A Web-based Bystander Education Program: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Web-based Bystander Education Program (NCT01903876), a randomized na study. The primary outcome is Prosocial Intervening Behavior.

2. Clinical Question

3. Trial Design

The registry describes a completed randomized, parallel-group study with no masking. It lists enrollment of 743 and identifies Emory University as the sponsor. The listed study start is 2010-02, and the primary completion date is 2011-05. The registry does not report a trial phase beyond NA.

Participants were male college students recruited online and assigned to one of two web-based behavioral programs. The Bystander & Sexual Violence Prevention program was a 3-hour intervention designed to teach bystander intervention. The General Health Promotion comparison was also 3 hours and addressed general mental health. Thus, the contrast is between two online programs, not between a program and no intervention.

The registry’s description says investigators asked about intervening and sexually coercive behaviors before the intervention, surveyed participants after the intervention, and assessed outcomes at 6 months. The posted primary outcome is at 6 months. The record does not give the timing or numerical results of the post-intervention assessment in the outcome information.

Random assignment can support a comparison between assigned groups, but interpretation still depends on outcome measurement, follow-up, analysis choices, and completeness of outcome data. The registry record does not report the randomization sequence, concealment procedures, exact eligibility criteria, recruitment flow, or reasons for missing outcome data. It also does not report the study’s original sample-size justification or its planned analysis details. The registry lists 743 enrolled, while the primary outcome counts are 92 and 123 in the respective arms; the record does not explain the difference between overall enrollment and the outcome-specific counts.

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: N=368 participants completed the eligibility screener and provided informed consent in the study, but failed to complete the baseline assessment. Because our randomization occurred following completion of baseline assessment, these participants were not randomized to groups and were not included.

Baseline to Post-Intervention Period

Reasons for not completing this period:

Six-Month Follow-up Period

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 Prosocial Intervening Behavior, measured over 6 months.

The primary endpoint is Prosocial Intervening Behavior at 6 months, measured as a percent score on the Reactions to Offensive Language and Behavior index. The registry describes using a 7-item self-behavior subscale plus an additional 8 items reflecting the content of RealConsent. Participants were presented with 15 potential situations involving offensive language or behavior. For situations they had experienced in the past 6 months, they were asked whether they intervened.

The scale ranges from 0%, meaning no intervention at any time, to 100%, meaning intervention every time. The posted result is a mean, so it describes the average score in each arm rather than the experience of every participant. It is a self-reported measure of whether men confronted inappropriate behavior by other men; it does not directly establish how often such behavior occurred in the wider population or independently verify reports of intervention.

This endpoint is relevant to the study’s stated question because the program was designed to teach bystander intervention, and the registry’s main hypothesis concerned prosocial intervening behavior. General measurement principle: an outcome should be interpreted in light of what its instrument records. Here, the score captures responses to specified situations and depends on participants’ recall and reports. The registry record does not report further information about validation, scoring rules beyond the description, or handling of unanswered items.

7. Results Posted in the Registry

• Bystander & Sexual Violence Prevention: mean 81.10 versus 72.20 on control; difference 8.90 (95% CI 8.82 to 8.98), standardized effect d = 29.18; z-test p = <0.001.

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.

Sexual Violence Perpetration

• 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:
  • Bystander & Sexual Violence Prevention: mean 0.26 versus 0.50 on control; difference -0.24 (95% CI -0.47 to -0.01), standardized effect d = -0.30; z-test p = 0.037.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Bystander & Sexual Violence Prevention than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The posted secondary outcome is Sexual Violence Perpetration, assessed at 6 months. Across the available secondary results, the descriptive direction is consistent with the stated study hypothesis: the Bystander & Sexual Violence Prevention arm has a lower mean score than the General Health Promotion arm. The posted unadjusted comparison and the sponsor’s ANCOVA result are both reported with p-values below .05. This directional pattern aligns with, but does not independently confirm, the primary finding of a higher average prosocial intervening score in the intervention arm.

