A PPG Sensor-Based Feedback Intervention for Heavy Drinking Young Adults: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A PPG Sensor-Based Feedback Intervention for Heavy Drinking Young Adults (NCT05090995), a randomized na study. The primary outcome is Total Drinks Consumed.

2. Clinical Question

3. Trial Design

The registry describes a randomized, single-masked, parallel-group behavioral trial, marked completed. It lists Yale University as sponsor and enrollment of 60. The record gives a start date of 2022-02-23 and a primary completion date of 2023-06-01. The phase is listed as not applicable.

The two arms share a self-monitoring component. Participants in each arm were to wear a PPG device for 6 weeks, monitor their health, and report sleep behaviors daily. The distinguishing component in the Self-Monitoring and Feedback arm was personalized feedback based on PPG data and sleep diaries, scheduled for weeks two, four, and six. The Self-Monitoring arm description does not include that feedback.

The primary outcome was assessed at baseline, Week 6, and Week 10 using the Time Line Followback Interview. The registry says the interview asks participants to report daily drinking over the preceding 4 weeks at baseline and over the period since the last assessment at Weeks 6 and 10. The primary-outcome population is described as all randomized participants who started either condition. The registry does not report further detail about masking, adherence, participant flow, or how missing observations were handled.

The record lists secondary outcomes covering drinking patterns, alcohol-related consequences, sleep, and heart rate variability. It also lists a range of outcome measures and analyses. A registered outcome or analysis does not, by itself, establish that a complete numerical result is available for every measure.

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

Analysis population for the primary outcome, as defined in the registry: All randomized participants who started either treatment condition.

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 Total Drinks Consumed, measured over up to Week 10.

The primary endpoint is total drinks consumed. It represents the sum of self-reported drinks over the preceding 4 weeks at baseline and over the interval since the previous assessment at Week 6 and Week 10. The Time Line Followback Interview is the stated measurement method, and the registry says higher scores indicate more drinks consumed.

This endpoint is used to describe drinking quantity over the assessment periods and to compare the outcome pattern between the two conditions over time. The registry’s analysis description includes condition, time, and their interaction, with sex and baseline total drinks as covariates. This makes the question broader than a comparison at only one assessment: the model is described as evaluating whether the pattern across assessment times differs by condition.

The registry says totals were square-root transformed because they were not normally distributed. That transformation matters when interpreting the reported values: the record does not explain whether the posted least-squares means have been converted back to the original drinks scale. The endpoint is labeled in drinks, but the transformation means the scale of the reported model estimates should be interpreted cautiously. The registry does not report a clinical threshold for interpreting a difference in total drinks.

7. Results Posted in the Registry

• Self-Monitoring: mean 7.07 versus 6.76 on control; difference 0.31 (95% CI -0.41 to 1.03), standardized effect d = 0.22; z-test p = 0.399.

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.

Drinks Per Drinking Day

• 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:
  • Self-Monitoring: mean 2.34 versus 2.26 on control; difference 0.08 (95% CI -0.06 to 0.22), standardized effect d = 0.29; z-test p = 0.258.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Self-Monitoring and the control. The interval gives the range of effects that remain compatible with the data.

Percent Heavy Drinking Days

• 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:
  • Self-Monitoring: mean 19.83 versus 18.62 on control; difference 1.21 (95% CI -2.91 to 5.33), standardized effect d = 0.15; z-test p = 0.565.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Self-Monitoring and the control. The interval gives the range of effects that remain compatible with the data.

Percent Abstinent Days

• 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:
  • Self-Monitoring: mean 65.56 versus 66.04 on control; difference -0.48 (95% CI -5.39 to 4.43), standardized effect d = -0.05; z-test p = 0.848.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Self-Monitoring and the control. The interval gives the range of effects that remain compatible with the data.

The listed secondary outcomes span drinking patterns, alcohol-related consequences, sleep measures, and heart rate variability. Numerical summaries and unadjusted recalculations are available for Drinks Per Drinking Day, Percent Heavy Drinking Days, and Percent Abstinent Days. Across these three recalculations, the estimated differences are modest relative to their intervals, and each interval includes zero. These unadjusted comparisons do not provide clear evidence of separation between the arms on those summaries.

