Survival and treatment patterns of patients with relapsed or refractory multiple myeloma in France — a cohort study using the French National Healthcare database (SNDS)

Over the past decade, several drugs have been approved for the treatment of relapsed or refractory multiple myeloma (RRMM). This retrospective study, using the French National Healthcare database (SNDS), describes the treatment patterns and outcomes of patients with RRMM treated in real-world clinical practice in France. Patients were adults, with a diagnosis of multiple myeloma, who initiated second-line (2L) treatment approved for use in France between 2014 and 2018; this included bortezomib, carfilzomib, daratumumab, ixazomib, lenalidomide, or pomalidomide. Data were analyzed overall, by first-line (1L) autologous stem cell transplant (ASCT) status and by lenalidomide treatment status at 2L. In total, 12987 patients with RRMM were included in the study (mean age 69.5 years); 27% received an ASCT at 1L, and 30% received a lenalidomide-sparing regimen at 2L. Overall, and among the ASCT and non-ASCT subgroups, most patients received a bortezomib-based regimen at 1L, whereas lenalidomide-based regimens were most common at 2L. Among patients who received a lenalidomide-sparing regimen at 2L, this was most often a proteasome inhibitor-based regimen. Mortality rate was 26.1/100 person-years, and median (95% confidence interval) survival from 2L initiation was 32.4 (31.2–33.6) months. Survival differed by various factors, shorter survival was reported in the non-ASCT group, those receiving a lenalidomide-sparing regimen at 2L, older patients (≥ 70 years), and those with multiple comorbidities. This analysis provides insight into the real-world use of approved novel MM treatments and highlights an ongoing unmet need to improve outcomes, particularly for selected patient groups. Supplementary Information The online version contains supplementary material available at 10.1007/s00277-021-04522-y.


Introduction
Multiple myeloma (MM), characterized by clonal plasma cell proliferation in the bone marrow, is the second most common hematologic malignancy [1]. In 2018, the MM agestandardized incidence per 100000 was estimated at 4.3 in Europe [2], and the number of new cases in France was estimated to be 5500 [3].
Given the rapidly evolving treatment landscape, the aim of this study was to use the latest available data (2014 to 2018) from the "Système National des Données de Santé" (SNDS, for French National Healthcare database) to describe the RRMM treatment patterns in France, including recently approved novel therapies. Survival outcomes were also assessed as an exploratory analysis.

Data source
This was a retrospective observational cohort study that used data obtained from the SNDS database, which includes claims data for more than 65 million individuals (including 50 million adults), from birth (or immigration) to death, and is highly representative of the French population, covering 99% of the total population [26]. The SNDS database contains several datasets linked together via a unique patient ID (social security number). These datasets include the "Système National d'Information Inter-Régime de l'Assurance Maladie" (SNIIRAM, national information system for health insurance dataset), which contains demographic and administrative patient data (e.g., age, sex, and place of residence); healthcare visits and procedures reimbursed (e.g., medicines, medical procedures, medical devices, lab tests); and date of death. Data from hospitals and other healthcare facilities are extracted from the "Programme de Médicalisation des Systèmes d'Information" (PMSI, French National Hospital Informatics database), which includes inpatient data such as medical information, related diagnosis (based on International Classification of Diseases, 10th Revision, Clinical Modification [ICD-10-CM] codes), medical procedures, imaging, external visits, external procedures performed, expensive medicines, and implantable devices. Causes of death are also extracted for both inpatients and outpatients from the Epidemiological Center for the Medical Causes of Death database (CépiDc).

Study design
Patients were eligible for inclusion if they were adults (≥ 18 years) with a diagnosis of MM and had received at least one dose of an approved novel MM treatment of interest (bortezomib, carfilzomib, daratumumab, ixazomib, lenalidomide or pomalidomide) between 2014 and 2018 (Supplementary material 1). Patients with evidence of second-line (2L) treatment (defined as at least one previous line of therapy, either unspecified or including any drug of interest) were considered to have RRMM (Supplementary material 1). Data were extracted for a 10-year period (01/01/ 2009-12/31/2018). The index date was defined as the date of initiation of the novel MM treatment of interest as a 2L treatment for RRMM. A baseline period was defined as the time period preceding the index date, up to study start or date of first diagnosis with MM (if MM was diagnosed after 01/01/ 2009). Patients were followed up from the index date until the last available information in the datasets, death or end of study (12/31/2018), whichever was first.

