← Clinical Trials
Advanced / Metastatic NSCLC Phase 3 PK Non-Inferiority NCT05388669

PALOMA-3: Complete Statistical Analysis of Lazertinib With Amivantamab in Advanced or Metastatic NSCLC

An independent statistical review of the randomized phase 3 PALOMA-3 study comparing lazertinib with subcutaneous amivantamab SC-CF versus lazertinib with intravenous amivantamab in participants with EGFR-mutated advanced or metastatic non-small cell lung cancer.

Trial start: 2022-08-05  ·  Primary completion: 2024-01-03  ·  Status: Active, not recruiting
Scope of this record

This page separates reported trial results from statistical interpretation. The numerical results on this page are restricted to the ClinicalTrials.gov data posted on ClinicalTrials.gov for PALOMA-3. The registry provides three posted primary-endpoint analyses, all concerning pharmacokinetic exposure or concentration measures.

Registry note: This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.

1. Trial at a Glance

PALOMA-3 is a randomized, parallel, open-label phase 3 study in participants with EGFR-mutated advanced or metastatic non-small cell lung cancer. The study compares lazertinib with amivantamab administered as a subcutaneous co-formulation with recombinant human hyaluronidase against lazertinib with intravenous amivantamab, with the registered primary analyses focused on amivantamab pharmacokinetic concentration and exposure measures.

418
Enrollment
2 treatment arms
3
Primary endpoints
All PK measures
1.032
AUC ratio
90% CI 0.976–1.090
1.427
Ctrough estimate
90% CI 1.266–1.610
FeaturePALOMA-3
Trial namePALOMA-3
ClinicalTrials.gov identifierNCT05388669
PhasePhase 3
Therapeutic areaOncology
ConditionAdvanced or Metastatic Non-small Cell Lung Cancer
Population descriptorParticipants with epidermal growth factor receptor (EGFR)-mutated advanced or metastatic non-small cell lung cancer
AllocationRandomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Enrollment418.0
Arms2
Lead sponsorJanssen Research & Development, LLC
Sponsor typeIndustry
StatusActive, not recruiting

2. Clinical Question

The central statistical question is whether administering amivantamab subcutaneously as a co-formulation with recombinant human hyaluronidase produces amivantamab pharmacokinetic exposure and concentration measures that are non-inferior to those observed with intravenous amivantamab when both treatment strategies include lazertinib.

Population

Participants with EGFR-mutated advanced or metastatic non-small cell lung cancer.

Intervention

Lazertinib with amivantamab subcutaneous and co-formulated with recombinant human hyaluronidase (SC-CF).

Comparator

Lazertinib with amivantamab administered by intravenous (IV) infusion.

Primary question

Are the amivantamab pharmacokinetic concentration and exposure measures with SC-CF sufficiently preserved relative to IV administration under the registered non-inferiority framework?

This distinction is important. The registered primary endpoints are not overall survival, progression-free survival, tumor response, or another clinical outcome. They are pharmacokinetic measures of amivantamab concentration and exposure. Consequently, the principal statistical interpretation concerns comparability of drug exposure, not a direct estimate of comparative cancer-treatment efficacy.

3. Trial Design

01
Randomize418 participants enrolled
02
Arm ALazertinib + amivantamab SC-CF
03
Arm BLazertinib + amivantamab IV
04
PK assessmentCtrough and AUC measures
05
Non-inferiorityPrimary PK comparisons
Allocation
Randomized
Design model
Parallel
Masking
None
Primary purpose
Treatment
ARM A

Lazertinib + Amivantamab SC-CF

  • Lazertinib
  • Amivantamab subcutaneous and co-formulated with recombinant human hyaluronidase (SC CF)
ARM B

Lazertinib + Amivantamab IV

  • Lazertinib
  • Amivantamab intravenous (IV infusion)
What randomization contributes: Randomization creates the basis for comparing the two administration strategies without assigning treatment according to observed participant characteristics. The registry identifies the allocation as randomized but does not provide a randomization ratio in the ClinicalTrials.gov record, so no particular allocation ratio should be inferred from the total enrollment.

