Search results and characteristics of included studies
The electronic search finally identified 15 articles met inclusion criteria, including eight RCTs 20-28 and seven non-RCTs 17, 29-35 (Figure. 1). A total of 4091 patients with locally advanced rectal cancer was included, of these patients, 1906 in TNT and 2185 in NCRT group. The characteristics of included studies were listed in Table. 1, and the chemotherapy regimens and doses of each study were shown in Table. 2. The quality assessment of RCTs was described in Figure. 2.
Grade 3/4 acute toxicities
Grade 3/4 acute toxicities were reported in 12 studies, including seven RCTs 20-26, 28 and five non-RCTs 29, 30, 32, 33, 35. Of these studies, diarrhea and hematologic toxicities were two most common adverse effects following TNT. A significant heterogeneity was observed, and random-effects model was used (I2=88%) for analysis. The results indicated that TNT did not increase toxicity profiles compared to NCRT (OR=0.95, 95% CI: 0.58~1.54, P=0.83) (Figure. S1). In subgroup analysis, the results regarding grade 3/4 acute toxicities were similar between RCTs and non-RCTs. (Figure. 3).
T-downstaging rate
Six studies reported tumor downstaging rate after TNT or NCRT, including three RCTs 24, 26, 28 and three non-RCTs 32, 33, 35. The fixed-effects model was applied for meta-analysis (P=0.26, I2=24%). The results showed that T-downstaging rate was significantly higher in TNT group (OR=2.16, 95%CI:1.63~2.87, P<0.00001) (Figure.S2). In subgroup analysis, though no significant difference was observed between TNT and NCRT in RCTs subgroup, the overall T-downstaging rate was obviously higher in TNT than that in NCRT in non-RCTs subgroup (OR=2.40, 95% CI:1.63~3.94, P=0.0005) (Figure.4).
pCR rate
All included studies but one reported pCR rate in TNT and NCRT group 17, 20-22, 25-35. A total of 3555 patients were included. The pooled rate of pCR was 24.2% in TNT group and 13.9% in NCRT group, respectively. The heterogeneity was low and fixed-effects model was then used for pooled analysis (P=0.13, I2=30%). The results showed that TNT significantly increased the rate of pCR (OR=1.90, 95% CI:1.60~2.27, P<0.00001) (Figure. S3).
In subgroup analysis, all included RCTs reported pCR rate, the results revealed that the pooled rate of pCR in TNT group was significantly higher than that in NCRT group (23.8% vs. 13.1%, OR=2.08, 95% CI:1.66~2.60, P<0.00001) in RCTs subgroup, and a similar result was also observed in non-RCTs subgroup (24.7% vs. 15.5%, OR=1.65, 95% CI:1.24~2.20, P=0.0005) (Figure. 5).
R0 resection rate
10 studies described R0 resection rate as primary outcome measure, including six RCTs 22, 25-28, 31 and four non-RCTs 29, 30, 34, 35. No significant heterogeneity was observed and fixed-effects model was then chosen for analysis (p=0.07, I2=45%). The results revealed that TNT obtained higher R0 resection rate than that in NCRT group (OR=1.44, 95% CI: 1.07~1.93, P=0.01) (Figure.S4). In subgroup analysis, the pooled R0 resection rate was significantly higher in TNT than that in NCRT in non-RCTs subgroup (96.7% vs. 89.8%, OR=2.08, 95% CI:1.66~2.60, P=0.002), while no significant difference was observed in RCTs subgroup (Figure. 6).
Post-operative complications
Nine studies described post-operative complications as primary endpoint, including five RCTs 21, 23-25, 27and four non-RCTs 29, 30, 33, 35. Of these studies, surgical site complications and anastomotic leak rates were similar in both groups. A fixed-effects model was chosen for pooled analysis. The results revealed that TNT was as safe as NCRT without increasing post-operative complications (OR=1.04, 95% CI: 0.84~1.29, P=0.89) (Figure. S5).
The local recurrence and distant metastasis
A total of four studies reported the local control of disease 22, 23, 28, 34, 35, No significant heterogeneity was found among studies and a fixed-effects model was applied for analysis (I2=0%), the results showed that no significant difference regarding local recurrence rate was observed (OR=0.89,95% CI: 0.47~1.69, P=0.73) (Figure. S6). A total of six studies reported distant metastasis of rectal cancer 20-22, 25, 28, 34. And random-effects model was applied for analysis (I2=69%), the pooled analysis revealed that TNT had no potential of improving distant metastasis rate (19.4% vs. 20.8%, OR=1.05, 95% CI: 0.64~1.73, P=0.84) (Figure. S7).
The OS between TNT and NCRT
A total of eight studies reported the long-term survival OS, including five RCTs 20-22, 25, 28 and three non-RCTs 32, 34, 35. The heterogeneity was low and fixed-effects model was used for analysis. The results revealed that OS in TNT was significantly superior to NCRT (HR=0.74, 95% CI: 0.62~0.89, P=0.001) (Figure.S8). In subgroup analysis, the pooled analysis revealed that the OS in TNT group was significantly better than that in NCRT group in both RCTs (HR=0.79, 95% CI: 0.65~0.96, P=0.02) and non-RCTs subgroup (HR=0.48, 95% CI: 0.29~0.79, P=0.004) (Figure. 7).
The DFS between TNT and NCRT
A total of seven studies reported DFS as major outcome, including four RCTs 20-22, 25and three non-RCTs 32, 34, 35. And random-effects model was chosen for analysis due to obvious heterogeneity (I2=67%). The results revealed that the pooled DFS in TNT was comparable to NCRT (HR=0.76, 95% CI: 0.56~1.02, P=0.06) (Figure.S9). In subgroup analysis, though a similar results was observed in non-RCTs subgroup, the DFS in TNT was significantly better than NCRT (HR=0.82, 95% CI: 0.70~0.96, P=0.01) (Figure. 8).
Evaluation of publication bias
According to funnel plots of included studies, no significant publication bias existed and the data were normally distributed without any bias. All included studies but one reported pCR as primary outcome and the funnel plot revealed that there existed no publication bias (Figure. 9).