Metformin in Non Small Cell Lung Cancer (NSCLC): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Metformin in Non Small Cell Lung Cancer (NSCLC) (NCT02285855), a randomized phase 2 study. The primary outcome is RECIST and PERCIST Tumor Response.

2. Clinical Question

The study title refers to whether tumor mutation status predicts metabolic response, while the brief summary describes a comparison of metformin plus radiation with radiation alone. The posted primary objective is framed as the effect of metformin on response in people with non-small cell lung cancer treated with hypofractionated radiation. The record does not give a mutation-specific analysis or report whether mutation status was used to define subgroups.

The study is listed as randomized, double-masked, and parallel-group. Participants were assigned to metformin or placebo alongside SBRT. The registry describes baseline and follow-up FDG-PET/CT imaging, with response assessed at 6 months after treatment. The question is therefore about tumor response measured by imaging criteria, not a reported comparison of survival or symptoms. Those outcomes are not identified as the primary endpoint in the record.

3. Trial Design

This was a randomized, double-masked, parallel-group Phase 2 study sponsored by M.D. Anderson Cancer Center. The registry lists lung cancer as the condition and identifies metformin, placebo, and stereotactic body radiotherapy (SBRT) as the interventions. The study began on 2015-02-20, had a primary completion date of 2019-01-05, and is listed as terminated. The recorded enrollment is 27; the brief summary had described a plan to enroll up to 70 participants.

The two treatment groups paired SBRT with either metformin or placebo. In the metformin arm, the registry describes treatment for 3 weeks before SBRT and for 1 week during SBRT, with 2000 mg daily by mouth in divided doses. The description begins to mention starting metformin to reduce gastrointestinal toxicity but is incomplete; the full starting schedule is not reported in the registry record. The placebo arm description specifies oral administration three times a day and SBRT according to standard-of-care practice. The masking description is double, but which people were masked is not reported in the registry record.

For imaging, participants were scheduled for FDG-PET/CT at baseline before metformin, before radiotherapy, and at 6 months after radiotherapy, with a stated window of +/- 30 days. The primary outcome timeframe is described as baseline to post-radiotherapy treatment, assessed up to 6 months. The registry says 3 participants did not complete treatment: 2 in the metformin arm and 1 in the placebo arm. It also reports 15 participants randomized and completing, comprising 14 in the metformin arm and 1 in the placebo arm, and states that 1 participant died before the post-treatment 6-month evaluation. These record statements do not provide a complete account of participant flow or outcome availability.

The registry does not report the randomization sequence method, allocation concealment, planned sample-size justification, or a detailed statistical analysis plan. The study status is terminated, but the reason for termination is not reported in the registry record.

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: A total of 27 patients were consented to this study, but 9 patients withdrew consent prior to protocol medication or treatment. 18 patients randomized and treated under this protocol.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Three participants did not complete the treatment ( Metformin Arm in 2 and Placebo Arm in 1). 15 participants randomized and completed ( Placebo Arm in 1 and Metformin Arm in 14). One participant died prior post treatment 6 months evaluation.

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: A total of 18 stage I non-small cell lung cancer (NSCLC) patients were consented and randomized. Metformin Arm 16 patients and Placebo Arm 2 patients.

6. Primary Endpoint

The registry lists the primary outcome as RECIST and PERCIST Tumor Response, measured over From baseline (prior to metformin start) to Post-Radiotherapy (RT) Treatment, assessed up to 6 months.

The primary outcome is titled “RECIST and PERCIST Tumor Response” and is recorded as a count of participants. It concerns change in tumor response from baseline, before metformin began, to after radiotherapy, with assessment up to 6 months. The record describes response categories based on two imaging frameworks. RECIST categories include complete response, partial response, and stable disease. PERCIST categories include stable metabolic disease, progressive metabolic disease, and complete metabolic response.

The imaging schedule includes FDG-PET/CT at baseline, before radiotherapy, and at the 6-month assessment. The response description specifies relative change from the pre-treatment tumor. In general, an imaging endpoint can provide a structured way to classify changes in measurable disease or metabolic activity; using defined criteria can make response assessments more consistent than an unstructured description. RECIST and PERCIST focus on different features of imaging, so the category labels should not be treated as interchangeable.