4. Primary Endpoints

ClinicalTrials.gov lists three primary endpoints. All three are pharmacokinetic measures, but they differ in whether they measure a steady-state trough concentration, an earlier trough concentration, or exposure over a specified interval.

Primary endpointTime frameDefinition
For All Regions Other Than the European Union (EU) and Others Accepting Cycle 2 Day 1: Observed Serum Concentration (Ctrough) of Amivantamab at Steady State Pre-dose on Cycle 4 Day 1 (each cycle of 28 days) Ctrough was the observed serum concentration of Amivantamab at steady state immediately prior to the next drug administration.
For EU and Any Applicable Region: Observed Serum Concentration (Ctrough) of Amivantamab Pre-dose on Cycle 2 Day 1 (each cycle of 28 days) Ctrough was the observed serum concentration of Amivantamab immediately prior to the next drug administration.
Area Under the Concentration (AUC) Time Curve of Amivantamab From Day 1 to Day 15 (AUC [Day 1-15]) of Cycle 2 Cycle 2: Arm A: pre-dose, 24, 48, 72, 96, 168, and 360 hours (hrs) post-dose on Day 1; Arm B: pre-infusion, end of infusion (EOI)+10 minutes, EOI+2, EOI+6, EOI+24, EOI+48, EOI+72, EOI+168, and EOI+360 hrs post dose on Day 1 AUC (Day 1-15) defined as area under the concentration time curve from Cycle 2 Day 1 to Day 15 were reported.

Why these endpoints are different

A trough concentration and an area-under-the-curve measure summarize different aspects of pharmacokinetics. A Ctrough is a concentration observed immediately before the next administration. It is therefore a point measurement at a prespecified dosing time. An AUC, by contrast, summarizes exposure across an interval by integrating concentration over time.

That distinction matters statistically because an AUC ratio near 1.0 addresses overall exposure over the specified interval, whereas a Ctrough ratio addresses the concentration immediately before the next administration. The three primary analyses therefore provide related but non-identical evidence about pharmacokinetic comparability.

5. Statistical Methodology

The registry identifies non-inferiority as the hypothesis type for all three primary analyses. The stated primary hypothesis tests were based on a one-sided test at significance level alpha = 0.05 to demonstrate non-inferiority of amivantamab SC-CF relative to amivantamab IV.

FeatureRegistry information
Hypothesis typeNon-inferiority
Primary comparisonArm A: Lazertinib With Amivantamab SC-CF vs Arm B: Lazertinib With Amivantamab Intravenous (IV) Infusion
Analysis direction as reportedArm B Vs Arm A
Primary testing frameworkOne-sided test at alpha = 0.05
Confidence intervals90%, two-sided
Formal method fieldNot reported
Effect measure listed for first two analysesCox regression model
Effect measure for AUC analysisGeometric mean ratio
Important registry-methodology distinction: The ClinicalTrials.gov record does not name a statistical method for these analyses, while the effect-measure field for the first two primary analyses says “Cox regression model.” A Cox regression model is ordinarily a model used to estimate hazard-related effects, whereas Ctrough is a continuous pharmacokinetic concentration. The safest interpretation is therefore to reproduce the registry's reported fields without treating “Cox regression model” as proof of a conventional time-to-event hazard-ratio analysis.

PK primary endpoint evaluable sets

The analysis populations are defined by pharmacokinetic evaluability rather than simply by enrollment. For the steady-state Cycle 4 Day 1 Ctrough analysis, the registry describes a PK primary endpoint evaluable set consisting of randomized participants who had received all doses in Cycles 1-3, without dose modifications, and provided a Cycle 4 Day 1 Ctrough. For the EU and applicable-region Cycle 2 Day 1 Ctrough analysis, the set included randomized participants who had received all doses in Cycle 1, without dose modifications, and provided the Cycle 2 Day 1 Ctrough.

