Valvular Dysfunction and Carotid, Subclavian, and Coronary Artery Disease in Survivors of Hodgkin Lymphoma Treated With Radiation Therapy

Context The majority of patients with Hodgkin lymphoma are young and highly curable. This necessitates concern for prevention, diagnosis, and optimal management of potential treatment-related complications.

Objective To identify and quantify the incidence of and factors contributing to long-term cardiac and vascular complications after radiation therapy for Hodgkin lymphoma.

Design and Setting Retrospective study comparing patients treated from 1962 to 1998 at a university–based referral center with a matched general population.

Patients Four hundred fifteen consecutive patients who fulfilled the inclusion criteria of a minimum 2-year follow-up (median, 11.2 years) and whose radiation fields included the heart or carotid or subclavian arteries.

Main Outcome Measures Multivariable analyses of potential risk factors and observed-to-expected ratios for cardiac valve surgery, coronary artery bypass graft surgery, percutaneous coronary intervention, or both based on Surveillance, Epidemiology, and End Results (SEER) and National Hospital Discharge Survey (NHDS) data.

Results Forty-two patients (10.4%) developed coronary artery disease at a median of 9 years after treatment, 30 patients (7.4%) developed carotid and/or subclavian artery disease at a median of 17 years after treatment, and 25 patients (6.2%) developed clinically significant valvular dysfunction at a median of 22 years. The most common valve lesion was aortic stenosis, which occurred in 14 valves. The observed-to-expected ratio for valve surgery was 8.42 (95% confidence interval [CI], 3.20-13.65) and the observed-to-expected ratio for coronary artery bypass graft surgery or percutaneous coronary intervention was 1.63 (95% CI, 0.98-2.28). At least 1 cardiac risk factor was present in all patients who developed coronary artery disease. The only treatment-related factor associated with the development of coronary artery disease was utilization of a radiation technique that resulted in a higher total dose to a portion of the heart (relative risk, 7.8; 95% CI, 1.1-53.2; P = .04) . No specific treatment-related factor was associated with carotid or subclavian artery disease or valvular dysfunction. Freedom from any cardiovascular morbidity was 88% at 15 years and 84% at 20 years.

Conclusions Among patients treated with radiation therapy for Hodgkin lymphoma, there are statistically higher than expected rates of valve surgery and coronary revascularization procedures over the next 10 to 20 years. Coronary vascular disease is associated with higher radiation doses and traditional coronary heart disease risk factors. Noncoronary vascular disease and clinically important valvular dysfunction are less well understood complications at 15 to 20 years after radiation, requiring surveillance and further study.

Hodgkin lymphoma is 1 of the first cancers in which high cure rates were achieved with both radiation and chemotherapy. With longer follow-up, it is apparent that survivors carry risks for treatment-related effects that may not manifest until many years later. The focus of therapeutic trials in most stages of the disease has now shifted from a search for more effective therapy to a search for less toxic therapy. Documentation of the incidence and severity of these late complications, as well as contributing factors, is crucial in surveillance and intervention for late treatment effects in patients already treated, as well as prevention of late effects in future patients through optimal individual treatment decisions and clinical trial design.

Several studies have reported an increase in premature death from myocardial infarction among patients treated for Hodgkin lymphoma with mediastinal radiotherapy.^1-8 Reports of valvular disease^9-14 and peripheral vascular disease^15,16 after radiotherapy have been sporadic.

Our study was designed to estimate the incidence, severity, and risk factors associated with 3 potential radiation treatment–related complications—coronary and noncoronary atherosclerotic vascular disease and clinically important valvular disease—in a group of patients who have survived Hodgkin lymphoma.

Four hundred fifteen patients treated from 1962 through 1998 at the University of Florida met the following criteria for study inclusion: treatment with curative intent; a minimum of 2 years of follow-up; and radiation therapy (RT) delivered at the University of Florida to fields including a portion of the heart, carotid, or subclavian arteries. Patients whose RT fields included heart tissue comprised the cardiac subgroup; those with RT fields including carotid and/or subclavian arteries comprised the carotid–subclavian artery subgroup. The cardiac subgroup and the carotid–subclavian artery subgroup each had a total of 404 patients, with 393 included in both subgroups. Information was obtained retrospectively through RT, hospital, and physician records and through direct contact with the majority of patients or their families within a year of the time of analysis. Median follow-up was 11.2 years (range, 2.1-36.3 years); 345 patients (83.1%) had follow-up for more than 5 years, and 100 patients (24.1%) for more than 20 years, for a total of 5588 person-years of follow-up.

