LUX-Lung 5: Afatinib Plus Weekly Paclitaxel Versus Investigator's Choice of Single Agent Chemotherapy Following Afatinib Monotherapy in Non-small Cell Lung Cancer Patients Failing Erlotinib or Gefitinib: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for LUX-Lung 5: Afatinib Plus Weekly Paclitaxel Versus Investigator's Choice of Single Agent Chemotherapy Following Afatinib Monotherapy in Non-small Cell Lung Cancer Patients Failing Erlotinib or Gefitinib (NCT01085136), a randomized phase 3 study. The primary outcome is Progression Free Survival (Part B).

2. Clinical Question

3. Trial Design

LUX-Lung 5 was a Phase 3 randomized, open-label, active-controlled, parallel-group trial in non-small cell lung cancer. The registry reports overall enrollment of 1154. That figure relates to the trial record as a whole, not just the Part B efficacy comparison: the primary Part B analysis lists 138 participants in the afatinib-plus-paclitaxel group and 68 in the investigator’s-choice chemotherapy group. The record describes Part A as afatinib monotherapy and Part B as the randomized comparison following that monotherapy.

The study began in 2010-02, reached primary completion in 2013-10, and is recorded as completed. The trial summary describes treatment and follow-up until death or loss to follow-up and says health-related quality of life information was collected. The primary endpoint timeframe is from randomization to progression or death, up to 32 months. The overall-survival timeframe is from randomization to death, also up to 32 months.

Masking is listed as none, so the design was open-label. The registry describes the comparison as randomized but does not report the allocation ratio or the detailed sequence-generation and concealment procedures. It also does not report the original design assumptions, the specific chemotherapy choices, or the protocol’s detailed analysis plan. The Part B primary analysis population includes all randomized participants, irrespective of whether they received treatment; this definition should not be conflated with the separate safety denominators.

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: PD = Progression Disease

Part A

Reasons for not completing this period:

Part B

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Randomised Set: This analysis set consist of all randomised patients irrespective of whether treated or not.

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: Treated set, i.e. all patients who were documented to have taken at least 1 dose of afatinib 50 mg in Part A of the trial and all patients who received at least 1 dose of trial medication in Part B.

6. Primary Endpoint

The registry lists the primary outcome as Progression Free Survival (Part B), measured over From randomization until disease progression or death; Up to 32 months.

Progression-free survival measures the time from randomization until disease progression or death. In this trial, progression was determined using RECIST Version 1.1. A participant’s follow-up for this endpoint ends when either event occurs; participants without a recorded event by the end of their observed follow-up contribute information up to that point under the time-to-event analysis.

This endpoint is used because it combines disease progression and death into a single time-to-event outcome. It can describe how long participants remain alive without documented progression, while accounting for the fact that participants may have different lengths of follow-up. The Kaplan-Meier method estimates the distribution of event times while accommodating incomplete follow-up, subject to its assumptions about censoring.

The registry reports the median progression-free survival for each randomized group and a hazard ratio from the stratified log-rank analysis. A median is the time by which the estimated event-free proportion reaches the midpoint of its distribution. It is one summary of the outcome and does not describe every point in the survival curve. The hazard ratio compares the event rates over follow-up under the model’s interpretation; it is not a ratio of the two medians. The timeframe is reported as up to 32 months, so the results describe observed follow-up within that limit rather than outcomes over an unlimited period.

7. Results Posted in the Registry

• Afatinib Plus Paclitaxel (Part B) vs Investigators Choice of Chemotherapy (Part B): hazard ratio 0.61 (95% CI 0.44–0.85), reported p = 0.003 (as reported in the registry).

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Overall Survival (Part B)

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Stratified log-rank test.); it was not recalculated.
• Result:
  • Afatinib Plus Paclitaxel (Part B) vs Investigators Choice of Chemotherapy (Part B): hazard ratio 1.04 (95% CI 0.76–1.44), reported p = 0.790 (as reported in the registry).
• Interpretation:
  • The 95% confidence interval includes no difference (a ratio of 1), so these data do not show a statistically significant difference between Afatinib Plus Paclitaxel (Part B) vs Investigators Choice of Chemotherapy (Part B) and the control. The interval gives the range of effects that remain compatible with the data.

