Over the past week a furious debate has broken out over the merits of delaying the second dose of the approved COVID vaccines, to increase the number of people who can be given a single dose.
The case against doing so is clear: the labels of the approved vaccines are unambiguous that two doses should be given within a certain time window. As is always the case with drug labels approved by regulators, this reflects the way in which the vaccines had been used during the large, pivotal trials that underpinned their approvals. Quite simply, if that’s how the trials were conducted, we do not have any direct evidence for how this vaccine (or any other drug) would perform if used differently.
Why, then, the proposal to now ignore that and vaccinate more people with single doses and delay (possibly indefinitely) given them the second dose prescribed on the label?
Because circumstantial evidence suggests that one dose may offer substantial, if incomplete, protection for a period, and because the faster we increase the number of people with at least some protective immunity the quicker we can suppress community transmission. If we already had unlimited supplies of vaccine doses, then it would be a moot point, but with bottlenecks affecting the entire vaccine delivery chain from manufacture through to immunising people, restricting people to a single dose will approximately double the number of people given some protection in a given period of time.
The benefit of doing that may be very substantial. Modelling from the University of Toronto predicts that increasing the number of people protected, by limiting individuals to a single dose, reduces severe COVID events (ICU stays and death) by between 30 and 40% over a 6 month period, which could amount to 20,000 lives saved and many more expensive, traumatic stays in ICUs that are struggling for capacity. And that’s even before considering the downstream benefits, with less disruption to care for other serious diseases such as cancer cases whose surgery is currently being delayed, and a quicker return to something approaching normal activity across the economy as a whole. That is no small prize, to be easily passed over.
Put like that, who could argue against such a policy? Following the guidance on the vaccine labels seems like imprudent adherence to regulatory red-tape that borders on the reckless. Yet the strongest voices criticising the move come from eminent scientists. Their principal concern is that we don’t know how much protection a single dose of any of the vaccines actually delivers, and we have absolutely no clue how long such protection might last. If the protection is weak or fades quickly we could use up all the available vaccine supplies and protect no-one. In the absence of data, the chorus of scientists squeal, it would be madness to take such a risk.
What we have is a classic case of decision-making in a data-poor environment. There simply isn’t enough information available to make a choice (one way or the other) with confidence. We encounter this a lot in fast-changing situations, such as pandemics, but it is also surprisingly common in every-day life in more normal times too.
And it turns out the very worst people to listen to, in such situations are scientists. At least in the modern world, science has achieved its de facto position as the final arbiter of truth precisely by demanding proof. Hypothesis only turns into accepted fact in the face of extensive experimental evidence. Conjecture is only useful to guide future experimentation – scientists at least attempt to deal in a currency of hard facts.
Trained to think in that way, many scientists are dismissively calling the idea of adopting a single-dosing strategy “guesswork”. There is no doubt that, compared to following the data and administering the vaccine according to the proven protocol, it does indeed have a shaky evidence-base – but far from wandering into the darkness, it is absolutely the rational choice.
Here’s why.
Firstly, while we do not have direct evidence from clinical trials designed to estimate the efficacy of a single dose of these approved vaccines, we are not operating without any inisghts. All of the vaccine Ph3 trials reported COVID events occurring in the vaccine and placebo groups from the day of the first dose. All of them show significantly lower levels of infections in the vaccine arm than the placebo arm in the period up to the point where the second dose was given. Where the data is available (for example for the Moderna vaccine), once the initial 14 day period after the first dose has passed, but before the second dose was administered, the efficacy of the vaccine was as good as in the whole trial period (there were 2 cases in the 14,000 or so people given one dose of the vaccine and 35 cases in the 14,000 who got placebo).
The trial wasn’t designed with this as an end-point (so the analysis of the numbers is, by definition, what statisticians call post hoc – much less powerful than a priori analysis according to the trial design). In addition the number of infections we are looking at in this truncated analysis is small, which means the confidence intervals around the estimate of efficacy are large. Yet, despite these caveats, the estimates of efficacy prior to the second dose from all the Ph3 vaccine trials paint a highly consistent picture – and although the confidence around any one of them is lower, taken together they are pretty compelling. Not proof, in the way scientists use that term, but too convincing to ignore nonetheless.
The issue of how long the protection might last is harder to address, because the trials administered a second dose to essentially every participant. So even the post hoc analysis only tells us that the protection was substantial for a few weeks. But even still, we are not blind as to what will likely happen. Protection will fade, rather than stop suddenly one day, so if (as the data we do have suggest) the protection is near complete for a few weeks, it is reasonable to conclude it will be sustained, even if at declining levels, for several months (if not longer).
