Efficacy and Safety Study of Satralizumab (SA237) as Monotherapy to Treat Participants With Neuromyelitis Optica (NMO) and Neuromyelitis Optica Spectrum Disorder (NMOSD): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Efficacy and Safety Study of Satralizumab (SA237) as Monotherapy to Treat Participants With Neuromyelitis Optica (NMO) and Neuromyelitis Optica Spectrum Disorder (NMOSD) (NCT02073279), a randomized phase 3 study. The primary outcome is Time to First Protocol-Defined Relapse (TFR) During the Double-Blind (DB) Period.

2. Clinical Question

In general, a randomized, placebo-controlled comparison is intended to estimate how outcomes differ between assigned treatment strategies under the conditions of the trial. The primary endpoint focuses on when a qualifying relapse first occurs, rather than counting every relapse or measuring symptom change at a single visit. Its meaning therefore depends on the protocol definition of relapse, the period in which events were counted, and the handling of participants who had not relapsed by the end of observation.

3. Trial Design

The registry describes a randomized, double-blind, parallel-group Phase 3 study sponsored by Hoffmann-La Roche. The study was completed. The record gives a start date of 2014-08-05 and a primary completion date of 2018-10-12. Enrollment was 95. Participants were assigned to satralizumab monotherapy or placebo monotherapy during the double-blind period.

The posted primary outcome compares the placebo, then satralizumab group with the satralizumab, then satralizumab group. The primary endpoint was time to first protocol-defined relapse during the double-blind period, with a timeframe described as up to Week 216. The reported analysis used the intention-to-treat population: all participants randomized to the treatment groups. For participants who had not relapsed, time to first relapse was censored on the date the double-blind period ended.

The record names secondary outcomes covering pain, fatigue, relapse-free rate, annualized relapse rate, and several health-status domains. These include bodily pain, general health, mental health, and physical functioning. The registry lists 31 secondary outcomes and 32 outcome measures overall. The reported outcome information does not establish how each secondary measure was analyzed or how those results should be prioritized relative to the primary endpoint.

Randomization can support a fair comparison by reducing systematic differences between assigned groups at the start of a trial. Double blinding is generally intended to limit the influence of treatment knowledge on participant experiences and outcome assessment. Those are general features of these design choices; the registry record does not report further operational details about how randomization or masking was implemented. It also does not report the original sample-size justification or the detailed protocol for the transition to open-label treatment.

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 with neuromyelitis optica (NMO) or NMO spectrum disorder (NMOSD) were randomized 2:1 to receive either satralizumab 120 mg or matching placebo.

Double-blind Period

Reasons for not completing this period:

Open-label Extension Period

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: ITT population included all participants randomized to the treatment groups. For participants who had not relapsed, the TFR was censored on the date of end of the DB period.

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 participants randomized to the treatment groups.

6. Primary Endpoint

The registry lists the primary outcome as Time to First Protocol-Defined Relapse (TFR) During the Double-Blind (DB) Period, measured over Up to Week 216.

The primary endpoint was time to first protocol-defined relapse during the double-blind period. Time was measured from randomization to the first qualifying relapse. A relapse required new or worsening neurological symptoms attributable to NMO or NMOSD that persisted for >24 hours and were not attributable to confounding clinical factors, such as fever, infection, injury, change in mood, or medication adverse reactions. New or worsening symptoms occurring <31 days after the onset of a protocol-defined relapse were counted as part of that same relapse; the onset date used in the analysis was the date of the first relapse.

A time-to-event endpoint captures both whether an event occurs and when it occurs during follow-up. In general, this can use information from participants who have different lengths of observation, including those who do not experience the event before follow-up ends. Such participants are censored rather than treated as if a relapse occurred at the end of follow-up. In this trial, participants without a relapse were censored on the date their double-blind period ended.

The registry reports median time to first relapse and a hazard ratio from a log-rank analysis. The median describes the event-time distribution, while the hazard ratio compares event rates over follow-up under the analysis model. These measures answer related but different questions: neither is simply a count of relapses or a measure of symptom severity. Interpretation rests on the event definition, censoring, and the reported analysis population.

7. Results Posted in the Registry

• Placebo, Then Satralizumab vs Satralizumab, Then Satralizumab: hazard ratio 0.45 (95% CI 0.23–0.89), reported p = 0.018 (as reported in the registry).

