Elective nodal irradiation versus involved-field irradiation in patients with esophageal cancer receiving neoadjuvant chemoradiotherapy: a network meta-analysis

Background To assess the comparative efficacy and safety of elective nodal irradiation (ENI) and involved-field irradiation (IFI) in patients with esophageal cancer (EC) receiving neoadjuvant chemoradiotherapy plus surgery (nCRTS). Material and methods PubMed, Embase, Cochrane Library, Web of Science and major meetings were searched for randomized controlled trials (RCTs) that compared at least two of the following treatment regimens: nCRTS, neoadjuvant chemotherapy plus surgery (nCTS), and surgery (S) alone. Overall survival (OS) was the primary outcomes of interest, reported as hazard ratio (HR) and 95% confidence intervals (CIs). A Bayesian network meta-analysis was performed to compare all regimens simultaneously. Results Twenty-nine RCTs with a total of 5212 patients were included in the meta-analysis. Both nCRTS adopting ENI (nCRTS-ENI) (HR = 0.63, 95% CI: 0.48–0.83) and nCRTS adopting IFI (nCRTS-IFI) (HR = 0.75, 95% CI: 0.66–0.86) significantly improved OS compared to S alone. No significant differences in OS, locoregional recurrence, distant metastases, R0 resection and postoperative mortality were observed between nCRTS-ENI and nCRTS-IFI. In subgroup analyses, nCRTS-IFI showed a significant OS advantage over nCTS (HR = 0.78, 95% CI: 0.63–0.96) and S alone (HR = 0.50, 95% CI: 0.38–0.68) for esophagus squamous cell carcinoma (ESCC), but nCRTS-ENI did not; nCRTS-ENI using three-dimensional radiotherapy (3D-RT) resulted in an improved OS compared to that with 2D-RT (HR = 0.58, 95% CI: 0.34–0.99). Based on treatment ranking in term of OS, nCRTS-IFI (0.90) and nCRTS-ENI (0.96) was ranked the most effective treatment for ESCC and esophagus adenocarcinoma (EAC), respectively. Conclusion Either adopting ENI or IFI, nCRTS is likely to be the optimal treatment for resectable EC, and nCRTS-IFI and nCRTS-ENI seem to be more effective for patients with ESCC and EAC, respectively. Future head to head comparison trials are needed to confirm these findings.

It should be noted that radiation fields used for patients receiving nCRTS are inconsistent in trials, which might affect the outcomes. Some trials adopted elective nodal irradiation (ENI, nodal target volume covering both metastatic lymph nodes and regional nodes) [17][18][19][20][21][22], and others adopted involved-field irradiation (IFI, nodal target volume including only the metastatic nodes) [5][6][7][8][9][10][11][12][13][14][15][16]. Efficacy of ENI and IFI has been compared in patients with locally advanced EC undergoing radical CRT in some retrospective studies [23][24][25][26], but with different results. At present, no trials have compared the two radiation fields directly in patients undergoing nCRTS, and therefore, there are still questions around which is more superior, and what is the suitable patient population for adopting ENI or IFI.
In light of these issues, we performed a network metaanalysis to assess the comparative effectiveness and safety of ENI and IFI, attempting to identify the best radiation field in patients receiving nCRT.

Literature search strategy
This meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) criteria [27] (Additional file 1: Tables S1). PubMed, Embase, Cochrane Library, Web of Science were searched for the available studies published before April 1, 2019, using the strategy as shown in Additional file 1: Tables S2. The reference lists of retrieved studies were manually scanned for relevant additional studies missed by the electronic search.

Inclusion and exclusion criteria
Studies were included if they met the following criteria: (1) types of studies: RCTs; (2) types of participants: resectable EC; (3) types of interventions: compared at least two of the following treatments: nCRTS, neoadjuvant chemotherapy plus surgery (nCTS), and S alone; and (4) outcomes: overall survival (OS), locoregional recurrence (LR), distant metastases (DM), R0 resection, and postoperative mortality (POM) data. Studies which failed to meet the above criteria were excluded from the network meta-analysis.

Data extraction
The data were extracted by two investigators independently. The following data were extracted from each study: first author or name of individual RCT, years of publication, duration of the study, country of origin, treatments, numbers of patients, pathologic type, and data of OS, LR, DM, R0 resection, and POM.

Quality assessment
The methodological quality of RCTs was assessed by Cochrane risk of bias tool [28], which consists of the following five domains: sequence generation, allocation concealment, blinding, incomplete data, and selective reporting. A RCT was finally rated as "low risk of bias" (all key domains indicated as low risk), "high risk of bias" (one or more key domains indicated as high risk), and "unclear risk of bias".

