Delish Study: Diabetes Education to Lower Insulin, Sugars, and Hunger: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Delish Study: Diabetes Education to Lower Insulin, Sugars, and Hunger (NCT03207711), a randomized na study. The primary outcome is Frequency of Eating in Response to Cravings (Primary Mechanistic Outcome).

2. Clinical Question

The study addresses whether adding mindfulness training to carbohydrate-restricted diet education is associated with a different change in eating in response to food cravings over 6 months. The registry describes this as a primary mechanistic outcome: a measure of eating behavior that may relate to dietary adherence, rather than a direct measure of glycemic control or another clinical outcome.

This framing separates two questions that are sometimes conflated. One is whether the recorded outcome changed within each group over time. The other is whether the change differed between groups. The registry provides estimates for the first question, with confidence intervals, but does not provide an estimable result for the second. The posted arm estimates therefore describe changes in the two assigned groups; they do not, by themselves, quantify the additional effect associated with mindfulness.

3. Trial Design

The Delish Study is a randomized, single-masked, parallel-group study of behavioral interventions for type 2 diabetes. The registry lists the phase as NA and the status as completed. The study sponsor is the University of California, San Francisco. The record gives a start date of 2017-02-17 and a primary completion date of 2018-04-30.

The registry lists enrollment of 60. For the primary outcome, the posted estimates include 28 participants in the Diet Education arm and 30 in the Diet Education + Mindfulness arm. These are the arm sizes attached to the reported endpoint estimates; the reason they do not add to the listed enrollment is not reported in the registry record. A participant flow account, including reasons for any differences between enrollment and endpoint numbers, is not reported in the registry record.

Both arms received education about a carbohydrate-restricted diet. The record describes the study diet as approximately 10% of calories from carbohydrate, typically 50 grams/day or fewer, not including fiber. The Diet Education + Mindfulness arm also received two integrated mindfulness components: use of a mindful eating app at home to learn and practice skills for food cravings and eating, and in-person group-based training. The record does not report enough detail to characterize the frequency or duration of those activities.

The randomized, parallel design assigns participants to one of two groups and follows the groups over time. Random allocation can support a fair comparison when the analysis preserves the randomized groups and handles missing observations appropriately, but the analysis population and missing-data approach are not reported in the registry record. Single masking is listed, but who was masked is not reported in the registry record. The posted primary outcome is change from baseline to 6 months. Other listed outcomes include measures related to eating, glycemic control, insulin resistance, weight, and dietary adherence.

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 Frequency of Eating in Response to Cravings (Primary Mechanistic Outcome), measured over change from baseline to 6 months.

The primary endpoint is titled Frequency of Eating in Response to Cravings. It measures the percentage of ecological momentary assessment (EMA) opportunities in which participants reported eating in response to food cravings over a 3 day period. Participants received EMA measures on cell phones 3 times each day. The final measure each day included a second question asking whether any craving-related eating had occurred but had not already been reported earlier that day. This created 4 potential reporting opportunities per day, or 12 potential measures over 3 days. The reported percentage uses the number of EMA responses received as its denominator.

This endpoint focuses on a specific behavior: eating in response to cravings, rather than craving intensity alone or overall food intake. Repeated prompts during daily life can capture reports close to when behaviors occur. As a general measurement principle, that may offer a different perspective from asking participants to recall behavior over a much longer period. It remains a self-reported measure, and the percentage is based on responses received, not simply on all potential prompts.

The endpoint is described as a primary mechanistic outcome. In general, a mechanistic outcome assesses a process that might help explain how an intervention relates to a later or broader outcome. The registry's rationale is that mindfulness may help participants experience cravings and difficult emotions without responding impulsively, potentially supporting dietary adherence. The registry does not establish that this mechanism changed dietary adherence or clinical outcomes. The posted timeframe is change from baseline to 6 months.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: it is expressed as a percentage, and the registry identifies the reported parameter as a mean. A mean summarizes the average value or average change across the participants contributing to an estimate. It does not imply that every participant had a change close to that average. Individual responses may vary, and the registry record does not provide the distribution of participant-level changes.

