A Study to Evaluate Chemotherapy Plus Osimertinib Against Chemotherapy Plus Placebo in Patients With Non-small Cell Lung Cancer (NSCLC): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Evaluate Chemotherapy Plus Osimertinib Against Chemotherapy Plus Placebo in Patients With Non-small Cell Lung Cancer (NSCLC) (NCT04765059), a randomized phase 3 study. The primary outcome is Progression Free Survival (PFS).

2. Clinical Question

The study asks whether adding osimertinib to platinum-pemetrexed chemotherapy is associated with a different time-to-event outcome than adding placebo to the same chemotherapy in this specified post-osimertinib setting. It is a comparison of treatment strategies, not a comparison of osimertinib alone with no treatment: both groups received pemetrexed and platinum chemotherapy.

The posted primary result is a hazard ratio comparing Arm A with Arm B. A value below 1 places the estimated relative hazard in the direction of fewer or later progression-or-death events in Arm A, as represented by the reported analysis. The confidence interval communicates uncertainty about that estimate. It does not, by itself, give the probability that an individual participant will benefit, nor does it establish that every participant experiences the same course.

The registry describes the population and treatment context but does not report a p-value for the primary comparison. The result should therefore be read through the reported hazard ratio and its confidence interval, alongside the median progression-free survival estimates and their intervals. The posted record does not establish how the findings apply outside the defined disease and treatment setting.

3. Trial Design

COMPEL is listed as a Phase III, randomized, double-blind, placebo-controlled, parallel-group study sponsored by AstraZeneca. The registry lists quadruple masking. Its stated purpose is to evaluate efficacy and safety for chemotherapy combined with osimertinib compared with chemotherapy combined with placebo in participants whose non-small cell lung cancer progressed extracranially after first-line osimertinib. The registry status is active, not recruiting.

Arm A combines osimertinib with pemetrexed and either cisplatin or carboplatin. The listed regimen gives osimertinib 80 mg QD and pemetrexed 500 mg/m^2, with either cisplatin 75 mg/m^2 or carboplatin AUC 5, administered on Day 1 of 21-day cycles for 4 cycles. It then describes osimertinib with pemetrexed maintenance. Arm B uses placebo for osimertinib alongside the same listed chemotherapy options and describes placebo with pemetrexed maintenance. The maintenance text in the registry record is truncated, so additional regimen detail is not reported in the registry record.

The record lists enrollment of 98. The primary results give 49 participants in each arm. The primary analysis population is identified as the full analysis set, consisting of all randomized participants. The primary endpoint is progression-free survival, and the registry also lists intracranial progression-free survival, extracranial progression-free survival, and overall survival among secondary outcomes.

The recorded study start is 2021-09-12 and the primary completion date is 2024-10-28. These are registry dates, not a description of follow-up completeness for every participant. The original design assumptions, including any target effect, formal sample-size justification, planned interim analyses, and detailed statistical analysis plan, are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: Participants meeting eligibility criteria predefined in protocol were enrolled in the study. Assessments performed as per the schedule of the assessments.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The Full analysis set consisted of all randomized participants.

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: The full analysis set consisted of all randomized participants.

6. Primary Endpoint

The registry lists the primary outcome as Progression Free Survival (PFS), measured over From date of first dose (Day 1) until date of progression at local site or death due to any cause or data cut-off date whichever occurred first (up to 3 years 1 month).

The primary endpoint is progression-free survival (PFS), reported as a median in months with a 95% confidence interval. The registry defines it as the time from randomization until progression, either intracranial or extracranial, whichever occurs first, under RECIST 1.1 for extracranial progression and CNS RECIST 1.1 for intracranial progression, as assessed by the investigator at the local site, or death due to any cause. The timeframe description refers to measurement from the first dose (Day 1) until progression at the local site or death, or data cut-off, whichever came first, up to 3 years 1 month.

PFS is a time-to-event endpoint: the analysis concerns how long participants remain alive without a recorded progression event. It is commonly used in trials of cancer because it brings progression and death into a single defined outcome and can be observed without waiting for every participant to die. That general rationale does not establish the details of follow-up or assessment procedures beyond those stated in the registry record.