The registry also reports p < 0.001 for the Mixed Models Analysis associated with each of these secondary comparisons, while providing no corresponding model estimate or interval. As with the primary outcome, the reported model results and the simple recalculations differ. The discrepancy means the recalculated summaries should not be used to override the sponsor’s analyses, but the missing model estimates make the reported results difficult to evaluate in terms of magnitude and precision.

The secondary outcomes address related aspects of drinking and health, but they do not replace the primary endpoint. A coherent interpretation would consider whether results across outcomes support a consistent pattern, while recognizing that different measures capture different constructs. Numerical results for all other listed secondary outcomes are not reported in the registry record, so an overall pattern across every secondary measure cannot be established.

Unadjusted secondary findings generally carry less weight than a prespecified primary comparison because multiple outcomes create more opportunities for chance findings. The registry record does not report whether multiplicity was addressed, whether a hierarchy of secondary hypotheses was specified, or whether the secondary analyses were exploratory. The listed outcomes should therefore be interpreted cautiously and in the context of the primary analysis, reporting gaps, and the discrepancy between sponsor analyses and recalculations.

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

• Drinks Per Drinking Day
• Percent Heavy Drinking Days
• Percent Abstinent Days
• Alcohol-related Consequences
• Sleep Quality
• Sleep-related Impairment
• Sleep Duration
• Heart Rate Variability (HRV)

9. Statistical Interpretation

10. Statistical Methods Explained

The trial is described as randomized, single-masked, and parallel-group. Random allocation is intended to make treatment groups comparable on average, while a parallel design assigns participants to one condition rather than having each participant receive both. The record does not describe who was masked or how masking was maintained. These design labels do not establish that groups were balanced on every characteristic in this particular trial.

What the primary model is described as doing

The registry says mixed effects models were used to evaluate the effect of condition on total drinks over time. Condition, time, and their interaction were included in the model, with sex and baseline total drinks as covariates. In general, a model with condition and time can account for how measurements vary across assessment occasions, while the interaction represents whether the pattern over time differs by condition. Mixed effects methods are commonly used for repeated measurements because observations from the same participant may be related. The record does not state the exact covariance structure, random-effects specification, estimation method, or handling of incomplete repeated measurements.

Adjustment for baseline total drinks and sex means the described model is not equivalent to a direct subtraction of the two posted arm summaries. Adjustment can change an estimate and its uncertainty. In general, a covariate-adjusted longitudinal model and an unadjusted comparison of summary means answer related but not identical statistical questions. The registry does not supply the model estimate or confidence interval, so the adjusted difference and its precision cannot be assessed from the reported p-value alone.

Transformation and summary statistics

The record states that total drinks were transformed using a square root because the totals were not normally distributed. In general, transformation can make a distribution more suitable for a model’s assumptions or reduce the influence of unusually large values. It also changes the scale on which model estimates are calculated. Unless the method of converting estimates back to the original scale is reported, a transformed-scale estimate should not automatically be interpreted as a difference in raw drinks. Such back-transformation details are not reported in the registry record.

The posted summaries identify least-squares means and standard errors. In general, a least-squares mean is a model-based mean, and a standard error describes uncertainty in an estimated mean. The registry’s arm summaries do not include confidence limits for the individual arms. The recalculated comparison is explicitly a simple unadjusted calculation from the posted summaries; it is not the sponsor’s mixed-model result.

Uncertainty and interpretation

A confidence interval describes a range of values compatible with the data and the calculation’s assumptions. A wide interval indicates that the estimate is not precise enough to distinguish a broad range of possible effects. A p-value measures how unusual the observed result would be under a specified null model; it is not the probability that the null hypothesis is true and does not measure effect size. General statistical interpretation should therefore consider the estimated effect and interval as well as the p-value. For this trial, the sponsor’s interval is not reported in the registry record.

The recalculated confidence interval is based on posted summary statistics, not participant-level data. The calculation uses an approximate pooled standard deviation derived from the reported standard errors, as indicated in the comparison information. General caution applies because this approximation cannot represent the full longitudinal model, covariate adjustment, transformation details, or any dependence among repeated observations. It is useful as a transparent summary-statistic comparison, but it is not a substitute for the analysis specified in the registry.

