The CAPTURE Study: Validating a Unique COPD Case Finding Tool in Primary Care (Aim 3): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for The CAPTURE Study: Validating a Unique COPD Case Finding Tool in Primary Care (Aim 3) (NCT03583099), a randomized na study. The primary outcome is Proportion of CAPTURE+ Participants Who Meet a Composite Endpoint for Diagnosis and Management of COPD.

2. Clinical Question

The composite includes referral for clinical spirometry testing, a new COPD diagnosis, newly prescribed respiratory medication, referral to a specialist for respiratory evaluation or treatment, or referral for or completion of pulmonary rehabilitation. The central statistical question is whether the group-level proportions differed under the two practice education approaches. The registry reports an estimated difference and its confidence interval from a generalized estimating equation contrast. The record does not identify the order in which the two arms enter that contrast, so the sign alone should not be used to assign the estimate to a particular arm.

3. Trial Design

This was a prospective, multicenter study in primary care settings, combining cross-sectional validation of CAPTURE with a cluster-randomized controlled clinical trial. The registry describes the validation aim as defining sensitivity and specificity for identifying previously undiagnosed patients with clinically significant COPD, alongside assessing impact on clinical care. The reported primary result addressed the care-impact component among CAPTURE+ participants.

The record lists randomized allocation, double masking, and a parallel model. It describes two practice-level approaches: basic COPD education alone, and basic COPD education with patient-level CAPTURE information and interpretation education. The registry does not report who was masked, how masking was implemented, or the number of practices assigned to each arm. The trial was listed as phase NA.

Enrollment was 2004. The recorded study start was 2018-10-12, primary completion was 2023-04-07, and status was completed. The listed sponsor was Weill Medical College of Cornell University. The registry describes the clinical setting as a broad range of primary care settings, but the specific sites and their characteristics are not reported in the registry record.

The primary endpoint was measured from baseline to 12 months and was defined as meeting any of five diagnosis or management events. The listed secondary outcomes include referrals or completion of spirometry, new COPD diagnosis, new respiratory medication, specialist referral, pulmonary rehabilitation, influenza vaccination recommendation or administration, and outcomes concerning participants who were smoking at baseline. The registry lists 150 outcome measures and 131 analyses. It does not explain how those counts correspond to the outcomes analyzed in the reported primary contrast.

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.

Population: Participants with a spirometry consistent with obstruction (FEV1/FVC \< 0.70); spirometry FEV1 \<80% predicted, as well as a 5% random sample of all other participants underwent longitudinal follow-up at 12-months.

6. Primary Endpoint

The registry lists the primary outcome as Proportion of CAPTURE+ Participants Who Meet a Composite Endpoint for Diagnosis and Management of COPD, measured over Baseline to 12 months.

The primary endpoint is the proportion of CAPTURE+ participants who met at least one component of a composite for COPD diagnosis and management between baseline and 12 months. A participant could meet the endpoint through referral for clinical spirometry, a new COPD diagnosis, a newly prescribed long-acting bronchodilator or anti-inflammatory medication for a respiratory condition, referral to a respiratory specialist, or referral for or completion of pulmonary rehabilitation.

A composite can capture several types of clinical response within one outcome, rather than treating each event as the sole measure of care impact. It is useful when the question concerns whether any of several specified actions occurred. Its interpretation depends on the listed components: the endpoint does not distinguish a participant who met one component from a participant who met more than one, and the reported proportion does not by itself show which component contributed most.

The population is restricted to CAPTURE+ participants, defined using CAPTURE score and peak expiratory flow criteria. The measure therefore concerns care following a positive screen under that definition; it is not a proportion for all enrolled participants. The registry names the measure as a number with a 95% confidence interval and gives the time frame as baseline to 12 months. It does not report component-specific primary results in the reported contrast.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry reports a generalized estimating equation contrast for the primary comparison. The trial was cluster randomized, so the assignment was made at the practice level rather than independently for each participant. Outcomes measured on participants from one practice may be more alike than outcomes from participants in different practices. Treating every participant as statistically independent can therefore give misleading uncertainty estimates.

What does a generalized estimating equation do?

