Evaluation of Green Dot a Primary Prevention Intervention: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Evaluation of Green Dot a Primary Prevention Intervention (NCT01878097), a randomized na study. The primary outcome is Average Number of Sexual Assault Events Used (Perpetrated) Per School..

2. Clinical Question

In statistical terms, the central comparison is between groups of schools rather than a simple comparison of individually randomized students. Students contributed anonymous reports, while the stated primary measure aggregates reports at the school level and adjusts for baseline and number of students. That distinction matters: the outcome is not described as the proportion of individual students who perpetrated an event, nor as a count of events for a single student. The target is an adjusted school-level average. The registry identifies the study as randomized and parallel, but does not report the number of schools, the allocation process in detail, or the precise statistical model. Those omissions limit how specifically the contrast can be reconstructed from the registry record.

3. Trial Design

The registry describes a randomized, parallel-group evaluation of Green Dot across high schools in Kentucky. The sponsor is the University of Kentucky. The study status is completed, the phase is listed as NA, and the registry dates are 2009-11 for study start and 2014-05 for primary completion. The listed enrollment is 89707. The record does not report the number of schools or explain whether enrollment refers only to surveyed students or includes another unit.

Schools are the unit emphasized in the study summary: schools receiving Green Dot were to be compared with comparable high schools randomized not to receive Green Dot. The intervention is described as behavioral, and the arms are Green Dot Bystander Training and Control, with the control intervention described as Awareness Education. Masking is listed as none. The record does not explain whether outcome assessors, analysts, or participants were aware of assignment, beyond that masking designation.

The primary endpoint description refers to anonymous survey data collected over 5 years and sorted by intervention. It states that data would be collected at baseline and annually for 4 years, and gives a primary-outcome timeframe of up to 5 years follow up from baseline intervention. The record does not explain the schedule in greater detail or reconcile these time descriptions. The stated intent-to-treat analysis included schools that dropped out, using imputed data. The method used to impute data and the number of schools or observations affected are not reported in the registry record.

This design can support a comparison of assigned groups if the analysis accounts appropriately for the way assignment and outcomes are organized. However, the record does not describe the randomization procedure, whether allocation was stratified or restricted, or the model used beyond stating that baseline and number of students were included as covariates. The registry lists 4 outcome measures and 2 analyses, but does not identify their full scope or explain how the primary result was selected from across outcomes and time points. Detailed design and analysis documentation is not reported in the registry record.

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: 46 demographically similar public high schools were identified as willing participate in the trial. Two schools were excluded by researchers because these were too small (\<100 per grade). Among the remaining 36 schools, two schools within each of the 13 regions were selected and randomized (n=26).

Baseline

Reasons for not completing this period:

Implementation Year 1

Reasons for not completing this period:

Implementation Year 2

Reasons for not completing this period:

Implementation Year 3

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Anonymous survey data collected over 5 years and sorted by intervention. To complete ITT analysis, schools that dropped out were included in analysis using imputed data.

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: Surveys were anonymous. The number of participants analyzed represents the number of analyzable surveys within each group.

6. Primary Endpoint

The registry lists the primary outcome as Average Number of Sexual Assault Events Used (Perpetrated) Per School., measured over up to 5 years follow up from baseline intervention.

The primary endpoint is titled “Average Number of Sexual Assault Events Used (Perpetrated) Per School.” It is based on students' self-reports of sexual assault perpetration, averaged at school level. The registry states that the analysis adjusts for baseline and number of students by including both as covariates in the model. It describes the reported quantity as a least-squares mean and its uncertainty as a 95% confidence interval. The stated timeframe is up to 5 years follow up from baseline intervention.

This endpoint is relevant to the study question because it places the reported outcome at the level of the school, where the intervention was applied and schools were randomized. A school-level average can describe the overall burden of reported events in each school while accounting for differences in baseline values and student counts, as described in the registry. General statistical context: adjusting for baseline can help distinguish a follow-up group contrast from pre-existing differences in measured baseline outcome, while including the number of students can account for differences in the size of the surveyed school population. The registry does not provide the model specification or the exact interpretation of these adjustments.

