Background In cervical cancer, the benefits of sentinel lymph node biopsy (SLNB) have long been confined by the lack of precise intraoperative pathological examination. Objective To determine the diagnostic performance and optimal protocol of frozen section examination (FSE) in SLNB for cervical cancer. Search Strategy PubMed, EMBASE, Web of Science, Cochrane Library, Wanfang Data and China National Knowledge Infrastructure were searched from inception to July 30, 2019. Selection Criteria Studies reporting the data of SLNB combined with FSE in cervical cancer were included. Data Collection and Analysis Two independent reviewers extracted the data. Bivariate mixed-effects regression model was applied for analyses. Sensitivity of FSE in detecting SLN metastasis was the primary diagnostic indicator for evaluation. Main Results The pooled sensitivity of FSE among 31 eligible studies (1887 patients) was 0.77 (95% CI 0.66–0.85) with high heterogeneity (I²=69.73%). Two representative FSE protocols were identified from 26 studies, described as equatorial (E-protocol, SLN was bisected) and latitudinal (L-protocol, SLN was cut at intervals). Meta-regression showed that FSE protocol was the only source of heterogeneity (p<0.001). The pooled sensitivity was 0.86 (95% CI 0.79–0.91, I²=0%) and 0.59 (0.46–0.72, I²=58.47%) for FSE using L- and E- protocol, respectively. The pooled sensitivity of FSE using L-protocol would reach 0.97 (0.89–0.99) if only marcometastases were considered. These findings were robust to sensitivity analyses. Conclusions With L-protocol, FSE can provide precise intraoperative pathology for SLNB, which enables immediate decision-making for individualized managements. Keywords Cervical cancer, sentinel lymph node, metastasis, frozen section, sensitivity, meta-analysis.