Project MARS: Mobile-Assistance for Regulating Smoking: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Project MARS: Mobile-Assistance for Regulating Smoking (NCT04376489), a randomized na study. The primary outcome is Proportion of Randomized Decision Points in Which Participant Reported Proximal Engagement by Randomized Assignment.

2. Clinical Question

3. Trial Design

Project MARS was a randomized behavioral study, with randomization occurring within participants rather than between separately assigned treatment groups. The registry describes assignment up to 6 times daily for 10 days, using a 2:1:1 scheme for no intervention prompt, a prompt to engage with a low-effort strategy, or a prompt to engage with a more effortful strategy. Participants could therefore contribute decision points under different prompt assignments.

The registry record lists the study as single-group and unmasked. Those labels should be read alongside the description of repeated within-participant randomization: the displayed comparison does not represent two independently enrolled groups. The primary outcome analysis included 99 participants. Although 114 people started the study and 111 were non-beta participants, the registry says sub-par fidelity to the study protocol led to use of 99 participants for analyzing results.

The study began on 2020-08-27, had a primary completion date of 2023-10-20, and is listed as completed. The sponsor was the University of Utah. The registry describes behavioral prompts aimed at capitalizing on states of vulnerability and receptivity, but further details about the protocol or its implementation are not reported in the registry record.

The main comparison contrasts intervention prompts with no prompt. The listed interventions include prompts recommending low-effort strategies under different combinations of negative affect and cigarette availability, as well as more effortful activities. The registry also lists secondary outcomes involving proximal cigarette use, nicotine-containing non-cessation product use, and tobacco product use. The primary results concern proximal engagement, not those secondary outcomes.

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:

Analysis population for the primary outcome, as defined in the registry: Of the 114 who started the study, 111 were non-beta participants. However, due to sub-par fidelity to the study protocol, 99 were used for analyzing results.

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.

Population: Of the 114 who agreed to participate following informed consent, 3 were beta testers whose data were used to verify study procedures and functionality of the software used to schedule and execute micro-randomization. As such, they were not included in the calculation of baseline characteristics, or estimates of treatment effects.

6. Primary Endpoint

The registry lists the primary outcome as Proportion of Randomized Decision Points in Which Participant Reported Proximal Engagement by Randomized Assignment, measured over 1 hour following micro-randomization.

The primary endpoint is the proportion of randomized decision points at which a participant reported proximal engagement with a strategy during the one hour following micro-randomization. Engagement was binary at each decision point: a participant was classified as engaged if they answered yes to at least one of three questions about using a recommended or other strategy in the last hour. A no answer to all three items counted as not engaged.

The registry says proportions were first determined for each participant and then averaged across participants. Thus, the reported outcome summarizes participant-level engagement proportions rather than simply describing a count of all affirmative decision points pooled together. Participants could have up to 60 decision points.

This endpoint is useful for assessing whether a prompt is followed by reported use of a strategy in the short term. It is proximal to the prompt and reflects the behavior the prompt was designed to encourage. It is not itself a measure of smoking cessation, reduced tobacco use over a longer period, or health outcomes. Its interpretation also depends on self-report: the recorded measure is participants’ answers to the three questions, not an independently verified observation of strategy use.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary measure is a proportion, but the randomization and data collection occurred repeatedly within participants. A participant could encounter different prompt assignments across decision points, and each decision point asked whether a strategy had been used in the preceding hour. This repeated structure matters: the observations from one participant are not equivalent to observations from unrelated people. Statistical methods for repeated measurements need to account for the fact that responses from the same participant may be related.

What does the participant-level proportion represent?

For each participant, the study determined a proportion of decision points with reported engagement, then averaged those proportions across participants. This gives a participant-level summary before the across-participant average is calculated. It differs conceptually from pooling every decision point and treating them all as independent observations. The registry does not report all details of the estimation procedure used to produce the risk ratio, so the exact handling of repeated observations beyond the stated participant-level averaging is not reported in the registry record.

