When is a Ghost Really Gone? A Systematic Review and Meta-analysis of the Accuracy of Imaging Modalities to Predict Complete Pathological Response of Colorectal Cancer Liver Metastases After Chemotherapy

Abstract

Administration of chemotherapy to patients with colorectal liver metastases may result in disappearing liver metastases (DLM). This poses a therapeutic dilemma due to the uncertainty of true complete (pathological) response. We aimed to examine the diagnostic performance of imaging modalities in detecting true complete response in patients with DLM after chemotherapy. We performed a systematic search for articles assessing the diagnostic performance of imaging modalities in evaluating DLM following chemotherapy. True complete response was defined as 1-year recurrence-free survival in non-resected patients or complete pathological response on histologic examination in resected patients. We calculated the negative predictive value (NPV) for detecting true complete response of each imaging modality using a random effects model. Thirteen studies comprising 332 patients with at least one DLM were included. The number of DLMs after chemotherapy was 955 with computed tomography (CT), 104 with positron emission tomography (PET), 50 with intraoperative ultrasound (IOUS), 585 with magnetic resonance imaging (MRI), and 175 with contrast-enhanced IOUS (CEIOUS). Substantial variation in study design, patient characteristics, and imaging features was observed. Pooled NPV was 0.79 (95% confidence interval [CI] 0.53–0.96), 0.73 (95% CI 0.58–0.85), 0.54 (95% CI 0.37–0.7), 0.47 (95% CI 0.34–0.61), and 0.22 (95% CI 0.11–0.39) for CEIOUS, MRI, IOUS, CT, and PET, respectively. After chemotherapy, MRI or CEIOUS are the most accurate imaging modalities for assessment of DLM and should be used routinely in this context. Given the high NPV of these two modalities, surgical resection of visible CRLM is warranted if technically possible, even if DLM remain.