This page separates reported trial information from statistical interpretation. The ClinicalTrials.gov record reports results for the registered objective-response endpoint, but the ClinicalTrials.gov record contains no formal statistical-analysis results with effect estimates, confidence intervals, or p-values.
1. Trial at a Glance
FIR was a phase 2, non-randomized, single-group study of atezolizumab in participants with PD-L1 positive locally advanced or metastatic non-small cell lung cancer. The study enrolled 138 participants across 3 arms.
| Feature | FIR |
|---|---|
| Trial name | FIR |
| ClinicalTrials.gov identifier | NCT01846416 |
| Phase | Phase 2 |
| Condition | Non-Small Cell Lung Cancer |
| Population | Participants with programmed death-ligand 1 (PD-L1) positive locally advanced or metastatic non-small cell lung cancer (NSCLC) |
| Allocation | NON_RANDOMIZED |
| Design model | SINGLE_GROUP |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Enrollment | 138.0 |
| Arms | 3 |
| Intervention | Atezolizumab (MPDL3280A) [TECENTRIQ], an engineered anti-PDL1 antibody |
| Lead sponsor | Genentech, Inc. |
| Sponsor type | INDUSTRY |
| Status | COMPLETED |
2. Clinical Question
The registered study evaluated atezolizumab in participants with PD-L1 positive locally advanced or metastatic NSCLC. Because the study is recorded as NON_RANDOMIZED with a SINGLE_GROUP design and no masking, the statistical question is fundamentally different from that of a randomized two-arm comparative trial.
Population
Participants with programmed death-ligand 1 (PD-L1) positive locally advanced or metastatic non-small cell lung cancer.
Intervention
Atezolizumab (MPDL3280A) [TECENTRIQ], an engineered anti-PDL1 antibody.
Comparator
The ClinicalTrials.gov record does not identify a randomized comparator arm. The study is recorded as non-randomized and single-group.
Primary question
What percentage of participants experienced an objective response according to modified RECIST during the registered assessment schedule?
3. Trial Design
The registry classifies FIR as a phase 2, non-randomized, single-group study with no masking. There were 3 atezolizumab arms in the ClinicalTrials.gov record. The registry enrollment field reports 138.0 participants.
Atezolizumab (MPDL3280A)
- Serious adverse events: 17/31 affected/at risk
- Registry intervention: Atezolizumab (MPDL3280A) [TECENTRIQ]
Atezolizumab (MPDL3280A)
- Serious adverse events: 46/93 affected/at risk
- Registry intervention: Atezolizumab (MPDL3280A) [TECENTRIQ]
Atezolizumab (MPDL3280A)
- Serious adverse events: 6/13 affected/at risk
- Registry intervention: Atezolizumab (MPDL3280A) [TECENTRIQ]
4. Study Timeline
Trial begins
The ClinicalTrials.gov record lists 2013-05-30 as the study start date.
Primary completion
The registry lists 2015-01-07 as the primary completion date.
Registry status
The ClinicalTrials.gov profile lists the overall study status as COMPLETED.
5. Primary Endpoint
| Endpoint | Registry definition | Type | Assessment schedule |
|---|---|---|---|
| Percentage of Participants With Objective Response According to Modified Response Evaluation Criteria in Solid Tumors (RECIST) | Objective response was defined as a complete response (CR) or partial response (PR), as determined by investigator according to modified RECIST criteria. Modified RECIST was derived from RECIST v1.1 conventions and immune related response criteria. CR was defined as disappearance of all tumor lesions (target lesion [TL] and non-target lesion [non-TL]) and no new measurable or unmeasurable lesions, all lymph node short axes must be less than 10 millimeters (mm), and PR was defined as at least 30 percent (%) decrease in sum of diameter of TLs and all new measurable lesions since baseline in absence of CR, and both confirmed by consecutive assessment greater than or equal to 4 weeks from date first documented. Participants not meeting these criteria, including participants without at least one post-baseline response assessment were considered as non-responders. | Binary | Baseline, and Day 1 of Cycle 1 (21-day cycle), then every 6 weeks for the first 12 months and then every 9 weeks thereafter until disease progression (up to 20 months) |
Why the endpoint is binary
At a defined assessment, each participant can be classified according to whether the prespecified objective-response definition is met. The resulting endpoint can therefore be represented as a binary indicator: response versus no response.
