This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
CALORIES evaluated whether adding mobile health components to Brenner mFIT standard care could affect the weight status of youth, as quantified by BMI percentile. The study was randomized, parallel-group, single-masked, and had a planned enrollment of 28 participants.
| Feature | CALORIES |
|---|---|
| Official title | Increased Monitoring of Physical Activity and Calories With Technology |
| Status | Completed |
| Phase | NA |
| Design | Randomized, parallel-group, single-masked |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 28 |
| Number of arms | 2 |
| Lead sponsor | Wake Forest University Health Sciences |
| Conditions | Weight Change, Body; Behavior, Health; Obesity, Childhood; Parent-Child Relations |
| Study dates | 2020-11-04 to 2022-12-01 |
| ClinicalTrials.gov | NCT03961061 |
2. Clinical Question
The clinical question is whether adding mobile health components to Brenner mFIT standard care changes the weight status of youth compared with Brenner mFIT standard care alone. The registry defines BMI percentile as the primary endpoint and specifies assessment at baseline, 3 months, and 6 months.
Population
Youth participating in a study addressing body weight, health behavior, childhood obesity, and parent-child relations.
Intervention
Brenner mFIT standard care plus mobile health components, described in the registry as a behavioral intervention.
Comparator
Brenner mFIT standard care, also described as a behavioral intervention.
Primary question
Does the addition of mobile health components affect BMI percentile over the specified assessment time frames?
3. Trial Design
Brenner mFIT — Standard Care
- Behavioral intervention
- Standard care
- Serious adverse events: 0/7 affected / at risk
Brenner mFIT — Standard Care Plus Mobile Health
- Behavioral intervention
- Standard care plus mobile health components
- Serious adverse events: 0/7 affected / at risk
The serious-adverse-event denominator reported for each arm is 7. That figure is the denominator for the posted serious-adverse-event measure; it should not be substituted for the overall enrollment of 28 or used to infer the number randomized to each arm.
4. Endpoints
BMI percentile is registered as the primary endpoint at three specified time frames: baseline, 3 months, and 6 months. The definition is repeated across those time points and is based on measured height and weight.
| Endpoint | Time frame | Registry definition |
|---|---|---|
| BMI Percentile | Baseline | The weight status of youth will be quantified through BMI derived from height and weight measured at intake and follow-up visits. Height (plus/minus 0.1 cm) and weight (plus/minus 0.5 kg) will be recorded twice and averaged. BMI will be calculated as kg/m2. BMI z-score will be calculated using CDC growth charts and converted to BMI-for-age percentile based on CDC growth charts for children and teens ages 2 through 19. |
| BMI Percentile | 3 months | The weight status of youth will be quantified through BMI derived from height and weight measured at intake and follow-up visits. Height (plus/minus 0.1 cm) and weight (plus/minus 0.5 kg) will be recorded twice and averaged. BMI will be calculated as kg/m2. BMI z-score will be calculated using CDC growth charts and converted to BMI-for-age percentile based on CDC growth charts for children and teens ages 2 through 19. |
| BMI Percentile | 6 months | The weight status of youth will be quantified through BMI derived from height and weight measured at intake and follow-up visits. Height (plus/minus 0.1 cm) and weight (plus/minus 0.5 kg) will be recorded twice and averaged. BMI will be calculated as kg/m2. BMI z-score will be calculated using CDC growth charts and converted to BMI-for-age percentile based on CDC growth charts for children and teens ages 2 through 19. |
How BMI percentile is defined
The registry describes BMI percentile categories using CDC growth charts for children and teens ages 2 through 19. A BMI percentile less than the 5th percentile is classified as underweight; between the 5th percentile and less than the 85th percentile as a healthy weight; over the 85th percentile to less than the 95th percentile as overweight; and above the 95th percentile as obesity.
5. Results
The registry marks results as posted and marks each of the three registered BMI Percentile endpoint records as having results posted. However, the registry record summarized here does not contain posted statistical analyses with numerical treatment-effect estimates, confidence intervals, or p-values for those endpoints.
BMI Percentile — Baseline
The registry identifies BMI Percentile at baseline as a primary endpoint. Baseline BMI percentile establishes the starting weight-status measure before subsequent follow-up assessments.
