Johnson & Johnson has reported the final overall-survival analysis from the Phase 3 PAPILLON trial of first-line RYBREVANT, or amivantamab, plus carboplatin-pemetrexed chemotherapy in advanced non-small cell lung cancer carrying EGFR exon 20 insertion mutations. Median overall survival reached 34.3 months with the RYBREVANT regimen compared with 27.9 months for chemotherapy alone.
Those numbers sound straightforwardly positive until the statistical result is examined. The protocol-specified intention-to-treat analysis produced a hazard ratio of 0.87, with a 95% confidence interval of 0.66 to 1.14 and a p-value of 0.307. In other words, the final unadjusted overall-survival difference did not meet conventional statistical significance.
The complication is what happened after patients in the control arm progressed. Around 76% of eligible chemotherapy patients subsequently crossed over to amivantamab, giving many control patients access to the active drug that the trial was originally designed to compare against chemotherapy.
A prespecified analysis attempting to adjust for that crossover estimated a 43% reduction in the risk of death, with a hazard ratio of 0.57 and nominal statistical significance.
That produces a sophisticated but important clinical question: when patients assigned to the control arm later receive the experimental treatment, which survival analysis best reflects the true value of using the drug first?
Why was PAPILLON able to establish RYBREVANT before the final survival result?
PAPILLON enrolled previously untreated patients with advanced or metastatic non-small cell lung cancer carrying EGFR exon 20 insertions. The Phase 3 study compared amivantamab plus carboplatin-pemetrexed with chemotherapy alone.
The trial previously met its primary progression-free-survival endpoint. Amivantamab plus chemotherapy reduced the risk of disease progression or death by approximately 60%, supporting regulatory adoption of the first-line regimen before mature overall-survival data became available.
That earlier result changed clinical practice while the survival experiment was still running. Once patients and physicians knew that amivantamab had substantial activity after progression, withholding the medicine indefinitely from control patients became increasingly difficult to justify.
The resulting crossover benefits patients but complicates statistics.
How does crossover dilute an overall-survival difference?
Imagine two groups beginning with different treatment strategies. If the control group must remain permanently on less effective therapy, any true survival benefit can remain clearly separated.
But if patients in the control group later receive the experimental drug, the survival curves can move closer together. The trial begins comparing “drug now versus no drug” but gradually becomes closer to comparing “drug now versus drug later.”
That is essentially the interpretive problem in PAPILLON. Of 128 chemotherapy-arm patients eligible to cross over after progression, 97, or approximately 76%, ultimately received amivantamab, according to the final analysis reported at the World Conference on Lung Cancer.
Those patients may have benefited from the therapy after their disease progressed, increasing survival in the control arm and making the original randomized comparison less able to measure the value of receiving the drug from the start.
This is not statistical trickery created after an inconvenient result. The crossover-adjusted analysis was prespecified, meaning the trial protocol anticipated the problem.
What did the adjusted analysis show?
Johnson & Johnson reported that the prespecified inverse probability of censoring weighting analysis produced a hazard ratio of 0.57, corresponding to an estimated 43% reduction in the risk of death after adjustment for crossover.
OncLive reported that the same modeling approach estimated median overall survival for the chemotherapy group at 22.1 months after adjustment, compared with the observed 27.9 months when post-progression crossover remained part of the unadjusted result.
A separate two-stage estimation method produced a similarly favourable adjusted hazard ratio, providing some consistency across statistical approaches.
Still, adjusted analyses are models rather than randomized observations. They require assumptions about what would have happened if patients had not crossed over.
The most rigorous interpretation therefore preserves both facts simultaneously: the unadjusted intention-to-treat overall-survival endpoint was not statistically significant, while prespecified analyses designed to account for extensive crossover suggest the treatment effect was substantially diluted by later amivantamab use.
Does PFS2 strengthen the case for treating patients earlier?
PAPILLON also measured progression-free survival through a second disease progression or death, known as PFS2. Median PFS2 was 28.3 months with initial amivantamab plus chemotherapy compared with 17.5 months for chemotherapy, corresponding to a hazard ratio of 0.59.
That outcome is useful because it asks whether starting amivantamab earlier merely shifts the timing of therapy or creates a durable advantage extending through subsequent treatment.
Time to subsequent therapy also favoured the amivantamab arm, at 16.9 months versus 9.9 months, while time to treatment discontinuation reached 14.1 months versus 7.6 months.
Together, those measures support the argument that first-line use influences the broader treatment trajectory rather than simply delivering the same outcome several months earlier.
They still do not replace the randomized overall-survival analysis, but they help explain why physicians may view the first-line regimen favourably despite the headline p-value.
