Rainbow Extension Study: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Rainbow Extension Study (NCT02640664), a non-randomized phase 3 study. The primary outcome is Visual Acuity (VA) of the Better-seeing Eye at the Participant's Fifth Birthday Visit - Comparison Between Treatment Arms.

2. Clinical Question

This is a long-term comparison following treatment in an earlier core study, rather than a description of a newly randomized treatment assignment. The registry’s stated purpose was to evaluate long-term efficacy and safety of ranibizumab compared with laser ablation therapy in participants previously treated for retinopathy of prematurity. The primary assessment was made at the fifth birthday visit, with the timeframe described as a maximum of 5 years and 4 months post core baseline visit.

The question is about a group-level visual-acuity score at that visit. It does not, by itself, describe every child’s vision, other aspects of eye health, or whether any observed difference was caused by the original treatment. Because allocation was non-randomized, differences between groups may reflect factors other than treatment. Details of how participants came to be in each group are not reported in the registry record.

3. Trial Design

The RAINBOW Extension Study is listed as a phase 3, parallel, non-randomized, unmasked study of long-term efficacy and safety in retinopathy of prematurity. Its purpose was to follow participants treated in the core study CRFB002H2301 and compare those previously treated with ranibizumab with those previously treated with laser therapy. The registry lists enrollment of 180 and status as completed.

The registry lists three groups: ranibizumab 0.2 mg, ranibizumab 0.1 mg, and laser therapy. The two ranibizumab descriptions specify one intravitreal injection in both eyes on Day 1 (Baseline), with up to 2 re-treatments allowed for each eye if required. The laser group is described as receiving laser therapy. Details about the conduct of laser treatment are not reported in the registry record.

The study started on 2016-06-16, and its primary completion date is 2022-04-21. These dates describe the registry record; they do not establish the follow-up duration for every participant. The primary outcome timeframe is at the participant’s fifth birthday visit, with a maximum of 5 years and 4 months post core baseline visit.

The non-randomized allocation matters for interpretation. Group differences cannot be read as though treatment had been assigned by chance, and the record explicitly cautions that the arms are not a like-for-like control comparison. Masking is listed as none. The registry record does not report the method used to assign participants to groups, the reasons for group assignment, or whether outcome assessors were masked. It also does not report a detailed statistical analysis plan or sample-size justification.

This extension draws on participants from the core study Safety Set who entered the extension. The primary outcome description calls this the Extension Safety Set and says that it comprises data from both core and extension studies. The outcome population also includes the number of participants contributing to the analysis, but the primary outcome contribution counts 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: This study was not randomized. During Epoch 1 (starting at the baseline visit for the extension study up to 40 weeks from the baseline visit in the core study) participants continued to receive treatment as in the core study (NCT02375971). During Epoch 2 (starting at the end of Epoch 1 up to participant's 5th birthday) participants no longer received treatment (the study was observational)

Epoch 1

Reasons for not completing this period:

Epoch 2

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Extension Safety Set: defined as the subset of the participants from the Safety Set of the core study who entered the extension study and comprised data from both core and extension studies. The outcome measure included number of participants contributing to analysis.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as Visual Acuity (VA) of the Better-seeing Eye at the Participant's Fifth Birthday Visit - Comparison Between Treatment Arms, measured over at the participant's fifth birthday visit (maximum 5 years and 4 months post core baseline visit).

The primary endpoint is visual acuity in the better-seeing eye at the participant’s fifth birthday visit, expressed as a score on a scale and summarized using a least-squares mean with a standard error. The visit’s timeframe is described as a maximum of 5 years and 4 months post core baseline visit.

Assessment followed Early Treatment Diabetic Retinopathy Study methodology. Testing was performed in a sitting position at an initial distance of 3 meters using Lea Symbols charts. The score counted the optotypes, or Lea symbols, the participant identified and ranged from 0 to 100; higher scores indicate better visual acuity. Each eye was tested using the child’s current refractive index. The eye with the higher score at the fifth birthday visit was defined as the better-seeing eye; if scores were the same, the right eye was assigned as the better-seeing eye.

