Drug class refractoriness, not number of prior lines of therapy, properly classify patients with relapsed and refractory multiple myeloma

Clinical trial eligibility and regulatory approvals of new treatments often define populations of patients with relapsed, refractory multiple myeloma (RRMM) based on the number of prior lines of therapy (LOTs)1-8 with the premise that the number of prior LOTs properly identifies patients with a similar prognosis undergoing existing treatments. This approach, also seen in other haematologic malignancies and solid tumours,9 is supported by scarce evidence, and is not validated in the current therapeutic landscape.5 Several recent analyses suggest the refractoriness to multiple classes of agent better identifies patients with poor response to therapy, as well as progression-free (PFS) and overall survival (OS).10-13 We hypothesise that drug class refractoriness, rather than the number of prior LOTs, better classifies patients with RRMM undergoing modern therapy. We retrospectively analysed patients starting a new, non-experimental regimen for RRMM at a single treatment centre between 2016 and 2022. For each regimen initiated (index regimen) we extracted patient and disease characteristics, number of prior LOTs and refractoriness to each three main classes of anti-myeloma agents: immunomodulatory drugs, proteasome inhibitors, and anti-CD38 monoclonal antibodies. We grouped prior LOTs in 1, 2/3 or 4+ and refractoriness in no/single-class refractory (N/SCR), double-class refractory (DCR) and triple-class refractory (TCR). For each regimen we captured the treatment response according to the International Myeloma Working Group criteria and progression-free survival (PFS). We evaluated the impact of prior LOT and class refractoriness on the overall response rate (ORR) and PFS by univariate and multivariable analysis. Since many of these patients received experimental, transformative new therapies (e.g. CAR-T cell therapy) in subsequent lines, we focused the analysis on PFS and did not evaluate OS. Therefore, each index regimen is considered individually even if applied to the same patient in subsequent LOTs. We performed a multivariate analysis for the ORR, utilising generalised linear mixed models with a binomial logit link accounting for multiple observations per patient. For multivariate analysis for PFS, we utilised the Andersen–Gill recurrent event model, with a robust variance estimator accounting for multiple PFS events per patient.14 All patients had provided written informed consent to have data collected in a database. We included 72 patients receiving 183 distinct regimens for RRMM (range 1–6 index regimens per patient). Median age at onset of index regimen was 57 years. Median number of prior LOTs was two (range 1–9), median time between MM diagnosis and initiation of the index regimen was 39.0 months for patients with N/SCR, 45.9 months for DCR, and 52.1 months for TCR MM. Patient, disease and treatment characteristics are outlined in Table 1. Most patients had DCR or TCR MM after two prior LOTs (Figure S1). An anti-CD38 monoclonal antibody was part 51% of regimens employed for N/SCR, 53% for DCR and 35% for TCR MM. Carfilzomib was utilised in 32%, 31% and 45%; pomalidomide in 32%, 40% and 36%; and alkylating agents in 22%, 20% and 32% of regimens for N/SCR, DCR and TCR MM, respectively. The overall response rate was 65%, 56% and 36% for patients with 1, 2/3 or 4+ prior LOT, and 74%, 53% and 23% for N/SCR, DCR and TCR MM, respectively. In a multivariate analysis, drug-class refractoriness was strongly associated with ORR (p < 0.001). When compared to regimens administered for N/SCR MM, the odds ratio (OR) for DCR was 0.23 (95% CI 0.08–0.63) and OR for TCR was 0.05 (95% CI 0.02–0.19). The number of prior LOT was not associated with ORR (p = 0.30) (Figure 1A and Table S1). Median PFS was 16.0 months for regimens administered after one LOT, 11.0 months for 2/3 LOT and 6.1 months for 4+ LOT (Figure S2). Regarding class refractoriness, median PFS was 17.8 months for regimens administered for N/SCR, 7.2 months for DCR and 3.0 months for TCR RRMM (Figure S2). In a multivariate analysis, the number of prior LOT was not associated with PFS (p = 0.46), while class refractoriness strongly predicted PFS (p < 0.001). Using regimens administered for N/SCR MM as a reference, the hazard ratio (HR) for DCR MM regimens was 1.62 (95% CI 0.94–2.79) and HR for TCR MM regimens was 3.77 (95% CI 1.95–7.30) (Figure 1B, Table S1). Our analysis has the limitation of a small sample size with data collected at a single centre. LOCOMMOTION,12 a recent prospective observational study, demonstrated poor ORR (25.1%) and PFS (median 3.9 months) in patients with triple-class exposed MM treated with conventional agents. These findings were similar to the retrospective MAMMOTH10, 13 study in the US. Compared to the present study, LOCOMMOTION and MAMMOTH included patients who had received more LOTs (median 4) before MM became refractory to anti-CD38 monoclonal antibodies, reflecting the introduction of these agents in later LOTs. The present study better reflects the current practice in the US with most patients receiving anti-CD38 monoclonal antibodies in early LOTs. Unlike most of the clinical trial-based literature, this analysis includes a large proportion of non-Hispanic Black and older patients, reflective of the real-world population of patients with RRMM. We observed that among patients receiving modern therapy for RRMM, it was class-refractoriness and not the number of prior LOTs which predicted ORR and PFS. We believe this observation has profound implications for clinical trial design and regulatory approvals. Clinical trials designed to test the role of a certain therapy in a specific disease aimed at including a relatively homogenous population with a similar treatment history, available options and expected outcomes. Defining eligibility by LOTs rather than class refractoriness defies such principles. Given the constantly changing therapeutic landscape, a therapy that demonstrates clinical benefit in clinical trials and is approved for use after, for example, four prior LOTs will end up being used in a population with far greater refractoriness by the time the fourth LOT fails. This structure leads to patients finding themselves in a ‘therapeutic vacuum’, where their disease has become refractory to all three main classes of agents after one or two LOTs, but they cannot access the next line until the fourth LOT fails, and are forced to use agents and regimens that are toxic and have low activity, often succumbing to the disease.15 We believe the present data indicates that clinical trial design and regulatory approvals should define RRMM populations based on drug-class refractoriness while disregarding the number of prior LOTs. LJC contributed essential patient information, designed the research study, performed the research, analysed the data, wrote the paper, approved the submitted and final version of the manuscript. SG contributed essential patient information, performed the research, analysed the data, reviewed the manuscript critically, approved the submitted and final version of the manuscript. GR contributed essential patient information, reviewed the manuscript critically, and approved the submitted and final version of the manuscript. SB contributed essential patient information, reviewed the manuscript critically, and approved the submitted and final version of the manuscript. KNG contributed essential patient information, reviewed the manuscript critically, and approved the submitted and final version of the manuscript. LC: Honoraria (Amgen, Celgene-Bristol Myers Squibb, AbbVie, Adaptive Biotechnologies, Janssen, Sanofi, Takeda), Research Funding (Amgen, Celgene-Bristol Myers Squibb, Janssen); The remaining authors did not have any relevant conflict of interest to disclose. Figure S1. Figure S2. Table S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.