ATS
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Ann Thorac Surg 2010;90:920-925. doi:10.1016/j.athoracsur.2010.06.024
© 2010 The Society of Thoracic Surgeons
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to Personal Folders
Right arrow Download to citation manager
Right arrow Permission Requests
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Grotenhuis, B. A.
Right arrow Articles by van Lanschot, J. J. B.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Grotenhuis, B. A.
Right arrow Articles by van Lanschot, J. J. B.
Related Collections
Right arrow Esophagus - cancer
Right arrowRelated Article

Original Articles: General Thoracic

Validation of a Nomogram Predicting Complications After Esophagectomy for Cancer

Brechtje A. Grotenhuis, MDa,*, Pieter van Hagen, MDa, Johannes B. Reitsma, MD, PhDb, Sjoerd M. Lagarde, MD, PhDc, Bas P.L. Wijnhoven, MD, PhDa, Mark I. van Berge Henegouwen, MD, PhDc, Hugo W. Tilanus, MD, PhDa, J. Jan B. van Lanschot, MD, PhDa

a Department of Surgery, Erasmus Medical Center, Rotterdam, The Netherlands
b Department of Clinical Epidemiology, Biostatistics and Bioinformatics, Academic Medical Center, Amsterdam, The Netherlands
c Department of Surgery, Academic Medical Center, Amsterdam, The Netherlands

Accepted for publication June 7, 2010.

* Address correspondence to Dr Grotenhuis, Erasmus Medical Center, Department of Surgery, PO Box 2040, 3000 CA Rotterdam, The Netherlands (Email: b.grotenhuis{at}erasmusmc.nl).


   Abstract
Top
 Abstract
Introduction
 Patients and Methods
 Results
 Comment
 Acknowledgments
 References
 
Background: A nomogram has been developed recently in order to predict the occurrence and severity of postoperative complications after esophagectomy for cancer. In the present study, we externally validated this nomogram in a new cohort of patients who underwent esophagectomy for cancer in a different high-volume center.

Methods: An independent dataset of 777 patients who underwent esophagectomy for cancer was used for validation. The discriminatory capability of the nomogram was determined by using the concordance index (C statistic). Calibration was evaluated by comparing the observed with the expected number of patients with complications, as predicted by the original nomogram across patients with different risk profiles. We also examined whether adjusting the value of the original coefficients of the predictors or adding new predictors would improve the fit of the nomogram.

Results: Discrimination of the original nomogram was similar in the validation cohort: the C statistic hardly decreased from 0.65 in the original cohort to 0.64 in the validation cohort. Observed and expected number of patients with complications were in close agreement, reflecting a good calibration (p = 0.84). Reestimation of the coefficients in the validation cohort did not lead to any significant changes of the original nomogram values.

Conclusions: External validation of a nomogram predicting the occurrence and severity of complications after esophagectomy showed that the model is applicable in other high-volume hospitals. Nevertheless, preoperative prediction of complications in individual patients remains difficult, most likely due to the complexity of mechanisms causing these complications.


   Introduction
Top
 Abstract
 Introduction
Patients and Methods
 Results
 Comment
 Acknowledgments
 References
 
Surgery is the primary curative therapy for patients with esophageal cancer. However, esophagectomy is associated with a high operative risk [1, 2]. Although operative mortality is below 5% in high volume centers [1, 3], esophageal resection is still accompanied by substantial morbidity. Early postoperative complication rates vary between 40% and 80%, depending on the applied criteria and on the extent of resection [1, 4, 5]. Complications can range from minor complications (eg, urinary tract infection) to major complications (eg, respiratory failure). Several previous studies focused on predisposing factors for complications after esophagectomy for cancer, but this did not result in reliable predictive models [6–10], except for the prediction of pulmonary complications [11].