Baseline characteristics
Demographics, clinical characteristics, and history of autologous stem cell transplant (ASCT) were assessed during the baseline period. Codes for identifying specific procedures and treatments are presented in Supplementary material 2. The Charlson comorbidity index (CCI) was retrospectively estimated according to an algorithm adapted from Bannay [27]. Frailty scores could not be estimated due to the unavailability of some key variables, such as laboratory-based biomarkers [28][29][30].

Exposures and outcomes
Definitions of treatment regimens and lines of therapy were based on the algorithm described by Palmaro et al. [31], using the French national healthcare insurance database (SNIIRAM). The treatment regimens were also in line with ESMO guidelines [4]. It was assumed that supporting therapies, such as dexamethasone and prednisone, were given in combination with each treatment even though they are not visible within the SNDS dataset. Treatments were assigned to mutually exclusive regimens (Supplementary material 3). Mortality was estimated as the number of patients who died by the end of follow-up and rate was reported per person-years of follow-up.

Statistical analysis
Results for treatment patterns and clinical characteristics were summarized by calculating the frequency and percentages for categorical variables and mean, standard deviation (SD), median, and interquartile ranges (IQR) for continuous variables. Overall survival was summarized using Kaplan-Meier methodology. The analyses were conducted overall, by ASCT status at first-line (1L) treatment and by lenalidomide status at 2L (defined as 2L regimens including lenalidomide, either doublets or triplets). Further analysis by lenalidomide status was conducted using the Cox regression model for all-cause mortality, adjusted by age at initiation of 2L treatment, time from MM diagnosis to initiation of 2L treatment, 1L treatment received, sex, ASCT status at 1L, and presence of comorbidities (hypertension, dementia, diabetes mellitus, moderate to severe renal disease or any tumor). All statistical analyses were conducted using SAS® Enterprise Guide version 7.15. Graphical representation was carried out using R version 3.6.2.

Patient disposition
Between 2014 and 2018, 12987 patients with RRMM were treated with at least one drug of interest, started a 2L treatment, and were included in the study cohort (Fig. 1).

Baseline characteristics
The mean (SD) age at index date was 69.5 (10.6) years; 54% of patients (n = 7028) were male.

Baseline characteristics by ASCT status at 1L
Twenty-seven percent of patients (n = 3454) received an ASCT at 1L (ASCT subgroup), and 73% of patients (n = 9533) did not (non-ASCT subgroup). Patients in the ASCT subgroup were younger than those in the non-ASCT subgroup (mean [SD] age 60.6 [7.6] years vs 72.7 [9.7] years) ( Table 1). The proportion of patients who had at least one comorbidity was higher in the non-ASCT subgroup than in the ASCT subgroup (47% vs 32%). Comorbidities commonly reported for patients with RRMM were more frequent in the non-ASCT subgroup than in the ASCT subgroup (hypertension, 48% vs 31%; diabetes mellitus, 14% vs 10%; moderate to severe renal disease, 12% vs 5%). Similarly, patients in the non-ASCT subgroup had a higher CCI than those in the ASCT subgroup (mean [SD] CCI: 1.6 [2.5] vs 0.9 [2.0]).

