← Back
NeuroTrials.ai
Neurology Clinical Trial Database

DELIVER-MS

Disease-modifying treatment preferences and decision-making in a multiple sclerosis randomized and observational clinical trial (DELIVER-MS)

Year of Publication: 2026

Authors: Tallantyre EC, Planchon SM, Howard H, et al; DELIVER-MS co-investigators

Journal: Multiple Sclerosis Journal

Citation: Mult Scler. 2026 Jun;32(7):722-736. DOI: 10.1177/13524585261449988

Link: https://doi.org/10.1177/13524585261449988


Clinical Question

In treatment-naive people with RRMS offered a pragmatic RCT of early highly-effective (EHT) vs escalation (ESC) DMT, what drives the decision to enter the RCT vs an observational cohort, and among those who choose their own therapy, which factors predict EHT vs ESC?

Bottom Line

DELIVER-MS enrolled 816 treatment-naive RRMS patients across 31 UK/US sites into a pragmatic RCT (n=393 mITT) and parallel observational study (n=374 mITT). Preference for a specific DMT (85%), efficacy concerns (20%), and safety concerns (9%) were the leading reasons for declining randomization; RCT participation was associated with older age, higher brain parenchymal fraction, and pre-COVID enrollment, while EHT choice in the observational cohort was associated with US enrollment, higher education, higher relapse rate, and greater T1 lesion volume. The RCT arms are well balanced; the definitive efficacy comparison (36-month brain atrophy) is still pending.

Major Points

  • Pragmatic 2-arm RCT (EHT vs ESC) plus a parallel observational cohort for those who declined randomization, at 31 US/UK sites; 816 enrolled, 767 in the mITT analysis (393 RCT, 374 OBS).
  • Reasons for declining RCT (n=371, multi-select): preference for a specific DMT 317 (85%), efficacy concerns 74 (20%), safety concerns 32 (8.6%), insurance 3 (0.8%), other 25 (6.7%).
  • Efficacy concerns were higher in the US (27%) than the UK (13%) (p=0.002); safety concerns trended higher in the UK (12%) than the US (5.7%) (p=0.061).
  • Independent predictors of RCT (vs OBS) participation in multivariate logistic regression: COVID-era enrollment (OR 0.15 vs pre-COVID, p<0.001), post-vaccine COVID era (OR 0.39, p<0.001), higher baseline brain parenchymal fraction (OR 1.34 per z-score, p=0.002), older age (OR 1.02/yr, p=0.037), lower SDMT (OR 0.98/point, p=0.009), and lower prior-year relapse rate (OR 0.79, p=0.043).
  • OBS cohort DMT selection: 125/374 (33%) chose ESC, 249/374 (67%) chose EHT; US 158/193 (82%) chose EHT vs UK 91/181 (50%) (p<0.001).
  • Independent predictors of choosing EHT (vs ESC) in the OBS cohort: UK site (OR 0.22 vs US, p<0.001), lower educational attainment (OR 0.33 for high school or less vs some college+, p<0.001), higher prior-year relapse rate (OR 1.55, p=0.025), and higher T1 lesion volume (OR 1.21, p<0.001).
  • RCT arms (n=193 ESC vs n=200 EHT) were well balanced at baseline including the primary endpoint (baseline brain parenchymal fraction 0.839 vs 0.836).
  • Among 317 people citing a specific-DMT preference for declining RCT, 100 (32%) ultimately started ESC and 217 (68%) started EHT; safety-driven decliners more often chose ESC (24/32 = 75%).
  • The definitive DELIVER-MS efficacy comparison (36-month whole-brain volume loss, EHT vs ESC) is expected mid-2027 and is not reported in this paper.

Design

Study Type: Multi-center, pragmatic, open-label, rater-blinded, phase 4 randomized controlled trial with a parallel prospective observational cohort (participants who declined randomization); baseline / decision-making analysis paper.

Randomization: 1

Blinding: Open-label for treating clinicians and participants; MRI raters (brain-volume analysis) are blinded to treatment allocation.

Enrollment Period: 10 January 2019 - 6 March 2023 (OBS cohort); 10 January 2019 - 31 March 2024 (RCT cohort).

Follow-up Duration: 36 months (primary endpoint of the parent trial); baseline analysis in this paper.