What the confidence intervals express

A 95% confidence interval gives a range of values compatible with the estimated mean under the statistical procedure used. In general, if the same study and procedure could be repeated many times, intervals constructed in this way would capture the target quantity at the stated rate over repeated samples. For an individual study, the interval is not a probability statement that the true value lies inside its specific bounds, and it is not a range containing 95% of participants. The posted intervals quantify uncertainty around the arm-specific mean changes; they do not replace a direct estimate of the between-arm difference.

The intervals are reported as the dispersion measure for the primary outcome. No standard deviation or standard error is posted for these endpoint estimates. The methods used to construct the intervals, including any adjustment for repeated observations or other features of the analysis, are not reported in the registry record. It is therefore safest to describe the intervals as posted rather than infer a particular model from their form.

Why the arm comparison matters

In a randomized study, the main comparative quantity is generally the outcome contrast between assigned groups. For a continuous endpoint, that could be a difference between group means, possibly adjusted for baseline values or other variables. A confidence interval and statistical test for that contrast would address whether the observed group changes are distinguishable from one another under the analysis. The registry record does not provide an estimable two-arm comparison for this endpoint. Separate within-arm changes cannot substitute for that contrast: each describes change in its own group, and neither accounts for the other group's change.

The record lists the comparison as unavailable because values are missing or not estimable for a two-arm comparison of a continuous endpoint. The reason a comparison could not be estimated is not reported in the registry record. No comparative estimate, confidence interval, or p-value is posted in the endpoint details. This absence should not be recast as evidence that the groups were equivalent, nor as evidence that they differed.

Repeated EMA observations and the denominator

EMA collects reports at multiple opportunities during ordinary daily routines. Here, the protocol description allows 12 potential measures over 3 days: 3 prompts each day, plus a second question included with the final daily measure. The primary outcome is the percentage of EMA responses received that report craving-related eating. This denominator is important. A response-based percentage describes the reports that were actually received; it is not automatically the percentage of all potential prompts or the percentage of all participants who ate in response to a craving.

As a general statistical issue, multiple observations from the same participant are usually related rather than independent. Analyses of repeated observations may need to account for this within-participant dependence. The registry record does not report how repeated EMA responses were modeled, whether a participant-level summary was formed first, or how uncertainty was calculated in light of the repeated measurements. The posted estimates should not be used to infer a particular repeated-measures method.

Uncertainty, missing responses, and over-dispersion

In general, over-dispersion occurs when observed variability is greater than a statistical model assumes. If a model underestimates variability, confidence intervals can be too narrow and tests can overstate precision. Repeated EMA measurements, differences in how often participants respond, and participant-to-participant variation can all be relevant considerations when analyzing this kind of endpoint. These are general methodological concerns, not findings about this trial. The registry record does not report a model, a diagnostic assessment of over-dispersion, or a method for handling unequal numbers of responses.

Because the denominator is responses received, missing prompts can affect both the amount of information available and what the percentage represents. For example, if response availability is related to eating behavior, a response-based summary might not represent unreported opportunities in the same way. This is a general limitation of response-dependent measurement; whether it affected these results is not reported in the registry record. The proportion of prompts answered, reasons for missing responses, and any sensitivity analyses are not reported in the registry record.

Planning illustrations are not observed power

Any sample-size illustrations based on the observed effect are planning scenarios, not evidence of the power achieved by the completed study. The stated assumptions for such illustrations are a two-sided significance level of 0.05, equal allocation, a normal approximation, and an effect defined as a stated fraction of the observed effect. The illustration is not the trial's original design justification. Its assumptions do not show that the study was planned using those values, and an illustration cannot establish whether the trial had adequate power for its actual analysis. The trial's own sample-size justification is not reported in the registry record.

More generally, power calculations describe the chance of detecting a prespecified effect under specified assumptions. They do not measure the probability that a result is true, and an after-the-fact calculation based on an observed effect is not observed power. The posted arm-specific changes alone cannot provide a valid comparative power assessment when the two-arm contrast is not estimable. The registry record does not report a planned interim analysis, statistical analysis plan, or details of any adjustment to the primary analysis; those details are not reported in the registry record.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists 11 outcome measures overall and 10 secondary outcomes posted. The secondary outcomes include measures of impulsivity, emotion-related and stress-related eating, HbA1c, fasting glucose, insulin resistance, weight change, and diet adherence assessed using fingerstick blood ketones. The registry record lists zero separate analyses. This does not establish that no analyses were performed; it indicates that no analysis records are listed in the information reported for this study.

