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
NEUGR-003 was a randomized, triple-masked phase 3 supportive-care trial asking whether Neugranin 40 mg was non-inferior to pegfilgrastim in limiting the duration of severe neutropenia during the first cycle of doxorubicin/docetaxel chemotherapy in participants with breast cancer.
| Feature | NEUGR-003 |
|---|---|
| Official title | Neugranin in Breast Cancer Participants Receiving Doxorubicin/Docetaxel |
| Phase | Phase 3 |
| Condition | Chemotherapy-induced neutropenia |
| Primary purpose | Supportive care |
| Design | Randomized, parallel-group, triple-masked, with a double-blind phase and an open-label phase |
| Arms | 3 (double-blind pegfilgrastim; double-blind Neugranin 40 mg; open-label Neugranin 40 mg) |
| Enrollment | 381 |
| Interventions | Neugranin (biological); pegfilgrastim (biological); chemotherapy (drug) |
| Primary hypothesis | Non-inferiority of Neugranin 40 mg to pegfilgrastim for cycle 1 duration of severe neutropenia |
| Start date | June 30, 2010 |
| Primary completion date | February 29, 2012 |
| Status | Completed; results posted |
| Lead sponsor | Teva Branded Pharmaceutical Products R&D, Inc. (industry) |
| ClinicalTrials.gov | NCT01126190 |
2. Clinical Question
Myelosuppressive chemotherapy such as doxorubicin/docetaxel commonly produces severe neutropenia, and granulocyte colony-stimulating factors are used to shorten it. The question in NEUGR-003 was not whether Neugranin was better than pegfilgrastim, but whether it was not unacceptably worse at shortening severe neutropenia in the first chemotherapy cycle.
Population
Participants with breast cancer receiving doxorubicin/docetaxel chemotherapy, at risk of chemotherapy-induced neutropenia.
Intervention
Neugranin 40 mg, given alongside chemotherapy (double-blind phase, with an additional open-label Neugranin 40 mg arm).
Comparator
Pegfilgrastim, an established long-acting granulocyte colony-stimulating factor, given alongside chemotherapy in the double-blind phase.
Primary question
Is the mean duration of severe neutropenia in cycle 1 with Neugranin 40 mg no more than 0.62 days longer than with pegfilgrastim?
3. Trial Design
Experimental arm
- Neugranin 40 mg
- Doxorubicin/docetaxel chemotherapy
- Masked to treatment assignment
- Contributes to the formal non-inferiority comparison
Active comparator arm
- Pegfilgrastim
- Doxorubicin/docetaxel chemotherapy
- Masked to treatment assignment
- Reference arm for the non-inferiority comparison
Masking
The registry classifies masking as triple, meaning participants, care providers/investigators and at least one further party (such as outcomes assessors) were masked in the double-blind phase. Masking matters especially in a non-inferiority trial: when the expected difference between active treatments is small, subtle differences in care or assessment could shift the estimate toward the margin, and masking protects against that.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Double-Blind Phase: Duration of Severe Neutropenia in Cycle 1 (primary) | Severe neutropenia was defined as Grade 4 neutropenia (absolute neutrophil count [ANC] <0.5 × 109/L). Duration was calculated by cycle as the number of days from the first day the ANC fell below 0.5 × 109/L after beginning a chemotherapy cycle until the participant had an ANC ≥0.5 × 109/L within the cycle. | Cycle 1 (cycle length = 21 days) |
| Open-Label Phase: Duration of Severe Neutropenia in Cycle 1 (primary) | Same definition as above. There was no planned statistical analysis for the open-label phase arm. | Cycle 1 (cycle length = 21 days) |
Duration of severe neutropenia (DSN) is a count of days. It is bounded below at zero (many participants receiving an effective growth factor may never fall below the Grade 4 threshold, or do so for only a day or two) and bounded above by the length of the cycle. Its distribution is therefore typically right-skewed and contains many ties and zeros, which has direct consequences for how a confidence interval for a mean difference should be computed.
5. Primary Result: Double-Blind Duration of Severe Neutropenia
The registry posts one formal statistical analysis: the non-inferiority comparison of cycle 1 DSN between double-blind Neugranin 40 mg and double-blind pegfilgrastim, in the per-protocol population.
Difference in mean DSN, cycle 1
95% CI (two-sided): −0.13 to 0.37 days
Non-inferiority margin: upper CI limit ≤ 0.62 days
| Item | Posted value |
|---|---|
| Groups compared | Double-Blind Phase: Pegfilgrastim vs Double-Blind Phase: Neugranin 40 mg |
| Effect measure | Mean difference (final values) |
| Estimate | 0.1 days |
| 95% CI (two-sided) | −0.13 to 0.37 |
| Hypothesis | Non-inferiority |
| Non-inferiority criterion | Upper endpoint of the CI for the difference in mean DSN ≤ 0.62 days |
| CI method | Stratified bootstrap resampling |
| Analysis population | Per protocol: all data from randomized participants obtained prior to experiencing major protocol violations |
| P-value | Not reported on ClinicalTrials.gov |
The upper limit of the 95% CI (0.37 days) is below the prespecified margin of 0.62 days, so the posted analysis meets the registry's stated criterion for non-inferiority. The interval also includes zero, so the data are compatible with no difference in mean DSN between the two growth factors.
