Hypothyroidism is a potential complication after radiation therapy when treatment fields include the thyroid gland. In patients with head and neck cancer and Hodgkin disease, the incidence of hypothyroidism is as high as 30% to 50% 5 years or more after treatment with radiation.1–6 In patients with breast cancer, a portion of the thyroid gland may also be included in the treatment fields, particularly when radiation is administered to supraclavicular lymph nodes. No prior study, however, has clearly determined whether breast cancer patients who receive radiation therapy to this field have an increased risk of developing hypothyroidism. Quantifying the magnitude of risk in these patients is important to help determine whether regular monitoring of thyroid function after treatment would be useful in breast cancer patients, particularly as breast cancer survival continues to improve and survivorship issues gain increasing importance.7–9
Accordingly, in a cohort of older breast cancer patients identified by using the Surveillance, Epidemiology, and End Results (SEER)-Medicare Linked Database, we sought to determine the incidence of hypothyroidism in patients treated with and without radiation, and we specifically sought to quantify the magnitude of risk in patients more likely to receive radiation therapy to a supraclavicular field. We, in addition, sought to compare the incidence of hypothyroidism in breast cancer patients with elderly women in the general population, thereby quantifying the excess burden of morbidity from hypothyroidism experienced by breast cancer survivors.
METHODS AND MATERIALS
The SEER-Medicare cohort is a population-based sample of Medicare beneficiaries with incident cancer identified in the SEER program tumor registries. SEER and Medicare linkage is available because of a collaborative effort between the National Cancer Institute (NCI) and Centers for Medicare and Medicaid Services (CMS). From 1992 through 1999, the SEER program included 11 registries accounting for 14% of the United States' population, and in 2000, the program expanded to include an additional 5 registries; geographic areas covered in these 16 registries accounted for 26% of the population in United States. Linkage to Medicare claims' data are available for 93% of patients in SEER aged ≥65 years.10 Additional details have been published elsewhere.11–13
Study Sample and Exclusions
The study population included 148,102 women aged >65 years who were, according to SEER-Medicare, diagnosed with breast cancer between January 1, 1992 and December 31, 2002. We excluded from this sample patients with any diagnosis claim of hypothyroidism in the year before cancer diagnosis (n = 33,048). In addition, we excluded women who had any of the following characteristics, a history of prior nonskin cancer (n = 9697), nonepithelial histology (n = 6303), unknown stage (n = 4325), unknown nodal status (n = 41,021), no pathologic confirmation (n = 2613), distant metastases (n = 6523), bilateral disease (n = 161), treatment with neoadjuvant chemotherapy or radiation therapy (n = 1187), or diagnosis of a second primary tumor within 9 months of the initial diagnosis date (n = 5852), leaving an initial sample of 62,987 patients who met the clinical inclusion criteria (as patients could have met more than 1 exclusion criteria). We then excluded patients missing Medicare fee-for-service coverage (Parts A and/or B) from 12 months before to 9 months after the initial diagnosis date24,732 (as these patients were effectively missing comorbidity or treatment information), leaving a total study sample of 38,255 patients.
The SEER-Medicare registry has identified a cancer-free comparison cohort (control group) representative of the general Medicare population from which the breast cancer sample was derived. Medicare randomly sampled 5% of patients living in SEER regions without a history of cancer or current cancer at the time of identification.10 To obtain a control group similar to the breast cancer patient group, we derived a study sample of 111,944 women from this comparison cohort population who 1) were never missing Medicare fee-for-service coverage (Parts A and/or B) from 12 months before to 9 months after the first day of the follow-up period, defined as January 1, 1992, as indicated by Medicare records; 2) were aged >65 by the first day of the follow-up period; 3) had no prior diagnosis claim for cancer; and 4) had no diagnosis claim for hypothyroidism in the year before entering the cohort.
