食管癌6

Lancet Oncol 2011; 12: 681–92Published Online June 17, 2011DOI:10.1016/S1470-2045(11)70142-5See Comment page 615National 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, PrincessAlexandra 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 SoftTissue Unit, Princess Alexandra Hospital, Brisbane, QLD, Australia (B M Smithers, A Barbour); and PeterMacCallum Cancer Centre, and Department of Medicine, University of Melbourne, Melbourne, VIC, Australia (Prof J R Zalcberg FRACP)Correspondence to:Prof Val Gebski, National Healthand Medical Research Council Clinical Trials Centre, LockedBag 77, Camperdown 1450 NSW, Australiaval@.auSurvival after neoadjuvant chemotherapy orchemoradiotherapy for resectable oesophageal carcinoma: an updated meta-analysisKatrin M Sjoquist, Bryan H Burmeister, B Mark Smithers, John R Zalcberg, R John Simes, Andrew Barbour, Val Gebski, for the Australasian Gastro-Intestinal Trials GroupSummaryBackground In a previous meta-analysis, we identifi ed a survival benefi t 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 benefi ts of preoperative neoadjuvant chemotherapy compared with neoadjuvant chemoradiotherapy.Methods To identify additional studies and published abstracts from major scientifi c 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 eff ects by tumour histology and relations between risk (survival after surgery alone) and eff ect 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 0·78 (95% CI 0·70–0·88; p<0·0001); the HR for squamous-cell carcinoma only was 0·80 (0·68–0·93; p=0·004) and for adenocarcinoma only was 0·75 (0·59–0·95; p=0·02). The HR for all-cause mortality for neoadjuvant chemotherapy was 0·87 (0·79–0·96; p=0·005); the HR for squamous-cell carcinoma only was 0·92 (0·81–1·04; p=0·18) and for adenocarcinoma only was 0·83 (0·71–0·95; p=0·01). The HR for the overall indirect comparison of all-cause mortality for neoadjuvant chemoradiotherapy versus neoadjuvant chemotherapy was 0·88 (0·76–1·01; p=0·07).Interpretation This updated meta-analysis provides strong evidence for a survival benefi t of neoadjuvantchemoradiotherapy 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.IntroductionOverall survival of patients with resectable oesophagealcancer remains poor, with a 5-year survival of 15–34%,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 diffi culties 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 signifi cant survival benefi t for neoadjuvant chemoradiotherapy and,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.4Most trials of chemotherapy or combined chemo-radiotherapy have used a doublet of cytotoxics, usually a platinum compound, most often cisplatin, and fl uorouracil,1,7–22 with varying doses and sequencing. Some small trials of neoadjuvant chemotherapy have used triplet chemotherapy, which resulted in increased toxicity without signifi cant survival benefi ts.23,24 Recent treatment modifi cations have included the use of more modern cytotoxic drugs, changes in chemotherapy sequencing, or changes in the dose and fractionation of radiotherapy.2We 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 benefi ts of neoadjuvant chemotherapy and neo-adjuvant chemoradiotherapy before surgery for resectable oesophageal carcinoma, and to assess whether any increase in survival benefi ts was off set by an increase in perioperative mortality.MethodsFor the fi rst section of the meta-analysis, we sought to assess the survival benefi ts 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 oesophago g astric junction carcinoma (squamous-cell carcinoma, adenocarcinoma, or mixed tumours) were eligible for inclusion, provided they were analysed by intention-to-treat.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 subgroupdata could be identified separately, with the same criteria for inclusion as described in the previous meta-analysis.6The primary outcome of interest for the fi rst section of the meta-analysis was all-cause mortality. The secondary endpoint was the eff ect 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 benefi ts of neoadjuvant chemo-therapy compared with neoadjuvant chemoradio t herapy. All randomised trials of this comparison that included adult patients with squamous-cell carcinoma, adeno-carcinoma, or mixed tumours were eligible. We did not identify any previous systematic reviews or meta-analysesof suffi cient quality, and so we searched Medline, Embase, and Central with the same terms as for the fi rst section. All trials published between January, 1980 and November, 2010, were included if the abstract or fullarticle was available in English.For both sections of the meta-analysis, we wrote toauthors of studies that had previously only appeared asabstracts 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.25–28 Trials that focused on primary gastric cancers were excluded if the results for oesophagogastric junction tumours were not available separately.Statistical analysesFor the fi 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 eff ects. 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 χ² test to assess heterogeneity, with the level of signifi cance set at 5%. We assessed publication bias by the methods described by G leser and Olkin,30 and we used Fisher’s failsafe N 31 to identify the number of studies with a p value of 0·5 (ie, an HR of 1·0) that would need to be added to those in the meta-analysis to produce a non-signifi cant result.Figure 1: Flow diagram showing inclusion and exclusion of studies*Includes 17 studies that were included in this meta-analysis and fi 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.See Online for webappendixWe 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 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 comparisonsto obtain estimates of the benefi t of neoadjuvant chemo-radiotherapy compared with neoadjuvant chemo t herapy. 