Reported analyses and missing methodological detail

The registry reports a Mixed Models Analysis and p < 0.001 for the primary comparison, but leaves the estimate and confidence limits unreported. The same kind of reporting appears for the listed secondary comparisons. The record does not describe a multiplicity adjustment, interim analysis, missing-data method, or a full statistical analysis plan. Their absence should not be interpreted as evidence that a particular method was or was not used; these details are not reported in the registry record.

General statistical practice distinguishes the intention-to-treat population, which generally retains randomized participants in their assigned groups, from a population defined by treatment initiation or other eligibility for analysis. The primary-outcome population is described as randomized participants who started either condition. The record does not state how many randomized participants did not start, whether all starters contributed to every assessment, or whether other analysis populations were used. These limits constrain evaluation of departures from the planned randomized comparison.

11. Confidence Intervals and Precision

For the primary outcome, the unadjusted recalculated mean difference is 0.31, with a confidence interval from -0.411 to 1.031. The interval spans zero and includes values on either side of it. This indicates that the summary-statistic comparison does not estimate the direction of a difference precisely. The arm-level standard errors are 0.26 for both groups, but those standard errors describe the reported arm means, not the full longitudinal model.

A confidence interval is useful because it displays the range of effect estimates compatible with the calculation and its assumptions, rather than reducing uncertainty to a single p-value. Here, the interval is from a simple unadjusted recalculation. It does not account for the primary model’s condition-by-time structure, covariate adjustment, or any repeated-measure correlation. It should not be read as the confidence interval for the sponsor’s model.

The registry reports p < 0.001 for the sponsor’s Mixed Models Analysis but does not report its estimate or confidence interval. As a result, the sponsor analysis’s effect magnitude and precision cannot be evaluated from the registry record. A small p-value alone cannot show whether the estimated difference is large or small, and it cannot supply the missing interval. Because the record flags a discrepancy, the sponsor analysis takes precedence over the recalculation; the missing estimate and interval nevertheless limit how fully that analysis can be interpreted.

The square-root transformation further cautions against treating the posted values as straightforward differences in raw drinks without knowing how model estimates were expressed. The registry does not report a transformed-scale confidence interval or a back-transformation method. Precision is therefore best described separately for the available unadjusted calculation and the sponsor-reported test.

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 sample-size figures are planning illustrations based on the observed effect, not the trial’s original design assumptions. They are stated as participants per arm for 80% power and participants per arm for 90% power. They are not total event counts. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; the target effect is the stated fraction of the observed effect. The trial’s own original design assumptions are not reported in the registry record.

The 80% power column gives the illustrative number of participants per arm that would be needed, under those assumptions, to have an 80% chance of detecting the specified effect if that effect were present. The 90% power column uses a higher chance of detection and therefore requires a larger planning sample under the same assumptions. These are planning calculations, not a retrospective measure of the trial’s success.

Rows for 100% of the observed effect use the full observed effect as the assumed target; rows for 75% and 50% use smaller fractions. Smaller effects are harder to distinguish from random variation, so the planning figures increase as the assumed effect gets smaller. For example, the primary-outcome illustration lists 332 participants per arm for 80% power at 100% of the observed effect and 1,326 participants per arm at 50% of the observed effect. These figures depend on treating the recalculated observed effect as a planning input and should not be mistaken for a recommended design.

This is not observed power. It does not tell whether the completed trial had adequate power for the sponsor’s model, and it does not replace the uncertainty interval for the result. Because the primary recalculation and sponsor analysis disagree, the observed-effect basis for the illustration also warrants caution.

