Method for assessing the radiation risks of the solid cancer incidence accounting for possible diagnostic errors

Abstract

A bias in radiation risk estimates may be caused by diagnostic errors. In this paper radiation risk estimates are examined with the use of the contingency table 2х2 of irradiated cancer cases with account of sensitivity and specificity indicators of diagnostic methods accuracy were examined. Radiation risk is estimated by metrics of the odds ratio (OR) and relative risk ratio (RR). It is shown that the estimation of radiation risk in the RR metric did not depend on the diagnostic sensitivity indicator, and therefore is preferable, compared to the OR metric. When the specificity of the diagno-sis is reduced, the RR value increases, compared to the risk estimated with the use of observed number of cancer cases regardless of the specificity. As a numerical example, data on trachea, bronchus and lung cancers in a cohort of the Chernobyl cleanup workers of Russia formed by using data monitored within frames of the National Radiation Epidemiological Register (NRER) from 1992 to 2020. During this period, 2,222 cancer cases were reported in the cohort of 67,587 people. The cohort members were divided into two groups, those who were unexposed to radiation (radia-tion doses less than 0.1 Gy) and other, exposed to radiation. If the specificity indicator is taken to be 100%, then the RR=1.07 at 95% CI (1.02; 1.13). The estimates of radiation risks of malignant neoplasms obtained directly from the observed number of cancer cases in exposed and unexposed groups of people, excluding the specificity and sensitivity indicators of diagnosis accuracy, are the lower estimates of radiation risk. As the probability of diagnostic errors increases (as the specificity and sensitivity of diagnostic tests decrease), and when these errors are taken into account, the es-timates of radiation risks increase. If the specificity is reduced to 98.7% CI radiation risk estimates for trachea, bronchus and lung cancer among cleanup workers of the Chernobyl accident increase to RR=1.13, but it is within 95% CI of the RR estimate if specificity is 100%. The radiation risk as-sessment approach used in the study, based on the comparison of exposed and unexposed groups of cleanup workers in OR or RR metrics, imposes high requirements to the specificity indicator of the complex of diagnostic methods in trachea, bronchus and lung (Sp>98.7%) in order to obtain unbiased estimates of radiation risks of cancer in the cleanup workers received dose below 1 Sv. These requirements are not supported by current screening methods, they can be fulfilled as a result of longterm medical monitoring within the system of the NRER.