Neoadjuvant chemoradiation therapy for resectable esophago-gastric adenocarcinoma: a meta-analysis of randomized clinical trials

To identify useful studies and published abstracts,we systematically searched electronic databases including the Cochrane Central Register of Controlled Trials (CENTRAL), PubMed, Excerpta Medica Database (EMBASE), the Cochrane Database of Systematic Reviews and the China National Knowledge Infrastructure (CNKI). There were no language restrictions. The medical subject headings were listed as follows: esophagus/gastroesophageal /gastric adenocarcinoma, preoperative chemoradiation therapy, and randomized controlled trials (RCT). The search included literature published from January 1990 to July 2014. We also reviewed all abstracts that were potentially relevant to our subject. Furthermore, other grey literature as well as unpublished work, ongoing studies and negative results were searched as well. Two investigators conducted the search independently, and their results were combined.

Two authors independently reviewed the study. The titles and abstracts were in agreement with the articles to be retrieved. To identify studies for the analysis, the inclusion criteria were designed as follows: (1) published RCTs that had a clear statement in the Materials and Methods section. (2) RCTs comparing preoperative CRT plus surgery with surgery alone or preoperative CRT plus surgery with preoperative chemotherapy plus surgery. (3) RCTs including patients with resectable, histologically proven adenocarcinoma of the esophagus, stomach or gastroesophageal junction without metastatic disease. (4) RCTs with a low risk of selection bias, performance bias, detection bias, attrition bias, reporting bias and other bias. Bias was assessed using Begg’s and Egger’s tests [22,23]. (5) Patient survival was used as the measureable outcome.

The primary outcome was overall survival, mostly based on an intention-to-treat analysis. The secondary outcomes were the R0 resection rate, which was defined by a tumor-free resection margin; positive rate of lymph node metastasis; postoperative recurrence rate; complete pathological response rate; and perioperative mortality.

Data analysis was performed using Review Manager 5.2.0 for Windows. Overall, survival was measured with a hazard ratio (HR), while the R0 resection rate, positive rate of lymph node metastasis, postoperative recurrence rate and perioperative mortality were measured using odds ratios (OR). Furthermore, intention-to-treat (ITT) analyses were conducted when possible. If permitted, HR and the corresponding standard errors were obtained directly from the article; otherwise, they were calculated using the methods of Parmar [24], Tierney [25], and Williamson [26]. These approaches use confidence intervals, log-rank p-values, number of events and Kaplan–Meier survival curves to estimate the HR and standard errors. Moreover, the measures of HR and OR were investigated for statistical heterogeneity by I² statistics, with a value of I² > 50% indicating substantial heterogeneity. Where there was evidence of heterogeneity, subgroup analysis or sensitivity analysis were performed to investigate possible bias and derived summary estimates according to the random effect model; otherwise, the Mantel-Haenszel fixed effect model was used to compute the results. All of the significance tests were two-sided, with p = 0.05 as the cutoff.

The results of the literature search are displayed in a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram (Figure 1). A total of 1522 studies were retrieved in the database, and 4 additional studies were found from other channels, such as conference reports, and so on. Among them, 1519 records were mostly unrelated to our subject, and only 7 RCTs met our inclusion criteria, which examined a total of 1085 patients. The main features of the trials included in the meta-analysis are shown in Table 1. The seven RCTs included 869 patients with esophago-gastric adenocarcinoma, 430 of whom received CRT before surgery. Approximately 5 RCTs focused on the topic of preoperative chemoradiation therapy (CRT) followed by surgery versus surgery alone [27-31], while 2 RCTs focused on the topic of CRT followed by surgery versus preoperative chemotherapy (CT) followed by surgery [32,33]. The cancer positions involved in the 6 RCTs were of the esophagus and gastro-esophageal junction, and 2 RCTs referred to the cardia. In addition, the median age of patients ranged from 56 to 65 year, and the proportion of females was 18.4%. Table 2 and Table 3 display some other characteristics regarding the RCTs included in our study. The total sample size of our meta-analysis was 1085, which contained 869 patients with adenocarcinoma, while the number of each RCTs varied greatly. The 3 year and 5 year overall survival (OS) rate are also displayed in Tables 2 and 3, where there was a certain difference between the CRT plus surgery group and the surgery alone or CT plus surgery groups. The treatment schedule is also listed (Tables 2 and 3). Additionally, no publication bias was found from the funnel plots (Figure 2).

The primary outcome examined in our study was overall survival and was reported in all seven RCTs. The meta-analysis gave the result that the pooled HR was 0.74 (95% CI 0.63–0.88) for the preoperative CRT plus surgery group compared to the preoperative CT plus surgery or surgery alone groups (Figure 3). It is noteworthy that four RCTs not only contained adenocarcinoma but also contained squamous cell carcinoma (SCC) [28-31]. Therefore, the individual HR of the CROSS trial and the TROG trial, excluding SCC, were calculated separately from the data given by the original article, and the results were 0.73 (95% CI 0.54-1.00) in the CROSS trial and 1.02 (95% CI 0.72-1.44) in the TROG trial. In the CALGB9781 trial and Urba’s trial, the individual data of SCC were not displayed; however, the number of SCC in these two studies was only 37. Compared to the total number of 869, the interference of these 37 SCC patients could be ignored. As the heterogeneity test was not statistically significant (I² = 13%), the fixed effect model was used to calculate the result for OS. Figure 3 shows Forest plots for OS. The individual HR ranged from 0.45 (95% CI 0.20-1.01) for the CALGB9781 trial to 1.02 (95% CI 0.72-1.44) for the TROG trial; only a single individual HR favored the preoperative CRT group. This result together with those of the pooled HR indicated that there was a survival advantage for patients with preoperative CRT followed by surgery. In addition, according to the results of the heterogeneity test, there was no need to perform sensitivity analysis or subgroup analysis for the primary outcome.

Figure 4, 5, 6, 7 show Forest plots for the secondary outcomes, including the R0 resection rate, positive rate of lymph node metastasis, postoperative recurrence rate, pathological complete response rate (pCR) and perioperative mortality. Five RCTs reported the R0 resection rate, indicating a statistically significant difference (OR 2.35, 95% CI 1.29–4.30, Figure 3). As the heterogeneity test was statistically significant (I² = 59%), random effect modeling and subgroup analysis were performed, revealing a significant difference in the comparison between preoperative CRT and surgery alone (OR 3.55, 95% CI 2.34–5.39), while no statistically significant difference was observed in the comparison between preoperative CRT and preoperative CT (OR 1.17, 95% CI 0.61–2.27). Five RCTs reported a postoperative recurrence rate that included local and distant failure, indicating a statistically significant difference (OR 0.51, 95% CI 0.38–0.68, Figure 2), and no statistical heterogeneity was detected (I² = 0%). Five RCTs reported the positive rate of lymph node metastasis according to the postoperative pathological report. The pooled OR was 0.30 (95% CI 0.23-0.39), revealing a significant difference, and no statistical heterogeneity was detected (I² = 49%). Five RCTs reported perioperative mortality, and there was no statistically significant difference between the two groups (OR 1.10, 95% CI 0.62–1.93, Figure 4) nor a statistical heterogeneity (I² = 0%). Six RCTs reported pCR after chemoradiation therapy, while two trials revealed pCR after chemotherapy, and the result of pCR in the chemoradiation therapy group was 21.56% (Table 1).