← Clinical Trials
Cervical Cancer Phase 1 Randomized NCT03738228

NRG-GY017: Complete Statistical Analysis of Atezolizumab in Cervical Cancer

An independent statistical analysis of the randomized phase 1 NRG-GY017 trial evaluating atezolizumab before and/or with chemoradiotherapy in patients with node-positive stage IB2, II, IIIB, or IVA cervical cancer.

Trial status: Completed  ·  Study start: January 7, 2019  ·  Primary completion: May 1, 2022
ClinicalTrials.gov record

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

NRG-GY017 was a randomized, parallel, open-label phase 1 treatment study involving 40 participants with node-positive stage IB2, II, IIIB, or IVA cervical cancer. The study evaluated atezolizumab before and/or with chemoradiotherapy, with immune response defined through T-cell receptor beta clonal expansion in peripheral blood.

40
Enrollment
Total participants
2
Arms
Randomized parallel design
1
Phase
Phase 1
2022
Primary completion
May 1, 2022
FeatureNRG-GY017
PhasePhase 1
StatusCompleted
PopulationPatients with node-positive stage IB2, II, IIIB, or IVA cervical cancer
ConditionsCervical adenocarcinoma; cervical adenosquamous carcinoma; cervical squamous cell carcinoma; stage IB2, II, IIA, IIA1, IIA2, IIB, IIIB, or IVA cervical cancer according to AJCC v8
DesignRandomized, parallel, open-label
AllocationRandomized
Primary purposeTreatment
Number of arms2
Enrollment40
Lead sponsorNational Cancer Institute (NCI)
Sponsor typeNIH
ClinicalTrials.govNCT03738228

2. Clinical Question

The registry describes NRG-GY017 as a study of atezolizumab before and/or with chemoradiotherapy in patients with node-positive cervical cancer. The principal statistical question is therefore centered on whether the randomized study groups differ in the specified measure of immune response after exposure to atezolizumab in the context of the study treatment strategy.

Population

Patients with cervical adenocarcinoma, cervical adenosquamous carcinoma, or cervical squamous cell carcinoma meeting the registry's node-positive stage criteria: IB2, II, IIIB, or IVA, with additional stage II subcategories listed in the record.

Intervention framework

Atezolizumab was studied in conjunction with a treatment program that included radiation therapy, brachytherapy, and cisplatin.

Comparator structure

The study used 2 randomized parallel arms. The registry describes the study as evaluating atezolizumab before and/or with chemoradiotherapy.

Primary question

How does the specified immune response, measured through total TCRB clonal expansion, compare across the randomized study groups at the prespecified assessment windows?

3. Trial Design

01
Enroll40 participants
02
Randomize2 parallel arms
03
TreatAtezolizumab and chemoradiotherapy framework
04
AssessTCRB clonal expansion
05
CompletePrimary completion May 1, 2022
Allocation
Randomized allocation to 2 parallel study arms.
Masking
None. The study was open-label.
Primary purpose
Treatment.
Study phase
Phase 1.

Study interventions

The registry lists four interventions: atezolizumab, brachytherapy, cisplatin, and radiation therapy. The record describes the overall study as evaluating atezolizumab before and/or with chemoradiotherapy. The registry information available here does not provide additional arm-specific dosing or scheduling details.

Open-label design: masking was listed as none. That is particularly relevant when interpreting subjective or behavior-sensitive outcomes, although the primary endpoint in this study is a laboratory-based immune-response measure.

4. Trial Population and Statistical Context

The enrolled population consisted of 40 participants. Because this was a phase 1 randomized study, the sample size is modest for between-group statistical estimation. A small randomized sample can be useful for characterizing an experimental biological signal, but estimates from such a study are generally less precise than estimates from a substantially larger confirmatory trial.

Population featureRegistry information
Total enrollment40
Study arms2
AllocationRandomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Study phasePhase 1

Randomization is statistically important even in a small phase 1 study because it creates a defined comparison between study groups. However, randomization does not make a small sample statistically equivalent to a large one: uncertainty around estimated treatment differences can remain substantial.

5. Primary Endpoint

EndpointRegistry definitionTime frame
Immune Response The immune response is measured by total T cell receptor beta (TCRB) clonal expansion in peripheral blood at day 21 from baseline using Adaptive Biotechnologies' immunoSEQ platform. The registry states that a higher number of total TCR clonal expansion indicates better immune response. Arm A: 42 days from the first dose of Atezolizumab

Arm B: 21 days from the first dose of Atezolizumab

How the endpoint is constructed

The endpoint is based on total TCRB clonal expansion measured in peripheral blood. The registry describes the measurement using the immunoSEQ platform and anchors the immune-response assessment to day 21 from baseline.

