A Multilevel Intervention in the Hunger Relief Network to Improve Diet Among Adults Experiencing Food Insecurity: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Multilevel Intervention in the Hunger Relief Network to Improve Diet Among Adults Experiencing Food Insecurity (NCT03421106), a randomized na study. The primary outcome is Overall Diet Quality of Clients.

2. Clinical Question

The study addresses whether a food-pantry-level change was accompanied by a difference in clients’ diet-quality scores over follow-up. The intervention was delivered at pantry level, while the primary outcome was measured in clients. That distinction matters: the intervention description concerns how pantries offered food, whereas the endpoint concerns clients’ scores.

The registry describes a randomized evaluation with intervention and control pantries. It also describes a sample of clients enrolled at baseline and followed for 1 year, plus a separate follow-up sample used to assess the nutritional quality of food selected at a pantry. The posted primary analysis is restricted to Sample A participants at 11 completed sites. COVID-19 disrupted intervention and post-evaluation activities in 5 sites.

The central statistical question is about the estimated difference in average pre-post change between groups, not whether every client changed in the same way. The posted analysis reports a net mean difference and a p-value, while the recalculation gives an approximate confidence interval for the difference. Neither by itself describes every aspect of the intervention or the variation among individual clients.

3. Trial Design

The registry identifies this as a randomized, parallel-group study with no masking. The planned intervention was implemented through food pantries: intervention pantries would transform food offerings, while control pantries would make no changes during the evaluation period. The investigator’s description refers to 8 intervention and 8 control pantries. The primary analysis population, however, is described as Sample A at the 11 completed sites. The record does not explain the site-level path from the planned evaluation to those 11 sites in further detail.

The registry reports enrollment of 504. That figure is not the same as the numbers in the posted primary outcome groups: the primary outcome reports 89 intervention clients and 104 control clients. The record describes COVID-19 disruption to intervention and post-evaluation activities in 5 sites, and says that the primary analysis includes participants in Sample A at the 11 completed sites. It does not report a participant-flow account that reconciles enrollment with each outcome-specific analysis group.

The primary outcome was assessed at baseline and 1 year. The brief description also mentions cardiovascular health and nutritional quality of food selected at the pantry. The food-selection outcome used a post-sample comparison at 1 year, whereas cardiovascular health used a baseline-to-1-year change measure. These endpoints therefore concern distinct measurements and, in the case of food selection, a distinct sample description.

The registry lists the study status as terminated. It does not report the reason for termination, the original statistical analysis plan, the trial’s original sample-size justification, or detailed procedures for randomization and allocation concealment. The record lists the phase as not applicable. These omissions limit what can be concluded about the design process beyond the details recorded.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Sample A (Followed for 1 Year)

Sample B (Evaluated at Post Only)

Analysis population for the primary outcome, as defined in the registry: COVID-19 disrupted intervention and post-evaluation activities in 5 sites. The analysis includes participants in Sample A at the 11 completed sites.

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 Overall Diet Quality of Clients, measured over Baseline, 1 year.

The primary endpoint is overall diet quality, assessed with Healthy Eating Index-2015 (HEI-2015) scores. The registry describes the measure as a pre-post change in score from baseline to 1 year. The score scale runs from 0 to 100, with 100 representing the best possible score, meaning nutritional quality best aligned with the Dietary Guidelines for Americans.

A change score expresses how a participant’s measurement differs between the specified assessments. The primary comparison then asks whether average change differed between clients in intervention and control pantry groups. In the posted results, the group estimates are least-squares mean changes, and their dispersion is reported as standard errors. The registry’s analysis labels the between-group quantity a net mean difference.

General statistical context: a continuous score can retain information across its range rather than reducing diet quality to a binary category. Comparing change can focus the question on movement over the follow-up period, rather than only comparing scores at one assessment. The score remains a measure of dietary quality as operationalized by HEI-2015; it is not itself a direct measure of every dimension of diet or cardiovascular health.

