Use of Non-Invasive Positive Pressure Ventilation in Patients With Severe Obesity Undergoing Upper Endoscopy Procedures: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Use of Non-Invasive Positive Pressure Ventilation in Patients With Severe Obesity Undergoing Upper Endoscopy Procedures (NCT03529461), a randomized na study. The primary outcome is Percentage of Participants With an Oxygen Desaturation Event ≤ 94%.

2. Clinical Question

3. Trial Design

The registry describes a randomized, double-masked, sequential trial of NIPPV during upper endoscopy procedures in patients with severe obesity. The recorded enrollment is 56, with 28 participants in each arm for the primary outcome summaries. The record lists the trial as completed and gives a phase of NA. Bristol Hospital is identified as the sponsor.

The experimental arm received a non-invasive positive pressure nasal mask connected to the machine once the patient was sedated. The control arm received a nasal cannula with 6L O2 and a non-invasive positive pressure nasal mask that was not connected to the machine. Rescue NIPPV through a nasal mask and secondary rescue maneuvers are also listed among the interventions. The primary endpoint covers the period from the start of the procedure, described as anesthesia induction, through completion, described as eyes open to verbal stimuli.

The record gives a start date of 2017-04-25 and a primary completion date of 2018-04-18. It does not report further detail about how randomization was generated or concealed, which individuals were masked, or how the sequential model was implemented. In particular, the word “sequential” alone does not establish that there was a particular stopping rule, interim analysis, or sequence of treatment assignment. Those details are not reported in the registry record.

The outcome summaries are arm-level percentages and event counts. A formal analysis plan, including the original sample-size assumptions and the statistical procedures chosen by the investigators, is not reported in the registry record. The posted comparison includes a risk difference, confidence interval, and unadjusted p-value, but these summaries do not fill in all the missing design and analysis 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.

Before assignment: 208 patients assessed for eligibility: * 56 patients enrolled * 152 patients excluded (reasons: 127 not meeting inclusion criteria or no show for procedure, 6 patients declined to participate, 19 patients excluded because of early trial closure)

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 Percentage of Participants With an Oxygen Desaturation Event ≤ 94%, measured over Time in seconds beginning with the start of procedure (anesthesia induction) ending with procedure completion (eyes open to verbal stimuli)..

The primary endpoint is the percentage of participants who developed a peripheral oxygen saturation at or below 94%, measured by pulse oximetry. Its timeframe begins with the start of the procedure, identified as anesthesia induction, and ends with procedure completion, identified as the point when the participant’s eyes are open to verbal stimuli. The endpoint is therefore a participant-level binary outcome: each participant either met the stated desaturation threshold during that period or did not.

A threshold-based endpoint gives a clinically interpretable way to summarize whether an oxygen saturation event occurred during the defined procedure period. Comparing the percentages between arms describes how often the event was recorded under the two approaches. It does not describe the duration or severity of each event, the lowest saturation observed, or the time until an event. Those measures are not reported for the primary outcome in the registry record.

The endpoint is specifically defined by the threshold at or below 94%; it should not be conflated with the separately listed secondary outcome, which concerns control participants with oxygen saturation below 90% who responded to rescue NIPPV. The registry record does not report additional detail about how pulse-oximetry readings were processed, how repeated readings were handled, or how the endpoint was adjudicated.

7. Results Posted in the Registry

• Experimental: 14.3% (4/28) versus 57.1% (16/28) on control; risk difference -42.9 percentage points (95% CI -65.3 to -20.4), risk ratio 0.25 (0.10–0.65); unadjusted two-proportion z-test p = <0.001.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is binary: participants are classified according to whether they developed oxygen saturation at or below 94% during the specified procedure period. For a binary outcome, an arm-level percentage describes the proportion with the event. The posted summaries report the percentages and event counts for each arm. The record does not report how the investigators handled missing readings or participants without an evaluable endpoint.

What does the risk difference describe?

The risk difference is the experimental-arm event probability minus the control-arm event probability, as reflected by the comparison’s stated arm order. Its posted estimate is -0.429. A value below zero means the observed event proportion was lower in the experimental arm on this scale. Unlike a relative measure, a risk difference describes an absolute contrast in event probabilities. The posted 95% confidence interval runs from -0.653 to -0.204. This interval expresses the uncertainty associated with the estimate under the analysis used; it is not a range containing a fixed proportion of individual participant outcomes.

