Abstract 89P
Background
Selinexor is a first-in-class oral selective inhibitor of nuclear export that is approved in the US for treatment of patients with relapsed multiple myeloma, and has demonstrated anti-cancer activity in patients with other haematological and solid tumor malignancies.
Methods
Diagnostic, demographic and baseline laboratory data from 1,936 cancer patients treated with selinexor on 13 clinical trials were used to predict tolerability (measured by days on selinexor) and best overall response. An integrated machine learning model averaging neural networks, random forests, logistic regression, and gradient boosting was used. To distinguish between predictive markers for selinexor and general prognostic markers, patients enrolled on the BOSTON study treated with selinexor, dexamethasone, and bortezomib (SVd), were compared to patients on the control arm (Vd).
Results
A combined machine learning model for predicting response that tested over 70 variables among selinexor treated patients achieved an area under the receiver operating characteristic curve of 0.674. Lower lactate dehydrogenase (LDH), higher hematocrit (HCT) and higher hemoglobin (HGB) were found to have the strongest associations with positive response. In a separate analysis of selinexor tolerability, higher HGB, lower LDH and higher HCT were also the three strongest predictors of longer time on study. When applying these models to the BOSTON study, variable correlations with treatment duration were stronger and more consistent with data from other trials for patients on the SVd arm compared to those on the Vd arm, indicating specificity for predicting tolerance to selinexor (Table). Table: 89P
Comparison of single variable correlations with time on study between two arms of the BOSTON study
| Variable (measured at baseline) | SVd arm | Vd arm | Overall correlation in selinexor-treated patients from all trials |
| hemoglobin | 0.14 | 0.09 | 0.18 |
| lactate dehydrogenase | -0.12 | -0.01 | -0.18 |
| hematocrit | 0.16 | 0.13 | 0.17 |
| red blood cell count | 0.16 | 0.10 | 0.16 |
| albumin | 0.25 | 0.10 | 0.15 |
| neutrophil/lymphocyte ratio | -0.04 | -0.19 | -0.12 |
| creatine kinase | 0.17 | 0.14 | 0.11 |
| ECOG performance status | -0.06 | 0.09 | -0.11 |
| pulse | -0.12 | -0.22 | -0.11 |
| neutrophil count | -0.10 | -0.18 | -0.10 |
Conclusions
Lower LDH and higher HCT/HGB may be useful for predicting patients who will respond safely and favorably to selinexor. Our combined models using common lab values that can easily be assessed may help in identifying patients most likely to derive clinical benefit from selinexor treatment.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
Karyopharm Therapeutic Inc.
Funding
Karyopharm Therapeutic Inc.
Disclosure
Y. Artstein: Full/Part-time employment: Karyopharm Therapeutics. C. Walker: Shareholder/Stockholder/Stock options, Full/Part-time employment: Karyopharm Therapeutics. F. Yang: Shareholder/Stockholder/Stock options, Full/Part-time employment: Karyopharm Therapeutics. D. Van Domelen: Shareholder/Stockholder/Stock options, Full/Part-time employment: Karyopharm Therapeutics. D. Borochov: Shareholder/Stockholder/Stock options, Full/Part-time employment: Karyopharm Therapeutics. I. Mercier: Shareholder/Stockholder/Stock options, Full/Part-time employment: Karyopharm Therapeutics. J. Shah: Leadership role, Shareholder/Stockholder/Stock options, Full/Part-time employment: Karyopharm Therapeutics. S. Shacham: Leadership role, Shareholder/Stockholder/Stock options, Full/Part-time employment, Officer/Board of Directors: Karyopharm Therapeutics. Y. Landesman: Leadership role, Shareholder/Stockholder/Stock options, Full/Part-time employment, Officer/Board of Directors: Karyopharm Therapeutics. S. Tang: Leadership role, Shareholder/Stockholder/Stock options, Full/Part-time employment, Officer/Board of Directors: Karyopharm Therapeutics. E. Shacham: Leadership role, Shareholder/Stockholder/Stock options, Full/Part-time employment: Karyopharm Therapeutics.