Breast cancer patients and control subjects were defined as having developed the outcome of hypothyroidism (International Classification of Diseases, Ninth Revision [ICD-9] codes 244; 244.0–3, 8–9) in the follow-up period if they had at least 1 diagnosis claim during a hospitalization (Medicare Part A claim) or at least 2 diagnosis claims more than 30 days apart during outpatient visits (Medicare Part B claim) in any of the diagnoses listed in claims. This algorithm for identifying comorbid conditions has been validated in a prior study that sought to maximize sensitivity and specificity of identifying comorbid conditions through the SEER-Medicare database.14
Treatment with radiation therapy was determined by using both SEER records and Medicare claims (ICD-9 Procedure codes 92.21–92.27 and 92.29; ICD-9 Diagnosis codes V58.0, V66.1, V67.1; Current Procedural Terminology (CPT) codes 77,401–77,525 and 77,761–77,799; or Revenue Center (RC) Codes 0330 and 0333). Patients were considered to have received radiation therapy if a claim was recorded within 9 months of breast cancer diagnosis.
Surgery was determined by using both SEER and Medicare claims (Mastectomy ICD-9 Procedure codes 85.41, 85.42, 85.43, 85.44, 85.45, 85.46, 85.47, 85.48; CPT codes 19,180, 19,182, 19,200, 19,220, 19,240; Lumpectomy ICD-9 Procedure codes 85.20, 85.21, 85.22, 85.23, 85.25; CPT codes 19,110, 19,120, 19,125, 19,160, 19,162). The most extensive surgical procedure reported within 9 months of breast cancer diagnosis was considered the definitive surgery.
Treatment with chemotherapy within 6 months of diagnosis was determined from Medicare claims (ICD-9 Procedure code 99.25; ICD-9 Diagnosis codes V58.1, V66.2, V67.2; CPT Codes 96,400–96,549; Healthcare Common Procedure Coding System codes J9000-J9999, Q0083-Q0085; and RC Codes 0331, 0332, 0335). As adjuvant endocrine therapy is not reported in either SEER or Medicare claims, this variable was not included in the analysis. These treatment definitions have been applied in prior published studies that used the SEER-Medicare database.11, 12
Patient characteristics obtained from SEER registry data included age and race (categorized as white, black, or other). Severity of comorbid disease was defined on the basis of a modified Charlson comorbidity index score validated in prior studies.14, 15 This index compiled a weighted score incorporating comorbid diseases coded on Medicare claims between 12 months and 1 month before cancer diagnosis. To enhance specificity, patients must have had at least 1 diagnosis claim during a hospitalization (Medicare Part A claim) or at least 2 diagnosis claims more than 30 days apart during outpatient visits (Medicare Part B claim) to have been considered to have a history of the disease.14 The modified Charlson comorbidity score was then categorized as 0 (no comorbidity), 1 (mild to moderate), or 2 (severe). Tumor characteristics included as covariates were size, number of positive lymph nodes, grade, histology, and receptor status as reported by the SEER registry. Margin status and lymph-vascular space invasion are not reported in SEER. Socioeconomic characteristics included median income of census tract, urban or rural residence, SEER geographic region, and year of treatment as reported by the SEER registry. To characterize the frequency of patient interaction with the healthcare system, we determined the number of physician visits on separate dates between 12 months and 1 month before cancer diagnosis from Medicare claims.
The incidence of hypothyroidism was determined by using Kaplan-Meier survival function estimates, and the incidence rates in patients treated with versus those treated without radiation were compared by using the log-rank test. The time to event was calculated from the date of cancer diagnosis in breast cancer patients and from the date of entry into the cohort for the control group. Patients were censored if they died or were lost to follow-up (earliest date that Medicare fee-for-service coverage lapsed or were lost to follow-up in the SEER registry, respectively) before any diagnosis of hypothyroidism.
We tested the unadjusted association between treatment with radiation and risk of subsequent hypothyroidism by using the Pearson chi-square test, and we tested the adjusted association using a multivariable proportional hazards regression model. Log-log complementary plots were used to assess proportionality assumptions. Covariate selection for the final multivariable model was based on clinical significance (clinical judgment) or statistical significance in univariate analyses and prior studies.11, 12 Continuous covariates were included in models based on evaluation of linearity in univariate analyses.
