Designing Food Voucher Programs to Reduce Disparities in Healthy Diets: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Designing Food Voucher Programs to Reduce Disparities in Healthy Diets (NCT02843178), a randomized na study. The primary outcome is Change in Cup-equivalents of Fruit and Vegetable Intake From Baseline to Month 6.

2. Clinical Question

3. Trial Design

CHIVES was a randomized, factorial trial sponsored by Stanford University. The registry records an enrollment of 359 and describes the allocation as randomized, the masking as none, and the model as factorial. The listed study period began in 2016-12, and primary completion was 2018-10-09. The status is completed.

The four arms combine two design features: when vouchers could be used and whether their use was restricted to fruits and vegetables. Distributed vouchers were issued as four $5 vouchers, each valid for a subsequent week of the month. Lump-sum vouchers were four $5 vouchers valid for an entire month. Targeted vouchers were restricted to fruits and vegetables. The registry says the lump-sum untargeted vouchers could be used for any food except tobacco, alcohol or prepared foods. The full purchase rules for distributed untargeted vouchers are not reported in the registry record.

A factorial design can, in general, evaluate more than one intervention feature within the same trial. Here, the four arms represent combinations of timing and targeting. The registry does not report whether the results were analyzed as factorial main effects, as an interaction between the features, or only through other comparisons. It also does not report the randomization procedure, allocation concealment, stratification, or whether assignment used equal group sizes.

The primary outcome was change in cup-equivalents of fruit and vegetable intake from baseline to month 6, assessed by 24-hour dietary recall. Three secondary outcomes are listed: change in fruit and vegetable intake from baseline to month 12, change in Healthy Eating Index from baseline to month 6, and voucher utilization rate. Although the record lists eight outcome measures, it does not report analysis estimates for the secondary outcomes. The number of analyses listed as posted is zero.

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 Change in Cup-equivalents of Fruit and Vegetable Intake From Baseline to Month 6, measured over Baseline and Month 6.

The primary endpoint is the change in fruit and vegetable intake, measured in cup-equivalents, between baseline and month 6. Intake at the end of month 6 was assessed using a 24-hour dietary recall. A change measure describes how intake at follow-up differs from the participant's baseline level; it is not the same as the intake level at month 6 alone.

Fruit and vegetable intake is relevant to the intervention question because the voucher arrangements were intended to alter access to, and purchasing of, these foods. In general, an outcome linked to the targeted behavior can help assess whether different program features are associated with changes in that behavior. A dietary recall is a way of collecting reported food intake over a specified period. The registry does not report further details about recall administration, the number of recalls per participant, or how intake was converted into cup-equivalents.

The record gives arm-level mean changes and 95% confidence intervals. A mean summarizes the average change for the participants contributing to an arm's estimate. The confidence interval describes statistical uncertainty around that estimate under the analysis used; it is not a range that contains every participant's change. The registry does not give the analysis population, missing-data method, or model used to produce these estimates, so the precise assumptions behind the intervals cannot be assessed from the record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a mean change describe?

For a continuous outcome, a mean summarizes the average value in a group. Here the reported quantity is a change in cup-equivalents of fruit and vegetable intake from baseline to month 6. A positive mean change corresponds to an increase on that measure, while a negative mean change corresponds to a decrease. This is a general interpretation of the direction of a change score, not a claim about the experience of each participant. Individual participants may have changes that differ from the group mean.

The reported arm means are 0.18 for distributed, targeted vouchers; 0.08 for lump-sum, targeted vouchers; 0.09 for distributed, untargeted vouchers; and 0.04 for lump-sum, untargeted vouchers. A numerical ordering of these means is not by itself evidence that the arms differ. In a randomized trial, the comparative question is generally addressed by estimating a contrast between assigned groups and describing its uncertainty. The registry record does not report those contrasts.

How should the confidence intervals be read?

A confidence interval is a statistical summary of uncertainty around an estimate. In general, it gives a range of values compatible with the observed data under the model and assumptions used to calculate it. A narrower interval indicates greater precision than a wider interval, but interval width alone does not indicate whether an intervention has a meaningful effect. A confidence interval is not a range intended to contain a specified proportion of individual participants' outcomes.

The registry reports 95% confidence intervals for the arm means. The interval for distributed, targeted vouchers runs from -0.004 to 0.36; the interval for lump-sum, targeted vouchers runs from -0.09 to 0.26; the interval for distributed, untargeted vouchers runs from -0.09 to 0.26; and the interval for lump-sum, untargeted vouchers runs from -0.13 to 0.21. All four intervals extend across zero. They therefore do not resolve whether the mean change in each arm is positive or negative with high precision. They cannot be used as substitutes for confidence intervals on differences between arms.

