Diabetes Prevention Program Lifestyle Intervention in the Marshallese Population: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Diabetes Prevention Program Lifestyle Intervention in the Marshallese Population (NCT03270436), a randomized na study. The primary outcome is Change in Mean Body Weight (kg).

2. Clinical Question

This is a comparative-effectiveness question: both groups received an active lifestyle program, and the comparison concerns their outcomes relative to one another. It is not a comparison of either program with no intervention or usual care. The trial also planned to evaluate HbA1c, blood pressure, physical activity, and dietary intake as secondary outcomes. The registry record does not report numerical results for those outcomes in the registry record for this analysis.

3. Trial Design

The registry describes a randomized, parallel-group trial using cluster randomization at the church level. In a cluster-randomized design, groups of people are assigned together rather than assigning each participant independently. Participants from the same church may share experiences or influences, so their outcomes may be more alike than outcomes from participants in different churches. That feature matters when estimating uncertainty: an analysis that treats every participant as fully independent can give an inappropriate assessment of precision.

The two study groups received behavioral lifestyle interventions. WORD DPP was faith-based; PILI DPP was family- and community-based. Both programs lasted 24 weeks and addressed healthy eating, physical activity, and healthy weight. The registry describes the study as unmasked, with a parallel model. It does not report further details about how the churches were allocated or whether allocation was balanced on particular characteristics.

The primary outcome was change in mean body weight from baseline, measured at immediate post-intervention and 6 months post-intervention. The study description also refers to data collection at baseline, immediate post-intervention (6 months post-initiation), and 6 months post-intervention (12 months post-initiation). Weight was measured without shoes, in light clothing, using a calibrated digital scale. The registry lists secondary measures that include HbA1c, systolic and diastolic blood pressure, sugar-sweetened beverage consumption, fruit and vegetable consumption, and sufficient physical activity.

The record lists enrollment of 380 and a completed status. It gives a start date of 2018-01-08 and a primary completion date of 2020-05-14. The statistical analysis plan, the number of churches, the planned allocation ratio, and the method used to account for church-level clustering are not reported in the registry record.

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.

Baseline

Immediate Post-Intervention

6 Months Post-Intervention

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: The total number of participants providing data at baseline data collection.Two enrolled participants failed to provide data at baseline.

6. Primary Endpoint

The registry lists the primary outcome as Change in Mean Body Weight (kg), measured over Baseline, immediate post-intervention, 6 months post-intervention.

The primary endpoint is change in mean body weight, measured in kilograms, from baseline to immediate post-intervention and to 6 months post-intervention. It is a continuous outcome: each participant’s weight can take values across a range, and the analysis concerns the group mean rather than a yes-or-no event. The registry says that weight was measured without shoes, in light clothing, with a calibrated digital scale at each time point.

Change from baseline expresses follow-up weight in relation to the starting measurement. Comparing changes between randomized groups can help distinguish differences in the course of weight over follow-up from differences that were already present at baseline. In general, continuous outcomes can retain information across the range of measurements rather than reducing them to categories. A mean change also has a direct measurement unit—in this case, kilograms—which can help readers interpret the scale of an estimated difference.

The registry identifies the outcome and its time points, but the posted results do not give the mean change in either group, the estimated difference between groups, or a confidence interval. Consequently, the direction and size of the estimated weight changes and the precision of those estimates cannot be assessed from the reported results. The p-values alone do not convey the size of a change in kilograms.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a mean change describe?

A mean is an average across the measurements in a group. For a change-from-baseline outcome, each participant’s follow-up measurement is considered in relation to that participant’s baseline measurement, and the group summary describes the average change. The clinical interpretation depends on both the size and direction of the change, as well as on the between-group contrast. The registry names mean body weight change as the primary outcome, but does not report the numerical group changes or contrast.

A standard error describes the statistical uncertainty of an estimated quantity, such as a mean. It is not the same as the spread of individual participants’ measurements. A confidence interval combines an estimate with its uncertainty to give a range of values compatible with the data and the specified method, subject to the model assumptions. The registry does not report confidence intervals for the primary outcome, so the precision of the estimated group contrast cannot be evaluated from the posted analysis entries.

What does linear mixed effects regression do?

Linear mixed effects regression is a model for continuous outcomes that can represent both average patterns and dependence among observations. In general, mixed effects models can include terms that account for groups or repeated measurements, alongside terms describing the average association of interest. The choice of terms and assumptions matters: the model must reflect the way participants were assigned and the structure of the measurements.

Here, assignment occurred at the church level, while weight was measured repeatedly over time. Those features create potential dependence both among participants from the same church and across measurements from the same participant. The registry reports the model family but does not specify the model terms, how church clustering was represented, how repeated measurements were handled, or whether covariates were included. Those details are not reported in the registry record.

Why does cluster randomization affect precision?

In an individually randomized study, the assignment unit is the participant. In a cluster-randomized study, people within an assigned group share the same allocation. If outcomes within a church are correlated, the amount of independent information is lower than it would be if every participant contributed wholly independent information. This is why an analysis that ignores clustering can misstate the standard error and confidence interval.

