← Clinical Trials
NSCLC Phase 1 Dose Finding NCT01346540

LUME-Lung 3: Complete Statistical Analysis of Nintedanib in Non-Small-Cell Lung Cancer

An independent statistical review of the randomized, double-blind LUME-Lung 3 trial evaluating continuous oral BIBF 1120, subsequently identified in the registry as nintedanib, added to standard gemcitabine/cisplatin therapy in first-line patients with non-small-cell lung cancer and squamous cell histology.

Completed  ·  Enrollment: 16  ·  Sponsor: Boehringer Ingelheim
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.

1. Trial at a Glance

LUME-Lung 3 was a randomized, double-blind, parallel-group phase 1 study of BIBF 1120 added to gemcitabine/cisplatin in first-line patients with non-small-cell lung cancer and squamous cell histology. The trial enrolled 16 participants and had two arms. The registry reports that the study was prematurely discontinued and that its planned phase II component was not conducted.

16
Enrollment
Total participants
2
Arms
Parallel design
1
Phase
Phase 1
2
Primary endpoints
DLT and MTD
FeatureLUME-Lung 3
Clinical trial identifierNCT01346540
AcronymLUME-Lung 3
PhasePhase 1
StatusCompleted
AllocationRandomized
Design modelParallel
MaskingDouble
Primary purposeTreatment
Enrollment16
Number of arms2
ConditionCarcinoma, Non-Small-Cell Lung
InterventionsBIBF 1120 and placebo
Lead sponsorBoehringer Ingelheim
Sponsor typeIndustry
Start date2011-04-14
Primary completion date2013-04-25
Trial discontinuation: The registry states that the trial was prematurely discontinued following the sponsor's decision not to continue the trial in this indication. The phase II component was not conducted.

2. Clinical Question

The clinical question was centered on dose finding: among first-line patients with non-small-cell lung cancer with squamous cell histology receiving standard gemcitabine/cisplatin therapy, what dose of BIBF 1120 could be administered while maintaining an acceptable frequency of dose-limiting toxicities during the first treatment cycle?

Population

First-line NSCLC patients with squamous cell histology.

Intervention

Continuous oral BIBF 1120 added to standard gemcitabine/cisplatin therapy.

Comparator

Placebo was the other registered drug intervention.

Primary question

Which nintedanib dose met the registry's maximum-tolerated-dose rule based on first-cycle DLTs?

3. Trial Design

01
Enroll16 participants
02
Randomize2 parallel arms
03
TreatStudy drug with standard therapy
04
Cycle 1Up to 21 days
05
EvaluateDLTs and MTD
Allocation
Randomized allocation to two parallel arms.
Masking
Double-masked study design.
Primary purpose
Treatment.
Enrollment
16 participants.

For a phase 1 dose-finding study, the statistical objective differs from the confirmatory efficacy objective of a large phase 3 trial. The key quantity is the observed pattern of treatment-related toxicity across dose levels. Here, the registry defines DLTs during the first 21 days as the basis for determining the maximum tolerated dose.

4. Treatment Arms

REGISTERED ARM · 4 PARTICIPANTS AT RISK

Nintedanib 150 Milligram

  • Registered intervention: BIBF 1120 / nintedanib.
  • Serious adverse events affected 2 of 4 participants at risk.
REGISTERED ARM · 12 PARTICIPANTS AT RISK

Nintedanib 200 Milligram

  • Registered intervention: BIBF 1120 / nintedanib.
  • Serious adverse events affected 5 of 12 participants at risk.

The registry identifies BIBF 1120 and placebo as the drug interventions, while the posted safety results identify the two treatment groups as nintedanib 150 Milligram and nintedanib 200 Milligram. The two groups therefore provide the dose-level structure relevant to the phase 1 MTD question.

5. Primary Endpoints

EndpointTime frameRegistry definition
Number of Participants With Dose Limiting Toxicities (DLTs) During First Cycle for the Determination of the Maximum Tolerated Dose (MTD) Up to 21 days from first drug administration The following drug-related adverse events qualified as a DLT: non-hematological toxicity consisting of CTCAE Grade ≥3 events excluding transient electrolyte abnormality, hyperuricemia and isolated elevation of gamma-glutamyl trans-peptidase; gastrointestinal adverse events including nausea, vomiting, diarrhoea and abdominal pain or hypertension of CTCAE Grade ≥3 despite optimal supportive care/intervention; alanine aminotransferase and/or aspartate aminotransferase elevation of CTCAE Grade ≥3; uncomplicated CTCAE Grade 4 neutropenia not associated with fever of ≥38.5° Celsius for >7 days, except during Cycle 1; CTCAE Grade 4 febrile neutropenia associated with fever ≥38.5º Celsius; platelet levels decreasing to CTCAE Grade 4 or CTCAE Grade 3 associated with bleeding or requiring transfusion; or inability to resume nintedanib dosing within 14 days of stopping because of a drug-related adverse event.
Maximum Tolerated Dose (MTD) of Nintedanib Added to Cisplatin/Gemcitabine Based on the Occurrence of DLTs During Treatment Cycle 1 Up to 21 days from first drug administration The MTD was defined as the dose of nintedanib administered with gemcitabine/cisplatin at which no more than 1 of 6 patients experienced DLT, or one dose tier below that dose at which 2 or more of 6 patients experienced DLT, during the first 21-day treatment cycle. Any DLTs experienced after the start of the second treatment period were considered separately.

