A comment on: ‘Absorbed radiation doses in the thyroid as estimated by UNSCEAR and subsequent risk of childhood thyroid cancer following the Great East Japan’

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Journal of Radiation Research, Vol. 62, No. 3, 2021, pp. 420–424 doi: 10.1093/jrr/rraa145 Advance Access Publication: 10 February 2021

A comment on: ' Absorbed radiation doses in the thyroid as estimated by UNSCEAR and subsequent risk of childhood thyroid cancer following the Great East Japan' Hidehiko Yamamoto 1

DEAR EDITOR
We were very interested in the article by Ohira et al. [1]. Whereas Tsuda et al. [2], Yamamoto et al. [3], Kato [4] and Toki et al. [5] found a significant association between the occurrence of thyroid cancer and radiation following the Fukushima nuclear accidents, Ohira et al. claim no association between thyroid doses and thyroid cancer risk. Ohira et al. [1] stratified the Fukushima prefecture into four regions defined by the quartiles of the absorbed thyroid dose distribution and assumed that the dose should have been avoided in the evacuation areas. The question arises of whether the evaluation of the thyroid dose including the evacuated municipalities can show a significant correlation. To this end, we considered the municipality-specific counts of thyroid cancers and the person-years in the Fukushima Health Management Survey (FHMS) as published in tables 1 and 2 of Yamamoto et al. [3]. Table 1 supplements this information with the total absorbed thyroid dose to 10-year-old children as estimated by UNSCEAR in the Attachments C-16 and C-18 of its 2013 Report [6]. These internal doses are compiled in the last column of Table 1, whereby the missing dose values in Attachment C-16 for the partly or completely evacuated prefectures were imputed by the dose values in Attachment C-18 taking the proportion of evacuees in the individual municipalities into account by linear interpolation.
Yamamoto et al. [3] found a considerably elevated detection rate per dose-rate of thyroid cancer below 2 μSv h -1 compared with the detection rate ratio from unrestricted data. We built on this finding by performing a segmented regression analysis [7] to determine an optimum dose (mGy) beyond which the slope of the detection rate by dose changes significantly. The dashed light blue elements in Fig. 1 present the corresponding change point analysis based on the deviance criterion [8]. The optimum thyroid absorbed dose of this change point is 21 mGy, 95% confidence interval (CI) 17-24. The detection rate ratio (DRR) below 21 mGy is 1.154 per mGy, 95% CI 1.044-1.277, P value 0.0053, and the residual DRR above 21 mGy is 1.003. The odds ratio and the P value for the interaction (change of slope) are 0.869, 95% CI 0.783-0.965, and 0.0083, respectively. This means that the overall effect is driven by the strong effect below 21 mGy. The solid blue line in Fig. 1 depicts this change point model. The solid black line in Fig. 1 indicates the overall association between the thyroid cancer occurrence and the thyroid absorbed dose in all 59 municipalities of Fukushima after the nuclear accidents. The DRR and the P value for this overall trend are 1.008, 95% CI 1.000-1.017, and 0.0445, respectively. The first-and second-order models are possible alternatives, which cannot be distinguished with certainty based on the data at hand. The presence of significant non-linearity does not mean that a simple linear overall model is inappropriate. If the simple linear model is not significant, this is not evidence of no effect [9].
The raw detection rate (DR r = cases/person-years) and of the adjusted detection rate (DR a = RR a × cases 0 /person-years 0 ), where superscript '0' means the counts of cases (n = 142) and person-years (n = 1 865 957) at zero dose can be determined using table 1 in Lubin et al. [10]. These data are compiled in Table 2 [11]. However, these estimates of the minimum latency are based on few observations and cannot entirely exclude the possibility of earlier disease onset in (unnoticed) highly exposed or particularly sensitive children, see also paragraph '2.2 Induction and latent period, point prevalence, incidence • 420 Table 1. FHMS basic data of the combined first and second screening rounds [3]: municipality, person-years, thyroid cancers, detection rate and UNSCEAR (2013) total thyroid absorbed dose of 10-year-old children (mGy) in the first year after Fukushima derived from the UNSCEAR 2013 Report Attachments C-16 and C018 [6]   proportion and incidence rate, and detection rate' in Yamamoto et al. [3]. In summary, our findings contradict the conclusion of Ohira et al. stating 'No dose-dependent pattern emerged from the geographical distribution of absorbed doses by municipality, as estimated by UNSCEAR, and the detection of thyroid cancer among participants within 4-6 years after the accident' [1]. We conjecture that the negative finding by Ohira et al. [1] may partly be due to a too coarse exposure  Table 2): thin red line and circles. Detection rate from the study of Yamamoto et al. [3]: thick black line and circles. Circle areas proportional to person-years for dose categories. The detection rate ratios (DRRs) per mGy and their 95% confidence intervals are 1.0067, 1.0046-1.0088, P value < 0.0001 for the study of Lubin et al. [7], and 1.0100, 1.0006-1.0196, P value 0.0379 for the FHMS [3].
stratification, the neglect of the evacuation areas and the disregard of the non-linearity of the association between radiation dose and thyroid cancer in the FHMS.