A statistical method for assessing change in immunologic parameters in a patient.

Serial measurement of in vitro immunologic parameters in patients is used to detect change in immune status over time due to disease progression and/or immunodulatory therapy. A statistical method is presented for looking at serial measurements on an individual to detect whether a change in a parameter is outside the bounds of expected within-individual variation. Analysis of variance is used, assuming a normal distribution, to obtain percentiles of the distribution of the absolute difference between consecutive values of immunologic parameters in a healthy population. The assumptions in this analysis are justified from a statistical point of view. We discuss how to use this statistical method to make judgments relevant to clinical immunology, including how to construct a table that can be used to determine quickly if an "interesting" change for some standard immunologic parameters has occurred, whether a linear (additive) or logarithmic (proportional) model for change might be more appropriate for a given parameter, and how to modify the calculations if change is expected in a certain direction or if multiple pre- and/or postevent (clinical change or intervention) measurements are available.