For the Cycle 2 AUC analysis, the registry describes a PK primary endpoint evaluable set consisting of randomized participants who had received all doses up to Cycle 2 Day 1, without dose modifications, and provided all necessary PK samples to derive the primary endpoint.

Why the analysis population matters: These are not simply “all 418 randomized participants.” A PK evaluable set requires adequate dosing and the required pharmacokinetic measurements. Consequently, the estimated exposure ratio describes the participants meeting the relevant PK-evaluability conditions, not necessarily every enrolled participant.

6. Results: Primary Pharmacokinetic Endpoints

ClinicalTrials.gov contains three formal statistical analyses corresponding to the three registered primary endpoints. Each provides an effect estimate and a two-sided 90% confidence interval. The ClinicalTrials.gov record does not provide p-values, so none are added to the results.

6.1 Steady-State Ctrough — Cycle 4 Day 1

Reported primary estimate

1.427

Two-sided 90% CI: 1.266–1.610

Comparison reported as Arm B Vs Arm A

FeatureReported result
EndpointFor All Regions Other Than the European Union (EU) and Others Accepting Cycle 2 Day 1: Observed Serum Concentration (Ctrough) of Amivantamab at Steady State
Time framePre-dose on Cycle 4 Day 1 (each cycle of 28 days)
UnitMicrograms per milliliters (mcg/mL)
Estimate1.427
Confidence interval90% two-sided CI 1.266–1.610
HypothesisNon-inferiority
Analysis directionArm B Vs Arm A
Clinical Biostats interpretation

The reported estimate of 1.427 is an arm-comparison estimate for the observed amivantamab Ctrough at steady state, with the registry specifying the comparison direction as Arm B versus Arm A. Numerically, an estimate above 1 means the reported quantity for Arm B is estimated to be higher relative to Arm A under the analysis as posted.

It does not mean that Arm B produces 42.7% greater clinical efficacy, nor does it mean that participants have a 42.7% greater probability of a clinical outcome. The endpoint is a pharmacokinetic concentration, not a clinical event endpoint.

The two-sided 90% confidence interval, 1.266–1.610, describes statistical uncertainty around the reported estimate under the analysis framework. It does not describe the range of Ctrough values for individual participants and does not by itself establish whether a prespecified non-inferiority margin has been met.

The registry states that non-inferiority was tested using a one-sided alpha of 0.05, but the ClinicalTrials.gov record does not state the numerical non-inferiority margin. Therefore, the confidence interval cannot be compared with a specific margin on this page. A non-inferiority conclusion requires the prespecified margin and the direction in which it is applied.

6.2 Ctrough — Cycle 2 Day 1

Reported primary estimate

1.145

Two-sided 90% CI: 1.040–1.261

Comparison reported as Arm B Vs Arm A

FeatureReported result
EndpointFor EU and Any Applicable Region: Observed Serum Concentration (Ctrough) of Amivantamab
Time framePre-dose on Cycle 2 Day 1 (each cycle of 28 days)
Unitmcg/mL
Estimate1.145
Confidence interval90% two-sided CI 1.040–1.261
HypothesisNon-inferiority
Analysis directionArm B Vs Arm A
Clinical Biostats interpretation

The reported estimate of 1.145 compares the observed amivantamab Ctrough between the two randomized administration strategies in the direction specified by the registry. The point estimate is above 1, so the reported Arm B quantity is estimated to be higher than the corresponding Arm A quantity under the posted analysis.

That numerical difference should not be translated directly into a difference in tumor response, survival, or patient benefit. Ctrough is a pharmacokinetic concentration measured at a specified time point.

The 90% confidence interval of 1.040–1.261 gives a measure of precision around the estimated comparison. It indicates that the estimate is not being presented as an exact population quantity. As with the other primary analyses, the interval is not an interval for individual participant concentrations.

