Assessment Of Prognostic Value

While adequate statistical methods are now applied by most investigators, the material used varies from one series to the other. Prognostic factors were never assessed from a population-based series that are available from cancer registries only. The main reason is that data would be incomplete and heterogeneous and highly dependent upon the institution where the patients have been treated. Hospital-based series are biased and very often limited in number since Hodgkin's lymphoma is a rare disease. When a large single institution series is used, it generally includes patients treated over a long period of time during which treatments may have changed dramatically, leading to the loss of predictive power for most factors tested.41 To overcome this difficulty, large homogeneous series are needed that may come from cooperative groups; alternatively, the combination of data from clinical trials with the same objectives (i.e., meta-analysis approach) has proved to be effective.

The prognostic value of scoring systems should be confirmed using independent series of patients. This approach can be applied within series when the analysis is made using a random sample of patients and the validation made using the remaining.24 Another possibility is to apply and/or compare various scoring systems using an independent series.42-44 Only factors or scoring systems that are consistently found of independent prognostic value in independent series should then be used.

0 0

Post a comment