coccal septic shock is still a large, hetero-geneous group with varying severity which needs some additional refinement. We hypothesized that the Malley et al. score could reliably identify patients with invasive meningococcal disease "at high-est risk" with a high degree of specificity without increasing the number of predic-tors and hence could be useful in the early selection of patients for trials assessing novel or aggressive therapies, as the au-thors state [1]. It is known that prognostic scores should not be used until the model has been independently validated in the target population. Interinstitutional com-parative analyses of these scores in larger groups of patients are essential to demon-strate that the model can perform well in other patients than those on whom the model was developed [4]. In our valida-tion sample the Malley et al. score per-formed better in identifying children at low risk than those at high risk. It defined a broad group of children with 100% sur-vival. However, only 35% of the survivors were predicted correctly. This low speci-ficity may be explained by the small num-ber of patients with septic shock included in their retrospective study. With regard to definitions of outcome, the majority of clinical trials in septic pa-tients use mortality as the primary study end point, but increasingly researchers are including morbidity in their trial design [5]. We do agree that limb amputation could have been included in addition to death as another primary outcome vari-able. However, when we considered death and skin graft or amputation as the depen-dent variable, the areas under the receiver operating characteristic curves provided by the Malley et al. score and the new score did not change significantly, mainly due to the low rate of this major morbidi-ty. In summary, it is important to recognize that the enrollment of patients at different stages of a heterogeneous disease repre-sents a limiting factor in demostrating sig-nificant treatment effects in randomized clinical trials. Clearly, more precise defini-tions and more accurate tools are needed for reliable comparison of patients, thera-pies, and outcomes. References 1. Malley R, Huskins WC, Kuppermann N (1996) Multivariable predictive models for adverse outcome of invasive menin-gococcal disease in children. J Pediatr 129:702–710 2. Castellanos-Ortega A, Delgado-Rodriguez M, Llorca J, et al (2002) A new prognostic scoring system for men-ingococcal septic shock in children. Comparison with other three scoring sys-tems. Intensive Care Med 28:341–351 3. Castellanos A, Gandarillas MA, Teja JL, et al (1996) Definitions for meningococ-cal sepsis in children. A review of 80 cases. An Esp Pediatr 44:219–224 4. Suter P, Armaganidis A, Beaufils F, et al (1994) Predicting outcome in ICU pa-tients. Consensus Conference organized by the ESICM and he SRLF. Intensive Care Med 20:390–397 5. (2000) Recombinant bactericidal/perme-ability-increasing protein (rBPI21) as adjunctive treatment for children with severe meningococcal sepsis: a random-ised trial. Lancet 356:961–967 Á. Castellanos-Ortega (✉)