Disappearing liver metastases: A systematic review of the current evidence.

Advances in systemic chemotherapy have resulted in a significant increase in the reported response rates of colorectal liver metastases (CRLM) over time. Although radiologic response is usually prognostic of favorable outcomes, complete shrinkage of CRLM after chemotherapy, namely "disappearing liver metastases" (DLMs) poses significant therapeutic dilemmas. A systematic review of the literature was conducted to evaluate the existing evidence on the imaging and management of patients with DLMs using the PubMed (Medline), Embase and Cochrane library through December 21st, 2018. The following algorithm was used: "(disappearing OR vanishing OR missing OR (residual tiny)) AND ((liver OR hepatic) AND (metastasis OR metastases OR metastatic OR secondary))." From the 225 records retrieved, 15 studies were finally deemed eligible. A total of 479 patients with DLMs with a median age of 59.5 years (range, 30-83) were identified. Median number of DLM per patient ranged from 1 to 8.8. Median size of LMs prior to chemotherapy was 1.07 cm (range 0.3-3.5). The systemic treatment used to achieve DLMs included systemic chemotherapy alone (only 2 studies) or in combination with targeted agents (11 studies). The median number of chemotherapy cycles in the included studies was 7.8 (range 6-12). Identified factors predisposing to the development of DLM were small size (<2 cm), increased number of treatment cycles, oxaliplatin-based therapy, increased number of CRLM (≥3) and synchronous CRLM. Baseline and preoperative MRI with iv contrast showed the highest sensitivity for DLM detection. Fiducial placement facilitated pre- and intra-operative identification of DLM. Although resection of DLM decreased the local recurrence risk, there was no clearly demonstrated survival benefit after resecting all sites of disappearing lesions. Future randomized clinical trials are highly encouraged to provide strict, evidence-based recommendations for the treatment of patients with DLM.