Dietary Assessment Study Via Digital Images: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Dietary Assessment Study Via Digital Images (NCT03267004), a randomized na study. The primary outcome is Weighed Food Intake.

2. Clinical Question

The study’s stated purpose was to validate a passive image-assisted dietary assessment method. Participants were assigned to one of two sequences of meals. Each meal included a regular-shaped single food, an irregular-shaped single food, a regular-shaped mixed food, and an irregular-shaped mixed food. Examples named in the registry include a cookie, ice cream, a sandwich, and a pasta dish.

The primary endpoint description focuses on food weight before and after each meal session. Comparing the weights was intended to assess the weight change attributable to food intake. The posted summaries report mean intake for each assigned order, not separate results for each food category. The registry record does not report image-derived portion estimates or a direct statistical comparison of image estimates with measured weights. That distinction matters: group means for weighed intake describe consumption in the study sessions, but do not by themselves quantify the accuracy of the image-analysis method.

3. Trial Design

The registry identifies this as a randomized, single-masked, factorial-model study, with enrollment of 30. It was marked completed. The listed start date is 2017-08-03, and the primary completion date is 2018-08-03. The study’s phase is listed as not applicable. The sponsor was The University of Tennessee, Knoxville.

Randomization assigned participants to one of two meal orders. In Meal Order 1, Meal A was served during meal session 1 and Meal B during meal session 2. In Meal Order 2, Meal B was served during session 1 and Meal A during session 2. Each arm contained 15 participants in the posted results. Thus, the contrast concerns assigned order, with Meal A and Meal B appearing in opposite sessions across the groups.

The registry labels the model factorial, but the factors and their levels are not reported in the registry record. The identity of the masked party is also not reported in the registry record. The record does not describe sequence-generation details, allocation concealment, or whether participants completed both sessions as planned. These details cannot be inferred from the labels alone.

The primary outcome was measured at meal sessions 1 and 2 within 4 weeks of the initial screening session. The record describes food items being weighed before and after each session with an electronic food scale, to the nearest tenth of a gram. It lists two outcome measures, but no secondary outcomes are posted. There are no analysis entries listed, although group summaries and a comparison for the primary outcome are available in the posted results. The registry record does not explain the relationship between the absent analysis entries and those posted summaries.

The design description presents an assessment study using passive images and automated image analysis, but the reported primary endpoint is weighed intake. The available results therefore permit a comparison of intake by assigned order; they do not establish a complete account of the image-validation analyses that may have been planned.

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

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 Weighed Food Intake, measured over Meal session 1 and 2 within 4 weeks of initial screening session.

The primary endpoint is weighed food intake, expressed in grams. The registry describes weighing each food item before and after a meal session and assessing the weight change. This is an objective measurement of food consumed during the study meals, as described in the outcome definition. The endpoint was recorded for meal sessions 1 and 2 within 4 weeks of the initial screening session.

Weighed intake serves as a measured reference quantity for the meal. In a dietary assessment validation context, such a measurement can be compared with an image-based estimate to evaluate whether the method identifies foods and estimates portions accurately. However, the posted primary results are group means for weighed intake by assigned meal order. No image-derived estimates, image-to-weight errors, or agreement statistics are reported in the registry record. Accordingly, the displayed endpoint does not itself show the accuracy of DietCam’s food identification or portion-size estimation.

The two groups had a posted mean of 354.4 grams, with a standard deviation of 126.2 grams in each group. Those summaries describe the distribution of measured intake within each order group. They do not indicate that every participant consumed the same amount, nor do they describe intake separately by meal, food type, or session. Such breakdowns are not reported in the registry record.

The endpoint is useful for describing measured intake and for assessing whether the randomized meal-order groups had different average intake. Its relevance to validating image-assisted dietary assessment depends on the corresponding image measurements and their comparison with weighed amounts. Those validation results are not reported in the registry record.

7. Results Posted in the Registry

• Meal Order 2: mean 354.40 versus 354.40 on control; difference 0.00 (95% CI -90.32 to 90.32), standardized effect d = 0.00; z-test p = 1.000.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: it records food intake as a quantity in grams. For a continuous measure, a group mean summarizes the average observed value, while a standard deviation describes how spread out individual observations are around that mean. In the posted results, both meal-order groups have a mean of 354.4 grams and a standard deviation of 126.2 grams. A shared mean does not imply that participants had identical intake. The standard deviations indicate variation among observations within each group.

What does the mean difference represent?

