Study of Pembrolizumab With Maintenance Olaparib or Maintenance Pemetrexed in First-line (1L) Metastatic Nonsquamous Non-Small-Cell Lung Cancer (NSCLC) (MK-7339-006, KEYLYNK-006): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Study of Pembrolizumab With Maintenance Olaparib or Maintenance Pemetrexed in First-line (1L) Metastatic Nonsquamous Non-Small-Cell Lung Cancer (NSCLC) (MK-7339-006, KEYLYNK-006) (NCT03976323), a randomized phase 3 study. The primary outcome is Progression-free Survival (PFS).

2. Clinical Question

The registry identifies PFS and overall survival as the study’s two primary hypotheses. The posted primary result concerns PFS, measured from randomization until documented progression or death from any cause, whichever occurred first. The comparison is specifically for the first pembrolizumab course in the Maintenance Phase. That distinction matters: this estimate is not an induction-phase PFS result, and it does not by itself answer the separate overall-survival hypothesis.

The clinical contrast is between two treatment strategies that share pembrolizumab and induction chemotherapy but differ in the maintenance drug. The statistical question is whether the time-to-event evidence supports the stated superiority hypothesis, and how precisely the treatment comparison is estimated. The registry record does not report the overall-survival result.

3. Trial Design

This was a Phase 3, randomized, parallel-group study with no masking reported. The registry lists an enrollment of 1003 and marks the study as completed. The listed start date is 2019-06-28, and the primary completion date is 2024-02-07. The overall study status is COMPLETED.

Both groups began with pembrolizumab, pemetrexed, and platinum chemotherapy in the Induction Phase. The study description names carboplatin or cisplatin as platinum choices. The strategies then differed in the Maintenance Phase: pembrolizumab with olaparib versus pembrolizumab with pemetrexed. The registry describes an induction course of 4 cycles, with a cycle lasting 3 weeks, in the arm descriptions. The primary PFS analysis is for the first pembrolizumab course in maintenance, not the induction period.

The reported PFS population is all randomized participants in the Maintenance Phase. The arm-level PFS estimates are reported for Pembrolizumab + Olaparib (Maintenance Phase), with 337.0 participants, and Pembrolizumab + Pemetrexed (Maintenance Phase), with 335.0 participants. These analysis counts should not be treated as a replacement for the registry’s overall enrollment figure; they describe the analysis population reported for this endpoint.

Randomization supports a comparison between assigned strategies, while the parallel design means the groups are followed as separate treatment arms. With no masking reported, the study is described as open-label. The primary endpoint was nevertheless assessed by blinded independent central review, according to the endpoint description. The allocation ratio, detailed randomization procedures, and the original 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: Per protocol, response or progression during the second pembrolizumab course were not counted towards patient reported outcomes (PROs) or efficacy outcome measures and adverse events during the second pembrolizumab course will not be counted towards safety outcome measures.

Induction Phase

Reasons for not completing this period:

Maintenance Phase

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All randomized participants in the Maintenance Phase. Per protocol, analysis for this outcome measure was not planned or conducted in the Induction Phase.

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: Pembrolizumab + Pemetrexed + Platinum Therapy (Induction Phase): All participants allocated to the pembrolizumab + Pemetrexed + Platinum Therapy arm in the Induction phase but not randomized into the maintenance phase; Pembrolizumab + Olaparib (Maintenance Phase): All participants randomized to the Pembrolizumab + Olaparib arm post Induction Phase; Pembrolizumab + Pemetrexed (Maintenance Phase): All participants randomized to the Pembrolizumab + Pemetrexed arm post Induction Phase.

6. Primary Endpoint

The registry lists the primary outcome as Progression-free Survival (PFS), measured over Up to ~31 months.

Progression-free survival is the time from randomization to the first documented progressive disease or death due to any cause, whichever occurs first. Progression was determined using RECIST 1.1 based on blinded independent central review. The registry defines progression as at least a 20% increase in the sum of diameters of target lesions, together with an absolute increase of at least 5 mm; the appearance of one or more new lesions also qualifies as progression.

PFS is a time-to-event endpoint. In general, it captures how long participants remain alive without documented progression, combining progression and death into a single event definition. It is useful when the clinical question concerns delaying progression or death, and it allows the analysis to account for differing follow-up times and participants whose event has not been observed by the analysis cutoff.