In general, generalized estimating equations estimate an average relationship between an outcome and comparison groups while allowing observations within a cluster to be correlated. For a clustered trial, the cluster is the unit within which responses may be related. A contrast then expresses the estimated difference between the groups under the chosen model. The method's behavior depends on choices such as outcome scale, link function, working correlation structure, and variance estimation. Those specific choices are not reported in the registry record.

The reported measure is a difference in proportion. In general, a difference compares the outcome frequency between groups on a common scale; a value of zero corresponds to no difference. The result is reported as 4.0, with a 95% confidence interval from -6.9 to 15.0. Because the order of arms in the contrast is not mapped to the arm descriptions in the registry result, the positive sign cannot establish which arm had the higher proportion. The interval communicates uncertainty around the estimated contrast rather than a range containing a fixed proportion of individual outcomes.

How should the confidence interval be read?

In general, a confidence interval describes the precision of an estimate under the statistical model and its assumptions. This interval includes zero and also includes values on either side of zero. Thus the posted estimate is compatible with no difference as well as with differences in either direction on the reported scale. The interval does not show that every value inside it is equally plausible, nor does it establish that the true effect must lie within it. Its interpretation relies on the analysis being appropriate for the design and data.

The reported p-value is 0.47. In general, a p-value measures how unusual data at least as discordant with a specified null model would be if that model were true, subject to the analysis assumptions. It is not the probability that the null hypothesis is true, nor the probability that the intervention works. A p-value alone does not describe the size or practical importance of an effect; the estimate and interval are needed for that context.

Why does clustering matter?

General statistical explanation: clustering affects the amount of independent information in a study. When participants share a practice, their responses may be correlated because they experience common clinicians, procedures, or local context. A method that recognizes the grouping can produce uncertainty estimates more appropriate to the randomized design. The registry names a generalized estimating equation contrast, which is consistent with an analysis that recognizes grouped observations. The exact implementation, number of clusters, and any adjustments are not reported in the registry record. A simple individual-level re-analysis would be misleading for this trial, so the sponsor's own analysis is the reported comparison.

The randomization and outcome levels also matter. Practices were randomized, while the primary endpoint is defined for CAPTURE+ participants. The registry does not state the number of randomized practices, the distribution of CAPTURE+ participants across them, or whether the analysis used any additional weighting. These details can affect how an average difference is interpreted and how precisely it is estimated.

What is the role of the composite endpoint?

In general, combining several events can provide a single outcome for a broad clinical-care question. Here, any one of five specified diagnosis or management actions qualifies. That makes the estimate a comparison of the proportion meeting at least one component, not a comparison of each action separately. Components may differ in frequency or clinical meaning; without component results, the aggregate does not reveal what drove the contrast. The record does not report a component-level breakdown for the primary result.

The CAPTURE+ definition sets the population for the primary endpoint. It includes participants with CAPTURE score ≥ 5, or a score of 2, 3, or 4 together with peak expiratory flow below the sex-specific thresholds stated in the endpoint description. The record says higher CAPTURE scores represent a higher likelihood of clinically significant COPD. The primary result is consequently conditional on this screen-positive definition, rather than an estimate across all enrolled participants.

What can planning illustrations establish?

Sample-size figures based on an observed effect are planning illustrations, not observed power and not a description of the trial's original design. The stated illustration assumptions are a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation, with the effect to detect expressed as a fraction of the observed effect. For non-time-to-event outcomes, the stated convention is participants per arm; time-to-event displays use total events. These assumptions do not establish what assumptions were used to plan this trial. The protocol's sample-size justification is not reported in the registry record.

More generally, an illustration based on an observed estimate is sensitive to uncertainty in that estimate. It should not be interpreted as a retrospective probability that the trial detected an effect, and it cannot replace a prospective design calculation. The registry record does not report the original target enrollment rationale, interim analyses, or a detailed statistical analysis plan.

The primary result should therefore be interpreted with the reported cluster-aware method, the confidence interval, the outcome definition, and the limits of the registry entries in view. The estimate is a group contrast for a composite outcome in a defined screened-positive population. It does not independently explain the component events or resolve unreported analysis choices.