Baseline characteristics by lenalidomide status at 2L
Seventy percent of patients (n = 8940) received a lenalidomide-based regimen at 2L (lenalidomide subgroup), and 30% of patients (n = 3872) did not (lenalidomide-sparing subgroup). Patients treated with a lenalidomide-based regimen at 2L were slightly younger than those treated with lenalidomide-sparing regimens (mean [

Treatment patterns by ASCT status at 1L
In both the ASCT and non-ASCT subgroups, the majority of patients received bortezomib-based regimens at 1L (84 and 64% of patients, respectively). These were bortezomib-based doublet or triplet regimens that excluded lenalidomide: triplet regimens were most common in the ASCT subgroup (61% of patients), whereas doublet regimens were more common in the non-ASCT subgroup (58% of patients; Fig. 2). The most frequently used 2L treatment was lenalidomide-based doublet regimens in both ASCT and non-ASCT subgroups (55 and 53% of patients, respectively). The second most common 2L regimen was a combination of bortezomib and lenalidomide for the ASCT subgroup (22% of patients) and a PI-based doublet (in particular, bortezomib-based doublet) for the non-ASCT subgroup (25% of patients). At third-line (3L) treatment, PI-based doublets (predominately bortezomib-based doublets) were the most common regimens used in the ASCT subgroup (19% of patients), followed by pomalidomide (11% of patients). In the non-ASCT subgroup, pomalidomide was the most common 3L treatment (13% of patients), followed by PI-based doublets (predominately bortezomib-based doublets) (12% of patients; Fig. 2).

Treatment patterns by lenalidomide status at 2L
At 1L, most patients (93%) in the lenalidomide subgroup received a bortezomib-based regimen; 61% of patients in this subgroup received a bortezomib-based doublet at 1L (Fig. 4). At 2L, most patients in the lenalidomide subgroup received a lenalidomide-based doublet regimen (78% of patients). In the lenalidomide-sparing subgroup, most patients received either a bortezomib-based or a lenalidomide-based regimen at 1L  (Fig. 4). The most common 2L treatment in the lenalidomidesparing subgroup was a PI-based doublet regimen (predominately bortezomib-based doublets), which was given to 67% of patients.  Fig. 5). Patients in the ASCT subgroup were younger and had fewer comorbidities than those in the non-ASCT subgroup. When considering mortality rate by age and/ or CCI, it remained higher for the non-ASCT subgroup and in particular among patients with a high CCI (> 1) ( Table 3).

