TD-0903 for ALI Associated With COVID-19: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for TD-0903 for ALI Associated With COVID-19 (NCT04402866), a randomized phase 2 study. The primary outcome is Part 2: Number of Respiratory Failure-free Days (RFDs) From Randomization to Day 28.

2. Clinical Question

This is a question about the distribution of a bounded clinical outcome over a defined follow-up period, not simply whether participants were alive at one time point. The endpoint incorporates both survival and the need for respiratory support: a day counted only when a participant was alive and did not require the specified forms of support. The primary comparison therefore concerns a combined measure of clinical status over time. It does not isolate the effect on mortality from the effect on respiratory support.

The registry also lists outcomes concerning oxygenation, ordinal clinical status, and being alive without respiratory failure. Those outcomes address related aspects of illness, but they do not replace the stated primary question. The registry record does not report a formal hypothesis statement beyond the named endpoint and its comparison.

3. Trial Design

NCT04402866 is a completed phase 2 study titled TD-0903 for ALI Associated With COVID-19. The official title describes a randomized, double-blind, placebo-controlled, parallel-group, multicenter study of inhaled TD-0903. The registry lists triple masking. It does not specify which participant, investigator, or other study roles were masked, so the masking label should not be interpreted more specifically than the record allows.

The study had two parts. In Part 1, dose cohorts evaluated TD-0903 multiple-ascending-dose regimens alongside placebo cohorts. The listed TD-0903 dose groups were 1 mg, 3 mg, and 10 mg in the safety results. In Part 2, the randomized comparison was TD-0903 at 3 mg against matching placebo, with both used alongside standard of care. The Part 2 arm descriptions state that 99 subjects were planned for randomization to each group; posted outcome denominators differ across analyses and are described with the relevant results.

The registry lists enrollment of 235. The study began on 2020-06-24, and primary completion was recorded as 2021-04-21; the overall status is completed. Theravance Biopharma is listed as sponsor. The brief summary says the study evaluated efficacy, safety, pharmacodynamics, and pharmacokinetics in hospitalized patients with confirmed COVID-19-associated acute lung injury and impaired oxygenation.

Randomization is intended to make treatment groups comparable on average, while masking can reduce some forms of bias in care and assessment. Placebo control gives the comparison a concurrent reference group. These design features do not, by themselves, establish that all potential sources of bias were prevented. Details such as the randomization procedure, allocation concealment, planned sample-size justification, and protocol deviations are not reported in the registry record.

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: 235 participants were enrolled and 230 participants were randomized and treated with study drug. Within each cohort in Part 1 (multiple-ascending dose design), participants were randomized 3:1, TD-0903 to placebo. During Part 2 (parallel-group design), participants were stratified by baseline age (≤ 60 versus \> 60 years), and by concurrent use of antiviral medications (yes or no) at baseline. Within each stratum, participants were randomized 1:1 to receive either TD-0903 or placebo.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: ITT analysis set (Part 2) - all participants with analyzable data who were randomized into the study.

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: Part 1: Safety analysis set - all participants who received at least one dose of study drug. Part 2: Intent-to-Treat (ITT) analysis set - all participants who were randomized into the study.

Only the first 6 of 7 groups are shown.

6. Primary Endpoint

The registry lists the primary outcome as Part 2: Number of Respiratory Failure-free Days (RFDs) From Randomization to Day 28, measured over Randomization to Day 28.

The primary endpoint is the number of respiratory failure-free days from randomization to Day 28, summarized by the median and interquartile range. A day counted as respiratory failure-free when the participant was alive and did not require invasive mechanical ventilation, non-invasive positive pressure ventilation, high-flow oxygen devices, or oxygen supplementation. Participants who used respiratory support for 28 days or longer, or who died on or before Day 28, were assigned zero days.

The registry links this measure to the clinical status scale: a score of 4 or less on a given day was equivalent to a respiratory failure-free day. The scale ranges from 1 to 8, with higher scores representing worse outcomes. A score of 4 describes a hospitalized participant not requiring supplemental oxygen but still receiving ongoing medical care, whether or not that care was related to COVID-19.

As a general statistical concept, this kind of endpoint is useful when both survival and respiratory support are clinically relevant. It captures more than a single status assessment because it counts qualifying days over the specified period. At the same time, the combined definition means that a difference in the endpoint would not identify which component—survival or respiratory support—accounted for it. The zero-day rule also groups together participants who died and those who remained on support for the specified duration. The registry does not report additional details about how assessments between recorded days were handled.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is a count of qualifying days within a fixed follow-up period, with a participant assigned zero days under the registry's rules for prolonged respiratory support or death. The record summarizes this endpoint using the median and interquartile range. The median is the midpoint of an ordered distribution: half the observations lie at or below it and half at or above it, subject to ties. The interquartile range describes the middle half of the observations. These summaries are often useful when data are skewed, bounded, or contain a substantial concentration at particular values, although the registry record does not state why this summary was selected.