Predicting the severity of complications after esophagectomy may supply important information for both patient and surgeon. The impact of postoperative complications on quality of life plays an important role in decision making whether to proceed with an operation in an individual patient with esophageal cancer. Furthermore, the prediction of severity of complications in the preoperative phase may help in choosing the extent of the operation and in informing patients. Recently, a nomogram has been developed to predict the occurrence and severity of complications in esophagectomy patients [12]. A nomogram gives a graphic representation of the predictive strength of specific predictors and enables clinicians to calculate an overall risk score for individual patients reflecting their personal risk. In this nomogram the severity of complications was predicted at three levels: no complications, minor complications, or severe complications. Specific patients' characteristics (ie, more advanced age, myocardial infarction, stroke or transient ischemic attack in the medical history, lower forced expiratory volume in one second, and presence of electrocardiographic changes) and the application of more extensive surgery (ie, transthoracic esophagectomy [TTE]) were associated with a higher risk of (more severe) postoperative complications.

This new nomogram has been validated in a second group of patients within the same institution (temporal validation), which did not reveal any statistically significant changes in the predictive strength of the included prognostic factors. However, the power of that validation study was only moderate given the limited sample size of 100 patients [13], and validation within the same institute may not reveal inherent problems of a prognostic model that can become apparent when differences in patient populations, surgeons, applied surgical techniques, and postoperative care between institutions are taken into account [14]. Therefore, the aim of the present study was to externally validate this nomogram in a new and large cohort of patients who underwent esophagectomy for cancer in a different high-volume center.


   Patients and Methods
Top
 Abstract
 Introduction
 Patients and Methods
Results
 Comment
 Acknowledgments
 References
 
Patients
Between January 1991 and September 2008, a consecutive series of 777 patients underwent a potentially curative esophagectomy for adenocarcinoma or squamous cell carcinoma of the esophagus or gastroesophageal junction in the Erasmus Medical Center, a tertiary referral center with wide experience in esophageal surgery. Patients who received neoadjuvant chemo(radio)therapy or induction chemotherapy were not excluded from the present study. In this hospital, patients mainly received neoadjuvant chemo(radio)therapy in the context of randomized controlled trials [15, 16]. Induction chemotherapy was given in patients with either a cT4-tumor without distant metastases or in patients with gross involvement of celiac trunk lymph nodes (M1a), who were not considered eligible for primary surgical therapy. Transhiatal esophagectomy was the preferred technique in this series of patients (n = 744). A minority of patients (n = 33) underwent TTE, mainly in the context of a randomized controlled trial [1, 17]. Surgical techniques have been described before [1, 17]. Clinicopathologic data of all patients had been collected routinely in an ongoing registry.

The Medical Ethics Committees of the participating hospitals that provided the derivation cohort (Academic Medical Center, Amsterdam) and the validation cohort (Erasmus Medical Center, Rotterdam) have approved the ongoing registry of data of esophageal cancer patients in prospective databases. They also approved the publication of the data used for the current study.

Definitions of Complications
The severity of postoperative complications was graded according to the morbidity scale proposed by Dindo and colleagues [18]. This classification system is based on the therapeutic consequences of complications and consists of five grades and two subgrades. Grade I complications do not need any medical or surgical intervention (eg, atelectasis, vocal cord paralysis, radiologic anastomotic leakage). Grade II complications need pharmacologic treatment (eg, pneumonia, chyle leakage, pulmonary embolus). Grade III complications need an intervention (grade IIIa: eg, anastomotic leakage requiring drainage of the neck, pneumothorax, or pleural empyema requiring drainage of the chest; grade IIIb: any reoperation). Grade IV complications are life threatening and represent single organ failure (grade IVa: eg, pulmonary dysfunction requiring artificial ventilation, heart failure requiring intravenous inotropic agents, renal insufficiency requiring dialysis) or multiorgan failure (grade IVb). Finally, grade V complications are complications leading to death. Grading of complications was performed according to the most severe complication in each patient. Similar to the original study, complications were categorized into three groups: no complications, minor to moderate complications (grades I to IIIb), and severe complications (grades IVa, IVb, and V).

