Identification and Validation of a Multigene Predictor of Recurrence in Primary Laryngeal Cancer Elena Fountzilas 1 *, Vassiliki Kotoula 2 , Nikolaos Angouridakis 3 , Ilias Karasmanis 3 , Ralph M. Wirtz 4¤ , Anastasia G. Eleftheraki 5 , Elke Veltrup 4¤ , Konstantinos Markou 3 , Angelos Nikolaou 3 , Dimitrios Pectasides 6 , George Fountzilas 1 1 Department of Medical Oncology, ‘‘Papageorgiou’’ Hospital, Aristotle University of Thessaloniki School of Medicine, Thessaloniki, Greece, 2 Department of Pathology, Aristotle University of Thessaloniki School of Medicine, Thessaloniki, Greece, 3 ENT Department, ‘‘AHEPA’’ Hospital, Aristotle University of Thessaloniki School of Medicine, Thessaloniki, Greece, 4 Siemens Healthcare Diagnostics, Cologne, Germany, 5 Section of Biostatistics, Hellenic Cooperative Oncology Group, Data Office, Athens, Greece, 6 Oncology Section, Second Department of Internal Medicine, ‘‘Hippokration’’ Hospital, University of Athens School of Medicine, Athens, Greece Abstract Purpose: Local recurrence is the major manifestation of treatment failure in patients with operable laryngeal carcinoma. Established clinicopathological factors cannot sufficiently predict patients that are likely to recur after treatment. Additional tools are therefore required to accurately identify patients at high risk for recurrence. This study attempts to identify and independently validate gene expression models, prognostic of disease-free survival (DFS) in operable laryngeal cancer. Materials and Methods: Using Affymetrix U133A Genechips, we profiled fresh-frozen tumor tissues from 66 patients with laryngeal cancer treated locally with surgery. We applied Cox regression proportional hazards modeling to identify multigene predictors of recurrence. Gene models were then validated in two independent cohorts of 54 and 187 patients (fresh-frozen and formalin-fixed tissue validation sets, respectively). Results: We focused on genes univariately associated with DFS (p,0.01) in the training set. Among several models comprising different numbers of genes, a 30-probe set model demonstrated optimal performance in both the training (log- rank, p,0.001) and 1 st validation (p = 0.010) sets. Specifically, in the 1 st validation set, median DFS as predicted by the 30- probe set model, was 34 and 80 months for high- and low-risk patients, respectively. Hazard ratio (HR) for recurrence in the high-risk group was 3.87 (95% CI 1.28–11.73, Wald’s p = 0.017). Testing the expression of selected genes from the above model in the 2 nd validation set, with qPCR, revealed significant associations of single markers, such as ACE2, FLOT1 and PRKD1, with patient DFS. High PRKD1 remained an unfavorable prognostic marker upon multivariate analysis (HR = 2.00, 95% CI 1.28–3.14, p = 0.002) along with positive nodal status. Conclusions: We have established and validated gene models that can successfully stratify patients with laryngeal cancer, based on their risk for recurrence. It seems worthy to prospectively validate PRKD1 expression as a laryngeal cancer prognostic marker, for routine clinical applications. Citation: Fountzilas E, Kotoula V, Angouridakis N, Karasmanis I, Wirtz RM, et al. (2013) Identification and Validation of a Multigene Predictor of Recurrence in Primary Laryngeal Cancer. PLoS ONE 8(8): e70429. doi:10.1371/journal.pone.0070429 Editor: John D. Minna, Univesity of Texas Southwestern Medical Center at Dallas, United States of America Received March 11, 2013; Accepted June 18, 2013; Published August 9, 2013 Copyright: ß 2013 Fountzilas et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: This work was supported by an internal Hellenic Cooperative Oncology Group (HeCOG) research grant (HE R_5G). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have the following interests: At the time of the study Ralph M. Wirtz and Elke Veltrup were employed by Siemens Healthcare Diagnostics. On behalf of the Hellenic Foundation for Cancer Research, Athens, Greece, the senior author (GF) has pending patent applications with Siemens Healthcare Diagnostics, Tarrytown, NY. There are no further patents, products in development or marketed products to declare. This does not alter the authors’ adherence to all the PLOS ONE policies on sharing data and materials, as detailed online in the guide for authors. * E-mail: elenafou@gmail.com ¤ Current address: STRATIFYER Molecular Pathology GmbH, Cologne, Germany Introduction Laryngeal cancer is the eleventh most common type of cancer in men worldwide. Every year 52,000 cases are newly diagnosed in Europe and 10,000 in the United States [1,2]. Despite the latest advances in diagnostic and therapeutic techniques, the majority of patients still recur after treatment [3]. Established clinicopatholog- ical factors cannot sufficiently predict patients that will recur. Additional factors are therefore required to accurately identify patients with poor prognosis. Expression profiling has been successfully used in the stratification of cancer patients with unfavorable prognosis [4,5,6,7]. Previous studies in head and neck cancer patients have linked gene expression profiles to nodal status [8,9,10], distant metastases [11,12] and disease-free survival [13,14,15,16]. While these studies provided great insight into the molecular complexity of head and neck cancer they did not identify a robust gene profile. The clinical use of these models has been limited by the large number of genes, the small-sized datasets and the lack of reproducibility and independent validation. Moreover, none of these studies focused exclusively on laryngeal cancer. PLOS ONE | www.plosone.org 1 August 2013 | Volume 8 | Issue 8 | e70429