A Study of Brigatinib Compared to Alectinib in Adults With Non-Small-Cell Lung Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study of Brigatinib Compared to Alectinib in Adults With Non-Small-Cell Lung Cancer (NCT03596866), a randomized phase 3 study. The primary outcome is Progression-free Survival (PFS) as Assessed by Blinded Independent Review Committee (BIRC) Per RECIST v1.1.

2. Clinical Question

3. Trial Design

This was a completed, Phase 3, randomized, open-label, parallel-group study sponsored by Takeda. The registry record lists 248 participants enrolled. Participants were assigned by chance to brigatinib or alectinib; the record does not give the allocation ratio as a separate protocol detail. Because masking was none, participants and study personnel knew which treatment was assigned. The primary outcome was nevertheless assessed by a blinded independent review committee, a distinction between treatment masking and outcome review.

The study concerned advanced ALK-positive non-small-cell lung cancer after progression on crizotinib. Brigatinib was administered as a 90 mg oral dose once daily for 7 days followed by 180 mg once daily. Alectinib was administered orally at 600 mg twice daily. The registry describes treatment continuation in relation to disease progression and tolerability, and also records a maximum duration for the alectinib regimen. A detailed description of all treatment-discontinuation procedures is not reported in the registry record.

The primary endpoint was committee-assessed progression-free survival under RECIST v1.1, with follow-up listed as up to 33.8 months. The registry reports 125 participants in the brigatinib analysis and 123 in the alectinib analysis. It also lists secondary outcomes including overall survival, investigator-assessed progression-free survival, response outcomes, and intracranial outcomes. The start date, primary completion date, and completed status are recorded, but the rationale for the enrollment target, the original design assumptions, and details of any interim analyses 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 with anaplastic lymphoma kinase positive (ALK+) non-small-cell lung cancer (NSCLC) who had progressed on crizotinib were administered either brigatinib or alectinib in this study.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: FAS included all participants randomized to each regimen regardless of whether they were ALK+ by an FDA approved test (Vysis ALK Break Apart FISH Probe Kit, Ventana ALK (D5F3) CDx Assay, Foundation Medicine's FoundationOne CDx) or a local test other than FISH and immunohistochemistry, or whether they received study drug or adhered to the assigned dose.

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: Full Analysis Set(FAS) included all participants randomized to each regimen regardless of whether they were ALK+ by an Food and Drug Administration(FDA) approved test (Vysis ALK Break Apart fluorescence in situ hybridization(FISH) Probe Kit, Ventana ALK (D5F3) Companion Diagnostics (CDx) Assay, Foundation Medicine's FoundationOne CDx) or a local test other than FISH and immunohistochemistry, or whether they received study drug or adhered to the assigned dose.

6. Primary Endpoint

The registry lists the primary outcome as Progression-free Survival (PFS) as Assessed by Blinded Independent Review Committee (BIRC) Per RECIST v1.1, measured over Up to 33.8 months.

Progression-free survival measures the interval from randomization until the first objectively documented disease progression or death from any cause, whichever occurs first. In this trial, progression was assessed by a blinded independent review committee using RECIST v1.1. Participants without documented progression or death were censored at their last valid tumor-response assessment, according to the endpoint description.

As a time-to-event endpoint, progression-free survival incorporates both whether an event occurred and when it occurred. The median is the time by which half of the analyzed group is estimated to have experienced progression or death, under the survival analysis. It is not an average duration for every participant, and it does not describe an individual participant’s likely course. A hazard ratio summarizes the relative event rate over follow-up under the analysis model; it is not a ratio of the reported medians.

Using a blinded committee for the primary assessment can reduce the influence of knowledge of assigned treatment on the classification and timing of progression. The open-label design still matters for other aspects of trial conduct and assessment. The registry states that censoring occurred at the last valid tumor assessment for participants without a recorded event. How censoring was handled beyond that description, and whether sensitivity analyses addressed alternative assumptions, is not reported in the registry record.

7. Results Posted in the Registry

• Brigatinib vs Alectinib: hazard ratio 0.97 (95% CI 0.66–1.42), reported p = =0.8672 (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 (OS)

• 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:
  • Brigatinib vs Alectinib: hazard ratio 1.59 (95% CI 0.96–2.65), reported p = =0.0713 (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 Brigatinib vs Alectinib and the control. The interval gives the range of effects that remain compatible with the data.

