Practical Telemedicine to Improve Control and Engagement for Veterans With Clinic-Refractory Diabetes Mellitus: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Practical Telemedicine to Improve Control and Engagement for Veterans With Clinic-Refractory Diabetes Mellitus (NCT03520413), a randomized na study. The primary outcome is Hemoglobin A1c.

2. Clinical Question

3. Trial Design

The registry describes NCT03520413 as a randomized, parallel-group trial with single masking. It lists an enrollment of 200 and a completed status. The record gives a start date of 2018-12-21 and a primary completion date of 2021-02-01. The listed phase is NA. The sponsor is the VA Office of Research and Development.

The two groups were PRACTICE-DM and Standard VA Home Telehealth. The registry lists 101 participants in the PRACTICE-DM group and 99 in the standard-care group. PRACTICE-DM was a comprehensive telemedicine intervention combining telemonitoring, self-management support, diet and activity support, medication management, and depression support. Standard VA Home Telehealth was described as care coordination and telemonitoring. The comparison is therefore between two telehealth approaches, with the intervention group receiving the described comprehensive bundle.

The primary outcome was Hemoglobin A1c, measured as % HbA1c, with the reported endpoint at 12 months. The stated measure was a mean, and the arm-level results were accompanied by 95% confidence intervals. The registry says mixed models were used and that the analysis used observations from baseline and follow-up timepoints. It reports 200 observations at baseline and 150 observations at 12 months.

The record does not describe the randomization procedure, allocation concealment, or the specific role of masking. Which participants or study personnel were masked is not reported in the registry record. Details such as intervention delivery schedules, adherence, and protocol deviations are also not reported in the registry record. These omissions limit what can be said about implementation and the safeguards used to reduce bias, even though the broad allocation and trial structure are recorded.

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:

Analysis population for the primary outcome, as defined in the registry: Mixed models were used to analyze outcomes. This methodology uses all observations from baseline and the follow-up timepoints. Therefore 200 observations were included at baseline and 150 observations at 12 months.

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: Randomized patients with persistently poorly controlled diabetes mellitus

6. Primary Endpoint

The registry lists the primary outcome as Hemoglobin A1c, measured over 12 months.

Hemoglobin A1c is a laboratory blood test used to measure diabetes control. The registry identifies it as the primary outcome and expresses it as % HbA1c. The reported statistic is the mean, with 95% confidence intervals, and the primary timeframe is 12 months. This outcome offers a laboratory measure that can be compared between the randomized groups rather than relying only on a participant's account of care or self-management.

The registry reports arm-level values of 8.58 for PRACTICE-DM and 9.19 for Standard VA Home Telehealth. Their respective 95% confidence intervals are 8.22 to 8.94 and 8.83 to 9.55. The study's mixed-model analysis is reported as a mean difference in final values. That model used observations from baseline and follow-up, rather than treating the endpoint as an isolated comparison of the 12-month observations alone.

As a general statistical point, a mean describes the average outcome in a group, while a confidence interval conveys uncertainty around an estimate under the assumptions of the analysis. Neither the mean nor its interval directly describes every participant's result. The registry's description does not give a threshold for a clinically important change, so the clinical importance of a particular difference cannot be determined from the endpoint definition alone.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

For a continuous outcome such as Hemoglobin A1c, a mean summarizes the central level of the measure in a group. A mean difference expresses the estimated separation between group means under a specified analysis. The registry reports a mean difference in final values for PRACTICE-DM and Standard VA Home Telehealth. Its estimate is -0.61. The sign indicates the direction of the reported contrast as recorded; the posted arm-level means are 8.58 and 9.19, respectively.

What does the confidence interval convey?

A confidence interval is an uncertainty interval produced by a statistical procedure. In repeated use under the procedure's assumptions, intervals constructed in the same way would cover the target parameter at the stated confidence level at the stated frequency. For this comparison, the reported 95% confidence interval runs from -1.12 to -0.11. It communicates uncertainty about the model-based difference, rather than uncertainty about an individual participant's value. It is not a guarantee that the true effect lies in the interval, and it does not account for every possible source of bias or model misspecification.

