Introduction

Multiple myeloma (MM) is a bone marrow cancer characterised by uncontrolled growth of monoclonal plasma cells that secrete non-functional immunoglobulin. Accumulation of monoclonal immunoglobulins and interaction of the malignant plasma cells with other cells within the bone marrow lead to high symptom burden and complications including hypercalcaemia, renal damage, anaemia, and bone destruction1,2,3. Patients with MM will experience periods of remission followed by relapse, and their disease will eventually become refractory to treatment over time1,4.

The current treatment landscape for MM is complex and characterised by multiple lines of therapy (LoTs), which most commonly include immunomodulatory drugs, proteasome inhibitors, monoclonal antibodies and corticosteroids5,6,7. Autologous stem cell transplantation (ASCT) is associated with improved duration of remission in patients with MM, but not all patients are eligible or fit to receive ASCT; suitability is assessed based on factors such as performance status and comorbidities2,7,8.

Systemic anti-cancer therapy (SACT) is an important part of disease management in MM. It is defined as treatment with one or more systemic anti-cancer agents, which can be administered alone, in combination, or in sequence6,9. Collection of SACT data by the National Health Service (NHS) began in England in 2012 and relates to all patients with cancer (both adult and paediatric) treated in acute inpatient, day-case, and outpatient settings or in the community. It covers treatment for patients with solid and haematological malignancies, including those in clinical trials10. SACT data are collected via the National Cancer Registration and Analysis System (NCRAS) Cancer Analysis System (CAS) database, a routinely reported resource of SACT activity in the patient population managed by NHS trusts in England10,11. Therefore, SACT data make a valuable contribution to understanding the evidence base for real-world treatment patterns and outcomes in patients with cancer. These data are also increasingly being used to support health technology assessment submissions and decisions10,12,13.

In England, the SACT combination of bortezomib with thalidomide and dexamethasone was approved by National Institute for Health and Care Excellence (NICE) in April 2014 for patients eligible for subsequent ASCT14. Other options for these patients include bortezomib or thalidomide in combination with cyclophosphamide and dexamethasone14. In March 2021, lenalidomide was approved for maintenance therapy after ASCT15. For patients ineligible for ASCT, NICE recommended thalidomide or bortezomib with an alkylating agent and corticosteroid16, and lenalidomide in combination with dexamethasone17. The introduction of anti-CD38 monoclonal antibodies in the front-line setting has changed the treatment landscape4,6, with NICE approvals of the combination of daratumumab, bortezomib, thalidomide, and dexamethasone (Dara-VTd) in February 2022 and the combination of daratumumab, lenalidomide, and dexamethasone (Dara-Rd) in October 2023 for transplant-eligible and ineligible patients, respectively18,19.

The complexity and rapid evolution of the MM treatment pathway drives a continuing need to capture real-world treatment data to: (1) quantify the effectiveness of long-term disease management in MM throughout the treatment pathway, and (2) assess the impact of treatment-management strategies on patient outcomes in clinical practice5.

The aim of the present study was to describe demographic and clinical characteristics of patients with MM at diagnosis and at the initiation of each LoT using real-world evidence from the CAS SACT database in England. The study also examined the longitudinal sequencing and duration of SACT treatments, as well as patient survival outcomes. We highlight the limitations and learnings of using the CAS database and suggest areas for improvement, given the growing importance of real-world evidence in supporting reimbursement decisions in the UK20.

Methods

Study design and data source

This was a descriptive, non-interventional, retrospective population-based cohort study (Supplementary Information Sect. 1: Fig. S1) covering data from 1 January 2014 to 31 August 2021. Data were extracted from the CAS linked to the Hospital Episode Statistics (HES) database. The CAS database incorporates the SACT and the Cancer Outcomes and Services Dataset11,21. Approval from institutional review board and informed consent were deemed unnecessary according to national legislation due to the retrospective nature of this study and because it did not involve individually identifiable patient data. All methods were performed in accordance with the relevant guidelines and regulations.

As LoTs are not directly recorded or defined in the CAS database, recorded SACT agents and regimens were grouped into LoTs according to an algorithm developed iteratively in collaboration with expert clinical input, to minimise LoT misclassification (Supplementary Information Sect. 2).

