3.1 Data Pooling

The PopPK and ERs analyses were conducted using data from 353 patients (199 and 154 from the dose-finding and ASCEMBL studies, respectively, including 188 patients following a 40-mg b.i.d. dosing regimen, 18 with 80 mg q.d., and 132 patients with a total daily dose >  80 mg), with 6603 asciminib PK concentrations. Laboratory abnormalities and vital sign analyses used data from 349 patients from both studies, while the QT/QTc analysis used data from 239 patients from the dose-finding study. For the ERe analysis, subsets of this pool were used for the full analysis (303 patients), the first subgroup analysis comparing the effects of the 40-mg b.i.d. and 80-mg q.d. dosing regimens (194 patients), and the second subgroup analysis studying the effect of asciminib 200 mg b.i.d. on patients with the T315I mutation (67 patients who harbored the T315I mutation) [22].

3.2 PopPK Analysis

Both 40-mg b.i.d. and 80-mg q.d. dosing regimens showed comparable steady-state AUC0-24h (12,638 ng*h/mL and 12,646 ng*h/mL, respectively), while the average steady-state Cmax and Cmin of 80 mg q.d. were 1.61-fold and 0.72-fold that of 40 mg b.i.d., respectively. Detailed results can be found in our previous publication [22].

3.3 Exposure–Response Analysis for Efficacy3.3.1 Dosing Regimen Analysis: 40 mg b.i.d. versus 80 mg q.d.

To evaluate whether asciminib dosing regimens of 80 mg q.d. and 40 mg b.i.d. led to similar effects on the time-course of BCR::ABL1IS, the final ER model [23] was slightly modified and refitted to better describe a subset of the original dataset, which consisted of patients with starting-dose regimens of either 80 mg q.d. or 40 mg b.i.d. Since the Cmin of asciminib is about 30% lower for the 80-mg q.d. dose compared with that of the 40-mg b.i.d. dose and given that this difference in Cmin may be a contributor for differences in efficacy between the two regimens, the effect of asciminib on BCR::ABL1IS was estimated as a function of Cmin. In an additional model, a function of the dosing regimen (ARM) treated as a categorical covariate was used to further assess the difference between the two dosing regimens. Equations describing the final models have previously been published and are summarized in the ESM (Methods S3) [23].

The results of the two pharmacodynamic models are presented in Table 1.

Table 1 Parameter estimates of the pharmacodynamic model for asciminib 80 mg q.d. versus 40 mg b.i.d. in the subset of patients without the T315I mutation

In the Cmin model, the effect of Cmin as estimated by the gamma parameter is small (0.0365), such that a lower Cmin by 30% would lead to a decrease in drug effect by < 2%. Varying Cmin magnitude on the drug killing effect of asciminib on the susceptible P cell population resulted in a plateau. In the regimen model, the Effmag of the 40-mg b.i.d. regimen was 39.8, while the estimated regimen effect was − 0.045 (not statistically significant), giving an Effmag for the 80-mg q.d. regimen close to 38. These two models thus estimated a negligible difference in the drug killing effect on the susceptible leukemic P-cell population between the two regimens. Since the regimen model is associated with better model diagnostics as it has a higher log likelihood and lower Bayesian information criteria (BIC), model diagnostics are presented only for the regimen model (Figs. S1 and S2, see ESM), and the regimen model was used to perform simulations.

As expected, the predictions from simulations were very similar for 80-mg q.d. and 40-mg b.i.d. dosing regimens (Fig. 1). The resulting predicted MMR rates at Week 24 and Week 48, with or without stratification by baseline BCR::ABL1IS levels and number of prior TKIs are displayed in Table 2.