6. Results

The registry contains posted results for both primary endpoints, but it does not contain posted formal statistical analyses. The available results are therefore best understood as dose-finding and safety observations rather than as a conventional hypothesis-tested efficacy comparison.

Primary Endpoint: Dose-Limiting Toxicities

The primary DLT endpoint counts participants who experienced protocol-defined dose-limiting toxicity during the first cycle. The registry specifies a first-cycle observation window of up to 21 days from first drug administration.

Registered treatment groupParticipants at riskReported DLT result
Nintedanib 150 Milligram4The registry posts the primary endpoint but does not provide a formal statistical analysis.
Nintedanib 200 Milligram12The registry posts the primary endpoint but does not provide a formal statistical analysis.
No formal statistical analyses were posted to ClinicalTrials.gov. For this type of phase 1 endpoint, the principal analysis would normally be descriptive: count DLTs by dose level, determine the proportion experiencing a DLT, and apply the prespecified MTD decision rule. With only 16 enrolled participants, the clinical interpretation depends heavily on the observed toxicity pattern at each dose rather than on conventional large-sample hypothesis testing.
Clinical Biostats interpretation

The DLT endpoint is fundamentally a toxicity-count endpoint, not a time-to-event efficacy measure. The important statistical object is the number and proportion of participants experiencing protocol-defined DLTs during the specified first-cycle window.

The absence of a posted formal analysis means that an estimated treatment effect, confidence interval, or p-value is not available for this endpoint in the registry. A p-value would also be poorly aligned with the central phase 1 decision: selecting a dose according to the prespecified toxicity rule.

The small sample size is particularly important. When only a few participants are observed at a dose level, one additional DLT can materially change the observed proportion. That makes the MTD rule more informative for dose selection than treating the two arms as if they were a conventional confirmatory randomized comparison.

Primary Endpoint: Maximum Tolerated Dose

The registry defines MTD using a six-patient dose-level rule. A dose qualifies when no more than 1 of 6 patients experiences a DLT during the first 21-day cycle; if 2 or more of 6 patients experience DLTs at a dose, the rule identifies one dose tier below that dose.

Registry MTD rule

≤ 1 of 6

DLTs during the first 21-day treatment cycle at the MTD dose

If 2 or more of 6 experience DLTs, the registry definition uses one dose tier below that dose.

Clinical Biostats interpretation

The MTD is a decision threshold, not an estimate of treatment efficacy. It translates observed first-cycle toxicity into a dose-selection rule. The threshold of no more than 1 of 6 DLTs means that the protocol is willing to accept some observed toxicity while seeking a dose with an acceptably limited frequency of serious early treatment-related toxicity.

The rule does not mean that the selected dose has a true DLT probability of exactly 1/6. The observed proportion in a small cohort is subject to substantial sampling uncertainty, and the MTD decision is conditional on the protocol's dose-escalation framework.

Because the registry does not post a formal statistical analysis for the MTD endpoint, there is no confidence interval or p-value accompanying the decision rule. In phase 1 dose finding, that distinction is important: the purpose is dose selection and characterization rather than demonstration of a statistically significant treatment effect.

7. Safety Results

The registry reports serious adverse events by registered treatment group. These data provide an additional safety perspective beyond the specific DLT definition.

Registered treatment groupSerious adverse events: affected / at risk
Nintedanib 150 Milligram2/4
Nintedanib 200 Milligram5/12
Serious adverse events: affected participants
Nintedanib 150 Milligram
2/4
Nintedanib 200 Milligram
5/12

These serious-adverse-event counts should not be substituted for the DLT endpoint. A serious adverse event and a protocol-defined DLT are different statistical and clinical classifications. The DLT endpoint uses the detailed first-cycle criteria specified in the registry, whereas the serious-adverse-event results summarize a broader safety category.

Small-cohort caution: The denominators are only 4 and 12 participants at risk. Direct comparison of the observed serious-adverse-event counts should therefore be treated as descriptive. The registry does not provide confidence intervals, a formal between-group hypothesis test, or an adjusted estimate for these safety results.