Objective Response (Part B)

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Investigators Choice of Chemotherapy (Part B): 13.2% (9/68) versus 31.2% (43/138) on control; risk difference -17.9 percentage points (95% CI -29.1 to -6.8), risk ratio 0.42 (0.22–0.82); unadjusted two-proportion z-test p = 0.005.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically lower for Investigators Choice of Chemotherapy (Part B) than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The posted secondary outcomes give a mixed pattern rather than a uniform extension of the primary result. Objective response favors the afatinib-plus-paclitaxel group in the reported comparison, while overall survival does not show a clear separation between groups. These endpoints measure different aspects of outcome: response records tumor response under the stated assessment schedule, whereas overall survival follows time until death. Agreement between endpoints is not guaranteed, and a difference in one does not imply the same effect in another.

The primary progression-free survival finding therefore stands as the principal result, while the secondary outcomes add context. The response result is consistent in direction with the primary comparison, but it is not a substitute for progression-free survival. The overall-survival estimate is less conclusive about a group difference because its interval includes the null value. The record also lists other secondary outcomes, including Part A progression-free survival, Part A objective response, and adverse events; not all listed outcomes have detailed results in the posted analysis information.

Secondary results commonly carry less inferential weight when many outcomes are examined. Testing multiple outcomes increases the chance that at least one result will appear statistically notable by chance if no multiplicity procedure is used. The registry record does not report whether the secondary analyses were adjusted for multiplicity, how their testing order was set, or whether they were intended as confirmatory or exploratory. Their p-values and confidence intervals should consequently be interpreted as outcome-specific results, not as independent confirmation of the primary conclusion.

The registry also does not report a complete account of missing observations for these endpoints or the detailed analysis populations for every secondary analysis. Those omissions limit how fully the secondary pattern can be assessed. The results are informative descriptions of the outcomes recorded, but they do not establish a single overall treatment effect across all endpoint types.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Progression Free Survival (Part A)
• Overall Survival (Part B)
• Objective Response (Part A)
• Objective Response (Part B)
• Intensity and Incidence of Adverse Events (AEs) for Part A & Part B.

9. Statistical Interpretation

10. Statistical Methods Explained

The primary endpoint is time-to-event data: each participant contributes a time from randomization until progression or death, or until their last relevant follow-up if neither event has been recorded. This format is useful when follow-up durations vary. It retains information from participants who have not yet had an event at the time their observation ends, rather than treating them as if their event time were known.

What does the Kaplan-Meier median represent?

The Kaplan-Meier method estimates the probability of remaining event-free over time while accounting for censored observations. The median is the time at which the estimated event-free probability reaches the midpoint. It is a compact description of the estimated distribution, but it does not show the full pattern of event timing. Two groups can have medians that summarize their curves without the median difference capturing every aspect of the comparison.

The reported progression-free survival medians are 5.55 months and 2.89 months. Their confidence intervals describe uncertainty around each group’s median estimate. Those intervals are not a confidence interval for the difference between medians. The registry does not report a direct median-difference analysis.

How should the hazard ratio be read?

The sponsor reports a Cox proportional-hazards estimate of 0.61 for afatinib plus paclitaxel versus investigator’s choice of chemotherapy. A hazard ratio compares the instantaneous event rates between groups among participants who have not yet had the event, as represented by the fitted model. A value below 1 points to a lower estimated hazard in the first-listed group. It is not an absolute risk difference, does not state how many participants experienced an event, and is not interchangeable with the ratio of the group medians.

The proportional-hazards interpretation is most straightforward when the relative hazards are reasonably stable over time. The registry names a Cox proportional-hazards analysis but does not report diagnostics or results assessing that assumption. It also does not provide enough detail to assess other model assumptions. The reported hazard ratio is therefore best read as the sponsor’s stated summary measure, alongside the medians and their intervals.

What does the stratified log-rank test contribute?

The stratified log-rank test compares event-time experience between groups while accounting for the strata used in the analysis. The registry names this test and reports p = 0.003 for the primary comparison. It does not identify the stratification factors in the record. The p-value is evidence against the test’s null hypothesis under its assumptions; it does not measure the size or clinical importance of the effect. The hazard ratio and its confidence interval give information about estimated magnitude and precision that the p-value alone cannot provide.

