Study Evaluating the Safety and Efficacy of AR-15512 (COMET-3): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Study Evaluating the Safety and Efficacy of AR-15512 (COMET-3) (NCT05360966), a randomized phase 3 study. The primary outcome is Percentage of Subjects Who Achieved Equal to or Greater Than 10 Millimeter Increase From Pre-drop at Baseline to Post-drop on Day 14 in Study Eye Unanesthetized Schirmer Score.

2. Clinical Question

The endpoint is a threshold response in tear production, not a direct measure of every aspect of dry eye symptoms. The registry describes the Schirmer test as filter paper placed on the lower eyelid, with wetting recorded on a scale from 0 millimeters to 35 millimeters. An increased score is designated a positive outcome. The comparison therefore concerns the proportion of subjects reaching the specified increase, rather than the average change in score.

The study was randomized, double-masked, parallel-group, and vehicle-controlled. Those design features frame the comparison, but the registry record does not report enough detail to reconstruct the randomization procedure, masking procedures, or all analysis decisions. The endpoint uses one study eye per subject, as specified in the record. It should not be read as an outcome measured independently in both eyes.

3. Trial Design

COMET-3, registered as NCT05360966, was a Phase 3 randomized, double-masked, parallel-group, vehicle-controlled study of 0.003% AR-15512 for dry eye disease. Aerie Pharmaceuticals is listed as the sponsor. The record lists enrollment of 467 and a completed study. The listed start date is 2022-07-18 and the primary completion date is 2023-10-17.

The two groups received either 0.003% AR-15512 ophthalmic solution or AR-15512 vehicle ophthalmic solution. Both regimens specified one drop in each eye twice a day for 90 days. The registry summary describes a multicenter study at approximately 20 sites in the United States. The planned visit schedule included screening at Day -14, baseline at Day 1, and follow-up visits at Day 7, Day 14, Day 28, and Day 90, identified as study exit. The summary states an individual participation duration of approximately 15 weeks.

The primary endpoint analysis is described as involving the intention-to-treat population, comprising all randomized subjects, and using multiple imputation methodology. The posted arm summaries list 232 for 0.003% AR-15512 and 234 for AR-15512 Vehicle. These are endpoint-analysis group figures and should not be substituted for the registry’s enrollment figure. The record does not explain the relationship between the enrollment count and the endpoint-analysis figures.

Some protocol and design details are not reported in the registry record. The original sample-size justification, the randomization method, the precise masking arrangements, and the trial’s prespecified statistical analysis plan are not reported in the registry record. The listed planning illustrations are based on the observed effect and are not evidence of the assumptions used when this trial was designed.

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: This reporting group includes all randomized participants.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: ITT population comprising all randomized subjects. This analysis uses multiple imputation methodology.

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: ITT Population

6. Primary Endpoint

The registry lists the primary outcome as Percentage of Subjects Who Achieved Equal to or Greater Than 10 Millimeter Increase From Pre-drop at Baseline to Post-drop on Day 14 in Study Eye Unanesthetized Schirmer Score, measured over Baseline (Day 1) pre-drop; Day 14 post-drop.

The primary endpoint is the percentage of subjects whose study-eye unanesthetized Schirmer score increased by equal to or greater than 10 millimeters from the pre-drop baseline on Day 1 to the post-drop measurement on Day 14. The registry describes the Schirmer test as a measurement of tear production: filter paper is placed on the lower eyelid, and the amount of wetting is recorded from 0 millimeters, indicating no tear production, to 35 millimeters, the maximum. An increased score is designated a positive outcome.

This is a binary, threshold-based endpoint. Each subject is classified according to whether the specified increase was achieved, and the group result is the percentage meeting that criterion. It therefore answers a different question from a comparison of average score changes. The threshold may make the result easy to interpret as a proportion reaching a defined response, while not conveying how far beyond the threshold individual scores may fall or how much change occurred among those who did not reach it.

