Guanfacine to Reduce Relapse Risk in Women With Alcohol Use Disorder (AUD): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Guanfacine to Reduce Relapse Risk in Women With Alcohol Use Disorder (AUD) (NCT03137082), a randomized phase 1, phase 2 study. The primary outcome is Percentage of Days Abstinent.

2. Clinical Question

The stated rationale was that guanfacine may preferentially reduce craving and improve cognitive control in women with alcohol use disorder compared with men, and that these behaviors are related to relapse. The stated objective was to examine effects on drinking measures. These statements describe the study rationale and objective; they are not findings from the trial. The record does not report a men’s comparison group, so it cannot establish a difference in response between women and men.

3. Trial Design

The record identifies this as a phase 1 and phase 2, randomized, triple-masked, parallel-group trial sponsored by Stony Brook University. It lists enrollment as 32. The enrollment figure does not, by itself, describe how many participants contributed data to the primary outcome analysis. Arm-level primary outcome summaries list 8 participants for guanfacine XR and 7 for placebo; the record does not explain the relationship between those figures and the listed enrollment.

The comparison was guanfacine XR 3mgs/Daily versus placebo for guanfacine. The record describes oral administration every 24 hours. It gives a titration period, a full-dose period, and a taper, and characterizes the intervention schedule as 12 weeks. The same schedule is listed for both groups. Details of sequence generation, allocation concealment, and how masking was maintained are not reported in the registry record.

The registry record contains different descriptions of trial duration: its brief summary calls the study a 10-week outpatient clinical trial, while the intervention schedules and primary outcome timeframe refer to 12 weeks. The record does not explain this difference, so the duration cannot be reconciled from the posted information. It lists a start date of 2017-07-12, a primary completion date of 2021-03-31, and a completed status.

The primary endpoint was percentage of days abstinent across the 12-week trial. The registry says this was calculated using the Timeline Follow Back. Mood, alcohol craving, and several emotion-regulation measures are also listed among secondary outcomes. The record does not provide the full visit schedule, a detailed participant flow, or the analysis population definition.

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.

Before assignment: Following consent, participants were scheduled for an eligibility intake (including clinical and physical assessments). If eligible, they were scheduled for a second intake which included completing baseline assessments of alcohol and substance use in addition to alcohol craving, mood, anxiety and emotion regulation. Of the 32 women consented, N=17 did not meet criteria and N=15 were eligible and randomized onto study medication.

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.

Population: 32 women were enrolled in the study. N=17 were not eligible, N=15 were randomized to study medication

6. Primary Endpoint

The registry lists the primary outcome as Percentage of Days Abstinent, measured over 12 weeks.

The primary endpoint is the percentage of days abstinent across the 12-week trial. It is a continuous measure: each participant’s abstinent days are expressed as a percentage of days in the assessment period, and the posted arm summaries describe the mean and standard deviation. The registry record says the measure was calculated using the Timeline Follow Back.

In general, an abstinence-day measure captures the frequency of abstinence over a specified period. It can distinguish patterns that a single status at one visit would not describe, because it reflects abstinence across multiple days. It does not, by itself, describe the timing of drinking episodes, the amount consumed on drinking days, or other dimensions of alcohol use. Those require separate measures.

The endpoint is relevant to the stated study objective of examining drinking measures and the rationale concerning relapse. However, the record does not provide a detailed endpoint definition beyond the percentage of days abstinent across the trial and the named assessment method. It does not specify how missing daily reports were handled, whether every participant had the same assessable period, or how the percentage was constructed when observations were incomplete. Those details matter when interpreting a mean based on repeated reports.

The registry reports a mean and standard deviation for each arm, but does not post a confidence interval for either arm or for the between-group contrast. The group summaries describe the observed outcome values; the mixed models p-value is the reported inferential result. The record does not provide enough model output to translate that p-value into a precise estimate of the size of the group contrast.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does the mean describe?

The primary outcome is continuous, and the registry identifies its summary parameter as a mean. A mean describes the central level of the observed values in a group. For this outcome, the posted values summarize percentage of days abstinent across the trial. A mean is useful for comparing groups on a common scale, but it can conceal variation among participants and does not show how values are distributed. The standard deviation adds information about that variation: a larger standard deviation indicates more spread around the group mean. It is not a measure of the precision of the group comparison.