For a single-group response endpoint, the central descriptive quantity is the proportion p of evaluable participants meeting the registered response definition. The ClinicalTrials.gov record does not provide the numerical response proportion or its confidence interval.
6. Assessment Schedule and Endpoint Interpretation
The registered assessment schedule begins at baseline and Day 1 of Cycle 1, where a cycle is defined as 21 days. Assessments then occur every 6 weeks for the first 12 months and every 9 weeks thereafter, according to the registry wording.
Repeated assessment
The same participant may have multiple tumor assessments over time. The primary endpoint is nevertheless expressed as a percentage of participants with an objective response.
Investigator determination
The registry-reported definition specifies that CR or PR was determined by the investigator according to modified RECIST criteria.
Modified RECIST
The registry describes modified RECIST as derived from RECIST v1.1 conventions and immune related response criteria.
Time frame
The registered schedule extends from baseline through repeated assessments, initially every 6 weeks and later every 9 weeks.
7. Results and What the Registry Reports
ClinicalTrials.gov reports that results are posted and that 14 outcome measures are posted. However, no formal statistical analyses were posted. Consequently, the ClinicalTrials.gov record does not contain a formal statistical-analysis record with an objective-response estimate, confidence interval, p-value, or treatment-effect comparison.
What can be concluded from the ClinicalTrials.gov record
- The primary endpoint is the percentage of participants with objective response according to modified RECIST.
- The endpoint is classified as binary.
- The endpoint was assessed at baseline and Day 1 of Cycle 1, then every 6 weeks for the first 12 months and every 9 weeks thereafter.
- The ClinicalTrials.gov record contains no formal estimate, confidence interval, or p-value for the primary endpoint.
- The study is non-randomized and single-group, so the registry-reported design does not provide a randomized control-group comparison.
8. Planned Analysis
Because the ClinicalTrials.gov record contains no formal statistical-analysis results, the appropriate educational approach is to distinguish what is registered from what would normally be done for a binary objective-response endpoint.
Descriptive response analysis
For a single-group binary endpoint, the primary descriptive analysis would normally report the number of participants meeting the objective-response definition and divide that count by the relevant analysis population. The resulting percentage estimates the observed objective-response proportion.
The denominator is analytically important. A response percentage based on all enrolled participants is not necessarily interchangeable with a percentage based on an evaluable response population.
Confidence interval for a response proportion
A confidence interval would normally accompany the response proportion to quantify statistical uncertainty. For a binary endpoint, an exact binomial interval or another prespecified binomial confidence-interval method may be used depending on the statistical analysis plan.
The registry-reported FIR data do not specify which confidence-interval method was used, nor do they provide the corresponding numerical interval. It would therefore be inappropriate to assign a particular method to the FIR analysis as though it were documented in the ClinicalTrials.gov record.
What a hypothesis test would require
A formal one-sample hypothesis test would require a prespecified reference response rate or other null hypothesis. No such reference value is included in the ClinicalTrials.gov record. Without that information, a p-value cannot be reconstructed from the ClinicalTrials.gov record.
9. Why a Randomized Treatment Comparison Is Not Available Here
FIR is explicitly classified as NON_RANDOMIZED with a SINGLE_GROUP design. This has major consequences for statistical interpretation.
| Question | What the FIR design can address | What it cannot establish from the registry-reported design alone |
|---|---|---|
| Observed response | How frequently objective response occurred under the registered response definition | Whether response was caused by treatment rather than other factors |
| Comparator difference | No randomized comparator is identified in the ClinicalTrials.gov record | A randomized treatment-versus-control effect |
| Randomization balance | Not applicable to the single-group design | Balance of prognostic factors between randomized groups |
| Causal inference | Descriptive evidence about the treated study population | The causal contrast provided by random assignment |
This does not make the objective-response endpoint statistically uninformative. It changes the question being answered. A single-group response proportion is primarily descriptive unless it is compared against a prespecified external benchmark under an appropriate statistical framework.
