The risk of death from heart disease in patients with nonsmall cell lung cancer who receive postoperative radiotherapy

Analysis of the Surveillance, Epidemiology, and End Results database




This study was designed to investigate whether the mortality from heart disease, a manifestation of intercurrent disease after postoperative radiotherapy (PORT), has decreased over time for patients with nonsmall cell lung cancer (NSCLC).


The 17-registry 1973 to 2003 dataset from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program was used to create a cohort of patients with NSCLC who had evidence of ipsilateral lymph node involvement diagnosed from 1983 to 1993 and who underwent pnuemonectomy/lobectomy (n = 6148 patients). Heart disease mortality was the primary endpoint: Deaths from other causes were censored, and surviving patients were censored at 10 years. The independent variable was PORT use, and adjustment variables included age at diagnosis, sex, race, year of diagnosis, laterality, location, histology, and the operation performed.


Multivariate analysis revealed that PORT use was associated with an increase in heart disease mortality (hazards ratio [HR], 1.30; 95% confidence interval [95% CI], 1.04–1.61; P = .0193) along with older age, male sex, African-American race, and earlier year of diagnosis. The association was confirmed in the cohort that was diagnosed from 1983 to 1988 (HR, 1.49; 95% CI, 1.11–2.01 [P = .0090]) but not for the cohort that was diagnosed from 1989 to 1993 (HR, 1.08; 95% CI, 0.79–1.48 [P = .6394]).


The results from this study demonstrated that the risk of heart disease mortality associated with PORT has declined in more recent years. This may be secondary to improvements in the treatment planning and delivery of thoracic radiotherapy. Properly designed, prospective, adjuvant trials will be needed to verify these findings. Cancer 2007; 110:911–7. © 2007 American Cancer Society.

It has been demonstrated in multiple randomized trials that postoperative radiation therapy (PORT) improves local control for patients with resected nonsmall cell lung cancer (NSCLC).1–6 The enthusiasm for PORT declined, however, after a meta-analysis published in 1998 demonstrated a 7% absolute increase in mortality associated with PORT, despite a 24% reduction in local recurrences.7 The authors of that report suggested that the addition of PORT may have had an adverse effect by virtue of acute or delayed radiation effects, such as radiation pneumonitis or cardiotoxicity. Heart disease mortality accounts for the major component of late radiation-induced toxicity8, 9; the discrepancy between an increase in local control and a decrease in survival for patients receiving PORT warrants further investigation.

There have been a number of concerns raised regarding the 1998 PORT meta-analysis.10–13 Trials that were included in the meta-analysis utilized radiation techniques that now are recognized as less than optimal, particularly lateral radiation beam designs and relatively large radiation fields.14 Treatment with cobalt-60 teletherapy units was allowed in 7 of the 9 trials, and 4 trials used relatively large daily radiation fractionation schedules [> 2 grays (Gy) per day]; both of these techniques have been associated with increased normal tissue toxicity.15, 16 Taken together, these critical components of radiation delivery may have deleterious consequences in a population of patients with already compromised cardiac and pulmonary reserve.

A recent analysis by Lally et al.17 of > 7000 postoperative NSCLC patients who were treated in relatively modern times (1988–2002) indicated that, for patients with N2 lymph node involvement, PORT use demonstrated both an overall survival benefit and an NSCLC-specific survival benefit. The survival benefit may have been related to the sterilization of residual microscopic disease, a decrease in PORT-associated toxicity, or both. It is clear that major advances in the technology used both to plan and to deliver thoracic radiation therapy have occurred. We hypothesized that these improvements have resulted in a reduction in deaths from intercurrent, radiation-related heart disease for patients receiving PORT.



Data for this study were obtained from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program using the 17-Registry dataset from 1973 to 2003 (November 2005 submission, released May 2006). SEER is a national cancer surveillance program that collects information on all incident cancer cases from population-based cancer registries that cover approximately 26% of the U.S. population; the data are broadly representative of national demographics. The SEER Program registries routinely gather data on patient demographics, primary tumor site, tumor morphology, stage at diagnosis, first course of treatment, and follow-up for vital status. The SEER Program is the only comprehensive source of population-based information in the U.S. that includes data on stage of cancer at the time of diagnosis and patient survival.

