Comparison of different chemoradiotherapy regimens for the preoperative treatment in patients with locally advanced rectal cancer: a network meta-analysis.

BACKGROUND: The efficacy of different neoadjuvant chemoradiotherapy regimens on locally advanced rectal cancer (LARC) remains confusing. We evaluated them together via a network meta-analysis in terms of survival benefits to find the optimal treatment.
METHODS: We searched, EMBASE, Cochrane Central Register of Controlled Trials and the ClinicalTrials website according to the selection criteria for eligible publications before Oct 25, 2019. Pathological complete response rate (pCR), disease-free survival (DFS), and overall survival (OS) were analyzed based on Bayesian methods in the meta-analysis.
RESULTS: Twenty-five articles containing 7,142 participants and 12 preoperative regimens were analyzed. In terms of pCR, radiation therapy plus 5-fluorouracil (RT+5-Fu), RT plus capecitabine (RT+CAPE), RT plus 5-fluorouracil and oxaliplatin (RT+FOLFOX), RT plus capecitabine and oxaliplatin (RT+XELOX), and RT plus S-1 and irinotecan (RT+IS) were better than RT alone [odds ratio (OR) =2.66, 95% credible interval [CrI], 1.38-5.01; OR =3.11, 95% CrI: 1.33-6.98; OR =4.03, 95% CrI: 1.77-9.47; OR =4.22, 95% CrI: 1.60-10.87; OR =4.55, 95% CrI: 1.11-18.88, respectively] and RT+FOLFOX and RT+XELOX were superior to FOLFOX (OR =4.58, 95% CrI: 1.57-14.19; OR =4.81, 95% CrI: 1.20-18.73), too. Benefits could be seen on comparing RT+CAPE, RT+FOLFOX, and RT+XELOX with RT (OR =0.84, 95% CrI: 0.73-0.97; OR =0.88, 95% CrI: 0.80-0.97; OR =0.79, 95% CrI: 0.66-0.95, respectively) in DFS. RT+XELOX seemed to have better effects on OS compared than RT+5-Fu and RT+CAPE (OR =0.78, 95% CrI: 0.61-1.00; OR =0.86, 95% CrI: 0.74-1.00, respectively). Moreover, according to surface under the cumulative ranking curve analysis, RT+XELOX had the best outcomes in terms of pCR (79.18%) and OS (83.49%) and RT plus capecitabine, irinotecan, and cetuximab (RT+XELIRI+CET) ranked first with respect to DFS (87.86%).
CONCLUSIONS: RT+XELOX is likely to be the best treatment with a comprehensive curative effect and the standard treatment of 5-fluorouracil-based chemoradiotherapy has some advantages, as well. More relevant evidence is needed for clinicians' guidance. 2020 Translational Cancer Research. All rights reserved.

Introduction

Locally advanced rectal cancer (LARC) is a specific rectal cancer with T3–4, N0 or T any, N1–2 stage estimated by magnetic resonance imaging (MRI). Fluorouracil-based chemotherapy and radiotherapy are currently recommended as the standard preoperative neoadjuvant therapies for LARC (1). To achieve maximum clinical benefits, clinicians have made several attempts to combine different chemotherapy drugs. However, no conclusion has been reached to date. Due to the lack of sufficient information to comprehensively and directly compare different schemes, we performed a Bayesian network meta-analysis to directly and indirectly compare different interventions in LARC by determining the pathological complete response rate (pCR), disease-free survival (DFS), and overall survival (OS). Moreover, we revealed the results of ranking different regimens for LARC to guide clinical practice. We present the following article in accordance with the PRISMA guideline checklist (available at http://dx.doi.org/10.21037/tcr-20-683).

Methods

This meta-analysis adheres to the requirements of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. An IRB approval and informed consent of the patients were not required for this network meta-analysis.