Why are EGFR exon 20 insertion mutations different from common EGFR mutations?
EGFR-mutated lung cancer is not one homogeneous disease. The common exon 19 deletion and L858R mutations respond well to established oral tyrosine kinase inhibitors, while exon 20 insertions historically have been far less sensitive to those therapies.
That created a substantial treatment gap despite the presence of an identifiable oncogenic driver. Amivantamab was developed to bind both EGFR and MET and became one of the first targeted approaches to show meaningful activity in this molecular subgroup.
PAPILLON moved the drug from post-treatment use into first-line combination therapy, changing the initial treatment decision for patients whose tumours carry exon 20 insertions.
The relatively rare mutation also reinforces the importance of comprehensive molecular testing. Patients cannot benefit from genotype-specific treatment if the alteration is missed at diagnosis.
Does the nearly three-year survival figure tell the whole story?
The 34.3-month median is the longest reported overall survival Johnson & Johnson cites for this population, but cross-trial comparisons require caution. Patient selection, subsequent therapies and supportive care evolve over time.
The chemotherapy arm’s 27.9-month median is itself unusually long relative to older historical outcomes, likely reflecting in part the opportunity for many patients to receive amivantamab later.
That is precisely why the raw six-month difference understates the complexity of the result.
Rather than asking whether 34.3 is simply “better” than 27.9, clinicians need to decide whether using amivantamab from the beginning delivers advantages that cannot be recovered fully by waiting until progression.
PFS, PFS2 and crossover-adjusted survival all point toward an early-treatment benefit, but the protocol-specified OS analysis remains statistically non-significant and should not be rewritten as though it were positive.
Could oral exon 20 inhibitors change the competitive landscape?
The market is already evolving beyond chemotherapy and intravenous amivantamab. Several companies are developing oral tyrosine kinase inhibitors specifically targeting EGFR exon 20 alterations.
That creates a potential convenience advantage because oral medicines can avoid repeated infusion visits and some antibody-related toxicities. Emerging first-line studies may therefore challenge amivantamab-based treatment if they show competitive efficacy with simpler administration.
PAPILLON sets a substantial efficacy benchmark, however. Any new entrant must compete not only against chemotherapy but against a regimen with established first-line progression-free-survival benefit and nearly three years of observed median survival.
The competitive question will increasingly include sequencing. Physicians may eventually need to choose between an antibody-based combination first, an oral targeted inhibitor first or a strategy that preserves one modality for resistance after the other.
Can subcutaneous RYBREVANT solve some of the treatment-burden problem?
Johnson & Johnson is separately developing and commercializing subcutaneous amivantamab formulations intended to reduce administration time and infusion-related reactions.
Recent COPERNICUS results in common EGFR-mutated lung cancer showed low rates of several treatment-related events when subcutaneous RYBREVANT FASPRO was combined with LAZCLUZE and proactive prophylactic strategies. Fewer than 1% of patients discontinued due to treatment-related adverse events at one year in that study.
Those findings come from a different patient population and should not be directly substituted for PAPILLON safety results. They nevertheless demonstrate Johnson & Johnson’s broader strategy of reducing the operational burden associated with amivantamab-based treatment.
If subcutaneous administration eventually becomes integrated broadly across relevant lung-cancer indications, some of the convenience advantage enjoyed by oral competitors could narrow.
What should physicians conclude from a negative p-value and positive adjusted analysis?
The most defensible interpretation is neither “PAPILLON failed” nor “PAPILLON proved a 43% survival reduction.”
The randomized intention-to-treat final overall-survival comparison was not statistically significant. That is an important result and must remain visible.
At the same time, three-quarters of eligible control patients crossed over to the active therapy after progression, making the trial poorly suited to answer a hypothetical question in which the control group never receives amivantamab. Prespecified crossover-adjusted analyses attempt to estimate that alternative scenario and strongly favour first-line treatment, but they rely on modeling assumptions.
Clinical decision-making can incorporate the totality of evidence: the previously positive PFS endpoint, the PFS2 advantage, time to subsequent therapy, observed median survival and crossover-adjusted analyses.
PAPILLON has therefore produced something more informative than a simple positive or negative survival headline. It demonstrates one of oncology’s recurring trial-design paradoxes: once an experimental treatment works well enough to become an ethical rescue option for control patients, proving that the earlier strategy extends overall survival can become statistically harder.
For Johnson & Johnson, that is a desirable problem commercially. For clinicians and readers, it is a reminder that a p-value rarely tells the entire story when treatment after progression fundamentally changes the experiment.
