Birgitta Versluijs

47 Association between respiratory virus and alloimmune lung syndromes until recovery, up to a maximum of 6 courses. Recovery was defined as normalization of PFTs and/or resolved symptoms, with no extra oxygen requirement. In between the subsequent courses of MP, prednisone 0.5 mg/kg/day was given. Other immunosup- pressive agents (usually cyclosporine) were continued. In addition, azythromycin was gi- ven, because of its suggested immunomodulatory effect. 24 Along with immunosuppres- sive therapy, supportive care was provided, with extra oxygen and mechanical ventilation when necessary. IgG level was maintained above 4 g/L. Endpoints The primary endpoint of this study was the development of acute and chronic allo-LS. The secondary endpoint was OS. Statistical analysis Differences between the RV-positive and RV-negative groups were tested using Pearson’s X 2 test. Results with a P value <.05 were considered statistically significant. The duration of follow-up was the time to the endpoints, the development of an allo-LS and death, or the last assessment for survivors. To analyze risk factors for outcomes, we considered variables associated with the recipient (age at transplantation, sex, CMV se- rology, RV positivity, single/multiple viruses), the disease (malignant vs nonmalignant), the donor/transplantation technique (cell source, HLA disparity, donor relationship, conditioning regimen), HSCT complications (allo-LS, acute GVHD [aGVHD], CMV and adenovirus plasma DNA positivity, venoocclusive disease), and relapse. To examine the influence of the various viruses on the primary endpoint, rhinovirus was compared with the other viruses, and multiple viral infection was compared with single viral infection. In the analyses, we tested for allo-LS as a group (IPS 1 BOS/BOOP) based on the hypo- thesis that early viral infection might be a trigger for both acute and chronic allo-LS. In addition, we tested both syndromes separately (ie, BOS/BOOP excluding IPS from the analyses, and IPS excluding BOS/BOOP). Associations between variables (including recipient, disease, and HSCT technique) and the primary endpoint were evaluated using Cox proportional hazard models. Dichoto- mous outcomes (eg, allo-LS: yes/no) were used as dependent variables, and predictors were used as independent variables. Univariate predictors of outcome with a P-value <.10 were used for multivariate analysis. Results are expressed as hazard ratios (HRs) and corresponding 95% confidence interval (CIs). CIs not including 1 were considered statistically significant. Analyses for the association between HSCT complications and the primary endpoint (allo-LS) as well as the secondary endpoint (OS) were done using logistic regression. Di- chotomous outcomes (eg, allo-LS or survival: yes/no) were used as dependent variables, 3