The importance of detecting cytomegalovirus infections in studies evaluating new therapies for severe sepsis.

Cytomegalovirus infection is present in the majority of the population and is associated with substantial morbidity and mortality in immunocompromised patients. Recently, this viral infection has also been reported in nonimmunocompromised critically ill patients. We hypothesize that active cytomegalovirus infection may confound survival outcomes in severe sepsis trials. Scenarios based on three published studies on activated protein C were used as real examples for statistical simulations looking for how much bias in favor of or against activated protein C would be observed in single-arm studies (scenario 1), phase II trials (scenario 2A/2B), and phase III trials (scenario 3A/3B). Scenario 1 simulations evaluated a single-arm study and demonstrated that confounding biases could cause an absolute decrease of 5.7% to an increase of 6.1% on mortality rates of similar studies. Scenarios 2A and 3A evaluated phase II and III trials assuming that activated protein C is truly effective; they showed that trials with an imbalance on the proportion of active cytomegalovirus infection between study arms could lead to false-negative rates (β errors) in up to 25% of phase II trials and up to 42% of phase III trials. Scenarios 2B and 3B assumed that activated protein C was truly ineffective; they showed that imbalanced cytomegalovirus infection between arms would cause false-positive rates (α errors) in up to 5% of phase II and up to 7% of all phase III trials. The inclusion of an imbalanced proportion of patients with active cytomegalovirus infection may severely compromise the reliability of outcome results of severe sepsis trials independent of their design. Even randomized trials could have a much higher probability of false-negative rates for a new therapy than designed. Future severe sepsis trials should consider including active cytomegalovirus infection as a prospective covariate.