Statistical Analysis
All analyses were performed using SAS software version 9.1 (SAS Institute Inc, Cary, NC). Clinicopathologic characteristics of the derivation cohort (Academic Medical Center, Amsterdam) and the validation cohort (Erasmus Medical Center, Rotterdam) were analyzed in a descriptive way. Data on daily alcohol intake and presence of preoperative dyspnea were not available for the Rotterdam patients. In Rotterdam, weight loss was classified as a categoric value rather than a continuous variable (Amsterdam).

Ordinal logistic regression was used to examine the association between predictors and the occurrence of complications classified in three categories of severity (no complications, minor complications, severe complications). The ordinal logistic regression (proportional odds model) is an extension of the binary logistic regression model that is used in case of three or more outcome states that are naturally ordered [19]. Univariate analyses of potential predictors for the severity of complications were performed with the data set of the Rotterdam cohort. Because missing data result in loss of statistical power and can lead to possible bias, multiple imputation techniques were applied [20]. All predictors, as well as the observed outcome, were used to impute missing values based on multivariate normal distributions using the Markov chain Monte Carlo method. The coefficients of ten rounds of imputations were combined to obtain the final estimates of odds ratios and their 95% confidence intervals.

The nomogram predicting the occurrence and severity of complications previously developed in Amsterdam was validated in several ways on the Rotterdam patient cohort. First, discrimination was evaluated by calculating a risk score for each patient in the validation cohort based on the original coefficients from the derivation cohort. The discriminatory properties of the model were examined by visualizing the distribution and overlap in risk scores of individual patients within and between the three outcome categories. Furthermore, the discriminative capability was quantified by using the concordance (C) statistic. The C statistic is a measure that can be interpreted as the probability among all possible pairs between patients from different outcome categories that the patient with the more severe complication also has the higher risk score. Values can range from 0.5 (due to chance, no discrimination) to 1.0 (perfect discrimination). We calculated C statistics for the three categories (no complications versus minor complications, minor complications versus severe complications, no complications versus severe complications) as well as the overall C statistic in the validation cohort. These data were compared with the original C values in the derivation cohort.

Second, calibration was evaluated by comparing the expected and observed number of patients in each of the three outcome categories across seven quantiles of expected risk (ie, seven groups of each 111 patients with an increasing mean risk score). Calibration was tested for significance by using an extension of the Hosmer-Lemeshow goodness-of-fit statistic [21].

The third analysis examined whether the importance of the individual predictors or intercepts within the original nomogram was different in the validation cohort. For each predictor a proportional odds model was fitted with only that predictor using the risk score based on the original coefficients as an offset variable. Ideally, if the weight (importance) of the predictor is comparable between the validation cohort and the derivation cohort, the coefficient of each predictor would be zero as its weight is already incorporated in the risk score. The likelihood ratio test was used to indicate whether reweighing of the predictor will significantly improve the model. Similarly, potential predictors that did not significantly improve the prognostic performance of the original nomogram were reevaluated in the derivation cohort whether they improved the prognostic performance of the model in the new cohort.


   Results
Top
 Abstract
 Introduction
 Patients and Methods
 Results
Comment
 Acknowledgments
 References
 
Clinicopathologic characteristics of the two study populations are shown in Table 1. The groups were comparable with regard to general patient characteristics such as age, sex, American Society of Anesthesiologists classification, body mass index, and comorbidities. More patients with a tumor located in the proximal or midesophagus underwent an esophagectomy in the Amsterdam group, which is also reflected in the higher proportion of patients who underwent an extended lymphadenectomy by means of a TTE. The median preoperative risk score as calculated by the original nomogram was significantly lower in Rotterdam than in Amsterdam (0.82 vs 1.07, p < 0.001), indicating that the case mix with respect to the presence of predictors included in the nomogram was more favorable in Rotterdam. The frequency and severity of complications in the validation cohort (Rotterdam) and the derivation cohort (Amsterdam) are shown in Table 2. More patients in Rotterdam did not develop any complications postoperatively (40% Rotterdam versus 30% Amsterdam). In-hospital mortality was comparable (5% Rotterdam versus 4% Amsterdam).


