This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
CRUISE was a randomized, double-masked, parallel-group phase 3 trial with three arms. It asked whether intravitreal ranibizumab, at either of two doses, improved best corrected visual acuity (BCVA) at 6 months compared with a sham injection in eyes with macular edema secondary to central retinal vein occlusion (CRVO).
| Feature | CRUISE |
|---|---|
| Phase | Phase 3 |
| Conditions | Macular edema; retinal vein occlusion (central retinal vein occlusion) |
| Design | Randomized, parallel-group, double-masked, sham-controlled |
| Arms | Sham injection; ranibizumab injection 0.3 mg; ranibizumab injection 0.5 mg |
| Enrollment | 392 |
| Primary endpoint | Mean change from baseline in BCVA score at 6 months |
| Primary analysis | ANOVA, difference in least squares means, stratified by baseline visual acuity; ITT with LOCF imputation |
| Study dates | Start July 2007; primary completion June 2009 |
| ClinicalTrials.gov | NCT00485836 |
| Sponsor | Genentech, Inc. (industry) |
2. Clinical Question
The central question was whether ranibizumab injections produced a larger mean gain in BCVA letters from baseline to month 6 than sham injection in patients with macular edema due to central retinal vein occlusion, and whether this held at both a lower (0.3 mg) and higher (0.5 mg) dose.
Population
Patients with macular edema secondary to central retinal vein occlusion, with one study eye assessed for visual acuity and retinal thickness.
Intervention
Ranibizumab injection at 0.3 mg or 0.5 mg, as two separately randomized dose arms.
Comparator
Sham injection, which mimics the injection procedure without delivering drug and supports masking of patients and assessors.
Primary question
Is the mean change in BCVA score from baseline to 6 months greater with each ranibizumab dose than with sham?
3. Trial Design
Ranibizumab 0.3 mg
- Ranibizumab injection 0.3 mg
- Compared pairwise with sham
Ranibizumab 0.5 mg
- Ranibizumab injection 0.5 mg
- Compared pairwise with sham
Sham injection
- Sham injection procedure without active drug
- Common control for both dose comparisons
One control, two comparisons
- Each dose is tested against the same sham group
- The registry posts dose-versus-sham analyses, not 0.5 mg versus 0.3 mg
The registry reports the number of participants at risk for adverse events in each arm (132, 129 and 129, totalling 390 of the 392 enrolled). The dosing schedule, number of injections and any rescue-treatment rules are not described in the ClinicalTrials.gov record.
4. Stratification and Analysis Populations
All posted analyses were stratified or adjusted by baseline visual acuity score in three categories: ≤ 34, 35–54, and ≥ 55 letters. Baseline acuity is strongly related to how many letters a patient can gain or lose, partly because of ceiling and floor effects on the ETDRS chart, so accounting for it improves precision and protects against chance imbalance.
| Analysis population | Definition / role |
|---|---|
| Intent to treat (randomized) | All randomized patients analysed by assigned arm; used for BCVA, letter-gain/loss and central foveal thickness endpoints. |
| Missing-data handling | Missing values imputed by last observation carried forward (LOCF). |
| NEI VFQ-25 analyses | Intent-to-treat population, but patients without a baseline subscale score were excluded. |
| Safety (adverse events) | Participants at risk: 129 sham, 132 ranibizumab 0.3 mg, 129 ranibizumab 0.5 mg. |
The ClinicalTrials.gov record does not report baseline demographic or disease characteristics by arm in the material summarised here, so no baseline table is presented.
5. Endpoints
Primary endpoint
| Endpoint | Definition | Time frame |
|---|---|---|
| Mean change from baseline in BCVA score at 6 months | BCVA score in the study eye based on Early Treatment Diabetic Retinopathy Study (ETDRS) visual acuity charts (number of correct letters), assessed at a starting distance of 4 meters. | Baseline and 6 months |
Secondary endpoints with posted analyses
| Endpoint | Type | Time frame | Method |
|---|---|---|---|
| Percentage gaining ≥ 15 letters in BCVA at month 6 | Binary | Baseline and 6 months | Cochran-Mantel-Haenszel χ² |
| Percentage losing < 15 letters in BCVA at month 6 | Binary | Baseline and 6 months | Cochran-Mantel-Haenszel χ² |
| Percentage with central foveal thickness ≤ 250 μm at month 6 | Binary | 6 months | Cochran-Mantel-Haenszel χ² |
| Mean absolute change in central foveal thickness at month 6 | Continuous (μm) | Baseline and 6 months | ANCOVA |
| Mean change in NEI VFQ-25 Near Activities subscale at month 6 | Continuous (points) | Baseline and 6 months | ANCOVA |
| Mean change in NEI VFQ-25 Distance Activities subscale at month 6 | Continuous (points) | Baseline and 6 months | ANCOVA |
6. Primary Endpoint Results
The registry lists each comparison as "Sham Injection vs Ranibizumab Injection". For the BCVA endpoints the posted differences are positive, which corresponds to a larger letter gain in the ranibizumab arm; a positive value therefore favours ranibizumab. Both analyses used ANOVA stratified by baseline visual acuity category, in the intent-to-treat population with LOCF imputation, under a superiority hypothesis.
