This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Rights & Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Steyerberg, E. W. Articles by Earle, C. C. Search for Related Content PubMed PubMed Citation Articles by Steyerberg, E. W. Articles by Earle, C. C. Social Bookmarking                            What's this?

Surgical Mortality in Patients With Esophageal Cancer: Development and Validation of a Simple Risk Score

Ewout W. Steyerberg, Bridget A. Neville, Linetta B. Koppert, Valery E.P.P. Lemmens, Hugo W. Tilanus, Jan-Willem W. Coebergh, Jane C. Weeks, Craig C. Earle

From the Departments of Public Health and Surgery, Erasmus MC, University Medical Center Rotterdam, Rotterdam; Eindhoven Cancer Registry, Comprehensive Cancer Center South, Eindhoven, the Netherlands; and the Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA

Address reprint requests to Ewout W. Steyerberg, PhD, Department of Public Health, AE-236, Erasmus MC, University Medical Center Rotterdam, PO Box 2040, Rotterdam, the Netherlands 3000 CA; e-mail: e.steyerberg{at}erasmusmc.nl

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
Purpose Surgery has curative potential in a proportion of patients with esophageal cancer, but is associated with considerable perioperative risks. We aimed to develop and validate a simple risk score for surgical mortality that could be applied to administrative data.

Patients and Methods We analyzed 3,592 esophagectomy patients from four cohorts. We applied logistic regression analysis to predict mortality occurring within 30 days after esophagectomy for 1,327 esophageal cancer patients older than 65 years of age, diagnosed between 1991 and 1996 in the linked Surveillance, Epidemiology and End Results (SEER) - Medicare database. A simple score chart for preoperative risk assessment of surgical mortality was developed and validated on three other cohorts, including 714 SEER-Medicare patients diagnosed between 1997 and 1999, 349 patients from a population-based registry in the Netherlands diagnosed between 1993 and 2001, and 1,202 patients from a referral hospital in the Netherlands diagnosed between 1980 and 2002.

Results Surgical mortality in the four cohorts was 11% (147 of 1,327), 10% (74 of 714), 7% (25 of 349), and 4% (45 of 1,202), respectively. Predictive patient characteristics included age, comorbidity (cardiac, pulmonary, renal, hepatic, and diabetes), preoperative radiotherapy or combined chemoradiotherapy, and a relatively low hospital volume. At validation, the simple score showed good agreement of predicted risks with observed mortality rates (calibration), but low discrimination (area under the receiver operating characteristic curve 0.58 to 0.66).

Conclusion A simple risk score combining clinical characteristics along with hospital volume to predict surgical mortality after esophagectomy from administrative data may form a basis for risk adjustment in quality of care assessment.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
Surgical resection offers a chance of long-term survival in patients with esophageal cancer.¹ However, even after careful staging, survival remains disappointing with less than 25% of patients surviving at 5 years after esophagectomy.² Better results may be achieved with the concomitant use of preoperative (neoadjuvant) chemotherapy and radiotherapy, although the benefits may be small.^3-5 The surgical risk in the short-term and the potential loss in quality of life have to be weighed against the long-term benefit, such as a longer survival.^6,7

Accurate prediction of surgical mortality is important not only for appropriate selection of candidates for esophagectomy,⁸ but also for evaluation of quality of care and policy decisions. Risk adjustment is particularly necessary when comparing surgical mortality rates between institutions.^9-11 It is well known that the short-term surgical risk of esophagectomy varies by clinical characteristics, such as age¹² and presence of concomitant diseases (comorbidity).^8,13 Further, esophagectomy is among the procedures where physician and hospital characteristics, especially volume, have been found to be strongly related to the surgical outcome.^14-18 Hence, patients at higher risk may most appropriately undergo surgery at high volume centers.^18-20

Patient characteristics have been combined in multivariable prognostic models for short-term mortality after esophagectomy. However, these models were often based on selected patient groups in specialized centers^8,21 thus limiting the generalizability of the results. Furthermore, validation on new patients was often not performed or showed unsatisfactory results.²² The aim of this study was to develop a simple and robust prediction model for surgical mortality in esophageal cancer patients. We first analyzed several previously identified predictive characteristics in a large population-based cohort, then developed a simple risk score, and finally validated this score in three other cohorts.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
We analyzed four cohorts: two population-based series of 1,327 and 714 patients from the linked Surveillance, Epidemiology and End Results (SEER) - Medicare database,²³ another population-based series of 349 patients from the Netherlands (Eindhoven),²⁴ and 1,202 patients from a referral hospital in the Netherlands (Rotterdam).^25,26 The larger SEER-Medicare cohort served as the model development set and the other three as validation sets.

