Real-world comparison of progression-free survival and time to next therapy in 761 patients with newly diagnosed multiple myeloma

TO THE EDITOR:

Progression-free survival (PFS) is the widely used efficacy end point in multiple myeloma (MM) trials and often required for regulatory approval.¹ It captures time from therapy start to disease progression or death, per International Myeloma Working Group (IMWG) criteria.² Outside trials, however, complete laboratory, imaging, or marrow assessments needed to operationalize IMWG progression are inconsistently available in real-world data sets.³ This inconsistency complicates the accurate measurement of PFS in clinical practice. To address this limitation, time to next treatment (TTNT) has emerged as a pragmatic alternative to PFS.⁴ TTNT is easier to capture using routinely available health data sets. However, TTNT can be influenced by nonbiologic factors, including treatment toxicity, patient/clinician preference, and drug access, raising questions about its correlation with PFS.⁵ Evidence comparing PFS and TTNT in newly diagnosed MM (NDMM), particularly at the individual patient level outside clinical trials, remains limited.^5,6 Understanding the degree to which TTNT mirrors PFS is essential for interpreting real-world outcomes and comparing them with clinical trial–based benchmarks. We therefore conducted a retrospective cohort study to evaluate the correlation between TTNT and PFS in patients with NDMM in the real world.

The study was conducted at 2 Canadian (University of Ottawa and McMaster University) and 1 US center (Mayo Clinic, AZ). Consecutive adults with NDMM initiating first-line therapy with at least 1 novel agent, proteasome inhibitor and/or immunomodulatory drug, between January 2015 and January 2024 were included. Patients treated on clinical trials were excluded. Clinical variables were abstracted manually from health records using a standardized template. PFS was defined as the time from first-line therapy initiation to IMWG-defined progression or death. TTNT was defined as the time from first-line therapy start to second-line start or death. We estimated PFS and TTNT by the Kaplan-Meier method, compared distributions via log-rank tests overall and in clinical subgroups, constructed 2 × 2 cross-tabulations of PFS/TTNT event status, summarized the within-patient lag (days) between PFS and TTNT among those with both events, and computed the Pearson correlation coefficient between times.

A total of 761 patients with NDMM were included. A total of 412 (54.1%) patients underwent autologous stem cell transplant (ASCT). Among ASCT recipients, induction was predominantly CyBorD (cyclophosphamide, bortezomib, and dexamethasone; n = 361 [87.6%]), followed by VRd (Velcade [bortezomib], Revlimid [lenalidomide], and dexamethasone; n = 44 [10.7%]), and D-VRd (daratumumab-VRd; n = 7 [1.7%]). Among non-ASCT (n = 349), first-line regimens were Rd (n = 131 [37.5%]), CyBorD (n = 115 [33.0%]), VRd (n = 67 [19.2%]), DRd (n = 27 [7.7%]), and KRd (carfilzomib-Rd; n = 9 [2.6%]). Full baseline details are provided in supplemental Table 1. The median follow-up of the cohort was 5.1 years.

The overall response rate in the cohort was 92% (complete response or better, 22.1%; very good partial response, 50.2%; partial response, 20.3%). Kaplan-Meier curves for PFS and TTNT are shown in Figure 1. Overall, 61.5% experienced a PFS event and 64.3% a TTNT event. Among those with progression (n = 468), 14 (3.0%) had not initiated a subsequent therapy by last follow-up. Conversely, among those without a PFS event (n = 293), 35 (12.0%) initiated second-line therapy. In patients with both events (n = 454), PFS typically preceded TTNT with a median lag of 26 days (range, −876 to 3006), and the times were strongly correlated (Pearson r = 0.82). Patient-level concordance is illustrated in supplemental Figure 1. Large gaps were uncommon: 2.6% had PFS of >1 year before TTNT (0.4% of >2 years), whereas 4.4% started the next therapy >1 year before documented progression (3.1%, >2 years). The distribution of TTNT-PFS lags and the frequency of large gaps are shown in supplemental Table 2. Given this, the overall survival distributions of PFS and TTNT were not significantly different (log-rank P = .24). Median PFS of 3.1 years (95% confidence interval, 2.9-3.4) and TTNT of 3.0 years (95% confidence interval, 2.8-3.2) were closely aligned. One-, 2-, and 5-year estimates were 84.3%, 65.8%, and 33.3% for PFS, and 82.0%, 63.4%, and 30.4% for TTNT, respectively. Patients with a larger discrepancy included 21 individuals who had a PFS of >3 years longer than their TTNT, due to the following reasons: side effects (n = 9), suboptimal response (n = 8), progressive disease not meeting IMWG criteria (n = 3), and unknown (n = 1). Conversely, 15 patients had a TTNT of >1 year longer than their PFS, for the following reasons: new bone lesion (n = 2), slow biochemical progression over time (n = 3), patient declined treatment (n = 1), and unknown (n = 9). Patient status at the last follow-up is summarized in supplemental Figure 2.

Figure 1.