The registry also gives different overall time frames for the two randomized groups: 42 days from the first dose of atezolizumab for Arm A and 21 days from the first dose for Arm B. Those different windows are important when interpreting the comparison because timing is part of the endpoint definition.

Measurement concept
Immune response  ∝  total TCRB clonal expansion

The registry explicitly states that a higher number of total TCR clonal expansion indicates a better immune response. The endpoint is therefore directional: larger measured expansion corresponds to a more favorable immune-response interpretation under the registry definition.

6. Planned Analysis

Results status: The ClinicalTrials.gov record indicates that results are posted, but no formal statistical analyses are posted in the registry information reviewed here. The primary endpoint is therefore described statistically below without presenting an estimated treatment effect, confidence interval, or p-value.

For a continuous laboratory-based immune-response endpoint such as total TCRB clonal expansion, a typical randomized comparison would begin by describing the distribution within each arm and then estimating the between-arm difference. If the endpoint were analyzed as a continuous change from baseline, a model such as an ANCOVA could compare treatment groups while adjusting for the baseline measurement. Alternatively, a two-sample comparison of changes could be used when its assumptions are appropriate.

The most important issue is that the statistical scale must match the endpoint actually analyzed. If clonal-expansion measurements are strongly skewed, a transformation such as the logarithm may be considered before modeling. If the distribution remains unsuitable for a standard parametric model, a rank-based comparison could be considered. These are general statistical approaches for this type of endpoint; the registry information does not identify which of them was used for the posted results.

Baseline adjustment

When a continuous biological measurement is available both before and after treatment, baseline adjustment can improve precision by accounting for differences in starting values. An ANCOVA framework would typically model the follow-up measurement as a function of treatment group and baseline measurement.

Illustrative statistical structure
Follow-up TCRB expansion = treatment group + baseline TCRB expansion + error

This equation describes a common analysis framework for a continuous randomized endpoint. It is an educational representation of a possible analysis, not a claim that this exact model was specified or used in the registry.

Effect estimate

For this endpoint, the treatment contrast could be expressed as an adjusted mean difference, an unadjusted mean or median difference, or another prespecified scale depending on the distribution and statistical analysis plan. The choice of effect measure matters because it determines how the magnitude of an immune-response difference is communicated.

7. Statistical Methods Explained

Why does randomization matter for an immune-response endpoint?

Randomization creates the comparison between the two study groups before the outcome is observed. In principle, this reduces systematic allocation differences as an explanation for a later difference in TCRB clonal expansion. With only 40 enrolled participants, however, randomization cannot guarantee that every baseline characteristic is numerically balanced.

Why is a continuous endpoint different from a survival endpoint?

Total TCRB clonal expansion is a laboratory measurement rather than a time-to-event outcome such as death or progression. Kaplan-Meier curves and Cox proportional-hazards models are therefore not the natural primary tools for the endpoint as it is defined in the registry. The central comparison is the distribution of the immune-response measurement between randomized groups.

Why might baseline adjustment be useful?

Suppose two patients begin with very different levels of TCRB activity. Comparing only their follow-up values can mix treatment effects with pre-existing differences. An analysis that incorporates baseline measurements can separate the treatment-group comparison from variation that was already present before treatment.

What does a higher total TCR clonal expansion mean?

According to the registry definition, a higher number of total TCR clonal expansion indicates a better immune response. This is a statement about the biological endpoint, not a direct statement that a patient lived longer, had better tumor control, or experienced a particular clinical outcome.

Why do the different assessment windows matter?

The registry specifies 42 days from the first dose of atezolizumab for Arm A and 21 days from the first dose for Arm B. A comparison between groups therefore has to respect the prespecified timing. Measurements obtained at different treatment intervals can represent different biological stages, so the timing cannot simply be ignored when interpreting the result.

Why does a small phase 1 sample require caution?

With 40 enrolled participants, random variation can have a substantial effect on estimated between-group differences. Confidence intervals, when available from a formal analysis, are therefore especially important because they communicate the precision of the estimated effect rather than only whether a hypothesis test crossed a threshold.

8. Statistical Interpretation of the Primary Endpoint

What the endpoint can establish

A comparison of total TCRB clonal expansion can describe whether the randomized study groups differed in the specified laboratory measure of immune response. The registry explicitly defines greater total TCR clonal expansion as indicating better immune response.

