A Study of the Anti-PD1 Antibody PDR001, in Combination With Dabrafenib and Trametinib in Advanced Melanoma: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study of the Anti-PD1 Antibody PDR001, in Combination With Dabrafenib and Trametinib in Advanced Melanoma (NCT02967692), a randomized phase 3 study. The primary outcome is Randomized (Part 3): Progression-Free Survival (PFS) as Per Investigator's Assessment by RECIST 1.1.

2. Clinical Question

The study was designed to evaluate the safety and efficacy of adding an anti-PD-1 antibody to dabrafenib and trametinib. The primary endpoint was progression-free survival measured from first dose until documented radiological progression or death from any cause. The registry reports median estimates for each randomized arm and a sponsor-reported hazard ratio from a log-rank analysis.

The randomized comparison is the most direct answer to the stated primary question. Part 1 was a safety run-in, and Part 2 was a biomarker cohort; neither is the randomized Part 3 comparison. Their results should not be treated as randomized estimates of the intervention-versus-placebo effect.

The registry records enrollment of 568 and reports 267 participants in the PDR001 arm and 265 in the placebo arm for the primary endpoint. These arm-specific primary analysis counts are the relevant denominators for reading the posted PFS estimates; the registry does not explain the difference between those counts and overall enrollment.

3. Trial Design

This was a randomized, quadruple-masked, parallel-group Phase III study sponsored by Novartis Pharmaceuticals. The official title describes a placebo-controlled comparison of PDR001, dabrafenib and trametinib against placebo, dabrafenib and trametinib in previously untreated participants with unresectable or metastatic BRAF V600 mutant melanoma. The registry lists melanoma as the condition.

The study had distinct parts. Part 1 was a safety run-in in which participants received different dose levels to identify a recommended Phase 3 regimen. Part 2 was a biomarker cohort receiving the combination. In Part 3, participants were randomized to the PDR001 combination or the matching-placebo combination. The Part 3 dose descriptions specify PDR001 400 mg Q4W, dabrafenib 150 mg BID and trametinib 2 mg QD in the active arm, with placebo substituted for PDR001 in the comparator arm.

The primary endpoint population is described as all Part 3 participants assigned treatment by randomization, regardless of whether treatment was administered. The registry reports 267 participants for the active-arm PFS estimate and 265 for the placebo-arm estimate. The safety listing gives a different placebo-arm at-risk count for Part 3; the registry record does not explain this difference, so safety counts should not be assumed to use precisely the same analysis population as the primary endpoint.

The registry lists the study status as terminated, with a start date of 2017-02-17 and a primary completion date of 2020-08-11. The reason for termination and its implications for the analysis are not reported in the registry record. The original design assumptions, including the intended sample-size justification, 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: The screening phase began once written informed consent was provided and ended after 28 days or when subject received the first dose (Part 1 and 2) or was randomized (Part 3), whichever came first.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Dose-Determining Set (DDS) including all subjects in Part 1 who received at least one dose of study treatment who either 1) met the minimum exposure criterion and had sufficient safety evaluations, or 2) experienced a DLT during the first 8 weeks (56 days) of spartalizumab in combination with dabrafenib and trametinib dosing.

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 Randomized (Part 3): Progression-Free Survival (PFS) as Per Investigator's Assessment by RECIST 1.1, measured over Up to disease progression or death due to any cause, whichever occurs first, assessed up to 2.8 years (Part 3).

Progression-free survival measures the time from the first dose to the first documented radiological progression, assessed by the investigator according to RECIST 1.1, or death from any cause. It is a time-to-event endpoint: each participant contributes follow-up until an event is recorded or follow-up is censored. The registry says that a participant without an event at data cutoff was censored at the date of the last adequate tumor assessment.

PFS is used in trials of advanced cancer because it captures the period during which a participant remains alive without documented progression. It differs from overall survival, which measures time to death, and from response rate, which records whether specified tumor shrinkage criteria are met. This endpoint combines progression and death, so the estimate concerns the time to either event rather than either outcome considered alone.

The registry estimated the PFS distribution with the Kaplan-Meier method and reports the median for each randomized arm. A median is the estimated time by which half of the relevant event-time distribution has experienced an event; it does not describe every participant's course or the average time free of progression. The median estimates were 16.2 months for the PDR001 combination and 12.0 months for the placebo combination.

