Original article

Risk factor analysis of post-operative mortality in oesophagectomy

J. F. Liu,1 D. I. Watson,1 P. G. Devitt,1 G. Mathew,1 J. Myburgh,2 G. G. Jamieson1

University of Adelaide 1Department of Surgery and 2Intensive Care Unit, Royal Adelaide Hospital, Adelaide,South Australia, Australia

SUMMARY. Oesophagectomy for cancer is associated with a signi®cant incidence of post-operative complica-tions and death, and so this study sought to determine objective criteria which could better predict operative risk.Clinical risk factors for oesophagectomy and the results of objective investigations were assessed prospectively byindependent surgical and intensive care specialists and a multivariate analysis was used to develop a scoring systemfor predicting operative risk. From September 1994 to June 1997, 32 patients from an overall experience of 70oesophagectomy procedures for cancer at the Royal Adelaide Hospital were entered into this study. Hypertension,a history of previous cigarette smoking and FEV1/FVC were identi®ed as independent predictors of the post-operative outcome. Age and FEV1/FVC were also signi®cantly associated with the occurrence of cardiovascularand pulmonary complications respectively. The average risk score was 4.8 � 4.5 (mean � SD) for patients whodied, 2.9 � 2.9 for patients who developed post-operative complications and 2.6 � 2.1 for patients who had anuncomplicated recovery. The likelihood of post-operative mortality and morbidity was highest in patients with ascore of 5 or more. Mortality rates of di�erent patient groups undergoing oesophagectomy by the same surgeonsduring the same time period were also compared, showing greatly di�erent mortality rates. Important risk factorscan be identi®ed preoperatively, and a scoring system can be used to provide objective criteria which can be used toidentify patients at an increased risk of post-operative complications and death. A prospective study of this scoringsystem is now needed to determine whether it proves useful in rejecting patients for surgery who would otherwisehave undergone oesophagectomy.

INTRODUCTION

Despite the recent development of non-operativetreatment methods for oesophageal cancer, forexample radiotherapy, chemotherapy, laser therapyand oesophageal stenting, surgical resection contin-ues to o�er the best means of restoring swallowingand sometimes e�ecting a cure. However, oesopha-geal cancer usually occurs in older patients who oftenhave coexisting diseases, particularly of the heart andlungs, and these co-morbidities can impair thetolerance of patients to the trauma of oesophagec-tomy. For this and other reasons, oesophagectomyhas a high 30-day mortality rate, ranging up to 16.0%in recent reports.1,2 We have recently published areport which detailed a 30-day mortality rate of 9%

and an in-hospital mortality rate of 13% for oeso-phagectomy.3

In recent years, there have been some analyses ofpreoperative risk factors.4,5 However, there remains aneed to develop clinically practical methods for theassessment of operative risk. The present studyprospectively evaluated possible risk factors foroesophagectomy and applied a multivariate analysisto determine which risk factors were important. Thisfacilitated the development of a composite scoringsystem and an overall risk index for the determin-ation of the likelihood of post-operative complica-tions and death after oesophagectomy.

MATERIAL AND METHODS

Patients

From September 1994 to June 1997, an oesophagec-tomy was performed in 70 patients with oesophagealcarcinoma at the Royal Adelaide Hospital. From thisoverall experience, 32 patients were entered into the

Address correspondence to: Mr. David I Watson, University ofAdelaide Department of Surgery, Royal Adelaide Hospital,Adelaide, South Australia 5000, Australia.Tel: (+61)8 82225516; Fax: (+61)8 82225896;E-mail: david.watson@adelaide.edu.au

130

Diseases of the Esophagus (2000) 13, 130±135Ó 2000 ISDE/Blackwell Science Asia

current study, with a tendency for higher risk patientsto be entered. The study group comprised 22 men and10 women, aged between 44 and 81 years (median66). An Ivor Lewis oesophagectomy was performedin 29 patients, a transhiatal oesophagectomy in 2 anda three-stage cervicothoracoabdominal oesophagealresection with right colon interposition in 1. In thestudy group, 11 patients recovered without compli-cations, 13 patients recovered after the developmentof complications of varying severity and eightpatients died in hospital (Table 1). This selectedexperience re¯ected a worse outcome than our overallexperience with oesophagectomy during the period(see later).

During the same period, the same group ofsurgeons undertook 26 oesophagectomies in anassociated hospital where the intensive care unit sta�were also from the Royal Adelaide Hospital. Thecriteria for selecting this group of patients foroesophagectomy was the same as for the 70 RoyalAdelaide Hospital patients, including the 32 patientsin this study.