A mean difference compares the average outcome in one group with the average in another. Here, the displayed comparison is between Meal Order 2 and Meal Order 1, and the estimated mean difference is 0.0 grams. Since the two posted means are both 354.4 grams, the point estimate indicates no observed difference between the group averages. The comparison concerns the assigned meal orders, not an estimate of the difference between image-based measurements and weighed amounts.

The difference is expressed in the original unit, grams, so it remains interpretable in the outcome’s measurement scale. A mean difference is distinct from a standardized effect measure: the former reports the contrast in grams, while a standardized measure expresses a contrast relative to variability. The posted standardized effect is 0.0. Neither measure provides a direct assessment of image-analysis accuracy without corresponding image estimates and a comparison against measured weights.

What does the confidence interval add?

A confidence interval gives a range of values compatible with the estimate under the statistical method and its assumptions. It conveys uncertainty that is not visible in a point estimate alone. The interval for the mean difference runs from -90.32 to 90.32 grams. Its inclusion of zero is consistent with the estimated difference of 0.0 grams, but the interval also includes positive and negative differences. The interval is not a range containing a specified proportion of individual participants’ intake values; individual variation is described by the standard deviation instead.

In general, the width of a confidence interval depends on the amount of information in the data and the variability of the outcome, as well as the method used to calculate it. Greater precision corresponds to a narrower interval. The registry comparison reports standard errors of 32.58 for each group and a pooled standard deviation of 126.2. The registry record does not describe a more detailed analysis model, adjustments, or assumptions beyond the information in the posted comparison.

How should the p-value be read?

The unadjusted p-value is 1.0. In general, a p-value describes how unusual data at least as incompatible with a specified null hypothesis would be if that hypothesis and the test assumptions were true. It is not the probability that the null hypothesis is true, and it does not measure the probability that the findings occurred by chance. A large p-value is not proof of no difference or equivalence. Interpretation should consider the estimated difference and its confidence interval, as well as the study design and reporting limitations.

The comparison is unadjusted in the posted result. Whether this was the prespecified primary analysis, whether other tests were conducted, and whether any multiplicity procedure was planned are not reported in the registry record. The registry lists no analysis entries and no posted secondary outcomes, so a broader account of analytic choices is unavailable.

Why can variability and over-dispersion matter?

In general, a statistical model should reflect the way its outcome varies. Over-dispersion means that observed variability is greater than a model’s assumptions allow. If it is ignored, uncertainty may be understated and inferential results may be too optimistic. This concept is relevant to count or rate outcomes and other settings where a model specifies a particular variance structure. The primary outcome in this study is a continuous weight measurement, and the registry record does not report an over-dispersion analysis. It would not be appropriate to assume that such an issue was assessed or present.

For a continuous outcome, the reported standard deviations and standard errors describe different forms of variability. The standard deviation of 126.2 grams summarizes the spread of participant measurements within each group. The standard error of 32.58 summarizes uncertainty in each group mean under the calculation used. A standard error is not the same as the spread among participants; the two quantities answer different questions.

What can planning illustrations tell readers?

Sample-size calculations are planning tools, not observations from the completed trial. The planning description states that its illustrations use the observed effect, a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation; the effect to detect is a stated fraction of the observed effect. Such illustrations describe hypothetical planning under those assumptions. They are not observed power and do not establish what the trial was designed to detect.

The registry record does not report its own sample-size justification or planning assumptions. In particular, it does not state whether the enrollment target was selected to detect a meaningful difference in weighed intake or to assess a particular level of image-estimation accuracy. The observed mean difference is 0.0 grams, so illustrations based on fractions of that observed effect should not be treated as evidence that the study had adequate power for another, clinically or scientifically meaningful effect. No additional calculation or design rationale can be inferred from the posted summaries.

Overall, the available statistical quantities answer a narrow question about mean weighed intake by assigned meal order. To interpret the study as a validation of the image-based method, readers would also need image-derived measurements and prespecified analyses comparing those estimates with the measured weights. Those results and analysis details are not reported in the registry record.

10. Confidence Intervals and Precision

The estimated mean difference in weighed intake is 0.0 grams, with a confidence interval from -90.32 to 90.32 grams. The interval expresses uncertainty around the estimated group contrast. It includes no difference, but it also includes differences in both directions. The posted point estimate is therefore more definite than the evidence about the possible underlying contrast.