The p-value is not included in the ClinicalTrials.gov record. More importantly for a non-inferiority analysis, a p-value alone would not determine the non-inferiority conclusion. The prespecified margin, the direction of the comparison, and the corresponding confidence-interval criterion are central to the decision rule.

6.3 AUC Day 1–15 — Cycle 2

Geometric mean ratio

1.032

Two-sided 90% CI: 0.976–1.090

Effect measure: Ratio of geometric means

FeatureReported result
EndpointArea Under the Concentration (AUC) Time Curve of Amivantamab From Day 1 to Day 15 (AUC [Day 1-15]) of Cycle 2
Time frameCycle 2: Arm A: pre-dose, 24, 48, 72, 96, 168, and 360 hours (hrs) post-dose on Day 1; Arm B: pre-infusion, end of infusion (EOI)+10 minutes, EOI+2, EOI+6, EOI+24, EOI+48, EOI+72, EOI+168, and EOI+360 hrs post dose on Day 1
UnitMicrograms*hour per milliliters
Effect measureGeometric mean ratio
Estimate1.032
Confidence interval90% two-sided CI 0.976–1.090
HypothesisNon-inferiority
Analysis directionArm B Vs Arm A
Clinical Biostats interpretation

The geometric mean ratio of 1.032 indicates that the geometric mean AUC in Arm B relative to Arm A was estimated at 1.032 under the reported comparison direction. On a ratio scale, 1.0 represents equal geometric means; 1.032 is therefore close to 1.0.

It does not mean that 3.2% more patients benefited from one administration strategy, nor does it describe a 3.2% change in a clinical outcome. It describes a ratio of pharmacokinetic exposure measures.

The two-sided 90% confidence interval of 0.976–1.090 provides the uncertainty around the geometric mean ratio. The interval crosses 1.0, which is the equality value for a ratio, but that fact alone does not answer the non-inferiority question. Non-inferiority is assessed against a prespecified margin rather than against the null value of equality alone.

The ClinicalTrials.gov record identifies the hypothesis as non-inferiority and specify a one-sided alpha of 0.05, but they do not provide the numerical margin. Consequently, this page can describe the estimate and its precision without assigning a formal non-inferiority conclusion based on an unstated margin.

7. How to Read the Three Primary Results Together

The three estimates should not be collapsed into one overall statistic. They address different pharmacokinetic summaries and different regional or cycle-specific conditions.

Primary measureEstimate90% two-sided CIStatistical meaning
Steady-state Ctrough, Cycle 4 Day 11.4271.266–1.610Relative comparison of observed steady-state trough concentration in the reported direction
Ctrough, Cycle 2 Day 11.1451.040–1.261Relative comparison of observed trough concentration in the reported direction
AUC Day 1–15, Cycle 21.0320.976–1.090Ratio of geometric mean exposure in the reported direction

The AUC estimate is particularly straightforward to interpret on a ratio scale because its effect measure is explicitly identified as a geometric mean ratio. An estimate of 1.032 is close to the equality value of 1.0. The Ctrough analyses are different: although the registry reports “Cox regression model” in the effect-measure field, the endpoints themselves are continuous concentration measures and the record does not name a statistical method. That discrepancy should be preserved rather than silently converting the results into conventional hazard ratios.

Ratio-scale interpretation
Ratio = geometric mean in Arm B / geometric mean in Arm A

For a ratio measure, 1.0 represents equality. Values above 1 indicate a higher quantity in the numerator arm, while values below 1 indicate a lower quantity. The clinical interpretation depends on what the underlying quantity measures and on the prespecified non-inferiority margin.

8. Non-Inferiority Logic

Non-inferiority trials ask a different question from superiority trials. The objective is generally not to show that the new strategy is better than the comparator. Instead, the design specifies an amount of loss that would be considered acceptable for the new strategy, and the observed data are evaluated against that prespecified margin.

Equality is not the margin

A ratio of 1.0 represents equality, but a non-inferiority design can permit a specified amount of deviation from equality.