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 diff erences between the eff ects obtained from the direct and indirect comparisons. Consistency between eff ects for indirect and direct comparisons provides confi dence about the additional information in the indirect comparisons. These approaches were described by Song and colleagues,32as was the comparison of the amount of discrepancy between the indirect comparisons and those of the randomised trials; a p value of greater than 0·05 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). Diff erences between outcomes in the control arms of the diff erent trials were measured against the logarithm of the HR to give an estimate of the relation between risk and benefi t of neoadjuvant chemotherapy or neoadjuvant chemoradiotherapy. The number of events (deaths) used was estimated from the sample size and median follow-up time reported for each trial and an assumption that about 50% of events had occurred within 18 months.We also did prespecified subgroup analyses to examine the effects of neoadjuvant chemotherapy or neo a djuvant chemoradiotherapy according to tumour histology (squamous-cell carcinoma or adenocarcinoma) where this information was available.Differences between treatment arms for 30-day postoperative or in-hospital mortality were calculated from the absolute diff erence in the proportion of deaths, with 95% CIs.Role of the funding sourceThe 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 final responsibility, with the agreement of all authors, for the decision to submit for publication.ResultsWe included 24 studies in total (fi gure 1),1,7–15,17–24,37,39–43 which consisted of all 17 trials from the previous meta-analysis8–10,12–14,16,17,19–24,38,39,40 and seven further studies.7,11,15,18,37,41,42 12 were randomised comparisons of neo a djuvant chemo r adio t herapy versus surgery alone (n=1854),10–14,17,22,39–43 nine were randomised comparisons of neoadjuvant chemotherapy versus surgery alone (n=1981),1,7,8,19–21,23,24,37 two compared neoadjuvant chemo-radiotherapy with neoadjuvant chemotherapy (n=194),15,18 and one study included two comparisons and was included in analyses of both neoadjuvant chemo-radiotherapy (n=78) and neoadjuvant chemo t herapy (n=81).9 This study was a 2 × 2 factorial study that simultaneously compared the effect of neoadjuvant chemo t herapy, neoadjuvant chemo r adio t herapy, and neoadjuvant radiotherapy on survival, and the control group was used for both the neoadjuvant chemotherapy and neoadjuvant chemo r adiotherapy 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.6The updated analysis contained 4188patients whereas the previous publication included 2933 patients. TheFigure 2: All-cause mortality for chemoradiotherapy compared with surgery aloneEff ects 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.updated work includes information on 3994patients 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 benefi ts of neo a djuvant treatment with either neoadjuvant chemo-therapy or neoadjuvant chemoradio t herapy,1,7,8–14,17,19–24,37,39–43 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),9,10,12–14,16,17,19–24,38–40 three of which have been updated,1,8,43 and fi ve new eligible studies were identifi ed.7,11,37,41,42 Four studies were excluded from the previous meta-analysis because ofFigure 3: All-cause mortality for chemotherapy compared with surgery aloneEff ects 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.unclear or unavailable information. We wrote to all these authors again but were unable to obtain suffi cient 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 insuffi cient 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 long-term results have recently been published and are now included.37 All patients had stages T0–3, N0–1 disease according to the 2002 AJCC staging.45 There were no major differences in the age of participants between studies or groups, although no formal statistical comparison was done. Three trials included only patients with early (stage I–II) 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 alone10–14,16,17,22,39–42 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 fi 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 identifi ed;33 however, it was not included because the information from the abstract was insuffi cient 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 s tudies f or n eoadjuvant c hemoradiotherapy, but 23 more studies with a HR of 1·0 would need to be added to these 13 to produce an overall non-signifi cant 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 chemoradiot-herapy followed by surgery with surgery alone. The pooled HR was 0·78 (95% CI 0·70–0·88; p<0·0001). This corresponds to an absolute survival benefi t at 2 years of 8·7% and a number needed to treat of 11. There was no evidence of signifi cant heterogeneity between the trials or between pooled results by tumour histology. The survival benefi ts for neoadjuvant chemoradiotherapy were similar in the tumour type subgroups: squamous-cell carcinoma (HR 0·80, 95% CI 0·68–0·93; p=0·004) and adeno-carcinoma (0·75, 0·59–0·95; p=0·02).Ten studies were included in the comparison of neoadjuvant chemotherapy with surgery alone (n=2062): nine were randomised comparisons of neoadjuvant chemotherapy with surgery alone1,7,8,19–21,23,24,37 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 adeno-carcinomas of gastric, oesophagogastric junction, and