Xenotransplantation — will tolerance be essential ?

kidney disease (CKD) adds to their prognostic evaluation in terms of mortality. They conclude that the determination of inflammatory markers is not helpful if age and comorbidity are taken into account. It would, however, be appropriate if some further analyses were performed before such a conclusion is advanced. The conclusion of Caravaca et al. [1] is based on a multivariate Cox-regression analysis in which the significance of CRP as a predictor of mortality is lost after adjustment for other variables. For this analysis, CRP and other inflammatory markers are dichotomized, while other variables, including age, are left unchanged with retention of full variance. Entrance of dichotomized and continuous variables together in one Cox-regression model is like amputating a swimmer’s legs before competition. If a predictor, like CRP, has a continuous distribution, dichotomization along the median results in a loss of predictive power by 35% [2,3]. Another issue is that of multicollinearity. When different domains of the same phenomenon, such as inflammation, are studied simultaneously, the variables tend to correlate with each other, thus violating the assumptions of multivariate regression models [4]. Caravaca et al. [1] do not report whether or not there were significant correlations between CRP, white blood cell counts and counts of polymorphonuclear leucocytes. However, significant correlations of all these parameters with the negative acute phase response marker, serum albumin, existed. The authors should thus at least repeat their multivariate analysis with inflammatory markers as continuous variables, in the absence of serum albumin as a variable in the model. If the other inflammatory markers are found to be correlated, this analysis should ideally be repeated for each of the markers separately. Then, if the authors still find the same loss of significance of each of the inflammatory markers as a predictor of mortality in the multivariate model(s), their conclusion is valid.