ORIGINAL ARTICLE

Hospital and surgeon volume in relation tolong-term survival after oesophagectomy: systematicreview and meta-analysisNele Brusselaers,1 Fredrik Mattsson,1 Jesper Lagergren1,2

▸ Additional material ispublished online only. To viewplease visit the journal online(http://dx.doi.org/10.1136/gutjnl-2013-306074).1Upper GastrointestinalSurgery, Department ofMolecular Medicine andSurgery, Karolinska Institutet,Stockholm, Sweden2Division of Cancer Studies,King’s College London, GeneralSurgery Offices, St Thomas’Hospital, London, UK

Correspondence toDr Nele Brusselaers, UpperGastrointestinal Surgery,Department of MolecularMedicine and Surgery, NorraStationsgatan 67, Level 2,Karolinska Institutet, StockholmSE-171 76, Sweden;nele.brusselaers@ki.se

Received 11 September 2013Revised 30 October 2013Accepted 10 November 2013Published Online First22 November 2013

To cite: Brusselaers N,Mattsson F, Lagergren J. Gut2014;63:1393–1400.

ABSTRACTBackground Centralisation of healthcare, especially foradvanced cancer surgery, has been a matter of debate.Clear short-term mortality benefits have been describedfor oesophageal cancer surgery conducted at high-volume hospitals and by high-volume surgeons.Objective To clarify the association between hospitalvolume, surgeon volume and hospital type in relation tolong-term survival after oesophagectomy for cancer, by ameta-analysis.Design The systematic literature search includedPubMed, Web of Science, Cochrane library, EMBASE andScience Citation Index, for the period 1990–2013.Eligible articles were those which reported survival (timeto death) as HRs after oesophagectomy for cancer byhospital volume, surgeon volume or hospital type. Fullyadjusted HRs for the longest follow-up were the mainoutcomes. Results were pooled by a meta-analysis, andreported as HRs and 95% CIs.Results Sixteen studies from seven countries met theinclusion criteria. These studies reported hospital volume(N=13), surgeon volume (N=4) or hospital type (N=4).A survival benefit was found for high-volume hospitals(HR=0.82, 95% CI 0.75 to 0.90), and possibly also, forhigh-volume surgeons (HR=0.87, 95% CI 0.74 to 1.02)compared with their low-volume counterparts. Noassociation with survival remained for hospital volumeafter adjustment for surgeon volume (HR=1.01, 95% CI0.97 to 1.06; N=2), while a survival benefit was foundin favour of high-volume surgeons after adjustment forhospital volume (HR=0.91, 95% CI 0.85 to 0.98; N=2).Conclusions This meta-analysis demonstrated betterlong-term survival (even after excluding early deaths)after oesophagectomy with high-volume surgery, andsurgeon volume might be more important than hospitalvolume. These findings support centralisation with fewersurgeons working at large centres.

INTRODUCTIONCentralisation of complex cancer surgery is a topicof debate in several countries. Such centralisationcan improve care by collating multidisciplinaryexpertise and experience, as well as specialisedequipment, within centres of excellence, and willaffect the healthcare budget. Treating more patientsshould improve the skills of the medical team, andadapting specific treatment procedures shouldfacilitate and improve patient-tailored care.However, the benefit of such centralisation shouldbe weighed up against the potential disadvantagesfor patients— for example, long travel times and

social isolation.1 Oesophageal cancer surgery is oneof the most complex surgical procedures, entailinga substantial risk of severe postoperative complica-tions, and a convincing benefit of centralisation hasbeen shown for short-term mortality (in-hospitaland 30-day postoperatively) for this operation.2–7

However, the existing data describing oesophagealcancer surgery volume in relation to long-term sur-vival are limited, and the results from individualstudies are contradictory.7 Moreover, the influenceof tumour stage has not always been taken into

Significance of this study

What is already known about this subject?▸ Oesophageal cancer has a very poor prognosis,

with a 5-year survival of only 30% aftercomplete oesophageal cancer resection(oesophagectomy).

▸ Oesophagectomy is one of the most complexsurgical procedures, entailing a substantial riskof severe postoperative complications.

▸ Short-term mortality (in-hospital and 30-daypostoperatively) has been shown to be lower ifoperations are carried out in high-volumehospitals and/or by high-volume surgeons. Yet,95% of all patients survive the first 30 daysafter surgery.

▸ Based on these differences in short-termmortality, centralisation of oesophagectomywithin centres of excellence has beenrecommended and adopted in many countries.

What are the new findings?▸ This meta-analysis demonstrated a better

long-term survival after oesophagectomy withhigh-volume surgery.

▸ Surgeon volume seems to be more importantthan hospital volume.

