Promoting Veteran-Centered Colorectal Cancer Screening: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Promoting Veteran-Centered Colorectal Cancer Screening (NCT02027545), a randomized na study. The primary outcome is Number of Participants With CRC Screening Ordered.

2. Clinical Question

The study addresses a decision-making challenge: screening may have different potential benefits and harms for different people, and the registry describes prior efforts to increase screening as increasing both appropriate and inappropriate use. The intervention was designed to support an informed choice that reflects an individual’s circumstances and discussion with a doctor. The primary endpoint, however, is an order placed after a visit. It does not directly measure the quality of the discussion, whether a patient’s choice was informed, or whether the order matched that person’s preferences.

The trial also listed concordance between screening orders and screening benefit, and the number of participants with colorectal cancer screening utilized, as secondary outcomes. Results for those outcomes are not reported in the registry record. Consequently, the reported primary analysis addresses ordering, not whether screening was ultimately completed or whether decisions were concordant with estimated benefit.

3. Trial Design

The registry describes a randomized, parallel-group study and identifies the design as cluster-randomized. Patients were grouped according to whether their primary care provider was assigned to the Decision Aid intervention or the pragmatic control. Randomization at the provider level means that the assignment unit was a provider cluster, not necessarily an individual patient. The registry enrollment is 436; it does not report the number of provider clusters or their sizes.

The Decision Aid intervention combined an individualized decision aid with provider education and a modified performance measure or reminder. The control also included provider education and the modified measure or reminder, but not the decision aid. This shared structure makes the comparison specifically relevant to the added decision aid within the broader approach described in the registry. A simple description of the arms should not imply that all other elements differed.

Masking is listed as none. In a behavioral intervention involving an individualized aid and provider education, the absence of masking is relevant when considering how the intervention was delivered and how outcomes could be influenced. The primary endpoint was assessed using manual electronic health record review, but the registry does not report who conducted that review or whether reviewers knew the assigned group.

The study is listed as completed. Its recorded start date is 2015-11-20, and its primary completion date is 2018-08-08. The registry gives the condition as colorectal cancer and lists the VA Office of Research and Development as sponsor. The phase is recorded as NA, consistent with the nature of the behavioral intervention rather than a drug-development phase.

The registry record does not report the number of clusters, the randomization procedure in further detail, whether any participants or providers changed groups, or the extent of variation in cluster size. It also does not state how clustering was handled in the logistic regression. Those details matter for evaluating uncertainty in a cluster-randomized study, so the reported p-value should be read alongside these gaps rather than as a complete account of the design or analysis.

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

Reasons for not completing this period:

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 Number of Participants With CRC Screening Ordered, measured over 2 weeks.

The primary endpoint is the number of participants with colorectal cancer screening ordered. The registry’s description clarifies the underlying measure: whether screening was ordered within 2 weeks after the clinic visit. This is a binary outcome, because each record is classified according to whether an order was present in the specified interval. Orders were identified by manual review of electronic health records.

An order is a practical measure of a clinical action following a visit. In this study, it can indicate whether a screening decision translated into an order in the health record. It is also closely connected to the intervention’s intended point of influence: personalized information and space for a patient and doctor to discuss whether screening is appropriate for that individual.

But an order is not the same as screening being completed. The registry separately lists screening utilized as a secondary outcome, underscoring that ordering and use are distinct events. Nor does the primary endpoint by itself establish whether the order reflected an informed preference, whether screening was appropriate for an individual, or whether the discussion was patient-centered. Concordance between orders and screening benefit was also listed as a secondary outcome, but its results are not reported in the registry record.

The short outcome window makes the endpoint specific to ordering after the visit; it does not capture longer-term outcomes. The registry record does not report the exact operational rules used to classify orders, how ambiguous records were handled, or whether reviewers were masked to group assignment. These details would help assess consistency of endpoint measurement.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The reported primary analysis is logistic regression. The endpoint is binary: screening was either ordered within 2 weeks after the clinic visit or it was not. In general, logistic regression models the probability of a binary outcome in relation to explanatory variables. It commonly expresses comparisons on an odds scale, but the registry does not report the fitted coefficient, odds ratio, other effect estimate, covariates, or confidence interval. It reports only p = 0.491 for the superiority analysis.

What does the p-value convey?

In general, a p-value describes how unusual the observed data, or more extreme data, would be under a specified null hypothesis and statistical model. A smaller value can indicate greater incompatibility with that null model, provided the model and analysis assumptions are suitable. The p-value is not the probability that the intervention has no effect, and it does not quantify the effect’s size. The registry’s p = 0.491 is not enough by itself to determine whether plausible effects are small, large, beneficial, or harmful. An estimate with a confidence interval would be needed to judge direction and precision more directly.

Why does clustering matter?

The study is described as cluster-randomized, with primary care providers assigned to intervention groups and patients belonging to those providers’ groups. In general, outcomes within a cluster can be correlated because people treated by the same provider may share care processes or other influences. This correlation means that observations within a cluster do not necessarily contribute as much independent information as observations from different clusters.