10. Analyses Reported by the Sponsor

The registry reports the sponsor's own generalized estimating equation contrast: a difference in proportion of 4.0, a 95% confidence interval from -6.9 to 15.0, and p = 0.47. Because allocation was cluster randomized, a simple individual-level re-analysis would be misleading. The comparison is therefore presented using the sponsor's cluster-aware analysis rather than an unadjusted re-calculation from summary statistics.

The record does not provide the group-specific primary proportions or enough model detail to reproduce the generalized estimating equation contrast independently. It also does not map the analysis contrast order to the named intervention and comparator arms. The numerical result can be described as the reported between-arm contrast, but its positive direction should not be attributed to a particular arm on the available record.

The confidence interval includes zero and spans values in either direction. The p-value is 0.47. These statistics indicate uncertainty about the direction and size of the difference; they do not establish equivalence or prove that no difference exists. No conflict between a sponsor result and a simple re-calculation is presented, because that individual-level re-analysis is not appropriate for this clustered design.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 150 outcome measures and 131 analyses, alongside multiple secondary outcomes. A large set of outcomes or analyses creates the general statistical issue of multiplicity: when many hypotheses are examined, the chance of finding apparently notable results by chance can increase. The record does not report whether multiplicity adjustments were used, which outcomes were included in any family of tests, or whether the reported primary result was protected by a prespecified testing hierarchy. Those details are not reported in the registry record.

The registry identifies a primary endpoint and reports a contrast for it. It does not provide enough information to determine whether the endpoint definition, analysis, and comparison were prespecified in a statistical analysis plan, or whether any changes were made. Interim analyses and stopping rules are not reported in the registry record. No assumptions about these procedures can be made from the listed outcome and analysis counts.

Missing-data handling is also not reported in the registry record. The entry does not state how participants without a recorded 12-month endpoint were handled, whether imputation was used, or how loss to follow-up was assessed. The number contributing to the primary analysis and the analysis population definitions are not reported in the registry record. Consequently, the reported estimate cannot be fully assessed for sensitivity to missing outcomes or differences between enrollment and analysis populations.

The primary result applies to CAPTURE+ participants, but the record does not give the number of such participants in each arm or explain whether the analysis followed an intention-to-treat principle at the cluster or participant level. It also does not report whether all randomized practices contributed data. These are relevant details for judging how the analysis population relates to the randomized design. Their absence limits evaluation, but does not by itself establish that the analysis was biased or inappropriate.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

The adverse-event record lists 1118 participants at risk in the COPD education plus CAPTURE education and patient-level screening-results arm, and 886 at risk in the COPD education arm. Serious adverse events are recorded as 0 in both groups, and other adverse events are also recorded as 0 in both groups. The death count is 2 in each group.

These figures are counts as entered in the registry and should be read with their listed denominators. The record does not provide event descriptions, timing, causes of death, relatedness assessments, or additional details about ascertainment. In particular, the death counts alone do not establish whether deaths were related to the study intervention. Zero recorded events in these fields should not be interpreted as proof that no events occurred outside the information recorded in the registry.

The adverse-event figures are separate from the primary endpoint comparison. They do not determine the interpretation of the generalized estimating equation contrast for the composite diagnosis and management outcome. The registry does not provide further safety analyses in the information reported.

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

The reported result is an estimated contrast between the randomized practice groups for the proportion of CAPTURE+ participants meeting at least one component of a specified COPD diagnosis and management endpoint within 12 months. The estimate is 4.0, with a 95% confidence interval from -6.9 to 15.0 and p = 0.47. The interval includes zero, so the posted result does not establish a clear difference between groups. At the same time, the interval extends in both directions and does not demonstrate that the groups are equivalent.

The endpoint reflects whether any one of several listed care actions occurred. It does not tell which action was most common or whether every component changed in the same direction. Without component-specific estimates, the overall contrast cannot identify whether spirometry referral, diagnosis, medication, specialist referral, or pulmonary rehabilitation contributed most. It also does not directly quantify CAPTURE's sensitivity or specificity; those belong to the study's validation aim, not the reported primary contrast.

