Avelumab in First-line NSCLC (JAVELIN Lung 100): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Avelumab in First-line NSCLC (JAVELIN Lung 100) (NCT02576574), a randomized phase 3 study. The primary outcome is Progression Free Survival (PFS) According to Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST v1.1) Assessed by Independent Review Committee (IRC) in High Programmed Death Ligand 1 (PD-L1) + Full Analysis Set (FAS).

2. Clinical Question

3. Trial Design

This was a Phase III, randomized, open-label, multicenter, parallel-group trial of avelumab versus platinum-based doublet chemotherapy as first-line treatment for recurrent or Stage IV PD-L1+ non-small cell lung cancer. The registry lists total enrollment of 1214 and a completed study. Participants were not masked to treatment assignment. The randomization ratio and details of the allocation process are not reported in the registry record.

The study included avelumab biweekly, avelumab weekly, and chemotherapy groups. The primary endpoint record reports the comparison between avelumab biweekly and chemotherapy in the high PD-L1+ Full Analysis Set. The registry’s stated study purpose refers to demonstrating superiority with respect to overall survival or progression-free survival, with progression-free survival assessed by an Independent Review Committee. The primary result discussed in this analysis is the reported progression-free survival result, not an overall survival primary result.

The chemotherapy intervention list includes pemetrexed, paclitaxel, gemcitabine, carboplatin, and cisplatin. The registry record does not specify which combinations participants received, nor does it provide descriptions of the individual arm regimens. It also does not report the trial’s original sample-size justification, randomization ratio, or detailed statistical analysis plan. Those omissions limit what can be inferred about the trial’s planned hypotheses and analysis procedures from the registry record alone.

The primary endpoint assessment used RECIST v1.1 and was based on the Independent Review Committee assessment. The stated timeframe runs from randomization to the data cutoff, with assessment up to approximately 71.5 months. This is a randomized comparison, but its open-label design means participants and treating teams were not masked; the registry does not report whether other parts of study conduct were masked.

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

Analysis population for the primary outcome, as defined in the registry: High PD-L1+ FAS included all high expression PD-L1+ participants who were randomized to study intervention. High expression PD-L1+ participants with greater than or equal to (\>=) 80 percent (%) of tumor cells deemed positive for PD-L1 expression at any staining intensity (1+, 2+, or 3+).

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.

6. Primary Endpoint

The registry lists the primary outcome as Progression Free Survival (PFS) According to Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST v1.1) Assessed by Independent Review Committee (IRC) in High Programmed Death Ligand 1 (PD-L1) + Full Analysis Set (FAS), measured over Time from date of randomization up to data cutoff (assessed up to approximately 71.5 months).

The primary endpoint is progression-free survival, measured in months from randomization until the first documentation of progressive disease or death from any cause in the absence of documented progression, whichever occurred first. Progression was defined using RECIST v1.1: an increase in the sum of the longest diameters or the appearance of new lesions. The registry says progression-free survival was estimated using the Kaplan–Meier method and assessed by an Independent Review Committee.

In general, a time-to-event endpoint records both whether an event occurred and the time until it occurred. Progression-free survival combines disease progression and death into one endpoint, so a participant’s time ends at whichever qualifying event occurs first. This is useful in trials where the timing of disease progression is an important outcome and where waiting for deaths alone could leave the clinical question unresolved for longer. That general rationale does not establish why this particular endpoint was selected; the protocol’s endpoint-selection rationale is not reported in the registry record.

The median is the estimated time by which half of the population has experienced the event, according to the time-to-event curve. It is not an average duration for every participant, and it does not describe the outcome of any one individual. The registry’s primary comparison also reports a hazard ratio, which summarizes the relative event rate over follow-up under the assumptions of the analysis. The median and hazard ratio therefore describe different aspects of the time-to-event data.

7. Results Posted in the Registry

• Avelumab Biweekly vs Chemotherapy: hazard ratio 0.71 (95% CI 0.54–0.93), reported p = 0.007 (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.

