[Self-reported weight and height for determining nutritional status of adults and elderly: validity and implications for data analysis].

This study evaluated the validity of self-reported weight and height for determining nutritional status and the implications of their use for analyzing associations with health outcomes. A population-based cross-sectional study in Pelotas, Rio Grande do Sul State, Brazil, in 2007 (n = 2,986) drew a sub-sample of 276 individuals aged ≥ 20 years. Mean self-reported weight was similar to measured weight; height was overestimated in men (1.4 cm) and women (2.5 cm); real body mass index (BMI) was underestimated by about 1 kg/m(2). Even with small mean differences, data variability was great. The results were influenced by gender, age, and schooling. The use of self-reported measures underestimated prevalence of overweight and obesity, and unpredictable errors were found in the analysis of association with health outcomes (underestimation, overestimation, and reversal of real effect measures). Correction equations reduced the mean differences but did not resolve variability of the differences, classification errors, or biases in the associations.