The arm-level means also have reported 95% confidence intervals: 8.22 to 8.94 for PRACTICE-DM and 8.83 to 9.55 for Standard VA Home Telehealth. Those intervals describe uncertainty around each posted group mean. They are not a substitute for the confidence interval for the between-group contrast. In general, uncertainty in a difference depends on the uncertainty and relationship between the group estimates, as well as on the analysis method.

Why use a mixed model?

The registry says mixed models were used to analyze outcomes and that the method uses observations from baseline and follow-up timepoints. In general, a mixed model can account for the fact that repeated measurements from the same participant are related, rather than treating every measurement as independent. It can use information across measurement occasions to estimate group patterns or contrasts. The particular model specification is not reported in the registry record, so details about its fixed effects, random effects, covariance structure, or adjustment variables cannot be described.

The record reports 200 observations at baseline and 150 observations at 12 months. These are observation counts at the stated timepoints; the entry does not give participant-level reasons for observations being absent at follow-up. A mixed model can use available observations under particular assumptions about missingness and the model. Its use does not, by itself, show that missing data are harmless or that those assumptions are met. The registry does not report sensitivity analyses for missing data or the missing-data assumptions used.

How should the p-value be read?

The reported p-value is 0.02. As a general explanation, a p-value measures how unusual data at least as incompatible with a specified null hypothesis would be under that hypothesis and the statistical model. It is not the probability that the null hypothesis is true, nor is it a measure of the chance that the result occurred by chance in an everyday sense. Its interpretation depends on the analysis plan, model assumptions, and whether the outcome was selected or tested among multiple outcomes.

The registry labels the analysis EQUIVALENCE, but does not report an equivalence margin or the decision rule used for that label. In general, equivalence testing asks whether the plausible difference is contained within prespecified bounds that represent differences small enough to be considered equivalent for the study's purpose. A nonzero estimate and a p-value do not alone establish equivalence. Without the margin and the testing details, the label cannot be independently evaluated from the registry record.

What can planning calculations tell readers?

Sample-size calculations are planning tools, not evidence that a trial had a particular probability of detecting its result after it was completed. The registry describes planning illustrations based on the observed effect, a two-sided significance level of 0.05, equal allocation, and a normal approximation; the effect to detect is described as a stated fraction of the observed effect. These are assumptions for illustrative planning, not the trial's own design justification. The record does not report the trial's sample-size justification. Any participant counts generated for planning should be read as prospective illustrations, not as observed power or as a guarantee about the precision of this trial's estimate.

In general, over-dispersion means that observed variability is greater than expected under a simpler statistical model. When relevant, failing to account for extra variability can make standard errors too small and confidence intervals too narrow. The registry's primary outcome reports confidence intervals and a mixed-model analysis, but the record does not describe an over-dispersion assessment or identify over-dispersion as a modeling issue. It would therefore be inappropriate to infer that it affected these results.

Finally, interpretation depends on more than the numerical estimate. The analysis population, missing-data procedures, multiplicity adjustment, and prespecified decision criteria shape how confirmatory a result is. Where those details are absent from the registry entry, statistical explanations can clarify the general concepts but cannot fill in the trial's unreported procedures.

10. Analyses Reported by the Sponsor

The registry's own reported comparison is a mixed-model analysis of mean difference in final values. It gives an estimate of -0.61, a 95% confidence interval from -1.12 to -0.11, and p = 0.02. That is the trial's reported model-based result and should be distinguished from a simple unadjusted calculation using posted arm-level summary statistics.

An unadjusted calculation from group summaries does not reproduce a model that uses observations from baseline and follow-up. The two approaches can differ because they use the data differently and may rely on different assumptions. The registry does not report a separate numerical recalculation, so no independent estimate or interval can be quoted as a recalculated result. The posted group means, 8.58 and 9.19, are descriptive arm-level values; they are not a substitute for the model's reported estimate and uncertainty.

The record does not flag a discrepancy between the reported analysis and any recalculation. More importantly, the details necessary to reproduce the mixed model are not reported in the registry record. The most defensible account is therefore to identify the model-based result as the registry's reported analysis and to treat any unadjusted summary-based contrast as a different, less fully specified comparison rather than as a replacement for the trial analysis.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists six outcome measures, five posted secondary outcomes, and six analyses. The primary outcome is Hemoglobin A1c. The posted secondary outcomes are Diabetes Distress Scale, Diabetes Self-Management Questionnaire, Perceived Competence Scale, Body Mass Index, and Patient Health Questionnaire-8. The record does not describe whether the analyses were prespecified in a protocol or statistical analysis plan, or whether adjustment was made for testing multiple outcomes or analyses. A multiplicity strategy is not reported in the registry record.

