SummaryBackgroundMultiple approaches based on cell-free DNA (cfDNA) have been applied to detect minimal residual disease (MRD) and to predict prognosis or recurrence. However, a comparison of the approaches used in different cohorts and studies is difficult. We aimed to compare multiple approaches for MRD analysis after neoadjuvant therapy (NAT) in patients with locally advanced rectal cancer (LARC).MethodsSixty patients with LARC from a multicentre, phase II/III randomized trial were included, with tissue and blood samples collected. For each cfDNA sample, we profiled MRD using 3 approaches: personalized assay targeting tumour-informed mutations, universal panel of genes frequently mutated in colorectal cancer (CRC), and low depth sequencing for copy number alterations (CNAs).FindingsPositive MRD based on post-NAT personalized assay was significantly associated with an increased risk of recurrence (HR = 27.38; log-rank P < 0.0001). MRD analysis based on universal panel (HR = 5.18; log-rank P = 0.00086) and CNAs analysis (HR = 9.24; log-rank P = 0.00017) showed a compromised performance in predicting recurrence. Both the personalized assay and universal panel showed complementary pattern to CNAs analysis in detecting cases with recurrence and the combination of the two types of biomarkers may lead to better performance.InterpretationThe combination of mutation profiling and CNA profiling can improve the detection of MRD, which may help optimize the treatment strategies for patients with LARC.