Articles

Survival after neoadjuvant chemotherapy or chemoradiotherapy for resectable oesophageal carcinoma: an updated meta-analysis
Katrin M Sjoquist, Bryan H Burmeister, B Mark Smithers, John R Zalcberg, R John Simes, Andrew Barbour, Val Gebski, for the Australasian Gastro-Intestinal Trials Group

Summary
Background In a previous meta-analysis, we identied a survival benet from neoadjuvant chemotherapy or chemoradiotherapy before surgery in patients with resectable oesophageal carcinoma. We updated this meta-analysis with results from new or updated randomised trials presented in the past 3 years. We also compared the benets of preoperative neoadjuvant chemotherapy compared with neoadjuvant chemoradiotherapy. Methods To identify additional studies and published abstracts from major scientic meetings, we searched Medline, Embase, and Central (Cochrane clinical trials database) for studies published since January, 2006, and also manually searched for abstracts from major conferences from the same period. Only randomised studies analysed by intention to treat were included, and searches were restricted to those databases citing articles in English. We used published hazard ratios (HRs) if available or estimates from other survival data. We also investigated treatment eects by tumour histology and relations between risk (survival after surgery alone) and eect size. Findings We included all 17 trials from the previous meta-analysis and seven further studies. 12 were randomised comparisons of neoadjuvant chemoradiotherapy versus surgery alone (n=1854), nine were randomised comparisons of neoadjuvant chemotherapy versus surgery alone (n=1981),and two compared neoadjuvant chemoradiotherapy with neoadjuvant chemotherapy (n=194) in patients with resectable oesophageal carcinoma; one factorial trial included two comparisons and was included in analyses of both neoadjuvant chemoradiotherapy (n=78) and neoadjuvant chemotherapy (n=81). The updated analysis contained 4188 patients whereas the previous publication included 2933 patients. This updated meta-analysis contains about 3500 events compared with about 2230 in the previous meta-analysis (estimated 57% increase). The HR for all-cause mortality for neoadjuvant chemoradiotherapy was 078 (95% CI 070088; p<00001); the HR for squamous-cell carcinoma only was 080 (068093; p=0004) and for adenocarcinoma only was 075 (059095; p=002). The HR for all-cause mortality for neoadjuvant chemotherapy was 087 (079096; p=0005); the HR for squamous-cell carcinoma only was 092 (081104; p=018) and for adenocarcinoma only was 083 (071095; p=001). The HR for the overall indirect comparison of all-cause mortality for neoadjuvant chemoradiotherapy versus neoadjuvant chemotherapy was 088 (076101; p=007). Interpretation This updated meta-analysis provides strong evidence for a survival benet of neoadjuvant chemoradiotherapy or chemotherapy over surgery alone in patients with oesophageal carcinoma. A clear advantage of neoadjuvant chemoradiotherapy over neoadjuvant chemotherapy has not been established. These results should help inform decisions about patient management and design of future trials. Funding Cancer Australia and the NSW Cancer Institute.
Lancet Oncol 2011; 12: 68192 Published Online June 17, 2011 DOI:10.1016/S14702045(11)70142-5 See Comment page 615 National Health and Medical Research Council Clinical Trials Centre, University of Sydney, Sydney, NSW, Australia (K M Sjoquist FRACP, Prof R J Simes MD, Prof V Gebski MStat); Division of Cancer Services, University of Queensland, Princess Alexandra Hospital, Brisbane, QLD, Australia (Prof B H Burmeister MD); Department of Surgery, University of Queensland, Brisbane, QLD, Australia (Prof B M Smithers FRACS, A Barbour FRACS); Upper Gastrointestinal and Soft Tissue Unit, Princess Alexandra Hospital, Brisbane, QLD, Australia (B M Smithers, A Barbour); and Peter MacCallum Cancer Centre, and Department of Medicine, University of Melbourne, Melbourne, VIC, Australia (Prof J R Zalcberg FRACP) Correspondence to: Prof Val Gebski, National Health and Medical Research Council Clinical Trials Centre, Locked Bag 77, Camperdown 1450 NSW, Australia val@ctc.usyd.edu.au

Introduction
Overall survival of patients with resectable oesophageal cancer remains poor, with a 5-year survival of 1534%,1 despite changes in management over the past 20 years. Most patients who undergo radical resection for oesophageal cancer will eventually relapse and die as a result of their disease.2 Because of diculties in administering chemotherapy or radiotherapy soon after a surgical procedure, high perioperative morbidity, and the disappointing results of trials of adjuvant chemotherapy,3 radiotherapy,4,5 or combination chemoradiotherapy,2 the focus of recent trials has been on neoadjuvant treatment. In our previous meta-analysis,6 we reported a signicant survival benet for neoadjuvant chemoradiotherapy and,
www.thelancet.com/oncology Vol 12 July 2011

to a lesser extent, neoadjuvant chemotherapy in patients with squamous-cell carcinoma or adenocarcinoma of the oesophagus. At present, there is no evidence supporting the use of neoadjuvant radiotherapy alone.4 Most trials of chemotherapy or combined chemoradiotherapy have used a doublet of cytotoxics, usually a platinum compound, most often cisplatin, and uorouracil,1,722 with varying doses and sequencing. Some small trials of neoadjuvant chemotherapy have used triplet chemotherapy, which resulted in increased toxicity without signicant survival benets.23,24 Recent treatment modications have included the use of more modern cytotoxic drugs, changes in chemotherapy sequencing, or changes in the dose and fractionation of radiotherapy.2
681

Articles

We aimed to assess whether the results of recently published or updated trials have changed the outcomes of our previous meta-analysis.6 We also sought to compare the benets of neoadjuvant chemotherapy and neoadjuvant chemoradiotherapy before surgery for resectable oesophageal carcinoma, and to assess whether any increase in survival benets was oset by an increase in perioperative mortality.

Methods
For the rst section of the meta-analysis, we sought to assess the survival benets of neoadjuvant treatment with either neoadjuvant chemotherapy or neoadjuvant chemoradiotherapy of any regimen. All randomised controlled trials that compared survival after neoadjuvant chemoradiotherapy or neoadjuvant chemotherapy followed by surgery with surgery alone in the initial management of resectable oesophageal or oesophagogastric junction carcinoma (squamous-cell carcinoma, adenocarcinoma, or mixed tumours) were eligible for inclusion, provided they were analysed by intention-to-treat.

See Online for webappendix

22 studies identied from previous meta-analysis*

388 studies identied from database searches

76 additional records identied from other sources

302 studies after duplicates removed Title and abstracts screened for eligibility

35 abstracts reviewed

3 excluded 1 (An, 2003): abstract not available in English and no further details provided by author 2 (Peng, 2008,33 and Wang, 200125): full text not in English, insucient detail in abstract, and no further details provided by author

32 full-text articles assessed for eligibility 8 excluded 2 (Cunningham, 2006,34 and Schuhmacher, 201035): results for gastro-oesophageal junction and oesophageal tumours not available separately; most patients received gastrectomy 2 (Bokhyan, 200936 [abstract] and Stidil, 2006 [abstract]; both abstracts describe the same study): insucient information available from abstract and no response from authors 1 (Boige, 2007 [abstract]): abstract not used; insucient data on subgroups; nal publication (Ychou7) used because additional information available to meet inclusion criteria 1 (Kok, 199728 [abstract]): abstract not used because updated results available (Boonstra37) after contacting authors 2 (Kelsen, 1998, and Medical Research Council, 200238): updated results available (Kelsen, 2007,8 and Allum, 2009,1 respectively)

