Study objectives: To provide a systematic overview of the literature investigating patient and tumor factors that are predictive of survival for patients with non-small cell lung cancer (NSCLC), and to analyze patterns in the design of these studies in order to highlight problematic aspects of their design and to advocate for appropriate directions of future studies.
Design: A systematic search of the MEDLINE database and a synthesis of the identified literature.
Measurements and results: The database search (January 1990 to July 2001) was carried out combining the MeSH terms prognosis and carcinoma, nonsmall cell lung. Eight hundred eighty-seven articles met the search criteria. These studies identified 169 prognostic factors relating either to the tumor or the host. One hundred seventy-six studies reported multivariate analyses. Concerning 153 studies reporting a multivariate analysis of prognostic factors in patients with early-stage NSCLC, the median number of patients enrolled per study was 120 (range, 31 to 1,281 patients). The median number of factors reported to be significant in univariate analyses was 4 (range, 2 to 14 factors). The median number of factors reported to be significant in multivariate analyses per study was 2 (range, 0 to 6 factors). The median number of studies examining each prognostic factor was 1 (range, 1 to 105 studies). Only 6% of studies addressed clinical outcomes other than patient survival.
Conclusions: While the breadth of prognostic factors studied in the literature is extensive, the scope of factors evaluated in individual studies is inappropriately narrow. Individual studies are typically statistically underpowered and are remarkably heterogeneous with regard to their conclusions. Larger studies with clinically relevant modeling are required to address the usefulness of newly available prognostic factors in defining the management of patients with NSCLC.