Fig. 1

Simulated time-course of log10-transformed BCR::ABL1IS for asciminib 40 mg b.i.d. and 80 mg q.d. in patients not harboring the T315I mutation: all patients (a) and patients stratified by number of prior TKIs and baseline BCR::ABL1IS levels (b). The black line represents the median over the 100 replicates of the 50th percentile of BCR::ABL1IS. The darker shaded area represents the median of 25th and 75th percentiles of BCR::ABL1IS, and the lighter shaded area is the median of 10th and 90th percentiles of BCR::ABL1IS. b.i.d. twice daily, q.d. once daily, TKI tyrosine kinase inhibitor

Table 2 MMR rate at Weeks 24 and 48 comparing 40-mg b.i.d. versus 80-mg q.d. asciminib regimens from simulations of 100 patients per trial in 100 trials based on a PD model for patients not harboring the T315I mutation

The overall predicted MMR rates for 40 mg b.i.d. versus 80 mg q.d. were 27.6% versus 24.8% and 32.3% versus 30.6% at Weeks 24 and 48, respectively. These rates are very similar for both regimens, though slightly lower for 80 mg q.d., and in close agreement with the observed rates in the ASCEMBL study (25.5% at Week 24) [17]. As baseline BCR::ABL1IS level was a significant covariate in the model, we further stratified the simulations (Fig. 1). While MMR rates were lower for patients with high baseline BCR::ABL1IS for all dosing regimens, the stratified results show that patients with higher baseline BCR::ABL1IS levels would still derive benefit from asciminib treatment. The predicted BCR::ABL1IS levels still decrease after 1 year, suggesting that these patients would benefit from a continuous long-term treatment to further decrease BCR::ABL1IS. Figure 1 and Table 2 show that patients with more lines of prior TKI therapy (≥ 3) would also benefit from asciminib treatment, with an efficacy similar to that predicted for patients treated with fewer prior TKIs.

3.4 Analysis for Patients Harboring the T315I Mutation

We used the final T315I model [23] to further evaluate the efficacy of asciminib in patients with the T315I mutation following a 200-mg b.i.d. dosing regimen, using an Emax model (Methods S3, see ESM). Compared with the dose of 200 mg b.i.d., this analysis indicated that patients treated with lower dose regimens were less likely to respond to treatment (Fig. 2). Only the three highest doses (120, 160, and 200 mg b.i.d.) showed a general decrease in BCR::ABL1IS levels over time, suggesting that these regimens achieved higher reductions in BCR::ABL1IS as compared with 40 mg b.i.d. and 80 mg q.d. in patients with the T315I mutation (Fig. 2). Findings of the ERe analysis also showed the benefit of using the highest clinical dose to treat patients harboring the T315I mutation: indeed, at 200 mg b.i.d., 99.4% of patients had asciminib exposure (measured as AUC) above the estimated 90% effective concentration (EC90). Lower doses of asciminib also resulted in a lower proportion of patients achieving EC90 than 200 mg b.i.d. (120 mg b.i.d. [84.5% patients] and 160 mg b.i.d. [96.8% patients]) [23].

Fig. 2

Simulated time-course of log10-transformed BCR::ABL1IS for asciminib 40 mg b.i.d., 80 mg q.d., 120 mg b.i.d., 160 mg b.i.d., or 200 mg b.i.d. in patients (with any number of prior TKI treatments) harboring the T315I mutation. The black line represents the median over the 100 replicates of the 50th percentile of BCR::ABL1IS. The darker shaded area represents the median of 25th and 75th percentiles of BCR::ABL1IS and the lighter shaded area is the median of 10th and 90th percentiles of BCR::ABL1IS. b.i.d. twice daily, q.d. once daily, TKI tyrosine kinase inhibitor

Based on the simulated BCR::ABL1IS, the predicted MMR rates for the 200-mg b.i.d. dose regimen at Weeks 24 and 48 were 20.7% and 23.7%, respectively. Similar MMR rates were obtained for 120 mg b.i.d. (19.3% and 22.9%) and 160 mg b.i.d. (20.2% and 23.8%).

3.5 Exposure–Response Analysis for Safety

The efficacy of 80 mg q.d., 40 mg b.i.d. and 200 mg b.i.d. was demonstrated in the earlier publication [22] and in the previous section. Evaluation of a potential link between exposure and safety was analyzed and results are described in this section to justify the benefit/risk between efficacy and safety.