Patients: SEER-Medicare
The SEER database is made up of 11 tumor registries covering approximately 14% of the United States population. It has been linked to the Centers for Medicare and Medicaid services Medicare database through the end of 2001. We identified two sets of patients diagnosed with pathologically confirmed esophageal cancer. The first set (SEER 91-96) included patients diagnosed between January 1, 1991 and December 31, 1996. It was used for development of the prediction model. The second set included patients diagnosed between January 1, 1997 and December 31, 1999 (SEER 97-99) for validation of the developed model. The selection criteria and definitions of variables were identical in both sets. We excluded patients for whom the date of death differed by more than 3 months between the SEER and Medicare database, patients who were diagnosed from death certificate or autopsy, and patients for whom the month of diagnosis was not available. We also excluded patients who were only eligible for Medicare on the basis of end-stage renal failure or disability; therefore, all patients were 65 years of age or older.

We considered combinations of surgery, radiation, and chemotherapy.²⁷ Surgery was identified from the Medicare database using the International Classification of Diseases, ninth revision (ICD-9; codes 42.0 to 43.99).²⁸ Information on radiation use was based on SEER records and Medicare data.²⁹ Information on chemotherapy was based on Medicare data only.³⁰

Comorbidity was determined based on Medicare claims between 13 months and 1 month before diagnosis.³¹ Missing values were assigned to patients without Medicare data from this time window if no comorbidity was registered. Missing values were statistically imputed to allow for analysis of the available information from other predictors.³² Exclusion of these patients in a sensitivity analysis did not affect results (data not shown). ICD-9 codes of both inpatient and outpatient bills were analyzed.^33,34 Comorbidities were grouped as cardiovascular (previous myocardial infarction, heart failure, peripheral arterial disease, cerebrovascular disease), diabetes (with or without complications), pulmonary (chronic obstructive pulmonary disease), renal (mild to severe), and hepatic (mild to severe).^8,35 We created a simple comorbidity score based on the presence of cardiac, pulmonary, renal, or hepatic comorbidity, or diabetes. For simplicity, each comorbidity was counted as one point, based on similar regression coefficients.³⁶

Patients were classified as having surgery performed in a teaching hospital versus not in a teaching hospital. Additional hospital characteristics included the hospital volume, which was based on the sum of esophagectomies per hospital using the unique hospital provider number in the Medicare data.³⁷ Low, intermediate and high volumes were defined by tertile of patients.

Patients: Eindhoven
As a second validation cohort, we considered 349 patients who underwent surgical resection for a primary tumor of the esophagus, diagnosed between January 1, 1993 and December 31, 2001 in the southeast part of the Netherlands.²⁴ The Eindhoven Cancer Registry covers approximately 2 million inhabitants who are served by 16 community hospitals and two large radiotherapy institutes (16% of the Netherlands). On notification by one of six pathological laboratories and the hospital medical records departments, registration clerks actively collect information on diagnosis, tumor stage, treatment, and comorbidities from the medical records. The Eindhoven Cancer Registry has been collecting detailed data on clinically relevant comorbidity for new cancer patients since 1993.³⁵

Patients: Rotterdam
As a third validation cohort, we considered 1,202 patients who underwent surgical resection for a primary tumor of the esophagus at the University Hospital Rotterdam, between January 1, 1980 and December 31, 2002. This hospital serves as a referral center for the southwest part of the Netherlands. A database system is maintained with detailed information on diagnosis, tumor stage, treatment, and comorbidity.^25,26 Information on hepatic and renal disease was not available for these patients.