How might it impact on clinical practice inthe foreseeable future?▸ Differences in long-term survival imply an ever

greater benefit of centralisation for the numberof patient-years than already presumed basedon short-term mortality, since the mortalityfollowed by such surgery is strongly dominatedby deaths from tumour recurrence rather thanprocedure-related complications.

Brusselaers N, et al. Gut 2014;63:1393–1400. doi:10.1136/gutjnl-2013-306074 1393

Oesophagus

group.bmj.com on August 6, 2014 - Published by gut.bmj.comDownloaded from

account in previous research, which is of major concern when-ever long-term mortality is the outcome. Thus, the impact ofsurgery volume on long-term survival after oesophageal cancerremains to be established. Such knowledge would be of clinicalimportance since tumour recurrence in oesophageal cancer iscommon, resulting in a risk of death of about 60–70% within5 years of surgery.8 9 The objective of this study was to clarifythe association between hospital volume, surgeon volume andhospital type in relation to long-term survival after oesophagect-omy for cancer, by means of a meta-analysis.

METHODSThis was a systematic review and meta-analysis analysing differ-ences in long-term survival between high-volume and low-volume hospitals and surgeons after oesophagectomy for cancer.The available literature was identified and examined by a sys-tematic review and survival meta-analysis. The results arereported in accordance with the PRISMA guidelines (PreferredReporting Items for Systematic Reviews and Meta-Analyses).10

The study followed an a priori established study protocol.

Exposure and outcomeThe main exposure was surgery performed in either low- orhigh-volume hospitals, as defined by the authors of the includedstudies. Other examined exposures were surgeon volume (lowor high) and type of hospital (eg, university or non-university).Both surgeon and hospital volume were measured as the averageannual number of oesophagectomies per year (eg, >10 or <10procedures/year). The group with the lowest volume or non-university was used as reference category. If an included studyreported HRs for different surgical volume groups, only thelowest and highest volume group were compared and reportedin the Forest plots.

The study outcome was the time to death after the oesopha-gectomy, defined by a minimal period of follow-up of the studycohort of 3 months.

Data sources and searchesThe primary data sources screened were PubMed and the Webof Science. The search string in these databases consisted of fourparts: (1) the anatomical location of interest (ie, oesophagus,oesophagectomy, and oesophageal), (2) surgery, surgical orcancer, (3) outcome, mortality, survival or prognosis and (4)volume, determinants or predictors. Different spelling wasaccounted for, and medical subheadings (MeSH) were incorpo-rated in the PubMed search. Complementary searches wereperformed through analyses of reference lists, the ScienceCitation index, Cochrane library, EMBASE and searching forrelevant publications of ‘expert’ authors (known or identified tohave published in the field of surgery volume).

Study selectionThe time period for publications was limited to January 1990–September 2013, a period which we considered to representmodern oesophageal cancer management. The search methodused to identify all relevant articles was discussed and developedby two authors (NB and JL) and the final search string wasapproved by all authors. The initial search was performed byone reviewer (NB), who eliminated clearly irrelevant articlesbased on the title and abstract as defined by the pre-set selectioncriteria. The final selection of articles was made by mutual con-sideration of all authors, based on the reporting of all necessarydata and in accordance with the predefined inclusion and exclu-sion criteria.

Inclusion and exclusion criteriaStudies that provided original data on survival of patients whounderwent oesophagectomy for malignancy were included.Abstracts or other conference proceedings, case reports, caseseries, intervention studies and review articles were excluded.Both prospective and retrospective studies were eligible. Articlesdescribing oesophagectomy for non-malignant reasons wereexcluded, as were studies reporting a subgroup of oesophagect-omy patients only. If studies also reported survival after gastriccancer surgery, survival for oesophageal cancer had to bereported separately, otherwise the study was excluded. Languagerestriction was applied only in the end stage of the search, toenable assessment of language selection bias. The languagesselected a priori as eligible were English, French, Dutch,German, Spanish, Swedish and Chinese. Studies were eligibleonly if HRs comparing survival after oesophagectomy by hos-pital or surgeon volume groups, or by hospital type werereported. The minimum reported follow-up time was 3 months.

Data extraction and quality assessmentThe following data were collected (if available): study and popu-lation characteristics, type of surgery and hospital characteristics.Assessment of quality and generalisability was based on the keydomains considered fundamental for observational studies.11

From each article, the crude HRs were extracted (if reported),and the HRs based on the most fully adjusted regression modelsfor the longest duration of follow-up. If several volume groupswere reported, the most extreme comparison —that is, highestversus lowest reported volume, was considered the primaryresult. If possible, the HR for hospital volume adjusted forsurgeon volume was extracted as well as the most fully adjustedmodel without adjustment for surgeon volume. The sameapproach was taken for surgeon volume and adjustment for hos-pital volume. If HRs were reported including and excluding‘early’ mortality, defined as within 3 months of surgery, bothHRs were extracted, but the HR including the full follow-upperiod was considered the main result.