Why do confidence intervals matter?

A confidence interval is a range of estimates compatible with the observed data and the method’s assumptions at its stated confidence level. It communicates uncertainty in a way a point estimate cannot. For the primary hazard ratio, the 95% confidence interval is 0.44 to 0.85. The interval does not include 1, the value corresponding to equal hazards in this comparison. Its endpoints still allow a range of possible relative effects, so the point estimate should not be treated as exact.

Confidence intervals for the medians describe uncertainty in each median separately. They should not be used as a substitute for an interval around a between-group contrast. Overlap or non-overlap of separate intervals is not itself the formal test of the group comparison; the registry’s reported comparison comes from the stratified log-rank analysis and Cox model.

Why can event rates and follow-up affect interpretation?

Time-to-event estimates depend on the observed event and censoring information over follow-up. Censoring means the event time is not observed within the participant’s available follow-up. Kaplan-Meier and related methods rely on assumptions about censoring, including that censoring does not systematically conceal different event risks in a way the analysis fails to address. The registry record does not describe the censoring rules or the amount and pattern of censoring.

More events generally provide more information for estimating a time-to-event comparison, while limited follow-up or few events can leave greater uncertainty. The primary result is reported through 32 months, but the record does not give the number of progression-or-death events for this analysis. It would therefore be inappropriate to infer event totals from the enrollment or analysis-group counts.

The randomized design supports a comparison of the assigned groups, but the open-label nature and the selected post-afatinib population remain important context. Randomization does not remove uncertainty, establish the effect in every subgroup, or answer outcomes not measured in the registry record. The registry also does not report the trial’s original sample-size assumptions, interim monitoring procedures, missing-data plan, or complete prespecified statistical analysis plan.

11. Confidence Intervals and Precision

The primary hazard-ratio estimate is 0.61, and its 95% confidence interval runs from 0.44 to 0.85. The interval excludes 1, while still spanning a range of possible relative hazard effects. This provides evidence of a group difference under the reported analysis, but the precise magnitude remains uncertain. The interval is more informative about that uncertainty than the point estimate alone.

The group-specific Kaplan-Meier median estimates also have uncertainty intervals: 5.06 to 6.31 months for the combination group and 1.87 to 3.94 months for chemotherapy. These are intervals around separate medians, not around the difference between groups. A reader should not subtract their endpoints or use their overlap as a replacement for the reported time-to-event comparison.

Overall survival has a 95% confidence interval for its hazard ratio from 0.76 to 1.44. That range includes 1 and is compatible with effects in either direction around the null value. The posted objective-response intervals are also estimates with uncertainty; their width reflects that the observed response data do not identify a single exact response rate or contrast.

Precision depends on the information available for each endpoint, including the number and timing of events for time-to-event analyses and the observations contributing to binary outcomes. The registry reports analysis-group sizes for the endpoints but does not report the primary event count, censoring pattern, or detailed missing-data information. Confidence intervals describe sampling uncertainty under the relevant methods; they do not account automatically for biases from outcome assessment, missing data, or limitations in the trial design.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures use assumptions stated in the registry analysis: two-sided significance level 0.05, equal allocation between two arms, and a normal approximation. They use the observed effect as a starting point and consider detecting 100%, 75%, or 50% of that observed effect. The original trial’s own design assumptions are not reported in the registry record, so these figures should be read as planning illustrations rather than as a reconstruction of the trial’s design.

For time-to-event outcomes, the entries are total events required, not participants per arm. For the binary outcome, the entries are participants per arm. The 80% power column represents the planned chance of detecting the stated effect under the listed assumptions if that effect is real; the 90% power column uses a higher target chance of detection. Neither column is a guarantee that a future study will produce a statistically significant result.

The smaller the effect chosen for planning, the more information is generally needed to distinguish it from the null under otherwise fixed assumptions. That is why the scenarios for a smaller fraction of the observed effect call for more events or participants. A sample-size curve expresses this planning relationship visually; its meaning depends on the same assumptions and should not be mistaken for observed trial outcomes.