View this table:
[in this window]
[in a new window]

  		Table 1 Clinicopathologic Characteristics and Risk Score of Patients who Underwent Esophagectomy for Cancer in Rotterdam (Validation Cohort) and Amsterdam (Derivation Cohort)
 

View this table:
[in this window]
[in a new window]

  		Table 2 Frequency and Severity of Complications in the Rotterdam Study Population (Validation Cohort) Versus the Amsterdam Patients (Derivation Cohort)
 
In the first validation step we examined the discriminatory capability of the original risk score in the validation cohort. The distribution and overlap in risk scores of individual patients within and between the three outcome categories are shown in Figure 1. The mean risk scores (±SD) in the three complication categories were significantly different: 0.74 (±0.48) in patients without complications, 0.94 (±0.59) in patients with minor complications, and 1.30 (±0.70) in patients with severe complications (p < 0.001). However, there was a substantial overlap in scores among the three categories (Fig 1). This was also reflected in the pairwise C statistics (overall measure of discriminatory capability): 0.61 for the discrimination between the group without complications versus patients with minor complications, 0.68 for the group with minor complications versus severe complications, and 0.77 for patients without complications versus severe complications. The overall C statistic of the model in the Rotterdam validation cohort was 0.64, which was only marginally lower than the C statistic (0.65) in the original Amsterdam derivation cohort [12], reflecting a moderate individual discriminatory capability of the model.


Figure 1
View larger version (28K):
[in this window]
[in a new window]

  		Fig 1. Distribution of risk score (as calculated with the original nomogram) in the patients of the validation cohort who experienced no complications, minor complications, or severe complications. (Concordance [C] statistic = a measure that can be interpreted as the probability among all possible pairs between patients from different outcome categories that the patient with the more severe complication also has the higher risk score.)

 
Secondly, we evaluated the calibration of the nomogram score in the validation cohort (ie, the closeness of predicted and observed frequency of outcomes). In Figure 2 the expected number of patients in each outcome category is depicted next to the observed number of patients in each category, across seven equally sized groups of 111 patients in the validation cohort ordered according to their risk score. In general, there was a tendency that complications in the validation cohort occurred less frequently or were less severe than expected (except for group 4 in which more severe complications were observed than expected). These discrepancies became smaller when the intercepts (ie, background risk for an individual hospital) were reestimated in the validation cohort while still using the original risk score (data not shown). The fit of the nomogram was evaluated by means of the goodness-of-fit test (p = 0.84), which indicated that the differences between the probabilities predicted by the model and the actual probabilities were small and nonsignificant.


Figure 2
View larger version (54K):
[in this window]
[in a new window]

  		Fig 2. Expected (exp) versus observed (obs) number of patients in each of the three outcome categories (no complications, minor complications, or severe complications) across seven quantiles of increasing expected risk, with a p value for goodness-of-fit of 0.84 (ie, differences between observed and expected number of patients are nonsignificant).

 
In the third validation step, we examined whether adjusting the original coefficients (weight) of the predictors would improve the fit of the nomogram model in the validation cohort. The outcome is shown in Table 3; only for the predictor age there was an indication (p = 0.07) that a change in coefficient would improve the model. Furthermore, the optimal coefficients in the validation cohort did not significantly differ from the coefficients of the original nomogram model. Finally, predictors which were not incorporated in the original nomogram were reevaluated in the derivation cohort in order to determine whether they might improve the prognostic performance of the nomogram in the new cohort. The addition of American Society of Anesthesiologists classification (p = 0.14), body mass index (p = 0.62), smoking (p = 0.47), or application of neoadjuvant chemo(radio)therapy (p = 0.17) to the existing model did not result in a significant improvement in predicting the occurrence and severity of postoperative complications in the validation cohort.


