Department of Radiation Oncology, Abramson Cancer Center, University of Pennsylvania, Philadelphia, Pennsylvania
Corresponding author: Smith Apisarnthanarax, MD, Department of Radiation Oncology, Perelman Center for Advanced Medicine, 4th Fl West, 3400 Civic Center Blvd, Philadelphia, PA 19104; Fax: (215) 349-8975; firstname.lastname@example.org
Abstract previously presented at the American Society for Clinical Oncology 2012 Gastrointestinal Cancers Symposium; January 19-21, 2012; San Francisco, CA.
The addition of chemoradiation (CRT) to surgery has been shown to improve survival in patients with esophageal cancer. In the current study, the authors determined whether the sequencing of CRT has an effect on survival and cardiopulmonary mortality in patients with esophageal cancer.
Patients with the following inclusion criteria were identified within 17 Surveillance, Epidemiology, and End Results registries from 1988 through 2007: adenocarcinoma or squamous cell carcinoma of the esophagus and having undergone esophagectomy. Patients who died within 90 days of surgery were excluded. Demographic, tumor, and survival data were compared between patients receiving preoperative and postoperative RT. Cox proportional hazards regression models were calculated to identify parameters associated with cause-specific survival and overall survival. A competing risk analysis was performed to account for death due to esophageal cancer in the calculation of cardiopulmonary mortality.
Of 5512 patients, 1881 received preoperative RT, 901 received postoperative RT, and 2730 did not receive RT. Patients receiving preoperative RT had improved 5-year cause-specific survival (41% vs 31%; P < .0001) and overall survival (33% vs 23%; P < .0001) compared with those receiving postoperative RT. No differences in adjusted cardiopulmonary mortality were found between patients who received RT versus those who did not (8% vs 10% at 10 years; hazards ratio [HR], 0.84 [95% confidence interval (95% CI), 0.64-1.12] [P = .24]) or between those treated with preoperative RT versus those treated with postoperative RT (HR, 0.70; 95% CI, 0.46-1.08 [P = .11]).
In patients with locally advanced esophageal cancer, the addition of chemoradiation (CRT) to surgery has shown improvements in overall survival (OS) when given preoperatively. In the subset of patients with tumors of the gastroesophageal junction, CRT delivered postoperatively has also demonstrated a survival benefit. To the best of our knowledge, no randomized controlled trials currently exist that compare the outcomes of the sequencing of CRT in relation to surgery in patients with esophageal cancer.
Preoperative CRT is preferred over postoperative treatment in patients with other gastrointestinal malignancies such as rectal cancer, in whom preoperative CRT has been demonstrated to have a superior locoregional control rate, toxicity profile, and survival rate.[8, 9] It is postulated that preoperative treatment is favored because of tumor downstaging and smaller radiation treatment fields. We hypothesized that preoperative RT is associated with decreased cardiopulmonary mortality from lower RT doses to the heart and lungs as well as improved survival outcomes compared with postoperative RT in patients with esophageal cancer.
Using the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database, we assessed differences in survival outcomes between patients treated with preoperative RT and those treated with postoperative RT and analyzed the effect of RT on heart and lung morbidity as defined by cardiopulmonary mortality in 2 comparisons: those patients receiving any RT versus those receiving no RT and those receiving preoperative versus postoperative RT.
MATERIALS AND METHODS
SEER Database and Queries
The SEER database is a national cancer surveillance program that collects information regarding the incidence and survival of cancer cases. It has been extensively used for survival and outcomes research in a variety of organ sites. The SEER database was queried for patients who were diagnosed with stage I to stage III (any T, any N, M0) (according the seventh edition of the American Joint Committee on Cancer) adenocarcinoma or squamous cell carcinoma of the esophagus and underwent esophagectomy between 1988 and 2007. Patients who died within 90 days of surgery were excluded to limit the confounding effect of perioperative mortality. Follow-up was defined as the time from the diagnosis to date of death or through December 31, 2009, whichever occurred first. SEER*Stat statistical software (version 7.0.4) was used to perform all queries.