3.5.1 Laboratory Abnormalities

The analysis of laboratory abnormalities showed no statistically significant relationship between asciminib exposure and the majority of the laboratory events analyzed. Figure 3 presents the results of the repeated measures logistic regression analysis. Regression coefficient estimates were negative for the exposure metrics for the majority of laboratory events (including amylase increase [any grade and grade ≥ 3], platelet decrease [grade ≥ 3], neutrophil decrease [grade ≥ 3], ALT increase [grade ≥ 2] and triglyceride increase [any grade]), suggesting no increase in the risk of laboratory events with increased exposure. Exposure metric regression coefficients were positive for several laboratory abnormalities (lipase increase of any grade and grade ≥ 3, hemoglobin decrease of grade ≥ 3, and bilirubin increase of grade ≥ 2), but they were not statistically significant, suggesting minimal impact of increased exposure on the risk of these laboratory abnormalities. Only AST increase of grade ≥ 2 showed a significant positive estimate for the exposure metric, with a p-value of 0.015 for Cmax and 0.088 for AUC. However, the frequency of these events was very low in the ASCEMBL study (1.3% of patients and 0.1% of cycles). For a 5-fold increase in the geometric mean Cmax with the 40-mg b.i.d. dose, the predicted probability of experiencing an event within a cycle increases from 0.3% to 0.9% for the dose-finding study and from 0.1 to 0.4% for the ASCEMBL study, which highlights the rarity of these events (Table S1, see ESM).

Fig. 3

Regression coefficient estimates of repeated-measures logistic regression model for asciminib exposure metrics versus laboratory events, vital signs (hypertension) and AEs of fatigue/asthenia. AE adverse event, AUC area under the concentration–time curve, CI confidence interval, Cmax maximum plasma concentration, Cmin minimum plasma concentration

3.5.2 Adverse Events and Vital Signs

The results of the repeated measures logistic regression analysis showed a negative coefficient estimate (range, − 0.075 to − 0.008) for the asciminib exposure metrics (AUC, Cmax, and Cmin) for hypertension or fatigue and asthenia (any grade), as well as for TEAEs of grade ≥ 3, (AUC, − 0.211; Cmax, − 0.256; Cmin, − 0.245) (Fig. 3). These results suggest no clinically relevant relationship between exposure and these safety events. TEAEs leading to dose reduction or dose interruption were reported in 152 of 351 patients (43.3%). A time-dependent Cox regression analysis provided positive coefficient estimates of 0.032 for AUC (p = 0.715), 0.074 for Cmax (p = 0.449), and 0.016 for Cmin (p = 0.834), but these results were not statistically significant.

3.5.3 QT/QTc Analysis

The estimated mean QTcF changes from baseline were 3.35 (90% CI 2.28–4.43), 3.64 (90% CI 2.60–4.68), 5.37 (90% CI 3.97–6.77), and 6.77 (90% CI 4.67–8.87) ms for Cmax at 40 mg b.i.d., 80 mg q.d., 200 mg b.i.d., and the HCRE, respectively, below the regulatory threshold of 10 ms and therefore not clinically significant according to the regulatory guidance (Fig. 4).

Fig. 4

Scatterplot, regression line (red), and 90% CI (blue shaded area) of change from baseline QTcF versus plasma asciminib concentration. Data are from 239 patients from the dose-finding study (patients with CML and Ph+ acute lymphoblastic leukemia). Each circle represents a data point (PK sample with time-matched QTc assessment) any time post-asciminib dose. Vertical lines represent median Cmax for 40 mg b.i.d., 80 mg q.d., 200 mg b.i.d. and HCRE. Scatterplot includes only the matched ECG records. The model was QTcF change from baseline = concentration + baseline QTcF are fixed effects, and patient is a random effect. b.i.d. twice daily, CI confidence interval, Cmax maximum plasma concentration, HCRE highest clinically relevant exposure, PK pharmacokinetic, q.d. once daily, QTcF QT interval corrected using the Fridericia method, TKI tyrosine kinase inhibitor

3.5.4 ECG Analysis From Clinical Data

In the dose-finding study, new QTcF > 500 ms was noted in 3/241 (1.2%) patients (of whom two also had increase in QTcF > 60 ms from baseline). However, none of these abnormalities were associated with cardiac-related symptoms. In the ASCEMBL study, new QTcF > 500 ms with increase in QTcF > 60 ms from baseline was noted in 1/156 (0.6%) patients, which was reported as a treatment-related grade 3 ECG QT prolongation. This event was managed with treatment interruption and dose reduction, and no subsequent episodes of QTcF > 500 ms were observed in this patient.