The median is the time by which half of the relevant event-time distribution has experienced the endpoint, under the analysis method used; it is not an average and does not describe each participant’s individual time. The hazard ratio adds a relative comparison of event rates over follow-up. Neither measure alone gives a complete account of symptoms, quality of life, treatment burden, or overall survival. Overall survival is listed as a secondary outcome, but its results are not included among the posted endpoint estimates in this record.

7. Results Posted in the Registry

• Treatment Arm A: Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin vs Treatment Arm B: Placebo for Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin: hazard ratio 0.43 (95% CI 0.27–0.70) (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.

Intracranial Progression Free Survival in Participants With Brain Metastases at Baseline

• 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:
  • Treatment Arm A: Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin: 81.8% (9/11) versus 58.3% (7/12) on control; risk difference 23.5 percentage points (95% CI -12.5 to 59.5), risk ratio 1.40 (0.81–2.44); unadjusted two-proportion z-test p = 0.221.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Treatment Arm A: Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin and the control. The interval gives the range of effects that remain compatible with the data.

Intracranial Progression Free Survival in Participants Without Brain Metastases at Baseline

• 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:
  • Treatment Arm A: Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin: 86.8% (33/38) versus 62.2% (23/37) on control; risk difference 24.7 percentage points (95% CI 5.7 to 43.6), risk ratio 1.40 (1.06–1.85); unadjusted two-proportion z-test p = 0.014.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for Treatment Arm A: Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Extracranial Progression Free-survival

• 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:
  • Treatment Arm A: Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin vs Treatment Arm B: Placebo for Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin: hazard ratio 0.43 (95% CI 0.26–0.71) (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 Treatment Arm A: Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin vs Treatment Arm B: Placebo for Osimertinib (AZD9291), Pemetrexed, Cisplatin or Carboplatin than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary outcomes broaden the picture beyond the primary PFS endpoint. They address intracranial progression-free survival according to brain-metastasis status at baseline, extracranial progression-free survival, and overall survival. Together, these outcomes concern different event locations, a subgroup distinction, and a later survival outcome. The registry lists overall survival but does not include an estimate for it among the posted results summarized in this record.

At a broad level, the posted secondary progression outcomes generally point in the same direction as the primary comparison, but the evidence is not equally precise across outcomes or subgroups. The sponsor-reported extracranial time-to-event estimate is aligned in direction with the primary estimate. Intracranial findings are less straightforward to summarize from a single statistic because one subgroup has a small analysis population and wide percentage intervals, while the time-to-event estimate reported for the other subgroup has an interval that includes 1. These results provide context, rather than an independent confirmation of the primary finding.

Unadjusted secondary results carry less weight because multiple outcomes and subgroup comparisons create more opportunities for apparently notable findings. The registry record does not report a multiplicity adjustment plan or a prespecified hierarchy for these outcomes. An unadjusted p-value does not account for the fact that several comparisons may have been examined, and the subgroup estimates can be especially uncertain when their analysis populations are limited. The reported secondary measures also differ in type: percentages at 6 months, hazard ratios, and medians do not answer precisely the same question.

Accordingly, secondary endpoints can be used to describe whether the pattern appears consistent across disease locations, but they should not be treated as equivalent to the primary endpoint or as definitive evidence for a subgroup. The registry does not report enough inferential detail to establish how the secondary analyses were selected, adjusted, or interpreted within the trial’s formal testing strategy.

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

• Intracranial Progression Free Survival in Participants With Brain Metastases at Baseline
• Intracranial Progression Free Survival in Participants Without Brain Metastases at Baseline
• Extracranial Progression Free-survival
• Overall Survival (OS)

9. Statistical Interpretation

10. Statistical Methods Explained

What does a hazard ratio describe?

A hazard ratio compares the instantaneous rates of experiencing an event between groups among participants who have not yet experienced that event. For this primary endpoint, the event is progression or death under the registry definition. The reported hazard ratio of 0.43 compares Arm A with Arm B. A value below 1 indicates that the estimated hazard is lower in Arm A relative to Arm B in the reported contrast; it does not mean that the risk or probability is lower by the same proportion at every point in time.

Time-to-event methods retain information about when events occur and can account for participants whose event has not been recorded by the end of their observed follow-up. The registry defines a data cut-off as one possible end of the endpoint observation window. The exact censoring rules and the method for handling participants lost to observation are not reported in the registry record. It is therefore not appropriate to infer those details from the summary estimates alone.

