E228: Perioperative and oncologic outcomes of hepatic arterial infusion pump chemotherapy for patients with unresectable colorectal liver metastases at an expanding HAI program

Hepatic arterial infusion (HAI) is an established treatment for patients with unresectable colorectal liver metastases (uCRLM). Until recently, HAI was only performed at a few centers with adequate expertise and infrastructure. We previously reported early outcomes suggesting that implementation of a new HAI program is effective. Here, we report perioperative and longer-term oncologic outcomes from an expanded series of patients with uCRLM treated with HAI.

Methods:

We analyzed outcomes from consecutive patients with uCRLM who underwent HAI pump (HAIP) placement at Duke University Hospital from 2018-2023.  Demographics, prior treatment, perioperative and oncologic outcomes were assessed. Overall survival (OS), hepatic- (hPFS) and extra-hepatic (ehPFS) progression free survival were calculated using Kaplan-Meier methods among patients with at least 12 months follow-up.

Results:

94 patients underwent HAIP placement for uCRLM. 58 (62%) were male, with a median age of 52(22,80). 78(83%) patients presented with synchronous liver metastases, 76(81%) with bilobar metastases, and median of 7 liver tumors. Most patients (92, 98%) received prior chemotherapy with median treatment cycle number of 12 (2,66). 43(45%) pumps were implanted robotically.

Post-operative HAI-specific complications occurred in 21 patients (22%), including 6(6%) patients with biliary sclerosis, and a 90-day mortality rate of 3%. Of 74 patients with at least 12 months of follow up, 70(95%) initiated HAI therapy and received a median 5 cycles of FUDR. 13(18%) achieved conversion to resection, and 2 underwent liver transplantation. 51(68%) experienced hepatic progression or death with a median hPFS of 16(12.2,19.8) months, while 53(72%) experienced extra-hepatic progression or death with a median ehPFS of 11.5(7.59, 19.8) months. Median OS was 35.9(30.6, N.E.) months. These outcomes were similar when stratified by extent of prior chemotherapy. Patients who converted to resection had improved OS when compared to those who did not (median OS NR vs. 30 months, p=0.02).

Conclusions:

We demonstrate that a new HAI center can recapitulate previously reported perioperative and survival outcomes of HAI treatment for uCRLM at established centers. Conversion to resection is achievable and yields improved survival. Granular analysis of our institutional data, which includes mutation status and primary tumor sidedness could inform selection and treatment strategies for new HAI programs.