A visual-acuity score provides a defined way to compare measured visual function between groups at a specified visit. The better-seeing-eye measure focuses on the eye with the higher measured score at that visit, rather than averaging both eyes. Because the eye is selected using the score from that same assessment, the endpoint does not describe the difference between the two eyes or the acuity of the worse-seeing eye. The registry also lists worse-seeing-eye acuity as a secondary outcome.

The endpoint definition is detailed, but the group least-squares means, standard errors, and treatment comparison are not reported in the registry record. Without those results, the magnitude and precision of the primary comparison cannot be evaluated.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry describes a continuous outcome summarized as a least-squares mean and standard error. It does not report the fitted model, covariates, contrast specification, or uncertainty interval for the primary comparison. The concepts below explain how those statistical summaries are generally used; they are not a claim about unreported details of this study’s analysis.

What does a least-squares mean represent?

For a continuous outcome, an ordinary group mean is the arithmetic average of observed values. A least-squares mean is generally an estimate from a statistical model. It can represent a group average adjusted for terms included in that model. Its interpretation therefore depends on which terms were fitted and how the model was specified. The registry does not report those details, so the precise adjustment represented by the primary endpoint summary is not reported in the registry record.

The standard error describes the estimated precision of a mean or contrast. It reflects uncertainty in the estimated quantity, rather than the range of scores among individual participants. Smaller standard errors generally correspond to more precise estimates, all else being equal. A standard error alone does not state the range of values compatible with the data as directly as a confidence interval would. Neither a standard error value nor a confidence interval for the primary outcome is reported in the registry record.

Why does the study design affect the comparison?

Allocation is listed as non-randomized, and the registry warns that the arms are not a like-for-like control comparison. Statistical adjustment can account for measured characteristics included in a model, but it cannot guarantee removal of bias from unmeasured factors or factors measured inadequately. Randomization and statistical adjustment are not interchangeable: adjustment depends on the variables measured and the assumptions of the analysis. The registry does not report which characteristics, if any, were included in the primary model.

The study is also listed as unmasked. Masking can help limit some influences on assessment or reporting, depending on how it is applied. The record does not describe who knew treatment group during the extension or whether any outcome assessment procedures were masked. These details are therefore not reported in the registry record.

How should the visual-acuity scale be read?

The endpoint uses a score from 0 to 100, with higher scores indicating better visual acuity. The assessment used Lea Symbols charts and counted the optotypes identified. The better-seeing eye was selected by comparing the two eye scores at the fifth birthday visit, with the right eye selected if both scores matched. These rules make the outcome definition explicit, but they do not provide the outcome values. Group means, the difference between groups, and the standard error are not reported in the registry record.

The endpoint is a continuous score, so a mean comparison describes an average group-level difference, not the pattern of every participant’s scores. Distributions can vary even when group averages are similar, and a mean does not show how many participants had a particular score. The registry record does not report the distribution of primary outcome scores.

What do the sample-size illustrations mean?

The sample-size material is described as a planning illustration based on the observed effect, not as the study’s own design. It assumes a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation; the effect to detect is a stated fraction of the observed effect. Such calculations illustrate how a future study’s required sample size changes with the effect size selected and other assumptions. They are not observed power and do not establish that the extension was designed with those assumptions.

For continuous outcomes, a planning calculation also depends on variability and the scale of the effect being targeted. The record does not provide the numerical planning results in the registry record for this explanation. Its sample-size justification, if any, is not reported in the registry record. The illustrations should not be interpreted as evidence that the observed comparison was sufficiently powered or that a non-significant result would demonstrate equivalence.

Why can over-dispersion matter for event counts?

In general, count data may vary more than a simple statistical model expects. This is called over-dispersion. If it is ignored, standard errors may be too small and intervals too narrow, making estimates appear more precise than they are. This general issue can matter in studies analyzing event counts, but the primary endpoint in this record is a continuous visual-acuity score. The registry does not report a count model for the primary analysis.

For adverse events, the registry gives counts of participants with serious and other events, not a model-based comparison of event rates. No adjustment, hypothesis test, confidence interval, or exposure-time analysis for those counts is reported in the registry record. The counts should therefore be read descriptively rather than as a formal test of treatment differences.