Centers: 31

Countries: USA, United Kingdom

Sample Size: 816

Power Calculation: DELIVER-MS was powered to detect a difference in 36-month brain volume loss between EHT and ESC; this paper reports baseline data only.

Analysis: Modified intention-to-treat (mITT) baseline cohort of 767 (excluded 49 who did not complete baseline MRI or start DMT within 90 days). Stepwise multivariable logistic regression predicting (a) RCT vs OBS participation and (b) EHT vs ESC choice within OBS. Continuous MRI metrics (BPF, GMF, WMF) scaled by z-score. Percentages calculated on non-missing data.


Inclusion Criteria

  • Adults aged 18-60 years
  • Relapsing-remitting multiple sclerosis meeting 2017 McDonald criteria
  • Treatment-naive to disease-modifying therapy
  • EDSS ≤ 6.5
  • Disease duration ≤ 5 years from symptom onset
  • Willing to receive either an early highly-effective (EHT) or an escalation (ESC) DMT
  • Able to attend follow-up visits and complete baseline MRI

Exclusion Criteria

  • Prior use of disease-modifying therapy for MS
  • Primary or secondary progressive MS
  • EDSS > 6.5
  • Disease duration > 5 years from symptom onset
  • Age < 18 or > 60 years
  • Inability to complete baseline MRI within 90 days of screening
  • Inability to commence assigned/chosen DMT within 90 days of screening

Baseline Characteristics

CharacteristicOverall mITT (n=767)RCT (n=393)OBS (n=374)RCT-ESC (n=193)RCT-EHT (n=200)OBS-ESC (n=125)OBS-EHT (n=249)
US n (%)411 (54%)218 (55%)193 (52%)35 (28%)158 (63%)
UK n (%)356 (46%)175 (45%)181 (48%)90 (72%)91 (37%)
Median age (IQR) yrs35 [29-43]
Female n (%)547 (71%)280 (71%)267 (71%)139 (72%)141 (71%)
White n (%)649 (85%)
Black n (%)64 (8.3%)
High school or less n (%)190 (25%)40 (32%)44 (18%)
Employed/Student/Homemaker n (%)690 (90%)
Median disease duration (IQR) yrs0.77 [0.33-1.87]
Median EDSS (IQR)2.0 [1.5-3.0]2.0 [1.5-3.0]2.0 [1.5-3.0]2.0 [1.5-3.0]2.0 [1.0-3.0]2.0 [1.0-2.75]2.0 [1.5-3.0]
Mean prior-year relapses (SD)1.25 (0.80)1.18 (0.80)1.33 (0.80)1.14 (0.88)1.22 (0.71)1.23 (0.80)1.38 (0.80)
Mean SDMT correct (SD)57.05 (14.46)55.68 (14.08)58.49 (14.72)
Median T2 lesion volume (IQR) mL8.28 [4.09-15.93]
Mean brain parenchymal fraction (SD)0.836 (0.020)
GdE lesion present n (%)250/696 (36%)
Median age (IQR)36 [29-43]34 [28-42]36 [29-43]35 [28.5-42.5]
Mean BPF (SD)0.837 (0.019)0.835 (0.020)0.839 (0.020)0.836 (0.019)
GdE present n (%)137/366 (37%)113/330 (34%)

Arms

FieldRCT - Early Highly-Effective Therapy (EHT)ControlObservational - EHTControl
InterventionFirst-line high-efficacy DMT (natalizumab, ocrelizumab, ofatumumab, alemtuzumab, rituximab, or ublituximab)First-line modest-efficacy DMT (interferons, glatiramer acetate, teriflunomide, dimethyl/diroximel fumarate, or S1P modulators [fingolimod, siponimod, ozanimod, ponesimod]) with escalation on breakthrough activityPatient-and-clinician-selected high-efficacy DMT (as above)Patient-and-clinician-selected modest-efficacy DMT (as above)
N200193249125