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 = 0.310; pooled standard deviation σ = 1.424.
2. n = 2 × (1.96 + 0.842)² × 1.424² / 0.310² = 331.3, rounded up to 332 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 <- 0.31                      # observed difference in means
sigma <- 1.42407864951343                     # 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(5090995)
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 = 332 (80% power), 444 (90% power)
 75% of observed effect: n per arm = 589 (80% power), 789 (90% power)
 50% of observed effect: n per arm = 1,326 (80% power), 1,774 (90% power)
power.t.test: 332.2 
Simulated power at n = 332 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 reports the sponsor’s primary analysis as a Mixed Models Analysis with p < 0.001. It does not report the model’s effect estimate, confidence interval, or further numerical detail. A separate unadjusted recalculation from the posted arm summaries gives a mean difference of 0.31, a confidence interval from -0.411 to 1.031, and p = 0.399.

These results do not agree. The recalculation compares posted summaries and does not reproduce the stated model, which includes condition, time, their interaction, sex, and baseline total drinks. It also cannot reconstruct the repeated-measures analysis or the effect of the square-root transformation. Those methodological differences may be relevant, but the registry does not explain the discrepancy, so no specific explanation can be assigned.

Because the discrepancy is flagged, the sponsor’s own analysis takes precedence over the simple recalculation. That does not remove the reporting limitation: without the sponsor’s estimate and interval, the magnitude and precision of the sponsor result cannot be assessed. The p-value alone cannot resolve those omissions. The recalculation remains a descriptive check on the posted summaries, not an alternative sponsor result.

15. Multiplicity, Missing Data and Analysis Populations

The registry lists multiple secondary outcomes and analyses in addition to the primary endpoint. When many outcomes or comparisons are tested, the chance of at least one apparently unusual result can increase. General approaches to multiplicity include adjusting significance thresholds or specifying a hierarchy of hypotheses in advance. Whether either approach was used for this trial is not reported in the registry record. The reported p-values should not be assumed to be adjusted, and the unadjusted recalculations should not be treated as multiplicity-controlled evidence.

The record does not report interim analyses or a statistical analysis plan with details of prespecification. It also does not state how missing observations were handled in the mixed models. In general, mixed models can use available repeated measurements under particular assumptions about missingness, but the analysis method alone does not establish which missing-data assumptions were made or whether sensitivity analyses were conducted. Those details are not reported in the registry record.

The primary-outcome population is described as all randomized participants who started either treatment condition. This is narrower in wording than a population that necessarily includes every randomized participant, because participants who did not start are not addressed by that definition. The registry does not report the number who started, the number contributing data at each assessment, reasons for missing observations, or whether an intention-to-treat analysis was also conducted. Although the arm summaries list 30 participants in each arm, that does not clarify completeness of follow-up or inclusion in each model.

These omissions matter because missingness and analysis-population choices can affect the interpretation of a randomized comparison. They do not establish that the analysis was biased, but they prevent a reader from checking the assumptions or tracing participants through each assessment. The sponsor’s estimates, confidence intervals, missing-data procedures, multiplicity approach, and participant-flow details are not reported in the registry record.

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 5% reporting threshold.

The registry lists 30 participants at risk for serious adverse events in each arm and zero serious adverse events in each arm. It also lists 30 participants at risk for other adverse events in each arm and zero other adverse events in each arm. Deaths are listed as zero in both arms, and no top serious-event terms are listed.

These figures describe what was recorded in the registry; they do not establish that no adverse experiences occurred outside the categories or reporting period captured. The record does not provide further detail about event definitions, surveillance procedures, reporting completeness, severity classifications beyond the serious-event category, or the time window for these safety summaries. With zero recorded events, the registry figures do not support a comparison of event rates between the conditions. They should be read as reported counts, not as proof of absence of risk.

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

The result concerns a behavioral intervention package combining PPG-based self-monitoring with personalized feedback, compared with PPG-based self-monitoring without that feedback. The primary endpoint is self-reported total drinks assessed over repeated intervals through Week 10. The registry reports a sponsor mixed-model p-value of p < 0.001, but does not provide the model estimate or confidence interval. The available unadjusted recalculation does not align with that result, and the record flags the discrepancy.

This evidence therefore does not provide a complete numerical account of the estimated effect under the sponsor’s model. A p-value indicates evidence against a specified null model under the analysis assumptions; it does not convey the size of the difference, its clinical importance, or its precision. The unadjusted estimate and interval provide a separate summary-statistic comparison but do not reproduce the longitudinal adjusted analysis.

