Variant | Patients (n) | Mean years | Absolute treatment effect (EDSS) | Relative treatment effect (EDSS) | |||
Follow-up | On treatment | Markov model | MLM | Markov model | MLM | ||
Primary analysis | 4862 | 8.7 | 7.0 | 0.12 (0.07, 0.17) | 0.61 (0.55, 0.66) | 93% (90%, 96%) | 72% (69%, 74%) |
RRMS-only subgroup | 4217 | 8.9 | 7.3 | 0.25 (0.20, 0.31) | 0.68 (0.62, 0.74) | 86% (83%, 89%) | 69% (67%, 72%) |
Supplementary/sensitivity analyses (including patients with SPMS at baseline except where noted) | |||||||
Excluding EDSS scores recorded after a patient has switched to a non-scheme DMT* | |||||||
Primary analysis cohort | 4799 | 8.0 | 6.4 | 0.13 (0.08, 0.19) | 0.59 (0.53, 0.64) | 91% (88%, 95%) | 70% (68%, 73%) |
RRMS-only subgroup† | 4157 | 8.1 | 6.7 | 0.26 (0.21, 0.32) | 0.66 (0.60, 0.72) | 84% (81%, 88%) | 68% (65%, 71%) |
Excluding EDSS scores recorded after a patient has switched to any other DMT* | |||||||
Primary analysis cohort | 4475 | 7.0 | 5.7 | 0.16 (0.11, 0.22) | 0.57 (0.52, 0.62) | 88% (84%, 92%) | 68% (65%, 70%) |
RRMS-only subgroup | 3871 | 7.0 | 5.9 | 0.29 (0.23, 0.35) | 0.64 (0.59, 0.70) | 80% (76%, 84%) | 65% (62%, 68%) |
Imputation—Markov model, using imputed values derived from the MLM | |||||||
‘on treatment’ assumption | 4862 | 9.9 | 7.7 | 0.12 (0.07, 0.17) | NA | 93% (91%, 96%) | NA |
‘off treatment’ assumption | 4862 | 9.9 | 7.0 | 0.10 (0.04, 0.15) | NA | 95% (92%, 98%) | NA |
Imputation—multilevel model | |||||||
‘on treatment’ assumption | 4862 | 9.9 | 9.9 | NA | 0.67 (0.62, 0.72) | NA | 72% (70%, 74%) |
‘off treatment’ assumption | |||||||
Single imputation | 4862 | 9.9 | 9.9 | NA | 0.65 (0.59, 0.70) | NA | 73% (70%, 75%) |
Multiple imputation | 4862 | 9.9 | 9.9 | NA | 0.64 (0.59, 0.70) | NA | 73% (71%, 75%) |
Single imputation +0.5 EDSS points for each imputed value | 4862 | 9.9 | 9.9 | NA | 0.48 (0.43, 0.53) | NA | 80% (78%, 82%) |
Supplement RSS data with imputed values derived from out-of-window year 10 scores‡ | 4862 | 8.9 | 7.0 | 0.12 (0.07, 0.18) | ‡ | 93% (90%, 96%) | ‡ |
Use transition matrices from BCMS data set supplemented by imputing additional data for patients with sparse follow-up‡ | |||||||
Including patients with SPMS at baseline | 4862 | 8.7 | 7.0 | 0.28 (0.23, 0.34) | ‡ | 84% (81%, 87%) | ‡ |
Excluding patients with SPMS at baseline | 4217 | 8.9 | 7.3 | 0.43 (0.37, 0.49) | ‡ | 78% (75%, 81%) | ‡ |
Alternative Markov model with time-varying natural history transition matrices | 4862 | 8.7 | 7.0 | 0.48 (0.42, 0.53) | NA | 76% (74%, 79%) | NA |
Year 1 baseline (see online supplementary appendix 8) | 4360 | 7.9 | 6.2 | 0.07 (0.02, 0.12) | 0.43 (0.38, 0.48) | 95% (92%, 99%) | 77% (74%, 79%) |
*Patients switching before year 1 are also excluded from the analysis.
†This was the primary analysis for the interim year 6 analysis.
‡These variants were only carried out for the Markov model.
DMT, disease-modifying treatment; EDSS, Expanded Disability Status Scale; MLM, multilevel model; NA, not applicable; RRMS, relapsing remitting multiplesclerosis; RSS, risk sharing scheme; SPMS, secondary progressive multiple sclerosis;BCMS, British Columbia Multiple Sclerosis.