GMS Medizinische Informatik, Biometrie und Epidemiologie (Feb 2008)
Systematic review about data quality and protocol compliance in clinical trials
Abstract
For Investigator Initiated Trials (IITs) alternative risk-adapted monitoring strategies are discussed in order to fulfill rules and regulations, taking into consideration the restricted resources. In this systematic review investigations, presenting quantitative data about data quality and protocol compliance in clinical trials, are analyzed. The results are discussed taking into account the quality assurance procedures implemented. Based on a systematic MEDLINE retrieval, 21 studies could be identified in which data and process quality in clinical trials were investigated and assessed by site visits with source data verification or review of copied source data in the study center and quantitative information about data quality and protocol compliance was available. The majority of investigations were performed by three organizations: European Organization for Research and Treatment of Cancer (EORTC) (n=7), National Cancer Institute (NCI) (n=7) and investigations of the Trans-Tasman Radiation Oncology Group (TROG) (n=4). In addition three investigations of other study groups were identified. The investigations were published between 1981 and 2003. In the majority of cases oncological trials were investigated (n=19) with a focus on radiotherapy trials (n=8). In the EORTC-trials an overall good data quality was assessed (80–90% correct data). Singular problems were found with respect to protocol compliance and reporting of adverse reactions and serious unexpected events. Good quality was also observed with respect to the correct inclusion of patients into trials. By the NCI a standardized audit system was introduced and implemented within cooperative study groups. In the context of these audits different criteria were assessed and a good data quality and protocol compliance were measured. Deficits occurred in about 5% of the cases with respect to informed consent, correct application of inclusion criteria, protocol compliance, assessment of outcome criteria, assessment of toxicity, adequate data reporting and data verification. In some investigations problems with treatment compliance (10-20%), drug dose deviations (10%) and drug accountability (15%) were identified. By the TROG a quality assurance procedure was implemented, based on central monitoring of copied source data. By this approach major problems with protocol compliance could be reduced to less than 10% together with a good data quality with an error rate under 5%. The correct handling of in- and exclusion criteria was a problem in individual cases. In summary we found out that good data quality and good to moderate protocol compliance were reported by the study groups that are included in the systematic review. Due to the authors this is mainly due to an established quality assurance system, taking into consideration that audits itself may be an influential factor. Generally, minor problems were observed with respect to informed consent, correct handling of in- and exclusion criteria and data quality, however, in some studies there were problems with protocol compliance. Overall, number and type of deficits depend on study type, quality management and organization of the study group. Fraud was detected only in very few cases. The available evidence refers mainly to established study groups; for data quality and protocol compliance outside these groups only few information is available. However, it should be taken into consideration, that the analysis was performed, at least partly, by the study groups themselves and is not based on independent audits (e.g. external audits). The analysis cannot answer the question whether the results would have been replicable if a strict review according to criteria of Good Clinical Practice (GCP) would have been performed. From the systematic review consequences have to be taken for planning a prospective controlled trial comparing different monitoring strategies. The main influence factor for data quality and protocol compliance in a clinical trial is the quality management system. Quality management covers several other measures apart from monitoring. In order to assess monitoring strategies, all quality assurance procedures within a clinical trial have to be taken into consideration. For the comparison of different monitoring strategies adequate outcome parameter have to be defined (e.g. severe deficits with respect to inclusion and exclusion criteria, safety). The analysis indicates that with good quality management and no extensive on-site monitoring severe errors were detected only at relative low frequency. It could well be that with an efficient quality management system differences between monitoring strategies would be small. In order to demonstrate statistical equivalence of monitoring strategies, only small differences can be accepted which again leads to consequences for the sample-size calculation. In addition, it must be taken into consideration that within a controlled trial an independent audit is necessary to assess the effect of different monitoring strategies on the safety of patients and the quality of data. Audits however may be a possible influence factor for data quality and protocol compliance and may complicate the evaluation of the benefit of a monitoring strategy.
