Outcome Results of the 1996-1999 Patterns of Care Survey of the National Practice for Patients Receiving Radiation Therapy for Carcinoma of
http://www.100md.com
《临床肿瘤学》
the University of Maryland School of Medicine, Baltimore, MD
American College of Radiology
Abramson Cancer Center at the University of Pennsylvania, Philadelphia, PA
Community Medical Center, Toms River, NJ
Boston University Medical Center
Massachusetts General Hospital, Boston, MA
Cancer Center at St Agnes, Fresno, CA
Memorial Sloan-Kettering Cancer Center, New York, NY.
ABSTRACT
PURPOSE: A Patterns of Care Study of patients treated from 1996 to 1999 evaluated the national practice for patients receiving radiation therapy for carcinoma of the esophagus in the United States.
METHODS: A national survey was conducted at 59 institutions in a stratified random sample selected from a master list of radiation therapy facilities throughout the United States. Patient, tumor, and treatment characteristics were evaluated. Multivariate comparisons of survival times were made using the Cox proportional hazards model.
RESULTS: Adenocarcinoma was diagnosed in 51% of patients and squamous cell carcinoma in 49% of patients. Sixteen percent of patients were clinical stage (CS) I (using the 1983 American Joint Committee on Cancer system), 39% were CS II, and 33% were CS III. Significant variables in the multivariate analysis of survival times included clinical stage, treatment approach, and facility size. Patients with CS III disease had a higher hazard risk of death as compared with CS I patients (hazard ratio [HR], 2.01; P = .001), whereas those treated with chemoradiotherapy followed by surgery (HR, 0.32; P < .0001) had a decreased risk of death compared with chemoradiotherapy-only patients. Patients at small centers had a higher risk of death (HR, 1.32; P = .03) compared with patients treated at larger facilities.
CONCLUSION: Concurrent chemoradiotherapy continued to be the most commonly utilized treatment approach during the time period studied. The observation that patients undergoing surgical resection following chemoradiation have a decreased HR or chance of death compared with other treatment schemes supports the need for a randomized trial comparing these strategies.
INTRODUCTION
The Patterns of Care Studies (PCS) are funded by the National Cancer Institute (NCI) to establish benchmarks for diseases managed with radiotherapy.1-3 In 1992, the PCS executive committee added esophageal cancer to its list of diseases studied based on the understanding of the fundamental role that radiation therapy plays in the management of patients diagnosed with this disease. The original PCS esophageal cancer survey examined the evaluation, management, and outcomes of patients treated at radiotherapy centers throughout the United States between 1992 and 1994.4 The results of this analysis helped to document the increasing utilization of concurrent chemoradiotherapy treatment strategies. In addition, the outcome results of this retrospective survey suggested that the long-term outcomes for patients treated at radiotherapy centers nationwide were similar to the results obtained in many of the clinical trials that had helped to set the national standards of practice.5
In 1999, the PCS executive committee reconvened and decided to support a second survey of the national practice processes of care and outcome for esophageal cancer. The esophageal cancer committee reviewed the records of patients treated between 1996 and 1999 throughout the United States to document any changes in national practice since the original survey, and to identify any differences in outcome based on patient factors, tumor characteristics, or treatment approach.
The PCS Esophageal Cancer Committee has previously reported the results of the portion of the 1996-1999 survey designed to examine the evaluation and management of this patient population.6 This study identified a significant rise in the use of endoscopic ultrasounds as part of the pretreatment staging evaluation in comparison to the findings of the original survey. In addition, this survey documented an increase in the use of concurrent chemoradiotherapy before planned surgical resection, and an escalation of the utilization of paclitaxel in combined modality treatment strategies. The purpose of this analysis is to report the outcome of patients treated during this time period, and to identify any possible factors associated with improved survival. These results will be compared to the outcomes of patients treated between 1992 and 1994.
METHODS
A national survey was conducted at 59 institutions in a stratified random sample selected from a master list of radiation therapy facilities located throughout the United States in 1998. Stratified two-step sampling was used to select institutions and cases for data collection.7 Radiation therapy facilities in the master list, 1,450 facilities nationally, were divided into four strata: large academic (main teaching hospital of a medical school or NCI-designated Comprehensive Cancer Center with 500 or more new patients per year), small academic (main teaching hospital of a medical school or NCI-designated Comprehensive Cancer Center with fewer than 500 new patients per year), large nonacademic (nonacademic facility with 500 or more new patients per year), and small nonacademic (nonacademic facility with fewer than 500 new patients per year). Eighty facilities were selected from the four strata and invited to voluntarily participate in the survey, though two of those facilities closed before the survey began. Site visits to collect data took place at 20 large academic centers (71 nationally), two small academic centers (26 nationally), 19 large nonacademic facilities (246 nationally), and 18 small nonacademic facilities (1,107 nationally), for a participation rate of 76%. In the second stage of sampling, cases for each disease site were randomly selected for review and data abstraction, based on lists generated by the facility according to eligibility criteria.
PCS research associates who performed on-site chart reviews at each facility participating in the study collected data from 2000 to 2002 for patients treated from 1996 to 1999. Data items on the survey form were determined by the PCS Esophageal Committee, which met before national site visits in order to identify key evaluation, management, and treatment issues in esophageal cancer. Data collected included the following: demographics, symptoms at presentation, extent of work-up (including study results), staging, treatment (planned and delivered), toxicities encountered, and outcome.
An unweighted sample size (uwss) of 414 patients (weighted sample size [wss], 11,340 patients) with esophageal cancer who received radiotherapy as part of definitive or adjuvant management was identified. To be eligible for participation in this study, patients had to meet several criteria. All cases eligible for review were required to have no evidence of distant metastases at presentation (defined as either palpable supraclavicular or computed tomography evidence of celiac nodes > 1 cm), squamous cell, or adenocarcinoma histology, Karnofsky performance status of more than 60, tumor in the thoracic esophagus with less than 2 cm extension into the stomach, and no prior malignancies within the last 5 years. All patients were staged according to the modified 1983 American Joint Committee on Cancer (AJCC) staging system, and those patients with stage I, II, or III disease were eligible. The decision to use this staging system was based on the reality that the current AJCC staging system is based on pathologic information, and had little utility for a majority of patients undergoing radiotherapy during the time period studied.