The registry lists 99 participants for each of the two displayed arms, while also describing within-participant randomization. That displayed arm count should not be read as 99 separately randomized people in each arm. Rather, participants contributed to assignments within the same study, and the primary population was 99 participants. The total number of decision points contributing to each arm is not reported in the registry record.

How should a risk ratio be read?

In general, a risk ratio compares the frequency or proportion of an outcome in one condition with the frequency or proportion in another. A ratio of 1 represents equal proportions; values above 1 indicate a larger proportion in the first condition, while values below 1 indicate a smaller proportion. Here the posted risk ratio is 1.11 for intervention prompt versus no intervention prompt. The interpretation is relative and does not by itself state the absolute difference in engagement proportions. No absolute difference is calculated in this explanation.

The registry reports a two-sided t-test for the primary comparison. A two-sided test considers departures in either direction rather than only a higher engagement proportion under prompts. The reported p-value is 0.03. In general, a p-value measures how unusual data at least as incompatible with a specified null hypothesis would be if that hypothesis and the analysis assumptions held. It is not the probability that the null hypothesis is true, and it does not quantify the practical importance of the effect.

What does the confidence interval add?

The primary risk ratio is accompanied by a 95% confidence interval. In general, a confidence interval expresses statistical uncertainty around an estimate under the method’s assumptions. The interval provides a range of effect sizes compatible with the observed data and the specified procedure. Its width reflects precision: narrower intervals indicate greater precision than wider intervals, all else being equal. The interval does not account automatically for systematic measurement error, protocol deviations, or incomplete reporting.

The arm-level proportions are also accompanied by 95% confidence intervals. These describe uncertainty in each participant-averaged arm summary, but comparing separate arm intervals is not the same as testing the contrast between arms. The direct risk-ratio estimate and its confidence interval are the registry’s reported inferential comparison.

Why does repeated randomization matter?

Within-participant randomization can make comparisons less dependent on differences between people, because the same participant may experience more than one assignment. At the same time, responses at nearby decision points may be related, and a prompt or strategy use at one point could potentially be associated with later responses. Such dependence and timing are general considerations for repeated-measures studies. The precise model used to address these features, including any adjustment for time or other variables, is not reported in the registry record.

Over-dispersion is a general statistical issue that can arise when observed variation is greater than a model expects. If a model assumes more similarity or less variability than the data contain, uncertainty estimates may be too small. Conversely, accounting for extra variation can make intervals wider. The registry record does not identify an over-dispersion assessment or report whether it was relevant to this analysis, so no conclusion about that issue can be drawn for this trial.

What do the additional estimates describe?

The registry includes interaction estimates on the log-risk-ratio scale and simple effects on the risk-ratio scale, each with 90% confidence intervals. An interaction, in general, asks whether an effect differs across levels of another factor. A simple effect describes a contrast within a specified level or condition. The registry labels these analyses but does not provide enough context to identify the factors or conditions for each estimate. Their substantive interpretation, and whether they were planned or exploratory, are not reported in the registry record.

The registry does not report the detailed statistical analysis plan, the rationale for the chosen sample size, interim analyses, or a multiplicity procedure. General planning calculations based on an observed effect are illustrations of possible sample sizes under stated assumptions, not evidence of observed power and not proof that the study was designed to detect a particular effect. Their assumptions do not replace the trial’s own design justification, which is not reported in the registry record.

10. Analyses Reported by the Sponsor

The registry’s reported primary analysis gives a risk ratio of 1.11, a 95% confidence interval from 1.01 to 1.22, and p = 0.03 for intervention prompt versus no intervention prompt. The displayed arm summaries give participant-averaged engagement proportions of 0.655 and 0.584, respectively, with their own 95% confidence intervals. These are the sponsor-reported results available in the registry record.

The registry record does not report a separate recalculated primary risk ratio or confidence interval. It therefore does not present an independent numerical recalculation that can be compared directly with the sponsor’s estimate. The record does not flag a discrepancy between sponsor and recalculated results. The reported analysis should be read as the registry’s primary comparison; the arm summaries alone do not recreate that analysis or establish how its uncertainty was estimated.