PFS as Assessed by Investigator Per RECIST v1.1

• 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:
  • Brigatinib vs Alectinib: hazard ratio 1.23 (95% CI 0.86–1.76), reported p = =0.2501 (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 Brigatinib vs Alectinib and the control. The interval gives the range of effects that remain compatible with the data.

Confirmed Intracranial Objective Response Rate (iORR) as Assessed by BIRC Per Modified RECIST v1.1

• 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:
  • Alectinib: 67.7% (21/31) versus 73.3% (22/30) on control; risk difference -5.6 percentage points (95% CI -28.4 to 17.2), risk ratio 0.92 (0.67–1.28); unadjusted two-proportion z-test p = 0.632.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Alectinib and the control. The interval gives the range of effects that remain compatible with the data.

The secondary outcomes extend the primary question to survival, investigator assessment, response, and intracranial disease. The posted numerical results do not form a uniformly precise pattern of differences between groups: the hazard-ratio confidence intervals for overall survival and investigator-assessed progression-free survival include 1, while the intracranial response estimates are imprecise. This broad pattern does not overturn the primary result; it adds outcomes measured in different ways and, in some cases, among smaller analysis groups.

Secondary outcomes are useful for exploring whether the primary finding is accompanied by signals in related clinical measures. They do not all measure the same event. Overall survival concerns death, investigator-assessed progression-free survival depends on investigator evaluation, and intracranial response focuses on response within the brain. Different definitions, assessors, follow-up, and analysis populations can lead to different estimates without implying that one result invalidates another.

The registry lists additional outcomes, including response rate, duration and timing of response, intracranial duration of response, and cumulative incidence of intracranial progression. Numerical results for several of these outcomes are not reported in the registry record summarized. Their presence in the outcome list does not permit conclusions about their observed values.

Multiplicity matters because examining many endpoints and analyses increases the chance that some results may appear notable by chance. Whether the trial adjusted for multiple comparisons, used a testing hierarchy, or designated particular secondary outcomes as confirmatory is not reported in the registry record. Accordingly, secondary p-values and estimates should be read as supportive or exploratory evidence unless a prespecified multiplicity strategy is documented. The secondary findings are context for the primary comparison, not independent proof that one regimen is superior or equivalent.

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

• Overall Survival (OS)
• PFS as Assessed by Investigator Per RECIST v1.1
• Objective Response Rate (ORR) as Assessed by BIRC and Investigator Per RECIST v1.1
• Duration of Response (DOR) as Assessed by BIRC and Investigator Per RECIST v1.1
• Time to Response as Assessed by Investigator and BIRC Per RECIST v1.1
• Confirmed Intracranial Objective Response Rate (iORR) as Assessed by BIRC Per Modified RECIST v1.1
• Intracranial Duration of Response (iDOR) as Assessed by the BIRC Per Modified RECIST v1.1
• Cumulative Incidence of Intracranial Disease Progression (iPD) as Assessed by BIRC Per Modified RECIST v1.1

9. Statistical Interpretation

10. Statistical Methods Explained

The main analysis is a time-to-event comparison. Such analyses are useful when participants have different lengths of follow-up or when some have not experienced the event by the end of observation. Rather than treating every participant as if follow-up ended at the same time, survival methods incorporate event timing and censoring. The endpoint description states that participants without documented progression or death were censored at their last valid tumor-response assessment.

What the median represents

A median time-to-event estimate marks the time at which the estimated event-free proportion reaches one half. It is a feature of the estimated survival curve, not the arithmetic average of observed times. The registry gives median progression-free survival of 19.25 months for brigatinib and 19.19 months for alectinib. The confidence-interval limits are reported for the lower end of each estimate; an upper limit is not reported in the registry record. A median can be estimated even when not every participant has an observed event, because the analysis uses the available follow-up and censoring information.

What the hazard ratio represents

The reported hazard ratio compares the event hazard between treatment groups over follow-up, according to the specified model. A hazard ratio of 1 represents no relative difference in hazards under that comparison. Values on either side of 1 indicate different estimated directions, but the estimate must be interpreted with its confidence interval and the analysis assumptions. A hazard ratio is not a comparison of median times, nor does it directly state the probability that one participant will experience progression before another.

