Anti-VEGF Treatment for Prevention of PDR/DME: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Anti-VEGF Treatment for Prevention of PDR/DME (NCT02634333), a randomized phase 3 study. The primary outcome is Development of PDR and/or DME (Whichever Came First).

2. Clinical Question

The clinical question is about prevention or delay of a vision-threatening complication in eyes considered at high risk, rather than treatment of a complication already present. The registry summary frames the uncertainty as whether earlier, less frequent anti-VEGF dosing would be associated with favorable anatomic outcomes and whether those outcomes would be accompanied by favorable visual-acuity outcomes. Those are related but distinct questions: an eye may meet an anatomic disease criterion without a corresponding change in a visual-acuity measure.

The study compared an aflibercept strategy initiated promptly with an observation strategy using sham initially and allowing deferred treatment under stated conditions. It therefore does not describe a simple contrast between receiving treatment and never receiving treatment. The possibility of deferred treatment is part of the comparator strategy and matters when interpreting the contrast. The record does not report how often deferred treatment was used.

The primary outcome combines first development of PDR and/or center-involved DME. The registry separately describes development of PDR and development of DME, but the primary title identifies the combined first-event outcome. The question is framed at the level of eyes. Because a participant could have two study eyes, eye-level counts should not automatically be read as participant-level counts or independent observations.

3. Trial Design

The registry describes a phase 3, randomized, double-masked, parallel-group trial sponsored by Jaeb Center for Health Research. The record lists enrollment of 399 and a completed status. The start is recorded as 2016-01 and primary completion as 2020-05. These dates describe the trial record and do not, by themselves, establish the follow-up available for each outcome.

The two named study strategies were prompt aflibercept and observation with prompt sham. The aflibercept schedule was at randomization and visits at 1, 2, and 4 months, then every 4 months thereafter. In the observation arm, sham was given on that schedule. The descriptions allow deferred aflibercept if center-involved DME or PDR develops, and describe the possible later addition of deferred laser. The record does not report how frequently those additional treatments occurred.

The primary endpoint is reported in eyes, not participants. Its population description says a participant could have two study eyes. For participants with two study eyes, the record describes random assignment of sham and aflibercept across the eyes, with assignment of which eye received which strategy linked to which eye had worse diabetic retinopathy. This means that some observations may be paired within a participant rather than arising from separate participants. The registry record does not report the statistical handling of this feature in an analysis of the primary endpoint.

The registry lists change in visual acuity from baseline and development of PDR and/or DME as secondary outcomes. The visual-acuity outcome is reported over 4 years, whereas the primary endpoint is framed over 2 years. The reported endpoint summaries therefore use different follow-up timeframes and measure different aspects of disease. The registry record does not report the protocol’s sample-size justification, its planned statistical analysis for the primary endpoint, or whether any interim analyses were conducted.

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: Units are in eyes. A participant could have 2 study eyes. For participants with two study eyes the participant was randomly assigned to receive sham in the eye with worse diabetic retinopathy and aflibercept in the other, or aflibercept in the eye with the worse diabetic retinopathy and sham in the other.

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: Units are in eyes. A participant could have 2 study eyes. For participants with two study eyes the participant was randomly assigned to receive sham in the eye with worse diabetic retinopathy and aflibercept in the other, or aflibercept in the eye with the worse diabetic retinopathy and sham in the other.

6. Primary Endpoint

The registry lists the primary outcome as Development of PDR and/or DME (Whichever Came First), measured over 2 years.

The primary endpoint counts eyes meeting criteria for first development of PDR and/or center-involved DME during the stated 2-year timeframe. These are complications the trial summary identifies as vision-threatening, and the endpoint is intended to capture whether either of the specified disease criteria is met. The registry defines the combined outcome so that an eye meeting any criterion is counted under the first PDR and/or DME criteria met, and is then censored from contributing to the next criterion. An eye not meeting the outcome is censored at its last completed visit.

This approach makes the endpoint a count of eyes meeting an event definition, rather than a mean change in a continuous measure such as visual-acuity letter score. The registry also gives descriptions for development of PDR and development of DME separately, regardless of whether the other condition developed. Those descriptions distinguish the combined first-event endpoint from the individual disease outcomes.