What the estimate means. A mean difference of 0.1 days says that, in the per-protocol population, average cycle 1 DSN differed by about one tenth of a day between the two arms. The non-inferiority rule is framed on the upper CI limit, which implies the difference is oriented so that positive values represent longer severe neutropenia with Neugranin; on that reading, Neugranin participants averaged roughly 0.1 days more DSN. The registry lists the comparison as "Pegfilgrastim vs Neugranin 40 mg", so the sign convention should be confirmed against the full report before quoting a direction.
What it does not mean. It does not show that Neugranin is superior, nor that the two products are equivalent in every respect. Non-inferiority is a one-directional claim: Neugranin is not worse than pegfilgrastim by more than 0.62 days. It also says nothing about individual patients; a mean difference of 0.1 days is fully compatible with many participants having zero days of severe neutropenia and some having several.
Precision. The interval from −0.13 to 0.37 days is narrow relative to the margin. The data rule out (at the stated confidence level) a mean excess DSN with Neugranin of more than about 0.37 days, and also rule out a mean advantage for Neugranin larger than 0.13 days.
Why no p-value is needed. The decision rule is the position of the CI relative to the margin. A p-value against a null of zero difference would be beside the point here, and even a non-inferiority p-value would only summarize how far the interval sits from the margin; neither measures the size of the effect. The effect size is the 0.1-day difference and its interval.
Cautions. The analysis used the per-protocol population, which in a non-inferiority trial is often paired with an intention-to-treat analysis because each can be biased in different directions; only the per-protocol analysis is posted. A two-sided 95% CI judged at its upper limit corresponds to a one-sided test at the 2.5% level. Finally, the conclusion is only as credible as the margin of 0.62 days, whose clinical and historical justification is not described in the registry.
6. Open-Label Phase
The open-label Neugranin 40 mg arm also measured cycle 1 DSN using the identical definition. Because the registry states there was no planned statistical analysis for this arm, and because it was neither randomized against a comparator nor masked, its role is descriptive: it shows how DSN behaves with Neugranin outside the blinded setting. Any informal comparison with the double-blind arms would lack the protection of randomization and masking and should not be read as a test of efficacy.
7. Safety: Serious Adverse Events
The registry reports serious adverse events (SAEs) by arm as the number of participants affected over the number at risk.
| Arm | Participants with ≥1 SAE / at risk |
|---|---|
| Double-blind phase: Pegfilgrastim | 7 / 150 |
| Double-blind phase: Neugranin 40 mg | 6 / 153 |
| Open-label phase: Neugranin 40 mg | 6 / 77 |
SAE counts in the two double-blind arms were similar and small. With single-digit event counts, any between-arm comparison has very little precision: a difference of one event can move a proportion noticeably, and confidence intervals for a risk difference would be wide relative to the observed rates. The trial was not powered for safety comparisons, and the open-label arm has a smaller denominator and no randomized comparator, so its counts are not directly comparable.
8. Statistical Methodology
Non-inferiority framework
The null hypothesis in a non-inferiority trial is that the experimental treatment is worse than the comparator by at least the margin. Here, with Δ denoting the difference in mean cycle 1 DSN (Neugranin minus pegfilgrastim, under the orientation implied by the decision rule):
H0 is rejected, and non-inferiority concluded, when the upper limit of the two-sided 95% CI for Δ is ≤ 0.62 days. Observed upper limit: 0.37 days.
Stratified bootstrap confidence interval
The 95% CI for the difference in mean DSN was calculated by stratified bootstrap resampling. In a bootstrap, the observed data are resampled with replacement many times; the difference in means is recomputed in each resample, and the spread of these recomputed differences is used to form the interval (for example from its 2.5th and 97.5th percentiles). "Stratified" means resampling is done within the randomization strata, so that each bootstrap sample preserves the stratum composition of the actual trial.
95% CI ≈ [ 2.5th percentile of Δ*, 97.5th percentile of Δ* ]
The registry does not state the number of bootstrap replicates, the stratification factors, or the specific bootstrap interval variant used.
Per-protocol analysis population
The primary analysis used the per-protocol population, defined in the registry as all data from randomized participants obtained prior to experiencing major protocol violations. This is a data-level rather than participant-level exclusion: participants contribute data up to the point of a major violation. Per-protocol analyses are emphasized in non-inferiority trials because protocol deviations, non-adherence and cross-contamination tend to make two treatments look more similar, which in an intention-to-treat analysis would bias toward a conclusion of non-inferiority.
Multiplicity
Only one formal hypothesis test is posted: the double-blind non-inferiority comparison. The open-label primary outcome had no planned statistical analysis, so there is no competing test that would require splitting the type I error. The registry does not describe any interim analyses or secondary endpoints with formal testing.
9. Statistical Methods Explained
Why was a bootstrap used instead of a standard t-based interval?