The sponsor reports a hazard ratio of 0.968 for brigatinib versus alectinib, with a 95% confidence interval from 0.658 to 1.424. Because the interval includes 1, the data remain compatible with no relative hazard difference as well as with effects on either side of that value. The reported p-value is 0.867. A p-value is the probability, under the null hypothesis and the test’s assumptions, of observing a result at least as incompatible with that hypothesis as the one recorded. It is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect.

Log-rank comparison and assessment

The primary analysis used a 2-sided stratified log-rank test. In general, the log-rank test compares event-time patterns across groups while accounting for censoring. A stratified version performs the comparison within strata and combines information across them. The registry does not report the stratification factors or enough detail to reconstruct the test. The record describes the primary progression assessment as performed by a blinded independent review committee under RECIST v1.1. Blinded review can limit assessment bias, while the overall study itself was open-label.

Investigator-assessed progression-free survival is a separate secondary outcome. Its event definition also refers to RECIST v1.1, but the assessor differs. Results from the investigator and committee assessments can differ because assessment procedures, timing, or judgments may not be identical. The registry’s secondary analysis reports a log-rank hazard ratio of 1.232, a 95% confidence interval from 0.862 to 1.761, and p =0.250.

Binary response outcomes

Intracranial objective response is a binary outcome: participants are classified according to whether a confirmed response occurred. The registry reports percentages with 95% confidence intervals, based on 30 brigatinib participants and 31 alectinib participants for this endpoint. The sponsor analysis used the Cochran-Mantel-Haenszel method and reported an odds ratio of 1.31, with a 95% confidence interval from 0.44 to 3.84 and p =0.625. An odds ratio compares odds rather than probabilities. It should not be read as a percentage-point difference in response rates.

An unadjusted re-calculation for this binary endpoint is a simple comparison based on the posted summary counts, without the sponsor’s adjustment method. Such calculations can be useful as a check on the direction and scale of the observed data, but they do not replace the sponsor’s analysis. The two approaches may differ because they use different methods and assumptions. For this trial, the discrepancy flag is false, and the sponsor’s reported analysis remains the principal result to interpret.

Assumptions and limits of the methods

Time-to-event methods rely on assumptions about the event process and censoring. In general, censoring is most straightforward to interpret when the reason follow-up ends is not related to the participant’s unobserved future event time, conditional on the analysis information. The registry record describes the censoring rule but does not report diagnostic checks or sensitivity analyses addressing this assumption. General statistical interpretation also depends on whether the hazard ratio provides a useful summary across follow-up; the registry does not report an assessment of proportional hazards.

The registry lists a 2-sided stratified log-rank analysis for the primary comparison, but the detailed statistical analysis plan, model diagnostics, covariate specifications, and any alternative analyses are not reported in the registry record. These omissions limit the ability to independently evaluate the full analysis strategy. They do not alter the numbers reported, but they matter when judging the robustness of the result beyond the principal estimate.

11. Confidence Intervals and Precision

Precision is conveyed by the confidence intervals, which show a range of values reasonably compatible with the data and the analysis assumptions. The primary hazard ratio is 0.968, and its 95% confidence interval runs from 0.658 to 1.424. The interval includes 1, so it is compatible with no relative hazard difference and with estimates in either direction. Its breadth indicates that the study result does not pin down the relative effect narrowly.

The median progression-free survival estimates are 19.25 months for brigatinib and 19.19 months for alectinib. The registry reports lower confidence limits of 15.67 months and 12.88 months, respectively, but no upper limits for these intervals. The absence of an upper limit in the record means the precision of the complete median intervals cannot be described from the posted information. Median confidence intervals and the hazard-ratio interval answer related but distinct questions.

For overall survival, the reported hazard ratio is 1.592, with a 95% confidence interval from 0.956 to 2.652. The interval is broad and includes 1. The median is not reported for either group, although a lower confidence limit of 38.47 months is recorded for brigatinib. These posted details do not permit a precise summary of median overall survival for both groups.