We included all trials from our previous meta-analysis.6 We searched Medline, Embase, and Central (Cochrane clinical trials database) for studies published since January, 2006. We used the following search terms: esophageal neoplasms, gastro-esophageal junction neoplasms, antineoplastic agents, chemotherapy, radiotherapy, surgery, and esophagectomy or gastrectomy. We also manually searched for abstracts from major conferences over the same period. Articles for which neither the abstract nor full text was available in English were excluded after review. We also included studies in an analysis of histological subtypes if subgroup data could be identied separately, with the same criteria for inclusion as described in the previous meta-analysis.6 The primary outcome of interest for the rst section of the meta-analysis was all-cause mortality. The secondary endpoint was the eect on all-cause mortality of treatment for each histological subtype (squamous-cell carcinoma or adenocarcinoma). For the second section of the meta-analysis, we sought to assess the survival benets of neoadjuvant chemotherapy compared with neoadjuvant chemoradiotherapy. All randomised trials of this comparison that included adult patients with squamous-cell carcinoma, adenocarcinoma, or mixed tumours were eligible. We did not identify any previous systematic reviews or meta-analyses of sucient quality, and so we searched Medline, Embase, and Central with the same terms as for the rst section. All trials published between January, 1980 and November, 2010, were included if the abstract or full article was available in English. For both sections of the meta-analysis, we wrote to authors of studies that had previously only appeared as abstracts and were not in the databases to clarify whether updated data had been published, including the authors of four studies that were excluded from the previous meta-analysis.2528 Trials that focused on primary gastric cancers were excluded if the results for oesophagogastric junction tumours were not available separately.

Statistical analyses
For the rst section of the meta-analysis, we used the same statistical methods as previously reported.6 We used the hazard ratio (HR) for the comparison in each trial to assess the treatment eects. Where possible, the HR and associated variance were obtained directly from each trial publication or from individual patient data. HRs not reported were calculated by the methods of Parmar and colleagues.29 Statistical tests were two-sided. Pooled estimates with 95% CIs were calculated by the weighted variance technique and we used the test to assess heterogeneity, with the level of signicance set at 5%. We assessed publication bias by the methods described by Gleser and Olkin,30 and we used Fishers failsafe N31 to identify the number of studies with a p value of 05 (ie, an HR of 10) that would need to be added to those in the meta-analysis to produce a non-signicant result.
www.thelancet.com/oncology Vol 12 July 2011

24 included in meta-analysis

Figure 1: Flow diagram showing inclusion and exclusion of studies *Includes 17 studies that were included in this meta-analysis and ve studies that could not be included. Includes abstracts that reported earlier results where full papers were subsequently published and multiple abstracts that reported the results of the same study. See webappendix p 1 for reference.

682

Articles

We calculated the overall 2-year survival estimate in the control group from a mean of individual 2-year survival rates weighted by the sample size of the control group. The estimate of the 2-year survival rate in the intervention was obtained by applying the HR to the
Year Radiotherapy schedule started Chemoradiotherapy vs surgery alone Nygaard9 Apinop39 Le Prise10 Urba40 Bosset12 Walsh13 Walsh14 Burmeister22 Tepper43 Lv41|| Lee17 Mariette11|| van der Gaast42|| 1983 1986 1988 1989 1989 1990 1990 1994 1997 1997 1999 2000 2004 35 Gy, 175 Gy per fraction over 4 weeks 40 Gy, 2 Gy per fraction over 4 weeks 20 Gy in 10 fractions over 12 days 45 Gy, 15 Gy per fraction over 3 weeks 37 Gy, 37 Gy per fraction over 2 weeks 40 Gy in 15 fractions over 3 weeks 40 Gy in 15 fractions over 3 weeks 35 Gy in 15 fractions over 3 weeks 504 Gy, 18 Gy per fraction over 56 weeks 40 Gy, 2 Gy per fraction over 4 weeks 456 Gy, 12 Gy per fraction over 28 days 45 Gy in 25 fractions over 5 weeks 414 Gy, 18 Gy per fraction over 46 weeks

estimated control rate. We then used these data to calculate the absolute risk reduction and number needed to treat. Because the individual trials could be deemed independent of each other, we used indirect comparisons
Concurrent Tumour type or sequential Sample Median size follow-up (months)*

Chemotherapy schedule

Two cycles: cisplatin 20 mg/m days 15; bleomycin 5 mg/m days 15 Two cycles: cisplatin 100 mg/m day 1; uorouracil 1000 mg/m days 14 Two cycles: cisplatin 100 mg/m day 1; uorouracil 600 mg/m days 25 and 2225 Two cycles: cisplatin 20 mg/m days 15; uorouracil 300 mg/m days 121; vinblastine 1 mg/m days 14 Two cycles: cisplatin 80 mg/m days 02 Two cycles: cisplatin 75 mg/m day 7; uorouracil 15 mg/kg days 15 Two cycles: cisplatin 75 mg/m day 7; uorouracil 15 mg/kg days 15 One cycle: cisplatin 80 mg/m day 1; uorouracil 800 mg/m days 25 Two cycles: cisplatin 60 mg/m day 1; uorouracil 1000 mg/m days 35

Sequential Concurrent Sequential Concurrent Sequential Concurrent Concurrent Concurrent Concurrent

SCC SCC SCC SCC and adenocarcinoma SCC SCC Adenocarcinoma SCC and adenocarcinoma SCC and adenocarcinoma SCC SCC SCC and adenocarcinoma SCC and adenocarcinoma SCC SCC SCC SCC SCC SCC SCC and adenocarcinoma SCC SCC and adenocarcinoma Adenocarcinoma

78 69 86 100 293 61 113 256 56 160 101 195 364

18 12 12 98 55 10 24 65 72 45 25 68 32

Two cycles: cisplatin 20 mg/m per day days 13 and 2225; Concurrent paclitaxel 135 mg/m starting on day 1 and day 22 of radiotherapy Two cycles: cisplatin 60 mg/m day 1; uorouracil 1000 mg/m days 35 Two cycles: cisplatin 75 mg/m day 1 or 2; uorouracil 800 mg/m days 14 5 weeks concurrent chemotherapy: carboplatin area under curve=2 and paclitaxel 50 mg/m on day 1 weekly Two cycles: cisplatin 120 mg/m day 1; vindesine 3 mg/m days 1, 8; bleomycin 10 U/m days 36 Two cycles: cisplatin 20 mg/m days 15; bleomycin 5 mg/m days 15 Three cycles: cisplatin 20 mg/m days 15; uorouracil 1000 mg/m days 15 Two cycles: cisplatin 100 mg/m day 1; bleomycin 10 mg/m days 38; vinblastine 3 mg/m days 1 and 8 Two cycles: cisplatin 100 mg/m day 1; uorouracil 1000 mg/m days 15 Two cycles: cisplatin 80 mg/m day 1; etoposide 200 mg/m days 15 Three cycles: cisplatin 100 mg/m day 1; uorouracil 1000 mg/m days 15 Two cycles: cisplatin 100 mg/m day 1; uorouracil 1000 mg/m days 15 Two cycles: cisplatin 80 mg/m day 1; uorouracil 1000 mg/m days 14 Planned six perioperatively: (two or three cycles before surgery plus four or three cycles after surgery) of intravenous cisplatin (100 mg/m) on day 1, and a continuous intravenous infusion of fluorouracil (800 mg/m per day) for 5 consecutive days (days 15) every 28 days Concurrent Concurrent Concurrent

Chemotherapy vs surgery alone Roth23 Nygaard9 Schlag19 Maipang24 Law20 Boonstra37|| Kelsen8 Ancona21 Allum1 Ychou7|| 1982 1983 39 81 46 46 147 169 467 96 802 169 20 18 75 17 17 145 106 24 72 684