When a study assesses multiple outcomes or multiple comparisons, the chance of at least one apparently unusual finding can increase if each result is interpreted separately. Statistical approaches can address multiplicity, but the appropriate approach depends on which outcomes and comparisons were planned and how they were prioritized. The registry does not report a multiplicity adjustment, a hierarchy among secondary outcomes, or a complete analysis plan. Whether any adjustment was used is not reported in the registry record. The primary mechanistic outcome should therefore be distinguished from the other registered outcomes, and isolated secondary findings should not be treated as confirmatory without information about the analysis plan.

The primary endpoint denominator is the number of EMA responses received. The record does not report how missing EMA responses were handled beyond defining that denominator, nor does it report response completeness, imputation, or sensitivity analyses. These omissions matter because response availability can influence the precision and interpretation of a response-based percentage. The analysis population is also not reported in the registry record: it is not stated whether the posted estimates include all randomized participants, only participants with endpoint data, or another defined group. A participant flow account and reasons for missing endpoint information are not reported in the registry record.

Randomization is listed, but the registry does not report whether analysis followed assigned groups regardless of intervention adherence, or whether any per-protocol or other population was used. Those details are not reported in the registry record. Without an analysis population definition, missing-data methods, and a prespecified multiplicity approach, the scope of inference and the confirmatory status of any additional outcome findings cannot be fully assessed.

11. Safety Results

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

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent non-serious adverse event terms

The posted serious adverse event counts are 0 of 28 participants in Diet Education and 0 of 30 in Diet Education + Mindfulness. The registry also reports 0 deaths in each arm. These counts describe the events recorded in the registry for the stated groups; they do not establish that no unrecorded or unobserved events occurred.

The record separately lists other adverse events affecting 12 of 28 participants in Diet Education and 11 of 30 in Diet Education + Mindfulness. These are counts of participants affected, not counts of all event episodes. The specific non-serious event terms are not described in the information, so the types, severity distribution, and timing of these events cannot be characterized from these figures.

Zero serious adverse events in the posted data is a descriptive finding, not proof that the risk of serious events is zero. Safety interpretation depends on what was monitored, how events were defined, how long participants were followed, and how completely events were collected. The safety monitoring details and follow-up period for adverse events are not reported in the registry record. These arm counts should not be used to make a general claim about safety beyond the events recorded for this study.

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

The primary outcome results show negative mean changes in both arms over the baseline-to-6-month timeframe: -20.39 for Diet Education and -25.27 for Diet Education + Mindfulness. Their respective 95% confidence intervals are -28.15 to -12.62 and -32.95 to -17.58. These figures describe the arm-specific changes in the percentage of EMA responses reporting eating in response to cravings. They do not directly estimate the effect of adding mindfulness to diet education.

The registry record does not provide an estimable comparison between arms for the primary continuous endpoint. A difference between two arm-specific point estimates is not, by itself, a completed comparative analysis: uncertainty in the contrast and the analysis method are also needed. Since no between-arm result is posted, the available primary results cannot establish whether the groups differed, nor the size or precision of any such difference.

The endpoint is mechanistic and behavioral. It should not be treated as a direct substitute for glycemic control, weight, insulin resistance, or adherence. Those measures appear among the secondary outcomes, but the registry information does not include their outcome estimates. The primary results also do not show that the proposed mindfulness mechanism caused any change. The study's rationale describes a hypothesized pathway involving cravings and emotional eating; the posted arm-level changes alone cannot verify that pathway.

In general, a change in a self-reported measure can be informative while still depending on when and how responses were collected. Here, the measurement period was 3 days, with repeated phone prompts and a denominator based on responses received. The findings therefore pertain to the recorded EMA measure under those conditions, not necessarily to every eating occasion or to behavior outside the measurement period.