Statistical methods for cluster-randomized trials should reflect the assignment structure and account for within-cluster dependence when estimating uncertainty. If an analysis instead treats all patients as independent, standard errors and p-values may not properly reflect the design. The registry says the sponsor used logistic regression, but the registry record does not describe the clustering adjustment, the number of clusters, or whether other design features were incorporated. The sponsor’s own analysis is therefore the relevant reported result; a simple unadjusted patient-level recalculation would be misleading.

What can confidence intervals add?

In general, a confidence interval gives a range of effect estimates compatible with the data under a particular statistical procedure. Its width communicates precision: a narrower interval indicates more precise estimation, while a wider interval indicates greater uncertainty. A confidence interval can also show whether the data are compatible with effects in more than one direction or with effects of differing practical importance. No confidence interval is reported for this primary analysis, so the precision of the estimated comparison cannot be assessed from the registry result.

Why distinguish orders from use?

The measured primary endpoint is an order, not completed screening. This distinction matters because an order is an intermediate clinical action; completion is a separate outcome. In general, an intervention can influence ordering without having the same effect on subsequent use, and the reverse may also be possible depending on care processes. The registry lists screening utilized as a secondary outcome, as well as concordance between screening orders and screening benefit. It does not report results for either. The primary p-value should not be interpreted as a finding about completion or concordance.

What analysis details are missing?

The registry record does not report the adjustment variables, the analysis population, the handling of missing or incomplete records, or whether the endpoint was analyzed according to randomized assignment. It also does not report whether the model included provider-level effects or another method to account for clustering. These details affect the assumptions behind the model and the interpretation of its uncertainty. The registry record does not report an interim-analysis plan, a statistical analysis plan, or a sample-size justification. None should be inferred from the reported result.

The registry lists three outcome measures and two analyses, but does not describe a multiplicity strategy or the relationship between each analysis and each outcome. In general, when several outcomes or analyses are considered, the chance of at least one apparently notable result can differ from what would be expected for a single prespecified test. Without a reported plan, it is not possible to determine how multiplicity was addressed. The available primary p-value remains a result for the reported analysis, not proof about all outcomes or comparisons.

10. Exact Estimates for Each Arm

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

R code and output

arm    <- c("Decision Aid", "No Decision Aid")
n      <- c(258, 173)
events <- c(162, 114)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  Decision Aid  258  162  62.8  56.6  68.7  56.7  68.5  56.7  68.5
ARM  No Decision Aid  173  114  65.9  58.3  72.9  58.6  72.5  58.5  72.5

11. Analyses Reported by the Sponsor

The registry reports the sponsor’s primary comparison as logistic regression with p = 0.491. It does not report the corresponding effect estimate, confidence interval, model details, or a numerical reanalysis. Because the trial is cluster-randomized, a simple re-calculation from individual-level summary statistics would not adequately represent the assignment structure and could give misleading uncertainty. The sponsor’s own analysis therefore takes precedence over such a simple re-calculation.

This does not make the reported p-value a complete description of the result. Without an estimate and confidence interval, readers cannot judge the size, direction, or precision of the comparison from the registry record. The registry also does not explain how the logistic regression accounted for clustering, so the model’s treatment of the design cannot be independently evaluated from the information recorded.

The p-value pertains to the primary outcome of screening ordered within 2 weeks after the clinic visit. It should not be treated as an estimate of screening completion, decision quality, or concordance between orders and screening benefit. Those secondary outcomes are listed, but their results are not reported in the registry record.

12. Multiplicity, Missing Data and Analysis Populations

The registry lists three outcome measures: the primary measure of participants with screening ordered, and secondary measures concerning concordance between screening orders and screening benefit, and screening utilized. It reports two analyses. The record does not describe how those analyses map to the listed outcomes, whether any outcome was designated for a particular analysis beyond the primary endpoint description, or how multiplicity was handled.

In general, testing multiple outcomes or conducting multiple analyses can increase the chance of observing a low p-value even when no underlying effect exists. A prespecified testing hierarchy or adjustment procedure can help clarify how evidence should be interpreted across several comparisons. The registry record does not report such a procedure, so the primary p-value should be understood as the result reported for that analysis rather than as a multiplicity-adjusted conclusion about all outcomes.

The analysis population is not reported in the registry record. It is also not reported how missing, incomplete, or ambiguous electronic health record data were handled, or whether participants were analyzed according to their provider’s randomized assignment. These omissions leave uncertainty about which records contributed to the logistic regression and how missing endpoint information could have affected the result.

The registry record does not report an interim-analysis plan, stopping rules, or the statistical analysis plan. It does not report a sample-size justification, so the enrollment figure alone does not establish what effect the study was designed to detect or how precisely the primary comparison could be estimated. The sample-size assumptions associated with planning illustrations are not the trial’s own design justification and should not be read as observed power.

These reporting gaps do not change the numerical result that the registry records, but they restrict its interpretation. More detail about outcome handling, the analysis population, multiplicity, and prespecified methods would help readers assess the risk of bias and the reliability of the statistical uncertainty for both primary and secondary outcomes.

