Cancer Mortality Following Treatment for Adult Hyperthyroidism FREE

HYPERTHYROID DISEASE is relatively common, particularly among women. The incidence of Graves disease, the cause of over 80% of hyperthyroidism in the United States, is between 0.02% and 1%.^1- 3 Hyperthyroidism can be treated with radioiodine (iodine 131), surgery (subtotal thyroidectomy), or antithyroid drugs, but ¹³¹I is the treatment of choice for most adults in the United States.⁴ The extensive use of ¹³¹I therapy has raised concern regarding its potential carcinogenic and leukemogenic effects, although there is little evidence to support this concern.^5- 9 Public interest in the late health effects of ¹³¹I exposure was rekindled by the 1986 Chernobyl accident and acknowledgment of past ¹³¹I releases from nuclear reactors and bomb testing.

The Cooperative Thyrotoxicosis Therapy Follow-up Study began in 1961. It included 35609 patients with hyperthyroidism treated between 1946 and 1964 at one of 26 study clinics. When the study ended in 1968, after a mean follow-up of 8.2 years, thyroid cancer incidence and mortality¹⁰ and leukemia incidence⁵ were not significantly elevated in the ¹³¹I-treated patients compared with other patients. In a more recent follow-up of patients from the Mayo Clinic, Rochester, Minn, and Massachusetts General Hospital, Boston, total cancer mortality was not associated with¹³¹ I therapy.^11- 12 However, the incidence of cancers in organs that concentrate iodine was elevated (relative risk [RR], 1.8; 95% confidence interval [CI], 1.1-3.2) among the ¹³¹ I-treated patients at the Mayo Clinic,¹³ and breast cancer incidence was slightly increased (RR, 1.3; 95% CI, 0.8-1.9) in the Massachusetts General Hospital patients.¹⁴ Although no association between ¹³¹I exposure and mortality from leukemia (standardized mortality ratio [SMR], 0.98; 95% CI, 0.64-1.42) or breast cancer (SMR, 0.86; 95% CI, 0.69-1.02) was seen among Swedish hyperthyroid patients, a significant excess of respiratory (SMR, 1.26; 95% CI, 1.04-1.49) and digestive (SMR, 1.14; 95% CI, 1.03-1.25) cancer mortality was observed. During the first 10 years of follow-up, there was an excess of thyroid cancer mortality (SMR, 3.22; 95% CI, 1.54-5.98), which completely disappeared after 10 years (SMR, 0.66; 95% CI, 0.08-2.37).¹⁵ To assess the long-term carcinogenic effects of treatment for hyperthyroidism, we conducted a mortality follow-up, through 1990, of the original Cooperative Thyrotoxicosis Therapy Follow-up Study population.

The original Cooperative Thyrotoxicosis Therapy Follow-up Study included all hyperthyroid patients treated with ¹³¹I, thyroid surgery, antithyroid drugs, or any combination of these treatments, between January 1, 1946, and December 31, 1964, at 25 US and 1 British medical center (Table 1). Comprehensive clinical data were abstracted from medical records and hyperthyroidism was classified as Graves disease in 91% of the patients, toxic nodular goiter in 8%, and unknown in 1%. Forty percent of the patients were known to have received some treatment for hyperthyroidism prior to study entry; 75% of these patients had received antithyroid drugs only, 22% had undergone thyroidectomy alone or in combination with antithyroid drugs, and 2% had received ¹³¹ I therapy with or without other treatments. Follow-up information was obtained from the treating physician or clinic or from the patient when no medical source was available. At the end of the first follow-up, 6228 patients (17.5%) had died and 1432 (4.0%) were lost to follow-up. The original study methods have been described by Saenger et al,⁵ Dobyns et al,¹⁰ Becker et al,¹⁶ and Tompkins.¹⁷

