Abstract
Background: Venetoclax (VEN) as a BCL-2 inhibitor, has shown promising results in treating adult acute myeloid leukemia (AML). However, studies on the relationship between VEN plasma concentrations and clinical efficacy and prognosis are still insufficient, significant inter-individual pharmacokinetic variability and controversy over the role of therapeutic drug monitoring (TDM) persist. According to the importance of plasma concentrations in clinical treatment, this study aims to analyze the impact of VEN plasma concentrations on the efficacy and prognosis of AML treatment.
Methods: This retrospective study included 132 adult AML patients treated with VEN-based regimens in West China Hospital from 2021 to 2024. The primary endpoint was clinical response. Logistic regression and restricted cubic splines (RCS) were used to assess linear and non-linear relationships. Maximally selected rank statistics (Mann-Whitney U) was used to identify the optimal concentration cutoff that best-differentiated responders from non-responders. A permutation test (n=1000) was performed to derive a robust p-value to correct for multiple-testing bias. Receiver operating characteristic (ROC) curve analysis was used to illustrate the relationship between sensitivity and specificity.
Results: In the overall cohort (N=132, 82 newly diagnosed, 50 relapsed/refractory), the overall response rate (ORR) was 58.3% (77/132), while 64.6% in ND and 48.0% in R/R. Responders had numerically higher median VEN concentrations than non-responders (1920.00 ng/ml vs. 1165.00 ng/ml), though the distributions were wide and overlapping. No significant linear relationship between VEN concentration and response was observed via logistic regression (p=0.124) or Spearman correlation (ρ=0.146, p=0.094). RCS analysis also failed to detect a significant non-linear association (p=0.615).For threshold analysis, using maximally selected rank statistics identified an optimal VEN plasma concentration threshold of 1685.00 ng/ml. After rigorous correction via permutation testing, this cutoff remained statistically significant difference between responders and non-responders (maximal U statistic = 2594.0, permutation-adjusted p=0.032), demonstrating modest discriminatory power (AUC: 0.586, Sensitivity: 0.597, Specificity: 0.600). In subgroup analyses, this threshold did not reach statistical significance after permutation correction, neither in ND patients (N=82; cutoff 1685.00 ng/ml, adjusted p=0.096) nor in R/R patients (N=50; cutoff 1887.00 ng/ ml, adjusted p=0.116). Meanwhile, stratification of patients by the identified threshold was not associated with a significant difference in overall survival in each group (p=0.096 for overall; p=0.097 for ND; p=0.753 for R/R).
Conclusion: The relationship between venetoclax plasma concentration and clinical response in AML appears to be characterized by a therapeutic threshold rather than a continuous linear association. Our analysis identifies a statistically robust efficacy threshold of approximately 1685 ng/ ml. Critically, achieving this concentration did not translate to improved overall survival, suggesting a narrow therapeutic window likely constrained by concentration-dependent toxicities. Although the threshold observed in the subgroup analyses was similar to that of the overall cohort, a statistically stable threshold was not identified, which may be related to smaller sample sizes. Notably, the potential threshold was higher in R/R patients, suggesting that this population may require higher therapeutic concentrations to achieve a response. These findings underscore the need for individualized strategies that integrate pharmacokinetics with biological markers to truly optimize venetoclax therapy in AML.