For investigator-assessed progression-free survival, the hazard ratio is 1.232, with a 95% confidence interval from 0.862 to 1.761. The interval also includes 1. For intracranial response, the group estimates have confidence intervals of 54.1% to 87.7% for brigatinib and 48.6% to 83.3% for alectinib. These intervals are wide, consistent with the relatively small endpoint groups recorded. Confidence intervals describe uncertainty in estimates; they are not ranges containing a fixed proportion of individual outcomes.

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 are illustrations calculated from the observed effect, not the trial’s original design calculations. For time-to-event outcomes, entries refer to total events required across the comparison, not participants per arm. For the binary response outcome, entries refer to participants per arm. The illustrations assume a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation. Each row considers an effect equal to the stated fraction of the observed effect.

The 80% power column describes the event count or per-arm participant count that would, under those planning assumptions and for the chosen effect size, give an 80% chance of rejecting the null hypothesis if that effect were truly present. The 90% power column represents a higher target probability of detecting the same assumed effect, and therefore requires more information. These are planning quantities, not observed power; they do not tell whether the completed study had a particular probability of success after its results were known.

The rows vary the effect to be detected. An effect set to 100% of the observed effect uses the full observed estimate as the planning target; the 75% and 50% rows use smaller targets. Smaller effects are harder to distinguish from random variation, so they require more events or participants to achieve the same target power. For time-to-event comparisons, the Schoenfeld method expresses the requirement in total events, making event accumulation and follow-up important. For the response comparison, the requirement is expressed as participants per arm.

The registry’s original sample-size justification and design assumptions are not reported in the registry record. The planning illustrations should therefore not be read as evidence about the actual enrollment target, the study’s prespecified power, or whether the trial met a design threshold. They are hypothetical translations of the observed effect under the stated assumptions.

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 = 0.968, so ln(HR) = -0.0325.
2. d = 4 × (1.96 + 0.842)² / [ln(0.968)]² = 29,681.2, rounded up to 29,682 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 <- 0.968                              # 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(3596866)
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 = 29,682 (80% power), 39,735 (90% power)
 75% of observed effect: events = 52,767 (80% power), 70,640 (90% power)
 50% of observed effect: events = 118,725 (80% power), 158,939 (90% power)
Simulated power with 29,682 events: 0.801 (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 58,410 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 248 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 <- 19.187          # control-arm median, months (converted from the posted unit)
hr          <- 0.968          # 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: 49412         p (1-sided)   0.0001
 Events: 13401        ~HR at bound   0.9379
     Month: 20    P(Cross) if HR=1   0.0001
               P(Cross) if HR=0.97   0.0338
     IA 2: 67%                   Z   2.5114
      N: 58410         p (1-sided)   0.0060
 Events: 26802        ~HR at bound   0.9698
     Month: 33    P(Cross) if HR=1   0.0060
               P(Cross) if HR=0.97   0.5603
         Final                   Z   1.9930
      N: 58410         p (1-sided)   0.0231
 Events: 40203        ~HR at bound   0.9803
     Month: 60    P(Cross) if HR=1   0.0250
               P(Cross) if HR=0.97   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 primary sponsor-reported analysis is a 2-sided stratified log-rank test, with hazard ratio 0.968, 95% confidence interval 0.658 to 1.424, and p =0.867. The time-to-event interpretation uses that reported hazard ratio rather than an independently reconstructed one. No discrepancy is flagged for the primary result, and the medians and their available confidence limits are reported separately from the hazard-ratio comparison.

For intracranial response, the sponsor used a Cochran-Mantel-Haenszel analysis and reported an odds ratio of 1.31, 95% confidence interval 0.44 to 3.84, and p =0.625. A simple unadjusted re-calculation from posted summary statistics is not the same analysis: it does not reproduce the sponsor’s adjustment method. It may provide a descriptive cross-check, but it should not displace the sponsor’s reported comparison. The record does not report the strata used for the sponsor’s adjusted analysis.

There is no flagged discrepancy between sponsor results and the posted analyses. The sponsor’s test and effect estimate are the primary inferential reference for each reported comparison. Any unadjusted calculation is supplementary and should be interpreted with its simpler assumptions in mind, rather than treated as a correction to the sponsor’s result.

