Pragmatic Investigation of optimaL Oxygen Targets Trial: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Pragmatic Investigation of optimaL Oxygen Targets Trial (NCT03537937), a randomized na study. The primary outcome is Ventilator-free Days (VFDs) to Study Day 28.

2. Clinical Question

The clinical question concerns the target used to guide oxygen titration during invasive ventilation, rather than a comparison of different ventilation devices or a comparison of oxygen with no oxygen. The registered lower, intermediate, and higher targets were distinct saturation ranges, and the intervention descriptions specify that the fraction of inspired oxygen was titrated to achieve the assigned target.

The primary outcome combines survival and liberation from invasive ventilation over the stated follow-up period. A participant who remained invasively ventilated through day 28, or who died before day 28, was assigned zero ventilator-free days under the endpoint definition. For someone liberated from ventilation and then ventilated again, the count was based on the final liberation before day 28. The record identifies 28-day, in-hospital mortality as a secondary outcome and lists other exploratory clinical and organ-function outcomes. It does not report numerical findings for those outcomes in the record details.

3. Trial Design

The Pragmatic Investigation of optimaL Oxygen Targets (PILOT) Trial was a randomized, single-masked, crossover trial. Its brief summary describes a cluster-randomized cluster-crossover design. The distinction matters: allocation occurred in a cluster-crossover framework, so the study was not simply described as assigning each participant independently to one unchanging strategy.

The three intervention conditions were lower, intermediate, and higher oxygen saturation targets. During invasive mechanical ventilation in a study location, the fraction of inspired oxygen was titrated toward the assigned arterial oxygen saturation target. The targets were 90% (range 88-92%), 94% (range 92-96%), and 98% (range 96-100%). The registry describes the intervention types as “other”; it does not identify a drug intervention.

The brief summary proposed a 2,250-patient cluster-randomized cluster-crossover trial, while the enrollment field records 2541. These are different registry entries: the first is the planned size stated in the summary, and the second is the recorded enrollment. The trial is marked completed. The recorded start date is 2018-07-01, and the primary completion date is 2022-01-31. The phase is listed as NA.

The registry identifies Vanderbilt University Medical Center as sponsor. It does not report which person or role was masked, the number or identity of clusters, the sequence and duration of crossover periods, or details of how participants entered clusters. Those features should not be inferred from the labels alone.

A crossover design can expose a cluster or setting to more than one strategy over time. In general, this can support comparisons within clusters, while also making period, sequence, and cluster structure relevant to analysis. Because observations within a cluster may be related, treating every participant as an independent observation can misrepresent uncertainty. The registry explicitly cautions that a simple individual-level re-analysis would be misleading and states that only the sponsor's own analysis is shown. The sponsor's numerical analysis is not reported in the registry record details.

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 Ventilator-free Days (VFDs) to Study Day 28, measured over 28 days.

The primary endpoint is ventilator-free days (VFDs) to study day 28. It measures the number of days alive and free from invasive mechanical ventilation between a participant’s final liberation from invasive ventilation before day 28 and study day 28. The registry specifies that a participant still receiving invasive ventilation at day 28, or a participant who died before day 28, receives zero VFDs. If a participant returned to invasive ventilation and was liberated again before day 28, the count begins from that final liberation.

The registry identifies the endpoint as a number of days, with the median as the summary measure and the inter-quartile range as the dispersion measure. The timeframe is 28 days. These choices describe how the outcome is to be summarized; the registry record does not give arm-specific medians or inter-quartile ranges.

VFDs combine survival and time free of invasive ventilation in one outcome. This can be useful when both remaining alive and leaving invasive ventilation are relevant to the clinical question. The zero rule makes the treatment of death and continued ventilation explicit rather than allowing those cases to contribute a positive count. At the same time, a single summary can reflect different underlying experiences: the same VFD value need not arise from the same timing of liberation or the same clinical course. Secondary outcomes can provide additional perspectives, but their numerical results are not reported in the registry record details.

In general, reporting a median and inter-quartile range describes the center and spread of observed values without assuming that the outcome follows a symmetric distribution. It does not, by itself, quantify uncertainty about a between-strategy contrast. The registry record does not report a confidence interval or a numerical primary-outcome comparison.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The trial is described as randomized, single-masked, and crossover, with a cluster-randomized cluster-crossover design in the brief summary. The registry does not give enough detail to reconstruct the randomization sequence, cluster structure, crossover periods, or the identity of the masked party. These are important design details, but they are not reported in the registry record.

