A Study of Atezolizumab in Participants With Programmed Death - Ligand 1 (PD-L1) Positive Locally Advanced or Metastatic Non-Small Cell Lung Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study of Atezolizumab in Participants With Programmed Death - Ligand 1 (PD-L1) Positive Locally Advanced or Metastatic Non-Small Cell Lung Cancer (NCT02031458), a na phase 2 study. The primary outcome is Percentage of Participants Achieving Objective Response (ORR) Per Response Evaluation Criteria In Solid Tumors (RECIST) Version (v) 1.1 as Assessed by Independent Review Facility (IRF).

2. Clinical Question

The study asked whether participants with PD-L1-positive locally advanced or metastatic non-small cell lung cancer (NSCLC) treated with atezolizumab achieved an objective response, and it also set out to evaluate safety. It was a Phase II, multicenter, single-arm study: all participants received the study drug, and there was no concurrent control group. Accordingly, the main efficacy question concerns the proportion of participants with a confirmed response under the specified assessment criteria, rather than a randomized comparison between treatments.

The primary endpoint is a response proportion, not a direct measure of how outcomes compare with another treatment or with no treatment. The registry record also lists outcomes based on investigator assessments, modified RECIST, duration of response, and progression-free survival. These outcomes broaden the study's descriptive scope, but their definitions alone do not establish a treatment effect. The numerical primary endpoint result is not reported in the registry record.

3. Trial Design

This was a Phase II, multicenter, single-arm study of atezolizumab in people with PD-L1-positive locally advanced or metastatic NSCLC. The registry record lists enrollment of 667 and identifies the study as completed. The study started on 2014-01-22, and primary completion was recorded as 2015-05-28. Hoffmann-La Roche is listed as the sponsor.

The intervention was atezolizumab 1200 mg given intravenously every 3 weeks. Treatment was described as continuing for as long as the participant was experiencing clinical benefit as assessed by the investigator, provided there was no unacceptable toxicity or symptomatic deterioration attributed to disease progression. The registry record lists masking as none and allocation as not applicable. These features are consistent with a design in which participants received the intervention without a concurrent randomized control group.

The primary endpoint was the percentage of participants achieving objective response, assessed by an IRF under RECIST v1.1. The endpoint description specifies a confirmed best overall response and identifies the efficacy evaluable population as the population analyzed. It also says results were reported by line of therapy and PD-L1 expression subgroup. The analyzed participant counts for those groups are not reported in the registry record.

The record lists secondary outcomes for objective response using investigator assessment and modified RECIST, along with duration of response and progression-free survival assessed by the IRF or investigator under the stated criteria. A single-arm study can describe outcomes among participants receiving the intervention, but without a concurrent control it cannot isolate the intervention's effect from other explanations for the observed outcomes. The registry record does not report a randomized comparison or a control group.

Although enrollment is recorded as 667, this is not identified as the denominator for every outcome. The primary endpoint description defines its population as efficacy evaluable and says that the number analyzed equals the number analyzed within the specified group; it does not provide those group counts. The record therefore does not establish that every enrolled participant contributed to every endpoint.

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: Screening was performed from Day -28 to Day -1.

May 8, 2015 to January 11, 2019

Reasons for not completing this period:

Baseline to May 8, 2015

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Efficacy evaluable population; Number (n) equals (=) number of participants analyzed within the specified group.

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.

Population: Efficacy evaluable population included all enrolled participants who received any dose of the study drug (atezolizumab) during the study treatment period. Since the primary CSR one patient in Cohort 1 has moved to Cohort 2 and one patient in Cohort 3 was moved to Cohort 2.

6. Primary Endpoint

The registry lists the primary outcome as Percentage of Participants Achieving Objective Response (ORR) Per Response Evaluation Criteria In Solid Tumors (RECIST) Version (v) 1.1 as Assessed by Independent Review Facility (IRF), measured over Screening, Every 6 weeks (± 3 days) for 12 months following Cycle 1, Day 1 and every 9 weeks (± 1 week) thereafter until disease progression, intolerable toxicity or death until data cut-off on 28 May 2015 (Up to 16 months).

The primary endpoint is the percentage of participants whose confirmed best overall response was complete response or partial response, based on IRF assessment under RECIST v1.1. A complete response requires disappearance of all target and non-target lesions; pathological lymph nodes must have a short axis < 10 millimeters (mm). A partial response requires a > or equal to 30 percent decrease from baseline in the sum of target-lesion diameters, non-progressive disease in non-target lesions, and no new lesions.