Search strategy

Databases including, EMBASE, Cochrane Central Register of Controlled Trials (CENTRAL2019), and the ClinicalTrials website (https://clinicaltrials.gov) were explored to identify eligible articles before Oct 25, 2019 with no language restrictions. The keywords included “(colorectal or rectum or rectal) and (cancer or carcinoma or tumor) and (neoadjuvant or preoperative) treatment”, “clinical trials” and the list of the relevant therapeutic drugs and measures. Two authors (ZYY and LYX) independently picked out relevant articles by screening the titles and abstracts according to the selection criteria. Only trials with full-text articles were enrolled. Different articles belonging to the same trials were reserved and the newest outcomes were extracted. Discrepancies were resolved by discussion.

Selection criteria

We analyzed the articles meeting the following inclusion criteria:

Randomized controlled trials (RCTs) or qualified cohort studies;

Comparison of different preoperative treatments for LARC;

Twelve preoperative regimens were enrolled, namely, radiation therapy plus 5-fluorouracil (RT+5-Fu); RT plus capecitabine (RT+CAPE); RT plus 5-fluorouracil and oxaliplatin (RT+FOLFOX); RT plus capecitabine and oxaliplatin (RT+XELOX); RT plus capecitabine and irinotecan (RT+XELIRI); RT plus S-1 and irinotecan (RT+IS); RT plus oral tegafur (RT+UFT); RT plus capecitabine, irinotecan, and cetuximab (RT+XELIRI+CET); 5-fluorouracil and oxaliplatin (FOLFOX); RT plus capecitabine and bevacizumab (RT+CAPE+BEV); RT; RT plus 5-fluorouracil and irinotecan (RT+FOLFIRI). At least two different regimens were compared and at least one agent should be chemoradiotherapy;

Each arm of different trials must apply only one kind of treatment;

Clinical outcomes included pCR and hazard ratio (HR) with 95% confidence interval (CI) of DFS and OS.

Studies would be excluded if:

The trial only had one treatment regimen or a single arm received more than one kind of regimen;

The publication type was a case-control study or matching research;

No sufficient data.

Data extraction and quality assessment

The following contents were extracted: first author’s name, publication year, treatment regimens, events and sample size of pCR, HR with 95% CI of DFS and OS, and median follow-up time if provided.

The quality of the eligible articles was assessed separately based on the different types of trials. The RCTs bias was evaluated according to the Cochrane Handbook for Systematic Reviews of Interventions (2). Only one cohort study was assessed using the Newcastle-Ottawa Scale (NOS) (3).

Statistical analysis

We performed a network meta-analysis based on Bayesian methods (4,5). Odds ratio (OR) and 95% credible interval (95% CrI) were calculated for pCR using GeMTC version-0.14.3. The consistency model was firstly chosen for analysis. There were totally 300,000 times simulated interactions (75,000 per chain). Annealing value was set at 100,000. The consistency was assessed by the node-splitting method. The method separated the evidence concerning certain comparison into direct and indirect evidence, and the inconsistency was reported as the Bayesian P value (6). If any evidence of inconsistency was observed, we would turn to use the inconsistency model.

HR and 95% CrI were calculated for DFS and OS using OpenBUGS321. The fixed-effects model was used for analysis. A total of 300,000 simulated iterations were updated (100,000 per chain). Annealing value was set at 100,000. The convergence of iterative simulations was estimated by Gelman-Rubin-Brook diagrams (7,8)). If the convergence was not satisfied, we would increase the operation time. If problems remained, we would reappraise the included data or switch to add the random-effects model.

The rank probabilities would be sorted in plots. And the global effectiveness of each treatment would be ranked by the surface under the cumulative ranking curve analysis (SUCRA) (9).

The network structure diagrams, funnel plots and Begg’s test were all completed with STATA14. If the publication bias existed, we would check the outcomes of all publications and find the source of bias. Trim and fill methods might be applied to solve the bias (10).