Such an endpoint is used to focus on the onset of specified disease complications rather than on a surrogate alone or on visual acuity alone. General statistical context: time-to-event methods can account for participants or eyes whose observed follow-up ends before the full outcome window, but their interpretation depends on the censoring rules and analysis method. The registry record does not report a time-to-event estimate, uncertainty interval, or inferential comparison for the primary endpoint. The posted arm values are counts, so they do not by themselves communicate the timing of events or the precision of a treatment contrast.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

General statistical context helps distinguish the different kinds of quantities in the registry record. A count records how many observations met a stated definition. A mean summarizes the central value of a quantitative outcome, while a standard deviation describes variability among observations. Neither kind of summary, by itself, establishes whether an observed arm difference is precise or compatible with a range of underlying effects. Precision is usually communicated with an uncertainty interval and an analysis that reflects the trial design.

What can be learned from the primary counts?

The primary endpoint is reported as a number of eyes meeting a disease criterion. The aflibercept and sham values are 27 and 75, respectively. A count can be useful as a direct description of the posted events, but counts alone are not a complete comparison when the amount of follow-up, censoring, or the number of eyes contributing to each analysis may differ. Even if arm sizes are displayed alongside counts, dividing values or treating all observations as having equal follow-up would create a new calculation that is not the registry’s reported analysis.

The endpoint description includes censoring: eyes that meet a criterion no longer contribute to the next criterion, while eyes without the outcome are censored at the last completed visit. General statistical context: censoring means that the event status is not observed for the full period in the same way for every eye. A valid comparison may need to account for event timing and the follow-up process, rather than treat all eyes as if they had identical observation time. The registry record does not report the primary analysis method, its estimate, or its uncertainty, so the posted counts cannot be used to reconstruct that analysis.

Why does the eye-level structure matter?

The outcome unit is eyes, and the registry notes that a participant could have two study eyes. For some participants, the two eyes were assigned opposite strategies. General statistical context: measurements from two eyes of one participant may be correlated because they share participant-level characteristics. If that dependence is ignored, a statistical analysis can misstate the uncertainty around a contrast. Appropriate approaches can account for clustering or paired observations, but which approach was used for the primary outcome is not reported in the registry record. The posted counts should therefore remain descriptions of eye-level outcomes, not be recast as independent participant outcomes.

How should the visual-acuity comparison be read?

The continuous secondary outcome is change in visual acuity from baseline, expressed in letter score. The registry reports group means and standard deviations. A mean difference represents a contrast between average changes, while a confidence interval expresses a range of estimates compatible with the data and the assumptions of the calculation. The recalculated unadjusted mean difference is -0.3, and its 95% confidence interval is -1.766 to 1.166. Because the interval crosses zero, the posted summary does not establish a clear direction of the unadjusted mean contrast. This does not prove that the arms are identical; the interval also reflects uncertainty.

The p-value for this recalculated comparison is 0.688. General statistical context: a p-value describes how unusual data at least as incompatible with a null hypothesis would be under that hypothesis and the analysis assumptions. It is not the probability that the null hypothesis is true, nor a measure of clinical importance. The value is unadjusted, and the registry record lists no reported analyses for the secondary outcome. It should not be treated as an adjusted trial conclusion.

What do the planning illustrations mean?

The planning illustrations use the observed visual-acuity effect as a starting point and show participant counts per arm for specified fractions of that effect and target power levels. They use a two-sided significance level of 0.05, equal allocation, and a normal approximation. The illustrations are not the trial’s original design calculation, and they are not observed power. Observed power is not an informative replacement for estimating uncertainty from the observed results; the planning values instead indicate how sample-size requirements change under the stated hypothetical effect assumptions. The registry record does not report the trial’s own sample-size justification, so the illustrations cannot be used to infer the original design rationale.

The illustration for 100% of the observed effect lists participants per arm for 80% power as 6,627 and for 90% power as 8,872. For 75% of the observed effect, the corresponding figures are 11,781 and 15,772; for 50% of the observed effect, they are 26,507 and 35,486. These are planning quantities, not observed enrollment, event counts, or evidence that the trial had a particular power for its primary endpoint. They are based on the secondary continuous outcome and should not be transferred to the eye-level count endpoint.

Why does over-dispersion matter in general?

General statistical context: count outcomes are sometimes analyzed with models that assume a particular relationship between the mean and variability. Over-dispersion occurs when observed variability is greater than that assumption allows. If present and unaddressed, it can lead to uncertainty estimates that are too narrow. The registry record does not report a count-model analysis or enough detail to assess whether over-dispersion was relevant to the primary outcome. It would therefore be inappropriate to claim that a particular count model was used or that a particular variance assumption was satisfied.

In sum, the posted evidence contains descriptive primary counts and a recalculated unadjusted comparison for visual-acuity change. The primary inferential analysis, handling of eye-level dependence, and methods for missing or censored observations are not reported in the registry record. Those absences limit what can be concluded about statistical precision, even when the descriptive values are clear.