Statistical Analysis
Descriptive statistics were used for univariable analyses, with cells with fewer than 5 patients in the SEER cohorts indicated as less than 5. We applied logistic regression analysis to relate patient and treatment characteristics to mortality within 30 days after surgery. Mortality was considered irrespective of the cause. Odds ratios (ORs) were calculated with 95% CIs. Potential predictive characteristics were chosen from the clinical literature and expert opinion.³⁸ The performance of the model was assessed with respect to calibration and discrimination.³⁹ Calibration refers to the agreement between observed outcomes and predicted probabilities and is the most important quality when trying to predict the expected mortality rate for a group of patients. Calibration was assessed graphically and tested with the Hosmer-Lemeshow goodness of fit test.⁴⁰ Discrimination refers to the ability to distinguish patients who will die from those who will survive. Discrimination was quantified by the area under the receiver operating characteristic curve (AUC), which is identical to the concordance statistic.³⁹ An AUC of 0.5 indicates no discriminative ability at all (ie, a coin flip) while an AUC of 1 indicates perfect discrimination (ie, a test with 100% sensitivity and 100% specificity). Prediction models with an AUC exceeding 0.8 have often been labeled as good to excellent, those with AUC between 0.7 and 0.8 as moderate, and those with AUC between 0.6 and 0.7 as providing low discrimination.

Multivariable models were internally and externally validated. Internal validation was performed with a standard bootstrap procedure.^38,39 Bootstrap samples were drawn with replacement of the same size as the original sample. Predictions from each bootstrap model were evaluated in the original sample. The difference in performance in the bootstrap sample and in the original sample quantifies the optimism that may be expected when the multivariable model is applied to new, but similar, patients. A score chart was derived from the multivariable regression coefficients. For simple application, the coefficients were multiplied by two and rounded.

For external validation, we constructed logistic regression models for each cohort, containing the same predictors as the multivariable model based on SEER 91-96. We studied whether the predictors had similar effects. Subsequently, we derived a combined model based on all patient data, with stratification for study.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
The SEER-Medicare patients were on average 73 and 74 years of age (Table 1). The patients in the Eindhoven and Rotterdam series were approximately 10 years younger on average (64 and 62 years, respectively), because these series included patients of all ages, not just patients 65 years or older. The majority of patients were male. Comorbidity was found in approximately 20% of the SEER-Medicare patients (19% and 23%, respectively, especially pulmonary [9% and 9%], cardiovascular [9% and 9%], and diabetes [8% and 10%]). Pulmonary and cardiovascular comorbidities were more often registered for the Rotterdam patients (15% and 16%, respectively), while cardiovascular comorbidity was more common in the Eindhoven patients (18%). Most patients had adenocarcinoma and pathologically confirmed locoregional disease. Neoadjuvant treatment was given in approximately 20% of the SEER-Medicare patients, 7% of the Eindhoven patients, and 38% of the Rotterdam patients. Most SEER-Medicare patients were treated in teaching hospitals, although the annual volumes of esophagectomies were relatively small. This information was not reliably available for the Eindhoven patients. The Rotterdam center is a referral hospital with more than 50 esophagectomies per year.

Univariable Analyses
In univariable analysis of the 1,317 SEER 91-96 patients, characteristics that were statistically significantly associated with mortality included age and comorbidity (pulmonary, diabetes; Table 2). Neoadjuvant treatment with radiotherapy was associated with a substantially higher surgical mortality risk (23%), as was neoadjuvant chemoradiotherapy (16%). In contrast, neoadjuvant chemotherapy alone was associated with a somewhat lower risk of mortality (6%). For robust additional analyses we combined the patients without any neoadjuvant treatment with those with neoadjuvant chemotherapy. Finally, higher hospital volume was clearly associated with lower surgical mortality (P value for trend .003).

Multivariable Analyses
Neoadjuvant treatment remained associated with an increased risk of surgical mortality in multivariable logistic regression analysis of the 1,317 SEER 91-96 patients, with adjusted ORs of 2.5 and 1.9 for radiotherapy and chemoradiotherapy, respectively (Table 3). Comorbidity and age were also highly predictive, with an OR of 1.6 per comorbid condition, and an OR of 1.6 per decade of age. Higher volume hospitals exhibited close to half the mortality of lower volume hospitals (OR, 0.59).

