Slow, slow, quick-quick, slow

In this case, the approval was based on a surrogate endpoint – a reduction in beta-amyloid burden – with a requirement for Biogen to conduct further studies to establish a clinical benefit. This approval has been considered by many observers to set a difficult precedent, as the many years it will take for Biogen to demonstrate any clinically meaningful benefit could lead to substantial patient exposure – and costs – while significant uncertainty on the risk/benefit profile remains.

The data to date on reducing cognitive and functional decline are widely considered marginal. The circumstances of the approval are particularly unusual as the linkage between amyloid burden, cognitive impairment, and downstream functional deficits remains a matter of significant debate. Further controversy was also ignited when it
became clear the label did not specify disease stage, despite restrictions within the study population. The outcry was amplified by its proximity to the FDA’s Oncology Drugs Advisory Committee’s conclusion in April that 2 of 6 scrutinised immuno-oncologic indications should be withdrawn due to the lack of positive confirmatory data.

Of course, the debate on accelerating patient access to pharmacotherapeutics is not new. Outside of deliberations concerning the use of ‘surrogate endpoints’, the question of
the magnitude and duration of patient exposure in clinical trials to provide evidence not only of clinical benefit but also the accompanying risks, has waxed and waned over the lifespan of my pharma career. I am sure many will remember the COX-2 inhibitor saga, with accumulating post-marketing evidence of significant cardiovascular risks associated with some – but not all – agents within the class. This, and other exemplars, remind us that even with approvals by the more traditional route, problems can emerge months or indeed years
post-approval. As the recent rapid approvals of SARS-CoV-2 vaccines also reminds us, studies in tens of thousands of subjects can be insufficient when adverse events (in this case thrombocytopenia, and venous and arterial thrombotic events) are relatively rare. Approval remains only one of the steps in the regulatory pathway, with the possibility of label modifications and even withdrawals a long-term possibility.

That the attitudes to the balance of risk/benefit have changed over time should be clear to all observers. Such changes are linked to evolving views not only of what constitutes ‘meaningful clinical benefit’ and ‘meaningful risks’ (the former has risen; the latter has reduced), but also the degree of urgency that drugs with therapeutic value should reach patients with a minimal delay. The drive has been to reduce what is considered ‘delays’ in granting patient access, particularly where ‘unmet need’ is considered high. Having reduced review times (which produce only marginal gains, given that they are a fraction of overall development time), the only other option was to reduce the burden of evidence required for registration, with deferral of confirmatory evidence beyond launch.

At a recent forum, Richard Pazdur, Director of the FDA's Oncology Center of Excellence, rebuffed recent critiques of the accelerated approval approach, citing the extensive list of agents that have benefitted from the program since its inception in the early 1990s (https://www.fda.gov/media/151146/download), with a very small number failing to maintain regulatory requirements regards confirmatory data. It is worth remembering that the overwhelming majority of accelerated approvals has been under his purview, and it could well be considered harsh to question the validity of the program based on the Aduhelm case: the first for some time that hasn’t involved an oncologic.

You may notice my liberal use of quotation marks in this article, in defiance of good practice on their use. I did this however to highlight where the shades of grey exist regards decisioning for accelerated approval, and why a ‘one size fits all’ (there I go again) philosophic viewpoint misses the nuances: after all, God is in the details:

• Surrogate endpoints: objective responses and progression-free survival have been shown to be sufficiently correlated with overall survival to be warranted as surrogates for some (but not all) cancers. However, such correlations aren't perfect, and there will always be disagreement on exactly how much correlation is acceptable (what r-squared value is enough?). Further, is overall survival all that matters in cancer, and is that true for all tumours? Beyond cancer, there are other good examples of correlations that – by consensus – allow surrogacy to be considered acceptable. Cue Aduhelm, where amyloid burden may not be one of them

• Meaningful clinical benefit: Long a subject of debate within and between disease areas. While some margins of benefit can be analysed based on patient perception (e.g. how much change on a visual analogue scale is perceptible to a patient in pain), others such as an extension to cancer survival or a reduction in disability may be harder to agree. In addition, ‘efficacy’ (sorry again) is rarely – if ever – unidimensional. Most disorders, and the impact of most therapeutics, cover multiple outcomes. 'Meaningful benefit' is therefore complex, and not as easy to measure as it might appear (for example, is two months additional overall survival without a reduction in severe cancer pain always a good outcome?)

• Meaningful risks: Risk is notoriously hard to conceptualise, even for physicians used to dealing with it. Just consider the aforementioned risks of serious side effects from SARS-CoV-2 vaccines in comparison with the potential benefits: acceptable to some, deeply unnerving to others. In weighing up this issue at a population level (as a regulator must) the boundaries of ‘meaningful’ are very far from fixed, and clearly highly contextual to the disorder, its stage, the age of the patients, etc…These then need to be considered against the nature, scale and duration of clinical benefits

• Delays: Whatever time it takes to get through development and licensing, it is clear that patients are suffering the consequences of their disease without the potential benefits a new option may bring (as well as its harms). It is clearly frustrating for patients, their carers and clinicians if these periods are longer than they need to be, but equally exposing patients to therapies with unproven benefits and excess risk is an obvious concern. In this regard, the Oncology Center of Excellence at the FDA has very largely called it right, with most accelerated approvals translating into therapeutics rightly remaining available for the long term

• Unmet need: This to me remains one of the most nebulous concepts in medicine. Until many of the disorders we treat (not least cancer) remain incurable, there will always be an unmet need. How do we rate lung cancer against rheumatoid arthritis? For me, there is no answer that is not pejorative. I would rather focus on what meaningful improvement means, especially to patients
  
  So, we are back to where we began: is accelerated approval a good idea, and if yes, was it the right call with Aduhelm? In my opinion, accelerated approval remains highly attractive, where we have good surrogates, and a good handle on what 'meaningful benefit' (last one I promise) looks like. In Stage IV cancer, it makes special sense. It is however highly contextual and needs to be considered on a case-by-case basis. For all the reasons above, one size cannot fit all. Regards Aduhelm, I may eat my words, but I would concur with many experts who feel the balance of benefit/risk, the lack of good surrogacy, and the extended time to unravel the clinical benefit profile post-launch, make me very uneasy. Having spent 15 years working on assets for cognitive impairment – including for Alzheimer's disease – I am acutely aware of the needs of patients (and a lack of options), but fear this time, the FDA did not call this one right.