Hyperbaric Oxygen for COVID-19 Patients: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Hyperbaric Oxygen for COVID-19 Patients (NCT04332081), a non-randomized na study. The primary outcome is Mortality.

2. Clinical Question

3. Trial Design

The registry describes an open-label, single-center study of emergency hyperbaric oxygen for respiratory distress in patients with COVID-19. It lists the allocation as non-randomized, masking as none, and the model as single-group. The phase is listed as NA. These design details matter because treatment assignment was not randomized and outcome assessment was not masked, leaving the study vulnerable to differences between groups and to measurement or reporting influences.

The brief summary describes HBOT as an adjunct to standard therapy for a cohort of 40 COVID-19-positive patients with respiratory distress at NYU Winthrop Hospital. It says the clinical team and a hyperbaric physician would evaluate patients before HBOT. After the intervention portion, a chart review was to compare intervention patients with patients who received standard of care. The record therefore describes a planned comparison, while its formal model field identifies a single-group study. It does not explain this difference in design description.

The registry lists an enrollment of 80 and gives the study status as terminated. The study start is listed as 2020-04-06, and primary completion as 2020-05-29. These dates describe the registry record; they do not by themselves explain why the study was terminated or whether follow-up was complete for every participant. The reason for termination and details of follow-up are not reported in the registry record.

The listed secondary outcome is need for mechanical ventilation. The record identifies two outcome measures and no analyses. It does not report randomization procedures, allocation concealment, a comparator-matching approach, or a statistical analysis plan. Without those details, the comparison can be described only as non-randomized and based on the planned chart review, not as an estimate protected from confounding.

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.

Overall 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.

6. Primary Endpoint

The registry lists the primary outcome as Mortality, measured over through study completion; an average of 50 days.

The primary endpoint is mortality, expressed as a percentage of participants. The stated timeframe is through study completion, with an average of 50 days. In general, a binary endpoint records whether each participant experienced an event, such as death, during the defined observation period; the percentage represents the share of the endpoint population with that event. The registry record does not further define the endpoint population or explain how follow-up time was handled.

Mortality is a clinically important endpoint because it captures a serious outcome directly. Its interpretation in a study depends on clear definitions of who was included, when observation began and ended, and how missing follow-up was treated. Those details are not reported in the registry record. Nor does the record include a primary mortality analysis, an estimate of the group contrast, or uncertainty around that estimate.

The adverse-event listing separately records deaths for the HBOT and standard-of-care groups. Those entries are safety-record fields and should not be treated as a substitute for a reported primary endpoint analysis. A primary outcome analysis would need to specify the endpoint population and follow-up rules and then present the comparison and its statistical uncertainty. The registry record does not report that analysis.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a binary endpoint represent?

Mortality is a binary outcome at the participant level: a participant either has the event during the defined observation period or does not. The registry labels the outcome as a percentage of participants with an event. In general, a percentage is a way to describe how common an event was in a defined group. Its meaning depends on the group included in the denominator and on the period during which events were counted. The record gives a timeframe of through study completion, with an average of 50 days, but does not provide further detail about individual follow-up or the endpoint analysis population.

A percentage alone does not show how precise an estimate is. A confidence interval is a general statistical tool that describes a range of values compatible with the data under a specified method. A narrower interval generally indicates greater precision, while a wider interval indicates greater uncertainty. No confidence interval for mortality is reported in the registry record, so the precision of a primary-outcome estimate cannot be assessed from a reported analysis.

What can a group comparison establish?

In general, a comparison of outcomes between groups can describe an association. A causal interpretation requires a design and analysis that address other factors related to both group membership and outcome. Random assignment can help balance such factors, on average, but this study is listed as non-randomized. The summary describes comparison with standard-of-care patients through chart review. The registry record does not report how patients entered either group, whether groups were matched, or whether statistical adjustment was used.

When groups differ at baseline, an unadjusted contrast can mix a possible treatment effect with differences in patient characteristics and care. Statistical adjustment can address measured differences when appropriately planned and implemented, but it cannot guarantee removal of confounding, particularly from factors that were not measured or recorded. No adjustment method or analysis is reported. Consequently, the group entries should be read descriptively rather than as an isolated estimate of the effect of HBOT.

Why does uncertainty matter?

General statistical interpretation considers both the estimated effect and its uncertainty. A point estimate alone can be misleading if the underlying data are sparse or the estimate is imprecise. A confidence interval can show whether the data are compatible with a broad range of effects, but it cannot correct biases in study design or data collection. Since the registry record reports no primary outcome estimate or interval, neither the size nor the precision of a mortality contrast is available from the posted analysis.