The registry record does not give participant counts in the response categories, a numerical definition of response for the analysis, or a single metformin-versus-placebo effect estimate. It also does not explain how RECIST and PERCIST findings were combined, if they were combined, or whether each was analyzed separately. As a result, the endpoint’s measurement framework is described, but the posted record does not show the response distribution or a statistical comparison between groups.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry classifies the primary outcome as a count of participants, but its response description has several categories and imaging criteria. In a simple binary outcome, each participant is classified into one of two states, and an analysis can compare the number or proportion in each state between groups. Here, the listed RECIST and PERCIST categories are not reduced to a single binary comparison. The registry explicitly does not provide a single comparison because the primary outcome has several categories or time points. Without the category-level observations and a stated analysis approach, a binary effect estimate would not represent the posted endpoint.

What randomization contributes

In general, random assignment is intended to make comparison groups more similar, on average, with respect to both measured and unmeasured characteristics at the start of a study. That feature can support a causal interpretation when assignment is implemented as planned and outcome assessment and analysis are appropriate. Randomization does not guarantee that groups will be balanced in a small trial, and it does not remove uncertainty in the estimated effect. The registry identifies the study as randomized, but the sequence-generation and allocation-concealment procedures are not reported in the registry record.

The study is also described as double-masked. In general, masking can reduce the chance that knowledge of assigned treatment influences care, reporting, or outcome assessment. The record does not identify who was masked or describe how masking was maintained. Because the primary outcome uses imaging criteria, information about blinded image review could be relevant, but such a procedure is not reported in the registry record and should not be assumed.

What an effect estimate would need to show

For a multi-category response outcome, an analysis would need to define how the categories are to be compared. A report might present counts for each category by arm and an overall comparison, or specify a different outcome definition. This is a general explanation, not a description of an analysis documented for this study. The registry gives the category names but no category counts, group comparison, effect estimate, or confidence interval. The analysis method and any adjustment for baseline characteristics are not reported in the registry record.

In general, a confidence interval communicates the precision of an estimated effect: a wider interval indicates greater uncertainty than a narrower interval, under the assumptions of the analysis. A p-value, when reported, describes how unusual the data would be under a specified null model; it does not give the probability that a treatment works. The registry record posts neither an estimate nor a p-value for the primary outcome, so neither statistical significance nor precision can be assessed.

Why sparse data need care

General statistical principle: when a study has few observations or very few events, estimates can be unstable, and conventional large-sample approximations may be unreliable. Sparse counts also make detailed subgroup analysis difficult. The recorded enrollment is 27, and the participant-completion information indicates unequal numbers in the reported arms. The primary outcome results are not posted, so the size of any response counts cannot be assessed. No specific model or small-sample method is reported in the registry record.

Over-dispersion is a general issue for count data when observed variation exceeds what a chosen statistical model expects. In such settings, a model that assumes less variability can understate uncertainty. The endpoint is described as a participant count across response categories, but no count model, dispersion assessment, or over-dispersion adjustment is reported. This concept therefore does not establish that over-dispersion occurred in this study; it explains why the model and assumptions matter when count outcomes are analyzed.

Planning calculations are not trial results

In general, sample-size planning uses assumptions about the effect to be detected, variability, significance level, allocation, and desired power. A planning calculation is not an observed treatment effect, and it does not show that a completed study had a particular probability of detecting an effect. The registry describes a planning illustration based on the observed effect, with a two-sided significance level of 0.05, equal allocation, and a normal approximation. It also states that the effect to detect is a stated fraction of the observed effect. These are planning assumptions, not reported design justifications. Because the primary outcome is not reduced to one comparison and no response effect is posted, those assumptions do not provide a directly interpretable primary-outcome sample-size result.

The registry does not report a prospective sample-size justification, the target effect used to plan enrollment, or an interim-analysis plan. The planned enrollment of up to 70 appears in the brief summary, while the recorded enrollment is 27. Without a documented explanation, it is not possible to determine how the original enrollment target related to study conduct or termination. Any planning illustration should be kept separate from the observed evidence, especially when the outcome results and analysis are not reported.

