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
ORIGINAL was a randomized, parallel-group, single-masked phase 3 trial comparing tenecteplase with alteplase in Chinese patients with stroke. The primary endpoint was the percentage of participants with a modified Rankin Scale (mRS) score of 0 or 1 at Day 90±7 days, analyzed as a binary outcome under a prespecified non-inferiority framework.
| Feature | ORIGINAL |
|---|---|
| Trial name | ORIGINAL |
| ClinicalTrials.gov identifier | NCT04915729 |
| Phase | Phase 3 |
| Condition | Stroke |
| Therapeutic area | Cardiovascular; Neurology |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Single |
| Primary purpose | Treatment |
| Enrollment | 1,489 |
| Arms | 2 |
| Interventions | Tenecteplase; alteplase |
| Status | Completed |
| Start | June 22, 2021 |
| Primary completion | October 8, 2023 |
| Lead sponsor | Boehringer Ingelheim |
2. Clinical Question
The central question was whether tenecteplase could provide a non-inferior rate of favorable functional outcome compared with alteplase, using the percentage of participants with an mRS score of 0 or 1 at Day 90±7 days as the primary endpoint.
Population
Chinese patients with stroke enrolled in the randomized phase 3 ORIGINAL trial.
Intervention
Tenecteplase.
Comparator
Alteplase, designated as the active control group.
Primary question
Is the risk ratio for favorable functional outcome with tenecteplase sufficiently high to satisfy the prespecified non-inferiority criterion relative to alteplase?
3. Trial Design
Tenecteplase
- Tenecteplase was the investigational treatment.
- Efficacy comparisons were made against the alteplase active control group.
- The primary efficacy analysis used treatment assignment as randomized.
Alteplase
- Alteplase was the active comparator.
- The registry identifies this as the active control group.
- The same randomized comparison framework was used for the primary endpoint.
4. Analysis Populations
The registry distinguishes the efficacy Full Analysis Set from the Safety Set. That distinction matters because the two populations serve different statistical purposes.
| Population | Definition and role |
|---|---|
| Full Analysis Set (FAS) | Included all randomized patients who received any dose of treatment. Treatment assignment was as randomized. This was the primary analysis set for presentation of efficacy according to the intention-to-treat principle. |
| Safety Set (SS) | Included all patients who were randomized and received any dose of treatment. It was the main analysis set for presentation of safety, with patients analyzed according to the actual treatment. |
This distinction preserves two different inferential questions. The FAS keeps the efficacy comparison anchored to randomized assignment, while the Safety Set describes safety according to the treatment actually received.
5. Endpoints
| Endpoint | Time frame | Type | Analysis |
|---|---|---|---|
| Percentage of Participants With Modified Rankin Scale (mRS) Score of 0 or 1 | At Day 90±7 days | Binary | Modified Poisson regression; risk ratio |
| Percentage of Participants With Major Neurological Improvement | At 24 hours | Binary | Linear regression; risk ratio |
| Percentage of Participants With mRS Score of 0-2 | At Day 90 | Binary | Linear regression; risk ratio |
| Change From Baseline of NIHSS Score | At baseline and at Day 90 | Continuous | Mixed models analysis; mean difference |
| Distribution of Modified Rankin Scale (mRS) | At Day 90 | Binary in registry classification | Logistic regression; odds ratio |
| Percentage of Participants With Barthel Index Score ≥95 | Up to 90 days | Binary | Linear regression; risk ratio |
| Percentage of Participants With Symptomatic Intracerebral Haemorrhage (sICH) Per ECASS III Definition During On-treatment Period | Up to 7 days | Binary | Suissa-Shuster test; risk ratio |
| Percentage of Participants Who Died by Day 90 | Up to 90 days | Binary | Chi-squared test; risk ratio |
| Percentage of Participants With mRS Score of 5 or 6 | At Day 90 | Binary | Linear regression; risk ratio |
Primary endpoint definition
The primary endpoint was the percentage of participants with an mRS score of 0 or 1 at Day 90±7 days. The mRS is a standardized measure of disability after stroke, increasing from 0, representing no symptoms at all, to 6, representing death. The registry states that the primary endpoint was measured at Day 90 and that percentages were rounded to the nearest digits.