1988** 1988 1989 1989 1990 1992 1992 1995

(Continues on next page)

www.thelancet.com/oncology Vol 12 July 2011

683

Articles

Year Radiotherapy schedule started (Continued from previous page) Chemoradiotherapy vs chemotherapy Stahl 2000 (chemoradiotherapy)18|| 30 Gy, 2 Gy per fraction over 3 weeks, commencing 2 weeks after last day of induction chemotherapy 35 Gy in 15 fractions commencing day 22

Chemotherapy schedule

Concurrent Tumour type or sequential

Sample Median size follow-up (months)*

Induction and Adenocarcinoma 12 weeks (induction): uorouracil 2000 mg/m over 24 h day 1; concurrent folinic acid 500 mg/m day 1; and cisplatin 80 mg/m biweekly. Followed by cisplatin 50 mg/m on day 1 and day 8 and etoposide 80 mg/m on days 35, concurrent with radiotherapy 1 cycle induction: cisplatin 80 mg/m day 1 and uorouracil 1000 mg/m infusion over 96 h day 1. Followed by cisplatin 80 mg/m day 1 and uorouracil 800 mg/m infusion over 96 h on day 1; concurrent with radiotherapy 15 weeks: uorouracil 2000 mg/m over 24 h day 1, folinic acid 500 mg/m day 1; cisplatin 80 mg/m biweekly Two cycles: cisplatin 80 mg/m day 1, uorouracil 1000 mg/m infusion over 96 h day 1 Induction and Adenocarcinoma concurrent

60

46

Burmeister 2000 (chemoradiotherapy)15||

36

70

Stahl (chemotherapy)18|| 2000 Burmeister (chemotherapy)15|| 2000

Adenocarcinoma Adenocarcinoma

59 36

46 70

SCC=squamous-cell carcinoma. AC=adenocarcinoma. NR=not reported. *As published. Estimated as the median survival. Unpublished thesis. Updated since previous meta-analysis. Does not include the 78 patients in this study who received postoperative chemoradiotherapy. ||Not included in previous meta-analysis. **Year of activation not reported, but imputed. Imputed, rather than published. Previously published as abstract only. Does not includes the 55 patients in this study who had primary gastric cancer. An additional seven patients were randomised but deemed ineligible. Survival analysis includes 119 patients in total.

Table 1: Chemoradiotherapy and chemotherapy regimens in randomised trials included in the meta-analysis

to obtain estimates of the benet of neoadjuvant chemoradiotherapy compared with neoadjuvant chemotherapy. Potential biases might occur in such comparisons, thus the amount to which the results of the indirect comparisons corroborate those of the randomised studies is also of interest. We assessed this by comparing dierences between the eects obtained from the direct and indirect comparisons. Consistency between eects for indirect and direct comparisons provides condence about the additional information in the indirect comparisons. These approaches were described by Song and colleagues,32 as was the comparison of the amount of discrepancy between the indirect comparisons and those of the randomised trials; a p value of greater than 005 would suggest that the results for the indirect and direct comparisons are statistically consistent. Data were analysed with software provided by the Cochrane Library (Rev Man 4.3, and 5.0 downloaded April, 2010). Dierences between outcomes in the control arms of the dierent trials were measured against the logarithm of the HR to give an estimate of the relation between risk and benet of neoadjuvant chemotherapy or neoadjuvant chemoradiotherapy. The number of events (deaths) used was estimated from the sample size and median followup time reported for each trial and an assumption that about 50% of events had occurred within 18 months. We also did prespecied subgroup analyses to examine the eects of neoadjuvant chemotherapy or neoadjuvant chemoradiotherapy according to tumour histology (squamous-cell carcinoma or adenocarcinoma) where this information was available. Dierences between treatment arms for 30-day postoperative or in-hospital mortality were calculated from the absolute dierence in the proportion of deaths, with 95% CIs.
684

Role of the funding source

The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, writing of this report, or the decision to publish the results. KS and VG had full access to all of the data in the study and had the nal responsibility, with the agreement of all authors, for the decision to submit for publication.

Results
We included 24 studies in total (gure 1),1,715,1724,37,3943 which consisted of all 17 trials from the previous meta-analysis810,1214,16,17,1924,38,39,40 and seven further studies.7,11,15,18,37,41,42 12 were randomised comparisons of neoadjuvant chemoradiotherapy versus surgery alone (n=1854),1014,17,22,3943 nine were randomised comparisons of neoadjuvant chemotherapy versus surgery alone (n=1981),1,7,8,1921,23,24,37 two compared neoadjuvant chemoradiotherapy with neoadjuvant chemotherapy (n=194),15,18 and one study included two comparisons and was included in analyses of both neoadjuvant chemoradiotherapy (n=78) and neoadjuvant chemotherapy (n=81).9 This study was a 2 2 factorial study that simultaneously compared the eect of neoadjuvant chemotherapy, neoadjuvant chemoradiotherapy, and neoadjuvant radiotherapy on survival, and the control group was used for both the neoadjuvant chemotherapy and neoadjuvant chemoradiotherapy comparisons. Accordingly, half the sample size for the control group was reported in each of the comparisons. The HR for the comparison of neoadjuvant chemotherapy with surgery alone in this study was used as calculated for the previous meta-analysis.6 The updated analysis contained 4188 patients whereas the previous publication included 2933 patients. The
www.thelancet.com/oncology Vol 12 July 2011

Articles

A
Nygaard9 Apinop39 Le Prise10 Urba40 Bosset12 Walsh (SCC)13 Walsh (adenocarcinoma)14 Burmeister22 Tepper43 Lv41 Lee17 Mariette11 van der Gaast42 Total Test for overall eect: Z=428 (p<00001)

Chemoradiotherapy (total) 53 35 45* 50 148 29 58 128 30 80 51 97 176 980

Surgery alone (total) 25 34 41 50 145 32 55 128 26 80 50 98 188 952 02 05 Favours chemoradiotherapy Surgery alone (total) 1 2 5 Favours surgery alone

Hazard ratio (95% CI) 076 (045128) 080 (048134) 085 (050146) 074 (048112) 096 (073127) 074 (046118) 058 (038088) 094 (070126) 035 (018068) 055 (036084) 088 (048162) 109 (074159) 067 (049091) 078 (070088)

Heterogeneity: 2=1804, df=12 (p=011); I2=33%

B
Squamous-cell carcinoma Nygaard9 Apinop39 Le Prise10 Urba40 Bosset12 Walsh13 Burmeister22 Lv41 Lee17 Subtotal Heterogeneity: 2=531, df=8 (p=072); I2=0% Test for overall eect: Z=290 (p=0004) Adenocarcinoma Urba40 Walsh14 Burmeister22 Subtotal

Chemoradiotherapy (total)

Hazard ratio (95% CI)

53 35 45 13 148 29 44 80 51 498

25 34 41 12 145 32 48 80 50 467

076 (045128) 080 (048134) 085 (050146) 083 (036189) 096 (073127) 074 (046118) 068 (040115) 055 (036084) 088 (048162) 080 (068093)

37 58 80 175

38 55 77 170

069 (042114) 058 (038088) 094 (066134) 075 (059095)

Heterogeneity: 2=311, df=2 (p=021); I2=36% Test for overall eect: Z=240 (p=002) Combined results (pooled SCC and adenocarcinoma) Tepper43 Mariette11 van der Gaast42 Subtotal Test for overall eect: Z=262 (p=0009) Total Test for overall eect: Z=455 (p<00001) Test for subgroup dierences: 2=035, df=2 (p=084), I2=0% 976 949 077 (069086) 30 97 176 303 26 98 188 312 035 (018068) 109 (074159) 067 (049091) 074 (059093)

Heterogeneity: 2=909, df=2 (p=001); I2=78%

Heterogeneity: 2=1787, df=14 (p=021); I2=22% 02 05 Favours chemoradiotherapy 1 2 5 Favours surgery alone