The endpoint is measured in one eye, designated the study eye, rather than treating both eyes as separate observations. The registry identifies the primary analysis population as all randomized subjects and says that multiple imputation methodology was used. The record does not report the detailed imputation model or the number of missing Day 14 measurements.

7. Results Posted in the Registry

• 0.003% AR-15512: 53.0% (123/232) versus 14.5% (34/234) on control; risk difference 38.5 percentage points (95% CI 30.6 to 46.3), risk ratio 3.65 (2.61–5.09); unadjusted two-proportion 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.

Least Squares Mean Change From Baseline in Global Symptom Assessment iN Dry Eye (SANDE) Score on Day 28

• 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:
  • 0.003% AR-15512: mean -22.30 versus -19.10 on control; difference -3.20 (95% CI -7.77 to 1.37), standardized effect d = -0.13; z-test p = 0.170.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between 0.003% AR-15512 and the control. The interval gives the range of effects that remain compatible with the data.

Least Squares Mean Change From Pre-drop Baseline Unanesthetized Schirmer Score on Post-drop Day 14 (Study Eye)

• 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:
  • 0.003% AR-15512: mean 10.90 versus 2.60 on control; difference 8.30 (95% CI 6.65 to 9.95), standardized effect d = 0.91; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for 0.003% AR-15512 than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Percentage of Subjects Who Achieved Equal to or Greater Than 10 Millimeter Increase From Pre-drop at Baseline to Post-drop on Day 1 in Study Eye Unanesthetized Schirmer Score

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • 0.003% AR-15512: 53.9% (125/232) versus 13.7% (32/234) on control; risk difference 40.2 percentage points (95% CI 32.4 to 48.0), risk ratio 3.94 (2.80–5.55); unadjusted two-proportion z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for 0.003% AR-15512 than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The listed secondary outcomes span symptom scores and tear-production measures, including change in SANDE scores and Schirmer measurements at different study visits. Taken together, the posted results include a pattern of larger Schirmer changes or response percentages in the AR-15512 group on the specified measures. The symptom-score comparison at Day 28 is less conclusive in the reported analysis: its confidence interval includes zero and its p-value is 0.132. This combination suggests that the posted secondary outcomes do not all convey the same degree of statistical evidence.

These endpoints relate to the primary result by examining other dimensions or time points. The primary endpoint asks whether a specified Day 14 increase threshold was reached; continuous Schirmer changes address average change, while SANDE outcomes concern symptom scores. Agreement across some tear-production measures can provide context for the primary finding, but endpoints are not interchangeable. A result on one measure does not establish the same result on another.

Secondary analyses generally carry less confirmatory weight when there are multiple outcomes or time points and the testing strategy is unknown. Testing multiple outcomes can increase the chance of at least one apparently favorable finding by chance. Multiplicity adjustment, a prespecified hierarchy, and the full analysis plan are not reported in the registry record. The reported secondary p-values should therefore be read as exploratory evidence unless the trial’s prespecified approach establishes otherwise.

The registry record lists 11 posted secondary outcomes, while the complete intended outcome set and analysis plan are not described in enough detail to assess selective reporting. The secondary findings can help characterize the trial’s results, but they do not independently resolve questions about clinical importance, missing data, or how well the results generalize beyond this study.

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

• Least Squares Mean Change From Baseline in Global Symptom Assessment iN Dry Eye (SANDE) Score on Day 28
• Least Squares Mean Change From Pre-drop Baseline Unanesthetized Schirmer Score on Post-drop Day 14 (Study Eye)
• Percentage of Subjects Who Achieved Equal to or Greater Than 10 Millimeter Increase From Pre-drop at Baseline to Post-drop on Day 1 in Study Eye Unanesthetized Schirmer Score
• Least Squares Mean Change From Pre-drop Baseline in Unanesthetized Schirmer Score on Post-drop Day 1 (Study Eye)
• Percentage of Subjects Who Achieved Equal to or Greater Than 10 Millimeter Increase From Pre-drop at Baseline to Post-drop on Day 90 in Study Eye Unanesthetized Schirmer Score
• Least Squares Mean Change From Pre-drop Baseline in Unanesthetized Schirmer Score on Post-drop Day 90 (Study Eye)
• Least Squares Mean Change From Baseline in Global SANDE Score on Day 90
• Least Squares Mean Change From Baseline in SANDE Frequency Score on Day 90