10. Statistical Methodology
Binary endpoint analysis
The registered primary endpoint is binary. Each participant's response status can be represented as 1 for objective response and 0 for no objective response, after applying the registered modified-RECIST definition.
The corresponding sample proportion is the number of participants with Y = 1 divided by the relevant denominator.
Binomial framework
If each participant contributes one primary binary response classification and the analysis population is defined in advance, the count of responders can be represented with a binomial model for educational purposes. This provides a natural framework for estimating the response proportion and its uncertainty.
Confidence intervals
A confidence interval around a response proportion addresses sampling uncertainty. It does not describe the range of responses an individual participant might experience, and it does not establish a causal treatment effect by itself.
Hypothesis testing
For a single-group phase 2 response endpoint, a formal hypothesis test would require a prespecified null response rate. The registry-reported FIR data do not identify such a null rate, so no p-value is presented here.
Analysis population
The choice of denominator is a critical part of the analysis. the ClinicalTrials.gov record gives overall enrollment of 138.0 and arm-specific serious-adverse-event denominators of 31, 93, and 13, but it does not supply the numerical denominator used for the objective-response result. A response percentage should therefore not be reconstructed from unrelated safety denominators.
Repeated tumor assessments
The registered assessment schedule creates longitudinal tumor measurements, but the primary endpoint itself is registered as a binary percentage of participants with objective response. The ClinicalTrials.gov record does not state whether additional time-to-response, duration-of-response, or longitudinal modeling methods were used for the primary analysis.
11. Statistical Methods Explained
Why is objective response treated as a binary endpoint?
The registry defines objective response in terms of achieving a complete response or partial response according to modified RECIST. Once the response classification is made, the primary endpoint can be summarized as the percentage of participants meeting that definition. This makes the endpoint naturally suited to binary-proportion analysis.
Why does the denominator matter?
A response percentage is always a numerator divided by a denominator. If the denominator changes, the percentage changes even when the number of responders remains the same. That is why the analysis population must be defined before interpreting a response rate.
What would a confidence interval tell us?
A confidence interval would quantify statistical uncertainty around the estimated response proportion under the chosen statistical framework. A narrower interval generally indicates greater precision, while a wider interval indicates greater uncertainty. It would not establish causation.
Why is a p-value not automatically required for a response rate?
A descriptive response proportion answers "what percentage responded?" A p-value answers a different question involving a prespecified null hypothesis. Without a documented reference response rate or another null hypothesis, there is no basis for presenting a one-sample p-value.
Why is this not the same as a randomized treatment effect?
Randomization creates a comparison in which treatment assignment is not determined by baseline prognostic characteristics. FIR is recorded as non-randomized and single-group, so its observed response proportion does not by itself provide the randomized contrast between two treatment strategies.
Why should repeated assessments be interpreted carefully?
Participants can undergo multiple assessments over time, but the primary endpoint is a participant-level objective-response classification. Repeated measurements can affect how response is determined and how missing or unevaluable assessments are handled. The ClinicalTrials.gov record does not specify the detailed missing-data or confirmation rules beyond the available registry definition.
12. Safety Results
the ClinicalTrials.gov record reports serious adverse events by arm using affected/at-risk counts. These are reported below exactly as provided.
| Atezolizumab arm | Serious adverse events affected | At risk | Reported affected/at risk |
|---|---|---|---|
| 1L Participan | 17 | 31 | 17/31 |
| 2L+ Participa | 46 | 93 | 46/93 |
| 2L+ Brain Met | 6 | 13 | 6/13 |
These serious-adverse-event counts are safety observations rather than efficacy estimates. The ClinicalTrials.gov record does not provide statistical-analysis entries comparing safety rates between the arms, nor do they provide confidence intervals or p-values for these safety measures.
13. Interpreting the Three Study Arms
The registry-reported profile identifies 3 arms, all under the atezolizumab intervention, with arm-specific serious-adverse-event counts labeled as 1L Participan, 2L+ Participa, and 2L+ Brain Met. Because the registry-reported statistical-analysis array is empty, no formal comparative efficacy analysis among these groups is reported here.
1L Participan
Serious adverse events affected 17 of 31 participants in the ClinicalTrials.gov record.
2L+ Participa
Serious adverse events affected 46 of 93 participants in the ClinicalTrials.gov record.