The registry record does not report details such as the exact score-construction procedures used in this analysis, the handling of missing baseline or follow-up measurements, or a clinically important difference for HEI-2015. Therefore, the posted estimate can be interpreted as a statistical difference in the stated score scale, but the record does not establish a threshold for judging its practical importance.

7. Results Posted in the Registry

• Intervention Food Pantries and Clients: mean 1.09 versus -1.42 on control; difference 2.51 (95% CI -1.61 to 6.63), standardized effect d = 0.17; z-test p = 0.232.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Nutritional Quality of Food Selected by Clients at the Food Pantry

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
• Result:
  • Intervention Food Pantries and Clients: mean 60.60 versus 64.00 on control; difference -3.40 (95% CI -6.57 to -0.23), standardized effect d = -0.31; z-test p = 0.036.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Intervention Food Pantries and Clients than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Cardiovascular Health

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard errors.
• Result:
  • Intervention Food Pantries and Clients: mean -0.09 versus 0.41 on control; difference -0.50 (95% CI -0.93 to -0.07), standardized effect d = -0.33; z-test p = 0.023.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Intervention Food Pantries and Clients than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary endpoints cover two different dimensions: the nutritional quality of food selected by clients at the pantry and cardiovascular health. Their measurement periods differ as well. Food selection is reported as a post-sample comparison at 1 year, while cardiovascular health is reported as a baseline-to-1-year change. These differences mean the endpoints should be interpreted separately rather than treated as interchangeable measures of one outcome.

Across the recalculated secondary contrasts, the numerical direction for intervention relative to control is negative, unlike the positive primary point estimate. This directional pattern alone does not establish a consistent overall effect or explain why results differ. The endpoints use distinct measures and analysis populations, and the registry does not report sponsor analyses for either secondary outcome.

The unadjusted recalculations for these endpoints include p-values and confidence intervals, but the registry record does not report a multiplicity procedure. General statistical context: when several outcomes are examined, the chance of at least one apparently unusual result can increase if each is tested separately without adjustment. Consequently, an unadjusted secondary p-value carries less weight than a prespecified primary analysis considered on its own, particularly when the number of planned tests and their hierarchy are not documented.

The secondary results can be useful descriptively and can help identify questions for further study, but they should not overrule the primary result or be combined into a single conclusion. The record does not state whether these outcomes were prespecified in a hierarchy, whether analyses were exploratory, or how missing observations were handled. The results therefore warrant cautious interpretation as distinct, unadjusted comparisons.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Nutritional Quality of Food Selected by Clients at the Food Pantry
• Cardiovascular Health

9. Statistical Interpretation

10. Statistical Methods Explained

General statistical context: the methods used to summarize a continuous outcome depend on what quantity the analysis targets. A group mean describes the average score or average change in that group. A difference between group means expresses the average contrast in the units of the outcome. Here, the primary outcome is a change in HEI-2015 score, and the registry’s reported analysis is a mixed models analysis of a net mean difference.

What does a least-squares mean represent?

A least-squares mean is a model-based estimate of a group mean. General statistical context: unlike a simple arithmetic mean, a model-based mean can account for terms included in the statistical model. Its interpretation depends on the model specification, including any covariates, repeated-measure structure, and assumptions. The registry record identifies the method as mixed models analysis but does not report the model formula, included covariates, covariance structure, or estimation details. Those aspects cannot be inferred from the method label alone.

The posted primary group results are least-squares mean changes, with standard errors. The standard error describes the estimated uncertainty of a mean, under the model or calculation used; it is not the same as the spread of individual participant measurements. A standard deviation describes variation among observations, while a standard error describes uncertainty in an estimated mean. Keeping those quantities distinct is important when reading outcome summaries.

What does the net mean difference compare?

The net mean difference is the contrast between the intervention and control groups’ estimated changes. The registry reports 2.53 for the mixed-model analysis. The recalculated unadjusted comparison is 2.51, using the posted group estimates. The recalculation does not reproduce the mixed model: it uses the posted summary statistics and does not incorporate unreported model adjustments or the study’s site structure. It is therefore a descriptive cross-check, not an alternative primary analysis.

