Study of Nivolumab in Combination With Ipilimumab Compared to the Standard of Care (Extreme Regimen) as First Line Treatment in Patients With Recurrent or Metastatic Squamous Cell Carcinoma of the Head and Neck: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Study of Nivolumab in Combination With Ipilimumab Compared to the Standard of Care (Extreme Regimen) as First Line Treatment in Patients With Recurrent or Metastatic Squamous Cell Carcinoma of the Head and Neck (NCT02741570), a randomized phase 3 study. The primary outcome is Overall Survival (OS) in Participants With Programmed Death-Ligand 1 (PD-L1) With a Combined Positive Score (CPS) ≥20.

2. Clinical Question

The clinical question is a comparison of treatment strategies in a defined disease setting, not a comparison of individual drugs in isolation. The intervention combines nivolumab and ipilimumab, while the comparator combines cetuximab with platinum chemotherapy and fluorouracil. The registry describes the study as first-line treatment for recurrent or metastatic disease.

The primary endpoint is restricted to randomized participants with PD-L1 CPS at least 20. That restriction matters when interpreting the result: the primary estimate describes this subgroup, rather than all randomized participants. A separate secondary overall-survival outcome concerns participants with CPS at least 1, a broader subgroup. Results for that secondary outcome should not be treated as though they were the primary analysis or as though they applied to every participant.

The practical statistical question is whether the observed survival patterns are compatible with a difference between the assigned strategies and how precisely that difference is estimated. The registry reports medians, confidence intervals, a hazard ratio, and a log-rank p-value. These statistics describe different features of the survival comparison and do not, individually or collectively, establish what would happen for every patient.

3. Trial Design

This was a randomized, open-label, parallel, two-arm phase III study. The registry reports enrollment of 947 participants and lists the study as completed. The study began on 2016-10-05, and primary completion was recorded as 2021-05-10. Masking was none, so the trial was open label.

One group received nivolumab plus ipilimumab. The other received the EXTREME regimen, described as cetuximab with cisplatin or carboplatin and fluorouracil. The registry gives both arm descriptions as treatment on specified doses and days, without providing those schedules in the record summarized by the posted results. It does not report the randomization ratio or the original sample-size justification.

The primary outcome is overall survival among randomized participants with PD-L1 CPS at least 20. The reported groups contain 185 participants in the nivolumab-plus-ipilimumab arm and 178 in the EXTREME arm. Those subgroup counts are not the same as the study-wide enrollment count, because the primary outcome is restricted to a biomarker-defined population.

The registry defines overall survival as time from randomization to death. For participants without documented death, the outcome is censored on the last date they were known to be alive. Participants randomized without follow-up are censored at randomization. The primary outcome timeframe is up to approximately 55 months. The record also lists secondary outcomes of overall survival in participants with CPS at least 1, progression-free survival, objective response rate, and duration of objective response. The registry record does not report the original design assumptions or a complete account of how all these outcomes were analyzed.

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: 947 participants randomized to receive study treatment. 909 participants received study treatment.

Pre-Treatment Period

Reasons for not completing this period:

Treatment Period

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All randomized PD-L1 CPS \>= 20 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.

6. Primary Endpoint

The registry lists the primary outcome as Overall Survival (OS) in Participants With Programmed Death-Ligand 1 (PD-L1) With a Combined Positive Score (CPS) ≥20, measured over From randomization to date of death or date the participant was last known to be alive (Up to approximately 55 months).

Overall survival measures the time from randomization until death. In the primary analysis, it is evaluated among randomized participants with PD-L1 CPS at least 20. Participants without documented death contribute information through the last date they were known to be alive and are then censored; randomized participants with no follow-up are censored at randomization. The registry describes the estimates as based on Kaplan-Meier methods and gives a follow-up timeframe of up to approximately 55 months.

General statistical context: overall survival is a time-to-event endpoint. It uses both whether an event occurred and the time until it occurred, while allowing participants whose event time is not observed to contribute information up to censoring. A median summarizes the estimated survival distribution: it is the time by which the estimated proportion surviving reaches one half. It is not an average survival time, nor does it describe the experience of every participant.

