoking category (reference = non-smoker), completed years of alcohol drinking, type of alcohol drinking (reference = non-drinker), family history of diabetes (reference = no) and diagnosis of hypertension (reference = normo-tensive) separately.

Socio-demographics and anthropometric measures
Among 269 randomly selected subjects (134 males and 135 females), mean age was 42

Discussion
In a representative sample of an urban adult population of West Bengal, the burden of AGM was found to be considerable. While FPG was found to be at impaired and diabetic level for 20.07% and 7.81% of individual s respectively, 26.02% had impaired and 9.67% had diabetic level of glycemic control. Based on the testing and standard diagnostic protocol, [18,19] 27.88% and 10.41% subjects were pre-diabetic and diabetic respectively. Including both previous diagnosis and test results during the study, the prevalence of diabetes was found to be 15.61% which was over 6% higher than the national level estimate of 9.01% by IDF for 2012. [6,11]Furthermore, among the previously diagnosed diabetes cases, 46.43% had uncontrolled AGM. These findings cumulatively indicated a large burden of AGM in the study population associated with known reversible factors. Therefore appropriately targeted intervention is necessary to reduce this burden. [35][36][37][38] Subjects in this study were recruited randomly from an urban population cohort with a very low (7.24%) non-response rate. Hence considering the representativeness of the sample, the results revealed high prevalence of obesity, smoking, alcohol-drinking, hypertension, family history of diabetes and low mean hours of physical exercise among urban adults of West Bengal. As evidenced elsewhere, [13,35,39,40] the observed burden of lifestyle factors seemed to have the potential for increasing the risk of diabetes and its complications among the urban residents in West Bengal.
In this study, compared to laborers, retired subjects had higher odds (independent of age) of being diabetic while business-holders and retired were more likely to have uncontrolled AGM. Service-holders seemed to have relatively better control over aAGM. While advancing age, less physical activity and other co-morbidities were likely to increase the risk of diabetes among retired subjects lack, of social support, age-induced self-neglect and forgetfulness regarding medicines might be the explanation for poor control of AGM. While a relatively more routine life probably helped service-holders and house-wives to keep their AGM under control; stress, anxiety and instability probably increased the chances of uncontrolled AGM among businesspersonnel and unemployed subjects. Apart from few conflicting results, [30] prior evidences did also show that the unemployed, the retired and workers in higher income jobs were frequently found to be diabetic (and more often uncontrolled) compared to those with other jobs. [36,40,43].
Diabetic patients having stressful jobs, sedentary lifestyle along with those who are retired and unemployed should be targeted for better control.
Compared to those with normal body weight, overweight subjects in this study were more likely to be diabetic. Although obesity is an established risk factor for diabetes [11,37,40,44], there was no association between risk of diabetes with obesity in this study. This observation could be explained by the possibility that obesity, being a known risk factor, when compared to overweight subjects, obese diabetics might have been diagnosed and commenced on treatment earlier. Because of their knowledge regarding harmful effects of obesity, they could also have been more compliant with treatment. Consistent with this observation, a higher waist circumference was also associated with being diabetic and having uncontrolled AGM. A positive association of abdominal obesity with diabetes was also observed in prior studies. [11,15,36,37,39,40] Thus abdominal obesity appeared to be a stronger correlate of overall and uncontrolled AGM among urban residents in West Bengal. Hence screening for abdominal obesity should be included in all diabetic risk assessment tools considering this to be an important lesson from the study.
As in other studies [45][46][47][48] the current study showed that, compared to non-smokers, smokers were more likely to have diabetes [35] while current smokers had higher odds of having better control. Awareness of the harmful effects of smoking might have increased the compliance with control of AGM among smokers.
Compared to non-drinkers, ex-drinkers were also more likely to be diabetic while longer duration of alcohol drinking and being an ex-drinker were found to be associated with uncontrolled AGM. Previous studies also revealed a correlation between alcoholism and diabetes. [41] Reverse causation could be the explanation for this observation (i.e. people with alcoholism might have quitted drinking alcohol because of uncontrolled AGM) for both ex-smokers and ex-drinkers having higher chance of having uncontrolled AGM. Persons with a history of current or past smoking and alcoholism probably required special attention regarding control of AGM.
Subjects with a family history of diabetes also seemed to be at higher odds of having impaired FPG in this study. A genetic predisposition of AGM has also been reported from previous studies conducted in different states of India. [11,13,15,36,37,39,44,49] The participants in this study had a high prevalence of hypertension as the proportions of subjects with pre-hypertension, stage-I hypertension and stage II hypertension were 33.09%, 17.84% and 12.64% respectively. A previous study in 2010, revealed a similar picture in West Bengal regarding the prevalence of hypertension (i.e. 46.5%). [41]Similar to previous studies, [11,15,39] in the current study, compared to normotensives, stage I and stage II hypertensives had higher odds of having AGM, while stage-II hypertensives were more likely to have uncontrolled AGM. Thus it is inferred that hypertensives need additional attention for good control of AGM.
This study has important limitations. Self-reported socio-demographic and behavioral information (e.g. smoking, drinking, exercise etc.) collected in this study had the potential to suffer from social desirability bias and the reporting of having previously been diagnosed as a diabetic might well have included some misclassification. Which also could be suspected as none of the female participants reported any history of smoking or alcohol drinking. However, as this study was nested in a long-term cohort, the trusted relationship between the research team and the participants could have helped in convincing the subjects adequately regarding the confidentiality of their information. Thus better recall and less chance of mis-reporting were likely to reduce the magnitudes of these information biases. Although we expect to have recruited a representative sample of an urban population of eastern India by virtue of our sampling design nested in a long-standing cohort, still the potential threat to generalizability owing to the emergence of new households in the study area over time and consequent missing of some eligible subjects should also be borne in mind.
Due to sample size issues (owing to lack of budget to conduct a larger study), some of the observations lacked power and multivariate analyses were not feasible. Thus we could only present the results of the bivariate analyses, but still consider them quite useful in developing important insight. However, while interpreting these results, the potential for confounding should always be borne in mind. Due to the cross-sectional nature of the study, causal interpretations of the results of the study were not possible. Temporal ambiguity is also an issue, although being nested in a population-based cohort, some amount of temporality was probably incorporated into any interpretation. Despite these limitations, by virtue of having a representative sample, nested in an urban cohort and having only 7% non-response, generalizability issues and selection biases were not likely to affect the study results.

Conclusion
In this urban population of India, a high burden of total and uncontrolled abnormal glucose metabolism was observed, which was greater than the so-far reported prevalence in other communities of Indian subcontinent. Results suggested that waist circumference should be included as an important screening tool for diabetes.
Control of risk factors especially among relatively older subjects, retired, smokers, ex-drinkers and those having higher waist circumference is required to curb the diabetes epidemic in urban India. For harm reduction, appropriate counseling of the retired, business personnel, exsmokers and ex-drinkers to ensure better control of AGM appears indicated.