Model Performance and Risk Score
The multivariable model based on the 1,317 SEER 91-96 patients showed low discrimination (AUC, 0.66). Internal validation of this model indicated a slight decrease in discriminative ability (AUC, 0.65). The multivariable model performed similarly in the 714 SEER 97-99 patients (AUC, 0.70) and Rotterdam patients (AUC, 0.66), but poor in the Eindhoven patients (AUC, 0.56).

A simple chart assigned 1 point per 15 years of age, 1 point per comorbidity, 1 point for neoadjuvant chemoradiotherapy, and 1.5 points for radiotherapy (Table 4). Hospital volume was scored as 0, –0.5, –1.5 and –2 for low, intermediate, high, and very high volume, respectively, based on the multivariable effects from Table 3. A summary score corresponds to a predicted probability of 30-day mortality (Fig 1). For example, a 65-year-old patient, who has pulmonary and cardiovascular comorbidity, has not received neoadjuvant treatment, and undergoes surgery in a low volume hospital has a score of 0 + 2 + 0 + 0 = 2 points. This score corresponds to a predicted mortality of 17% (95%CI, 14% to 21%). If this patient were to be treated in a very high volume hospital (score –2, sum score = 0), the predicted mortality would be 7% (95% CI, 5.7% to 8.5%). The performance of the risk score was similar to the original model for each cohort (Table 3).

View larger version (9K): [in this window] [in a new window] [PowerPoint Slide for Teaching]   Fig 1. Estimated surgical mortality in relation to the sum score that can be obtained from Table 4. The 95% CIs are based on analysis of four cohorts, containing 3,592 patients undergoing surgery for esophageal cancer.

View larger version (14K): [in this window] [in a new window] [PowerPoint Slide for Teaching]   Fig 2. Calibration of predictions of 30-day mortality after esophagectomy (n = 3,592).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

Age predicts surgical risk for a wide range of procedures. For esophagectomy, one major study found that 30-day mortality increased from 10.7% for patients between 65 and 69 years of age to more than 20% for those older than 80 years of age.¹² We confirmed this trend in our data, especially in the SEER-Medicare cohorts, where we observed a relative increase in mortality of 40% per decade in adjusted analysis. The presence of comorbidity is known to affect outcome in many cancers. A number of scoring systems have been developed to measure it, including the Charlson score⁴¹ and the Adult Comorbidity Evaluation index (ACE-27).¹³ We used a simple count of comorbid cardiovascular, pulmonary, renal, hepatic conditions, and diabetes, and found that each point was on average associated with a 50% increase in surgical risk (OR, 1.5). Comorbidity scoring was claims based in the SEER-Medicare data and chart based in the other two cohorts. Despite these and other differences in definitions, the comorbidity-mortality relationship was similar across the four study cohorts in line with findings in a previous study.⁴²

Preoperative unimodality radiotherapy was clearly associated with higher surgical mortality. This treatment strategy has largely been abandoned in recent years,⁴³ so a more relevant finding is that neoadjuvant chemoradiotherapy nearly doubled mortality compared with no neoadjuvant treatment or neoadjuvant chemotherapy alone. It is possible that these results may be somewhat confounded by selection of patients with more advanced tumors for neoadjuvant treatment. However, our findings are consistent with recent meta-analyses of randomized trials, showing that neoadjuvant chemoradiotherapy was associated with a 1.72-fold higher surgical mortality (95% CI, 0.96 to 3.1; P = .07)³, while neoadjuvant chemotherapy alone was not (OR, 1.08; 95% CI, 0.45 to 2.6; P = .87).⁴⁴ This implies that part of the benefit of neoadjuvant chemoradiotherapy (for example, on small metastases which would not be resected by surgery) may be offset by higher surgical mortality. This issue requires further detailed evaluation in randomized trials. Our results suggest that measures to reduce surgical risk should especially be considered for patients with neoadjuvant chemoradiotherapy.

