The analysis was based on cross-sectional data collected in the 1990–1994 National Health Survey of Pakistan, conducted by the Pakistan Medical Research Council, under the technical guidance and support of the US National Center for Health Statistics. The overall design of the survey was a modification of the Third National Health and Nutrition Examination Survey (NHANES III), conducted by the US National Center for Health Statistics, tailored to the needs of Pakistan. The details of sampling, design, components, survey instruments and quality control have previously been reported.12
Ethical approval for the survey was obtained from the Institutional Review Board at the Pakistan Medical Research Council.13
In brief, the survey was conducted on a nationally representative sample of 18 135 people aged 6 months to 110 years from 2400 urban and rural households who gave informed consent. A 2-stage stratified design was used.14
The urban and rural areas of each of the 4 provinces of Pakistan were taken as strata. There were 80 urban or rural primary sampling units; 30 households were drawn into the sample from each unit, and all residents of the household were included in the study. The overall individual response rate was 92.6%.
Data on demographic, socioeconomic and health-related variables were collected with a questionnaire validated in local languages. Dietary data were collected with a food-frequency questionnaire. All women aged 40 years or under were asked whether they were currently pregnant. Physicians at mobile examination centres performed a standardized physical examination that included 2 blood pressure readings obtained at least 20 minutes apart from the right arm by means of a mercury sphygmomanometer with the subject sitting. Trained technicians performed anthropometric examinations. Weight and height were recorded while the subject was in light clothing and without shoes. BMI was calculated as weight (in kilograms) divided by height (in metres) squared. Blood samples were obtained at least 1 hour after the subject arrived at the examination centre; fasting was not required. Blood glucose and serum cholesterol concentrations were determined with the use of the Reflotron, a multiphasic biochemical analyzer.15
Random midstream urine samples were tested with Multistix 10 SG urinalysis strips (Bayer HealthCare, Diagnostics Division, Tarrytown, NY; reagent, tetrabromophenol blue) in subjects aged 5 years or over. Quality-control measures included a field visit by expert consultants, duplicate examination by field supervisors, calibration protocols and retraining exercises.16
The category of overweight and obesity was defined as a BMI of 23 kg/m2
or greater. This definition, based on the revised criteria for Asian populations,11
was lower than the conventional cutoff value of 25 kg/m2
for populations of European origin and reflects the higher ratio of body fat to muscle mass in the former. Obesity was defined as a BMI of 27 kg/m2
Ethnicity was reported as “mother tongue,” which is specific for each of the 5 major ethnic subgroups of Pakistan: Muhajir, Punjabi, Baluch, Sindhi and Pashtun. Literacy was defined as whether an individual could read or not. Three levels of economic status were defined by simply counting owned items with use of a list of household electrical items and owned transport vehicles; this measure has been validated.17
High intake of meat, fruit, milk, eggs, rice and potatoes was defined as consumption of these items at least every other day. Use of tobacco (cigarettes, “beddies” [hand-rolled, often filterless cigarettes wrapped in temburni leaf or tendu leaf that are available in a variety of candy-like flavours], and chewing tobacco or snuff) was dichotomized as current use or not. Hypertension was defined as a systolic blood pressure of 140 mm Hg or greater or a diastolic blood pressure of 90 mm Hg or greater (based on the mean of the 2 readings) or current therapy with antihypertensive medication. Diabetes was defined as a nonfasting blood glucose concentration of 140 mg/dL (7.8 mmol/L) or greater or a history of diabetes; this definition, based on old criteria,18
diverges from the recent standard criterion of a fasting blood glucose concentration of more than 126 mg/dL (7.0 mmol/L).19
Proteinuria was defined as a urine protein concentration of 30 mg/dL (0.3 g/L) or greater as measured by a reagent strip (reading of 1+ or greater).20
Prevalence and 95% confidence intervals (CIs) were calculated with weighting that reflected the oversampling of urban areas and the 3 smaller provinces. Multivariable models were built for the primary outcome of overweight. The candidate predictor list included sociodemographic variables (age, sex, urban v. rural dwelling, literacy status, economic status, high intake of meat, fruit, milk, eggs, rice and potatoes, and current cigarette use). Variables associated with the primary outcome with p
< 0.1 in the stepwise multiple regression analysis were considered for selection in the multivariable model. For the primary outcome, we performed logistic regression analysis specific for complex survey designs21
that accounted for the clusters (primary sampling units), strata (provinces) and data weighted to the general population of Pakistan in 1990, with weights calculated by the Pakistan Federal Bureau of Statistics and confirmed by the US National Center for Health Statistics for that purpose. The final model included variables that were associated with the outcomes at p
< 0.05. The presence of each of hypertension, diabetes and raised serum cholesterol concentration was then entered into the final model to explore the independent association of BMI with these conditions.
The sensitivity and specificity of the BMI for an association with hypertension or diabetes were expressed by receiver operating characteristic (ROC) curves.