Methods

Patients in the study were seen and assessed preop-eratively by an experienced oesophageal surgeon (notthe operating surgeon) and an intensive care special-ist. Both of these clinicians independently assessedeach patient's general ®tness for surgery and identi-®ed potential preoperative risk factors. The clinicianswere asked to rate independently each patient's riskof complications and death after surgery, with theirassessment based initially on this clinical assessmentand then subsequently on the outcome of theobjective investigations. Preoperative pulmonaryfunction studies and echocardiography (to measurecardiac ejection fraction) were performed routinely.Patients were only included in this study if bothclinical assessments had been performed and a fullcomplement of preoperative investigations had beenundertaken. This protocol resulted in a cohort ofhigher-risk patients being selectively included in thestudy.

Seventeen parameters were selected as possible riskfactors on the basis of our previous experience withoesophagectomy for cancer or because they had beenreported to be important by others. These are listed inTable 2. Data were analysed by multivariate analysisto determine which individual risk factors wereimportant. Data were also collated and scoredaccording to the scoring criteria outlined in Table 2to provide a composite risk assessment score foroesophagectomy.

The operative outcome was classi®ed into threegroups for subsequent analysis: uncomplicatedrecovery, complicated recovery and death (Table 3).This information was used to determine an outcomescore based on the scoring system described in Table 3.

Table 1. Post-operative complications and death (32 patientsin study)

Complications 21Respiratory failure/ARDS 10Pneumonia 8Cardiac failure or myocardial infarction 3Cardiac arrhythmia 3Anastomotic leak 3Bronchopleural ®stula 1Polymyopathy 1Panophthalmitis 1Necrotizing pancreatitis 1

Mortality 8ARDS/multiorgan failure 4Cardiac failure/myocardial infarction 3Pancreatitis 1

Table 2. Preoperative and operative criteria used for scoring overall risk (the sum of all positive factors entailed the `risk score')

Score

0 1 2 3

Age (years) <70 70±75 75±80 >80

Coexisting diseaseCardiovascular disease Nil Mild arrhythmia

HypertensionValve disease withoutsymptomsAortic stenois

AnginaOld infarction

Respiratory disease Nil SmokerMild problem, i.e.occasional cough

Heavy smokerAsthma Chronicbronchitis

Chronic infectious disease,i.e. tuberculosis

Other disease Nil Mild problems, i.e.previous operation,alcohol, obesity

Moderate problems, i.e.signi®cant weight loss,diabetes, general debility

Severe problem, i.e.anasarca, cachexia

FEV1/FVC >80% 71±80% 60±70% <60%

FEV1 >70% 61±70% 50±60% <50%

Cardiac ejection >55% 50±55% 40±49% <40%

ECG Normal Mild arrhythmia Ischaemia Old infarction

Liver function (Alb) >35 31±35 21±30 <20

Operation Transhiatal resection Ivor Lewisprocedure

Three-stageoesophagectomy

Post-operative mortality in oesophagectomy6 131

This outcome score was then used to facilitatesubsequent comparison with the preoperative riskassessment. The di�erences between the risk scores foreach of the three outcome groups were also compared,and the relationship between individual preoperativerisk factors and outcome were determined.

Data are reported as means � standard deviation.Relationships between preoperative and operativefactors and the outcome of surgery were determinedby multivariate linear regression analysis. One-wayanalysis of variance (ANOVA) was used to assess thedi�erences between the three groups, and the rela-tionship between the preoperative risk score andsubsequent outcome was estimated using the chi-squared test. A linear correlation was used to analysethe ability of the clinicians to predict outcome. Allstatistical analyses were performed using STATVIEW

software (Abacus Concepts, CA, USA). A p-value ofless than 0.05 was considered to be statisticallysigni®cant.

RESULTS

Of the overall experience of 70 oesophagectomiesduring the study period, ®ve patients died within 30days of surgery (30-day mortality 7%) and a furtherfour patients died in hospital more than 30 days aftersurgery (hospital mortality 13%). The study groupwas a subset of 32 patients from this overall experi-

ence, and eight of the nine hospital deaths occurredwithin the study group. In the 26 additional oeso-phagectomies performed in the associated hospital,there was no in-hospital mortality.

The results of multivariate analysis demonstrateda highly signi®cant association between hyperten-sion and a poor outcome. It was also associatedwith a high risk of post-operative cardiac andrespiratory complications, as well as non-cardiacand non-respiratory complications. A history ofcigarette smoking and an abnormal FEV1/FVCratio were also associated with poor outcome. TheFEV1/FVC ratio was also associated with thelikelihood of post-operative pulmonary complica-tions. Age was associated with cardiovascularcomplications only. Overall results for the multi-variate analysis are summarized in Table 4.