We specifically characterized the adjusted risk of hypothyroidism in patients treated with radiation who had 4 or more (4+) positive lymph nodes (These patients more likely to receive radiation therapy to a supraclavicular field.), and we compared their risk with patients treated with radiation who had 0 positive lymph nodes and with patients who were not treated with radiation who had 0 positive lymph nodes. Also, to identify whether any other high-risk patient subgroup existed, we tested the interaction between radiation therapy and other covariates selected a priori, including older age (age ≥75 years), race, treatment with chemotherapy, and estrogen receptor (ER) and progesterone receptor (PR) status.
Finally, we compared this cohort of breast cancer patients with the cancer-free control sample to determine whether there was an excess burden of developing hypothyroidism in all breast cancer patients compared with elderly women in the general population. The incidence of hypothyroidism in the control sample was determined by using Kaplan-Meier survival function estimates. In addition, we calculated the unadjusted population attributable risk percentage (PAR percentage) of hypothyroidism in breast cancer patients versus control patients. Finally, we used a proportional hazards model to determine the relative risk of developing hypothyroidism in breast cancer patients compared with control patients over time in models adjusted for age, year of entry into the cohort, and number of physician visits.
In the multivariable model that assessed the adjusted risk of hypothyroidism associated with radiation therapy, we conducted a subsidiary analysis in the subset of patients who were diagnosed with breast cancer before the year 2000 (n = 30,323) to ensure that results were not affected by the shorter follow-up time of patients diagnosed after 2000.
Analyses were conducted by using SAS statistical software, version 9.1.3 (SAS Institute Inc, Cary, NC), and all statistical tests assumed a 2-tailed α of .05. The University of Texas M. D. Anderson Cancer Center Institutional Review Board approved use of the SEER-Medicare database.
In 38,255 breast cancer patients, median follow-up time was 4.6 years (interquartile range, 3.0–7.8 years). Mean age was 75 ± 6 years, and 86% of patients were white. Ninety-five percent had stage I–III disease, 5% had stage 0 disease, 70% had ductal histology, and 9% had 4+ positive lymph nodes. For the initial course of treatment, 44% of patients were treated with radiation therapy, 16% were treated with chemotherapy, 39% with breast conserving surgery, and 61% with mastectomy.
Of patients treated without radiation, 74% had 0 positive lymph nodes, and 7% had 4+ positive lymph nodes. Of patients treated with radiation, 74% had 0 positive lymph nodes and 10% had 4+ positive lymph nodes. Compared with patients with 0 positive lymph nodes treated with or without radiation, patients with 4+ positive lymph nodes treated with radiation were more likely to be younger, have less comorbidity, have larger and poorly differentiated tumors, and were more likely to be treated with chemotherapy (Table 1).
Table 1. Characteristics of Selected Comparison Groups of Breast Cancer Patients
| ||n = 15,743||n = 12,544||n = 1774|| |
|Demographics|| || || || |
|Mean age ± SD, y||76 ± 6||74 ± 5||74 ± 6||<.001|
|Race|| || || ||<.001|
|Tumor characteristics|| || || || |
|Size|| || || || |
|2 to <5 cm||25||15||49|| |
|≥5 cm||3||2||18|| |
|Histology|| || || || |
|Grade|| || || ||<.001|
|Moderately differentiated||35||35||32|| |
|Poorly or undifferentiated||25||26||48|| |
|ER status|| || || ||<.001|
|PR status|| || || ||<.001|
|Other clinical characteristics|| || || || |
|Charlson comorbidity index score|| || || ||<.001|
|None (0)||71||77||68|| |
|Mild to moderate (1)||17||14||16|| |
|Severe (2)||7||4||6|| |
|Treatment|| || || || |
|Surgery|| || || ||<.001|
|Breast conserving surgery||6||90||29|| |
|Socioeconomic characteristics|| || || || |
|Location|| || || ||<.001|
|Large metropolitan area||50||59||57|| |
|Metropolitan area||29||28||27|| |
|Urban area||7||6||6|| |
|Less-urban area||11||5||8|| |
|Rural area||3||1||2|| |
|Income (quartile)|| || || ||<.001|
|1st (median, $25,730)||27||17||23|| |
|2nd (median, $36,529)||26||21||22|| |
|3rd (median, $47,054)||24||27||27|| |
|4th (median, $69,094)||23||35||28|| |
|No. of doctor visits in last year|| || || ||<.001|
|Mean ± SD||12 ± 10||12 ± 10||11 ± 10|| |
Incidence and Predictors of Hypothyroidism
In all breast cancer patients, the 1-year incidence of hypothyroidism was 4%, and the 5-year incidence was 14% after cancer diagnosis. Patients who developed hypothyroidism were more likely to be older and white. In general, with the exception of smaller tumor size at diagnosis, tumor characteristics were not predictive of subsequent hypothyroidism. However, subsequent hypothyroidism was significantly associated with socioeconomic factors, including residence in large metropolitan areas (in contrast to rural areas), lower median income, and increased number of physician visits.