Figure 2: All-cause mortality for chemoradiotherapy compared with surgery alone Eects of chemoradiotherapy compared with surgery alone on survival in patients with oesophageal cancer (A) and in subgroups of patients with oesophageal carcinoma (squamous-cell carcinoma, adenocarcinoma or pooled histology type subgroups; B). SCC=squamous-cell carcinoma. *Includes all randomised patients. Includes four patients whose histology was unknown or who had mixed tumours. Includes three patients whose histology was unknown or who had mixed tumours.

www.thelancet.com/oncology Vol 12 July 2011

685

Articles

A
Roth23 Nygaard9 Schlag19 Maipang24 Law20 Boonstra37 Kelsen8 Ancona21 Allum1 Ychou7 Total Test for overall eect: Z=283 (p=0005)

Chemotherapy (total) 19 56 22 24 74 85 233* 48 400 85 1046

Surgery alone (total) 20 25 24 22 73 84 234 48 402* 84 1016 02 05 Favours chemotherapy 1 2 5 Favours surgery alone

Hazard ratio (95% CI) 071 (036143) 122 (082181) 097 (060157) 161 (079327) 073 (053100) 071 (051098) 105 (086128) 085 (050144) 084 (072098) 063 (045089) 087 (079096)

Heterogeneity: 2=1577, df=9 (p=007); I2=43%

Chemotherapy (total)

Surgery alone (total)

Hazard ratio (95% CI)

Squamous-cell carcinoma Roth23 Nygaard9 Schlag19 Maipang24 Law20 Boonstra37 Kelsen8 Ancona21 Allum1 Total Test for overall eect: Z=134 (p=018) Adenocarcinoma Kelsen8 Allum1 Ychou7 Subtotal Test for overall eect: Z=258 (p=001) Total Test for overall eect: Z=271 (p=0007) Test for subgroup dierences: 2=114, df=1 (p=029); I2=124% 1024 1006 088 (080096) Heterogeneity: 2=283, df=2 (p=024); I2=29% 120 265 85 470 124 268 84 476 090 (069118) 086 (071105) 063 (045089) 083 (071095) 19 56 22 24 74 85 103 48 123 554 20 25 24 22 73 84 110 48 124 530 071 (036143) 122 (082181) 097 (060157) 161 (079327) 073 (053100) 071 (051098) 125 (094167) 085 (050144) 081 (061107) 092 (081104)

Heterogeneity: 2=1470, df=8 (p=007); I2=46%

Heterogeneity: 2=1868, df=11 (p=007); I2=41% 02 05 Favours chemotherapy 1 2 5 Favours surgery alone

Figure 3: All-cause mortality for chemotherapy compared with surgery alone Eects of chemotherapy compared with surgery alone on survival in patients with oesophageal cancer (A) and in subgroups of patients with oesophageal carcinoma (squamous-cell carcinoma or adenocarcinoma histology; B). *Includes ten patients whose histology was unknown or who had mixed tumours. Includes 12 patients whose histology was unknown or who had mixed tumours.

updated work includes information on 3994 patients for neoadjuvant treatments compared with surgery alone and 194 patients for comparisons of neoadjuvant chemoradiotherapy with neoadjuvant chemotherapy. This updated meta-analysis contains about 3500 events compared with 2230 in the previous meta-analysis (estimated 57% increase). 22 randomised trials were suitable for inclusion in the quantitative analysis of the survival benets of
686

neoadjuvant treatment with either neoadjuvant chemotherapy or neoadjuvant chemoradiotherapy,1,7,814,17,1924,37,3943 and two were suitable for inclusion in the quantitative analysis of neoadjuvant chemotherapy with neoadjuvant chemoradiotherapy.15,18 17 of these were included in the previous publication (webappendix p 1),9,10,1214,16,17,1924,3840 three of which have been updated,1,8,43 and ve new eligible studies were identied.7,11,37,41,42 Four studies were excluded from the previous meta-analysis because of
www.thelancet.com/oncology Vol 12 July 2011

Articles

Chemoradiotherapy (total) Individual trials Stahl18 Burmeister15 Subtotal Heterogeneity: 2=084, df=1 (p=036); I2=0% Test for overall eect: Z=136 (p=017) Pooled trials (indirect) Indirect Subtotal Heterogeneity: not applicable Test for overall eect: Z=142 (p=015) Total Test for overall eect: Z=183 (p=007) Test for subgroup dierences: 2=053, df=1 (p=046); I2=0% 1079 980 980 60 39 99

Chemotherapy (total)

Hazard ratio (95% CI)

59 36 95

067 (041108) 096 (053174) 077 (053112)

1046 1046

090 (077104) 090 (077104)

1141

088 (076101)

Heterogeneity: 2=138, df=2 (p=050); I2=0% 02 05 1 Favours chemoradiotherapy 2 5 Favours chemotherapy

Figure 4: Indirect comparison of all-cause mortality for chemoradiotherapy and chemotherapy

unclear or unavailable information. We wrote to all these authors again but were unable to obtain sucient information for three of the four studies to allow inclusion in this analysis. Reasons for exclusion were: unclear randomisation method,26 analysis not by intention to treat,44 and full text not in English and insucient information available in English abstract.25 The remaining study was excluded from the previous quantitative analysis because it had only been published in abstract form28 at the time of the previous publication. The longterm results have recently been published and are now included.37 All patients had stages T03, N01 disease according to the 2002 AJCC staging.45 There were no major dierences in the age of participants between studies or groups, although no formal statistical comparison was done. Three trials included only patients with early (stage III) disease, all of which compared neoadjuvant chemoradiotherapy with surgery alone.10,11,46 Table 1 summarises details of all trials included in the meta-analysis. 13 studies were included in the comparison of neoadjuvant chemoradiotherapy with surgery alone in patients with resectable oesophageal carcinoma (n=1932): 12 were randomised comparisons of neoadjuvant chemoradiotherapy followed by surgery with surgery alone1014,16,17,22,3942 and one study included this comparison within a 2 2 factorial study.9 Ten had been included in the previous meta-analysis;6 of these, one (Tepper16) had been published only in abstract form, and the data included in this version is the updated nal publication.43 Two additional trials (Mariette11 and van der Gaast42) were presented at the American Society of Clinical Oncology meeting in June, 2010, and one study was published since the previous meta-analysis.41 One further publication from China with an English abstract was identied;33 however, it was not included
www.thelancet.com/oncology Vol 12 July 2011

because the information from the abstract was insucient to calculate an HR and we were unable to obtain further details from the author. A test for potential publication bias suggested ten potentially unpublished studies for neoadjuvant chemoradiotherapy, but 23 more studies with a HR of 10 would need to be added to these 13 to produce an overall non-signicant result, suggesting that the meta-analysis is robust to publication bias. Figure 2 shows pooled estimates for all-cause mortality for the trials that compared neoadjuvant chemoradiotherapy followed by surgery with surgery alone. The pooled HR was 078 (95% CI 070088; p<00001). This corresponds to an absolute survival benet at 2 years of 87% and a number needed to treat of 11. There was no evidence of signicant heterogeneity between the trials or between pooled results by tumour histology. The survival benets for neoadjuvant chemoradiotherapy were similar in the tumour type subgroups: squamous-cell carcinoma (HR 080, 95% CI 068093; p=0004) and adenocarcinoma (075, 059095; p=002). Ten studies were included in the comparison of neoadjuvant chemotherapy with surgery alone (n=2062): nine were randomised comparisons of neoadjuvant chemotherapy with surgery alone1,7,8,1921,23,24,37 and one study included this comparison within a 2 2 factorial study.9 Eight studies were included in the previous meta-analysis, two of which had been updated since our previous publication.1,8 One study37 had only been published in abstract form28 at the time of the previous publication and was previously excluded from the quantitative analysis. One study included adenocarcinomas of gastric, oesophagogastric junction, and oesophageal origin, but was eligible for the quantitative analysis because the nal publication reported outcomes by tumour location.7 Only the outcomes for the
687