The registry reports median PFS and a 95% confidence interval for each maintenance group, as well as a hazard ratio from a log-rank analysis. The medians describe the estimated time by which half the analyzed group has experienced an event, under the Kaplan-Meier method. They are not averages, and they do not alone describe the whole event-time pattern. The registry gives the primary PFS timeframe as up to ~31 months and specifies that the final protocol-planned PFS analysis is reported for the first pembrolizumab course in maintenance.

7. Results Posted in the Registry

• Pembrolizumab + Olaparib (Maintenance Phase) vs Pembrolizumab + Pemetrexed (Maintenance Phase): hazard ratio 1.12 (95% CI 0.92–1.36), reported p = 0.872 (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.

Time to True Deterioration (TTD) in EORTC QLQ-C30 Global Health Status / Quality of Life (Items 29 & 30) Scale Score

• 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:
  • Pembrolizumab + Olaparib (Maintenance Phase) vs Pembrolizumab + Pemetrexed (Maintenance Phase): hazard ratio 0.97 (95% CI 0.76–1.24), reported p = 0.812 (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 Pembrolizumab + Olaparib (Maintenance Phase) vs Pembrolizumab + Pemetrexed (Maintenance Phase) and the control. The interval gives the range of effects that remain compatible with the data.

TTD in EORTC QLQ-LC13 Cough (Item 1) Scale Score

• 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:
  • Pembrolizumab + Olaparib (Maintenance Phase) vs Pembrolizumab + Pemetrexed (Maintenance Phase): hazard ratio 0.88 (95% CI 0.66–1.17), reported p = 0.378 (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 Pembrolizumab + Olaparib (Maintenance Phase) vs Pembrolizumab + Pemetrexed (Maintenance Phase) and the control. The interval gives the range of effects that remain compatible with the data.

TTD in EORTC QLQ-LC13 Chest Pain (Item 10) Scale Score

• 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:
  • Pembrolizumab + Olaparib (Maintenance Phase) vs Pembrolizumab + Pemetrexed (Maintenance Phase): hazard ratio 1.08 (95% CI 0.78–1.51), reported p = 0.627 (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 Pembrolizumab + Olaparib (Maintenance Phase) vs Pembrolizumab + Pemetrexed (Maintenance Phase) and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists secondary outcomes spanning adverse events, treatment discontinuation due to adverse events, and patient-reported quality-of-life measures. The quality-of-life outcomes include global health status or quality of life and lung-cancer symptoms such as cough and chest pain, assessed through change from baseline or time to true deterioration. The posted time-to-deterioration comparisons have hazard-ratio confidence intervals that include 1, and their reported p-values do not indicate evidence of a group difference under those individual comparisons.

Across these secondary time-to-event findings, the estimates do not form a clear pattern of decisive separation between maintenance strategies. Some group-level median estimates are reported, while median values are not reported for cough and chest-pain deterioration. These secondary outcomes address related but distinct questions from PFS: they concern patient-reported deterioration rather than the primary progression-or-death event definition.

Secondary outcomes can add context about symptoms, quality of life, and tolerability, but they do not replace the primary PFS comparison or establish the separate overall-survival hypothesis. Their interpretation also depends on the analysis plan, the completeness of questionnaire data, and how many outcomes are tested. The registry record does not report a multiplicity adjustment for the secondary analyses or a detailed missing-data approach for patient-reported outcomes.

In general, testing many outcomes increases the chance that at least one result may appear notable by chance if each is interpreted without adjustment. For that reason, unadjusted secondary results generally carry less confirmatory weight than prespecified primary comparisons. The results are best treated as supportive or exploratory unless the registry reports a specific confirmatory testing strategy. The record does not establish such a strategy for these secondary outcomes.