National estimates were calculated from the survey data using SUDAAN statistical software (Research Triangle Institute, Research Triangle Park, NC), which incorporates the design elements (the two stages of stratification) and weights that reflect the relative contribution of each institution and each patient in the analysis of this complex survey. The weights were the product of the following two factors:
All figures calculated using SUDAAN are national estimates for the patient population defined by the eligibility criteria listed in the previous section. For these analyses the strata were combined into two strata representing large institutions (more than 500 new patients per year) and small institutions (fewer than 500 new patients per year; see Table 1 for patient breakdown in these strata). The weights were adjusted before analyzing the combined strata.
This analysis was performed using three SUDAAN procedures: one for percentages and tests on categoric variables (PROC CROSSTAB), one for descriptive statistics and tests on continuous variables (PROC DESCRIPT), and one for survival analysis that produced hazard ratios (HR) from fitting a Cox proportional hazards model to the sample data (PROC SURVIVAL). Tests for associations were performed using the Pearson 2 statistic. Differences were deemed significant if their P values were less than .05. A Bonferroni-adjusted P value was used if there were multiple comparisons, in order to control for experimental error that can result from multiple testing. Multiple prognostic factors were evaluated in univariate and multivariate Cox proportional hazards models. Variables with P values of less than .10 in the univariate Cox models were considered statistically significant, and were then included in multivariate Cox models. The significance level of P < .10 was chosen so that the multivariate models would include all possible variables that have a potential effect on our outcome end points. We realize, however, that in using a significance level of P < .10, we increased the chance of including variables in the model that truly have no association with our end points. Differences in outcomes are expressed as HRs. For specific analyses in which the interest was to assess strata (size of radiation therapy facility) as a covariate, the design was necessarily simplified; these analyses did not include stratification as a design factor. To permit valid comparisons between the results of this survey with the 1992-1994 survey, the data from the 1992-1994 Patterns of Care Study was restratified and then analyzed using SUDAAN. The restratification was necessary since the strata for this earlier survey was facility type (academic, research, and nonresearch) not facility size (< 500 new patients per year and 500 new patients per year).
RESULTS
The 1996-1999 Survey
The median age of patients included in this analysis was 64 years (Table 1). Seventy-seven percent were male. A review of the histologic diagnoses revealed adenocarcinoma/adenosquamous in 50.5%, and squamous cell carcinoma in 48.7% (unknown in 0.8%). Eighty-five percent had a Karnofsky performance status 80, and 29% had a 10% weight loss at presentation. Seventy-five percent were white, 21% were African American, 3% were Asian, and 1% were Hispanic (other/unknown, 0.3%). When staging was performed according to the clinically based 1983 AJCC system, 16% were stage I (< 5 cm in length, noncircumferential, nonobstructing), 39% stage II (> 5 cm in length, circumferential, or causing obstruction) and 33% were stage III (clinical evidence of extraesophageal spread). Clinical stage was unable to be determined in 12% of cases.
Ninety-nine percent of the patients had follow-up information. This was the group used for this outcome analysis. The median follow-up was 8 months (range, 1 day to 80 months). At the time of this analysis, 42% of the patients had died (wss, 4,632; uwss, 140) and 50% of these patients had died from esophageal cancer. Locoregional failure was the first end point examined in this outcomes analysis. Thirty-seven percent of patients had a locoregional failure (wss, 4,133; uwss, 141), including patients with persistent disease. A univariate Cox proportional hazards model tested each of the following variables for an association with an increased risk of locoregional failure: age, race, sex, clinical stage, treatment approach, institution size (strata), and histology (Table 2). This analysis revealed that treatment approach and center size were statistically significant predictors of locoregional failure. A multivariate Cox model tested these two covariates to determine if they were still predictive after adjusting for each other (Table 3). This analysis also showed that patients receiving chemoradiotherapy followed by surgical resection had a greater than 50% reduction in the risk of locoregional failure (HR, 0.40; P < .0001) compared with definitive chemoradiotherapy patients (reference level [RL]). In addition, patients treated at small centers had approximately a 40% higher risk of locoregional failure (HR, 1.36; P = .01) than those receiving therapy at large centers (RL).
A similar univariate model incorporated these same variables to examine overall survival (Tables 2 and 4). In this analysis, clinical stage, histology, treatment approach, and facility size were statistically significant predictors of survival. When these four covariates were combined in a multivariate model, histology was no longer a statistically significant predictor of survival after adjusting for the other three covariates. A more intuitive result, however, was revealed with this model. Patients with clinical stage III disease had a significantly higher risk of death (HR, 2.01; P = .001) as compared with stage I patients (RL). The analysis also indicated that patients treated with trimodality therapy were three times more likely to survive their disease (HR, 0.32; P < .0001) than those patients treated with chemoradiotherapy (RL). As previously observed in the analysis for locoregional failure, this multivariate analysis also identified that the size of the treatment center was a significant predictor of overall survival. Patients treated at small centers had a 30% higher hazard ratio of death (HR, 1.32; P = .03) than patients treated at large centers (RL).
Analysis of the 1992-1994 Survey
The outcome results from the original esophageal cancer survey were also analyzed using the SUDAAN statistical software and Cox proportional hazard models were generated with the same variables as used in the above analysis. In the univariate analysis of locoregional failure, race, clinical stage, treatment approach, and facility size were shown to be statistically significant predictors associated with locoregional failure. A multivariate analysis was performed using these variables. Nonwhite patients (HR, 1.65; P < .0001) appeared to have a greater risk of locoregional failure than white patients (RL). Clinical stage III disease patients (HR, 1.66; P < .0001) showed an increased risk for death than patients who were clinically stage I (RL). Patients who were treated with radiation alone or in conjunction with surgery (HR, 1.79; P < .0001) also had a greater risk of experiencing a locoregional failure than patients treated with chemoradiotherapy (RL). Facility size was not statistically significant in this multivariate model.
