Quality control in multicentric clinical trials. An experience of the EORTC Gynecological Cancer Cooperative Group G. Favalli a , J.B. Vermorken b, *, K. Vantongelen c , J. Renard d , A.T. Van Oosterom c , S. Pecorelli a a Department of Gynecologic Oncology, University of Brescia, Spedali Civili Brescia, Italy b Department of Medical Oncology, University Hospital, Antwerp, Wilrijkstraat 10, 2650 Edegem, Belgium c Department of Oncology, University Hospital Gasthuisberg, Leuven, Belgium d EORTC Data Center, Brussels, Belgium Received 31 August 1999; received in revised form 26 January 2000; accepted 10 March 2000 Abstract Data Quality is a central requirement of scienti®c research and external monitoring is essential in multicentric clinical trials (MCT). A quality control (QC) study was conducted in the main Institutions participating in EORTC-GCCG Protocol number 55863 Ð randomised phase III trial of vindesine, cisplatin, bleomycin and mitomycin-C (BEMP) versus cisplatin (P) in dis- seminated squamous cell carcinoma of the uterine cervix Ð in order to assess the impact of variations in data quality on the con- clusions of the trial. The reliability of the dierent centres in following the protocol was investigated by a questionnaire covering drug prescription, local facilities and the procedure for preparation and administration of chemotherapy. The `treatment protocol adherence' was evaluated by recalculation of the ideal protocol dose and its comparison with the actual delivered dosage at each cycle of chemotherapy. `Data quality control' was assessed by comparison of data on case report forms (CRFs) with the corre- sponding items in the medical records. Eleven centres participating in the trial were visited by the same team of reviewers. Striking dierences were noted in the chemotherapy administration procedures and between the type and quality of hospital ®les. Overall, there was an acceptable level of data quality and protocol compliance. Data accuracy was 81.8% (range: 65.6±97%) of the 4424 items checked. Incorrect data were found in 7.0% (2.3±14.5%), data were missing on the form in 3.6% of cases (0±12%) and data was on the form but not in the ®le in 7.6% of cases (0.7±17.5%). Causes of inaccuracy were analysed. Both problems in data management but also in a lack of clarity of the protocol and/or CRFs were to blame. Training and supervision of data managers, precision in writing protocols, standardisation of some aspects of CRFs and the use of a checklist for chemotherapy data and treatment toxicities would have avoided many of these errors. The need for QC in all collaborative groups performing MCT is emphasised. A literature review on QC in MCT dealing with chemotherapy is included. # 2000 Published by Elsevier Science Ltd. All rights reserved. Keywords: Quality control; Multicentric trials; Chemotherapy 1. Introduction Scienti®c quality is an ethical obligation of clinical research [1]. Clinical studies are aimed at improving the quality of management of the patients and, ideally, the treatment results of a certain disease. The way in which clinical research protocols are handled in the collabor- ating centres of multi-institutional studies can vary greatly and major inadequacies in treatment execution may alter the outcome of the study. In the 1980s the European Organization for Research and Treatment of Cancer (EORTC) received several grants from the Eur- opean Community to perform quality control (QC) projects at dierent levels. Many variables were investi- gated in QC studies including diagnostic techniques, treatment modalities, tumour response assessment and, more recently, quality of data management [2±14]. Since 1988, the EORTC Study Group on Data Man- agement (SGDM) has focused its eorts on establishing 0959-8049/00/$ - see front matter # 2000 Published by Elsevier Science Ltd. All rights reserved. PII: S0959-8049(00)00090-3 European Journal of Cancer 36 (2000) 1125±1133 www.ejconline.com * Corresponding author. Tel.: +32-3-821-3375; fax: +32-3-825- 0564. E-mail address: jan.b.vermorken@uza.uia.ac.be (J.B. Vermorken).