No abstract

e15743 Background: NG is a standard 1st line treatment for APC. Although recommended in 2nd line after FOLFIRINOX, there is little evidence of its efficacy, tolerability and of markers of efficacy. Methods: We performed a multicenter retrospective cohort study, including patients (pts) with APC from 5 centers in Portugal treated with 2nd line NG after 1st line FOLFIRINOX from 01/2013-12/2016. We collected demographic, clinicopathological characteristics and treatment data. We used descriptive statistics, Kaplan-Meier methods and Cox proportional hazards analysis. Results: 30 pts were included; median age was 64 years (range 45–78); the majority had stage IV (90%) disease, an ECOG Performance Status of 0 (76.7%) and had received a median of 8.5 cycles of FOLFIRINOX (range 1–18). A median of 6 cycles of NG were administered (range 1–13). Median progression free survival (PFS) and overall survival (OS) were 6.4 months (CI 95% 3.0-8.5) and 11.4 months (CI 95% 8.4–16.5), respectively, and did not differ by age < 65 or ≥65 (p = 0.87; p = 0.57 respectively). The most frequent toxicity was fatigue (66.6%, any grade). Grade 3-4 events occurred in 40% of pts – thrombocytopenia in 16.7%, neutropenia in 10.0%; anemia, sensorial neuropathy, fatigue and diarrhea each occurred in 3.3% of patients. No febrile neutropenia events or toxic deaths occurred. Median CA 19.9 at the beginning of NG was 1254U/mL (IQR: 207–6775); the median decrease of CA19.9 at 3 months was 45U/mL (IQR:-1373– +174). CA 19.9 variation at 3 months did not correlate with PFS (p = 0.53) or OS (p = 0.09) in multivariate analysis (adjusted for age and stage at diagnosis). Neutrophil to Lymphocyte ratio (NLR) was high ( > 3.0) in 37.5% of patients before 1st line treatment and in 27.6% at the beginning of NG. In multivariate analysis NLR before 1st or 2nd chemotherapy lines were not associated with PFS (p = 0.39; p = 0.14 respectively) or OS (p = 0.44; p = 0.12, respectively). Conclusions: In this cohort of pts with APC, NG was an effective and well tolerated 2nd line regimen after FOLFIRINOX failure, even in pts ≥65 years. Neither CA19.9 variation at 3 months nor NLR were markers of NG clinical benefit.

Cervical cancer (CC) is one of the acquired immunodeficiency syndrome (AIDS) defining diseases and the human immunodeficiency virus (HIV) infection is thought to relate with increased acute toxicity of chemoradiotherapy (CRT).We investigated the effect of HIV status in the incidence of neutropenia associated with cisplatin-based CRT for CC and its impact in treatment completion.This is a single-center retrospective cohort study. Data collection was performed for all the consecutive stage Ib-IV CC women treated with cisplatin-based CRT from 2012 to 2016, and with known HIV status.Sixty-one patients were included, 6 were HIV+. HIV+ patients had a higher risk of neutropenia at any cycle during cisplatin CRT [adjusted odds ratio (OR) 7.3, 95% confidence interval (95% CI) 1.02–52.3; P = .05]. Despite the absolute differences, mean neutrophil count was nonsignificantly lower in HIV+ women, both at baseline [4455/μL (interquartile range, IQR: 1830–6689) vs 6340 (IQR: 1720–18,970) for HIV−, P = .98] and at the end of treatment [1752/μL (IQR: 1100–2930) vs 3147/μL (IQR: 920–18,390) in HIV−; P = .06]. Moreover, when considering the effect of time, CRT seems to induce a consistent drop of neutrophils in both groups (P = .229). No febrile neutropenia events occurred.In HIV+ women, there were more CT cycle delays (P = .013), patients were more prone to use granulocyte colony-stimulating factor (G-CSF; HIV+ 40.0% vs HIV− 4.0%; P = .04) and less likely to complete at least 5 cycles of cisplatin (P = .02). All patients received adequate dose of pelvic RT, regardless of HIV status.HIV+ patients have a significantly increased risk of neutropenia during CRT treatment for CC and are less likely to complete chemotherapy with cisplatin.

We propose a cross-lingual framework for fine-grained opinion mining using bitext projection. The only requirements are a running system in a source language and word-aligned parallel data. Our method projects opinion frames from the source to the target language, and then trains a system on the target language using the automatic annotations. Key to our approach is a novel dependency-based model for opinion mining, which we show, as a byproduct, to be on par with the current state of the art for English, while avoiding the need for integer programming or reranking. In cross-lingual mode (English to Portuguese), our approach compares favorably to a supervised system (with scarce labeled data), and to a delexicalized model trained using universal tags and bilingual word embeddings.