16. Multiplicity, Missing Data and Analysis Populations

The registry identifies one primary endpoint and several secondary outcomes, but it does not report whether a formal multiplicity adjustment was used. It also does not describe a hierarchy for testing secondary endpoints, identify which secondary comparisons were confirmatory, or state how repeated analyses were controlled. When multiple outcomes are tested, the chance of at least one apparently unusual result can increase. Without the analysis plan, the evidential weight of individual secondary p-values is difficult to determine. The reported estimates remain informative descriptions, but they should not automatically be treated as definitive confirmatory findings.

The primary population is described as the full analysis set, including all randomized participants assigned to each regimen regardless of specified ALK test status. The available population description is truncated, so further eligibility or analysis-population details are not reported in the registry record. The endpoint description says participants without documented progression or death were censored at their last valid tumor-response assessment. The record does not detail missing assessment patterns, reasons for missingness, sensitivity analyses, or how missing data outside the stated censoring rule were handled.

For survival analyses, censoring permits participants with incomplete follow-up to contribute information up to their last assessment. General statistical interpretation depends on censoring being sufficiently independent of the future event process, given the information used in analysis. Whether that assumption was assessed is not reported. The registry also does not report missing-data imputation procedures for the response outcomes.

Details of interim analyses, stopping boundaries, the full statistical analysis plan, and the original sample-size justification are not reported in the registry record. These details would help clarify which analyses were prespecified and how the totality of outcomes was intended to be interpreted. Their absence does not change the posted estimates, but it limits assessment of selective reporting, analysis flexibility, and the intended control of false-positive findings.

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: dyspnoea (4.0%), interstitial lung disease (4.0%), pneumonia (2.5%), non-small cell lung cancer (2.4%), respiratory failure (2.4%).

The registry records serious adverse events in 38 of 125 participants at risk in the brigatinib group and 24 of 122 at risk in the alectinib group. These are counts of participants with at least one serious adverse event in the recorded analysis, not a measure of event severity beyond the registry’s serious classification and not a direct comparison adjusted for follow-up. The registry also records 37 deaths in the brigatinib group and 25 in the alectinib group; the cause categories and relationship to treatment are not posted in this summary.

Other adverse events were recorded in 122 of 125 participants at risk with brigatinib and 118 of 122 at risk with alectinib. The listed serious-event terms include dyspnoea and interstitial lung disease at 4.0% each, pneumonia at 2.5%, non-small cell lung cancer and respiratory failure at 2.4% each, and anaemia at 1.6%. These percentages are the registry’s reported figures; the record excerpt does not provide all term-specific counts or explain whether categories overlap.

Serious adverse-event summaries are descriptive and do not, on their own, establish that treatment caused an event or permit a complete benefit-risk assessment. Exposure duration, event timing, severity gradations, and detailed attribution are not reported in the registry information. These figures should be interpreted alongside the trial’s population and follow-up, not as a standalone comparison of safety.

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

The primary result describes the progression-free survival comparison observed in this randomized study. The median estimates are 19.25 months for brigatinib and 19.19 months for alectinib, and the reported hazard ratio is 0.968 with a 95% confidence interval from 0.658 to 1.424. The interval includes 1, and the reported p-value is 0.867. Taken together, these results do not provide a statistically clear difference in the primary endpoint, while the interval remains compatible with a range of relative effects.

This finding does not show that the treatments are identical. A non-significant result is not an equivalence or non-inferiority conclusion unless an appropriate design, margin, and analysis support that claim. The registry lists a superiority analysis and does not report a non-inferiority margin. Nor does the result establish that either treatment is beneficial or ineffective relative to no treatment, because the trial compared two active regimens rather than including an untreated control.

The results also do not determine which option is appropriate for any individual. Trial averages and group comparisons cannot account for every clinical circumstance, and the registry record is not a basis for medical advice. The open-label design, endpoint assessment procedures, censoring, and available follow-up all shape what can be inferred. Blinded independent review was used for the primary progression assessment, but the study overall was not masked.