Radiation Therapy and Risk of Hypothyroidism
On unadjusted analysis, the incidence of hypothyroidism did not differ by radiation therapy status (P = .52). For example, the 5-year unadjusted incidence of hypothyroidism was 14% in patients treated with and without radiation. Incidence was also 14% in patients with 4+ positive lymph nodes treated with radiation (Table 2). On adjusted analysis in all breast cancer patients, the association between treatment with radiation and risk of subsequent hypothyroidism remained marginal (hazard ratio (HR) 1.10; 95% confidence interval (CI), 1.00–1.20) (Table 3). In the 30,323 patients who were diagnosed with breast cancer in SEER before the year 2000, the association between radiation therapy and subsequent hypothyroidism was not substantially changed (HR = 1.10; 95% CI, 1.00–1.22).
Table 2. Unadjusted Incidence of Hypothyroidism in Breast Cancer Patients and Control Subjects
|Patients with RT||4%||14%||19%|
|RT and 4+ positive lymph nodes||4%||14%||17%|
|RT and 0 positive lymph nodes||4%||14%||19%|
|Patients with no RT||4%||14%||19%|
|No RT and 0 positive lymph nodes||4%||13%||19%|
Table 3. Adjusted Risk of Subsequent Hypothyroidism Associated With Radiation Therapy: Sequential Effects of Various Confounding Variables
|Any RT (unadjusted)||1.02||0.85, 1.08||.52|
|Any RT (adjusted)†|| || || |
|+Demographic variables||1.02||0.96, 1.07||.60|
|+Tumor characteristics||1.02||0.97, 1.08||.42|
|+Chemotherapy and surgery||1.06||0.97, 1.17||.19|
|+Socioeconomic characteristics||1.10||1.00, 1.20||.05|
|RT and 4+ positive lymph nodes‡|| || || |
|vs No RT and 0 positive lymph nodes||1.14||0.98, 1.34||.09|
|vs RT and 0 positive lymph nodes||1.04||0.89, 1.23||.61|
Patients with 4+ positive lymph nodes who were treated with radiation also had a marginally higher risk of subsequent hypothyroidism when compared with patients with 0 positive lymph nodes who were treated without radiation (HR = 1.14; 95% CI, 0.98–1.34). However, no significant difference was found when comparing these patients against those with 0 positive lymph nodes also treated with radiation (HR = 1.04; 95% CI, 0.89–1.23) (Table 3).
When other potentially high-risk subgroups were evaluated, patients treated with radiation who were white, older, and who had estrogen receptor (ER)-positive disease exhibited a trend toward higher risks of hypothyroidism. However, the interaction term between radiation therapy and these covariates was statistically significant only for ER status (P = .05). Treatment with chemotherapy along with radiation therapy, even in patients with 4+ positive lymph nodes, was not associated with any additional increase in risk of hypothyroidism (Table 4).