The registry also classifies the outcome kind as an event rate per unit of exposure, while the primary endpoint description gives an average number of events per school. The exposure unit and any conversion from school average to rate are not reported in the registry record. No numerical group means, adjusted contrast, confidence-interval bounds, or p-value for this endpoint are reported in a single comparison. The endpoint definition therefore identifies what was intended to be measured, but it does not by itself establish the observed group difference.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry identifies an adjusted school-level endpoint and several elements of the intended analysis. It states that student self-reports were averaged at school level and that baseline and number of students were included as covariates. The endpoint is labelled a least-squares mean, with a 95% confidence interval as its dispersion measure. The specific model, assumptions, and numerical estimates are not reported in the registry record. The concepts below explain how such methods generally work; they do not establish which unreported procedures were used.

What does a least-squares mean represent?

General statistical context: a least-squares mean, often described as an adjusted mean, is calculated from a statistical model rather than simply averaging every observed value without adjustment. When covariates are included, the estimate reflects the model's comparison after accounting for those covariates. Here, the registry says that baseline and number of students were included. This can make the group estimates more comparable with respect to those measured characteristics, but it does not guarantee that all differences between schools are removed. It also does not explain whether additional covariates, transformations, or weighting were used. The registry does not provide enough detail to reconstruct the adjusted means or the exact conditions under which they should be interpreted.

Adjustment for baseline is commonly used in follow-up studies because schools may differ in their starting outcome levels. Including baseline in a model can improve the precision of a treatment comparison when baseline is related to follow-up outcome, and can account for measured starting differences. These are general properties, not a claim about the magnitude of any precision gain in this study. The record does not report baseline values by arm, the association between baseline and follow-up, or the model form.

Why does the school level matter?

General statistical context: when participants are grouped in schools, reports from students within the same school may be more alike than reports from students in different schools. If so, treating every student report as statistically independent can understate uncertainty. A school-level analysis or a model that accounts for clustering can address the assignment and outcome structure, depending on the details. In this trial, the endpoint description explicitly says reports were averaged at school level, and the summary says high schools were randomized. However, the record does not state the number of schools, how clustering was handled in the model, or whether any additional dependence structure was used. It is therefore not possible to assess those aspects of the analysis from the registry record.

The stated adjustment for number of students is also relevant because school sizes may differ. General statistical context: a school average based on a smaller number of respondents may have different sampling variability from an average based on a larger number. The way school size is included can affect both the estimate and its uncertainty. The registry says number of students was a covariate, but does not state whether this refers to enrollment, survey respondents, or another count, and does not describe how school averages were weighted. Those details are not reported in the registry record.

How should a confidence interval be read?

General statistical context: a confidence interval is a range of values generated by a procedure intended to have a stated long-run coverage rate under its assumptions. A 95% confidence interval communicates uncertainty about an estimate; it is not a guarantee that the true value lies within a particular reported interval. Its width depends on the amount and variability of data and on the analysis. The registry identifies 95% Confidence Interval as the dispersion measure for the least-squares mean, but does not report interval endpoints for a single primary comparison. A confidence interval for each arm's mean would also not, by itself, replace an interval for the between-arm contrast. The contrast and its uncertainty are needed to assess the primary comparison directly.

What does an event-rate outcome require?

General statistical context: rates relate counts of events to an amount of exposure, such as time or another defined opportunity for events to occur. The denominator matters because equal counts can imply different rates under different exposure amounts. A rate analysis may also need to account for extra variability beyond what a simple count model assumes; this is often called over-dispersion. Allowing for over-dispersion can produce more realistic uncertainty when observations vary more than the model's basic assumptions allow. These are general considerations, not claims about the analysis used in this trial. The registry classifies the outcome as an event rate per unit of exposure, but the primary endpoint is described as an average number of events per school. It does not define the exposure unit, identify a count distribution, or report whether over-dispersion was assessed or addressed.

What can be concluded from the stated analysis description?