Data synthesis and analysisThis survival meta-analysis pooled the HRs based on hospitalvolume, surgeon volume and hospital type. Subanalysis wasbased on duration of follow-up, inclusion or exclusion of earlymortality and reported regression models. Random-effectmeta-analyses were performed with STATA (StataCorp, V.12·1/MP4), and were based on the HRs and SEs. The values werereported by a Forest plot, and uncertainty about the pooled esti-mates was quantified by 95% CI. If no SE was reported, it wascalculated based on the 95% CI, number of patients or reportedp values.12 13 The presence of small study effects and publica-tion bias was evaluated by funnel plots and Egge’s regressionasymmetry analysis.14 Statistical heterogeneity was assessed bymeans of Cochran’s Q test and I2 test. I2 represents the percent-age of variation attributable to heterogeneity, which is usuallycategorised as low (25–50%), moderate (51–75%) or high(>75%).15

RESULTSDescription of the included studiesThe search was finalised on 10 September 2013, and included a totalof 2392 publications as shown in the online supplementary appendix.The search resulted in 16 eligible studies. Only one potentially eligibleJapanese study was excluded because of the language restriction.16

The eligible studies were from the USA (N=4),1 17–19 Sweden

1394 Brusselaers N, et al. Gut 2014;63:1393–1400. doi:10.1136/gutjnl-2013-306074

Oesophagus

group.bmj.com on August 6, 2014 - Published by gut.bmj.comDownloaded from

Table 1 Study and population characteristics of all 16 included studies dealing with survival after oesophageal cancer surgery

Author Year Region Cohort

Studyinclusionperiod

Studyduration,years

Age,years

Male,%

Adeno-carcinoma, %

Squamous cellcarcinoma, % Exposure Mortality

Survivalafter

Maximumfollow-uptime, years

Hospitals,N

Oesophagectomies,annual maximumper hospital

Surgeons,N

Oesophagectomies,annual maximalper surgeon

Coupland28 2013 England Population-based 2004–2008 5 nr 72* – – HV All cause Surgery >1 144 132 nr nr

Derogar20 2013 Sweden Population-based 1987–2005 15 Median66

74 36 60 SV, HV All cause Surgery 23 nr ≥17 nr ≥10

Dikken (1)24 2012 Netherlands Population-based 1989–2009 21 nr 76 75 23 HV All cause Diagnosis 3 44 >20 nr nr

Dikken (2)25 2012 Netherlands Population-based 1989–2009 21 nr 76 75 23 HT All cause Diagnosis 3 44 >20 nr nr

van dePoll-Franse26

2011 South Netherlands Population-based 1995–2006 12 Median66

73–75 66–71 22–29 HV All cause Diagnosis 3 >12 15–20 nr 15–20

Cheung19 2010 Florida, USA Population-based 1998–2002 5 Median69†

75† 48† 45† HT All cause nr 12? nr nr nr nr

Stavrou29 2010 NewS-Wales, Australia

Population-based 2000–2005 6 nr 74 nr nr HV Fromcancer

Diagnosis 3 nr >20 nr nr

Bilimoria1 2008 USA Population-based 1994–1999 6 Median64–65

nr nr nr HV All cause Surgery 5 1154 >15 nr nr

Sundelof22 2008 Sweden Population-based 1994–1997 4 nr 83 nr nr SV, HV All cause Surgery 5 33 ≥10 nr ≥10Birkmeyer17 2007 USA Hospital/

NCI-centre based1992–2002 11 nr 74–79 nr nr HV All cause Surgery 5 206 107 nr nr

Ioka31 2007 Osaka, Japan Population-based 1994–1998 5 nr nr nr nr HV All cause Diagnosis 5 143 >43 nr nr

Rouvelas21 2007 Sweden Population-based 1987–2000 14 Mean65–66

71–72 26–29 59–66 HV All cause Surgery 5 22–30 ≥10 nr nr

Simunovic30 2006 Ontario, Canada Population-based 1990–2000 11 Median63–65

nr nr nr HV, HT All cause Admissionfor surgery

10? 68 ≥44 nr nr

Birkmeyer18 2005 USA Hospital/NCI-centre based

1994–1999 6 nr 75–77 nr nr HT All cause Surgery 5 102 107 nr nr

Wenner23 2005 Sweden Population-based 1987–1996 10 Median66–67†

75–77†

77† 23† HV All cause Surgery 5 74 >20 nr 34

Bachmann27 2002 South and WestEngland

Hospital-based 1996–1997 1 nr nr nr nr SV, HV All cause Surgery 3 23 83 nr nr

*Also including patients with gastric carcinoma.†Also including non-operated patients.HT, hospital type; HV, hospital volume; nr, not reported; SV, surgeon volume .

