Percutaneous Insertion of Hepatic Arterial Infusion Ports by Interventional Radiology for Colorectal Liver Metastases

Background: Hepatic arterial infusion (HAI) chemotherapy for colorectal liver metastases (CRLM) is delivered through a port–catheter system implanted percutaneously by interventional radiology or surgically. Previous studies have pooled device complications across access methods as event counts, but have not synthesised the technical performance and durability of fully percutaneous insertion at the patient level. Methods: Medline, EMBASE, Cochrane Library, Scopus and Web of Science were searched for studies published from 2000 onward reporting adults with CRLM undergoing HAI port insertion. Primary analyses were restricted to fully percutaneous implanted ports, using outcome-specific, patient-level numerators and denominators; a study using surgical arterial cutdown was added in sensitivity analyses. Risk of bias in the device outcomes was assessed at item level for patient selection, completeness of data, outcome ascertainment and selective reporting, with Cochrane RoB 2 used additionally for the single randomised trial. Random-effects proportion meta-analyses used the Freeman–Tukey double arcsine transformation and the DerSimonian–Laird estimator, with a binomial generalised linear mixed model (GLMM) as an alternative. Outcomes with inconsistent definitions or denominators were synthesised narratively. Results: Fifteen studies contributed 811 patients, 654 of whom underwent percutaneous insertion. One study was randomised, five prospective and nine retrospective; three reported a surgical comparator arm. Risk of bias in the device outcomes was low in one study, moderate in six and high in eight. Eleven studies (423 patients) of fully percutaneous implanted ports contributed to the primary analyses. Technical success was reported explicitly by eight and achieved in 276 of 278 patients (99.3%, exact 95% CI 97.4–99.9%); with only two failures, pooled models added little. Device-attributable definitive failure, meaning permanent discontinuation of HAI because of the device, occurred in 39 of 374 patients (crude 10.4%) over observation periods that varied widely; the random-effects estimate was 6.9% (95% CI 2.1–13.7%; I2 = 74.5%; GLMM 6.9%, 3.0–15.2%), and 11.6% (95% CI 6.3–18.0%) in the six studies at low or moderate risk of bias. Catheter dislodgement affected a pooled 5.8% (95% CI 1.6–11.6%), and 14.8% where the port was checked angiographically before every course or every four weeks; hepatic arterial occlusion affected 3.4% (95% CI 0.9–7.0%; four studies). Revision was required in 56% and 59% of percutaneous patients in the two series reporting it. Conclusions: Percutaneous insertion of HAI ports rarely failed technically in these series, but reintervention was common and permanent loss of HAI to a device problem within the follow-up reported was estimated at 6.9% in the primary analysis and 11.6% in studies at lower risk of bias, with more problems detected where the device was monitored more closely. Catheter-tip technique, follow-up duration and intensity of device surveillance are possible, unproven contributors to heterogeneity. These predominantly observational data support the feasibility of percutaneous access, not patient suitability for HAI chemotherapy.