Using computer lists, microfiche, microfilm cassettes, and handwritten material, a master file of 35630 patients from the original study was assembled in 1984 (Figure 1). The National Cancer Institute worked with 4 regional study centers, Harvard University, Boston, Memorial Sloan-Kettering Cancer Center, New York, NY, University of Southern California, Los Angeles, and Research Triangle Institute, Research Triangle Park, NC, to conduct a second mortality follow-up of the study cohort. Follow-up was not attempted for 4010 patients, including 1414 patients from 3 small treatment centers not in the geographical areas of the other centers and 2596 patients whose names had been removed from the medical records by the treatment clinics after completion of the first study. The exclusion of these patients from further follow-up was due to technical reasons and was in no way related to patient characteristics. With these exclusions, 31620 patients were eligible for further follow-up. Since 6228 patients were known to have died during the first follow-up, 25392 patients remained to be traced. After supplemental identification data were abstracted, study subjects were traced using records from the National Death Index, Health Care Financing Administration, Social Security Administration, state vital statistics offices, state motor vehicle administrations, local tracing resources, and commercial tracing companies. Copies of death certificates were obtained for deceased patients and were coded according to the International Classification of Diseases, Eighth Edition (ICD-8).¹⁸ Causes of death from the first study were recoded from the International Classification of Diseases, Seventh Edition (ICD-7)¹⁹ to ICD-8 by trained nosologists. After final editing of the data tape, 35593 patients were included in the study cohort.

In the main analyses, we compared the observed number of deaths with the expected number based on US national mortality rates. Person-years at risk were computed for each patient from the date of study entry (the first time a patient visited a participating study clinic during the study enrollment period) until the date of death, the date last known to be alive for those lost to follow-up, or the end of follow-up (December 31, 1990) for those known to be alive. Standardized mortality ratios (the ratio of observed to expected number of deaths) and 95% CIs were computed assuming a Poisson distribution for the observed number. Cause-specific expected numbers of deaths were calculated by applying the age-, sex-, race-, and calendar-year–specific mortality rates to the appropriate person-years at risk.²⁰ Because we were using US mortality rates as the comparison population, the 1699 patients from England had to be excluded from these analyses. In addition, we excluded 146 patients for whom information on sex, date of birth, or date of study entry was missing, leaving 33748 patients.

The SMR analyses were performed separately for patients with Graves disease and toxic nodular goiter and for 7 treatment categories: ¹³¹I treatment only; ¹³¹ I treatment and surgery; ¹³¹I treatment and drugs; ¹³¹ I treatment, surgery, and drugs; surgery only; surgery and drugs; and drugs only. Because results were virtually the same for the 4 groups of patients treated with ¹³¹I and for the 2 surgery groups, we present results for combined treatment groups. Dates of onset of treatment were unknown for patients receiving drug therapy; however, since 98% of the ¹³¹I-treated patients and 78% of patients who underwent surgery received their first treatment within 2 years of study entry, study entry date was used as a surrogate date for evaluation of latency. Only 231 patients had surgery before ¹³¹ I treatment. Since deaths occurring within 1 year of study entry are not likely to be caused by treatment for hyperthyroidism, these deaths were excluded in all analyses related to treatment. As in all mortality analyses using US national mortality rates as a comparison, patients with preexisting cancers were not excluded because they are included in the national rates.

Doses from ¹³¹I to 17 organs were estimated for study subjects by multiplying the amount of administered activity by the dose factors (age [1, 5, 10, 15, and >15 years old], and 24-hour thyroid uptake [0%, 25%, 35%, 45%, and 55%]) provided for each organ in current International Commission on Radiological Protection tables.²¹ Fewer than 500 patients were aged 15 years or younger and most patients (70%) had at least one 24-hour thyroidal ¹³¹ I uptake greater than 55%. Because many clinics treated all patients with a standard amount of administered activity, thyroid uptake measurements were not performed before treatment. We estimated these uptakes using the experience of the 3162 uptakes performed within 1 month prior to treatment, modeled on sex, age at treatment, number of treatments, calendar year of treatment, clinic, type of hyperthyroidism, and administered activity of ¹³¹I. We did not estimate dose to the thyroid because gland size, dose distribution, and effective half-life must be known. However, at least 50 to 70 Gy delivered to the thyroid is usual when treating hyperthyroidism.²² Only 3 other organs received an average ¹³¹ I dose above 100 mGy: stomach (178 mGy), bladder (128 mGy), and small intestine (108 mGy). The average estimated dose to the red bone marrow was 42 mGy and to the breast was 32 mGy. Using a different method of estimation, Saenger et al⁵ reported the mean bone marrow dose to be between 80 and 160 mGy. Their estimate is 2 to 4 times higher than ours, probably because based on information from the literature at that time,^23- 24 they assumed that the blood dose was 1.7 times the administered activity and that the bone marrow dose was between 46% and 86% of the blood dose. More recently, Zanzonico et al²⁵ reconstructed doses for 468 patients from the original study group and their results were similar to ours: for patients with Graves disease, average doses to the bladder, red bone marrow, and breast were 190 mGy, 50 mGy, and 30 mGy, respectively. To ensure that the imputed uptakes did not significantly affect the analyses, we also estimated doses assuming all unknown uptakes to be 25% and 55% (the lower and upper limits used in the International Commission on Radiological Protection tables). Dose estimates based on 55% uptakes were similar to those based on the imputed uptakes. When 25% uptakes were assumed, dose estimates were somewhat lower, but few hyperthyroid patients would have uptakes below 50%.

