PRediction of Opioid-induced Respiratory Depression In Patients Monitored by capnoGraphY: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for PRediction of Opioid-induced Respiratory Depression In Patients Monitored by capnoGraphY (NCT02811302), a na na study. The primary outcome is Determine Number of Subjects With RD While on Opioid Therapy.

2. Clinical Question

PRODIGY was designed to develop a score for identifying risk, rather than to compare two assigned treatment strategies. Its target population was patients receiving opioids in the hospital ward under capnography monitoring. The primary objective was to use a derivation cohort to construct the score and a validation cohort to assess it internally.

The primary endpoint concerned respiratory depression episodes during the 48-hour monitoring timeframe. The registry describes several possible physiological thresholds and respiratory opioid-related adverse events that could qualify. An independent Clinical Endpoint Committee assessed the endpoint using device-memory and clinical information.

The registry record does not report the primary endpoint count, the score's performance, or a comparison between patients classified at different risk levels. It therefore documents the intended clinical question and endpoint definition, but not enough results to judge how accurately the score identified risk. The cohort design also does not establish a causal contrast between capnography monitoring and another monitoring strategy.

3. Trial Design

PRODIGY is described as a prospective, multicenter, international, post-market cohort study. The record lists the study as completed, with enrollment of 1495. The recorded start date is 2017-04-07 and the primary completion date is 2018-05-07. The phase is listed as not applicable. The sponsor is Medtronic - MITG.

The study used a single-group model, with allocation listed as not applicable and masking listed as none. These features fit the stated objective of collecting monitoring and clinical data to derive and internally validate a risk assessment score. They do not describe a randomized treatment comparison. The registry record does not report a separate control group or a planned allocation between interventions.

Patients were monitored on the hospital ward using capnography and pulse oximetry. Monitoring data were to be collected for up to 48 hours while patients were receiving opioid therapy, followed by a 1-month follow-up. The named device intervention is the Capnostream Monitor. The primary endpoint was based on Capnostream 20p device-memory data combined with clinical data, and an independent Clinical Endpoint Committee validated endpoint determinations.

The registry describes use of a modified dataset, called mFAS, for derivation and validation. Patients with major deviations or consent withdrawals, and those without continuous monitoring data, were excluded. The record states that 1336 patients were included in the derivation and validation of the tool; it does not give the size of either cohort separately.

The registry record does not report the details of cohort assignment, the score-building procedure, the validation metrics, or a statistical analysis plan. It also does not report a sample-size justification. The study's prospective cohort design and endpoint definition are described, but those details alone do not establish how well the eventual score performed.

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

Analysis population for the primary outcome, as defined in the registry: A modified dataset (mFAS) was used to derive and validate the risk assessment tool. Subjects with major deviations or consent withdrawals; subjects which had no continuous monitoring data were also excluded. A total of 1336 patients were included in the derivation and validation of the tool.

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: Consented subjects that met eligibility criteria.

6. Primary Endpoint

The registry lists the primary outcome as Determine Number of Subjects With RD While on Opioid Therapy, measured over 48 hours.

The primary endpoint was the number of participants with respiratory depression while receiving opioid therapy during the 48-hour timeframe. Continuous monitoring captured end tidal carbon dioxide (etCO2) and oxygen saturation (SpO2), with respiratory rate and apnea criteria also used in endpoint assessment. A positive determination could be made for etCO2 at or below 15 or at or above 60 mmHg for at least 3 minutes; respiratory rate at or below 5 breaths for at least 3 minutes; SpO2 at or below 85% for at least 3 minutes; an apnea episode lasting more than 30 seconds; or any respiratory opioid-related adverse event.

An independent Clinical Endpoint Committee assessed positive respiratory depression determinations, using Capnostream 20p device-memory data together with clinical data. This adjudication is relevant because the endpoint depends on interpreting monitored signals and clinical events against defined criteria. Having a committee assess endpoint status can make the endpoint determination more systematic, though the registry record does not report the adjudication procedures or agreement between assessors.

