This is an Extension Study of the Roche P-trials to Investigate Safety and Effectiveness of Ocrelizumab in Participants With Multiple Sclerosis (MS): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for This is an Extension Study of the Roche P-trials to Investigate Safety and Effectiveness of Ocrelizumab in Participants With Multiple Sclerosis (MS) (NCT03599245), a na phase 3 study. The primary outcome is Time to Onset of Confirmed Disability Progression (CDP) Sustained for at Least 24 Weeks From CASTING Baseline to LIBERTO Week 96.

2. Clinical Question

The clinical question concerns time to confirmed disability progression among participants with multiple sclerosis who entered LIBERTO after participating in the CASTING study. The registry record specifies a primary endpoint of time from CASTING baseline to onset of progression confirmed for at least 24 weeks, with follow-up through LIBERTO Week 96, up to 4 years. It describes treatment with ocrelizumab, but there is no concurrent control arm in this extension study.

This is an extension-study question about outcomes during follow-up in a defined rollover population, not a randomized estimate of the difference between ocrelizumab and another treatment. Without a concurrent comparison group, the endpoint can describe the observed time-to-event experience in the enrolled population, but it cannot by itself separate treatment effects from other influences on disease course or follow-up.

3. Trial Design

NCT03599245 is described as a Phase 3, single-arm, open-label, multicentre extension study. Its stated purpose was to evaluate the effectiveness and safety of ocrelizumab in people with multiple sclerosis who had previously participated in a Roche-sponsored ocrelizumab phase IIIb/IV trial. The registry record identifies the sponsor as Hoffmann-La Roche.

The study began on 2018-07-12, had primary completion on 2025-07-17, and is listed as completed. Enrollment was 1055. Participants were assigned no separate randomized comparator: the allocation entry is NA, masking is none, and the study model is single group. Consequently, comparisons between treatment arms are not part of the reported design.

The intervention description states that participants received a single 600-mg infusion of ocrelizumab every 24 weeks up to Week 72. The primary endpoint focuses on the CASTING rollover cohort and measures time from CASTING baseline to LIBERTO Week 96. The primary endpoint population was defined as all participants who rolled over from CASTING to LIBERTO and received any dose of ocrelizumab in LIBERTO, including participants who withdrew prematurely for any reason. This endpoint population is therefore specifically defined by prior CASTING participation and rollover; the registry record does not establish that it is identical to the full enrollment population.

Secondary outcomes include time to a first protocol-defined event of disease activity, time to first relapse, annualized protocol-defined relapse rate, relapse-free status, and several measures involving no evidence of disease activity or progression in named cohorts. The record lists 51 outcome measures and 0 analyses. It does not report further design detail about how secondary outcomes were prioritized or how their results were analyzed.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: Participants continued to receive ocrelizumab during treatment phase and subsequently entered a safety follow-up in LIBERTO study.

Treatment Period

Reasons for not completing this period:

Safety Follow-up Period

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: ITT population included all participants who rolled over from the CASTING study to LIBERTO and received any dose of ocrelizumab in LIBERTO, including participants who prematurely withdrew from the study for any reason.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: Intent-to-Treat (ITT) population included all enrolled participants who received any dose of ocrelizumab.

6. Primary Endpoint

The registry lists the primary outcome as Time to Onset of Confirmed Disability Progression (CDP) Sustained for at Least 24 Weeks From CASTING Baseline to LIBERTO Week 96, measured over From CASTING Baseline to LIBERTO Week 96 (up to 4 years).

The primary endpoint is time to onset of confirmed disability progression, sustained for at least 24 weeks, from CASTING baseline to LIBERTO Week 96. It is expressed in weeks, and the planned summary measure listed in the registry record is the median with a 95% confidence interval. The endpoint is restricted to the CASTING rollover population described in the record.

Progression was based on change in the Expanded Disability Status Scale (EDSS). For a baseline EDSS score of 5.5 or less, progression was defined as an increase of at least 1.0 point not attributable to another cause, such as fever, concurrent illness, or concomitant medication. For a baseline score above 5.5, the specified increase was at least 0.5 points. For participants with an EDSS of 0, progression was defined as a change of at least 1.5 points. An increase counted as confirmed when a regularly scheduled visit at least 24 weeks after the initial documentation of neurological worsening confirmed it.