Because of particular interest in thyroid, hematopoietic, and lymphoproliferative malignancies, we used Poisson regression methods for analysis of grouped cohort time-to-exposure data to evaluate the relationship between deaths from hematopoietic and lymphoproliferative malignancies and ¹³¹I administered activity, as well as estimated absorbed bone marrow dose. For thyroid mortality, we could only assess administered activity of ¹³¹I because we were unable to estimate organ doses. Since these analyses did not use an external comparison, we included the patients from England, but excluded patients with cancers diagnosed before study entry. The data were cross-classified by sex, age at treatment (0-29, 30-39, 40-49, 50-59, and ≥60 years), time since treatment (0-4, 5-9, 10-19, and ≥20 years), type of hyperthyroidism (Graves disease, toxic nodular goiter), administered ¹³¹I activity (0, 0.01-4, 5-6, 7-9, 10-14, 15-19, and ≥20 mCi), and for the analysis of hematopoietic and lymphoproliferative malignancies, red bone marrow dose estimates (0, 0.1-14, 15-19, 20-24, 25-29, 30-39, 40-49, 50-79, and ≥80 mGy). We fit a linear excess RR dose-response model (the background disease rate among the nonexposed×[1+the excess RR]), taking the described covariates into consideration. Background rates were stratified by sex, age at treatment, time since treatment, and type of hyperthyroidism. Maximum likelihood parameter estimates were obtained, and likelihood ratio tests were used to assess the significance of the dose response.²⁶ Trend tests were calculated using ¹³¹I exposure as a continuous variable. EPICURE software (HiroSoft International Corp, Seattle, Wash) was used for these analyses.²⁷

At the end of follow-up, the 35593 study subjects (79% female) had 738831 person-years of follow-up. The mean age at study entry was 46 years and 3% were younger than 20 years. There were 10760 patients (30.2%) still alive, 17959 (50.5%) were deceased, and 6874 (19.3%) were lost to follow-up. Eighty-eight percent of the study population had 5 to 10 years of follow-up and 76% had 10 years or more. The mean length of follow-up from first clinic visit was 21 years (range, 1-44 years). Overall follow-up was better for men (83% vs 79% for women), older patients (86% for those ≥50 years at study entry vs 77% for those <50 years), and patients treated with surgery (88% vs 78% for ¹³¹I and 75% for drug treatment). The difference by treatment group is due to the low proportion of surgery patients (13%) in the group of 4010 patients for whom follow-up was not conducted for technical reasons. There were no differences in follow-up by treatment among the 31620 patients for whom follow-up was conducted in the current study (92%, 92%, and 90% for surgery, ¹³¹I treatment, and drug treatment, respectively).

Approximately 65% of the study subjects had been treated with ¹³¹I; 25% had received ¹³¹I treatment alone and 39% had received ¹³¹I in addition to surgery and/or drugs (Table 2). The average ¹³¹ I administered activity was 6.1 mCi (SD, 5.2) per treatment. Patients treated with ¹³¹I received 1.8 treatments on average, resulting in a cumulative mean total administered activity of 10.4 mCi (5th-95th percentile=3, 27 mCi). Both the mean number of treatments and cumulative administered activity were higher in patients with toxic nodular goiter (2.1 treatments and 17.0 mCi) compared with patients with Graves disease (1.7 treatments and 10.0 mCi). The mean ¹³¹ I thyroidal uptake was 62.4% among the 20639 patients for whom an uptake value was recorded at first examination. Slightly more than 40% of the patients underwent surgery, almost always in conjunction with antithyroid drugs. Only 4% of the patients received drug therapy alone. Women were younger and had surgery somewhat more frequently than men, and patients receiving surgery or drug therapy alone were slightly younger than those treated with ¹³¹ I. Patients with Graves disease were more likely to receive ¹³¹I treatment (65% of the Graves patients) than surgery (30.7%) or antithyroid drugs (3.8%). In contrast, 43.2% of the patients with toxic nodular goiter received ¹³¹I therapy compared with 54.5% treated surgically and 2.3% treated with drugs.