Objective response is used to describe tumor response according to prespecified criteria. Combining complete and partial responses into a binary endpoint makes the outcome a proportion: participants either meet the response definition or do not. The registry record specifies a 95% confidence interval for this percentage, but the numerical percentage and interval are not reported in the registry record.

Assessments were scheduled from screening, every 6 weeks (± 3 days) for 12 months following Cycle 1, Day 1, and every 9 weeks (± 1 week) thereafter until disease progression, intolerable toxicity or death, with data cut-off on 28 May 2015 (up to 16 months). The endpoint is reported by line of therapy and PD-L1 expression subgroup. These features describe how response was defined and assessed; they do not, by themselves, show how many participants responded.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is a binary response endpoint summarized as a percentage. The registry record defines a responder as a participant whose confirmed best overall response was complete response or partial response by IRF assessment under RECIST v1.1. A participant who does not meet either response category is not counted as an objective responder for this endpoint. The record identifies the efficacy evaluable population as the analysis population, but does not report the number analyzed or the numerical response percentage.

What a response percentage describes

General statistical context: a proportion describes the share of an analyzed group meeting a defined outcome. Its interpretation depends on both the outcome rule and the denominator. Here, the outcome rule is specified, but the registry record does not give the analyzed denominator or the number of responders. Enrollment of 667 should not automatically be substituted for the efficacy evaluable denominator, because the endpoint description specifies an efficacy evaluable population and subgroup-specific analyzed numbers are not reported.

The registry record specifies a 95% confidence interval for the primary percentage. General statistical context: an interval estimate expresses the uncertainty associated with an estimated quantity under a statistical method and its assumptions. A narrower interval generally indicates greater precision than a wider one, but interval width depends on the data and method. The interval is not reported numerically, and its construction method is not reported in the registry record. Thus, neither its width nor the precision of the response estimate can be assessed.

Assessment criteria and timing

RECIST v1.1 supplies operational rules for classifying tumor response. In this record, complete response requires disappearance of target and non-target lesions, with pathological lymph nodes reduced to a short axis < 10 mm. Partial response requires a > or equal to 30 percent decrease from baseline in the sum of target-lesion diameters, non-progressive disease in non-target lesions, and no new lesions. The endpoint requires a confirmed best overall response, so a response category is not simply a single unconfirmed observation. The registry does not describe the confirmation procedure beyond this endpoint wording.

Assessments were scheduled at screening and then at stated intervals: every 6 weeks (± 3 days) for 12 months following Cycle 1, Day 1, and every 9 weeks (± 1 week) thereafter until disease progression, intolerable toxicity or death, through the data cut-off on 28 May 2015 (up to 16 months). General statistical context: the timing and completeness of repeated assessments can affect which response observations are available and how best overall response is determined. The registry record does not explain how missed visits, unevaluable scans, or incomplete follow-up were handled.

Independent and investigator assessment

The primary response endpoint uses assessment by an IRF. Secondary outcome descriptions include objective response assessed by the investigator, including under modified RECIST. They also list duration of response and progression-free survival under IRF or investigator assessment. These are different outcome definitions or assessment sources, not interchangeable readings of one result. General statistical context: independent assessment can reduce reliance on treating-team judgment, while investigator assessment reflects the investigator's classification. The registry record does not report the numerical results needed to compare these assessment approaches, nor does it describe blinding or adjudication procedures for the IRF.

Subgroups and outcome breadth

The primary outcome description says results were reported by line of therapy and PD-L1 expression subgroup: TC3 or IC3, TC3 or IC2/3, and TC2/3 or IC2/3. The record also identifies first-line, second-line, and third-line-and-beyond cohorts in the safety information. Subgroup estimates can describe whether outcomes differ across defined groups, but they generally involve smaller denominators than an overall estimate. The registry record does not report the analyzed subgroup counts or subgroup response estimates. It also does not state whether subgroup analyses were prespecified or how subgroup comparisons were statistically handled.

The registry record lists 23 outcome measures and 0 analyses. It therefore gives outcome definitions but does not provide inferential analysis outputs in this record. Statistical analysis details such as the confidence-interval construction method, missing-data rules, and any adjustment for multiple outcome testing are not reported in the registry record. They should not be inferred from the endpoint titles.

Planning illustrations are not trial design results

The planning illustration assumptions are explicitly distinct from the trial's own design. They describe a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation, with the effect to detect expressed as a stated fraction of the observed effect. For time-to-event outcomes, the illustration concerns total events; for other outcomes, it concerns participants per arm. These assumptions do not turn this single-arm study into a two-arm comparison. A planning illustration is not observed power, and the registry record does not report the study's own sample-size justification or a numerical observed effect for interpreting such an illustration.