The additional interaction and simple-effect estimates are also posted, but their specific contexts and the full statistical model are not reported in the registry record. They should not be treated as a second, fully specified analysis of the overall prompt effect. The primary risk-ratio result remains the clearest reported contrast, interpreted within the limits of the endpoint, analysis population, and available methods information.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists five outcome measures and 16 analyses. It identifies proximal engagement as the primary outcome and lists proximal cigarette use, proximal nicotine-containing non-cessation product use, and proximal tobacco product use among the secondary outcomes. The primary comparison reports a two-sided p-value and a 95% confidence interval. Several additional interaction and simple-effect estimates are also posted with 90% confidence intervals.

When multiple outcomes, contrasts, or subgroup analyses are examined, the chance of finding an apparently notable result can increase. In general, multiplicity procedures are used to manage the interpretation of a family of statistical tests, for example by adjusting thresholds or intervals. The registry record does not report whether a multiplicity procedure was used, how the 16 analyses were grouped, or which analyses were considered confirmatory versus exploratory. The additional estimates should therefore be read cautiously and not as a collection of independent confirmations.

The analysis population is described as 99 participants, following 114 people who started the study and 111 non-beta participants. The registry attributes the use of 99 participants to sub-par fidelity to the study protocol. It does not explain the specific fidelity criteria, identify the number or reasons for exclusions in more detail, or report how excluded participants’ data were handled. Missing-data procedures and sensitivity analyses are not reported in the registry record.

The study repeatedly randomized decision points within participants, but the registry does not report the detailed model or its handling of missing responses at individual decision points. The primary endpoint’s questions depend on participant responses about strategy use in the last hour; the extent of unanswered items or how such responses were classified is not reported in the registry record. These omissions limit assessment of the robustness of the estimates. They do not negate the reported result, but they leave uncertainty about how alternative analysis choices or missing-data assumptions might affect it.

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.

Most frequent non-serious adverse event terms

The registry reports serious adverse events for the current-smoker group with 114 participants at risk and 0 affected by a serious adverse event. It reports 2 affected by other adverse events and 1 death. No serious adverse event terms are listed. These figures describe what the registry record reports for the study; they do not, by themselves, establish whether an event was caused by a prompt or another aspect of study participation.

A count of zero serious adverse events in this record should be read as no serious adverse events reported among the listed participants at risk, not as proof that no serious event could occur. The other adverse-event count and death count are reported separately from the serious-event count. The record does not provide event descriptions, timing, severity details for the other events, or relatedness assessments, so their clinical circumstances cannot be characterized from the available information.

These safety figures are not an efficacy outcome and do not change the meaning of the proximal-engagement comparison. The registry does not report further safety-analysis methods or additional detail about follow-up for adverse events.

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

The primary result describes reported engagement with self-regulatory strategies during the hour after a randomized decision point. The posted estimate favors a higher proportion of engagement at intervention-prompt decision points than at no-prompt decision points, with a risk ratio of 1.11 and a 95% confidence interval from 1.01 to 1.22. The estimate is statistically distinguishable from 1 under the reported comparison, but its interpretation remains specific to the defined, self-reported proximal endpoint.

This result does not establish that prompts caused people to stop smoking, reduced their longer-term cigarette use, or improved health. Those are different outcomes from answering yes to at least one of the strategy-use questions within the preceding hour. The registry lists product-use outcomes separately as secondary outcomes, and the primary engagement result cannot be substituted for those outcomes.

The reported proportion is an average of participant-level proportions. It is not a percentage of all participants who ever used a strategy, nor a count of unique people who responded to a prompt. The within-participant design also means the two displayed arms are not separate cohorts. Participants could contribute decision points under different randomized assignments.

The interpretation is further bounded by the analysis population and reporting detail. The registry says 99 participants were used for analyzing results after 114 started and 111 were non-beta participants, due to sub-par protocol fidelity. The criteria and consequences of that exclusion are not described in detail. The exact statistical model and the handling of missing decision-point responses are also not reported in the registry record.