At the time of study entry, 1082 patients had a history of 1338 prior cancers. Twenty-three percent of these patients died of a malignancy during the study period. Patients treated with drug therapy alone had the highest crude rate (39.3 in 1000 patients) of previous cancers, those treated with ¹³¹I had an intermediate rate (14.2 in 1000 patients), and those treated surgically had the lowest rate (5.5 in 1000 patients), except for patients with a history of prior thyroid cancer who most often were treated surgically for their hyperthyroidism.

All Patients. Overall, 16683 deaths (49.4%) occurred among the 33748 patients from the United States with known date of birth, date of study entry, and sex. Cause of death was known for all but 111 of the deceased subjects. Cancer was the primary cause of death for 2950 patients, whereas 2857.6 deaths from cancer (SMR, 1.03) were expected (Table 3). The risk of death from cancer varied by years of follow-up, but was significantly elevated only during the first 4 years (SMR, 1.31; 95% CI, 1.19-1.43), particularly during the first year (SMR, 1.52; 95% CI, 1.25-1.84). During the entire study period, deaths from cancers of the lung (SMR, 1.11), breast (SMR, 1.17), kidney (SMR, 1.31), and thyroid (SMR, 2.77) were significantly elevated compared with the US population. For all 4 cancer types, the SMRs were highest within 5 years of study entry, and only thyroid cancer mortality remained significantly higher than expected 10 or more years after study entry (SMR, 2.01; 95% CI, 1.10-3.37).

Mortality from cancers of the liver, larynx, uterus, and prostate occurred significantly below expectation. No excess deaths from lymphoma, leukemia, or hematopoietic and lymphoproliferative malignancies as a group were apparent over the entire study period, but leukemia mortality was increased 5 to 9 years after study entry (SMR, 1.62; 95% CI, 1.01-2.45).

Because 79% of the study population was female, their mortality experience dominates the overall evaluations. Among women, 2252 cancer deaths were observed compared with 2101.8 expected (SMR, 1.07; 95% CI, 1.03-1.12), and an excess of lung (SMR, 1.28; 95% CI, 1.14-1.44), breast (SMR, 1.17; 95% CI, 1.07-1.28), kidney (SMR, 1.37; 95% CI, 1.00-1.83), and thyroid (SMR, 2.66; 95% CI, 1.70-3.95) cancer mortality was observed. A deficit of mortality from cancers of the uterus (corpus and cervix) also was seen (SMR, 0.67; 95% CI, 0.55-0.80). Five or more years after study entry, only mortality from lung and thyroid cancer remained in excess, but the deficit of deaths from cancer of the uterus was still observed. For men, the overall risk of cancer mortality was lower than expected (SMR, 0.92; 95% CI, 0.86-0.99). Based on only 5 deaths, the risk of thyroid cancer was elevated (SMR, 3.51; 95% CI, 1.13-8.18), largely because of a 14-fold excess (95% CI, 2.35-33.72) during the first 4 years of follow-up. Deficits of laryngeal (SMR, 0.29; 95% CI, 0.006-0.83) and prostate (SMR, 0.73; 95% CI, 0.56-0.94) cancer mortality were evident.

Patients With Graves Disease and Toxic Nodular Goiter. An increase in overall cancer mortality was not associated with Graves disease (SMR, 1.02; 95% CI, 0.98-1.06), but was 16% higher than expected among patients with toxic nodular goiter (SMR, 1.16; 95% CI, 1.03-1.30) (Table 4). Among Graves disease patients, mortality from cancers of the breast, kidney, and thyroid was elevated, but again the excess risks were observed only during the first 4 years of follow-up. Deaths from cancers of the liver, larynx, uterus, and prostate as well as multiple myeloma were lower than expected. Among patients with toxic nodular goiter, excess deaths from lung, breast, and thyroid malignancies were observed and were seen 5 or more years after study entry (SMR for lung cancer, 1.46 [95% CI, 1.01-2.06]; for breast cancer, 1.45 [95% CI, 1.03-1.98]; and for thyroid cancer, 4.3 [95% CI, 1.16-11.01]).