Investigator assessment is the stated basis for progression. The registry record does not report whether an independent central review was also used, nor does it give details about the timing or frequency of tumor assessments. Those details matter because the observation schedule and assessment process can affect when progression is recorded.

7. Results Posted in the Registry

• Part 3- Arm 1: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD vs Part 3- Arm 2: Placebo + Dabrafenib 150 mg BID + Trametinib 2 mg QD: hazard ratio 0.82 (95% CI 0.66–1.03), reported p = 0.042 (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 Response Rate (ORR) as Per Investigator's Assessment by RECIST 1.1

• 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:
  • Part 2- Biomarker Cohort: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD: 70.4% (19/27) versus 64.2% (170/265) on control; risk difference 6.2 percentage points (95% CI -11.9 to 24.4), risk ratio 1.10 (0.85–1.42); unadjusted two-proportion z-test p = 0.519.
  • Part 3- Arm 1: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD: 68.5% (183/267) versus 64.2% (170/265) on control; risk difference 4.4 percentage points (95% CI -3.6 to 12.4), risk ratio 1.07 (0.95–1.21); unadjusted two-proportion z-test p = 0.284.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Part 2- Biomarker Cohort: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Part 3- Arm 1: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD and the control. The interval gives the range of effects that remain compatible with the data.

Disease Control Rate (DCR) as Per Investigator's Assessment by RECIST 1.1

• 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:
  • Part 2- Biomarker Cohort: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD: 92.6% (25/27) versus 86.4% (229/265) on control; risk difference 6.2 percentage points (95% CI -4.5 to 16.9), risk ratio 1.07 (0.95–1.20); unadjusted two-proportion z-test p = 0.363.
  • Part 3- Arm 1: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD: 84.3% (225/267) versus 86.4% (229/265) on control; risk difference -2.1 percentage points (95% CI -8.2 to 3.9), risk ratio 0.98 (0.91–1.05); unadjusted two-proportion z-test p = 0.484.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Part 2- Biomarker Cohort: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Part 3- Arm 1: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD and the control. The interval gives the range of effects that remain compatible with the data.

Randomized (Part 3): Time to 10 Point Definitive Deterioration in European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ) C30- Global Health Status

• 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:
  • Part 3- Arm 1: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD vs Part 3- Arm 2: Placebo + Dabrafenib 150 mg BID + Trametinib 2 mg QD: hazard ratio 1.18 (95% CI 0.86–1.62), reported p = 0.297 (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 Part 3- Arm 1: PDR001 400 mg Q4W + Dabrafenib 150 mg BID + Trametinib 2 mg QD vs Part 3- Arm 2: Placebo + Dabrafenib 150 mg BID + Trametinib 2 mg QD and the control. The interval gives the range of effects that remain compatible with the data.

The secondary outcomes cover response, disease control, survival, duration of response and quality-of-life measures. The posted Part 3 response and disease-control estimates do not form a uniform pattern in the same direction: the response estimate is numerically higher in the PDR001 arm, while disease control is numerically higher in the placebo arm. The quality-of-life deterioration comparison also has a hazard-ratio point estimate above 1, with an interval that includes 1. These are different outcomes and should be interpreted according to their definitions, populations and uncertainty rather than treated as interchangeable confirmations of PFS.

The registry lists overall survival, duration of response, response rate, disease control, quality-of-life changes and time to deterioration among the secondary outcomes. The posted results include 20 secondary outcomes, but complete numeric results and inferential analyses for all those outcomes are not reported in the registry record. The outcome list alone does not establish whether results were favorable, unfavorable or inconclusive.

Secondary findings can complement the primary endpoint by addressing other aspects of disease course and participant-reported health. They do not overturn or independently validate the primary result simply because they point in a similar direction. Each outcome may have a different analysis population, missing-data pattern, time horizon and measurement process.

Unadjusted secondary comparisons carry less weight for confirmatory claims when many outcomes or comparisons are assessed. The more analyses performed, the greater the chance that some results appear noteworthy by chance, even if no true effect exists for those outcomes. Adjustment procedures, hierarchical testing or another prespecified strategy could address this issue, but the multiplicity plan is not reported in the registry record. The recalculated binary comparisons are expressly unadjusted and should be read as exploratory context, not as definitive proof of an effect.