In England, some therapies are funded through the Cancer Drugs Fund (CDF), which allows patients early access to promising new treatments22. Due to restrictions on reporting of treatment data for drugs funded through the CDF at the time of the study, patients receiving or who had received a drug that was on the CDF treatment list at the time of data extraction (1 January 2014 to 31 March 2021) were excluded from all treatment- and outcome-related analyses. Patients who received drugs that were previously on the CDF list were included. This was in agreement with NHS Digital requirements22 in place at the time this study was conducted (Supplementary Information Sect. 3: Table S3.1).

Study population

Adult patients with an incident diagnosis of MM between 1 January 2014 and 31 December 2019 were identified in the Cancer Outcomes and Services Dataset according to the International Classification of Diseases, 10th Revision (ICD-10) code C90.0. Patients were excluded if they had a record of SACT more than 30 days before the date of incident MM diagnosis or a record of MM diagnosis before 1 January 2014 (this was to exclude patients with prior MM diagnoses from the study population). To ensure SACT records were for MM, patients were excluded if they had a prior record of any other malignancy to ensure that treatments received were for MM and not any other malignancies (except for non-melanoma skin cancer [ICD-10 code C44.x]; these patients were not excluded as it is a relatively common malignancy for which non-SACT-based treatments are preferred)23.

The main study cohort consisted of all eligible patients with MM, except those receiving CDF treatment (used for descriptions of baseline characteristics only; Supplementary Information Sect. 3). From the overall cohort, a subgroup of patients who received at least one SACT were included in the SACT-treated sub-cohort. Only treated patients with MM were included in the outcomes analyses to ensure that these represented patients with active MM.

Follow-up

Patients were followed up from a relevant index date as specified below, to the earliest occurrence of death, last record of the patient being ‘alive’ in the Cancer Outcomes and Services Dataset, loss to follow-up, or end of data availability. The last availability date was 31 May 2021 for treatment data and 31 August 2021 for mortality data (Supplementary Information Sect. 1: Fig. S1).

Study variables and outcome measures

Baseline demographic and clinical characteristics were described for each study cohort. Variables evaluated at the time of diagnosis included: age, sex, year of diagnosis, ethnicity, Charlson Comorbidity Index (CCI), Eastern Cooperative Oncology Group (ECOG) score, and International Staging System (ISS) stage. Age, weight and ECOG score were also evaluated at treatment initiation for all patients from start of the 1 L to the fourth (4 L). In the SACT-treated sub-cohort, demographic and clinical characteristics were described at the start of 1 L and stratified by ASCT status (received/not received in any LoT during follow-up). Survival outcomes were described in the SACT-treated sub-cohort from the start of 1 L, and stratified first by ASCT status (received/not received in 1 L only) and second by most commonly occurring SACT regimens. Survival outcomes included time to next treatment or death (TTNTD; a common proxy for real-world progression-free survival) and overall survival (OS). As CAS does not capture longitudinal disease progression, TTNTD for each LoT was defined as the time (in months) from LoT initiation until the earliest date of the next LoT initiation or death. OS was defined as time (in months) from the start of any given LoT until the date of death.

The distribution of treatment regimens was described for the SACT-treated sub-cohort overall, stratified by LoT (1–4 L) and by ASCT status. The longitudinal sequencing of SACT regimens of interest (defined as health technology assessment/NICE-recommended regimens) at 1 L was created as part of the analysis of each LoT; this was described in the SACT-treated sub-cohort at 1 L and by ASCT status. The reported treatment regimens did not include steroids, as these were not consistently recorded in CAS.