How do the relative measures differ?

The risk ratio compares the event probability in the experimental arm with the event probability in the control arm. Its reported estimate is 0.25, with a 95% confidence interval from 0.096 to 0.654. A ratio of 1 would correspond to equal event probabilities; values below 1 indicate a lower probability in the numerator arm than in the denominator arm. The odds ratio compares event odds rather than event probabilities. The reported odds ratio is 0.125, with a 95% confidence interval from 0.034 to 0.457. Odds and probabilities are different quantities, so the odds ratio should not be read as though it were the risk ratio.

What does the p-value say?

The unadjusted p-value is <0.001. In general, a p-value describes how incompatible the observed data, or data more extreme, are with a specified null model if that model and the test assumptions hold. It is not the chance that the finding occurred by chance, and it is not a measure of the endpoint’s importance. The record does not report the named test or a fuller statistical analysis plan. The word “unadjusted” indicates that this posted value is not adjusted for multiplicity; no multiplicity adjustment procedure is reported.

Random allocation is intended to support a fair comparison by assigning participants to arms through chance rather than by treatment choice. It does not guarantee that every baseline characteristic is balanced, particularly in a trial with a limited enrollment. Baseline characteristics and any adjusted analyses are not reported in the registry record. The record identifies the trial as double-masked, but does not identify the masked groups or describe how masking was maintained.

In general, over-dispersion is a concern when observed variation is greater than a statistical model expects. It matters for models that estimate uncertainty using a particular distributional assumption, because underestimated variation can make intervals too narrow or tests too optimistic. This endpoint is summarized as a binary participant-level outcome, but the registry record does not identify a model or report a dispersion assessment. It would not be justified to infer a particular over-dispersion problem or correction from the posted summaries.

The registry’s model designation is “sequential.” In general, sequential methods can involve accumulating observations or making decisions in stages, but that general description does not identify the procedures used in this trial. The record does not report a stopping boundary, interim analyses, or a sequential testing method. The label alone should not be taken as evidence that the reported p-value incorporates any particular sequential adjustment.

No recalculated comparison is needed to interpret the posted risk difference, relative measures, confidence intervals, and p-value. The registry record does not report a sponsor-versus-recalculation discrepancy. More broadly, the posted statistical summaries quantify the observed comparison, while important analytic details—including analysis population, missing-data handling, and adjustment choices—are not reported in the registry record. Those omissions limit how fully the estimates’ assumptions and robustness can be evaluated.

10. Confidence Intervals and Precision

The risk difference estimate is -0.429, and its 95% confidence interval extends from -0.653 to -0.204. The interval gives a range of values compatible with the data under the statistical procedure used, subject to that procedure’s assumptions. It communicates uncertainty about the estimated arm contrast; it does not mean that 95% of participants had outcomes inside the interval, nor that there is a 95% probability that the true value lies within this particular calculated interval.

The interval does not include zero, the value representing no risk difference. This is consistent with the posted unadjusted p-value of <0.001. The two summaries address related questions: the interval emphasizes the range and precision of possible effect values, while the p-value summarizes compatibility with a null model. Neither alone describes clinical importance, and neither resolves limitations in the endpoint definition or analysis reporting.

The registry also reports intervals for the risk ratio and odds ratio. The risk-ratio interval is 0.096 to 0.654 around an estimate of 0.25; the odds-ratio interval is 0.034 to 0.457 around an estimate of 0.125. Both intervals lie below 1, their respective no-relative-difference value. Because these measures use different scales, their numerical widths should not be compared directly with the risk-difference interval.

Precision depends in part on the amount of information available and on the number of events, but the registry record does not provide a precision target or the trial’s original sample-size justification. The confidence intervals describe uncertainty for the posted analysis, not uncertainty about unreported outcomes or the extent to which results generalize to other settings.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are participants per arm, not total events. They illustrate the group sizes that would be needed under the stated planning assumptions for different fractions of the observed effect. For 100% of the observed effect, the figures are 19 participants per arm for 80% power and 25 participants per arm for 90% power. For 75% of the observed effect, they are 36 and 48 participants per arm, respectively. For 50% of the observed effect, they are 84 and 112 participants per arm, respectively.

