Background The need for radical surgery followed by adjuvant chemoradiation may be reduced by abandoning radical surgery in patients in whom lymph node involvement is detected intra-operatively.

Objectives To analyze, in a retrospective cohort study, the efficacy of the algorithm using intra-operative pathological assessment of sentinel lymph nodes.

Methods A retrospective single-institution study was carried out, which analyzed data from all consecutive patients with cervical cancer who were referred for primary surgical treatment between May 2005 and December 2015. Inclusion criteria were as follows: (1) TNM stage T1a1 with lymphovascular space invasion, T1a2, T1b, T2a, and selected T2b with incipient parametrial invasion; (2) adenocarcinoma, squamous cell carcinoma, or adenosquamous carcinoma; (3) no evidence of enlarged suspicious nodes or distant metastases on pre-operative imaging; (4) primary surgery with curative intent; (5) successful detection of sentinel lymph node, at least, unilaterally. All patients had at least one sentinel lymph node detected and submitted for frozen section evaluation. When sentinel lymph node involvement was detected intra-operatively, the cervical procedure was abandoned and the patient was referred for definitive chemoradiation. Radical surgery was completed in patients with intra-operative negative sentinel lymph nodes. The reliability of intra-operative sentinel lymph node assessment was evaluated by calculating the sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio, and negative likelihood ratio.

Results The study included a total of 309 patients. Sentinel lymph nodes were detected bilaterally in 86% of the patients. Lymph node positivity was detected intra-operatively in 18 (6%) patients in whom the cervical procedure was abandoned. Adjuvant radiotherapy after completed radical surgery was given to 29 (9%) patients, including 20 patients with macrometastases (8) or micrometastases (12) reported from the final histology, eight patients with positive parametria (all ≤3 mm), and one patient with a positive vaginal resection margin. The sensitivity, specificity, positive predictive value, and negative predictive value for the intra-operative detection of lymph node positivity (macrometastases or micrometastases) was 47% (95% CI 31% to 64%), 100%, 100%, and 93% (95% CI 90% to 96%), respectively. A total of 18 (6%) patients were spared combined treatment owing to the intra-operative sentinel lymph node triage; 29 patients (9%) received combined treatment with both radical surgery and adjuvant radiotherapy

Conclusions Of 47 patients with high-risk prognostic risk factors (lymph node, parametria, or surgical margin involvement), combined treatment was successfully avoided in 18 (38%). Despite an effort to triage the patients intra-operatively, 9% received a combination of cervical procedure and adjuvant chemoradiation, mostly owing to the low sensitivity of the frozen section in the detection of micrometastases and macrometastases.