In the analysis of overall survival, clinical stage and facility size were found to be statistically significant predictors of overall survival on univariate analysis. A multivariate analysis for survival was carried out using clinical stage and facility size (Table 5). Patients with either clinical stage III disease (HR, 2.16; P < .0001) or clinical stage II disease (HR, 1.45; P < .0001) showed an increased risk of death compared with patients who were clinically stage I (RL). Patients treated at small facilities (HR, 0.82; P < .0001) had a lower risk of dying than patients treated at large centers (RL).
Analysis of the Combined Data From Both Surveys
Data from the 1992-1994 survey and the current 1996-1999 survey were combined to evaluate and determine if age, race, sex, clinical stage, histology, treatment approach, and facility size had any affect on locoregional failure or survival for this larger cohort of patients with esophageal cancer. Significant predictors in the univariate models for locoregional failure were race, clinical stage, treatment, and facility size (Table 6). In the multivariate model, which makes adjustments for these covariates, patients receiving trimodality therapy had a 50% lower risk of locoregional failure compared with patients receiving chemoradiotherapy without surgery (HR, 0.55 v RL; P < .0001). In contrast, patients receiving radiation alone or with surgery had a 50% higher risk of failure compared with chemoradiotherapy patients (HR, 1.46 v RL; P = .0001). In addition, nonwhite patients (HR, 1.17; P = .07), clinical stage III disease patients (HR, 1.47; P = .001), and patients treated at small facilities (HR, 1.18; P = .02) all demonstrated an increased hazard ratio of locoregional failure in the combined surveys.
When the end point of survival was evaluated, clinical stage and treatment approach predicted statistically significant differences in the risk of death in the univariate models (Table 6). In the multivariate model (Table 7), patients with stage III disease had more than twice the hazard rate of death compared with patients with stage I (HR, 2.10 v RL; P = .0002), and patients receiving chemoradiotherapy and surgery had a hazard rate of death that was nearly half (HR, 0.52; P = .0001) of the rate for patients treated with chemoradiotherapy (RL).
DISCUSSION
The Patterns of Care Studies were initiated in 1973 as a national effort to improve the quality and accessibility of radiation oncology care in the United States. These surveys are based on a model of quality assessment that incorporates three fundamental components: structure, process, and outcome.3,8 The structure of this survey was defined by the Facilities Master List, which described the size, leadership, and location of all radiation oncology departments in the country.2 The processes of care have measured the evaluation and treatment approaches utilized throughout these centers, which have then been evaluated against the current options for care as defined by the results of randomized trial data. The value of the PCS can be defined by its ability to examine the interrelationship between these two components and outcome.2
For example, this analysis identified treatment approach as an important determinant of locoregional control. The addition of surgical resection following chemotherapy and radiation was associated with a significant decrease in locoregional failure when compared with either definitive chemoradiotherapy or radiotherapy alone. A review of the chemoradiotherapy literature does reveal a local failure risk between 25% and 40%, and it is therefore not surprising to identify improved local control associated with the addition of a planned surgical resection.9-14 While this finding may be intuitive, only a prospective trial would be able to determine whether the improvement in local control associated with surgical resection following chemoradiotherapy would directly result in an overall survival advantage. Multiple randomized trials have been conducted comparing trimodality therapy to surgical resection alone, however, it is important to note that there is little phase III data to date that evaluates the value of surgical resection following concurrent chemotherapy and radiation.15-19
The original PCS outcomes analysis on esophageal cancer identified that patients treated at nonacademic centers had more favorable outcomes than those patients treated at centers identified as either NCI-designated comprehensive cancer centers or major teaching hospitals associated with US medical schools.5 An evaluation of this finding led the PCS executive committee to refocus this portion of this analysis to determine whether these differences were driven by volume of new cancer cases seen each year, rather than type of academic affiliation. This decision was, in part, based on the fact that no significant differences were seen in the performance status, age, or stage of disease of patients seen at the two types of centers. With the assumption that patient volumes are reflective of clinical experience and quality of support, the current analysis lends support to the contention that the size of the treatment center has a more direct bearing on outcome than the type of academic affiliation. This finding is important to recognize in light of the fact that no significant differences were noted in the type of therapy delivered between large and small institutions. The influence of patient volumes on outcomes has been previously reported for various other malignancies including lung, pancreas,and rectal cancer.20-23 This PCS outcomes analysis is the first report to note a difference in survival for patients with esophageal cancer based on this factor.
Both the current (1996-1999) and original (1992-1994) surveys identified the stage of disease as important predictors of overall survival. The stages were assigned based on the 1983 AJCC clinical staging system, since a majority of patients had not undergone either surgical resection or endoscopic ultrasounds to determine depth of invasion. These data confirmed the utility of the clinically based staging system. These findings, coupled with the fact that few patients had the depth of penetration determined before therapy, highlights the need to reconsider incorporating elements of the clinically based system into future staging algorithms.6 While stage of disease was an important determinant of survival for both time periods studied, this analysis failed to identify any stage-specific improvements in survival when the 1996-1999 survey results were compared with the 1992-1994 survey.
Perhaps the most intriguing finding of this analysis is the improved survival associated with surgical resection following chemoradiotherapy. This result mirrors the outcomes analysis of the original survey; however, it must be interpreted with great caution. The PCS surveys are retrospective analyses of patterns of practice throughout the United States. There are obvious inherent biases built into any retrospective analysis, and the outcome results cannot be considered equivalent to prospective randomized trial data. Patient selection factors and physician preferences greatly influence treatment approaches, and these variables were largely unaccounted for in this analysis.
However, these results do provide some very valuable insights that can lead to future clinical trial design. The phase III randomized trials that have evaluated the role of trimodality therapy in this disease have yet to identify a clear and reproducible benefit of trimodality therapy compared with surgery alone.16,17,19,24 Despite this fact, it is clear from both the original and current PCS surveys, trimodality therapy is increasing in popularity across the country. The data from these surveys emphasized the need to consider a national trial designed to compare definitive chemoradiotherapy with chemoradiotherapy followed by surgical resection in order to appropriately determine a standard of care for the management of these patients.
