Model-based estimation of thyroid cancer incidence from ultrasound examinations in the Fukushima Health Management Survey: estimated results considering the non-examinees in the first, second and third rounds of the cohort study

Abstract

Background: In October 2011, the Fukushima prefectural government started a thyroid ultrasound examination (TUE) as part of the Fukushima Health Management Survey following the Fukushima Daiichi Nuclear Power Plant accident. The proportion of examinees is an important factor when interpreting the results.

Objectives: To construct models that assess the relationship between the proportion of non-examinees and the characteristics of eligible participants in the first-round to third-round TUEs. Using these models, estimate the number of thyroid cancer cases in the entire population for each survey, considering non-examinees.

Design: Model-based estimation using cohort survey data.

Settings: Fukushima Health Management Survey from 2011 to 2017.

Participants: Children and adolescents aged 18 years or younger (363 342 individuals) who were identified through resident registration records, resided in Fukushima Prefecture at the time of the accident, and were eligible for each TUE.

Outcome measures: Modelling the relationship between non-examinee status and individual characteristics, and estimating the number of cases in the entire population for each survey round.

Results: The area under the receiver operating characteristic curve of the constructed models ranged from 0.815 to 0.905. In the first-round, second-round and third-round TUEs, 115, 70 and 30 cases were observed among 294 921, 258 771 and 208 955 examinees, respectively, whereas the estimated number of cases in the entire population including the non-examinees was 177.3 (95% CI 167.0 to 188.0), 126.3 (95% CI 106.3 to 150.2) and 49.7 (95% CI 35.8 to 71.9), respectively.

Conclusions: These estimates were higher than the actual number of observed cases because they considered non-examinees. Our model for non-examination showed a high discriminant accuracy and was considered to capture well the factors that resulted in non-examinees. This study's findings provide valuable information for studies considering the number of potential thyroid cancer cases among non-examinees and may facilitate appropriate interpretation of reports and prospective survey outcome management.

Keywords: EPIDEMIOLOGY; Mass Screening; PUBLIC HEALTH; STATISTICS & RESEARCH METHODS; Thyroid disease.