These figures are not observed power. They are hypothetical planning calculations based on the observed effect and the specified assumptions. They do not establish that the completed trial had a particular probability of success, and they do not replace a prospective design calculation using assumptions chosen before a new study begins.

13. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

d is the total number of events needed (Schoenfeld's formula) with 1:1 allocation and proportional hazards.

Worked calculation for this trial

1. Hazard ratio HR = 0.610, so ln(HR) = -0.4943.
2. d = 4 × (1.96 + 0.842)² / [ln(0.610)]² = 128.5, rounded up to 129 events in total for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
hr <- 0.61                              # observed hazard ratio
events_needed <- function(power, frac = 1) ceiling(round(4 * (z_a + qnorm(power))^2 / log(hr^frac)^2, 6))   # Schoenfeld, 1:1 allocation
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: events = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(events_needed(0.80, f), big.mark = ","), format(events_needed(0.90, f), big.mark = ",")))

# Simulation: given d total events under proportional hazards, events in the treatment arm are Binomial(d, hr / (1 + hr)).
# The score (log-rank type) statistic is (D1 - d / 2) / sqrt(d / 4).
set.seed(1085136)
d    <- events_needed(0.80)
reps <- 20000
D1 <- rbinom(reps, d, hr / (1 + hr))
z  <- (D1 - d / 2) / sqrt(d / 4)
cat(sprintf("Simulated power with %s events: %.3f (%d replicates)\n", format(d, big.mark = ","), mean(abs(z) > z_a), reps))

Output from running the code

100% of observed effect: events = 129 (80% power), 173 (90% power)
 75% of observed effect: events = 229 (80% power), 306 (90% power)
 50% of observed effect: events = 514 (80% power), 689 (90% power)
Simulated power with 129 events: 0.798 (20000 replicates)

The simulation uses the fact that, with d events in total and proportional hazards, the number of events in the treatment arm is binomial. It is an approximation that conditions on the total number of events, and it should land close to the target power.

14. Group-Sequential Design Analysis

Many time-to-event trials, especially in oncology, stop or report early when interim data are overwhelming, so a single fixed-sample number understates the real design problem. The calculation below is an illustrative group-sequential design built from this trial's posted control median and hazard ratio. The registry record does not state the planned number of looks or the spending function, so the design is a planning illustration, not a reconstruction of the protocol. Assumed: one-sided α = 0.025, 90% power, three equally spaced looks on information, Lan-DeMets O'Brien-Fleming spending, 24 months of accrual, 60 months in total and a 1% monthly dropout hazard.

Under these assumptions the design needs about 184 participants in total, with the boundaries below.

Fixed design versus group-sequential alternatives

Group-sequential designs cost a small increase in maximum sample size and events compared with the fixed design, in exchange for the chance to stop early for overwhelming efficacy. O'Brien-Fleming spending keeps early boundaries very conservative and the final boundary close to the fixed-design value, whereas Pocock spending spends more alpha early.

Sensitivity to a smaller true effect

Planning for exactly the observed effect is optimistic, because observed effects in published trials tend to be somewhat larger than the true effect. The table repeats the main design (three looks, O'Brien-Fleming spending, 90% power) for smaller true hazard ratios.

For context, this trial enrolled 1,154 participants.

Cross-check, upper boundaries: gsDesign 3.71, 2.511, 1.993 | rpact 3.71, 2.511, 1.993. Agreement between the two packages confirms the boundary calculation.

R code and output

suppressWarnings(suppressMessages({ library(gsDesign); library(rpact) }))
med_control <- 2.89          # control-arm median, months (converted from the posted unit)
hr          <- 0.61          # hazard ratio used for planning
lambdaC     <- log(2) / med_control
eta         <- 0.01               # dropout hazard per month (assumption)
acc_months  <- 24                 # accrual duration (assumption)
study_months<- 60                 # total study duration (assumption)

design <- function(k, sfu, beta, h = hr) gsSurv(k = k, test.type = 1, alpha = 0.025, beta = beta, sfu = sfu,
  lambdaC = lambdaC, hr = h, hr0 = 1, eta = eta, gamma = 1, R = acc_months, T = study_months, minfup = study_months - acc_months)