The Patterns of Care esophageal cancer data adds to the body of literature that supports the essential role of radiotherapy in the management of this disease. At the same time, it also serves to highlight the debate regarding several key issues such as the appropriate role of surgical resection, optimal chemotherapy combinations, and the value of a clinically based staging system. Whereas this PCS analysis suggests that combined modality approaches have led to improved outcomes, perhaps most importantly, this data emphasizes the need to conduct well-designed clinical trials aimed at addressing the questions that still remained unanswered.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank the principal investigators during this survey, Gerald E. Hanks, MD, and J. Frank Wilson, MD. We also thank radiation oncologists and staff at participating facilities for their support and cooperation, which is essential to the PCS.
We thank PCS research associates, Thomas Iarocci, MD, MS, Debora Grant, RN, MSN, Ann Mann, MHSA, RT(T), Andrea Trent, RN, MSN, and Kimberly Roy, RN, BSN, for collecting the survey data, Lisa Morabito for administrative support, and Dolores Malone for database design and administration.
NOTES
Supported by grant No. CA 65435 from the National Cancer Institute, National Institutes of Health.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Kramer S, Hanks GE, Diamond JJ: Summary results from the fourth Facilities Master List survey conducted by the Patterns of Care Study. Int J Radiat Oncol Biol Phys 9:1881-1883, 1983
Owen JB, Coia LR, Hanks GE: The structure of radiation oncology in the United States in 1994. Int J Radiat Oncol Biol Phys 39:179-185, 1997
Hanks GE, Coia LR, Curry J: Patterns of Care Studies: Past, Present, and Future. Semin Radiat Oncol 7:97-100, 1997
Coia LR, Minsky BD, John MJ, et al: The evaluation and treatment of patients receiving radiation therapy for carcinoma of the esophagus: Results of the 1992-1994 Patterns of Care Study. Cancer 85:2499-2505, 1999
Coia LR, Minsky BD, Berkey BA, et al: Outcome of patients receiving radiation for cancer of the esophagus: Results of the 1992-1994 patterns of care study. J Clin Oncol 18:455-462, 2000
Suntharalingam M, Moughan J, Coia LR, et al: The national practice for patients receiving radiation therapy for carcinoma of the esophagus: Results of the 1996-1999 Patterns of Care Study. Int J Radiat Oncol Biol Phys 56:981-987, 2003
Owen JB, Sedransk J, Pajak TF: National averages for process and outcome in radiation oncology: Methodology of the Patterns of Care Study. Semin Radiat Oncol 7:101-107, 1997
Kramer S, Hanks GE, Herring DF, et al: Summary results from the Facilities Master List surveys conducted by the Patterns of Care Study. Int J Radiat Oncol Biol Phys 8:883-888, 1982
Bates BA, Detterbeck FC, Bernard SA, et al: Concurrent radiation therapy and chemotherapy followed by esophagectomy for localized esophageal carcinoma. J Clin Oncol 14:156-163, 1996
Coia L, Engstrom P, Paul A: Long-term results of infusional 5-FU mitomycin-C and radiation as primary management of esophageal cancer. Int J Radiat Oncol Biol Phys 20:29-36, 1991
al-Sarraf M, Martz K, Herskovic A, et al: Progress report of combined chemoradiotherapy versus radiotherapy alone in patients with esophageal cancer: An Intergroup study. J Clin Oncol 15:277-284, 1997 [Erratum: J Clin Oncol 15:866, 1997]
Cooper JS, Guo MD, Herskovic A, et al: Chemoradiotherapy of locally advanced esophageal cancer: Long-term follow-up of a prospective randomized trial (RTOG 85-01)—Radiation Therapy Oncology Group. JAMA 281:1623-1627, 1999
Forastiere AA, Urba SG: Combined Modality Therapy for Cancer of the Esophagus, in DeVita VT, Hellman S, Rosenberg SA (eds): Principles and Practices of Oncology (ed 4). Philadelphia, PA, Lippincott-Raven, 1996, pp 1-15
Herskovic A, Martz K, al-Sarraf M, et al: Combined chemotherapy and radiotherapy compared with radiotherapy alone in patients with cancer of the esophagus. N Engl J Med 326:1593-1598, 1992
Forastiere AA, Orringer MB, Perez-Tamayo C, et al: Preoperative chemoradiation followed by transhiatal esophagectomy for carcinoma of the esophagus: Final report. J Clin Oncol 11:1118-1123, 1993
Bosset JF, Gignoux M, Triboulet JP, et al: Chemoradiotherapy followed by surgery compared with surgery alone in squamous-cell cancer of the esophagus. N Engl J Med 337:161-167, 1997
Urba SG, Orringer MB, Turrisi A, et al: Randomized trial of preoperative chemoradiation versus surgery alone in patients with locoregional esophageal carcinoma. J Clin Oncol 19:305-313, 2001
Urschel JD, Vasan H, Blewett CJ: A meta-analysis of randomized controlled trials that compared neoadjuvant chemotherapy and surgery to surgery alone for resectable esophageal cancer. Am J Surg 183:274-279, 2002
Walsh TN, Noonan N, Hollywood D, et al: A comparison of multimodal therapy and surgery for esophageal adenocarcinoma. N Engl J Med 335:462-467, 1996 [Erratum: N Engl J Med 341:384, 1999]
Begg CB, Cramer LD, Hoskins WJ, et al: Impact of hospital volume on operative mortality for major cancer surgery. JAMA 280:1747-1751, 1998
Begg CB, Scardino PT: Taking stock of volume-outcome studies. J Clin Oncol 21:393-394, 2003
Lieberman MD, Kilburn H, Lindsey M, et al: Relation of perioperative deaths to hospital volume among patients undergoing pancreatic resection for malignancy. Ann Surg 222:638-645, 1995
Lee JS, Scott CB, Komaki R, et al: Impact of institutional experience on survival outcome of patients undergoing combined chemoradiation therapy for inoperable non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 52:362-370, 2002
Le Prise E, Etienne PL, Meunier B, et al: A randomized study of chemotherapy, radiation therapy, and surgery versus surgery for localized squamous cell carcinoma of the esophagus. Cancer 73:1779-1784, 1994(Mohan Suntharalingam, Jen)
American College of Radiology
Abramson Cancer Center at the University of Pennsylvania, Philadelphia, PA
Community Medical Center, Toms River, NJ
Boston University Medical Center
Massachusetts General Hospital, Boston, MA
Cancer Center at St Agnes, Fresno, CA
Memorial Sloan-Kettering Cancer Center, New York, NY.