Why does cluster randomization matter?

In general, cluster randomization assigns groups or settings rather than treating each participant as an independently randomized unit. Participants within a cluster may share care practices or other features, so their outcomes can be correlated. If an analysis ignores this correlation, it may give uncertainty estimates that are too narrow. The registry specifically warns that a simple individual-level re-analysis would be misleading. That warning is particularly relevant when interpreting any analysis that treats participant records as independent observations.

The crossover structure also has statistical implications. In general, when clusters experience more than one strategy over time, the analysis may need to account for which strategy was used in each period and for changes over time. The design can allow comparisons that reflect repeated exposure of clusters, but period and sequence effects can matter. The registry record does not report the sponsor's model, how it handled periods or sequences, or other analytical details. Those choices cannot be inferred from the design label.

What does the primary summary measure say?

The primary endpoint is ventilator-free days to study day 28, reported by the median and inter-quartile range. In general, the median describes the middle of an ordered set of observations, while the inter-quartile range describes the spread of the central part of that distribution. This approach does not require the outcome to be normally distributed. It is a descriptive summary, however: to interpret a difference between strategies, readers also need an inferential comparison and a measure of its uncertainty. Arm-specific summaries and inferential results are not reported in the registry record details.

The endpoint definition assigns zero ventilator-free days to participants who die before day 28 or remain on invasive ventilation at day 28. It also specifies how to count for participants who are re-ventilated and later liberated. These rules define how the outcome is constructed, but the registry entries do not state how missing endpoint measurements were handled or how the sponsor analyzed the distribution.

Why not re-analyze the posted counts as if participants were independent?

The registry's explicit caution is that a simple individual-level re-analysis would be misleading because of the cluster-randomized design. In general, a re-analysis based only on group totals may omit the dependence introduced by clustering and the structure introduced by crossover. Such a calculation would not necessarily reproduce the sponsor's estimand, model, or uncertainty. The record states that only the sponsor's own analysis is shown, but the sponsor's numerical analysis and hazard ratios are not reported in the registry entries used for this account. Consequently, no recalculated comparison is presented.

The registry entries identify zero analyses in the analysis count. They do not describe an adjustment set, a statistical model, a hypothesis-testing procedure, an interim analysis, or a sample-size justification. Each is not reported in the registry record. The lack of those details limits independent assessment of the analysis, but it does not establish that a particular method was or was not used by the investigators.

How should planning illustrations be understood?

The record describes a planning illustration based on an observed effect, using a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. It also says that the effect to detect is a stated fraction of the observed effect. No fraction-specific sample-size values or observed effect estimate are reported in the registry record details. Therefore, a reader cannot infer a numeric planning requirement from those assumptions alone.

In general, sample-size calculations are planning exercises: they describe how much information a future study might require under specified assumptions. They do not describe the power achieved by the completed trial, and they are not evidence that an observed result is reliable. Where a planning illustration uses a fraction of an observed effect, changing that fraction changes the target effect in the calculation. Any such figures should be read as hypothetical planning scenarios, not as observed trial outcomes or observed power.

For time-to-event outcomes, the registry's general note says the tables use total events under the Schoenfeld approach; other sample-size tables give participants per arm. Those conventions do not provide the missing event count, effect estimate, or planning outputs for this trial. They also do not change the fact that the registered primary endpoint is described as a median number of ventilator-free days. The registry identifies time-to-event as the interpretation for its analysis kind and says only sponsor-reported hazard ratios are used, but no such hazard ratios are reported in the registry entries. The endpoint and analysis descriptions therefore should not be conflated.

Finally, a confidence interval and a p-value answer different questions. In general, a confidence interval conveys uncertainty around an estimated effect, while a p-value summarizes how incompatible the observed data are with a specified null model, conditional on the analysis assumptions. Neither measure can be reconstructed responsibly without the relevant sponsor analysis and design-aware methods. Since the registry record details do not report the sponsor's estimates or inferential statistics, statistical significance and precision cannot be assessed from them.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists a primary outcome and several secondary and exploratory outcomes. The primary endpoint is ventilator-free days to study day 28. The secondary outcome is 28-day, in-hospital mortality. Exploratory outcomes include intensive care unit mortality, vasopressor-free days, renal replacement therapy-free days, intensive care unit-free days, hospital-free days, and acute kidney injury. The record also reports 14 outcome measures, but the numerical findings for these outcomes are not included in the registry entries used.