Why does the factorial structure matter?

In general, a factorial design combines levels of multiple intervention features. It can support estimates of the average effect associated with one feature across levels of another, and it can assess whether the features' effects depend on each other. The four CHIVES arms combine distributed or lump-sum timing with targeted or untargeted vouchers. To interpret results as effects of timing, targeting, or an interaction between them, the analysis needs to define and estimate those quantities.

The registry identifies the model as factorial, but it does not report the analysis specification or factorial effect estimates. It is therefore not possible to tell from the record whether the principal inference used a timing effect, a targeting effect, an interaction, selected pairwise comparisons, or another approach. Nor is the statistical model, any covariate adjustment, or any method for accounting for repeated measurements reported.

What can the posted estimates establish?

The record contains arm-level primary outcome estimates and confidence intervals, but no posted two-arm comparison estimates. The number of analyses listed as posted is zero, and the record notes that values are missing or not estimable for a two-arm comparison for this continuous endpoint. Consequently, there is no reported contrast estimate from which to assess the magnitude or precision of a difference between voucher strategies.

The sample-size assumptions describe planning illustrations based on the observed effect, a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. In general, a planning illustration asks what sample size would be needed under stated assumptions to detect an effect. It is not a measure of the probability that this completed trial detected an effect, and it is not observed power. Numerical sample-size illustrations are not reported in the registry record. The record also does not report the trial's own sample-size justification, so the planning assumptions should not be mistaken for its original design rationale.

The record gives no standard deviations or other dispersion estimates for participant-level change. It reports confidence intervals, but does not describe the method used to calculate them. As a result, readers cannot verify from the registry whether the intervals came from a particular model, whether any clustering or repeated observations were incorporated, or whether the analysis accounted for the factorial assignment in a specific way.

For a continuous measure, variability among participants matters because participants need not have similar changes. In general, greater variability makes an average harder to estimate precisely for a given amount of information. This is one reason confidence intervals are important alongside means. The arm-level intervals communicate uncertainty about the means, but the missing model details limit evaluation of the assumptions behind that uncertainty.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists three secondary outcomes: change in fruit and vegetable intake from baseline to month 12, change in Healthy Eating Index from baseline to month 6, and voucher utilization rate. It also lists eight outcome measures in total. Results for the secondary outcomes are not reported in the registry record. The record does not describe which outcomes were designated as primary or secondary in a statistical analysis plan beyond the outcome labels presented, nor does it report a hierarchy for testing them.

Multiplicity arises in general when several outcomes, time points, or group comparisons are examined. The chance of at least one apparently unusual result can increase across multiple tests if no adjustment or prespecified interpretation strategy is used. The registry does not report whether any multiplicity adjustment was planned or applied, how many comparisons were conducted, or whether analyses were prespecified. No conclusion about multiplicity control can therefore be drawn from the posted record.

Missing-data handling and analysis populations are also not described. The arm sample sizes attached to the primary estimates are 86, 90, 92, and 91, while enrollment is 359. The registry does not state whether the outcome estimates include all randomized participants, only participants with available month 6 data, or another group. It does not report the amount or pattern of missing endpoint data, reasons for missingness, imputation methods, or sensitivity analyses.

These omissions matter because an analysis based only on participants with observed outcomes can be affected if the availability of follow-up data is related to outcome values or assignment. This is a general concern, not evidence that such bias occurred in this trial. Without the analysis population and missing-data approach, readers cannot assess how the reported estimates relate to the full randomized groups.

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.

The registry reports no serious adverse events in any of the four arms. In the listed serious-event groups, the numbers at risk were 86 for distributed, targeted vouchers; 90 for lump-sum, targeted vouchers; 92 for distributed, untargeted vouchers; and 91 for lump-sum, untargeted vouchers. The number affected was zero in each arm. The registry also reports zero other adverse events in each arm and zero deaths in each arm.

These figures mean that the registry record lists no events in the reported categories during the recorded safety assessment. They should not be interpreted as proof that no adverse experiences occurred in any participant or at any time. The record does not report the safety collection period, event definitions, severity grading, relatedness assessment, or whether all participants were assessed for every category. The safety figures describe the posted reporting and do not establish a comparative safety difference between the voucher arrangements.

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

The primary outcome estimates describe average changes in fruit and vegetable intake within each randomized arm from baseline to month 6. The reported means range from 0.04 to 0.18 cup-equivalents, and each arm has a 95% confidence interval that extends across zero. This indicates uncertainty about the direction and size of each arm's mean change. The registry does not report estimates comparing one arm with another, so the numerical differences among the means cannot establish that one voucher arrangement produced a larger change than another.