Overall survival is used in clinical trials because death is a clear event, but its interpretation still depends on follow-up, censoring, the population analyzed, and the comparison method. A median alone does not capture the full pattern over time. The hazard ratio provides a relative comparison across the observed follow-up under its model or testing framework, while the confidence interval expresses uncertainty around the estimate. These measures answer related but distinct questions.

7. Results Posted in the Registry

• Nivolumab + Ipilimumab vs EXTREME Regimen: hazard ratio 0.78 (95% CI 0.59–1.03), reported p = 0.047 (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) in Randomized Participants With Programmed Death-Ligand 1 (PD-L1) With a Combined Positive Score (CPS) ≥ 1

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor; it was not recalculated.
• Result:
  • Nivolumab + Ipilimumab vs EXTREME Regimen: hazard ratio 0.80 (95% CI 0.68–0.95) (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 Nivolumab + Ipilimumab vs EXTREME Regimen than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary outcomes broaden the assessment beyond the primary overall-survival population. They include overall survival among randomized participants with PD-L1 CPS at least 1, progression-free survival, objective response rate, and duration of objective response. The registry includes a reported overall-survival comparison for the CPS-at-least-1 population; numerical results for the other listed secondary outcomes are not reported in the registry record summarized for this analysis.

The broader overall-survival comparison has an estimated hazard ratio below 1, and its reported confidence interval lies below 1. That direction is consistent with the direction of the primary hazard-ratio estimate. This directional resemblance does not make the two analyses interchangeable: the populations differ, the primary and secondary status differs, and the reported methods and confidence levels are not identical. It also does not supply results for progression-free survival, response, or duration of response.

General statistical context: secondary endpoints can help describe whether findings extend across different outcomes, but each additional test creates more opportunities for a nominally notable result to occur by chance. If many outcomes are examined without a planned multiplicity strategy, the usual interpretation of an isolated p-value can be overly confident. The registry record does not report a multiplicity procedure or a hierarchical testing plan. For that reason, secondary findings should generally carry less inferential weight than a prespecified primary outcome with a clearly documented analysis plan.

The secondary results are best treated as supporting context, not as a substitute for the primary comparison. Their interpretation depends on population definitions, endpoint-specific methods, missing outcome results, and whether analyses were adjusted for multiple testing. Those details are not all reported in the registry record. A consistent direction across the reported survival estimates can be noted, while the evidence remains bounded by the incomplete statistical-plan information and the absence of reported numerical results for some listed outcomes.

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

• Overall Survival (OS) in Randomized Participants With Programmed Death-Ligand 1 (PD-L1) With a Combined Positive Score (CPS) ≥ 1
• Progression Free Survival (PFS)
• Objective Response Rate (ORR)
• Duration of Objective Response (DOR)

9. Statistical Interpretation

10. Statistical Methods Explained

General statistical context: a randomized parallel-group trial assigns participants to one of two treatment strategies and compares outcomes between the assigned groups. Randomization is intended to support a fair comparison by reducing systematic differences in treatment assignment. The registry identifies this study as randomized and parallel, but does not report the allocation ratio or additional details about the randomization procedure. Open-label means masking was none; the registry does not specify how that feature was addressed in outcome assessment or analysis.

What does a median survival estimate describe?

The primary endpoint is time from randomization to death, estimated using Kaplan-Meier methods. The Kaplan-Meier approach estimates the survival distribution over follow-up while accounting for censoring. Censoring occurs when a participant’s death is not documented during the observed period; the registry says survival is censored at the last known date alive. A censored participant is not treated as though death occurred at that point. Instead, their observed time contributes information up to censoring, subject to the method’s assumptions.

The median is a point on the estimated survival curve, rather than a comparison of every participant’s survival time. It gives a compact description of the distribution, but it does not show how the curves behave at all points in follow-up. The confidence interval around a median conveys uncertainty in that estimated point. General statistical context: confidence intervals are generated by a procedure that, under its assumptions, has a stated coverage rate over repeated studies. They are not a guarantee that the true value lies inside any one calculated interval.