View this table:
[in this window]
[in a new window]

  		Table 3 Comparison of the Importance of Each Predictor in the Validation Cohort With That Within the Derivation Cohort (Original Nomogram). Coefficients Are Presented as Odds Ratios With Their 95% Confidence Intervals in Brackets
 

   Comment
Top
 Abstract
 Introduction
 Patients and Methods
 Results
 Comment
Acknowledgments
 References
 
In the present study we externally validated a previously published nomogram [12] predicting the occurrence and severity of complications in an independent cohort of patients who underwent esophagectomy for cancer in a different high-volume center. This nomogram had been developed to assist surgeons in assessing preoperatively the risk of complications after esophagectomy. In the original nomogram, there was a considerable overlap in nomogram scores between patients with no, minor, or severe complications (ie, moderate discriminative capability), despite the identification of several factors associated with postoperative complications [12]. It was concluded that the complication nomogram could be of value if an adequate performance of the model would be examined in other settings (hospitals) with possible differences in case mix, surgical procedures, and perioperative care.

The clinicopathologic characteristics of patients undergoing esophagectomy for cancer in the validation cohort in Rotterdam and the derivation cohort in Amsterdam were comparable with regard to potential prognostic factors such as age, American Society of Anesthesiologists classification, body mass index, and comorbidities (Table 1). However, the median preoperative risk score as calculated by the original nomogram was significantly lower in Rotterdam than in Amsterdam. Apparently, the preoperatively determined risk factors were more prevalent in Amsterdam. Another explanation could be the fact that a TTE was performed more often in the original cohort (36% Amsterdam versus 4% Rotterdam). The weight or importance of the predictor operation technique in the validation cohort did not significantly differ from that in the original nomogram, but it can be questioned whether the power of this parameter in the validation cohort was sufficient to evaluate this issue adequately. Overall, no large differences in grades of complications impeded an adequate evaluation of the two patient cohorts in this study (Table 2), although the comparison of occurrence and severity of complications between hospitals is subject to bias because of variable definitions of complications and different scoring systems.

When analyzing the discriminatory capability of the original risk score in the validation cohort, it appeared that the mean risk scores in the three outcome categories were significantly different, but a substantial overlap in scores among the three categories was noted (Fig 1). This was also reflected in the pairwise C statistics among the three complication categories, and the overall moderate discriminative power of the model (0.64 in the validation cohort, 0.65 in the derivation cohort). Apparently, patient-related factors are not the only determinators responsible for developing complications after esophageal cancer surgery. The intraoperative course (eg, fluid management by the anesthetist [22, 23] and surgical complications resulting in a longer operative time or more blood loss [2]), as well as the early postoperative phase (eg, application and effectiveness of epidural analgesia [24–26]), will also influence the final outcome of patients who undergo esophagectomy for cancer. However, these factors cannot be predicted in a preoperative risk assessment.

Nevertheless, the overall fit of the model and the calibration of the nomogram score in the validation cohort (ie, the closeness of predicted and observed frequency of outcomes) were good; the predictions for groups of patients with a similar risk profiles matched the observed probabilities (Fig 2). The expected frequency of patients with no or minor complications was higher than expected on the basis of the nomogram, while these differences improved after adjusting the intercept (ie, background risk for the new cohort) suggesting that the general level of complications was lower in the validation cohort. One could hypothesize that this may reflect differences in postoperative care between the two hospitals (eg, more specialized intensive care unit-care and faster detection of complications, preventing an increase in complication grade). On the other hand it can be questioned whether the complication registration in the validation cohort was as efficient and complete as in the derivation cohort.

Despite the complexity of mechanisms that can lead to the development of complications (preoperative, intraoperative, and early postoperative factors, as well as intrinsic [patient-related] and extrinsic [hospital-related] elements) and the subjectivity in the registration of complications, the optimal coefficients of the prognostic factors in the validation cohort did not significantly differ from the coefficients of the original nomogram model. This indicates that the current nomogram is applicable in other high-volume hospitals performing esophagectomies for cancer. The model can give a preoperatively estimated risk of the occurrence and severity of postoperative complications. Furthermore, as we have shown that adequate model performance can be achieved, this model can be used to adjust for case-mix when comparing hospital performances; the nomogram can play a role in the risk-adjusted audit of morbidity after esophagectomy for cancer.