The margin matters

The numerical margin determines what range of results is compatible with the prespecified non-inferiority claim.

Confidence intervals matter

Non-inferiority is commonly assessed using a confidence bound relative to the prespecified margin, rather than by asking only whether a p-value is below 0.05.

Direction matters

The registry identifies the comparison as Arm B Vs Arm A while stating the non-inferiority hypothesis for SC-CF relative to IV. The direction of the reported estimate therefore has to be read carefully.

Margin not reported: The ClinicalTrials.gov record identifies the hypothesis as non-inferiority and the one-sided alpha as 0.05, but they do not provide a numerical non-inferiority margin. It would therefore be inappropriate to invent a margin or declare a formal non-inferiority result from the confidence intervals alone.

Why the 90% confidence intervals are notable

The primary analyses report two-sided 90% confidence intervals. A two-sided 90% interval corresponds to a 5% probability allocated to each tail under the usual confidence-interval construction. The registry separately describes the hypothesis test as one-sided with alpha = 0.05. In a non-inferiority framework, this pairing is commonly connected through the equivalence between a one-sided 5% test and a two-sided 90% confidence interval, provided the interval and test are constructed from the same statistical model and estimator.

That connection does not remove the need to know the non-inferiority margin. The confidence interval supplies the uncertainty around the estimate; the margin supplies the prespecified boundary against which that uncertainty is evaluated.

9. Statistical Methods Explained

Why use a geometric mean ratio for AUC?

Pharmacokinetic exposure measures such as AUC can be analyzed on a multiplicative scale. A geometric mean ratio compares the typical exposure level between two groups in a way that is naturally suited to ratio-based interpretation. An estimate of 1.032 therefore means the estimated geometric mean exposure in the numerator arm is 1.032 times that of the denominator arm, under the comparison direction specified by the registry.

The ratio is not the same as an arithmetic mean difference. A difference of 10 units and a ratio of 1.10 answer different statistical questions and have different interpretations.

Why is 1.0 important for a ratio?

For a ratio, 1.0 is the equality value. If two geometric means were exactly equal, their ratio would be 1.0. A ratio of 1.032 is therefore close to equality on the multiplicative scale. But a ratio being close to 1.0 is not, by itself, the formal definition of non-inferiority.

Why is non-inferiority judged against a margin rather than simply a p-value?

A superiority p-value typically addresses a null hypothesis such as equality. Non-inferiority instead asks whether the new strategy is sufficiently close to the comparator according to a prespecified clinically or scientifically justified margin. A result can fail to demonstrate superiority while still satisfying a non-inferiority criterion, depending on the margin and the confidence interval.

For PALOMA-3, the ClinicalTrials.gov record explicitly identify a non-inferiority hypothesis and a one-sided alpha of 0.05, but the numerical margin is not reported. That means the posted estimates can be interpreted descriptively without reconstructing an unstated decision threshold.

What does the 90% confidence interval tell us?

The confidence interval quantifies uncertainty around the estimated treatment-group comparison under the specified statistical framework. For example, the AUC ratio is 1.032 with a two-sided 90% confidence interval of 0.976–1.090. The interval is narrower than it would be for a two-sided 95% confidence interval under the same estimator and standard-error structure, although the actual width depends on the analysis.

The interval does not tell us that 90% of individual participants will have ratios inside that range. It is a statement about uncertainty in the estimated population-level comparison under repeated-sampling reasoning.

Why does the analysis population differ from total enrollment?

The pharmacokinetic primary endpoint evaluable sets require specific dosing histories and the necessary concentration samples. A participant can therefore be randomized into the trial without contributing to every primary PK analysis. This is different from an efficacy analysis that might be defined simply by randomized assignment.

Why should the “Cox regression model” field be interpreted cautiously?