Progression Free Survival (PFS) According to Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST v1.1) Assessed by Independent Review Committee (IRC) in Moderate and High Programmed Death Ligand 1 (PD-L1)+ Full Analysis Set (FAS)

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Log Rank); it was not recalculated.
• Result:
  • Avelumab Biweekly vs Chemotherapy: hazard ratio 0.78 (95% CI 0.62–0.98), reported p = 0.015 (as reported in the registry).
• Interpretation:
  • The 95% confidence interval excludes no difference (a ratio of 1), so the hazard of the event is statistically lower for Avelumab Biweekly vs Chemotherapy than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Progression Free Survival (PFS) According to Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST v1.1) Assessed by Independent Review Committee (IRC) in Moderate and High Programmed Death Ligand 1 (PD-L1)+ Modified Full Analysis Set (mFAS)

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Log Rank); it was not recalculated.
• Result:
  • Avelumab Weekly vs Chemotherapy: hazard ratio 0.88 (95% CI 0.67–1.15), reported p = 0.175 (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 Avelumab Weekly vs Chemotherapy and the control. The interval gives the range of effects that remain compatible with the data.

Overall Survival (OS) in Moderate and High Programmed Death Ligand 1 (PD-L1)+ Full Analysis Set (FAS)

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Log Rank); it was not recalculated.
• Result:
  • Avelumab Biweekly vs Chemotherapy: hazard ratio 0.82 (95% CI 0.66–1.00), reported p = 0.026 (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 Avelumab Biweekly vs Chemotherapy and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists secondary outcomes spanning progression-free survival, overall survival, and confirmed objective response, with analyses across different PD-L1-defined populations and analysis sets. The available secondary comparisons do not all present the same pattern: some reported hazard-ratio estimates favor the avelumab group, while another progression-free survival comparison has an interval that includes 1. This variation is a reason to interpret secondary findings individually and in context rather than treating them as one consistent result across all populations and endpoints.

The secondary results relate to the primary result by examining related time-to-event outcomes, broader or modified populations, and other measures of tumor response. They can help describe whether the primary pattern is seen across different analyses, but they do not change the definition or population of the primary endpoint. Differences in populations, treatment groups, and endpoints mean that the results are not interchangeable.

There are 28 secondary results posted. When many endpoints, populations, and comparisons are examined, the chance of finding apparently notable results by chance increases unless multiplicity is addressed. The registry does not report whether secondary p-values were adjusted for multiple comparisons, or how any testing hierarchy was defined. Unadjusted secondary results therefore carry less weight than a prespecified primary result with a defined analysis, especially when they are used to make claims beyond their individual populations.

The secondary results are most informative as context for the primary finding and as a record of the range of outcomes assessed. They should not be selected only by whether an individual p-value crosses a threshold. The registry does not report enough detail to determine whether each secondary analysis was prespecified in the final analysis plan, whether all planned analyses are posted, or how missing data were handled for each endpoint.

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

• Progression Free Survival (PFS) According to Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST v1.1) Assessed by Independent Review Committee (IRC) in Moderate and High Programmed Death Ligand 1 (PD-L1)+ Full Analysis Set (FAS)
• Progression Free Survival (PFS) According to Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST v1.1) Assessed by Independent Review Committee (IRC) in Moderate and High Programmed Death Ligand 1 (PD-L1)+ Modified Full Analysis Set (mFAS)
• Overall Survival (OS) in Moderate and High Programmed Death Ligand 1 (PD-L1)+ Full Analysis Set (FAS)
• Overall Survival (OS) in Moderate and High Programmed Death Ligand 1 (PD-L1)+ Modified Full Analysis Set (mFAS)
• Overall Survival (OS) in Full Analysis Set (FAS)
• Overall Survival (OS) in Modified Full Analysis Set (mFAS)
• Percentage of Participants With Confirmed Objective Response According to Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST v1.1) as Assessed by Independent Review Committee (IRC) in High PD-L1+ Full Analysis Set
• Percentage of Participants With Confirmed Objective Response According to Response Evaluation Criteria in Solid Tumors Version 1.1 (RECIST v1.1) as Assessed by Independent Review Committee (IRC) in High PD-L1+ Modified Full Analysis Set

9. Statistical Interpretation

10. Statistical Methods Explained

Time-to-event analysis is used when the outcome is the time until an event, such as documented disease progression or death. Some participants may not have experienced the event by the data cutoff. In general, these participants contribute information up to the time they were last known to be event-free; this is called censoring. The exact censoring rules and details of handling missing or incomplete follow-up are not reported in the registry record.