Statistical analysis

All recorded SACT treatments were combined into LoTs according to the algorithm (see Supplementary Sect. 2 for details of how the algorithm was developed and implemented), and regimens were grouped according to agents identified in the first 28 days of LoT start. For LoT-specific stratification by common SACT regimens, the five most frequent grouped regimens were identified. Continuous variables were described by the mean, standard deviation, median, and first and third quartiles (Q1, Q3); categorical variables were described by the number and percentage of patients in each category. Sankey diagrams were used to depict longitudinal treatment sequences related to SACT treatments. Time-to-event analyses were performed using the Kaplan–Meier product-limit method. Censoring events for the TTNTD and OS analyses were based on the earliest of date of loss to follow-up or end of treatment data availability (31 May 2021). No imputation was performed for missing data. Analyses were performed using R version 4.2.1. The analyses were conducted on anonymised data by Health Data Insight in collaboration with the authors, and only aggregated data were shared with the authors.

Results

After application of the inclusion and exclusion criteria, the main cohort included 24,329 patients. Of these, 4,089 patients had a record of receiving a drug that was on the CDF list (CDF cohort) and were excluded from all treatment- and outcome-related analyses due to data reporting restrictions. Of the remaining 20,240 patients, 12,095 had a record of receiving at least one SACT. After the exclusion of 211 patients with at least one ASCT-only LoT, survival outcomes were determined for the remaining 11,884 patients from the SACT-treated sub-cohort (Fig. 1). From the overall cohort of 20,240 patients, those with a LoT that contained only an ASCT (n = 211) and those with an ASCT record but no SACT (n = 182) were included in the analysis of demographic and clinical characteristics but were excluded from further analyses, as these patients likely had missing SACT information or received ASCT treatment for a non-MM diagnosis (Fig. 1).

Fig. 1
figure 1

Study attrition. ASCT, autologous stem cell transplantation; CDF, Cancer Drugs Fund; LoT, line of therapy; MM, multiple myeloma; SACT, systemic anti-cancer therapy. *Demographic characteristics were reported at diagnosis in n = 4,089 patients who had received at least one drug on the CDF register (Supplementary Sect. 2). †Patients may have received SACT therapies, but these were either not recorded (due to poor capture in the database or due to treatment received outside the NHS) or were recorded outside the time window set out in the algorithm rules (and, therefore, could not be reliably linked to ASCT). Note: Of the overall cohort, 211 patients had a LoT which contained only ASCT, and 182 had an ASCT record but no SACT. This suggests that these patients had missing SACT information, or that ASCT was received as treatment for a non-MM diagnosis. These patients were therefore removed from all survival-related objectives.

Full size image

Patient demographics and baseline characteristics

Selected patient demographics and clinical characteristics for the overall cohort at diagnosis are shown in Table 1. More than half the patients were male (56.0%), with a median age of 73 years. Few comorbidities were recorded, and 50.1% of patients had a CCI score of 0. ECOG score and disease stage were poorly recorded, with more than half of the patients (58.6%) having a missing/unknown ECOG score and 70.2% of patients having a missing/unknown disease stage.

Table 1 Demographic and clinical characteristics of the overall cohort at diagnosis (all eligible patients excluding those taking CDF drugs).
Full size table

The demographic and clinical characteristics available at initiation of 1 L for patients in the SACT sub-cohort (overall and stratified by ASCT status) are shown in Table 224. In total, 3,419 patients (28.3%) received ASCT during the follow-up period, and weight and ECOG score were poorly recorded in patients receiving an ASCT.

Table 2 Demographic and clinical characteristics of patients in the SACT sub-cohort (overall and stratified by ASCT status).
Full size table

The demographic and clinical characteristics at diagnosis for patients in the CDF cohort and for all patients including those treated with drugs from the CDF (n = 24,329) are shown in the Supplementary Information Sect. 4: Table S4.1.1. Patients in the CDF cohort were younger, had a lower ECOG performance status, and fewer recorded comorbidities than the overall population.

SACT treatment distribution

The distribution of the most common SACT treatments overall and at 1–4 L is shown in Table 3. The five most common regimens of interest across all LoTs are highlighted in grey. Of the 12,095 patients in the SACT-treated sub-cohort, 9,101 (75.2%) received one of the five most common regimen groups at 1 L (highlighted in green).