What does the log-rank method contribute?

The primary comparison is reported using a log-rank method. In general, the log-rank test compares event experience across groups over follow-up rather than comparing only their medians. The registry reports the hazard ratio alongside this method, but does not report a p-value for the primary analysis. The statistical plan, any stratification variables, and any adjustment for covariates are not reported in the registry record.

The posted hazard ratio is the sponsor’s reported time-to-event estimate. It should not be reconstructed from the two medians: medians contain only a summary point from each group, while the hazard ratio uses event-time information. The reported confidence interval also depends on the sponsor’s analysis and cannot be recovered from the median values alone.

How should confidence intervals be read?

A confidence interval expresses the precision of an estimate under the analysis method. The primary 95% confidence interval runs from 0.27 to 0.70 around the reported hazard ratio of 0.43. The endpoints indicate a range of values compatible with the data and method at the stated confidence level. The interval is not a guarantee that the true value lies within it, and it does not assign probabilities to particular values after the data have been observed.

The interval’s position relative to 1 is informative for a hazard ratio because 1 represents equal hazards in the two groups. The primary interval lies below 1. This helps describe the statistical direction and precision, but it does not address clinical importance, generalizability, or whether the same relative pattern holds throughout follow-up.

Why are the binary secondary comparisons different?

The intracranial progression-free survival results are posted as percentages of participants at 6 months, separately for those with and without brain metastases at baseline. The registry also posts unadjusted risk-difference recalculations based on the group summaries. A risk difference describes an absolute difference between proportions; it is not the same measure as a hazard ratio. The recalculated confidence intervals and p-values are unadjusted summary-statistic analyses, not a substitute for a sponsor-reported time-to-event analysis.

For the baseline brain-metastasis subgroup, the reported percentages have wide confidence intervals, reflecting limited precision in the subgroup data. In the subgroup without brain metastases at baseline, the posted unadjusted risk-difference interval does not include zero and its unadjusted p-value is 0.014. For the subgroup with brain metastases at baseline, the unadjusted p-value is 0.221 and the interval for the recalculated risk difference spans zero. These unadjusted values should not be taken as definitive subgroup evidence, particularly because the registry does not report adjustment for multiple comparisons or a prespecified subgroup testing plan.

What does over-dispersion mean in general?

Over-dispersion arises when observed variability is greater than a statistical model expects. In general, ignoring over-dispersion can make uncertainty appear smaller than it is, depending on the model. The primary PFS analysis is described through time-to-event methods, and no over-dispersion model is reported. It should not be assumed that an over-dispersion adjustment was used or needed for the primary analysis.

What is not documented?

The registry record identifies the full analysis set as all randomized participants for the primary endpoint, and it gives outcome summaries and reported analyses. It does not report the detailed statistical analysis plan, the handling of missing endpoint assessments, the censoring rules, interim analyses, multiplicity control, or the original sample-size justification. These are consequential distinctions: an endpoint definition says what is measured, but does not supply every operational or inferential decision used to analyze it. Interpretation should therefore remain tied to the reported results and the methods explicitly named in the registry record.

11. Confidence Intervals and Precision

The primary estimate is a hazard ratio of 0.43, with a 95% confidence interval from 0.27 to 0.70. The interval gives a sense of the estimate’s precision: it allows a range of relative hazard values rather than identifying a single exact effect. Its endpoints both lie below 1, while the width shows that the effect magnitude is not known with pinpoint precision. The registry does not report a primary p-value.

The median PFS estimates also have confidence intervals. Arm A’s median is 8.4 months, with an interval from 5.75 to 11.83; Arm B’s median is 4.4 months, with an interval from 3.52 to 5.55. These intervals describe uncertainty around each arm’s median separately. They are not a confidence interval for the difference between medians, and overlapping or non-overlapping intervals should not be used as a replacement for the reported hazard-ratio analysis.

Precision is more limited for some secondary outcomes. The intracranial endpoint among participants with brain metastases at baseline has wide percentage intervals in both arms. The arm estimates are based on 11 and 12 participants, respectively. In the subgroup without brain metastases at baseline, the estimates are based on 38 and 37 participants, and the intervals are narrower than those in the smaller subgroup but still leave uncertainty about the true proportions.

