In 1999, the demographic and social characteristic profile of the general population of Chandra, was described as rural, without urban facilities. The main livelihood of the population was agricultural and other agrarian activities. Ten villages were randomly selected from five areas. The total population of these villages was approximately 20,000 aged ≥20 years. During the recruitment period all persons aged ≥20 years, willing to participate and able to communicate were included in the study. Pregnant women and those with a diagnosed acute physical or mental illness were excluded. Description of the population and recruitment process were described earlier in brief [10]. In 1999, 5,000 subjects were recruited and of these 4,757 (95%) agreed to participate. In 2004, 5,000 subjects were again recruited and of these 3,981 (80%) subjects participated. In 2009, 3,000 of same age group subjects were randomly selected and among them 2,376 (79.2%) participated. After exclusion of 83 participants with missing glucose values, 2,293 participants were included for the analysis.

Verbal consent was obtained from each participant prior to their inclusion in the study as the majority of participants were illiterate. This procedure was adopted to avoid selection bias and individual discrimination. Information, explaining the study objectives and protocol were read aloud to the participants, including, their right to refuse and withdraw at any stage of the study or to exclude their data from analyses. The protocol was approved by the Norwegian National Committee for Medical and Health Research Ethics and the Ethical Review Committee of the Diabetic Association of Bangladesh.

On completion of the selection procedure, participants were requested to visit a nearby field centre. In the 1999 and 2004 studies, capillary whole blood was used to measure the fasting blood glucose (FBG) after an overnight fast of 8 to 12 hours, using a Hemocue blood glucose analyzer. Internal quality control tests were conducted each day, using recommended glucose solutions, to ensure accuracy and reliability of the results. DM and impaired fasting glucose (IFG) were diagnosed using World Health Organization (WHO) criteria [11]. Diagnosis of DM was made if the FBG was ≥6.1 mmol/L. IFG was diagnosed if the FBG was between ≥5.6 to <6.1 mmol/L. In 2009, venous plasma glucose (VPG) samples were collected on the morning following an overnight fast. DM was diagnosed if the FBG value was ≥7.0 mmol/L. IFG was diagnosed if the FBG was between ≥6.1 to <7.0 mmol/L [11]. In addition, known DM was defined by the use insulin or oral antidiabetic medication(s) and self-reported DM. VPG samples were collected in a tube containing sodium fluoride and were centrifuged on site within 30 minutes of collection. Separated plasma samples were sent to the Bangladesh Institute of Research and Rehabilitation in Diabetes, Endocrine and Metabolic Disorders (BIRDEM) in ice gel packed cooling boxes and stored at minus 70℃ until laboratory assays. Plasma glucose was measured by the glucose oxidase method using Dimension RxL Max (Siemens AG, Erlangen, Germany) on the same day. In the 2009 study, quality control on the blood glucose measurement was checked by measuring the fasting plasma glucose values using the glucose oxidase method in every 10th sample. The intra-assay coefficient of variation was 1.24% at a mean of 5.86 mmol/L, and the inter-assay coefficient of variation was 2.10% at a mean of 5.23 mmol/L. The correlation between capillary whole blood (using point of care meter) and venous plasma blood was calculated using Pearson's correlation methods (r=0.94, P<0.001). All study participants were informed of their glucose tolerance status as soon as the results were available.

Socioeconomic and demographic information, parental and personal health history were verbally obtained using standardised questionnaires and anthropometric measurements were recorded. These were carried out through trained researchers.

Measurements of height, weight, and waist circumference (WC), and hip circumference (HC) were taken with participants in light clothes and without shoes. For height, the participant stood erect against a wall, with the occiput, back, hip, and heels in contact with the wall while looking straight ahead and keeping the tragus and lateral orbital margin in the same horizontal plane. WC was measured by placing a plastic tape horizontally midway between 12th rib and iliac crest on the midaxillary line. Similarly, HC was measured by placing the tape measure at the level of maximum posterior extension of the buttocks and horizontal around the body. Body mass index (BMI) was calculated as weight divided by height squared (kg/m²). Obesity was defined as a BMI of 25 kg/m² or greater for Asian Indians [12,13]. Central obesity indicated by WC ≥90 cm for male and ≥80 cm for female and waist hip ratio (WHR) ≥0.90 for male, ≥0.80 for female. Blood pressure (BP) was measured by 1) ensuring the participant rested for 10 minutes prior to the BP being measured, 2) using standard cuffs for adults fitted with a standard mercury sphygmomanometer, and 3) placing the stethoscope bell lightly over the pulsatile brachial artery on the right arm. The average of two BP readings taken at a 5-minute interval in both sitting and standing positions was recorded. Hypertension (HTN) was defined as a systolic BP (SBP) of ≥140 mm Hg and/or diastolic BP (DBP) of ≥90 mm Hg or diagnosed HTN by a physician prior to the survey and reported regular in place of daily use of antihypertensive medication(s).

All the continuous variables were presented as mean (95% confidence interval [CI]) and categorical data as percentage (95% CI). Differences between the three study cohorts of means and proportions adjusted for age was tested by analysis of covariance (ANCOVA) and logistic regression respectively. Multiple comparisons were performed using the Bonferroni method. The prevalence of DM and IFG were age standardised by the direct standardization method. Age standardised prevalence was estimated on the basis of 1991 census data adjusted in 2000 for all three surveys [14]. Multiple logistic regression models were used to test for linear trends in the age standardised prevalence and estimate adjusted odds ratios with 95% CIs for all three studies. Two-sided P values were given and the significance level was set at 0.05. STATA 11 for Windows (STATA Co., College Station, TX, USA) was used to estimate the age standardized prevalence and creating the figures. Otherwise PASW statistics version 18 for Windows (SPSS Inc., Chicago, IL, USA) was used.

Capillary whole blood was used in the first two studies and venous plasma blood in the third study. These methods were used to confirm the diagnosis of DM and IFG. WHO cutoff glucose values used should reflect equivalent glucose values for capillary and venous plasma blood. However, the use of point of care blood glucose meters in a multi operator setting in large populations may have adversely influenced the inter-rater reliability of the researchers' technique and glucose values in the first two studies. Health questionnaires were administered by researchers who obtained self-report information verbally and recorded it as the majority of the participants were illiterate. Therefore, the interpretation and recording of the results is a potential bias in the studies. The studies were conducted in a rural area of Bangladesh which during the 10-year period had undergone rapid urbanization and associated environmental change. Therefore the extent to which the results are generalizable or representative of other populations in Bangladesh is not known.

The prevalence of DM has continued to increase in the past 10 years in a rural population of Bangladesh. The higher prevalence of DM may also be indicative of other changing environmental factors where rapid urbanization as a pretext for economic development may have influenced individual lifestyles and dietary behaviors. The increased prevalence of risk factors for DM, amongst the study populations includes obesity and central obesity, as indicated by BMI, WC, and WHR. The significant increase of these risk indicators demonstrates the potential impact on health in a developing country. More large scale studies with control populations may be helpful in recognizing the confounding factors responsible for increased prevalence of DM and IFG in rural Bangladesh.