Statistical analysis
The primary outcome was OS, and the secondary outcomes were LR, DM, R0 resection, and POM. Hazard ratios (HRs) or odds ratios (ORs) and their 95% confidence intervals (CIs) were used as summary statistics. For direct comparisons, standard pairwise meta-analysis was performed. A statistical test for heterogeneity was performed using the chi-square (χ 2 ) and I-square (I 2 ) tests with the significance set at I 2 > 50% or P < 0.10. If significant heterogeneity existed, a random-effects analysis model was used; otherwise, a fixed-effects model was used.
The Bayesian network-meta analysis (NMA) was performed in a random-effect model using Markov chain Monte Carlo methods [29,30] in JAGS and the GeMTC package in R (https://drugis.org/software/r-packages/ gemtc). For each outcome measure, four independent Markov chains were simultaneously run for 20,000 burn-ins and 100,000 inference iterations per chain to obtain the posterior distribution. The traces plot and Brooks-Gelman-Rubin method were used to assess the convergence of model [31]. Treatment effects were estimated by HR/OR and corresponding 95% CI. Network consistency was assessed with node-split models by statistically testing between direct and indirect estimates within treatment loop [32]. To rank probabilities of all available treatments, the surfaces under the cumulative ranking curve (SUCRAs) were calculated [33]. SUCRA equals one if the treatment is certain to be the best and zero if it's certain to be the worst [33]. In addition, we conducted subgroup analyses according to histologic type, RT dose, and RT technique. Lastly, comparisonadjusted funnel plot was used to detect the presence of small-study effects or publication bias [34].

Assessment of included trial
The risk of bias in included RCTs was summarized in Additional file 1: Figure S1. Seven trials [13-16, 21, 22, 48, 49] were judged to be unclear risk of bias, as they had more than three domains indicating as unclear risk. The remaining trials were rated with a low risk of bias. Funnel plot analysis in term of OS did not indicate any evident risk of publication bias (Additional file 1: Figure S2).

Conventional pairwise meta-analysis
Results of direct comparison meta-analysis are shown in Network meta-analysis

Inconsistency assessment and treatment ranking
There were two independent closed loops in the network for OS, LR, DM, and R0 resection: nCRTS-ENI/nCTS/S alone and nCRTS-IFI/nCTS/S alone; one independent closed loop for POM: nCRTS-ENI/nCTS/S alone. Analysis of inconsistency showed that the NMA results were similar to the PWMA results for the five outcomes, which suggested the consistency between the direct and indirect evidence (Additional file 1: Figure S3). Results of the treatment rankings based on SUCRA are shown in Table 4a. In term of OS, nCRTS-ENI (0.93) was ranked the most effective treatment in term of OS, followed by nCRTS-IFI (0.71). nCRTS-IFI (0.95) was ranked the most effective treatment in term of LR, followed by nCRTS-ENI (0.62). With regard to DM, POM, and R0 resection, SUCRA values were similar between nCRTS-ENI and nCRTS-IFI.

Subgroup analyses
NMA results of subgroup analyses are shown in Table 3b (SUCRA values are shown in Table 4b). Subgroup analyses for esophagus squamous cell carcinoma (ESCC) and esophagus adenocarcinoma (EAC) were conducted in 23 trials with 3164 patients and 11 trials with 1997 patients, respectively. With regard to ESCC, nCRTS-IFI showed significant OS advantage over S alone and a trend OS advantage over nCTS, and was ranked the most effective treatment (0.90); nCRTS-ENI had a trend OS benefit over S alone. As for EAC, both nCRTS-ENI and nCRTS-IFI significantly improved OS compared to S alone, and nCRTS-ENI was ranked the best treatment (0.96).
In subgroup analysis according to RT dose (18 trials with 2860 patients), nCRTS-IFI with dose of ≥40Gy significantly improved OS compared to S alone, while nCRTS-IFI with dose of <40Gy did not; both nCRTS-  ENI with dose of ≥40Gy and < 40Gy showed a significant OS advantage over S alone; and nCRTS-ENI with dose of ≥40Gy was ranked the most effective regimen (0.86).
In subgroup analysis according to RT technique (16 trials with 2774 patients), nCRTS-ENI adopting threedimensional radiotherapy (3D-RT) significantly improved OS compared to nCRTS-ENI adopting 2D-RT, nCTS, and S alone, and was ranked the most effective regimen (0.99); nCRTS-IFI was more effective than S alone regardless RT technique adopted.

