Preventing Squamous Cell Skin Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Preventing Squamous Cell Skin Cancer (NCT02347813), a randomized phase 2 study. The primary outcome is Mean Number of Squamous Cell Carcinomas.

2. Clinical Question

The brief summary explains that the study was proposed to examine tumor counts, tumor histologic grade, and AKR1C3 expression in biopsy specimens. The registry lists mean number of squamous cell carcinomas as the primary outcome and mean number of other skin cancers as a secondary outcome. It does not post results for AKR1C3 expression or histologic grade in the outcome information. The question is therefore best read as a comparison of reported tumor-count summaries, not as a complete assessment of every biological or clinical aim described in the study plan.

3. Trial Design

The registry identifies this as a Phase 2 randomized trial with no masking and a crossover model. The brief summary describes two sequences: observation followed by pioglitazone, and pioglitazone followed by observation. The arm descriptions specify 24 weeks in each period. Tumors were to be observed and treated according to standard care in both periods. The primary outcome timeframe is listed as 1 year, consistent with the two-period schedule described in the record.

A crossover design allows participants to contribute observations under more than one condition, with the sequence determining which condition comes first. That design can make within-participant comparisons useful, but the statistical analysis needs to account for the repeated observations and the order or period in which they occur. The registry record does not report whether the primary comparison used paired observations, a model accounting for period and sequence, or another approach. It also does not report whether any washout interval was used. The brief summary says the study would examine washout effects, but that aim does not establish how washout was handled analytically.

The registry lists enrollment as 12. Separately, the brief summary describes a proposal to enroll 40 participants. The record does not explain the difference between the proposed enrollment and the recorded enrollment. The endpoint summaries list 8 participants for each of the “No Intervention” and “Pioglitazone” labels. The record does not explain how those summary counts relate to the enrollment figure or to the two randomized sequences. They should not be assumed to represent complete crossover pairs.

The trial is marked completed. The sponsor is the University of Rochester. Although the record names randomized allocation, it does not report details of sequence generation, allocation concealment, adherence, or departures from assigned treatment. Those details are 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: Thirteen subjects consented, 1 subject screen failed, 2 subjects from the Delayed Intervention Arm were Early Terminations, 2 subjects from the Immediate Intervention Arm were Early Terminations, resulting in 4 subjects being randomized in to the Delayed Intervention Arm and 4 subjects being randomized in to the Immediate Intervention arm.

First Intervention, 24 Weeks

Reasons for not completing this period:

Second Intervention, 24 Weeks

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 Mean Number of Squamous Cell Carcinomas, measured over 1 year.

The primary endpoint is the mean number of squamous cell carcinomas, with the unit described as the number of squamous cell carcinomas. The registry says these tumors were counted by visual inspection of all skin on the body. The stated timeframe is 1 year. A count is a measure of how many events or findings occurred for each participant over a defined period; reporting the mean summarizes the average count in a group or condition.

Tumor counts are relevant to the study’s stated prevention question because they directly record the occurrence of the cancer type named in the primary outcome. The measure is also clinically interpretable as a count rather than an indirect laboratory marker. However, counts can be unevenly distributed: some participants may have no tumors while others have several. In general, this feature can make means sensitive to a small number of higher counts, and the spread of the data helps readers judge how variable the observations are.

The registry posts a mean and standard deviation for each endpoint label. It does not report participant-level counts, the distribution of counts, or details about how the visual assessments were standardized. It also does not give period-specific outcomes for the crossover sequences. Consequently, the posted summaries describe average counts but do not show the full pattern of tumor occurrence or establish how the endpoint was analyzed across periods.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The statistical information in the registry has three distinct components: descriptive summaries, a reported hypothesis test, and an indication that a recalculated two-arm comparison could not be estimated. Each addresses a different question, and they should not be treated as interchangeable.

What do the means and standard deviations describe?