The record reports a mean and standard deviation for each arm, but no confidence interval for those arm summaries. More importantly for the randomized comparison, it does not report a model-based contrast with a confidence interval. In general, a confidence interval gives a range of values compatible with the data and the assumptions of the analysis, expressing uncertainty in an estimate. Its width helps readers assess precision. A p-value cannot replace that information because it does not show the range of plausible effect sizes.

What does a mixed models analysis do?

The registry names the sponsor’s method as a mixed models analysis but does not provide its specification. In general, mixed models can account for repeated observations on the same participant by representing both group-level patterns and within-participant correlation. They may also use observations collected at more than one time point, depending on the model. Such models require decisions about the outcome structure, fixed effects, random effects, covariance pattern, and handling of incomplete observations. Those decisions are not reported in the registry record, so it is not possible to determine which features were included in this analysis.

The endpoint is described as a percentage across the trial, but the record does not say whether the sponsor’s model used a single participant-level summary or repeated measurements over time. Nor does it report the model’s estimated treatment contrast, uncertainty estimate, degrees of freedom, or diagnostics. The phrase “mixed models analysis” identifies a broad method family, not enough detail to reproduce or fully assess the analysis.

How should the p-value be read?

The sponsor reports p < 0.03 for the primary comparison. In general, a p-value measures how incompatible the observed data, or more extreme data, are with a specified null model, assuming that model and the analysis assumptions hold. A smaller p-value is not a measure of clinical importance, and it is not the probability that the null hypothesis is true. Because the registry gives only a threshold rather than an exact value, and omits the estimated contrast and confidence interval, the p-value does not communicate the magnitude or precision of the between-group result.

The record does not state the null hypothesis in detail, the test used within the mixed model, or the planned significance threshold for the primary analysis. A two-sided significance level of 0.05 appears among the assumptions for a planning illustration, not as a reported description of the trial’s own analysis plan. It should not be treated as evidence that the trial used that threshold for its primary inference.

Why can over-dispersion matter?

In general, over-dispersion occurs when observed variability is greater than a statistical model expects. It can make standard errors too small and confidence intervals too narrow if the model does not account for the extra variation. That concern is relevant in principle when analyzing bounded or percentage outcomes, especially when observations vary across participants or time. The registry does not report a model diagnostic or identify over-dispersion as an issue in this trial. The posted standard deviations describe spread in the arm summaries, but are not enough to diagnose whether the sponsor’s model handled variability appropriately.

What can the recalculation and planning illustrations establish?

The record does not support a recalculated two-arm comparison from the posted summary statistics; the comparison is described as not estimable for this continuous endpoint. Consequently, there is no recalculated contrast, confidence interval, or p-value to compare numerically with the sponsor’s mixed-model result. The sponsor’s own analysis remains the available inferential analysis in the registry.

Any sample-size figures associated with a planning illustration are not observed power and are not a description of the trial’s actual design justification. In general, such illustrations ask how many participants per arm might be needed under stated assumptions to detect a chosen fraction of an observed effect. The record describes the illustration as using a two-sided significance level of 0.05, equal allocation, and a normal approximation. It does not report the trial’s own sample-size justification or establish that its design used these assumptions. Planning calculations cannot repair uncertainty about the model, missing data, or the observed effect’s precision.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Guanfacine XR 3mgs/Daily", "Placebo (for Guanfacine)")
n   <- c(8, 7)
m   <- c(74.7, 50.6)
sp  <- c(27.4, 26.5)
sd  <- sp      # posted standard deviations
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Guanfacine XR 3mgs/Daily  8  74.700  27.400  51.793  97.607
CARM  Placebo (for Guanfacine)  7  50.600  26.500  26.092  75.108

11. Analyses Reported by the Sponsor

The registry’s inferential result is the sponsor-reported mixed models analysis, with p < 0.03. The posted arm summaries give means and standard deviations, but the record does not provide a model-based effect estimate or confidence interval. A separate unadjusted recalculation is not available: the two-arm comparison is not estimable from the posted summary statistics for this continuous outcome.

Because there is no recalculated estimate, the sponsor result cannot be numerically checked against a reconstructed contrast or p-value from the arm summaries. The mixed-model result may incorporate data structure or adjustments that cannot be recovered from the posted summary statistics, but the registry does not describe the model sufficiently to say what it incorporated. The sponsor’s reported analysis is therefore the available primary inferential result, while its magnitude and precision remain unclear in the registry record.