In general, a confidence interval is useful because it shows both the location of an estimate and its precision. A narrow interval indicates that the data and analysis constrain the plausible values more tightly; a broad interval leaves greater uncertainty. The reported interval is wide relative to the zero point estimate and spans differences that may matter for interpretation of average intake. The registry record does not define a threshold for a meaningful difference, so the interval cannot be classified against a prespecified acceptability margin.

There were 15 participants in each meal-order group. The standard deviation was 126.2 grams in each group, and the standard error for each group mean was 32.58. These quantities help contextualize the interval: participant intake varied, and the uncertainty in the group means remains relevant despite identical observed averages. The standard deviation describes participant-level spread; it should not be confused with the confidence interval for a difference in group means.

The confidence interval does not show where most individual intake values lie, nor does it provide a probability that the true difference falls inside the stated limits. It also does not establish equivalence. A conclusion about equivalence would require a prespecified margin and an appropriate analysis, neither of which is reported in the registry record.

Finally, precision for the meal-order comparison is not precision for the image-validation question. No interval for image-based portion-estimation error or agreement with weighed intake is reported in the registry record. The available interval should be interpreted only as uncertainty around the displayed comparison of group mean weighed intake.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists no analysis entries and no posted secondary outcomes. The available primary outcome comparison is reported with an unadjusted p-value of 1.0. The registry record does not report a statistical analysis plan, a hierarchy of outcomes, or whether any adjustment for multiple comparisons was planned or performed. Consequently, it is not possible to determine from the registry record whether the displayed comparison was the only analysis intended or whether other analyses were considered.

Multiplicity arises when several outcomes, comparisons, time points, or subgroup analyses are tested. In general, conducting many tests can increase the chance of obtaining at least one apparently noteworthy result, so prespecified priorities and suitable adjustment methods can matter. That general concern does not establish that multiplicity affected this study. The registry record does not describe a set of tests from which the extent of any multiplicity could be assessed.

Missing-data handling is also not reported in the registry record. The posted group summaries include 15 participants in each arm, but the record does not explain whether this represents all randomized participants with outcome measurements, how missing sessions or measurements were handled, or whether any imputation was used. The analysis population is not reported in the registry record. These details matter because exclusions or incomplete outcome measurements can affect the interpretation of a randomized comparison.

The record describes randomization but does not specify whether the displayed analysis followed intention-to-treat principles, used only participants who completed both meal sessions, or applied another analysis population. These approaches can differ in how they preserve the randomized comparison and how they address incomplete data. No particular approach should be attributed to the study without a registry description.

Interim analyses and stopping rules are not reported in the registry record. This absence does not establish that none occurred; it means the record does not provide information to evaluate them. In combination, the lack of analysis-plan, missing-data, and analysis-population details limits assessment of how the reported comparison was selected and conducted.

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’s adverse-event summaries list 15 participants at risk in each meal-order group. In Meal Order 1, the record lists 0 serious adverse events, 0 other adverse events, and 0 deaths. In Meal Order 2, it likewise lists 0 serious adverse events, 0 other adverse events, and 0 deaths. No serious adverse-event terms are listed.

These figures describe the adverse-event entries posted for the study. A count of zero in the registry results means no events of that type are recorded in those entries; it should not be expanded into a claim about all possible symptoms, discomforts, or experiences. The registry record does not report further detail on event definitions, collection procedures, observation periods, or completeness of safety ascertainment.

The absence of posted events does not establish that an intervention or assessment method is free of risk. The study enrolled 30 participants, and the posted safety counts alone do not characterize uncommon events or experiences outside the recorded categories. The available figures can be read as the registry’s reported counts for the two groups, with no further safety conclusions warranted from the record.

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

The posted results show identical mean weighed food intake in the two randomized meal-order groups: 354.4 grams in each. The estimated difference is 0.0 grams, with a confidence interval from -90.32 to 90.32 grams and an unadjusted p-value of 1.0. These statistics describe the comparison of average weighed intake by assigned order in the enrolled participants.

They do not show that the meal orders are equivalent, that the true difference is exactly zero, or that individual participants had the same intake. The confidence interval retains uncertainty around the estimate, and the registry record does not report an equivalence margin. The results also do not establish that meal order has no influence on intake outside the conditions and participants represented in the posted data.

The study was described as a validation investigation using passive images from Sony Smarteyeglass and automated analysis software, DietCam. However, the posted primary outcome is weighed intake. The record does not report food-identification accuracy, portion-size-estimation accuracy, agreement between image estimates and weights, or an image-based error measure. Identical group averages for weighed intake cannot answer those validation questions.