In general, power is the probability that a specified analysis will detect an effect of a chosen size under its assumptions. The 80% power column and 90% power column therefore represent different planning targets: the latter aims for a higher chance of detecting the effect used in the calculation. The listed assumptions are a two-sided significance level of 0.05, equal allocation between the arms, and a normal approximation. Each calculation uses the stated fraction of the observed effect as the effect to detect.

Smaller effects are harder to distinguish from random variation, so the planning illustrations require more participants per arm as the target effect becomes smaller. These calculations are illustrations based on the observed effect, not the trial’s original design assumptions; those original assumptions are not reported in the registry record. They are not observed power.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.571; treatment proportion p₁ = 0.143; average p̄ = 0.357.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.6776 = 1.3281.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.6061 = 0.5101.
4. n = (1.3281 + 0.5101)² / (-0.4286)² = 18.4, rounded up to 19 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
p0 <- 0.571428571428571                          # control proportion
p1 <- 0.142857142857143                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(3529461)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 19 (80% power), 25 (90% power)
 75% of observed effect: n per arm = 36 (80% power), 48 (90% power)
 50% of observed effect: n per arm = 84 (80% power), 112 (90% power)
power.prop.test: 18.4 
Simulated power at n = 19 per arm: 0.841 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Multiplicity, Missing Data and Analysis Populations

The registry record lists one secondary outcome: the percentage of control participants with oxygen saturation below 90% who responded to rescue NIPPV. It also records three outcome measures, but the result for that listed secondary outcome is not reported in the registry record. The primary endpoint is the percentage of participants with oxygen saturation at or below 94% during the procedure period. These endpoints use different thresholds and describe different questions, so their definitions should not be treated as interchangeable.

Multiplicity arises when more than one outcome, comparison, subgroup, or analysis is examined. In general, repeated testing can increase the chance of at least one apparently notable result even when no underlying effect exists. A prespecified testing hierarchy or adjustment can address some multiplicity concerns. The registry record does not report whether the investigators prespecified a hierarchy, how many analyses were planned, or whether any adjustment was used. The posted primary comparison gives an unadjusted p-value; no multiplicity-adjusted result is reported.

The handling of missing data is not reported in the registry record. It is also not reported whether all randomized participants contributed to the primary endpoint, whether any readings were unavailable, or how such cases would have been treated. The analysis population—such as an intention-to-treat population or a population defined by treatment received—is not reported. These details matter because excluding participants or handling missing outcomes differently can affect both the estimated arm percentages and uncertainty.

The registry record does not describe baseline adjustment, subgroup analyses, sensitivity analyses, or the statistical test used for the primary comparison. It reports zero posted analyses in the analysis listing, while also giving arm-level outcome summaries and a comparison with an unadjusted p-value. The available summaries therefore do not amount to a complete statistical analysis plan. No conclusion about omitted procedures or decisions should be inferred from the absence of those details.

14. 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.

The adverse-event summaries report zero serious adverse events, zero other adverse events, and zero deaths in both the control and experimental groups. Each group has 28 participants at risk for the serious and other adverse-event summaries. These are the counts recorded in the registry, not evidence that no event could have occurred outside the collection or reporting process described by the record.

Zero recorded events do not establish that the underlying risk is zero. When event counts are absent, the observed data provide limited information about how uncommon events might be, and they cannot exclude events that were not captured or reported. The registry record does not provide further detail about adverse-event definitions, ascertainment methods, follow-up duration for these summaries, or how events were judged to be related to an intervention.

The safety summaries should be read alongside the limited information about outcome collection and analysis. They describe the posted counts for the two arms, but do not establish a comparative safety profile or resolve what the study could detect for uncommon events. No serious adverse-event terms are listed in the record.

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

The posted primary outcome results describe a lower observed percentage of participants with oxygen saturation at or below 94% in the experimental arm than in the control arm during the specified upper endoscopy period. The experimental-arm percentage was 14.3%, compared with 57.1% in the control arm. The estimated risk difference was -0.429, with a 95% confidence interval from -0.653 to -0.204 and an unadjusted p-value of <0.001. These results give evidence of a difference in the recorded primary endpoint in this trial.

The result is specific to the endpoint definition, the study population, the procedure period, and the two approaches described in the registry. It concerns whether a threshold-based saturation event was recorded, not how long an event lasted, how low oxygen saturation fell, or every aspect of the procedure. Those additional outcomes are not reported for the primary endpoint.