The estimate is therefore evidence about a near-term, self-reported behavior in the analyzed study population under the reported randomization and analysis. It is not a complete account of the intervention’s effects, and it does not show whether any engagement was sustained or translated into changes in tobacco use. Broader claims would require evidence from the relevant outcomes and a fuller account of the analysis and follow-up.

14. Why This Trial Matters Statistically

Project MARS raises a statistical question that differs from a conventional comparison of two groups: how to evaluate randomized prompts delivered repeatedly to the same participants. Such a design can study momentary responses to interventions as circumstances change. It also requires careful interpretation because observations from one person may be related, and participants may experience several prompt conditions over time.

The endpoint makes the time scale explicit. It asks whether participants reported using a strategy in the hour after a decision point, which closely matches the intended immediate response to a prompt. Averaging participant-level proportions also makes the participant, rather than each decision point alone, central to the summarized outcome. That choice is important to understand when interpreting the reported engagement proportions.

The registry’s reporting illustrates why a statistically interpretable headline estimate is only part of the evidence. A risk ratio and confidence interval are available for the main prompt-versus-no-prompt comparison, but details about the model, missing responses, multiplicity, and the contexts of secondary interaction estimates are not reported in the registry record. Those details are needed to assess how the estimates were generated and how much weight to place on additional analyses.

The record also distinguishes an analyzed population from the number of people who started or were classified as non-beta participants. Protocol fidelity led to a smaller analysis population, but the criteria are not described in detail. For repeated-measures research, transparent reporting of participant inclusion and decision-point data handling helps readers understand which responses inform the estimate.

Statistically, the study is useful as an example of how a proximal behavioral endpoint and within-participant randomization can address a focused question. Its primary result is about engagement with strategies, not a final clinical outcome. Keeping that distinction clear prevents a short-term process measure from being mistaken for evidence about smoking behavior or health.

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 registry provides limited detail about the statistical model used for repeated within-participant observations. It reports a two-sided t-test for the primary comparison, but the full estimation procedure and the handling of dependence among decision points are not reported in the registry record. The number of decision points contributing to each condition is also not reported.

The analysis population was reduced to 99 participants because of sub-par fidelity to the study protocol. The record does not specify the fidelity thresholds, the number excluded for each reason, or how exclusions related to response patterns. The possibility that the analyzed participants differ from those not included cannot be assessed from the registry description alone.

The endpoint relies on participant self-report about strategy use in the preceding hour. The record describes the three items and the binary coding rule, but it does not report the frequency of unanswered items or the method used to handle them. It also does not provide independent confirmation of reported engagement.

The record lists multiple outcome measures and analyses but does not report a multiplicity plan or identify the inferential status of each additional estimate. The contexts of the posted interaction and simple-effect estimates are not sufficiently described for confident interpretation. The protocol, sample-size justification, interim analyses, and detailed analysis plan are not reported in the registry record.

Finally, proximal engagement is not a direct measure of cessation or longer-term tobacco use. The primary comparison should be interpreted as evidence about the specific short-term self-reported outcome, within the analyzed participants and the reported study design. The registry’s results do not resolve whether engagement translated into later behavioral or clinical changes.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Project MARS: Mobile-Assistance for Regulating Smoking — NCT04376489.
• Registry record last updated 2026-08-31; status last verified 2026-07.
• Results first posted 2026-07-29.
• 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

Project MARS: Mobile-Assistance for Regulating Smoking was a completed randomized behavioral study sponsored by the University of Utah. It began on 2020-08-27 and had a primary completion date of 2023-10-20. The study used repeated within-participant randomization among no prompt, a low-effort strategy prompt, and a more effortful strategy prompt.

The primary endpoint was participant-level proportion of randomized decision points with self-reported proximal engagement during the one hour following micro-randomization. The analysis used 99 participants after 114 people started and 111 were non-beta participants; the registry attributes the smaller analysis population to sub-par protocol fidelity.

The reported primary risk ratio for intervention prompt versus no intervention prompt was 1.11, with a 95% confidence interval from 1.01 to 1.22 and p = 0.03. This is a short-term engagement result, not a direct measure of cessation or long-term tobacco use. Detailed statistical-plan and missing-data information are not reported in the registry record.