The following patient and tumor characteristics were collected: age at diagnosis, sex, race, marital status, median income of the patient's county of residence, year of diagnosis, histology, T classification, number of positive lymph nodes, and the total number of lymph nodes examined. Patient characteristics were compared between patients who received preoperative RT and those receiving postoperative RT using the chi-square test. The use of RT and sequencing of RT in relation to esophagectomy (preoperative or postoperative) was recorded. No data were available concerning the use of chemotherapy.
Endpoints and Statistical Analyses
Kaplan-Meier curves (along with log-rank statistics) were estimated to compare preoperative versus postoperative RT on both OS and cancer-specific survival (CSS). Furthermore, a multivariable Cox proportional hazards regression analysis was used to estimate the effect of radiation sequencing on both OS and CSS with adjustment for age, sex, race, marital status, median income in the patient's county of residence, radiation sequence, T classification, lymph node status, total number of lymph nodes examined, and tumor histology. Tumor staging was derived by taking data regarding the available extent of disease and lymph node status and recoding them as per the American Joint Committee on Cancer guidelines (seventh edition). Due to the nature of the SEER database, information regard the T and N classifications reflects pathologic staging after surgery, and clinical staging at the time of the initial presentation was not available. A sensitivity analysis was performed to address the effect of unmeasured clinical staging as a potential confounder on clinical outcomes.[15, 16]
Cardiopulmonary mortality was defined as death because of “diseases of the heart”; “pneumonia and influenza”; “other diseases of arteries, arterioles, and capillaries”; “atherosclerosis”; “lung and bronchus”; and “chronic obstructive pulmonary disease and allied conditions.” A competing risks analysis was performed in which death due to esophageal cancer was considered as a competing risk for cardiopulmonary mortality. The analysis was performed for 2 comparisons: 1) patients who received RT at any time during their treatment of esophageal cancer versus those who received no RT; and 2) patients who received preoperative versus postoperative RT.
All statistical analyses were performed using Stata statistical software (version 11.1; (StataCorp LP, College Station, Tex). Unless otherwise specified, all tests were 2-sided with a P value < .05 considered to be statistically significant. Sensitivity analyses were performed using R statistical software (version 2.14; The R Foundation for Statistical Computing, Vienna, Austria).
Patient and Tumor Characteristics
The study cohort comprised 5512 patients diagnosed with esophageal cancer and treated with esophagectomy during the period from 1988 to 2007 (Table 1). The median age at the time of diagnosis was 63 years, with an interquartile range of 56 years to 71 years. The median survival was 28 months, with a median follow-up of 73 months. There was a strong predilection for male sex, compromising 80% of patients in the analysis. Furthermore, approximately 86% of patients in the analysis were white. More patients had histologically proven adenocarcinoma (67%) compared with squamous cell carcinoma (33%); however, this percentage was not consistent over the study period (Fig. 1A). There was a slight preponderance (58% vs 42%) of squamous cell carcinoma in patients diagnosed in 1988. The percentage of patients with adenocarcinoma increased over time; in patients diagnosed in 2006, 76% had adenocarcinoma, with only 24% of patients found to have squamous cell carcinoma.
A total of 2782 patients (50%) received external beam RT, with an approximate 2:1 ratio (1881 vs 901) of patients receiving preoperative to postoperative RT. There were statistically significant differences noted with regard to race, tumor histology, T classification, lymph node status, and number of lymph nodes examined between patients treated with preoperative compared with those treated with postoperative RT (Table 1). Approximately 90% of patients treated with preoperative RT were white compared with 83% of the patients treated with postoperative RT (P < .001). Adenocarcinomas comprised 69% of cases in the preoperative RT group versus 61% in the postoperative RT group (P < .001). T3 and T4 tumors were more frequent in the patients receiving postoperative RT compared with those treated with preoperative RT (76% vs 63%; P < .001). Likewise, positive lymph nodes were more frequent in the patients treated with postoperative RT (70% vs 46%; P < .001). However, because T and N classifications reflect pathologic staging, the effect of preoperative RT on primary tumor and lymph node downstaging must be taken into account when interpreting these differences between T and N classifications. The percentage of patients receiving preoperative versus postoperative RT was 1:1 until 1998, at which point significantly more patients began receiving preoperative RT (Fig. 1B). In the treatment era between 2001 and 2007, the ratio was 3:1 of patients receiving preoperative versus postoperative RT.