General statistical context: if outcomes are measured before and after an intervention, comparing group changes asks whether the average movement differed between groups. A positive contrast means the intervention-group change estimate is higher than the control-group estimate on the HEI-2015 change scale. That direction describes the numerical contrast only; interpretation of practical importance also depends on uncertainty, measurement context, and whether the analysis population represents the intended population.

How should the confidence interval be understood?

A confidence interval gives a range of estimates compatible with the observed data under the calculation’s assumptions and procedure. In repeated use of the same procedure under its assumptions, intervals at the stated confidence level would cover the target quantity at the stated rate. It is not a statement that there is a particular probability that this one fixed interval contains the true value. The recalculated primary interval is an unadjusted interval based on the posted summary statistics. It is not reported as the confidence interval from the sponsor’s mixed model.

General statistical context: an interval that includes zero for a difference is compatible with no between-group difference, but that does not prove the groups are identical. An interval’s width conveys precision: a broader interval leaves more values compatible with the data. For the primary endpoint, the interval spans values below and above zero, so the direction and size of the population contrast are not pinned down precisely by the recalculation.

Why is the recalculation approximate?

The primary endpoint is reported as least-squares mean changes with standard errors, while the recalculation uses posted group summaries to estimate an unadjusted difference and interval. The standard deviations used for the recalculation are described as approximate. The analysis therefore does not reproduce the original model-based approach, which may reflect features of the data and design not available in the summary record. The registry record does not report enough model detail to independently reconstruct the mixed-model analysis.

General statistical context: observations from participants within the same pantry may be more alike than observations from participants in different pantries. This is called clustering. If clustering is present, treating all participants as fully independent can understate uncertainty, because the effective information depends partly on the number and distribution of clusters. A mixed model can be used to represent structured variation, but the registry’s method label alone does not reveal how pantry clustering was handled. The original model specification is not reported in the registry record.

Over-dispersion is a general concern when observed variability exceeds what a chosen statistical model assumes. For a continuous outcome, an incorrect variance model can affect standard errors and confidence intervals, even when the mean estimate is similar. Whether over-dispersion was assessed or addressed in this trial is not reported in the registry record. It would be inappropriate to assume either that the issue arose or that it was ruled out.

What assumptions underlie planning illustrations?

The planning illustration uses a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. It uses the observed effect or a stated fraction of that effect as the effect to detect. These are planning assumptions for an illustration derived from the observed result, not known assumptions from the trial’s original design. The original sample-size justification and its assumptions are not reported in the registry record.

General statistical context: a two-sided test considers departures in either direction from a null difference. Equal allocation means the planning exercise assumes the same number of participants in each arm. A normal approximation uses a mathematical approximation for the distribution of the estimate or test statistic. Actual trial design may involve other features, such as clustering or loss of outcome data; the illustration does not establish that those features were incorporated into the original design.

Power is the probability, under a specified alternative and assumptions, of rejecting the null hypothesis using the planned test. The planning figures describe participant counts associated with assumed effects and target power levels. They are not observed power and do not quantify the probability that the completed trial’s result is correct. The trial’s own prespecified target effect and power are not reported in the registry record.

What analysis choices are not documented?

The record does not report the full statistical analysis plan, the missing-data method, whether covariates were prespecified, whether any interim analyses were planned or conducted, or how multiplicity was addressed. It does not report adjustment details for the sponsor’s mixed model. These omissions matter because analytic choices can affect estimates, uncertainty, and interpretation. The findings can be described from the posted analyses, but unreported procedures should not be filled in by assumption.

11. Confidence Intervals and Precision

The recalculated primary contrast is 2.51 HEI-2015 change-score points, with a 95% confidence interval from -1.607 to 6.627. The interval covers zero and includes values on both sides of it. This means the unadjusted recalculation does not pin down the direction of the population contrast with high precision. The posted point estimate is only one value within a wider range of estimates compatible with the calculation.

The interval is based on a simple unadjusted recalculation from posted group summaries. It is not a confidence interval reported for the sponsor’s mixed-model analysis; the registry does not provide that analysis’s interval. The recalculation also uses approximate standard deviations. Therefore, its precision should be read in the context of those limitations and not as a complete reconstruction of the trial’s primary model.