What the endpoint does not establish by itself

A higher immune-response measurement is not itself a measure of overall survival, progression-free survival, tumor response, or cure. Those are distinct clinical outcomes and cannot be substituted for the TCRB endpoint.

Why the effect scale matters

If the analysis reports a difference in TCRB clonal expansion, the numerical scale should be interpreted in the units of the underlying laboratory measurement. A statistically detectable difference and a clinically meaningful difference are not automatically the same thing.

Why a p-value would not measure effect size

If a formal analysis reports a p-value, that value would address evidence against a specified null hypothesis under the statistical model. It would not tell the reader how large the immune-response difference is. The effect estimate and its confidence interval provide the information about magnitude and precision.

9. Analysis Population and Missing Data Considerations

The registry identifies the enrollment of 40 participants but does not provide an analysis-population definition for the primary immune-response endpoint in the information summarized here. That distinction matters because a randomized analysis, an evaluable-patient analysis, and an analysis restricted to participants with usable laboratory measurements can produce different denominators and different uncertainty.

Laboratory endpoints can also be affected by missing specimens or measurements that fail quality-control requirements. The statistical consequence depends on why measurements are missing and how the prespecified analysis handles them. The registry information does not specify an imputation method for the immune-response endpoint.

Randomized population

Randomization establishes the original comparison between the 2 study arms.

Endpoint-evaluable population

A laboratory endpoint may require an available and valid TCRB measurement. The registry information does not define a separate evaluable population here.

Missing measurements

The registry information does not specify an imputation strategy for missing TCRB measurements.

Interpretation

The denominator used in any formal analysis should therefore be read together with the endpoint definition and analysis population.

10. Safety Results

The registry reports serious adverse events by arm using affected participants over participants at risk. The reported counts are:

ArmParticipants with serious adverse eventsParticipants at risk
Arm A619
Arm B717
Serious adverse events by arm
Arm A
6 / 19
Arm B
7 / 17

The reported denominators for serious adverse events are 19 in Arm A and 17 in Arm B. They do not equal the overall enrollment of 40, so the safety result should be interpreted using the registry's stated at-risk denominators rather than substituting the total enrollment.

How to interpret the safety counts

The registry reports 6 affected participants among 19 at risk in Arm A and 7 among 17 at risk in Arm B. These are counts of participants with serious adverse events, not counts of individual adverse-event episodes. They also should not be interpreted as evidence of a difference in overall safety without considering the underlying exposure, event definitions, follow-up, and statistical analysis.

11. Timing and Longitudinal Structure

2019-01-07

Study start

The registry lists January 7, 2019 as the study start date.

Primary endpoint timing

Immune-response assessment

The endpoint is anchored to day 21 from baseline for the TCRB clonal-expansion measurement, with the registry specifying 42 days from the first dose of atezolizumab for Arm A and 21 days for Arm B.

2022-05-01

Primary completion

The registry lists May 1, 2022 as the primary completion date.

Current registry status

Completed

The ClinicalTrials.gov record identifies the study status as completed.

12. Design Features That Affect Statistical Interpretation

Randomized parallel design

The study was randomized and parallel, so each participant was assigned to one of two study arms rather than being repeatedly assigned to both treatment strategies. This provides a direct between-group framework for the primary immune-response comparison.

Open-label treatment

Masking was listed as none. For a laboratory endpoint such as TCRB clonal expansion, the lack of masking may be less consequential than it would be for a subjective clinical assessment, but it remains a design characteristic that should be recorded when interpreting the trial as a whole.

Phase 1 context

The phase 1 designation places the study in an early clinical-development setting. Statistical interpretation should therefore emphasize estimation, biological signal characterization, uncertainty, and the relationship between the immune endpoint and later clinical-development questions rather than treating the study as equivalent to a large confirmatory efficacy trial.

No reported non-inferiority framework

The registry information does not identify a non-inferiority hypothesis or margin. The primary endpoint is an immune-response measure rather than a registry-defined non-inferiority outcome.

No reported crossover design

The registry information does not identify crossover between randomized study arms. The design is described as parallel.

No reported factorial design

The registry identifies 2 randomized parallel arms and does not describe a factorial allocation structure.

No reported Bayesian analysis

The registry information does not identify a Bayesian statistical method for the primary endpoint.

Multiplicity

The registry identifies one registered primary endpoint: immune response. No additional primary endpoints or multiplicity-adjustment strategy are identified in the information summarized here.