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is a time-to-event measure: time from randomization until the first protocol-defined relapse during the double-blind period. This differs from an analysis of a simple event count. A participant contributes information about the time at risk until a relapse is observed or follow-up ends. Participants who have not relapsed when observation ends contribute censored information rather than an observed relapse time. The registry states that participants without a relapse were censored at the end of their double-blind period.

What does censoring mean?

In general, censoring occurs when the event of interest has not been observed by the end of a participant’s available follow-up, or when observation otherwise stops before the event. The participant’s exact event time is then unknown, though it is known to extend beyond the last observed time. A time-to-event method can retain this partial information. Its validity depends on assumptions about the relationship between censoring and event risk. The registry record does not describe diagnostic checks or additional censoring assumptions for this analysis.

The endpoint definition also determines which events count. In this trial, a relapse required new or worsening neurological symptoms attributable to NMO or NMOSD, persistence for >24 hours, and no attribution to listed confounding clinical factors. Symptoms arising <31 days after relapse onset were treated as part of the same relapse. These rules aim to distinguish qualifying relapses from transient symptoms or other causes and to avoid counting closely spaced symptom episodes as separate events. In general, endpoint definitions improve consistency, but they also limit what the result says: the estimate pertains to events meeting this protocol definition.

How should the hazard ratio be read?

The registry reports a hazard ratio of 0.45 for placebo, then satralizumab versus satralizumab, then satralizumab, using a log-rank analysis. In general, the hazard is the rate at which an event occurs among participants who have not yet experienced it, at a given point in follow-up. A hazard ratio compares these rates over time. It is not the same as a risk ratio, an odds ratio, or the ratio of the groups’ median times. Its interpretation depends on which group is the numerator and the assumptions about how the relative hazards behave over time.

The registry names the order of the compared groups but does not explain the numerical reference coding. The estimate should consequently be quoted with the group labels, rather than described as favoring one group without confirming the coding. The posted result gives a 95% confidence interval from 0.23 to 0.89 and a p-value of 0.018. The sponsor’s reported analysis is the relevant result for this comparison; the record does not provide a separate adjusted estimate or an alternative model to interpret alongside it.

What does the log-rank method contribute?

In general, the log-rank test compares the observed event pattern between groups over follow-up, taking account of the timing of events and censoring. It tests a null hypothesis of no difference in the event-time distributions under its assumptions. The associated hazard ratio provides an estimate of relative hazards, while the test’s p-value summarizes how compatible the observed data are with the null hypothesis. A p-value is not the probability that the null hypothesis is true, nor does it measure the size or clinical importance of an effect.

Time-to-event summaries can be useful when participants are followed for different periods or when some have not experienced the event by study end. In general, this approach relies on the censoring process not distorting the comparison and on the analysis being suitable for the event patterns. The registry record does not report diagnostic assessments of proportional hazards, an alternative model, or details of how ties or other technical features were handled. The log-rank method is named, but further statistical implementation details are not reported in the registry record.

Why are the medians not complete on both sides?

The placebo, then satralizumab group has a reported median of 128.3 weeks and lower confidence limit of 29.9 weeks, while its upper confidence limit is not reported. The satralizumab, then satralizumab group has no numeric median reported; its lower confidence limit is 135.7 weeks, and the upper limit is not reported. In general, a median event time is the point at which the estimated event-time distribution reaches its midpoint. If the data do not support estimating that point or one side of its interval with sufficient information, the corresponding value may be unavailable. The record does not state why these particular median entries or interval limits are absent, so no more specific explanation can be made.

These median summaries and the hazard ratio should be considered complementary, not interchangeable. The medians describe selected points on the event-time distributions, whereas the hazard ratio summarizes a relative event-rate comparison over follow-up. Neither should be interpreted without the censoring rule, time window, and event definition.

10. Confidence Intervals and Precision

The primary hazard-ratio estimate is 0.45, with a 95% confidence interval from 0.23 to 0.89. In general, a confidence interval gives a range of values compatible with the data and the statistical procedure, under the model’s assumptions. It is not a guarantee that the true effect lies inside the interval, nor does it describe the spread of individual participants’ outcomes. The interval is the key indication of precision for the reported hazard ratio.