When a trial evaluates multiple outcomes or multiple comparisons, the chance of at least one apparently notable result can increase. In general, a prespecified multiplicity strategy can clarify which analyses are confirmatory and how interpretation accounts for multiple testing. The multiplicity procedure, testing hierarchy, and distinction between confirmatory and exploratory analyses beyond the listed labels are not reported in the registry record.

The registry entries do not report how missing primary-outcome data were handled, whether imputation was used, or how participants with incomplete follow-up were treated. The analysis populations—such as the exact populations included in the sponsor's primary analysis—are not reported in the registry record. These omissions prevent assessment of how exclusions, missing measurements, or deviations from assigned strategies might affect the reported comparisons.

The record does not report whether interim analyses were conducted, whether stopping rules were specified, or whether any adjustment was made for interim looks. These details are not reported in the registry record and should not be guessed. The registry also cautions that a simple individual-level re-analysis would be misleading because the design was cluster-randomized and crossover; the sponsor's analysis would be necessary to interpret the treatment comparisons appropriately. Its estimates and methods are not reported in the registry entries used.

11. 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 non-serious adverse event terms

The registry adverse-event entries distinguish serious adverse events from other adverse events. For the lower target, the record lists 0 serious events among 808 at risk and 1 other adverse event among 808 at risk. For the intermediate target, it lists 0 serious events among 859 at risk and 0 other adverse events among 859 at risk. For the higher target, it lists 0 serious events among 874 at risk and 0 other adverse events among 874 at risk.

The record also lists deaths of 285 in the lower-target group, 296 in the intermediate-target group, and 297 in the higher-target group. These entries are presented separately from the serious and other adverse-event counts. The registry entries do not say that these deaths were attributed to an intervention, and the counts alone do not establish whether groups differ. They should not be interpreted as a treatment effect or as an adjudicated safety comparison.

In general, adverse-event counts depend on how events were defined, collected, and classified, as well as on the analysis period and population. The registry entries do not report the event definitions, collection procedures, seriousness criteria, or a statistical comparison. The zero serious-event entries therefore mean that the record lists no serious events in those categories; they do not establish the absence of all possible harms. Likewise, the single other event listed for the lower target should not be treated as evidence of a causal relationship.

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

The registry record establishes that PILOT was designed to compare lower, intermediate, and higher oxygen saturation targets during invasive mechanical ventilation in a randomized cluster-crossover framework. It defines ventilator-free days to study day 28 as the primary endpoint and specifies how death, continued ventilation, and return to ventilation affect that endpoint. It also records enrollment, completion status, secondary outcomes, and adverse-event counts.

The registry entries do not report the primary outcome values by group, a between-group estimate, a confidence interval, or a p-value. The sponsor's own hazard ratios are not reported in the registry record details. Thus, the record details do not show the magnitude or direction of the primary comparison, nor how precisely it was estimated. They cannot support a conclusion that any target improved or worsened outcomes, or that the strategies produced equivalent outcomes.

That limitation is important because the study's design calls for analysis that respects clustering and crossover. A simple re-analysis at the individual level may not reflect the sponsor's intended comparison or the dependence among observations. The registry itself cautions against that approach and states that only the sponsor's own analysis is shown. Without its estimates and methods, the effect and uncertainty cannot be independently interpreted from these registry entries.

The listed adverse-event counts are also descriptive record entries, not proof of cause or comparative safety. The death entries are not stated to be intervention-related. Secondary and exploratory outcomes are named, but their numerical results are not reported in the registry record details. Accordingly, the record details support understanding the question and registered outcome definitions, not a claim about the clinical consequences of selecting any oxygen target.

In general, a result should be interpreted using its estimate, uncertainty, analysis population, and design-aware methods, alongside the outcome definition. Those elements are not available for the primary comparison in the registry entries. No medical decision should be inferred from these incomplete statistical results.