The distribution of the most common treatments in the SACT-treated sub-cohort by ASCT status received at any LoT is shown in Table 4. Differences in the treatment regimens at 1 L were evident for patients who received an ASCT at any LoT, compared with those who did not. The greatest difference was seen in patients on the grouped regimen bortezomib plus thalidomide who received ASCT (42.1%), compared with those who did not (10.0%), as would be expected based on NICE recommendations. Some treatment regimens (Supplementary Information Sect. 4: Table 4.2.1) were received only by patients who did not receive any ASCT at any LoT during the follow-up period, including bortezomib plus melphalan, thalidomide plus melphalan, and daratumumab.

Table 3 Distribution of SACT treatments: most common combination regimens (overall and by LoT).
Full size table
Table 4 Most common SACT regimen groups at 1 L in the SACT-treated sub-cohort, overall and stratified by transplant status (ASCT received vs. not received at any LoT).
Full size table

SACT treatment sequencing

Longitudinal sequencing of SACT treatments was reported in patients receiving specific regimens of interest at 1 L and by ASCT status. Sankey diagrams depicting this can be seen in Fig. 2. In total, 8,062 patients (from the SACT-treated sub-cohort of 12,095) received a regimen of interest at 1 L, of whom 2,209 (27.4%) received an ASCT. Treatment sequencing differed according to ASCT status, with bortezomib plus thalidomide, bortezomib plus cyclophosphamide, or bortezomib monotherapy received by 99.3% of patients who received an ASCT compared with 83.5% of patients who did not.

Fig. 2
figure 2figure 2figure 2

Sankey diagrams for patients receiving regimens of interest by ASCT status (n = 12,095). LoT, line of therapy; ASCT, autologous stem cell transplantation. Sankey diagrams are presented to assess longitudinal treatment sequencing for patients receiving any of the regimens of interest at 1 L: (a) overall, (b) for patients who received ASCT in any LoT, and (c) for patients who did not receive ASCT in any LoT. Treatment sequencing for patients receiving regimens of interest at 1 L differed according to whether the patients did or did not receive ASCT during the follow-up period. Bortezomib plus thalidomide, bortezomib plus cyclophosphamide, or bortezomib monotherapy were received by 99.3% of patients who received ASCT in any LoT (i.e. 2,194 of 2,209 patients); however, these regimens were received by 83.5% of patients who did not receive ASCT (i.e. 4,885 of 5,853 patients). In addition, 100 patients (4.5%) who received ASCT in any LoT died after receiving 1 L, while 2,078 patients (35.5%) who did not receive ASCT in any LoT died after receiving 1 L.

Full size image

Time to next treatment or death

TTNTD was assessed from the initiation of each LoT (1–4 L) in the sub-cohort of SACT-treated patients excluding those with at least one ASCT-only LoT (n = 11,884). Overall, TTNTD decreased by an average of 31% at each LoT; the estimated median was 17.3 months (95% confidence interval [CI]: 16.7–17.9) from 1 L, 12.0 months (95% CI: 11.5–12.6) from 2 L, 8.4 months (95% CI: 7.9–9.4) from 3 L, and 5.6 months (95% CI: 4.8–6.1) from 4 L. By transplant status, median TTNTD from 1 L was 47.4 [43.4, 51.7] months in patients who received ASCT and 13.4 [12.9, 13.9] in those who did not (Supplementary Information Sect. 4: Fig. S4.3.1, Table S4.3.1). As with previous results, there was variation in TTNTD from each LoT by regimen group. Patients receiving treatment through clinical trials across 2 L to 4 L had better TTNTD compared with patients receiving all other common regimens.

Across the most common regimens in 1 L, the median TTNTD from initiation of 1 L was 47.4 months (95% CI: 43.4–51.7) among patients who received an ASCT during follow-up (Fig. 3a and Supplementary Information Sect. 4: Table S4.3.2) and 13.4 months (95% CI: 12.9–13.9) among those who did not (Fig. 3b and Supplementary Information Sect. 4: Table S4.3.3). This difference was greatest in patients receiving bortezomib plus cyclophosphamide or bortezomib plus thalidomide, for which the median TTNTD was approximately 50 months higher in patients who received ASCT.