Effect of Radiation Timing on Survival
Preoperative RT was associated with a statistically significantly increased overall survival compared with postoperative RT (HR, 0.78; 95% CI, 0.71-0.85 [P < .0001]). The median survival was 27 months versus 20 months, the 5-year OS rate was 33% versus 23%, and the 10-year OS rate was 22% versus 15% for preoperative and postoperative RT, respectively (Fig. 2A). Preoperative RT remained statistically significantly associated with improved OS after adjustment for covariates (HR, 0.88; 95% CI, 0.78-0.98 [P = .029]) (Table 2). This also held true across treatment eras with an HR of 0.81 (95% CI, 0.69-0.95; P = .008) for the treatment era between 1988 and 1998, and an HR of 0.87 (95% CI, 0.77-0.98; P = .02) for patients treated after that period. CSS was also found to be superior for preoperative RT compared with postoperative RT, with a median CSS of 34 months versus 23 months, 5-year CSS rates of 39% versus 28%, and 10-year CSS rates of 33% versus 24% (HR, 0.75; 95% CI, 0.67-0.83 [P < .0001) (Fig. 2B).
Table 2. Univariable and Multivariable Analyses of Overall Survival
Because data were not available regarding pretreatment clinical tumor staging, the potential imbalance of clinical T and N stage distribution between patients treated with preoperative RT versus those treated with postoperative RT (with the preoperative RT group having a higher prevalence of clinically overstaged patients) may have led to an overestimation of the estimated HR for preoperative RT versus postoperative RT. A sensitivity analysis was performed to account for unmeasured clinical stage as a confounder on the estimated treatment effect HR. In these analyses, a range of 10% to 20% was used for the percentage of patients treated with preoperative RT being clinically overstaged and 5% to 10% was used for the patients treated with postoperative RT.[18, 19] In addition, an assumption was made that the HR associated with a lower clinical stage was between 0.60 and 0.80 (protective effect). Under plausible scenarios, the HR adjusted for clinical stage remained similar to our original estimate (HR, 0.88), falling within the range of 0.89 to 0.92.
Factors found to be associated with poor prognosis on univariable analysis were T3/T4 tumor classification, positive lymph node status, greater than 10 lymph nodes examined, squamous cell carcinoma histology, age > 65 years, male sex, nonwhite race, unmarried status, and a median county income of < $50,000 (Table 2). On multivariable analysis, tumor stage, lymph node status, number of lymph nodes examined, age, sex, and median county income remained statistically significant whereas marital status, race, and tumor histology were not found to be statistically significant. Among the tumor characteristics, lymph node status and T classification were associated with a worse OS (HRs of 1.60 and 1.40, respectively).
Effect of RT on Cardiopulmonary Mortality
Cardiopulmonary mortality was compared in all patients (n = 5512) who were treated with surgery alone with those who received RT (preoperative or postoperative) using a competing risk model to account for the risk of death due to esophageal cancer. This analysis demonstrated no difference in the cohorts (HR, 0.84; 95% CI. 0.64-1.12 [P = .24]), with 5-year and 10-year estimated rates of cardiopulmonary mortality of 6% and 10%, respectively, in the group receiving no RT and 5% and 8%, respectively, in the group who received RT (Fig. 3A). Because it is conceivable that preoperative treatment could negatively impact cardiopulmonary mortality immediately occurring after surgery, cardiopulmonary mortality was also compared when including patients who died within 90 days of surgery (309 additional patients); similarly, we found no significant difference: the 5-year and 10-year estimated rates of cardiopulmonary mortality were 7% and 10%, respectively, in the group receiving no RT and 5% and 8%, respectively, in the group treated with RT. Using multivariable competing risk regression (Table 3), age > 65 years was found to be statistically significantly associated with poorer cardiopulmonary mortality, while being married and a having a negative lymph node examination were associated with favorable cardiopulmonary mortality. A subgroup analysis of patients aged < 65 years demonstrated no difference in cardiopulmonary mortality (HR, 1.00; 95% CI, 0.71-1.41 [P = .98]). To account for the potential effects of improved surgical and RT techniques over time, we assessed whether there were differences with regard to cardiopulmonary mortality during different time eras in the current study. Although a trend toward decreased cardiopulmonary mortality over the study period was noted, the differences were not statistically significant between treatment eras: HR of 1.21 (95% CI, 0.92-1.58; P = .18) for RT versus no RT in 1988 through 1999 and HR of 0.99 (95% CI, 0.72-1.39 [P = .99]) in 2000 through 2007.
