Association of Radioactive Iodine, Antithyroid Drug, and Surgical Treatments With Solid Cancer Mortality in Patients With Hyperthyroidism

This cohort study compares the association of radioactive iodine or antithyroid drug treatments for hyperthyroidism vs surgical management with solid cancer mortality in patients without cancer.


Introduction
Overt hyperthyroidism is diagnosed in approximately 0.5% of the US population. 1 The most common diagnosis is Graves disease, with a lifetime prevalence of approximately 3% for women and less than 1% for men. Radioactive iodine (RAI), antithyroid drugs (ATDs), and thyroidectomy alone or in combination are the 3 main modalities used in the management of hyperthyroidism; however, treatment practices vary substantially over time and by geographic region. For toxic nodular goiter (TNG), RAI and thyroidectomy are the primary treatment options, with long-term ATD treatment offered less frequently. 1 For uncomplicated Graves disease, RAI has been the treatment of choice in the US, whereas ATDs have been the more commonly preferred option in most other regions of the world. [2][3][4] However, in the US and Europe, RAI treatment of Graves disease has been decreasing in recent decades in favor of ATDs. 2 Evidence of long-term risks of leukemia and other secondary malignant tumors after postsurgical adjuvant treatment of differentiated thyroid cancer with RAI may have contributed to this trend. [5][6][7][8][9] Studies evaluating the long-term adverse effects of RAI and other treatment options for hyperthyroidism may help to better inform clinical treatment practices and guidelines. Because of the lack of large randomized clinical trials, information on long-term associations of treatment with cancer and other outcomes has, thus far, relied on results from observational studies. However, results from the few large observational studies [10][11][12][13][14][15][16] on this topic have been inconsistent, and summarizing and comparing the findings across studies are challenging because of differences in sample size, duration of follow-up, information on treatment types and doses, statistical methods, and ability to control for confounding and other sources of bias. Confounding is especially of concern for studies that rely on an external comparison group (eg, general population rates) because patients can differ in many important ways from the general public in terms of genetic, environmental, and behavioral factors and access to care. In addition, the underlying disease itself may be directly associated with risk of the outcome under study independently of any treatment effects. 17,18 Another source of bias is confounding by indication (or treatment allocation bias) in which the reason or choice of one treatment vs another is associated with the baseline risk of the outcome under study (eg, patients' age, sex, general health status, and disease origin and severity). 19,20 Thus, although some of these studies have found increased incidence or mortality from cancers of the stomach, 11-13 kidney, 12 breast, 12 and thyroid 10,14 among RAI-treated patients, the direct effects of treatment are difficult to discern in the presence of uncontrolled confounding.
With use of data from the Cooperative Thyrotoxicosis Therapy Follow-up Study (CTTFUS), a large cohort study of patients treated for hyperthyroidism between the 1940s and 1960s in the US and UK, a positive association between organ-absorbed dose and mortality from solid cancers (including breast cancer) was previously reported. 21 The focus on RAI-treated patients was a strength because it helped to minimize confounding caused by imbalances in risk factors across treatment groups and provided a direct assessment of the risk per unit dose. 22 Nonetheless, comparisons of risks across treatment groups are of interest for practitioners involved in the management of hyperthyroidism who must select the treatment modality and the extent of treatment.
Thus, for the current study, we used data from the CTTFUS cohort to evaluate the association between treatment received (RAI, ATDs, surgery, and various combinations of these) and risk of solid cancer death. We also investigated these risks across levels of RAI-administered activity (as a more intuitive measure and proxy for organ-absorbed doses). Results presented in the current article informed the methodologic approach used in a recent analysis 21 of organ-absorbed dose. With an additional 24 years of follow-up, the current study also provides an update of the standardized mortality ratios (SMRs) for solid cancer published by Ron et al 10 more than 2 decades ago. This cohort study included data from the original CTTFUS cohort, which included 35 593 patients   with hyperthyroidism treated with RAI, ATDS, surgery, or some combination of these at 25 US and 1 British medical centers between January 1, 1946, andDecember 31, 1964. 10 Comprehensive clinical data were abstracted from medical records, and early follow-up information was obtained from treating physicians, clinics, or patients when no medical sources were available. In 1984, the National Cancer Institute and 4 regional study centers reassembled patient and treatment data from computer lists, microfiche, microfilm cassettes, and handwritten material; all but 4010 patients (n = 31 583) were eligible for subsequent tracing and mortality follow-up through 1990. 10 Follow-up of US patients was recently extended through December 31, 2014. 21 Person-time for all patients was defined as the date of the first study examination through the date of death, date of loss to follow-up or last known vital status, or end of study, whichever occurred first. This study was approved by the institutional review board of the National Cancer Institute of the National Institutes of Health.