BrusselaersN,etal.G

ut2014;63:1393–1400.doi:10.1136/gutjnl-2013-306074

1395

Oesophagus

group.bm

j.com on A

ugust 6, 2014 - Published by

gut.bmj.com

Dow

nloaded from

(N=4),20–23 the Netherlands (N=3),24–26 the UK (N=2),27 28

Australia (N=1),29 Canada (N=1)30 and Japan (N=1).31 The longestreported follow-up was 23 years.20 Thirteen studies reported HRsfor survival by hospital volume groups (n=39 761),1 17 20–24 26–31

four by surgeon volume (n=2874),20 22 26 27 and four by hospitaltype (n=13 433).18 19 25 26 Two of the three Dutch studies describedthe same nationwide cohort, once for hospital volume and once forhospital type,24 25 and there was some overlap with the third(regional) study.26 The studies from the UK did not overlap.27 28

Some overlap was possible in the four American studies.1 17–19 Thestudy periods of the three oldest (nationwide) Swedish studies partlyoverlapped.21–23 An overview of the main study characteristics is pre-sented in table 1, and a quality assessment is shown in figure 1. Allbut three studies 17 18 27 were population-based, and 10 studies werenationwide. Eight studies also described other cancer types. Mean ormedian age of patients undergoing the operation ranged from 63 to66 years, as reported in four studies.1 20 21 30 The proportion of malepatients, as reported in nine studies, ranged from 71% to 83%. Fourstudies reported the proportion of the main histological types: adeno-carcinoma (26–75%) and squamous cell carcinoma (22–66%).20 21

24 25

Hospital volume and long-term survivalAll 13 studies dealing with hospital volume reported adjustedHRs. One study considered hospital volume as a continuous vari-able (HR for an increase of 10 oesophagectomies per year) andreported only HRs adjusted for surgeon volume.27 Another studyreported HRs both including and excluding adjustment forsurgeon volume.20 Ten studies reported HRs for the complete sur-vival period, two of which also reported HRs excluding mortalitywithin the first 2 months of surgery,1 or as ‘survived surgery’ (notspecified).17 Three studies reported only HRs excluding early mor-tality— that is, the first 2,23 3,20 or 624 postoperative months.20

Six studies adjusted for tumour stage.1 17 20 22 24 31

The pooled adjusted HRs of mortality in 12 studies (exclud-ing the study adjusting for surgeon volume)27 was 0.82 (95% CI0.75 to 0.90) in favour of high-volume hospitals (figure 2). Thestatistical heterogeneity was moderate (I2=68.0%). Subanalysesare presented in table 2, showing a pooled HR of 0.76 (95% CI0.68 to 0.84) in studies adjusting for tumour stage. The seven

studies with the longest complete follow-up (over 3 years)showed an HR of 0.77 (95% CI 0.69 to 0.87), and a survivalbenefit remained after exclusion of early mortality (HR=0.85,95% CI 0.75 to 0.95) (table 2). In the two studies reportingcomplete follow-up, and follow-up without early mortality,1 17

the HRs were 0.75 (95% CI 0.69 to 0.81) and 0.78 (95% CI0.71 to 0.85), respectively, and I2=0%. In the two studies thatadjusted for surgeon volume, the HR was 1.01 (95% CI 0.97 to1.06, I2=0%) (figure 2).

There was no evidence of publication bias or small-studyeffects bias (p=0.313) (funnel plot not shown).

Surgeon-volume and long-term survivalThe four studies that reported HRs for surgeon volume alsoreported hospital volume.20 22 26 27 In one study, surgeon andhospital volume were equivalent since all oesophagectomieswere performed by one surgical team led by one surgeon ateach specialised regional centre.26 One study reported the HRfor surgeon volume adjusted for hospital volume as a continu-ous variable.27 Another study reported HRs with and withoutadjustment for hospital volume.20 One study reported both hos-pital and surgeon volume, but did not conduct any mutualadjustment for these variables.22 Only one study excluded earlymortality (within 3 months of surgery).20 The follow-up timeranged from 2 to 23 years.20 22 26 27 Only two studies adjustedfor tumour stage.20 22

As presented in figure 3, the pooled adjusted HR for surgeonvolume was 0.87 (95% CI 0.74 to 1.02) in favour of highsurgeon volumes. After further adjustment for hospital volume,the pooled HR was 0.91 (95% CI 0.85 to 0.98). Statistical het-erogeneity in both analyses was low (I2=0%).