The endpoint was chosen to support the study's primary objective: deriving a score to identify patients at risk of respiratory depression episodes and internally validating that score. The registry record does not report the observed number of participants meeting the endpoint or the score's predictive performance. Thus, the endpoint definition and intended use are available, but the primary result cannot be assessed from the record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry describes PRODIGY as a prospective, multicenter, international cohort study with a single-group model. The primary objective was score derivation and internal validation. The study did not list a randomized allocation or masking. These design features matter when interpreting the data: the cohort can describe respiratory depression events under the study's monitoring and clinical conditions, but it does not create a randomized contrast for estimating the effect of one monitoring strategy versus another.

What does a binary endpoint mean?

The primary outcome is a participant count: whether a participant had a respiratory depression episode meeting the endpoint definition during the 48-hour timeframe. In general, a binary endpoint reduces each participant's outcome to a category such as event or no event for the specified assessment. The registry record does not report the observed count or the number without an event, so the event frequency cannot be interpreted from the primary endpoint listing alone.

The endpoint definition uses several possible triggers: thresholds for etCO2, respiratory rate and SpO2 maintained for a minimum duration, a prolonged apnea episode, or a respiratory opioid-related adverse event. In general, such a composite definition can capture different clinical and monitored manifestations under one outcome label. Its interpretation depends on the exact criteria and how consistently they are assessed. Here, an independent Clinical Endpoint Committee assessed positive determinations using device-memory and clinical data. The record does not describe committee procedures, blinding during adjudication, or agreement between assessors.

What is derivation and internal validation?

In general, derivation uses observed data to identify predictors and construct a prediction rule. Internal validation evaluates the rule using data from the same study framework, with some method of separating or resampling observations to assess performance beyond the data used to build the rule. Internal validation is useful for identifying possible overfitting, but it is not the same as validation in an independent external population. The registry says the score was to be derived in a derivation cohort and internally validated using a validation cohort. It does not report how participants were assigned to those cohorts, the model-building method, or the validation results.

Prediction performance has multiple aspects. In general, discrimination describes how well a model distinguishes participants who experience an event from those who do not. Calibration describes agreement between predicted risks and observed event frequencies. A useful evaluation may also consider the performance of risk categories and the consequences of classification errors. The registry record does not report these metrics, so no conclusion about the score's accuracy or reliability can be drawn from the available primary result information.

Why does the analysis population matter?

The registry states that a modified dataset, mFAS, was used for derivation and validation. It excludes participants with major deviations or consent withdrawals and those without continuous monitoring data. The record gives 1336 patients included in the tool's derivation and validation, while enrollment is 1495. In general, restricting an analysis to participants with usable monitoring data can affect how representative the analyzed group is of the enrolled cohort. If missing monitoring is related to patient characteristics or event risk, estimates and model performance could differ from those in the full enrolled group. The registry does not report enough detail to determine whether such differences occurred.

How should planning illustrations be understood?

The registry's sample-size assumptions describe a planning illustration based on the observed effect, a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The effect to detect is expressed as a stated fraction of the observed effect. For time-to-event calculations, the stated output is a total number of events; other calculations use participants per arm. These are planning assumptions, not a description of PRODIGY's actual single-group cohort design. In general, a sample-size calculation estimates the amount of information a hypothetical study might need under selected assumptions. It is not observed power and does not demonstrate that the study's prediction score performed well.

The record does not report the score's model form, predictor selection procedure, handling of nonlinearity, missing-data method beyond the stated exclusions, or performance statistics. Statistical concepts such as calibration, discrimination, overfitting, and external validation are general explanations of what would inform a prediction model assessment; the record does not establish that particular procedures were used. The available information therefore supports interpretation of the study's design and endpoint, but not a detailed audit of its modeling methods.

