← Clinical Trials
Stroke Phase 3 Double-Blind NCT02072226

PRISMS: Complete Statistical Analysis of Alteplase in Mild Stroke

An independent statistical review of the randomized phase 3 PRISMS trial evaluating alteplase in participants with mild stroke, with emphasis on the Day 90 modified Rankin Scale endpoint, absolute risk difference, odds-ratio analyses, and reported safety outcomes.

Trial status: Terminated  ·  Enrollment: 313  ·  Primary completion: March 22, 2017
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.

1. Trial at a Glance

PRISMS was a randomized, double-blind, parallel phase 3 trial evaluating alteplase in participants with mild stroke. The registry reports 313 enrolled participants, two arms, a binary primary endpoint at Day 90, and six posted statistical analyses.

313
Enrolled
Randomized trial
2
Arms
Parallel design
-1.10
Primary risk difference
95% CI -9.44 to 7.25
Day 90
Primary endpoint
mRS 0 or 1
FeaturePRISMS
Trial namePRISMS
Brief titleA Study of the Efficacy and Safety of Alteplase in Participants With Mild Stroke
PhasePhase 3
ConditionStroke
DesignRandomized, parallel, double-masked, treatment-focused trial
Enrollment313
Primary endpoint typeBinary
Results postedYes
Outcome measures posted9
Statistical analyses posted6
Lead sponsorGenentech, Inc.
Sponsor typeIndustry
Trial statusTerminated
ClinicalTrials.govNCT02072226

2. Clinical Question

The statistical question is whether the randomized comparison between alteplase plus aspirin placebo and alteplase placebo plus aspirin differs in the percentage of participants achieving a favorable Day 90 functional outcome defined by a modified Rankin Scale (mRS) score of 0 or 1.

Population

Participants with mild stroke enrolled in the PRISMS phase 3 trial.

Intervention

Alteplase with aspirin placebo.

Comparator

Alteplase placebo with aspirin.

Primary question

What is the difference between randomized groups in the percentage of participants with an mRS score of 0 or 1 at Day 90?

3. Trial Design

01
Randomize313 participants
02
Parallel arms2 randomized groups
03
Double maskDouble-masked design
04
Day 90mRS assessment
05
AnalysisITT and safety populations
ARM A

Alteplase + Aspirin Placebo

  • Alteplase
  • Aspirin placebo
  • Serious adverse events: 40/154
ARM B

Alteplase Placebo + Aspirin

  • Alteplase placebo
  • Aspirin
  • Serious adverse events: 20/153
Why the double-masked structure matters. Blinding is especially relevant when an outcome has an element of investigator assessment. The primary endpoint was based on an mRS score determined by the investigator. A double-masked design is therefore an important design feature when interpreting the possibility of assessment-related differences between treatment groups.

4. Trial Timing and Early Termination

May 31, 2014

Study start

The registry lists May 31, 2014 as the study start date.

March 22, 2017

Primary completion

The registry lists March 22, 2017 as the primary completion date.

Terminated

Study status

The trial is listed as terminated. The registry caveat states that the study was terminated early due to low enrollment, leading to a smaller number of participants analyzed.

Early termination is statistically important. The registry specifically identifies low enrollment as the reason for early termination. A smaller-than-intended randomized population generally means less statistical information and wider uncertainty around estimated treatment differences. In PRISMS, the primary 95% confidence interval extends from a negative to a positive risk difference, illustrating the importance of considering precision rather than looking only at the point estimate.

5. Primary Endpoint

EndpointRegistry definitionTime frameType
Percentage of Participants With a Modified Rankin Scale (mRS) Score of 0 or 1 at Day 90 mRS score was determined by the investigator. The mRS is a 7 point scale (0-6) with 0: No symptoms at all, 1: No significant disability despite symptoms, able to carry out all usual duties and activities. Day 90 Binary

The primary endpoint converts the seven-level mRS into a binary outcome: whether the participant has a score of 0 or 1 at Day 90. This creates a clinically interpretable responder-type endpoint, but it also discards information contained in mRS scores above 1. The registry's posted primary analysis therefore focuses on the between-group difference in the percentage meeting this binary criterion.

6. Results: Primary Endpoint

The registry reports an intention-to-treat analysis including all randomized participants. The comparison was between alteplase plus aspirin placebo and alteplase placebo plus aspirin.