What does a hazard ratio compare?

A hazard ratio compares the estimated event rates between groups over time, according to the analysis approach. The primary reported comparison is a hazard ratio of 0.78 for nivolumab plus ipilimumab versus the EXTREME regimen. A value of 1 is the no-relative-difference reference; a value below 1 points toward a lower estimated hazard in the first-named group. A hazard ratio is not a ratio of median survival times, does not directly state how many months an individual gains, and does not mean that every participant experiences the same relative change.

Interpreting a single hazard ratio also calls for care about how the relative hazards behave during follow-up. General statistical context: a single summary may be less descriptive if the relative event rates change over time. The registry record identifies the primary comparison method as log rank and reports the hazard ratio, but does not give further details about checking assumptions or the time pattern of hazards. The registry’s secondary CPS-at-least-1 analysis is labeled Cox proportional hazard; its method is not specified for the primary entry beyond the log-rank designation.

How should the test and interval be read together?

The log-rank test compares survival experience between groups over follow-up. The reported primary p-value is 0.047. In general, a p-value describes how unusual data at least as incompatible with a specified null model would be if that model and the test assumptions held. It is not the probability that the null hypothesis is true, nor a measure of the size or clinical importance of a difference.

The registry reports a 97.51% confidence interval for the primary hazard ratio, from 0.59 to 1.03. Since this interval contains 1, it includes the no-relative-difference value. The p-value and interval are both part of the posted analysis, but the registry record does not explain their relationship or why the interval has that confidence level. They should not be silently converted to a common confidence level or treated as interchangeable summaries. The reported analysis is labeled as superiority; that designation does not change the numerical interval.

For the secondary overall-survival analysis in participants with CPS at least 1, the registry reports a Cox proportional hazard estimate and a 95% confidence interval. It gives no p-value in the record. A missing p-value is not evidence of no difference; it means that a p-value is not reported in the registry record. Likewise, numerical results for the other listed secondary outcomes are not reported in the registry record summarized for this analysis.

Why can over-dispersion matter in other outcome types?

General statistical context: over-dispersion occurs when observed variation is greater than a model assumes. It can make standard errors too small and intervals too narrow if a model does not allow for that extra variation. This is a general issue in count or rate analyses, not a reported finding about the primary time-to-event analysis in this trial. The registry entries summarized for the primary endpoint do not identify an over-dispersion assessment.

More broadly, a statistical method should match the endpoint and the data structure. Overall survival is a time-to-event outcome, so its censoring and timing matter. The registry reports Kaplan-Meier estimates, a log-rank comparison, and a hazard ratio. It does not report all modeling diagnostics, analysis-plan details, or sensitivity analyses. Those absences limit how fully an outside reader can assess the analytical choices, but do not change the values actually reported.

11. Confidence Intervals and Precision

The primary median estimates are 17.58 months for nivolumab plus ipilimumab and 14.59 months for the EXTREME regimen. Their respective confidence intervals, 13.77 to 21.98 months and 12.32 to 15.97 months, express uncertainty around each estimated median. The intervals overlap, but comparing median intervals alone is not the same as testing the time-to-event distributions; the registry’s reported group comparison is the hazard ratio and log-rank analysis.

For the hazard ratio, the reported estimate is 0.78 and the interval is 0.59 to 1.03 at a 97.51% confidence level. This interval extends on both sides of 1, so the values compatible with the interval include the no-relative-difference reference as well as values below it. The interval therefore signals uncertainty about the magnitude and direction at that stated confidence level. Its upper bound is close to 1, but the estimate remains imprecise enough that the interval includes 1.

General statistical context: a confidence interval is more informative than a point estimate alone because it shows a range of values compatible with the data under the interval procedure and its assumptions. Width reflects precision: narrower intervals indicate less sampling uncertainty than wider ones, all else equal. The interval does not account automatically for every source of uncertainty, such as outcome measurement, model limitations, or multiplicity.