# Main design: 3 looks, Lan-DeMets O'Brien-Fleming spending, one-sided alpha 0.025, 90% power
x <- design(3, sfLDOF, 0.10)
cat(sprintf("N_TOTAL|%s\n", format(ceiling(sum(x$gamma * x$R)), big.mark = ",")))
for (i in 1:3) cat(sprintf("GS|%d|%.1f|%.0f|%.3f|%.5f|%.5f\n", i, x$T[i], x$n.I[i], x$upper$bound[i], pnorm(-x$upper$bound[i]), cumsum(x$upper$spend)[i]))
cat("\n"); print(gsBoundSummary(x))

# Comparison of designs
cmp <- list(list("Fixed design, one analysis", 1, sfLDOF), list("Two looks, O'Brien-Fleming", 2, sfLDOF),
            list("Three looks, O'Brien-Fleming", 3, sfLDOF), list("Three looks, Pocock", 3, sfLDPocock))
cat("\n")
for (b in c(0.20, 0.10)) for (d in cmp) {
  y <- design(d[[2]], d[[3]], b)
  cat(sprintf("DES|%s|%d|%.0f|%s|%.1f\n", d[[1]], round(100 * (1 - b)), tail(y$n.I, 1), format(ceiling(sum(y$gamma * y$R)), big.mark = ","), tail(y$T, 1)))
}

# Sensitivity: what if the true effect is smaller than the observed one? (3 looks, O'Brien-Fleming, 90% power)
cat("\n")
for (fr in c(1, 0.75, 0.5)) {
  y <- design(3, sfLDOF, 0.10, exp(log(hr) * fr))
  cat(sprintf("SENS|%.2f|%.3f|%.0f|%s\n", fr, exp(log(hr) * fr), tail(y$n.I, 1), format(ceiling(sum(y$gamma * y$R)), big.mark = ",")))
}

# Independent cross-check of the boundaries with rpact
r <- getDesignGroupSequential(kMax = 3, alpha = 0.025, beta = 0.10, sided = 1, typeOfDesign = "asOF", informationRates = x$timing)
cat(sprintf("\nCross-check, upper boundaries: gsDesign %s | rpact %s\n", paste(round(x$upper$bound, 3), collapse = ", "), paste(round(r$criticalValues, 3), collapse = ", ")))

Method: LachinFoulkes 
    Analysis               Value Efficacy
   IA 1: 33%                   Z   3.7103
       N: 98         p (1-sided)   0.0001
  Events: 58        ~HR at bound   0.3771
   Month: 13    P(Cross) if HR=1   0.0001
             P(Cross) if HR=0.61   0.0338
   IA 2: 67%                   Z   2.5114
      N: 162         p (1-sided)   0.0060
 Events: 116        ~HR at bound   0.6271
   Month: 21    P(Cross) if HR=1   0.0060
             P(Cross) if HR=0.61   0.5603
       Final                   Z   1.9930
      N: 184         p (1-sided)   0.0231
 Events: 174        ~HR at bound   0.7390
   Month: 60    P(Cross) if HR=1   0.0250
             P(Cross) if HR=0.61   0.9000



Cross-check, upper boundaries: gsDesign 3.71, 2.511, 1.993 | rpact 3.71, 2.511, 1.993

15. Analyses Reported by the Sponsor

The sponsor-reported primary analysis gives a Cox proportional-hazards estimate of 0.61, a 95% confidence interval from 0.44 to 0.85, and p = 0.003 using a stratified log-rank test. The posted primary endpoint summaries use the same group comparison and report the Kaplan-Meier medians. No discrepancy is flagged between the reported primary analysis and the posted results.

For objective response, the registry reports a sponsor analysis using logistic regression, with an odds ratio of 2.980 and a 95% confidence interval from 1.357 to 6.543. The simple unadjusted re-calculation expresses the contrast as a risk difference, which is a different measure from an odds ratio. These measures use the same underlying response outcome but answer the comparison on different scales, so their numerical values should not be compared as if they were interchangeable. The sponsor’s own analysis takes precedence if a discrepancy is identified; the registry does not flag one.

For overall survival, the sponsor-reported hazard ratio is 1.04, with a 95% confidence interval from 0.76 to 1.44 and p = 0.790. The registry names a stratified log-rank test. These sponsor analyses are the trial’s reported inferential results; simple recalculations from posted summaries do not reproduce all model details or replace the sponsor’s methods.