8. Statistical Methodology

Descriptive analysis of DLTs

The most direct statistical analysis of the first primary endpoint is a dose-level frequency table. For each dose, the analysis would report the number of participants with at least one protocol-defined DLT and the number of participants evaluated during the first cycle.

Basic dose-level summary
DLT proportion = Number of participants with a DLT ÷ Number of participants evaluated

This proportion describes observed first-cycle toxicity at a dose. It is not a model-based estimate of the long-term probability of toxicity.

Rule-based MTD determination

The second primary endpoint is governed by a prespecified decision rule rather than a conventional superiority test. The rule uses cohorts of 6 and classifies a dose according to whether 0 or 1 of 6, or 2 or more of 6, participants experience DLTs.

MTD decision structure
0 or 1 DLT among 6 → dose can satisfy the stated MTD criterion
2 or more DLTs among 6 → use one dose tier below that dose

The registry also states that DLTs occurring after the start of the second treatment period were considered separately.

Why a conventional p-value is not the central statistic

A phase 1 dose-finding study is typically designed to identify a dose suitable for further development rather than to establish a statistically significant difference in efficacy between randomized groups. A hypothesis test comparing two treatment arms would not answer the main MTD question as directly as the prespecified dose-toxicity rule.

Randomization and masking

The registry classifies the study as randomized, parallel, and double-masked. Randomization can reduce systematic allocation differences between groups, while masking can reduce the influence of knowledge of treatment assignment on treatment administration, assessment, and reporting. In a small phase 1 study, however, randomization does not eliminate sampling variability.

Safety summaries

Serious adverse events are naturally summarized using counts and denominators. When event counts are small, exact confidence intervals can be more appropriate than relying on large-sample normal approximations. The registry does not post such intervals for the reported serious-adverse-event results.

9. Statistical Methods Explained

Why is the DLT endpoint analyzed as a count?

DLT is defined as whether a participant experiences a qualifying toxicity during a specified first-cycle period. That makes the basic outcome binary at the participant level: DLT or no DLT. Summing those outcomes within a dose cohort gives the DLT count used by the MTD rule.

Why is the first 21-day cycle so important?

The MTD endpoint specifically defines its observation window as up to 21 days from first drug administration. Restricting the primary dose-finding assessment to a common early treatment period makes toxicity observations more comparable across dose cohorts and aligns the analysis with the prespecified decision rule.

What does “no more than 1 of 6” mean?

It means that, within a six-patient cohort, zero or one participant may experience a protocol-defined DLT while satisfying that part of the MTD definition. It is an observed cohort criterion, not a statement that the underlying population DLT probability is known with precision.

Why is the MTD rule different from a conventional hypothesis test?

The MTD question is operational: which dose should be considered tolerable enough for further development under the study's predefined toxicity rule? A p-value asks whether data are inconsistent with a null hypothesis. Those are different questions, so the decision rule is the more direct statistical mechanism for this phase 1 objective.

Why should serious adverse events not be treated as DLTs?

The registry defines DLT using a specific collection of treatment-related toxicity criteria, including CTCAE grades, selected laboratory abnormalities, specific gastrointestinal and hypertension criteria, hematologic toxicity, and inability to resume nintedanib dosing within 14 days. Serious adverse events are a broader safety classification and therefore cannot simply be substituted for DLT counts.

Why does the small sample size matter?

With 16 participants overall, the observed toxicity frequency is based on a small amount of information. A single additional event can substantially change the observed proportion within a small dose cohort. Confidence intervals, when calculated, would consequently be wide, and estimates would remain uncertain even if the observed event pattern appears clear.

10. Interpreting the Randomized Design in a Phase 1 Trial

The randomized and double-masked design is notable because many dose-finding studies use sequential rather than randomized structures. Here, randomization provides a framework for assigning participants to the registered treatment groups, while the primary statistical objective remains dose and toxicity assessment.

Randomization

Randomized allocation helps prevent investigators from deliberately assigning participants with particular characteristics to one treatment group.

Double masking

Double masking can reduce bias associated with knowledge of treatment assignment during treatment and assessment.

Parallel groups

Participants remain in their assigned treatment pathway rather than being deliberately exposed to both randomized groups in a crossover design.

Dose finding

The principal statistical task is characterization of toxicity across dose levels rather than estimation of a definitive efficacy effect.

The presence of randomization should therefore not be mistaken for evidence that the study was powered as a conventional comparative efficacy trial. The registry describes a phase 1 study with an MTD-focused primary objective, and the study was prematurely discontinued before its planned phase II component.

11. Understanding the Posted Safety Numbers

Observed counts

The registry reports serious adverse events affecting 2/4 participants in the nintedanib 150 Milligram group and 5/12 participants in the nintedanib 200 Milligram group.

These are raw affected/at-risk counts. They do not provide a formal estimate of a dose-response relationship, and they should not be interpreted as proof that the difference between dose groups is statistically significant.