Ranibizumab 0.3 mg versus sham
Difference in least squares mean BCVA change at 6 months
95% CI: 7.7–15.3 · P < 0.0001
ANOVA stratified by baseline visual acuity (≤ 34, 35–54, ≥ 55 letters)
What it means: after accounting for baseline visual acuity category, the estimated mean improvement in BCVA from baseline to month 6 was 11.5 letters greater with ranibizumab 0.3 mg than with sham. On the ETDRS chart, 5 letters correspond to one line, so this is a between-group difference of a little over two lines in average change.
What it does not mean: it does not mean each patient gained 11.5 letters, or that 11.5 letters is the gain in the ranibizumab arm alone. It is a difference between two group means; individual responses vary widely around those means.
Precision: the 95% CI of 7.7 to 15.3 letters excludes zero by a wide margin. Even the lower bound represents a mean difference of more than one ETDRS line, so the data are compatible with a benefit ranging from moderate to large, not merely with a marginal effect.
The p-value: P < 0.0001 says that a difference this large would be very unlikely if the true mean difference were zero. It does not measure how big the effect is; that information is in the estimate and its interval.
Cautions: the estimate relies on LOCF imputation, which assumes a patient's last recorded acuity would have persisted to month 6. Two dose arms are tested against one sham group, and the registry record does not describe how type I error was controlled across these comparisons.
Ranibizumab 0.5 mg versus sham
Difference in least squares mean BCVA change at 6 months
95% CI: 10.3–17.4 · P < 0.0001
ANOVA stratified by baseline visual acuity (≤ 34, 35–54, ≥ 55 letters)
What it means: the estimated mean BCVA gain at month 6 was 13.8 letters greater with ranibizumab 0.5 mg than with sham, adjusting for baseline acuity stratum. This corresponds to almost three ETDRS lines of difference in mean change.
What it does not mean: the 0.5 mg estimate (13.8) being numerically larger than the 0.3 mg estimate (11.5) does not show that 0.5 mg is superior to 0.3 mg. The two confidence intervals overlap substantially, and no direct 0.5 mg versus 0.3 mg analysis is posted in the registry.
Precision: the 95% CI of 10.3 to 17.4 letters lies entirely above zero and above one ETDRS line, indicating a precisely estimated benefit relative to sham.
The p-value: P < 0.0001 is reported as a threshold, not an exact value. It establishes that the difference is statistically distinguishable from zero but carries no information about clinical importance on its own.
Cautions: both primary comparisons share the same sham group, so their results are correlated rather than two independent confirmations. LOCF and the superiority framework apply here as for the 0.3 mg comparison.
| Comparison | Method | LS mean difference | 95% CI | P-value |
|---|---|---|---|---|
| Ranibizumab 0.3 mg vs sham | ANOVA, stratified | 11.5 | 7.7 to 15.3 | <0.0001 |
| Ranibizumab 0.5 mg vs sham | ANOVA, stratified | 13.8 | 10.3 to 17.4 | <0.0001 |
The registry posts the between-group differences; arm-level mean changes are not part of the analyses summarised here, so the size of the change within each arm is not presented.
7. Secondary Endpoint Results
All secondary analyses were superiority comparisons of each ranibizumab dose versus sham in the intent-to-treat population with LOCF imputation (NEI VFQ-25 analyses excluded patients without a baseline score).