9. Statistical Interpretation

10. Statistical Methods Explained

The primary endpoint is binary: a subject either met the specified increase threshold or did not. A percentage summarizes the share of subjects in a group meeting that criterion. Comparing the group percentages answers how frequently the threshold response was recorded in one group relative to the other. It does not retain the full range of score changes that would be used in a continuous-outcome analysis.

The registry reports Pearson's Chi-squared test for the primary comparison and labels its parameter as a difference in percentages. In general, a Chi-squared test assesses whether the pattern of categorical outcomes is compatible with no association between group and response, under the assumptions of the test. The reported p-value is evidence against that null model; it is not the probability that the null hypothesis is true, nor does it measure the size or importance of the response. The reported difference and confidence interval carry information about magnitude and precision that the p-value alone does not.

The recalculated risk difference is based on the posted response percentages and group figures. In general, a risk difference expresses the contrast between two proportions on the proportion scale. A value above zero in this comparison indicates a greater response proportion in the AR-15512 group. The recalculation is unadjusted and should be distinguished from the registry’s reported test and estimate. It is not a replacement for the sponsor’s analysis.

For continuous secondary outcomes, the registry reports least squares mean changes and standard errors, with ANCOVA as the reported method for the listed continuous analyses. In general, an ANCOVA estimates a group contrast while accounting for model covariates. The registry record does not specify the covariates or model details, so it is not possible to assess those choices from the summary alone. A standard error describes the sampling uncertainty of an estimate under the fitted model; it is not the same quantity as the standard deviation of individual observations.

Several recalculations use an approximate pooled standard deviation derived from standard errors and group sizes. Such an approximation can help express a continuous contrast on a standardized scale, but it is not the same as access to participant-level data or the original fitted model. The reported ANCOVA estimates and the simple unadjusted recalculations may differ because they represent different analyses. No discrepancy is identified in the registry record.

The intention-to-treat population is described as comprising all randomized subjects. In general, intention-to-treat analysis retains subjects according to randomized group, regardless of later treatment adherence, preserving the comparison created by randomization. The registry says the primary analysis used multiple imputation. In general, multiple imputation replaces missing values through repeated plausible imputations based on an analysis model, with uncertainty across those imputations incorporated into inference. The registry does not report the model, variables, assumptions, or sensitivity analyses, so those features cannot be evaluated from the posted summary.

Confidence intervals express a range of estimates compatible with the data and the statistical procedure under its assumptions. A 95% confidence interval is a procedure that, across repeated comparable studies, would cover the target parameter at the stated rate when its assumptions hold. It is not a guarantee that the particular interval contains the true value, and it does not account for uncertainties outside the model or design. Intervals that are narrower indicate greater statistical precision for that estimate, while still not resolving possible bias or limitations in measurement.

Multiplicity arises when several outcomes, time points, or contrasts are tested. In general, conducting many tests increases the chance that at least one result will appear statistically notable by chance, unless the analysis plan addresses that issue. The registry lists multiple secondary outcomes, but multiplicity adjustment or a testing hierarchy is not reported in the registry record. Consequently, secondary p-values should be interpreted more cautiously than a single prespecified primary comparison, particularly when the full testing plan is unavailable.

Over-dispersion is relevant to some count or proportion models when observed variability exceeds what a simple model assumes. In general, ignoring over-dispersion can make uncertainty estimates too small. The primary test is reported as Pearson's Chi-squared test, and the registry record does not describe an over-dispersion assessment. It would not be appropriate to infer that such an assessment was or was not performed from the test label alone.

11. Confidence Intervals and Precision

The reported primary difference in percentages is 38.8, with a 95% confidence interval from 30.8 to 46.8. This interval describes the statistical precision of the estimated group contrast under the sponsor’s reported analysis. It is centered on the reported estimate in the units used by the registry. The interval excludes zero, in line with the reported p < 0.001 for the test of no difference.