Cancer mortality relative to treatment was evaluated excluding the first year of follow-up (Table 5). Although almost 21000 people with over 385000 person-years of follow-up were studied, the overall cancer risk for patients treated with ¹³¹I was close to that expected using national cancer mortality rates (SMR, 1.02; 95% CI, 0.98-1.07). Mortality from thyroid cancer (SMR, 3.94; 95% CI, 2.52-5.86) was raised, but mortality from cancers of the uterus and prostate was below expectation. Among the 8054 patients treated with ¹³¹I only, findings were similar. To ensure that the elevated thyroid cancer mortality was not due to the 4 patients who had thyroid cancer prior to entering the study, we also analyzed the data excluding them. The SMR was still substantially increased (SMR, 3.28).

Between 1 and 5 years after follow-up, significantly elevated SMRs (Table 6) were observed for all cancer mortality (SMR, 1.24; 95% CI, 1.10-1.40) and mortality from colorectal cancer (SMR, 1.42; 95% CI, 1.04-1.90). Deaths from lung cancer (SMR, 1.44; 95% CI, 1.13-1.81) and non–chronic lymphatic leukemia (non-CLL) (SMR, 1.88; 95% CI, 1.18-2.84) were in excess within the first 9 years. After 10 years of follow-up only thyroid cancer mortality remained significantly increased. There were no observed or expected thyroid cancer deaths among the 365 patients treated with ¹³¹I before age 20 years.

Among patients with Graves disease treated with ¹³¹I (Table 7), thyroid cancer mortality was increased (SMR, 2.84; 95% CI, 1.62-4.61), but it was not significantly elevated 5 or more years after follow-up (SMR, 1.88; 95% CI, 0.86-3.57). An excess of total cancer deaths (SMR, 1.31; CI, 1.07-1.58) and mortality from cancers of the esophagus, uterus (particularly corpus uteri), and thyroid were observed among toxic nodular goiter patients receiving ¹³¹ I therapy (Table 7). Mortality from cancers of the esophagus (SMR, 5.26; CI, 1.42-13.5), corpus uteri (SMR, 3.0; CI, 1.09-6.53), and thyroid (SMR, 11.1; CI, 2.26-32.5) remained high in patients with toxic nodular goiter more than 5 years after study entry.

Significant relationships between the number of ¹³¹I treatments and overall cancer mortality (SMR, 1.00, 1.09, and 0.99 for 1, 2-4, and ≥5 treatments, respectively) or individual cancer sites, except thyroid cancer, were not observed (data not shown). Restricting the analysis to either 5 years or 10 or more years after study entry did not modify these conclusions. The SMRs for thyroid cancer deaths increased with increasing number of treatments during the study period (2.74, 6.21, and 6.61 for 1, 2-4, and ≥5 treatments, respectively), as well as for the period beginning 10 years after study entry (2.66, 4.48, and 5.43 for 1, 2-4, and ≥5 treatments, respectively).

Table 7 provides some evidence of a relationship between administered ¹³¹ I activity and total cancer mortality, as well as mortality from lung and thyroid cancer. However, at 10 or more years after study entry, the relationship with administered activity was less notable; the SMRs for less than 7 mCi, 7-15 mCi, and 15 mCi or more administered activity were 0.90, 1.03, and 1.06, respectively, for all cancer mortality; 0.83, 1.03, and 1.21, respectively, for lung cancer mortality; and 1.85, 4.11, and 1.59, respectively, for thyroid cancer mortality (data not shown). An evaluation of ¹³¹I activity by type of hyperthyroidism indicated that thyroid cancer mortality increased with increasing administered activity for patients with Graves disease. Among patients with toxic nodular goiter, there was no relationship with amount of activity, but the number of cancers was small and the CIs were wide.

The Poisson regression analyses of the relationship between administered activity and subsequent thyroid cancer mortality, taking type of hyperthyroidism into account, excluded patients treated surgically (because their risk of thyroid cancer is reduced by decreasing the volume of tissue) and patients who had cancer before study entry, leaving 16 thyroid cancer deaths (Table 8). The RR in the highest administered activity group was 2.3 (95% CI, 0.97-9.5), but the exposure-response trend was not statistically significant (P=.12). Similarly, the excess RR point estimate result (excess RR at 1 mCi=0.12) was positive, but not significant. Five or more years after study entry, no relationship was seen between administered activity and subsequent thyroid cancer mortality.