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Reasons for declining randomization to the RCT (multi-select from Table 4), among 371/374 OBS participants who provided a reason.PrimaryNot applicable (single-arm descriptive outcome).OBS cohort n=371: preference for a specific DMT 317 (85%); efficacy concerns 74 (20%); safety concerns 32 (8.6%); insurance issue 3 (0.8%); other 25 (6.7%).
Multivariate predictor of RCT (vs OBS) participation: COVID-pandemic era (Mar 2020 - Dec 2020) vs pre-COVID (Jan 2019 - Mar 2020)SecondaryPre-COVIDCOVID era0.15<0.001
Multivariate predictor of RCT participation: post-vaccine COVID era (Jan 2021 - Mar 2023) vs pre-COVIDSecondaryPre-COVIDPost-vaccine COVID0.36<0.001
Multivariate predictor of RCT participation: baseline brain parenchymal fraction (per 1-SD z-score)Secondaryper +1 z-score BPF1.340.002
Multivariate predictor of RCT participation: age (per year)Secondaryper +1 year1.020.037
Multivariate predictor of RCT participation: prior-year relapse rateSecondaryper +1 relapse0.790.043
Multivariate predictor of RCT participation: SDMT (per point correct)Secondaryper +1 SDMT point0.980.009
DMT choice within OBS cohortSecondaryESC 125/374 (33%)EHT 249/374 (67%)
EHT choice by country (OBS cohort)SecondaryUK 91/181 (50%) EHTUS 158/193 (82%) EHT0.22<0.001
Multivariate predictor of EHT (vs ESC) choice in OBS: lower educational attainment (high school or less vs some college+)SecondarySome college+High school or less0.33<0.001
Multivariate predictor of EHT choice in OBS: prior-year relapse rateSecondaryper +1 relapse1.550.025
Multivariate predictor of EHT choice in OBS: baseline T1 lesion volumeSecondaryper +1 mL T1 lesion volume1.21<0.001
Reasons for declining RCT - US vs UK: efficacy concernsSecondaryUK 23/179 (13%)US 51/192 (27%)0.002
Reasons for declining RCT - US vs UK: safety concernsSecondaryUK 21/179 (12%)US 11/192 (5.7%)0.061
No treatment safety outcomes reported in this baseline/decision-making paper; treatment-emergent adverse events will be reported with the final 36-month efficacy analysis.Safety

Subgroup Analysis

Country was the dominant subgroup effect: US participants were far more likely than UK participants to choose EHT in the OBS cohort (82% vs 50%, p<0.001). US decliners cited efficacy concerns more often (27% vs 13%, p=0.002); UK decliners cited safety concerns more often (12% vs 5.7%, p=0.061). COVID era was a strong subgroup: 68% of pre-COVID recruits entered the RCT vs 29% during the pandemic and 49% post-vaccine. Sensitivity analysis restricted to enrollments up to March 2023 (OBS closure) did not change the multivariate results.


Criticisms

  • Baseline/design paper only - the definitive EHT-vs-ESC efficacy comparison (36-month brain volume loss) is not yet available (expected 2027), so no conclusions about which strategy is better can be drawn from this publication.
  • Overrepresentation of academic MS centers, particularly in the US - EHT uptake in the OBS cohort (US 82%) is higher than reported in US population-level cohorts and limits generalizability.
  • Upper age limit of 60 years excludes older RRMS patients where the safety/benefit balance of EHT is more uncertain.
  • Reasons for declining the RCT were captured via a single questionnaire allowing multiple responses; qualitative interviews are ongoing but not yet reported.
  • COVID-19 pandemic dramatically shifted RCT vs OBS participation (pre-COVID 68% RCT vs pandemic 29%), which may confound comparisons and reduce statistical power for the parent trial's ITT analysis.
  • Open-label, patient-and-clinician-selected DMTs in the OBS cohort will require propensity-score adjustment for the observational comparisons and cannot fully eliminate confounding.
  • The observed association between higher education and EHT choice raises concerns about health-literacy-based disparities in DMT selection but the study cannot separate education from race, deprivation, or clinician bias.

Funding

Patient-Centered Outcomes Research Institute (PCORI) Award MS-1610-37047.

Based on: DELIVER-MS (Multiple Sclerosis Journal, 2026)

Authors: Tallantyre EC, Planchon SM, Howard H, et al; DELIVER-MS co-investigators

Citation: Mult Scler. 2026 Jun;32(7):722-736. DOI: 10.1177/13524585261449988

Content summarized and formatted by NeuroTrials.ai.