Secondary outcomes generally carry less evidential weight than a clearly prespecified primary outcome. General: when several outcomes or analyses are examined, the chance of obtaining at least one apparently notable result can increase unless the analysis plan accounts for multiplicity. The registry lists one secondary outcome and does not report a multiplicity adjustment, a complete hierarchy of hypotheses, or a statistical analysis plan. It is therefore not possible to determine from the registry record whether this secondary analysis was one of several originally planned tests or how the investigators intended to interpret multiple results.

The registry describes a secondary score, not an independently verified count of sexual violence events. Its meaning depends on how the scale was defined, completed, and analyzed. The record does not provide additional scoring detail, the ANCOVA estimate, its confidence interval, or missing-data methods. The secondary finding should thus be treated as supportive and exploratory in relation to the primary question, rather than as a standalone basis for broad conclusions.

The primary and secondary endpoints also have different outcome-specific arm counts, indicating that the posted analyses do not contain the same number of observations in each arm. The reason is not reported in the registry record. That difference reinforces the need to consider outcome availability and analysis populations when comparing the overall pattern. The results can be described as directionally aligned, but the registry does not establish why observations are missing or whether the same participants contributed to each outcome.

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

• Sexual Violence Perpetration

9. Statistical Interpretation

10. Statistical Methods Explained

What does a mean difference describe?

General: a mean is the arithmetic average of the observed scores in a group. A mean difference compares the averages between groups, with the direction depending on which arm is subtracted from which. For the primary outcome, the posted means are 72.2 and 81.1, and the recalculated difference is 8.9 percentage points in favor of the higher-scoring Bystander & Sexual Violence Prevention arm. The endpoint is already expressed as a percent score, so the difference is most naturally read in percentage points, not as a relative percentage change.

A group mean does not show the full distribution of individual scores. Participants can vary around the mean, and the same average can arise from different patterns of individual responses. The registry gives standard errors alongside the arm means, but does not report the individual-level score distributions. Consequently, the means alone do not show how much the groups’ individual scores overlap or whether the response was similar across participants.

What is the role of the unadjusted re-calculation?

General: an unadjusted comparison uses the group summaries without controlling for other variables. It can provide a transparent description of the difference between the posted means, but it does not reproduce an analysis that adjusts for baseline measures or other covariates. The registry’s reported analysis method is ANCOVA, a model that generally compares outcomes while accounting for specified covariates. The registry does not identify the covariates used, the model specification, or the precise analysis population.

The recalculated result is therefore useful as a simple summary-statistic comparison, not as a replacement for the sponsor’s analysis. Its unadjusted p-value is <0.001 for the primary outcome. The registry’s ANCOVA p-value is .044. Different methods can produce different p-values because they may use different information and assumptions. The registry record does not provide enough detail to attribute the difference to any particular modeling choice.

What does a p-value tell the reader?

General: a p-value measures how incompatible the observed data are with a specified null model, under the assumptions of the analysis. It is not the probability that the null hypothesis is true, nor does it measure the size or practical importance of an effect. The unadjusted primary p-value and the sponsor-reported ANCOVA p-value arise from different analyses and should be read in their respective contexts. Neither p-value answers questions about generalizability or measurement accuracy.

The registry labels its ANCOVA analysis as non-inferiority or equivalence. Such analyses require a defined margin that sets the boundary for a non-inferiority or equivalence conclusion. The registry record does not report that margin, nor does it report an ANCOVA estimate or confidence interval. As a result, the label and p-value alone are not enough to evaluate the formal claim. The posted unadjusted mean difference should not be treated as an equivalence margin or a formal non-inferiority result.

Why can over-dispersion matter?

General: for outcomes that are counts, the variation can be greater than a simple statistical model expects; this is called over-dispersion. It can make standard errors too small if the model does not account for the extra variation. The primary endpoint in this record is a continuous percent score, not a count of events, so count-model over-dispersion is not the central issue for the posted primary comparison. The registry describes the secondary outcome as a score on a scale as well. It does not report a count model or a model-based dispersion assessment.

For continuous outcomes, interpretation still depends on the distribution of scores, the validity of model assumptions, and the precision of estimates. The registry provides means and standard errors, but does not report diagnostics or individual-level distributions. General statistical explanation cannot resolve those missing details; they require the analysis specification and data.

What can the endpoint-specific counts establish?