DSN is a short, bounded count with many zeros and a long right tail. A t-interval relies on approximate normality of the sample mean; with a few hundred participants per arm this is often adequate, but the bootstrap avoids depending on that approximation and on equal variances. It builds the sampling distribution of the mean difference directly from the observed data, which is attractive when the decision hinges on a single CI limit being compared with a fixed margin.
Why is non-inferiority judged against the margin rather than a p-value?
The claim being tested is "not worse by more than 0.62 days", not "different from zero". The CI shows the full range of mean differences consistent with the data; if every value in that range is below 0.62 days, the trial supports non-inferiority. A conventional p-value for difference would be irrelevant: a large p-value would not demonstrate non-inferiority, and a small one would only show the arms differ.
What does a mean difference of 0.1 days with a CI of −0.13 to 0.37 tell us?
The point estimate is close to zero, and the interval straddles zero, so the trial did not detect a difference in either direction. More importantly for this design, the interval's upper end sits well inside the margin: even the least favourable value compatible with the data (0.37 days) is less than the maximum excess of 0.62 days the investigators deemed acceptable.
Why use a per-protocol population for the primary analysis?
In a superiority trial, the intention-to-treat analysis is conservative because dilution makes it harder to show a difference. In a non-inferiority trial the same dilution makes it easier to show similarity, so intention-to-treat is no longer conservative. A per-protocol analysis restricts attention to data collected while the protocol was being followed, where any real difference is more likely to be visible. Regulators usually expect consistent conclusions from both populations.
Why does a two-sided 95% CI correspond to a one-sided 2.5% test?
Non-inferiority is decided using only the upper limit. A two-sided 95% interval leaves 2.5% in each tail, so checking the upper limit alone controls the one-sided type I error at 2.5%, a common standard for non-inferiority claims.
What does the margin of 0.62 days represent?
The margin is the largest increase in mean cycle 1 DSN with Neugranin that would still be considered clinically acceptable and that preserves a meaningful share of pegfilgrastim's known benefit over no growth factor. It is a design choice made before the trial; the registry states the value but not its derivation, so its clinical justification has to be assessed separately.
10. Limitations
- Per-protocol only: the registry posts the per-protocol analysis; a concordant intention-to-treat analysis, often expected for non-inferiority claims, is not reported there.
- Direction of the difference: the registry lists the groups as "Pegfilgrastim vs Neugranin 40 mg" while the decision rule refers to the upper CI limit; the sign convention is implied rather than stated explicitly.
- Arm-level summaries: ClinicalTrials.gov does not report baseline characteristics, the number of participants in the per-protocol analysis, or the per-arm mean DSN alongside the posted analysis, so the absolute level of severe neutropenia in each arm cannot be described here.
- Margin justification: the clinical and historical basis for the 0.62-day margin is not described in the registry.
- Bootstrap details: the number of resamples, stratification factors and interval type are not reported, which limits reproducibility.
- Single cycle: the primary endpoint covers cycle 1 only; whether results hold over later cycles is not addressed by the posted analysis.
- Surrogate endpoint: DSN is a laboratory-based measure; clinically important outcomes such as febrile neutropenia, infections or chemotherapy dose delays are not part of the posted primary analysis.
- Open-label arm: unmasked and without a planned analysis, it cannot support comparative conclusions.
- Safety precision: SAE counts are small, so between-arm differences are estimated imprecisely.
11. Why This Trial Matters Statistically
NEUGR-003 is a compact teaching example of a modern active-controlled non-inferiority trial with a non-normal continuous endpoint.
| Concept | How it appears in NEUGR-003 |
|---|---|
| Non-inferiority design | Neugranin 40 mg tested against pegfilgrastim with a 0.62-day margin on mean cycle 1 DSN |
| Confidence-interval decision rule | Non-inferiority concluded from the upper CI limit (0.37) relative to the margin |
| Bootstrap | Stratified bootstrap resampling used to form the 95% CI for a skewed count outcome |
| Per-protocol analysis | Primary analysis restricted to data obtained before major protocol violations |
| Masking | Triple masking in the double-blind phase protects a small expected difference from assessment bias |
| Descriptive vs inferential arms | Open-label arm reports the same endpoint with no planned statistical analysis |
| Bounded duration endpoint | DSN is measured in days within a 21-day cycle and is anchored at an ANC threshold |
12. Related Tutorials
Learn more about the methods used in this trial:
13. Related Calculators
14. Sources
- ClinicalTrials.gov: NCT01126190 — Neugranin in Breast Cancer Participants Receiving Doxorubicin/Docetaxel.
- PubMed: PubMed record 24485296.
Explore more trial statistics
Each trial page links its endpoints and methods to tutorials and calculators so the statistical reasoning can be followed step by step.
15. Record Summary
NEUGR-003 compared Neugranin 40 mg with pegfilgrastim in a randomized, triple-masked non-inferiority design. In the per-protocol population, the difference in mean cycle 1 duration of severe neutropenia was 0.1 days (95% CI −0.13 to 0.37), with the upper limit below the prespecified 0.62-day margin. The most useful reading combines the point estimate, the interval relative to the margin, the analysis population, and the justification of the margin itself, while recognising the descriptive role of the open-label arm and the limited precision of the safety counts.