Binary visual-acuity and anatomical endpoints
These endpoints were analysed with Cochran-Mantel-Haenszel (CMH) χ² tests stratified by baseline visual acuity category. The effect measure is a difference in percentages, with CMH weights used to combine stratum-specific differences into one adjusted estimate.
| Endpoint | Comparison | Difference in % | 95% CI | P-value |
|---|---|---|---|---|
| Gained ≥ 15 letters at month 6 | 0.3 mg vs sham | 29.3 | 18.8 to 39.7 | <0.0001 |
| 0.5 mg vs sham | 30.3 | 19.6 to 40.9 | <0.0001 | |
| Lost < 15 letters at month 6 | 0.3 mg vs sham | 11.3 | 4.3 to 18.2 | 0.0019 |
| 0.5 mg vs sham | 13.6 | 7.2 to 20.1 | <0.0001 | |
| Central foveal thickness ≤ 250 μm at month 6 | 0.3 mg vs sham | 51.9 | 41.6 to 62.3 | <0.0001 |
| 0.5 mg vs sham | 54 | 44.0 to 64.1 | <0.0001 |
A difference of 29.3 percentage points in the proportion gaining ≥ 15 letters means that, after CMH weighting across baseline-acuity strata, the share of patients achieving a three-line gain was estimated to be 29.3 percentage points higher with ranibizumab 0.3 mg than with sham. This is an absolute risk difference, not a relative increase. The "lost < 15 letters" endpoint measures avoidance of substantial vision loss; its differences are smaller (11.3 and 13.6 points). Because the outcome counts patients who did not deteriorate substantially, the room for a between-arm difference depends on how often such loss occurs without treatment, so this endpoint and the ≥ 15-letter gain endpoint capture different parts of the outcome distribution.
Continuous anatomical and patient-reported endpoints
These endpoints were analysed with pairwise ANCOVA models adjusted for baseline visual acuity category and the baseline value of the outcome itself.
| Endpoint | Comparison | LS mean difference | 95% CI | P-value |
|---|---|---|---|---|
| Change in central foveal thickness (μm) | 0.3 mg vs sham | -272.2 | -329.9 to -214.5 | <0.0001 |
| 0.5 mg vs sham | -283.8 | -337.8 to -229.8 | <0.0001 | |
| NEI VFQ-25 Near Activities (points) | 0.3 mg vs sham | 5.8 | 2.1 to 9.4 | 0.0019 |
| 0.5 mg vs sham | 4.9 | 1.2 to 8.6 | 0.0099 | |
| NEI VFQ-25 Distance Activities (points) | 0.3 mg vs sham | 6.3 | 3.1 to 9.5 | 0.0002 |
| 0.5 mg vs sham | 4.1 | 0.7 to 7.6 | 0.0199 |
The central foveal thickness differences are negative because ranibizumab produced a greater reduction in retinal thickness than sham; for this endpoint, a negative value favours ranibizumab. The NEI VFQ-25 differences are positive, indicating greater improvement in self-reported near and distance visual function with ranibizumab. The patient-reported intervals are wider relative to their estimates than the anatomical ones, and the 0.5 mg Distance Activities interval (0.7 to 7.6) approaches zero, illustrating that statistical significance and estimate precision differ across endpoints even when all p-values fall below 0.05.
8. Safety: Serious Adverse Events
| Arm | Participants with serious adverse events | Participants at risk |
|---|---|---|
| Sham injection | 13 | 129 |
| Ranibizumab injection 0.3 mg | 16 | 132 |
| Ranibizumab injection 0.5 mg | 13 | 129 |
Serious adverse event counts were similar across the three arms. No formal statistical comparison of adverse events is posted, and with event counts in the teens per arm, a trial of this size has limited power to detect differences in uncommon serious events. Safety findings are therefore descriptive and should be read alongside, not combined with, the efficacy estimates.
9. Statistical Methodology
ANOVA for the primary endpoint
The primary endpoint, change in BCVA letters from baseline to month 6, is a continuous outcome. The registry reports an analysis of variance stratified by baseline visual acuity category, with the treatment effect expressed as a difference in least squares (LS) means. LS means are model-based group means that balance over the stratification factor, so each arm's mean is estimated as if the arms had identical distributions of baseline-acuity strata.
The treatment effect for each dose is the difference τranibizumab − τsham, estimated as a difference in LS means, with stratum terms γ for baseline acuity ≤ 34, 35–54 and ≥ 55 letters.
