Relations Among Factors

For a variable to qualify as a useful prognostic factor, it must be significant, independent, and clinically important.10 Many, if not all, variables are potentially of prognostic significance and many proved to be so in univariate analysis. However, variables are likely to be highly interrelated and some may be partial substitutes for one another, and few present with independent prognostic value. Moreover, factors exist that predict for a given therapy only; some for a given stage and some in the context of the presence of other factors. The use of multivariate statistical analyses is needed to assess which factors are independently significant and which only correlate with known prognostic factors but are without independent prognostic significance. This last point dramatically highlights the limits of prognostic studies, either retrospective or prospective, since the results highly depend on the variables available for all cases enrolled and the statistical model used.

0 0

Post a comment