Secondary outcomes offer additional context but have distinct definitions and limitations. The registry’s numerical secondary findings do not produce a conclusive pattern that can be substituted for the primary analysis, and multiple comparisons may affect their interpretation. Some listed secondary outcomes lack posted numerical results in the registry record. The most defensible reading is therefore bounded: the reported primary comparison did not clearly distinguish the groups, and the uncertainty interval does not rule out potentially different relative effects.

19. Why This Trial Matters Statistically

This trial is statistically informative because it compares two active regimens in a randomized design and reports a time-to-event primary endpoint assessed by a blinded independent review committee. Randomization supports a structured between-group comparison, while blinded outcome review can reduce the influence of treatment knowledge on progression classification. The open-label design remains relevant to trial conduct and outcomes assessed without blinded review.

The primary result also illustrates why point estimates, confidence intervals, and p-values should be read together. A hazard ratio close to 1 and a high p-value alone do not prove that effects are the same. The confidence interval shows that the estimate remains uncertain across values on both sides of 1. The result is best described as not demonstrating a clear difference in the primary analysis, rather than as proof of equivalence.

Different outcomes answer different questions. Committee-assessed progression-free survival, overall survival, investigator-assessed progression-free survival, and intracranial response are not interchangeable. Their assessment methods, event definitions, follow-up, and analysis groups differ. The record’s secondary outcomes help frame the broader evidence, while the lack of a reported multiplicity strategy limits how strongly isolated secondary findings can be interpreted.

Finally, the distinction between a sponsor’s analysis and an unadjusted re-calculation matters. The sponsor’s reported methods define the principal inferential results; simpler calculations from posted summaries can serve as checks but may not reproduce adjustments or other design features. Planning illustrations based on observed effects are likewise not evidence of the trial’s original design assumptions or observed power. Together, these points show why transparent reporting of estimands, analysis populations, missing data, multiplicity, and design assumptions is essential to interpreting a clinical trial.

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 registry record leaves several analysis details unspecified. The primary analysis is identified as a 2-sided stratified log-rank test, but the stratification factors and the full statistical analysis plan are not reported. Proportional-hazards diagnostics and sensitivity analyses for the time-to-event assumptions are also not reported. These omissions limit independent assessment of the model and the robustness of the hazard-ratio summary.

The primary endpoint description specifies censoring at the last valid tumor-response assessment for participants without documented progression or death. Missing assessment patterns, reasons for loss to follow-up, and sensitivity analyses for informative censoring are not reported. The full analysis-set description is truncated, which limits understanding of the complete population definition. The registry also does not report the original sample-size justification, interim analyses, or a multiplicity strategy.

The available median confidence intervals are incomplete: upper limits are not reported for the primary medians. Overall survival medians are not reported for either group, although a lower confidence limit is recorded for brigatinib. These omissions constrain interpretation of the full survival distributions. Several listed secondary outcomes have no numerical result in the record summarized, so their contribution cannot be evaluated from the posted information.

The study was open-label, even though the primary progression assessment was blinded and independent. Secondary assessments made by investigators may be more vulnerable to knowledge of assigned treatment. Some secondary outcomes were based on smaller subsets than the primary analysis, making their estimates less precise. Finally, serious adverse-event summaries are not a full safety analysis: detailed timing, attribution, exposure-adjusted rates, and complete event-level information are not reported in the registry record.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: A Study of Brigatinib Compared to Alectinib in Adults With Non-Small-Cell Lung Cancer — NCT03596866.
• Registry record last updated 2025-09-12; status last verified 2025-09.
• Results first posted 2025-02-19.
• 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

NCT03596866 is a completed Phase 3 randomized, open-label, parallel-group comparison of brigatinib and alectinib in adults with advanced ALK-positive non-small-cell lung cancer after progression on crizotinib. The registry records 248 participants enrolled. The primary endpoint was blinded independent review committee-assessed progression-free survival under RECIST v1.1, with follow-up listed as up to 33.8 months.

Median progression-free survival was 19.25 months for brigatinib and 19.19 months for alectinib. The reported hazard ratio was 0.968, with a 95% confidence interval from 0.658 to 1.424 and p =0.867. The analysis does not show a statistically clear difference, while the confidence interval leaves uncertainty about the relative effect. Interpretation is limited by unreported details concerning the analysis plan, multiplicity, missing data, and original design assumptions.