13. A Statistical Framework for TCRB Clonal Expansion

Total TCRB clonal expansion is a quantitative biological measurement. The first statistical task is to characterize its distribution. That includes examining central tendency, variability, and the presence of unusually large or small observations. These steps matter because immune-repertoire measurements can behave differently from approximately symmetric clinical laboratory variables.

Location

The mean or median describes the typical level of clonal expansion, depending on the distribution and prespecified analysis.

Variation

Standard deviation, interquartile range, or another dispersion measure describes how heterogeneous the measurements are within an arm.

Between-arm effect

A treatment contrast quantifies how the immune-response distributions differ between randomized groups.

Precision

A confidence interval communicates the statistical uncertainty surrounding the estimated treatment contrast.

Difference versus ratio

Two common effect scales for quantitative outcomes are a difference and a ratio. A difference preserves the original measurement scale. A ratio instead expresses one group's value relative to another group's value. Which scale is most informative depends on the biological interpretation and the prespecified statistical model.

For skewed biological measurements, a logarithmic transformation can make a ratio interpretation especially natural because differences on the log scale correspond to multiplicative differences on the original scale. The registry information does not specify that such a transformation was used.

14. Why This Trial Matters Statistically

NRG-GY017 is a useful statistical teaching case because it links randomized trial methodology to a biological endpoint that is different from the time-to-event outcomes commonly seen in oncology trials. The central challenge is not simply whether the two arms differ numerically, but how to define, measure, compare, and interpret a quantitative immune-response endpoint within a small randomized phase 1 design.

Statistical conceptHow it appears in NRG-GY017
RandomizationParticipants were randomized to 2 parallel study arms.
Continuous endpointImmune response is measured using total TCRB clonal expansion.
Baseline timingThe endpoint is defined around day 21 from baseline.
Unequal assessment windowsThe registry specifies 42 days from the first atezolizumab dose for Arm A and 21 days for Arm B.
Laboratory measurementTCRB clonal expansion is measured in peripheral blood using the immunoSEQ platform.
Small randomized sampleTotal enrollment was 40 participants.
Safety denominatorSerious adverse events are reported as affected participants over participants at risk by arm.
Open-label designMasking was listed as none.
Phase 1 developmentThe trial is registered as phase 1.

15. Important Limitations and Interpretation Issues

16. What a Complete Statistical Report Would Need to Show

For a quantitative primary endpoint such as total TCRB clonal expansion, a statistically complete results presentation would ideally identify the analysis population, the number of observations contributing to the endpoint, the distribution within each randomized arm, the treatment contrast, its confidence interval, and the hypothesis-testing framework if one was prespecified.

ComponentWhy it matters
Analysis denominatorDefines which participants contributed to the primary endpoint analysis.
Arm-specific distributionShows the magnitude and variability of immune-response measurements.
Effect estimateQuantifies the between-arm difference on the chosen statistical scale.
Confidence intervalCommunicates precision around the estimated effect.
P-value, if prespecifiedAddresses evidence against a specified null hypothesis; it is not a measure of effect size.
Analysis modelAllows readers to understand how baseline values, distributional assumptions, and other covariates were handled.
Missing-data handlingShows how incomplete laboratory measurements were incorporated into the analysis.

These elements are particularly important in a small phase 1 randomized study. A numerical difference between groups is only one part of the statistical story; the uncertainty around that difference and the population from which it was estimated are equally important.

17. Statistical Concepts in This Trial

18. Sources

Continue through Clinical Biostats

Explore additional clinical-trial analyses and statistical educational resources.

19. Record Summary

NRG-GY017 is a completed randomized phase 1 study with 40 enrolled participants and 2 parallel, open-label arms. The registry describes atezolizumab before and/or with chemoradiotherapy in patients with node-positive cervical cancer and identifies immune response as the registered primary endpoint. That endpoint is defined through total TCRB clonal expansion in peripheral blood, measured using the immunoSEQ platform, with the registry specifying day 21 from baseline as the measurement point and different overall time frames from the first atezolizumab dose for the two arms.

Statistically, the study illustrates the distinction between a randomized comparison and the interpretation of a quantitative biological endpoint. The appropriate analysis must account for the endpoint's measurement scale, baseline values, variability, assessment timing, analysis population, and missing observations. The registry reports serious adverse events by arm as 6 of 19 participants in Arm A and 7 of 17 participants in Arm B, providing a separate safety perspective from the immune-response endpoint.

Clinical Biostats methodology: The statistical interpretation of an early randomized trial should distinguish the biological measurement being studied from downstream clinical outcomes. For NRG-GY017, total TCRB clonal expansion is the registered primary endpoint, while safety is reported separately through serious adverse events by study arm.