What does the Kaplan–Meier median describe?

The registry reports Kaplan–Meier estimates of median progression-free survival. In general, the Kaplan–Meier method estimates the probability of remaining event-free over time while accounting for participants whose event times are not observed during the available follow-up. The median is the time at which the estimated event-free probability reaches one half. It summarizes the curve at a single point, whereas the curve contains information across follow-up. A median alone does not show the full shape of the event-time distribution or whether the relative event rates changed over time.

The reported medians are 8.4 months for avelumab biweekly and 5.6 months for chemotherapy in the primary population. Their confidence intervals describe uncertainty in each median estimate. These median estimates should not be confused with the hazard ratio: the medians compare selected points on the Kaplan–Meier curves, while the hazard ratio summarizes a relative event rate over the analysis period.

How should the hazard ratio be read?

The registry reports a hazard ratio of 0.71 for avelumab biweekly versus chemotherapy, estimated using a log-rank analysis, with a 95% confidence interval from 0.54 to 0.93. In general, the hazard is the instantaneous event rate among participants who have not yet experienced the event. A hazard ratio compares those rates between groups. A value below 1 indicates a lower estimated event rate for the first group relative to the comparator under the analysis; it is not a ratio of median survival times, a percentage of participants who avoided an event, or a direct prediction for an individual.

The log-rank method is a general method for comparing time-to-event experience between groups. The registry identifies it as the method for the reported comparison. It does not report the full model specification, assessment of the proportional-hazards assumption, covariate adjustment, or any sensitivity analyses. These details should not be inferred from the hazard ratio alone.

What do the p-value and confidence interval contribute?

The p-value of 0.007 describes how compatible the observed data are with a null comparison under the statistical test’s assumptions. It is not the probability that the null hypothesis is true, nor is it a measure of the clinical importance of the result. The 95% confidence interval from 0.54 to 0.93 gives a range of values compatible with the data under the method used. In general, a confidence interval conveys both the direction and precision of an estimate: a narrower interval indicates greater precision than a wider interval, while an interval’s position relative to 1 is relevant for a hazard ratio.

Statistical significance and precision are related but distinct. A p-value can indicate evidence against a null value without showing whether the range of plausible effects is narrow enough to support a precise estimate. Interpretation should consider the hazard ratio, its interval, the endpoint definition, the study population, and the analysis context together.

Why can over-dispersion matter in other analyses?

Over-dispersion is a general issue in analyses of counts or rates when the observed variability is greater than the model assumes. If it is ignored, standard errors may be too small and confidence intervals or tests may appear more certain than warranted. The primary endpoint is time-to-event, and the registry identifies a log-rank comparison rather than a count model. Over-dispersion is not reported as part of this primary analysis, so it should not be treated as a documented feature of the result.

What analysis population is used?

The primary High PD-L1+ Full Analysis Set included all high-expression PD-L1+ participants randomized to study intervention, with high expression defined as at least 80% of tumor cells positive at any staining intensity. The primary comparison includes 151.0 participants in the avelumab biweekly group and 216.0 in the chemotherapy group. The registry also lists modified Full Analysis Set analyses among secondary outcomes, but the definition of the modified set is not reported in the registry record. The primary result therefore should be interpreted using its stated population definition, rather than assuming that all analysis populations include the same participants.

11. Confidence Intervals and Precision

The primary hazard ratio is 0.71, with a 95% confidence interval from 0.54 to 0.93. The interval indicates that the estimate is not exact: values across that range remain compatible with the observed data under the reported method. Because the interval lies below 1, the reported comparison favors a lower estimated event rate in the avelumab biweekly group under the hazard-ratio interpretation. The interval does not establish that the true value equals the point estimate, and it does not quantify the probability that a particular participant will experience progression or death.

