Purpose: To assess the validity of the cohort study participants' self-reported cancer history via data linkage to a cancer registry database.
Methods: We included 143,965 participants from the Health Examinees (HEXA) study recruited between 2004 and 2013 who gave informed consent for record linkage to the Korean Central Cancer Registry (KCCR). The sensitivity and the positive predictive value of self-reported histories of cancer were calculated and 95% confidence intervals were estimated.
Results: A total of 4,860 participants who had at least one record in the KCCR were included in the calculation of sensitivity. In addition, 3,671 participants who reported a cancer history at enrollment were included in the calculation of positive predictive value. The overall sensitivity of self-reported cancer history was 72.0%. Breast cancer history among women showed the highest sensitivity (81.2%), whereas the lowest sensitivity was observed for liver cancer (53.7%) and cervical cancer (52.1%). The overall positive predictive value was 81.9%. The highest positive predictive value was observed for thyroid cancer (96.1%) and prostate cancer (96.1%), and the lowest was observed for cervical cancer (43.7%).
Conclusion: The accuracy of self-reported cancer history varied by cancer site and may not be sufficient to ascertain cancer incidence, especially for cervical and bladder cancers.
Keywords: Epidemiologic studies; Health surveys; Neoplasms; Registries; Reproducibility of results; Self report; Sensitivity and specificity.
Copyright © 2017. Published by Elsevier Ltd.