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

• Number of Participants Who Experience One or More Adverse Events (AEs)
• Number of Participants Who Discontinue Study Treatment Due to an AE
• Change From Baseline in European Organization for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire-Core 30 (QLQ-C30) Global Health Status / Quality of Life (Items 29 and 30) Scale Score
• Time to True Deterioration (TTD) in EORTC QLQ-C30 Global Health Status / Quality of Life (Items 29 & 30) Scale Score
• Change From Baseline in EORTC Quality of Life Questionnaire Lung Cancer Module 13 (QLQ-LC13) Cough (Item 1) Scale Score
• TTD in EORTC QLQ-LC13 Cough (Item 1) Scale Score
• Change From Baseline in EORTC QLQ-LC13 Chest Pain (Item 10) Scale Score
• TTD in EORTC QLQ-LC13 Chest Pain (Item 10) Scale Score

9. Statistical Interpretation

10. Statistical Methods Explained

The primary PFS analysis uses the Kaplan-Meier method, a non-parametric approach for estimating the event-time distribution. In general, Kaplan-Meier estimation uses the observed event times while retaining information from participants whose event time is not observed during the available follow-up. Those participants are censored at their last known event-free time under the analysis rules. The registry record does not report the detailed censoring rules or missing-data procedures for this analysis.

The Kaplan-Meier curve represents the estimated probability of remaining alive without documented progression over time. Its median is the time at which the estimated event-free probability reaches one half, if that point can be estimated. The registry reports median PFS for both maintenance groups. A median is one summary of the curve; it does not show every change in the estimated event-free probability or describe the full shape of the curves.

The treatment comparison is reported as a hazard ratio from a log-rank analysis. In general, the hazard is the instantaneous event rate among participants who have not yet had the event. A hazard ratio compares these rates between groups over the analysis period. It is not a direct comparison of the proportion who experienced an event, nor is it a ratio of the reported medians. Interpretation is tied to which group is named first, the event definition, and the analysis population. The registry reports the comparison as pembrolizumab plus olaparib versus pembrolizumab plus pemetrexed.

The log-rank test is a standard method for comparing time-to-event experience between groups. In general, it evaluates whether the observed event-time patterns differ more than would be expected under a null hypothesis of no difference, taking account of event timing and censoring. Its p-value is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect. The sponsor reports p = 0.872 for the PFS comparison.

A 95% confidence interval gives a range of values compatible with the data and the analysis method, under repeated-sampling assumptions. In general, it is a measure of estimation uncertainty rather than a guarantee that the true value lies in the particular interval. The interval around the PFS hazard ratio extends from 0.92 to 1.36. Because it includes 1, the result is compatible with no relative hazard difference as well as with values in either direction. The median confidence intervals likewise reflect uncertainty around each group’s estimated median, rather than a direct interval for the between-group difference.

The registry lists a superiority hypothesis for the PFS comparison. A superiority analysis asks whether the data are inconsistent with the null benchmark of no group difference in the specified direction and framework. A failure to meet a superiority criterion does not demonstrate equivalence or non-inferiority; those conclusions require their own hypotheses and analysis plans. The registry does not report an equivalence or non-inferiority hypothesis for this comparison.

The endpoint population is described as all randomized participants in the Maintenance Phase. The record explicitly states that PFS analysis was not planned or conducted in the Induction Phase. That population definition is important because a reader should not assume the PFS estimate uses every participant in the overall enrollment. The registry does not provide further detail on the exact handling of participants who did not enter maintenance or on the analysis-set rules beyond the stated population.

In general, proportional hazards is an assumption often relevant to interpreting a single hazard ratio as a common relative effect over time. The registry description identifies Kaplan-Meier estimation and a log-rank method but does not report a proportional-hazards diagnostic or an assessment of whether the relative hazard changed over time. The curve and median summaries can help describe event-time patterns, but the record does not report a separate time-varying treatment effect analysis.

Over-dispersion is mainly a concern for count outcomes when observed variability exceeds what a simple count model expects. In general, time-to-event data are analyzed with methods that explicitly account for event timing and censoring rather than with a basic count model. The registry’s stated PFS methods are Kaplan-Meier and log-rank; it does not report a count-model analysis for this endpoint.

The posted result includes the sponsor’s reported hazard ratio, confidence interval, method, and p-value. No discrepancy is flagged between the sponsor analysis and the re-calculated presentation. In general, re-calculations from summary statistics are unadjusted and may not reproduce participant-level modeling or the complete prespecified analysis. For this outcome, the sponsor’s stated analysis is the primary reported comparison, and the registry does not report a distinct adjusted model in the endpoint description.