Fig. 3
figure 3

Time to next treatment or death in patients receiving 1 L. (a) Patients who received ASCT in any LoT. (b) Patients who did not receive ASCT in any LoT. 1 L, first line of therapy; ASCT, autologous stem cell transplantation; LoT, line of therapy. Because no patients who received bortezomib + melphalan in 1 L subsequently received ASCT in any LoT, this regimen is not displayed in panel a. Patients with ASCT-only LoTs are excluded from this output. See Supplementary Tables S4.3.2 and S4.3.3 for medians and confidence intervals.

Full size image

Overall survival

OS was assessed from initiation of each LoT (1–4 L) in the sub-cohort of SACT-treated patients (n = 11,884), after excluding those with at least one ASCT-only LoT. OS decreased by an average of 36% at each LoT; the median OS was estimated to be 44.5 months (95% CI: 43.3–46.0) from 1 L (not reached for those who underwent ASCT and 29.6 [28.6, 30.9] for those who did not), 28.0 months (95% CI: 26.3–29.6) from 2 L, 17.6 months (95% CI: 15.8–19.3) from 3 L, and 11.5 months (95% CI: 9.8–13.4) from 4 L (Supplementary Information Sect. 4: Fig. S4.3.2 and Table S4.3.4).

As with TTNTD, OS from each LoT varied by regimen. From 1 L, the median OS for patients receiving bortezomib monotherapy was lower (22.8 months [95% CI: 20.8–25.0]) compared with those receiving bortezomib in combination with other chemotherapy agents (e.g. 41.8 months [95% CI: 39.2–45.1] for bortezomib plus cyclophosphamide). Patients receiving pomalidomide in 3 L and 4 L had the lowest median OS (9.6 months [95% CI: 6.9–12.9] and 6.3 months [95% CI: 4.6–7.9], respectively) compared with those receiving other common regimens at these LoTs (Supplementary Information Sect. 4: Fig. S4.3.2 and Table S4.3.5). Patients receiving therapy through clinical trials across 2 L to 4 L also had better median OS compared with those receiving other common SACT regimens at these LoTs, with median OS not reached in the trial population at 2 L, 3 L, and 4 L.

Differences in OS were also observed according to ASCT status (Fig. 4a and b). Approximately 77% of patients (95% CI: 75–78) who received ASCT during the follow-up period survived beyond 5 years from the initiation of 1 L (median value not reached) compared with approximately 28% (95% CI: 27–29) among those who did not.

Fig. 4
figure 4

Overall survival in patients receiving the five most common regimen groups at 1 L. (a) Patients who received ASCT in any LoT. (b) Patients who did not receive ASCT in any LoT. 1 L, first line of therapy; ASCT, autologous stem cell transplantation; LoT, line of therapy. Because no patients who received bortezomib + melphalan in 1 L subsequently received ASCT in any LoT, this regimen is not displayed in panel a. Patients with ASCT-only LoTs are excluded from this output.

Full size image

Discussion

This large retrospective study of patients with MM in England provided new insights into real-world treatment patterns in this population, using the CAS database.

Analysis of demographic data showed that patients receiving CDF treatments were younger than those in the main study cohort, with lower ECOG performance status scores and fewer recorded comorbidities. These patients could not be included in the SACT database8,25,26 analysis during the study period due to the previously mentioned restrictions on CDF reporting and were, therefore, treated as a separate analysis group, which may have skewed the findings by excluding healthier patients and therefore underestimating median OS. As the restrictions on CDF drug analysis are currently being lifted, future studies may provide a more comprehensive understanding of the real-world population of patients with MM, their clinical management, and treatment outcomes.

In the present study, 28.3% of patients on SACT received ASCT, which is in line with the figures of 26%25 and 31%8 reported in the literature. We reported differences between patients receiving ASCT and non-ASCT treatment regimens at 1 L, with the greatest difference among common regimen groups seen in the 42.1% of patients in the ASCT treatment pathway who received bortezomib plus thalidomide compared with 10% in the non-ASCT pathway. This is consistent with NICE recommendations for this combination at the time of the study16.