The interval includes values representing different magnitudes of relative hazard. Because its limits are separated, the estimate is not known with high precision. Both the estimate and its uncertainty matter: reporting only 0.45 would give an incomplete impression, while the interval shows that the data are compatible with a range of relative effects. The lower and upper confidence limits lie on either side of 1, the value corresponding to equal hazards in a conventional ratio interpretation.

The reported median estimates are also incomplete. For placebo, then satralizumab, the median is 128.3 weeks, with a lower limit of 29.9 weeks and no upper limit reported. For satralizumab, then satralizumab, the median is not reported as a number; the lower limit is 135.7 weeks and the upper limit is not reported. These missing interval endpoints restrict what can be said about the precision of the medians. The registry record does not explain why the values are unavailable.

Confidence intervals account for sampling uncertainty under the analysis, but they do not address every possible source of uncertainty. In general, they do not correct for limitations in the outcome definition, follow-up, missingness, model assumptions, or the conduct of the study. The registry record does not report sensitivity analyses or diagnostics that would clarify how robust the primary result is to such choices.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are based on the observed effect and are illustrations, not the trial’s original design calculations. For this time-to-event outcome, the figures represent total events, not participants per arm. The listed totals are 50 and 66 events for planning around 100% of the observed effect, 88 and 118 events for 75% of the observed effect, and 197 and 264 events for 50% of the observed effect, for the respective 80% and 90% power targets.

The stated illustration assumes a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. It uses an effect equal to the stated fraction of the observed effect. In general, the 80% power column means the planning target is to have an 80% chance of detecting the specified effect under those assumptions if that effect is present. The 90% column uses a higher chance of detection, and therefore calls for a larger event target for the same assumed effect.

The figures are not observed power. They describe planning targets under assumptions, rather than the probability that the completed trial detected its actual effect. For time-to-event planning, the displayed target is the number of events needed, not a direct count of people to enroll. How many participants would be needed depends on event occurrence and follow-up, among other factors. Smaller effects need more participants or events to detect than larger effects, all else equal.

The original sample-size justification and design assumptions are not reported in the registry record. The planning figures should therefore not be treated as evidence of the trial’s planned enrollment rationale or as a reconstruction of its original protocol.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

d is the total number of events needed (Schoenfeld's formula) with 1:1 allocation and proportional hazards.

Worked calculation for this trial

1. Hazard ratio HR = 0.450, so ln(HR) = -0.7985.
2. d = 4 × (1.96 + 0.842)² / [ln(0.450)]² = 49.2, rounded up to 50 events in total for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
hr <- 0.45                              # observed hazard ratio
events_needed <- function(power, frac = 1) ceiling(round(4 * (z_a + qnorm(power))^2 / log(hr^frac)^2, 6))   # Schoenfeld, 1:1 allocation
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: events = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(events_needed(0.80, f), big.mark = ","), format(events_needed(0.90, f), big.mark = ",")))

# Simulation: given d total events under proportional hazards, events in the treatment arm are Binomial(d, hr / (1 + hr)).
# The score (log-rank type) statistic is (D1 - d / 2) / sqrt(d / 4).
set.seed(2073279)
d    <- events_needed(0.80)
reps <- 20000
D1 <- rbinom(reps, d, hr / (1 + hr))
z  <- (D1 - d / 2) / sqrt(d / 4)
cat(sprintf("Simulated power with %s events: %.3f (%d replicates)\n", format(d, big.mark = ","), mean(abs(z) > z_a), reps))

Output from running the code

100% of observed effect: events = 50 (80% power), 66 (90% power)
 75% of observed effect: events = 88 (80% power), 118 (90% power)
 50% of observed effect: events = 197 (80% power), 264 (90% power)
Simulated power with 50 events: 0.816 (20000 replicates)

The simulation uses the fact that, with d events in total and proportional hazards, the number of events in the treatment arm is binomial. It is an approximation that conditions on the total number of events, and it should land close to the target power.

13. Analyses Reported by the Sponsor

The registry’s reported analysis is a log-rank comparison with a hazard ratio of 0.45, a 95% confidence interval from 0.23 to 0.89, and a p-value of 0.018. The same hazard-ratio estimate and interval appear in the posted comparison information. The record does not flag a discrepancy between the sponsor-reported analysis and the listed comparison.

The reported analysis is the primary source for interpreting the trial’s result. Recalculated values, where available, are simple unadjusted calculations from posted summary statistics and may not reproduce the sponsor’s analysis choices. In this record, no discrepancy is flagged, so there is no stated conflict requiring a competing recalculated result to be discussed. The sponsor’s method is identified as log rank; the registry does not give further implementation details or an adjusted result.

