Invasive mold infection is a major cause of morbidity and mortality among severely immunocompromised individuals. We discuss the challenges involved in the design and interpretation of salvage antifungal trials, focusing on mold infection. We suggest that patients with refractory fungal infection be analyzed separately from those with intolerance to standard regimens because of the poorer prognosis of the former group. We propose a composite outcome assessment in which refractory infection is defined as infection associated with the worsening of at least 2 of the following 3 types of criteria: clinical, radiologic, and mycologic. Confounding variables, including heterogeneity in host factors, initial antifungal therapy, and selection bias, are discussed. Although randomized studies would provide the most credible results, the lack of an adequate number of patients to meet prespecified stratification criteria for all confounding variables makes such studies impractical. Given that randomized studies are unrealistic, studies involving carefully selected, matched, contemporaneous control subjects are likely to be the most useful alternative.