Utility of ctDNA in predicting relapse in solid tumors after curative therapy: a meta-analysis

BACKGROUND: Presence of circulating tumor DNA (ctDNA) is prognostic in solid tumors treated with curative intent. Studies have evaluated ctDNA at specific “landmark” or multiple “surveillance” time points. However, variable results have led to uncertainty about its clinical validity. METHODS: A PubMed search identified relevant studies evaluating ctDNA monitoring in solid tumors after curative intent therapy. Odds ratios for recurrence at both landmark and surveillance time points for each study were calculated and pooled in a meta-analysis using the Peto method. Pooled sensitivity and specificity weighted by individual study inverse variance were estimated and meta-regression using linear regression weighted by inverse variance was performed to explore associations between patient and tumor characteristics and the odds ratio for disease recurrence. RESULTS: Of 39 studies identified, 30 (1924 patients) and 24 studies (1516 patients) reported on landmark and surveillance time points, respectively. The pooled odds ratio for recurrence at landmark was 15.47 (95% confidence interval = 11.84 to 20.22) and at surveillance was 31.0 (95% confidence interval = 23.9 to 40.2). The pooled sensitivity for ctDNA at landmark and surveillance analyses was 58.3% and 82.2%, respectively. The corresponding specificities were 92% and 94.1%, respectively. Prognostic accuracy was lower with tumor agnostic panels and higher with longer time to landmark analysis, number of surveillance draws, and smoking history. Adjuvant chemotherapy negatively affected landmark specificity. CONCLUSIONS: Although prognostic accuracy of ctDNA is high, it has low sensitivity, borderline high specificity, and therefore modest discriminatory accuracy, especially for landmark analyses. Adequately designed clinical trials with appropriate testing strategies and assay parameters are required to demonstrate clinical utility.