Confidence intervals quantify sampling uncertainty under the analysis method; they do not capture every source of uncertainty. They do not resolve questions about missing data, measurement, generalizability, or multiplicity. A narrow interval can indicate greater statistical precision without establishing clinical importance, while a wide interval can be compatible with multiple effect sizes. The registry’s reported intervals should be read alongside the endpoint definitions and the limits of the documented analysis methods.

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 sample-size figures are planning illustrations based on the observed effect, not a description of the trial’s original design justification. The assumptions listed for the illustrations are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is expressed as a stated fraction of the observed effect. The registry’s original planning assumptions are not reported in the registry record.

For time-to-event outcomes, the figures are total events, not participants per arm. For the primary PFS comparison, planning for 100% of the observed effect corresponds to 45 total events for 80% power and 60 total events for 90% power. Planning for 75% of the observed effect corresponds to 79 and 105 total events, respectively. Planning for 50% of the observed effect corresponds to 177 and 237 total events, respectively. The extracranial time-to-event planning illustration has the same event counts.

For the binary secondary outcome, the figures are participants per arm. For the baseline brain-metastasis subgroup, planning for 100% of the observed effect corresponds to 59 participants per arm for 80% power and 78 for 90% power; planning for 75% corresponds to 111 and 148; planning for 50% corresponds to 261 and 349. The corresponding figures for the subgroup without baseline brain metastases are 48 and 64 for 100% of the observed effect, 93 and 124 for 75%, and 222 and 297 for 50%.

The 80% and 90% columns describe the probability of detecting the specified effect under the listed planning assumptions if that effect were the true effect. Smaller effects require more participants or events to detect with the same power. These figures are planning illustrations, not observed power.

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 = 0.430, so ln(HR) = -0.8440.
2. d = 4 × (1.96 + 0.842)² / [ln(0.430)]² = 44.1, rounded up to 45 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 <- 0.43                              # 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(4765059)
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 = 45 (80% power), 60 (90% power)
 75% of observed effect: events = 79 (80% power), 105 (90% power)
 50% of observed effect: events = 177 (80% power), 237 (90% power)
Simulated power with 45 events: 0.743 (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 66 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 98 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 <- 4.4          # control-arm median, months (converted from the posted unit)
hr          <- 0.43          # 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: 42         p (1-sided)   0.0001
 Events: 20        ~HR at bound   0.1874
  Month: 15    P(Cross) if HR=1   0.0001
            P(Cross) if HR=0.43   0.0338
  IA 2: 67%                   Z   2.5114
      N: 66         p (1-sided)   0.0060
 Events: 40        ~HR at bound   0.4487
  Month: 24    P(Cross) if HR=1   0.0060
            P(Cross) if HR=0.43   0.5603
      Final                   Z   1.9930
      N: 66         p (1-sided)   0.0231
 Events: 59        ~HR at bound   0.5950
  Month: 60    P(Cross) if HR=1   0.0250
            P(Cross) if HR=0.43   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 is a log-rank hazard ratio of 0.43, with a 95% confidence interval from 0.27 to 0.70. That is the primary inferential result to use when describing the registry’s PFS comparison. No p-value is reported for it. The posted median estimates provide a complementary summary of each arm’s event-time distribution, but they are not a replacement for the sponsor’s hazard-ratio analysis.

The registry record does not flag a discrepancy between the sponsor’s reported primary analysis and the posted summary. The binary secondary comparisons include unadjusted re-calculations from posted summary statistics. Those calculations are distinct from the sponsor’s reported time-to-event analyses and should not be described as the sponsor’s test of the primary endpoint.

For secondary time-to-event outcomes, the registry reports sponsor hazard ratios where available. The comparison for extracranial progression-free survival is reported as 0.43, with a 95% confidence interval from 0.26 to 0.71. For intracranial progression-free survival among participants without brain metastases at baseline, a log-rank hazard ratio of 0.56 is reported with a 95% confidence interval from 0.27 to 1.13. The record does not provide p-values for these reported hazard ratios. Keeping sponsor-reported analyses separate from unadjusted recalculations preserves the distinction between the trial’s reported analysis and descriptive calculations made from the posted summaries.

