This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the results posted in the ClinicalTrials.gov record; calculations added by Clinical Biostats are labelled as such.
1. Trial at a Glance
This page covers the registry results for A Study to Evaluate the Clinical Performance of VeriSee GLC for Screening of Glaucoma Suspect Using Color Fundus Images (NCT07550712), a registered phase not stated study. The primary outcome is Sensitivity and Specificity of the VeriSee GLC for Screening Glaucoma Suspects.
| Feature | Registry value |
|---|---|
| ClinicalTrials.gov identifier | NCT07550712 |
| Phase | Phase not stated |
| Condition | Glaucoma |
| Allocation | Not Stated |
| Design model | Not Stated |
| Masking | Not Stated |
| Enrollment | 773 |
| Arms in the primary outcome | n/a |
| Lead sponsor | Acer Medical Inc. |
| Results posted | Yes |
2. Clinical Question
Population
Adult participants whose previously collected color fundus images were used in this retrospective observational study. The registry reports 773 included cases and says 24 were excluded from the primary analysis because of insufficient image quality or insufficient clinical information.
Intervention
Automated image analysis with VeriSee GLC. The system evaluates fundus images for glaucomatous signs, including retinal nerve fiber layer defects, vertical cup-to-disc ratio enlargement, and rim-to-disc ratio changes. Its stated role is to assist identification of glaucoma suspects for referral, rather than provide a definitive diagnosis.
Comparator
The reference standard is the majority consensus of three independent evaluators. The registry does not report details about their qualifications, instructions, masking, or how disagreements were resolved beyond the majority-consensus definition.
Primary question
How sensitive and specific is VeriSee GLC for identifying glaucoma suspects from color fundus images when its classifications are compared with the evaluator consensus? The registry specifies a 95% confidence interval for these measures, but the estimates and intervals are not reported in the registry record.
This is a diagnostic-performance question, not a comparison of treatments or a test of whether the system changes health outcomes. Sensitivity concerns identification of cases designated as glaucoma suspects by the reference standard; specificity concerns correct classification of cases not designated as suspects by that standard. The reference standard therefore defines the labels against which the automated screening output is assessed. Its consensus process is part of the interpretation: the performance estimates would describe agreement with that reference, not agreement with an independently established biological truth. The registry also lists secondary measures involving predictive values, accuracy, detection of glaucomatous signs, and disc and cup contours. Their results are not reported in the registry record.
3. Trial Design
This is a completed observational study, not a randomized trial. The registry describes it as retrospective and says it uses previously collected fundus images. VeriSee GLC analyzes color fundus images, and its screening output is compared with a reference standard based on the majority consensus of three independent evaluators. The intended use described in the record is to help healthcare professionals identify glaucoma suspects for referral; it is not described as a definitive diagnostic process.
The registry lists enrollment of 773 and reports that 24 cases were excluded from the primary analysis because image quality or clinical information was insufficient. It does not report the resulting analysis denominator, the number classified as suspects by the evaluators, or the distribution of classifications. The primary outcome timeframe is listed as 1 Day. Study status is completed, and both the listed start date and primary completion date are 2024-08-26.
There is no randomized allocation or treatment-arm comparison described. The registry does not report a control arm, recruitment process, study setting, image-acquisition procedures, or the clinical characteristics of the included participants beyond their being adults. It also does not report whether the evaluators were masked to the automated output, or whether VeriSee GLC operators had access to the reference assessments. These details matter when assessing possible influences on classification, but they cannot be inferred from the retrospective design alone.
The registry identifies Acer Medical Inc. as the sponsor. It lists five outcome measures, including the primary sensitivity and specificity measure and secondary measures of predictive value, accuracy, sign detection, and contour similarity. No analysis results are reported in the registry record. Consequently, the design description establishes what was evaluated and against which reference, but does not establish the observed level or precision of diagnostic performance.