10. Multiplicity, Missing Data and Analysis Populations

The registry record does not report a multiplicity plan, such as a prespecified hierarchy among outcomes or an adjustment for testing multiple hypotheses. It lists multiple secondary outcomes and describes several disease-related outcomes, but it does not state how the probability of false-positive findings across those analyses was controlled. General statistical context: when many outcomes or comparisons are examined, an unadjusted p-value for an individual result does not account for the full set of analyses. This is one reason unadjusted secondary results generally provide weaker confirmatory evidence than a prespecified primary analysis.

The record also does not report the primary statistical analysis plan, the analysis population for each endpoint, or whether analyses followed intention-to-treat principles. It does not state how missing visual-acuity assessments were handled, how incomplete follow-up was treated beyond the primary endpoint’s censoring description, or whether sensitivity analyses were performed. For the primary outcome, eyes without the event are censored at the last completed visit; the record does not explain how that censoring was incorporated into an inferential comparison.

These omissions matter because missingness and censoring can affect both the available information and the interpretation of an estimate. General statistical context: an analysis can be sensitive to whether missing outcomes are related to observed health status, treatment assignment, or reasons for leaving follow-up. Different assumptions can lead to different estimates, so a transparent report normally describes the missing-data approach and any checks of its robustness. The registry record does not report those details.

The record identifies an eye-level endpoint and notes that some participants could contribute two eyes with different assignments. It does not report how this dependence was addressed in the primary analysis. It also does not report interim analyses or stopping rules. No conclusions about those procedures should be inferred from the completed status or the posted counts. These reporting gaps limit assessment of the analysis’s precision and robustness; they do not, on their own, demonstrate that a particular method was or was not used.

11. 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.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: congestive heart failure (4.7%), myocardial infarction (4.7%), pneumonia (3.9%), diabetic ketoacidosis (2.8%), hyperglycemia (2.8%).

The registry reports serious adverse-event figures separately for the aflibercept group, the sham group, and bilateral participants. For aflibercept, 129 were at risk and 52 were reported as affected; for sham, 128 were at risk and 50 were affected. The bilateral-participant group lists 71 at risk and 24 affected. The record also lists deaths as 9, 10, and 5 in those respective groups. The bilateral group is a separate listing and should not be added to the other groups or treated as a directly comparable third randomized arm.

For other adverse events, the posted at-risk and affected counts are 129 and 116 for aflibercept, 128 and 112 for sham, and 71 and 62 for bilateral participants. Serious-event terms listed with percentages include congestive heart failure and myocardial infarction at 4.7% each, pneumonia at 3.9%, and diabetic ketoacidosis, hyperglycemia, and chest pain at 2.8% each. The record does not specify denominators for those term percentages in the posted details.

General statistical context: adverse-event counts describe reported safety experience; they are not automatically comparative risks. Interpretation depends on definitions, follow-up duration, exposure, and whether events are counted by participant or eye. The registry record does not provide a comparative safety estimate, uncertainty interval, or testing method. These figures should therefore be read descriptively, not as proof of a safety difference or equivalence.

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

The primary posted values describe how many eyes met the combined PDR and/or DME outcome in each arm: 27 for aflibercept and 75 for sham. They indicate a numerical difference in the reported counts, but the registry record does not post the primary effect estimate or its uncertainty. Without that inferential information, the size and precision of the primary contrast cannot be established from the counts alone. The result should not be converted into an event rate or a treatment effect by making an unreported calculation.

The outcome is eye-based, and some participants could contribute two eyes assigned to different strategies. That design feature affects how observations may be related and how uncertainty should be calculated. The registry record does not report the primary analysis method, the handling of paired eyes, or an inferential comparison for this endpoint. It is therefore not possible to assess the primary result’s statistical precision from the posted count values alone.

The secondary visual-acuity result is a distinct piece of evidence. Its recalculated unadjusted mean difference is -0.3 letter score, with a 95% confidence interval from -1.766 to 1.166 and p = 0.688. This interval includes zero and does not establish a clear unadjusted difference in mean change. It also does not establish equivalence, because the registry record does not report an equivalence margin or a design intended to test equivalence. The visual-acuity comparison does not replace the missing inferential analysis of the primary disease-development outcome.

General statistical context: a trial result is not a universal statement about every person, eye, setting, or treatment schedule. The record describes a particular randomized comparison in eyes at high risk and reports specified treatment schedules, follow-up windows, and outcomes. The available summaries do not establish whether the findings generalize beyond that context, how deferred treatment influenced the contrast, or how different dosing schedules would compare. The registry record does not report the frequency of deferred treatment or detailed outcome data for each component of the combined endpoint.