We found that some regimens that are now common, such as daratumumab, were received only by patients who did not receive an ASCT at any LoT during the follow-up period. This is likely due to NICE recommendations not overlapping the time frame of this study and the exclusion of patients receiving treatments via the reporting restrictions on CDF drugs (which are currently being lifted)18. Longitudinal treatment sequences in patients receiving health technology assessment/NICE-recommended regimens in 1 L were generally in accordance with treatment guidance, both overall and by ASCT status, with most patients receiving bortezomib plus thalidomide, bortezomib plus cyclophosphamide, or bortezomib monotherapy7. Treatment sequencing differed according to whether patients received ASCT during the follow-up period. The grouped SACT treatment regimens were generally in accordance with the NICE treatment pathway7, providing evidence supporting the good performance of the algorithm developed for this study and used to combine recorded treatments into LoTs. Other studies have also shown clinical practice reflecting treatment availability, with exposure to treatments such as daratumumab increasing in later LoTs25,26. With the increasing use of quadruplet regimens that include anti-CD38 MAbs in recent years, and their establishment as SoC at first LoT, it is expected that these regimens will positively impact PFS and OS outcomes.

Our analysis showed that among SACT-treated patients, survival outcomes decreased from the initiation of each subsequent LoT. This is aligned with other published findings showing that the depth and duration of response diminish over time27,28. We found that TTNTD was improved in patients who received the regimen groups bortezomib plus thalidomide and bortezomib plus cyclophosphamide; however, this may not have a causal interpretation, as differences between specific regimens may also be impacted by the differences in baseline characteristics. Furthermore, in this study, survival was longer in patients who received ASCT than in those who did not receive ASCT, regardless of the type of SACT received. This aligns with previous studies showing that ASCT is associated with improved survival and delayed disease progression compared with SACT alone, as well as with delayed disease progression regardless of other treatments29,30, but is confounded in our analysis by the likely selection of younger, fitter patients as suitable for ASCT treatment.

These results reflect clinical practice at the time of the study. Some regimens appeared to have been administered outside of NICE recommendations, but it is not possible to determine if this is due to individual variance in practice or a limitation of the algorithm used to identify the LoTs correctly. The MM treatment pathway in the UK has also evolved since this study took place, with recent and future progress expected to be seen in non-ASCT patients at diagnosis.

Learnings and limitations

This large database study contributes to the evidence base for MM in England. Study strengths include the highly granular nature of the NCRAS CAS database, which is thought to provide near-comprehensive capture of MM diagnoses (see Supplementary Information Sect. 5 for more information on the SACT database). The Cancer Outcomes and Services Dataset has been the national standard for reporting cancer since January 2013, while the SACT database provides comprehensive treatment capture10,11. The results obtained from analysis of these databases can, therefore, be considered broadly generalisable to adult patients in England over this time period.

Our study identified several gaps in data from the CAS SACT database that could impede the execution of fully representative analyses. These included inconsistent or non-comprehensive capture of important prognostic factors or treatment-effect modifiers such as performance status, comorbidities, and tumour characteristics (including cytogenetic risk) at baseline and later LoTs. There were variations in the consistency of steroid treatment capture in CAS, and data on steroid prescribing were lacking, which impacted treatment grouping and interpretation of results, as steroids are used in almost all combination regimens with other anti-myeloma classes of therapy. CAS does not directly capture therapy sequence, and ASCT records had to be identified separately within HES. Baseline characteristics are sometimes missing, or details of characteristics, such as weight and ECOG score, are not reported at treatment initiation or by LoT. SACT regimens are captured in CAS, but reliable data on treatment cycle, end dates, and reasons for treatment discontinuation (e.g. disease progression or treatment toxicity) are also lacking. Therefore, treatment pattern assessment in this study relied on the use of a specifically developed algorithm to combine recorded SACTs into LoTs (Supplementary Information Sect. 2). Although this algorithm was developed using national and international treatment guidance, some degree of LoT misclassification is still expected.

MM diagnosis codes may also include patients with smouldering MM, which does not require treatment. To ensure that the outcomes were for patients with active MM, only treated patients were included in the study objectives. A higher-than-expected number of patients with an MM diagnosis did not undergo treatment. This may reflect the number of smouldering MM cases, but could also indicate missing SACT data or incorrectly assigned diagnoses.