The sponsor's primary comparison is reported as a Van Elteren test with a common odds ratio. The Van Elteren test is a stratified rank-based comparison, a form of the Wilcoxon-Mann-Whitney approach that accounts for stratification. In general, rank-based methods compare the ordering of observations rather than relying on a normal-distribution model for the raw outcome. A common odds ratio summarizes the comparative ordering across strata under the analysis framework. The strata used, the precise odds-ratio convention, and further details of the model are not reported in the registry record. The result should therefore be read using the registry's reported estimate and test, without inferring undocumented modeling choices.

The 95% confidence interval around the common odds ratio communicates uncertainty in that estimate under the fitted analysis and its assumptions. For odds ratios, the null value is 1. The primary interval, 0.706 to 1.846, includes that value. A confidence interval is not a range guaranteed to contain the true effect, nor does it give the probability that a particular effect is correct. Its width reflects the precision available from the data and method. The registry reports p = 0.614; a p-value measures how unusual data at least as inconsistent with the null would be if the null and model assumptions held. It does not measure clinical importance or prove no difference.

For the oxygenation outcome, the sponsor reports a mixed model repeated measures analysis of the change from baseline in SaO2/FiO2 ratio on Day 7. Such models are generally used to analyze repeated measurements while accounting for the correlation among observations from the same participant. The registry reports an adjusted least-squares mean difference of -0.51, its 95% confidence interval, and a p-value. It does not report the model's covariance structure, included repeated visits, adjustment variables, or handling of missing repeated measurements. Those omissions matter because model estimates can depend on these choices and assumptions.

For the Day 28 alive and respiratory failure-free outcome, the sponsor reports a Cochran-Mantel-Haenszel chi-square test and a risk difference. In general, this test compares categorical outcomes while accounting for strata. The record does not identify those strata or give further details of the analysis specification. The posted sponsor risk difference is 5.06, with its 95% confidence interval and p-value. A risk difference expresses the difference between group event probabilities; the reported interval includes the null value of zero. The registry's specific reporting scale should be retained when quoting this estimate.

Some unadjusted re-calculations are included as descriptive comparisons from posted summary statistics. These calculations do not reproduce the sponsor's stratified or repeated-measures models. For the oxygenation comparison, the unadjusted re-calculation reports a mean difference of -0.51, a 95% interval from -21.89 to 20.87, and p = 0.963. For the Day 28 binary outcome, it reports a risk difference of 0.051, with an interval from -0.048 to 0.149 and p = 0.314. The methods and displayed scales differ from the sponsor analyses, so the re-calculations should not replace those analyses. The primary median endpoint is not re-calculated because the posted summary statistics do not support an unadjusted two-group comparison of that type.

The planning illustrations based on the secondary outcomes use the observed effects as inputs, with equal allocation, a two-sided significance level of 0.05, and a normal approximation. They show participants per arm for 80% and 90% power under the stated fractions of the observed effect. These are planning illustrations, not observed power and not the trial's original design justification. The registry does not report the trial's sample-size justification. In general, sample-size projections are highly sensitive to the assumed effect, variability, event rate, and analysis method; projections based on an observed effect can be especially unstable when that effect is imprecisely estimated.

Over-dispersion is a general concern when outcome variability exceeds what a statistical model assumes. If ignored, standard errors and confidence intervals can be too narrow, making the results appear more precise than they are. This concern is not specifically reported as an issue in this trial, and the registry does not describe a dispersion assessment for the primary endpoint. It is a general reason to assess whether the model fits the outcome structure rather than an assertion that the reported analysis failed.

10. Analyses Reported by the Sponsor

The sponsor's primary analysis is the appropriate reported result for the primary endpoint: the Van Elteren test gave a common odds ratio of 1.142, a 95% confidence interval from 0.706 to 1.846, and p = 0.614. The primary outcome is reported as medians and interquartile ranges, but those summaries are not sufficient to reproduce the sponsor's stratified rank-based analysis. The registry record does not report a compatible unadjusted re-calculation for this endpoint.