The same principle applies to adverse-event comparisons. Counts can describe reported events, but interpretation also depends on the time each group was observed, the rules used to classify events, and whether all relevant events were collected consistently. Those details are not reported in the registry record. A count of zero does not establish that an event could not occur; it describes only what was recorded for the listed group and observation conditions.

How should planning illustrations be understood?

In general, a sample-size calculation estimates how many participants would be needed under specified assumptions to detect a chosen effect with a chosen statistical procedure. The record describes planning illustrations based on the observed effect, with a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation; the effect to detect is a stated fraction of the observed effect. For time-to-event calculations, the illustrations are expressed as total events, while other calculations are expressed as participants per arm.

These calculations are planning illustrations, not results from the enrolled study and not evidence that the study had a particular probability of finding an effect. They are not observed power. Their interpretation depends on the assumptions, including the effect used as a reference and the statistical approximation. The registry record does not report the trial’s own sample-size justification or explain whether these assumptions were part of the original design. Because the study is non-randomized and the registry reports no primary analysis, a planning calculation cannot resolve the limitations of the outcome comparison.

What does the analysis count tell readers?

The record lists two outcome measures and zero analyses. This means the registry does not provide a reported statistical analysis for the outcomes in the record. It does not show whether analyses were conducted outside the registry. No model, test, adjustment strategy, or inferential result for mortality is reported. General concepts such as confidence intervals, hypothesis tests, or regression cannot be attributed to this study unless the registry reports their use.

For a binary outcome, common analytical approaches can estimate a group difference or another contrast and quantify uncertainty. Which approach is suitable depends on the design, the endpoint definition, the follow-up structure, and the analysis population. Here, the record does not specify those analytical choices. It would therefore be inappropriate to infer a method or to reconstruct a formal result from the adverse-event counts.

10. Exact Estimates for Each Arm

No two-arm comparison is shown for this trial, but each arm's result can still be summarized with exact interval estimates computed from the posted counts. The Clopper-Pearson interval is exact and conservative; the Wilson interval is a common closer-to-nominal alternative; the Bayesian credible interval uses a uniform Beta(1, 1) prior. For small groups the three can differ noticeably.

R code and output

arm    <- c("Hyperbaric Oxygen Therapy (HBOT)", "Standard of Care")
n      <- c(20, 60)
events <- c(2, 13)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  Hyperbaric Oxygen Therapy (HBOT)  20  2  10.0  1.2  31.7  2.8  30.1  3.0  30.4
ARM  Standard of Care  60  13  21.7  12.1  34.2  13.1  33.6  13.2  33.7

11. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when several outcomes, time points, or analyses are tested. In general, conducting many tests can increase the chance of a false-positive finding unless the analysis plan addresses that issue. The registry lists mortality as the primary outcome and need for mechanical ventilation as a secondary outcome, and identifies two outcome measures. It does not report a multiplicity strategy, a hierarchy of tests, or an analysis plan. With no analyses reported, the registry record does not show how any multiplicity concern was handled.

Missing data can affect both the estimated event frequency and the comparison between groups. For a mortality endpoint, incomplete follow-up or unavailable records can leave outcome status unknown. The record does not state whether data were missing, how missing outcomes were handled, or whether every participant was followed through the stated study-completion timeframe. The average follow-up description does not establish that each participant had the same observation period.

Analysis populations also require clear definitions. A report may specify which participants were included in each outcome analysis and how withdrawals, protocol deviations, or incomplete records were handled. The registry does not report an analysis population for mortality or mechanical ventilation. The adverse-event listings do give at-risk counts for the two named groups, but those entries do not define the population used for the primary endpoint analysis.

The lack of these details limits reproducibility and makes it difficult to assess whether exclusions or incomplete observation could have influenced the reported entries. These omissions should not be filled in by assumption; the registry record does not report them.

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 0% reporting threshold.

Most frequent serious adverse event terms

Serious events with the highest arm-level rate: hypoxic arrest (5.0%).

The adverse-event listing gives separate at-risk counts for HBOT and standard of care. In the HBOT group, 1 serious event is listed among 20 at risk, and 0 other events are listed among 20 at risk. In the standard-of-care group, 0 serious events and 0 other events are listed among 60 at risk. The listing also gives death counts of 2 for HBOT and 13 for standard of care. These entries should be interpreted as registry-recorded counts, not as a complete comparative safety analysis.

The most frequently listed serious term is Hypoxic Arrest, at 5.0%. The record does not give further detail about that event in the registry record for this analysis. Serious-event counts and death counts are distinct fields in the listing; the death entries should not be assumed to be additional serious events or to use the same event-classification rules.

In general, safety counts depend on how events were defined, collected, and followed over time. The record does not report exposure duration by group, event adjudication, or whether collection was equally complete across the groups. The non-randomized comparison also means that group differences cannot be attributed to HBOT from these counts alone. Zero recorded events in a group does not prove that no risk exists.