There was no evidence of publication bias or small-studyeffects bias (p=0.150) (funnel plot not shown).

Hospital type and long-term survivalFour studies analysed hospital type in relation to survival afteroesophagectomy.18 19 25 26 It was not possible to calculate apooled HR because of clinical heterogeneity between thesestudies (figure 4).

Figure 1 Characteristics and quality assessment of all 16 included studies dealing with survival after oesophageal cancer surgery.

1396 Brusselaers N, et al. Gut 2014;63:1393–1400. doi:10.1136/gutjnl-2013-306074

Oesophagus

group.bmj.com on August 6, 2014 - Published by gut.bmj.comDownloaded from

DISCUSSIONThis meta-analysis indicates a survival benefit in favour of high-volume hospitals and high-volume surgeons compared with thelow-volume equivalents. No independent risk reductionremained for high hospital volume after adjustment for surgeonvolume, while high surgeon volume remained beneficial afteradjustment for hospital volume.

Strengths of meta-analyses include the fact that they facilitateobjective evaluation and pooling of different study populations,enable analyses of large and diverse cohorts of patients and

summarise the available evidence up to a certain time. Inherentlimitations of meta-analyses are that the results depend on theavailability, quality and methods of the published studies, andthey might be hampered by publication bias and clinical andstatistical heterogeneity. However, there was no evidence of pub-lication bias in this study, and the statistical heterogeneity waslow to moderate. Language bias cannot be ruled out completelysince we based our search on English-language dominatedsources. Only one potentially eligible study was excludedbecause of the language,16 but this Japanese study was con-ducted by the same group as in an included study.31 Moreover,even after implementing strict inclusion criteria, a considerableclinical heterogeneity remained between study populations.Moreover, meta-analyses assessing long-term survival are morecomplex than those evaluating short-term mortality, because ofthe need to take duration of survival and variation in follow-uptime into account.

Despite the heterogeneity between studies, this study providesevidence of improved long-term survival when the oesophagect-omy is conducted at high-volume hospitals and by high-volumesurgeons. This distinction between early mortality and long-term mortality enables comparison of the underlying mechan-isms; when patients die early after surgery it is usually owing tocomplications, while later deaths are typically related to cancerrecurrence. This study showed that even after excluding the firstmonths after surgery, a 15% benefit in favour of high-volumehospitals remained, indicating that surgery volume influencesthe risk of tumour recurrence. These findings support centralisa-tion of oesophageal cancer surgery to fewer surgeons workingat centres of excellence in this field.32–34 The discussion aboutwhich hospital and surgeon volume should be recommendedcannot be answered by this study. As in other meta-analyses 2 6

Figure 2 Forrest plot of survival benefit based on annual hospital volume of oesophageal cancer surgery, in 13 studies. The adjusted HRs arebased on the most fully adjusted models as reported for each study, excluding or including adjustment for surgeon volume. ¤Hospital volume wasreported as average per quartile (range per quartile not reported). *Hospital volume was considered a continuous variable and HRs were reportedfor an increase of 10 units.

Table 2 Subanalyses for survival by hospital volume of the 12studies reporting HRs without adjustment for surgeon volume ofoesophageal cancer surgery

Hospital volume (N=12) HR 95% CI N studies I2

All 0.82 0.75 to 0.90 12 68.0Adjusted for tumour stage 0.76 0.68 to 0.84 6 67.0Not adjusted for tumour stage 0.91 0.83 to 1.00 6 11.8Early mortality excluded 0.85 0.75 to 0.95 5 68.3Complete follow-up* 0.78 0.70 to 0.87 9 61.4Long follow-up (>3 years) 0.80 0.71 to 0.89 8 70.1Long and complete follow-up (>3 years)* 0.77 0.69 to 0.87 7 67.1Short follow-up (max 3 years) 0.90 0.72 to 1.14 4 71.4Short and complete follow-up(max 3 years)*

0.97 0.55 to 1.73 2 50.1

*Complete follow-up includes only studies reporting survival including mortalitywithin the early postoperative period.Values are presented as pooled HRs and 95% CIs, with low-volume hospital groupsas reference category. The I2 represents the percentage of variation attributable toheterogeneity, categorised as low (25–50%), moderate (51–75%) or high (>75%).

Brusselaers N, et al. Gut 2014;63:1393–1400. doi:10.1136/gutjnl-2013-306074 1397

Oesophagus

group.bmj.com on August 6, 2014 - Published by gut.bmj.comDownloaded from

we compared the largest volume group with the lowest volumegroup when different thresholds were used, so based on thisstudy design we cannot define a recommended quantity ofannual oesophagectomies by hospital or surgeon.