ABSTRACT
PURPOSE: A Patterns of Care Study of patients treated from 1996 to 1999 evaluated the national practice for patients receiving radiation therapy for carcinoma of the esophagus in the United States.
METHODS: A national survey was conducted at 59 institutions in a stratified random sample selected from a master list of radiation therapy facilities throughout the United States. Patient, tumor, and treatment characteristics were evaluated. Multivariate comparisons of survival times were made using the Cox proportional hazards model.
RESULTS: Adenocarcinoma was diagnosed in 51% of patients and squamous cell carcinoma in 49% of patients. Sixteen percent of patients were clinical stage (CS) I (using the 1983 American Joint Committee on Cancer system), 39% were CS II, and 33% were CS III. Significant variables in the multivariate analysis of survival times included clinical stage, treatment approach, and facility size. Patients with CS III disease had a higher hazard risk of death as compared with CS I patients (hazard ratio [HR], 2.01; P = .001), whereas those treated with chemoradiotherapy followed by surgery (HR, 0.32; P < .0001) had a decreased risk of death compared with chemoradiotherapy-only patients. Patients at small centers had a higher risk of death (HR, 1.32; P = .03) compared with patients treated at larger facilities.
CONCLUSION: Concurrent chemoradiotherapy continued to be the most commonly utilized treatment approach during the time period studied. The observation that patients undergoing surgical resection following chemoradiation have a decreased HR or chance of death compared with other treatment schemes supports the need for a randomized trial comparing these strategies.
INTRODUCTION
The Patterns of Care Studies (PCS) are funded by the National Cancer Institute (NCI) to establish benchmarks for diseases managed with radiotherapy.1-3 In 1992, the PCS executive committee added esophageal cancer to its list of diseases studied based on the understanding of the fundamental role that radiation therapy plays in the management of patients diagnosed with this disease. The original PCS esophageal cancer survey examined the evaluation, management, and outcomes of patients treated at radiotherapy centers throughout the United States between 1992 and 1994.4 The results of this analysis helped to document the increasing utilization of concurrent chemoradiotherapy treatment strategies. In addition, the outcome results of this retrospective survey suggested that the long-term outcomes for patients treated at radiotherapy centers nationwide were similar to the results obtained in many of the clinical trials that had helped to set the national standards of practice.5
In 1999, the PCS executive committee reconvened and decided to support a second survey of the national practice processes of care and outcome for esophageal cancer. The esophageal cancer committee reviewed the records of patients treated between 1996 and 1999 throughout the United States to document any changes in national practice since the original survey, and to identify any differences in outcome based on patient factors, tumor characteristics, or treatment approach.
The PCS Esophageal Cancer Committee has previously reported the results of the portion of the 1996-1999 survey designed to examine the evaluation and management of this patient population.6 This study identified a significant rise in the use of endoscopic ultrasounds as part of the pretreatment staging evaluation in comparison to the findings of the original survey. In addition, this survey documented an increase in the use of concurrent chemoradiotherapy before planned surgical resection, and an escalation of the utilization of paclitaxel in combined modality treatment strategies. The purpose of this analysis is to report the outcome of patients treated during this time period, and to identify any possible factors associated with improved survival. These results will be compared to the outcomes of patients treated between 1992 and 1994.
METHODS
A national survey was conducted at 59 institutions in a stratified random sample selected from a master list of radiation therapy facilities located throughout the United States in 1998. Stratified two-step sampling was used to select institutions and cases for data collection.7 Radiation therapy facilities in the master list, 1,450 facilities nationally, were divided into four strata: large academic (main teaching hospital of a medical school or NCI-designated Comprehensive Cancer Center with 500 or more new patients per year), small academic (main teaching hospital of a medical school or NCI-designated Comprehensive Cancer Center with fewer than 500 new patients per year), large nonacademic (nonacademic facility with 500 or more new patients per year), and small nonacademic (nonacademic facility with fewer than 500 new patients per year). Eighty facilities were selected from the four strata and invited to voluntarily participate in the survey, though two of those facilities closed before the survey began. Site visits to collect data took place at 20 large academic centers (71 nationally), two small academic centers (26 nationally), 19 large nonacademic facilities (246 nationally), and 18 small nonacademic facilities (1,107 nationally), for a participation rate of 76%. In the second stage of sampling, cases for each disease site were randomly selected for review and data abstraction, based on lists generated by the facility according to eligibility criteria.
PCS research associates who performed on-site chart reviews at each facility participating in the study collected data from 2000 to 2002 for patients treated from 1996 to 1999. Data items on the survey form were determined by the PCS Esophageal Committee, which met before national site visits in order to identify key evaluation, management, and treatment issues in esophageal cancer. Data collected included the following: demographics, symptoms at presentation, extent of work-up (including study results), staging, treatment (planned and delivered), toxicities encountered, and outcome.
An unweighted sample size (uwss) of 414 patients (weighted sample size [wss], 11,340 patients) with esophageal cancer who received radiotherapy as part of definitive or adjuvant management was identified. To be eligible for participation in this study, patients had to meet several criteria. All cases eligible for review were required to have no evidence of distant metastases at presentation (defined as either palpable supraclavicular or computed tomography evidence of celiac nodes > 1 cm), squamous cell, or adenocarcinoma histology, Karnofsky performance status of more than 60, tumor in the thoracic esophagus with less than 2 cm extension into the stomach, and no prior malignancies within the last 5 years. All patients were staged according to the modified 1983 American Joint Committee on Cancer (AJCC) staging system, and those patients with stage I, II, or III disease were eligible. The decision to use this staging system was based on the reality that the current AJCC staging system is based on pathologic information, and had little utility for a majority of patients undergoing radiotherapy during the time period studied.