The registry record does not report a threshold for a meaningful change in total drinks, longer-term outcomes beyond the stated assessment period, or results for every listed secondary measure. It also does not supply details needed to evaluate missing observations, multiplicity, or the sponsor model’s assumptions. The record consequently supports a description of the interventions, endpoint, and reported analyses, but leaves important questions about effect magnitude and robustness unresolved.

These findings should not be generalized automatically to populations or intervention settings that are not described in the record. The registry entries concern heavy-drinking young adults in this specific randomized comparison. They do not establish that the intervention is effective or ineffective in other settings, nor do they establish the mechanism through which any observed difference might occur.

18. Why This Trial Matters Statistically

This trial is statistically informative because it compares two active behavioral conditions that share device-based self-monitoring, while the feedback arm adds personalized information. That structure focuses the comparison on the added feedback component rather than contrasting the complete package with no intervention. Interpretation still depends on the randomized design, follow-up, outcome measurement, and analysis details available in the registry.

The primary outcome is measured repeatedly and the registry describes a model that includes condition, time, and their interaction. This emphasizes a key statistical distinction: a longitudinal treatment question is not necessarily captured by comparing two arm summaries at a single point or by subtracting summary means. The discrepancy between the reported mixed-model p-value and the unadjusted recalculation makes transparent reporting of estimates and confidence intervals especially important.

The record also illustrates why significance alone is not enough. Without the sponsor model’s effect estimate and interval, readers cannot evaluate its magnitude or precision. Conversely, a simple recalculation can be useful for checking posted summaries but cannot replace a model that adjusts for baseline values and uses repeated observations. The distinction between those analyses is central to responsible interpretation.

Finally, multiple drinking, sleep, and physiological outcomes create a broad outcome landscape. Clear reporting of prespecified hypotheses, multiplicity handling, missing-data methods, and complete outcome estimates would help readers judge how much weight to place on individual findings and whether results form a consistent pattern.

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 are specific to the information in this trial record. The sponsor’s mixed-model analysis is reported with a p-value but without an estimate or confidence interval. This prevents assessment of the sponsor model’s effect magnitude and precision, and the reason it differs from the unadjusted recalculation is not reported in the registry record.

The primary outcome is based on self-reported drinking assessed retrospectively using the Time Line Followback Interview. The registry describes the measure and assessment intervals but does not report validation details for this trial, completeness of diary or interview data, or whether outcome assessors were masked. The masking field says single, but who was masked is not reported in the registry record.

The record states that total drinks were square-root transformed, yet does not explain how the reported least-squares means relate to the original drinks scale or whether any back-transformation was applied. Details of the mixed model, missing-data handling, model diagnostics, and sensitivity analyses are also not reported. These gaps limit reproducibility and make it difficult to determine whether the model assumptions were appropriate.

The primary population is described as randomized participants who started a condition, while participant flow and follow-up completeness are not reported. Multiplicity procedures and interim analyses are also not reported in the registry record. Finally, numerical results are unavailable for several listed secondary outcomes, limiting assessment of consistency across drinking, sleep, and physiological measures.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A PPG Sensor-Based Feedback Intervention for Heavy Drinking Young Adults — NCT05090995.
• Registry record last updated 2024-12-13; status last verified 2024-11.
• Results first posted 2024-12-13.
• 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

The registry describes a completed randomized, single-masked, parallel-group behavioral trial sponsored by Yale University, with enrollment of 60. It compared PPG-based self-monitoring with the same self-monitoring plus personalized feedback. The primary outcome was total drinks consumed, assessed at baseline, Week 6, and Week 10 and analyzed, according to the record, with a mixed effects model adjusting for sex and baseline total drinks.

The sponsor analysis is reported as p < 0.001, but its estimate and confidence interval are not reported. An unadjusted recalculation from the posted summaries gives a mean difference of 0.31 with a confidence interval from -0.411 to 1.031 and p = 0.399. The record flags this discrepancy, so the sponsor analysis takes precedence, while its missing estimate and interval limit interpretation. Safety summaries list zero serious adverse events, other adverse events, and deaths in both arms.