Birgitta Versluijs

72 Results allo-LS and nonallo-LS patients. This score was also significantly different between the groups of diseases (Figure 1). A logistic regression analysis was performed to ascertain the effects of the individual severity scores of ground-glass pattern and airtrapping on the likelihood that the patients had allo-LS. Increasing scores in ground-glass pattern and airtrapping were associated with increased likelihood of allo-LS, with odds ratios of 1.04 and 1.06 per score point. The regression coefficients were transformed to weight factors according to their relative contribution to the risk estimation to reach a final weighed allo-score: ground-glass pat- tern + 1.5 airtrapping (Table 6). The capacity of the weighed and non-weighed allo-scores to predict allo-LS was calcula- ted using an ROC curve model. There was no significant difference between the weighed score and the non-weighed score, both with an AUC of 0.82, which is considered good for a diagnostic test. Therefore, we further studied the simple nonweighed score consis- ting of the sum of ground-glass pattern and air trapping scores only. Test characteristics for the different cutoff values of this allo-score are shown in Table 7. FIGURE 1. Box and whisker chart shows median, interquartile range, and range for “alloscore” (ground-glass pattern + airtrap- ping) in alloimmune-mediated lung syndromes (allo-LS) and non-alloimmune lung syndromes (nonallo-LS) early after hema- topoietic cell transplantation. Median allo-score for nonallo-LS patients was 6.9 (25–75%, 0–16.7). Median allo-score for allo-LS patients was 29.2 (25–75%, 19.4–33.3). Circle indicates outlier. Median Allo-Score 60 50 40 30 20 10 0 0 Nonallo-LS 1 Allo-LS 4