BMI Percentile — 3 months
The registry identifies BMI Percentile at 3 months as a primary endpoint. The record does not provide a numerical between-group estimate, confidence interval, or p-value for this assessment.
BMI Percentile — 6 months
The registry identifies BMI Percentile at 6 months as a primary endpoint. The record does not provide a numerical between-group estimate, confidence interval, or p-value for this assessment.
What analysis would normally be used?
Because BMI percentile is a quantitative outcome measured repeatedly over time, a randomized comparison would ordinarily focus on the difference between intervention groups while accounting for the longitudinal structure of the measurements. Depending on the prespecified statistical analysis plan, approaches could include an analysis of covariance at a specified follow-up time with baseline BMI percentile as a covariate, or a repeated-measures or mixed-effects model using the available measurements over time.
The choice between these approaches matters because baseline adjustment can improve precision, while a longitudinal model can use information from multiple follow-up assessments and account for within-participant correlation. With a small randomized study, model assumptions and the amount of missing follow-up information become particularly important to interpretation.
There is no treatment-effect estimate to interpret numerically in the registry record. Consequently, an effect size, confidence interval, or p-value should not be attributed to the CALORIES comparison on the basis of the posted endpoint entries alone.
For a continuous outcome such as BMI percentile, a treatment effect would generally be expressed as a difference between randomized groups, potentially adjusted for baseline BMI percentile and, if a longitudinal model were used, for the repeated nature of the observations. A confidence interval would describe the statistical precision of that estimated difference; it would not describe the range of individual participants' responses.
A p-value, if available, would address evidence against a prespecified null hypothesis under the chosen model. It would not measure the size or clinical importance of an effect.
6. Statistical Methodology
Randomization as the basis for comparison
Randomized allocation creates the framework for comparing the two behavioral interventions. In a randomized trial, the treatment assignment rather than the observed outcome determines the primary comparison. This is important because post-randomization characteristics can be affected by the intervention or by follow-up processes and therefore should not automatically replace the randomized treatment comparison.
Continuous outcome analysis
BMI percentile is a quantitative outcome rather than a binary event or a time-to-event endpoint. A direct statistical comparison would therefore generally estimate a difference in BMI percentile between the randomized groups rather than a hazard ratio or odds ratio.
The sign of the estimated difference depends on the direction in which the groups are defined. The magnitude describes the separation between the group summaries on the BMI-percentile scale.
Baseline adjustment
Because BMI percentile is measured at baseline and again during follow-up, baseline adjustment can be statistically useful. An ANCOVA-style analysis, for example, can compare follow-up BMI percentile between treatment groups while accounting for baseline BMI percentile. This can reduce residual variability when baseline status is strongly related to the follow-up measurement.
Repeated measurements
The presence of BMI percentile assessments at baseline, 3 months, and 6 months creates a longitudinal data structure. Measurements from the same participant are correlated, so treating every observation as if it came from a different independent participant would generally be inappropriate.
Small-sample considerations
The planned enrollment of 28 makes estimation precision especially important. In a study of this size, even moderate amounts of missing follow-up information can materially reduce the amount of information available for a longitudinal comparison. Confidence intervals, rather than a p-value alone, are therefore particularly useful for communicating what the study can estimate.
7. Statistical Methods Explained
Why is BMI percentile a quantitative endpoint?
BMI percentile places a youth's BMI in the context of age- and sex-specific growth references. The registry specifies calculation of BMI, conversion to BMI z-score using CDC growth charts, and then conversion to BMI-for-age percentile for children and teens ages 2 through 19. This produces a continuous measure that can be compared between randomized groups.
Why measure height and weight twice?
The registry specifies that both height and weight are recorded twice and averaged. Repeated measurement can reduce the influence of an isolated measurement error. The specified tolerances are plus/minus 0.1 cm for height and plus/minus 0.5 kg for weight.
Why adjust for baseline BMI percentile?
If baseline BMI percentile is strongly associated with follow-up BMI percentile, adjustment can improve statistical precision by explaining some of the variation in the follow-up outcome. The resulting treatment comparison is still anchored to randomized assignment when the model is specified appropriately.
Why might a mixed-effects model be useful?
A mixed-effects model can represent repeated BMI-percentile observations from the same participant while accounting for within-participant correlation. It can also provide a framework for estimating treatment differences over multiple follow-up times rather than treating each time point as an unrelated analysis.