Multiplicity matters because examining multiple outcomes can increase the chance that at least one result appears notable even when none reflects a reliable underlying effect. This is a general statistical concern, not a claim that the findings in this trial are false. Without the analysis plan and any adjustment procedure, the p-value for the primary analysis should be interpreted as the reported value for that comparison, not as a complete account of all testing across the study.

The primary outcome description says the mixed model used observations from baseline and follow-up. It reports 200 observations at baseline and 150 observations at 12 months. The registry does not explain why fewer observations are reported at the later timepoint, whether observations were missing for particular participants, or what happened to participants who did not contribute a 12-month observation. Reasons for missing observations, missing-data assumptions, and sensitivity analyses are not reported in the registry record.

The record also does not identify the primary analysis population, such as whether the analysis followed randomized assignment regardless of intervention receipt, or whether a different population was used. The listed enrollment and arm counts do not by themselves establish the analysis population. Although mixed models may use available repeated observations, that fact alone does not establish how missing data were handled or whether the results are robust to missing-data assumptions. These reporting gaps limit assessment of bias and the reproducibility of the analysis.

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.

Most frequent serious adverse event terms

Serious events with the highest arm-level rate: hospitalization (4.0%), important medical event (2.0%), persistent or significant illness (1.0%).

The registry reports serious adverse events among 15 of 101 participants at risk in the PRACTICE-DM group and 14 of 99 participants at risk in the Standard VA Home Telehealth group. It reports other adverse events as 0 among 101 at risk in PRACTICE-DM and 0 among 99 at risk in Standard VA Home Telehealth. The registry also records 0 deaths in PRACTICE-DM and 1 death in Standard VA Home Telehealth.

The most frequently listed serious-event terms are Hospitalization at 4.0%, Important Medical Event at 2.0%, and Persistent or Significant Illness at 1.0%. These are the listed terms and percentages; the registry entry does not provide enough detail to determine their overlap, duration, severity beyond the category, or relationship to the intervention. Serious adverse events are not necessarily caused by a study intervention. Attribution is not reported in the registry record.

These figures should be read as counts recorded during the study, not as evidence that either care approach caused or prevented an event. The record does not provide the event-level clinical narratives or further detail about follow-up time for safety reporting. The group counts alone do not establish comparative safety, particularly without context about the types of events, their timing, and how they were assessed.

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

The posted primary outcome results show different group-level Hemoglobin A1c means at the 12-month endpoint: 8.58 for PRACTICE-DM and 9.19 for Standard VA Home Telehealth. The registry's mixed-model analysis reports a mean difference in final values of -0.61, with a 95% confidence interval from -1.12 to -0.11 and p = 0.02. These statistics describe the estimated contrast and its uncertainty under the reported analysis.

They do not reveal the outcome for every participant, explain why individual results varied, or establish that the difference is clinically important. The registry does not report a threshold for a meaningful Hemoglobin A1c difference. Statistical evidence and clinical importance are related but distinct questions: a model estimate and p-value do not answer the clinical-importance question on their own.

The registry labels its analysis EQUIVALENCE. A formal equivalence interpretation depends on a prespecified margin and testing procedure; those details are not reported in the registry record. The reported interval and p-value can be described, but the equivalence label cannot be verified from those values alone without the relevant criteria. Nor should the reported p-value be interpreted as the probability that the groups are equivalent or different.

The findings pertain to the randomized groups and follow-up described in the record. The trial compared a comprehensive telemedicine package with Standard VA Home Telehealth, so it does not isolate the effect of each intervention component. The registry's description of the mixed model indicates that baseline and follow-up observations were used, but details needed to assess missingness, model specification, and analysis population are not reported.

Accordingly, the results support a careful description of the observed and modeled group comparison, not a broad conclusion about all veterans, every telemedicine program, or the effect of any single component. Further interpretation would require context about clinical importance, the prespecified analysis criteria, missing data, and the complete study methods.