The trial's design addresses two practical program features: whether voucher value is made available in weekly installments or as vouchers valid for an entire month, and whether the vouchers are restricted to fruits and vegetables. The four-arm structure can, in general, support evaluation of these features together. However, the posted record does not give estimates for the effects of timing, targeting, or an interaction between them. It therefore does not quantify the comparative evidence for the central design questions.

The outcome is a dietary measure collected by 24-hour dietary recall, and the registry does not provide detail on the number of recalls, measurement procedures, missing responses, or the analysis population. The secondary outcomes are listed but their results are not reported in the registry record. These gaps limit interpretation of how the primary estimates relate to broader dietary changes or voucher use.

These results should not be read as proof that the voucher approaches are equivalent, nor as proof that any approach caused an increase or decrease in intake. Establishing a comparative conclusion would require the relevant contrast estimates, their uncertainty, and sufficient information about the analysis methods and data included. The reported arm summaries alone do not provide that evidence. The record documents a completed randomized factorial trial and its arm-level primary outcome results, while leaving the comparative statistical findings unreported.

13. Why This Trial Matters Statistically

This trial is statistically notable because it uses a factorial arrangement to study more than one feature of a food voucher program. Timing and food restrictions are distinct design choices, and combining them into four randomized arms creates the possibility of evaluating each feature and whether their effects depend on one another. In general, factorial designs can address several related questions within one study, but interpretation depends on how those questions are translated into prespecified contrasts and models.

The posted primary results illustrate why arm means are not enough for comparative inference. The means are 0.18, 0.08, 0.09, and 0.04 cup-equivalents across the four arms, but the registry does not report between-arm estimates or factorial effects. The confidence intervals describe uncertainty around the arm means, not uncertainty around differences between strategies. Without the comparative estimates, readers cannot assess whether the observed numerical ordering is compatible with a meaningful difference or is consistent with sampling variation.

The registry also makes visible a distinction between enrollment and the sample sizes attached to outcome estimates. Enrollment was 359, and the primary estimates have arm sample sizes of 86, 90, 92, and 91. Since the analysis population and missing-data methods are not reported, the meaning of that relationship cannot be fully evaluated.

For biostatistical interpretation, the record is a reminder that design labels alone do not specify an analysis. A randomized factorial trial may be analyzed in different ways, and the assumptions, contrast definitions, multiplicity approach, and missing-data strategy affect what conclusions can be drawn. The available arm summaries support descriptive reading, but the registry does not report enough comparative analysis to answer the trial's central statistical questions.

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 main trial-specific limitation for interpreting the posted result is that the registry gives arm-level means and confidence intervals but no two-arm contrast estimates. It reports that values are missing or not estimable for a two-arm comparison for this continuous endpoint. Without those estimates, it is not possible to quantify the comparative effect of voucher timing or targeting from the posted results.

The record does not describe the statistical model or explain how the factorial design was represented in the analysis. It does not report whether the analysis estimated main effects, an interaction, or particular pairwise contrasts. No p-values or factorial effect results are reported. The stated sample-size assumptions are planning illustrations, and the trial's own sample-size justification is not reported in the registry record.

Although the arm sample sizes are listed, the outcome analysis population is not defined. Missing-data amounts, patterns, methods, and sensitivity analyses are not reported. The registry also does not give the number of recalls per participant, details of recall procedures, or information about how the continuous outcome was summarized from dietary reports. These details would help assess measurement and analysis uncertainty.

The secondary outcomes are listed, but their results are not reported in the registry record. The record does not state how multiple outcomes or comparisons were handled. Further details about eligibility, participant characteristics, the randomization procedure, allocation concealment, and the full rules for distributed untargeted voucher use are also not reported in the registry record. These limitations concern the completeness of the posted information; they do not show that any particular design or analysis procedure was inadequate.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Designing Food Voucher Programs to Reduce Disparities in Healthy Diets — NCT02843178.
• Registry record last updated 2019-08-28; status last verified 2019-07.
• Results first posted 2019-08-28.
• 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

CHIVES (NCT02843178) was a completed randomized factorial trial sponsored by Stanford University, with enrollment of 359 and no masking. Its four arms combined distributed or lump-sum vouchers with targeted or untargeted voucher rules.

The primary outcome was change in fruit and vegetable intake, measured in cup-equivalents from baseline to month 6 using 24-hour dietary recall. The arm-level mean changes were 0.18, 0.08, 0.09, and 0.04, respectively; each reported 95% confidence interval extended from a negative to a positive value. The registry does not report two-arm comparison estimates or factorial effect estimates, and secondary outcome results are not reported in the registry record.

The safety record lists zero serious adverse events, zero other adverse events, and zero deaths in all four arms. Analysis populations, missing-data methods, and the statistical analysis plan are not reported in the registry record.