Discussion
This study describes a diverse treatment landscape for RRMM in a nationwide cohort of 12987 patients treated with approved novel therapies between 2014 and 2018 in France. A few treatments, however, emerged as the preferred choices at each treatment line. At 1L, most patients were treated with bortezomib, whereas lenalidomide was the most commonly used agent at 2L. PI-based doublet regimens, predominately bortezomib-based doublets, and pomalidomide were the most common 3L treatments. In this cohort study, 40% of the patients died during the follow-up period; mortality rate was 26.1/100 person-years, with a median survival time from 2L initiation of 32.4 months. Compared with the overall cohort, survival was shorter for patients who did not receive an ASCT at 1L, those receiving a lenalidomide-sparing regimen at 2L, older patients (≥ 70 years) and those with multiple comorbidities.
Although there continues to be no standard of care for RRMM in Europe [32], recent studies show that lenalidomide is often the most common treatment, and PI-based regimens, including triplets, are increasingly being prescribed [21,33]. Practice patterns in France differ from the rest of Europe, and only a few studies have reported on recent treatment patterns in RRMM that are specific to the French population [19,20]. These studies align with the findings of the current study in terms of the most common 2L treatments; however, treatment patterns appear to diverge at 3L treatment. [19]Treatment choice in the RRMM landscape can be influenced by many  [1]. In our study, treatment regimens were varied, particularly at 3L and beyond, potentially reflecting the wide variability in patient characteristics and prior response. Transplantation at 1L also appears to influence subsequent treatment choices: triplet regimens were more frequently prescribed to patients who received an ASCT at 1L than to those who did not. Studies have shown that younger patients, with fewer comorbidities, are likely to tolerate more lines of treatment (including ASCT) and more aggressive regimens [1,22,34,35]. Access to and reimbursement of treatments also influences treatment patterns [1]. In our study, some recent therapies were not available in France except in "Temporary Authorization for Use" (ATU). Carfilzomib became fully available in July 2018 (previously in ATU and compassionate use thereafter since 2016) and ixazomib in October 2018. During the study, daratumumab and elotuzumab were not reimbursed, while pomalidomide was. Despite several agents not being reimbursed, regimens including these more recently available drugs have been captured in our study, particularly at 3L and beyond. Our findings suggest a potential gap in treatment for patients in France, given that recent clinical trials have shown that agents such as carfilzomib, daratumumab, elotuzumab, and ixazomib can result in substantially improved progression-free survival in RRMM [7-12, 15, 16]. Furthermore, there is evidence that use of these drugs in earlier treatment lines can produce a greater depth of response and further improve outcomes [36].
Our study evaluated treatments for RRMM in a real-world setting, where patient populations are more heterogeneous than clinical trial populations; routine clinical care may differ from the more rigorous protocols followed in a clinical trial setting [37]; and patient outcomes have been observed to differ from those achieved in clinical trial settings for MM [25,38]. Patients in our cohort study were similar in age to clinical trial populations [7, 9-12, 15, 16], and 40% of patients died during the follow-up period, similar to other real-world French studies [20]. Patients in our study who received a lenalidomide-sparing regimen at 2L had a lower median survival time than those receiving a lenalidomide-based regimen; adjusted analyses confirmed this relationship between use of lenalidomide-sparing regimens at 2L and increased risk of death. Among patients treated with lenalidomide-sparing regimens at 2L, half received lenalidomide-based regimens at 1L and may have had lenalidomide-refractory RRMM and a poor prognosis. These findings highlight the need for new effective therapeutic strategies for patients who cannot receive lenalidomide-based regimens at 2L [39].
Limitations of this study include those inherent to the use of retrospective administrative claims data and limitations associated with missing data. However, we had access to multiple linked data sources (primary care, hospital, pharmacy data, and central death registrations), which greatly improves exposure and outcome ascertainment. Furthermore, the SNDS includes data on 99% of the French population, enabling the analysis of a large and representative cohort of patients with RRMM. For patients receiving a newly approved treatment available from 2018 onwards, sample sizes were low and exposure to such treatments and follow-up were limited [39]. The current study also did not capture clinical trial participation and information on the line of therapy for such patients may not be precise. This real-world analysis illustrates the dynamic MM treatment paradigm and provides useful information for payers making decisions about reimbursement options and providers evaluating choice of treatment regimens in order to optimize the management of RRMM. Survival data from this study suggest that there remains a need to improve real-world outcomes, particularly among selected patient populations.

Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1007/s00277-021-04522-y. and the data privacy laws, for their involvement during the data application process. The authors acknowledge the help of Luyuan Qi, who was involved in data management and analysis, as well as the help of Karina Berenson, who assisted with the development of the first draft of the manuscript. The authors acknowledge Kelly Soady, PhD from PharmaGenesis London, London, UK, who provided editorial assistance (funded by Amgen [Europe] GmbH) during the preparation of the manuscript.
Author's contribution GD, NQ, and AK designed the study. CT, NQ, AK, GD, JM, HJ, VT, JVC, and MS drafted the article or revised it critically for important intellectual content. JM, HJ, CT, NQ, GD, AK, VT, and MS were involved in data analysis and interpretation. All authors made a substantial contribution to analysis or interpretation of data and approved the version to be published. All authors read and approved the final manuscript.
Funding This study was funded by Amgen Europe GmbH.
Availability of data and materials Data were shared by the French national health insurance: "Caisse Nationale d'Assurance Maladie" (CNAM).
Code availability All statistical analyses were conducted using SAS® Enterprise Guide version 7.15. Graphical representation was carried out using R version 3.6.2. Specific code may be available upon request.

Declarations
Ethics approval This study has been approved by the French 'Commission Nationale Informatique et Libertés' (CNIL) governing the data access and the data privacy laws.
Consent to participate Not applicable. As this study involved anonymized structured data, which according to applicable legal requirements do not contain data subject to privacy laws, obtaining informed consent from patients was not required.

Consent for publication Not applicable
Conflict of interest CT received consulting fees from Amgen GmbH; NQ, JM, HJ, and AK are full-time employees of Certara and received consulting fees from Amgen GmbH to conduct the study. JVC and GD are full-time employees at Amgen France SAS and own stock. MS and VT are full-time employees at Amgen Ltd.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.