Accordingly, the posted information supports description of the recorded counts and secondary summaries, but not a definitive claim about the precision of the primary treatment contrast. It also does not support a broad claim of benefit, lack of benefit, safety, or equivalence. Those conclusions would require the relevant analyses and uncertainty estimates, interpreted alongside the trial design and its reporting limitations.

13. Why This Trial Matters Statistically

This trial is statistically notable because its main outcome is defined at the level of eyes, while the allocation description allows some participants to contribute two eyes receiving different strategies. That structure creates a design and analysis issue beyond comparing two independent groups of participants. An analysis needs to respect any within-participant relationship when estimating uncertainty. The registry record explains the eye-level assignment feature but does not report how the primary analysis handled it.

The primary endpoint also involves time to first criteria being met and censoring, while the posted result is a count. The endpoint description indicates that eyes without an outcome were censored at their last completed visit and that an eye meeting a criterion stopped contributing to the next criterion. Those rules distinguish event occurrence from observation time. Without a reported inferential analysis, readers cannot assess how event timing and censoring were reflected in the primary comparison.

The record also places an anatomic disease endpoint alongside a visual-acuity outcome. Statistically, these represent different domains and timeframes, and the existence of a secondary visual-acuity summary cannot stand in for a primary disease-endpoint analysis. The unadjusted secondary comparison is useful as a descriptive estimate with an uncertainty interval, but it is not a reported sponsor analysis and the record lists no reported analyses.

Finally, the planning illustrations make an important distinction between design planning and results. They use the observed visual-acuity effect under stated assumptions and show hypothetical participants per arm, not observed power and not the trial’s own design justification. Keeping that distinction clear prevents a planning calculation from being mistaken for evidence about the precision of the primary endpoint. The registry record does not report the trial’s original sample-size rationale or primary analysis plan, so those aspects cannot be reconstructed from the posted summaries.

14. Limitations

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

The main limitation of the posted primary result is that it consists of counts without a primary effect estimate, confidence interval, or p-value. The outcome is defined in eyes, and the record notes that some participants could contribute two eyes assigned to different strategies. The registry record does not report whether or how the primary analysis accounted for that dependence. It also does not report how censoring was incorporated into an inferential comparison.

The combined endpoint includes PDR and/or center-involved DME, but the posted information does not provide a primary inferential comparison for the combined outcome or clarify the contribution of each component to the posted count. The record describes separate development outcomes, but does not list reported analyses for them. This limits assessment of whether one component drove the combined count.

The trial strategies also allow deferred treatment under stated conditions. Because the frequency of deferred aflibercept or laser is not reported, the extent to which later treatment affected the comparison cannot be assessed. The registry record does not report adherence, actual treatment exposure, or detailed follow-up by arm.

For the visual-acuity outcome, the comparison is recalculated and unadjusted, and the record lists no reported analyses. It does not report an adjusted estimate, a clinically important threshold, a missing-data method, or a multiplicity plan. The planning illustrations use a secondary outcome’s observed effect and specified assumptions; they are not the study’s own design justification. Finally, the adverse-event summaries are descriptive and do not include a comparative inferential safety analysis. These limitations are specific to what the registry record reports and should not be taken as proof that unreported analyses were never conducted.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Anti-VEGF Treatment for Prevention of PDR/DME — NCT02634333.
• Registry record last updated 2023-06-18; status last verified 2023-05.
• Results first posted 2021-11-24.
• 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.

18. Record Summary

NCT02634333 is a completed, randomized, double-masked, parallel-group phase 3 trial sponsored by Jaeb Center for Health Research. Enrollment is listed as 399. The trial compared prompt aflibercept with prompt sham in eyes at high risk of vision-threatening diabetic retinopathy, with deferred treatment allowed under specified conditions.

The primary endpoint was the number of eyes meeting criteria for first development of PDR and/or DME over 2 years. The posted counts are 27 for aflibercept and 75 for sham. No primary inferential comparison or uncertainty interval is reported. The secondary visual-acuity outcome is reported over 4 years; its recalculated unadjusted mean difference is -0.3 letter score, with a 95% confidence interval from -1.766 to 1.166 and p = 0.688.

The record notes that a participant could have two study eyes and describes different assignments across eyes for some participants. It does not report the primary analysis method, multiplicity plan, or missing-data methods. Safety information is posted as descriptive counts, without a comparative inferential analysis.