What can be concluded from the posted methods

The record supports a description of a randomized, double-masked, parallel-group design and an imaging-based response endpoint. It does not support reproducing or evaluating an inferential analysis. Statistical details such as the primary analysis population, treatment of missing assessments, handling of multiple response criteria, covariate adjustment, and sensitivity analyses are not reported in the registry record. Without those details and the outcome data, a reader cannot check whether the analysis matched the study objective or quantify uncertainty in a metformin-versus-placebo comparison.

10. Multiplicity, Missing Data and Analysis Populations

The primary outcome description includes several response categories under RECIST and PERCIST and multiple imaging time points. The registry does not report whether these were treated as separate endpoints, combined into a single endpoint, or arranged in a hierarchy. When several outcomes or comparisons are examined, the chance of at least one apparently notable result can increase. In general, statistical methods may address multiplicity through a prespecified hierarchy, adjustment, or a clearly defined primary comparison. The registry record does not report such a plan, and no primary comparison is posted.

Missing-data handling is also not described. The record says 3 participants did not complete treatment and that 1 participant died before the post-treatment 6-month evaluation. It does not say which response assessments were available for each participant, how the death affected endpoint classification, or whether missing assessments were excluded, imputed, or handled in another way. These choices can affect the interpretation of an imaging endpoint, particularly when the follow-up assessment is central to outcome classification.

The analysis population is not defined. The registry does not state whether results were intended to include all randomized participants, only those who completed treatment, or another population. The participant information is not a substitute for a prespecified analysis-population definition. The record also does not report sensitivity analyses that might test whether conclusions depend on missing-data assumptions.

Because the response-category results and analysis are absent, it is not possible to assess the impact of multiplicity, missing assessments, or population selection on a treatment comparison. These are not merely reporting details: they define which observations contribute to the estimate and how the uncertainty should be understood.

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% reporting threshold.

Most frequent serious adverse event terms

Serious events with the highest arm-level rate: diarrhea (7.1%).

The registry lists serious adverse events by arm using participants at risk and participants affected. In the metformin arm, 14 participants are listed as at risk for serious adverse events and 1 as affected; in the placebo arm, 1 participant is listed as at risk and 0 as affected. These are the registry’s recorded figures. The number at risk differs between arms, so the raw affected counts should not be read as if the groups had equal observation totals.

For other adverse events, the registry lists 14 participants at risk and 0 affected in the metformin arm, and 1 at risk and 0 affected in the placebo arm. Deaths are listed as 1 in the metformin arm and 0 in the placebo arm. The registry also identifies diarrhea at 7.1% among the top serious adverse-event terms. It does not state the denominator or provide enough detail to connect that term to a specific participant count. It should therefore be read as a posted percentage, not converted into a count or treated as a complete description of serious events.

These safety figures are descriptive. The registry does not provide event timing, duration, relatedness to study treatment, or a complete event-level account in the record summarized. With the small and unequal at-risk groups, the counts do not by themselves establish a comparative safety rate. The death listed in the metformin arm is also reported as occurring before the post-treatment 6-month evaluation; the record does not provide a cause of death. These data describe what is posted, not a causal assessment.

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

The record documents a randomized Phase 2 study comparing metformin plus SBRT with placebo plus SBRT, and it describes an imaging-based primary response endpoint. It also records enrollment of 27, a terminated status, participant completion information, and selected adverse-event counts. These registry entries establish the study’s general design and the outcome it intended to assess; they do not establish the size or direction of a response difference.

No participant counts for the RECIST or PERCIST response categories are posted, and there is no treatment-effect estimate, confidence interval, or statistical test. The data available in the registry record therefore do not show whether tumor responses differed between the arms, how precise such a difference would be, or whether any pattern varied by tumor mutation status. The trial title’s reference to mutation status cannot substitute for results from a mutation-specific analysis.

In general, a statistically significant result, if reported, would not by itself establish clinical importance; effect magnitude, uncertainty, outcome definition, study conduct, and other evidence would matter. Conversely, absence of a posted test result is not evidence that the treatments had identical outcomes. Here, the issue is that the primary response results and inferential analysis are not reported, so neither similarity nor difference can be estimated from the registry record.

The enrollment and completion statements should also be kept distinct. The brief summary says up to 70 participants would be enrolled; the recorded enrollment is 27. The record does not explain the difference or the reason for termination. It reports that 3 participants did not complete treatment, that 15 randomized participants completed, and that 1 participant died before the 6-month evaluation. Those statements do not describe every participant’s outcome status. They should not be used to infer unreported response counts or to reconstruct an analysis population.