From the SEER database, we assembled a cohort of patients aged ≥21 years who had pathologically confirmed, nonmetastatic NSCLC diagnosed from 1983 to 1993 and who were followed through 2005. We chose to examine the period form 1983 to 1993 because 1) prior to 1983, SEER did not contain adequate information on the surgery performed, a potential confounding factor, and 2) we wanted at least 10 years of follow-up for the assessment of heart disease mortality (a potential long-term toxicity of radiation therapy). Patients who were selected had to have evidence of ipsilateral hilar/mediastinal adenopathy. Patients who had mainstem bronchus tumors were excluded. Further inclusion criteria were having undergone either a lobectomy or pneumonectomy, having received either postoperative external-beam irradiation or no irradiation, and having survived for ≥3 months to account for perioperative mortality. The following NSCLC histologic types were included: large cell carcinoma (large cell carcinoma, large cell neuroendocrine carcinoma, and giant cell carcinoma); squamous cell carcinoma (papillary squamous cell carcinoma, squamous cell carcinoma, squamous cell carcinoma [keratinizing], squamous cell carcinoma [large cell, nonkeratinizing], and squamous cell carcinomas [small cell, nonkeratinizing]); adenocarcioma (adenocarcinoma and acinar cell carcinoma); bronchioalveolar adenocarcinoma (bronchioloalveolar adenocarcinoma, alveolar adenocarcinoma, bronchioloalveolar carcinoma [nonmucinous], and bronchioloalveolar carcinoma [mucinous]); adenocarcinoma with mixed subtypes (adenocarcinoma with mixed subtypes, papillary adenocarcinoma, clear cell adenocarcinoma, mucinous adenocarcinoma, mucin-producing adenocarcinoma, signet ring cell carcinoma, and adenosquamous carcinoma).

SEER recodes the cause of death based on International Classification of Disease (ICD) 9th revision codes (available at URL: Patients were considered to have died of heart disease when their SEER cause of death recode was 50060. This corresponds with the following ICD 9th revision codes: 390 through 398, 403, 404, or 410 through 429. Together, this includes rheumatic heart disease, hypertensive heart disease, hypertensive heart and kidney disease, ischemic heart disease, disease of pulmonary circulation, and other forms of heart disease. Patients who received PORT were compared with patients who did not receive PORT. Potential covariates included patient age at diagnosis, sex, race, year of diagnosis, laterality (right vs left), location (upper lobe, middle lobe, lower lobe, or main bronchus), histology, and the operation performed (lobectomy vs pneumonectomy). Data on margin status, performance status, use of adjuvant chemotherapy, and thoracic radiotherapy treatment details (dose, fractionation, beam energy, etc) were not available within the SEER database and are not included in this analysis.

Statistical Analysis

We used the chi-square test to compare the prevalence of covariates among patients who did and did not receive PORT and then used proportional hazards models to examine the adjusted association of PORT and potential covariates with heart disease mortality. Patients contributed person-time from the date of their NSCLC diagnosis until they died of heart disease, or they were censored because of death from other causes or because they achieved 10 years of survival. Based on our a priori understanding of PORT and heart disease, all models included patient age at diagnosis, sex, race, year of diagnosis, laterality, location, histology, and the operation performed as either stratification or adjustment variables. First, we created an overall model. Then, to examine possible temporal trends more specifically, we created a model that was stratified by the first (1983–1988) and second (1989–1993) 5 years of our study period. Finally, to provide additional evidence of the association between PORT and heart disease, within each period, we created a model stratifying by the upper right lobe, in which radiation exposure would be minimal, versus all other lobes, in which there would have been a risk of heart irradiation.

All data were analyzed using the SAS version 8.02 (SAS Institute, Cary, NC) statistical software package. Results were considered statistically significant when P < .05. This study was approved by the Wake Forest University Health Sciences Institutional Review Board and was conducted in full compliance with federal, state, and institutional regulations and guidelines.