6. Statistical Methodology
Primary analysis: modified Poisson regression
The primary endpoint is binary, so the analysis focuses on the probability of achieving the favorable mRS category rather than on a continuous change in score. The registry reports a modified Poisson regression model and expresses the treatment effect as a risk ratio.
The analysis specified treatment as the main effect and included baseline NIHSS, age, and time to drug administration since stroke symptoms onset as continuous covariates.
Covariate adjustment
The primary analysis adjusted for baseline NIHSS, age, and time to drug administration since stroke symptoms onset. Covariate adjustment can improve precision by accounting for prognostic variables that are related to the endpoint, while the randomized treatment comparison remains the principal contrast.
Intention-to-treat principle
The registry describes the Full Analysis Set as the primary efficacy population and explicitly states that treatment assignment was analyzed as randomized. This is consistent with the intention-to-treat principle: the treatment comparison remains tied to the original randomization rather than being redefined according to subsequent treatment behavior.
Other binary-outcome analyses
Several secondary binary endpoints were analyzed with regression-based methods and expressed as risk ratios. These included major neurological improvement, mRS 0-2, Barthel Index ≥95, and mRS 5 or 6. Mortality was analyzed using a chi-squared test, while symptomatic intracerebral haemorrhage was analyzed using the Suissa-Shuster test.
Continuous outcome analysis
Change from baseline in NIHSS was analyzed using a mixed models analysis. The registry identifies baseline NIHSS, age, and time to administration since stroke symptoms onset as linear covariates, with treatment as a fixed effect. The treatment contrast was the difference in least-squares mean between tenecteplase and alteplase.
Ordinal mRS analysis
The distribution of the mRS at Day 90 was analyzed with logistic regression, with an odds ratio reported as the effect measure. The registry describes this as an assumption-free ordinal analysis comparing tenecteplase with alteplase.
7. Primary Result
The registry reports a formal statistical analysis for the primary endpoint: percentage of participants with an mRS score of 0 or 1 at Day 90±7 days.
Risk ratio for favorable functional outcome
95% CI: 0.9678–1.0915
Prespecified non-inferiority margin for the risk ratio: 0.937
Analysis population: Full Analysis Set
| Feature | Primary analysis |
|---|---|
| Endpoint | Percentage of participants with mRS score of 0 or 1 |
| Time frame | Day 90±7 days |
| Population | Full Analysis Set |
| Comparison | Tenecteplase vs alteplase |
| Method | Modified Poisson regression model |
| Effect measure | Risk ratio |
| Estimate | 1.0278 |
| 95% CI | 0.9678–1.0915 |
| Non-inferiority margin | 0.937 |
| Covariates | Baseline NIHSS, age, and time to drug administration since stroke symptoms onset |
The risk ratio of 1.0278 means that the estimated probability of achieving an mRS score of 0 or 1 was 1.0278 times as high with tenecteplase as with alteplase under the fitted adjusted model. Expressed as a relative comparison, this corresponds to an estimated 2.78% higher probability on the risk-ratio scale.
The estimate itself is not a statement that 2.78% more patients necessarily achieved the endpoint. A risk ratio is relative; an absolute percentage-point difference would require the corresponding event percentages in each group.
The 95% confidence interval of 0.9678 to 1.0915 describes uncertainty around the estimated risk ratio. It includes 1, so the interval is compatible with both a modest relative reduction and a modest relative increase in the probability of favorable outcome.
The non-inferiority question is different from a conventional superiority question. The prespecified margin was 0.937. The lower confidence limit, 0.9678, is above that margin, so the reported interval is consistent with the prespecified non-inferiority criterion. A non-inferiority conclusion does not mean that the two treatments have identical effects, nor does it establish superiority.
No p-value is reported for this primary analysis in the registry's posted statistical analysis. A p-value would address evidence against a specified null hypothesis; it would not quantify the magnitude of the treatment effect or replace the non-inferiority margin.
Why the non-inferiority margin matters
For a risk-ratio non-inferiority analysis in which values below the margin would indicate an unacceptably lower treatment effect, the lower confidence bound is the critical comparison. The observed lower bound remains above the prespecified boundary.