Articles

Intervention

Control

Neoadjuvant chemoradiotherapy versus surgery alone Nygaard9 Apinop39 Le Prise10 Urba40 Bossett12 Walsh13 (SCC) Walsh14 (adenocarcinoma) Burmeister22 Tepper16 Lv41* Lee17 Mariette11 van der Gaast42 Roth23 Nygaard9 Schlag19 Maipang24 Law20 Boonstra37 Kelsen8 Ancona21 Allum1 Ychou7 Stahl18 Burmeister15 8/53 (15%) 3/35 (9%) 4/45 (9%) 1/50 (2%) 17/148 (11%) 5/29 (17%) 4/58 (7%) 5/128 (4%) 0/30 (0%) 3/80 (4%) 1/51 (2%) 7/97 (7%) 7/176 (4%) 2/19 (11%) 6/56 (11%) 5/34 (15%) NR/24 5/74 (7%) 4/85 (5%) 10/233 (4%) 1/48 (2%) 36/400 (9%) 5/113 (4%) 6/60 (10%) 0/39 (0%) 5/50 (10%) 5/34 (15%) 3/41 (7%) 2/50 (4%) 5/145 (3%) 6/32 (19%) 1/55 (2%) 6/128 (5%) 1/26 (4%) 0/80 (0%) 1/50 (2%) 1/98 (1%) 7/188 (4%) 0/20 (0%) 5/50 (10%) 4/41 (10%) NR/22 6/73 (8%) 3/84 (4%) 13/234 (6%) 2/48 (4%) 40/402 (10%) 5/111 (5%) 2/59 (3%) 0/36 (0%)

Neoadjuvant chemotherapy versus surgery alone

Neoadjuvant chemoradiotherapy versus neoadjuvant chemotherapy

Data are n/N (%). NR=not reported. SCC=squamous-cell carcinoma. *Treatment-related death. Intervention=chemoradiotherapy, control=chemotherapy. In-hospital mortality.

Table 2: Perioperative mortality (absolute 30-day or in-hospital mortality) after chemotherapy or chemoradiotherapy

169 patients (85 neoadjuvant chemotherapy and 84 surgery alone) with tumours arising in the oesophagogastric junction and oesophagus were used. The HRs for oesophageal and oesophagogastric junction tumours from this study were pooled to give an HR of 063 (95% CI 045089; gure 3). We identied two studies that included oesophagogastric junction tumours and primary gastric adenocarcinomas, the UK MAGIC trial34 and the European Organisation for Research and treatment of Cancer (EORTC) 40954 study.35 The numbers of patients with lower oesophageal and oesophagogastric cancers were small (n=73 and n=58, respectively in MAGIC34 and n=76 for oesophagogastric in EORTC35), and most patients received a gastrectomy. Also, these studies did not publish outcomes by tumour site and so were excluded from the quantitative meta-analysis. In the UK MAGIC trial,34 there was a signicant survival advantage for patients who received perioperative chemotherapy. However, the EORTC study,35 which examined preoperative treatment, was closed because of
688

poor accrual and did not show a survival advantage for this strategy. Using the HR for the primary oesophageal cancer subgroups from the MAGIC study (HR 075, 95% CI 042133; p=033), as provided by the investigators (personal communication) we did a sensitivity analysis. This analysis did not aect the overall estimate of benet from neoadjuvant chemotherapy (HR 087, 95% CI 087095; p=0001). One additional study had nal results published in abstract form as part of conference proceedings.36 The authors reported a disease-free survival advantage for the chemotherapy arm; however, information about censored observations was not reported. Attempts to obtain further information from the authors were unsuccessful and hence the study was not included in the main analysis. We did a sensitivity analysis with an HR calculated by estimating the relative risk from the overall survival rates reported at years 3 and 5 in each treatment arm, by a method described by Parmar.29 The inclusion of this study did not substantially change the pooled HR for neoadjuvant chemotherapy compared with surgery alone (HR 086, 95% CI 079094; p=0001). A test for potential publication bias suggested that there are zero potentially unpublished studies on neoadjuvant chemotherapy compared with surgery alone, and 15 further studies with an HR of 10 would need to be added to these ten studies to produce an overall non-signicant result. Figure 3 shows the pooled estimate for all-cause mortality for trials that compared neoadjuvant chemotherapy followed by surgery with surgery alone. The pooled HR was 087 (079096; p=0005). This corresponds to an absolute survival dierence at 2 years of 51%, which is equivalent to a number needed to treat of 19. A subgroup analysis by histological type for those studies where histology data were available gave an HR for squamous-cell carcinoma of 092 (95% CI 081104; p=018) and for adenocarcinoma of 083 (071095; p=001). There was no evidence of signicant heterogeneity overall or between subgroups by tumour histology. For the second component of the meta-analysis, assessing the survival benets of neoadjuvant chemotherapy compared with neoadjuvant chemoradiotherapy, we identied two studies, one published before18 and one published after15 the database search (n=194). The German study18 included only patients with oesophagogastric junction adenocarcinoma and used 15 weeks of chemotherapy in each arm and 18% had a gastrectomy only. The subsequent Australian trial compared these modalities, but used two cycles of cisplatin and uorouracil and radiotherapy of 35 Gy in 15 fractions starting on day 22.15 The outcomes in the chemoradiotherapy arm were similar to those of the German study, but the preoperative chemotherapy group had higher median and 3-year survivals.15 Neither trial showed an advantage for neoadjuvant chemoradiotherapy over neoadjuvant chemotherapy, although both trials
www.thelancet.com/oncology Vol 12 July 2011

Articles

closed prematurely and were consequently underpowered to detect a signicant survival advantage. We combined these two studies with the pooled results of the neoadjuvant chemoradiotherapy and neoadjuvant chemotherapy arms from the other analysed studies to compare the survival benets of neoadjuvant chemoradiotherapy and neoadjuvant chemotherapy (n=2220). Figure 4 shows the estimates for all-cause mortality for the two individual studies and the pooled indirect comparison. The HR from the randomised comparisons was 077 (95% CI 053112; p=017), in favour of neoadjuvant chemoradiotherapy. From the indirect comparison, the HR was 090 (077104; p=015). Combining the indirect and direct comparisons yielded an overall HR of 088 (076101; p=007). Compared with the direct comparison, the discrepancy of the indirect comparison (ie, ratio of the indirect and direct HRs) was 17% larger (HR 117, 078175; p=045). We also examined whether the potential benet of the neoadjuvant treatment regimens was oset by a higher mortality rate by comparing the benet (HR) with the absolute 30-day or in-hospital mortality dierence (as a percentage) between the control and intervention arms (table 2 and gure 5). There was little association between risk of postoperative mortality (in-hospital or 30-day postoperative death) and the neoadjuvant interventions. We examined the relation between the benets achieved by the addition of neoadjuvant treatment and the outcome in the surgery alone group by plotting the logarithm of the HR against the absolute 2-year survival in the control group for each trial. The webappendix p 2 shows this nding for chemotherapy and chemoradiotherapy. As the 2-year survival in the control group increased (a better prognosis for patients in the trial), there seemed to be a smaller benet from the intervention, but this was not statistically signicant (neoadjuvant chemoradiotherapy p=084; neoadjuvant chemotherapy p=082).