The record marks no discrepancy between a recalculation and the sponsor analysis, but there is no recalculated result to compare. That should not be interpreted as independent confirmation of the mixed-model result. A fuller reading would require the model estimate, confidence interval, model specification, analysis population, and missing-data approach, none of which is reported in the registry record.

12. Multiplicity, Missing Data and Analysis Populations

The record lists 13 outcome measures, 12 analyses, and 10 posted secondary outcomes. The secondary measures include anxiety, depression, alcohol craving, and several difficulties related to emotion regulation. The record does not explain why the numbers of outcome measures, analyses, and posted secondary outcomes differ, nor does it specify which comparisons were included in each analysis.

No multiplicity adjustment is reported in the registry record. When several outcomes or comparisons are tested, the chance of at least one apparently unusual result can rise if each is interpreted without accounting for the family of tests. In general, multiplicity procedures define how to control that risk, but the appropriate approach depends on which outcomes are confirmatory and how the analysis plan was structured. The record does not identify a testing hierarchy, distinguish confirmatory from exploratory secondary outcomes, or say whether any adjustment was made. The primary p-value should therefore be understood as the reported result for the stated primary comparison, with its relationship to the other tests unknown.

Missing-data methods are also not reported. The record does not describe the amount or pattern of missing primary outcome data, how incomplete Timeline Follow Back reports were handled, or whether the mixed models analysis used all available observations under a specified assumption. It does not state whether sensitivity analyses were conducted. These omissions matter because the primary endpoint summarizes abstinent days across a period, and incomplete reporting could affect the denominator or the observed outcome, depending on the method.

The analysis population is not defined. The record lists enrollment as 32, while the primary outcome summaries give arm-specific counts of 8 and 7. It does not say whether these summaries are based on all randomized participants, participants with observed outcome data, or another population. A participant flow and reasons for any exclusions or missing outcomes are not reported. As a result, the reader cannot assess the extent of attrition or determine how closely the analyzed groups correspond to the randomized groups.

13. 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 serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pelvic pain (12.5%).

The registry reports serious adverse events affecting 1 participant in the guanfacine XR group and 0 in the placebo group. It lists 8 participants at risk for serious adverse events in the guanfacine group and 7 in the placebo group. The listed top serious adverse event term is Pelvic Pain, with 12.5%; the record does not clearly assign that term to an arm in the posted summary, so it should not be attributed to one.

For other, non-serious adverse events, the record reports 2 affected participants in the guanfacine XR group and 1 in the placebo group. It lists 8 and 7 participants at risk, respectively. Deaths are reported as 0 in both groups. These are registry counts, not a complete account of event timing, severity, duration, relatedness, or recurrent events. The record does not provide the event-level detail needed to interpret those dimensions.

In general, safety counts are descriptive and should be read in light of the number of participants observed, the period of follow-up, and how events were collected and classified. The record does not describe those collection procedures in detail. The counts alone do not establish whether an event was caused by an intervention, and they do not provide a precise comparative estimate of safety. They should not be used in isolation to make an individual treatment decision.

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

The posted result has two distinct parts. First, the descriptive primary outcome means are higher in the guanfacine XR arm than in the placebo arm: 74.7 versus 50.6 percentage of days abstinent. The reported standard deviations are 27.4 and 26.5, and the arm summaries are based on 8 and 7 participants, respectively. Second, the registry reports a sponsor-conducted mixed models analysis with p < 0.03. These summaries and the sponsor’s model-based test are related but are not interchangeable: the record does not give the model’s estimated contrast or its confidence interval.

The result does not establish the size or precision of the primary comparison. A p-value below 0.03 is not an estimate of how much the groups differed, and it does not show the range of effect sizes compatible with the analysis. The registry does not provide enough information to reconstruct that range. The posted arm summaries also do not explain why their outcome counts differ from the listed enrollment of 32.

The result does not establish that guanfacine reduces relapse, improves cognitive control, or reduces craving. Those ideas appear in the study rationale, while the primary endpoint is percentage of days abstinent. Although craving and emotion-regulation measures are listed as secondary outcomes, the registry does not provide results for every listed secondary outcome in the registry record. The record also does not include a comparison with men, so it cannot establish that effects differ by sex.