The ClinicalTrials.gov record lists “Cox regression model” as the reported effect measure for the two Ctrough analyses, while no statistical method is named. Cox regression is conventionally associated with time-to-event outcomes and hazard-based effects. Ctrough, however, is a continuous pharmacokinetic concentration. Because the ClinicalTrials.gov record does not provide enough detail to reconstruct the exact model specification, the appropriate educational approach is to report the field exactly as posted without relabeling the estimate as a conventional hazard ratio.

Why can a confidence interval cross 1.0 and still be compatible with non-inferiority?

Because equality and non-inferiority are different hypotheses. For the AUC analysis, the interval is 0.976–1.090 and therefore includes 1.0. That means exact equality cannot be ruled out in the usual two-sided sense represented by that interval. But non-inferiority does not require demonstrating superiority or even necessarily ruling out equality. What matters is whether the confidence bound satisfies the prespecified non-inferiority margin. The margin registry-reported to investigators is therefore essential to the formal interpretation.

10. Understanding the Analysis Direction

The statistical-analyses data state “Arm B Vs Arm A”. Arm A is lazertinib with amivantamab SC-CF, while Arm B is lazertinib with amivantamab IV infusion. The interpretation of a ratio therefore depends on keeping the numerator and denominator straight.

ArmInterventionRole in reported comparison
Arm ALazertinib With Amivantamab SC-CFReference side of the reported Arm B Vs Arm A comparison
Arm BLazertinib With Amivantamab Intravenous (IV) InfusionComparison side of the reported Arm B Vs Arm A comparison

This matters especially for non-inferiority. Saying that an estimate is “above 1” is incomplete unless the reader knows which arm is in the numerator. The registry's explicit comparison direction prevents the estimate from being casually interpreted in the opposite direction.

Interpretation principle

A ratio is not self-interpreting. Always identify what quantity is being compared, which arm is the numerator, which arm is the denominator, and what boundary defines non-inferiority. Omitting any of these can reverse the practical interpretation of a numerical result.

11. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized arm. These are presented as affected participants over participants at risk.

Safety measureArm A
Lazertinib + Amivantamab SC-CF
Arm B
Lazertinib + Amivantamab IV
Serious adverse events59/20664/210
Serious adverse events by arm
Arm A · SC-CF
59/206
Arm B · IV
64/210

The ClinicalTrials.gov record is limited to serious adverse events by arm. It should therefore not be expanded into conclusions about overall adverse-event incidence, grade-specific toxicity, treatment discontinuation, or individual adverse-event categories.

The denominators also illustrate an important distinction from the overall trial enrollment of 418. The serious-adverse-event summary uses 206 participants in Arm A and 210 in Arm B as the affected/at-risk denominators reported in the ClinicalTrials.gov record. Those denominators sum to 416, not 418, so the safety table should not be silently presented as if it were a complete accounting of all 418 enrolled participants.

Safety versus PK: The pharmacokinetic primary endpoints and serious-adverse-event counts answer different questions. PK endpoints describe drug concentration or exposure. Serious adverse events describe an important category of participant safety events. Neither should be used as a surrogate for the other without additional evidence.

12. What the Primary Estimates Do — and Do Not — Mean

The 1.427 Ctrough estimate

The reported 1.427 estimate describes the comparison of observed steady-state amivantamab Ctrough in the reported Arm B-versus-Arm-A direction. It does not mean a 42.7% improvement in cancer outcomes, a 42.7% higher probability of response, or a 42.7% difference in survival.

The 1.145 Ctrough estimate

The reported 1.145 estimate describes the Cycle 2 Day 1 Ctrough comparison in the reported direction. It is a pharmacokinetic comparison at a specified pre-dose time point, not a measure of relative treatment efficacy.

The 1.032 AUC ratio

The 1.032 geometric mean ratio indicates that the reported Arm B-to-Arm-A exposure ratio is close to 1.0. The 90% confidence interval of 0.976–1.090 expresses uncertainty around that estimate. Formal non-inferiority interpretation still requires the prespecified margin.