10. Multiplicity, Missing Data and Analysis Populations

The registry record lists 2 outcome measures and 2 analyses, but it does not report a multiplicity adjustment or explain how the analyses were prioritized beyond identifying the primary endpoint. In general, testing multiple outcomes or conducting repeated analyses can increase the chance of a chance-positive finding if each is interpreted as confirmatory. Whether adjustment is necessary depends on the study questions, analysis plan, and intended interpretation. Those details are not reported in the registry record, so it is not possible to determine how multiplicity was addressed.

The record describes a modified dataset for derivation and validation. It says that participants with major deviations or consent withdrawals, and participants without continuous monitoring data, were excluded. The analysis population included 1336 patients. These are reported exclusions, but the registry record does not give a complete accounting of missing values, reasons for absent monitoring, or any imputation strategy. It does not report how incomplete clinical data were handled.

In general, excluding participants with missing data can be appropriate for defined analysis purposes, but it may also change the population represented by the analysis. The direction and size of any effect depend on why data are missing and how missingness relates to outcomes and predictors. The record does not provide enough information to evaluate those relationships.

The registry record also does not provide the sizes of the derivation and validation cohorts separately, the full analysis populations for each outcome, or a detailed protocol for handling participants with multiple qualifying events. These omissions limit reproducibility and assessment of the score development process. The available information describes the analysis population at a high level, but does not establish the complete missing-data or multiplicity strategy.

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 1% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pneumonia (0.5%), ileus (0.4%), postoperative wound infection (0.4%), abdominal pain (0.2%), device related infection (0.2%).

The safety listing describes a blinded capnography monitoring group. It reports 1495 participants at risk for the serious-event listing, with 145 participants affected, and 1495 at risk for the other adverse-event listing, with 181 affected. The record also reports 2 deaths. These are counts as recorded; they should not be combined or converted into new totals.

The serious adverse event terms listed most prominently are pneumonia (0.5%), ileus (0.4%), postoperative wound infection (0.4%), abdominal pain (0.2%), device related infection (0.2%), and sepsis (0.2%). These percentages are reported for the terms in the registry record. The record does not provide enough detail to infer whether the listed terms overlap within participants or to attribute an event to the monitoring device.

In general, adverse-event counts describe events recorded during a study and do not by themselves establish causation. A causal safety comparison would require an appropriate comparator and sufficiently detailed methods for ascertainment and attribution. PRODIGY is described as a single-group cohort, not a randomized comparison; therefore these figures are descriptive rather than a comparative estimate of safety. The registry record does not report further details on event timing, severity grading beyond the serious-event classification, or the investigators' assessment of relatedness.

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

The central contribution described in the registry record is a study framework for developing and internally validating a score intended to identify patients at risk of respiratory depression during opioid therapy on a hospital ward. The study combined continuous respiratory monitoring with clinical information and used an independent Clinical Endpoint Committee to assess the primary endpoint. This clarifies how the investigators intended to define respiratory depression and use the information for prediction.

However, the registry record does not report the primary endpoint count, the score itself, or its predictive performance. Without those results, readers cannot determine how many participants met the endpoint, how well the score separated higher- and lower-risk patients, or whether predicted risks corresponded to observed event frequencies. The described objective is not itself evidence that the score achieved useful prediction.

The study's single-group cohort design also defines the limits of interpretation. The record does not describe random assignment to capnography versus another monitoring strategy. Thus, the cohort does not provide a randomized estimate of whether capnography changes event rates, nor does it isolate the effect of monitoring from other aspects of care. The record should be read as a prediction-score study, not as a treatment-effect comparison.

The registry lists safety events and deaths, but those counts do not establish that monitoring caused or prevented any event. The record also does not provide a comparative safety analysis. Similarly, planning illustrations based on an observed effect are not observed power and are not a substitute for reported primary outcomes.

In general, a prediction model's usefulness depends not only on how it performs in development data, but also on validation, calibration, applicability to the intended clinical setting, and the consequences of decisions based on its predictions. Internal validation is informative about performance within the study framework, but external performance would require evaluation in an independent setting. The record does not report external validation. The study's intended use is clear; the available results do not allow a judgment about accuracy, generalizability, or downstream clinical impact.

