A Study of Erlotinib (Tarceva) in Participants With Non-Small Cell Lung Cancer (NSCLC): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study of Erlotinib (Tarceva) in Participants With Non-Small Cell Lung Cancer (NSCLC) (NCT02774278), a na phase 2 study. The primary outcome is Number of Differentially Expressed Genes Associated With Clinical Benefit.

2. Clinical Question

3. Trial Design

MERIT was a phase II marker-identification trial of erlotinib in second-line non-small cell lung cancer. The registry lists the condition as non-squamous non-small cell lung cancer, and describes the study as a single-group design with allocation recorded as not applicable and no masking. Enrollment was 264. The registry records the study as completed, with a start date of 2005-07 and primary completion in 2009-06.

Participants received erlotinib orally each day. The planned treatment period continued until disease progression, unacceptable toxicity, or death. The study aim was to assess potentially predictive markers of efficacy in participants receiving erlotinib, rather than to estimate a treatment contrast against a concurrent control group. The sponsor was Hoffmann-La Roche.

The primary outcome concerned the number of differentially expressed genes associated with clinical benefit. Affymetrix gene-expression profiles were analyzed using a multivariate linear model. The primary analysis set was defined as participants with evaluable tissue samples who were included in the primary Affymetrix efficacy analysis. The size of this analysis set is not reported in the registry record.

The registry also lists outcomes for overall response, defined as complete or partial response using RECIST, and clinical benefit, defined as complete response, partial response, or stable disease for at least 12 weeks after study entry. There were 5 outcome measures recorded, but the registry lists 0 analyses. The primary endpoint result, gene-level findings, and detailed results for the other outcomes are not reported in the registry record. Because the study has no concurrent control group, the design can describe outcomes and marker associations within the treated group but cannot isolate a comparative treatment effect.

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.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Primary analysis set (PAS) included all participants who provided evaluable tissue samples and were included in the primary Affymetrix efficacy analysis.

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: Analysis population included all enrolled participants.

6. Primary Endpoint

The registry lists the primary outcome as Number of Differentially Expressed Genes Associated With Clinical Benefit, measured over Baseline until disease progression, unacceptable toxicity or death, evaluated up to 4 years.

The primary endpoint is the number of differentially expressed genes associated with clinical benefit. It was intended to summarize how many genes showed expression differences between participants classified as deriving clinical benefit and those who did not. Clinical benefit was defined using RECIST as complete response, partial response, or stable disease lasting at least 12 weeks after study entry. The analysis used baseline gene-expression profiles, with the stated timeframe extending from baseline until disease progression, unacceptable toxicity, or death, evaluated up to 4 years.

This endpoint addresses the trial’s marker-identification objective: searching for biological measurements that may be associated with a clinical outcome among participants receiving erlotinib. A gene-expression association is not, by itself, proof that a marker predicts who benefits specifically from erlotinib. Demonstrating treatment-specific prediction generally requires evidence that the marker modifies a treatment contrast, which this single-group design does not provide.

The analysis used a false discovery rate criterion below 0.3 to select differentially expressed genes. The registry record does not report the resulting gene count, the genes identified, or estimates of the differences in expression. It also does not report confidence intervals or other measures of precision for the primary endpoint. Consequently, the endpoint’s intended meaning can be described, but the registry record does not permit interpretation of its observed value.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary analysis used Affymetrix gene-expression profiles to look for differences between participants classified as deriving clinical benefit and those who did not. This is an association analysis: it asks whether measured expression varies with the clinical-benefit grouping. It does not by itself test whether erlotinib performs better than a comparator, because the trial is recorded as a single-group study without a concurrent control group.

What does the multivariate model do?

A multivariate linear model estimates relationships while including multiple predictors in the same model. In this analysis, the listed predictors were clinical-benefit status, histology, ethnicity, sex, RNA integrity number (RIN), smoking status, and stage. Including these variables allows the analysis to account for their modeled relationships when assessing expression differences associated with benefit status. It does not guarantee that the model captures every relevant factor, nor does it establish causation. The registry record does not report the model coefficients, uncertainty estimates, diagnostics, coding choices, or gene-level results.

The analysis population also matters. The primary analysis set included participants who provided evaluable tissue samples and were included in the primary Affymetrix efficacy analysis. This is not necessarily the same population as everyone enrolled or everyone included in the safety summaries. The number of participants in the primary analysis set, the reasons tissue samples were not evaluable, and the handling of missing clinical or expression data are not reported in the registry record.

Why use a false discovery rate criterion?

Gene-expression studies can examine many genes. When many statistical tests are carried out, some apparent findings may arise by chance even when there is no real association. False discovery rate methods address the expected proportion of false findings among the findings selected under a specified procedure. The trial description states a false discovery rate criterion below 0.3 as the cutoff for identifying differentially expressed genes.

This criterion is a selection rule, not a measure of the probability that any particular gene is truly associated with benefit. It also does not show how large or clinically meaningful an expression difference is. To judge the findings, readers would need the observed gene list, effect estimates, uncertainty, and enough information about the analysis procedure to understand how the cutoff was applied. Those details are not reported in the registry record. The stated cutoff should therefore be described as the trial’s selection criterion, not as proof of a validated marker.