The unadjusted recalculation gives a risk difference of 0.385, with a 95% confidence interval from 0.306 to 0.463. This interval is based on the posted arm summaries and the simple recalculation, rather than the sponsor’s reported Chi-squared analysis. It is important not to treat the two intervals as interchangeable: the reported estimate is in percentage units, while the recalculated contrast is on the proportion scale.

In general, a confidence interval conveys a range of values reasonably compatible with observed data and the method’s assumptions. It is not a range containing every plausible outcome for an individual, nor does it capture uncertainties from unreported analysis choices, outcome measurement, or study conduct. A statistically precise estimate can still be limited in applicability, and precision does not establish clinical importance.

The registry record does not report the detailed uncertainty calculation for the primary sponsor analysis beyond its method, interval, and p-value. Nor does it provide the imputation model or missing-data sensitivity results. Those omissions limit independent assessment of how robust the reported precision is to analysis assumptions. The interval should therefore be read as conditional on the reported analysis rather than as a complete account of every source of uncertainty.

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 planning figures are participants per arm, not total participants and not total events. They are illustrations based on the observed effect, rather than the trial’s original design assumptions. The listed assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. Each planning row uses the stated fraction of the observed effect as the effect to detect. The trial’s own original sample-size assumptions are not reported in the registry record.

The 80% power column gives the illustrative number of participants per arm for a design that would have an 80% chance of detecting the specified effect under the planning assumptions, if that effect were the true effect and the assumptions held. The 90% power column expresses the corresponding planning target for a 90% chance. These are design calculations, not a statement about the probability that the observed result is true or that the trial has succeeded.

Rows planning for a smaller fraction of the observed effect require more participants per arm in the displayed illustrations. This reflects a general feature of power calculations: smaller differences are harder to distinguish from random variation, so larger planned samples are needed when other assumptions are held fixed. The numbers are illustrative because the true effect in a future study may differ from the observed estimate, and practical design considerations may differ from the simple assumptions used.

These values are not observed power. Observed power is not a useful substitute for interpreting the estimate and its confidence interval after a study has been completed. The planning figures are best read as scenarios about future study design, not as evidence about the original trial’s planning process.

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, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.145; treatment proportion p₁ = 0.530; average p̄ = 0.338.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.6688 = 1.3109.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.6110 = 0.5142.
4. n = (1.3109 + 0.5142)² / (0.3849)² = 22.5, rounded up to 23 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
p0 <- 0.145299145299145                          # control proportion
p1 <- 0.530172413793103                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^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 (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(5360966)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 23 (80% power), 30 (90% power)
 75% of observed effect: n per arm = 38 (80% power), 50 (90% power)
 50% of observed effect: n per arm = 77 (80% power), 102 (90% power)
power.prop.test: 22.5 
Simulated power at n = 23 per arm: 0.817 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

14. Analyses Reported by the Sponsor

The registry reports the sponsor’s primary analysis as a Pearson's Chi-squared test, with a difference in percentages of 38.8, a 95% confidence interval from 30.8 to 46.8, and p < 0.001. The posted group percentages are 53.2 for 0.003% AR-15512 and 14.4 for AR-15512 Vehicle. The registry’s reported estimate and inference are the results to cite as the sponsor’s analysis.

The unadjusted recalculation from the posted group summaries is expressed as a risk difference of 0.385, with a 95% confidence interval from 0.306 to 0.463 and p < 0.001. This is a separate calculation from summary values and is not a reconstruction of the sponsor’s full analysis. The registry record does not identify a discrepancy between the reported and recalculated results. Their units and analysis presentations differ, so the estimates should not be treated as exact duplicates.

The recalculation is useful as a transparent descriptive check on the direction and scale of the group contrast, but it does not replace the registry’s reported analysis. Participant-level data and the complete analysis specifications are not reported in the registry record, so the sponsor’s result cannot be independently reproduced in full from these summaries alone.