Table 3. Multivariable Competing Risk Analysis for Cardiopulmonary Morbidity
Because our hypothesis was that preoperative RT may lead to a decrease in cardiopulmonary toxicity and subsequent cardiopulmonary mortality because of a reduction in the normal tissue volumes being irradiated (Fig. 4), we compared cardiopulmonary mortality between patients treated with preoperative and postoperative RT using a similar competing risk model. No statistical difference in cardiopulmonary mortality was found for patients who received preoperative RT versus postoperative RT on both univariable and multivariable analysis (HR, 0.70; 95% CI, 0.46-1.08 [P = .11]) (Table 3). The 5-year and 10-year rates of cardiopulmonary mortality were 5% and 8%, respectively, for patients receiving preoperative RT versus 6% and 8%, respectively, for patients receiving postoperative RT (Fig. 3B). A subgroup analysis containing only patients aged < 65 years also indicated no difference in cardiopulmonary mortality (HR, 1.05; 95% CI, 0.63-1.76 [P = .85]).
In our analysis of patients with esophageal cancer in the SEER database who were treated with a combined modality treatment, we found that preoperative RT resulted in superior OS and CSS compared with postoperative RT without increasing cardiopulmonary mortality versus surgery alone. The benefits of preoperative RT over postoperative RT with regard to OS and CSS are both statistically and clinically significant, with an approximately 10% absolute benefit noted at 5 years. These benefits were especially prominent in patients with T3/T4 or lymph node-positive disease, suggesting that these patients may benefit the most from a course of preoperative RT. Our finding that the use of RT, either preoperatively or postoperatively, does not significantly increase cardiopulmonary mortality compared with surgery alone is important when justifying the addition of a potentially morbid treatment modality.
Preoperative RT has the theoretical advantages of tumor downstaging and smaller treatment volumes. In the current study, it was not possible to ascertain whether the survival benefits observed with preoperative RT were attributable to tumor downstaging because the tumor staging information in the SEER database reported only pathologic staging data. Because preoperative treatment fields in patients with esophageal cancer (5 cm superiorly and inferiorly from macroscopic tumor disease) are classically smaller than postoperative treatment fields (coverage of the entire postoperative esophagectomy surgical bed), they should result in less radiation dose to the heart and lungs (Fig. 4) and be less toxic. We hypothesized that the survival benefits of preoperative RT may be related to decreased cardiopulmonary mortality.
Several studies in the breast cancer literature have demonstrated that cardiac-related mortality decreases over time in patients treated with RT after lumpectomy, presumably because of improved, modern radiation techniques that minimize the radiation dose to the heart. However, the published literature concerning late cardiopulmonary mortality in patients with esophageal cancer who are treated with RT is scarce. Morota et al investigated 74 patients who were treated with definitive concurrent CRT using an extended field from the supraclavicular fossa to the gastroesophageal junction, and concluded that the risk of grade 3 or greater cardiopulmonary toxicities was greater in patients aged > 75 years (29% vs 3%). In Radiation Therapy Oncology Group (RTOG) 85-01, 28% of patients had grade 3 or higher toxicity, but only 1 of those cases was related to cardiopulmonary disease. Ishikura et al have shown that large-field RT with concurrent chemotherapy has a high rate (12%) of grade 3 or higher cardiopulmonary toxicity. Recent data published in patients with esophageal cancer have also indicated a direct correlation between cardiac radiation volume/dose and symptomatic cardiac toxicity. However, the results in the population of patients with esophageal cancer in the current study demonstrated no difference in cardiopulmonary mortality between patients treated preoperatively versus those treated postoperatively.