Reported surgical risks vary widely in the literature. Much of this variation can be explained by differences in hospital volume.^14-19,45,46 Hospital volume remained important after adjustment for case-mix, which is generally an important methodological consideration in such analyses of observational data.^9-11 Many authors suggest that a policy of concentrating care in high-volume centers should be considered especially for esophagectomy, where outcome varies substantially between low-volume and high-volume providers.^20,47 As illustrated, a patient could have a 17% or 7% predicted mortality risk depending on surgery in a low-volume or in a high volume center.

Our study has some limitations. Our four cohorts were of considerable size, but the Eindhoven and Rotterdam series had only few events, which makes firm conclusions on external validity difficult.⁴⁸ We included all patients undergoing esophagectomy. Approximately 10% had pathologically distant disease, and we cannot exclude that a few patients had clinically known distant disease before surgery. The inclusion of these patients may have led us to overestimate risk for patients with true locoregional disease. In contrast, we considered 30-day mortality, and in-hospital mortality can be substantially higher. We further note that the model was mainly based on patients older than 65 years of age; validity may be best for this patient category. Next, the exact limits to define low-, medium-, and high-volume centers are hard to determine. We used rather low annual volumes (< 1, 1 to 2.5, 2.6 per year), based on tertiles of patients, while higher limits may be better defendable. Finally, we did not have information on physiologic variables such as performance status or American Society of Anesthesiologists (ASA) score, which have been used in clinical models.^45,49 Our risk model hence had only a low discriminative ability, and can therefore only be a relatively rough, though evidence based, basis for the surgical risk of individual patients. Our score is easy to calculate from existing, readily available data, however, and so could well serve for case-mix adjustment when comparing surgical mortality rates between institutions.

In conclusion, we found substantial mortality after esophagectomy, which was related to patient, neoadjuvant therapy, and hospital characteristics. We developed and externally validated a simple risk score, which provides an admittedly rough estimate of surgical mortality with which to compare actual outcomes. Further validation and extension of this score is mandatory.

	Authors' Disclosures of Potential Conflicts of Interest

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
The authors indicated no potential conflicts of interest.

	Author Contributions

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 

Conception and design: Ewout W. Steyerberg, Bridget A. Neville, Linetta B. Koppert, Jane C. Weeks, Craig C. Earle Financial support: Ewout W. Steyerberg, Jane C. Weeks, Craig C. Earle Collection and assembly of data: Bridget A. Neville, Linetta B. Koppert, Valery E.P.P. Lemmens, Hugo W. Tilanus, Jan-Willem W. Coebergh Data analysis and interpretation: Ewout W. Steyerberg, Bridget A. Neville, Craig C. Earle Manuscript writing: Ewout W. Steyerberg, Bridget A. Neville, Jane C. Weeks, Craig C. Earle Final approval of manuscript: Ewout W. Steyerberg, Bridget A. Neville, Linetta B. Koppert, Valery E.P.P. Lemmens, Hugo W. Tilanus, Jan-Willem W. Coebergh, Jane C. Weeks, Craig C. Earle

 

	GLOSSARY

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 

Bootstrap procedure:

A nonparametric statistical method to estimate sampling distributions of an estimator by resampling with replacement from the original sample. In prognostic research, the bootstrap helps to obtain an impression of the validity of predictions in new but similar patients.

Calibration:

Agreement of predicted risks with observed outcomes; for example, mortality rates.

Comorbidity:

Having two or more diseases at the same time.

Discrimination:

The ability to distinguish between patients with good and poor outcomes.

Logistic regression analysis:

A multivariable regression model in which the log of the odds of a time-fixed outcome event (eg, 30-day mortality) is related to a linear equation.

Optimism:

The decrease in model performance in new patients compared with performance in the sample under study.

Prognostic model:

A combination of patient, tumor, and treatment characteristics that predict outcome of individual patients.

Risk adjustment:

Risk-adjustment aims to allow for fair comparison of outcomes of different patient samples by statistical compensation for risk factor differences between the samples (eg, case-mix in different hospitals).

Risk score:

A simplified version of a prognostic model, where scores are assigned to each risk factor (eg, based on rounded regression coefficients).