16. Multiplicity, Missing Data and Analysis Populations

Multiplicity concerns arise when a study assesses several endpoints, subgroups, or comparisons. In general, the more statistical tests that are conducted, the greater the chance that at least one result will appear notable by chance if every test is interpreted using the same unadjusted threshold. A prespecified testing hierarchy or adjustment procedure can address this issue, but neither is reported in the registry record.

The registry lists a primary PFS endpoint and several secondary outcomes, including intracranial outcomes divided by baseline brain-metastasis status, extracranial progression-free survival, and overall survival. The existence of these outcomes does not reveal whether they were formally tested in a particular order, whether any analysis was designated exploratory, or whether multiplicity was controlled. Those details are not reported in the registry record. The unadjusted p-values accompanying recalculated binary comparisons therefore should not be interpreted as multiplicity-adjusted confirmatory findings.

The primary full analysis set is defined as all randomized participants. This is useful information about the stated analysis population, but it does not answer how missing outcome assessments, withdrawals, incomplete follow-up, or censoring were handled. The registry record does not report those methods. It also does not report whether any sensitivity analyses examined alternative missing-data assumptions.

Similarly, the endpoint definition identifies progression or death as events and notes a data cut-off, but does not specify the detailed operational rules for censoring or the management of assessments that could not be evaluated. These decisions can affect time-to-event estimates, so they should not be inferred from the reported medians or hazard ratios.

The reported log-rank method and hazard ratios identify part of the primary analysis approach. The full statistical analysis plan, including any covariate adjustment, stratification, interim analyses, or other safeguards, is not reported in the registry record. The absence of these details does not establish that a procedure was or was not used; it means the record does not document it. Conclusions should therefore be confined to the outcome summaries and analysis details that are explicitly reported.

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 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: platelet count decreased (6.2%), cellulitis (4.2%), pancytopenia (4.2%), pulmonary embolism (4.2%), appendicitis (2.1%).

The registry’s adverse-event summary lists serious adverse events for 18 participants at risk in Arm A and 15 at risk in Arm B, with 48 at risk in each arm. It lists other adverse events affecting 46 participants in Arm A and 47 in Arm B, also with 48 at risk in each arm. The record separately lists deaths as 26 and 28 in the two arms. These are descriptive counts in the adverse-event record; they do not establish that the events or deaths were caused by a study treatment.

Among the serious adverse-event terms listed, platelet count decreased is reported at 6.2%; cellulitis, pancytopenia, and pulmonary embolism are each reported at 4.2%; appendicitis and bronchopulmonary aspergillosis are each reported at 2.1%. These terms describe events recorded in the registry. They do not provide a complete account of severity, duration, treatment relationship, or clinical course.

Safety comparisons require caution because the registry summary does not explain attribution rules, exposure-adjusted rates, follow-up time by arm, or how repeated events were counted. The listed at-risk numbers are the denominators recorded for these safety summaries and should not be assumed to match every efficacy analysis population. The counts describe what the registry reports, not an individualized assessment of risk.

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

The posted primary result describes a time-to-event comparison in a specific clinical setting: EGFRm, locally advanced or metastatic non-small cell lung cancer that progressed extracranially after first-line osimertinib. The comparison is between osimertinib plus platinum-pemetrexed chemotherapy and placebo plus the same chemotherapy. The primary hazard ratio is 0.43, with a 95% confidence interval from 0.27 to 0.70, and the reported median PFS values are 8.4 months and 4.4 months in the respective arms.

These statistics describe group-level outcomes under the trial’s reported analysis. They do not predict an individual participant’s outcome, establish the same effect for every patient, or provide a direct probability of benefit. A hazard ratio is a relative time-to-event measure; it is not an absolute risk difference, and it should not be read as a direct percentage change in survival time. The medians do not indicate how long any particular person will remain progression-free.

The primary endpoint includes progression or death and is not the same as overall survival. Overall survival is listed as a secondary outcome, but a result for it is not reported in the posted findings summarized in the registry record. The PFS result therefore should not be presented as an overall-survival finding. Nor does the endpoint alone capture symptoms, quality of life, treatment burden, or all aspects of safety.