10. Multiplicity, Missing Data and Analysis Populations

The registry record lists 23 outcome measures, including the primary objective response endpoint and secondary outcomes for response, duration of response, and progression-free survival. Several outcomes use different assessment sources or criteria. When many outcomes are considered, the chance of at least one apparently notable result can increase if statistical testing is conducted repeatedly without an appropriate strategy. General statistical context: multiplicity methods can control or describe this issue, but the choice depends on the analysis plan and the purpose of the comparisons.

A multiplicity adjustment, testing hierarchy, or other strategy for handling multiple outcomes is not reported in the registry record. The record lists 0 analyses, so it does not provide p-values or an analysis plan from which a multiplicity procedure could be identified. No claims about adjustment, prespecification, or confirmatory status should be inferred from the order of outcome descriptions.

The primary endpoint is defined in the efficacy evaluable population, and the record states that the analyzed number equals the number analyzed within the specified group. The number analyzed is not reported. Handling of missing assessments, participants who discontinue treatment, participants without confirmation, and other unevaluable outcomes is not reported in the registry record. It is also not clear from this record whether the efficacy evaluable population was defined before analysis or how it related to all enrolled participants.

These omissions matter because endpoint estimates rely on the analyzed population and the treatment of incomplete data. General statistical context: if the chance of having an evaluable outcome is related to participant characteristics or outcomes, excluding participants without evaluable data can affect interpretation. The registry record does not provide enough information to assess whether that issue arose or how it was addressed. It does not report imputation procedures or sensitivity analyses for missing outcome data.

The record notes subgroup reporting by line of therapy and PD-L1 expression, but subgroup-specific analysis counts and statistical methods are not reported in the registry record. Without those details, the stability of subgroup estimates and whether comparisons were intended as formal tests cannot be assessed.

11. 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 0.05% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pneumonia (6.3%), dyspnoea (4.4%), pyrexia (3.3%), pneumonitis (2.4%), hypercalcaemia (2.2%).

The registry record gives adverse-event summaries for three cohorts: Cohort 1, first-line atezolizumab; Cohort 2, second-line atezolizumab; and Cohort 3, third line and beyond atezolizumab. For serious adverse events, the numbers affected and at risk are 47 of 138 in Cohort 1, 117 of 269 in Cohort 2, and 113 of 252 in Cohort 3. Other adverse events affected 116 of 138, 229 of 269, and 229 of 252, respectively. Death counts are 89, 193, and 198, respectively. The registry record does not state in these counts that deaths were caused by treatment.

The listed serious-event terms and percentages are pneumonia, 6.3%; dyspnoea, 4.4%; pyrexia, 3.3%; pneumonitis, 2.4%; hypercalcaemia, 2.2%; and anaemia, 1.6%. The record does not specify a cohort breakdown for these listed percentages, so they should not be assigned to a particular cohort. These figures describe the serious-event terms recorded, not the probability that an individual participant will experience an event in another setting.

General statistical context: adverse-event counts are descriptive and depend on the population, observation period, and event definitions. Without a control group, they do not establish whether event frequency differs from another treatment or from background risk. The registry record does not report causal attribution for these event counts.

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

The registry record establishes that a Phase II, multicenter, single-arm study evaluated atezolizumab in participants with PD-L1-positive locally advanced or metastatic NSCLC. It specifies the treatment schedule, the primary objective response definition, the assessment source, the planned assessment timing, and the stated efficacy evaluable population. It also lists secondary outcome definitions and descriptive safety information.

It does not report the numerical primary response percentage or its 95% confidence interval. As a result, the registry information does not establish the observed frequency of objective response or how precisely that frequency was estimated. The endpoint definition indicates what counted as a response, not how many participants met that definition. The analyzed participant count for the primary endpoint and subgroups is also not reported.

The design supports description of outcomes among participants who received atezolizumab, but it lacks a concurrent control group. General statistical context: an outcome observed in a treated group is not, by itself, an estimate of the difference that treatment caused. Without randomized or otherwise adequate comparison information, other factors that influence outcomes cannot be separated from treatment effects through this design. This limitation remains even if a response estimate is precise.

The safety counts and listed serious-event percentages describe what the registry record reports for the study cohorts. They do not show a comparative safety effect, and the death counts should not be read as treatment-related deaths because causal attribution is not stated. The available record also does not report missing-data procedures, a multiplicity strategy, or the statistical analysis details needed to evaluate those issues.