Kaplan-Meier curves for PFS and TTNT in the overall cohort.

Specific subgroups were also analyzed for PFS and TTNT distribution (Figure 2). As shown, PFS and TTNT distributions did not differ (P > .05) by specific subgroups, including site, transplant eligibility, cytogenetic risk, depth of response, regimen used, or maintenance strategy. Median values and within-group log-rank P values were similar to the overall cohort, suggesting that the observed concordance is not restricted to a particular practice environment or disease-risk criteria.

Figure 2.

Median PFS and TTNT across clinical subgroups. CI, confidence interval; CR, complete response; PR, partial response; Tx, treatment; VGPR, very good partial response.

In this multicenter, real-world NDMM cohort, TTNT aligned closely with PFS both at the population and patient levels. These findings support pragmatic use of TTNT as a real-world comparison with PFS in MM when rigorous IMWG progression ascertainment is infeasible.

Previous research exploring the relationship between PFS and TTNT has been limited and often conflicting. In solid organ malignancies, TTNT has been shown to be potentially shorter than the PFS.⁷ Although the aforementioned study is often quoted in hematological malignancies literature, it is important to note that the cohort comprised patients with breast cancer, for which different biology and treatment modalities can affect TTNT and PFS. In a different study with more rigorous data collection, TTNT and PFS were, in fact, similar in metastatic solid malignancies.⁸

In MM specifically, a study conducted by the Swedish group between 2000 and 2011 highlighted that PFS (median PFS, 8.5 months) was shorter than TTNT (median TTNT, 12.3 months), particularly during first-line treatment.⁹ Although this was a large study, it was conducted among patients receiving more finite treatment with minimal usage of novel agents. In more recent years, a retrospective chart review of 207 transplant-ineligible patients treated in the United States between 2012 and 2017 was reported. In the aforementioned study, the median TTNT (10.4 months) was slightly shorter than the median PFS (12.3 months), but it is important to note that this was limited to only transplant-ineligible patients and not all patients were treated with novel agents.¹⁰ Additional database studies like Flatiron have tried to report PFS using a combination of structured and unstructured data variables¹¹; but without rigorous patient-level chart review, it is difficult to evaluate whether PFS is accurately being measured using these algorithms. Additionally, many of these studies use a variation of IMWG PFS (termed real-world progression-free survival [RW-PFS]) defining PFS differently from the IMWG as any event of disease progression, death, or any change in antimyeloma treatment, whichever occurred first.¹²

Our study significantly complements these earlier analyses by providing extensive individual-patient-level evidence from a multicenter, rigorously collected retrospective chart review of a real-world cohort of 761 patients with NDMM. Our work provides evidence that TTNT aligns closely with PFS at both the population and patient levels, demonstrating highly similar survival distributions (median PFS, 3.1 years; and TTNT, 3.0 years). This concordance was observed across various clinical subgroups, suggesting that the consistency is widely applicable within the NDMM population and not restricted to a particular practice environment or disease-risk criteria. From a regulatory perspective, agencies such as the US Food and Drug Administration and the European Medicines Agency have acknowledged the potential utility of real-world end points in complementing trial data, particularly for postauthorization safety and effectiveness evaluations.^13,14 Because MM treatments continue to evolve rapidly, TTNT may serve as a valuable end point for capturing longitudinal treatment patterns, comparative effectiveness, and real-world disease burden, facilitating more inclusive and representative evidence generation for regulatory and health technology assessment purposes.

Limitations of our study include limited sites for this chart review as well as the evaluation of PFS and TTNT limited to first-line settings. Although we observed consistent PFS-TTNT alignment across the NDMM setting, generalizability, particularly to the relapsed settings, in which nonbiologic triggers for switching therapy may be more diverse, is unknown. Broader linkage with patient-reported outcomes could further clarify circumstances in which TTNT diverges from progression for clinically meaningful reasons. Lastly, with more novel therapies and more sensitive disease assessment modalities such as minimal residual disease testing, it is possible that in the future TTNT and PFS may be discrepant. Future studies should focus on directly comparing PFS and TTNT within these large clinical trial data sets as a study outcome in both newly diagnosed and relapsed settings. Such analyses would provide valuable insights into the concordance between these end points in different settings and with different treatment modalities and monitoring approaches.

In conclusion, our data support that among patients with NDMM treated outside trials, TTNT mirrors PFS closely and therefore can serve as a pragmatic end point for real-world studies in frontline NDMM when robust IMWG ascertainment is not available or possible.

This retrospective chart-review study was approved by the research ethics/institutional review boards at McMaster University (Hamilton, ON, Canada), the University of Ottawa (Ottawa, ON, Canada), and the Mayo Clinic institutional review board (Phoenix, AZ). A waiver of informed consent was granted at all sites, and the study was conducted in accordance with the Declaration of Helsinki.

Conflict-of-interest disclosure: H.M. is supported by an early career award from Hamilton Health Sciences and serves as the ArcelorMittal Dofasco Cancer Therapeutics research chair. The remaining authors declare no competing financial interests.