oesophageal origin, but was eligible for the quantitative analysis because the fi nal publication reported outcomes by tumour location.7 Only the outcomes for theFigure 4: Indirect comparison of all-cause mortality for chemoradiotherapy and chemotherapy169 patients (85 neoadjuvant chemo t herapy 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 0·63 (95% CI 0·45–0·89; fi gure 3). We identifi ed two studies that included oesophagogastric junction tumours and primary gastric adenocarcinomas, the UK MAG IC 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 MAG IC trial,34 there was a signifi cant survival advantage for patients who received perioperative chemotherapy. However, the EORTC study,35 which examined preoperative treatment, was closed because of poor accrual and did not show a survival advantage for this strategy. Using the HR for the primary oesophageal cancer subgroups from the MAG IC study (HR 0·75, 95% CI 0·42–1·33; p=0·33), as provided by the investigators (personal communication) we did a sensitivity analysis. This analysis did not aff ect the overall estimate of benefit from neoadjuvant chemotherapy (HR 0·87, 95% CI 0·87–0·95; p=0·001).One additional study had final 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 0·86, 95% CI 0·79–0·94; p=0·001). A test for potential publication bias suggested that there are zero potentially unpublished studies on neoadjuvant chemo-therapy compared with surgery alone, and 15 further studies with an HR of 1·0 would need to be added to these ten studies to produce an overall non-signifi cant 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 0·87 (0·79–0·96; p=0·005).This corresponds to an absolute survival diff erence at 2 years of 5·1%, 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 0·92 (95% CI 0·81–1·04; p=0·18) and for adeno c arci n oma of 0·83 (0·71–0·95; p=0·01). There was no evi d ence of signi fi cant heterogeneity overall or between sub g roups by tumour histology.For the second component of the meta-analysis, assessing the survival benefits of neoadjuvant chemo-therapy compared with neoadjuvant chemo r adiotherapy, we identifi ed two studies, one published before18 and one published after15 the database search (n=194).TheG erman 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 fl 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 trialsclosed prematurely and were consequently underpowered to detect a signifi cant survival advantage. We combined these two studies with the pooled results of thechemotherapy arms from the other analysed studies to compare the survival benefits of neoadjuvant chemo-radiotherapy 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 0·77 (95% CI 0·53–1·12; p=0·17), in favour of neoadjuvant chemoradiotherapy. From the indirect comparison, the HR was 0·90 (0·77–1·04; p=0·15). Combining the indirect and direct comparisons yielded an overall HR of 0·88 (0·76–1·01; p=0·07). Compared with the direct comparison, the discrepancy of the indirect comparison (ie, ratio of the indirect and direct HRs) was 17% larger (HR 1·17, 0·78–1·75; p=0·45).We also examined whether the potential benefi t of the neoadjuvant treatment regimens was off set by a higher mortality rate by comparing the benefit (HR) with the absolute 30-day or in-hospital mortality diff erence (as a percentage) between the control and intervention arms (table 2 and fi 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 benefi ts 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 finding 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 benefi t from the intervention, but this was not statistically signifi cant (neoadjuvant chemoradiotherapy p=0·84; neoadjuvant chemotherapy p=0·82). DiscussionSurvival benefi ts of neoadjuvant chemoradiotherapy and neoadjuvant chemotherapy were shown in the previous meta-analysis by our group.6 This updated analysisincluded updated data on previously published studies and additional studies, with 43% more patients and about 57% more events compared with the previous meta-analysis. The additional information has strengthened the evidence of a survival advantage of neoadjuvant therapy compared with surgery alone. In the present meta-analysis, there is evidence in favour of both neoadjuvant chemoradiotherapy and neoadjuvant chemotherapy before surgery. The evidence supporting the use of chemo-radiotherapy is not only stronger than previously reported but is also clear for both squamous-cell carcinoma and adenocarcinoma histologies. Neoadjuvant chemo t herapy also seemed to be associated with improvements in each histological subtype compared with surgery alone, although the treatment effects were not as large as for chemoradiotherapy. Both treatment strategies causetoxicities that are well known and that potentially increasethe risk of surgical morbidity.There is evidence that neoadjuvant chemoradiotherapyincreases the rate of complete resection,15,47 particularlyfor patients with locally advanced disease, although thisincrease has not always translated into a survival benefi tin individual studies. The two most recent trials thatassessed neoadjuvant chemoradiotherapy have beenreported as abstracts only.11,42 In the larger Dutch trial,42there was a survival benefit for patients who had neoadjuvant chemoradiotherapy with paclitaxel andcarboplatin weekly for 5 weeks with 41·4 Gy radiotherapycompared with those who had surgery alone.In theFrench trial,11 in which patients received the typicalFigure 5: Effect of chemoradiotherapy or chemotherapy on 30-day or in-hospital absolute mortality Eff ect of chemoradiotherapy (A) and chemotherapy (B) compared with surgery alone in patients with oesophageal carcinoma.。