Forty-nine primary and three follow-up publications were included in systematic review after screening (see Fig 1).

Fig 1: PRISMA chart of literature search.

Prescribed standard radiation doses ranged from 37.8-54Gy, with boost dosing of 5-22.2Gy, resulting in a total radiotherapy dose of 54.6-65Gy administered in 23-38 fractions (EQD2₁₀ 52.95-66.25Gy, median 56.42Gy). The proportion of patients completing a prescribed treatment course >54Gy was high (range 89-100%, median 100%) but was not reported in 11 publications.

Surgical interval varied (median: 8 weeks, range: 2-14 weeks) and was not specified in five publications. Surgical complication data of patients treated to >54Gy was available in 25 publications, with median morbidity of 15% (range: 0-100%), median mortality of 0% (range: 0-6%) and median rate of anastomotic leakage 0% (range: 0-6%).

Survival data and late oncological outcomes were available in 22 primary and three follow-up publications. Functional assessment or patient-reported outcome measures were reported in only five publications. Late toxicity was reported in eight studies, four of which used standardised grading. The highest reported rate of late grade ≥3 toxicity was 14%.

Pooled estimates of pCR, toxicity and R0 resection across 37 eligible publications eligible for meta-analysis (n = 1817 patients) were 24.1% (95% CI: 21.2-27.4%; see Fig 2), 11.2% (95% CI: 7.2-17.0%, see Fig 3) and 90.7% (95% CI: 87.9-93.8%, see Fig 4). Within inverse planning studies (17 publications, n = 959 patients), these rates were 25.7% (95% CI: 21.0-31.1%), 9.8% (95% CI: 4.6-19.7%) and 95.3% (95% CI: 91.6-97.4%).

Fig 2: Forest plot of pathological complete response (pCR) across all available...

Fig 3: Forest plot of acute grade ≥3 non-haematological toxicity across all...

Fig 4: Forest plot of R0 resection across all available studies (n=30) using a...

Regression analysis did not identify any significant predictor of pCR (p > 0.05) nor account for between-study heterogeneity (R² = 0%).  Similarly, acute toxicity was not significantly related to any of the evaluated covarites (p > 0.05), nor account for significant heterogeneity (R² = 8.02%). Rate of R0 resection positively correlated with the use of inverse-planning techniques (r = 1.0783, p = 0.0391) but no other cofactor (R² = 24.90%), which may represent confounding by multifactorial improvements in outcomes in more recent trials.