13. Why This Trial Matters Statistically

PRODIGY raises a statistical question that is different from the usual comparison of assigned treatments: can routinely collected respiratory-monitoring and clinical information support a score that identifies patients at risk? Answering that question requires a clearly defined event outcome, a suitable development population, and validation that limits overly optimistic assessment of model performance.

The registry provides useful design context for interpreting that objective. It describes prospective monitoring on hospital wards, a specified 48-hour primary endpoint timeframe, and committee assessment of endpoint determinations. It also documents that the tool's derivation and validation used a modified dataset, with exclusions for major deviations or consent withdrawals and absence of continuous monitoring. Those details help readers understand how the intended analysis population differs from enrollment.

Statistically, development and validation should not be conflated. A model can appear to perform well in the data used to create it because it has adapted to features of those observations. Internal validation is intended to assess that concern within the study, while external validation addresses performance in a separate population or setting. The registry describes internal validation, but does not report the method or results. That makes the objective meaningful to recognize, while leaving its success unevaluable from the recorded primary results.

The single-group cohort design is also important. It supports describing and modeling observed risk in the monitored cohort, but it does not establish a randomized contrast between monitoring strategies. This distinction prevents a prediction question from being mistaken for a causal treatment question.

Finally, the safety counts and planning illustrations require careful reading. Descriptive event counts are not proof of causation, and hypothetical sample-size illustrations are not observed power. The record's most important statistical limitation is that the primary event result and model-performance findings are not reported, so the intended prediction contribution cannot be judged.

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 primary limitation for interpreting this study is that the registry record does not report the primary endpoint result or the derived score's performance. It provides the endpoint definition and states that the score was to be derived and internally validated, but does not report event counts, discrimination, calibration, or validation findings. The main prediction objective therefore cannot be evaluated from the record.

The analysis used a modified dataset that excluded participants with major deviations or consent withdrawals and participants without continuous monitoring data. The record states that 1336 patients were included in derivation and validation, but does not provide a detailed flow of exclusions or explain whether excluded patients differed in event risk. This creates uncertainty about how representative the analyzed population was of all enrolled patients.

The record does not report the separate sizes or characteristics of the derivation and validation cohorts, the cohort assignment procedure, model specification, predictor selection, or methods for handling missing clinical information. It also does not describe validation metrics or external validation. These omissions constrain reproducibility and assessment of overfitting and generalizability.

The endpoint uses multiple physiological and clinical criteria. The record identifies an independent Clinical Endpoint Committee, but does not describe its procedures, how conflicting information was resolved, or whether assessors were blinded to other information. Endpoint consistency therefore cannot be assessed in detail.

Finally, the single-group design does not support a randomized comparison of capnography with another monitoring approach. The safety counts are descriptive, and the record does not report a comparative attribution analysis. The study's cohort design is appropriate to the stated prediction objective, but its findings should not be interpreted as a causal effect of monitoring.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: PRediction of Opioid-induced Respiratory Depression In Patients Monitored by capnoGraphY — NCT02811302.
• Registry record last updated 2019-06-12; status last verified 2019-04.
• Results first posted 2019-06-12.
• 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

PRODIGY (NCT02811302) is a completed, prospective, multicenter, international, post-market cohort study sponsored by Medtronic - MITG. The record lists enrollment of 1495, a start date of 2017-04-07, and a primary completion date of 2018-05-07. Its stated objective was to derive and internally validate a score for identifying risk of respiratory depression among patients receiving opioid therapy on hospital wards.

Patients were monitored with capnography and pulse oximetry for up to 48 hours, with a 1-month follow-up. The primary endpoint was respiratory depression assessed using device-memory and clinical information, with determinations validated by an independent Clinical Endpoint Committee. The modified dataset used for derivation and validation included 1336 patients after the exclusions described in the record.

The registry record does not report the primary endpoint result or the score's predictive performance. Safety listings include 145 participants affected by serious adverse events, 181 affected by other adverse events, and 2 deaths. These are descriptive figures from a single-group cohort, not a randomized comparison.