The results support a cautious descriptive statement: both arms have reported negative mean changes, with the stated confidence intervals. They do not support a claim that mindfulness was superior, that the interventions were equivalent, or that either intervention caused a clinical benefit. The missing comparative estimate, unreported missing-data handling, and limited endpoint measurement details constrain conclusions about the treatment contrast.

13. Why This Trial Matters Statistically

This trial raises several statistical questions common to behavioral studies. The primary outcome is repeated, self-reported, and measured through a sequence of daily prompts. The endpoint's response-based denominator makes the completeness of those prompts relevant to both interpretation and precision. Clear reporting of the number of completed responses, the analysis population, and the handling of missing observations would help readers understand what the estimated percentage represents.

The distinction between within-group change and between-group difference is especially important. The registry posts a mean change and confidence interval for each arm, but the comparative estimate is not estimable. Readers can describe how the recorded outcome changed within each group, but cannot infer from those two separate estimates whether adding mindfulness changed the outcome relative to diet education alone. This illustrates why randomized trials are chiefly interpreted through their planned contrasts rather than by asking whether each group changed on its own.

The endpoint is identified as mechanistic, while the study also registers clinical and adherence-related outcomes. Keeping these outcome types distinct helps avoid treating a behavioral measure as a clinical endpoint or a hypothesized mechanism as an established explanation. A complete analysis plan would also clarify how the primary question is prioritized relative to multiple secondary outcomes and how uncertainty is managed across those analyses. Such planning information is not reported in the registry record.

Finally, planning illustrations based on an observed effect should be separated from the original design justification and from the actual comparative evidence. They can show how assumptions affect hypothetical sample-size planning, but they do not establish the power achieved in this completed study. Together, these issues make transparent reporting of estimands, missingness, repeated-measure methods, and analysis populations central to evaluating behavioral trial results.

14. Limitations

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

The principal limitation of the posted primary results is that the registry does not provide an estimable two-arm comparison. The arm-level changes and confidence intervals are descriptive, but they do not answer whether the change differed between diet education alone and diet education plus mindfulness. The reason the contrast could not be estimated is not reported in the registry record.

The endpoint uses EMA responses received as the denominator. The record describes 12 possible reporting opportunities over 3 days, but does not report how many responses were completed, whether response rates differed across arms, or how missing responses were handled. A response-based percentage can be difficult to interpret if the availability of responses is related to the behavior being measured. Whether this occurred in this study is not reported in the registry record.

The outcome relies on self-report and captures a short measurement window, even though the endpoint is evaluated as change through 6 months. It measures reported eating in response to cravings, not all eating, craving severity, or a clinical outcome. The registry record does not report validation details for this measure, or the statistical method used to account for repeated reports from the same participant.

Further details that would support interpretation are not reported in the registry record: the participant flow, the analysis population, the missing-data approach, the identity of the masked parties, and the statistical analysis plan. The listed arm sizes for the endpoint differ from the registry enrollment of 60, but the reason is not reported in the registry record. The registry lists secondary outcomes, yet their results are not included in the primary outcome information; the outcome-specific estimates and their uncertainty therefore cannot be evaluated from these results.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Delish Study: Diabetes Education to Lower Insulin, Sugars, and Hunger — NCT03207711.
• Registry record last updated 2025-12-08; status last verified 2025-12.
• Results first posted 2025-12-08.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

18. Record Summary

NCT03207711 is a completed, randomized, single-masked, parallel-group study sponsored by the University of California, San Francisco. It enrolled 60 participants and compares diet education with diet education plus mindfulness for type 2 diabetes. The primary outcome is change from baseline to 6 months in the percentage of EMA responses reporting eating in response to cravings over a 3 day period.

The posted mean changes are -20.39 for Diet Education, with a 95% confidence interval from -28.15 to -12.62, and -25.27 for Diet Education + Mindfulness, with a 95% confidence interval from -32.95 to -17.58. The registry record does not provide an estimable two-arm comparison for this continuous endpoint, so the arm-specific changes do not establish a comparative effect.

The registry lists 10 secondary outcomes posted and 11 outcome measures overall. Serious adverse events and deaths are reported as 0 in each arm; other adverse events affected 12 participants in Diet Education and 11 in Diet Education + Mindfulness. Details of the analysis population, missing-data approach, and statistical plan are not reported in the registry record.