The finding does not by itself establish how the same comparison would perform in different populations, with other procedures, or under other clinical practices. The registry record does not report detailed information about participant characteristics, missing data, analysis populations, or the full statistical plan. These omissions constrain assessment of how robustly the outcome was measured and analyzed.

The posted result also does not answer every question about rescue treatment or safety. A secondary outcome concerning response to rescue NIPPV in control participants with saturation below 90% is listed, but its result is not reported. The adverse-event summaries record zero serious events, zero other events, and zero deaths in each group, but those counts do not prove that the underlying risks are absent. The result should therefore be understood as a comparison of the posted primary endpoint in this completed trial, not as a complete account of all possible outcomes.

16. Why This Trial Matters Statistically

This trial illustrates why the definition of a binary endpoint matters. “Oxygen desaturation” is made operational through a stated pulse-oximetry threshold of at or below 94% and a defined observation period, from anesthesia induction to procedure completion. That precision makes the endpoint interpretable, while also limiting the conclusion to events meeting that threshold within that time window.

The arm-level percentages and risk difference answer related but distinct descriptive questions. The percentages communicate how often the event was recorded in each arm; the risk difference expresses the absolute contrast between those proportions. The risk ratio and odds ratio put the comparison on relative scales. Reporting more than one measure helps readers see that effect magnitude depends on the scale, rather than relying on a p-value alone.

The confidence interval is important because an estimate is not exact. Its range provides information about uncertainty in the risk-difference estimate, while the unadjusted p-value addresses compatibility with a null model. Neither is a substitute for examining the endpoint, the design, or the reporting of analysis choices. In this record, randomization and masking are identified, but details about implementation and the analysis plan are not reported.

The sample-size illustrations also show a general planning principle: detecting a smaller effect requires more participants under the stated assumptions. Those illustrations are based on the observed effect and should not be confused with the trial’s original design justification or with evidence about the trial’s achieved power. Together, these statistical distinctions help keep the interpretation focused on what was measured and how precisely the comparison was estimated.

17. 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.
• Participant counts were reconstructed from posted percentages and arm sizes, so small rounding differences are possible.

The registry record does not report detailed baseline characteristics, so the composition of the enrolled population beyond the stated severe obesity and procedure context cannot be assessed from these registry entries. It does not report how randomization was generated or concealed, or identify which individuals were masked. The randomized and double-masked labels describe the stated design but do not supply those implementation details.

The primary endpoint is based on a fixed saturation threshold, and the record does not describe how readings were sampled, processed, or adjudicated. It does not report duration, nadir saturation, or other measures of desaturation for the primary outcome. The result is consequently limited to the recorded occurrence of the stated threshold event during the specified period.

Missing-data handling, the analysis population, the primary statistical test, and the full analysis plan are not reported. The record also does not give a result for the listed secondary outcome concerning response to rescue NIPPV. Although the posted comparison includes a confidence interval and unadjusted p-value, the absent analysis details constrain evaluation of assumptions and robustness.

The registry’s sequential model designation is not accompanied by details of any sequential procedure, stopping rule, or interim analysis. The original sample-size assumptions are also not reported; the planning illustrations instead use fractions of the observed effect. Finally, the safety summaries list zero recorded events but give no additional details about ascertainment or follow-up. These limits should be considered when interpreting the precision and scope of the reported results.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Use of Non-Invasive Positive Pressure Ventilation in Patients With Severe Obesity Undergoing Upper Endoscopy Procedures — NCT03529461.
• Registry record last updated 2019-05-31; status last verified 2019-05.
• Results first posted 2019-05-31.
• 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.

21. Record Summary

The registry record describes a completed, randomized, double-masked, sequential trial sponsored by Bristol Hospital. It lists enrollment of 56 and compares a control approach using a nasal cannula with 6L O2 plus a nasal mask not connected to a machine against NIPPV through a nasal mask connected once the participant was sedated. The population was patients with severe obesity undergoing upper endoscopy procedures.

The primary endpoint was the percentage of participants with oxygen saturation at or below 94% during the procedure period. The recorded percentages were 57.1% in the control arm and 14.3% in the experimental arm. The posted risk difference was -0.429, with a 95% confidence interval from -0.653 to -0.204 and an unadjusted p-value of <0.001.

Further details about the statistical plan, missing data, analysis population, and implementation of masking and the sequential model are not reported in the registry record.