The average preoperative risk score for patientswho died in hospital after oesophagectomy was 4.8,compared with 2.9 in patients who had a complicatedrecovery and 2.6 in those without any complications(p � 0.02, Table 5). Post-operative mortality andmorbidity were higher in the patients with a scoreof 5 or more (p � 0.008, Table 6).

The clinicians, both the surgeon and the intensivecare specialist, performed poorly as assessors ofperioperative risk. No correlation existed betweentheir clinical assessment and the likelihood of post-operative complications, including death (r � 0.12,p � 0.52).

Table 3. Classi®cation of post-operative results

Score Overall results Cardiovascular complications Pulmonary complications Other complications

0 Uncomplicated recovery Recovered withoutcardiac complications

Recovered withoutpulmonary complications

Recovered without othercomplications

1 Complicated recovery Recovered with cardiaccomplications

Recovered with pulmonarycomplications

Recovered with othercomplications

2 Hospital death Died of cardiaccomplications

Died of pulmonarycomplications

Died of othercomplications

5Table 4. Multivariant analysis of the relationship between preoperative and operative `risk factors' and post-operative outcome

p-value

Risk factors Overall results Cardiac complications Respiratory diseases Other complications

Age 0.45 0.02 0.07 0.50Artrial ®brillation 0.07 0.92 0.08 0.78Hypertension <0.0001 <0.0001 0.0008 0.0008Smoker 0.046 0.51 0.33 0.44Asthma 0.27 0.11 0.08 0.30Weight loss 0.56 0.07 0.17 0.09Diabetes 0.27 0.11 0.08 0.30Generalized debility 0.40 0.67 0.44 0.83Obesity 0.27 0.11 0.08 0.30Alcohol 0.53 0.97 0.48 0.65FEV1 0.91 0.13 0.65 0.89FEV1/FVC 0.049 0.50 0.01 0.76ECG 0.81 0.19 0.66 0.25Liver function 0.51 0.31 0.67 0.43Cardiac ejection 0.07 0.30 0.29 0.29Previous HH operation 0.61 0.37 0.37 0.55Operation type 0.12 0.32 0.12 0.69

HH, hiatus hernia.

132 Diseases of the Esophagus

DISCUSSION

The perioperative mortality for patients undergoingoesophagectomy for cancer has been reduced signi-®cantly in recent years, with improvements in preop-erative preparation, operative and anaesthetictechniques and post-operative care. Nevertheless,mortality after oesophagectomy remains a signi®cantproblem, with hospital mortality re¯ecting bothnumbers of procedures being undertaken in aninstitution and the degree of selection applied indi�erent hospitals. We have used a liberal selectionpolicy, believing that oesophagectomy provides thebest palliation and sometimes leads to cure. The onlypatients not considered for oesophagectomy are thoseolder than age 85 years, those with a cardiac ejectionfraction of less than 40% and those with an FEV1 ofless than 1 l. Therefore, we were well placed toexamine factors which in¯uence the risk of mortalityafter oesophagectomy. A number of analyses ofoperative risk factors have been reported previ-ously,2,4±7 although there has been some variationbetween the outcomes of these reports. A possiblereason for this is the fact that multiple factors, ratherthan a single factor, a�ect the risk of mortality in thispatient group, and this can make di�cult preopera-tive assessment of risk. Furthermore, some patientsdie from problems which are di�cult to predictbefore surgery, e.g. anastomotic leakage. Therefore,we cannot rely on a single factor to determineaccurately the risk of a poor outcome after oeso-phagectomy.

Age has been reported to correlate strongly withthe outcome of surgery in some reports,2,5,6 but otherreports suggest that it is not an important riskfactor.7±10 A possible reason for this variation is thatclinicians may apply di�erent criteria to screen olderpatients before surgery than others.8 If, in a centre,only very ®t elderly patients are selected for oeso-phagectomy, then age is unlikely to be a risk factor if

outcome data are analysed from such a centre.Because we have deliberately used a relatively unse-lective approach to oesophagectomy, in the presentstudy age correlated signi®cantly with the risk ofcardiovascular complications and death. This selec-tion policy no doubt also contributed to the highmortality in the patient group assessed in this study,although the 30-day mortality after oesophagectomyin our overall experience at the Royal AdelaideHospital was an acceptable 7%.