Cochran-Mantel-Haenszel analysis for binary endpoints
For the proportion of patients gaining ≥ 15 letters, losing < 15 letters, or reaching central foveal thickness ≤ 250 μm, the CMH χ² test compares arms within each baseline-acuity stratum and combines the stratum results. The adjusted difference in percentages is a weighted average of stratum-specific differences using CMH weights.
where k indexes baseline-acuity strata and p1k, p0k are the response proportions in the ranibizumab and sham arms within stratum k.
ANCOVA for central foveal thickness and NEI VFQ-25
For these continuous secondary endpoints, pairwise ANCOVA models included the baseline value of the outcome as a covariate alongside baseline visual acuity category. Adjusting for the baseline value of the same measurement typically reduces residual variance substantially, giving narrower confidence intervals than an unadjusted comparison of change scores.
Intention-to-treat and LOCF
Efficacy analyses used the intent-to-treat (randomized) population, analysing patients according to their assigned arm. Missing month-6 values were filled using last observation carried forward. For the NEI VFQ-25 endpoints, patients without a baseline score were excluded, so those analyses depart slightly from a complete ITT population.
Hypothesis framework
Every posted analysis is labelled as a superiority test. The ClinicalTrials.gov record does not describe the sample-size calculation, the multiplicity strategy across the two dose comparisons and multiple secondary endpoints, or any interim analyses.
10. Missing Data and the LOCF Assumption
LOCF replaces a missing month-6 value with the patient's most recent observed value. Its implicit assumption is that the outcome would have remained unchanged after the last visit.
When LOCF can favour treatment
If patients on sham drop out early while vision is declining, carrying forward an earlier, better value can understate deterioration in the control arm, which would shrink the observed difference rather than enlarge it.
When LOCF can favour control
If treated patients drop out before reaching their maximal gain, carrying forward an early value understates their benefit. The direction of LOCF bias depends on who drops out and when.
LOCF also understates uncertainty, because imputed values are treated as if they had been observed. Multiple imputation or mixed models for repeated measures are modern alternatives that incorporate uncertainty about missing values. The registry does not report the number of patients with imputed values, so the potential influence of LOCF cannot be gauged from the record alone.
11. Multiplicity
CRUISE involves two dose-versus-sham comparisons for the primary endpoint and twelve posted secondary comparisons. Without adjustment, testing many hypotheses increases the chance that at least one reaches P < 0.05 by chance.
| Analysis | Role | Interpretation |
|---|---|---|
| BCVA change, 0.3 mg and 0.5 mg vs sham | Primary | Main confirmatory comparisons; both P < 0.0001 |
| ≥ 15-letter gain; < 15-letter loss | Secondary | Responder-type summaries of the same BCVA measurement |
| Central foveal thickness | Secondary | Anatomical outcome supporting the mechanism of effect |
| NEI VFQ-25 subscales | Secondary | Patient-reported function; smallest margins from zero |
Most posted p-values are <0.0001, so they would remain significant under nearly any conventional correction. The less extreme results (for example, P = 0.0199 for the 0.5 mg Distance Activities comparison) are the ones whose interpretation would depend on the multiplicity strategy, which the registry does not describe.
12. Statistical Methods Explained
Why was ANOVA used for the primary endpoint rather than ANCOVA?
The primary analysis is described as an ANOVA stratified by baseline visual acuity category. Because the outcome is already a change from baseline and the categorical baseline-acuity stratum enters the model, the analysis accounts for baseline acuity in a coarse, three-level form. ANCOVA with continuous baseline BCVA would adjust more finely; the secondary continuous endpoints did use ANCOVA with the baseline value as a covariate.
What does a "difference in least squares means of 13.8" mean?
It is the model-estimated difference in mean BCVA change between the ranibizumab 0.5 mg and sham arms, balanced across the three baseline-acuity strata. It equals almost three ETDRS lines of difference in average improvement. It is not an individual patient's expected gain.
Why use a Cochran-Mantel-Haenszel test for the ≥ 15-letter endpoint?
The likelihood of gaining 15 letters depends heavily on starting acuity: patients with good baseline vision have less room to improve. The CMH approach compares arms within baseline-acuity strata and pools the results, so the treatment comparison is not distorted by differences in stratum composition between arms.
What is the difference between a risk difference of 30.3 points and a relative risk?
A difference of 30.3 percentage points is absolute: it says how many more patients per 100 achieved the outcome with ranibizumab 0.5 mg than with sham. A relative risk would divide the two proportions and could look very large if the sham proportion were small. The registry reports the absolute difference, which is the more direct measure of how many additional patients benefit.