Study Population
Because mortality follow-up was based on linkages with available databases and involved no direct contact with study participants, the requirement for informed consent was waived. All data were deidentified. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
For the current analyses, we additionally excluded patients with no follow-up time (n = 206), those with missing entry or exit dates (n = 8), and those whose exit dates occurred on or before the study entry date (n = 6) ( Figure). Of the 31 363 remaining patients, death records were available for 25 445 (81.1%), whereas 2829 (9.0%) were censored at their last known vital status date and the others were found or presumed to be alive by the end of follow-up on December 31, 2014 (3089 [9.8%]). Data analysis was performed from August 1, 2019, to April 23, 2020.

Statistical Analysis
For external comparisons, SMRs and 95% CIs were used to compare the observed number of solid cancer (including breast and nonbreast solid cancer) deaths in the cohort with the expected number of deaths based on age-, sex-, and calendar year-specific mortality rates in the general populations of the US and UK. The SMRs were calculated separately by treatment group. Influence of possible sources of bias were examined through exclusion of specific patient groups (eg, those with comorbidities) or person-time (to account for potential exposure-disease latency periods).
For internal cohort comparisons, Cox proportional hazards regression models were used to estimate multivariable-adjusted hazard ratios (HRs) and 95% CIs for the association between treatment groups (or categories of total administered activity) and risk of solid cancer death. All Figure. Description of the Cooperative Thyrotoxicosis Therapy Follow-up Study and Inclusion and Exclusion Criteria for the Current Analysis models excluded patients with prior cancer at study entry (n = 1013) or between entry and date of last RAI treatment (n = 36), used attained age as the time scale (incorporating a 5-year latency period), and were adjusted for sex and underlying diagnosis (Graves disease, TNG, and other or unknown). Statistical analyses were conducted with Stata, version 16.0 (StataCorp) and Epicure, version 2.00.02 (Risk Sciences International). A 2-sided P < .05 was considered to be statistically significant.

External Comparisons
The SMRs by treatment group are given in Table 3

Internal Comparisons
After excluding prior cancers and adjusting for patient age, sex, and underlying diagnosis, we found no significant differences in the HR of solid cancer death across treatment groups ( Table 4). None of the findings above changed notably after restricting the findings to patients with Graves disease.