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

• Overall Survival (OS)
• Overall Response Rate (ORR) as Per Investigator's Assessment by RECIST 1.1
• Duration of Response (DOR) as Per Investigator's Assessment by RECIST 1.1
• Disease Control Rate (DCR) as Per Investigator's Assessment by RECIST 1.1
• Randomized (Part 3): Change From Baseline in European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ) C30- Global Health Status Scores
• Randomized (Part 3): Change From Baseline in European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ) C30- Physical Functioning Scale Scores
• Randomized (Part 3): Change From Baseline in European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ) C30- Pain Symptom Scale Scores
• Randomized (Part 3): Time to 10 Point Definitive Deterioration in European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ) C30- Global Health Status

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is a time-to-event measure. Such outcomes account for both whether an event occurs and how long participants are followed before an event or censoring. The registry defines PFS from first dose until investigator-assessed radiological progression under RECIST 1.1 or death from any cause. Participants without an event by the data cutoff were censored at their last adequate tumor assessment. Censoring is not the same as an event; it indicates that the endpoint event was not observed during the participant's available follow-up.

What does the Kaplan-Meier median represent?

The Kaplan-Meier method estimates the event-free distribution over follow-up while accommodating participants whose event time is not observed because their follow-up ends first. The median is the point at which the estimated event-free distribution reaches its midpoint. It is a compact summary, but it does not describe the full shape of the time-to-event curves. Two arms can have different medians even if their curves vary in ways that are not captured by that single summary.

The registry reports median PFS of 16.2 months in the PDR001 combination arm and 12.0 months in the placebo combination arm, each with a 95% confidence interval. It does not post the event counts or Kaplan-Meier estimates at specific follow-up times in the registry record for this explanation. The extent of censoring and the number still at risk at different times are therefore not reported in the registry record.

How should the hazard ratio be read?

The sponsor used a log-rank analysis and reported a hazard ratio of 0.820. A hazard ratio compares the instantaneous event rates between groups over follow-up, subject to the model and interpretation used for the analysis. A value below 1 points toward a lower event rate in the first-listed arm relative to the comparator; it is not a ratio of the median times and should not be translated into a percentage change in median PFS. It also does not say that every participant had a lower risk at every point in time.

The proportional-hazards interpretation is most straightforward when the relative hazard is reasonably stable over follow-up. Whether that assumption was assessed, and what diagnostics or alternative analyses were used, are not reported in the registry record. The log-rank test is a comparison of event-time distributions, while the hazard ratio gives an effect-size summary. Their results are related but are not identical statistical quantities.

Why do confidence intervals matter?

A confidence interval expresses uncertainty about an estimated quantity under the analysis assumptions. The primary hazard-ratio interval is 0.655 to 1.027. It includes 1, the value corresponding to equal hazards, and is compatible with a range of relative effects on either side of that point. A confidence interval is not a guarantee that the true value lies within its endpoints, nor does it indicate the probability that a particular value is true.

The registry reports p = 0.042 for the log-rank analysis. A p-value describes how unusual data at least as discordant with a specified null hypothesis would be if that hypothesis and the test assumptions held. It is not the probability that the null is true, and it does not measure the clinical importance of an estimate. The apparent difference between the p-value and the interval's inclusion of 1 should be noted; the registry does not explain it. The sponsor-reported analysis is the source for the primary hazard ratio, interval and test result.

What does randomization contribute?

Random allocation is intended to make the treatment groups comparable on average at assignment, including for factors that were not measured. The registry calls the study randomized, quadruple-masked and parallel. Masking can reduce the influence of knowledge about assignment on participant experience, care and outcome assessment. The record does not detail who was included in the quadruple masking or how masking was maintained.

The primary population includes all Part 3 participants assigned treatment by randomization, regardless of whether treatment was administered. This definition preserves the randomized comparison as the registry describes it. It is different from restricting an analysis to participants who received or adhered to treatment, which could undermine the advantages of randomization. The record does not provide further details on analysis exclusions or deviations from this population definition.

Why can time-to-event data be difficult?

Time-to-event analyses rely on appropriate handling of censoring and on the quality and timing of event assessment. If censoring is related to participants' underlying event prospects, the Kaplan-Meier and hazard-ratio interpretations may be affected. The record gives the censoring rule for participants without an event at cutoff but does not report the amount or reasons for censoring, or analyses testing sensitivity to it.

The endpoint is based on investigator assessment under RECIST 1.1. Differences in assessment timing can affect when progression is detected, and any differences in assessment practice could influence recorded times. The registry record does not report independent review procedures or assessment schedules. The hazard ratio and medians therefore need to be understood in the context of the stated investigator-assessed endpoint, not as independent measurements of tumor progression.