Coronary artery disease (CAD) was defined as a history of documented myocardial infarction, coronary artery bypass graft surgery, percutaneous coronary intervention, or more than 75% diameter stenosis on coronary angiography or autopsy. Noncoronary atherosclerotic disease was defined as 40% or more stenosis of the carotid or subclavian artery by ultrasound or angiography, transient ischemic attack (TIA), or stroke. Clinically important valvular dysfunction was defined as moderate or severe stenosis or insufficiency on echocardiogram or angiogram, or marked abnormality on autopsy. The presence or absence of hypertension, diabetes, hypercholesterolemia (total cholesterol, ≥200 mg/dL [5.19 mmol/L]), family history of CAD (at least 1 first-degree relative), history and quantity of cigarette smoking, and hypothyroidism were researched in each patient; diagnoses were based on medical records and self-reports during patient contacts.

Our cohort consisted of 251 men (60%) and 164 women (40%) with a median age at diagnosis of 25 years (range, 4-75 years). The Ann Arbor stages were 99 patients (24%), stage I; 184 patients (44%), stage II; 105 patients (25%), stage III; and 27 patients (7%), stage IV. Two hundred fifty-seven patients (62%) received chemotherapy and RT and 158 patients (38%) received RT alone. The initial treatment in the chemotherapy and RT subgroup was combined RT and chemotherapy for 188 patients (73%), initial RT alone followed by chemotherapy for recurrence for 48 patients (19%), and chemotherapy alone followed by RT for recurrence for 21 patients (8%; Table 1).

Two hundred nine patients (50%) received chemotherapy as part of the initial treatment plan; 90 (43%) of the chemotherapy regimens were doxorubicin-based. The RT fields were mantle alone for 54 patients (13%), mantle and subdiaphragmatic fields for 339 patients (81%), primarily subdiaphragmatic treating only the inferior portion of the heart for 11 patients (3%), and involved field for 11 patients (3%).

Treatment factors investigated for their impact on the late effects of interest were: (1) total dose delivered to various structures, (2) a field-matching technique resulting in a higher (up to 50% more than prescribed) total dose over the base of the heart beneath the junction of mantle and subdiaphragmatic radiation fields,¹⁷ (3) a technique used before 1975 in which the entire daily mantle dose was given through the anterior field 3 days a week and the posterior field 2 days a week, delivering a higher incremental dose to the heart and coronary arteries 3 days a week, (4) an anteriorly weighted radiation approach in which the majority of the mantle dose was delivered through the anterior field, and (5) the use of chemotherapy and particular chemotherapeutic agents (Table 1).

All radiation doses were calculated prospectively using Clarkson calculations¹⁸ and confirmed retrospectively. The calculated mid-mediastinal dose, located near the base of the heart, was used to estimate the radiation dose delivered to both coronary arteries and valves. The calculated low-mediastinal dose was not used because of the difficulty in precisely determining the actual delivered dose in the area of potential field overlap from abutting mantle and subdiaphragmatic fields. The calculated low-cervical dose was used to estimate the dose delivered to the carotid and subclavian arteries.

For all analyses, SAS software was used.¹⁹ Actuarial incidence estimates were computed using the Kaplan-Meier method.²⁰ The Kruskal-Wallis κ-sample test was used to compare ages and doses across event groups. The procedure incidence data from the National Hospital Discharge Survey (NHDS) from 1999 were accessed to estimate a baseline age- and sex-stratified national utilization rate for valve surgery (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 35.00 to 35.28), percutaneous intervention (ICD-9-CM codes 36.01 to 36.09), and coronary artery bypass graft surgery (ICD-9-CM codes 36.10 to 36.19).²¹ This utilization rate was applied to the 1999 US population estimate obtained from the Surveillance, Epidemiology, and End Results (SEER) database to establish an expected incidence for these procedures.²² The incidence of utilization of these procedures was taken as a surrogate for clinically significant valvular dysfunction and CAD. The incidence of these procedures in our database was compared with the expected national incidence to generate observed-to-expected ratios (OERs).²³

Cox multiple regression analyses²⁴ were used to evaluate the previously described treatment- and patient-related covariates as potential predictors of incidence of CAD, noncoronary arteriovascular disease, or clinically significant valvular dysfunction. Before initiating the analyses, the proportional hazards assumption was evaluated by testing each covariate with a corresponding time-dependent covariate in the same model. The assumption was found to be reasonable in all cases. A bootstrapping procedure was then implemented (1000 simple random samples were generated from the original dataset), in order to objectively select a reduced number of covariates to go into each final model and thereby prevent overfitting the model. Each of the 1000 bootstrapped samples was then analyzed with imputation to impute any randomly missing data. Each multiply imputed, bootstrapped sample was then analyzed via Cox regression. The results were summarized, and those covariates that entered a given model most frequently were entered into a Cox model to analyze the complete dataset. Backward selection was implemented to build the final model. A value of P ≤.05 was considered to be statistically significant.