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

The registry record documents a study of HBOT for patients with COVID-19 and respiratory distress, with mortality named as the primary endpoint and need for mechanical ventilation listed as a secondary outcome. It describes the intervention as an adjunct to standard therapy and a planned chart review comparing HBOT recipients with standard-of-care patients. The adverse-event listing records events and deaths for both groups.

What the record does not provide is a primary mortality analysis, a statistical comparison, an estimate of uncertainty, or a description of adjustment for differences between groups. The descriptive entries cannot establish whether HBOT caused any difference in mortality or other outcomes. Because allocation was non-randomized, measured and unmeasured differences between patients could influence the observed group outcomes. The registry’s single-group model field and its summary’s planned comparator review also leave the exact analytic structure unclear.

The results should not be read as proof that HBOT is beneficial or harmful. The listed deaths and adverse events are not a substitute for a clearly defined, adjusted outcome analysis with follow-up and missing-data procedures. Nor does a lack of reported analysis imply that no analysis was performed outside the registry; it means the registry record does not report one.

Planning calculations, if considered, answer a different question from study results. They illustrate sample sizes under stated assumptions and do not show observed power or resolve bias in a non-randomized comparison. The most defensible interpretation is therefore limited: the record describes the study design and posts selected event counts, but does not provide enough analyzed outcome information to estimate a treatment effect or its precision.

14. Why This Trial Matters Statistically

This study illustrates why a registry design label and the narrative description both matter. The record identifies a non-randomized, single-group model, while the summary describes comparing HBOT patients with standard-of-care patients through chart review. These descriptions signal that the comparison is not a randomized trial contrast, and the record does not explain how the comparison was formed or analyzed. Readers should distinguish a planned comparison from a reported, statistically analyzed treatment effect.

It also shows the importance of separating endpoint results from safety listings. Mortality is identified as the primary endpoint, but the death counts appear in adverse-event entries and are not accompanied by a primary endpoint analysis. Without a stated analysis population, follow-up accounting, and uncertainty estimate, those counts cannot answer the full statistical question implied by the primary endpoint.

The record’s lack of reported analyses makes transparency especially important. A study record can describe outcomes and event counts without reporting the methods needed to interpret group differences. In general, uncertainty, confounding, and missing-data handling are part of the result, not optional details. Here, those matters are not reported, so conclusions must remain descriptive.

Finally, the terminated status and the stated study dates provide context but not an explanation of the study’s conduct or outcome completeness. The reason for termination is not reported. This makes the record a useful example of why registry status, outcome definitions, analysis plans, and posted analyses should be considered separately rather than treated as interchangeable evidence.

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.

This study has limitations specific to its design and registry record. It is listed as non-randomized, open-label, and single-center. The planned chart-review comparator may differ from the HBOT group in patient characteristics or care pathways, and the record does not describe a method for reducing such differences. The formal single-group model designation also does not align clearly with the summary’s description of a comparator review; the registry record does not explain the relationship between these descriptions.

The primary endpoint is mortality, but no primary analysis is reported. There is no estimate of the group contrast, confidence interval, or statistical test. The record does not define the mortality analysis population in detail or report how follow-up differences and missing outcome data were handled. The average 50-day timeframe does not establish identical observation for each participant.

The registry lists need for mechanical ventilation as a secondary outcome but does not report its analysis. It also does not report an analysis plan, adjustment methods, multiplicity procedures, or the reason for termination. The listed adverse events provide counts but not event narratives, exposure duration, or the completeness of event collection. These omissions limit the ability to assess both efficacy-related outcomes and safety comparisons.

The enrollment entry and the planned cohort description should not be treated as evidence that every enrolled participant contributed to each outcome. The record does not report participant flow or outcome-specific denominators. These limitations make causal interpretation and precision assessment particularly uncertain.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Hyperbaric Oxygen for COVID-19 Patients — NCT04332081.
• Registry record last updated 2021-06-22; status last verified 2021-06.
• Results first posted 2021-06-22.
• 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

The registry describes an open-label, non-randomized, single-center study of emergency HBOT for respiratory distress in patients with COVID-19. The summary presents HBOT as an adjunct to standard therapy and describes a chart-review comparison with standard-of-care patients. The formal model is listed as single-group. The primary endpoint is mortality through study completion, with an average timeframe of 50 days; need for mechanical ventilation is listed as a secondary outcome.

The registry lists an enrollment of 80, a terminated status, and no analyses. Its adverse-event entries include serious-event, other-event, and death counts for HBOT and standard-of-care groups. The record does not report a primary outcome analysis, statistical comparison, uncertainty estimate, missing-data approach, or reason for termination. Accordingly, the posted information supports a descriptive account of the study record, not a causal estimate of HBOT’s effect.