Childhood exposure to radiation is a well known cause of thyroid cancer, but until recently there were limited data at the high doses used to treat childhood cancers (5
). The improvement in childhood cancer survival (8
) and relatively high risk for developing subsequent treatment-related thyroid cancers (7
) highlighted the need to better understand the role of radiotherapy and the joint effects of radiotherapy and chemotherapy in the development of thyroid cancer. Our pooling used original data from the four studies that had individual thyroid dose estimates and information on chemotherapy. The CCSS-US was the largest individual study, with 57% of all thyroid cancers. However, the inclusion of the LESG, CCSS-Fr/UK and CCSS-Nordic studies increased our ability to evaluate patterns of radiation and chemotherapy risks and interactions between treatment types over a range of time periods (1936 to 2000) and populations (North American and European) beyond those demonstrated in the CCSS-US study alone (7
In this first pooled analysis of childhood cancer survivors, which included 187 thyroid cancer cases, we observed an approximately linear increase in RR of thyroid cancer with dose through about 10 Gy. Above 10 Gy, RRs leveled, then declined above 30 Gy. Study-specific RR estimates at 10 Gy ranged from 6.8 to 12.9. For the pooled data, the best fitting ERR model was linear exponential (linear) with an estimated RR at 10 Gy of 13.7 (95% CI: 8.0–24.0). The maximum fitted RR was 14.9 (95% CI: 8.2–24.5), which occurred at 15.9 Gy.
At 1 Gy, our estimated RR was 3.2 (95% CI: 2.2–5.3), which was lower than the estimated 8.7 at 1 Gy (95% CI: 3.1, 29.7) from the pooled analysis of five cohort studies of populations irradiated primarily at low to moderate doses (5
). However, after adjustment for (non-significant) differences between exposed and non-exposed subjects, the Ron et al
. estimate was 4.8 (95% CI: 2.4, 11.7) and very similar to our estimate.
While the leveling and downturn of the ERR at high doses likely reflected cell killing, it is noteworthy that the ERR remained elevated even above 40 Gy. For lung and breast cancer risks after high-dose radiotherapy for Hodgkin lymphoma, Sachs and Brenner proposed that stem cell repopulation during and after fractionated exposure can compensate for cell killing (25
). They suggested that accelerated repopulation can explain why risks for solid cancer after radiotherapy remains substantially greater than predictions based solely on initiation and cell-killing mechanisms.
Both the ERR and EAR increased with decreasing age at exposure and showed no diminution with time since exposure. Although risks remained elevated throughout the follow-up period, risks appeared to peak 15–19 years after exposure in both the ERR and EAR models. As in previous studies, we found that gender modified the EAR/PY-Gy, with females having double the risk of males, but not the ERR/Gy. This agreed with the pooled analysis of Ron et al
., where females had a higher ERR/Gy, but the difference was not statistically significant and individual studies were inconsistent (5
). Attained age and calendar year of diagnosis both modified the EAR, with risk increasing with increasing age and calendar period of incidence. These results generally were similar to those reported for the CCSS-US study as it was the larger study in the EAR analysis. We found that the effect modification by number of treatments differed by study, with a statistically significantly higher ERR/Gy with greater number of treatments for the CCSS-Fr/UK study, and a non-statistically significantly lower ERR/Gy in the CCSS-US and CCSS-Nordic studies. These findings were consistent with dose rate having only a weak effect on thyroid cell line survival (26
) and were compatible with risk estimates from studies of persons receiving external radiation, or I-131 radiation from Chornobyl during childhood (27
). Finally, we observed lower ERR/Gy estimates among Hodgkin lymphoma patients compared with “other” first cancers (). The lower ERR/Gy persisted for Hodgkin lymphoma patients after including age at exposure as a modifier in the model. The reason for this difference was unknown. Radiation dosimetry would have accounted for treatment-related dose, and our evaluation of ERR/Gy did not vary by tumor size, which suggested comparable rates of thyroid screening among Hodgkin lymphoma patients.
A particularly interesting finding was the joint radiotherapy and chemotherapy relationship. Veiga et al
. recently reported that chemotherapy increased the risk of developing a secondary thyroid cancer after a childhood cancer only if the radiotherapy dose to the thyroid was less than 20 Gy (15
). While overall results were similar, our pooling expanded that case group by 80% and allowed a more detailed examination of RRs for chemotherapy with increasing radiation dose. We observed RRs of thyroid cancer of 3.2 to 4.5 for patients with chemotherapy (alkylating agents, anthracyclines and bleomycin) relative to those without chemotherapy, among those who received no radiation, RRs of 1.4 to 2.6 among those who received doses up to 5 Gy, RRs of 1.1 to 2.4 among those who received doses between 5 to 20 Gy, and no excess RRs among those who received doses >20 Gy. For both alkylating agents and bleomycin, decreasing trends were statistically significant as radiation dose increased, with a similar but non-significant pattern for anthracyclines. Consistent with Veiga et al
., among patients who did not have chemotherapy, we observed enhanced radiation effects in the two cohort studies. However, we did not find a similar pattern in the CCSS-Nordic case-control study, which included only 13 thyroid cancer patients.
Combinations of chemotherapies were common, and data were insufficient to disentangle the effects of individual drugs or of differences in the radiation and chemotherapy interactions. While observed RRs were slightly greater in patients with multiple chemotherapies, differences were nonsignificant. These results were consistent wit Veiga et al
., who found little evidence of differential RRs for treatment combinations, either overall or ≤20 Gy (15
). Although we could not harmonize a consistent metric for drug dose across studies, Veiga et al
. found increased RRs with increasing drug score for alkylating agents.
As in all second cancer studies, we were evaluating older treatment regimens because the occurrence of a second cancer required a minimum follow-up period. With the majority of treatments occurring before 1985, we could not evaluate conformal radiotherapy, intensity-modulated radiation therapy (IMRT) and proton therapy. While these new techniques offer a great advantage in delivering the radiation to a more concentrated target volume and thereby minimize dose to nearby normal tissue, the larger total body dose may increase risk of second cancers due to increased leakage radiation and exposure of a larger tissue volume (28
The treatment of childhood cancer patients is one of oncology’s greatest successes, but second cancer risks from high-dose radiotherapy remain a concern. From this perspective, the ideal treatment plan would give a very high dose to the tumor while minimizing radiation dose to surrounding and distant normal tissue and thus second cancer risks. A better understanding of the shape of the dose-response relationship for radiation-induced cancers potentially allows optimization of treatment plans to minimize second cancer risks from radiotherapy. Our pooled analysis of thyroid cancer adds to the growing literature showing that certain chemotherapeutic agents can increase risks of solid tumors, but may interact with radiotherapy in complex ways. We also show that, while risks decline after high-dose radiotherapy, elevated risks continue for potentially many decades and therefore indicates the need for continued follow up of childhood cancer survivors.