Work Should Not Hurt You: Reduction of Hazardous Exposures in Small Businesses Through a Community Health Worker Intervention: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Work Should Not Hurt You: Reduction of Hazardous Exposures in Small Businesses Through a Community Health Worker Intervention (NCT03455530), a randomized na study. The primary outcome is Change in Volatile Organic Chemical Concentrations From Pre- to Post-Intervention.

2. Clinical Question

The trial asks whether an industrial-hygiene-enhanced community health worker intervention is associated with changes in average total volatile organic chemical exposures in small businesses. The focus is on exposure measurement at the business level, rather than a separate outcome for each worker. The registry describes a delayed-intervention group, so the study also compares businesses receiving the intervention earlier with businesses scheduled to receive it later.

The record describes the study as cluster-randomized, with small businesses as clusters. That matters to the question: workers in the same business share a setting, and the primary outcome is constructed at the business level. The registry reports two mixed-model estimates, but the available descriptions do not clearly map each estimate to a named arm comparison. Nor do they state p-values. The question can therefore be framed from the design and endpoint, while the exact contrast and direction of the reported estimates require caution.

3. Trial Design

This was a randomized, parallel-group trial with quadruple masking recorded in the registry. The trial is marked as not applicable for phase, and the reported enrollment is 268. The study status is completed. Its listed sponsor is the University of Arizona. The registry records a start date of 2020-02-21 and a primary completion date of 2024-01-22.

The allocation unit was the small business: the registry explicitly describes a cluster-randomized design with businesses as clusters. The primary outcome's unit of analysis was the average total volatile organic chemical exposure for each business at each analysis. There were 1-4 workers at each shop per assessment. Consequently, the enrollment count should not be read as the number of independent business-level observations; the number of businesses is not reported in the registry record.

One group was scheduled to receive the industrial-hygiene-enhanced community health worker intervention before the other group. The delayed-intervention group was scheduled to receive it after the first group, with the registry describing the timing as approximately three months. The primary outcome timeframe is 6 months. The record does not give enough detail to reconstruct every assessment date or explain exactly how the 6-month timeframe relates to the intervention sequence.

The project description centers on hazardous chemical exposures in small businesses and the use of community health workers alongside industrial hygiene. It describes barriers to occupational health and pollution-prevention information for workers in these settings, and presents the intervention as a way to connect small businesses with relevant stakeholders. Those statements describe the motivation and intended approach; they do not establish that exposure changed.

Details about the randomization procedure, the number of clusters assigned to each group, allocation concealment, masking roles, and the handling of the delayed intervention beyond its approximate timing are not reported in the registry record. The record also does not report a protocol or a detailed statistical analysis plan. Such omissions limit the extent to which the design and analysis can be independently reconstructed.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: This is a cluster-randomized trial where the clusters were small businesses. There were 1-4 workers at each shop per assessment. The unit of analysis here is the average total volatile organic chemicals exposures for each business at each analysis.

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: This is a cluster-randomized trial where the clusters were businesses. 1-4 workers participated per business.

6. Primary Endpoint

The registry lists the primary outcome as Change in Volatile Organic Chemical Concentrations From Pre- to Post-Intervention, measured over 6 months.

The primary endpoint is the change in average total volatile organic chemical exposure concentrations for each business from pre- to post-intervention. Concentrations are expressed in ppb, and the registry identifies the outcome parameter as a geometric mean with a 95% confidence interval. The stated timeframe is 6 months. The business-level measure is consistent with the cluster design: each business, rather than each individual worker, contributes the unit of analysis described for the outcome.

General statistical context: concentration measurements can be strongly right-skewed, with some observations much higher than others. A geometric mean is often used for positive, skewed measurements because it summarizes values on a multiplicative scale. Its interpretation is not the same as an arithmetic average. The record does not describe transformations, treatment of values below detection limits, or the measurement procedures, so those details should not be assumed.

The endpoint is relevant to the stated research aim because it measures workplace chemical exposure rather than a downstream health outcome. A change in measured exposure can help characterize whether the workplace environment differed across assessments, but it is not itself a measure of illness, injury, or long-term health. The registry's reported analyses are labeled as mean differences in final values, while the primary endpoint description emphasizes pre-to-post change. The record does not explain the relationship between that endpoint definition and the final-value estimates, so the two descriptions should not be treated as interchangeable without further information.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry identifies the primary endpoint as a business-level geometric mean of total volatile organic chemical concentrations, expressed in ppb, with a 95% confidence interval. It describes the target as change from pre- to post-intervention. The reported analyses are mixed models, with estimates labeled mean difference in final values. These labels convey the general approach and reported result type, but the record does not include enough model detail to reproduce or fully interpret the analysis.