15. Multiplicity, Missing Data and Analysis Populations

The primary population is described as all randomized subjects, and the primary analysis is stated to use multiple imputation. This identifies the broad analysis population and missing-data approach, but it does not provide enough information to assess the assumptions behind the imputation or their effect on the estimate. The imputation model, variables included, number of imputations, and sensitivity analyses are not reported in the registry record.

The registry record does not report the number of subjects with missing primary endpoint measurements, the reasons for missingness, or the amount of missing data in each group. Without those details, readers cannot judge whether missingness was similar between groups or how strongly the results may depend on the imputation assumptions. Multiple imputation can address missing outcomes under specified assumptions; it does not by itself ensure that missing-data bias has been eliminated.

The record lists several secondary outcomes measured at different visits and does not report a multiplicity adjustment or a hierarchical testing procedure. In general, as the number of statistical comparisons increases, so does the chance of obtaining at least one apparently notable result by chance. The absence of a reported multiplicity strategy means secondary findings should be interpreted with caution and should not be treated as if each were the sole prespecified test.

The registry record does not report the full statistical analysis plan, detailed model specifications, or whether interim analyses were planned or conducted. These omissions do not establish that such procedures were absent; they mean they cannot be assessed from the registry record. The primary result is therefore best read alongside the stated analysis method, interval, and limitations in reporting.

16. Safety Results

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

Deaths

Other (non-serious) adverse events

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

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: acute myocardial infarction (0.4%), angina unstable (0.4%), covid-19 (0.4%), ankle fracture (0.4%), prostate cancer (0.4%).

The adverse-event record separates ocular and nonocular categories and lists serious and other adverse events. For serious adverse events, the record reports 0 in the pretreatment category, 0 in the 0.003% AR-15512 ocular category, 3 in the 0.003% AR-15512 nonocular category, 0 in the AR-15512 Vehicle ocular category, and 6 in the AR-15512 Vehicle nonocular category. Deaths are listed as 0 in each category. The denominators at risk are displayed separately for each category in the registry record.

The listed serious-event terms each have a reported frequency of 0.4%: acute myocardial infarction, angina unstable, COVID-19, ankle fracture, prostate cancer, and squamous cell carcinoma. These entries describe events recorded during the study; they do not by themselves establish whether an event was caused by study treatment. The registry record does not provide event-level timing, clinical context, or a causal assessment in the information.

Serious events are distinct from other adverse events, and event counts should not be interpreted as a complete measure of comparative safety. Differences in exposure, follow-up, reporting, and event definitions can affect counts. The registry record’s safety figures are descriptive and do not establish that either regimen is safe or unsafe.

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

The primary finding is a difference in the proportion of subjects reaching the registry-defined Day 14 Schirmer response threshold. The posted response percentages are 53.2 for 0.003% AR-15512 and 14.4 for AR-15512 Vehicle. The sponsor’s reported comparison gives a difference in percentages of 38.8, with a 95% confidence interval from 30.8 to 46.8 and p < 0.001. The data therefore show a higher observed response proportion in the AR-15512 group for this endpoint in this study.

That conclusion is specific to a threshold on unanesthetized Schirmer score in one study eye, measured from pre-drop baseline to post-drop Day 14. It does not establish a particular change in symptoms, a result for both eyes considered separately, or an outcome beyond the measured follow-up. Secondary symptom and tear-production measures address related but distinct questions, and their results vary in strength.

A small p-value does not measure the size of a clinical benefit, the probability that a finding is true, or the likelihood of an individual response. The confidence interval describes statistical uncertainty under the reported analysis, not uncertainty from every possible design or measurement limitation. The registry does not report a threshold for clinical importance, so statistical evidence should not be equated with established clinical importance.

The trial’s randomized, double-masked design supports a controlled comparison, but the registry record does not provide all details needed to evaluate conduct, missingness, or analysis choices. The primary analysis used multiple imputation, and its assumptions and sensitivity analyses are not reported. The result should therefore be understood as the estimate reported for this trial, rather than a guarantee of the result in other populations, settings, or durations.