The high mortality in the study group initiallypreplexed us. However, the fact that 8 of 32 patientsin the study group died compared with only 1 of 38patients not in the study group suggests selectionbias. Anecdotally at least, we believe that in youngerand ®tter patients, in whom problems were notexpected, the surgeon sometimes had less opportunityto enter these patients into the study as they tended tobe admitted closer to the time of their surgery.Additionally, the surgeons tended to forget to entersome of these patients into the study. When patientswere older and less ®t, mortality was uppermost inthe surgeon's minds and patients were usually enteredin the assessment study. The problem of makingcomparisons of mortality rates across series isunderlined by the fact that in an associated hospitalduring the same time period 26 cases underwentoperation without hospital mortality, with the samesurgeons undertaking the operations. So, thein-hospital mortality rate for the group of surgeonsinvolved can be recorded in several ways, dependingon which patients are included. For this study, anin-hospital mortality rate of 25% can be derived.However, if we take all patients undergoing oeso-phagectomy by this surgical group at the RoyalAdelaide Hospital, the in-hospital mortality rate was13%. Furthermore, if we take all patients undergoingoesophagectomy by this group for this time period,the in-hospital mortality was 9%, and if the groupwished to publish its associated hospital experienceonly for the same period the in-hospital mortality ratewas 0%.

The results of this analysis demonstrated a signi-®cant association between preoperative hypertensionand smoking and post-operative complications.These ®ndings di�er from the results of some otherreports.2,7 Bartels et al7 reported that tobacco use didnot in¯uence post-operative outcomes, whereas

Table 5. Risk scores for di�erent outcome groups

n Accumulated scores

Died 8 4.9 � 2.2Complicated recovery 13 2.8 � 1.7Uncomplicated recovery 11 2.6 � 1.4

p-value 0.02

Table 6. The relationship between risk score and outcome

ScoreNumber ofpatients

Number ofdeaths (%)

Complication(including deaths) (%)

Uncomplicatedrecovery (%)

>5 8 4 (50) 8 (100) 0 (0)3 ± 4 11 3 (27) 7 (64) 4 (36)1 ± 2 13 1 (8) 6 (46) 7 (54)

Total 32 8 21 11p-value (v2 for trend) 0á03 0á01 0á01

Post-operative mortality in oesophagectomy6 133

Ferguson et al2 failed to show any relationshipbetween pre-existing hypertension and post-operativecomplications. In our experience, however, smokerswere more likely to have obstructive respiratorydiseases, and this can result in post-operative respi-ratory problems, including bronchospasm and spu-tum retention. Hypertension often coexists with othercardiovascular diseases such as myocardial ischaemiaand cardiac arrhythmias. These pre-existing problemsmay impair the tolerance of the patients to extensiveoperations such as oesophagectomy. Of the eightpatients who died in our study, three had pre-existinghypertension and six were heavy smokers. For thisreason, pre-existing co-morbidities should be carefullyassessed preoperatively, and the risks of surgeryshould be carefully balanced against the potentialbene®ts before a ®nal decision is made to proceed tooesophagectomy.

To improve selection for surgery further, thefunctional reserve of some organs can be objectivelyassessed. Our study demonstrated that the FEVI/FVC ratio correlated signi®cantly with the likelihoodof post-operative complications and death. This issimilar to results reported by Law et al4 More than60% of the post-operative deaths after oesophagec-tomy result from pulmonary and cardiac complica-tions.6,11±13 Hence, routine preoperative pulmonaryand cardiac function tests should be performed forthe patients who are being assessed for ®tness foroesophagectomy.

Transhiatal oesophagectomy is considered bysome to be an operative procedure which isassociated with a reduced operative risk,14 althoughour study did not include su�cient patients under-going a transhiatal procedure to demonstrate this.Rindani et al15 and Berdejo16 reported a loweroperative mortality in the patients undergoingtranshiatal oesophagectomy compared with trans-thoracic resection, although other studies have notreported a reduction in the incidence of cardiovas-cular and pulmonary complications after transtho-racic oesophagectomy.17,18 Recently, however, theuse of thoracoscopic surgery has been advocated toreduce the risk of oesophagectomy.19±21 Whetherthis will provide a better solution remains uncer-tain.

In the present study, we asked experiencedspecialists, a surgeon and an intensive care special-ist, to assess independently the likelihood of a pooroutcome and complications following surgery. Wehypothesized that the assessment of an experiencedspecialist would be adequate for the preoperativeassessment of patients undergoing oesophagectomy.Unfortunately, the study did not validate this,con®rming the importance of using more objectivecriteria to assess the suitability of patients for thissurgical procedure. Hence, the impetus for thedevelopment of an objective scoring system, and

this proved to be more reliable for the prediction ofpost-operative results. In particular, a patient witha preoperative risk score of 5 or more is a poorrisk for surgery and alternative non-operativetherapies should perhaps be given stronger consid-eration than is our current practice, in which weview oesophagectomy as essentially a procedure toallow patients to swallow normally until death,accepting that approximately 10% will die as aresult of their operation and about 25% will bealive at 5 years.