8. Secondary Efficacy Results
| Endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Major neurological improvement at 24 hours | Risk ratio | 1.0671 | 0.9573–1.1895 | 0.2412 |
| mRS score 0-2 at Day 90 | Risk ratio | 1.0078 | 0.9614–1.0564 | 0.7480 |
| Change from baseline in NIHSS at Day 90 | Mean difference | -0.45 | -1.40–0.50 | 0.3511 |
| Distribution of mRS at Day 90 | Odds ratio | 1.0418 | Not reported | 0.4806 |
| Barthel Index score ≥95 up to 90 days | Risk ratio | 1.0189 | 0.9635–1.0774 | 0.5116 |
| mRS score 5 or 6 at Day 90 | Risk ratio | 0.9215 | 0.6578–1.2908 | 0.6345 |
Major neurological improvement at 24 hours
Risk ratio
95% CI: 0.9573–1.1895 · P = 0.2412
Endpoint: major neurological improvement at 24 hours
The risk ratio of 1.0671 corresponds to an estimated 6.71% higher probability of major neurological improvement on the relative-risk scale. The confidence interval ranges from 0.9573 to 1.1895 and includes 1. The registry reports this endpoint as analyzed with linear regression and describes a log-binomial regression model in the analysis notes, with baseline NIHSS, age, and time to drug administration since stroke symptoms onset as covariates.
The estimate describes a relative probability, not an absolute percentage-point increase. The confidence interval is more informative about precision than the point estimate alone: it includes values below and above 1, so the reported estimate should not be treated as establishing a definite relative increase in the endpoint.
The P-value of 0.2412 does not measure how large the risk ratio is. It describes the evidence against the statistical null hypothesis used for this analysis. The size and direction of the estimated effect are conveyed by the risk ratio and its confidence interval.
mRS score 0-2 at Day 90
Risk ratio
95% CI: 0.9614–1.0564 · P = 0.7480
The estimated risk ratio of 1.0078 is close to 1, indicating little relative separation between treatment groups in the fitted estimate for the mRS 0-2 endpoint. The 95% confidence interval of 0.9614–1.0564 is relatively narrow around 1 compared with several of the other secondary binary outcomes.
The P-value of 0.7480 should not be interpreted as a probability that the treatments are equivalent. It is a hypothesis-testing quantity, whereas the confidence interval describes the range of effect estimates compatible with the statistical model and data.
Change from baseline in NIHSS
Adjusted mean difference in final values
95% CI: -1.40 to 0.50 · P = 0.3511
The reported mean difference was -0.45 for tenecteplase versus alteplase. The registry describes this as the difference in least-squares mean, with treatment as a fixed effect and baseline NIHSS, age, and time to administration since stroke symptoms onset as linear covariates.
The confidence interval spans -1.40 to 0.50, crossing zero. Thus the interval includes both a modestly lower and a modestly higher adjusted mean final NIHSS value for tenecteplase relative to alteplase. The P-value of 0.3511 is a hypothesis-testing measure and should not be read as an effect-size measure.
Distribution of mRS at Day 90
Odds ratio
P = 0.4806
Analysis: logistic regression
An odds ratio of 1.0418 is a relative comparison of odds rather than probabilities. It should therefore not be interpreted as a 4.18% increase in the probability of a particular mRS outcome.
The registry reports a P-value of 0.4806 but does not provide a confidence interval for this posted analysis. The analysis is described as an assumption-free ordinal analysis comparing tenecteplase with alteplase.
Barthel Index score ≥95
Risk ratio
95% CI: 0.9635–1.0774 · P = 0.5116
The risk ratio of 1.0189 represents an estimated 1.89% higher probability on the relative-risk scale. The 95% confidence interval, 0.9635–1.0774, includes 1 and therefore encompasses both a modest relative reduction and a modest relative increase.
The registry states that no imputation was performed for patients missing the Barthel Index score at Day 90. Consequently, this endpoint analysis is based on the observed cases available for the analysis rather than an imputed endpoint dataset.
mRS score 5 or 6 at Day 90
Risk ratio
95% CI: 0.6578–1.2908 · P = 0.6345
The estimated risk ratio of 0.9215 corresponds to an estimated 7.85% lower probability on the relative-risk scale for the tenecteplase group. The confidence interval is considerably wider than the point estimate, extending from 0.6578 to 1.2908.
Because the interval includes 1, the estimate alone does not establish a treatment-group difference for this endpoint. The P-value of 0.6345 is consistent with the same statistical interpretation but does not quantify the magnitude or clinical importance of the effect.