As a general statistical concept, a time-to-event endpoint incorporates both whether an event occurs and how long it takes to occur. Requiring confirmation over time can distinguish a sustained change from a worsening that is not confirmed at the required later visit. The EDSS combines ratings across seven functional systems; the registry describes its scale as ranging in 0.5-point steps from 0, normal, to 10, death.

The registry record names the Kaplan-Meier method for estimating the median, but does not report the resulting median or its confidence interval. The observed endpoint result therefore cannot be characterized numerically from the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is a time-to-event measure. Each participant’s event time is measured from the stated origin, CASTING baseline, until the qualifying onset of confirmed disability progression, or until follow-up ends without such a recorded event. The endpoint definition requires worsening on the EDSS and confirmation at a regularly scheduled visit at least 24 weeks after the initial documentation of neurological worsening. This confirmation rule is part of the outcome definition; it should not be confused with the separate statistical handling of participants who have incomplete follow-up.

What does Kaplan-Meier estimation do?

As a general statistical explanation, the Kaplan-Meier method estimates the probability of remaining free of the event over time. It can use information from participants whose follow-up ends before an event is observed, provided their event status up to the end of their follow-up is known. Such observations are commonly called censored. Censoring is not the same as an event, and a participant who leaves follow-up is not automatically counted as having experienced progression.

The method relies on assumptions about how censoring relates to the event process. In general, the estimate is most interpretable when participants whose follow-up ends are not systematically more or less likely to experience the event, conditional on the information used in the analysis. The registry record does not report the number censored, reasons for censoring, censoring distribution, or checks of assumptions. It therefore does not allow assessment of how these features affected the estimate.

How is a median time interpreted?

In general, a median from a Kaplan-Meier curve is the time at which the estimated event-free probability reaches one half, if the curve reaches that point during observed follow-up. It is a summary of the estimated distribution, not the average follow-up time and not a comparison of two groups. If the curve does not reach that point, a median may not be estimable within the observed data. The registry lists the median as the endpoint parameter, but gives no numerical estimate and does not say whether it was estimable.

What does a confidence interval contribute?

In general, a confidence interval describes statistical uncertainty around an estimated quantity under the analysis assumptions. For a median time, the interval indicates a range of values compatible with the estimation procedure and observed data at the stated confidence level. It does not guarantee that the true value lies within a particular interval, and it does not correct for design limitations such as lack of a concurrent randomized comparator. The registry specifies a 95% confidence interval, but its endpoints are not reported in the registry record.

Why does event definition matter?

The EDSS-based definition sets thresholds that depend on baseline score: an increase of at least 1.0 point for a baseline score of 5.5 or less, at least 0.5 for a baseline score above 5.5, and at least 1.5 for participants whose baseline score is 0. The record also excludes increases attributed to another etiology, with examples including fever, concurrent illness, and concomitant medication. Confirmation at a later scheduled visit helps define the event as sustained according to the protocol’s stated rule. Differences in threshold by baseline score mean that the event is not defined by one identical numerical change for every participant.

The EDSS is described as incorporating seven functional systems rated and scored as Functional Systems Scores. Each system uses an ordinal clinical rating scale ranging from 0 to 5 or 6, and the overall EDSS runs from 0, normal, to 10, death in 0.5-point steps. As a general measurement concept, ordinal scales rank levels but do not imply that every step represents the same underlying change. The endpoint therefore depends on a standardized clinical assessment and the stated threshold and confirmation rules.

What can a single-arm estimate establish?

In general, an outcome estimate from one group can describe the observed experience of that group under its study conditions. It does not provide the counterfactual outcome that the same population would have had under a different treatment or no treatment. A randomized comparator can support a between-group estimate by balancing, on average, measured and unmeasured factors at assignment; this study’s single-group design does not make that comparison. Open-label status also means masking was absent, although the registry record does not describe how this affected outcome assessment.