Adjusted risk difference

-1.10 percentage points

95% CI: -9.44 to 7.25   ·   Two-sided

Analysis population: Intent-to-Treat population

Primary endpointEffect measureEstimate95% CI
mRS score 0 or 1 at Day 90 Adjusted Risk Difference -1.10 -9.44 to 7.25
Clinical Biostats interpretation

The reported adjusted risk difference of -1.10 means that the estimated percentage of participants achieving an mRS score of 0 or 1 at Day 90 was 1.10 percentage points lower in the alteplase-plus-aspirin-placebo group than in the alteplase-placebo-plus-aspirin group, according to the posted analysis.

The estimate does not mean that treatment reduced every participant's probability by 1.10 percentage points, nor does it describe an individual patient's expected outcome. It is a population-level between-group effect estimate.

The 95% confidence interval of -9.44 to 7.25 is wide enough to include both negative and positive risk differences. Thus, the point estimate alone provides an incomplete description of the evidence. The interval indicates substantial statistical uncertainty around the estimated difference, which is particularly relevant given the registry's statement that the study was terminated early because of low enrollment.

The registry does not report a p-value for this primary analysis. A p-value, when available, would address evidence against a specified null hypothesis; it would not measure the magnitude or clinical importance of the treatment effect. The risk difference and its confidence interval provide the more direct description of the estimated effect and its precision.

Reading the confidence interval

Risk-difference scale
Risk Difference = Risk in Alteplase Group − Risk in Control Group

On this scale, 0 represents no estimated difference between groups. Negative values favor a higher percentage in the control group for this particular subtraction; positive values favor a higher percentage in the alteplase group.

The reported interval extends across zero. That matters because zero corresponds to no between-group difference on the risk-difference scale. The interval therefore does not restrict the plausible treatment-effect range to only one direction under the stated confidence framework.

This is a binary endpoint rather than a time-to-event endpoint. Kaplan-Meier curves, hazard ratios, and proportional-hazards assumptions are therefore not required to describe the posted primary analysis.

7. Secondary Endpoint Results

The registry posts five secondary statistical analyses in addition to the primary analysis. These results use the intention-to-treat population for efficacy outcomes and the safety population for the intracranial hemorrhage outcomes.

Distribution of Participants Across the Ordinal mRS

Odds ratio

0.810

95% CI: 0.527 to 1.244   ·   Two-sided

Analysis population: ITT

The odds ratio compares the odds associated with the reported mRS distribution between the randomized groups under the posted analysis. The registry does not report the statistical method used to obtain this odds ratio.

Global Favorable Recovery on mRS, NIHSS, BI, and GOS

Odds ratio

0.858

95% CI: 0.529 to 1.393   ·   Two-sided

Analysis population: ITT

This endpoint combines favorable recovery across the mRS, NIHSS, BI, and GOS measures. The posted estimate is an odds ratio of 0.858 with a 95% confidence interval from 0.529 to 1.393. The registry does not report the formal statistical method used to generate the estimate.

Symptomatic Intracranial Hemorrhage

Difference in percentages

3.25 percentage points

95% CI: 0.75 to 7.38   ·   Two-sided

Analysis population: Safety population

The reported difference in percentages was 3.25 percentage points for symptomatic intracranial hemorrhage within 36 hours after study drug administration on Day 1. The safety population included all participants who received any amount of study drug.

Any ICH — Site Report

Difference in percentages

4.53 percentage points

95% CI: -0.34 to 10.07   ·   Two-sided

Within 36 hours after study drug administration on Day 1

For any intracranial hemorrhage reported by site, the estimated difference in percentages was 4.53 percentage points, with a 95% confidence interval from -0.34 to 10.07.

Any ICH — Central Reader

Difference in percentages

3.87 percentage points

95% CI: -1.23 to 9.49   ·   Two-sided

Within 36 hours after study drug administration on Day 1

The corresponding analysis using any ICH reported by a central reader produced a difference in percentages of 3.87 percentage points, with a 95% confidence interval from -1.23 to 9.49.