16. Multiplicity, Missing Data and Analysis Populations

The registry identifies three reported analyses for the primary comparison and lists several secondary outcomes, including overall survival, objective response, and adverse events. It does not report a multiplicity-control strategy, a hierarchy for testing endpoints, or which analyses were designated confirmatory. Without that information, the risk of false-positive findings across multiple statistical tests cannot be assessed from the record. The primary result should be distinguished from secondary findings, which may have been examined among several outcomes.

The primary analysis population is defined as all randomized participants, whether or not they were treated. This is an important population definition, but the registry record does not give a detailed account of how missing endpoint data were handled, the censoring rules, or the number of participants with incomplete follow-up. It also does not report sensitivity analyses for missing data or departures from analysis assumptions. These details matter because time-to-event estimates rely on how incomplete follow-up is represented, while response analyses depend on how participants without an evaluable assessment are treated.

The reported primary endpoint uses progression or death and identifies RECIST Version 1.1 for assessing progression. The record does not specify whether progression was assessed by investigators alone or independently reviewed, nor does it give the full assessment procedures. With an open-label trial, those details would help readers assess how outcome measurement might have been influenced by knowledge of assigned treatment. This is a limitation in what can be evaluated from the registry record, not evidence that a particular bias occurred.

The registry does not report the original statistical analysis plan, interim analysis procedures, stopping rules, or the trial’s own sample-size justification. These elements should not be inferred from the posted results or from the planning illustrations. Their absence leaves uncertainty about prespecification and the handling of multiple analyses, but does not alter the fact that the primary comparison and sponsor-reported analysis are stated in the record.

17. 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: dyspnoea (5.8%), general physical health deterioration (5.2%), diarrhoea (5.1%), pneumonia (5.0%), pleural effusion (4.2%).

The adverse-event record separates serious adverse events from other adverse events and gives distinct denominators by treatment period and group. For Part A afatinib monotherapy, the record lists 1154 participants at risk for serious adverse events and 471 affected; it lists 1154 at risk for other adverse events and 1129 affected. For Part B, the afatinib-plus-paclitaxel group has 134 at risk for serious adverse events and 54 affected, and 134 at risk for other adverse events and 128 affected. The investigator’s-choice chemotherapy group has 60 at risk for serious adverse events and 19 affected, and 60 at risk for other adverse events and 50 affected.

These are counts of participants recorded as affected, not measures of event severity beyond the registry’s serious-versus-other classification. The Part B safety denominators differ from the efficacy analysis group sizes, so the two sets of figures should not be treated as if they describe identical analysis populations. The registry lists dyspnoea, general physical health deterioration, diarrhoea, pneumonia, pleural effusion, and malignant neoplasm progression among the leading serious-event terms, with percentages attached to the terms. The record does not give a treatment-group breakdown for those listed percentages. Death counts are not reported in the adverse-event fields.

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

The primary result is evidence of a difference in progression-free survival between the randomized Part B groups in the population studied. The sponsor-reported hazard ratio and its confidence interval indicate a lower estimated event hazard for afatinib plus paclitaxel relative to investigator’s choice of chemotherapy under the reported analysis. The Kaplan-Meier medians give additional descriptions of the group-specific event-time distributions.

This result does not establish an exact effect for every participant. A hazard ratio is a group-level summary over follow-up, not an individual prediction, and the confidence interval shows uncertainty in its magnitude. It also does not mean that every participant in one group had a better outcome than every participant in the other group. Outcomes vary within groups, and the posted summary statistics do not describe all that variation.

The result does not establish the same pattern for people outside this trial’s defined setting. Part B followed afatinib monotherapy in participants whose disease had progressed, and the trial description identifies prior failure of erlotinib or gefitinib. The findings should therefore be understood in that treatment-history context rather than generalized automatically to other populations or treatment sequences.

The secondary outcomes do not all point to the same level of evidence for a group difference. Objective response favors the combination in its reported comparison, while the overall-survival analysis is not conclusive about a difference. Because secondary analyses may involve multiplicity and the registry does not report a correction strategy, they warrant more cautious interpretation than a prespecified primary endpoint.