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: \[Not Specified\]
Overall Study
| Stage | Test Group |
|---|---|
| STARTED | 773 |
| COMPLETED | 749 |
| NOT COMPLETED | 24 |
Reasons for not completing this period:
| Reason | Test Group |
|---|---|
| Insufficient image quality | 6 |
| Insufficient information | 18 |
Analysis population for the primary outcome, as defined in the registry: Of the 773 included cases, 24 were excluded from the analysis due to insufficient image quality or insufficient clinical information.
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: Intended use population
| Characteristic | Unit | Test Group |
|---|---|---|
| Participants analysed | 773 | |
| Age, Continuous | years | 60 (SD 14.7) |
| Sex: Female, Male: Female | Participants | 466 |
| Sex: Female, Male: Male | Participants | 307 |
| Race and Ethnicity Not Collected | Participants | — |
| Spherical Equivalent | Diopter (D) | -2.7 (SD 3.1) |
6. Primary Endpoint
The registry lists the primary outcome as Sensitivity and Specificity of the VeriSee GLC for Screening Glaucoma Suspects, measured over 1 Day.
| Element | Registry entry |
|---|---|
| Result type | Number |
| Measure of spread | 95% Confidence Interval |
| Unit | percent |
| Endpoint classification | not reduced to a single comparison |
The primary endpoint is sensitivity and specificity of VeriSee GLC for identifying glaucoma suspects from color fundus images, compared with the majority consensus of three independent evaluators. The registry specifies that results are to be expressed as a number with a 95% confidence interval and lists the timeframe as 1 Day. The actual estimates, confidence limits, classification counts, and analysis denominator are not reported in the registry record.
These measures address different aspects of screening performance. Sensitivity describes how often the system classifies an evaluator-designated suspect as a suspect. Specificity describes how often the system classifies an evaluator-designated non-suspect as a non-suspect. Reporting both helps distinguish missed reference-positive classifications from false-positive classifications relative to the chosen reference standard. Neither measure alone describes the probability that a particular person classified as positive is truly a suspect; that question is addressed by positive predictive value and depends in part on the frequency of suspects in the population being screened.
The endpoint is relevant to a screening purpose because the stated aim is to identify glaucoma suspects for referral, rather than make a definitive diagnosis. However, diagnostic performance against the evaluator consensus is not itself evidence about referral decisions, subsequent clinical outcomes, or the benefit of using the system in practice. Those questions are not reported as primary outcomes in the registry record.
7. Why No Re-analysis Is Shown
8. Statistical Interpretation
The registry record describes the evaluation question and planned measures, but it does not report sensitivity, specificity, confidence intervals, or the underlying classification counts. There is therefore no posted diagnostic estimate to interpret for the primary outcome. The fact that the study is marked completed does not establish the direction, magnitude, or precision of its findings.
In general, sensitivity and specificity are complementary measures: sensitivity concerns reference-positive classifications, while specificity concerns reference-negative classifications. A screening system may perform differently on these two dimensions. A sensitivity estimate without specificity, or the reverse, would give an incomplete picture of classification performance. Here, neither estimate is reported in the registry record, so it is not possible to determine whether the system more often missed evaluator-designated suspects or classified evaluator-designated non-suspects as suspects.
The 95% confidence interval specified for the primary outcome is intended to communicate uncertainty around an estimate. In general, a confidence interval gives a range of values compatible with the observed data under the statistical procedure used; it is not a guarantee that the true value lies in that particular interval. Its width depends on factors including the amount of information and the number of relevant reference-positive and reference-negative observations. The interval limits are not reported, and the registry also does not provide the denominators needed to assess how much information supports either measure.
Interpretation is also conditional on the reference standard. The primary comparison is with the majority consensus of three independent evaluators, not an independently verified definitive diagnosis. Any disagreement, systematic classification tendencies, or limitations in the evaluator process could affect the apparent performance relative to that reference. The registry does not report enough detail to assess those possibilities. A result against this consensus would describe agreement with the registry-defined reference, rather than prove that the automated system identifies disease without error.