9. Safety Results
The registry reports serious adverse events by treatment arm and provides a separate safety analysis for symptomatic intracerebral haemorrhage and death.
| Safety measure | Tenecteplase | Alteplase |
|---|---|---|
| Serious adverse events, affected / at risk | 116 / 732 | 115 / 736 |
| Symptomatic intracerebral haemorrhage, sICH, up to 7 days | RR 1.005; 95% CI 0.370–2.701; P = 1.000 | |
| Death by Day 90 | RR 0.795; 95% CI 0.513–1.232; P = 0.303 | |
Serious adverse events
Serious adverse events affected 116 of 732 participants in the tenecteplase treatment group and 115 of 736 participants in the alteplase active control group. These counts describe the number affected and the corresponding at-risk population reported for each arm.
The two arm-level counts are close in absolute terms, but a direct comparison of the counts alone does not provide the full statistical analysis. Safety interpretation also depends on the denominator, the event definition, follow-up, and whether the endpoint is analyzed according to actual treatment exposure.
Symptomatic intracerebral haemorrhage
Risk ratio
95% CI: 0.370–2.701 · P = 1.000
Suissa-Shuster test; safety set
The estimated risk ratio of 1.005 is close to 1, but the 95% confidence interval is very wide, extending from 0.370 to 2.701. This means the point estimate is substantially less informative than it might appear without the interval: the data are compatible with a broad range of relative effects.
The P-value of 1.000 does not mean that the treatment effects are proven identical. It is a result of the statistical test used for this comparison. The wide confidence interval is particularly important when interpreting an uncommon safety endpoint.
Death by Day 90
Risk ratio
95% CI: 0.513–1.232 · P = 0.303
Chi-squared test; safety set
The estimated risk ratio of 0.795 corresponds to an estimated 20.5% lower relative risk of death by Day 90 in the tenecteplase group on the reported scale. This is a relative measure, not an absolute mortality reduction.
The 95% confidence interval of 0.513–1.232 includes 1 and is wide enough to encompass both a relative reduction and a relative increase. The P-value of 0.303 does not resolve that uncertainty or establish equivalence between treatments.
10. Statistical Methods Explained
Why was a risk ratio used for the primary endpoint?
The primary endpoint is binary: each participant either had an mRS score of 0 or 1 at Day 90±7 days or did not. A risk ratio directly compares the probability of the favorable outcome between the two treatment groups. This makes the primary effect measure naturally interpretable as a relative probability of favorable functional outcome.
Why was modified Poisson regression used?
The registry reports a modified Poisson regression model for the primary endpoint. For a binary outcome, this approach can estimate a risk ratio while allowing adjustment for baseline covariates. In this trial, treatment was the main effect and baseline NIHSS, age, and time to drug administration since stroke symptoms onset were included as continuous covariates.
Why is non-inferiority judged against a margin?
A non-inferiority trial is not asking only whether the observed estimate differs from 1. The key question is whether the evidence excludes a treatment effect worse than a prespecified clinically acceptable boundary. Here, the risk-ratio margin was 0.937. The reported lower 95% confidence bound was 0.9678, which is above that margin.
What does a risk ratio of 1.0278 mean?
A risk ratio of 1.0278 means that the estimated probability of the favorable primary outcome was 1.0278 times the probability under alteplase, according to the adjusted model. It corresponds to an estimated 2.78% relative increase in probability, not a 2.78 percentage-point increase.
What does the confidence interval tell us?
The 95% confidence interval describes uncertainty around the estimated treatment effect under the specified statistical framework. For the primary endpoint, the interval was 0.9678–1.0915. It therefore includes 1, while remaining entirely above the non-inferiority margin of 0.937. These are two different reference points answering two different questions.
Why were mixed-effects models used for NIHSS?
NIHSS is a quantitative score measured over time, with the endpoint defined using baseline and Day 90 assessments. A mixed-model framework can represent repeated measurements and account for the correlation between observations from the same participant. The registry specifically reports a mixed models analysis and identifies treatment as a fixed effect with baseline NIHSS, age, and time to administration as linear covariates.
Why does the analysis population matter?
The FAS preserves the randomized treatment comparison for efficacy, whereas the Safety Set is used for safety and analyzes participants according to actual treatment. Switching from randomized assignment to an exposure-based population can answer a different question. Keeping the populations explicit helps distinguish efficacy inference from exposure-based safety description.
