A new time-dependent approach for assessment of the impact of invasive aspergillosis shows effect on short- but not on long-term survival of patients with AML or high-risk MDS

Abstract

Invasive aspergillosis (IA) has been reported to yield high mortality rates. Patients with an unfavourable prognostic haematological disease not only have a higher probability of developing IA but are also more likely to die due to causes directly related to the underlying disease. This complexity of risk mechanisms confounds the causal interpretation of IA occurrence and mortality. Full consideration of the changing patient characteristics over time is necessary to obtain reliable estimates of the correlation of IA with mortality. We studied the effect of IA on mortality in 167 consecutive patients starting with remission-induction therapy for AML or of whom most patients continued to haematopoietic stem cell transplantation (HSCT). No standard antifungal prophylaxis was administered in the period before HSCT. Survival analyses were performed to determine risk estimates of IA for different phases of treatment before and after HSCT. Time-dependent adjustment for confounding variables was performed using Cox proportional hazards models. In 55 of 167 enroled patients, IA was diagnosed. Before HSCT, adjusted hazard ratios and 95% confidence intervals on mortality after the diagnosis of IA were 3.5 (1.7-7.5), 2.0 (0.69-5.9), 2.3 (0.79-6.8) and 0.80 (0.49-1.4) within 30 days, between 30 and 60 days, between 60 and 90 days or more than 90 days, respectively. A similar pattern was observed after HSCT. The occurrence of IA did not significantly influence the decision to follow through with HSCT. The results provide new insights in short- and long-term survival of patients diagnosed with IA. A significantly increased mortality risk was only observed in the first month after diagnosis of IA. No unfavourable association with mortality was observed in the later course of treatment. The occurrence of IA did not affect the probability of attaining HSCT in our population.