The mean is the average value of a measured outcome. Here, the outcome is a count of squamous cell carcinomas over the stated timeframe. A mean gives a compact summary, but it does not show whether most participants had similar counts or whether the average reflects a smaller number of higher counts. The standard deviation describes variation around the mean. In general, a larger standard deviation indicates more spread in the observations, though it does not describe the shape of that spread or identify individual values.

The registry posts means of 0.75 and 1.0, with standard deviations of 0.71 and 1.31, for the “No Intervention” and “Pioglitazone” labels, respectively. Each summary lists 8 participants. These descriptive statistics are not themselves a test of a treatment difference. In a crossover study, they also do not show whether the same participants contributed data under both conditions or how their within-participant changes were distributed.

What does the reported ANOVA result mean?

The registry describes the reported method as ANOVA and the parameter as a mean difference in final values. In general, analysis of variance is a family of methods for comparing means. The exact model matters: in a crossover trial, an analysis may need to account for repeated measurements within a participant and for period or sequence effects. The registry record does not describe the model specification, the observations included, or whether those features were handled.

The reported mean difference is 0.50, with p = 0.750. The record identifies a 95% confidence level but does not give the confidence interval limits. A confidence interval, in general, represents a range of estimates compatible with the data and the method’s assumptions; its width conveys precision. Since the limits are not reported, the precision of the stated difference cannot be judged from an interval. A p-value addresses compatibility with a specified null model and assumptions, not the probability that a treatment works or the chance that the finding is due to chance.

Why is no recalculated comparison shown?

The record marks the two-arm continuous comparison as having values missing or not estimable. A recalculated comparison would be an unadjusted calculation from posted summary statistics, rather than the sponsor’s analysis. The record does not support such a calculation for this endpoint, so no recalculated effect or confidence interval should be inferred from the group summaries. The reported means alone are insufficient to reconstruct the crossover analysis, particularly without information about pairing, period-level results, or the covariance between repeated measurements.

What can planning calculations tell readers?

Sample-size calculations based on an observed effect are planning illustrations, not evidence about the study’s achieved power. They ask how many participants per arm might be needed under stated assumptions to detect a chosen fraction of an observed effect, using a specified significance level, allocation, and approximation. Such calculations depend on the validity of the estimated effect and on assumptions about the design. For a crossover study, a planning method based on two independent arms may not capture the within-participant structure. No numerical planning results are included in the registry information, so no target sample size can be interpreted from it.

Why does over-dispersion matter for counts?

In general, count outcomes may show over-dispersion when their variability is greater than a simple count model would expect. If over-dispersion is present but ignored, standard errors and confidence intervals can be misleading, affecting the apparent precision of an estimate. Count models can be used when their assumptions fit the data, but the registry record does not report a count-model analysis or diagnostics for the distribution. The posted means and standard deviations do not by themselves establish whether over-dispersion was present.

Overall, the reported analysis and descriptive summaries answer only part of the statistical question. Missing model details and the absence of a recalculated comparison mean that the comparison’s direction, uncertainty, and handling of the crossover design cannot be fully evaluated from the registry record.

10. Confidence Intervals for Each Arm

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

R code and output

arm <- c("No Intervention", "Pioglitazone")
n   <- c(8, 8)
m   <- c(0.75, 1)
sp  <- c(0.71, 1.31)
sd  <- sp      # posted standard deviations
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  No Intervention  8  0.750  0.710  0.156  1.344
CARM  Pioglitazone  8  1.000  1.310  -0.095  2.095

11. Analyses Reported by the Sponsor

The registry reports an ANOVA with a mean difference in final values of 0.50 and p = 0.750. It identifies a 95% confidence level, but does not report confidence limits. The record does not provide enough detail to reconstruct the calculation, including how the crossover periods or repeated observations were handled. The reported difference also cannot be derived from the posted group summaries alone using the available information.

No simple recalculated comparison is available: the registry marks the two-arm continuous comparison as having values missing or not estimable. This means the descriptive means and standard deviations should not be used to create an alternative inferential result. The sponsor-reported analysis remains the only reported comparison, but its precision and model assumptions cannot be assessed from the record. The record does not flag a discrepancy between the reported analysis and a recalculation; rather, a recalculation is not available. The p-value should be read as a result of the stated analysis, not as proof of equivalence or as a direct measure of clinical importance.