The registry also lists a second primary hypothesis for overall survival, along with secondary outcomes. It does not give the multiplicity procedure governing those tests, the missing-data strategy for the secondary measures, or the full statistical analysis plan. Those omissions limit what can be concluded about family-wise error control, analysis populations beyond the stated endpoint population, and the treatment of incomplete observations.

11. Confidence Intervals and Precision

The PFS estimate for pembrolizumab with maintenance olaparib is a median of 7.1 months, with a 95% confidence interval from 5.6 to 8.7 months. For pembrolizumab with maintenance pemetrexed, the median is 8.3 months, with a 95% confidence interval from 6.9 to 11.5 months. The intervals indicate that the median estimates are not known without uncertainty; they are not ranges containing every participant’s PFS time.

For the relative comparison, the reported hazard ratio is 1.12 and its 95% confidence interval is 0.92 to 1.36. The interval includes 1, the benchmark for equal relative hazards. It therefore does not isolate a direction of effect with high precision. The interval’s endpoints also indicate that the estimate remains compatible with relative hazards below or above that benchmark. The p-value, 0.872, is consistent with the lack of evidence for a group difference in the reported superiority analysis, but it should not be read as evidence that the strategies are identical.

In general, confidence intervals communicate both the estimated effect and the degree of uncertainty around it. A narrower interval indicates greater precision for the quantity being estimated, while a wider one indicates more uncertainty. Precision depends on information in the analysis, including the number and timing of events and the amount of censoring; the registry does not provide all those details for interpreting the interval’s width.

The confidence intervals around the two medians are separate group-level intervals, not a confidence interval for the difference between medians. The hazard-ratio interval is the direct reported interval for the relative comparison. It is therefore the most relevant interval for judging the uncertainty in the sponsor’s treatment-effect estimate.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning illustrations use the observed effect as a starting point and show event targets for time-to-event analyses. These figures are total events across the two comparison groups, not participants per arm. The registry gives illustrations for detecting 100% of the observed effect, 75% of the observed effect, and 50% of the observed effect, with separate columns for 80% and 90% power.

The stated assumptions for these illustrations are a two-sided significance level of 0.05, equal allocation between the two arms, a normal approximation, and an effect to detect equal to the stated fraction of the observed effect. The total-event calculations use the Schoenfeld method. They are planning illustrations computed from the observed effect, not the trial’s original design justification. The trial’s own original sample-size assumptions and justification are not reported in the registry record.

In general, 80% power means that, if the assumed effect and other planning assumptions were correct, the design would have an 80% probability of rejecting the null hypothesis at the stated significance level. The 90% power column describes a higher target probability under the same type of assumptions. These are prospective planning concepts, not probabilities that the observed finding is correct.

Smaller effects are harder to distinguish from random variation than larger effects, so planning to detect a smaller effect generally requires more events or participants, depending on the outcome type. The illustrations therefore show increasing event requirements as the targeted fraction of the observed effect becomes smaller. This is not observed power: it does not measure the study’s achieved power after the results are known, and it does not establish what effect the trial was originally designed to detect.

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 = 1.120, so ln(HR) = 0.1133.
2. d = 4 × (1.96 + 0.842)² / [ln(1.120)]² = 2,444.5, rounded up to 2,445 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 <- 1.12                              # 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(3976323)
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 = 2,445 (80% power), 3,273 (90% power)
 75% of observed effect: events = 4,346 (80% power), 5,818 (90% power)
 50% of observed effect: events = 9,778 (80% power), 13,090 (90% power)
Simulated power with 2,445 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. Analyses Reported by the Sponsor

The registry reports the sponsor’s PFS comparison as a log-rank hazard ratio of 1.12, with a 95% confidence interval of 0.92 to 1.36 and p = 0.872. The arm-level medians are 7.1 months for pembrolizumab with maintenance olaparib and 8.3 months for pembrolizumab with maintenance pemetrexed. These are the posted endpoint results and should be distinguished from any simple re-calculation based on summary values.

Re-calculations from posted summary statistics are unadjusted and may not reproduce the sponsor’s participant-level analysis, censoring rules, or other analysis details. When the registry indicates a discrepancy, the sponsor’s own analysis takes precedence. No discrepancy is flagged for this result, so the reported sponsor comparison is consistent with the posted analysis information.