The 24 exclusions for insufficient image quality or clinical information are relevant to the population represented by an analysis. The registry does not report the characteristics or classifications of excluded cases, or the number remaining in the primary analysis. Thus the effect of these exclusions on the estimates cannot be assessed. If image quality or availability of clinical information is related to classification difficulty, estimates among analyzed images may not represent performance in all images encountered in routine screening; whether that applies is not reported.
Secondary outcomes include positive predictive value, negative predictive value, accuracy, accuracy for specified glaucomatous signs, and Dice coefficients for disc and cup contours. Their results are not reported in the registry record. Without those results, it is not possible to assess the broader profile of screening classifications or contour agreement. Nor does the registry report evidence about clinical impact, referral consequences, or performance in settings beyond the retrospective image evaluation. The record supports describing the study question and design, but not a conclusion that the system does or does not perform well.
9. Statistical Methods Explained
What do sensitivity and specificity measure?
General statistical explanation: Diagnostic performance is commonly described by comparing a test classification with a reference classification. Sensitivity is the proportion of reference-positive observations classified positive by the test. Specificity is the proportion of reference-negative observations classified negative by the test. The two measures use different denominators, so each needs its own estimate and uncertainty interval. For this study, the reference classification is the majority consensus of three independent evaluators, and the test classification is generated by VeriSee GLC from color fundus images.
Neither measure is interchangeable with overall accuracy. Accuracy combines correct positive and correct negative classifications relative to all evaluated observations, and can be affected by the balance of reference-positive and reference-negative cases. The registry lists accuracy among secondary outcomes, but does not report its value or the underlying classification counts. It is therefore not possible to determine how accuracy relates to sensitivity and specificity in this study.
What does the confidence interval add?
General statistical explanation: An estimate describes the observed data, while a confidence interval conveys its statistical uncertainty under a specified method. A narrow interval indicates greater precision than a wide interval, all else being equal. For diagnostic proportions, precision is influenced by the number of observations in the relevant reference category. The registry specifies 95% confidence intervals for the primary endpoint but does not report interval limits, methods for constructing them, or category-specific denominators. The precision of the primary estimates therefore cannot be evaluated from the registry record.
A confidence interval is not a measure of clinical importance and does not correct for a biased reference standard, selective exclusions, or differences between the evaluated images and other settings. Its interpretation depends on the data and statistical procedure used. The registry record does not provide enough detail to assess the interval method or any assumptions applied to the primary analysis.
How should the evaluator consensus be understood?
General statistical explanation: A reference standard supplies the classifications against which a test is evaluated. When the standard is a consensus, the result is agreement with that consensus definition. Consensus may provide a practical way to create reference labels, but it does not by itself show that all evaluators would classify every image identically or that the labels correspond to a definitive diagnosis. The registry identifies majority consensus among three independent evaluators, but does not report individual evaluator agreement, adjudication procedures, evaluator expertise, or masking. Those omissions limit assessment of the reproducibility and independence of the reference process.
What do the secondary measures contribute?
General statistical explanation: Positive predictive value is the proportion of test-positive classifications that are reference-positive; negative predictive value is the proportion of test-negative classifications that are reference-negative. Unlike sensitivity and specificity, predictive values depend on how common the reference-positive classification is in the evaluated population. Consequently, predictive values from a selected image set may not transfer directly to a population with a different frequency of glaucoma suspects. The registry lists PPV and NPV, but does not report estimates or the relevant population distribution.
Dice coefficients are used to quantify overlap between segmented regions. In this study, the registry lists a Dice coefficient for disc and cup contours. This measure concerns contour similarity rather than the same classification question addressed by sensitivity and specificity. Its interpretation would depend on the reference contours and analytic procedures, which are not described in detail in the registry record. Results for this outcome are also not reported.
What can planning calculations establish?
General statistical explanation: Sample-size calculations depend on design assumptions and a target effect or precision. The planning illustration associated with this analysis is described as using the observed effect, a two-sided significance level of 0.05, equal allocation between two arms, a normal approximation, and effects expressed as a stated fraction of the observed effect. The record does not report that these assumptions were used to design this observational diagnostic evaluation. The illustration is therefore not evidence about the study's original design, and it cannot be interpreted as observed power. No numerical planning results are reported in the registry record.