Why account for the cluster design?

General statistical context: in a cluster-randomized study, assignment occurs to groups rather than independently to each person. People or measurements within one cluster may resemble one another because they share a workplace or other environment. Treating every observation as independent can understate uncertainty and distort inference. Here, the registry identifies small businesses as clusters and says the primary outcome's unit of analysis is the business's average exposure. It also says a simple individual-level re-analysis would be misleading, so only the sponsor's own analysis is shown.

A mixed model can represent more than one source of variation, including variation among businesses and repeated measurements within a business, depending on its specification. General statistical context: the model's usefulness depends on the variables included, the structure used for correlated observations, and assumptions about the data. The registry does not report those details. It is therefore possible to identify the method family, but not to evaluate the model specification or reconstruct the fitted analysis from the record.

What does the geometric mean convey?

General statistical context: the geometric mean is formed on a multiplicative scale and is often used for positive measurements whose distribution is skewed. It can be less influenced by very large values than an arithmetic mean, but it answers a different descriptive question. The record names a geometric mean for the primary endpoint but labels the reported analysis parameter as a mean difference in final values. It does not state whether the model operates on the original concentration scale or a transformed scale, nor whether the reported estimates are differences, ratios, or another contrast. It would be inappropriate to relabel the estimates as ratios or to infer their direction from their values alone.

How should the confidence intervals be read?

General statistical context: a confidence interval communicates sampling uncertainty around an estimate, conditional on the analysis method and its assumptions. It does not describe the range of individual participants' measurements, and it does not guarantee that the true value lies within the particular interval. The reported intervals are 95% confidence intervals. One estimate is 0.73, with limits 0.45 and 1.19; the other is 1.34, with limits 0.55 and 3.24. The latter interval is wider, which indicates less precision on the reported scale. Since the scale and comparison are not fully described, interval width can be discussed, but substantive direction should not be assigned.

What can be concluded from the analysis labels?

The registry calls both results mixed-model analyses and identifies them as superiority analyses. These labels do not give the full statistical plan. The record does not report the model formula, adjustment variables, how repeated measurements were handled, the number of businesses analyzed, or how missing measurements were treated. It does not report a p-value for either estimate. A missing p-value should not be reconstructed from the confidence limits, particularly when the estimate's scale and method are not fully specified.

General statistical context: multiplicity arises when multiple outcomes, contrasts, or analyses create more than one opportunity to find an apparently notable result. The registry lists one primary outcome and two reported analyses, but it does not explain whether the analyses were prespecified, how the two estimates relate to the randomized groups, or whether any multiplicity procedure was planned. No adjustment or unadjusted status can therefore be assumed. Similarly, the record does not describe interim analyses, a sample-size justification, or the prespecified approach to missing data.

The registry's stated design is the reason not to replace the sponsor analysis with a simple calculation based on individual-level summaries. General statistical context: when clusters are the assignment units, an analysis that ignores the cluster structure can give misleading uncertainty estimates. The sponsor's mixed-model results are therefore the available inferential results in the record. Their limitations are not a reason to invent a different analysis; they are a reason to describe the reported results narrowly and note which methodological details are not reported.

Finally, the endpoint description and analysis description are not fully aligned in wording: one emphasizes pre-to-post change, while the other specifies final values. The record does not report baseline and follow-up summaries that would allow the change to be independently checked, nor does it state whether the final-value model adjusts for baseline. This distinction matters because an endpoint defined as change and a model comparing final values are not automatically the same analysis. The registry does not provide enough information to resolve the distinction.

10. Analyses Reported by the Sponsor

The two posted estimates are the sponsor's mixed-model analyses: 0.73 with a 95% confidence interval from 0.45 to 1.19, and 1.34 with a 95% confidence interval from 0.55 to 3.24. The record does not report p-values, and it does not clearly map each estimate to a named intervention comparison. Both are labeled mean differences in final values, but the primary outcome description refers to pre-to-post change. The registry does not explain how those descriptions correspond.

No simple re-calculation is presented. The record explicitly describes a cluster-randomized design with small businesses as clusters and warns that a simple individual-level re-analysis would be misleading. Without the underlying cluster-level data and fuller model details, a summary-statistic calculation would not reproduce the sponsor's analysis or provide a sound substitute for it. The sponsor's model-based estimates are therefore the available results, but their interpretation remains limited by the unclear contrast and scale.