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

The posted results report a logistic-regression p-value of 0.491 for the primary comparison of whether colorectal cancer screening was ordered within 2 weeks after a clinic visit. This is a result for the endpoint and analysis described in the registry. It is not an effect estimate, and the record provides no confidence interval. The magnitude, direction, and precision of the comparison therefore cannot be determined from the reported result.

A p-value alone does not establish that the intervention and control had the same effect. Nor does it establish that the intervention changed ordering. It describes the data’s compatibility with a specified null model, conditional on the model and its assumptions. In this case, important details of the logistic regression, including how it addressed provider-level clustering, are not reported.

The result is limited to ordering. It does not show whether screening was completed, whether a patient’s choice was informed, whether the order reflected personal preferences, or whether the order was concordant with screening benefit. The registry lists screening utilization and concordance as secondary outcomes but does not report their results. No conclusion about those outcomes follows from the primary p-value.

The study’s central idea was to personalize information about screening benefits and harms and support discussion between patient and doctor. The primary endpoint captures a subsequent record of an order, not the quality of that conversation or the appropriateness of the decision for a particular person. Those concepts require different measures and cannot be inferred from the ordering endpoint alone.

Finally, the cluster-randomized design and limited analysis reporting warrant caution. The record does not specify the number of provider clusters, the handling of missing data, the analysis population, or the model’s treatment of clustering. The registry result should therefore be read as a reported statistical test for one endpoint, with important uncertainty about effect size, precision, and broader clinical meaning.

14. Why This Trial Matters Statistically

This trial is statistically notable because randomization occurred at the primary-care-provider level while the primary outcome was measured for patients. That distinction between the assignment unit and the outcome unit is central to the analysis. In general, clustering can make patient outcomes within the same provider more alike, which affects how much independent information the data contain and how uncertainty should be calculated.

The intervention also illustrates why an endpoint should be interpreted in relation to the question it measures. Screening orders are a concrete clinical action, but the study’s rationale concerns individualized decisions in the context of benefits, harms, health, and screening history. An order may be relevant to that process, but it does not capture informed choice or completion by itself. The registry’s separate listing of utilization and concordance outcomes reflects these distinctions; their results are not reported.

There is also a practical comparison embedded in the design. Both groups received provider education and a modified performance measure or reminder, while the Decision Aid group additionally received an individualized aid. The primary comparison therefore concerns the added decision aid in the context of shared intervention elements. Treating the control as no intervention would misstate the design.

Finally, the available registry result illustrates the limits of a p-value without an effect estimate and confidence interval. The reported p-value is interpretable as a test result, but it does not communicate the size or precision of an effect. For a cluster-randomized study, reporting the estimate, interval, model details, and treatment of clustering would be especially useful for judging the evidence.

15. Limitations

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

The registry record provides limited detail for evaluating the primary statistical result. It reports a logistic-regression p-value but no effect estimate or confidence interval. It does not describe the model covariates, the exact method used to account for provider-level clustering, or the number and sizes of clusters. These gaps constrain assessment of both the estimated comparison and its uncertainty.

The record does not report the analysis population, how missing or ambiguous electronic health record information was handled, or whether the endpoint review was blinded to assignment. The outcome was identified by manual record review, but the review procedures and classification rules are not described in further detail. These issues matter for understanding whether endpoint measurement was consistent across groups.

The primary outcome captures orders placed within 2 weeks after a visit, not screening completion or the quality of the decision process. Secondary outcomes address utilization and concordance, but their results are not reported in the registry record. The available result therefore does not cover all the outcomes relevant to the intervention’s stated purpose.

The registry record does not report a statistical analysis plan, sample-size justification, or multiplicity procedure. It also lists masking as none. Together, these reporting limits make it difficult to assess how outcomes and analyses were selected, how potential missingness was addressed, and how much confidence to place in the reported statistical test. They do not establish that the analysis was incorrect, but they leave important questions unanswered.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Promoting Veteran-Centered Colorectal Cancer Screening — NCT02027545.
• Registry record last updated 2023-05-03; status last verified 2023-04.
• Results first posted 2020-01-10.
• 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

This completed, randomized, parallel-group, cluster-randomized study evaluated a patient-centered approach to colorectal cancer screening among older adults ages 70 to 75 who were due for screening. The registry records enrollment of 436 and lists the VA Office of Research and Development as sponsor.

The Decision Aid group received an individualized decision aid, provider education, and a modified performance measure or reminder. The pragmatic control received provider education and the modified measure or reminder, without the aid. Masking is listed as none.

The primary endpoint was whether screening was ordered within 2 weeks after the clinic visit, assessed through manual review of electronic health records. The registry reports logistic regression with p = 0.491, but no effect estimate or confidence interval. It also lists secondary outcomes for concordance between orders and screening benefit and for screening utilized; their results are not reported in the registry record.

Because assignment was clustered by provider, a simple individual-level recalculation would be misleading. The registry does not report key analysis details, including how clustering and missing data were handled.