Why are missing 3-month and 6-month observations important?
A follow-up outcome cannot contribute its full information to an analysis when it was not collected. The registry specifically states that participants did not provide three-month anthropometric, physical activity, or dietary data because of the COVID-19 pandemic and that participation in six-month follow-up data collection was extremely low. Missingness therefore directly affects the amount and potentially the representativeness of follow-up information.
Why should the analysis not rely on a p-value alone?
A p-value does not quantify the magnitude of a treatment difference. For a continuous outcome such as BMI percentile, an estimated between-group difference together with its confidence interval provides information about both direction and precision. A small study can produce a wide interval even when the estimated difference itself is not close to zero.
8. Missing Follow-Up Data and COVID-19
The registry identifies a major limitation affecting the longitudinal endpoint assessment. Because of the COVID-19 pandemic, participants did not provide three-month anthropometric, physical activity, or dietary data, and participation in six-month follow-up data collection was extremely low.
Three-month follow-up
The registry states that three-month anthropometric, physical activity, and dietary data were not provided because of the COVID-19 pandemic.
Six-month follow-up
The registry states that participation in six-month follow-up data collection was extremely low.
This is more than a routine missing-data issue. BMI percentile is the primary endpoint at both follow-up time frames, so missing measurements affect the ability to compare the randomized groups at precisely the times specified by the endpoint definitions.
For longitudinal clinical-trial analysis, the appropriate statistical treatment of missing outcomes depends on why observations are missing, the pattern of missingness, the prespecified analysis model, and the assumptions used to connect observed and unobserved outcomes. A complete-case comparison can discard substantial information, while model-based approaches can retain participants with partial follow-up under stated assumptions. Sensitivity analyses can be important when those assumptions are uncertain.
9. Safety Results
The registry reports serious adverse events by arm. For the listed safety population, no serious adverse events were reported in either arm.
| Safety measure | Brenner mFIT — Standard Care | Brenner mFIT — Standard Care Plus Mobile |
|---|---|---|
| Serious adverse events | 0/7 affected / at risk | 0/7 affected / at risk |
Within the reported denominator of 7 participants at risk in each listed arm, the registry records zero affected participants for serious adverse events. This is a descriptive safety result, not a formal comparative test of safety between the two interventions.
Zero observed serious adverse events does not establish that the probability of a serious adverse event is zero. In a small safety population, an uncommon event can remain statistically plausible even when no event is observed. The appropriate interpretation is that no serious adverse events were recorded among the 7 participants at risk in each listed arm.
10. Timeline
Study start
The CALORIES study began on 2020-11-04.
Primary completion
The registry lists 2022-12-01 as the primary completion date.
11. Limitations
- Small enrollment: the study enrollment is 28 participants, limiting the precision with which between-group differences can generally be estimated.
- Missing three-month data: the registry states that participants did not provide three-month anthropometric, physical activity, or dietary data because of the COVID-19 pandemic.
- Very low six-month participation: participation in six-month follow-up data collection was extremely low, directly affecting the availability of the registered primary endpoint at that time frame.
- Limited formal statistical reporting: the registry contains no posted statistical analyses with numerical treatment-effect estimates, confidence intervals, or p-values for the primary BMI-percentile endpoints.
- Safety denominator: the serious-adverse-event result is reported with 7 participants at risk in each listed arm and should not be treated as the overall enrollment denominator.
- Single masking: the study is classified as single-masked, but the summarized registry information does not identify which study participants or personnel were masked.
- Endpoint interpretation: BMI percentile is a derived growth-reference measure rather than a direct measure of body weight alone, so interpretation should remain tied to the specified BMI and CDC growth-chart methodology.
