Our results indicate that exposure to lower doses of therapeutic chest radiation (median 0.95 Gy, mean 1.29 Gy) during childhood is associated with an increased cumulative incidence of thyroid cancer, even up to a median follow-up of 57.5 years. This association remained even after adjusting for suspected thyroid cancer risk factors and factors previously associated with thyroid cancer in our cohort. Furthermore, our analysis of thyroid cancer incidence by time since radiation exposure indicates that although relative risk decreases with time since exposure, chest irradiation is still associated with an increased incidence of new thyroid cancers four to six decades after exposure.
Given that only two thyroid cancers occurred in the irradiated siblings after another malignancy had been diagnosed, the increased thyroid cancer incidence cannot be attributed solely to the secondary effects of chemotherapy or secondary radiotherapy. Our findings clearly confirm the results of other cohorts exposed to radiation during childhood, such as the Swedish Hemangioma cohort (30
), the Israeli tinea capitis study (31
), the Michael Reese Hospital benign head and neck cohort (32
), and the data on exposure during childhood from the atomic bomb survivor cohort (33
In comparing these findings to the results from the last follow-up of this cohort, which ended in 1987, it is notable that the unadjusted ERR/Gy fell from 9.0 (90% CI: 4.2–21.7) (16
) to the current 4.3 (95% CI: 2.2–8.7). This is most likely due to increasing time since exposure (25
) and slight differences in dose estimates used. Although revised dose estimates were used for this paper but not prior studies of this cohort, the new dosimetry tended to flatten the dose range by increasing the exposure estimates at the bottom end and decreasing dose estimates at the high end of the exposure range. Modeling using the old dosimetry revealed that the ERR/Gy estimate would be even lower, at 3.4 (95% CI: 1.7 to 6.8).
We again found that a linear model of dose fit the excess relative risk model better than a linear-quadratic or a quadratic model of dose. The linear model also fits better than a more complex radiobiological model, which included a term that allowed for a downturn at high doses as a result of “cell sterilization” (i.e., ERR(D) = (β1
D + β2
) * [exp(−β3
D − β4
). Other studies (25
) suggest that at thyroid doses less than 6.00 Gy, the linear dose model appears to fit thyroid cancer incidence data as well as or better than the other three alternatives above. In our cohort, less than 1% of the irradiated subjects with a known dose had an estimated thyroid exposure greater than 6.00 Gy.
Our study confirms the findings of an earlier pooled analysis of five cohort studies of external radiation exposure during childhood, one of which was our cohort with its follow-up until 1987 (25
). This pooled analysis demonstrated that ERR began to decline after 30 years of exposure but was still elevated at 40 years after exposure. Our study adds to the literature by showing that, at least in children exposed at a very young age, the increased risk continues at least up to a median 57.5 years after exposure. The pooled analysis, as well as a recent analysis of thyroid cancer incidence in the Childhood Cancer Survivors Study (37
), also found that risk decreased significantly with increasing age at exposure, with little risk apparent after 20 years of age. However, we could not assess the effect of age at exposure on ERR, because more than 95% of the thymus cohort was exposed at less than 1 year of age.
In terms of other potential risk factors, only Jewish religion and history of goiter appeared to have an independent association with thyroid cancer incidence when radiation was included in the risk model. The previous report from this cohort found that Jewish religion was an independent risk factor for thyroid cancer but that goiter was not after accounting for concurrent diagnoses of goiter and thyroid cancer (16
). Unfortunately, we were unable to adjust for concurrent diagnoses, because records of the date of goiter diagnosis were no longer available. Nevertheless, ERR/Gy estimates were not more than 10% different when we did not include history of goiter in our models, further suggesting that goiter is not a true independent risk factor for thyroid cancer. In terms of modifying the risk of radiation, in this analysis neither Jewish religion nor goiter appeared to have an interactive effect with radiation dose () or exposure, although in the prior report on thyroid cancer in this cohort Jewish religion interacted with radiation dose to modify ERR/Gy (16
) but not all studies (39
) of nonirradiated populations have suggested that older maternal age at first birth increases thyroid cancer risk. The last follow-up of this cohort found that maternal age at first birth ≥25 years was an independent risk factor for thyroid cancer after adjusting for thyroid radiation dose and attained age (16
). In the current follow-up, we found that the relative risk of age at first birth ≥25 years was of only borderline significance after adjusting for attained age and radiation dose, but it was no longer significant after adjusting for Jewish religion and history of goiter as well. There was no evidence for interaction between maternal age at first birth and radiation dose. Unfortunately, no other study of a medically irradiated cohort has evaluated this question.