A
20 Favours surgery alone Nygaard9 Walsh adenocarcinoma14 10 5 0 5 10 Tepper43 Lv41 Mariette11 van der Gaast42 Burmeister22 Urba40 Lee17 Le Prise10 Bossett12

Absolute mortality dierence (%)

Favours chemoradiotherapy

20

Walsh SCC13

Apinop39

Hazard ratio for chemoradiotherapy

B
20 Favours surgery alone Roth23 Schlag19 Ancona21 Nygaard9 10 5 0 5 Favours chemotherapy 10 Boonstra37 Ychou7 Allum1 Kelsen8 Law20

Absolute mortality dierence (%)

20

03

04

05

06 07 08 09 Hazard ratio for chemotherapy

10

11

12

13

Discussion
Survival benets of neoadjuvant chemoradiotherapy and neoadjuvant chemotherapy were shown in the previous meta-analysis by our group.6 This updated analysis included updated data on previously published studies and additional studies, with 43% more patients and about 57% more events compared with the previous metaanalysis. The additional information has strengthened the evidence of a survival advantage of neoadjuvant therapy compared with surgery alone. In the present metaanalysis, there is evidence in favour of both neoadjuvant chemoradiotherapy and neoadjuvant chemotherapy before surgery. The evidence supporting the use of chemoradiotherapy is not only stronger than previously reported but is also clear for both squamous-cell carcinoma and adenocarcinoma histologies. Neoadjuvant chemotherapy also seemed to be associated with improvements in each histological subtype compared with surgery alone, although the treatment eects were not as large as for
www.thelancet.com/oncology Vol 12 July 2011

Figure 5: Eect of chemoradiotherapy or chemotherapy on 30-day or in-hospital absolute mortality Eect of chemoradiotherapy (A) and chemotherapy (B) compared with surgery alone in patients with oesophageal carcinoma.

chemoradiotherapy. Both treatment strategies cause toxicities that are well known and that potentially increase the risk of surgical morbidity. There is evidence that neoadjuvant chemoradiotherapy increases the rate of complete resection,15,47 particularly for patients with locally advanced disease, although this increase has not always translated into a survival benet in individual studies. The two most recent trials that assessed neoadjuvant chemoradiotherapy have been reported as abstracts only.11,42 In the larger Dutch trial,42 there was a survival benet for patients who had neoadjuvant chemoradiotherapy with paclitaxel and carboplatin weekly for 5 weeks with 414 Gy radiotherapy compared with those who had surgery alone. In the French trial,11 in which patients received the typical
689

Articles

combination of 45 Gy concurrent with two cycles of cisplatin and uorouracil for early-stage disease (stage 1 or 2a, 68%), there was weak evidence of higher perioperative mortality with no dierence in survival compared with patients who had surgery alone. The dierences in outcomes between these two studies might be because of the tumour stage, choice of chemotherapy, or both. In another study of neoadjuvant chemoradiotherapy, taxanes were used.41 Compared with the Dutch trial, a higher dose of radiotherapy (40 Gy in 2 Gy fractions over 4 weeks) and dierent schedule of paclitaxel (135 mg/m every 3 weeks for two cycles) was used. This trial also reported a survival benet compared with surgery alone and supports the further development of taxane-based chemoradiotherapy schedules. The previous meta-analysis showed a proportionately larger benet for chemoradiotherapy prior to surgery than for chemotherapy prior to surgery, although both were associated with a survival benet compared with surgery alone. Many centres use the favourable results of large trials of neoadjuvant or perioperative chemotherapy compared with surgery alone (UK MAGIC,34 MRC OEO2,1 and FFCD 97037) to justify the exclusion of preoperative radiotherapy in patients with adenocarcinoma of the lower oesophagus and oesophagogastric junction. In updating this meta-analysis, we sought to assess both potential benets from the addition of radiotherapy to chemotherapy prior to surgery, and potential increases in mortality from the combined approach. Only two studies that met the inclusion criteria for our meta-analysis have directly compared neoadjuvant chemoradiotherapy with neoadjuvant chemotherapy. We wanted to quantify the relative survival benets when these two treatment modalities were compared directly to better inform decisions on future trial design and treatment decisions for individual patients. Therefore, to quantify any dierence in outcome from adding radiation to preoperative chemotherapy, we compared the treatment arms of the trials we have analysed. The indirect comparison of neoadjuvant chemoradiotherapy with neoadjuvant chemotherapy showed weak evidence in favour of neoadjuvant chemoradiotherapy, but this comparison might also have been prone to selection bias. Previous meta-analyses of chemoradiotherapy have been criticised because of small sample sizes; heterogeneity of the tumour types, radiation doses, and chemotherapy regimens; methods of preoperative staging and treatment delivery; and adequacy of surgical resection.48 Assessment of the early neoadjuvant chemotherapy trials might attract the same criticisms. The benet of adding radiotherapy might be at least in part due to improved locoregional control when suboptimal surgical resection has occurred. Evidence from multicentre data shows that lymph node retrieval at resection aects outcome.49,50 Future trials should include modern staging methods to facilitate appropriate stratication of patients and measures for assessing the quality of surgery.
690

We took into account the possibility that any survival gains from the addition of either neoadjuvant chemoradiotherapy or neoadjuvant chemotherapy before surgery might be oset by surgical mortality. The view that the overall survival advantage gained from neoadjuvant treatment is oset by higher 30-day postoperative or in-hospital mortality was not supported (table 2, gure 5, and webappendix p 2). Surgical morbidity was inconsistently reported across trials and direct comparisons were not possible. In trying to assess the relevant published work, another issue is the overlap between primary gastric, oesophagogastric junction, and lower oesophageal cancers. Assessment of the eects of neoadjuvant treatments on survival by tumour site was not possible in this analysis (oesophagogastric junction vs lower oesophageal) for patients with adenocarcinoma because most of the included studies did not provide this information. The role of preoperative radiotherapy for oesophageal and oesophagogastric junction cancer is further confused by the meta-analysis from Germany, reported in abstract form.51 This meta-analysis included all studies that assessed patients with gastric, oesophageal, and oesophagogastric junction adenocarcinoma who received preoperative chemotherapy with or without radiotherapy. Postoperative therapy was not assessed. The authors concluded that neoadjuvant chemotherapy is benecial, but the subgroup analysis that assessed the role of preoperative radiotherapy was underpowered because this modality was not used in the patients with adenocarcinoma conned to the stomach. These patients reported the highest benet for oesophagogastric junction cancers, but the statistical power of this subgroup analysis was low. This nding highlights the need to dierentiate and analyse separately the sites of adenocarcinomas in future trials. Building on the results of the previous publication,6 this meta-analysis has several strengths. Only minor publication bias was noted, and sensitivity analyses suggested that unpublished results would not alter conclusions for primary resectable oesophageal cancers. This study also compared results of two neoadjuvant treatment strategies (neoadjuvant chemoradiotherapy and neoadjuvant chemotherapy), although most of the weighted result came from the indirect comparison of pooled outcomes. In the indirect comparison, the surgery alone arms of all the trials were assumed to be homogeneous, and although there was no statistical inconsistency between the results from the direct and indirect comparisons, this comparison might have been aected by selection bias and must be interpreted with caution. Investigation of the eects of heterogeneity of tumour stage was not possible; the benets of neoadjuvant chemoradiotherapy (and even neoadjuvant radiotherapy) might exist only for locally advanced tumours (T2+ and N1+) where downstaging is likely to be greatest, with minimum or no benet in early cancers (T1). Three trials restricted enrolment to early-stage tumours,10,11,46 all
www.thelancet.com/oncology Vol 12 July 2011