The estimate should be read with the group order stated in the registry: placebo, then satralizumab versus satralizumab, then satralizumab. The record does not separately define the hazard-ratio reference coding, so the comparison label should accompany any interpretation of direction. The reported p-value and confidence interval describe statistical uncertainty under the sponsor’s analysis; they do not establish clinical importance or resolve limitations in the record.

14. Multiplicity, Missing Data and Analysis Populations

The primary endpoint is identified as time to first protocol-defined relapse during the double-blind period. The registry also lists 31 secondary outcomes, including measures of pain, fatigue, relapse-free rate, annualized relapse rate, and several SF-36 health domains. When a study evaluates multiple outcomes, time points, or analyses, the chance of at least one apparently unusual result can increase. In general, multiplicity procedures can control or describe this issue, but the record does not report a multiplicity adjustment strategy or a hierarchy for interpreting secondary outcomes.

The record does not report whether interim analyses were planned or conducted, whether the primary analysis was prespecified in the posted protocol, or whether any changes were made to the analysis plan. These details are not reported in the registry record. Their absence means the reported primary result can be described, but the record alone does not allow a full assessment of how the analysis plan was set or whether all planned analyses are represented.

The primary analysis population is described as the intention-to-treat population, including all participants randomized to the treatment groups. Participants without a relapse were censored at the end of the double-blind period. The registry does not report further missing-data methods, sensitivity analyses, reasons for censoring beyond that rule, or handling of missing covariates. In general, censoring and other missing data can affect interpretation if their occurrence is related to event risk or treatment experience.

The record identifies a log-rank method for the primary comparison, but does not report diagnostics for model assumptions or details of any supporting analyses. It also does not report a statistical plan for the secondary measures. These omissions do not negate the posted result; they limit the independent evaluation of analysis choices and the broader family of outcomes.

15. 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: urinary tract infection (3.6%), spinal compression fracture (3.6%), ovarian adenoma (3.6%), uterine leiomyoma (3.6%), migraine (3.6%).

The serious adverse event figures are reported separately for the double-blind and open-label periods. During the double-blind period, serious adverse events affected 5 of 32 participants in the placebo group and 12 of 63 in the SA237 group. During the open-label period, serious adverse events affected 3 of 28 in the placebo group and 7 of 56 in the SA237 group. No deaths are reported for any of these groups.

The listed serious adverse event terms include urinary tract infection, spinal compression fracture, ovarian adenoma, uterine leiomyoma, and migraine, each reported at 3.6%, and influenza at 3.2%. These are the terms identified in the registry’s serious adverse event listing; they do not constitute a complete description of all safety outcomes or establish causation.

In general, adverse-event counts depend on the population at risk, duration of observation, event definitions, and whether participants can contribute to more than one period. The registry record does not provide exposure-adjusted rates or enough detail to compare the periods as if they were equivalent. The figures should be read as reported counts and percentages within their named periods, not as a causal comparison or a complete assessment of safety.

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

The primary result is an estimate of time to first protocol-defined relapse during the double-blind period among the intention-to-treat population. The registry reports a hazard ratio of 0.45 for placebo, then satralizumab versus satralizumab, then satralizumab, with a 95% confidence interval from 0.23 to 0.89. The log-rank p-value is 0.018. These values describe the sponsor-reported statistical comparison, not an individual participant’s chance of relapse or a direct measure of how long any one person will remain relapse-free.

The estimate must be interpreted with the specified event definition and censoring rule. Relapses had to meet neurological symptom criteria, persist for >24 hours, and not be explained by listed confounding factors. Participants without a relapse were censored at the end of their double-blind period. The result therefore applies to the protocol-defined endpoint and follow-up conditions, not to every possible neurological symptom or outcome.

The hazard ratio’s direction depends on its reference coding. Although the group order is given, the registry does not explain the coding convention in its description. The estimate should be quoted with its comparison labels rather than assigned a directional conclusion beyond what the record states. The confidence interval also shows uncertainty around the estimate, and the incomplete median summaries limit interpretation of median event times.

The primary finding does not answer every question raised by the trial. Secondary outcomes include pain, fatigue, relapse-free rate, annualized relapse rate, and health-status domains, but their results and multiplicity handling are not fully described in the record. The safety figures are reported separately and should not be conflated with the primary efficacy endpoint.