Table 4. Adjusted Risk of Subsequent Hypothyroidism Associated With Radiation Therapy: Modifying Effects in Patient Subgroups
|In age <75 y||1.08||0.98, 1.20||.13||.58|
|In age ≥75 y||1.11||1.00, 1.24||.05|| |
|In white patients||1.13||1.03, 1.25||.01||.77|
|In black patients||1.08||0.82, 1.41||.54|| |
|In patients also treated with chemotherapy||1.04||0.89, 1.21||.66||.35|
|In patients not treated with chemotherapy||1.11||1.01, 1.23||.03|| |
|In patients with ER-positive disease||1.14||1.03, 1.26||.01||.05|
|In patients with ER- and PR-negative disease||1.05||0.88, 1.26||.61|| |
Excess Risk of Hypothyroidism Compared With Cancer-free Older Women
In 111,944 control subjects, mean age at entry into the cohort was similar to that of breast cancer patients at the time of diagnosis (76 ± 7 years). The 1-year incidence of hypothyroidism in older women without a history of cancer was 2%, and the 5-year incidence was 11% (Table 2), which converted to a PAR percentage of 9% for the 5-year incidence of hypothyroidism in older breast cancer patients. In addition, the unadjusted relative risk of developing hypothyroidism was significantly higher in older breast cancer patients than in cancer-free older women from the general population. (HR = 1.13; 95% CI, 1.10–1.35). In a model adjusted for age and limited to the breast cancer patients diagnosed in 1992, risk was also increased (HR = 1.26; 95% CI, 1.18–1.35). When we further adjusted for the number of physician visits in the year before cancer diagnosis or entry into the cohort (to account for frequency of patient interaction with the healthcare system), the risk remained similar (HR = 1.35; 95% CI, 1.26–1.44).
Development of hypothyroidism after diagnosis was common in older breast cancer patients, with a 5-year incidence in this patient population as high as 14%. Importantly, the adjusted risk for breast cancer patients significantly exceeded the risk in a cancer-free control population of elderly women. However, our results further suggested that patients who were more likely (4+ positive lymph nodes) to receive radiation therapy to a supraclavicular field did not have substantially amplified risk compared with those less likely (0 positive lymph nodes) to receive radiation therapy to this field. To our knowledge, this is the first study in the literature to address this question in a large, representative cohort of elderly breast cancer patients.
Although 1 previous study suggested no association between radiation and development of thyroid cancer in breast cancer patients,16 few prior studies have directly addressed whether radiation therapy could be a risk factor for subsequent hypothyroidism in breast cancer patients. Cutuli et al reported a recent case series of 80 breast cancer patients who underwent radiation therapy, chemotherapy, and surgery and found that 6.2% had clinically symptomatic hypothyroidism on short-term follow-up after primary treatment.17 However, their analysis was limited by a small sample size, lack of a comparison group (ie, patients not treated with radiation), and lack of data on history of hypothyroidism. In addition, their study did not report long-term outcomes. Other earlier studies of breast cancer patients had similar limitations and did not establish adequate evidence of hypothyroidism risk specifically associated with radiation therapy to a supraclavicular field.18, 19
In contrast, evidence of the risk of hypothyroidism secondary to external neck radiation in Hodgkin disease and head and neck cancer patients is well established, showing a high incidence, particularly in the long term.3, 5, 20 For example, in a cohort study of 504 head and neck cancer patients, biochemical hypothyroidism was detected in at least 50% of patients with a follow-up time of up to 10 years. Similarly, a 43% incidence of hypothyroidism was found in Hodgkin disease patients with median follow-up time of 11.3 years.2, 4
Studies in these patient populations have helped to establish the pathophysiology that underlies the development of hypothyroidism. Parenchymal thyroid cell injury and capsular fibrosis secondary to radiation have been implicated as mechanisms likely to underlie the development of hypothyroidism. Furthermore, radiation-induced damage to small thyroid vessels and atherosclerosis of larger vessels may also contribute. Lastly, evidence suggests that immune-mediated damage may further play a role in outcome.21
Previous investigators have posited that chemotherapy and hormonal therapy could have additional modifying effects.18, 19, 22 We did not find evidence of a synergistic effect between radiation therapy and chemotherapy in our cohort of breast cancer patients. However, our exploratory subgroup analyses did suggest that patients with ER-positive disease (who were more likely to be treated with tamoxifen) were at higher risk than those with ER- negative and PR-negative disease. In addition, our subgroup analyses suggested that older white patients could be at higher risk than others, although these differences were not statistically significant. Interestingly, this modifying effect by race is consistent with prior studies of the association between radiation therapy and development of hypothyroidism. Specifically, prior studies have reported a higher risk of subsequent hypothyroidism in white patients who were diagnosed with both Hodgkin disease4 and head and neck cancer.20 No physiologic explanation of racial differences for these risks has yet been identified. Given that our findings were not only that breast cancer patients are generally at higher risk for developing hypothyroidism but also that certain patient characteristics may identify a higher-risk group, we suggest further study with longer follow-up.