Why the p-value is not the effect size

A p-value is a measure associated with a hypothesis-testing procedure; it is not a measure of the magnitude of pharmacokinetic similarity or difference. The ClinicalTrials.gov record does not report p-values for these three analyses. Even if they did, the p-value would not replace the effect estimate, confidence interval, and non-inferiority margin when interpreting a non-inferiority PK study.

13. Limitations

14. Why This Trial Matters Statistically

PALOMA-3 is a useful statistical teaching case because the central problem is not a conventional superiority comparison of a clinical outcome. Instead, the trial illustrates how randomization, pharmacokinetic endpoints, ratio-scale estimation, confidence intervals, and non-inferiority testing fit together.

ConceptHow it appears in PALOMA-3
RandomizationParticipants are allocated randomly to two parallel treatment strategies.
Two-arm comparisonLazertinib with amivantamab SC-CF is compared with lazertinib with amivantamab IV infusion.
Pharmacokinetic endpointPrimary endpoints include observed amivantamab Ctrough and AUC.
Ratio-scale analysisThe AUC endpoint uses a geometric mean ratio.
Confidence intervalsAll three posted primary analyses provide two-sided 90% confidence intervals.
Non-inferiorityThe registered hypothesis is non-inferiority with a one-sided alpha of 0.05.
Analysis populationPK primary endpoint evaluable sets require specified dosing and PK data.
Analysis directionThe posted analyses specify Arm B Vs Arm A.
Method reportingThe statistical method is not reported; the first two analyses list “Cox regression model” as the reported effect measure.
Safety analysisSerious adverse events are reported by arm as affected participants over participants at risk.

15. Non-Inferiority Versus Superiority: A Statistical Distinction

It is tempting to read every randomized comparison as a search for a treatment effect greater than zero or a ratio different from 1.0. That framework is not sufficient for a non-inferiority trial.

Superiority question

Is the new strategy statistically distinguishable from the comparator in the favorable direction?

Non-inferiority question

Is the new strategy sufficiently close to the comparator according to a prespecified acceptable margin?

For PALOMA-3, the registry explicitly identifies the hypothesis as non-inferiority. This changes how the confidence intervals should be read. The AUC interval of 0.976–1.090 includes 1.0, but that alone neither proves nor disproves non-inferiority. The relevant question is whether the confidence bound lies within the prespecified acceptable range.

This is one reason why reporting only a p-value can be particularly misleading in non-inferiority work. A statistically significant result against equality can coexist with an estimate that is unfavorable in magnitude, while a nonsignificant superiority test can coexist with a valid non-inferiority conclusion. The estimand, margin, confidence interval, and analysis population must be considered together.

16. Pharmacokinetic Endpoint Interpretation

Ctrough

Ctrough is the observed serum concentration immediately before the next drug administration. In PALOMA-3, the first primary endpoint specifies steady-state Ctrough before dosing on Cycle 4 Day 1, while the second specifies Ctrough before dosing on Cycle 2 Day 1 for the EU and applicable regions.

These are point measurements. They do not summarize the entire concentration-time profile. A participant's Ctrough can therefore provide information about residual concentration immediately before another administration without providing the same information as total exposure over an interval.

AUC

AUC summarizes concentration over time. The registered PALOMA-3 AUC endpoint covers Cycle 2 Day 1 to Day 15, with the registry specifying different sampling descriptions for the two arms.

Conceptual exposure measure
AUC = ∫ C(t) dt

The area under the concentration-time curve integrates concentration across a specified time interval. The resulting quantity reflects cumulative exposure over that interval rather than concentration at a single instant.

For the AUC endpoint, the posted effect measure is a geometric mean ratio. This makes the analysis naturally interpretable on a multiplicative scale. The estimate of 1.032 can therefore be read as a ratio close to unity, while the 90% confidence interval provides the corresponding uncertainty range.