12. Multiplicity, Missing Data and Analysis Populations

The registry lists two outcome measures and identifies one secondary outcome: mean number of other skin cancers. The primary outcome is mean number of squamous cell carcinomas. The brief summary also describes examining AKR1C3 expression and tumor histologic grade, but the outcome information does not give posted results for those aims. The number of analyses listed is 1. The registry record does not explain how multiple outcomes or analyses were considered in the statistical plan.

When a study examines multiple outcomes or performs multiple tests, the chance of at least one apparently notable result can increase. In general, a prespecified hierarchy or multiplicity adjustment can address this issue, but the record does not report whether either was used. It also does not report an analysis plan, interim analyses, or a sample-size justification. The brief summary describes the study as a pilot designed to detect a statistically significant change in tumor numbers, but that statement does not supply the details needed to evaluate the statistical plan.

Missing-data handling is not reported in the registry record. The record does not say whether all randomized participants were analyzed, whether incomplete observations occurred, or how missing outcome values were handled. It does not identify an intention-to-treat, per-protocol, or other analysis population. Because this is a crossover study, information about participants with observations in only one period would be important to interpret the comparison, but such information is not reported.

The registry lists an enrollment of 12, while each primary-outcome label lists 8 participants. It does not explain the relationship between those figures or whether the listed summaries include all randomized participants. The brief summary describes a proposal to enroll 40, but the record does not explain the difference between the planned and recorded enrollment. These gaps limit assessment of completeness and the population represented by the posted analysis.

13. Safety Results

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

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: metastatic squamous cell carcinoma (8.3%).

The registry lists adverse-event groups labelled “No Intervention” and “Pioglitazone,” each with 12 participants at risk for serious adverse events and 12 at risk for other adverse events. It reports 1 serious adverse event in the no-intervention group and none in the pioglitazone group. For other adverse events, it reports none in the no-intervention group and 3 in the pioglitazone group. No deaths are reported in either group.

The record lists “Metastatic squamous cell carcinoma” among the top serious adverse-event terms, with 8.3%. That entry does not, by itself, establish when the event occurred, its relationship to study treatment, or how it was classified relative to the cancer outcome. The adverse-event summaries do not report event narratives, severity beyond the serious-event classification, causality assessments, or period-specific timing. The record also does not clarify how the adverse-event groups map to the crossover sequences.

These figures are descriptive counts from the registry. They should not be interpreted as a comparative safety estimate or as evidence that an event was caused by, or unrelated to, pioglitazone. The adverse-event risk groups also differ in listed size from the primary-outcome summaries, and the record does not explain the relationship.

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

The registry reports average squamous cell carcinoma counts of 0.75 under “No Intervention” and 1.0 under “Pioglitazone,” each based on a listed group size of 8. It also reports an ANOVA mean difference of 0.50 and p = 0.750. Those are the central posted results for the primary outcome, but the record does not give confidence interval limits or enough methodological detail to connect the analysis fully to the crossover design.

The result does not establish that pioglitazone reduced tumor counts, nor does it establish that the two conditions were equivalent. A non-small p-value is not evidence that there is no effect; it can occur when data are imprecise, when an effect is small, or for other reasons. Since the confidence interval is not reported and the comparison cannot be independently recalculated, the range of effects compatible with the data is not clear.

The study is marked Phase 2 and completed, and the brief summary describes it as a pilot. The record’s enrollment is 12, while the posted primary-outcome summaries list 8 participants per label; the relation between these figures is not explained. The brief summary’s proposed enrollment of 40 is not the recorded enrollment. These points limit how confidently the results can be generalized or how precisely they can be interpreted.