Results

Search results and study characteristics

By reviewing the titles and abstracts and assessing full articles based on the selective criteria, 25 articles containing a total of 20 trials and a total of 7,142 participants were included in the network meta-analysis. The characteristics of the 20 studies included are listed in (11-35). Among them, 19 studies were RCT and one was a cohort study. The main reasons for including only one cohort study were multiple-arm study and strong relevance to this topic. There were no significant risks of bias found in the articles included. The network structure diagrams are detailed in , which is the reflection of the relationships of different preoperative treatment regimens. Meanwhile, the number of comparisons was expressed proportionally by the thicknesses of the lines, and the number of treatments was reflected proportionally in the diameters of the circles. provides the ORs and the corresponding 95% CrIs of pCR as well as the HRs and the corresponding 95% CrIs of DFS and OS. The rank probability of each regimen is displayed in . Value and plots of SUCRA are summarized in . We would introduce these results from three aspects: pCR, DFS and OS, respectively.

Table 1

Characteristics of the studies included

Author’s name/year	Patients	Study arms	pCR	OS	DFS	Median follow-up time (months)
			Events/sample size	HR (95% CI)	HR (95% CI)
Fokas 2017 (11-13)	613	RT+FOLFOX	113/591	0.95 (0.71, 1.27)	0.85 (0.76, 1.08)	50
	623	RT+5-Fu	83/606
Azria 2017 (14-16)	299	RT+XELOX	55/283	0.71 (0.50, 1.01)	0.86 (0.66, 1.15)	60.2
	299	RT+CAPE	40/282
Jung 2015 (17)	71	RT+5-Fu	11/66	1.15 (0.13, 9.95)	0.78 (0.35, 1.72)	43.8
	70	RT+IS	17/67
Wong 2014 (18,19)	52	RT+XELIRI	5/52	0.83 (0.19, 3.64)	0.89 (0.40, 1.96)	45.2
	52	RT+XELOX	11/52			47.6
Hofheinz 2012 (20)	80	RT+5-fu	4/74	1.28 (0.69, 2.37)	1.4 (1.02, 2.02)	64.8
	81	RT+CAPE	10/73
Martijnse 2011 (21)	106	RT	5/106	2.73 (1.73, 4.34)	1.6 (1.00, 2.56)	62.4
	137	RT+5-Fu	18/137	1.85 (1.18, 2.89)	1.62 (1.10, 2.49)
	106	RT+CAPE	8/106	1.85 (1.01, 3.41)	1.55 (0.89, 2.68)
	155	RT+XELOX	22/155
Braendengen 2008 (22)	98	RT+5-Fu	16/98	0.88 (0.43, 1.80)	0.57 (0.34, 0.94)	61
	109	RT	8/109
de la Torre 2008 (23)	78	RT+UFT	10/76	1.39 (0.66, 2.93)	1.14 (0.64, 2.02)	22
	77	RT+5-Fu	10/76
Gérard 2006 (24)	375	RT+5-Fu	41/375	0.96 (0.73, 1.27)	0.96 (0.77, 1.20)	81
	367	RT	13/367
Boulis-Wassif 1984 (25)	126	RT+5-Fu	6/126	1.47 (0.95, 2.27)	0.97 (0.54, 1.73)	55.2
	121	RT	3/121			75.6
Bosset 2006 (26)	506	RT+5-Fu	NR	1.02 (0.83, 1.26)	0.84 (0.78, 1.13)	64.8
	505	RT	NR
Kim 2013 (27)	38	RT+XELIRI+CET	8/38		0.58 (0.16, 2.05)	NR
	44	RT+XELIRI	11/44
Singh 2017 (28)	14	RT+5-Fu	4/12			NR
	16	RT+CAPE	1/15
Wiśniowska 2016 (29)	136	RT+FOLFOX	15/136			NR
	136	RT+5-Fu	8/136
Deng 2016 (30)	155	RT+5-Fu	20/143			NR
	157	RT+FOLFOX	41/149
	163	FOLFOX	10/152
Salazar 2015 (31)	44	RT+CAPE+BEV	7/44			NR
	46	RT+CAPE	5/46
Saha 2015 (32)	21	RT+XELOX	5/21			NR
	21	RT+5-Fu	3/21
Aschele 2011 (33)	379	RT+5-Fu	62/379			NR
	368	RT+FOLFOX	59/368
Kim 2006 (34)	105	RT+5-Fu	12/105			NR
	90	RT+CAPE	20/90
Mohiuddin 2006 (35)	50	RT+5-Fu	13/50			NR
	53	RT+FOLFIRI	14/53