Accordingly, the defensible reading is limited: the record describes the study question, design, endpoint criteria, assessment schedule, and safety summaries, but it does not supply numerical efficacy results. It neither establishes a comparative treatment benefit nor supports a conclusion that the intervention does or does not work. Any such conclusion would require outcome results and suitable comparative evidence that are not reported in the registry record.

13. Why This Trial Matters Statistically

This study illustrates why trial design and endpoint results need to be read together. The primary endpoint is a clearly described response proportion, with response defined by confirmed best overall response under RECIST v1.1 and assessed by an IRF. Those details make the intended measurement interpretable. Yet the registry record does not report the resulting proportion, its confidence interval, or the analyzed denominator. A detailed endpoint definition cannot substitute for the result needed to judge the observed response frequency and its precision.

The single-arm design is also central to statistical interpretation. It allows outcomes among treated participants to be described, but it does not create a like-for-like untreated or alternative-treatment comparison. This distinction is important whenever a reported outcome is discussed as evidence of a treatment effect. The registry record lists no randomized comparison and no control arm.

The record further shows the practical importance of keeping outcome populations distinct. Enrollment is 667, while the primary endpoint is defined for the efficacy evaluable population and the subgroup denominators are not reported. Treating enrollment as though it were automatically the analysis denominator would blur an important distinction. Likewise, IRF and investigator assessments, RECIST v1.1 and modified RECIST, response and time-to-event outcomes are separate measurement choices, not interchangeable results.

Finally, the record lists multiple outcomes and subgroup reporting, but does not report a multiplicity strategy, missing-data approach, or inferential analyses. These are not minor technical details: they affect how readers assess uncertainty and the credibility of claims across outcomes. The case therefore demonstrates that registry entries can document a study's intended questions and methods while leaving essential numerical results and analysis details unavailable.

14. 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 main trial-specific limitation is the absence of numerical efficacy results in the registry record. Although the primary endpoint is defined as objective response by IRF under RECIST v1.1 and a 95% confidence interval is specified, neither the response percentage nor the interval limits are reported. The analyzed number in the efficacy evaluable population is also not reported, so the response denominator cannot be established from this record.

The study is single-arm, with no concurrent control group. Its outcomes therefore cannot be interpreted as a randomized comparison with another treatment or with no treatment. The record lists no allocation to comparative arms, and masking is listed as none. The registry description does not report details that would support a causal comparison.

The primary endpoint description calls for reporting by line of therapy and PD-L1 expression subgroup, but subgroup sample sizes and numerical estimates are not reported. The record does not state how subgroup analyses were prespecified or interpreted. It also does not report the method used to construct the confidence interval, how missing or unevaluable response assessments were handled, or whether sensitivity analyses were performed.

There are additional limits to reading the safety information. The adverse-event summaries are organized by cohort, but the listed serious-event percentages do not have a cohort breakdown in the record. The death counts are not accompanied by a causal attribution. No comparative safety group is present.

Finally, although the record lists 23 outcome measures, it lists 0 analyses. Outcome definitions and schedule information are available, but the record does not provide analysis outputs needed to assess efficacy, precision, multiplicity, or the influence of missing data.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: A Study of Atezolizumab in Participants With Programmed Death - Ligand 1 (PD-L1) Positive Locally Advanced or Metastatic Non-Small Cell Lung Cancer — NCT02031458.
• Registry record last updated 2020-01-06; status last verified 2019-12.
• Results first posted 2017-03-23.
• 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.

18. Record Summary

The registry record describes a completed Phase II, multicenter, single-arm study of atezolizumab in participants with PD-L1-positive locally advanced or metastatic NSCLC. Enrollment was 667. The study started on 2014-01-22, primary completion was 2015-05-28, and Hoffmann-La Roche is listed as sponsor.

Participants received atezolizumab 1200 mg intravenously every 3 weeks while the investigator assessed clinical benefit, subject to the stated toxicity and disease-progression conditions. The primary endpoint was the percentage achieving confirmed objective response by IRF assessment under RECIST v1.1, in the efficacy evaluable population. The numerical primary endpoint result and analyzed denominator are not reported in the registry record.

The record lists secondary response and time-to-event outcomes, safety summaries across three line-of-therapy cohorts, 23 outcome measures, and 0 analyses. It does not report a concurrent control group, numerical efficacy analysis, or key details on missing data and multiplicity.