Statistical Analysis

Survival time was partitioned into 3 health states: time with any treatment‐emergent grade 3 or higher adverse events (TOX); time without symptoms of disease progression or toxicity (TWiST); and time after treatment discontinuation due to insufficient clinical response, relapse, or death (REL). For the REL definition, patients who discontinued treatment for other reasons (including AEs) were censored at the date of discontinuation unless death occurred within 28 days of discontinuation. For TOX, the duration of an AE was calculated as the difference between the AE start and end dates. If the end date occurred on the same date as or after REL, the end date for the AE was imputed as the REL date and was counted as an event in order to not double‐count time after REL. A patient in this circumstance would have no TWiST by definition. AEs with overlapping duration were also truncated; only unique AE days were counted toward TOX to avoid redundancy or double counting.

Each health state's restricted mean duration was obtained by calculation of the area under the Kaplan‐Meier curve. A 20‐month cutoff for the maximum follow‐up was used to estimate restricted means based on the median OS in the trial. Differences in mean health state durations between treatment arms were tested with log‐rank tests. We then calculated the Q‐TWiST by summing up the time in each health state multiplied by its respective utility weight (range, 0‐1) to reflect patient preferences for time spent in each health state according to the following equation:

where U_TWiST is the utility weight for TWiST, U_TOX is the utility weight for TOX, and U_REL is the utility weight for REL. In the base‐case scenario, U_TWiST was assumed to be equal to 1 and represented a “perfect” or “best” state of health/quality of life (QOL), and the utility weight was 0.5 for TOX and REL, which was consistent with the utility weights often used within the Q‐TWiST literature. 6 To assess the precision of the mean restricted time in each health state, the overall Q‐TWiST, and the difference in Q‐TWiST, 95% CIs were computed with 1000 bootstrapped samples (with replacement) of trial patients. Finally, we calculated the relative Q‐TWiST gain (ie, difference in Q‐TWiST between arms divided by the mean OS of the LDAC group). According to Revicki et al's criteria, 16 a relative Q‐TWiST gain ≥10% was considered “clinically important,” and ≥15% was considered “clearly clinically important.”

A threshold analysis was conducted whereby the utility values for TOX and REL were varied between 0 and 1 (ie, to cover the range of possible values) in steps of 0.5.

Several TOX definitions were included for sensitivity analyses. The calculation of TOX was conducted with 2 separate definitions beyond the base case (ie, grade 3 or higher AEs before progression): the first limited TOX time included only symptomatic grade 3 or higher AEs, which excluded AEs that were listed as laboratory investigations (eg, “blood fibrinogen decreased” or “red blood cells urine positive”) according to the Medical Dictionary for Regulatory Activities preferred category term, and the second definition included any all‐cause AE, regardless of grade (grades 1‐5).

The BRIGHT AML 1003 trial followed up patients receiving glasdegib plus LDAC for a median of 21.7 months and patients receiving LDAC for a median 20.1 months. 2 For this analysis, we used a base‐case follow‐up time of 20 months. In a sensitivity analysis, the follow‐up time was modified to 6, 12, 18, and 24 months to understand the impact of various time horizons on overall Q‐TWiST.

In the BRIGHT AML 1003 trial, patients were stratified by cytogenetic risk based on their profile at study entry as either poor risk (ie, inv(3), t(6;9), 11q23, −5, −5q, −7, abnormal [17p], or a complex karyotype [3 or more clonal abnormalities]) or good/intermediate risk (ie, lacking the features of poor‐risk patients). 2 Analyses were also performed for the following prespecified subgroups: age (<75, ≥75, or ≥65 years), Eastern Cooperative Oncology Group (ECOG) performance status (PS; 0‐1 vs ≥2), AML (de novo vs secondary), and bone marrow blast count at the baseline (20%‐30% vs ≥30%).