This analysis included 6148 patients: 3589 patients (58%) who received PORT and 2559 patients (42%) who did not receive PORT. The median patient age at diagnosis within the cohort was 64 years (range, 24–88 years). The median follow-up for all patients was 2.1 years, but it was 10 years for those patients who remained alive at the time of censoring. Patient sex, race, tumor laterality, and year of diagnosis were not associated significantly with the use of adjuvant therapy. Younger patients were more likely to receive PORT; significant differences also were observed in terms of tumor histology and the surgical procedure performed (Table 1).

Table 1. Characteristics of Patients With Resected, Lymph Node-positive Nonsmall Cell Lung Cancer and the Frequency of Death from Heart Disease*
Variable PORT (%)No-PORT (%)P
  • PORT indicates postoperative radiotherapy; BAC, bronchioalveolar adenocarcinoma; AMST, adenocarcinoma with mixed subtype.

  • *

    P values demonstrate the significance of differences in frequency (according to the chi-square test) in distribution of patients between the PORT group and the no PORT group.

Age<60 years1294 (36%)702 (27%)<.001
60–70 years1565 (44%)1102 (43%)
>70 years730 (20%)755 (30%)
SexMale2225 (62%)1577 (62%)0.769
Female1364 (38%)982 (38%)
RaceCaucasian3080 (86%)2233 (87%)0.196
African American322 (9%)215 (8%)
Other187 (5%)111 (4%)
HistologyLarge cell273 (8%)202 (8%)0.040
Squamous cell1355 (38%)1036 (40%)
Adenocarcinoma1500 (42%)969 (38%)
BAC165 (5%)133 (5%)
AMST296 (8%)219 (9%)
LateralityRight1854 (52%)1265 (49%)0.086
Left1735 (48%)1294 (51%)
Subsite locationUpper lobe2246 (63%)1484 (58%)0.001
Middle lobe165 (5%)123 (5%)
Lower lobe1178 (33%)952 (37%)
SurgeryLobectomy2731 (76%)1821 (71%)<0.001
Pneumonectomy858 (24%)738 (29%)
Year of diagnosis1983–19881829 (51%)1279 (50%)0.448
1989–19931760 (49%)1280 (50%)
Death from heart disease207 (6%)147 (6%)0.969
Death from all causes3139 (87%)2161 (85%)0.001

The primary focus of our investigation was to determine whether the heart disease mortality associated with PORT in the past has decreased with time. In multivariate analysis for all patients, PORT use was associated with an increase in the hazard for heart disease mortality (hazards ratio [HR], 1.30; 95% confidence interval [95% CI], 1.04–1.61 [P = .019]) (Table 2). PORT was associated with a significant increase in the hazard for heart disease mortality for patients who were diagnosed with NSCLC from 1983 to 1988 (HR, 1.49; 95% CI, 1.11–2.01 [P = .009]) but not for patients who were diagnosed from 1989 to 1993 (HR, 1.08; 95% CI, 0.79–1.48 [P = .639]). Figure 1 presents the heart disease-specific survival curves for these 2 subsets stratified by PORT use. Older age, male sex, and year of diagnosis were associated with an increase in the hazard for heart disease mortality in all 3 models. Interaction terms investigating the relation between PORT use and age were tested in a multivariate model and were not significant.

Figure 1.

Plot of heart disease mortality free survival for 2 different time eras stratified by postoperative radiotherapy (PORT) use.

Table 2. Multivariate Analysis of Heart Disease Morality Hazard for Patients With Resected, Lymph Node-positive Nonsmall Cell Lung Cancer*
VariableAll Years1983–19881989–1993
  • HR indicates hazards ratio; 95% CI, 95% confidence interval; PORT, postoperative radiotherapy; Ref, reference; BAC, bronchioalveolar adenocarcinoma; AMST, adenocarcinoma with mixed subtype.

  • *

    An HR > 1 was associated with increased mortality. The model includes all variables listed in the table.

