In the previous issue of Critical Care, we read with interest the reaction of Girbes and Zijlstra [1] to our article on the role of autopsy in critically ill patients [2]. The authors believe that the declining autopsy rate is acceptable since current medicine is based on guidelines. However, guidelines can be driven by findings in large series of autopsies. Candida pneumonia occurs rarely in patients in whom Candida species are isolated in respiratory specimens; this argues against treating mechanically ventilated patients with antifungal drugs solely on the basis of a positive respiratory culture [3]. The recently published guidelines of the Infectious Diseases Society of America are also against such a practice [4].

We are convinced that the sensitivity and specificity of autopsy decline because of a lack of routine. Only pathologists who frequently perform autopsies are able to reveal rare pathologies. Good sensitivities and specificities of a test can be achieved only with a large sample size. Moreover, the autopsies should be performed in the presence of the treating intensivist in order to improve the yield of the autopsy. Innovative techniques also arise and might improve diagnostic performance (for example, molecular analysis in sudden death [5]).

Finally, we believe that autopsy is not always a nonrandom sample from a small selected population. Roosen and colleagues [6] found an autopsy rate of 93% in the medical intensive care unit. Some firm conclusions were drawn (for example, fungal pneumonia is among the most frequently missed diagnoses in a medical intensive care unit) [6]. Although we do realize that such high autopsy rates belong to the past rather than to the future, we think that autopsies remain valuable even in the era of modern medicine.