Background: The conduct and clinical applicability of systematic reviews of prognostic studies are affected by the poor quality of reporting in primary prognostic studies in the medical literature. Reports of primary prognostic studies of tumour markers are often compromised by poor reporting1. Concerns about poor reporting led to the development of the REMARK reporting guidelines for tumour marker prognostic studies2. In additional to the usual concerns about transparent and complete reporting of study methods, of particular concern is the adequate reporting of the flow of participants through a study. In the CONSORT guidelines for randomized controlled trials (RCTs) a flow diagram is recommended. Similar concerns apply to prognostic studies. In addition, as for any observational study, data-dependent components are important in the analysis of prognostic studies. The analysis is typically not (fully) pre-specified and is often more complex than for RCTs, particularly those including multiple analyses and subgroup analysis. Item 12 of the REMARK guidelines identifies the need to report the flow of patients and events included in the study and reasons for dropout. As a minimum a study should report the number of patients originally in the sample, the number of events observed, the number remaining after exclusions, and the numbers incorporated into univariate and multivariable analyses, including specific subgroups. A study 'profile' template has recently been developed to summarise this information. Our research will extract information on participant flow from published articles and discover the extent to which publications supply the information recommended in the REMARK guidelines.
Methods: We identified a sample of 50 prognostic tumour marker studies published in five high impact cancer journals in 2006. Our inclusion criteria included: outcome is survival or disease free survival from cancer; impact of single biological tumour marker studied; multivariate analysis. Articles on DNA microarray studies were excluded. A data extraction sheet was designed and piloted on 5 articles. A descriptive analysis will be presented on the reporting of patients, events and outcomes for studies. We will also extract information on the reporting of study and patient characteristics, methods and presentation of analyses, including the use of subgroup analyses.
Results and Discussion: We will summarise the completeness of reporting of key information in published prognostic studies, and highlight those areas needing particular improvement. We will describe in detail the reporting of the key information n participant flow needed to complete the study profile. We will illustrate the construction of study profiles in different table and diagram formats. Good reporting in primary prognostic studies is important for future systematic reviews of prognostic studies. The findings from our study will directly inform the detailed explanatory document that is being prepared to accompany the REMARK guideline.

References:
1. Kyzas PA et al. J Natl Cancer Inst 2007;99: 236.
2. McShane LM et al. J Natl Cancer Inst 2005;97:1180.