RT+5-Fu, radiation therapy plus 5-fluorouracil; RT+CAPE, radiation therapy plus capecitabine; RT+FOLFOX, radiation therapy plus 5-fluorouracil and oxaliplatin; RT+XELOX, radiation therapy plus capecitabine and oxaliplatin; RT+XELIRI, radiation therapy plus capecitabine and irinotecan; RT+IS, radiation therapy plus S-1 and irinotecan; RT+UFT, radiation therapy plus oral tegafur; RT+XELIRI+CET, radiation therapy plus capecitabine, irinotecan and cetuximab; FOLFOX, 5-fluorouracil and oxaliplatin; RT+CAPE+BEV, radiation therapy plus capecitabine and bevacizumab; RT, radiation therapy; RT+FOLFIRI, radiation therapy plus 5-fluorouracil and irinotecan; NR, not reported; HR, hazard ratio; CI, confidence interval.

Figure 1

Network of comparisons of pCR, (A), DFS (B) and OS (C) for meta-analysis. The size of the circle indicates the number of participants in each treatment. The width of lines indicates the number of direct comparisons between two treatments. pCR, pathological complete response rate; DFS, disease-free survival; OS, overall survival.

Figure 2

ORs for pCR (A). HRs for DFS (B) and OS (C). Results with statistical significance are bold. pCR, pathological complete response rate; DFS, disease-free survival; OS, overall survival.

Figure 3

Rankograms for pCR (A), DFS (B) and OS (C). The figure shows the probability of each treatment from the best to the worst. The height of the columns for “rank 1, rank 2, etc.” refers to the probability of each rank. If the height of the column belonging to intervention X painted in the color of “rank 1” was 0.4, it means the probability of interval X to rank first was 40%. Treatment legend: A: RT+5-Fu, B: RT+CAPE, C: RT+FOLFOX, D: RT+XELOX, E: RT+XELIRI, F: RT+IS, G: RT+UFT, H: RT+XELIRI+CET, I: FOLFOX, J: RT+CAPE+BEV, K: RT, L: RT+FOLFIRI. pCR, pathological complete response rate; DFS, disease-free survival; OS, overall survival.

Figure 4

Value of SUCRA (A) and plots of SUCRA for pCR (B), DFS (C) and OS (D). SUCRA, surface under the cumulative ranking curve analysis; pCR, pathological complete response rate; DFS, disease-free survival; OS, overall survival.

pCR

All studies except the NCT0002523 trial conducted by Bosset et al. (26) reported pCR of all 12 different treatment regimens and a total of 6,000 participants were enrolled in the analysis of pCR. The pCR ranged from 2.5% to 27.5%. The details of the comparison between all 12 treatments are displayed in . The ORs and the corresponding 95% CrIs of pCR are shown in . The patients receiving RT+5-FU, RT+CAPE, RT+FOLFOX, RT+XELOX, and RT+IS were significantly superior than RT in pCR (OR =2.66, 95% CrI: 1.38–5.01; OR =3.11, 95% CrI: 1.33–6.98; OR =4.03, 95% CrI: 1.77–9.47; OR =4.22, 95% CrI: 1.60–10.87; OR =4.55, 95% CrI: 1.11–18.88; respectively). And the pCR of RT+FOLFOX and RT+XELOX statistically preceded FOLFOX (OR =4.58, 95% CrI: 1.57–14.19; OR =4.81, 95% CrI: 1.20–18.73). The summarized possibility value of the rankings for these treatments in is a direct plot of rank probability. is the plot of SUCRA for each intervention and its detailed values. According to the results above, the regimens from best to worst were RT+XELOX, RT+IS, RT+FOLFOX, RT+CAPE, RT+FOLFIRI, RT+UFT, RT+5-FU, RT+CAPE+BEV, RT+XELIRI, RT+XELIRI+CET, RT, and FOLFOX.