National estimates were calculated from the survey data using SUDAAN statistical software (Research Triangle Institute, Research Triangle Park, NC), which incorporates the design elements (the two stages of stratification) and weights that reflect the relative contribution of each institution and each patient in the analysis of this complex survey. The weights were the product of the following two factors:
All figures calculated using SUDAAN are national estimates for the patient population defined by the eligibility criteria listed in the previous section. For these analyses the strata were combined into two strata representing large institutions (more than 500 new patients per year) and small institutions (fewer than 500 new patients per year; see Table 1 for patient breakdown in these strata). The weights were adjusted before analyzing the combined strata.
This analysis was performed using three SUDAAN procedures: one for percentages and tests on categoric variables (PROC CROSSTAB), one for descriptive statistics and tests on continuous variables (PROC DESCRIPT), and one for survival analysis that produced hazard ratios (HR) from fitting a Cox proportional hazards model to the sample data (PROC SURVIVAL). Tests for associations were performed using the Pearson 2 statistic. Differences were deemed significant if their P values were less than .05. A Bonferroni-adjusted P value was used if there were multiple comparisons, in order to control for experimental error that can result from multiple testing. Multiple prognostic factors were evaluated in univariate and multivariate Cox proportional hazards models. Variables with P values of less than .10 in the univariate Cox models were considered statistically significant, and were then included in multivariate Cox models. The significance level of P < .10 was chosen so that the multivariate models would include all possible variables that have a potential effect on our outcome end points. We realize, however, that in using a significance level of P < .10, we increased the chance of including variables in the model that truly have no association with our end points. Differences in outcomes are expressed as HRs. For specific analyses in which the interest was to assess strata (size of radiation therapy facility) as a covariate, the design was necessarily simplified; these analyses did not include stratification as a design factor. To permit valid comparisons between the results of this survey with the 1992-1994 survey, the data from the 1992-1994 Patterns of Care Study was restratified and then analyzed using SUDAAN. The restratification was necessary since the strata for this earlier survey was facility type (academic, research, and nonresearch) not facility size (< 500 new patients per year and 500 new patients per year).
RESULTS
The 1996-1999 Survey
The median age of patients included in this analysis was 64 years (Table 1). Seventy-seven percent were male. A review of the histologic diagnoses revealed adenocarcinoma/adenosquamous in 50.5%, and squamous cell carcinoma in 48.7% (unknown in 0.8%). Eighty-five percent had a Karnofsky performance status 80, and 29% had a 10% weight loss at presentation. Seventy-five percent were white, 21% were African American, 3% were Asian, and 1% were Hispanic (other/unknown, 0.3%). When staging was performed according to the clinically based 1983 AJCC system, 16% were stage I (< 5 cm in length, noncircumferential, nonobstructing), 39% stage II (> 5 cm in length, circumferential, or causing obstruction) and 33% were stage III (clinical evidence of extraesophageal spread). Clinical stage was unable to be determined in 12% of cases.
Ninety-nine percent of the patients had follow-up information. This was the group used for this outcome analysis. The median follow-up was 8 months (range, 1 day to 80 months). At the time of this analysis, 42% of the patients had died (wss, 4,632; uwss, 140) and 50% of these patients had died from esophageal cancer. Locoregional failure was the first end point examined in this outcomes analysis. Thirty-seven percent of patients had a locoregional failure (wss, 4,133; uwss, 141), including patients with persistent disease. A univariate Cox proportional hazards model tested each of the following variables for an association with an increased risk of locoregional failure: age, race, sex, clinical stage, treatment approach, institution size (strata), and histology (Table 2). This analysis revealed that treatment approach and center size were statistically significant predictors of locoregional failure. A multivariate Cox model tested these two covariates to determine if they were still predictive after adjusting for each other (Table 3). This analysis also showed that patients receiving chemoradiotherapy followed by surgical resection had a greater than 50% reduction in the risk of locoregional failure (HR, 0.40; P < .0001) compared with definitive chemoradiotherapy patients (reference level [RL]). In addition, patients treated at small centers had approximately a 40% higher risk of locoregional failure (HR, 1.36; P = .01) than those receiving therapy at large centers (RL).
A similar univariate model incorporated these same variables to examine overall survival (Tables 2 and 4). In this analysis, clinical stage, histology, treatment approach, and facility size were statistically significant predictors of survival. When these four covariates were combined in a multivariate model, histology was no longer a statistically significant predictor of survival after adjusting for the other three covariates. A more intuitive result, however, was revealed with this model. Patients with clinical stage III disease had a significantly higher risk of death (HR, 2.01; P = .001) as compared with stage I patients (RL). The analysis also indicated that patients treated with trimodality therapy were three times more likely to survive their disease (HR, 0.32; P < .0001) than those patients treated with chemoradiotherapy (RL). As previously observed in the analysis for locoregional failure, this multivariate analysis also identified that the size of the treatment center was a significant predictor of overall survival. Patients treated at small centers had a 30% higher hazard ratio of death (HR, 1.32; P = .03) than patients treated at large centers (RL).
Analysis of the 1992-1994 Survey
The outcome results from the original esophageal cancer survey were also analyzed using the SUDAAN statistical software and Cox proportional hazard models were generated with the same variables as used in the above analysis. In the univariate analysis of locoregional failure, race, clinical stage, treatment approach, and facility size were shown to be statistically significant predictors associated with locoregional failure. A multivariate analysis was performed using these variables. Nonwhite patients (HR, 1.65; P < .0001) appeared to have a greater risk of locoregional failure than white patients (RL). Clinical stage III disease patients (HR, 1.66; P < .0001) showed an increased risk for death than patients who were clinically stage I (RL). Patients who were treated with radiation alone or in conjunction with surgery (HR, 1.79; P < .0001) also had a greater risk of experiencing a locoregional failure than patients treated with chemoradiotherapy (RL). Facility size was not statistically significant in this multivariate model.