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("Ranibizumab 0.2 mg", "Ranibizumab 0.1 mg", "Laser Therapy")
n   <- c(45, 43, 36)
m   <- c(66.8, 64.6, 62.1)
sp  <- c(1.95, 2, 2.18)
sd  <- sp * sqrt(n)      # standard errors converted to SDs
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  Ranibizumab 0.2 mg  45  66.800  13.081  62.870  70.730
CARM  Ranibizumab 0.1 mg  43  64.600  13.115  60.564  68.636
CARM  Laser Therapy  36  62.100  13.080  57.674  66.526

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 25 outcome measures and 6 analyses. It also names multiple secondary outcomes, including ocular and non-ocular adverse events, visual acuity in the worse-seeing eye, absence of active retinopathy of prematurity at specified visits, and ocular structural abnormalities. When many outcomes or comparisons are considered, the chance of at least one apparently unusual result can increase. Statistical approaches to multiplicity can address this, but the registry record does not report whether a multiplicity adjustment was used, which outcomes were included in any adjustment, or how the analyses were prioritized.

The primary outcome is clearly identified as better-seeing-eye visual acuity at the participant’s fifth birthday visit. However, the registry does not report the estimate, standard error, confidence interval, or test result for that endpoint. The relationship between the listed analyses and specific outcomes or comparisons is not reported in the registry record. No claim about statistical significance or adjustment can be made from the outcome titles alone.

The primary population is defined as the Extension Safety Set: participants from the core study Safety Set who entered the extension, with data from both the core and extension studies. The registry says that the outcome includes the number contributing to analysis, but does not report that number for the primary outcome. It also does not report how missing fifth birthday assessments were handled, whether participants were excluded from the analysis, or whether any imputation method was used.

Missing data can affect both precision and interpretation. If the chance of having a recorded visit relates to visual acuity or other participant characteristics, the analyzed observations may not represent all eligible participants. The registry record does not report the extent or pattern of missing primary outcome data, so the potential impact cannot be evaluated.

12. Safety Results

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

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 4% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: bronchitis (9.3%), pneumonia (7.4%), bronchiolitis (6.6%), pyrexia (4.9%), dehydration (4.9%).

The adverse-event figures are counts of participants at risk and participants affected, reported regardless of whether the event was considered related to study treatment or a procedure. For serious events, the group counts are 21 affected among 61 at risk in the ranibizumab 0.2 mg group, 27 among 65 in the ranibizumab 0.1 mg group, and 26 among 54 in the laser therapy group. The registry’s total row reports 74 affected among 180 at risk. These are descriptive counts, not estimates of a causal treatment effect.

The serious-event terms listed most often include bronchitis at 9.3%, pneumonia at 7.4%, bronchiolitis at 6.6%, pyrexia at 4.9%, dehydration at 4.9%, and vomiting at 4.6%. The registry record does not report the term-specific denominators or whether one participant could contribute to more than one term. The percentages should not be added or treated as mutually exclusive categories.

For other adverse events, the record reports 41 affected among 61 at risk for ranibizumab 0.2 mg, 46 among 65 for ranibizumab 0.1 mg, and 40 among 54 for laser therapy; the total row reports 127 among 180. Death counts are 0, 0, and 2 in the respective groups, with 2 in total. The record does not attribute these deaths to treatment. Because allocation was non-randomized, these group counts should not be read as a like-for-like safety comparison.

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

The registry record establishes that the extension was intended to compare long-term outcomes after prior treatment for retinopathy of prematurity. It defines a specific primary outcome: a least-squares mean score for visual acuity in the better-seeing eye at the participant’s fifth birthday visit. The score ranges from 0 to 100, and higher scores indicate better visual acuity. These details clarify what the primary comparison was meant to measure.

However, an endpoint definition is not an endpoint result. The primary group estimates, their standard errors, the between-group contrast, and its uncertainty are not reported in the registry record. Therefore, the record does not establish whether the group scores differed, how large any difference was, or how precisely it was estimated.

The design also sets limits on causal interpretation. The study is listed as non-randomized and unmasked, and the registry cautions that its arms are not a like-for-like control comparison. Even if a numerical difference were available, a group contrast could reflect differences between participants or their treatment histories as well as treatment. The record does not provide the details needed to assess how such differences were handled.

The adverse-event counts show that participants in each group were reported with serious and other adverse events. They are not, on their own, evidence that treatment caused an event or that one group had greater treatment-related harm. The event counts are reported regardless of relationship to treatment or procedure, and the non-randomized design further limits direct comparison.