Secondary endpointEffect measureEstimate95% CIPopulation
Distribution of participants across the ordinal mRSOdds Ratio0.8100.527 to 1.244ITT
Global favorable recovery on mRS, NIHSS, BI, and GOSOdds Ratio0.8580.529 to 1.393ITT
Symptomatic ICHDifference in percentages3.250.75 to 7.38Safety
Any ICH, site reportedDifference in percentages4.53-0.34 to 10.07Safety
Any ICH, central readerDifference in percentages3.87-1.23 to 9.49Safety
Do not treat these secondary estimates as interchangeable. The endpoints use different outcome definitions, effect measures, and analysis populations. In particular, the functional outcomes use the ITT population, whereas the ICH analyses use the safety population. The registry also does not report formal statistical methods for these analyses.

8. Safety Results

The registry provides serious adverse event counts by randomized arm in terms of affected participants over the corresponding number at risk.

Safety measureAlteplase + Aspirin PlaceboAlteplase Placebo + Aspirin
Serious adverse events40/15420/153
Serious adverse events: affected / at risk
Alteplase + Aspirin Placebo
40 / 154
Alteplase Placebo + Aspirin
20 / 153

The serious-adverse-event figures should be kept separate from the efficacy analysis. Safety describes events observed among participants who received study drug, while the primary efficacy comparison is based on randomized assignment in the ITT population. The registry does not provide a formal statistical comparison for these serious adverse event counts in the ClinicalTrials.gov record.

Population distinction: the posted ICH analyses explicitly identify the safety population as all participants who received any amount of study drug. The primary efficacy analysis instead uses the ITT population, which included all randomized participants.

9. Statistical Methodology

What the registry reports

The posted primary analysis uses an adjusted risk difference for the binary Day 90 mRS 0-or-1 endpoint and identifies the ITT population as the analysis population. The registry does not name the statistical method used to produce the adjusted estimate.

The secondary analyses likewise identify their effect measures—odds ratios for two functional outcomes and differences in percentages for ICH outcomes—but the ClinicalTrials.gov record does not report the formal methods used to calculate those estimates.

Primary effect measure
Adjusted Risk Difference = Estimated Risk in Alteplase Group − Estimated Risk in Control Group

The posted estimate is -1.10, with a two-sided 95% confidence interval from -9.44 to 7.25.

Why risk difference is useful

Risk difference is an absolute effect measure. Unlike an odds ratio, it remains directly expressed in percentage points of the outcome probability. For a binary endpoint such as mRS 0 or 1 at Day 90, this makes the measure straightforward to connect to the proportion of participants achieving the specified outcome.

The word adjusted is important. An adjusted risk difference is not necessarily the same as a simple subtraction of two raw observed percentages. Because the registry does not report the adjustment model or covariates in the ClinicalTrials.gov record, the precise calculation of the -1.10 estimate cannot be reconstructed from the posted information alone.

Odds ratios

Two secondary analyses use odds ratios. An odds ratio compares odds rather than probabilities. The distinction becomes important when an outcome is not rare: an odds ratio should not automatically be read as though it were a risk ratio or a percentage-point difference.

Conceptual odds ratio
OR = (Odds in Alteplase Group) / (Odds in Control Group)

An OR of 1 represents equal odds. Values below 1 and above 1 indicate different directions depending on how the outcome and treatment contrast are coded.

Confidence intervals

Each posted statistical estimate is accompanied by a two-sided 95% confidence interval. The interval provides information about precision and sampling uncertainty under the analysis framework. It should not be interpreted as the range of outcomes that individual participants will experience.

Intention-to-treat analysis

The primary analysis population is explicitly the Intent-to-Treat population, defined in the ClinicalTrials.gov record as including all randomized participants. ITT analysis preserves treatment assignment as the basis of comparison and therefore maintains the principal randomized structure of the trial.

Safety population

The ICH analyses use a safety population consisting of participants who received any amount of study drug. This population definition is exposure-based rather than purely assignment-based. That distinction is methodologically important because the population used to assess treatment exposure and safety can differ from the population used for randomized efficacy comparisons.

10. Statistical Methods Explained

Why use an intention-to-treat population?

ITT analysis evaluates randomized participants according to their randomized assignment. This maintains the treatment comparison created by randomization. In PRISMS, the registry explicitly states that all randomized participants were included in the ITT population. The approach is particularly relevant when interpreting an estimate as a randomized treatment comparison rather than as a comparison of people who actually completed or adhered to treatment.

What does a risk difference of -1.10 mean?