General statistical context: precision is about how tightly an estimate is determined by the data and assumptions, whereas statistical significance concerns a test decision under a specified threshold. A confidence interval can be more informative than a p-value because it displays a range of plausible contrasts and their direction. It does not, by itself, establish whether any value in that range is practically important. The registry record does not state a threshold for a meaningful HEI-2015 change.

The primary analysis includes 89 intervention clients and 104 control clients, while enrollment is reported as 504. The record describes the analysis as limited to Sample A at 11 completed sites after COVID-19 disrupted intervention and post-evaluation activities in 5 sites. These registry entries provide context for the available analysis, but the registry does not report participant flow or the missing-data process in enough detail to assess their effect on precision.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are participants per arm, not total participants and not total events. This is a continuous-outcome setting, and the illustration uses equal allocation between two arms. It gives separate participant counts for target power of 80% and 90% at each stated fraction of the observed effect. For example, the 80% column means the planned count per arm associated with an 80% chance of detecting the assumed effect under the stated planning assumptions; the 90% column gives the corresponding count for the higher target power.

The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The assumed effect is the stated fraction of the observed effect: 100%, 75%, or 50%. When the assumed effect is smaller, more participants per arm are needed to reach the same target power. For the primary outcome, planning for 100% of the observed effect gives 538 participants per arm for 80% power and 720 participants per arm for 90% power; planning for 75% or 50% of the observed effect requires larger counts.

These figures are illustrations derived from the observed effect, not the trial’s own original sample-size justification. The original design assumptions and target effect are not reported in the registry record. This is not observed power: it does not measure the probability that the completed study’s result is correct or establish the probability of detecting a true effect after the trial has been completed. It is a prospective-style planning calculation using stated assumptions, applied to an effect estimated from the posted data.

13. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = 2.510; pooled standard deviation σ = 14.690.
2. n = 2 × (1.96 + 0.842)² × 14.690² / 2.510² = 537.7, rounded up to 538 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
delta <- 2.51                      # observed difference in means
sigma <- 14.6897736270733                     # pooled standard deviation
n_per_arm <- function(power, frac = 1) {
  ceiling(round(2 * (z_a + qnorm(power))^2 * sigma^2 / (delta * frac)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (t-based, so very slightly larger)
cat("power.t.test:", round(power.t.test(delta = delta, sd = sigma, power = 0.80)$n, 1), "\n")

# Simulation: exact two-sample t-test using simulated means and variances
set.seed(3421106)
n    <- n_per_arm(0.80)
reps <- 20000
m_diff <- rnorm(reps, delta, sigma * sqrt(2 / n))             # difference in sample means
s2     <- sigma^2 * rchisq(reps, 2 * (n - 1)) / (2 * (n - 1))   # pooled sample variance
tstat  <- m_diff / sqrt(s2 * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(tstat) > qt(0.975, 2 * (n - 1))), reps))

Output from running the code

100% of observed effect: n per arm = 538 (80% power), 720 (90% power)
 75% of observed effect: n per arm = 956 (80% power), 1,280 (90% power)
 50% of observed effect: n per arm = 2,151 (80% power), 2,880 (90% power)
power.t.test: 538.6 
Simulated power at n = 538 per arm: 0.799 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

14. Analyses Reported by the Sponsor

The registry’s own primary analysis is identified as a mixed models analysis. It reports a net mean difference of 2.53 and p = 0.23 for intervention versus control. The registry does not report a confidence interval for that model-based estimate.

The recalculation from posted summary statistics gives an unadjusted mean difference of 2.51, a 95% confidence interval from -1.607 to 6.627, and an unadjusted p-value of 0.232. It is a simple recalculation and does not reproduce the sponsor’s mixed model. In particular, the record does not give the model specification needed to reconstruct that analysis. Its standard deviations are approximate for the primary outcome.