Finally, these statistical results do not by themselves establish a complete benefit-harm profile. The adverse-event record supplies counts and selected serious-event terms, but does not resolve every safety question or provide all details needed to compare event patterns. The findings describe the registry-reported outcomes and their statistical uncertainty; they are not medical advice.

19. Why This Trial Matters Statistically

This trial is statistically notable because it compares a combination strategy with investigator’s choice of chemotherapy in a randomized Part B population selected after afatinib monotherapy. The primary endpoint is time-to-event, so the analysis must account for varying follow-up and participants whose progression or death was not observed before follow-up ended. Kaplan-Meier summaries and a hazard-ratio comparison address related but distinct features of that endpoint.

The results also illustrate why a point estimate, an interval, and a p-value should be read together. The hazard ratio communicates a relative event-rate summary, the confidence interval conveys uncertainty about its magnitude, and the p-value reflects the result of a specified test under its assumptions. None of these statistics alone describes the full clinical experience or establishes the importance of an effect.

There is also a useful distinction between the primary and secondary evidence. The primary progression-free survival analysis shows a group difference under the reported method, while the secondary outcomes do not all show the same pattern. That contrast makes multiplicity and endpoint-specific interpretation important: response, progression-free survival, and overall survival are not interchangeable outcomes, and multiple tests can complicate interpretation when their relationship is not prespecified or reported.

Finally, the registry record makes visible both what can be interpreted and what remains unknown. It reports the primary comparison, analysis population definition, and several secondary and safety results, but does not provide the complete analysis plan, missing-data procedures, or original design assumptions. For biostatistical readers, that boundary is part of the result: sound interpretation depends not only on estimates but also on the methods and reporting detail available to evaluate them.

20. 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.

Several limitations are specific to the information recorded for this trial. The study was open-label, and progression-free survival includes disease progression assessed using RECIST Version 1.1. The registry does not report whether progression assessments underwent independent review or provide detailed procedures for those assessments. This leaves readers with limited information for judging the potential influence of knowledge of treatment assignment on outcome ascertainment.

The randomized Part B comparison concerns participants whose disease progressed after afatinib monotherapy, in the context of prior erlotinib or gefitinib failure described in the trial record. That selection narrows the setting to which the comparison directly applies. The registry also gives different group sizes for the Part B efficacy and safety records; it does not explain the difference in those denominators, so the safety counts should be interpreted within their own reported populations.

For the primary endpoint, the record reports medians, confidence intervals, and a sponsor hazard ratio, but not the event count, censoring distribution, or detailed follow-up completeness. The maximum timeframe is up to 32 months. The registry does not report diagnostics for the proportional-hazards assumption or details of the stratification factors used in the test.

Important design and analysis information is also absent: the original sample-size justification, allocation procedures beyond randomization, interim monitoring details, missing-data handling, and multiplicity strategy are not reported in the registry record. The absence of these details limits independent assessment of prespecification and robustness. It does not justify assumptions about how the investigators handled those issues.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: LUX-Lung 5: Afatinib Plus Weekly Paclitaxel Versus Investigator's Choice of Single Agent Chemotherapy Following Afatinib Monotherapy in Non-small Cell Lung Cancer Patients Failing Erlotinib or Gefitinib — NCT01085136.
• Registry record last updated 2017-04-04; status last verified 2017-02.
• Results first posted 2014-10-03.
• 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.

24. Record Summary

LUX-Lung 5 is recorded as a completed, randomized, open-label Phase 3 trial in non-small cell lung cancer. Its Part B comparison evaluated afatinib plus weekly paclitaxel against investigator’s choice of chemotherapy after afatinib monotherapy. The primary endpoint was progression-free survival from randomization until progression or death, assessed using RECIST Version 1.1 and summarized with Kaplan-Meier medians.

The sponsor-reported primary hazard ratio was 0.61, with a 95% confidence interval from 0.44 to 0.85 and p = 0.003. Secondary results include overall survival and objective response, with different patterns across endpoints. The registry also reports adverse-event counts, but does not provide all details needed to assess multiplicity, missing data, outcome review, or the original design assumptions. Interpretation is therefore strongest when kept to the randomized Part B population and the methods explicitly recorded.