Adjuvant therapy
 No PORT1.00 (Ref)  1.00 (Ref)  1.00 (Ref)  
 Increasing (continuous)1.051.04–1.07<.0011.061.04–1.08<.0011.051.03–1.07<.001
 Men1.00 (Ref)  1.00 (Ref)  1.00 (Ref)  
 African American1.521.08–2.14.0151.240.75–2.05.4131.851.17–2.95.009
 Caucasian1.00 (Ref)  1.00 (Ref)  1.00 (Ref)  
 Large cell0.560.32–0.98.0410.490.21–1.13.0950.600.29–1.26.174
 Squamous cell1.260.98–1.61.0681.390.98–1.97.0631.130.79–1.62.492
 Adenocarcinoma1.00 (Ref)  1.00 (Ref)  1.00 (Ref)  
 Right1.00 (Ref)  1.00 (Ref)  1.00 (Ref)  
Subsite location
 Upper lobe0.840.67–1.05.1170.880.65–1.20.4210.760.55–1.06.104
 Middle lobe0.890.52–1.54.6860.930.40–2.18.8640.760.37–1.56.456
 Lower lobe1.00 (Ref)  1.00 (Ref)  1.00 (Ref)  
 Lobectomy1.00 (Ref)  1.00 (Ref)  1.00 (Ref)  
Year of diagnosis
 Decreasing (continuous)1.051.02–1.09.0021.141.05–1.24.0031.141.03–1.27.013

When we examined the association of PORT and heart disease mortality by anatomic location, we observed that PORT was not associated with heart disease mortality for patients who had tumors located in right upper lobe regardless of their period of diagnosis. For all other locations, PORT was associated with increased heart disease mortality for theearlier cohort (HR, 1.69; 95% CI, 1.18–2.41 [P = .004]), but not for the later cohort (HR, 0.91; 95% CI, 0.62–1.32 [P = .611]) (Table 3).

Table 3. Multivariate Analysis of Heart Disease Morality Hazard for Patients With Resected, Lymph Node-positive Nonsmall Cell Lung Cancer Stratified by Risk Group for Cardiac Irradiation*
CohortRUL tumorsAll other tumors
  • RUL indicates right upper lobe; HR, hazards ratio; 95% CI, 95% confidence interval; PORT, postoperative radiotherapy; Ref, reference.

  • *

    The model also included variables for age, histology, sex, race, surgery, and year of diagnosis.

 No PORT1.00 (Ref)  1.00 (Ref)  
 No PORT1.00 (Ref)  1.00 (Ref)  


Since its publication in The Lancet in 1998, the meta-analysis7 of 9 European trials has become a primary source of evidence regarding the use PORT for patients with NSCLC; the meta-analysis demonstrated a benefit in local control for PORT but raised concerns because of a decrease in overall survival. In the current, large SEER analysis of patients who were treated in the U.S. during a similar era, PORT was associated with an increase in heart disease mortality independent of the expected effects of age, sex, and race. It appears that this increase was limited to patients who were diagnosed from 1983 to 1988 who had tumors that were not located in the right upper lobe. The magnitude of this effect for this subgroup is consistent with that reported in the 1998 meta-analysis.7 Although we acknowledge that the overall management of heart disease has improved with time,18, 19 we speculate that our results represent improvements in radiotherapy techniques that minimize irradiation of the heart.

Technical improvements in the application of thoracic radiation therapy for lung cancer, particularly with the integration of computed tomography (CT)-based treatment planning, have greatly improved our ability to identify tumors and, relevant to this report, to delineate normal tissues. CT was introduced into clinical use in the U.S. in 1974,20 and its usefulness in radiation treatment planning was recognized immediately.21, 22 The incorporation of CT imaging for patients in the first cohort (1983 to 1988) would have been limited, because only 3.3 million CT scans (most of the head/brain) were performed in the U.S. in 1980. This number increased to 18.3 million CT scans by 1993 (unpublished results).