A limitation of the study is the fairly low response rate, which was differential between the irradiated and nonirradiated siblings. We evaluated whether this response pattern might lead to nonresponse bias, threatening internal validity. However, differences in potential thyroid cancer risk factors among responders and nonresponders by treatment group are minimal and would tend to cancel each other out (). Certain factors (e.g., female sex, Jewish religion, age at first birth) differed similarly between responders and nonresponders in both groups, which would not introduce analytic bias. Other factors were differential between responders and nonresponders for only one study group (e.g., smoking, hypothyroidism and oral contraceptive use), but those factors have little or no association with thyroid cancer risk, so no impact on the radiation risk estimates is expected.
This study depended on self-report and death-record identification of incident thyroid cancers. Although self-report may lead to bias, we confirmed 75% of the thyroid cancers in the cohort by pathology reports or medical records and attempted to confirm them all. Only one self-reported thyroid cancer turned out to be a benign tumor (which was not included in the current analysis as a thyroid cancer). In collecting medical records for other diagnoses, we found no thyroid cancers that had not been self-reported.
Although irradiated subjects may have known that they were at increased risk, and in the late 1970s to early 1980s free thyroid screening was offered to those who still lived in the area, ascertainment bias is unlikely to have markedly altered our results. First, analyses performed during the last follow-up in this cohort showed that even after adjusting for screening frequency, the estimated excess relative risk was not changed substantially. This analysis was done after the free screening had been offered. Second, since the last article on the cohort was published, the study had been dormant for more than 10 years, so there were no systematic reminders to subjects to have thyroid cancer screening (16
), nor has there been a consensus on the screening procedure for those exposed to irradiation of the thyroid.
Our study has several strengths. First, to our knowledge, the median time since irradiation is longer than that of any other radiation-exposed cohort followed for thyroid cancer, other than the atomic bomb survivor cohort (25
). Among other medically irradiated cohorts, the Israeli tinea capitis study previously had the longest follow-up for thyroid cancer incidence, with a median of 46 years and maximum of 54 years (31
). Second, within the cohort, the sibling comparison group helps control for confounding from family history and potential risk factors that were not collected but are related to upbringing. Third, although the radiation received by our cohort differs from that used today in terms of dose distribution and less precise techniques, this exposure is more similar to the therapeutic and diagnostic radiation received by patients today than is the whole-body radiation received by atomic bomb survivors. Fourth, in our cohort, radiation was not administered in response to cancer, so our findings are not confounded by the possibility that an initial malignancy may be a marker of cancer susceptibility or by chemotherapy effects.
Finally, thyroid doses in our cohort overlap with those from current medical practices that expose the thyroid to radiation during early childhood. The higher doses in this cohort overlap with exposures to the thyroid from radiotherapy protocols for various childhood cancers (42
). The lower doses in this cohort are similar to those to the thyroid of infants from chest and neck CT. In a recent dosimetry study, the thyroid radiation dose from a single torso CT was calculated to be 0.011 Gy in a newborn and 0.014 Gy in a 1-year-old (43
). In another study using various pediatric-sized phantoms, the scattered dose to the thyroid from neck CT examinations varied from 0.015 to 0.052 Gy (18
). Doses can be twice as high if the settings are not changed from adult levels, and a substantial fraction of CT-scanned patients receive multiple scans (2
). In fact, exposures at the lower end of the dose range in our cohort were unfractionated, and thus may be more like those from pediatric chest CT than in any other studied cohort.