In the analysis of overall survival, clinical stage and facility size were found to be statistically significant predictors of overall survival on univariate analysis. A multivariate analysis for survival was carried out using clinical stage and facility size (Table 5). Patients with either clinical stage III disease (HR, 2.16; P < .0001) or clinical stage II disease (HR, 1.45; P < .0001) showed an increased risk of death compared with patients who were clinically stage I (RL). Patients treated at small facilities (HR, 0.82; P < .0001) had a lower risk of dying than patients treated at large centers (RL).
Analysis of the Combined Data From Both Surveys
Data from the 1992-1994 survey and the current 1996-1999 survey were combined to evaluate and determine if age, race, sex, clinical stage, histology, treatment approach, and facility size had any affect on locoregional failure or survival for this larger cohort of patients with esophageal cancer. Significant predictors in the univariate models for locoregional failure were race, clinical stage, treatment, and facility size (Table 6). In the multivariate model, which makes adjustments for these covariates, patients receiving trimodality therapy had a 50% lower risk of locoregional failure compared with patients receiving chemoradiotherapy without surgery (HR, 0.55 v RL; P < .0001). In contrast, patients receiving radiation alone or with surgery had a 50% higher risk of failure compared with chemoradiotherapy patients (HR, 1.46 v RL; P = .0001). In addition, nonwhite patients (HR, 1.17; P = .07), clinical stage III disease patients (HR, 1.47; P = .001), and patients treated at small facilities (HR, 1.18; P = .02) all demonstrated an increased hazard ratio of locoregional failure in the combined surveys.
When the end point of survival was evaluated, clinical stage and treatment approach predicted statistically significant differences in the risk of death in the univariate models (Table 6). In the multivariate model (Table 7), patients with stage III disease had more than twice the hazard rate of death compared with patients with stage I (HR, 2.10 v RL; P = .0002), and patients receiving chemoradiotherapy and surgery had a hazard rate of death that was nearly half (HR, 0.52; P = .0001) of the rate for patients treated with chemoradiotherapy (RL).
DISCUSSION
The Patterns of Care Studies were initiated in 1973 as a national effort to improve the quality and accessibility of radiation oncology care in the United States. These surveys are based on a model of quality assessment that incorporates three fundamental components: structure, process, and outcome.3,8 The structure of this survey was defined by the Facilities Master List, which described the size, leadership, and location of all radiation oncology departments in the country.2 The processes of care have measured the evaluation and treatment approaches utilized throughout these centers, which have then been evaluated against the current options for care as defined by the results of randomized trial data. The value of the PCS can be defined by its ability to examine the interrelationship between these two components and outcome.2
For example, this analysis identified treatment approach as an important determinant of locoregional control. The addition of surgical resection following chemotherapy and radiation was associated with a significant decrease in locoregional failure when compared with either definitive chemoradiotherapy or radiotherapy alone. A review of the chemoradiotherapy literature does reveal a local failure risk between 25% and 40%, and it is therefore not surprising to identify improved local control associated with the addition of a planned surgical resection.9-14 While this finding may be intuitive, only a prospective trial would be able to determine whether the improvement in local control associated with surgical resection following chemoradiotherapy would directly result in an overall survival advantage. Multiple randomized trials have been conducted comparing trimodality therapy to surgical resection alone, however, it is important to note that there is little phase III data to date that evaluates the value of surgical resection following concurrent chemotherapy and radiation.15-19
The original PCS outcomes analysis on esophageal cancer identified that patients treated at nonacademic centers had more favorable outcomes than those patients treated at centers identified as either NCI-designated comprehensive cancer centers or major teaching hospitals associated with US medical schools.5 An evaluation of this finding led the PCS executive committee to refocus this portion of this analysis to determine whether these differences were driven by volume of new cancer cases seen each year, rather than type of academic affiliation. This decision was, in part, based on the fact that no significant differences were seen in the performance status, age, or stage of disease of patients seen at the two types of centers. With the assumption that patient volumes are reflective of clinical experience and quality of support, the current analysis lends support to the contention that the size of the treatment center has a more direct bearing on outcome than the type of academic affiliation. This finding is important to recognize in light of the fact that no significant differences were noted in the type of therapy delivered between large and small institutions. The influence of patient volumes on outcomes has been previously reported for various other malignancies including lung, pancreas,and rectal cancer.20-23 This PCS outcomes analysis is the first report to note a difference in survival for patients with esophageal cancer based on this factor.
Both the current (1996-1999) and original (1992-1994) surveys identified the stage of disease as important predictors of overall survival. The stages were assigned based on the 1983 AJCC clinical staging system, since a majority of patients had not undergone either surgical resection or endoscopic ultrasounds to determine depth of invasion. These data confirmed the utility of the clinically based staging system. These findings, coupled with the fact that few patients had the depth of penetration determined before therapy, highlights the need to reconsider incorporating elements of the clinically based system into future staging algorithms.6 While stage of disease was an important determinant of survival for both time periods studied, this analysis failed to identify any stage-specific improvements in survival when the 1996-1999 survey results were compared with the 1992-1994 survey.
Perhaps the most intriguing finding of this analysis is the improved survival associated with surgical resection following chemoradiotherapy. This result mirrors the outcomes analysis of the original survey; however, it must be interpreted with great caution. The PCS surveys are retrospective analyses of patterns of practice throughout the United States. There are obvious inherent biases built into any retrospective analysis, and the outcome results cannot be considered equivalent to prospective randomized trial data. Patient selection factors and physician preferences greatly influence treatment approaches, and these variables were largely unaccounted for in this analysis.
However, these results do provide some very valuable insights that can lead to future clinical trial design. The phase III randomized trials that have evaluated the role of trimodality therapy in this disease have yet to identify a clear and reproducible benefit of trimodality therapy compared with surgery alone.16,17,19,24 Despite this fact, it is clear from both the original and current PCS surveys, trimodality therapy is increasing in popularity across the country. The data from these surveys emphasized the need to consider a national trial designed to compare definitive chemoradiotherapy with chemoradiotherapy followed by surgical resection in order to appropriately determine a standard of care for the management of these patients.