11. Confidence Intervals and Precision

The median PFS estimate is 16.2 months in the PDR001 combination arm, with a 95% confidence interval from 12.7 to 23.9 months. In the placebo combination arm, the median is 12.0 months, with a 95% confidence interval from 10.2 to 15.4 months. These intervals show that the estimated medians are uncertain; they are not ranges containing the times for most participants.

The sponsor-reported hazard ratio is 0.820, with a 95% confidence interval from 0.655 to 1.027. The interval is compatible with a relative hazard below 1 and also extends slightly above 1. Its width indicates that the data do not pin down the relative hazard to a single precise value. The interval is particularly important because the point estimate alone gives no account of uncertainty.

Confidence intervals are generally constructed by a procedure that, over repeated comparable studies under its assumptions, would cover the underlying value at the stated rate. They do not assign a probability to the true effect after seeing these data. The interval also does not reflect every possible source of uncertainty, such as limitations in measurement, censoring assumptions, selective reporting or choice of analysis.

The confidence intervals for the arm-specific medians overlap. Comparing those intervals informally is not a substitute for the formal between-arm analysis. The hazard ratio and log-rank result address the randomized comparison using time-to-event information, while each median is a separate summary of its arm's estimated distribution. The registry record does not report confidence bands for the full survival curves or additional precision measures for the primary endpoint.

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 illustrations use the observed effect as a planning reference, then consider planning for 100%, 75% and 50% of that observed effect. For time-to-event outcomes, the figures are total events, not participants per arm. For other outcomes, the figures are participants per arm. The displayed 80% power column describes the illustrative number of events or participants associated with an 80% chance of detecting the stated effect under the planning assumptions. The 90% power column represents a higher target chance of detection and consequently requires a larger illustrative information size.

The stated assumptions are a two-sided significance level of 0.05, equal allocation between the two arms and a normal approximation. The calculation uses the stated fraction of the observed effect as the effect to detect; the time-to-event illustrations use the Schoenfeld approach. These assumptions describe the planning illustration, not the trial's original design. The original trial design assumptions and sample-size justification are not reported in the registry record.

Smaller effects are harder to distinguish from random variation, so the illustrations require more participants or events when planning for a smaller fraction of the observed effect. The figures should be read as conditional scenarios rather than forecasts: they depend on the effect chosen and the assumptions used, and may not reflect other design features or analysis choices. The table and curve are not estimates of the number actually enrolled or the number of events observed. This is not observed power; it is an illustration of planning requirements under 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 = 0.820, so ln(HR) = -0.1985.
2. d = 4 × (1.96 + 0.842)² / [ln(0.820)]² = 797.2, rounded up to 798 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 <- 0.82                              # 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(2967692)
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 = 798 (80% power), 1,068 (90% power)
 75% of observed effect: events = 1,418 (80% power), 1,898 (90% power)
 50% of observed effect: events = 3,189 (80% power), 4,269 (90% power)
Simulated power with 798 events: 0.797 (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 1,362 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 568 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 <- 12          # control-arm median, months (converted from the posted unit)
hr          <- 0.82          # 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: 1052         p (1-sided)   0.0001
  Events: 360        ~HR at bound   0.6761
    Month: 19    P(Cross) if HR=1   0.0001
              P(Cross) if HR=0.82   0.0338
    IA 2: 67%                   Z   2.5114
      N: 1362         p (1-sided)   0.0060
  Events: 719        ~HR at bound   0.8292
    Month: 29    P(Cross) if HR=1   0.0060
              P(Cross) if HR=0.82   0.5603
        Final                   Z   1.9930
      N: 1362         p (1-sided)   0.0231
 Events: 1079        ~HR at bound   0.8857
    Month: 60    P(Cross) if HR=1   0.0250
              P(Cross) if HR=0.82   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 uses a log-rank method and gives a hazard ratio of 0.820, a 95% confidence interval from 0.655 to 1.027, and p = 0.042. These are the primary comparison results to use when describing the registry's inferential analysis. The record labels the reported analysis as a superiority analysis.

No separate recalculated primary hazard ratio or p-value is presented. For time-to-event outcomes, the posted analysis relies on the sponsor's reported hazard ratio rather than a simple re-calculation from arm-level summary statistics. The primary result therefore should not be described as though an independent calculation replaced the sponsor's model or test.