General statistical context: a re-calculation from posted summary statistics, when appropriate, can serve as a check on a reported result, but it cannot replace a prespecified analysis that accounts for the design. In this trial, the cluster structure is central. There is no recalculated estimate to compare with the sponsor's results, so there is no discrepancy between a sponsor analysis and a recalculation to adjudicate. The appropriate reading is to retain the sponsor's figures as reported and make the unresolved reporting details explicit.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary outcome and two reported analyses. It does not state whether the two estimates represent separate planned contrasts, different populations, or another distinction. Their group descriptions are not rendered as named comparisons in the analysis results, so the precise relationship between the estimates is not clear. No multiplicity adjustment or rationale for treating the analyses as a family is reported in the registry record. This leaves uncertainty about how to interpret the two results together and whether any confirmatory inference was planned across them.

Missing-data methods are not reported in the registry record. The record does not say how businesses or measurements with incomplete follow-up were handled, whether any observations were excluded, or whether the mixed model used all available measurements under a stated assumption. The number of businesses contributing to each analysis is also not reported. These details matter because the primary outcome is measured at the business level and follow-up is part of the stated endpoint.

Analysis populations are not described beyond the cluster-level outcome definition and the reported enrollment. The record does not say whether analyses followed an intention-to-treat approach, whether all randomized businesses were included, or whether any per-protocol or other population was used. Randomization is reported, but the allocation procedure and the analyzed cluster counts are not given.

Details about interim analyses, stopping rules, and the statistical analysis plan are not reported in the registry record. Nor does the record describe a sample-size justification tied to the primary endpoint. Those omissions do not establish that such procedures were absent; they mean their existence and role cannot be determined from the registry information. The sample-size assumptions stated for a planning illustration do not document the trial's own design justification, and no numerical planning result is reported in the material available for interpretation.

General statistical context: multiplicity, missingness, and analysis-population choices can affect the precision and meaning of trial estimates. Without these details, the registered confidence intervals should be read as the results of the reported mixed models, not as a complete account of all design and analysis decisions. The registry does not allow an assessment of whether alternative reasonable choices would materially change the estimates.

12. Safety Results

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

Deaths

Other (non-serious) adverse events

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

The registry's adverse-event group descriptions cover the intervention and delayed-intervention groups for beauty salons and auto shops. The reported number at risk is 77 for the intervention beauty-salon group, 70 for the delayed-intervention beauty-salon group, 69 for the intervention auto-shop group, and 52 for the delayed-intervention auto-shop group. Each of these four rows reports zero serious adverse events and zero deaths. Each also reports zero other adverse events.

These are the recorded counts in the registry, not proof that no harm occurred. The record does not provide definitions of serious or other adverse events, the surveillance period, collection procedures, or the completeness of reporting. It also does not explain how these category-specific groups relate to the overall enrollment. The figures should therefore be read as reported event counts for the listed groups, rather than as a comprehensive measure of safety.

General statistical context: when no events are recorded, the observed count alone does not establish that the underlying event risk is zero. The precision of a safety estimate depends on the amount of observation and how events were collected, among other factors. The registry provides no interval estimates for these counts. The available information supports a factual statement about what was recorded, but not a broader conclusion about safety.

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

The registry records a randomized comparison of an industrial-hygiene-enhanced community health worker intervention delivered to small businesses, with a delayed-intervention group. Its primary endpoint is business-level average total volatile organic chemical exposure, described as change from pre- to post-intervention over a 6-month timeframe. The study therefore addresses workplace exposure measurement, not a direct clinical health outcome.

The posted results include two mixed-model estimates and their 95% confidence intervals. The registry does not report p-values, clearly identify the arm comparison associated with each estimate, or explain the scale and direction convention for the reported mean differences in final values. It also does not reconcile the final-value analysis wording with the pre-to-post change endpoint. For those reasons, the estimates cannot be translated into a clear statement that exposure rose or fell in a particular group.

General statistical context: an estimate is not a complete result without knowing the contrast it represents, the scale on which it is expressed, and the analysis assumptions behind its uncertainty interval. An interval expresses uncertainty about an estimate, not the range of individual workers' exposures and not a guarantee about the true effect. The wider of the two reported intervals indicates less precision on the stated scale, but does not by itself determine the practical importance of the result.

The cluster-randomized design also shapes what the results mean. Businesses were the clusters, and the outcome was summarized at the business level. The sponsor used mixed models, while the registry cautions that a simple individual-level re-analysis would be misleading. The reported analysis should therefore be understood as the available model-based result, not as a collection of independent worker-level comparisons.

The registry's safety figures record zero serious adverse events and zero deaths across the listed group categories. Those counts do not demonstrate that risk was absent, especially because event definitions and ascertainment details are not reported. Nor do the exposure estimates establish effects on asthma, cancer, cardiovascular disease, neurological disease, or any other health outcome. The project description discusses those health concerns as motivation; they are not the primary endpoint reported in the results.