The sponsor’s hazard ratio and p-value should not be replaced by an estimate inferred from the medians. The hazard ratio uses the time-to-event comparison, while medians summarize the event-time distribution separately by group. The registry identifies the log-rank method and supplies the sponsor’s confidence interval and p-value, but it does not report additional participant-level model details or a separate adjusted estimate.

15. Multiplicity, Missing Data and Analysis Populations

The registry identifies two primary hypotheses: superiority for PFS and superiority for overall survival. It reports the PFS analysis but does not report the overall-survival result in the material described for this analysis. The procedure for controlling the chance of false-positive findings across the primary hypotheses is not reported in the registry record. The record also does not state whether testing was sequential, whether significance thresholds were divided, or whether another multiplicity procedure was specified.

This omission matters because a study with multiple primary hypotheses needs a clear account of how evidence across those hypotheses is interpreted. In general, testing multiple hypotheses without adjustment can increase the chance of at least one false-positive result. The reported PFS p-value should therefore be read as the result of its stated analysis, not as proof that all trial-level multiplicity questions have been addressed.

The registry lists multiple secondary outcomes, including adverse-event counts and patient-reported measures. A multiplicity plan for those secondary comparisons is not reported. Their p-values and intervals may be useful descriptively, but without a stated testing hierarchy or adjustment, they should not be treated as a set of independent confirmatory conclusions.

The PFS analysis population is described as all randomized participants in the Maintenance Phase, and the record states that PFS analysis was not planned or conducted during induction. The detailed criteria for entering the maintenance analysis population and the treatment of participants who do not contribute follow-up are not reported. Also not reported are the missing-data rules for the PFS analysis, the censoring details, and the handling of missing patient-reported questionnaire responses.

In general, missing observations and censoring can affect the information available for an analysis, and assumptions about them can influence interpretation. The registry’s endpoint description specifies Kaplan-Meier estimation and log-rank testing but does not provide enough detail to assess the underlying missing-data assumptions or reproduce the analysis. The full statistical analysis plan, including interim-analysis details and analysis-population definitions beyond those stated, is not reported in the registry record.

16. 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: respiratory tract infection (7.7%), covid-19 pneumonia (5.1%), pneumonia (4.2%), pneumonitis (3.0%), anaemia (2.7%).

The registry reports serious adverse events during induction for 235 of 1003 participants at risk. In the first-course maintenance groups, serious adverse events were reported for 118 of 337 participants at risk with pembrolizumab and olaparib, and for 125 of 332 participants at risk with pembrolizumab and pemetrexed. These figures are counts of participants with one or more serious adverse events, not event rates per person-time.

The registry’s listed leading serious adverse-event terms include respiratory tract infection at 7.7%, COVID-19 pneumonia at 5.1%, pneumonia at 4.2%, pneumonitis at 3.0%, anaemia at 2.7%, and pulmonary embolism at 2.1%. These listed terms should not be added together: an individual may experience more than one event term, and the registry excerpt does not establish that the categories are mutually exclusive.

The death counts are separately reported by registry group; the record does not establish in this summary that these deaths were caused by adverse events. Safety counts should be interpreted in light of the different treatment phases and analysis populations, rather than as a direct efficacy comparison. These data describe reported events and do not determine what treatment is appropriate for an individual.

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

The primary result is an estimate of PFS for two maintenance strategies following pembrolizumab, pemetrexed, and platinum induction. The reported medians are 7.1 months for pembrolizumab with olaparib and 8.3 months for pembrolizumab with pemetrexed. The sponsor’s reported hazard ratio is 1.12, with a 95% confidence interval of 0.92 to 1.36 and p = 0.872. Taken together, the posted results do not establish superiority of the olaparib maintenance strategy for PFS.

This conclusion is limited to the specified endpoint, the maintenance analysis population, the follow-up reflected in the registry result, and the sponsor’s reported comparison. It is not a statement that the treatment strategies are equivalent. The confidence interval includes the no-relative-hazard benchmark and remains compatible with estimates in either direction. A non-significant p-value does not show that a difference is impossible or that both strategies have the same effect.

PFS does not answer every question about treatment. It combines documented progression and death into one endpoint and is not the same as overall survival, symptom control, quality of life, or adverse effects. The trial lists overall survival as another primary hypothesis, but its result is not reported in the registry record. The secondary outcomes address some quality-of-life and safety questions, but their unadjusted status and missing-data details limit confirmatory interpretation.

