FIG. 2.

POP for occurrence of AIDS. This graphic method allows ready comprehension and comparison of several surrogate markers. Calculations are based on survival analysis regression methods with Gaussian error distribution, as described in detail in reference 25. Lower values are better; a perfect predictor would be one that had zero deviation and could predict future occurrence of AIDS to a precise day. This would have a scale value of zero. (A) Illustrative diagram of three hypothetical surrogate markers, X, Y, and Z. In this example, assay Y is not as good as X but is better than Z. (B) Graphic presentation of several surrogate markers. A group of 659 seropositive men with mean CD4 T-cell levels of 514/mm³ were evaluated for prediction of AIDS occurrence within 3 years (25). SP, prognostic precision for the study population as a whole without surrogate markers; T, plasma sTNF-α–RII levels; 4, CD4 T-cell levels; H, plasma HIV concentration; C₁, combined value for CD4 T cells plus HIV load; C₂, combined value for CD4 T cells plus sTNF-α–RII.