What can a confidence interval add?

In general, a confidence interval expresses the precision of an estimated quantity under a statistical model and its assumptions. A narrow interval indicates greater precision than a wide interval, while an interval compatible with a range of effects signals uncertainty about the exact magnitude. The registry record does not provide confidence intervals for the primary endpoint or marker associations, so their precision cannot be assessed from the posted information.

Likewise, a p-value, if reported, would address compatibility of data with a specified null model, rather than the size or clinical importance of an association. The registry record does not list posted analyses or primary gene-level test results. No inferential conclusion can be reconstructed from the endpoint description alone.

Why does the single-group design matter?

In a randomized comparative trial, allocation can support estimation of a treatment contrast by creating groups intended to be comparable at baseline. Here, allocation is recorded as not applicable and the model as single group. The described analysis compares participants according to clinical-benefit status, not according to randomized treatment assignment. Differences in expression between those outcome groups may be associated with outcome, but cannot establish that a gene predicts a differential effect of erlotinib relative to another option.

For a marker to be specifically predictive of treatment effect, evidence generally concerns whether the treatment contrast differs across marker values. A treated-only association cannot estimate that interaction. A marker could instead be prognostic, associated with outcome regardless of treatment, or related to participant features that also correlate with outcome. This is a general distinction in biomarker interpretation, not a finding reported for any particular gene in this trial.

How should the analysis timeframe be read?

The primary endpoint description gives a timeframe from baseline until disease progression, unacceptable toxicity, or death, evaluated up to 4 years. This identifies the stated observation window; it does not report how many participants were followed for the full period or the timing of tissue collection beyond the baseline expression profiles. The registry record does not describe interim analyses, a detailed missing-data strategy, or a protocol for dealing with incomplete follow-up. These details should not be inferred from the timeframe alone.

The registry’s 0 listed analyses means the record does not provide registered analysis results in the information described. It does not replace a detailed statistical analysis plan, and the registry record does not report the sample-size justification, any interim-analysis plan, or a complete multiplicity strategy across all outcomes. Planning illustrations, if displayed with the results, should not be interpreted as observed power or as the trial’s original design rationale. The record does not supply an effect estimate for the primary endpoint on which to base an observed-power interpretation.

10. Multiplicity, Missing Data and Analysis Populations

The gene-expression analysis involved a false discovery rate criterion below 0.3. This criterion addresses multiplicity among the gene-level comparisons as described in the registry. It is not a report of how many genes passed the criterion, and the registry record does not provide the resulting gene list or the underlying test results.

The record lists 5 outcome measures, including the primary gene-count endpoint and clinical outcomes for overall response and clinical benefit. It does not explain whether or how multiplicity across these distinct outcomes was controlled, or whether any outcome was treated as confirmatory beyond the stated primary endpoint. A false discovery rate procedure for gene-level screening should not be assumed to resolve multiplicity across every outcome or analysis.

Missing-data details are also limited. The primary analysis set required evaluable tissue samples, but the registry record does not report how many participants met that requirement, why samples were unavailable or unevaluable, or how missing covariates and outcomes were handled. It does not describe imputation, sensitivity analyses, or assumptions about missingness. Because the tissue-based analysis may involve a subset of enrolled participants, the scope of the primary result cannot be assessed without those details.

The registry lists 0 analyses and does not report a detailed statistical analysis plan, sample-size justification, or interim-analysis plan. These items are not reported in the registry record and should not be inferred from the endpoint description or study duration.

11. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pneumonia (4.6%), dyspnoea (4.2%), haemoptysis (1.9%), asthenia (1.1%), general physical health deterioration (1.1%).

The safety figures describe the erlotinib group, with 261 participants listed as at risk for both serious and other adverse events. The record lists 65 participants affected by serious adverse events and 209 affected by other adverse events. These are separate registry categories; the record does not provide enough detail to determine overlap between them or to reconstruct event-level totals.

The serious adverse-event terms with listed percentages are pneumonia (4.6%), dyspnoea (4.2%), haemoptysis (1.9%), asthenia (1.1%), general physical health deterioration (1.1%), and pyrexia (1.1%). These percentages should be read as descriptive entries in the registry, not as comparative estimates against another group. The registry record does not report a death count in these safety data.

Adverse-event reports do not by themselves establish that the study drug caused an event. Without a concurrent control group, the figures cannot distinguish events attributable to treatment from events related to the underlying condition or other factors. The record also does not provide full event-level detail, severity grading, timing, or the number of repeated events for these terms. The listed affected-participant counts and percentages therefore describe reported safety experience but do not support a controlled causal comparison.

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

The registry record establishes that MERIT was a phase II, single-group study of oral erlotinib in non-squamous non-small cell lung cancer, designed to investigate markers associated with clinical benefit. It specifies a clinical-benefit definition based on RECIST: complete response, partial response, or stable disease lasting at least 12 weeks after study entry. It also describes a multivariate gene-expression analysis and a false discovery rate criterion below 0.3.