The equal-allocation, two-arm assumptions do not describe the registry's stated diagnostic comparison, which evaluates an automated screening output against a consensus reference standard. Planning calculations based on those assumptions should not be treated as a substitute for the study's actual analysis population, classification counts, confidence intervals, or statistical plan. The registry does not report its sample-size justification or a detailed statistical analysis plan.
What is the role of recalculation?
General statistical explanation: A recalculation from posted summary statistics can sometimes check an unadjusted result, but it cannot recreate an analysis that depends on unavailable individual-level data, adjustment variables, or unreported procedures. The registry record does not report primary outcome summary statistics from which a recalculation could be made. No independent numerical estimate can therefore be derived from the information reported in the registry record.
10. Multiplicity, Missing Data and Analysis Populations
The registry lists a primary outcome and several secondary outcomes, but does not report whether or how multiplicity was addressed. Multiplicity arises when several outcomes, measures, or comparisons are assessed: the chance of at least one apparently noteworthy result can increase as the number of statistical tests increases. General approaches include prespecifying a primary measure, adjusting significance thresholds, or treating secondary findings as exploratory. The registry record does not state which approach, if any, was used. No inferential results are reported, so the effect of any multiplicity procedure cannot be evaluated.
Missing-data handling is also not reported. The primary outcome description says 24 of the 773 included cases were excluded because image quality or clinical information was insufficient. The record does not describe whether those exclusions were handled before analysis, whether other data were incomplete, or whether any imputation or sensitivity analysis was used. It does not report the analysis denominator or the characteristics of excluded cases. The extent to which exclusions could affect diagnostic estimates therefore remains unclear.
The analysis populations are not fully specified beyond the stated exclusion. The registry does not say whether every available image was analyzed, whether multiple images could come from a participant, or how any such clustering was handled. It does not report the number of reference-positive and reference-negative cases, the handling of indeterminate assessments, or whether additional exclusions were applied. These details are important for understanding denominators and uncertainty in diagnostic measures.
The record also does not report interim analyses, a statistical analysis plan, or procedures for managing discordance between evaluators beyond the majority-consensus definition. None of these procedures should be assumed from the study's completed status or its listed endpoints.
11. What the Result Does, and Does Not, Mean
The study asks whether VeriSee GLC classifications from color fundus images correspond to a reference standard based on the majority consensus of three independent evaluators. That is a diagnostic-performance question in a retrospective observational setting. It is not a randomized comparison of treatments, and it does not test whether use of the system improves health outcomes.
The registry describes the system as a screening aid intended to help healthcare professionals identify glaucoma suspects for referral, not as a definitive diagnosis. Accordingly, even diagnostic performance estimates, if reported, would need to be interpreted in relation to that screening role and the evaluator-defined reference. Sensitivity and specificity would characterize classification relative to that reference; they would not by themselves establish the accuracy of a diagnosis in every clinical context.
The registry record does not report the primary sensitivity and specificity estimates, their 95% confidence intervals, or the classification counts. Secondary outcome results are not reported either. No conclusion about the observed performance, the balance of false-positive and false-negative classifications, or the precision of estimates can be drawn from the registry information available.
The record reports exclusions for insufficient image quality or clinical information, but does not report their characteristics or the size and composition of the final analysis population. This limits assessment of how broadly any eventual estimates would apply. The registry also does not report external validation, real-world referral outcomes, or effects on clinical decision-making. These are distinct questions from agreement with a consensus reference standard and should not be inferred from the study description.
In general, a screening result is one component of an assessment process; it is not equivalent to a definitive clinical conclusion. The registry's stated purpose is assistance with identifying suspects for referral. The reported design and endpoints alone do not establish clinical benefit, and the absence of posted results means even the specified diagnostic-performance measures cannot be judged.
12. Why This Trial Matters Statistically
Diagnostic studies require clear definitions of both the test and the reference standard. This registry identifies VeriSee GLC as an automated image-analysis system and defines its comparator as a majority consensus among three independent evaluators. Making that comparison explicit matters because sensitivity and specificity are conditional on the reference labels: they describe test performance relative to the chosen standard, not in isolation.