Validation:

The process that tests the performance of a previously defined classifier or prognostic model on a new set of patients. For example, a gene expression signature classifier developed using data from one set of patients might be validated on another, independent set of patients.

	NOTES

 
Terms in blue are defined in the glossary, found at the end of this article and online at www.jco.org.

Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.

	REFERENCES

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION Authors' Disclosures of... Author Contributions GLOSSARY REFERENCES

 
1. Leonard GD, McCaffrey JA, Maher M: Optimal therapy for oesophageal cancer. Cancer Treat Rev 29:275-282, 2003[CrossRef][Medline]

2. Wu PC, Posner MC: The role of surgery in the management of oesophageal cancer. Lancet Oncol 4:481-488, 2003[CrossRef][Medline]

3. Urschel JD, Vasan H: A meta-analysis of randomized controlled trials that compared neoadjuvant chemoradiation and surgery to surgery alone for resectable esophageal cancer. Am J Surg 185:538-543, 2003[CrossRef][Medline]

4. Kaklamanos IG, Walker GR, Ferry K, et al: Neoadjuvant treatment for resectable cancer of the esophagus and the gastroesophageal junction: A meta-analysis of randomized clinical trials. Ann Surg Oncol 10:754-761, 2003[CrossRef][Medline]

5. Geh JI: The use of chemoradiotherapy in oesophageal cancer. Eur J Cancer 38:300-313, 2002[CrossRef][Medline]

6. Weeks J: Overview of outcomes research and management and its role in oncology practice. Oncology (Huntingt) 12:11-13, 1998

7. Blazeby JM, Farndon JR, Donovan J, et al: A prospective longitudinal study examining the quality of life of patients with esophageal carcinoma. Cancer 88:1781-1787, 2000[CrossRef][Medline]

8. Bartels H, Stein HJ, Siewert JR: Preoperative risk analysis and postoperative mortality of oesophagectomy for resectable oesophageal cancer. Br J Surg 85:840-844, 1998[CrossRef][Medline]

9. DeLong ER, Peterson ED, DeLong DM, et al: Comparing risk-adjustment methods for provider profiling. Stat Med 16:2645-2664, 1997[CrossRef][Medline]

10. Shahian DM, Blackstone EH, Edwards FH, et al: Cardiac surgery risk models: A position article. Ann Thorac Surg 78:1868-1877, 2004[Abstract/Free Full Text]

11. Halm EA, Lee C, Chassin MR: Is volume related to outcome in health care? A systematic review and methodologic critique of the literature. Ann Intern Med 137:511-520, 2002[Abstract/Free Full Text]

12. Finlayson EV, Birkmeyer JD: Operative mortality with elective surgery in older adults. Eff Clin Pract 4:172-177, 2001[Medline]

13. Piccirillo JF, Tierney RM, Costas I, et al: Prognostic importance of comorbidity in a hospital-based cancer registry. JAMA 291:2441-2447, 2004[Abstract/Free Full Text]

14. Begg CB, Cramer LD, Hoskins WJ, et al: Impact of hospital volume on operative mortality for major cancer surgery. JAMA 280:1747-1751, 1998[Abstract/Free Full Text]

15. Swisher SG, Deford L, Merriman KW, et al: Effect of operative volume on morbidity, mortality, and hospital use after esophagectomy for cancer. J Thorac Cardiovasc Surg 119:1126-1132, 2000[Abstract/Free Full Text]

16. Dimick JB, Cattaneo SM, Lipsett PA, et al: Hospital volume is related to clinical and economic outcomes of esophageal resection in Maryland. Ann Thorac Surg 72:334-339, 2001; discussion 72:339-341, 2001[Abstract/Free Full Text]

17. Bachmann MO, Alderson D, Edwards D, et al: Cohort study in South and West England of the influence of specialization on the management and outcome of patients with oesophageal and gastric cancers. Br J Surg 89:914-922, 2002[CrossRef][Medline]

18. Finlayson EV, Goodney PP, Birkmeyer JD: Hospital volume and operative mortality in cancer surgery: A national study. Arch Surg 138:721-725, 2003; discussion 138:726, 2003[Abstract/Free Full Text]

