Obesity is a growing problem in India, the dietary determinants of which have been studied using an ‘individual food/nutrient’ approach. Examining dietary patterns may provide more coherent findings, but few studies in developing countries have adopted this approach. The present study aimed to identify dietary patterns in an Indian population and assess their relationship with anthropometric risk factors.

FFQ data from the cross-sectional sib-pair Indian Migration Study (IMS; n 7067) were used to identify dietary patterns using principal component analysis. Mixed-effects logistic regression was used to examine associations with obesity and central obesity.

The IMS was conducted at four factory locations across India: Lucknow, Nagpur, Hyderabad and Bangalore.

The participants were rural-to-urban migrant and urban non-migrant factory workers, their rural and urban resident siblings, and their co-resident spouses.

Three dietary patterns were identified: ‘cereals–savoury foods’ (cooked grains, rice/rice-based dishes, snacks, condiments, soups, nuts), ‘fruit–veg–sweets–snacks’ (Western cereals, vegetables, fruit, fruit juices, cooked milk products, snacks, sugars, sweets) and ‘animal-food’ (red meat, poultry, fish/seafood, eggs). In adjusted analysis, positive graded associations were found between the ‘animal-food’ pattern and both anthropometric risk factors. Moderate intake of the ‘cereals–savoury foods’ pattern was associated with reduced odds of obesity and central obesity.

Distinct dietary patterns were identified in a large Indian sample, which were different from those identified in previous literature. A clear ‘plant food-based/animal food-based pattern’ dichotomy emerged, with the latter being associated with higher odds of anthropometric risk factors. Longitudinal studies are needed to further clarify this relationship in India.

To validate questionnaire-based physical activity level (PAL) against accelerometry and a 24 h physical activity diary (24 h AD) as reference methods (Protocol 2), after validating these reference methods against the heart rate–oxygen consumption (HRVO2) method (Protocol 1).

Cross-sectional study.

Two villages in Andhra Pradesh state and Bangalore city, South India.

Ninety-four participants (fifty males, forty-four females) for Protocol 2; thirteen males for Protocol 1.

In Protocol 2, mean PAL derived from the questionnaire (1·72 (sd 0·20)) was comparable to that from the 24 h AD (1·78 (sd 0·20)) but significantly higher than the mean PAL derived from accelerometry (1·36 (sd 0·20); P < 0·001). Mean bias of PAL from the questionnaire was larger against the accelerometer (0·36) than against the 24 h AD (−0·06), but with large limits of agreement against both. Correlations of PAL from the questionnaire with that of the accelerometer (r = 0·28; P = 0·01) and the 24 h AD (r = 0·30; P = 0·006) were modest. In Protocol 1, mean PAL from the 24 h AD (1·65 (sd 0·18)) was comparable, while that from the accelerometer (1·51 (sd 0·23)) was significantly lower (P < 0·001), than mean PAL obtained from the HRVO2 method (1·69 (sd 0·21)).

The questionnaire showed acceptable validity with the reference methods in a group with a wide range of physical activity levels. The accelerometer underestimated PAL in comparison with the HRVO2 method.