For the two secondary outcomes with sponsor comparisons, the registry also gives re-calculations from posted summary statistics. These use unadjusted methods and do not reproduce the sponsor's mixed model repeated measures or Cochran-Mantel-Haenszel analyses. Their estimates and p-values are therefore not interchangeable with the sponsor's results. For example, the sponsor reports p = 0.962 for the oxygenation comparison and p = 0.300 for the Day 28 binary comparison, while the unadjusted re-calculations report p = 0.963 and p = 0.314, respectively. Small differences can arise because the analyses use different methods and assumptions.

The sponsor's own analyses take precedence when describing the trial's reported inferential results. The re-calculations are supplemental checks based on posted summaries, not corrections to the sponsor analyses. The registry does not report enough model detail to independently reproduce the sponsor analyses in full.

11. Multiplicity, Missing Data and Analysis Populations

The registry record identifies one primary endpoint and lists secondary outcomes, but it does not report a multiplicity strategy. It is not reported in the registry record whether secondary tests were adjusted, ordered in a formal hierarchy, or treated as exploratory. Without such information, a nominal p-value for an individual secondary outcome should not be read as if it were the sole planned test. When many outcomes, time points, or comparisons are examined, the chance of at least one apparently unusual result can increase. Multiplicity procedures are designed to control that risk, but no specific procedure can be attributed to this trial from the registry record.

The primary analysis population is described as the Part 2 ITT analysis set: all participants with analyzable data who were randomized into the study. This wording combines an intention-to-treat label with an analyzable-data qualification. The registry does not report the number excluded for lack of analyzable primary data, the reasons for missingness, or whether any randomized participants were omitted from this analysis. The arm denominators also vary across posted outcomes, and the record does not explain every difference in denominator.

Missing-data methods can affect estimates when outcomes are not observed for every randomized participant. As a general statistical principle, complete-case analyses rely on assumptions about why data are missing; imputation and model-based methods also require assumptions. The registry record does not report the primary endpoint's missing-data strategy, sensitivity analyses, or assumptions about missing outcomes. For the repeated-measures oxygenation outcome, the model is named, but missing-measurement handling is not described. No particular method should be inferred.

The registry does not report interim analyses, stopping rules, or a sample-size justification. These omissions do not establish that such plans were absent from the protocol; they mean that their details are not available in the registry record. Likewise, the stated Part 2 randomization plans and overall enrollment do not substitute for an account of the final analysis population or protocol-defined estimand.

12. 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: acute respiratory distress syndrome (33.3%), covid-19 (16.7%), multiple organ dysfunction syndrome (16.7%), cardiac arrest (16.7%), acute respiratory failure (16.7%).

The registry reports adverse-event counts separately for Part 1 dose cohorts and Part 2 treatment groups. In Part 2, serious adverse events were reported for 16 of 102 participants at risk in the matching-placebo group and 10 of 103 at risk in the TD-0903 3 mg group. Deaths were reported for 13 and 6 participants, respectively. Other adverse events were reported for 18 of 102 and 15 of 103 participants. These are descriptive counts, not adjusted comparisons, and the registry information given does not establish why individual events occurred.

Part 1 cohorts were small: the serious-event counts were 3 of 6 for matching placebo, 1 of 6 for TD-0903 1 mg, 1 of 7 for TD-0903 3 mg, and 0 of 6 for TD-0903 10 mg. The corresponding death counts were 2, 1, 0, and 0. With such small at-risk groups, individual events substantially affect the apparent pattern, and descriptive counts should not be treated as stable estimates of comparative risk.

The listed serious-event terms include acute respiratory distress syndrome at 33.3%, with COVID-19, multiple organ dysfunction syndrome, cardiac arrest, acute respiratory failure, and ventricular fibrillation each listed at 16.7%. The registry excerpt does not define the denominator for these term percentages, so they should not be interpreted as percentages of all participants or as treatment-specific rates. Adverse-event counts alone do not determine causation or establish a comparative safety profile.

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

The primary posted result describes two groups with the same median number of respiratory failure-free days: 21.0 days in each group. The interquartile ranges overlap substantially, and the sponsor-reported common odds ratio has a confidence interval that includes the null value. Taken together, these data do not provide a clear statistical separation on the primary endpoint. They also do not prove that the treatments have identical effects.

A failure to show a clear difference is not the same as evidence of equivalence. Equivalence or non-inferiority requires a prespecified margin and an analysis designed to evaluate that question. Such a margin is not reported in the registry record. The primary confidence interval is compatible with a range of effects on either side of the null, which means the estimate remains imprecise. The record does not support concluding that all clinically meaningful differences have been ruled out.