The primary outcome arm counts are 92 and 123. The secondary outcome counts are 87 and 115. The counts differ by endpoint, so the analyses do not use identical outcome-specific groups. The registry record does not explain why the counts differ, whether missingness was related to treatment assignment or outcomes, or how missing observations were handled. The overall enrollment of 743 should not be substituted for the number contributing to a particular endpoint analysis.

General: when outcome availability differs across participants, the estimated group contrast can be affected if the missingness process is related to participant characteristics or outcomes. A complete description would usually state the analysis population, missing-data approach, and sensitivity analyses. Those details are not reported in the registry record. The posted summary statistics alone cannot establish that missing data had no effect.

11. Confidence Intervals and Precision

A confidence interval describes a range of values compatible with an estimate under the assumptions of the method used to construct it. General: a narrower interval indicates greater statistical precision for that estimate, but does not guarantee that the estimate is unbiased, clinically important, or free of measurement limitations. Interval interpretation also depends on the analysis model and the quality of the observed data.

For the primary unadjusted mean difference, the posted recalculation gives 8.9, with an interval from 8.818 to 8.982. This interval is very narrow around the recalculated estimate. It reflects the calculation from the posted group summaries and should not be mistaken for the interval from the sponsor’s ANCOVA model: no ANCOVA interval is reported. The registry gives arm-level standard errors of 0.031 and 0.028 for the primary means, but does not provide an adjusted estimate or its precision.

The secondary unadjusted mean difference is -0.24, with an interval from -0.465 to -0.015. This interval is wider relative to the estimated difference than the primary interval, and its endpoints remain on the same side of zero. The sponsor’s ANCOVA interval is not reported for the secondary outcome either. Neither interval captures uncertainty due to possible missing-data bias, self-report measurement, or limited generalizability.

Confidence intervals are not guarantees that the true value lies within the stated range, and they do not convey all sources of uncertainty. They are conditional on the calculation and data. In particular, the recalculated intervals are unadjusted, while the reported sponsor analyses use ANCOVA. Readers should keep those analyses separate and consider the incomplete information on model specification, missing observations, and multiplicity when assessing how precisely the results answer the trial question.

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 effects, not the trial’s original recruitment plan. For these continuous outcomes, the displayed figures are participants per arm, not total events. The registry’s planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect targeted in each row is the stated fraction of the observed effect. Time-to-event planning tables would instead give total events using the Schoenfeld method, but that does not apply to these outcomes.

The 80% power column gives the illustrated participants per arm for a design that would have an 80% chance of detecting the specified effect, under the stated assumptions. The 90% column makes the same kind of planning calculation for 90% power. General: power is a design property calculated under a particular assumed effect and analysis; it is not a probability that a particular observed result is true. Smaller target effects generally require more participants to detect, all else equal. The secondary outcome illustration follows that pattern as the target fraction of the observed effect decreases.

The primary planning illustration reports 1 participant per arm across the listed effect fractions. This result follows from using the observed primary effect in the stated approximation and should not be treated as a practical recruitment recommendation. The planning values are based on the observed effect, and an unusually large observed contrast can yield very small illustrative requirements. No trial-specific justification for the original enrollment is reported in the registry record. These figures are not observed power.

The curve should be read as a planning display under the same assumptions: it relates a chosen effect size to the corresponding sample requirement. It does not show enrollment or outcome results from the trial. The target effect is an assumption in such a calculation, and estimates based on observed effects may be unstable or overly optimistic. A prospective design would need a justified target effect and the trial’s intended analysis, neither of which is reported in the registry record.

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 = 8.900; pooled standard deviation σ = 0.305.
2. n = 2 × (1.96 + 0.842)² × 0.305² / 8.900² = 0.0, rounded up to 1 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 <- 8.89999999999999                      # observed difference in means
sigma <- 0.30496821969424                     # 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(1903876)
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 = 1 (80% power), 1 (90% power)
 75% of observed effect: n per arm = 1 (80% power), 1 (90% power)
 50% of observed effect: n per arm = 1 (80% power), 1 (90% power)
power.t.test: 1.4 
Simulated power at n = 1 per arm: NA (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 sponsor’s reported analysis for the primary outcome is ANCOVA, with p = .044, and the registry labels the analysis as non-inferiority or equivalence. For the secondary outcome, the reported method is also ANCOVA, with p = .042 and the same analysis label. The registry record does not give an estimate, confidence interval, covariate list, or non-inferiority or equivalence margin for either analysis.