In conclusion, a recently developed nomogram predicting the occurrence and severity of complications was externally validated in a new cohort of patients who underwent esophagectomy for cancer in a different high-volume center. The model showed good overall calibration when applied in the validation cohort. Reestimating the coefficients of the prognostic factors within the nomogram in the validation cohort did not reveal significant improvement compared with the original values. Nevertheless, preoperative prediction of complications in individual patients remains difficult, most likely due to the complexity of mechanisms causing these complications.


   Acknowledgments
Top
 Abstract
 Introduction
 Patients and Methods
 Results
 Comment
 Acknowledgments
References
 
The authors are indebted to Mrs C.M. Vollebregt-Uiterwijk for her dedicated prospective collection of data.


   References
Top
 Abstract
 Introduction
 Patients and Methods
 Results
 Comment
 Acknowledgments
 References
 

  1. Hulscher JB, van Sandick JW, de Boer AG, et al. Extended transthoracic resection compared with limited transhiatal resection for adenocarcinoma of the esophagus N Engl J Med 2002;347:1662-1669.[Medline]
  2. Whooley BP, Law S, Murthy SC, Alexandrou A, Wong J. Analysis of reduced death and complication rates after esophageal resection Ann Surg 2001;233:338-344.[Medline]
  3. Rüdiger Siewert J, Feith M, Werner M, Stein HJ. Adenocarcinoma of the esophagogastric junction: results of surgical therapy based on anatomical/topographic classification in 1,002 consecutive patients Ann Surg 2000;232:353-361.[Medline]
  4. Hulscher JB, Tijssen JG, Obertop H, van Lanschot JJ. Transthoracic versus transhiatal resection for carcinoma of the esophagus: a meta-analysis Ann Thorac Surg 2001;72:306-313.[Abstract/Free Full Text]
  5. 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 2000;119:1126-1132.[Abstract/Free Full Text]
  6. Abunasra H, Lewis S, Beggs L, Duffy J, Beggs D, Morgan E. Predictors of operative death after oesophagectomy for carcinoma Br J Surg 2005;92:1029-1033.[Medline]
  7. Gockel I, Exner C, Junginger T. Morbidity and mortality after esophagectomy for esophageal carcinoma: a risk analysis World J Surg Oncol 2005;3:37.[Medline]
  8. Law S, Wong KH, Kwok KF, Chu KM, Wong J. Predictive factors for postoperative pulmonary complications and mortality after esophagectomy for cancer Ann Surg 2004;240:791-800.[Medline]
  9. 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 2005;201:253-262.[Medline]
  10. Avendano CE, Flume PA, Silvestri GA, King LB, Reed CE. Pulmonary complications after esophagectomy Ann Thorac Surg 2002;73:922-926.[Abstract/Free Full Text]
  11. Ferguson MK, Durkin AE. Preoperative prediction of the risk of pulmonary complications after esophagectomy for cancer J Thorac Cardiovasc Surg 2002;123:661-669.[Abstract/Free Full Text]
  12. Lagarde SM, Reitsma JB, Maris AK, et al. Preoperative prediction of the occurrence and severity of complications after esophagectomy for cancer with use of a nomogram Ann Thorac Surg 2008;85:1938-1945.[Abstract/Free Full Text]
  13. Vergouwe Y, Steyerberg EW, Eijkemans MJ, Habbema JD. Substantial effective sample sizes were required for external validation studies of predictive logistic regression models J Clin Epidemiol 2005;58:475-483.[Medline]
  14. Altman DG, Royston P. What do we mean by validating a prognostic model? Stat Med 2000;19:453-473.[Medline]
  15. Kok TC, van Lanschot JJ, Siersema PD, van Overhagen H, Tilanus HW. Neoadjuvant chemotherapy in operable esophageal squamous cell cancer: final report of a phase III multicenter randomized controlled trial Proc Am Soc Clin Oncol 1997;17:984.
  16. van Heijl M, van Lanschot JJ, Koppert LB, et al. Neoadjuvant chemoradiation followed by surgery versus surgery alone for patients with adenocarcinoma or squamous cell carcinoma of the esophagus (CROSS) BMC Surg 2008;8:21.[Medline]
  17. Omloo JM, Lagarde SM, Hulscher JB, et al. Extended transthoracic resection compared with limited transhiatal resection for adenocarcinoma of the mid/distal esophagus: five-year survival of a randomized clinical trial Ann Surg 2007;246:992-1000.[Medline]
  18. Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey Ann Surg 2004;240:205-213.[Medline]
  19. Hosmer DW, L S. Applied logistic regressionNew York: Wiley; 2000.
  20. Arnold AM, Kronmal RA. Multiple imputation of baseline data in the cardiovascular health study Am J Epidemiol 2003;157:74-84.[Abstract/Free Full Text]
  21. Pulkstenis E, Robinson TJ. Goodness-of-fit tests for ordinal response regression models Stat Med 2004;23:999-1014.[Medline]
  22. Kita T, Mammoto T, Kishi Y. Fluid management and postoperative respiratory disturbances in patients with transthoracic esophagectomy for carcinoma J Clin Anesth 2002;14:252-256.[Medline]
  23. Neal JM, Wilcox RT, Allen HW, Low DE. Near-total esophagectomy: the influence of standardized multimodal management and intraoperative fluid restriction Reg Anesth Pain Med 2003;28:328-334.[Medline]
  24. Flisberg P, Törnebrandt K, Walther B, Lundberg J. Pain relief after esophagectomy: Thoracic epidural analgesia is better than parenteral opioids J Cardiothorac Vasc Anesth 2001;15:282-287.[Medline]
  25. Rudin A, Flisberg P, Johansson J, Walther B, Lundberg CJ. Thoracic epidural analgesia or intravenous morphine analgesia after thoracoabdominal esophagectomy: a prospective follow-up of 201 patients J Cardiothorac Vasc Anesth 2005;19:350-357.[Medline]
  26. Cense HA, Lagarde SM, de Jong K, et al. Association of no epidural analgesia with postoperative morbidity and mortality after transthoracic esophageal cancer resection J Am Coll Surg 2006;202:395-400.[Medline]