No relationship between level of ¹³¹I administered activity or estimated bone marrow dose and CLL, non-CLL, Hodgkin disease, non–Hodgkin lymphoma, or multiple myeloma was demonstrated (Table 9). Because the latency period for radiation-induced non-CLL is short, we also analyzed the non-CLL data for only the first 10 years of follow-up, but again, no association with estimated bone marrow dose was seen (P=.46). For multiple myeloma, the RRs were increased in the 2 highest administrated activity and estimated bone marrow dose groups, but the numbers were small and the dose-response trend was not significant.

Among the 10876 patients receiving surgical treatment without ¹³¹ I, the SMR for total cancer mortality was 0.99 (Table 5). Excluding the first year of follow-up, lung and breast cancer mortality was significantly elevated among surgical patients, but not thyroid cancer mortality (SMR, 1.07). A 50% reduction in multiple myeloma mortality (SMR, 0.48; 95% CI, 0.19-0.99) was observed.

One year or more after study entry, there was a significant increase in the number of cancer deaths from all causes, as well as buccal and brain cancer among patients treated with antithyroid drugs (Table 5). Because patients treated with antithyroid drugs had an especially high rate of cancers diagnosed before study entry, an SMR analysis excluding patients with cancers diagnosed prior to study entry was performed. In this analysis, only brain cancer deaths remained elevated; however, excluding patients with cancers prior to study entry will underestimate the SMR.

We evaluated cancer mortality following 3 treatment modalities for hyperthyroidism. We put special emphasis on ¹³¹I therapy because it is the treatment of choice in the United States, and the role of ¹³¹I in carcinogenesis remains unclear. The investigators of the first evaluation of the Cooperative Thyrotoxicosis Therapy Follow-up Study cohort reported no increased risk of either leukemia or thyroid cancer incidence among patients treated with ¹³¹ I compared with patients treated by other methods.^5- 6,10,16 In contrast, Williams²⁸ interpreted the thyroid data as showing a slight excess of anaplastic thyroid cancer among the group treated with ¹³¹I, which he hypothesized might be caused by radiation-induced progression from undetected differentiated carcinomas to undifferentiated carcinomas. Additional follow-up of 2 subgroups of female patients from the original cohort suggested that there might be a slight increased risk of cancers occurring in organs known to concentrate ¹³¹ I^11,13 or of breast cancer incidence.¹⁴

In the present follow-up, ¹³¹I treatment did not result in a larger-than-expected number of deaths due to cancer when compared with US age-, sex-, and calendar-year–specific mortality rates. No excess mortality was observed when cancer in organs that concentrate ¹³¹ I (salivary, esophagus, stomach, colon, rectum, liver, bladder, and kidney) were grouped together (SMR, 1.03). A detailed analysis of hematopoietic malignancies revealed that neither ¹³¹ I administered activity nor estimated bone marrow dose were associated with mortality from non-CLL. This lack of positive results is not unexpected because the mean doses to all organs, except for the thyroid gland, which received extremely large doses, were below 200 mGy. At such low doses the statistical power to detect effects is limited. Because the study population was almost entirely adult, the results, however, cannot be generalized to childhood exposure.

Mortality studies are not well suited for studying thyroid cancer. Nonetheless, we did observe a significantly increased rate of thyroid cancer mortality among men and women and among patients with toxic nodular goiter or Graves disease. The elevated risk of thyroid cancer mortality was seen only for patients treated with ¹³¹I. Using either an external (SMR analysis) or internal (Poisson analysis) comparison, the risk appeared to increase with increasing amounts of ¹³¹ I administered activity. The trend was not significant more than 10 years after study entry, or in the analysis using the internal comparison.

Several points should be noted in interpreting these results. The correlation between the amount of administered activity and thyroid radiation dose generally is poor in patients with hyperthyroidism²²; patients receiving higher ¹³¹ I activity generally had more severe disease and frequently had relapses; patients dying from thyroid cancer were more likely to have had toxic nodular goiter than Graves disease (30% vs 8%); the major increase in mortality was seen in the first 5 years after treatment; goiter, particularly nodular goiter, has been reported as a risk factor for thyroid cancer^29- 31; and some of the patients with toxic nodular goiter may have had an undiagnosed preexisting thyroid cancer when they entered the study.