These limitations do not make the recorded design or outcome definitions uninformative. They identify the scope of what can be learned from the record: the intended long-term comparison, the way visual acuity was measured, and descriptive safety counts. Claims about efficacy, equivalence, safety differences, or treatment causation require comparative results and supporting analysis details that are not reported in the registry record.

14. Why This Trial Matters Statistically

Long-term follow-up raises statistical questions that differ from those in a short-term treatment assessment. The primary endpoint is scheduled at the participant’s fifth birthday visit, with a maximum of 5 years and 4 months post core baseline visit. Interpreting an outcome at a later visit requires knowing which participants contributed data and how incomplete follow-up was handled. Those primary analysis counts and missing-data details are not reported in the registry record.

The endpoint definition is another important statistical feature. It uses a standardized visual-acuity assessment, specifies the scoring scale, and defines how the better-seeing eye is chosen. Clear definitions reduce ambiguity about what the analysis target represents. At the same time, selecting the better-scoring eye at that visit means the endpoint is not a summary of both eyes or of the worse-seeing eye. The registry lists the worse-seeing eye as a secondary outcome, which addresses a different aspect of visual acuity.

The study is also an example of why design must accompany numerical comparison. A non-randomized extension can provide follow-up information, but group differences may be entangled with how participants came to be in each group. The registry explicitly warns that the arms are not a like-for-like control comparison. A group estimate without its design context could be overinterpreted as a treatment effect.

Finally, the record illustrates the distinction between an outcome being registered and its result being reported. The primary outcome is described in detail, yet its numerical comparison is not reported in the registry record. For statistical interpretation, an outcome definition, an estimate, its uncertainty, the analysis population, and the design all matter. When some are unavailable, the appropriate conclusion is limited to what the record actually documents.

15. Limitations

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

The most consequential limitation for interpreting the primary question is that the numerical visual-acuity comparison is not reported in the registry record. Without group estimates, uncertainty measures, or a test result, the primary endpoint cannot be evaluated for direction, magnitude, or precision.

Non-randomized allocation limits causal interpretation. The registry itself cautions that the arms are not a like-for-like control comparison. The method of group assignment, baseline comparability, and reasons for treatment choice are not reported in the registry record. The record also does not specify the model or covariates used for the least-squares mean, so the adjustment behind that summary cannot be assessed.

Masking is listed as none, but the record does not explain which people involved in treatment or outcome assessment knew group assignment. The primary analysis contribution counts and handling of missing fifth birthday assessments are not reported. This leaves uncertainty about the analyzed population and whether incomplete follow-up could affect the comparison.

The extension’s relationship to the core study is described through the Extension Safety Set, which includes participants from the core Safety Set who entered the extension. Results for this population may not describe people outside that defined group. The registry does not report details that would support broader generalization beyond the extension participants.

For safety, the record provides descriptive adverse-event counts regardless of treatment or procedure relationship. It does not provide causal attribution, event-specific denominators for the listed serious terms, or comparative uncertainty estimates. These limitations mean the counts should not be treated as formal evidence of a difference in treatment-related risk.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Rainbow Extension Study — NCT02640664.
• Registry record last updated 2023-02-13; status last verified 2023-01.
• Results first posted 2023-02-13.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

The RAINBOW Extension Study is a completed phase 3, parallel, non-randomized, unmasked study of long-term outcomes after treatment for retinopathy of prematurity in the core study CRFB002H2301. The registry lists enrollment of 180, a start date of 2016-06-16, and a primary completion date of 2022-04-21.

Its primary outcome is better-seeing-eye visual acuity at the participant’s fifth birthday visit, summarized as a least-squares mean score with a standard error. The scale ranges from 0 to 100, with higher scores indicating better visual acuity. The primary comparative result and its uncertainty are not reported in the registry record.

The registry describes ranibizumab 0.2 mg, ranibizumab 0.1 mg, and laser therapy groups, and cautions that the non-randomized arms are not a like-for-like control comparison. It reports descriptive adverse-event counts, including 74 participants with serious adverse events among 180 at risk in total. Those counts are reported regardless of relationship to treatment or procedure and do not establish causation.