Anthropometry and mortality : a cohort study of rural Bangladeshi women
Many authors suggest that low anthropometric levels are associated with higher mortality risk in adults, In developing countries however there have been few opportunities to test this hypothesis. In addition, there is increasing interest in the role of women's nutritional status in their own health and survival as distinct from its impact on infant outcomes. This thesis describes the results obtained from a longitudinal historical follow-up of a cohort of 2,314 rural Bangladeshi women over a period of 19 years (1975-1993). The demographic, socio-economic, and anthropometric characteristics of the study cohort are described with reference to the methods of data extraction, preparation and validation. The risk of mortality associated with different levels of the anthropometric indicators (height, weight, arm circumference and body mass index) were analysed using Cox's proportional hazards models. In addition to the basic survival models, the effects of confounding, early mortality, missing data, and young subjects, on the estimates are discussed. A significant association between BMI and mortality (p=0.009) was found in adjusted analyses which used categories that distinguished the women in the highest and lowest 10% of the cohort BMI distribution. Women with BMI levels between 10% and 90% and >90% had hazard ratios of 0.45 (95% confidence intervals 0.27,0.73) and 0.55 (0.25,1.22) respectively, when compared to women with BMI <10%. The strength of the association between BMI and mortality risk was reduced after adjusting the models for early mortality (<4 years), (p=0.068). No significant associations were found between height, arm circumference and mortality risk. In conclusion, these data provide no evidence that these anthropometric indicators would be useful in population-based screening programmes in rural Bangladesh to identify women at higher mortality risk. The findings are considered with respect to the study's methodological constraints and comparisons with other studies in order to produce recommendations for those working in research and health programmes in women's nutrition.