Articles

of which examined neoadjuvant chemoradiotherapy compared with surgery alone. Two of these trials only included squamous-cell carcinomas and used earlier staging systems,10,46 and the other, more recent trial11 included both adenocarcinoma and squamous-cell carcinoma. None of these trials reported survival benets for the intervention, and the study by Mariette and colleagues11 reported a signicant survival benet in favour of the control arm. These trials used dierent radiation doses, chemotherapy schedules, and staging systems. None used endoscopic ultrasound or PET routinely; therefore rm conclusions as to whether the absence of benet in these trials is a result of disease stage or the treatment given are dicult. An individual patient data meta-analysis that examined the role of neoadjuvant therapy over surgery alone in more detail according to histology and tumour site would be of value in investigating these questions. Other potential limitations have been discussed previously.6 The optimum neoadjuvant treatment regimen has not been established, because included trials used dierent drugs, doses, and schedules of chemotherapy and radiotherapy. The inclusion of chemoradiotherapy arms in future trials of neoadjuvant treatments is supported by the results of this meta-analysis. Treatment decisions for individual patients should take into account the risk of morbidity from neoadjuvant therapy, and the eects of each treatment on quality of life should be considered, which was not done in these analyses. The focus of future trials should be on identication of the optimum regimen of neoadjuvant therapy and should aim to minimise treatment toxicities and eect on quality of life, as well as attempt to identify and select those patients most likely to benet from specic treatment options. Based on present eveidence, neoadjuvant chemoradiotherapy improves survival compared with surgery alone for patients with operable oesophageal cancer. Neoadjuvant chemotherapy is also associated with improvements in survival compared with surgery alone and can be deemed a standard treatment. Although the benet for neoadjuvant chemotherapy was not as great as for neoadjuvant chemoradiotherapy, a clear advantage has not been established and further randomised trials comparing these two strategies directly are warranted.
Contributors KMS, BHB, BMS, JRZ, RJS, and VG designed the study. KMS and VG did the literature search, created the gures, and extracted data. KMS, BHB, BMS, AB, and VG collected and contributed data. All authors interpreted and analysed the data and wrote the manuscript. Conicts of interest JRZ is a board member of Australasian Gastro-Intestinal Trials Group (AGITG) who pay for his travel to meetings. JRZ and RJS receive grant funding from Cancer Australia and Cancer Institute (NSW) to the National Health and Medical Research Council Clinical Trials Centre and AGITG as program grants or infrastructure funding. All other authors declare that they have no conicts of interest. Acknowledgments This meta-analysis was funded by grants from Cancer Australia and the NSW Cancer Institute to the National Health and Medical Research

Council Clinical Trials Centre and the Australasian Gastro-Intestinal Trials Group. We thank Rhana Pike, who is funded by the Australian National Health and Medical Research Council Clinical Trials Centre, for her valuable assistance with the report; Kathryn Winter for providing results on behalf of the RTOG from their study RTOG 8911; Jurjien Boonstra for providing results from his updated manuscript; Sally Stenning and David Cunningham for providing information on the oesophageal cancer subgroup from the MAGIC trial on behalf of the Medical Research Council; and Stephen Thompson for advice on the concept. References 1 Allum WH, Stenning SP, Bancewicz J, Clark PI, Langley RE. Long-term results of a randomized trial of surgery with or without preoperative chemotherapy in esophageal cancer. J Clin Oncol 2009; 27: 506267. 2 Mariette C, Piessen G, Triboulet J-P. Therapeutic strategies in oesophageal carcinoma: role of surgery and other modalities. Lancet Oncol 2007; 8: 54553. 3 Ando N, Iizuka T, Ide H, et al, for the Japan Clinical Oncology Group. Surgery plus chemotherapy compared with surgery alone for localized squamous cell carcinoma of the thoracic esophagus: a Japan Clinical Oncology Group StudyJCOG9204. J Clin Oncol 2003; 21: 459296. 4 Arnott SJ, Duncan W, Gignoux M, et al. Preoperative radiotherapy for esophageal carcinoma. Cochrane Database Syst Rev 2005; 4: CD001799. 5 Zieren HU, Mller JM, Jacobi CA, Pichlmaier H, Mller RP, Staar S. Adjuvant postoperative radiation therapy after curative resection of squamous cell carcinoma of the thoracic esophagus: a prospective randomized study. World J Surg 1995; 19: 44449. 6 Gebski V, Burmeister B, Smithers BM, et al. Survival benets from neoadjuvant chemoradiotherapy or chemotherapy in oesophageal carcinoma: a meta-analysis. Lancet Oncol 2007; 8: 22634. 7 Ychou M, Boige V, Pignon J-P, et al. Perioperative chemotherapy compared with surgery alone for resectable gastroesophageal adenocarcinoma: an FNCLCC and FFCD multicenter phase III trial. J Clin Oncol 2011; 29: 171521. 8 Kelsen DP, Winter KA, Gunderson LL, et al. Long-term results of RTOG trial 8911 (USA Intergroup 113): a random assignment trial comparison of chemotherapy followed by surgery compared with surgery alone for esophageal cancer. J Clin Oncol 2007; 25: 371925. 9 Nygaard K, Hagen S, Hansen HS, et al. Pre-operative radiotherapy prolongs survival in operable esophageal carcinoma: a randomized, multicenter study of pre-operative radiotherapy and chemotherapy. The second Scandinavian trial in esophageal cancer. World J Surg 1992; 16: 110409. 10 Le Prise E, Etienne PL, Meunier B, et al. A randomized study of chemotherapy, radiation therapy, and surgery versus surgery for localized squamous cell carcinoma of the esophagus. Cancer 1994 ; 73: 177984. 11 Mariette C, Seitz JF, Maillard E, et al. Surgery alone versus chemoradiotherapy followed by surgery for localized esophageal cancer: analysis of a randomized controlled phase III trial FFCD 9901. Proc Am Soc Clin Oncol 2010; 28 (15 suppl): 4005 (abstr). 12 Bosset JF, Gignoux M, Triboulet JP, et al. Chemoradiotherapy followed by surgery compared with surgery alone in squamous-cell cancer of the esophagus. N Engl J Med 1997; 337: 16167. 13 Walsh T. The role of multimodality therapy in improving survival: a prospective randomised trial. In: Predicting, dening and improving outcomes for oesophageal carcinoma [MD thesis]. Dublin: Trinity College, University of Dublin, 1995: 12450. 14 Walsh TN, Noonan N, Hollywood D, Kelly A, Keeling N, Hennessy TP. A comparison of multimodal therapy and surgery for esophageal adenocarcinoma. N Engl J Med 1996; 335: 46267. 15 Burmeister B, Thomas JM, Burmeister EA, et al. Is concurrent radiation therapy required in patients receiving preoperative chemotherapy for adenocarcinoma of the oesophagus? A randomised phase II trial. Eur J Cancer 2011; 47: 35460. 16 Tepper JE, Krasna M, Niedzwiecki D, et al. Superiority of trimodality therapy to surgery alone in esophageal cancer: results of CALGB 9781. Proc Am Soc Clin Oncol 2006; 24 (18 suppl): 4012 (abstr). 17 Lee JL, Park SI, Kim SB, et al. A single institutional phase III trial of preoperative chemotherapy with hyperfractionation radiotherapy plus surgery versus surgery alone for resectable esophageal squamous cell carcinoma. Ann Oncol 2004; 15: 94754.