The most likely explanation for this lack of difference is that patients are far more likely to die of esophageal cancer before reaching the point at which RT may have an increased effect on cardiopulmonary mortality, as supported by the few total number of cardiopulmonary events noted in the current study (n = 405) compared with esophageal cancer-related deaths (n = 3124). The competing risk analysis was performed to account for this possible confounding issue, but failed to demonstrate a higher incidence of cardiopulmonary mortality in patients treated with RT. Another possible explanation for this lack of difference may be related to improved treatment techniques, which may translate into decreased cardiopulmonary toxicity and offset the deleterious effects of postoperative RT. This possibility is supported by the suggestion in the current study data that there may be a trend toward decreased cardiopulmonary mortality over time. A third possibility is that patients in the SEER cohort who were treated with postoperative RT may not have been consistently treated with classic, large treatment volumes that encompassed the entire surgical bed. However, details of radiation treatments are not available in the SEER database to investigate this issue further.
Although the results of the current study indicate no difference in cardiopulmonary mortality between the patients receiving preoperative RT and those receiving postoperative RT, the SEER database does not allow us to address the potentially increased risk of cardiopulmonary morbidity that is not ultimately the cause of death. Patients who receive postoperative RT may be more likely to experience adverse side effects that are not severe enough to cause an increase in mortality. It is possible that patients treated with preoperative RT are more likely to tolerate additional interventions after surgery such as adjuvant chemotherapy and that this is the source of the survival difference between the preoperative and postoperative RT groups. However, the limitations of the SEER database prevent us from performing this analysis. There is a broad range of late cardiopulmonary mortality including but not limited to acute myocardial infarction, pericarditis, pericardial or pleural effusion, heart failure, and radiation pneumonitis.
We acknowledge limitations in the current study. First, because patients who receive preoperative RT have a risk of being clinically overstaged (approximately 15%),[18, 19] the group treated with preoperative RT may have included patients with a more favorable diagnosis, resulting in overinflated survival outcomes and a misclassification bias. This bias is compounded by the inability to directly compare patients stage-for-stage in the SEER database because tumor staging is reported as postsurgical pathologic staging. However, the sensitivity analysis as performed in the current study suggests that the survival advantages of preoperative RT remained clinically significant when taking into account various scenarios of clinical overstaging in the patient population.
Second, an important caveat of the current study is that the SEER database does not provide data regarding chemotherapy. However, both preoperative and postoperative RT are almost always administered concurrently with chemotherapy,[1, 7, 26, 28] and therefore it is less likely that the lack of chemotherapy data is a source of bias in the concurrent setting. In addition, chemotherapy itself can be a possible cause of cardiopulmonary mortality. Although there has been some research regarding cardiopulmonary mortality in patients with esophageal cancer who are treated with both chemotherapy and RT, to the best of our knowledge there have been no studies performed to date to determine whether there is a synergistic effect between the 2 modalities on cardiopulmonary mortality. Finally, the addition of adjuvant chemotherapy after trimodality treatment could also be a potential confounder. However, none of the randomized trials evaluating trimodality included the use of adjuvant chemotherapy, and therefore it is unlikely that a significant percentage of patients received adjuvant chemotherapy after undergoing trimodality treatment.
In summary, preoperative RT appears to provide superior survival compared with postoperative RT and does so without increasing cardiopulmonary mortality over surgery alone as studied in this population-based analysis of patients with esophageal cancer from the SEER database. With the lack of randomized controlled trials addressing the optimal sequencing of RT in relation to surgery, these data support the use of preoperative RT as the standard of care for patients with locally advanced esophageal cancer and should not be withheld out of concern for cardiopulmonary mortality.