Discussion
Using data from a large cohort of patients treated for hyperthyroidism between the mid-1940s and mid-1960s, we evaluated the association between treatment received (RAI, ATDs, surgery, or combinations) and long-term risk of solid cancer death. After accounting for known sources of bias, we observed no significant differences in risk across treatment groups based on external and internal cohort comparisons, and notably, we found no evidence of an increased risk of solid cancer death associated with use of ATDs. Among RAI-treated patients, however, total administered activity was positively associated with risks of all solid cancers, female breast cancer, and nonbreast solid cancer death, with increases of 8% to 10% per 370 MBq. The direction and magnitude of these associations were consistent with previous findings based on estimated organ-absorbed doses. 21 Caution is needed when comparing risks across treatment groups using observational study data, particularly when relying on general population rates as the comparison group. SMRs may be less than, equal to, or greater than unity (1.0) for reasons other than the treatment (eg, genetic factors, the disease, environmental and behavioral factors, and access to care) and should be interpreted in the context of other SMRs. Internal comparisons (eg, HR estimates) provide greater reassurance against bias by restricting comparisons to individuals with the same disease and allowing for covariate adjustment. Nonetheless, these approaches yielded similar results after accounting for major known sources of confounding.
Although some have argued that the lack of an unexposed comparison group was an important limitation of the previous analysis of organ and tissue absorbed doses and cancer mortality, [23][24][25] the results of the current analysis suggest that the focus on RAI-treated patients was actually a strength because their inclusion could have biased our estimates of the radiation-associated risks. 26 For instance, the surgery-only group included a higher proportion of TNG than Graves disease diagnoses (44.5% vs 7.9% in the full cohort), and the drug-only group included a high proportion of patients with baseline comorbidities, including prior cancer diagnoses, raising the possibility that these groups may have had different baseline risks of cancer death for reasons unrelated to the treatments.
The elevated solid cancer mortality rate in the drug-only group, observed previously by Ron et al, 10 has been suggested to indicate direct association of ATDs with cancer risk [23][24][25] ; however, confounding by indication seems to be the more likely explanation. As was demonstrated in the earlier study, 10 this association was no longer apparent after excluding patients with prior cancer diagnoses, and no association was observed after the first 5 years of follow-up. In addition, similar elevations in solid cancer mortality were not observed in the drugs and surgery, drugs and RAI, or drugs, surgery, and RAI treatment groups. Although numerous changes have occurred in the management of hyperthyroidism since the 1940s to 1960s, it seems plausible that patients presenting with multiple comorbidities or short life expectancies were more likely to have been managed using only ATDs rather than RAI and/or surgery. Although excluding prior cancers helped minimize this bias, it also negatively biased all the SMRs to some extent (as can be observed in Table 3) because the same exclusion could not be performed for the expected (general population) rates.
More generally, the CTTFUS cohort is not well suited to the evaluation of long-term risks associated with ATDs. The ATD formulations have changed over time, as methimazole has replaced propylthioracil as the preferred option for most patients with Graves disease because of evidence of adverse effects of propylthioracil on the liver. 27,28 Propylthioracil was likely the most commonly prescribed ATD at the start of cohort follow-up. As described previously, 10 no information was abstracted or retained from the medical records of these patients that would allow us to distinguish between the specific ATD types or to reconstruct the prescribed doses, and we had incomplete (and sometimes inconsistent) information on drug treatment starting and stopping dates. Considering the temporal changes in ATD treatment, more contemporary cohort studies are needed to evaluate the long-term adverse effects of methimazole specifically.
Ionizing radiation, however, is an established carcinogen. 29 The breast is 1 of the most radiosensitive cancer sites, with epidemiologic studies 29-34 finding consistent positive associations between absorbed dose to the breast, even in the low-dose range, and subsequent risk of breast cancer. In contrast to ATDs, data on RAI treatment dates and administered activity were nearly complete for this cohort. A previous study 21  Previous estimates of risk per unit organ dose were consistent with those from other observational studies of radiation-exposed populations. [30][31][32][33][34] In the current study, the lack of discernible differences in solid cancer risk across treatment groups may be partly explained by the low and narrow range of estimated that 20 to 30 excess solid cancers may occur per 1000 patients treated as a direct result of the radiation exposure, although most excess deaths would likely occur more than 20 years after initial treatment. Thus, for many patients, the benefits of RAI may outweigh these risks.

Limitations
This study has limitations. Those not already mentioned include lack of information on some potential confounding factors, such as cigarette smoking, obesity, and reproductive factors 21,26 ; however, we found no association between treatment type (or radiation-absorbed dose to the lung 21 ) and lung cancer mortality, suggesting that a major smoking-related bias did not occur. We lacked laboratory measures of thyroid function to assess severity of the underlying disease. The reliance on cancer mortality, as opposed to incidence, follow-up was another limitation because we could not distinguish between factors associated with cancer development vs survival. In addition, mortality follow-up is not ideal for capturing cancers with high survival rates (eg, thyroid or breast) or for studying cancer subtypes. Findings related to RAI treatment may not be generalizable to patients with thyroid cancer, who typically receive higher administered activities (>3700 MBq) in the context of postsurgical remnant ablation or therapy for presumed or known advanced or metastatic disease, potentially yielding higher or lower doses to individual organs or tissues, depending on the size and location of the residual thyroid tissue. 36

Conclusions
This study found positive associations between total administered activity of RAI and risks of death from total solid cancer, female breast cancer, and nonbreast solid cancers. These findings were consistent with earlier results from this cohort based on organ-absorbed doses. 21 The low, narrow range of total administered activity and the potential for residual confounding by risk factors not captured in this study may have hindered our ability to detect differences in risk between RAI-treated and non-RAI-treated patients. No evidence supporting an association between ATD treatment and risk of solid cancer death was found in this cohort.