ESTRO 37 Abstract book

S199

ESTRO 37

OC-0394 Pretreatment bone marrow SUVmax in locally advanced cervical cancer: a novel prognostic biomarker ? R.D. Seban 1 , C. Robert 2 , L. Dercle 3 , R. Yeh 3 , S. Reuze 4 , A. Schernberg 2 , J. Lumbroso 1 , M. Schlumberger 1 , C. Haie- Meder 2 , C. Chargari 2 , E. Deutsch 2 1 Institut Gustave Roussy, Nuclear Medicine and Endocrine Oncology, Villejuif, France 2 Institut Gustave Roussy, Radiotherapy, Villejuif, France 3 Columbia University Medical Center, Radiology, New- York, USA 4 Institut Gustave Roussy, Medical Physics, Villejuif, France Purpose or Objective We investigated the prognostic value of pretreatment Bone Marrow maximum standardized uptake value (BM SUVmax) at baseline on 18F-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) in patients with locally advanced cervical cancer. Material and Methods We recruited retrospectively patients with locally advanced cervical cancer who underwent staging 18F-FDG PET/CT and had baseline blood tests then treated by chemoradiation therapy (CRT), followed by image-guided adaptive brachytherapy (IGABT). Bone marrow and tumor maximum standardized uptake value (BM and tumor SUVmax) were calculated using a manual segmentation method. Neutrophilia was defined as a count over 7 x10 9 /L at the time of diagnosis. Prognostic value of BM SUVmax, tumor SUVmax and neutrophilia for predicting locoregional recurrence-free survival (LRFS) and overall survival (OS) was assessed using a Cox hazard regression analysis. Cut-off values for imaging biomarkers were obtained from receiver operating characteristic (ROC) curves and DeLong tests were used to test the statistical significance of the differences between the areas under ROC curves. Results One-hundred and seventeen patients with FIGO stage Ib1- IVb cervical cancer, treated between 2005 and 2014, were included. The median follow-up was 41.6 months (range 5.0-100.4). BM SUVmax, tumor SUVmax and neutrophilia were significant univariate prognostic factors for poorer locoregional failure-free survival (p<.001; p =.002 and p=.005 respectively) and poorer overall survival (p<.001; p =.003 and p<.001 respectively). Cut off values maximizing the sum of sensitivity and specificity but also accuracy for LRFS were 2.8 for BM SUVmax (AUC 0.76, accuracy 0.74) and 15.2 for tumor SUVmax (AUC 0.66, accuracy 0.69).Comparing the AUC with the DeLong method, BM SUVmax tended to predict locoregional recurrence better than tumor SUVmax did, although the difference was not statistically significant (p=0.112), despite a consequent gap between AUCs (0.096). In multivariate analysis, LRFS and OS were significantly poorer in patients with high BM SUVmax, defined as over 2.8, compared to those with lower BM SUVmax (respectively p < .001; HR 4.3; CI 95 % [1.8-9.9] and p = .001; HR 4.3; CI 95 % [1.9-9.8]). Neutrophilia remained an independent statistically significant prognostic factor for LRFS (HR 2.0, p=0.033) and for OS (HR 2.4, p=.009). Conclusion Pretreatment BM SUVmax on 18F-FDG PET is a significant prognostic factor for LRFS and OS in locally advanced cervical cancer treated with CRT plus IGABT. Strikingly, baseline BM SUVmax tended to be a better predictor of locoregional recurrence than tumor SUVmax. This PET biomarker could help identifying patients with higher risk of locoregional recurrence beyond tumor staging.

chemotherapy (Washington, 1499 patients; French cooperative study, 1875 patients) and the Vale meta- analysis of 13 randomised trials of chemo-RT + conventional BT (3128 patients). Results Figure 1 shows the number of patients with different types of failure. The total number of failures (single and synchronous) during the observation period was 325. 9% of patients had local failure, 6% regional failure and 13% pelvic failure. 24% of patients had distant failure; of these, 9% had PAN failure and 21% had systemic failure. 24% of the patients with PAN failure had PAN RT (i.e. in- field recurrence) while in the remaining 76%, recurrence was outside the irradiated field.

A comparison of patterns of failures in the Washington, French, Vale and RetroEMBRACE series is shown in Figure 2. Overall, 30% of the RetroEMBRACE cohort experienced treatment failure compared to 35% in the Vale chemo-RT series and 38% and 39% in the Washington and French RT alone series. 36% and 33% of the Washington and French cohorts had synchronous pelvic and distant failure compared to 22% and 23% of the Vale and RetroEMBRACE cohorts. In the Vale chemo-RT series, the predominant form of failure was in the pelvis alone. In the RetroEMBRACE IGBT series, distant-alone relapse was the predominant form of failure.

Conclusion Our analysis has shown an interesting shift in patterns of recurrence with evolving treatment. In the RT alone era, both loco-regional and distant control were problematic except for patients with early-stage disease with a low propensity for metastatic spread. The addition of concomitant chemotherapy improved survival for early- stage patients but advanced-stage patients continued to succumb to uncontrolled pelvic disease. The implementation of IGBT has resulted in distant failure being the predominant problem. Improved methods to identify patients with high individual risk of different types of failure are needed to facilitate personalisation of treatment strategies. OC-0393 Validating tumour extension by MRI with histoPAthology in cervical cancer; the MPAC- study Abstract withdrawn