19. Birkmeyer JD, Siewers AE, Finlayson EV, et al: Hospital volume and surgical mortality in the United States. N Engl J Med 346:1128-1137, 2002[Abstract/Free Full Text]

20. Birkmeyer JD, Stukel TA, Siewers AE, et al: Surgeon volume and operative mortality in the United States. N Engl J Med 349:2117-2127, 2003[Abstract/Free Full Text]

21. Saito T, Shimoda K, Kinoshita T, et al: Prediction of operative mortality based on impairment of host defense systems in patients with esophageal cancer. J Surg Oncol 52:1-8, 1993[Medline]

22. Zafirellis KD, Fountoulakis A, Dolan K, et al: Evaluation of POSSUM in patients with oesophageal cancer undergoing resection. Br J Surg 89:1150-1155, 2002[CrossRef][Medline]

23. Potosky AL, Riley GF, Lubitz JD, et al: Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care 31:732-748, 1993[Medline]

24. Koppert LB, Janssen-Heijnen ML, Louwman MW, et al: Comparison of comorbidity prevalence in oesophageal and gastric carcinoma patients: A population-based study. Eur J Gastroenterol Hepatol 16:681-688, 2004[CrossRef][Medline]

25. Tilanus HW, Hop WC, Langenhorst BL, et al: Esophagectomy with or without thoracotomy. Is there any difference? J Thorac Cardiovasc Surg 105:898-903, 1993[Abstract]

26. Polee MB, Hop WC, Kok TC, et al: Prognostic factors for survival in patients with advanced oesophageal cancer treated with cisplatin-based combination chemotherapy. Br J Cancer 89:2045-2050, 2003[CrossRef][Medline]

27. Steyerberg EW, Earle CC, Neville BA, et al: Racial differences in surgical evaluation, treatment, and outcome of locoregional esophageal cancer: A population-based analysis of elderly patients. J Clin Oncol 23:510-517, 2005[Abstract/Free Full Text]

28. Cooper GS, Virnig B, Klabunde CN, et al: Use of SEER-Medicare data for measuring cancer surgery. Med Care 40:43-48, 2002

29. Virnig BA, Warren JL, Cooper GS, et al: Studying radiation therapy using SEER-Medicare-linked data. Med Care 40:49-54, 2002

30. Warren JL, Harlan LC, Fahey A, et al: Utility of the SEER-Medicare data to identify chemotherapy use. Med Care 40:55-61, 2002

31. Klabunde CN, Warren JL, Legler JM: Assessing comorbidity using claims data: An overview. Med Care 40:26-35, 2002[CrossRef][Medline]

32. Arnold AM, Kronmal RA: Multiple imputation of baseline data in the cardiovascular health study. Am J Epidemiol 157:74-84, 2003[Abstract/Free Full Text]

33. Deyo RA, Cherkin DC, Ciol MA: Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 45:613-619, 1992[CrossRef][Medline]

34. Klabunde CN, Potosky AL, Legler JM, et al: Development of a comorbidity index using physician claims data. J Clin Epidemiol 53:1258-1267, 2000[CrossRef][Medline]

35. Coebergh JW, Janssen-Heijnen ML, Post PN, et al: Serious co-morbidity among unselected cancer patients newly diagnosed in the southeastern part of The Netherlands in 1993-1996. J Clin Epidemiol 52:1131-1136, 1999[CrossRef][Medline]

36. Wang PS, Walker A, Tsuang M, et al: Strategies for improving comorbidity measures based on Medicare and Medicaid claims data. J Clin Epidemiol 53:571-578, 2000[CrossRef][Medline]

37. Schrag D, Bach PB, Dahlman C, et al: Identifying and measuring hospital characteristics using the SEER-Medicare data and other claims-based sources. Med Care 40:96-103, 2002[CrossRef][Medline]

38. Steyerberg EW, Eijkemans MJ, Harrell FE Jr, et al: Prognostic modeling with logistic regression analysis: A comparison of selection and estimation methods in small data sets. Stat Med 19:1059-1079, 2000[CrossRef][Medline]

39. Harrell FE Jr., Lee KL, Mark DB: Multivariable prognostic models: Issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 15:361-387, 1996[CrossRef][Medline]

