Page 130 - Haematologica March 2020

M. Hoeks et al.
was a difference by country as to whether a patient was treated with ICT; patients in the UK were less likely to be treated.
The overlap of propensity scores of both groups (chelated and non-chelated), which is essential for PS matching, was good for the majority of the patients (Online Supplementary Figure S2). The matched MI dataset included 197 of 199 chelated cases and identified 591 non-chelated controls. There were no differences by sex, RBCT intensity, cumulative RBCT units, serum ferritin levels, comorbidity, performance status, IPSS-R, presence of ringed sideroblasts, quality of life (QoL), and country between both groups (Table 2). Figure 3 shows the unad- justed survival plot by ICT status with receiving ICT as a time-dependent variable for the matched patients. A mul- tivariate Cox proportional hazard model was used to
adjust for potential confounders (age, sex, comorbidity, performance status, monthly RBCT intensity, number of RBC units transfused, IPSS-R, and presence of ringed sideroblasts). The estimated crude and adjusted hazard ratios were 0.70 (95% CI: 0.51-0.95) and 0.42 (0.27-0.63), respectively (Table 2) and the adjusted survival curve is shown in Figure 4. When we again restricted the analysis to the deferasirox-treated patients, the crude HR for OS was 0.63 (95%CI: 0.45-0.88) and the adjusted HR was 0.34 (95%CI: 0.22-0.53).
The distribution of erythropoiesis-stimulating agent (ESA) and lenalidomide-treated patients among chelated and non-chelated patients at time of eligibility were sim- ilar in the unmatched and matched sample. A sensitivity analysis excluding the treatment of ESA and lenalidomide showed similar results.
P<0.001
Figure 2. Overall survival by iron chela- tion therapy as a time-dependent vari- able in unmatched patients.
P<0.001
Figure 3. Overall survival by iron chela- tion therapy as a time-dependent vari- able in matched patients.
644
haematologica | 2020; 105(3)