Discussion
Currently, nCRTS has been the most common treatment approach for patients with resectable EC, but the optimal radiation field remains unidentified. EC is characterized as an aggressive disease, and lymph node metastasis, particularly regional lymph node involvement, usually occurs early. Taking into consideration microscopic spread, some trials adopted ENI instead of IFI for patients receiving nCRTS. In CALGB 9781 trials [18], nCRTS adopting ENI followed by surgery showed a long-term survival advantage over S alone for patients with EC. Nevertheless, there are also trials of a series of cases treated with IFI. Recently, two large phase III trials [5][6][7] also showed that nCRTS improved survival over surgery alone among patients with esophageal or junctional cancer, while IFI was adopted in RT. To date, there are still no trials that have compared efficacy of the two radiation fields directly in EC patients receiving nCRTS, and which is more effective remains unclear.
To our knowledge, this is the first network metaanalysis assessing the comparative efficacy and safety of nCRTS-ENI and nCRTS-IFI for patients with EC. It showed that both nCRTS-ENI and nCRTS-IFI significantly improved OS compared to S alone. nCRTS-ENI also showed significant OS advantage over nCTS. No significant difference in OS, LR, DM, and POM was observed between nCRTS-ENI and nCRTS-IFI. Based on treatment ranking in term of OS, nCRTS-ENI had the highest probability of being the most effective treatment (93%), followed by nCRTS-IFI (71%) and nCTS (36%).
However, in subgroup analysis according to pathologic type, nCRTS-IFI (90%) was ranked the most effective treatment for ESCC, followed by nCRTS-ENI (68%). nCRTS-IFI showed significant and a trend OS advantage over S alone and nCTS, respectively. While nCRTS-ENI only had a trend OS benefit compared to S alone. In the CROSS trial [6, 7], nCRTS-IFI resulted in improved OS for both ESCC and EAC, but the magnitude of this benefit was greater for ESCC patients (HR for ESCC vs. EAC were 0.48 vs. 0.73 respectively). These results suggested that nCRTS-IFI seemed to be more effective than nCRTS-ENI for patients with ESCC. Future head to head comparison trials are needed to confirm this finding and explore the mechanism.
RT dose and technique used in individual trials were various, which might also affect the outcomes. In our NMA, although nCRTS-ENI and nCRTS-IFI with dose of ≥40Gy seemed to be superior to those with dose of <40Gy based on treatment ranking, there were no significant difference in OS between the two dose group. Moreover, common dose in subgroup of ≥40Gy was only 40-41.4Gy. With developments in RT technique, whether a rather higher dose might be more reasonable needs further investigation.
In subgroup analysis of RT technique, we found that nCRTS-ENI adopting 3D-RT had a significant OS benefit compared to nCRTS-ENI adopting 2D-RT. Compared with 2D-RT, 3D-RT delivered a high dose to the tumor target volume while potentially minimizing the dose to the organ at risk. The results suggested that 3D-RT was more important for EC patients receiving nCRTS-ENI.
Treatment-related toxicities between ENI and IFI have been compared for EC patients receiving radical CRT in several retrospective studies. Results of two small metaanalysis [51,52] showed that the incidences of esophageal and lung toxicities were significantly higher in ENI group. However, most of trials comparing nCRTS with S alone did not reported CRT-related toxicities in detail, and therefore, indirect comparison of CRT-related toxicities between nCRTS-ENI and nCRTS-IFI could not be performed. In our NMA, nCRTS seemed to had a higher POM than S alone, but no significant difference was observed between nCRTS-ENI and nCRTS-IFI.
There are several limitations in our meta-analysis. Firstly, in common with other meta-analyses, data were collected and analyzed in aggregate on the basis of results reported from trials, instead of individual patient data. Secondly, different operative techniques and CT regimens were adopted in individual trials, which might lead to heterogeneity. Thirdly, most of the studies included patients with mixed stage and tumor location and could not be extracted separately, subgroup analyses according to stage and tumor location could not be performed. Finally, majority of trials comparing nCRTS with surgery alone did not reported RT related toxicities. Thus, the comparison of RT related toxicities between nCRTS-ENI and nCRTS-IFI could not be performed.

Conclusions
Either adopting ENI or IFI, nCRTS is likely to be the optimal treatment for resectable EC, and nCRTS-IFI and nCRTS-ENI seem to be more effective for patients with ESCC and EAC, respectively. 3D-RT seems to be more important for patients receiving nCRTS-ENI. nCRTS with RT dose of ≥40Gy seems to be superior to that with radiation dose of <40Gy, while the optimal dose remains unclear. Future head to head comparison trials are needed to confirm these findings.