Reliability of cancer family history reported by parents in a case–control study of childhood leukemia
BACKGROUND AND AIMS Family history is used extensively to estimate the risk of colorectal cancer but there is considerable potential for recall bias and inaccuracy. Hence we systematically assessed the accuracy of family history reported at interview compared with actual cancer experience in relatives. METHODS Using face to face interviews, we recorded family history from 199 colorectal cancer cases and 133 community controls, totalling 5637 first and second degree relatives (FDRs/SDRs). We linked computerised cancer registry data to interview information to determine the accuracy of family history reporting. RESULTS Cases substantially underreported colorectal cancer arising both in FDRs (sensitivity 0.566 (95% confidence interval (CI) 0.433, 0.690); specificity 0.990 (95% CI 0.983, 0.994)) and SDRs (sensitivity 0.271 (95% CI 0.166, 0.410); specificity 0.996 (95% CI 0.992, 0.998)). There was no observable difference in accuracy of reporting family history between case and control interviewees. Control subjects similarly underreported colorectal cancer in FDRs (sensitivity 0.529 (95% CI 0.310, 0.738); specificity 0.995 (95% CI 0.989, 0.998)) and SDRs (sensitivity 0.333 (95% CI 0.192, 0.512); specificity 0.995 (95% CI 0.991, 0.995)). To determine practical implications of inaccurate family history, we applied family history criteria before and after record linkage. Only two of five families reported at interview to meet surveillance criteria did so after validation, whereas only two of six families that actually merited surveillance were identified by interview. CONCLUSIONS This study has quantified the inaccuracy of interview in identifying people at risk of colorectal cancer due to a family history. Colorectal cancer was substantially underreported and so family history information should be interpreted with caution. These findings have considerable relevance to identifying patients who merit surveillance colonoscopy and to epidemiological studies.