A risk difference of -1.10 means the estimated probability of the specified favorable binary outcome was 1.10 percentage points lower in the alteplase group under the posted treatment contrast. The value is an absolute difference, not a relative percentage change. It also does not mean that an individual participant's outcome probability changes by exactly 1.10 percentage points.

Why does the confidence interval matter more than the point estimate alone?

The point estimate is only one summary of the data. The 95% confidence interval from -9.44 to 7.25 demonstrates that the estimate is imprecise enough that the interval includes both negative and positive values. A reader who considered only -1.10 would miss much of the uncertainty conveyed by the analysis.

What does an odds ratio of 0.810 mean?

An odds ratio of 0.810 indicates that the estimated odds under the treatment contrast were 0.810 times the comparison odds, according to the posted analysis. It does not mean that the probability was 19.0% lower, because odds and probabilities are different quantities. The accompanying 95% confidence interval, 0.527 to 1.244, is essential for assessing the precision of the estimate.

Why are the ICH analyses in the safety population?

Safety outcomes are naturally linked to exposure to study treatment. The registry defines the safety population for the ICH analyses as participants who received any amount of study drug. This differs from the ITT population used for the primary efficacy analysis because safety analysis asks a somewhat different question: what adverse events were observed among participants exposed to study treatment?

Why is the early termination statistically important?

The registry states that low enrollment caused early termination and resulted in a smaller number of participants analyzed. Smaller information sets generally produce less precise estimates. In PRISMS, the primary confidence interval extends from -9.44 to 7.25, so the uncertainty around the estimated risk difference is an important part of the result rather than a secondary footnote.

11. Understanding the Primary Result

Effect estimate

The adjusted risk difference of -1.10 is an absolute, population-level comparison for the percentage of participants with an mRS score of 0 or 1 at Day 90. Its interpretation depends on the direction of the treatment contrast specified by the registry: alteplase plus aspirin placebo versus alteplase placebo plus aspirin.

Precision

The 95% CI of -9.44 to 7.25 spans 16.69 percentage points from its lower to upper endpoint. This is a wide interval relative to the point estimate, indicating that the observed estimate should not be treated as a highly precise measurement of the underlying treatment difference.

What the result does not establish

The posted estimate does not by itself establish a treatment recommendation, does not quantify an individual patient's probability of recovery, and does not provide a measure of benefit or harm for every participant. It summarizes the randomized comparison for one prespecified binary functional endpoint.

Why no p-value is reported here

The ClinicalTrials.gov record contains an estimate and two-sided 95% confidence interval but no p-value. Therefore, a p-value should not be reconstructed from the confidence interval or introduced from another source. The confidence interval is the appropriate reported measure of uncertainty for this page.

12. Interpreting the Secondary Odds Ratios

EndpointOR95% CIInterpretive point
Distribution of participants across the ordinal mRS 0.810 0.527 to 1.244 The interval spans 1, the null value for an odds ratio.
Global favorable recovery on mRS, NIHSS, BI, and GOS 0.858 0.529 to 1.393 The interval spans 1, so the point estimate should not be interpreted without its uncertainty interval.

For odds ratios, 1 is the null value. Both posted intervals include 1. That does not mean the true effect is exactly 1; rather, the confidence intervals encompass the no-difference value under the stated statistical framework.

There is also a conceptual distinction between these secondary endpoints. The first describes the distribution of participants across the ordinal mRS, whereas the second defines global favorable recovery using mRS, NIHSS, BI, and GOS. They should not be collapsed into a single efficacy measure merely because both are reported as odds ratios.

13. Interpreting the ICH Results

ICH endpointDifference in percentages95% CI
Symptomatic ICH3.250.75 to 7.38
Any ICH, site reported4.53-0.34 to 10.07
Any ICH, central reader3.87-1.23 to 9.49

These three results illustrate why endpoint definition and ascertainment method matter. The symptomatic ICH analysis is a more specific safety endpoint, while the two any-ICH analyses differ according to whether the hemorrhage was reported by the site or by a central reader.

The two any-ICH confidence intervals cross zero, whereas the reported confidence interval for symptomatic ICH is entirely above zero. The registry supplies these estimates and intervals but does not supply p-values or a formal multiplicity framework for these secondary analyses. Accordingly, the results should be described as the posted statistical estimates rather than converted into an unsupported hierarchy of findings.