Coronary artery disease was diagnosed in 42 patients (30 men and 12 women) at a median of 9 years after RT (range, 1-32 years). The actuarial incidence of CAD was 3% at 5, 6% at 10, and 10% at 20 years. Those who developed CAD were somewhat older (median, 34 years; range, 16-67 years) at initiation of RT compared with the remainder of the cardiac subgroup (median, 24 years). The mid-mediastinum radiation dose was similar in patients who subsequently developed CAD (median, 35 Gy; range, 25-42 Gy) to patients who did not develop CAD (median, 33 Gy; range, 10-47 Gy).

Twenty patients had coronary artery revascularization procedures, with coronary artery bypass graft surgery in 13 patients, percutaneous coronary intervention in 11, and both procedures in 5. The OERs for these procedures were as follows: coronary artery bypass surgery, 2.42 (observed, 13; expected, 5.30; 95% CI, 1.11-3.74); percutaneous coronary intervention, 0.86 (observed, 11; expected, 12.77; 95% CI, 0.04-1.37), and total procedures, 1.63 (observed, 24; expected, 14.75; 95% CI, 0.98-2.28). Available coronary angiography reports described localized stenosis of more than 75% in the left main (3), left anterior descending (12), right coronary (13), and circumflex (5) arteries.

All traditional cardiac risk factors tested (male sex, hypertension, hypercholesterolemia, and age) were significantly associated with the development of CAD. All 42 patients with CAD had at least 1 of these cardiac risk factors. The only treatment-related covariate significantly associated with development of CAD was the field matching technique (P = .04). Neither the use of chemotherapy nor any specific chemotherapeutic agents were associated with the development of CAD (Table 2).

Thirty patients had at least 1 of the following events: 10 patients had stroke; 7, TIA; 14, carotid artery stenosis; and 7, subclavian artery stenosis. Two of the 14 patients with image-documented carotid artery disease were among the 7 patients diagnosed as having subclavian artery stenosis. The actuarial incidence of noncoronary atherosclerotic disease was 2% at 5, 3% at 10, and 7% at 20 years. Six patients underwent carotid endarterectomy; 1 patient, carotid artery stent placement; 1, subclavian artery stent placement; and 1, subclavian artery bypass graft surgery. The median age for all patients who developed noncoronary vascular disease was 34 years when undergoing RT, and the median time from therapy to event was 17 years. However, among those who experienced TIA or stroke, the median age when undergoing RT was 51 years and the median time from therapy to event was only 5.6 years. In contrast, the median age of patients with isolated subclavian or carotid artery stenosis was 20 years when undergoing RT (range, 5-58 years) and the median time from therapy to event was 21 years (range, 5.4-33 years). The median low-cervical radiation dose was 44 Gy (range, 37-48 Gy) for those who developed subclavian stenosis and 36 Gy (range, 13-76 Gy) for those who did not (P = .002). The median low-cervical radiation dose was 38 Gy (range, 30-57 Gy) for those who developed carotid artery diseases and 36 Gy (range, 13-76 Gy) for those who did not (P = .05). Hypertension and diabetes were the only patient- or treatment-related covariates associated with the development of noncoronary atherosclerotic vascular disease (Table 3).

Clinically important valvular dysfunction developed at a median of 22 years after RT (range, 6-31 years) in 25 patients, a total of 29 valves were involved: 15 aortic, 11 mitral, and 3 tricuspid. The actuarial incidence of clinically important valvular dysfunction was 1% at 10, 4% at 15, and 6% at 20 years. None of these patients had a known history of rheumatic fever; however, 1 patient had subacute bacterial endocarditis and a second patient had a bicuspid aortic valve. The median age at initiation of RT was 22 years (range, 5-48 years) for patients with clinically important valvular dysfunction, which was at a significantly younger age than for those who developed CAD or noncoronary vascular disease (P<.001). The median mid-mediastinum radiation dose was 37 Gy (range, 23-44 Gy) for patients who developed and 33 Gy (range, 10-47 Gy) for patients who did not develop clinically important valvular dysfunction (P = .01). The dominant lesions were aortic stenosis in 14 valves, mitral insufficiency in 8, mitral stenosis in 3, tricuspid insufficiency in 3, and aortic insufficiency in 1. Valve surgery was required in 7 (47%) of 15 patients with dysfunctional aortic valves and 3 (27%) of 11 patients with dysfunctional mitral valves. The OER for valve surgery was 8.42 (observed, 10; expected 1.19; 95% CI, 3.20-13.65).