The interval includes 1 even though the reported p-value is 0.042. The registry does not explain the difference between these summaries or provide enough detail to reconcile it. Both should be stated accurately: the reported test result is p = 0.042, and the confidence interval for the hazard ratio ranges from 0.655 to 1.027. The p-value does not erase the uncertainty reflected in the interval, and the interval should not be selectively omitted.

For selected binary secondary outcomes, simple unadjusted re-calculations are available alongside arm estimates. Those calculations are not the sponsor's reported analyses; no sponsor-reported comparisons are listed for those outcomes. Their unadjusted status and the absence of a reported multiplicity approach limit how strongly they can support confirmatory conclusions.

16. Multiplicity, Missing Data and Analysis Populations

The registry record does not report a statistical analysis plan describing how multiplicity was handled across the primary and secondary endpoints, or whether any hierarchy, gatekeeping strategy or adjustment procedure was used. It also does not say whether analyses were prespecified before results were examined. Without those details, the reported p-values should not automatically be interpreted as family-wise confirmatory evidence across all outcomes.

Multiplicity matters because testing several outcomes or comparisons increases the chance of at least one apparently unusual result under a global null. A nominal p-value can be useful as a measure for an individual test, but its interpretation changes when many tests are considered together. The posted binary secondary comparisons are unadjusted re-calculations, and the registry lists no sponsor-reported analysis for them. They therefore warrant more caution than a prespecified, multiplicity-controlled primary analysis.

The registry defines the primary population as all Part 3 participants assigned treatment by randomization, regardless of whether treatment was administered. For PFS, it specifies that participants without an event at data cutoff were censored at their last adequate tumor assessment. The record does not report the number or reasons for censored observations, missing endpoint assessments, missing-data imputation procedures, or sensitivity analyses addressing missingness.

For the secondary endpoints, the record does not consistently provide enough information in the registry record to establish the analysis population, missing-data handling or the effect of incomplete follow-up. The registry's outcome descriptions give timeframes, but they do not replace a full account of censoring, questionnaire completion or analytic decisions.

The sponsor-reported primary interval includes 1 while its reported p-value is 0.042. The registry does not explain this discrepancy in interpretation, nor does it provide sufficient test-statistic details to investigate it. These unreported design and analysis details limit assessment of confirmatory strength and robustness; they should not be filled in by assumption.

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: pyrexia (44.4%), hypophysitis (11.1%), diarrhoea (11.1%), umbilical hernia (11.1%), cholelithiasis (11.1%).

The registry's Part 3 safety listing reports serious adverse events affecting 150 of 267 at-risk participants in the PDR001 combination group and 123 of 264 at-risk participants in the placebo combination group. Other adverse events are reported for 262 of 267 and 250 of 264, respectively. These are the registry's group counts; the figures should be read using their stated at-risk denominators and not combined with the primary endpoint denominators.

Serious adverse events and other adverse events are separate registry categories, and the listed counts do not by themselves describe severity, duration, causal attribution or whether an event led to treatment discontinuation. The registry also lists serious-event terms, including Pyrexia at 44.4%, Hypophysitis at 11.1%, Diarrhoea at 11.1%, Umbilical hernia at 11.1%, Cholelithiasis at 11.1% and Hepatitis at 11.1%. The relevant denominator and grouping context for these percentages are not reported in the registry record.

Separate extended follow-up rows report deaths, with zero at-risk counts in those rows. They should not be added to the initial group rows or interpreted as a complete comparative mortality analysis without further definitions. Safety comparisons also require context on exposure duration, event ascertainment and attribution, which is not reported in the registry record.

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

The posted primary results describe a randomized comparison of investigator-assessed PFS for the PDR001 combination and the placebo combination. The median estimates are 16.2 months and 12.0 months, and the sponsor-reported hazard ratio is 0.820. The point estimates are directionally consistent with longer PFS in the PDR001 arm, while the confidence interval for the hazard ratio, 0.655 to 1.027, includes 1. The reported log-rank p-value is 0.042.

These statistics do not establish that every participant benefited, predict an individual participant's outcome or show that the same result would occur in other settings. A median describes the estimated distribution for a group, not the experience of every participant. A hazard ratio summarizes relative event rates over follow-up under its analysis framework; it is not a direct comparison of median durations and does not describe an absolute change for an individual.

The results also do not by themselves settle the interpretation of all secondary outcomes, establish an overall-survival effect or resolve safety trade-offs. The registry lists those outcomes, but complete analyses and relevant multiplicity details are not reported in the registry record. Nor does the record explain why the study was terminated or whether termination affected the completeness of follow-up.