40. Hosmer DW, Hosmer T, Le Cessie S, et al: A comparison of goodness-of-fit tests for the logistic regression model. Stat Med 16:965-980, 1997[CrossRef][Medline]

41. Charlson ME, Pompei P, Ales KL, et al: A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 40:373-383, 1987[CrossRef][Medline]

42. Newschaffer CJ, Bush TL, Penberthy LT: Comorbidity measurement in elderly female breast cancer patients with administrative and medical records data. J Clin Epidemiol 50:725-733, 1997[CrossRef][Medline]

43. Arnott SJ, Duncan W, Gignoux M, et al: Preoperative radiotherapy for esophageal carcinoma. Cochrane Database Syst Rev CD001799, 2000

44. 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[CrossRef][Medline]

45. McCulloch P, Ward J, Tekkis PP: Mortality and morbidity in gastro-oesophageal cancer surgery: Initial results of ASCOT multicentre prospective cohort study. BMJ 327:1192-1197, 2003[Abstract/Free Full Text]

46. van Lanschot JJ, Hulscher JB, Buskens CJ, et al: Hospital volume and hospital mortality for esophagectomy. Cancer 91:1574-1578, 2001[CrossRef][Medline]

47. Shahian DM, Normand SL: The volume-outcome relationship: From Luft to Leapfrog. Ann Thorac Surg 75:1048-1058, 2003[Abstract/Free Full Text]

48. Vergouwe Y, Steyerberg EW, Eijkemans MJ, et al: Substantial effective sample sizes were required for external validation studies of predictive logistic regression models. J Clin Epidemiol 58:475-483, 2005[CrossRef][Medline]

49. Sauvanet A, Mariette C, Thomas P, et al: Mortality and morbidity after resection for adenocarcinoma of the gastroesophageal junction: Predictive factors. J Am Coll Surg 201:253-262, 2005[CrossRef][Medline]

Submitted November 23, 2005; accepted May 2, 2006.

CiteULike    Complore    Connotea    Del.icio.us    Digg    Facebook    Reddit    Technorati    Twitter    What's this?

This article has been cited by other articles:

				E. C. Paulson, J. Ra, K. Armstrong, C. Wirtalla, F. Spitz, and R. R. Kelz Underuse of Esophagectomy as Treatment for Resectable Esophageal Cancer Arch Surg, December 1, 2008; 143(12): 1198 - 1203. [Abstract] [Full Text] [PDF]

				S. M. Lagarde, J. B. Reitsma, A.-K. D. Maris, M. I. van Berge Henegouwen, O. R.C. Busch, H. Obertop, A. H. Zwinderman, and J. J. B. van Lanschot Preoperative Prediction of the Occurrence and Severity of Complications After Esophagectomy for Cancer With Use of a Nomogram Ann. Thorac. Surg., June 1, 2008; 85(6): 1938 - 1945. [Abstract] [Full Text] [PDF]

				E. Internullo, J. Moons, P. Nafteux, W. Coosemans, G. Decker, P. De Leyn, D. Van Raemdonck, and T. Lerut Outcome after esophagectomy for cancer of the esophagus and GEJ in patients aged over 75 years Eur. J. Cardiothorac. Surg., June 1, 2008; 33(6): 1096 - 1104. [Abstract] [Full Text] [PDF]

				E. W. Steyerberg, B. Neville, J. C. Weeks, and C. C. Earle Referral Patterns, Treatment Choices, and Outcomes in Locoregional Esophageal Cancer: A Population-Based Analysis of Elderly Patients J. Clin. Oncol., June 10, 2007; 25(17): 2389 - 2396. [Abstract] [Full Text] [PDF]

				M. Extermann and A. Hurria Comprehensive Geriatric Assessment for Older Patients With Cancer J. Clin. Oncol., May 10, 2007; 25(14): 1824 - 1831. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Rights & Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Steyerberg, E. W. Articles by Earle, C. C. Search for Related Content PubMed PubMed Citation Articles by Steyerberg, E. W. Articles by Earle, C. C. Social Bookmarking                            What's this?