The 1980s also marked a change in the treatment used to deliver radiation therapy as the linear accelerator (LINAC) began to replace cobalt-60 teletherapy.15 The early 1990s were associated with the combining of the LINAC with CT imaging, resulting in 3-dimensional (3-D) conformal radiotherapy. Improvements in these various components of radiation therapy have resulted in an improved ability to deliver safe radiation doses to primary tumors and, at the same time, greater sparing of normal tissues.23, 24 We suspect that some (but not all) patients in our analysis likely had care that involved 3-D conformal radiotherapy; however, the billing codes for that treatment did not appear until 1994.25 Nevertheless, in centers in which examination of CT images was available but 3-dimensional conformal treatment planning was not, we suspect that radiation oncologists manually transferred target volumes and normal tissue volumes from CT images to simulation films.

In the report by Van Houtte et al.1 (1966–1975), radiotherapy was given using a 3-field technique, and the target volume included the mediastinum from the sternal notch down to 5 cm below the carina. The standard field size was 15 cm × 9 cm. The authors believed that the treated volume and the dose used were too excessive to show any improvement. In a report by the Lung Cancer Study Group6 (1978–1985), a similar target delineation of the mediastinum was used. It was not until the report by Dautzenberg et al. (1986–1994) that computer dosimetry was recommended. Furthermore, simulator films were checked with additional films that were taken with the treatment beam. In the report by Keller et al. (1991–1997), although the mediastinum was defined by anatomic landmarks, postoperative CT scans were required to document surgical changes and limit > 50% cardiac volume to not more than 35 Gy. It is clear that radiotherapy was evolving from nonsite-specific techniques using bony anatomy and hand-drawn blocking toward specialized planning that incorporated CT-based images and computer optimization algorithms over the time frame of this analysis (1983–1993).

The chronic toxicity of heart disease mortality from thoracic radiation has been studied in patients with lymphoma and breast cancer.26–28 Of particular relevance to our results is a report by Giordano et al.29 suggesting that patients with breast cancer who were treated with older technology (1973–1979) experienced an increase in radiotherapy-associated heart disease mortality compared with patients who were treated in more recent years. This finding did not begin to emerge until 10 years of follow-up had been established.

Today, for patients with stage II and IIIA NSCLC, adjuvant chemotherapy is considered the standard of care.30–34 More follow-up and analysis are needed to determine how the use of cytotoxic chemotherapy impacts the patterns of failure and/or late toxicity. If chemotherapy is able to control micrometastatic disease such that there is a proportional shift toward an increase in local failures, then PORT may have an increasingly greater impact. Adjuvant chemotherapy is unlikely to have confounded the findings in this study, because adjuvant chemotherapy was not standard during the period of the study.

The lack of detailed information on specific radiotherapy techniques and patient comorbidities constitutes a major limitation of our analysis. Detailed individual dose and volume information would have allowed us to identify the specific techniques that increase heart disease mortality. Comorbid conditions present at NSCLC diagnosis likely are associated both with the treatment received and heart disease mortality, but we could not adjust for this potential confounding variable. Nevertheless, the SEER database is one of the few options for rigorously collected data that can be used to examine the late effects of cancer treatments in a large population that is representative of all patients diagnosed with NSCLC in the U.S.

A prospective, adjuvant chemoradiation therapy trial should be designed to confirm our observations. This trial will need to be designed properly with acceptable endpoints and adequate power. Radiotherapy dose- and treatment-planning techniques must be the current standard with proper, centralized quality assurance of the radiation technique utilized. Failure to use such a design not only will result in more controversy35, 36 but also will fail to improve the standard of care for lung cancer patients after surgery.

In conclusion, our analysis suggests that PORT-associated heart disease mortality, the main source of death from intercurrent disease, has decreased over time, particularly for patients who have tumors located where heart irradiation is more likely to occur. We speculate that this is secondary to improvements in radiotherapy techniques. A properly designed, prospective, adjuvant (chemo)radiation therapy adjuvant trial should be initiated to confirm this finding. Until the results of such a trial are available, analysis of the available data suggests that the benefits of PORT with available technology outweigh the risks.


We thank Dr. Carolyn Ferree and Dr. Stacy Wentworth for their critical review of the manuscript. We also thank Dr. Luther Brady and Dr. Theodore Yaeger for their assistance in understanding the history of radiotherapy technology.