In general, evidence from a randomized comparison estimates an average contrast between assigned strategies in the analyzed study population. It does not predict an individual participant’s outcome, and it does not establish that a particular choice is suitable for a person. The registry results are statistical evidence about the specified comparison, not medical advice or a complete assessment of the benefits and harms relevant to treatment decisions.

18. Why This Trial Matters Statistically

This trial presents a useful statistical contrast because the two strategies share pembrolizumab and induction chemotherapy while differing in the maintenance agent. That structure focuses the clinical question on maintenance olaparib versus maintenance pemetrexed, while preserving the broader treatment sequence specified in the study. The primary endpoint is time-to-event, so both event timing and incomplete follow-up matter to the analysis.

The reported PFS medians alone might invite a simple comparison, but the sponsor’s analysis uses a hazard ratio and log-rank test across the time-to-event information. This illustrates why endpoint summaries should be interpreted according to the estimand and analysis method rather than by comparing medians as if they were direct treatment-effect estimates. The hazard-ratio interval also makes uncertainty explicit and includes the no-difference benchmark.

The record distinguishes the Maintenance Phase PFS analysis from the Induction Phase and identifies the analyzed population accordingly. This is important for understanding which participants and period the estimate describes. The trial also has a second primary hypothesis for overall survival, and the registry lists quality-of-life and safety outcomes, demonstrating that a clinical trial can pose multiple related but nonidentical statistical questions.

Finally, the registry record illustrates the limits of interpreting posted summary results without a full statistical analysis plan. Multiplicity procedures, detailed missing-data rules, and original design assumptions are not reported. Those details influence how strongly one can generalize from a p-value or treat secondary outcomes as confirmatory. Careful interpretation therefore requires attention to both the reported estimates and the information that remains unavailable.

19. 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 does not report the original sample-size justification or the trial’s design assumptions for the primary analysis. The planning illustrations based on observed effects are not a substitute for that original rationale. The record also does not report details of any interim analyses or how the two primary hypotheses were handled for multiplicity.

The PFS result applies to participants described as randomized in the Maintenance Phase, and the protocol did not plan or conduct PFS analysis during induction. The registry does not provide further detail on the rules defining entry into this analysis population, the reasons participants may not contribute to it, or the detailed censoring and missing-data procedures. Those details would help assess how follow-up and incomplete observations were handled.

The registry specifies Kaplan-Meier estimation and a log-rank analysis but does not report proportional-hazards diagnostics or a time-varying treatment effect analysis. In general, a single hazard ratio may be less informative if relative hazards change over time. The posted medians provide additional descriptive context, but they do not answer whether the relative treatment effect was constant.

Blinding is listed as none, although PFS was assessed by blinded independent central review. The registry does not describe other safeguards against bias in treatment delivery, follow-up, or patient-reported outcome collection. Further, the overall-survival result and the full statistical analysis plan are not reported in this record. These limitations affect the scope and detail of interpretation; they do not by themselves invalidate the reported estimates.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Study of Pembrolizumab With Maintenance Olaparib or Maintenance Pemetrexed in First-line (1L) Metastatic Nonsquamous Non-Small-Cell Lung Cancer (NSCLC) (MK-7339-006, KEYLYNK-006) — NCT03976323.
• Registry record last updated 2026-09-28; status last verified 2026-09.
• Results first posted 2025-02-06.
• 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.

23. Record Summary

NCT03976323 is a completed randomized Phase 3 parallel-group study of pembrolizumab with maintenance olaparib versus pembrolizumab with maintenance pemetrexed after pembrolizumab, pemetrexed, and platinum induction for first-line metastatic nonsquamous non-small-cell lung cancer. The registry lists enrollment of 1003 and no masking.

The primary PFS analysis is for all randomized participants in the Maintenance Phase and uses Kaplan-Meier estimation, with a log-rank hazard-ratio comparison. Median PFS was 7.1 months with olaparib and 8.3 months with pemetrexed. The reported hazard ratio was 1.12, with a 95% confidence interval of 0.92 to 1.36 and p = 0.872. The result does not establish PFS superiority for the olaparib strategy.

The registry identifies overall survival as a second primary hypothesis but does not report its result in this record. Multiplicity procedures, detailed missing-data rules, and original sample-size assumptions are also not reported.