The registry does not flag a discrepancy between the recalculation and the sponsor’s result. The estimates and p-values are close, but they come from different analytic approaches and should retain their labels. The sponsor’s mixed-model analysis is the trial’s reported analysis; the recalculation is a supplementary, unadjusted comparison based on the posted summaries. The missing model details and confidence interval limit direct assessment of the sponsor’s estimate’s precision.

15. Multiplicity, Missing Data and Analysis Populations

The registry identifies overall diet quality as the primary outcome and lists two secondary outcomes. It reports one primary analysis and does not report sponsor analyses for the secondary outcomes. It does not describe a testing hierarchy, multiplicity adjustment, or a plan for controlling the chance of false-positive findings across multiple outcomes. The approach to multiplicity is therefore not reported in the registry record.

General statistical context: multiplicity arises when several hypotheses or outcomes are examined. If each is tested separately at a conventional threshold without adjustment, the chance of at least one false-positive result across the set can exceed the chance for a single test. Adjustments or prespecified outcome hierarchies can address this issue, but the registry does not say whether such methods were used in the full analysis. The secondary recalculations should consequently be treated as unadjusted rather than as confirmatory evidence.

The registry notes that COVID-19 disrupted intervention and post-evaluation activities in 5 sites and that the primary analysis includes Sample A participants at the 11 completed sites. Enrollment is 504, while the primary outcome groups contain 89 and 104 clients. The record does not provide a complete participant-flow account, explain why observations are absent from the reported group summaries, or describe how missing data were handled. It does not say whether an intention-to-treat population was used or how the analysis population was defined beyond the Sample A and completed-site description.

General statistical context: missing outcomes can affect both precision and the validity of a comparison, particularly if the likelihood of missingness is related to outcome values or group assignment. Different methods rely on different assumptions about those missing observations. Without the analysis plan and missing-data details, the impact cannot be determined from the registry record.

The record also does not report interim analyses, stopping rules, or a protocol for changes to analysis populations or endpoints. The status is listed as terminated, but the reason is not reported in the registry record. These gaps do not negate the posted results; they limit the ability to evaluate how analysis decisions were prespecified and how complete the outcome data were.

16. 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 serious adverse events for 0 of 158 participants at risk in the intervention group and 0 of 159 participants at risk in the control group. It reports 0 other adverse events for 158 at risk in the intervention group and 0 for 159 at risk in the control group. The listed deaths are 1 in each group.

These are descriptive counts in the registry record. The record does not report serious adverse event terms, since no top serious terms are listed, or provide details about the deaths, their causes, timing, or relationship to the intervention. It also does not report whether events were actively collected using the same procedures across groups.

General statistical context: a count of zero recorded events is not proof that no event could occur, and small event counts provide limited information about differences between groups. The figures should not be interpreted as establishing a safety advantage or disadvantage. They describe the adverse-event data posted in the registry, without attributing events to the intervention.

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

The results describe a randomized comparison of clients associated with intervention and control food pantries, focused on change in HEI-2015 diet-quality score. The primary group estimates and net mean difference provide an estimate of the average contrast in the analyzed population. The recalculated confidence interval is broad enough to include zero and differences in either direction, while the sponsor’s reported analysis gives a p-value of 0.23 and no confidence interval in the registry record.

That pattern does not establish that the groups were identical, and it does not establish a clear, precise difference. A non-small p-value is not proof of no effect; it indicates that the reported data do not provide strong evidence against the tested null under that analysis. Conversely, a point estimate alone does not establish a meaningful or reproducible effect. The practical importance of a score difference cannot be judged from the registry record because it does not state a threshold for meaningful change.

The findings apply most directly to the Sample A participants at the 11 completed sites included in the primary analysis. COVID-19 disrupted intervention and post-evaluation activity in 5 sites. The record does not provide enough information about participant flow, missing data, or site-level analysis to establish how closely the analyzed group reflects all participants or the original planned evaluation.

The secondary outcomes are different measures and have unadjusted recalculations without sponsor analyses reported. Their results should not be used to turn an uncertain primary result into a single overall conclusion. Likewise, the adverse-event counts are descriptive and do not establish a causal relationship to the intervention or a comparative safety conclusion.