None of the potential patient- or treatment-related risk factors was significantly associated with the development of clinically important valvular dysfunction (Table 4).

In our study, the actuarial freedom from any cardiovascular morbidity was 88% at 15 years and 84% at 20 years. An increased risk of CAD among patients successfully treated for Hodgkin lymphoma has been well documented,²⁵ but the extent and severity of subsequent cardiac valve dysfunction is unclear. Most evidence for potential radiation-related valvular dysfunction comes from scattered reports and screening echocardiography in asymptomatic patients.^9-13 To the best of our knowledge, this is the first study to quantify a significant risk of clinically important valvular dysfunction in Hodgkin lymphoma survivors (5% at 20 years post-RT). Twenty-six of the 29 clinically significant valvular lesions were left-sided with the dominant dysfunction more likely to be stenosis (17 of 29) rather than insufficiency. The substantial incidence of valvular dysfunction in this study may be due to the extended follow-up; more than 24% of patients had more than 20 years of follow-up after RT. Given the median of a 22-year interval post-RT to diagnosis, clinically important valvular dysfunction is emerging as an important source of long-term morbidity among Hodgkin lymphoma survivors. Patients at risk for valve dysfunction differ from those at risk for CAD because of the absence of other risk factors, earlier age of RT, and longer post-RT latency period, suggesting the possibility that the mechanism of action in radiation–induced valve dysfunction may be different from that in CAD.

The 10.4% (42 of 404 patients) incidence of clinical CAD (corresponding to a 20-year actuarial incidence of 9.9%) with a median 11.2-year follow-up reported herein confirms and extends findings from previous reports.^26,27 Although a comparison group for incidence is not available, the OER for surgical or percutaneous revascularization procedures suggests a trend toward an increased risk of CAD. A typical mantle radiation field includes at least the right coronary and left anterior descending coronary arteries, which were the arteries most commonly affected in our series as well as in 2 other series.^23,27 The typical patterns of presentation in Hodgkin lymphoma preclude complete elimination of these vessels from the treatment field. Coronary artery disease, in our series and others,^7,9,23,27 occurs almost exclusively in patients with known cardiac risk factors. The correlation between radiation treatment technique and/or dose and the risk of pericarditis and myocarditis is clear, but the correlation between treatment technique and/or dose and CAD is less clear.^28-30 In our study, a previous irradiation technique used before 1990 that resulted in a 50% or more increase in total dose over a small section of cardiac tissue was significantly associated with the development of CAD, suggesting that radiation dose may be a factor.

Most publications reporting radiation-related non–coronary atherosclerotic vascular disease describe patients treated for carcinomas with higher doses of radiation than those used for Hodgkin lymphoma.^15,16,31-37 Based on our observations, we believe there were 2 distinct subgroups of Hodgkin lymphoma survivors who developed non–coronary atherosclerotic vascular disease. The first group is an older population with probable preexisting disease either not affected or only accelerated by radiation. These patients experienced strokes and TIAs, were older at RT exposure (median age, 51 years), and had a relatively short time interval (median, 5.6 years) to development of the vascular disease. The second group differs in that the patients were younger (median, 20 years) at RT exposure, had a longer latency period before diagnosis (median, 20.8 years), and had lesions not commonly seen in the general population that are more likely to be related to cell loss from radiation exposure at an early age. No national database is available to estimate the incidence of non–coronary atherosclerotic vascular disease in a matched cohort of patients, but an overall incidence of 7.4% (including a 3.5% incidence of documented carotid artery stenosis and a 1.7% incidence of subclavian artery stenosis) seems high in this relatively young patient population. Given the long latency period, it is probable that the frequency of these problems will increase as more patients, who were treated before the radiation dose reductions in the 1990s, reach their 20-year survival milestone.

Because most radiation effects are dose-related, it is probable that there is actually a dose effect for both noncoronary artery disease and clinically important valvular dysfunction not identified in this study because of the uniformity of doses used over the study period. It is probable that modern techniques using lower radiation doses and smaller treatment volumes combined with chemotherapy may reduce these risks. We believe that there should be a low threshold for evaluation of potential cardiac and vascular symptoms in survivors of Hodgkin lymphoma and that routine counseling and surveillance are indicated after Hodgkin lymphoma therapy to ensure timely interventions. New evaluation tools such as electron beam computed tomography imaging or magnetic resonance coronary angiography may facilitate better screening in high-risk asymptomatic subgroups of patients who survived Hodgkin lymphoma.^38,39