The description establishes that the intended primary endpoint was adjusted for baseline and number of students, and that least-squares means and 95% confidence intervals were relevant reporting quantities. It does not give the model equation, link function, standard-error method, covariate values, handling of repeated annual measurements, or the method for combining multiple time points. Those methods are not reported in the registry record. The record also identifies a primary-outcome structure with several categories or time points rather than one reduced comparison. Without the corresponding estimates, the stated method cannot be used to infer a direction or magnitude of effect.

The registry's planning assumptions describe illustrations based on the observed effect, with a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation. These are planning illustrations, not evidence about the study's design target or the probability that the trial detected an effect. The record does not report the trial's own sample-size justification. Any sample-size illustration based on an observed effect is conditional on that effect and on its assumptions; it is not observed power and does not replace the result estimate or its confidence interval.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists 4 outcome measures and 2 analyses. The primary outcome is described as having several categories or time points and is not reduced to one comparison. The record does not identify all outcome measures, specify which comparisons were primary at each time point, or state whether any adjustment was made for multiple comparisons. It also does not report a hierarchy of tests or a multiplicity-control procedure. When several outcomes or time points are examined, the chance of a statistically notable result can depend on how many analyses were conducted and how they were selected. This is general statistical context; without the analysis plan and results, the implications for this study cannot be quantified.

The population description states that schools that dropped out were included in an intent-to-treat analysis using imputed data. This indicates an intention to retain those schools in the analysis rather than omit them solely because they dropped out. However, the registry does not report how many schools dropped out, how much data were missing, which variables were imputed, the imputation method, or whether uncertainty from imputation was incorporated into the analysis. It also does not describe sensitivity analyses comparing alternative assumptions about missing data. These details are not reported in the registry record.

General statistical context: imputation fills in missing values under explicit assumptions, and different assumptions can yield different estimates. An intent-to-treat label does not by itself reveal the missing-data method or establish that all randomized units contributed observed outcome data. The record's statement about imputed data is useful, but does not allow assessment of the plausibility of its assumptions or the robustness of the estimates.

The record does not provide details of the analysis population beyond its statement about schools that dropped out and imputed data. It does not describe exclusions, adherence definitions, or a separate per-protocol analysis. The reported masking designation is none, but the record does not specify how knowledge of assignment may have affected survey collection or other procedures. The full multiplicity strategy, missing-data approach, and analysis-population definitions are therefore not reported in the registry record.

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

The study asks whether school assignment to Green Dot is associated with a different adjusted average number of student-reported sexual assault perpetration events per school than assignment to Awareness Education. The registry defines the intended outcome and identifies baseline and number of students as adjustment variables. It does not report a single numerical estimate for the primary comparison. As a result, the registry record does not establish whether the groups differed, in which direction they differed, or how precisely any difference was estimated.

This absence of a single comparison must not be interpreted as a finding of no effect. A result that is not reported cannot be treated as a zero difference, a non-significant finding, or evidence of equivalence. Conversely, the study hypothesis that Green Dot would be associated with lower rates is not evidence that such a reduction occurred. The group estimates, contrast, interval, and test result needed to evaluate that hypothesis are not reported in the registry record.

The endpoint is school-level and based on anonymous self-reports. It therefore describes an aggregate of reported behavior, not a verified count of all events and not an individual student's personal probability. The registry does not report the proportion of students who responded, how reports were validated, or the exact exposure denominator used for the event-rate classification. Interpretation should remain tied to the measure as described rather than extend to outcomes it does not directly quantify.

The summary's proposed mediation through violence-acceptance norms and bystander behaviors is a hypothesis about possible pathways. The registry does not report mediation estimates or the relevant analytical details, so the proposed pathway cannot be treated as established from this record. Similarly, the completed status indicates that the trial is no longer listed as ongoing, but completion alone says nothing about the direction or size of the outcome contrast.

In general, an estimate and its confidence interval are needed to understand both the direction and precision of a group comparison. A point estimate alone would not convey uncertainty, and a confidence interval alone would still need to be interpreted in relation to the endpoint and analysis. Neither is available as a single primary comparison in the registry record. The record supports a description of the research question, design, and intended analysis, but not a conclusion about the observed primary-outcome difference.