DFS

Twelve studies including nine preoperative treatment regimens (RT+5-FU, RT+CAPE, RT+FOLFOX, RT+XELOX, RT+XELIRI, RT+IS, RT+UFT, RT+XELIRI+CET, RT) were allocated for the analysis of DFS (5,052 participants). Their relationship to different treatments is detailed in and the pooled HRs and corresponding 95% CrIs of DFS is shown in . RT+CAPE, RT+FOLFOX, and RT+XELOX had a better DFS compared with RT (HR =0.84, 95% CrI: 0.73–0.97; HR =0.88, 95% CrI: 0.80–0.97; HR =0.79, 95% CrI: 0.66–0.95, respectively). According to the rank probabilities and SUCRA, RT+XELIRI+CET, RT+XELOX, RT+XELIRI, RT+CAPE, RT+FOLFOX, RT+5-FU, RT+UFT, RT+IS, and RT indicated a decreasing tendency of DFS ().

OS

The results of OS were calculated from 11 studies with a total of 4,970 participants including eight kinds of preoperative treatment regimens (RT+5-FU, RT+CAPE, RT+FOLFOX, RT+XELOX, RT+XELIRI, RT+IS, RT+UFT, RT). The relationship of different regimens is detailed in and the efficacy of different therapeutic paradigms for HRs and corresponding 95% CrIs is detailed in . As indicated in the results, no strong survival benefit could be seen in the direct comparisons. The RT+XELOX seemed to have better effects in OS compared with RT+5-Fu and RT+CAPE (OR =0.78, 95% CrI: 0.61–1.00; OR =0.86, 95% CrI: 0.74–1.00, respectively). Based on the results of rank probabilities and SUCRA, the treatment regimens from best to worst were RT+XELOX, RT+XELIRI, RT+CAPE, RT+IS, RT+FOLFOX, RT, RT+5-FU, and RT+UFT ().

Publication bias

The funnel plots are displayed in . The funnel plots were symmetric. The publication bias was not statistically significant as evaluated using Begg’s test (pCR, z =0.71, P=0.481; DFS, z =0.33, P=0.743; OS, z =0.98, P=0.360).

Figure 5

Funnel plots for pCR (A), DFS (B) and OS (C). pCR, pathological complete response rate; DFS, disease-free survival; OS, overall survival.

Discussion

This study is a network meta-analysis that compared the efficacy of single preoperative treatments for LARC. Twenty studies with a total of 7,142 participants were included in the final analysis and 12 regimens were involved in the comparison as described previously. We compared the efficacy of different regimens from three aspects (pCR, DFS, OS). Based on the rank probabilities and SUCRA, RT+XELOX had superiority on pCR and OS. This treatment regimen captured preponderance in DFS with a second place likewise.

At present, radiotherapy combined with chemotherapy is a consensus on the preoperative treatment of LARC. A Cochrane review including 6 RCTs found that chemotherapy added to preoperative radiation in patients with LARC reduced the risk of local recurrence, but had no effect on OS, 30-day mortality, sphincter preservation, and late toxicity (36). Our results revealed that the benefit of radiotherapy alone in terms of pCR and PFS was obviously inferior to preoperative concurrent chemoradiotherapy and provided an evidence to support standard treatment.

In the development of clinical practice, clinicians found that long-course and short-course radiation in neoadjuvant treatment had different advantages. Long-course radiation mainly had a higher pCR than short-course radiation. Further, they suspected that the reason for this was the duration between surgical interval and neoadjuvant therapy. Scholars also found that a surgical interval longer than 6 to 8 weeks from the end of neoadjuvant CRT and surgery significantly improved the pCR from 13.7% to 19.5% in a meta-analysis (37,38). In the researches included in our analysis, only few patients in Wiśniowska’s study underwent short-course radiation (29). Long-course radiation was used in most studies but the dose was a little lower in an individual early study (34.5 Gy) (25). In the aspect of surgical intervention, some articles did not mention it (11-16,20,21,27,29,30,32). Moreover, the rest of the studies reported the interval to be 3–10 weeks. These data were not specific in the subgroup analysis and the influence on the outcomes was assessed as unknown. With regard to preoperative chemotherapy regimens, the standard treatment is based on fluorouracil. 5-FU, capecitabine, UFT, and S-1 all belong to fluorouracil drugs. According to the results of our network analysis, RT+CAPE is superior to RT+5-FU and RT+UFT in terms of all three aspects. However, NSABP R-04 has shown that capecitabine was equivalent to 5-FU in perioperative chemoradiotherapy. There were no differences in locoregional events, DFS, OS, pCR, and surgical downstaging (39,40). Differently, in Hofheinz’s study, capecitabine had better results than 5-FU. However, the original non-inferiority design made it unable for capecitabine to replace 5-FU totally (20). Further superiority trials would not be carried out. These facts could also explain the reasons for RT+XELOX possibly being the best regimen. Commonly, RT+XELOX was slightly better than RT+CAPE, but there was no obvious statistical difference each time. After being processed by the pass-along effect of the chains in meta-analysis, the advantages could finally become more obvious.