The Kaplan–Meier median for avelumab biweekly is 8.4 months, with a 95% confidence interval from 5.4 to 12.6 months. The chemotherapy median is 5.6 months, with a 95% confidence interval from 5.4 to 6.8 months. These intervals describe uncertainty around separate median estimates. They are not a confidence interval for the difference between the medians, and the registry does not report a direct interval for that difference.

Precision depends on the available event information, follow-up, and variability, among other factors. The primary analysis group sizes are 151.0 and 216.0, but group size alone does not determine the precision of a time-to-event estimate. The registry states that assessment extended up to approximately 71.5 months, while the number of primary endpoint events is not reported in the registry record. The confidence intervals are therefore the direct reported guide to uncertainty for these estimates; they should be read alongside the population definition and follow-up timeframe.

A confidence interval is not a boundary that guarantees the true value falls inside it. In general, the stated confidence level describes the long-run behavior of the method across repeated applications under its assumptions. The interval also does not address all possible sources of uncertainty, such as bias from study conduct or limitations in the analysis plan. The registry record does not report enough detail to assess every such source.

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 illustrations translate an assumed effect and target power into a required amount of information. For the time-to-event comparisons, the displayed requirement is the total number of events across the two groups, not participants per arm. The event requirement is calculated using the Schoenfeld approach. For other endpoint types, the planning convention is participants per arm; those participant counts are not totals across both groups.

The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. Each illustration uses a stated fraction of the observed effect as the effect to detect. The 80% power column describes a planning target under which the assumed effect would be detected with 80% probability if the assumptions and model were correct. The 90% power column uses a higher detection probability target. These are planning quantities, not guarantees about a particular future study.

Rows based on 100%, 75%, and 50% of the observed effect illustrate how requirements change when the assumed effect is smaller than the observed estimate. Smaller effects are harder to distinguish from no effect, so they require more events for time-to-event outcomes or more participants for other outcome types. The illustrations are computed from the observed effect and are not the trial’s original design justification. They are not observed power.

For time-to-event outcomes, the event total is the relevant planning quantity because the number of events contains key information for estimating a hazard ratio. It is not interchangeable with the number enrolled: the event count depends on follow-up and how many participants experience the endpoint. The registry’s original planning assumptions and sample-size justification are not reported in the registry record.

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 = 0.710, so ln(HR) = -0.3425.
2. d = 4 × (1.96 + 0.842)² / [ln(0.710)]² = 267.7, rounded up to 268 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 <- 0.71                              # 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(2576574)
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 = 268 (80% power), 359 (90% power)
 75% of observed effect: events = 476 (80% power), 637 (90% power)
 50% of observed effect: events = 1,071 (80% power), 1,434 (90% power)
Simulated power with 268 events: 0.786 (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 400 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 1,214 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 <- 5.6          # control-arm median, months (converted from the posted unit)
hr          <- 0.71          # 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: 252         p (1-sided)   0.0001
 Events: 121        ~HR at bound   0.5088
   Month: 15    P(Cross) if HR=1   0.0001
             P(Cross) if HR=0.71   0.0338
   IA 2: 67%                   Z   2.5114
      N: 400         p (1-sided)   0.0060
 Events: 242        ~HR at bound   0.7237
   Month: 24    P(Cross) if HR=1   0.0060
             P(Cross) if HR=0.71   0.5603
       Final                   Z   1.9930
      N: 400         p (1-sided)   0.0231
 Events: 362        ~HR at bound   0.8109
   Month: 60    P(Cross) if HR=1   0.0250
             P(Cross) if HR=0.71   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 registry reports the sponsor’s primary hazard ratio as 0.71, with a 95% confidence interval from 0.54 to 0.93 and p = 0.007. The reported method is log-rank analysis. These are the figures to use when describing the sponsor’s reported primary comparison.

Any simple recalculation from posted summary statistics is unadjusted and should not replace the sponsor’s reported analysis. No discrepancy is identified between the reported analysis and the posted comparison in this record. A summary-statistic recalculation cannot reproduce details that depend on individual participant event times, censoring, or the full analysis procedure.

The registry does not report the complete analysis code, individual participant data, or the full statistical analysis plan. Consequently, a recalculation based on summary information is best understood as a check or illustration, not as an alternative primary analysis. For interpreting the trial result, the sponsor’s reported estimate and its stated uncertainty remain the central analysis.

16. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when several endpoints, populations, treatment comparisons, or analysis choices are tested. In general, conducting many tests without an adjustment increases the chance that at least one result will appear statistically notable even when the underlying effects are absent. This matters in this trial because the registry lists secondary progression-free survival, overall survival, and objective response outcomes across several analysis populations, in addition to the primary endpoint.

The registry’s stated study purpose refers to superiority with respect to overall survival or progression-free survival. However, the detailed testing hierarchy, allocation of significance levels, and any multiplicity adjustment are not reported in the registry record. It is therefore not possible from the record alone to determine how the primary and secondary hypotheses were ordered or whether the reported p-values account for all planned comparisons. Secondary results should consequently be viewed as supporting context rather than treated as independently confirmatory without further information.

The primary analysis population is defined as the High PD-L1+ Full Analysis Set, consisting of randomized participants with high expression in at least 80% of tumor cells at any staining intensity. The registry also lists modified Full Analysis Set results, but the modified set’s inclusion and exclusion rules are not reported in the registry record. The meaning of a difference between the Full Analysis Set and modified Full Analysis Set results cannot be fully assessed without those definitions.

Missing-data procedures are also not reported in the registry record. In time-to-event analysis, censoring is expected when a participant’s event time is not observed through the data cutoff, but the registry does not provide the detailed censoring rules, handling of loss to follow-up, or sensitivity analyses. The number of missing endpoint assessments and the approach to other missing outcome information are not reported. These gaps limit assessment of how robust the findings are to missingness or analysis choices.

The record gives the primary endpoint definition, analysis population, timeframe, Kaplan–Meier method, and reported log-rank comparison, but not the full protocol or statistical analysis plan. The original sample-size justification, interim-analysis procedures, and detailed prespecification of analyses are not reported in the registry record. Their absence should be treated as a limitation of the public record, not evidence that a particular procedure was or was not used.

17. 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

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: disease progression (9.1%), pneumonia (7.8%), anaemia (3.8%), pleural effusion (3.0%), infusion related reaction (3.0%).

The registry reports serious adverse event counts for three groups. In the avelumab biweekly group, 181 participants were affected among 361 at risk; in the avelumab weekly group, 143 were affected among 318 at risk; and in the chemotherapy group, 195 were affected among 500 at risk. These are group-level counts as recorded in the registry, not a comparison of the primary endpoint population. The registry also reports other adverse events affecting 331 of 361 at risk in the avelumab biweekly group, 298 of 318 in the weekly group, and 474 of 500 in the chemotherapy group.

The most frequently listed serious adverse event terms include disease progression at 9.1%, pneumonia at 7.8%, anaemia at 3.8%, pleural effusion at 3.0%, infusion related reaction at 3.0%, and febrile neutropenia at 2.6%. The record reports deaths of 279 in the avelumab biweekly group, 237 in the avelumab weekly group, and 405 in the chemotherapy group. These registry summaries do not by themselves establish whether an event was caused by treatment or provide a complete account of timing, duration, or clinical context. The definitions and attribution rules for these safety summaries are not reported in the registry record, so the figures should be read as descriptive counts rather than as a causal comparison.

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

The primary result describes progression-free survival in a specific randomized population: participants with high PD-L1 expression, defined as at least 80% of tumor cells positive at any staining intensity. It compares avelumab biweekly with platinum-based doublet chemotherapy, using the Independent Review Committee’s RECIST v1.1 assessment. The reported hazard ratio, medians, confidence intervals, and p-value summarize that comparison in the registry record.

The result does not describe every participant enrolled in the trial, every avelumab schedule, or all people with non-small cell lung cancer. It also does not directly describe overall survival, despite overall survival being named among the study’s aims and secondary outcomes. A progression-free survival result is not a substitute for a separately analyzed overall survival result.

In general, a group-level hazard ratio is not an individual prediction, and a median is not a promise about how long any one participant will remain free of progression or death. A statistically notable result does not alone establish the practical importance of an effect, nor does it address treatment decisions for an individual. Such judgments require information beyond this statistical comparison. No medical advice follows from the registry statistics.