In practical terms, the record supports describing the study design, endpoint, model family, estimates, confidence intervals, and recorded event counts. It does not support a firm directional or causal interpretation of the numerical exposure estimates without clearer comparison and scale information. Additional outcome detail, group mapping, and methods reporting would be needed for a more complete interpretation.

14. Why This Trial Matters Statistically

This trial is statistically notable because randomization occurred at the business level, while workers were measured within those businesses. Such designs are common when an intervention is delivered through a workplace or community setting, but they require analysis that respects the clustering. The registry explicitly identifies small businesses as the clusters and warns that a simple individual-level re-analysis would be misleading. This makes the unit of assignment and the unit of analysis essential parts of interpreting the result.

The primary endpoint also illustrates why outcome definition and analysis measure need to be read carefully. The endpoint description concerns change in a business-level geometric mean from pre- to post-intervention, while the posted analyses are labeled mean differences in final values. These descriptions may reflect choices made in the analysis, but the record does not explain their relationship. Statistical interpretation depends on knowing whether estimates are changes, final-value contrasts, or another model-based quantity.

The reported confidence intervals communicate differing levels of precision, but the registry leaves the estimate scale and comparison unclear. This is a useful reminder that numerical estimates alone are not self-interpreting. A reader needs the contrast, outcome scale, model specification, and analysis population to connect an estimate to a study question.

The trial also addresses measurement in small businesses, a setting that the project description says can involve hazardous solvents and barriers to occupational health information. That context explains why a business-level exposure endpoint was selected. It does not turn exposure measurement into evidence of a health benefit, and the reported record should be read within the boundaries of its stated outcome.

Finally, the posted result demonstrates the value of complete registry reporting. Clear naming of comparisons, explicit scales and direction, and descriptions of missing-data and multiplicity procedures would make the estimates more interpretable. The available results can still be reported accurately, but the unanswered methodological questions constrain how far the statistical narrative can go.

15. Limitations

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

Several limitations arise specifically from the information recorded for this trial. The most important interpretive limitation is that the two mixed-model estimates are not clearly tied to named arm comparisons in the result descriptions. The record lists estimates of 0.73 and 1.34, but a reader cannot confidently determine which comparison each represents from the accompanying plain-language descriptions.

The analysis parameter is labeled mean difference in final values, while the primary endpoint is described as change from pre- to post-intervention. The record does not report baseline and follow-up summaries, explain whether baseline was included in the model, or show how the final-value estimates answer the stated change question. It also does not clearly define the scale or sign convention of the reported estimates. These gaps prevent a reliable interpretation of direction and magnitude.

The study is cluster-randomized, but the number of randomized businesses and the number analyzed are not reported. Enrollment is reported as 268, and the registry describes 1-4 workers at each shop per assessment; neither detail supplies the cluster count. Model specification, adjustment variables, covariance structure, and handling of repeated measurements are not reported. These details would be needed to assess the analysis and its uncertainty more fully.

The record does not report p-values, missing-data procedures, analysis populations, multiplicity procedures, a sample-size justification, or interim analyses. The absence of those details in the registry record does not establish what was or was not done, but it limits independent evaluation. The adverse-event counts are also reported without event definitions, ascertainment methods, or confidence intervals.

Finally, the primary endpoint is exposure concentration rather than a clinical health outcome. The project description refers to possible health consequences of hazardous chemicals, but the reported endpoint cannot establish changes in those outcomes. These limitations call for a cautious account of the results rather than a broader conclusion about health effects or intervention effectiveness.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Work Should Not Hurt You: Reduction of Hazardous Exposures in Small Businesses Through a Community Health Worker Intervention — NCT03455530.
• Registry record last updated 2025-07-10; status last verified 2024-03.
• Results first posted 2025-07-10.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

NCT03455530 is a completed, randomized, parallel-group, quadruple-masked cluster trial sponsored by the University of Arizona. It reports enrollment of 268 and evaluates an industrial-hygiene-enhanced community health worker intervention in small businesses, with a delayed-intervention group. The primary outcome is the business-level change in average total volatile organic chemical exposure, expressed in ppb, with a 6-month timeframe.

The registry reports two mixed-model estimates labeled mean differences in final values: 0.73 (95% confidence interval 0.45 to 1.19) and 1.34 (95% confidence interval 0.55 to 3.24). P-values are not reported, and the record does not clearly map the estimates to named comparisons or reconcile the final-value analysis with the change endpoint. The safety listings report zero serious adverse events and zero deaths in each of four group categories. Important model, missing-data, multiplicity, and analysis-population details are not reported in the registry record.