The findings of this study complement previous meta-analysesevaluating short-term mortality after oesophagectomy, withpooled ORs of 0.20–0.40 for high-volume hospitals comparedwith low-volume hospitals.2 4 6 35 However, deaths fromtumour recurrence after oesophageal cancer surgery are farmore common (60–70%) than deaths occurring during theinitial postoperative period (<5%), emphasising the relevanceof evaluating surgery volume in relation to longer-term survival.

Yet, this issue has been dealt with in only a few systematicreviews,4 6 36–38 and one meta-analysis based on four studies.6

The previous meta-analysis showed an OR of survival of 1.17(95% CI 1.05 to 1.31) in favour of high-volume hospitals, but itdid not take the time to death into account.6 Nevertheless, thefinding of that meta-analysis is in line with the results of ourstudy.

An important, yet complex, question is the underlying mech-anism of the findings of this study, and if hospital and surgeonvolumes should be considered separate entities or merelyproxies of each other. The surgery volume effect might be dueto the total package of multidisciplinary teams, advanced

Figure 3 Forrest plot of survival benefit based on annual surgeon volume of oesophageal cancer surgery, based on four studies. The adjusted HRsare based on the most fully adjusted models as reported for each study, excluding or including adjustment for hospital volume. *Surgeon volumewas considered a continuous variable and HRs were reported for an increase of 10 units.

Figure 4 Forrest plot of survival benefit of hospital type of oesophageal cancer surgery, based on four studies. The adjusted HRs are based on themost fully adjusted models as reported for each study. AdenoCa, adenocarcinoma; NCI, National Cancer Institute (USA); SCC, squamous cellcarcinoma. University and teaching hospitals were assumed to have a higher volume than non-university and non-teaching hospitals; NCI hospitalshad similar hospital volumes to the control hospitals.

1398 Brusselaers N, et al. Gut 2014;63:1393–1400. doi:10.1136/gutjnl-2013-306074

Oesophagus

group.bmj.com on August 6, 2014 - Published by gut.bmj.comDownloaded from

diagnostics, treatment and care, or alternatively, it might mainlybe due to the experience and expertise of the surgeon and thesurgical team. This study found no association between hospitalvolume and survival when considering only the two studiesadjusting for surgeon volume, but interestingly, the influence ofsurgeon volume remained after adjusting for hospital volume.Although based on a small number of studies, this might suggesta more important role for surgeon volume than hospitalvolume.

It is important to point out that survival depends not only onsurgery volume.39 Substantial logistical and case-mix differencesexist between countries, especially for geographical distancesand tumour incidence. A related question is if the patient popu-lation and case-mix, particularly for tumour stage and socio-economic status, are similar between volume groups.40 41 In anattempt to adjust for such confounding, we included only themost fully adjusted regression models. The six studies includingadjustment for tumour stage, the strongest prognostic factor,showed a larger effect than those which did not adjust (24% vs9%), which indicates robustness of the findings of this study.

The size of the effect of high surgery volume in relation tolong-term survival after oesophageal cancer surgery is not negli-gible. The effect size is similar to the effect of preoperativeoncological therapy that has been demonstrated in recent largemeta-analyses,42 and neoadjuvant therapy has therefore becomeroutine clinical practice for these patients in most countries. Ittherefore seems logical that centralisation of this surgery shouldbe a prioritised measure to improve the prognosis in oesopha-geal cancer.

To conclude, this meta-analysis of long-term survival afteroesophageal cancer surgery showed an 18–25% and 9–13%improved survival for high-volume hospitals and high-volumesurgeons, respectively, compared with their low-volume counter-parts. This difference in survival was not solely due to adecreased early postoperative mortality, since even after exclu-sion of early deaths, a 15% benefit was found. Surgeon volumeappears to be more strongly related to survival than hospitalvolume.

Contributors All three authors contributed to the conception and design of thestudy. NB and FM performed the analysis and interpreted the data. NB drafted thearticle, which was revised critically for important intellectual content by all authors.All authors approved the final version of the manuscript.

Funding This study was funded by the Swedish Research Council (SIMSAM): 2008–7496, and Swedish Cancer Society (grant number 120748).

Competing interests None.

Provenance and peer review Not commissioned; externally peer reviewed.

Data sharing statement All data are presented in the article.

REFERENCES1 Bilimoria KY, Bentrem DJ, Feinglass JM, et al. Directing surgical quality improvement

initiatives: comparison of perioperative mortality and long-term survival for cancersurgery. J Clin Oncol 2008;26:4626–33.

2 Markar SR, Karthikesalingam A, Thrumurthy S, et al. Volume-outcome relationshipin surgery for esophageal malignancy: systematic review and meta-analysis 2000–2011. J Gastrointest Surg 2012;16:1055–63.

