Screening effects in risk studies of thyroid cancer after the Chernobyl accident

Abstract

In this article scenarios have been developed, which simulate screening effects in ecological and cohort studies of thyroid cancer incidence among Ukrainians, whose thyroids have been exposed to ¹³¹I in the aftermath of the Chernobyl accident. If possible, the scenarios were based on directly observed data, such as the population size, dose distributions and thyroid cancer cases. Two scenarios were considered where the screening effect on baseline cases is either equal to or larger than that of radiation-related thyroid cancer cases. For ecological studies in settlements with more than ten measurements of the ¹³¹I activity in the human thyroid in May–June 1986, the screening bias appeared small (<19%) for all risk quantities. In the cohort studies, the excess absolute risk per dose was larger by a factor of 4 than in the general population. For an equal screening effect on baseline and radiation-related cancer (Scenario 1) the excess relative risk was about the same as in the general population. However, a differential screening effect (Scenario 2) produced a risk smaller by a factor of 2.5. A comparison with first results of the Ukrainian–US-American cohort study did not give any indication that a differential screening effect has a marked influence on the risk estimates. The differences in the risk estimates from ecological studies and cohort studies were explained by the different screening patterns in the general population and in the much smaller cohort. The present investigations are characterized by dose estimates for many settlements which are very weakly correlated with screening, the confounding variable. The results show that under these conditions ecological studies may provide risk estimates with an acceptable bias.

Appendix

Summation of (3) and division by $N_{\rm{pop}}$ yield

$$ h_{\rm{pop}}=\frac{{1}}{{N_{\rm{pop}}}} \sum_{ijk} h_{ijk}=\frac{{1}}{{N_{\rm{pop}}}} \sum_{ijk} \left(1+ \eta_{ijk}\right) h_{{0}k} + \left(1+\kappa_{ijk}\right) \beta D_{ijk}, $$

(21)

the thyroid cancer risk in the general population. With

$$ D_k=\frac{{1}}{{N_k}}\sum_{ij} D_{ij} $$

(22)

and with $\eta_{ijk}$ and $\kappa_{ijk}$ being independent of i and j, it follows

$$ h_{\rm{pop}}=\frac{{1}}{{N_{\rm{pop}}}}\sum_{k} \left(1+ \eta_{k}\right) N_k h_{{0}k} + \left(1+\kappa_{k}\right) N_k\beta D_{k}. $$

(23)

With

$$ D_{\rm{pop}}=\frac{{1}}{{N_{\rm{pop}}}} \sum_k N_k D_k $$

(24)

and with the definition of $h_0$ (4), (23) may be rewritten as

$$ h_{\rm{pop}} = h_0 + \beta D_{{\rm {pop}}}+\frac{{1}}{{N_{\rm{pop}}}}\sum_k \eta_{k} N_k h_{{0}k} + \kappa_{k} N_k\beta D_{k}. $$

(25)

Solving this equation for β yields

$$ \beta= \frac{h_{\rm{pop}}-h_0 -{\frac{1}{N_{\rm{pop}}}}\sum_k \eta_{k} N_k h_{{0}k}}{{D_{\rm{pop}}+ \frac{{1}}{{N_{\rm{pop}}}}\sum_k \kappa_{k} N_k D_{k}}}. $$

(26)