The observed comparison is also specific to the studied treatment context and eligibility description. The registry record does not establish applicability to people outside that population, to other treatment sequences, or to regimens that differ from those listed. Secondary outcomes add context but have their own uncertainty, and the multiplicity plan is not reported. The sound interpretation is limited to the posted estimates, intervals, and documented analysis details, without converting them into medical advice or a broad claim about treatment effectiveness.

19. Why This Trial Matters Statistically

Statistically, COMPEL is informative because its primary endpoint is time-to-event and its central comparison is reported as a hazard ratio with a confidence interval. That lets readers distinguish the estimated direction of the relative event hazard from the uncertainty around its magnitude. The median PFS estimates offer an additional, more readily interpretable summary of when events accumulated in each arm, while addressing a different aspect of the event-time distribution.

The randomized, placebo-controlled design provides a direct comparison of two treatment strategies in the defined post-osimertinib setting. Both arms include platinum-pemetrexed chemotherapy, which makes the comparison specifically about adding osimertinib versus placebo to that chemotherapy regimen. It is not a comparison of chemotherapy with no chemotherapy. The registry lists double blinding and quadruple masking, but further implementation details are not reported.

The record also illustrates why precision and analysis populations matter. The primary arm estimates are reported for 49 participants in each arm, while secondary subgroup summaries have smaller analysis populations and correspondingly broad uncertainty in some estimates. Different endpoint types—hazard ratios, medians, and percentages at a stated time—must not be treated as interchangeable.

Finally, interpretation depends on what the registry documents as well as what it does not. The sponsor-reported primary hazard ratio and interval are available, but a primary p-value, the detailed statistical plan, multiplicity procedures, and missing-data methods are not reported in the registry record. Recognizing those boundaries helps keep the statistical reading precise: the result can be described clearly without overstating certainty, scope, or what the analysis can answer.

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 posted record. The primary result includes a hazard ratio and confidence interval but no reported p-value. The detailed statistical analysis plan is not reported, so the registry record does not establish whether the primary analysis used stratification, covariate adjustment, or additional prespecified procedures. The record names the log-rank method but does not document the full analytic implementation.

The endpoint timeframe refers to follow-up up to 3 years 1 month and a data cut-off, but the record does not give the data cut-off date, participant-level follow-up, or detailed censoring rules. Missing outcome assessment methods and sensitivity analyses are not reported. These gaps matter for assessing how incomplete follow-up could affect time-to-event estimates.

Multiplicity control for the several secondary outcomes and subgroup analyses is not reported. The binary secondary subgroup summaries have limited analysis populations, particularly among participants with brain metastases at baseline, and their confidence intervals are wide. Their unadjusted comparisons should therefore be interpreted cautiously and do not independently establish a subgroup effect.

The registry lists overall survival as a secondary outcome but does not post an estimate for it in the reported results. The adverse-event record provides counts and selected serious event terms, but does not report event attribution, exposure-adjusted rates, or enough detail to characterize each event’s course. Finally, the maintenance regimen descriptions are truncated in the record. Together, these limits constrain interpretation to the endpoints, estimates, and treatment details actually documented.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: A Study to Evaluate Chemotherapy Plus Osimertinib Against Chemotherapy Plus Placebo in Patients With Non-small Cell Lung Cancer (NSCLC) — NCT04765059.
• Registry record last updated 2026-07-07; status last verified 2026-07.
• Results first posted 2026-01-13.
• 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

The registry describes COMPEL, a Phase III, randomized, double-blind, placebo-controlled, parallel-group study of osimertinib versus placebo added to platinum-pemetrexed chemotherapy after extracranial progression following first-line osimertinib in EGFRm locally advanced or metastatic non-small cell lung cancer. It lists enrollment of 98 and active, not recruiting status.

The primary endpoint is PFS. The sponsor-reported hazard ratio is 0.43, with a 95% confidence interval from 0.27 to 0.70; the median estimates are 8.4 months in Arm A and 4.4 months in Arm B. The primary p-value and detailed statistical plan are not reported in the registry record.

Secondary findings include intracranial and extracranial progression outcomes. Overall survival is listed as an outcome, but an estimate is not posted in the record. The secondary comparisons, safety counts, and confidence intervals require interpretation within their stated populations and documented limitations.