Those details explain what the investigators intended to measure, but they do not supply the primary result. The number of differentially expressed genes, their identities, their expression differences, and estimates of precision are not reported in the registry record. The record also does not provide the detailed outcome results needed to describe the observed response or clinical-benefit experience.

The endpoint is an association between gene expression and outcome among participants receiving erlotinib. It is not a direct estimate of erlotinib’s effect relative to another treatment. Without a comparator group, a marker associated with clinical benefit in this study could reflect general prognosis, participant characteristics, or other influences; the design cannot establish that the marker predicts a treatment-specific benefit.

The safety figures are similarly descriptive. They report affected participants and selected serious-event terms in the erlotinib group, but do not provide a controlled comparison or establish that the drug caused each event. The record’s safety counts should therefore be understood as reported observations, not as estimates of excess risk attributable to treatment.

In short, the registry record documents the study question, analysis approach, and selected safety information. It does not report enough efficacy or marker results to determine the magnitude or precision of the primary association, nor does the design support a randomized treatment-effect conclusion. Conclusions beyond those stated would require information not reported in the registry record.

13. Why This Trial Matters Statistically

This trial is statistically important as an example of the difference between a biomarker association study and a comparative treatment trial. MERIT examined gene-expression profiles in relation to clinical benefit among participants receiving erlotinib. That can be useful for generating hypotheses about markers, but it answers a different question from whether treatment outcomes differ between erlotinib and a control treatment.

The primary endpoint also illustrates why analysis populations and outcome definitions matter. The gene-expression analysis was limited to participants with evaluable tissue who entered the primary Affymetrix efficacy analysis. Clinical benefit combined complete response, partial response, and stable disease lasting at least 12 weeks, whereas the separately listed overall-response outcome included complete or partial response. A result would need to be interpreted in light of both the selected tissue-analysis population and the specific outcome definition.

The use of a false discovery rate criterion highlights a common challenge in high-dimensional analyses: testing many genes raises the chance of false-positive findings. The registry states a cutoff below 0.3, but does not report which genes passed it or the estimates needed to judge effect size and precision. A selection criterion alone is not equivalent to independent validation or clinical usefulness.

The record also underscores the importance of transparent reporting. The design, endpoint, model covariates, and selected safety figures are described, but the registry lists 0 analyses and does not report the primary gene-count result. Without results and missing-data details, readers cannot assess the robustness or reproducibility of the marker findings. The study’s statistical contribution can therefore be understood from its question and methods, while its empirical marker conclusions remain unavailable in the registry record.

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 principal limitation is that the registry record does not report the primary endpoint result. It gives no number of differentially expressed genes, gene identities, expression effect estimates, confidence intervals, or gene-level test results. The clinical outcome results are also not reported, so the observed response and clinical-benefit distributions cannot be assessed.

The single-group design limits causal interpretation. There is no concurrent control group, and the study cannot estimate a comparative treatment effect or establish that a marker predicts differential benefit from erlotinib. Associations among treated participants may reflect prognosis or other participant characteristics. No treatment-by-marker comparison is described.

The primary analysis used only participants with evaluable tissue who were included in the Affymetrix efficacy analysis. The size and characteristics of that analysis set, the availability and quality of tissue, and the reasons for exclusion are not reported in the registry record. Missing-data handling and sensitivity analyses are not reported either, limiting assessment of selection bias and robustness.

The model lists several covariates and the false discovery rate cutoff, but the record does not provide coding details, model diagnostics, coefficients, or a full account of multiplicity across outcomes. It does not report the sample-size justification or interim-analysis plan. The safety information is also limited to selected counts and terms, with no controlled comparator and no death count reported in the adverse-event data. These limitations mean the record supports description of the planned analysis and selected safety observations, but not a complete assessment of efficacy, marker validation, or comparative safety.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: A Study of Erlotinib (Tarceva) in Participants With Non-Small Cell Lung Cancer (NSCLC) — NCT02774278.
• Registry record last updated 2016-08-08; status last verified 2016-08.
• Results first posted 2016-08-02.
• 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

NCT02774278, titled MERIT, was a completed phase II, single-group marker-identification study sponsored by Hoffmann-La Roche. The registry records enrollment of 264 participants with non-squamous non-small cell lung cancer. Participants received oral erlotinib daily until disease progression, unacceptable toxicity, or death; allocation was not applicable and masking was none.

The primary endpoint was the number of differentially expressed genes associated with clinical benefit, based on an Affymetrix expression analysis using a multivariate linear model and a false discovery rate criterion below 0.3. Clinical benefit was defined as complete response, partial response, or stable disease lasting at least 12 weeks after study entry. The registry lists 5 outcome measures and 0 analyses; the primary gene-count result and detailed efficacy findings are not reported in the registry record.

Safety entries list 261 participants at risk, 65 affected by serious adverse events, and 209 affected by other adverse events. These data are descriptive and do not provide a controlled comparison.