The finding should be interpreted within the limits of a small arm-level outcome summary, incomplete reporting of the statistical model, and missing information about analysis populations, missing data, and multiplicity. It describes what the registry reports for this trial; it is not a recommendation for care and does not settle questions about efficacy or safety beyond those posted data.

15. Why This Trial Matters Statistically

This trial illustrates why a reported p-value is only one part of statistical interpretation. The registry presents arm means and standard deviations for a clinically relevant continuous endpoint and reports a mixed-model p-value. But without the model-based contrast and confidence interval, readers cannot assess the estimated magnitude and precision of the primary comparison directly. The distinction between a descriptive summary and an adjusted or repeated-measures model result is important: they answer related questions but may not be identical estimands.

It also illustrates how participant counts need context. The registry lists enrollment as 32, whereas the arm-level primary outcome summaries list 8 and 7 participants. Without a participant flow and a definition of the analysis population, those numbers cannot reveal how much outcome information was missing or why. The completed status does not fill that reporting gap.

The record lists several secondary outcomes as well as a primary endpoint, but does not describe multiplicity handling or the analysis plan. That leaves uncertainty about the intended interpretation of secondary findings and about whether the primary analysis was prespecified in the form reported. Clear outcome hierarchies and transparent missing-data methods are central to assessing a trial’s statistical evidence.

Finally, the record’s duration descriptions do not align: the brief summary calls the study 10 weeks, while the endpoint and intervention schedules refer to 12 weeks. This makes careful attention to endpoint timing especially important. Together, these features show why registry entries are useful starting points for understanding design and reported results, while fuller statistical interpretation depends on complete estimates, uncertainty measures, and analysis details.

16. 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 record constrain interpretation. The study rationale refers to women with alcohol use disorder, but detailed eligibility criteria and population characteristics are not reported. The registry therefore does not establish how closely the participants represented the broader population described in the objective. It also does not report a men’s group, so the rationale’s comparison with men is not directly tested by this two-arm comparison.

The primary outcome summaries include 8 participants in the guanfacine XR arm and 7 in the placebo arm, while the record lists enrollment as 32. The reason for the difference is not reported in the registry record. There is no participant flow, analysis-population definition, or explanation of exclusions and missing data. This limits interpretation of both the group means and the sponsor’s model-based analysis.

Statistical reporting is incomplete for assessing the primary result. The registry gives a mixed models p-value but no estimated contrast, confidence interval, exact p-value, or detailed model specification. It does not describe the model’s assumptions, outcome structure, missing-data handling, or diagnostic checks. No separate recalculated comparison is estimable from the posted summary statistics. The relation of the primary test to the listed outcome measures is also unclear because multiplicity procedures are not reported.

The record gives different trial-duration descriptions: a 10-week outpatient trial in the brief summary and a 12-week timeframe for the intervention schedule and primary endpoint. It does not reconcile them. The adverse-event reporting is also limited: counts are available, but event timing, relatedness, collection procedures, and full event-level details are not. These limitations concern the information reported in the registry and should not be mistaken for proof that the underlying trial lacked additional procedures or data.

17. Related Tutorials

18. Related Statistical Calculators

19. Sources

• ClinicalTrials.gov: Guanfacine to Reduce Relapse Risk in Women With Alcohol Use Disorder (AUD) — NCT03137082.
• Registry record last updated 2024-06-05; status last verified 2024-05.
• Results first posted 2024-06-05.
• 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.

20. Record Summary

The registry describes a completed, randomized, triple-masked, parallel-group phase 1 and phase 2 trial sponsored by Stony Brook University. The comparison was guanfacine XR 3mgs/Daily versus placebo in women with alcohol use disorder. The listed enrollment is 32. The primary endpoint was percentage of days abstinent across the 12-week trial, calculated using the Timeline Follow Back.

The arm summaries report means of 74.7 for guanfacine XR and 50.6 for placebo, with standard deviations of 27.4 and 26.5. The corresponding outcome-summary counts are 8 and 7. The sponsor-reported mixed models analysis has p < 0.03, but the registry does not report a model-based contrast or confidence interval. It also does not describe the analysis population, missing-data methods, multiplicity handling, or full model specification. The brief summary’s 10-week description differs from the 12-week endpoint and intervention timeframe, and the record does not resolve that inconsistency.