Reliability of self-reported family history of cancer in a large case-control study of lymphoma.

BACKGROUND Case-control studies of familial cancer risk traditionally rely on self-reported family history of cancer, which may bias results due to differential recall between case patients and control subjects. To evaluate the reliability of self-reported data, we analyzed questionnaire and registry-based data on familial cancer from a population-based case-control study of malignant lymphoma. METHODS All 1508 lymphoma case patients and 1229 control subjects completed a telephone interview assessing cancer in family members. Participants were linked to the Swedish Multi-Generation Register and Cancer Register to identify confirmed cancer diagnoses in first-degree relatives. The sensitivity and specificity of self-reported familial cancer were calculated among case patients and control subjects and were compared using logistic regression. All statistical tests were two-sided. RESULTS Lymphoma case patients reported a family history of any cancer with statistically significantly higher sensitivity than control subjects (0.85, 95% confidence interval [CI] = 0.83 to 0.87 and 0.80, 95% CI = 0.77 to 0.82, respectively) but with marginally lower specificity (0.89, 95% CI = 0.87 to 0.91 and 0.92, 95% CI = 0.90 to 0.94, respectively). The sensitivity of self-reporting familial cancers by site ranged from less than 0.20 for rare malignancies to nearly 0.75 for more common types, whereas specificity was generally 0.98 or greater. For most sites, the reliability of self-report was similar in patients and control subjects. However, patients reported familial hematopoietic cancer with statistically significantly higher sensitivity (0.60, 95% CI = 0.57 to 0.62) than control subjects (0.38, 95% CI = 0.35 to 0.40). Odds ratios for the association between familial cancer and risk of non-Hodgkin lymphoma were consistently higher when based on self-reported, compared with registry data-based, family history of any cancer or of hematopoietic cancer. CONCLUSIONS Reliability of self-reported family history of cancer varies between case patients and control subjects. Recall bias may thus produce biased results in case-control studies of familial cancer risk.