Introduction: Despite improvement in treatments, multiple myeloma (MM) remains an incurable malignancy. There is currently interest within the MM community to explore the validity of using predictors of overall survival (OS) such as minimal residual disease (MRD). Best clinical response (BCR) to treatment is a well-defined and standardised ordinal outcome scale captured in MM clinical trials and quality care registries. The aim of this study was to determine whether BCR to treatment could be used to predict OS when modelling real-world MM disease and treatment pathways.
Methods: Risk equations controlling for age, sex, and disease severity were estimated using the Australia and New Zealand Myeloma & Related Diseases Registry data on approximately 3,600 MM patients diagnosed between 2011 and 2022. Multinominal logit regression including patient’s complete treatment history was used to assign chemotherapy regimens to patients. Parametric survival analysis was used to estimate OS, chemotherapy duration, and treatment-free intervals for up to six lines of therapy. Ordered logit regression was used to predict BCR to treatment and sequential logit regressions were used to predict receipt of stem cell transplant and/or maintenance therapy after induction therapy. These were incorporated into a discrete-event individual simulation model which was then used to predict OS based on BCR to treatment.
Results: The results show that the extremes of BCR are strong predictors of OS in MM, with patients achieving Complete Remission or Very Good Partial Response having the longest OS and patients with Progressive Disease the shortest OS. However, the discriminatory power of Partial Response versus Minimal Response versus Stable Disease does not appear to be as strong as the other categories. As expected, predicted OS to all BCR levels decreases with each additional line of therapy.
Conclusions: Our modelling study suggests that BCR to treatment is able to strongly predict OS in real-world MM clinical practice. Modelling MM in this way will provide clinicians, patients and funders with accurate evidence of the impacts of treatment and policy decisions. Importantly, such modelling could be explored further to facilitate more rapid determinations in regulatory authority drug approval processes so enabling patients earlier access to more effective therapies.
