PMC

Literature search initially revealed 883 studies of which 9 were included in the final analysis (10 including own data).

(A) Overall survival (n = 259; p = 0.051) and (B) disease-free survival (n = 258; missing = 1; p = 0.078) in own cohort of patients undergoing curative-intent liver resection for colorectal liver metastases.

(A) Overall survival (n = 258; missing = 1; p = 0.005) and (B) disease-free survival (n = 257; missing = 2; p = 0.027) in own cohort of patients undergoing curative-intent liver resection for colorectal liver metastases. Each factor (right-sided primary-tumour, positive lymph-nodes and extrahepatic disease) was accounted for one point in this score). (C) Overall survival (n = 517; missing = 10; p = 0.001) and (D) disease-free survival (n = 517; missing = 10; p = 0.011) in the international validation cohort.

Meta-analysis of published studies including own data (Heterogeneity: I² = 72.4% / p<0.001). (A) Forrest plot: right-sided primary tumour location is associated with worse disease-free survival after resection for liver metastases, however not significant in the random effects model (p = 0.077). (B) Funnel plot analysis does indicate only a minimal publication bias with two large study outliers outside the 95%CI (at the tip of the pyramid) but no small trial bias (two smaller studies on the left side of the overall effects line have larger counterparts on the right side). CI = confidence interval; HR = hazard ratio; KRAS^mut/wt = KRAS mutated or wild-type subgroup; with / w/o CTX = with or without preoperative chemotherapy.

Meta-analysis of published studies including own data (Heterogeneity: I² = 50.8% / p = 0.0216). (A) Forrest plot: right-sided primary tumour location is significantly associated with worse overall survival after resection for liver metastases (random effects model: p<0.001) (B) Funnel plot analysis does not indicate a relevant publication bias but minor small trial bias: While the number of larger studies (tip of the pyramid) on both sides of the total effects line are evenly distributed, two smaller studies with an overestimating positive effect have only one negative study as counterparts (base of the pyramid). CI = confidence interval; HR = hazard ratio; KRAS^mut/wt = KRAS mutated or wild-type subgroup; with or w/o CTX = with or without preoperative chemotherapy.