The endpoint itself combines survival with respiratory-support status, and its zero-day rule groups participants who died with those requiring support for the specified duration. Therefore, the primary summary cannot separately answer whether the groups differed in mortality, time on support, or another component. The secondary outcomes provide additional perspectives, but they are not interchangeable with the primary measure and have their own uncertainty and analysis considerations.

These results describe the trial population and analyses recorded for a phase 2 study. They do not establish effects in populations not represented by the registry description, and the registry does not provide enough detail to assess all design and analysis choices. Statistical estimates should be interpreted as evidence with uncertainty, not as a guarantee about outcomes for an individual or a recommendation about care.

14. Why This Trial Matters Statistically

This trial illustrates why endpoint definition matters in clinical research. Respiratory failure-free days combine survival and respiratory-support status across a follow-up period, while the median compresses the distribution into a central summary. The same median in both groups does not imply that every part of the distributions is the same, and the reported common odds ratio answers a different statistical question from a difference between medians.

It also shows why the analysis method and the summary statistics need to be read together. A stratified rank-based primary comparison cannot be reproduced from medians and interquartile ranges alone. Likewise, repeated-measures and stratified categorical analyses can differ from unadjusted calculations based on posted summaries. The re-calculations are useful as checks, but they do not supersede the sponsor's analyses or disclose the full assumptions behind them.

The secondary outcomes demonstrate the importance of precision and multiplicity. Confidence intervals convey the range of effects compatible with an analysis, and multiple secondary tests can make isolated findings less persuasive without a stated testing strategy. The registry does not report that strategy, missing-data procedures, or the sample-size justification, limiting how fully the statistical plan can be assessed.

Finally, safety counts from small cohorts are descriptive and should be read with their denominators and context. The record presents serious events and deaths by group, but the counts alone do not establish causes or comparative risks. For biostatisticians and clinical researchers, this study is a useful case in reading an endpoint, an inferential method, uncertainty, and registry reporting gaps as parts of one coherent statistical interpretation.

15. 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 leaves several trial-specific details unavailable. It does not describe the randomization sequence or allocation-concealment procedure, identify which roles were masked, or report the protocol's sample-size justification. It also does not report the primary analysis's detailed stratification, the common odds-ratio convention, or a complete statistical analysis plan. Those gaps limit independent assessment of how closely the posted analysis follows the intended protocol.

The primary analysis population is defined as randomized participants with analyzable data, but the registry does not explain the analyzability criterion, missing primary outcomes, or exclusions. Posted denominators vary across outcomes. The record does not explain all denominator differences or provide enough missing-data detail to evaluate possible bias. For secondary analyses, model specifications and assumptions are incomplete, including the repeated-measures covariance structure and the strata used in the Cochran-Mantel-Haenszel analysis.

Multiplicity handling is not reported, making the inferential status of secondary outcomes uncertain. The ordinal clinical status outcome is listed, but its results are not reported in the registry record. The registry also does not report interim analyses, stopping rules, or sensitivity analyses. These absences should be understood as reporting limitations, not proof that the protocol did not address the issues.

Finally, the primary endpoint combines death and respiratory-support experience, so the summary does not isolate those components. The Part 1 safety cohorts are small, and the denominator for the listed serious-event term percentages is not specified. The posted results offer useful descriptive evidence, but they do not supply every detail needed for a complete appraisal of efficacy, safety, or robustness.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: TD-0903 for ALI Associated With COVID-19 — NCT04402866.
• Registry record last updated 2022-03-17; status last verified 2022-03.
• Results first posted 2022-03-17.
• 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.

19. Record Summary

NCT04402866 is a completed phase 2 randomized, double-blind, placebo-controlled, parallel-group, multicenter study sponsored by Theravance Biopharma. It evaluated inhaled TD-0903 with standard of care against matching placebo with standard of care in hospitalized participants with COVID-19-associated acute lung injury and impaired oxygenation. The registry lists triple masking and enrollment of 235.

The primary Part 2 outcome was respiratory failure-free days from randomization through Day 28. The posted medians were 21.0 days in both groups. The sponsor-reported common odds ratio was 1.142, with a 95% confidence interval from 0.706 to 1.846 and p = 0.614. The result does not establish a clear between-group difference, and the interval leaves substantial uncertainty.

The registry lists secondary outcomes involving oxygenation, ordinal clinical status, and being alive and respiratory failure-free. It also reports adverse-event counts by group. Important details about multiplicity, missing-data handling, sample-size justification, and several analysis specifications are not reported in the registry record.