17. Analysis Populations and the Meaning of “Evaluable”

The ClinicalTrials.gov record distinguishes the PK primary endpoint evaluable sets according to the amount of dosing and PK information available.

EndpointPK evaluable-set requirement described in the registry
Cycle 4 Day 1 CtroughRandomized participants who received all doses in Cycle 1-3, without dose modifications, and provided Cycle 4 Day 1 Ctrough.
Cycle 2 Day 1 CtroughRandomized participants who received all doses in Cycle 1, without dose modifications, and provided Cycle 2 Day 1 Ctrough.
Cycle 2 AUCRandomized participants who received all doses up to Cycle 2 Day 1, without dose modifications, and provided all necessary PK samples to derive the primary endpoint.

The progressively different requirements illustrate why an analysis estimate is always conditional on an analysis population. The Cycle 4 Ctrough analysis requires more treatment exposure and later sampling than the Cycle 2 analyses. Therefore, the three estimates should not automatically be interpreted as if they were calculated from identical sets of participants.

General statistical lesson: “Randomized” and “evaluable for a particular endpoint” are different concepts. Randomization defines treatment assignment; endpoint evaluability determines whether a participant contributes the measurements needed for a particular analysis.

18. Safety and Efficacy Should Remain Separate

The registry-reported PALOMA-3 data contain both PK primary results and serious-adverse-event counts. These should be presented as separate evidence streams rather than combined into an informal benefit-risk score.

Pharmacokinetics

Answers questions about observed concentration and drug exposure under the specified dosing and sampling framework.

Safety

Describes serious adverse events among the participants included in the registry-reported at-risk denominators.

Non-inferiority

Provides a formal framework for assessing whether a specified new strategy remains within an acceptable margin.

Clinical outcomes

Would address outcomes such as response or survival, but such results are not part of the ClinicalTrials.gov record for this page.

This separation is particularly important because a PK non-inferiority result does not automatically establish clinical non-inferiority for every possible clinical endpoint. Pharmacokinetic comparability and clinical outcome comparability are related questions but are not mathematically identical estimands.

19. Record Summary

PALOMA-3 is a randomized phase 3, parallel, open-label trial with 418.0 enrolled participants and two treatment arms. The registered primary endpoints are pharmacokinetic: two observed amivantamab Ctrough measures and one Cycle 2 AUC measure. The registry identifies non-inferiority as the hypothesis type and states that the primary hypothesis tests used a one-sided alpha of 0.05.

The three posted primary estimates are 1.427 for steady-state Ctrough with a two-sided 90% confidence interval of 1.266–1.610, 1.145 for Cycle 2 Day 1 Ctrough with a two-sided 90% confidence interval of 1.040–1.261, and a 1.032 geometric mean ratio for Cycle 2 AUC with a two-sided 90% confidence interval of 0.976–1.090. The analysis direction is reported as Arm B Vs Arm A.

The central statistical lesson is that these numbers must be interpreted in the context of pharmacokinetic estimands, ratio-scale effects, endpoint-specific evaluable populations, confidence intervals, and non-inferiority margins. The ClinicalTrials.gov record does not provide the numerical non-inferiority margin or p-values, so this page does not manufacture either one. The first two analyses also contain a reporting inconsistency in which “Cox regression model” appears as the reported effect measure while no statistical method is named; that should be acknowledged rather than converted into an assumed conventional hazard-ratio analysis.

Clinical Biostats methodology: A rigorous trial-results page should distinguish what the registry reports from what statistical theory allows us to infer. For PALOMA-3, the most defensible interpretation is therefore built around the posted PK estimates and confidence intervals while preserving the non-inferiority framework, the analysis-population definitions, the comparison direction, and the limits of the registry-reported statistical-method detail.

20. Related Tutorials

Learn more about the methods used in this trial:

21. Related Calculators

22. Sources

Continue through Clinical Biostats

Explore statistical tutorials, calculators, and additional clinical-trial analyses covering the methods used to design, analyze, and interpret randomized studies.