TY - JOUR T1 - 960 The annual recurrence risk model for tailored surveillance strategy in cervical cancer patients JF - International Journal of Gynecologic Cancer JO - Int J Gynecol Cancer SP - A57 LP - A59 DO - 10.1136/ijgc-2021-ESGO.82 VL - 31 IS - Suppl 3 AU - D Cibula AU - J Jarkovsky AU - L Van Lonkhuijzen AU - H Falconer AU - A Fagotti AU - A Ayhan AU - S Kim AU - D Isla Ortiz AU - J Klat AU - A Obermair AU - F Landoni AU - J Rodriguez AU - R Manchanda AU - J Kostun AU - R Dos Reis AU - MM Meydanli AU - D Odetto AU - R Laky AU - M Borčinová AU - A Lopez Y1 - 2021/10/01 UR - http://ijgc.bmj.com/content/31/Suppl_3/A57.abstract N2 - Introduction/Background*Current guidelines for surveillance strategy in cervical cancer are rigid, recommending the same strategy for all survivors. The aim of this study was to develop a robust and comprehensive model allowing for individualised surveillance strategy based on risk profile of early-stage cervical cancer patients that were referred for surgical treatment with curative intent.Methodology The data of 4,343 cervical cancer patients with pathologically confirmed early-stage cervical cancer treated between 2007 and 2016 were obtained from SCANN consortium centres of excellence (Surveillance in Cervical CANcer). Only patients with complete key predictor variables and a minimum of one-year follow-up data availability were included. Based on the prognostic markers, a multivariable Cox proportional hazards model predicting disease-free survival (DFS) was developed and internally validated. A risk score, derived from regression coefficients of the model, stratified the cohort into significantly distinctive risk groups. On its basis, the annual recurrence risk model (ARRM) was calculated by conditional survival analysis.Result(s)*Five variables significant in multivariable analysis of recurrence risk were included in the prognostic model: maximal pathologic tumour diameter, tumour histotype, tumour grade, the number of positive pelvic lymph nodes, and lymphovascular space invasion (table 1). The model was ten-fold internally cross-validated with the average AUC of 0.732. Five risk groups significantly differing in prognosis were identified: with five-year DFS of 97.5%, 94.7%, 85.2%, and 63.3% in consecutive increasing risk groups, while two-year DFS in the highest risk group equalled 15.4%. Based on ARRM, the annual recurrence risk in the lowest risk group was below 1% in the first year of follow-up and declined below 1% at years three, four, and >5 in the three medium-risk groups (figure 1). The proportion of pelvic recurrences declined in groups with the growing risk. In the whole cohort, 26% of recurrences appeared at the first year of the follow-up, 48% by year two, and 78% by year five.View this table:Abstract 960 Table 1 Multivariate model for risk of recurrence predictionAbstract 960 Figure 1 ARRM: Landmark analysis of the annual probability of recurrence after surgery. N/A not analysed.Conclusion*ARRM represents a powerful tool for tailoring the surveillance strategy in early-stage cervical cancer patients based on the patient´s risk status and respective annual recurrence risk. It can easily be utilised in routine clinical settings internationally. ER -