www.thelancet.com/oncology Vol 12 July 2011

691

Articles

18

19

20

21

22

23

24

25

26

27

28

29

30 31 32

33

34

35

Stahl M, Walz MK, Stuschke M, et al. Phase III comparison of preoperative chemotherapy compared with chemoradiotherapy in patients with locally advanced adenocarcinoma of the esophagogastric junction. J Clin Oncol 2009; 27: 85156. Schlag PM. Randomized trial of preoperative chemotherapy for squamous cell cancer of the esophagus. The Chirurgische Arbeitsgemeinschaft Fuer Onkologie der Deutschen Gesellschaft Fuer Chirurgie Study Group. Arch Surg 1992; 127: 144650. Law S, Fok M, Chow S, Chu KM, Wong J. Preoperative chemotherapy versus surgical therapy alone for squamous cell carcinoma of the esophagus: a prospective randomized trial. J Thorac Cardiovasc Surg 1997; 114: 21017. Ancona E, Ruol A, Santi S, et al. Only pathologic complete response to neoadjuvant chemotherapy improves signicantly the long term survival of patients with resectable esophageal squamous cell carcinoma: nal report of a randomized, controlled trial of preoperative chemotherapy versus surgery alone. Cancer 2001; 91: 216574. Burmeister BH, Smithers BM, Gebski V, et al, for the Trans-Tasman Radiation Oncology Group; Australasian Gastro-Intestinal Trials Group. Surgery alone versus chemoradiotherapy followed by surgery for resectable cancer of the oesophagus: a randomised controlled phase III trial. Lancet Oncol 2005; 6: 65968. Roth JA, Pass HI, Flanagan MM, Graeber GM, Rosenberg JC, Steinberg S. Randomized clinical trial of preoperative and postoperative adjuvant chemotherapy with cisplatin, vindesine, and bleomycin for carcinoma of the esophagus. J Thorac Cardiovasc Surg 1988; 96: 24248. Maipang T, Vasinanukorn P, Petpichetchian C, et al. Induction chemotherapy in the treatment of patients with carcinoma of the esophagus. J Surg Oncol 1994; 56: 19197. Wang C, Ding T, Chang L. [A randomized clinical study of preoperative chemotherapy for esophageal carcinoma]. Zhonghua Zhong Liu Za Zhi 2001; 23: 25455 (in Chinese). Baba M, Natsugoe S, Shimada M, et al. Prospective evaluation of preoperative chemotherapy in resectable squamous cell carcinoma of the thoracic esophagus. Dis Esophagus 2000; 13: 13641. Natsugoe S, Okumura H, Matsumoto M, et al. Randomized controlled study on preoperative chemoradiotherapy followed by surgery versus surgery alone for esophageal squamous cell cancer in a single institution. Dis Esophagus 2006; 19: 46872. Kok T, Lanschot J, Siersema P, et al. Neoadjuvant chemotherapy in operable esophageal squamous cell cancer: nal report of a phase III multicenter randomized controlled trial. Proc Am Soc Clin Oncol 1997; 16: 277a (abstr 984). Parmar MK, Torri V, Stewart L. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med 1998; 17: 281534. Gleser LJ, Olkin I. Models for estimating the number of unpublished studies. Stat Med 1996; 15: 2493507. Fisher R. Statistical methods for research workers, 4th edn. London: Oliver and Boyd, 1932. Song F, Altman DG, Glenny A-M, Deeks JJ. Validity of indirect comparison for estimating ecacy of competing interventions: empirical evidence from published meta-analyses. BMJ 2003; 326: 472. Peng L, Xie TP, Han YT, et al. Randomized controlled study on preoperative concurrent chemoradiotherapy versus surgery alone for esophageal squamous cell carcinoma. Zhongliu 2008; 28: 62022. Cunningham D, Allum WH, Stenning SP, et al. Perioperative chemotherapy versus surgery alone for resectable gastroesophageal cancer. N Engl J Med 2006; 355: 1120. Schuhmacher C, Gretschel S, Lordick F, et al. Neoadjuvant chemotherapy compared with surgery alone for locally advanced cancer of the stomach and cardia: European Organisation for Research and Treatment of Cancer Randomized Trial 40954. J Clin Oncol 2010; 28: 521018.

36

37

38

39 40

41

42

43

44

45

46

47

48

49

50

51

Bokhyan V, Stilidi I, Malikhova O, Tryakin A. 6553 Neoadjuvant chemotherapy followed by transthoracic resection for locally advanced carcinoma of the esophagus: a randomized study. EJC Suppl 2009; 7: 377. Boonstra JJ, Kok T, Wijnhoven BPL, et al. Chemotherapy followed by surgery versus surgery alone in patients with oesophageal squamous cell carcinoma: long-term results of a randomized controlled trial. BMC Cancer 2011; published online May 19. DOI:10.1186/1471-2407-11-181. Medical Research Council Oesophageal Cancer Working Group. Surgical resection with or without preoperative chemotherapy in oesophageal cancer: a randomised controlled trial. Lancet 2002; 359: 172733. Apinop C, Puttisak P, Preecha N. A prospective study of combined therapy in esophageal cancer. Hepatogastroenterology 1994; 41: 39193. Urba SG, Orringer MB, Turrisi A, Iannettoni M, Forastiere A, Strawderman M. Randomized trial of preoperative chemoradiation versus surgery alone in patients with locoregional esophageal carcinoma. J Clin Oncol 2001; 19: 30513. Lv J, Cao XF, Zhu B, Ji L, Tao L, Wang DD. Long-term ecacy of perioperative chemoradiotherapy on esophageal squamous cell carcinoma. World J Gastroenterol 2010; 16: 164954. van der Gaast A, van Hagen P, Hulshof M, et al. Eect of preoperative concurrent chemoradiotherapy on survival of patients with resectable esophageal or esophagogastric junction cancer: results from a multicenter randomized phase III study. Proc Am Soc Clin Oncol 2010; 28 (15 suppl): 4004 (abstr). Tepper J, Krasna MJ, Niedzwiecki D, et al. Phase III trial of trimodality therapy with cisplatin, uorouracil, radiotherapy, and surgery compared with surgery alone for esophageal cancer: CALGB 9781. J Clin Oncol 2008; 26: 108692. Rosa AR, Gurski RR, Schirmer CC, Brentano L, Kruel CD. Survival and prognostic factors in patients with resected epidermoid oesophageal carcinoma. Int Surg 1999; 84: 19398. Greene FL, Page DL, Fleming ID, et al. Esophagus. In: AJCC Cancer Staging Manual, 6th edn. American Joint Committee on Cancer. New York, NY: Springer, 2002: 9198. Bosset J-F, Gignoux M, Triboulet J-P, et al. Chemoradiotherapy followed by surgery compared with surgery alone in squamous-cell cancer of the esophagus. N Engl J Med 1997; 337: 16167. Staal E, Aleman BMP, Boot H, van Velthuysen MLF, van Tinteren H, van Sandick JW. Systematic review of the benets and risks of neoadjuvant chemoradiation for oesophageal cancer. Br J Surg 2010; 97: 148296. Wijnhoven BP, van Lanschot JJ, Tilanus HW, Steyerberg EW, van der Gaast A. Neoadjuvant chemoradiotherapy for esophageal cancer: a review of meta-analyses. World J Surg 2009; 33: 260614. Smit JK, Pultrum BB, van Dullemen HM, Van Dam GM, Groen H, Plukker JT. Prognostic factors and patterns of recurrence in esophageal cancer assert arguments for extended two-eld transthoracic esophagectomy. Am J Surg 2010; 200: 44653. Solomon N, Zhuge Y, Cheung M, Franceschi D, Koniaris LG. The roles of neoadjuvant radiotherapy and lymphadenectomy in the treatment of esophageal adenocarcinoma. Ann Surg Oncol 2010; 17: 791803. Ronellentsch U, Schwarzbach M, Hofheinz R, et al. Meta-analysis of preoperative chemotherapy (CTX) versus primary surgery for locoregionally advanced adenocarcinoma of the stomach, gastroesophageal junction, and lower esophagus (GE adenocarcinoma). Proc Am Soc Clin Oncol 2010; 28 (suppl): abstr 4022.

692

www.thelancet.com/oncology Vol 12 July 2011