13. Why This Trial Matters Statistically

PILOT addresses a practical statistical problem: how to compare target strategies when care is organized in clusters and the design includes crossover. Cluster-randomized trials can be relevant when the strategy is implemented at a group or setting level, but the resulting observations may not be independent. The registry explicitly notes that a simple individual-level re-analysis would be misleading. This makes the design description essential context for interpreting any estimate.

The endpoint also illustrates the importance of a precise outcome definition. Ventilator-free days combine survival and time off invasive ventilation, and the registry specifies zero values for death before day 28 and continued ventilation at day 28. It further defines how to count after re-ventilation. Without these rules, different analysts could construct different outcomes from the same clinical course.

The design highlights why descriptive summaries alone are not enough. A median and inter-quartile range can characterize the endpoint distribution, but a comparative interpretation also requires an analysis that reflects clustering and crossover, together with uncertainty around the estimated contrast. The registry entries do not report those comparative results, so the statistical importance of the trial's question should not be mistaken for evidence of a particular result.

The record also distinguishes a proposed enrollment of 2,250 in its brief summary from recorded enrollment of 2541. Such entries describe different aspects of the study record and should not be merged into a new total or treated as a treatment effect. More broadly, planning assumptions and observed outcomes answer different questions. Planning calculations can illustrate design requirements, while analysis of trial data addresses the comparison actually observed. The registry entries give some planning assumptions but not numerical planning outputs or observed effect estimates.

For readers assessing a clinical trial, PILOT therefore underscores the need to read outcome definitions, design structure, and analysis reporting together. Its registry record makes the primary question and key design caution clear, while leaving the primary comparative results and important analytical details not reported in the registry record details.

14. 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 main limitation for interpreting this account is that numerical primary-outcome results are not reported in the registry record details. Arm-specific medians, inter-quartile ranges, sponsor estimates, confidence intervals, and p-values are not available. The record details therefore cannot support an estimate of the effect of one oxygen target relative to another.

Although the brief summary describes a cluster-randomized cluster-crossover trial, the registry entries do not give cluster counts, crossover sequences, periods, or details of how the design was implemented. The party masked is not identified. The sponsor's analysis model and its handling of clustering and crossover are not reported in the registry entries. The registry cautions against simple individual-level re-analysis, and that warning limits what can be inferred from aggregate event counts.

The record details also do not report eligibility criteria beyond the respiratory failure condition and invasive ventilation context described in the summary. The analysis population, handling of missing data, protocol deviations, multiplicity procedures, interim analyses, and sample-size justification are not reported in the registry record. These matters may affect interpretation, but their absence from the record details does not establish how the investigators handled them.

The endpoint combines survival with days free from invasive ventilation under specific zero and re-ventilation rules. That definition is clear, but numerical results for the endpoint and listed secondary or exploratory outcomes are not available in the record details. Adverse-event entries lack details about definitions, ascertainment, and attribution. As a result, the listed counts are descriptive and cannot establish comparative safety or causality.

Finally, the record contains a proposed enrollment of 2,250 in the brief summary and recorded enrollment of 2541. The discrepancy between these entries should be retained as stated rather than resolved by assumption. The completed status does not itself supply missing outcome results.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Pragmatic Investigation of optimaL Oxygen Targets Trial — NCT03537937.
• Registry record last updated 2023-09-25; status last verified 2022-11.
• Results first posted 2023-09-25.
• 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.

18. Record Summary

The Pragmatic Investigation of optimaL Oxygen Targets (PILOT) Trial, NCT03537937, was a randomized, single-masked crossover trial sponsored by Vanderbilt University Medical Center. The brief summary describes a cluster-randomized cluster-crossover design. The trial is marked completed, with a recorded start date of 2018-07-01 and primary completion date of 2022-01-31. Recorded enrollment was 2541; the brief summary proposed a 2,250-patient trial.

The trial compared lower, intermediate, and higher arterial oxygen saturation targets of 90% (range 88-92%), 94% (range 92-96%), and 98% (range 96-100%) during invasive mechanical ventilation. The primary outcome was ventilator-free days to study day 28, summarized by median and inter-quartile range. The registry record details do not report numerical primary results, sponsor effect estimates, or confidence intervals. It cautions that simple individual-level re-analysis would be misleading because of the cluster-randomized design. The record lists adverse-event counts and other outcomes, but does not provide the numerical comparative analysis needed to interpret treatment effects.