Article Text
Abstract
Introduction/Background*Up to 26% of early-stage cervical cancer patients relapse after primary surgical treatment. However, little is known about the factors affecting prognosis following disease recurrence. Hence, the aim of this study was to evaluate post-recurrence disease-specific survival (PR-DSS) and to identify respective prognostic factors.
Methodology Data from 528 early-stage cervical cancer patients who relapsed after primary surgical treatment performed between 2007 and 2016 were obtained from the SCCAN study (Surveillance in Cervical CANcer). Parameters related both to primary disease and recurrence diagnosis were combined to develop a multivariable Cox proportional hazards model predicting PR-DSS.
Result(s)*Five-year PR-DSS reached 39.1% (95% confidence interval: 22.7% – 44.5%) with median disease-free survival between primary surgery and recurrence diagnosis (DFI1) of 1.5 years and median survival after recurrence of 2.5 years. Six variables significant in multivariable analysis were included in the PR-DSS prognostic model; two related to the primary disease characteristics: maximal diameter of the tumour and lymphovascular space invasion; and four related to the recurrence diagnosis: DFI1, age, presence of symptoms, and recurrence localization (table 1). C-statistics of the final model after 10-fold internal validation equalled 0.701 (95% CI: 0.675 – 0.727). Five risk groups significantly differing in prognosis were identified, with 5-year DSS after recurrence of 85.6%, 62.0%, 46.7, 19.7%, and 0% in the highest risk group (figure 1).
Conclusion*We have developed the first robust model of disease-specific survival after recurrence stratifying relapsing cervical cancer patients according to their risk profile using six traditional prognostic markers. The strongest factor related to the length of post-recurrence survival was the largest size of the primary tumour, followed by the presence of symptoms at the time of diagnosis, which remained significant even after correction for lead-time bias.