The Patterns of Care esophageal cancer data adds to the body of literature that supports the essential role of radiotherapy in the management of this disease. At the same time, it also serves to highlight the debate regarding several key issues such as the appropriate role of surgical resection, optimal chemotherapy combinations, and the value of a clinically based staging system. Whereas this PCS analysis suggests that combined modality approaches have led to improved outcomes, perhaps most importantly, this data emphasizes the need to conduct well-designed clinical trials aimed at addressing the questions that still remained unanswered.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank the principal investigators during this survey, Gerald E. Hanks, MD, and J. Frank Wilson, MD. We also thank radiation oncologists and staff at participating facilities for their support and cooperation, which is essential to the PCS.
We thank PCS research associates, Thomas Iarocci, MD, MS, Debora Grant, RN, MSN, Ann Mann, MHSA, RT(T), Andrea Trent, RN, MSN, and Kimberly Roy, RN, BSN, for collecting the survey data, Lisa Morabito for administrative support, and Dolores Malone for database design and administration.
NOTES
Supported by grant No. CA 65435 from the National Cancer Institute, National Institutes of Health.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Kramer S, Hanks GE, Diamond JJ: Summary results from the fourth Facilities Master List survey conducted by the Patterns of Care Study. Int J Radiat Oncol Biol Phys 9:1881-1883, 1983
Owen JB, Coia LR, Hanks GE: The structure of radiation oncology in the United States in 1994. Int J Radiat Oncol Biol Phys 39:179-185, 1997
Hanks GE, Coia LR, Curry J: Patterns of Care Studies: Past, Present, and Future. Semin Radiat Oncol 7:97-100, 1997
Coia LR, Minsky BD, John MJ, et al: The evaluation and treatment of patients receiving radiation therapy for carcinoma of the esophagus: Results of the 1992-1994 Patterns of Care Study. Cancer 85:2499-2505, 1999
Coia LR, Minsky BD, Berkey BA, et al: Outcome of patients receiving radiation for cancer of the esophagus: Results of the 1992-1994 patterns of care study. J Clin Oncol 18:455-462, 2000
Suntharalingam M, Moughan J, Coia LR, et al: The national practice for patients receiving radiation therapy for carcinoma of the esophagus: Results of the 1996-1999 Patterns of Care Study. Int J Radiat Oncol Biol Phys 56:981-987, 2003
Owen JB, Sedransk J, Pajak TF: National averages for process and outcome in radiation oncology: Methodology of the Patterns of Care Study. Semin Radiat Oncol 7:101-107, 1997
Kramer S, Hanks GE, Herring DF, et al: Summary results from the Facilities Master List surveys conducted by the Patterns of Care Study. Int J Radiat Oncol Biol Phys 8:883-888, 1982
Bates BA, Detterbeck FC, Bernard SA, et al: Concurrent radiation therapy and chemotherapy followed by esophagectomy for localized esophageal carcinoma. J Clin Oncol 14:156-163, 1996
Coia L, Engstrom P, Paul A: Long-term results of infusional 5-FU mitomycin-C and radiation as primary management of esophageal cancer. Int J Radiat Oncol Biol Phys 20:29-36, 1991
al-Sarraf M, Martz K, Herskovic A, et al: Progress report of combined chemoradiotherapy versus radiotherapy alone in patients with esophageal cancer: An Intergroup study. J Clin Oncol 15:277-284, 1997 [Erratum: J Clin Oncol 15:866, 1997]
Cooper JS, Guo MD, Herskovic A, et al: Chemoradiotherapy of locally advanced esophageal cancer: Long-term follow-up of a prospective randomized trial (RTOG 85-01)—Radiation Therapy Oncology Group. JAMA 281:1623-1627, 1999
Forastiere AA, Urba SG: Combined Modality Therapy for Cancer of the Esophagus, in DeVita VT, Hellman S, Rosenberg SA (eds): Principles and Practices of Oncology (ed 4). Philadelphia, PA, Lippincott-Raven, 1996, pp 1-15
Herskovic A, Martz K, al-Sarraf M, et al: Combined chemotherapy and radiotherapy compared with radiotherapy alone in patients with cancer of the esophagus. N Engl J Med 326:1593-1598, 1992
Forastiere AA, Orringer MB, Perez-Tamayo C, et al: Preoperative chemoradiation followed by transhiatal esophagectomy for carcinoma of the esophagus: Final report. J Clin Oncol 11:1118-1123, 1993
Bosset JF, Gignoux M, Triboulet JP, et al: Chemoradiotherapy followed by surgery compared with surgery alone in squamous-cell cancer of the esophagus. N Engl J Med 337:161-167, 1997
Urba SG, Orringer MB, Turrisi A, et al: Randomized trial of preoperative chemoradiation versus surgery alone in patients with locoregional esophageal carcinoma. J Clin Oncol 19:305-313, 2001
Urschel JD, Vasan H, Blewett CJ: A meta-analysis of randomized controlled trials that compared neoadjuvant chemotherapy and surgery to surgery alone for resectable esophageal cancer. Am J Surg 183:274-279, 2002
Walsh TN, Noonan N, Hollywood D, et al: A comparison of multimodal therapy and surgery for esophageal adenocarcinoma. N Engl J Med 335:462-467, 1996 [Erratum: N Engl J Med 341:384, 1999]
Begg CB, Cramer LD, Hoskins WJ, et al: Impact of hospital volume on operative mortality for major cancer surgery. JAMA 280:1747-1751, 1998
Begg CB, Scardino PT: Taking stock of volume-outcome studies. J Clin Oncol 21:393-394, 2003
Lieberman MD, Kilburn H, Lindsey M, et al: Relation of perioperative deaths to hospital volume among patients undergoing pancreatic resection for malignancy. Ann Surg 222:638-645, 1995
Lee JS, Scott CB, Komaki R, et al: Impact of institutional experience on survival outcome of patients undergoing combined chemoradiation therapy for inoperable non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 52:362-370, 2002
Le Prise E, Etienne PL, Meunier B, et al: A randomized study of chemotherapy, radiation therapy, and surgery versus surgery for localized squamous cell carcinoma of the esophagus. Cancer 73:1779-1784, 1994(Mohan Suntharalingam, Jen)