The registry reports a 97.51% interval for the primary hazard ratio, whereas the secondary CPS-at-least-1 hazard-ratio interval is reported at 95%. These are distinct reported confidence levels and should not be compared as if they were identical procedures. The registry record does not explain the primary interval level or provide additional precision diagnostics. The p-value and confidence interval should both be retained in interpretation rather than selecting only the one that appears more decisive.

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 concern the total number of events for these time-to-event comparisons, not participants per arm. They illustrate how many events would be required under stated planning assumptions for 80% or 90% power at different fractions of the observed effect. The figures do not describe how many events were actually observed in the primary analysis.

The planning assumptions 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. For example, planning for 75% of the observed effect means using a smaller target effect than the full observed effect. The registry record does not report the original trial’s sample-size justification or its original design assumptions, so these planning figures should not be taken as a reconstruction of the trial’s protocol.

The 80% power column gives the total event count associated with an 80% chance of detecting the specified effect under those assumptions, if that effect is real and the assumptions hold. The 90% power column describes the corresponding event count for a 90% chance. Power is a property of a planned design under assumed conditions, not a measure of whether the observed result is true.

Across the planning rows, a smaller target effect requires more events than a larger target effect to attain the same stated power. This is because smaller group differences are harder to distinguish from random variation. These counts are planning illustrations, not observed power: observed power is not established by substituting the observed estimate into a design calculation after the results are known. The figures also do not establish that the trial was designed around the observed effect.

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 = 0.780, so ln(HR) = -0.2485.
2. d = 4 × (1.96 + 0.842)² / [ln(0.780)]² = 508.6, rounded up to 509 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 <- 0.78                              # 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(2741570)
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 = 509 (80% power), 681 (90% power)
 75% of observed effect: events = 905 (80% power), 1,211 (90% power)
 50% of observed effect: events = 2,035 (80% power), 2,724 (90% power)
Simulated power with 509 events: 0.800 (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 936 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 947 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 <- 14.59          # control-arm median, months (converted from the posted unit)
hr          <- 0.78          # 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: 762         p (1-sided)   0.0001
 Events: 230        ~HR at bound   0.6125
   Month: 20    P(Cross) if HR=1   0.0001
             P(Cross) if HR=0.78   0.0338
   IA 2: 67%                   Z   2.5114
      N: 936         p (1-sided)   0.0060
 Events: 459        ~HR at bound   0.7909
   Month: 31    P(Cross) if HR=1   0.0060
             P(Cross) if HR=0.78   0.5603
       Final                   Z   1.9930
      N: 936         p (1-sided)   0.0231
 Events: 688        ~HR at bound   0.8590
   Month: 60    P(Cross) if HR=1   0.0250
             P(Cross) if HR=0.78   0.9000



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

15. Analyses Reported by the Sponsor

The registry’s own primary analysis reports a hazard ratio of 0.78, with an interval from 0.59 to 1.03 at 97.51% confidence and a log-rank p-value of 0.047. The posted result is the sponsor-reported analysis for the primary comparison. The corresponding secondary CPS-at-least-1 analysis is reported as a Cox proportional hazard estimate of 0.80, with a 95% confidence interval from 0.68 to 0.95; no p-value is reported for that analysis in the registry record.

The statistical summaries do not constitute a separate recalculation that replaces the sponsor’s analysis. The registry record indicates no discrepancy for the comparison, so there is no conflicting recalculated result to resolve. The sponsor’s own analysis takes precedence whenever a discrepancy is present; in this record, the reported sponsor estimate and interval remain the basis for interpretation.

For the primary result, the estimate, interval, and p-value need to be read together. The estimated hazard ratio is below 1 and the p-value is 0.047, while the reported 97.51% confidence interval includes 1. The registry labels the analysis as superiority but does not explain this combination or give additional context about the interval level. The appropriate account preserves each reported statistic and avoids presenting only the estimate or only the p-value as a complete summary.