14. Why This Trial Matters Statistically

This trial is statistically interesting because its comparison is between two telehealth approaches rather than between telemedicine and no telemedicine. PRACTICE-DM bundled telemonitoring, self-management support, diet and activity support, medication management, and depression support. Standard VA Home Telehealth was described as care coordination and telemonitoring. The primary comparison therefore concerns a comprehensive program against an existing telehealth approach.

The primary outcome is a continuous laboratory measure, and the registry reports both group means and a model-based difference with a confidence interval. Those elements allow readers to distinguish the estimated size and direction of the contrast from the uncertainty around it. They also illustrate why a p-value alone is not a full interpretation: the estimate, interval, clinical context, and analysis assumptions all matter.

The use of mixed models is relevant because the outcome was measured at baseline and follow-up. The registry states that the analysis used observations across those timepoints, making it important to distinguish the reported model result from a simple endpoint-only comparison. At the same time, the absence of details about model specification and missing-data handling constrains independent assessment of how the model produced its estimate.

The EQUIVALENCE label highlights a further reporting issue. Equivalence is a defined statistical objective, not a synonym for finding no difference. A margin and a decision procedure are needed to interpret such a claim, and they are not reported in the registry record. The case therefore illustrates how a numerical result can be reported clearly while an important interpretive element remains unavailable.

More broadly, the record demonstrates why outcome multiplicity, analysis populations, and missing observations matter when interpreting trial results. The available information supports a precise account of the primary comparison, while leaving important questions about confirmatory testing and reproducibility unresolved.

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 provides a concise account of the intervention and comparator, but not the complete protocol or statistical analysis plan. Specific eligibility criteria, participant characteristics, the randomization procedure, and allocation concealment are not reported in the registry record. The record says single masking but does not identify which people were masked. These omissions limit evaluation of selection processes, blinding, and the extent to which the enrolled group represents the target population.

The intervention description identifies a bundle of program components, but does not provide detailed information about how consistently the components were delivered, participant engagement, adherence, or variation in delivery. The comparison is with Standard VA Home Telehealth, which is also an active telehealth approach. The result cannot isolate the contribution of any one component of PRACTICE-DM from the effects of the bundle as a whole.

For the primary analysis, the record states that mixed models used baseline and follow-up observations, and reports 200 observations at baseline and 150 observations at 12 months. It does not describe the model specification, missing-data assumptions, reasons for absent follow-up observations, or sensitivity analyses. The analysis population is not reported in the registry record. These gaps limit reproducibility and assessment of robustness.

The registry reports six outcome measures, five secondary outcomes, and six analyses, but does not report a multiplicity procedure or the prespecified analysis hierarchy. It labels the primary comparison EQUIVALENCE without reporting the equivalence margin or decision rule. The record also does not give a clinical-importance threshold for Hemoglobin A1c. Consequently, the magnitude and formal equivalence interpretation cannot be fully assessed from the registry entry alone.

Safety reporting gives serious-event counts and selected terms, but event narratives, attribution, and additional context are not reported. These limitations are specific to what is documented in the record and should not be read as evidence that the underlying trial did not address the issues.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Practical Telemedicine to Improve Control and Engagement for Veterans With Clinic-Refractory Diabetes Mellitus — NCT03520413.
• Registry record last updated 2023-07-27; status last verified 2023-07.
• Results first posted 2023-01-26.
• 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

NCT03520413 is the completed PRACTICE-DM randomized, parallel-group trial, sponsored by the VA Office of Research and Development. The registry lists enrollment of 200, single masking, and a phase of NA. It gives a start date of 2018-12-21 and a primary completion date of 2021-02-01.

The comparison was PRACTICE-DM, a comprehensive telemedicine intervention, versus Standard VA Home Telehealth. The primary outcome was mean Hemoglobin A1c at 12 months. The reported mixed-model mean difference in final values was -0.61, with a 95% confidence interval from -1.12 to -0.11 and p = 0.02. The registry labels the analysis EQUIVALENCE, but does not report an equivalence margin. It also does not report several details needed to assess the analysis fully, including model specification, missing-data handling, the analysis population, and multiplicity procedures.