11. Missing Data and Observed-Case Analyses
The registry explicitly describes missing-data handling for two secondary endpoints.
| Endpoint | Missing-data approach reported |
|---|---|
| Major neurological improvement at 24 hours | Analysis based on observed cases; no imputation performed for patients missing NIHSS score at 24 hours. |
| Barthel Index score ≥95 up to 90 days | No imputation performed for patients missing Barthel Index score at Day 90. |
An observed-case analysis uses the outcome values that are available for the specified analysis rather than replacing missing values through an imputation procedure. This is straightforward operationally, but the interpretation depends on why observations are missing. If missingness is related to prognosis or treatment, the analyzed subset can differ systematically from the randomized population.
For the primary endpoint, the registry identifies the Full Analysis Set as the primary efficacy population but does not provide a separate imputation method in the posted primary statistical analysis.
12. Covariate Adjustment
Several efficacy analyses adjusted for the same core baseline variables: baseline NIHSS, age, and time to drug administration since stroke symptoms onset.
| Endpoint | Adjusted variables reported |
|---|---|
| Primary mRS 0-1 | Baseline NIHSS, age, and time to drug administration since stroke symptoms onset |
| Major neurological improvement | Baseline NIHSS, age, and time to drug administration since stroke symptoms onset |
| mRS 0-2 | Baseline NIHSS, age, and time to drug administration since stroke symptoms onset |
| NIHSS change from baseline | Baseline NIHSS, age, and time to administration since stroke symptoms onset |
| Barthel Index ≥95 | Baseline NIHSS, age, and time to drug administration since stroke symptoms onset |
| mRS 5 or 6 | Baseline NIHSS, age, and time to drug administration since stroke symptoms onset |
Covariate adjustment does not turn an observational comparison into a randomized trial; the randomization remains the foundation of the treatment comparison. Its role here is to incorporate prespecified baseline information into the statistical model and potentially improve precision.
The variables have different statistical roles. Baseline NIHSS measures neurological severity before treatment, age is a demographic prognostic factor, and time to drug administration represents an important treatment-timing characteristic. Adjusting for them can make the reported treatment contrast conditional on these measured baseline variables.
13. Comparing the Effect Measures
| Effect measure | Used in ORIGINAL | Interpretation |
|---|---|---|
| Risk ratio | Primary endpoint; major neurological improvement; mRS 0-2; Barthel Index ≥95; sICH; death; mRS 5 or 6 | Relative probability of an event or favorable outcome between treatment groups. |
| Mean difference | Change from baseline in NIHSS | Difference in the adjusted mean outcome between treatment groups. |
| Odds ratio | Distribution of mRS | Relative odds under the logistic regression analysis; not equivalent to a risk ratio when outcomes are common. |
Using different effect measures is not inherently inconsistent. Binary outcomes can naturally be expressed as risk ratios, continuous outcomes as mean differences, and logistic regression analyses as odds ratios. The key is to interpret each measure according to its mathematical definition.
14. Multiplicity and the Secondary Endpoints
The registry reports one primary endpoint and eight additional posted statistical analyses. The secondary analyses address neurological improvement, functional outcomes, NIHSS, the mRS distribution, disability measured by the Barthel Index, symptomatic intracerebral haemorrhage, mortality, and severe disability or death represented by mRS 5 or 6.
Primary endpoint
The mRS 0-1 endpoint is the explicitly designated primary endpoint and is evaluated under the non-inferiority framework.
Secondary endpoints
The other posted analyses provide additional information about neurological recovery, functional status, and safety.
P-values
Secondary P-values should be interpreted in the context of their endpoint role rather than as independent evidence of a separate confirmatory claim.
Effect estimates
Confidence intervals and point estimates provide information about magnitude and precision that a P-value alone cannot convey.
The registry does not report a multiplicity adjustment procedure in the posted statistical analyses. Accordingly, the secondary P-values should be read as the results of their respective analyses rather than as a single family of independently confirmatory tests.