12. Why This Trial Matters Statistically
CALORIES is a useful teaching example because the central statistical problem is not a time-to-event endpoint or a large phase 3 comparison. Instead, it illustrates how randomized allocation, a quantitative pediatric growth measure, repeated follow-up, and substantial missing data interact in a small behavioral-intervention study.
| Concept | How it appears in CALORIES |
|---|---|
| Randomization | Participants were assigned through randomized allocation. |
| Parallel-group design | The study has 2 parallel arms. |
| Quantitative endpoint | BMI Percentile is the registered primary endpoint. |
| Baseline assessment | BMI Percentile is measured at baseline. |
| Repeated follow-up | BMI Percentile is also specified at 3 months and 6 months. |
| Baseline adjustment | An ANCOVA-style analysis is one conventional approach for a quantitative follow-up outcome with a baseline measurement. |
| Longitudinal modeling | Repeated BMI-percentile observations can motivate a repeated-measures or mixed-effects framework. |
| Missing data | COVID-19 affected three-month data collection, while six-month participation was extremely low. |
| Small-sample inference | Enrollment of 28 makes precision and uncertainty especially important. |
| Safety description | Serious adverse events are reported descriptively as affected / at risk by arm. |
13. Planned Analysis
The registry identifies BMI Percentile as the primary endpoint at baseline, 3 months, and 6 months. For this type of randomized quantitative outcome, a typical analysis would compare the two intervention groups on BMI percentile while accounting for baseline BMI percentile and, if all available longitudinal measurements are analyzed together, the correlation among repeated observations within participants.
A baseline-adjusted model could estimate the treatment-group difference at a specified follow-up time. A repeated-measures or mixed-effects model could instead estimate differences across the longitudinal follow-up period while accommodating within-participant correlation. The preferred model would depend on the prespecified analysis plan and its assumptions about the outcome and missing observations.
The substantial COVID-19-related loss of follow-up is particularly important for model selection and interpretation. Any longitudinal analysis would need to state how incomplete observations are handled and what assumptions are made about the relationship between observed and missing BMI-percentile measurements.
Here, β1 represents the adjusted treatment-group difference under the specified model. A longitudinal model would extend this framework to account for multiple observations per participant.
14. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The study uses randomized parallel-group allocation and a quantitative primary endpoint measured at baseline, 3 months, and 6 months. The principal statistical challenge is estimating a between-group difference while accounting for baseline measurement, repeated observations, small enrollment, and substantial missing follow-up.
Clinical interpretation
BMI percentile is intended to quantify youth weight status using BMI, CDC growth charts, and age-specific percentile conversion. The meaning of a change therefore depends on the growth-reference framework as well as the magnitude and timing of any observed between-group difference.
15. How to Read a Future Treatment Effect
If a formal analysis were reported, the most informative presentation would include the estimated difference in BMI percentile between the two randomized groups, its 95% confidence interval, and the corresponding p-value if hypothesis testing were prespecified.
The estimated difference would quantify how far apart the groups were on the BMI-percentile scale under the chosen statistical model. Its direction would depend on which treatment group was subtracted from which.
A 95% confidence interval would communicate statistical uncertainty around the estimated treatment difference. A narrow interval would indicate greater precision than a wide interval, but the interval would not describe the range of individual responses.
A p-value would address the compatibility of the observed result with a specified null hypothesis under the statistical model. It would not tell readers how large or clinically important the treatment effect is.
For this study in particular, the missing three-month and extremely low six-month follow-up make the confidence interval especially important. A statistical estimate should always be interpreted together with the amount and pattern of information contributing to it.
16. Sources
- ClinicalTrials.gov: CALORIES — NCT03961061.
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical-trial analyses and statistical resources across the Clinical Biostats site.
17. Record Summary
CALORIES is a randomized, parallel-group, single-masked study with 28 participants and two behavioral intervention arms: Brenner mFIT standard care and Brenner mFIT standard care plus mobile health components. Its registered primary endpoint is BMI Percentile, assessed at baseline, 3 months, and 6 months using repeated height and weight measurements, BMI calculation, BMI z-score calculation from CDC growth charts, and conversion to BMI-for-age percentile for children and teens ages 2 through 19.
Statistically, the central issue is the analysis of a quantitative longitudinal endpoint in a small randomized study. A suitable analysis would ordinarily account for baseline BMI percentile and the correlation created by repeated measurements. The registry's COVID-19 limitation is crucial: three-month anthropometric data were not provided, and participation in six-month follow-up data collection was extremely low. Those features substantially affect the information available for longitudinal inference.
The registry reports zero serious adverse events among 7 participants at risk in each listed arm. It does not provide formal statistical analyses with numerical treatment-effect estimates, confidence intervals, or p-values for the primary BMI-percentile endpoints. The most important statistical lesson is therefore methodological: randomized design alone does not guarantee precise longitudinal inference when follow-up information is severely limited.