Why is the central foveal thickness estimate negative?
Macular edema thickens the retina, and successful treatment reduces thickness. A difference of -283.8 μm indicates that thickness fell by 283.8 μm more, on average, with ranibizumab 0.5 mg than with sham, after adjusting for baseline thickness and acuity stratum. The sign depends on the outcome's direction, not on whether the result is favourable.
Does a smaller p-value mean a larger effect?
No. The 0.3 mg Near Activities comparison (P = 0.0019) and the 0.3 mg "lost < 15 letters" comparison (P = 0.0019) share a p-value but measure entirely different quantities on different scales. P-values reflect both effect size and precision; the estimate and confidence interval are the appropriate measures of magnitude.
13. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
Both ranibizumab doses showed mean BCVA gains at month 6 that exceeded sham, with confidence intervals well away from zero. Binary, anatomical and patient-reported secondary endpoints all pointed in the same direction, with varying precision.
Clinical interpretation
Differences of roughly two to three ETDRS lines in mean change, and 29.3 to 30.3 percentage points more patients gaining three or more lines, describe effects of a magnitude that patients would notice. Durability beyond 6 months is not addressed by the posted primary analysis.
The anatomical endpoint (central foveal thickness), the functional endpoint (BCVA) and the patient-reported endpoints (NEI VFQ-25) all favour ranibizumab. Coherence across different measurement types strengthens the overall inference beyond what any single p-value conveys, although the endpoints are correlated and are not independent replications.
14. Important Limitations and Interpretation Issues
- LOCF imputation: single imputation assumes no change after the last observation and understates uncertainty; its direction of bias depends on dropout patterns not reported in the registry.
- Shared control arm: both dose comparisons use the same sham group, so the two primary results are correlated.
- No dose-versus-dose analysis: numerical differences between 0.3 mg and 0.5 mg estimates should not be read as evidence of a dose response.
- Multiplicity: the registry does not describe how type I error was controlled across two doses and multiple secondary endpoints.
- Coarse baseline adjustment: the primary ANOVA adjusts for baseline acuity in three categories rather than as a continuous covariate.
- Patient-reported outcomes: NEI VFQ-25 analyses excluded patients without a baseline score, departing from full ITT.
- Time horizon: the primary endpoint is measured at 6 months; longer-term effects are outside the posted primary analysis.
- Safety precision: serious adverse event counts are small in each arm, limiting the ability to detect differences in rare events.
15. Why This Trial Matters Statistically
CRUISE is a useful teaching case for analysing continuous and binary outcomes in a multi-arm, sham-controlled ophthalmology trial.
| Concept | How it appears in CRUISE |
|---|---|
| Randomization and masking | Three-arm, double-masked design with a sham injection control |
| ITT analysis | Efficacy analysed in the intent-to-treat (randomized) population |
| ANOVA | Primary BCVA change analysis stratified by baseline acuity |
| ANCOVA | Central foveal thickness and NEI VFQ-25, adjusted for baseline values |
| Cochran-Mantel-Haenszel test | Stratified comparison of ≥ 15-letter gain, < 15-letter loss, and thickness ≤ 250 μm |
| Risk difference | Binary endpoints reported as CMH-weighted differences in percentages |
| Confidence intervals | 95% intervals for every posted comparison |
| Missing data | LOCF single imputation for month-6 values |
| Multiple comparisons | Two doses against one control across many endpoints |
16. Related Tutorials
Learn more about the methods used in this trial:
17. Related Calculators
18. Sources
- ClinicalTrials.gov: NCT00485836.
- PubMed: PMID 23699977.
- PubMed: PMID 23415775.
Continue exploring clinical biostatistics
Connect this trial's endpoints and methods to deeper statistical tutorials and to calculators for analysing similar data.
19. Record Summary
CRUISE illustrates the statistical toolkit of a modern sham-controlled ophthalmology trial: a continuous primary endpoint analysed by stratified ANOVA, binary responder endpoints analysed by CMH-weighted risk differences, baseline-adjusted ANCOVA for anatomical and patient-reported outcomes, intention-to-treat analysis with LOCF imputation, and two active doses compared with a single control. The most informative reading combines mean differences in letters, absolute differences in responder proportions, confidence intervals, and an understanding of how missing data and multiple comparisons shape inference.