The record identifies 51 outcome measures and 0 analyses, but gives no detailed statistical analysis plan, multiplicity procedure, or endpoint-specific analysis output. The stated Kaplan-Meier method is the only primary-endpoint estimation method described. Further details about covariate adjustment, sensitivity analyses, subgroup analyses, censoring rules, or assessment of proportional hazards are not reported in the registry record. No hazard-ratio result is stated for this primary endpoint. These omissions limit how fully the analysis can be reconstructed or its robustness assessed.

10. Multiplicity, Missing Data and Analysis Populations

The registry record lists 51 outcome measures, spanning the primary disability-progression endpoint and secondary outcomes involving disease activity, relapse, and progression. It lists 0 analyses. The record does not report whether the outcomes were organized into a hierarchy, whether any multiplicity adjustment was planned, or how the interpretation of multiple statistical tests was controlled. Without those details, any set of outcome findings would need to be interpreted with caution: when many outcomes are examined, the chance of at least one apparently notable result can increase. That is a general statistical consideration, not a claim about an unreported analysis in this study.

Missing-data procedures are not reported in the registry record. The primary endpoint population is described as all participants who rolled over from CASTING to LIBERTO and received any dose of ocrelizumab in LIBERTO, including those who withdrew prematurely for any reason. This definition indicates that withdrawal did not by itself remove a participant from the stated population, but it does not explain how follow-up after withdrawal, missing EDSS assessments, or incomplete event confirmation were handled in the analysis.

The record does not report the number of participants with missing endpoint assessments, their reasons for missingness, any imputation approach, or sensitivity analyses addressing missing data. These details matter because missing follow-up can affect the risk set and the interpretation of time-to-event estimates, depending on when and why follow-up ends. The record also does not report whether the analysis population differed from the stated endpoint population in practice. The stated population definition and the missing-data handling are distinct pieces of information; the former is available, while the latter is not reported in the registry record.

11. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: covid-19 pneumonia (2.6%), covid-19 (1.6%), abortion spontaneous (1.0%), sinusitis (0.7%), pneumonia (0.6%).

The safety record lists serious adverse event information separately for the CASTING and ENSEMBLE cohorts. In the CASTING cohort, 439 participants were at risk and 41 were reported as affected by serious adverse events; deaths were 0. In the ENSEMBLE cohort, 616 participants were at risk and 109 were reported as affected; deaths were 2. These are cohort-level counts, not randomized comparisons, and the record does not report an unexposed control group.

The listed serious adverse event terms include COVID-19 PNEUMONIA at 2.6%, COVID-19 at 1.6%, ABORTION SPONTANEOUS at 1.0%, SINUSITIS at 0.7%, PNEUMONIA at 0.6%, and PYELONEPHRITIS at 0.5%. These percentages describe the entries as reported in the registry; the record does not supply additional detail about event timing, recurrence, severity grading, or causal attribution. A listed event does not by itself establish that the study treatment caused it.

General interpretation requires attention to the population at risk, follow-up duration, and how events were collected and classified. The registry record does not report those details for these cohort summaries. The counts should therefore be read as descriptive safety information, not as a comparative estimate or a measure of treatment effect.

12. What the Result Does, and Does Not, Mean

The registry record establishes the design and definitions for an extension study of ocrelizumab in participants with multiple sclerosis previously enrolled in Roche-sponsored ocrelizumab trials. It specifies a single-group, open-label design, an infusion schedule, and a primary endpoint based on sustained, confirmed EDSS worsening among participants rolling over from CASTING to LIBERTO. It also describes a Kaplan-Meier method for the median time-to-event summary.

It does not report the primary endpoint’s median, 95% confidence interval, or numerical event result. Therefore, the record does not establish the observed timing or precision of disability progression in this population. The completed study status and the presence of a detailed endpoint definition do not substitute for those numerical results. Secondary outcome names are listed, but numerical outcomes and analysis details are not reported in the registry record.

The design also limits what any single-group result could mean. In general, a follow-up estimate can characterize the participants observed under the study conditions, but without a concurrent randomized comparator it cannot quantify how outcomes would differ under another treatment strategy. Participants entered as rollover cohorts from prior studies, and the primary endpoint population is restricted to those who rolled over from CASTING and received any LIBERTO dose. The result, if reported, would thus pertain to that defined population rather than automatically applying to all people with multiple sclerosis.