16. Multiplicity, Missing Data and Analysis Populations

The registry lists one primary overall-survival outcome for randomized participants with PD-L1 CPS at least 20 and several secondary outcomes, including overall survival in participants with CPS at least 1, progression-free survival, objective response rate, and duration of objective response. It also reports eight outcome measures and seven analyses. These counts indicate that multiple outcomes and analyses are present, but they do not establish how the statistical testing was organized.

The registry record does not report whether the primary and secondary outcomes were handled through a multiplicity adjustment, a hierarchical testing strategy, or another prespecified procedure. It does not report interim analyses or the original sample-size justification. Without those details, the family of hypotheses being tested and the intended error-rate control cannot be reconstructed from the posted results. A nominal p-value should therefore be read as the result of its reported test, not as a complete statement about evidence across every endpoint.

For missing data, the primary outcome description specifies censoring rules for participants without documented death and for randomized participants with no follow-up. It does not provide a comprehensive missing-data plan for every endpoint, explain whether sensitivity analyses were performed, or report how missing measurements for response-related outcomes were handled. The registry identifies the primary population as all randomized participants with CPS at least 20, but does not give a full account of analysis populations for every secondary endpoint.

These omissions do not show that the trial lacked a plan or that analyses were necessarily inappropriate. They mean the relevant details are not reported in the registry record. General statistical context: prespecified endpoint hierarchy, missing-data methods, interim testing, and analysis-population definitions help readers distinguish confirmatory tests from exploratory analyses and assess how robust conclusions may be. The posted information supports description of the reported estimates, but not a complete audit of these design and analysis choices.

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: malignant neoplasm progression (22.7%), pneumonia (7.0%), febrile neutropenia (5.4%), tumour haemorrhage (4.5%), anaemia (4.1%).

The registry reports serious adverse events among 468 participants at risk in the nivolumab-plus-ipilimumab group, with 313 affected, and among 441 at risk in the EXTREME group, with 290 affected. For other adverse events, it reports 420 affected among 468 at risk in the nivolumab-plus-ipilimumab group and 429 among 441 at risk in the EXTREME group. Death counts in these adverse-event groups are 388 and 406, respectively.

The most frequently listed serious adverse-event terms include malignant neoplasm progression at 22.7%, pneumonia at 7.0%, febrile neutropenia at 5.4%, tumour haemorrhage at 4.5%, anaemia at 4.1%, and dyspnoea at 3.0%. These are registry-reported percentages for the listed terms. The record does not provide further detail about event timing, severity grading, attribution, or whether participants could contribute more than one event.

These counts describe the posted safety data and should not be interpreted as a causal comparison on their own. The groups at risk differ in size, and the adverse-event summaries are separate from the primary survival subgroup counts. The numbers also do not establish that an event was caused by treatment. Interpreting safety requires the event definitions, observation period, exposure, and analysis rules; details beyond those stated are not reported in the registry record.

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

The primary result describes overall survival in randomized participants with PD-L1 CPS at least 20. The median estimates and hazard ratio give complementary summaries of the observed time-to-event data. The hazard-ratio estimate is below 1, while its reported 97.51% confidence interval includes 1 and the log-rank p-value is 0.047. The statistics therefore express a direction of the point estimate alongside uncertainty that should not be hidden by a single threshold-based label.

The result does not establish that every participant would have the same outcome, nor does it translate the hazard ratio into a fixed number of months for an individual. A group-level median is not a prediction for a particular person. The estimates also apply to the specified primary subgroup and should not automatically be generalized to the full study population or to other populations not described in the registry.

The result does not resolve the contribution of each component of either treatment strategy, since the comparison is between combinations. It does not establish outcomes for every secondary endpoint: numerical results for progression-free survival, objective response rate, and duration of objective response are not reported in the registry record. Nor does the reported survival result alone describe all aspects of safety or quality of life.

General statistical context: a p-value is not the probability that a treatment effect is real, and a confidence interval is not a guarantee of the range of individual outcomes. Interpretation also depends on the analysis plan, follow-up, censoring, and multiplicity. Several of these details are not reported in the registry record. The appropriate conclusion is therefore a measured description of the reported comparison and its uncertainty, rather than a definitive statement that one strategy works or does not work.