The recalculated comparisons use posted arm-level means and standard errors to produce unadjusted mean differences, confidence intervals, and p-values. For the primary outcome, that p-value is <0.001; for the secondary outcome it is 0.037. These recalculations are not the sponsor’s ANCOVA results. Adjustment can change estimates and their uncertainty, and the available record does not provide the specification needed to reproduce the sponsor’s models.

The different p-values should therefore not be treated as a direct contradiction or combined into a single result. They summarize different analyses. The sponsor’s analysis addresses the model the investigators reported, although its numerical estimate and key design details are unavailable in the registry record. The unadjusted calculation gives a transparent comparison of the posted means but cannot establish the sponsor’s formal non-inferiority or equivalence conclusion. The registry does not report enough information to assess that conclusion against a margin.

15. Multiplicity, Missing Data and Analysis Populations

The registry record lists a primary outcome and one secondary outcome, but does not report a full statistical analysis plan or whether any adjustment for multiple testing was used. General: testing several outcomes, time points, subgroups, or model specifications can increase the chance of a nominally small p-value arising by chance. Multiplicity procedures can control or describe that risk, but the appropriate approach depends on the prespecified hypotheses and testing strategy. Without that plan, the degree to which the reported p-values account for multiple comparisons is not reported in the registry record.

The posted outcome counts vary between the primary and secondary analyses. The registry does not report reasons for the different counts, the number or characteristics of participants with missing outcomes, or the method used to handle missing data. It does not report whether analyses used all randomized participants, participants with observed outcomes, or another analysis population. These omissions limit assessment of how well the comparisons preserve the benefits of random assignment.

General: missing data can affect a result when the probability of an outcome being observed is related to participant characteristics, treatment assignment, or the unobserved outcome itself. An analysis restricted to participants with recorded outcomes can be biased under some missingness patterns. Sensitivity analyses can examine whether conclusions change under alternative assumptions. The registry record does not report such analyses, so the posted group summaries cannot establish that missingness had no influence.

The record also does not report the original sample-size justification, interim-analysis plan, covariates in the ANCOVA models, or detailed model diagnostics. It does not provide a formal equivalence or non-inferiority margin despite using that analysis label. These are distinct limitations: the missing margin limits interpretation of the sponsor’s stated analysis type, while the missing analysis plan and missing-data information limit evaluation of the reported estimates and p-values. The available information supports description of the posted results, but not a full audit of the trial’s statistical conduct.

16. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

The registry lists serious adverse events and other adverse events by arm. For General Health Promotion, 367 were at risk for each category, and 0 serious and 0 other adverse events are reported as affected. For Bystander & Sexual Violence Prevention, 376 were at risk for each category, and 0 serious and 0 other adverse events are reported as affected. No serious adverse event terms are listed. Deaths are not reported in the registry record.

These figures are event-reporting entries, not the primary efficacy outcomes. A recorded zero means that no affected participants are listed for that category in the registry record; it does not prove that no adverse experiences occurred or that all events would necessarily have been detected and reported. The record does not provide further information on adverse-event collection methods, the reporting period, severity definitions, or completeness of safety follow-up.

General: safety data are best interpreted alongside the procedures used to solicit, classify, and report events. The available registry figures allow a description of what is recorded, but do not support conclusions beyond those recorded counts. The figures should not be used to infer comparative safety from the absence of reported events alone.

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

The results describe a randomized comparison of two web-based programs among male college students. At 6 months, the posted mean on the self-reported prosocial intervening scale is higher in the Bystander & Sexual Violence Prevention arm than in the General Health Promotion arm. The posted secondary score is lower in the intervention arm. The primary unadjusted mean difference and the sponsor’s ANCOVA p-value are different kinds of summaries and should be interpreted separately.