The safety counts are descriptive cohort summaries. They do not establish that an event was caused by treatment, and without a comparable control group they do not provide a randomized estimate of excess risk. The registry record also does not report enough detail about missing follow-up, event ascertainment, or analysis procedures to evaluate those aspects fully. The appropriate interpretation is consequently bounded: the record describes a study framework and specified outcomes, while key numerical efficacy results and several analytic details are not reported in the registry record.

13. Why This Trial Matters Statistically

This study illustrates why design and endpoint definitions must be considered alongside numerical estimates. Time to confirmed disability progression is not simply a count of worsening scores: the registry defines a change in EDSS relative to baseline, excludes worsening attributed to other causes, and requires confirmation at a later scheduled visit. These rules determine what qualifies as an event and when the event time is assigned.

The extension design also raises a central interpretive distinction. A single-arm estimate can describe the time-to-event experience in the defined rollover population, but it does not supply a randomized contrast. This matters when readers encounter terms such as effectiveness or safety in a study title: the design determines whether the reported data support description of one group or estimation of a comparative effect.

The record’s population definition is similarly important. The study enrollment was 1055, while the primary endpoint population is defined by prior CASTING participation, rollover to LIBERTO, and receipt of any dose. A general enrollment figure should not be assumed to be the denominator for every endpoint. Distinguishing study enrollment, endpoint eligibility, and the number actually contributing follow-up is essential for understanding an analysis.

Finally, an endpoint definition and a stated statistical method do not amount to a result. The record specifies Kaplan-Meier estimation of a median and a 95% confidence interval, but the numerical estimate and interval are not reported in the registry record. This makes the distinction between planned or described methods and reported findings especially visible: statistical interpretation depends on both the design and the actual estimates, with their uncertainty, being available.

14. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

Several limitations specific to this registry record constrain interpretation. First, the study is single-arm, open-label, and has no randomized comparator. Any reported outcome would therefore describe the observed cohort rather than estimate a randomized difference between treatment strategies. The record does not report masking, which is consistent with the listed masking value of none.

Second, the primary endpoint result itself is not reported numerically. Although the median and 95% confidence interval are identified as the summary, neither is stated. The registry record also does not report an endpoint event count, a hazard ratio, or a p-value. Precision and the amount of observed endpoint information consequently cannot be evaluated.

Third, the primary endpoint population is a defined rollover subset: participants from CASTING who entered LIBERTO and received any dose, including those who withdrew early. The full enrollment figure of 1055 does not establish the analyzed population size. The number included in the endpoint analysis is not reported in the registry record.

Fourth, the record does not report details of censoring, missing assessments, endpoint ascertainment, or sensitivity analyses. It also does not report the statistical plan for the 51 listed outcome measures, including whether multiplicity was addressed. These omissions limit assessment of robustness and of how secondary findings should be interpreted.

Finally, safety summaries are reported by cohort, without a randomized control group or further detail about exposure time and event attribution. The counts and listed terms are informative as descriptive entries but cannot establish comparative risk or causation on their own.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: This is an Extension Study of the Roche P-trials to Investigate Safety and Effectiveness of Ocrelizumab in Participants With Multiple Sclerosis (MS) — NCT03599245.
• Registry record last updated 2026-09-25; status last verified 2026-08.
• Results first posted 2026-09-25.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

18. Record Summary

NCT03599245 is a completed, Phase 3, single-arm, open-label, multicentre extension study sponsored by Hoffmann-La Roche. It enrolled 1055 participants with multiple sclerosis who had previously participated in a Roche-sponsored ocrelizumab trial. The study began on 2018-07-12, with primary completion on 2025-07-17.

The intervention was a single 600-mg ocrelizumab infusion every 24 weeks up to Week 72. The primary endpoint measured time to confirmed disability progression sustained for at least 24 weeks, from CASTING baseline to LIBERTO Week 96, up to 4 years. The specified summary was the Kaplan-Meier median with a 95% confidence interval in participants rolling over from CASTING who received any LIBERTO dose.

The registry record lists 51 outcome measures and 0 analyses. It does not report the primary endpoint’s numerical median or confidence interval, or detailed missing-data and multiplicity methods. Serious adverse event counts are listed for the CASTING and ENSEMBLE cohorts.