Accordingly, the posted information is useful for understanding the planned comparison and the limits of the available results. It is not enough to determine treatment efficacy, to assess the precision of a treatment effect, or to draw a mutation-specific conclusion. No medical recommendation follows from these registry entries.

13. Why This Trial Matters Statistically

This trial illustrates why a registry entry should be read as a record of both design and reporting, rather than as a complete statistical report. The study identifies a clinically framed comparison—metformin or placebo alongside SBRT—and specifies an imaging endpoint using RECIST and PERCIST response categories. Yet the endpoint’s category counts and analysis are not posted. Without them, the intended question cannot be answered quantitatively from the registry information.

The gap between the brief summary’s planned enrollment of up to 70 and the recorded enrollment of 27 is also important context. Enrollment size affects how much information a study can contribute, but the record does not report the enrollment rationale, the reason for termination, or a final analysis plan. A reader therefore cannot determine from these entries whether recruitment stopped for operational reasons, whether the study met a revised target, or how the available observations were intended to be analyzed.

The participant-flow statements emphasize another statistical issue: outcome availability matters. The record notes treatment non-completion and a death before the 6-month evaluation, while not giving a complete accounting of follow-up or missing response assessments. For an endpoint tied to a scheduled imaging evaluation, missingness and the analysis population can affect both the estimate and its uncertainty. Those methods are not reported.

Finally, the title invokes tumor mutation status, but the posted outcome and summary do not include mutation-stratified results. That difference between a study title and the analysis actually reported is a useful reminder to distinguish a motivating question from documented evidence. The registry provides design details and selected safety counts, but no primary response comparison. Statistically, the most important conclusion is not an estimated treatment effect; it is the limit imposed by absent outcome and analysis results.

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 primary limitation for interpretation is that the registry does not report the primary outcome results. It names RECIST and PERCIST categories but gives no response counts by category, group-level summary, treatment-effect estimate, confidence interval, or test. The stated outcome has multiple categories and time points, and the registry does not specify how those components were combined or compared.

The study stopped with recorded enrollment of 27, while the brief summary describes enrollment of up to 70. The reason for termination and any change to the enrollment target are not reported in the registry record. A prospective sample-size justification and the intended analysis plan are also not reported, so the planned precision and statistical rationale cannot be evaluated.

Participant accounting is incomplete for statistical interpretation. The record reports 3 participants who did not complete treatment, 15 randomized participants who completed, and 1 death before the 6-month assessment, but it does not provide a full disposition for all enrolled participants or identify who had evaluable imaging. It does not define the analysis population or specify how missing assessments and the death were handled.

Details relevant to bias assessment are also limited. The record identifies randomization and double masking but does not describe sequence generation, allocation concealment, who was masked, or whether imaging was independently assessed. It does not report baseline characteristics or tumor mutation categories, despite the title’s emphasis on mutation status. Selected adverse-event summaries lack timing and relatedness information, and the serious-event term percentage has no stated denominator. These limits mean that neither a primary treatment comparison nor a detailed assessment of potential bias can be reconstructed from the registry record.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Metformin in Non Small Cell Lung Cancer (NSCLC) — NCT02285855.
• Registry record last updated 2020-01-06; status last verified 2019-12.
• Results first posted 2020-01-06.
• 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 NCT02285855, a Phase 2, randomized, double-masked, parallel-group study sponsored by M.D. Anderson Cancer Center. It compared metformin plus SBRT with placebo plus SBRT in people with lung cancer. The study began on 2015-02-20, lists primary completion on 2019-01-05, is marked terminated, and records enrollment of 27.

The primary outcome is RECIST and PERCIST tumor response, assessed from baseline through post-radiotherapy follow-up up to 6 months. The record describes imaging schedules and response categories but does not report response-category results or an inferential comparison. It reports treatment non-completion for 3 participants, 15 randomized participants who completed, and 1 death before the 6-month evaluation, without a complete participant-flow or missing-data account.

Selected serious and other adverse-event counts are posted, but event-level context is limited. The reason for termination, the statistical analysis plan, the sample-size justification, the analysis population, and the primary response results are not reported in the registry record.