Denominators matter

The denominators are small and unequal. A count of 2 events in 4 participants represents a different amount of statistical information from 5 events in 12 participants. Comparing counts alone would therefore be misleading; even comparing proportions would leave substantial uncertainty because of the small sample sizes.

Serious adverse events versus DLTs

The reported serious-adverse-event results cannot be used to reconstruct the primary DLT endpoint because the registry's DLT definition is narrower and specifically tied to the first 21-day cycle. The two safety concepts should remain analytically distinct.

12. Planned Analysis

The registry's primary endpoints establish a dose-finding analysis centered on DLTs during the first 21 days and an MTD decision based on the observed number of DLTs within dose cohorts. For an endpoint of this type, the usual analysis is descriptive: tabulate DLT counts and proportions at each dose and then apply the prespecified MTD rule.

Statistical questionTypical analysis for this endpoint
How many participants experienced a DLT?Descriptive count of participants with at least one protocol-defined DLT.
What fraction experienced a DLT?DLT count divided by the number of participants evaluated during the defined first-cycle window.
Which dose meets the MTD criterion?Apply the registry's rule of no more than 1 of 6 DLTs at a dose, with one dose tier below a dose at which 2 or more of 6 experience DLTs.
How uncertain is the observed DLT frequency?For a fuller statistical analysis, an exact binomial confidence interval could be used because of the small cohort sizes.
How should serious adverse events be summarized?Report affected participants and denominators by treatment group, with appropriate small-sample caution.

Results have been posted on ClinicalTrials.gov, but formal statistical analyses are not posted in the registry. The registry does not report a formal p-value or confidence interval for either primary endpoint.

13. Trial Timeline

2011-04-14

Study start

The registry lists 2011-04-14 as the study start date.

2013-04-25

Primary completion

The registry lists 2013-04-25 as the primary completion date.

Completed

Premature discontinuation

The registry states that the trial was prematurely discontinued following the sponsor's decision not to continue the trial in this indication. Phase II was not conducted.

14. Limitations

15. Why This Trial Matters Statistically

LUME-Lung 3 is a useful teaching example because it illustrates a different statistical problem from the large confirmatory trials that dominate clinical-trial reporting. The central question is not whether a treatment produces a statistically significant improvement in a survival endpoint. Instead, the study uses early toxicity observations to guide dose selection.

Statistical conceptHow it appears in LUME-Lung 3
Phase 1 dose findingThe registry identifies the study as phase 1 and defines MTD as a primary endpoint.
RandomizationThe study is classified as randomized.
Double maskingThe study is classified as double masked.
Parallel designThe registry specifies a parallel design model with two arms.
Binary toxicity endpointDLT status is determined during a defined first-cycle period.
Small-sample inferenceOnly 16 participants were enrolled, making conventional large-sample approximations less reliable.
Rule-based decision makingThe MTD definition uses a specific DLT threshold among 6 patients.
Safety analysisSerious adverse events are reported as affected participants divided by participants at risk.
Trial discontinuationThe planned phase II component was not conducted after premature discontinuation.

16. Statistical Concepts in This Trial

LUME-Lung 3 demonstrates how clinical-trial statistics can be built around a prespecified safety decision rule rather than a conventional efficacy hypothesis test. The most important concepts are the definition of a DLT, the first-cycle observation window, small-cohort toxicity estimation, and the MTD threshold.

DLT classification

The detailed protocol definition converts multiple clinical and laboratory toxicity criteria into a binary dose-finding outcome.

MTD threshold

The MTD definition uses a cohort-based rule of no more than 1 DLT among 6 patients.

Small-sample inference

With small dose cohorts, observed proportions can be unstable and uncertainty can be substantial.

Safety interpretation

Serious adverse-event counts provide complementary safety information but are not interchangeable with DLTs.

17. Sources

Continue through the Clinical Biostats clinical-trial library

Explore additional clinical-trial analyses and statistical methods presented in a consistent educational framework.

18. Record Summary

LUME-Lung 3 was a randomized, double-blind phase 1 study enrolling 16 participants to evaluate BIBF 1120 added to gemcitabine/cisplatin therapy in first-line NSCLC patients with squamous cell histology. Its primary statistical objectives were the assessment of dose-limiting toxicities during the first 21 days and determination of the maximum tolerated dose using a prespecified cohort rule. The registry reports serious adverse events affecting 2/4 participants at the nintedanib 150 Milligram dose and 5/12 at the nintedanib 200 Milligram dose. The study was prematurely discontinued, and the planned phase II component was not conducted.

Statistical perspective: The key lesson from LUME-Lung 3 is that phase 1 dose-finding statistics are primarily about structured toxicity assessment and dose selection. The interpretation depends on the prespecified DLT definition, the first-cycle observation window, the cohort-based MTD rule, and the uncertainty inherent in small samples.