Systematic review of prognostic factor studies, with particular emphasis on measures of angiogenesis in non-small cell lung cancer
Marialena Trivella and Doug Altman with Francesco Pezzella, Adrian Harris (Oxford) and Ugo Pastorino (Milan)
The vast number of publications of prognostic factors has raised more controversy than certainty. There is considerable concern about the methodological weaknesses of many studies, and it is also believed that publication bias is rife. Studies of prognosis have received rather little attention by those carrying out systematic reviews, but it is already apparent that reviews based only on published data are severely hampered by poor reporting of primary studies. Systematic reviews with individual patient data meta-analysis offer a way to get more complete data. This project was set up to investigate the feasibility of collecting such data in the prognostic context.
The exploration of angiogenesis (assessed by microvessel density counting) as a potential prognostic factor in non-small cell carcinomas was used as a model project. The Prognosis in Lung Cancer project (PILC) obtained data for about 3000 patients from 16 centres around the world. During lengthy data cleaning and standardisation it became apparent that laboratory techniques vary considerably with potential influence on study outcomes. In particular, two rather different measurement techniques had been used. Meta-analysis techniques are being applied to investigate the hypothesis that angiogenesis is a prognostic factor in this disease, considering each main measurement method separately.
While we have demonstrated that such a study is feasible, it proved to be extremely time-consuming, raising concerns about cost-effectiveness. In particular, it is clear that while access to the raw data is very valuable in clarifying the nature of the evidence it cannot overcome deficiencies within studies nor inconsistencies between them.