In general, statistical evidence is not the same as evidence of clinical importance, and a p-value does not establish that a result is true or important. The posted result is best understood as a comparison under the trial’s design, endpoint definition, and analysis, with uncertainty and reporting limitations kept in view. It is not medical advice and does not determine what treatment is appropriate for any individual.

17. Why This Trial Matters Statistically

This trial illustrates the interpretive demands of time-to-event analysis in a randomized clinical study. The primary outcome is not simply whether a participant had a relapse; it is the time until the first qualifying relapse, with follow-up ending for some participants before any event was observed. That structure makes censoring central to the estimand and to how the results should be read.

The record also distinguishes several statistical summaries that are easy to conflate. The median event time describes a point in an estimated event-time distribution. The hazard ratio compares event rates over time under the reported analysis. The log-rank p-value measures compatibility with a null hypothesis under its assumptions. The confidence interval communicates uncertainty around the hazard-ratio estimate. None of these is interchangeable with a simple relapse count, and none alone captures clinical importance.

The report contains an informative but incomplete set of results: the hazard ratio and its confidence interval are available, while median estimates and interval limits are not all fully reported. This offers a practical example of why uncertainty should be assessed from the full result rather than from a point estimate alone. It also highlights why group ordering and reference coding matter when interpreting a ratio.

Finally, the record includes multiple secondary measures and safety periods, while key details about multiplicity, missing-data methods, and analysis assumptions are not reported. Statistical interpretation therefore involves not only reading the available estimate but also identifying what the registry record does not establish. Keeping the endpoint, analysis population, censoring, and reporting gaps explicit makes the conclusion more faithful to the evidence recorded.

18. 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.

The registry record gives limited detail on the primary analysis beyond naming the log-rank method and reporting a hazard ratio. It does not explain the hazard-ratio reference coding, describe diagnostics for the analysis assumptions, or report sensitivity analyses. This limits independent evaluation of how robust the estimate is to alternative analysis choices.

The median time-to-relapse entries are incomplete: the placebo, then satralizumab group has no upper confidence limit reported, and the satralizumab, then satralizumab group has no numeric median or upper limit reported. The record does not explain these unavailable values. This constrains interpretation of the median summaries in addition to the uncertainty shown by the hazard-ratio interval.

The registry describes censoring of participants without relapse at the end of the double-blind period, but does not report additional missing-data procedures, reasons for other loss of observation, or sensitivity analyses. It also does not report the original sample-size justification, interim analysis plans, or a multiplicity strategy for the primary and secondary outcomes. These details are not reported in the registry record and should not be inferred.

The outcome is specific to protocol-defined relapse during the double-blind period, rather than all symptoms or all aspects of NMO or NMOSD. The registry lists many secondary outcomes, but the result information does not support a complete assessment of their findings or their relationship to the primary analysis. Safety data cover double-blind and open-label periods separately, and the record does not provide exposure-adjusted rates or further comparative context.

These limitations describe the boundaries of interpretation of the registry entry. They do not by themselves demonstrate that the trial was conducted improperly or that the reported analysis is invalid.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Efficacy and Safety Study of Satralizumab (SA237) as Monotherapy to Treat Participants With Neuromyelitis Optica (NMO) and Neuromyelitis Optica Spectrum Disorder (NMOSD) — NCT02073279.
• Registry record last updated 2023-03-01; status last verified 2023-01.
• Results first posted 2020-12-31.
• 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.

22. Record Summary

NCT02073279 is a completed randomized, double-blind, parallel-group Phase 3 study sponsored by Hoffmann-La Roche. It evaluated satralizumab monotherapy against placebo in participants with NMO or NMOSD. Enrollment was 95, and the registry records study dates from 2014-08-05 to primary completion on 2018-10-12.

The primary endpoint was time to first protocol-defined relapse during the double-blind period, analyzed in the intention-to-treat population. The reported log-rank hazard ratio for placebo, then satralizumab versus satralizumab, then satralizumab was 0.45, with a 95% confidence interval from 0.23 to 0.89 and a reported p-value of 0.018. The record also includes median time summaries, secondary outcomes, and adverse-event results. Further details on multiplicity, missing-data methods, and analysis assumptions are not reported in the registry record.