Some treatment sequencing did not reflect expected clinical practice; for example, 182 patients appeared to receive ASCT without any SACT record. This may reveal how accurately data are recorded or whether patients may receive treatments that are not recorded in SACT. In addition, some data did not reflect real-world treatment patterns; for example, bortezomib use at 2 L was lower than expected from treatment guidelines. However, the exclusion of records for patients who had received a treatment available through the CDF is thought to have impacted the overall results, especially as CDF-funded treatments represent a considerable part of routine clinical practice in myeloma.

Exclusion of patients receiving CDF treatments from analyses of treatment patterns and clinical outcomes in this study may have affected data representativeness and interpretation and may limit the generalisability of these findings to the real-world MM population in England. This is particularly important in patients with MM, where the treatment pathway is rapidly evolving with increased access to novel therapies and combinations via the CDF. Survival outcomes were only evaluated among SACT-treated patients and are potentially biased towards those with better survival. The overall SACT-treated population was also not fully reflected in this study since the study population did not receive CDF regimens, which are associated with longer survival outcomes. Therefore, survival analysis findings need to be interpreted with caution.

The CAS database does not include information on disease progression, so progression-free survival was proxied by TTNTD in the present study. However, this will lead to an overestimation of time to progression, which occurs before administration of the next treatment, and there are usually gaps between stopping and starting a new treatment in clinical practice. Reasons for treatment discontinuation are incompletely captured in SACT and were, therefore, not included in this study. Other data gaps included fewer characteristics recorded at the time of treatment progression, and no treatment information provided for patients receiving privately accessed treatment. ECOG score and weight were missing for more patients in the ASCT group because these variables are recorded in the SACT data set but not in HES, where ASCT procedures are recorded. Therefore, these variables are missing when a LoT starts with an ASCT procedure. Additionally, staging data were missing for 70% of the overall cohort in this study. This was because staging in CAS is mainly based on TNM classification, which is only used for solid tumours, and the ISS/R-ISS classification is not well recorded within this variable. Masking by the data owners to prevent patient identification also limited some interpretation of analyses.

In addition to key learnings about the impact of real-world evidence on treatment patterns and outcomes in patients with MM, this database study has highlighted limitations in current real-world data collection practice. We show where improvements could be made in recording of disease progression, treatment patterns, and reasons for regimen modification/discontinuation, as well as more comprehensively including regimen end dates. Baseline characteristics should be captured in greater detail, with ECOG score being recorded at the point of initiation of all LoTs. ASCT could be added as a variable within CAS, potentially as part of a curated cancer surgery data set (which may cover all cancers), similar to the NHS Digital radiotherapy data set31. There should be greater alignment and consistency of regimen reporting, with consistent treatment grouping and recording of steroid use. It would also be optimal to see greater transparency on the use of data collected within SACT for research as well as publication of more NCRAS-curated case study examples. The recent changes in accessibility of CDF treatment data will help to refine our understanding of management strategies and survival outcomes for all patients with cancer, including those with MM.

The learnings from this study have informed suggestions on how to improve the database outputs as part of a National Disease Registration Service review of the SACT v3 data set32. If implemented, future analyses using the updated database will form the basis for capture of additional data on MM management in England, filling some of the real-world data and knowledge gaps that remain.

Conclusion

This study describes the real-world treatment landscape for patients with MM in England and provides insights into survival outcomes associated with treatment and disease progression in these patients. The regimens received by patients were broadly aligned with NICE recommendations, and important differences in clinical outcomes were observed between patients who received ASCT and those who did not.

These findings reflect a treatment landscape in which patients with MM in England receive multiple therapies and treatment combinations over the course of their disease. The NHS England CAS database is a valuable source of information on routine clinical practice that can be used to inform health technology assessment decision-making. Using the current analysis as an example, however, we have highlighted several potential gaps and limitations in the data. These limitations present potential challenges to developing data analyses that are fully representative of the patient populations in England, and removal of the restrictions on including data from patients receiving therapies (including anti-myeloma therapies) accessible via the CDF is an important development. Additional refinement of data capture is needed for more comprehensive, data-rich, and robust real-world evidence collection in patients receiving SACT32.