The endpoint also illustrates why reporting precision and denominators is essential. Sensitivity and specificity are based on different reference categories, so a single enrollment figure does not show how much information supports either estimate. Confidence intervals would help readers judge uncertainty, while category-specific counts would help interpret their precision. Although the registry specifies 95% confidence intervals, it does not report estimates, intervals, or classification counts.
The study is framed as screening support rather than definitive diagnosis. That distinction helps keep statistical performance separate from claims about clinical outcomes. Measures such as predictive values can be influenced by the frequency of suspects in the evaluated population, while accuracy can conceal differing performance across reference categories. The listed secondary measures address complementary aspects of classification and image analysis, but their results are not reported in the registry record.
The stated image-quality and clinical-information exclusions also highlight the importance of describing analysis populations. Readers need to know which observations contribute to each estimate and whether excluded images differ in ways related to classification. The registry gives a reason for the specified exclusions, but does not report the final analysis denominator or excluded-case characteristics. These reporting gaps constrain interpretation even when the study purpose and endpoint are clear.
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 registry record does not report diagnostic results, confidence interval limits, classification counts, or the final analysis denominator. This prevents assessment of both performance and precision. The enrollment figure alone cannot establish how many reference-positive or reference-negative cases contributed to the primary measures.
The reference standard is defined as majority consensus among three independent evaluators, but procedures that could clarify its reliability are not reported. These include evaluator qualifications, masking, individual agreement, instructions for assessing images, and handling of ties or indeterminate judgments. The consensus may be a practical comparator, but performance against it should not be conflated with agreement against a definitive diagnosis.
The registry reports that 24 cases were excluded for insufficient image quality or clinical information. It does not describe the excluded cases, report whether other missingness occurred, or explain how exclusions were incorporated into each outcome analysis. If excluded images differ systematically in interpretability, estimates among analyzed cases may not reflect the full set of images intended for screening; the registry does not provide data to evaluate that possibility.
The retrospective use of previously collected images may limit what can be inferred about performance in other image-acquisition conditions or clinical settings. The record does not report recruitment details, sites, camera or acquisition procedures, participant characteristics beyond adulthood, or external validation. It also does not report whether the study assessed downstream referral decisions or clinical outcomes. These are limitations of the available record, and their practical importance cannot be quantified from the reported information.
Finally, the registry does not report the detailed statistical analysis plan, multiplicity handling, sample-size justification, or methods for constructing the specified confidence intervals. These omissions prevent evaluation of analytic choices that could affect interpretation. They do not establish that a particular method was or was not used.
14. Related Tutorials
15. Related Statistical Calculators
16. Sources
- ClinicalTrials.gov: A Study to Evaluate the Clinical Performance of VeriSee GLC for Screening of Glaucoma Suspect Using Color Fundus Images — NCT07550712.
- Registry record last updated 2026-06-25; status last verified 2026-04.
- Results first posted 2026-06-25.
- 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.
Continue through the Clinical Biostats statistical pathway
Use this trial's endpoint as a starting point for deeper study of effect measures, confidence intervals, and sample-size planning.
17. Record Summary
NCT07550712 is a completed retrospective observational study sponsored by Acer Medical Inc. It evaluates VeriSee GLC screening for glaucoma suspects using previously collected color fundus images from adults. The registry lists enrollment of 773 and reports that 24 cases were excluded from the primary analysis because of insufficient image quality or clinical information.
The primary outcome is sensitivity and specificity compared with the majority consensus of three independent evaluators, with a specified 95% confidence interval and a timeframe of 1 Day. Secondary outcomes include predictive values, accuracy, detection of glaucomatous signs, and Dice coefficients for disc and cup contours. The registry record does not report outcome estimates or confidence intervals.
The listed start date and primary completion date are 2024-08-26. Detailed evaluator procedures, the final analysis denominator, missing-data methods, multiplicity handling, and a statistical analysis plan are not reported in the registry record.