These statistics do not provide medical advice or a basis for an individual treatment decision. They describe what the registry reports for the study outcomes and the limits of what can be concluded from those summaries.

18. Why This Trial Matters Statistically

This trial illustrates why endpoint definition matters. A threshold response, an average change in a score, and a symptom measure each summarize a different aspect of an outcome. The primary result concerns the percentage reaching a specified increase in study-eye Schirmer score; it cannot be substituted for the continuous score changes or symptom outcomes. Statistical interpretation begins with identifying exactly what was measured and when.

It also demonstrates why estimates should be read with their uncertainty. The sponsor reports a difference in percentages and a confidence interval, while an unadjusted recalculation from posted summaries uses a risk-difference scale. Each representation requires attention to its method and units. A p-value alone would not show the magnitude or precision of the observed contrast.

The registry describes an intention-to-treat primary population and multiple imputation, but does not report the imputation assumptions or missingness details. These are important because a missing-data method can influence the estimated comparison. Likewise, the record lists multiple secondary outcomes without describing a multiplicity strategy. This matters when distinguishing a primary confirmatory result from secondary patterns that may be more exploratory.

Finally, the planning illustrations clarify the difference between a trial’s observed result and assumptions for a future study. Their figures are based on the observed effect and stated planning assumptions, not the original design justification and not observed power. Keeping those concepts separate helps prevent retrospective calculations from being mistaken for evidence about how the trial was designed.

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.
• Participant counts were reconstructed from posted percentages and arm sizes, so small rounding differences are possible.

The registry record does not report the full statistical analysis plan, detailed randomization procedure, masking implementation, or original sample-size justification. Without these details, readers cannot fully evaluate the prespecified analysis decisions or compare them with the posted results. The planning illustrations based on the observed effect do not fill that gap.

The primary endpoint uses a threshold response in one study eye. Dichotomizing a continuous measurement can discard information about the size of changes on either side of the threshold, and the record does not report the clinical rationale or validation for the specified threshold. The endpoint description explains the measurement, but does not establish how well the threshold corresponds to meaningful symptom improvement.

The primary analysis uses multiple imputation, but the model, assumptions, missing-data counts, and sensitivity analyses are not reported in the registry record. It is therefore not possible to assess from the summary whether the result is robust to alternative missing-data assumptions. Details of the model covariates for reported ANCOVA analyses are also not reported.

The study summary describes a multicenter US study and a participation duration of approximately 15 weeks, with follow-up through Day 90. The record does not establish how well the participants represent other populations, or what might occur beyond the study follow-up. Multiple secondary measures are listed, but a multiplicity plan is not reported. These limits counsel against interpreting secondary results as definitive or extrapolating the primary finding beyond its defined endpoint and study context.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Study Evaluating the Safety and Efficacy of AR-15512 (COMET-3) — NCT05360966.
• Registry record last updated 2025-07-23; status last verified 2025-07.
• Results first posted 2025-07-23.
• 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

NCT05360966, COMET-3, is a completed Phase 3 randomized, double-masked, parallel-group, vehicle-controlled study of 0.003% AR-15512 in dry eye disease. The registry lists enrollment of 467 and Aerie Pharmaceuticals as sponsor. The two regimens were one drop in each eye twice a day for 90 days.

The primary endpoint was the percentage achieving equal to or greater than 10 millimeters increase in study-eye unanesthetized Schirmer score from Day 1 pre-drop to Day 14 post-drop. The posted response percentages were 53.2 for AR-15512 and 14.4 for vehicle. The reported difference in percentages was 38.8, with a 95% confidence interval from 30.8 to 46.8 and p < 0.001.

The primary analysis is described as intention-to-treat with multiple imputation. Details of the imputation assumptions, multiplicity strategy, and original sample-size justification are not reported in the registry record. Secondary outcomes and adverse-event counts provide additional context but do not alter the endpoint-specific nature of the primary result.