Statistical uncertainty is central to interpretation. The interval permits a range of relative effects, including values on either side of 1. The p-value summarizes compatibility with a null hypothesis under the test assumptions, not the probability that the treatment effect is real. Together, the reported results describe evidence from this trial; they do not constitute a medical recommendation or a guarantee of effect.

19. Why This Trial Matters Statistically

This trial is statistically informative because it used a randomized, masked, placebo-controlled comparison to evaluate an added treatment component while both groups received dabrafenib and trametinib. Randomization makes the between-group contrast more interpretable than a comparison between unrelated cohorts, and masking can reduce some forms of assessment and care bias. The primary endpoint is a time-to-event outcome, so its analysis incorporates event timing and censoring rather than only counting whether progression occurred.

The record also illustrates why effect estimates, intervals and test results need to be read together. The hazard-ratio point estimate favors the PDR001 arm under the usual direction of comparison, but its confidence interval includes 1, while the log-rank p-value is 0.042. Reporting both gives a fuller account than emphasizing either the point estimate or the p-value alone. The registry does not explain the apparent tension between the interval and test result.

Statistical interpretation is also shaped by what is missing. Multiplicity handling, detailed missing-data procedures, the original sample-size rationale and the reason for termination are not reported in the registry record. These omissions make it harder to assess how strongly the primary result was intended to support a confirmatory claim and how robust the findings are to analysis choices.

Finally, the secondary outcomes underscore that trial evidence is multidimensional. Response, disease control, quality of life, survival and safety are not substitutes for one another. A clear statistical reading preserves their distinct meanings and avoids turning a collection of estimates into a single unqualified conclusion.

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 states that the study was terminated but does not report the reason. It also does not explain whether termination affected planned follow-up, data collection or analysis. That context is important when interpreting a time-to-event endpoint whose precision depends on observed events and follow-up.

The primary endpoint relies on investigator-assessed progression under RECIST 1.1. The registry record does not report the tumor assessment schedule, independent central review, adjudication procedures or checks on assessment consistency. It also does not provide the event count, censoring count or numbers at risk over follow-up in the registry record. These omissions prevent a fuller assessment of the time-to-event data and censoring pattern.

The registry reports a hazard-ratio confidence interval that includes 1 alongside a log-rank p-value of 0.042, but does not explain the difference. It does not report proportional-hazards diagnostics, alternative model results or other sensitivity analyses. The original sample-size assumptions and statistical analysis plan are also not reported in the registry record.

For safety, the Part 3 placebo at-risk denominator differs between the primary endpoint and safety listings. The registry does not explain the difference. The listed serious-event term percentages lack clear denominator context in the summarized record, and extended follow-up death rows have zero at-risk counts. These details limit interpretation of the safety summaries as comparative rates or a complete account of mortality.

Secondary findings are limited by incomplete posted inferential detail and the absence of a reported multiplicity strategy. The binary recalculations are unadjusted, and not all listed secondary outcomes have full numeric results in the registry record for this account. These limitations argue for describing individual estimates and their uncertainty rather than making broad claims from a selective subset.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: A Study of the Anti-PD1 Antibody PDR001, in Combination With Dabrafenib and Trametinib in Advanced Melanoma — NCT02967692.
• Registry record last updated 2025-09-18; status last verified 2025-09.
• Results first posted 2023-07-24.
• 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

NCT02967692 is a randomized, quadruple-masked, parallel-group Phase III study of PDR001, dabrafenib and trametinib versus placebo, dabrafenib and trametinib in previously untreated participants with unresectable or metastatic BRAF V600 mutant melanoma. The registry records enrollment of 568 and lists the study status as terminated.

The primary endpoint was investigator-assessed PFS under RECIST 1.1, with death from any cause also counted as an event. Median PFS was 16.2 months in the PDR001 arm and 12.0 months in the placebo arm. The sponsor-reported hazard ratio was 0.820, with a 95% confidence interval from 0.655 to 1.027 and p = 0.042 from a log-rank analysis.

The interval includes 1, and the registry does not explain its apparent difference from the reported p-value. The reason for termination, original design assumptions, multiplicity plan and detailed missing-data procedures are not reported in the registry record. Secondary and safety findings provide additional context but have their own limitations and should not be treated as substitutes for the primary comparison.