Over-dispersion is a general statistical concern when observed variability is greater than a model expects. In clustered data, failing to represent dependence can make uncertainty look too small; other forms of model misspecification can also affect estimates and uncertainty. The registry’s warning against a simple individual-level recalculation is therefore relevant. The analysis reported by the sponsor takes precedence over an unadjusted re-analysis from summary statistics, although the sparse model description limits independent appraisal.

How should the p-values be read?

The registry reports p-values of 0.360 and 0.321 for two superiority analyses. In general, a p-value is calculated under a null hypothesis and model assumptions; it summarizes how unusual data at least as incompatible with that null would be under those assumptions. It is not an effect estimate and does not show the amount of weight change. Without the estimated contrast and its uncertainty, the p-values cannot establish which group had a larger mean change or whether the estimate was precise.

The record does not state the null hypothesis in detail, the time point associated with each p-value, or the decision threshold used for the outcome analyses. A significance level of 0.05 appears in the planning illustration assumptions, but that does not establish the analysis plan’s threshold. The two reported p-values should not be treated as a complete presentation of the primary outcome.

What can be recalculated?

A simple recalculation can sometimes use posted summary statistics to derive an unadjusted comparison. Such a calculation would not automatically reproduce a model-based analysis, especially when assignment is clustered and outcomes may be correlated within churches. For this trial, the registry notes that an individual-level re-analysis would be misleading and that only the sponsor’s own analysis is shown. The required numerical group summaries are also not reported in the entries. No recalculated effect estimate, confidence interval, or planning illustration based on an observed effect is therefore presented.

10. Analyses Reported by the Sponsor

The sponsor’s reported analyses use linear mixed effects regression, and the registry lists p-values of 0.360 and 0.321. The registry does not report the corresponding estimates, confidence intervals, standard errors, or group-specific mean changes. The analysis entries also do not identify which follow-up time corresponds to each p-value. These omissions prevent a direct comparison of the sponsor’s numerical effect estimates with any independent recalculation.

The registry specifically warns that this was a cluster-randomized trial and that a simple individual-level re-analysis would be misleading. That warning matters because church-level assignment can create dependence among participants in the same church. A recalculation that treats participants as independent would not necessarily estimate uncertainty in the same way as a model that accounts for clustering. For this reason, the sponsor’s own analysis takes precedence over a simple unadjusted re-analysis.

That precedence does not make the reported analysis fully reproducible. The model specification and its handling of church-level clustering are not reported in the registry record. Nor are the estimated effects and confidence intervals available to evaluate direction, magnitude, or precision. The posted p-values are the only numerical results for the primary analyses in the record; they should be interpreted as incomplete evidence about the outcome rather than as substitutes for the missing estimates.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists multiple outcome measures: the primary body-weight outcome and secondary outcomes involving HbA1c, systolic and diastolic blood pressure, sugar-sweetened beverage consumption, fruit and vegetable consumption, and sufficient physical activity. It also records 10 outcome measures and 20 analyses. When several outcomes, time points, or statistical comparisons are examined, the chance of observing at least one result that appears unusual under a null model can increase. Statistical approaches to multiplicity can address that issue, but the appropriate method depends on the analysis plan and the questions being asked.

Multiplicity adjustment methods, any hierarchy among outcomes, and any prespecified approach to interpreting multiple analyses are not reported in the registry record. The two primary-outcome p-values should therefore be understood in the context of an analysis record that does not describe how multiplicity was handled. The registry’s planning illustration mentions a two-sided significance level of 0.05, but this is a planning assumption and does not establish the inferential threshold used for the reported analyses.

The record does not describe missing-data amounts, reasons for missingness, imputation methods, sensitivity analyses, or how participants with incomplete follow-up were handled. It also does not report the analysis population definition, such as whether the analysis followed an intention-to-treat principle or used another population. These details matter because missing measurements can affect estimated changes, and conclusions can depend on assumptions about why measurements are absent and how they are treated.

There is also no reported account of adherence to the interventions or of exclusions from the analysis. The registry lists enrollment and the sponsor’s model, but that alone does not establish which participants contributed measurements at each time point. Without these details, readers cannot assess the potential influence of missing outcome data or determine whether the analysis population matched the randomized population.

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 0% reporting threshold.

The registry reports zero serious adverse events, zero other adverse events, and zero deaths in both intervention groups. The serious-event and other-event risk counts are 213 for WORD DPP and 167 for PILI DPP. These are the denominators listed in the adverse-event record; they should not be treated as proof that every enrolled participant was assessed at every time point.

A recorded count of zero means that no events of the reported category appear in the registry entry for the stated group. It does not establish that the true risk is zero or that the interventions are free of risk. In general, safety interpretation depends on how events were defined, how they were solicited and monitored, the observation period, and the completeness of reporting. Those details are not reported in the registry record. The serious-event terms list is empty.

The adverse-event figures describe the recorded safety outcomes; they do not establish a comparative safety difference between the interventions. The registry does not provide further event descriptions or an analysis of between-group safety differences. These counts should therefore be read as the reported event record, with the limits of the available detail in mind.

