Medically reviewed by Dr. Samuel Sarmiento, M.D., MPH on August 3, 2023
A real-world data-derived algorithm was determined to be a feasible and accurate method for assessing treatment response in patients with multiple myeloma. Findings suggest that the derived response algorithm can consistently reproduce response assessments in patients with multiple myeloma in a clinical trial setting and assess response in a real-world cohort.
Real-world data is important for understanding the treatment course and response of multiple myeloma (MM) patients. However, in clinical practice, patients’ treatment responses noted in medical records are often not standardized and pose the problem of missing data.
A pilot study in PLOS One proposed and evaluated a real-world derived response (dR) algorithm, adapted from International Myeloma Working Group response criteria, able to reliably assess the response of MM patients from incomplete data using less stringent criteria.
dR Algorithm Response Assessment
Data from the 291 patients in the BELLINI trial, a randomized 2:1, double-blind, multicentre, phase III trial that evaluated venetoclax or placebo in combination with bortezomib and dexamethasone in patients with relapsing–remitting MM, were included in the study. The dR algorithm response assessment was in concordance with that of the independent review committee (IRC, gold standard), with 275 out of 291 (Cohen’s Kappa 0.83) assignments of responders classified as confirmed partial response or better (≥PR) being in agreement.
There were 16 cases in which the dR algorithm and IRC assigned different responses. When depths of response were evaluated separately (i.e., partial response (PR), very good PR (VGPR), complete response (CR), and stringent CR) instead of only ≥PR, concordance was lower between the IRC and dR algorithm’s response assignments (Cohen’s Kappa 0.56) due to the omission of bone marrow data.
Overall Response Rate, VGPR, and CR
The odds ratio between the overall response rate (ORR) (≥PR) of the intervention and placebo arms using IRC response assessment in the BELLINI trial was 2.10, suggesting treatment benefit. The odds ratio between the ORR of intervention and placebo arms using the algorithm assessment was 2.31. The assignment of ≥VGPR by the algorithm was in agreement with the IRC. Characterization of ≥CR resulted in lower (but directionally in agreement) treatment efficacy estimates.
Despite removing 50% of laboratory measurements and 75% of urine monoclonal protein measurements, the algorithm accurately replicated the ORR between the intervention and placebo arms. However, the algorithm’s estimation of treatment response for specific categories (≥VGPR and ≥CR) was lower than the IRC assessment. Additionally, dR assessment of a real-world dataset demonstrated a significant association between dR and overall survival at both an individual level and at various treatment landmarks.
Xu, T., Roose, J., Williamson, M., Sawas, A., Hong, W., Jin, H., Maignan, K., Rocci, A., Yousefi, K., Kumar, S., & Tyanova, S. (2023). RWD-derived response in multiple myeloma. PLOS ONE, 18(5), e0285125. https://doi.org/10.1371/journal.pone.0285125