External validation of the SUCCOR Risk score for predicting recurrence in early-stage cervical cancer patients treated with radical hysterectomy

OBJECTIVE: To externally validate the Surgery in Cervical Cancer, Observational, Retrospective (SUCCOR) Risk score for predicting recurrence in early-stage cervical cancer treated with radical hysterectomy in a Thai population. METHODS: A retrospective cohort of patients with the International Federation of Gynaecology and Obstetrics (FIGO) 2009 stage IB1 cervical cancer who underwent radical hysterectomy at our institution between 2003 and 2019 was analyzed. The SUCCOR Risk score was calculated using original model specifications. Progression-free survival (PFS) and overall survival were estimated by Kaplan-Meier analysis and compared across risk groups. Model performance was evaluated in terms of discrimination, calibration, and overall accuracy. RESULTS: Of 1,309 eligible cases, 954 met the inclusion criteria. After a median follow-up of 79 months, 83 relapses (8.7%) and 66 deaths (6.9%) were observed. Five-year PFS was 95.3% in the low-risk, 88.0% in the medium-risk, and 88.0% in the high-risk groups (log-rank p=0.002). Medium-risk patients had a significantly higher relapse risk than low-risk patients (HR=2.62; 95% CI=1.51-4.53). The SUCCOR Risk score demonstrated moderate discrimination for 5-year relapse (area under the receiver operating characteristic curve=0.65; Harrell's c-statistic=0.68). Calibration analyses showed good average agreement but an attenuated calibration slope of 0.72, reflecting overdispersion of risk predictions. Overall model performance was useful, with a 5-year Brier score of 0.068 and an integrated Brier score (IBS) of 0.080, outperforming the null model (IBS=0.084). CONCLUSION: The SUCCOR Risk score maintained moderate discriminative ability and acceptable calibration in this Thai cohort, with modest improvement over baseline prediction. These findings support its generalizability across diverse clinical settings, though recalibration may enhance accuracy for local application.