3 Lauder CI, Marlow NE, Maddern GJ, et al. Systematic review of the impact ofvolume of oesophagectomy on patient outcome. ANZ J Surg 2010;80:317–23.

4 Gruen RL, Pitt V, Green S, et al. The effect of provider case volume on cancermortality: systematic review and meta-analysis. CA Cancer J Clin 2009;59:192–211.

5 Tol JA, van Gulik TM, Busch OR, et al. Centralization of highly complex low-volumeprocedures in upper gastrointestinal surgery. A summary of systematic reviews andmeta-analyses. Dig Surg 2012;29:374–83.

6 Wouters MW, Gooiker GA, van Sandick JW, et al. The volume-outcome relation inthe surgical treatment of esophageal cancer: a systematic review and meta-analysis.Cancer 2012;118:1754–63.

7 Lagergren J, Lagergren P. Recent developments in esophageal adenocarcinoma. CACancer J Clin 2013;63:232–48.

8 Rutegard M, Charonis K, Lu Y, et al. Population-based esophageal cancer survivalafter resection without neoadjuvant therapy: An update. Surgery 2012;152:903–10.

9 Koppert LB, Lemmens VE, Coebergh JW, et al. Impact of age and co-morbidity onsurgical resection rate and survival in patients with oesophageal and gastric cancer.Br J Surg 2012;99:1693–700.

10 Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviewsand meta-analyses: the PRISMA statement. PLoS Medicine 2009;6:e1000097.doi:10/1371/journal.pmed.1000097

11 Sanderson S, Tatt ID, Higgins JP. Tools for assessing quality and susceptibility tobias in observational studies in epidemiology: a systematic review and annotatedbibliography. Int J Epidemiol 2007;36:666–76.

12 Michiels S, Piedbois P, Burdett S, et al. Meta-analysis when only the mediansurvival times are known: a comparison with individual patient data results. Int JTechnol Assess Health Care 2005;21:119–25.

13 Parmar MK, Torri V, Stewart L. Extracting summary statistics to perform meta-analysesof the published literature for survival endpoints. Stat Med 1998;17:2815–34.

14 Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by asimple, graphical test. BMJ 1997;315:629–34.

15 Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency inmeta-analyses. BMJ 2003;327:557–60.

16 Hattori M, Fujita M, Nakamura Y, et al. [Estimation of the effects of centralizationof cancer treatment on mortality reduction by in Fukui Prefecture]. Nihon KoshuEisei Zasshi 2010;57:263–70.

17 Birkmeyer JD, Sun Y, Wong SL, et al. Hospital volume and late survival after cancersurgery. Ann Surg 2007;245:777–83.

18 Birkmeyer NJ, Goodney PP, Stukel TA, et al. Do cancer centers designated by theNational Cancer Institute have better surgical outcomes? Cancer 2005;103:435–41.

19 Cheung MC, Koniaris LG, Yang R, et al. Do all patients with carcinoma of theesophagus benefit from treatment at teaching facilities? J Surg Oncol2010;102:18–26.

20 Derogar M, Sadr-Azodi O, Johar A, et al. Hospital and surgeon volume in relationto survival after esophageal cancer surgery in a population-based study. J ClinOncol 2013;31:551–7.

21 Rouvelas I, Lindblad M, Zeng W, et al. Impact of hospital volume on long-termsurvival after esophageal cancer surgery. Arch Surg 2007;142:113–17; discussion 8.

22 Sundelof M, Lagergren J, Ye W. Surgical factors influencing outcomes in patientsresected for cancer of the esophagus or gastric cardia. World J Surg2008;32:2357–65.

23 Wenner J, Zilling T, Bladstrom A, et al. The influence of surgical volume on hospitalmortality and 5-year survival for carcinoma of the oesophagus and gastric cardia.Anticancer Res 2005;25:419–24.

24 Dikken JL, Dassen AE, Lemmens VE, et al. Effect of hospital volume onpostoperative mortality and survival after oesophageal and gastric cancer surgery inthe Netherlands between 1989 and 2009. Eur J Cancer 2012;48:1004–13.

25 Dikken JL, Wouters MW, Lemmens VE, et al. Influence of hospital type on outcomesafter oesophageal and gastric cancer surgery. Br J Surg 2012;99:954–63.

26 van de Poll-Franse LV, Lemmens VE, Roukema JA, et al. Impact of concentration ofoesophageal and gastric cardia cancer surgery on long-term population-basedsurvival. Br J Surg 2011;98:956–63.

27 Bachmann MO, Alderson D, Edwards D, et al. Cohort study in South and WestEngland of the influence of specialization on the management and outcome ofpatients with oesophageal and gastric cancers. Br J Surg 2002;89:914–22.