The reported p-value and confidence interval do not remove uncertainty. They characterize the estimate under the stated analysis and assumptions, while other uncertainties can arise from population selection, study conduct, missing-data handling, and analysis choices. Some of these details, including the full multiplicity plan and the modified population definition, are not reported in the registry record. The result should therefore be interpreted as evidence from the specified primary analysis rather than as a universal conclusion about treatment or all outcomes.

19. Why This Trial Matters Statistically

This trial illustrates several important features of clinical-trial interpretation. First, a trial can have a broad enrollment population while a primary result concerns a narrower biomarker-defined group. Here, the primary progression-free survival comparison is for the High PD-L1+ Full Analysis Set, defined using PD-L1 expression in tumor cells. Keeping the analysis population attached to the result prevents the estimate from being generalized beyond the group it describes.

Second, time-to-event results are not captured by one number alone. The Kaplan–Meier medians describe selected points in the estimated event-free curves, while the hazard ratio summarizes a relative event rate over follow-up. Confidence intervals add information about precision, and the endpoint definition determines what counts as an event. Reading these elements together is more informative than relying on a p-value alone.

Third, the registry offers several secondary outcomes and populations. Their variation highlights why prespecified analysis plans and multiplicity procedures matter: multiple comparisons can complicate interpretation, particularly when their adjustment status is unclear. The record’s missing detail about multiplicity, modified analysis populations, and missing-data procedures limits the conclusions that can be drawn from secondary results.

Finally, planning illustrations based on an observed effect are distinct from the trial’s original design. They can help readers understand how event requirements change under different assumed effects and power targets, but they cannot establish the original rationale for enrollment or demonstrate the power actually achieved. This distinction is useful whenever planning calculations accompany reported trial results.

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.

The primary result applies to the high PD-L1 expression population, not automatically to the full enrolled population or to all people with non-small cell lung cancer. The registry defines high expression for the primary population, but the degree to which the trial population represents other clinical settings is not reported. The primary efficacy comparison also concerns the biweekly avelumab group, so it should not be read as a result for the weekly schedule.

The trial was open-label. The primary endpoint was assessed by an Independent Review Committee, but the registry does not describe the full safeguards used for endpoint assessment or other aspects of study conduct. Open-label design can be relevant to some study procedures and decisions, even where an endpoint is independently assessed; the available record does not allow a more detailed evaluation.

The registry does not report the complete statistical analysis plan, the original sample-size justification, interim analyses, multiplicity procedures, or details of missing-data and censoring rules. It also does not define the modified Full Analysis Set used for some secondary analyses. These omissions prevent a full assessment of prespecification, robustness, and the relationship among reported analyses.

The primary record reports group sizes and summary estimates but not individual participant data or the total number of primary endpoint events. The chemotherapy intervention list names several drugs, but the specific combinations used are not reported. Safety summaries are descriptive in the record and do not provide detailed event attribution or context. These limitations do not negate the reported estimates; they define the boundaries of what can be evaluated from the registry record alone.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: Avelumab in First-line NSCLC (JAVELIN Lung 100) — NCT02576574.
• Registry record last updated 2025-01-31; status last verified 2025-01.
• Results first posted 2023-01-04.
• 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

NCT02576574 is a completed, Phase III, randomized, open-label, multicenter, parallel-group trial with enrollment of 1214. It evaluated avelumab and platinum-based doublet chemotherapy as first-line treatment for recurrent or Stage IV PD-L1+ non-small cell lung cancer. The primary endpoint record reports Independent Review Committee-assessed progression-free survival by RECIST v1.1 in the High PD-L1+ Full Analysis Set.

For avelumab biweekly versus chemotherapy, the registry reports medians of 8.4 and 5.6 months, respectively, and a hazard ratio of 0.71 with a 95% confidence interval from 0.54 to 0.93 and p = 0.007. These estimates apply to the defined primary population and should be read with their uncertainty and the registry’s limitations in mind. The record lists additional outcomes and safety summaries, but does not report several details needed for a complete evaluation of multiplicity, missing-data handling, and the full statistical plan.