Agreement between self-reported birth weight and birth certificate weights

Birth weight is emerging as a potentially important risk factor for several chronic diseases with adult onset, including breast cancer. Because participant recall is frequently used to gather data on early life exposures, it is essential that the accuracy of recall be assessed and validated. Self-reported birth weights and birth certificate weights were compared in women aged 35-51 years from the Western New York Exposures and Breast Cancer (WEB) Study, a population-based case-control study. A total of 180 participants had both birth certificate and interview data on birth weight. Participants reported birth weight to one of six categories (<5, 5-5.5, 5.6-7, 7.1-8.5, 8.6-10 and >10 lbs). The Spearman correlation for self-reported and birth certificate weights was 0.67. Sixty percent of participants reported weights with exact agreement with birth certificate; unweighted and weighted kappas (κ) were 0.39 and 0.68, respectively. Spearman correlations were similar for cases (0.67) and controls (0.68). Controls exhibited a significantly higher unweighted κ (0.51) than cases (0.27; P = 0.03), but weighted κ were not statistically different [controls, 0.73; cases, 0.64 (P = 0.32)]. Demographic and anthropometric characteristics were not different between participants who underreported, overreported, or correctly reported their birth weight for either cases or controls. Overall, the level of agreement for report of birth weight and actual birth weight was fair to moderate.