The findings are limited to the reported group summaries and comparison. The registry record does not provide participant characteristics, food-specific or session-specific results, missing-data methods, an analysis population, or a full statistical plan. Those omissions restrict assessment of generalizability and of how completely the reported analyses reflect the study design.

In general, a trial result is best interpreted by matching the conclusion to the outcome actually measured and the contrast actually analyzed. Here, the statistical evidence concerns the difference in mean weighed intake between meal-order groups. It does not provide a direct estimate of the digital assessment method’s validity. No medical or dietary recommendation follows from this comparison alone.

14. Why This Trial Matters Statistically

This study illustrates an important distinction in validation research: measuring the target quantity and evaluating the method used to estimate it are related but separate statistical tasks. Weighed food intake provides a measured quantity for the meal sessions. Evaluating an image-assisted method requires reporting image-derived quantities and comparing them with the measured reference. The registry’s posted primary result addresses the former, while the stated study purpose concerns the latter.

The comparison also demonstrates why a point estimate and a confidence interval should be read together. The group means are identical and the estimated difference is 0.0 grams, but the interval from -90.32 to 90.32 grams shows that the estimated contrast is uncertain. A p-value of 1.0 does not convert this result into proof of no difference. Readers need the interval and a definition of what differences would matter to assess precision.

Randomized meal order offers a way to compare average intake across the two assigned sequences. At the same time, the registry labels the study factorial without describing the factors, and it does not report key analysis details such as the analysis population or missing-data handling. Clear reporting of those elements would help readers understand precisely which comparisons the design supports.

The study is also a reminder that planning illustrations based on an observed effect should not be confused with the trial’s original design justification or with observed power. The registry record does not report its own sample-size rationale. Planning assumptions and validation targets should be stated directly when the aim is to evaluate measurement accuracy.

Statistically, the record supports a narrow reading of the available outcome: the posted averages for weighed intake did not differ between the assigned meal-order groups. It does not provide the evidence needed to judge the performance of the image-analysis method itself.

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 most direct limitation is that the stated research aim concerns validation of passive image-assisted dietary assessment, while the posted primary outcome results concern weighed food intake by meal order. The registry record does not report image-derived food identifications, portion estimates, accuracy measures, or comparisons between image estimates and the scale-based measurements. The central validation question therefore cannot be answered from the posted primary results.

The reported comparison has 15 participants per group and a confidence interval from -90.32 to 90.32 grams. That interval leaves substantial uncertainty about the difference in mean intake. The registry record does not state what difference would be considered important, so the precision cannot be judged against a prespecified threshold. Nor does the record describe an equivalence design that would support a conclusion of similarity.

Results are reported as group means and standard deviations, without food-specific, meal-specific, or session-specific estimates. Since the meals included different food shapes and compositions, the aggregate means do not reveal whether image assessment or intake differed across those categories. Such results are not reported in the registry record.

Several design and analysis details are also absent. The identity of the masked party, the factors underlying the factorial-model label, the analysis population, missing-data procedures, sample-size justification, interim analyses, and statistical analysis plan are not reported in the registry record. The record does not provide participant characteristics or detailed eligibility criteria, limiting assessment of who the results represent.

Finally, the posted adverse-event counts are zero in both groups, but the record does not describe ascertainment methods or observation details. The counts should be read as the recorded registry entries rather than as evidence that all possible adverse experiences were ruled out. These limitations do not negate the reported summaries; they define the boundaries of what those summaries can support.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Dietary Assessment Study Via Digital Images — NCT03267004.
• Registry record last updated 2025-04-08; status last verified 2025-04.
• Results first posted 2024-12-09.
• 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

NCT03267004, titled “Dietary Assessment Study Via Digital Images,” is a completed randomized, single-masked study with enrollment of 30. Its stated purpose was to validate passive image-assisted dietary assessment using Sony Smarteyeglass images and DietCam automated image analysis. Participants were assigned to one of two meal orders, with Meal A and Meal B served in opposite sequences across two sessions.

The primary outcome was weighed food intake in grams. Each arm had 15 participants, and both had a mean of 354.4 grams and a standard deviation of 126.2 grams. The estimated mean difference was 0.0 grams, with a confidence interval from -90.32 to 90.32 grams and an unadjusted p-value of 1.0.

The registry lists no analysis entries or posted secondary outcomes. Image-estimation accuracy, analysis-plan details, missing-data handling, and the study’s sample-size justification are not reported in the registry record. The available comparison describes weighed intake by assigned meal order; it does not directly report validation performance of the image-analysis method.