Survival in breast cancer patients continues to improve,7–9 and evidence to support strategies for screening and management of chronic comorbidities in these patients has become increasingly important as long-term survivorship issues continue to emerge. Given the paucity of prior studies of hypothyroidism risk for breast cancer patients, it is not surprising that no formal recommendations exist regarding post-treatment screening for hypothyroidism in this group. National Comprehensive Cancer Network guidelines suggest routine screening of thyroid function within the first year after radiation therapy in Hodgkin disease and head and neck cancer patients, if radiation treatment fields include the neck.23
In the general population, controversy exists regarding guidelines for thyroid function screening, even in elderly women, who have been recognized as having a high risk even if they are otherwise healthy. Prior evidence from the National Health and Nutrition Examination Survey suggested that the prevalence of hypothyroidism is as high as approximately 20% in this group,24 which is consistent with our finding of an 8-year incidence of 19% in controls. However, prior studies and proposed guidelines have debated whether routine thyroid screening in this population is cost beneficial, and evidence for the cost-effectiveness of screening in elderly populations is controversial.25–27 At present, Medicare does not reimburse for routine thyroid function screening in asymptomatic elderly patients. Given our findings that breast cancer survivors represent a higher-risk group compared with cancer-free older women in the general population, these patients could potentially benefit from screening, especially within the first 5 to 8 years after cancer diagnosis. Further studies on the cost-effectiveness of regular thyroid function screening in this patient population may help to inform future guidelines.
Our study was limited to an elderly population, so results cannot necessarily be extrapolated to all breast cancer patients. Of note, however, elderly patients have a particularly high risk of developing hypothyroidism and, thus, represent an appropriate group within which to assess the potential benefits of screening. Given the much lower baseline risk for hypothyroidism in younger women, it is uncertain whether young breast cancer patients would represent a similarly appropriate target group for post-treatment screening. Our analysis may have had residual confounding due to misclassification of the outcome, hypothyroidism, as it was based on ICD-9 diagnosis. Therefore, a diagnosis of hypothyroidism in our study could have reflected either a clinical diagnosis or a laboratory-confirmed diagnosis. However, we sought to emphasize specificity in our definition by using a previously validated algorithm for identifying comorbidities.14 In addition, hypothyroidism could have been diagnosed more frequently in breast cancer patients because of more frequent interactions with the healthcare system, although adjusting for the number of physician visits did not decrease the effect size in multivariable models. We used the number of positive lymph nodes as a surrogate measure for radiation therapy to a supraclavicular field. In addition, specific radiation field data, actual radiation dose to the thyroid, and the potential use of corner blocks that may shield thyroid exposure were not available data for our cohort. Although prior studies report high compliance with guidelines for delivery of radiation among patients who receive radiation therapy28 and a strong association between the number of positive lymph nodes and supraclavicular radiation and extremely low utilization of supraclavicular field radiation in patients with 0 positive lymph nodes,29 future studies that focus on the effect of these potential modifiers to radiation dose to the thyroid may be warranted. Finally, the majority of breast cancer patients in our study did not have follow-up beyond 8 years. Although our results preliminarily suggest that hypothyroidism risk for cancer patients and control patients may be similar beyond 5 to 8 years, future studies with additional follow-up are required to clarify the long-term hypothyroidism risk in breast cancer patients.
Development of hypothyroidism is fairly common in older breast cancer survivors. The risk for older breast cancer patients exceeds the risk found nationwide for cancer-free older women in the general population, regardless of whether patients received radiation therapy to a supraclavicular field. Given the concern for the elevated risk in this patient population, future studies on efficacy and cost benefit of routine post-treatment screening may be warranted, particularly as guidelines for routine monitoring after breast cancer treatment have not been defined.
This study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute (NCI); the Office of Research, Development and Information, Centers for Medicare & Medicaid Services (CMS); Information Management Services (IMS), Inc; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database.