BackgroundPregnancy is a time of optimum motivation for many women to make positive behavioural change to improve their own health and the health of their unborn child. We aimed to identify and describe groups of pregnant women with similar patterns of self-reported health behaviours with use of latent class analysis, and to establish the association of these behaviours with birthweight. MethodsThe Born in Bradford (BiB) birth cohort is a longitudinal study of more than 13–000 infants born in the city of Bradford, in northern England, between 2007 and 2010. About half the study participants were of Pakistani ethnic origin. This study used data from 10–850 mother-pairs for whom the mother completed a baseline questionnaire and had a live term singleton birth. Data were available for smoking and alcohol consumption at three different timepoints, representing before pregnancy, first trimester, and second trimester. Data for other health-related exposures, including caffeine consumption, vitamin D supplementation, use of other tobacco products (including paan), and exposure to second-hand smoke and other illegal drugs, were available for the second trimester of pregnancy. We used latent class analysis to identify groups of individuals with similar patterns of health behaviours with Mplus 6.0. This analysis classified participants into groups on the basis of conditional probabilities. We then used linear regression to examine the association between group membership and birthweight and to adjust for maternal sociodemographic and birth characteristics. FindingsFive classes of behaviours groupings fitted these data best. Class one members (7142, 65·8%) were the most healthy, being more likely to take vitamin D and were unlikely to report unhealthy behaviours. Class two members (416, 3·8%) were unlikely to smoke but were heavier drinkers before or during pregnancy and had low levels of vitamin D consumption. Babies born to mothers in this group were on average 20 g (95% CI −66·5 to 28·4; p=0·42) heavier than were those born to women in class one, although we noted no evidence of a difference between classes. Class three members (1514, 14·0%) were most likely to be ex-smokers or be cutting back on smoking and drinking, and were otherwise healthy. Babies born to these mothers were on average 21 g (–5·6 to 51·3; p=0·15) heavier than were those born to mothers in class one, although we noted no evidence of a difference between classes. Class four members (907, 8·4%) reported cutting back on smoking and drinking but were still partaking in other unhealthy behaviours and were also more likely to be exposed to passive smoke than were mothers in other classes. Babies born to these mothers were on average 97 g (–59·7 to −130·1; p<0·0001) lighter than were those born to mothers in class one. Class five members (871, 8·0%) were most likely to report high levels of unhealthy behaviours and low levels of healthy behaviours, in addition to being exposed to passive smoke. Babies born to these mothers were on average 227 g (–187·1 to 262·1; p<0·0001) lighter were than those born to mothers in class one. InterpretationThis latent class analysis identified five subgroups of pregnant women with different patterns of behaviour; generally those with more healthy behaviours had heavier babies at birth. Further work will explore which of the behaviours drives the association with birthweight. FundingNone.