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

The study addresses a practical comparison between two lifestyle programs for overweight and obese Marshallese adults in Arkansas and Oklahoma. One program was faith-based and the other family- and community-based. Both were designed around healthy eating, physical activity, and healthy weight, and both lasted 24 weeks. The primary outcome was change in mean body weight over the specified follow-up periods.

The registry reports two p-values from the sponsor’s linear mixed effects regression analyses, but does not report the associated mean changes, between-group estimates, or confidence intervals. As a result, the posted results do not show the direction or size of the difference in weight change, nor how precisely it was estimated. The p-values alone cannot answer those questions. They also do not show whether any observed difference would matter to participants or researchers.

The record’s warning about cluster randomization is important. Churches were randomized, so the statistical analysis needs to account for the assignment structure. The sponsor’s analysis is the appropriate analysis to refer to instead of a simple individual-level recalculation, but the registry does not give enough model detail to assess exactly how clustering was addressed.

The registry also does not report results for the listed secondary outcomes in the registry record for this analysis, or details about missing data, multiplicity, and analysis populations. These gaps limit how broadly the findings can be interpreted. The reported safety counts are zero for serious and other adverse events in both groups, with zero deaths, but the record does not provide further monitoring details.

Accordingly, the posted record establishes the trial’s design, interventions, primary outcome, and two sponsor p-values, but not the numerical primary-outcome effects needed for a full assessment. It does not support a conclusion about which program produced greater weight change or the precision of such a comparison.

14. Why This Trial Matters Statistically

This trial illustrates why the unit of randomization is not a technical detail. Churches, rather than individual participants, were randomized. Participants within a church may have shared experiences, settings, or social influences, so an analysis must reflect the cluster structure when estimating uncertainty. Treating all participants as independent could make the analysis appear more precise than the design supports. The registry itself cautions against a simple individual-level re-analysis.

The study also compares two active programs that use different social frameworks: one faith-based and one family- and community-based. A comparative-effectiveness question of this kind asks how outcomes differ between two intervention approaches in the enrolled population. It does not estimate either program’s effect relative to no intervention. Keeping that distinction clear helps prevent overinterpretation of a two-group comparison.

Statistically, the record shows the gap between a p-value and a fully interpretable result. Without group changes, an estimated between-group effect, and a confidence interval, readers cannot assess the direction, magnitude, or precision of the primary outcome. The model family is named, but further details on clustering and repeated measurements are absent. This makes transparent reporting of estimates and methods especially important.

The trial’s stated focus on Marshallese adults in Arkansas and Oklahoma also clarifies the population to which its results would be most directly relevant. The registry record does not report enough numerical outcome information to characterize the comparative result for that population. Its value for statistical interpretation lies in showing both the importance of cluster-aware analysis and the need to report effects and uncertainty alongside p-values.

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 main limitation of the posted statistical results is the absence of numerical estimates for the primary endpoint. The registry defines mean body weight change and reports two p-values, but does not give the group-specific changes, between-group differences, or confidence intervals. The direction, size, and precision of the primary comparison therefore cannot be assessed.

The design is cluster-randomized at the church level, but the registry does not report the number of churches, the allocation details, or the model specification. Although linear mixed effects regression is named, the record does not explain how the model accounted for church clustering, repeated measurements, or possible covariates. This limits independent assessment of the analysis.

The record does not identify which follow-up time corresponds to each reported p-value. It also does not describe the analysis population, missing-data handling, intervention adherence, or sensitivity analyses. The listed secondary outcomes and analysis counts raise questions about multiplicity, but no multiplicity strategy is reported. Numerical results for those secondary outcomes are not included in the registry record for this analysis.

Safety reporting is also limited: the event counts are zero, but event definitions, monitoring procedures, and observation details are not reported. These limitations concern the information available in the registry record; they do not establish that the study did not collect additional information. The planning illustration is described as based on an observed effect, but no effect estimate or planning sample-size values are included in the registry record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Diabetes Prevention Program Lifestyle Intervention in the Marshallese Population — NCT03270436.
• Registry record last updated 2020-12-09; status last verified 2020-11.
• Results first posted 2020-12-09.
• 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

NCT03270436 is a completed randomized, parallel-group cluster trial sponsored by the University of Arkansas. It enrolled 380 overweight and obese Marshallese adults living in Arkansas and Oklahoma. Churches were the units of randomization, and the study compared the 24-week faith-based WORD DPP with the 24-week family- and community-based PILI DPP.

The primary outcome was change in mean body weight from baseline to immediate post-intervention and 6 months post-intervention. The registry reports two sponsor analyses using linear mixed effects regression, with p-values of 0.360 and 0.321, but does not report the associated effect estimates or confidence intervals. It also lists secondary outcomes involving HbA1c, blood pressure, diet, and physical activity.

The registry reports zero serious adverse events, zero other adverse events, and zero deaths in each group. Further details on the statistical model, missing data, analysis populations, and multiplicity are not reported in the registry record.