15. Results in Statistical Context
The nine posted statistical analyses form a coherent set of complementary endpoints. The primary analysis focuses on favorable functional independence, while the secondary analyses examine earlier neurological improvement, a broader favorable mRS category, change in NIHSS, the mRS distribution, Barthel Index function, intracerebral haemorrhage, mortality, and severe disability or death.
| Domain | Endpoint | Effect |
|---|---|---|
| Primary functional outcome | mRS 0-1 at Day 90±7 days | RR 1.0278; 95% CI 0.9678–1.0915 |
| Early neurological recovery | Major neurological improvement at 24 hours | RR 1.0671; 95% CI 0.9573–1.1895 |
| Functional outcome | mRS 0-2 at Day 90 | RR 1.0078; 95% CI 0.9614–1.0564 |
| Neurological score | Change from baseline in NIHSS | Mean difference -0.45; 95% CI -1.40 to 0.50 |
| mRS distribution | Distribution of mRS at Day 90 | OR 1.0418; P = 0.4806 |
| Activities of daily living | Barthel Index ≥95 | RR 1.0189; 95% CI 0.9635–1.0774 |
| Intracranial bleeding | sICH up to 7 days | RR 1.005; 95% CI 0.370–2.701 |
| Mortality | Death by Day 90 | RR 0.795; 95% CI 0.513–1.232 |
| Severe disability or death | mRS 5 or 6 at Day 90 | RR 0.9215; 95% CI 0.6578–1.2908 |
The important statistical distinction is between the primary non-inferiority question and the descriptive interpretation of the secondary endpoints. The primary risk ratio was 1.0278, with a lower confidence limit of 0.9678 that remained above the 0.937 non-inferiority margin. The secondary analyses provide additional perspectives but use different endpoints, models, and effect measures.
Several secondary confidence intervals include the null value for their respective effect measure. That does not mean the endpoints are uninformative. It means that the posted estimates have varying degrees of statistical precision and that the intervals do not exclude the null value. The width of the interval is often as important as whether it crosses the null.
16. Why This Trial Matters Statistically
ORIGINAL is a useful teaching case because it combines a non-inferiority hypothesis with binary functional outcomes, covariate-adjusted regression, multiple effect measures, longitudinal analysis, and a separate safety population.
| Concept | How it appears in ORIGINAL |
|---|---|
| Randomization | Randomized allocation to two parallel treatment groups. |
| Intention-to-treat analysis | Primary efficacy analysis uses the Full Analysis Set with treatment assignment as randomized. |
| Non-inferiority design | Primary endpoint evaluated against a prespecified risk-ratio margin of 0.937. |
| Covariate adjustment | Baseline NIHSS, age, and time to drug administration included in several efficacy models. |
| Risk ratio | Primary and several secondary binary endpoints use risk ratios. |
| Linear models | Several binary secondary endpoints were analyzed with regression described as linear in the registry, while the analysis notes identify Poisson or log-binomial models. |
| Mixed-effects model | Change from baseline in NIHSS was analyzed with a mixed models analysis. |
| Logistic regression | Used for the Day 90 mRS distribution, with an odds ratio reported. |
| Chi-squared test | Used for death by Day 90. |
| Specialized safety test | Suissa-Shuster test used for symptomatic intracerebral haemorrhage. |
| Missing data | Observed-case analyses without imputation were specified for selected secondary endpoints. |
| Safety population | Safety analyses use the Safety Set and actual treatment received. |
17. Important Limitations and Interpretation Issues
- Non-inferiority is not equivalence: the primary analysis evaluates whether the lower confidence bound remains above the prespecified margin. It does not demonstrate that the two treatments have exactly the same effect.
- Relative versus absolute effects: the primary risk ratio of 1.0278 cannot be translated into an absolute percentage-point difference without the underlying event percentages by treatment group.
- Confidence-interval width: several secondary endpoints have intervals that span a broad range of plausible effects. This is particularly important for symptomatic intracerebral haemorrhage, whose reported interval is 0.370–2.701.
- Secondary endpoint multiplicity: the registry reports multiple secondary analyses, but no multiplicity-adjustment procedure is identified in the posted statistical analyses.
- Observed-case analyses: no imputation was performed for missing NIHSS values at 24 hours or missing Barthel Index values at Day 90. The interpretation therefore depends in part on the pattern of missingness.
- Analysis population: efficacy and safety use related but distinct analysis populations, and the safety set analyzes patients according to actual treatment.
- Model dependence: regression estimates depend on the statistical model and covariate specification. The treatment effect is therefore a model-based adjusted comparison rather than a simple unadjusted ratio.