19. Why This Trial Matters Statistically

This trial is statistically informative because it compares two multi-component treatment strategies in a randomized, parallel-group phase III study and reports an overall-survival analysis in a biomarker-defined population. The primary endpoint is time-to-event, so the analysis must account for both event timing and censoring rather than reducing every participant’s outcome to a simple event count.

The primary report also illustrates why a point estimate, confidence interval, and p-value should be considered together. The hazard-ratio estimate is below 1, the log-rank p-value is 0.047, and the reported 97.51% interval extends across 1. A reader who focuses only on the point estimate or only on a thresholded p-value misses part of the statistical picture. The registry’s superiority label does not remove the need to inspect the size and precision of the estimate.

The biomarker-defined primary population makes the scope of inference especially important. The primary result pertains to participants with CPS at least 20, while a separate secondary survival outcome concerns participants with CPS at least 1. These populations cannot be collapsed into a single result. The trial also lists outcomes beyond survival, but not all numerical results are reported in the registry record summarized for this analysis.

Finally, the record demonstrates the value of transparent reporting. Clear information on multiplicity, interim analyses, the sample-size rationale, missing-data methods, and analysis populations would help readers assess the planned inferential framework. Where these details are absent, careful interpretation means describing what is reported and keeping uncertainty visible rather than filling gaps with assumptions.

20. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The primary analysis is restricted to randomized participants with PD-L1 CPS at least 20. Its estimates should not be treated as estimates for all 947 enrolled participants. Although the registry also lists an overall-survival outcome for the CPS-at-least-1 population, that secondary analysis has a distinct population and should be interpreted separately.

The primary hazard-ratio interval is reported at 97.51% confidence and includes 1, while the p-value is 0.047. The registry record does not explain this pairing, the interval level, or whether it reflects a particular testing plan. It also does not report multiplicity control, interim analyses, the original sample-size justification, or the randomization ratio. These missing details constrain interpretation of the nominal p-value and the planned role of the primary result.

The primary endpoint description gives censoring rules, but the registry record does not report a complete missing-data strategy or sensitivity analyses. It also does not provide detailed diagnostics for the time-to-event analysis, including information about the behavior of relative hazards over follow-up. The open-label design is recorded, but the record does not explain how knowledge of treatment assignment was addressed for outcome assessment.

Several listed secondary outcomes do not have numerical results in the registry record summarized for this analysis. The safety information gives affected counts and selected serious adverse-event terms, but the record as summarized does not provide further detail on exposure, event attribution, or event-level timing. These limitations do not negate the posted estimates; they define what can and cannot be assessed from the registry information available for this statistical reading.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: Study of Nivolumab in Combination With Ipilimumab Compared to the Standard of Care (Extreme Regimen) as First Line Treatment in Patients With Recurrent or Metastatic Squamous Cell Carcinoma of the Head and Neck — NCT02741570.
• Registry record last updated 2023-09-21; status last verified 2023-08.
• Results first posted 2022-05-03.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

24. Record Summary

NCT02741570 is a completed, randomized, open-label, parallel, two-arm phase III study comparing nivolumab plus ipilimumab with the EXTREME regimen as first-line treatment for recurrent or metastatic squamous cell carcinoma of the head and neck. The registry reports enrollment of 947 participants.

The primary outcome is overall survival among randomized participants with PD-L1 CPS at least 20. Median overall survival was 17.58 months for nivolumab plus ipilimumab and 14.59 months for the EXTREME regimen. The reported hazard ratio was 0.78, with a 97.51% confidence interval from 0.59 to 1.03 and a log-rank p-value of 0.047.

The registry also lists overall survival in the CPS-at-least-1 population, progression-free survival, objective response rate, and duration of objective response as secondary outcomes. The record does not report several design and analysis details needed for a complete assessment, including the original sample-size justification and multiplicity plan.