2L+ Brain Met
Serious adverse events affected 6 of 13 participants in the ClinicalTrials.gov record.
Interpretive boundary
The ClinicalTrials.gov record does not provide arm-specific objective-response estimates or formal between-arm statistical comparisons.
14. Missing Statistical Details
Several statistical features commonly included in a detailed clinical-trial report are not present in the registry-reported FIR trial data. These omissions matter because they limit how far the statistical interpretation can go.
| Statistical feature | Supplied FIR information | Interpretive consequence |
|---|---|---|
| Formal efficacy estimate | Not reported | No objective-response percentage can be reported from the ClinicalTrials.gov record. |
| Confidence interval | Not reported | Precision of the response estimate cannot be quantified from the ClinicalTrials.gov record. |
| P-value | Not reported | No formal hypothesis-test result can be reported. |
| Null response rate | Not reported | A one-sample hypothesis test cannot be reconstructed. |
| Missing-data method | Not reported | The effect of missing or unevaluable tumor assessments cannot be quantified. |
| Multiplicity procedure | Not reported | No multiplicity-adjusted interpretation can be assigned. |
| Interim-analysis method | Not reported | No interim alpha-spending or stopping-rule analysis can be described as documented. |
| Bayesian methods | Not reported | No Bayesian posterior analysis should be inferred. |
| Stratification factors | Not reported | No stratified efficacy model can be attributed to the registry record. |
This distinction is important for reproducibility. Statistical methods should be attributed to the trial only when they are documented in the ClinicalTrials.gov record. Otherwise, they are best presented as general methods that would commonly be considered for the endpoint.
15. Non-Randomization and Causal Interpretation
The non-randomized design is one of the most important statistical characteristics of FIR. In a randomized trial, the treatment contrast is generated by random assignment. In a single-group study, there is no randomized concurrent control against which to calculate a treatment effect.
An observed objective-response proportion describes the frequency of response among participants analyzed under the study's response definition. It does not, by itself, quantify the difference between atezolizumab and another treatment.
If a phase 2 single-group study were designed around a prespecified historical or external response benchmark, the response proportion could potentially be tested against that benchmark. Such a comparison requires the benchmark, its justification, and the prespecified statistical framework. None of those quantities is reported here.
Without randomization, differences between the study population and an external population can affect observed response rates. Even a precise estimate of the observed proportion would not remove this design limitation.
16. Clinical Biostats Interpretation of the Primary Endpoint
The primary endpoint is conceptually straightforward but statistically dependent on several layers of definition: who is included in the denominator, how objective response is determined, when tumor assessments occur, and how incomplete assessments are handled.
| Layer | Question | Why it matters |
|---|---|---|
| Population | Which participants contribute to the analysis? | Determines the denominator of the response percentage. |
| Response definition | What qualifies as CR or PR? | Determines which participants are classified as responders. |
| Assessment timing | When is response evaluated? | Determines when a participant can meet the endpoint. |
| Missing assessments | What happens when response cannot be determined? | Can materially change the observed response proportion. |
| Inference | What uncertainty or hypothesis is being tested? | Determines how the response percentage should be interpreted statistically. |
| Design | Is there a randomized concurrent comparator? | Determines whether a causal comparative treatment effect can be estimated. |
The registry-reported FIR data answer some of these questions, particularly the response definition and assessment schedule, but do not provide enough information to reconstruct the full inferential analysis.
17. Limitations
- No randomized comparator: the registry classifies the study as non-randomized and single-group, limiting causal comparative interpretation.
- No formal statistical-analysis results reported: no formal statistical analyses were posted, so no numerical efficacy estimate, confidence interval, or p-value is reported here.
- Incomplete endpoint definition in the ClinicalTrials.gov record: the registry definition ends at "no new measurable or unmeasurable lesions, all lymph node short axes must be less than 10 millimeters (mm), and PR was defined as at least 30 percent (%) decrease in sum of diameter of TLs and all new measurable lesions since baseline in absence of CR, and both confirmed by consecutive assessment greater than or equal to 4 weeks from date first documented. Participants not meeting these criteria, including participants without at least one post-baseline response assessment were considered as non-responders.", so the remainder is not reconstructed from outside information.
