Fresh questions over DLL3

Judging by the enthusiasm Merck & Co showed in acquiring Harpoon and then teaming up with Daiichi Sankyo on its anti-DLL3 lead last week, you might be excused for thinking that approval of Amgen’s Imdelltra had swept away doubts about targeting the DLL3 antigen.

But recently published documents detailing the FDA’s review have laid bare the problems thrown up by Imdelltra’s registrational Dellphi-301 study, some relating to protocol deviations and the under-reporting of adverse events. In giving Imdelltra accelerated approval the FDA appears to have made exceptions, but this now puts the pressure on Amgen’s confirmatory Dellphi-304 trial.

Some of the harshest language in the documents, published on the FDA’s website last month, relates to "the large number of underreported AEs by [South] Korean sites, not captured during the applicant’s monitoring activities". This, the FDA states, "is a major concern and put into question the integrity of the safety data submitted to support the approval of tarlatamab for the proposed indication".

Lead author

One site in particular showed that such under-reporting related to 14 of 19 patients, and amounted to 61 adverse events, of which two were deaths due to underlying disease.

The Dellphi-301 results were published in the NEJM, and the FDA audits postdated this, but a precise comparison of the safety datasets is complicated by the NEJM paper including other trial sites. The final Imdelltra label includes the FDA-audited numbers, but adds another dose from Dellphi-301 as well as patients from Dellphi-300, a dose-expansion trial. As such, the NEJM paper alone might not give the full picture.

Amgen told ApexOnco that it audited Dellphi-301 as part of an independent programme, and stressed that the FDA’s own assessment concluded that the study was consistent with regulatory requirements.

Indeed, the documents make clear that the agency weighed up the pros and cons, and decided that approving Imdelltra was warranted. The FDA also called out unreported adverse events that weren’t likely to affect overall data reliability, and Amgen subsequently acknowledged the deficiencies and moved to conduct 100% source data verification.

One important factor is that the indication in question, SCLC, is highly intractable, and Amgen stressed that before Imdelltra’s approval “there had been few significant breakthroughs in the treatment of SCLC over the last three decades”. But it’s notable that accelerated approval leaves the door open for the FDA to change its mind if new evidence emerges.

Special measures?

The agency’s considerations extended to the implementation of a risk evaluation and mitigation strategy for Imdelltra. But, the documents reveal, such a measure was deemed unnecessary as the FDA decided that Imdelltra’s toxicity profile could be adequately conveyed through labelling, which talks extensively about the risks of cytokine release and neurotoxicity.

The FDA’s multi-discipline review document reveals that in Dellphi-301 there was at least one important protocol deviation in 81 of the trial’s 220 patients. Dellphi-300, meanwhile, saw at least one such deviation in 158 of 205 patients. The FDA decided that these deviations either didn’t affect results or were unlikely to have done so meaningfully.

Audits on Dellphi-301 revealed 393 previously unreported adverse events in 111 patients. 371 of these events were deemed treatment related, and included three serious events and 28 grade 3 events. The review also states that Dellphi-301 enrolled patients who were at least third line, but Amgen was seeking – and ultimately received – a second-line label.

Though no unreported grade 4 or 5 events were deemed treatment related, the various inconsistencies make it surprising that no adcom was convened to debate Imdelltra’s approval. By comparison, last October’s adcom over the controversial Codebreak-200 trial of Amgen’s Lumakras showed the type of savaging that such panels can inflict.

Pariah no more

Any revelation that Imdelltra, and by implication targeting DLL3, is associated with greater risk of toxicity than previously thought will be of note to other companies also seeking to develop drugs that hit DLL3.

DLL3 had become something of a pariah in the wake of AbbVie’s 2019 discontinuation of rova-T, an anti-DLL3 ADC acquired in the $5.8bn takeover of Stemcentrx. But more recently development has picked up, with OncologyPipeline currently listing 13 clinical-stage anti-DLL3 assets, including naked MAbs, bispecifics, Car-Ts and ADCs.

A special focus fell last week on Merck & Co’s MK-6070, acquired through January’s $680m takeout of Harpoon, and now licensed for $170m to Daiichi Sankyo.

The recent view has been that rova-T’s problem wasn’t the DLL3 target but the poor design of the ADC molecule. However, the FDA documents suggest that future developers of DLL3-targeting assets might continue grappling with toxicity.