Fluorouracil-based chemotherapy and radiotherapy could decrease the local recurrence rate, but it had no benefit to DFS and OS. Hence, the mixture of chemotherapy drugs is likely to have the potential to control the distant metastasis for enhanced survival. In our meta-analysis, RT+XELOX regimen had striking rankings in pCR and OS compared with RT+CAPE and RT+5-Fu. However, based on the results of ACCORD-12, STAR-01, NSABP R-04, CAO/ARO/AIO-04, and PTEACC-6 trials, RT+XELOX did not show significantly better outcomes for except a slight benefit. For example, only the CAO/ARO/AIO-04 trial found that the addition of oxaliplatin had benefits in 3-year DFS (HR =0.79, 95% CI, 0.64–0.98, P=0.03) (12). The pCR of the addition of oxaliplatin was slightly higher in the ACCORD-12 trial (19% vs. 14%, P=0.09) and CAO/ARO/AIO-04 trial (17% vs. 13%, P=0.04) (13,16). The advantages of RT+XELOX in our study may be due to the accumulation of sample size.

With the development of target therapy, attempts to combine epidermal growth factor receptor (EGFR) and vascular endothelial growth factor (VEGF) monoclonal antibodies with chemoradiotherapy were made for better efficacy. Our study showed that RT+XELIRI+CET ranked first in DFS but inferior in pCR. In some practice, addition of anti-EGFR in chemoradiotherapy confers a pCR of 19–25% (41,42). In addition, bevacizumab-containing chemoradiotherapy had a pCR of 18.4–19.2% (43,44). Nonetheless, there was no powerful evidence about adding monoclonal antibodies to chemoradiotherapy in RCTs.

Owing to the side effects of radiotherapy, chemotherapy alone may be an option. In the FOWARC trial, clinicians made an attempt to get rid of radiotherapy (30). Besides, immunotherapy in deficient-mismatch repair (d-MMR) and microsatellite instability-high (MSI-H) patients, proved to have desirable outcomes, and potentially may be used to treat LARC patients. In the NICHE study, ipilimumab and nivolumab as neoadjuvant therapies for nonmetastatic colon cancer achieved a major response rate of 100% and pCR of 57% (4/7) in seven d-MMR patients, indicating that neoadjuvant immunotherapy may be tried in LARC patients with d-MMR in the future (45).

There are several limitations of the network meta-analysis. Firstly, in the randomized, multicenter, non-inferiority, phase 3 trial conducted by Hofheinz et al., HR of DFS was calculated by the data of both adjuvant and neoadjuvant cohorts and specific data about the neoadjuvant cohort were not mentioned in the article (20). It may lead to some bias in the final results. Secondly, the toxicity of these treatments was not analyzed on account of the insufficient reports. Thirdly, the use of the fixed-effects model ignored the heterogeneity between studies, which may lead to overestimated results. Some outcomes without full-text manuscripts were not included.

Conclusions

This network meta-analysis aimed to assess the regimens of preoperative chemoradiotherapy in LARC, and a combination of oxaliplatin, capecitabine together with radiotherapy (RT+XELOX) is likely to be the best treatment option with a comprehensive curative effect currently. Despite the regimen being not recommended, it may have a potential in the future based on more evidence.