The finding is specific to the CAPTURE+ population and the baseline-to-12-month period. It should not be generalized from the primary endpoint to every enrolled participant without additional results. The registry does not report group-specific proportions, the number of CAPTURE+ participants analyzed, missing-data handling, or detailed model specifications. Those omissions limit assessment of the result's precision and reproducibility.

Statistical uncertainty is not the same as evidence of no effect. The p-value does not express the probability that either approach is superior, and the confidence interval is not a guarantee about the true difference. The result should be read as a sponsor-reported, cluster-aware estimate with substantial uncertainty about direction and size, rather than as proof for or against a particular clinical approach.

14. Why This Trial Matters Statistically

This trial raises an important design issue for clinical research conducted in primary care: the intervention is delivered through practices, while outcomes are recorded for participants. When the unit of randomization and the unit of measurement differ, analysis must reflect the grouping. A generalized estimating equation contrast is reported for the primary result, and the cluster-randomized structure is a reason not to rely on a simple individual-level re-analysis.

It also illustrates how a screening and care pathway can be evaluated using a composite endpoint. The components represent different actions, from diagnostic testing to referral and treatment-related management. Combining them provides a broad measure of whether any defined response occurred, while making component-level results important for understanding what the overall contrast means. Those component-specific primary results are not reported in the registry entry.

The result demonstrates why an estimate should be read alongside its uncertainty. A point estimate of 4.0 may sound definite in isolation, but its 95% confidence interval, from -6.9 to 15.0, includes zero and spans effects in either direction. The p-value of 0.47 does not resolve that uncertainty. The result is not proof of equivalence, and it is not a basis for claiming that one arm was better.

Finally, this record shows the distinction between a registry summary and a full statistical report. The record identifies the design, outcome, and analysis method, but important details about the model, analysis population, missing data, and multiplicity are not reported in the registry record. Readers need those details to fully appraise how the analysis was implemented and how broadly the findings apply.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The registry record provides limited detail for independently assessing the primary analysis. It names a generalized estimating equation contrast but does not report the model's link function, working correlation structure, variance estimator, covariate adjustment, or the number of randomized practices. These details matter in a cluster-randomized trial because they describe how the analysis represents within-practice correlation and uncertainty.

The reported result does not include group-specific primary proportions, the number of CAPTURE+ participants contributing to the comparison, or a breakdown by the five endpoint components. As a result, readers cannot determine the component contributions or verify the contrast from arm-level summaries. The contrast order is not mapped to the named arms, which prevents attributing its positive sign to a particular group.

The record does not report missing-data methods, analysis population definitions, multiplicity procedures, interim analyses, or the original sample-size justification. The listed enrollment alone does not establish the number of participants included in the primary analysis. The registry also lists many outcome measures and analyses but does not explain their relationship to the reported primary comparison.

The trial's primary result concerns CAPTURE+ participants and a composite care endpoint over baseline to 12 months. It is not, by itself, a result for all enrolled participants or a direct estimate of the screening tool's sensitivity and specificity. The entry describes a broad range of primary care settings but does not provide site-level characteristics. These limits are specific to the reported registry information and do not establish that the full study report lacks further detail.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: The CAPTURE Study: Validating a Unique COPD Case Finding Tool in Primary Care (Aim 3) — NCT03583099.
• Registry record last updated 2024-09-19; status last verified 2024-09.
• Results first posted 2024-09-19.
• 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 a completed, prospective, multicenter, cluster-randomized parallel trial in primary care, with enrollment of 2004. Practices were assigned to basic COPD education alone or basic COPD education plus patient-level CAPTURE information and interpretation education. The primary endpoint was the proportion of CAPTURE+ participants meeting at least one of five COPD diagnosis and management events from baseline to 12 months.

The reported generalized estimating equation contrast is a difference in proportion of 4.0, with a 95% confidence interval from -6.9 to 15.0 and p = 0.47. The interval includes zero and values in either direction; the result does not establish equivalence. The contrast order is not mapped to the named arms in the registry entry.

The record does not report full model details, primary analysis population size, missing-data handling, component-specific primary results, or multiplicity procedures. The posted adverse-event counts are 0 serious events and 0 other events in each listed group, with 2 deaths in each group.