General statistical context: a randomized design can support a causal comparison when allocation, follow-up, analysis, and assumptions are appropriately implemented. The registry describes randomization, but it does not provide enough detail to evaluate all those conditions. The reported results are evidence about the posted analyses, not medical advice, and they do not show that the intervention works or does not work in other settings.

18. Why This Trial Matters Statistically

Statistically, this trial illustrates the distinction between a point estimate and the precision with which it is known. The primary contrast is positive as a point estimate, but the unadjusted confidence interval spans zero and values in both directions. Reporting both the estimate and its uncertainty gives a more informative account than relying on a p-value alone.

The registry also distinguishes a model-based sponsor analysis from an unadjusted recalculation of summary statistics. The reported mixed-model result is the trial’s own analysis; a simple recalculation can serve as a cross-check but cannot recover model choices that are not described. This is especially relevant when intervention is organized at pantry level and outcomes are measured in clients, because the structure of the data can matter for uncertainty estimates.

The record makes the analysis population and site disruption visible: the primary outcome uses Sample A at 11 completed sites, following COVID-19 disruption in 5 sites. At the same time, details needed to assess missingness, clustering, multiplicity, and the original design assumptions are not reported. The combination shows why transparent reporting of population definitions, analysis methods, and uncertainty is important to interpreting a randomized study.

Finally, the planning illustrations emphasize that required sample size depends on the effect one aims to detect and the desired power. Smaller assumed effects require more participants per arm under the stated assumptions. Because those illustrations use an observed effect and do not represent the trial’s original design rationale, they demonstrate a general planning relationship rather than a retrospective verdict on whether the trial was adequately designed.

19. 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.

Several limitations specific to this trial affect interpretation. The registry describes a plan involving 8 intervention and 8 control pantries, but the primary analysis includes participants in Sample A at 11 completed sites. COVID-19 disrupted intervention and post-evaluation activities in 5 sites. The record does not explain the site-level disposition in greater detail or show whether the completed sites differed from those disrupted.

Enrollment is reported as 504, while the primary endpoint groups contain 89 and 104 clients. The registry does not provide participant flow by assessment, a reason for the difference between enrollment and analysis counts, or a missing-data analysis. It therefore is not possible from the record to assess how exclusions or unavailable measurements may have influenced the estimate.

The primary sponsor analysis is identified as mixed models, but its model formula, covariates, site or pantry structure, and confidence interval are not reported. The unadjusted recalculation uses approximate standard deviations and cannot reproduce that model. The record also does not report the trial’s original sample-size rationale, full statistical analysis plan, multiplicity strategy, interim analyses, or missing-data procedures.

The study status is terminated, but the reason is not reported in the registry record. The secondary outcomes have no sponsor analyses reported, and their recalculated p-values are unadjusted. The adverse-event record gives counts but not event details, causes, or attribution. These limitations do not change the posted values, but they restrict assessment of robustness, generalizability, and the completeness of the trial’s analytic record.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A Multilevel Intervention in the Hunger Relief Network to Improve Diet Among Adults Experiencing Food Insecurity — NCT03421106.
• Registry record last updated 2023-02-09; status last verified 2022-05.
• Results first posted 2023-02-09.
• 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.

23. Record Summary

NCT03421106 is a randomized, non-masked, parallel-group study of a multilevel food-pantry intervention for adults experiencing food insecurity. Intervention pantries were to offer healthier and more appealing food; control pantries were to make no changes during the evaluation period. The registry lists enrollment of 504 and study status as terminated.

The primary endpoint is change in HEI-2015 overall diet-quality score from baseline to 1 year. The primary analysis includes Sample A participants at 11 completed sites; COVID-19 disrupted intervention and post-evaluation activities in 5 sites. The registry reports a mixed-model net mean difference of 2.53 with p = 0.23 and does not report a confidence interval for that analysis. An unadjusted recalculation gives a difference of 2.51 with a 95% confidence interval from -1.607 to 6.627.

Two secondary outcomes are posted, but sponsor analyses are not reported for them. The registry record does not report the original sample-size justification, missing-data procedures, multiplicity approach, or detailed model specification.