Acknowledgements

J.F.L. was supported by a scholarship from theInternational Society for Diseases of the Esophagus,and G.M. was supported by a grant from theBurnside War Memorial Hospital, Burnside, SouthAustralia.

References

1. Patti M G, Corvera C U, Glasgow R E, Way L M. A hospital'sannual rate of oesophagectomy in¯uences the operative mor-tality rate. J Gastrointest Surg 1998; 2: 186±192.

2.4 Ferguson M K, Martin T R, Reeder L B, et al. Mortality afteroesophagectomy: risk factor analysis. World J Surg 1997; 21:599±603.

3. Bessell J R, Devitt P G, Gill P G, Goyal S, Jamieson G G.Prolonged survival follows resection of oesophageal SCCdownstaged by prior chemoradiotherapy. Aust NZ J Surg1996; 66: 214±217.

4. Law S Y, Fok M, Wong J. Risk analysis in resection ofsquamous cell carcinoma of the esophagus. World J Surg 1994;18: 339±346.

5. Zhang G H, Fujita H, Yamana H, et al. Preoperativepredication of mortality following surgery for esophagealcancer. Kurume Med J 1992; 39: 159±165.

6. Tsutsui S, Moriguchi S, Morita M, et al. Multivariate analysisof postoperative complications after esophageal resection. AnnThorac Surg 1992; 53: 1052±1056.

7. Bartels H, Stein H J, Siewert J R. Preoperative risk analysisand postoperative mortality of oesophagectomy for resectableoesophageal cancer. Br J Surg 1998; 85: 840±844.

8. Alexious C, Beggs D, Salama F D, et al. Surgery foresophageal cancer in elderly patients: the view from Notting-ham. J Thorac Cardiovasc Surg 1998; 116: 545±553.

9. Jougon J B, Ballester M, Du�y J, et al. Oesophagectomy forcancer in the patient aged 70 years and older. Ann Thorac Surg1997; 63: 1423±1427.

10.3 Kuwano H, Sumiyashi K, Sonoda K, et al. Relationshipbetween preoperative assessment of organ function and post-operative morbidity in patients with oesophageal cancer. Eur JSurg 1998; 164: 581±586.

11. Isono K, Onoda S, Ishikawa T, Sato H, et al. Studies on thecauses of death from esophageal carcinoma. Cancer 1982; 49:2173±2179.

12. Nishi M, Hiramatsu Y, Hioki K, et al. Pulmonary complica-tions after subtotal oesophagectomy. Br J Surg 1988; 75: 527±530.

13. Postethwaite R W. Complications and deaths after operationsfor esophageal carcinoma. J Thorac Cardiovasc Surg 1983; 85:827±831.

14. Hirai T, Kuwashara M, Yoshida K, et al. Clinical results oftranshiatal oesophagectomy for carcinoma of the lower tho-racic esophagus according to biological markers. Dis Esoph1998; 11: 221±225.

15.11 Rindani R, Martin C J, Cox M R. Transhiatal versus Ivor-Lewis oesophagectomy: is there a di�erence? Aust NZ J Surg1999; 69: 187±194.

134 Diseases of the Esophagus

16.2 Berdejo L. Transhiatal versus transthoracic oesophagectomyfor clinical stage I esophageal carcinoma. Hepatogastroen-terology 1995; 42: 789±791.

17. Gluch L, Smith R C, Bambach C P, et al. Comparison ofoutcomes following transhiatal or Ivor-Lewis oesophagectomyfor esophageal carcinoma. World J Surg 1999; 23:271±275.

18. Hostmann O, Verreet R R, Becker H, et al. Transhiataloesophagectomy compared with transthoracic resection and

systemic lymphadenectomy for the treatment of oesophagealcancer. Eur J Surg 1995; 161: 557±567.

19. Dexter S P, Martin I G, McMahonM J. Radical thoracoscopicoesophagectomy for cancer. Surg Endosc 1996; 10: 147±151.

20. Swanstrom L L, Hansen P. Laparoscopic total oesophagec-tomy. Arch Surg 1997; 132: 943±949.

21. Yahata H, Sugino K, Takiguchi T. Laparoscopic transhiataloesophagectomy for advanced esophageal cancer. Surg Lapa-rosc Endosc 1997; 7: 16±26.

Post-operative mortality in oesophagectomy6 135