The findings do not establish that every participant changed behavior, that all reported situations were independently verified, or that the outcomes apply to people beyond the study population. The primary index records reported responses to potential situations participants had experienced. It is not a direct measure of all bystander actions in a college setting. The secondary endpoint is a score on a scale, and the registry does not provide enough detail to translate its mean difference into a count of events.

The record also does not establish whether the observed differences persist beyond the listed 6-month assessment, whether the results generalize to other settings or populations, or whether outcomes differ in groups not described in the record. Although the registry lists overall enrollment of 743, the posted outcome-specific analyses include smaller counts in each arm. The reasons for those counts and the handling of missing observations are not reported.

The sponsor’s analyses are labelled non-inferiority or equivalence, but the registry does not report the margin or model estimate needed to evaluate such claims. The posted results therefore support a careful description of group scores and reported statistical tests, not an independent conclusion about equivalence or non-inferiority. Statistical evidence is also not a measure of practical importance by itself; the registry provides no threshold for a meaningful difference on the scale.

18. Why This Trial Matters Statistically

This trial illustrates why a registry result needs to be read at the level of the measurement and analysis, not only by looking at a p-value. The primary endpoint is a self-reported percent score on a scale designed around potential situations involving offensive language or behavior. That operational definition determines what the mean difference represents and what it cannot establish.

It also shows how different analyses of the same trial can answer related but non-identical questions. The recalculated unadjusted comparison describes the posted group means. The sponsor reports ANCOVA, but does not provide the adjusted estimate, confidence interval, covariates, or margin associated with the non-inferiority or equivalence label. Reporting the distinction prevents an unadjusted calculation from being mistaken for the sponsor’s formal analysis.

Outcome availability is another important part of interpretation. The primary and secondary analyses have different arm counts, while the registry reports overall enrollment of 743. Without an account of missing outcomes and analysis populations, readers cannot determine how the analyzed groups relate to all randomized participants or whether incomplete outcome data could influence estimates.

Finally, planning calculations based on observed effects are not substitutes for the original design rationale. The registry record does not report its sample-size justification. The planning illustrations can show how assumptions about effect size and power affect sample requirements, but they do not explain why the trial enrolled the number it did. Together, these issues make clear that statistical interpretation depends on transparent methods as well as numerical results.

19. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

Several limitations specific to this trial constrain interpretation. The primary endpoint is based on participants’ self-reports about whether they experienced specified situations and intervened. Recall and reporting may affect these scores, and the registry does not report independent verification. The scale’s description identifies its components and range, but the record does not give further psychometric or validation information.

The study compared two web-based programs, so the observed contrast is between the bystander intervention program and a general health promotion program. The posted results do not isolate the effect of each component of RealConsent or compare the program with no program. The registry does not report whether participants engaged with all program content or how adherence was assessed.

The overall enrollment is 743, while the arm counts for the primary and secondary outcomes are smaller and differ across endpoints. Reasons for those differences, the missing-data approach, and the analysis populations are not reported. The registry also does not report detailed eligibility criteria, recruitment flow, randomization implementation, or whether follow-up differed between arms.

The sponsor reports ANCOVA p-values but not model estimates, confidence intervals, covariates, or the stated non-inferiority or equivalence margin. The full analysis plan, multiplicity approach, and interim-analysis details are not reported in the registry record. The unadjusted recalculations cannot fill those gaps. Generalizability is also uncertain because the study concerns male college students recruited online, and the record does not establish applicability to other populations or settings.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A Web-based Bystander Education Program — NCT01903876.
• Registry record last updated 2014-07-23; status last verified 2014-07.
• Results first posted 2014-07-23.
• 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, parallel-group study with no masking, sponsored by Emory University. It compared a 3-hour web-based bystander and sexual violence prevention program with a 3-hour general health promotion program for male college students. Enrollment was 743, and the primary outcome was prosocial intervening behavior at 6 months.

The posted primary means are 72.2 and 81.1, with an unadjusted recalculated mean difference of 8.9 and p < 0.001. The sponsor reports ANCOVA p = .044 and labels the analysis non-inferiority or equivalence; the estimate and margin are not reported in the registry record. The secondary outcome shows a directionally aligned pattern, but multiplicity and missing-data methods are not reported. Interpretation is limited by self-report measurement, differing outcome-specific counts, and incomplete analysis details.