Multiplicity caution. PRISMS has multiple posted outcome analyses, but the ClinicalTrials.gov record does not state a multiplicity-adjustment procedure or an alpha allocation across these secondary endpoints. Multiple estimates should therefore not automatically be interpreted as if each represented a separate confirmatory hypothesis test.

14. Analysis Populations

PopulationDefinition / role
Intent-to-Treat Included all randomized participants. Used for the primary efficacy analysis and the two posted functional secondary analyses.
Safety Population Included all participants who received any amount of study drug. Used for the posted symptomatic ICH and any-ICH analyses.

This distinction is more than a technical labeling issue. The ITT analysis estimates the effect associated with assignment to the randomized strategy, whereas the safety population is defined by actual exposure to study drug. Mixing these populations could create confusion about what a reported treatment difference represents.

15. Limitations and Interpretation Issues

16. Why This Trial Matters Statistically

PRISMS is a useful teaching case because it shows how the statistical interpretation of a randomized trial depends not only on a point estimate but also on endpoint construction, analysis population, confidence intervals, safety definitions, and the amount of information actually accumulated.

ConceptHow it appears in PRISMS
RandomizationParticipants were assigned in a randomized parallel phase 3 design.
BlindingThe trial used a double-masked design.
Intention-to-treat analysisThe primary efficacy analysis included all randomized participants.
Binary endpointThe primary endpoint was the percentage with an mRS score of 0 or 1 at Day 90.
Risk differenceThe primary effect measure was an adjusted risk difference of -1.10.
Confidence intervalThe primary estimate was accompanied by a two-sided 95% CI of -9.44 to 7.25.
Odds ratioTwo secondary functional analyses reported odds ratios.
Safety populationICH analyses included participants who received any amount of study drug.
Early terminationLow enrollment led to early termination and a smaller number of participants analyzed.
Multiple analysesSix statistical analyses were posted across the primary and secondary outcomes.

17. A Practical Framework for Reading PRISMS

Start with the endpoint

The primary question is binary: whether the participant had an mRS score of 0 or 1 at Day 90. This determines the natural effect measures and the null value for the risk difference.

Then read the estimate

The adjusted risk difference is -1.10 percentage points. This is the estimated absolute contrast between the randomized groups under the posted analysis.

Then read the interval

The 95% CI extends from -9.44 to 7.25. The width and crossing of zero are central to understanding the uncertainty.

Then inspect the population

The primary analysis is ITT, whereas the ICH analyses use the safety population. The population definition changes the interpretation of the estimate.

This sequence is broadly useful in clinical-trial statistics: endpoint → analysis population → effect measure → point estimate → confidence interval → design limitations. It reduces the risk of treating a single number as if it were the entire statistical result.

18. Statistical Concepts in This Trial

Learn more about the methods and concepts used in this trial:

19. Related Statistical Calculators

20. Sources

Continue through Clinical Biostats

Use the related tutorials and statistical calculators to explore the methods behind randomized clinical-trial results, binary endpoints, confidence intervals, odds ratios, and absolute treatment effects.

21. Record Summary

PRISMS provides a compact example of how to interpret a randomized clinical trial when the central result is a binary functional endpoint and the available sample is limited by early termination. The primary Day 90 outcome was analyzed in the ITT population using an adjusted risk difference of -1.10, with a two-sided 95% confidence interval of -9.44 to 7.25. The registry also reports odds-ratio analyses for functional outcomes and percentage differences for intracranial hemorrhage outcomes, with the latter based on the safety population.

The most important statistical lesson is that the point estimate cannot be separated from its uncertainty and design context. The primary confidence interval spans zero, while the study was terminated early because of low enrollment. Secondary analyses use different endpoints and, for safety outcomes, a different analysis population. The ClinicalTrials.gov record does not report the formal statistical methods behind the adjusted estimate, nor does it provide a p-value for the primary analysis or a multiplicity-adjustment strategy for the posted secondary analyses.

Clinical Biostats methodology: A trial-results page should distinguish reported evidence from statistical interpretation. For PRISMS, that means preserving the registry's endpoint definitions, effect estimates, confidence intervals, analysis populations, and early-termination caveat while explaining what each statistical measure does—and does not—tell the reader.