28 Coupland VH, Lagergren J, Luchtenborg M, et al. Hospital volume, proportionresected and mortality from oesophageal and gastric cancer: a population-basedstudy in England, 2004–2008. Gut 2013;62:961–6.

29 Stavrou EP, Smith GS, Baker DF. Surgical outcomes associated withoesophagectomy in New South Wales: an investigation of hospital volume.J Gastrointest Surg 2010;14:951–7.

30 Simunovic M, Rempel E, Theriault ME, et al. Influence of hospital characteristics onoperative death and survival of patients after major cancer surgery in Ontario. Can JSurg 2006;49:251–8.

31 Ioka A, Tsukuma H, Ajiki W, et al. Hospital procedure volume and survival of cancerpatients in Osaka, Japan: a population-based study with latest cases. Jpn J ClinOncol 2007;37:544–53.

32 Chang KH, McAnena OJ, Smith MJ, et al. Surgery for oesophageal cancer atGalway University Hospital 1993–2008. Ir J Med Sci 2010;179:521–7.

33 Forshaw MJ, Gossage JA, Stephens J, et al. Centralisation of oesophagogastriccancer services: can specialist units deliver? Ann R Coll Surg Engl 2006;88:566–70.

34 Wouters MW, Karim-Kos HE, le Cessie S, et al. Centralization of esophageal cancersurgery: does it improve clinical outcome? Ann Surg Oncol 2009;16:1789–98.

35 Metzger R, Bollschweiler E, Vallbohmer D, et al. High volume centers foresophagectomy: what is the number needed to achieve low postoperative mortality?Dis Esophagus 2004;17:310–14.

36 Wouters MW, Krijnen P, Le Cessie S, et al. Volume- or outcome-based referral toimprove quality of care for esophageal cancer surgery in The Netherlands. J SurgOncol 2009;99:481–7.

Brusselaers N, et al. Gut 2014;63:1393–1400. doi:10.1136/gutjnl-2013-306074 1399

Oesophagus

group.bmj.com on August 6, 2014 - Published by gut.bmj.comDownloaded from

37 Killeen SD, O’Sullivan MJ, Coffey JC, et al. Provider volume and outcomes foroncological procedures. Br J Surg 2005;92:389–402.

38 Holscher AH, Metzger R, Brabender J, et al. High-volumecenters–effect of case load on outcome in cancer surgery. Onkologie2004;27:412–16.

39 Meguid RA, Weiss ES, Chang DC, et al. The effect of volume on esophagealcancer resections: what constitutes acceptable resection volumes for centers ofexcellence? J Thorac Cardiovasc Surg 2009;137:23–9.

40 Boudourakis LD, Wang TS, Roman SA, et al. Evolution of the surgeon-volume,patient-outcome relationship. Ann Surg 2009;250:159–65.

41 Wouters MW, Wijnhoven BP, Karim-Kos HE, et al. High-volume versuslow-volume for esophageal resections for cancer: the essential role of case-mixadjustments based on clinical data. Ann Surg Oncol 2008;15:80–7.

42 Sjoquist KM, Burmeister BH, Smithers BM, et al. Survival after neoadjuvantchemotherapy or chemoradiotherapy for resectable oesophageal carcinoma: anupdated meta-analysis. Lancet Oncol 2011;12:681–92.

1400 Brusselaers N, et al. Gut 2014;63:1393–1400. doi:10.1136/gutjnl-2013-306074

Oesophagus

group.bmj.com on August 6, 2014 - Published by gut.bmj.comDownloaded from

doi: 10.1136/gutjnl-2013-306074 2014 63: 1393-1400 originally published online November 22, 2013Gut

 Nele Brusselaers, Fredrik Mattsson and Jesper Lagergren systematic review and meta-analysislong-term survival after oesophagectomy: Hospital and surgeon volume in relation to

http://gut.bmj.com/content/63/9/1393.full.htmlUpdated information and services can be found at:

These include:

Data Supplement http://gut.bmj.com/content/suppl/2013/11/22/gutjnl-2013-306074.DC1.html

"Supplementary Data"

References http://gut.bmj.com/content/63/9/1393.full.html#ref-list-1

This article cites 42 articles, 8 of which can be accessed free at:

serviceEmail alerting

the box at the top right corner of the online article.Receive free email alerts when new articles cite this article. Sign up in

CollectionsTopic

(333 articles)Oesophageal cancer   � Articles on similar topics can be found in the following collections

Notes

http://group.bmj.com/group/rights-licensing/permissionsTo request permissions go to:

http://journals.bmj.com/cgi/reprintformTo order reprints go to:

http://group.bmj.com/subscribe/To subscribe to BMJ go to:

group.bmj.com on August 6, 2014 - Published by gut.bmj.comDownloaded from