Related Article

Invited Commentary
Mark K. Ferguson
Ann. Thorac. Surg. 2010 90: 926. [Extract] [Full Text] [PDF]



This article has been cited by other articles:


Home page
 Ann. Thorac. Surg.Home page
M. E. van Genderen, A. Lima, H. de Geus, E. Klijn, B. Wijnhoven, D. Gommers, and J. van Bommel
Serum C-Reactive Protein as a Predictor of Morbidity and Mortality in Intensive Care Unit Patients After Esophagectomy
Ann. Thorac. Surg., June 1, 2011; 91(6): 1775 - 1779.
[Abstract] [Full Text] [PDF]

Home page
 Ann. Thorac. Surg.Home page
M. K. Ferguson
Invited Commentary
Ann. Thorac. Surg., September 1, 2010; 90(3): 926 - 926.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to Personal Folders
Right arrow Download to citation manager
Right arrow Permission Requests
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Grotenhuis, B. A.
Right arrow Articles by van Lanschot, J. J. B.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Grotenhuis, B. A.
Right arrow Articles by van Lanschot, J. J. B.
Related Collections
Right arrow Esophagus - cancer
Right arrowRelated Article

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
ANN THORAC SURG ASIAN CARDIOVASC THORAC ANN EUR J CARDIOTHORAC SURG
J THORAC CARDIOVASC SURG ICVTS ALL CTSNet JOURNALS