Page 330 - 2022_01-Haematologica-web

Letters to the Editor
post hoc test for multiple comparisons was used to deter- mine statistical significance. Unsupervised analysis was performed to identify immune cell populations: data were clustered in the Vortex package7 using the x-shift algorithm. Elbow-point validation was used to affirm the correct cluster number. Differences in cluster frequency between groups were assessed by Mann-Whitney test for statistical significance. Cluster phenotypes were determined and validated via multiple visualization approaches; individualized clusters were visualized using brick plots.8
Comparisons of the maintenance arms were restricted to a modified ITT (mITT) set which excluded patients randomized in error. Secondary clinical time-to-event outcomes (PFS/OS) were compared between random- ized treatments using log-rank tests and estimates of the survival distributions were calculated using the Kaplan-
Meier method. Two-tailed P-values were used for all comparisons, and, unless otherwise stated, were per- formed using a significance level of 5%. Toxicity was assessed according to CTCAE version 4.0.
154 patients were enrolled (baseline characteristics listed in Table 1). The estimated median potential fol- low-up (by reverse Kaplan-Meier) for all registered patients for OS was 27.8 month (mo). Eighty-one patients were randomized, however, three were ran- domized in error, therefore a mITT analysis included 78 (51%) patients who achieved SD or better with POM- LoDEX induction: POM n=40, POM-LoDEX n=38. Median PFS (from time of randomization) was 2.6 mo (95% confidence interval [CI]: 1.8-3.0) for POM versus 5.7 mo (95% CI: 4.5-7.5) for POM-LoDEX (log-rank P=0.051; hazard ration [HR]: 0.63, 95% CI: 0.40-1.00) (Figure 1A). Median OS (from time of randomization)
A
B
Figure 1. Kaplan-Meier survivor functions for modified Intention to treat population (from time of randomization) (mITT: pomalidomide [POM] n=40; LoDEX: pomalidomide low-dose dexamethasone [POM-LoDEX] n=38). In anticipa- tion of early or late differences between the maintenance treatment arms in their time-to-event outcomes, 6 compar- isons between the arms were planned at 3, 6, 9, 12, 15 and 18 months (mo) from randomization. To account for multiplic- ity of comparisons, a Bonferroni adjustment to the alpha-level of each test was implemented, namely a comparison between the treatment arms at one of these time points was judged to be statistically significant if the associated P-value was ≤0.0083. The test was based on the complementary log- log transformation of the survival function. (a) Progression free survival: POM arm 2.6 mo (95% confidence interval [CI]: 1.8-3.0) vs. 5.7 mo (95% CI: 4.5-7.5) for POM-LoDEX (log-rank P=0.051; hazard ratio [HR]: 0.63, 95%CI: 0.40-1.00), early PFS favored POM-LoDEX, however late survival favored POM: a comparison of PFS at 6 3-monthly intervals favored POM- LoDEX (3-12 mo, P<0.001) however, at 18 mo, POM was favored (P=0.018). (B) Overall survival: POM arm 25.7 mo (95% CI: 16.7-42.2) vs. 17.4 mo (95% CI: 12.5-NA) for POM- LoDEX (P=0.356; HR: 1.36, 95%CI: 0.70-2.64). Like the pro- gression-free survival (PFS) analysis, comparisons of overall survival (OS) at 6 3-monthly intervals demonstrated no differ- ence between the arms at 3-12 mo, however at 15 mo and 18 mo, OS favored POM (P=0.006, P=0.021 respectively).
322
haematologica | 2022; 107(1)