If it is assumed that the observed excess of thyroid cancer is causal, the risk would be smaller than the risks observed following external radiation. Some of the difference may be explained by ¹³¹I therapy-induced cell killing. Although thyroid doses could not be estimated directly, the aim of ¹³¹I treatment for hyperthyroidism is to destroy the thyroid.³² The effectiveness of ¹³¹I in killing thyroid cells presumably reduces the likelihood of malignant cell transformation. Another explanation may be related to age at exposure. Since the risk of developing thyroid cancer after external irradiation decreases with increasing age at exposure, with little risk demonstrated after age 20 years, data from this study of adults are insufficient to allow relevant comparisons between ¹³¹I and external radiation. To date, almost all human data on ¹³¹I are from populations exposed as adults,^15,33- 37 whereas most data from external radiation exposure to the thyroid are from individuals treated during childhood.^38- 39 Finally, some patients treated with ¹³¹I in this study had full or partial thyroidectomies later, which might have lowered the risk of subsequent thyroid cancer by removing radiation-damaged tissue.

Compared with our study, the patients in a Swedish mortality series¹⁵ were slightly older, had 4 fewer years of follow-up, and had a higher percentage of patients with toxic nodular goiter and consequently a larger mean total ¹³¹I administered activity. The SMR for total cancer mortality was close to expected in both studies and, with the exception of thyroid cancer, the SMRs for other cancers were between 0.8 and 1.3. Although the increased risk of digestive tract and respiratory cancer mortality reached significance in the Swedish study, the risks were only 10% and 30% higher than expected, respectively. In our study, we did observe an increased risk of respiratory cancer mortality among patients with toxic nodular goiter treated with ¹³¹I. Differences between the 2 studies easily could be caused by the higher proportion of Swedish patients with toxic nodular goiter or due to chance.

This study is unique because of its large size and its inclusion of patients not treated with ¹³¹I. In the present follow-up, small but significant increases in breast, lung, and kidney cancer mortality were observed in the entire cohort of hyperthyroid patients compared with the general population. Smoking, lifestyle, and reproductive histories were not known, and these factors could confound the results. Because much of the elevated mortality occurred during the first few years of follow-up, underlying disease may be responsible for some of the excess. Although the group of patients treated with antithyroid drugs was small, the observed cancer mortality for several sites (total cancer, buccal, stomach, breast, and brain) was higher than expected compared with national rates. Analyses excluding patients with previous cancers suggested that some, but not all, of the excess could be attributed to the selection of patients for drug therapy. Patients treated with drugs mainly received thiourea derivatives, iodine compounds, or a combination of the 2, but sometimes received various other drugs. The type, quantity, and dates of drug use generally were not available from the medical charts and could not be taken into consideration in the analysis. Because the quality of the drug data in this study is clearly limited, long-term effects of antithyroid drugs need to be evaluated further.

Although the study has several strengths, it also has limitations. First, thyroid doses could not be estimated and exposure response was based on ¹³¹I administered activity. Second, mortality is not the ideal study end point. Causes of death, as listed on death certificates, frequently are inaccurate and rarely mention the histological type of cancer. In addition, information on cancers with high survival rates such as thyroid and breast is limited. Third, about 20 cancer sites were evaluated by latency, type of hyperthyroidism, treatment group, and for patients treated with ¹³¹ I, amount of administered activity. By conducting multiple comparisons, the probability of incorrectly rejecting a null hypothesis increases and suggests cautious interpretation of these data. On the other hand, the small number of patients and deaths in many of the subgroup analyses and the low radiation doses to most organs, other than the thyroid, limit the power to detect significant risks. Finally, patients are selected for a specific treatment based on their medical condition. For example, patients undergoing surgery were less likely to have had a previous cancer (except of the thyroid) than patients treated with drugs or ¹³¹I. Thus, comparing risks related to treatment is problematic.

In summary, in a large series of hyperthyroid patients followed up for nearly a lifetime, cancer mortality was not significantly elevated. High-dose therapeutic ¹³¹I administration, the most frequent treatment, also was not linked to an excess of total cancer mortality. Although a small number of thyroid cancer deaths might be attributed to ¹³¹I therapy, overall, it appears to be a safe therapy.