- Unknown analysis denominator: overall enrollment is 138.0, while the registry-reported serious-adverse-event denominators are 31, 93, and 13. The denominator for objective response is not reported.
- Missing-data methods are not reported: the ClinicalTrials.gov record does not identify how unevaluable or missing tumor assessments were handled.
- No documented null response rate: a formal one-sample p-value cannot be reconstructed without a prespecified reference value or other null hypothesis.
- No documented multiplicity strategy: the ClinicalTrials.gov record does not describe multiplicity adjustment across the 14 posted outcome measures.
- No documented subgroup efficacy results: the three safety denominators do not provide sufficient information to infer arm-specific response rates.
- Registry-level evidence: the page is constrained to the ClinicalTrials.gov record and the linked publication identifier rather than reconstructing methods from additional publications.
18. Why This Trial Matters Statistically
FIR is a useful teaching case because it illustrates a common phase 2 statistical structure: a binary response endpoint evaluated in a non-randomized study. The central statistical task is therefore different from the hazard-ratio-centered analysis of a randomized survival trial.
| Concept | How it appears in FIR |
|---|---|
| Binary endpoint | Percentage of participants with objective response. |
| Response definition | CR or PR according to investigator-determined modified RECIST. |
| Repeated assessment | Baseline and Day 1 of Cycle 1, then every 6 weeks for the first 12 months and every 9 weeks thereafter. |
| Single-group design | Registry design model is SINGLE_GROUP. |
| Non-randomization | Allocation is NON_RANDOMIZED. |
| Phase 2 development | Trial phase is 2. |
| Safety denominators | Serious-adverse-event affected/at-risk counts are reported separately for the three named arms. |
| Statistical reporting boundary | Results are posted, but no formal statistical analyses were posted. |
19. Statistical Methods Explained in More Depth
Observed response versus treatment effect
An observed response proportion is a description of what happened in the study population. A treatment effect is a comparison between what happened under one treatment and what would have happened under a specified alternative. Because FIR is non-randomized and single-group, the ClinicalTrials.gov record supports the former but do not provide the latter.
Precision versus validity
A large sample can produce a relatively precise estimate of a response proportion, but precision does not correct a design's lack of randomization. Conversely, a randomized estimate can have substantial uncertainty when the sample is small. Statistical precision and causal validity are distinct properties.
Why confidence intervals should accompany proportions
A point estimate alone does not show how much sampling uncertainty surrounds it. A confidence interval places the estimate in an inferential context. For a binary endpoint, the interval is naturally linked to binomial sampling assumptions or another prespecified method for estimating uncertainty.
Why response assessment timing matters
The endpoint is not simply a count of tumor measurements. The registry specifies a schedule beginning at baseline and continuing at defined intervals. The timing determines when response can be observed and can influence how rapidly or persistently responses are detected.
Why safety and efficacy use different denominators
The registry-reported serious-adverse-event data use arm-specific "affected/at risk" counts. An efficacy endpoint may use a different analysis population. Mixing those denominators would produce a statistic that does not correspond to either endpoint.
20. Related Statistical Concepts
Learn more about the methods used in this trial:
21. Related Statistical Calculators
22. Sources
- ClinicalTrials.gov: FIR — NCT01846416.
- Linked publication: PubMed record for PMID 33241650: PubMed PMID 33241650.
Continue through Clinical Biostats
Explore statistical tutorials, calculators, and clinical trial analyses covering the methods used to design, analyze, and interpret biomedical studies.
23. Record Summary
FIR is a completed phase 2, non-randomized, single-group study of atezolizumab in participants with PD-L1 positive locally advanced or metastatic non-small cell lung cancer. The registered primary endpoint is the percentage of participants with objective response according to modified RECIST, assessed from baseline and Day 1 of Cycle 1 through repeated follow-up assessments. The ClinicalTrials.gov record reports that results are posted, but the accompanying statistical-analysis array contains no formal efficacy estimates, confidence intervals, or p-values. The appropriate statistical interpretation is therefore centered on the binary response endpoint, the importance of the analysis denominator and confidence interval, and the limitations of drawing causal conclusions from a non-randomized single-group design.