The primary outcome measures tumor counts, but the registry record does not provide period-specific results, participant-level patterns, or results for the secondary outcome in the information described. It also does not report the planned analysis population, missing-data approach, or model details. The findings should therefore be understood as limited registry summaries rather than a complete account of the trial’s evidence. They do not support medical decisions for an individual or a broad conclusion about treatment effectiveness.

15. Why This Trial Matters Statistically

This trial is statistically notable because it uses a crossover design for a clinical outcome measured as a count. Crossover studies can make efficient use of within-participant comparisons, since each participant may be observed under more than one condition. That advantage depends on having clear period-specific measurements and an analysis that accounts for repeated observations, sequence, and any relevant carryover. The registry identifies the crossover model but does not report how those design features entered the analysis.

The endpoint is a count of tumors, and count data can vary considerably among participants. The posted standard deviations show variation in both endpoint summaries, while the registry does not give distributions or model diagnostics. In general, these characteristics make it important to assess whether a mean-based analysis adequately represents the data and whether its uncertainty is estimated appropriately.

The record also illustrates why descriptive summaries and inferential results need to be read separately. The group-labelled means and standard deviations describe the posted observations, while the reported ANOVA supplies a test result. The unavailable confidence limits, unexplained analysis structure, and absence of a recalculated comparison prevent a fuller appraisal of precision. The discrepancy between the proposed enrollment in the brief summary and the recorded enrollment is another point requiring clarification rather than speculation.

These limitations do not make the registry information useless. It identifies the outcome, design, treatment sequences, and reported summary statistics, which can guide careful questions about analysis and reporting. The main statistical lesson is that a treatment comparison in a crossover trial requires more than group means: readers need to know how observations were paired and modeled, who was included, and how missing outcomes were handled.

16. Limitations

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

Several limitations are specific to the information reported for this study. First, the registry identifies a crossover design but does not explain how the reported “No Intervention” and “Pioglitazone” outcome summaries relate to the immediate and delayed sequences. It does not provide period-specific means or describe whether the analysis accounted for within-participant pairing, period, sequence, or carryover. That makes it difficult to interpret the reported mean difference as a crossover treatment contrast.

Second, the enrollment and outcome-summary figures are not reconciled. The registry reports enrollment of 12, while each primary-outcome label lists 8 participants. The brief summary describes a proposed enrollment of 40. The record does not explain how these figures relate, whether observations were missing, or which participants contributed to each analysis. The analysis population and missing-data methods are not reported.

Third, the reported mean difference of 0.50 cannot be reconstructed from the posted summaries using the information in the record. The analysis gives p = 0.750 but no confidence interval limits. Model details, assumptions, and diagnostics are not reported. The outcome is a count, but no count-specific model or assessment of dispersion is described.

Finally, the registry lists a secondary outcome and describes additional biological and histologic aims in the brief summary, but results for those aims are not reported in the endpoint information. The record does not describe multiplicity handling, interim analyses, or a statistical analysis plan. These reporting gaps limit interpretation of both the primary comparison and the broader study objectives.

17. Related Tutorials

18. Related Statistical Calculators

19. Sources

• ClinicalTrials.gov: Preventing Squamous Cell Skin Cancer — NCT02347813.
• Registry record last updated 2020-06-23; status last verified 2020-02.
• Results first posted 2020-06-23.
• 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.

20. Record Summary

NCT02347813 is a completed, randomized, unmasked Phase 2 crossover study sponsored by the University of Rochester. It examined pioglitazone in a high-risk skin cancer setting, with an observation period and a treatment period described as 24 weeks each. The registry lists enrollment of 12 and a primary endpoint of mean squamous cell carcinoma count over 1 year, assessed by visual inspection of the skin.

The posted primary summaries list means of 0.75 for “No Intervention” and 1.0 for “Pioglitazone,” with 8 participants in each summary. The reported ANOVA gives a mean difference of 0.50 and p = 0.750; confidence interval limits are not reported. A recalculated comparison is unavailable because the two-arm continuous comparison is marked as missing or not estimable. The record does not explain the relationship between enrollment, outcome-summary counts, and crossover sequences, or report sufficient model details to assess the analysis fully.