- mRS complexity: a binary threshold such as mRS 0-1 compresses a six-point disability scale plus death into two categories. The separate mRS distribution analysis provides a different perspective.
- Registry reporting: the registry contains formal analyses for all nine posted outcome measures, but the amount of numerical detail differs across analyses. For example, the mRS distribution analysis reports an odds ratio and P-value without a posted confidence interval.
18. Interpreting the Primary Non-Inferiority Result
The primary risk ratio was 1.0278, with a two-sided 95% confidence interval of 0.9678–1.0915. The prespecified non-inferiority margin was 0.937. Because the lower confidence limit of 0.9678 is above 0.937, the reported confidence interval satisfies the stated risk-ratio non-inferiority boundary.
The result does not mean that tenecteplase and alteplase produce identical outcomes. It also does not mean that 2.78% more participants achieved favorable functional outcome. The estimate is a relative risk ratio, and the absolute treatment-group percentages are not reported in the posted statistical analysis.
In a non-inferiority analysis, the confidence interval is directly connected to the margin. The point estimate of 1.0278 alone is insufficient: the lower bound of 0.9678 is the quantity that establishes how far the plausible treatment effect extends toward the unfavorable side of the scale.
The registry does not report a P-value for the primary non-inferiority analysis. This is not a missing substitute for the margin-based decision rule. Non-inferiority is fundamentally a comparison of the estimated treatment effect and its confidence bound with a prespecified clinically acceptable margin.
19. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary adjusted risk ratio was 1.0278, with a 95% confidence interval of 0.9678–1.0915. The lower confidence bound was above the prespecified non-inferiority margin of 0.937.
Clinical interpretation
The endpoint represents functional status after stroke, with mRS 0 or 1 defining the favorable category. The statistical result therefore concerns the relative probability of reaching that functional outcome at Day 90±7 days.
Safety interpretation
Serious adverse events were reported in 116/732 participants with tenecteplase and 115/736 with alteplase. Separate analyses addressed symptomatic intracerebral haemorrhage and death.
Uncertainty
The primary confidence interval is sufficiently precise to remain above the non-inferiority margin, while several secondary endpoints have wider intervals that cross their respective null values.
20. Trial Timeline
Trial start
The ORIGINAL trial began on June 22, 2021.
Primary completion
The registry lists October 8, 2023 as the primary completion date.
Registry status
The ClinicalTrials.gov record lists the study status as completed.
21. A Practical Framework for Reading the Results
A useful way to read ORIGINAL is to move through the statistical hierarchy rather than beginning with the secondary P-values.
For the primary endpoint, this sequence leads to a clear statistical structure: the endpoint is binary; the effect measure is a risk ratio; the estimate is 1.0278; the two-sided 95% confidence interval is 0.9678–1.0915; and the lower bound is above the prespecified margin of 0.937. The secondary results then add information about other functional, neurological, and safety outcomes.
This approach also prevents a common statistical error: treating every reported P-value as though it answered the same question. The primary endpoint has a non-inferiority margin, whereas the secondary endpoints have their own estimates and hypothesis tests.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: ORIGINAL, NCT04915729.
- PubMed: PubMed record 41585404.
- PubMed: PubMed record 39264623.
Continue through Clinical Biostats
Connect the statistical methods used in ORIGINAL with deeper tutorials and practical statistical tools.
25. Record Summary
ORIGINAL provides a useful example of how a randomized clinical trial can be structured around a non-inferiority question while incorporating several complementary statistical approaches. The primary endpoint was a binary functional outcome measured at Day 90±7 days, analyzed in the Full Analysis Set using a modified Poisson regression model with baseline NIHSS, age, and time to drug administration since stroke symptoms onset as covariates.
The primary risk ratio was 1.0278, with a two-sided 95% confidence interval of 0.9678–1.0915. The prespecified non-inferiority margin was 0.937, placing the lower confidence bound above the specified boundary. The secondary analyses extended the statistical picture across neurological improvement, mRS outcomes, NIHSS, the Barthel Index, symptomatic intracerebral haemorrhage, mortality, and severe disability or death.
The broader statistical lesson is that the primary estimate, its confidence interval, and the non-inferiority margin must be interpreted together. Secondary P-values and effect estimates then provide additional information, but they answer different questions and use different statistical models. Safety interpretation likewise requires attention to the separate Safety Set and the specific endpoint definition.