12. Why This Trial Matters Statistically

This trial is statistically notable for the alignment between the intervention setting and the outcome's unit of analysis. The study summary describes randomization of high schools, while the primary endpoint averages anonymous student reports at school level. That makes clear why the number of schools and the handling of school-level variation are important for interpreting uncertainty. The registry does not report the number of schools or the detailed model, so these features cannot be evaluated fully from the record.

The endpoint description also illustrates why outcome definitions need to be read closely. It names an average number of events per school, while the registry classifies the outcome as an event rate per unit of exposure. A rate requires a denominator, and that denominator is not specified in the registry record. This difference in wording does not establish an error, but it leaves an important element of the measure unclear and makes it difficult to interpret a numerical result even if one were reported.

Adjustment for baseline and number of students is explicitly described, allowing readers to see which measured characteristics were included in the intended comparison. At the same time, a label such as least-squares mean does not convey the full model. The missing details about school counts, model structure, repeated time points, uncertainty estimation, and imputation show why a concise registry description cannot replace a complete statistical analysis plan and results report.

The study summary sets out a theory involving sexual-violence reports, social norms, and bystander behavior. Statistically, those are distinct outcome and mediation questions. A primary outcome contrast would address only the defined endpoint; a mediation claim would require additional estimates and assumptions. The registry does not report those results, so the distinction between a stated mechanism and evidence supporting it is essential.

Finally, the record underscores the importance of reporting an interpretable primary contrast. Several categories or time points are acknowledged, but no single comparison is reported. Without a defined estimate and its precision, readers cannot judge the scale or uncertainty of the primary result. Clear reporting would connect the school-level estimand, the analysis, missing-data handling, and the numerical findings.

13. Limitations

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

The main limitation of the registry record for statistical interpretation is the absence of a single numerical primary-outcome comparison. The record does not report adjusted estimates by arm, a between-arm contrast, confidence-interval limits, or a p-value for one primary comparison. It states that the primary outcome has several categories or time points and is not reduced to one comparison. This prevents a direct assessment of direction, magnitude, and precision.

The unit structure is incompletely described. The summary refers to high schools randomized to conditions and the endpoint averages reports at the school level, but the number of schools is not reported in the registry record. The listed enrollment is 89707, but its unit and distribution across schools and arms are not specified. These gaps prevent readers from judging the amount of independent school-level information or the balance of the randomized groups.

The endpoint is described both as an average number of events per school and as an event rate per unit of exposure. The exposure definition and the handling of the annual observations are not reported in the registry record. The model is also not described beyond the covariate adjustment and least-squares mean labels. Details about clustering, uncertainty estimation, model assumptions, and the construction of the adjusted comparison are absent.

Although the record states that imputed data were used for schools that dropped out in the intent-to-treat analysis, it does not identify the imputation method, amount of missingness, or sensitivity analyses. It also does not report a multiplicity strategy despite listing multiple outcomes and analyses. These omissions limit assessment of robustness and selective interpretation. The registry lists zero entries in the adverse-event count categories for both arms, but does not provide further context about ascertainment or reporting. These record-specific limitations mean the design can be described more fully than its results can be evaluated.

14. Related Tutorials

15. Related Statistical Calculators

16. Sources

• ClinicalTrials.gov: Evaluation of Green Dot a Primary Prevention Intervention — NCT01878097.
• Registry record last updated 2019-03-15; status last verified 2019-02.
• Results first posted 2017-08-09.
• 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.

17. Record Summary

The registry describes a completed randomized, parallel evaluation of Green Dot Bystander Training compared with Awareness Education in high schools. It lists no masking, sponsor University of Kentucky, enrollment 89707, and study dates 2009-11 to 2014-05. The primary endpoint is the adjusted average number of student-reported sexual assault perpetration events per school, with adjustment for baseline and number of students and a stated 95% confidence interval. The record says that anonymous surveys were collected over follow-up and that imputed data were used for schools that dropped out in the intent-to-treat analysis.

The primary outcome includes several categories or time points and is not reduced to one numerical comparison. Adjusted group estimates, a contrast, confidence-interval bounds, and a p-value for a single primary comparison are not reported in the registry record. The record therefore supports an account of the design and intended measure, but not a quantitative interpretation of the primary result.