But even the doubts around this assumption are not an insurmountable problem. We plan to vaccinate hundreds of thousands of people a week – so within a just a handful of days we will have more people to study with a single dose than in all the clinical trials reported to date. If we follow the rates of infection in that first cohort of people receicing the single dose, then we will quickly see if infection rates (which should initially be vanishingly low, once the first 14 day period has passed) start to rise as protection wanes. There are now two scenarios. If the protection is very short lived, as the scientists worry, then we will know very quickly and can revert to administering the vaccine according to the approved label. If, however, the protection is much more sustained, it will take a while to collect the data – but equally, it doesn’t matter because the people who received the single dose are being protected.
However you look at it, the cost of trying the single dose strategy is acceptably small.
There are more subtle, but nevertheless relevant, reasons to be optimistic about the single dose approach. If the initial protection from the priming shot is incomplete or shorter lived, then those individuals might be susceptible to catching a natural infection – but, the data suggest, this will be milder and may, in many cases, be completely asymptomatic. And this natural infection will act as the best possible boost (far superior, one imagines, to the planned second dose of the vaccine). As the very first vaccination story, using live cowpox to protect against smallpox, showed us – mild, natural infection is the very best protection. And epidemiology shows us how true that is for COVID – the rates of symptomatic re-infection are vanishingly low, even many months after the first wave. Natural immunity to real infection is strong and long-lived, just as you would predict from immunology textbooks.
On the one hand, therefore, you have the conservative approach favoured by many of the most vocal scientists – follow the data and do what you know for certain works; and on the other you have a strategy, supported by convincing albeit circumstantial arguments, that may well save tens of thousands of lives (and have a plethora of other benefits too).
Which way should we go?
Making this kind of decision is something we face endless in business, and its useful to formulate and quantitate the inputs that we have – because, quite frequently we have to make a binary decision (to go with the single-dose strategy or not; to make an investment in a particular company or not and so on) based on continuous probability distributions. For example, I may believe a given drug candidate has a 10% chance to work in a large Phase 3 trial – should I invest? The answer of course is “it depends” (in this case on the cost of making the investment, the value I will retain if it fails the trial, and the value created if it succeeds).
Coming back to dose sequencing for the COVID vaccine, therefore, I can create a similar formulation: is the risk-adjusted benefit exceed the cost plus the risk-adjusted harms? (You could, if you wish, invert the formulation to address the question of whether to follow a two-dose strategy).
This formulation teaches us something of general importance in business, as well as how to best use our limited doses of COVID vaccine. It tells us that when the cost in very low, and the risk-adjusted harms are negligible, we don’t need much confidence that we will deliver the benefit in order to take the action.
This is exactly the same reasoning that has led to the widespread adoption of masks – it’s a very low cost intervention, with essentially no risk of harms, so even if the benefit is modest it is a no-brainer to adopt. Of course, as with the dose sequencing for the vaccine, such a thought-process is vulnerable to the “there’s no direct evidence for the benefit” argument. But as for the vaccine, this demand for evidence leads to a dangerous missed opportunity.
Any time you might get a benefit, with little risk of harm, then if the cost is low you should do it. That isn’t a scientific way of thinking, because it leads you to do things for which there is scant evidence. But waiting until there is hard evidence means the opportunity to win has been lost. This applies in investing (once there’s proof it works, everyone wants to buy it and the price leaves no room for profit) but it also applies to vaccine use… if we wait for evidence that the single doses are “sufficiently effective” the opportunity to save lives and economic ruin will have been passed up.
So to arrive at a recommendation as to how governments should sequence vaccine doses, the place we really need to focus our attention is on the risk-adjusted harms component of the equation (because if that’s low, given that the cost of the action is negligible, we don’t need to be very convinced of the benefits to take the action). Why is DrugBaron so confident that giving a single dose wont cause harm?
Firstly, because we can mitigate that risk by collecting data, while we implement the strategy and change course if the data tells us to (as set out above).
Secondly, the vaccine only has to be 50% effective in two people to reduce the overall risk of infection that is achieved with 95% protection in one person (if I protect 10,000 people out of a population of 20,000 with 1% risk of infection at 95% efficacy, I get 5 cases from the protected subgroup and 95 cases in the unprotected subgroup for a total of 100 infections; if I protect all 20,000 people with 50% efficacy, again I get 100 infections). The existing dta strongly suggest that a singl dose will deliver at least 50% protection for at least a few months.
It is interesting to contrast this situation with another, where a binary decision has to be made with incomplete information: approval of aducanumab for treatment of Alzheimer’s Disease. The exact same rational framework makes it abundantly clear why the decision has to be refusal: weighing risk-adjusted benefit against cost plus risk-adjusted harms looks very different to the COVID vaccine decision. Now, the cost of approval is enormous (it will cost billions of dollars to treat even a subset of Alzheimer’s patients), and while the risk of harm is relatively small (at least at the lower dose), the maximum possible benefit is also small. We can argue about the detail of the trial design and the statistical analysis as to whether there is any evidence of benefit at all, but what is absolutely unarguable is that even if there is a benefit, it is very small.
Learning to think about how we make decisions in the absence of sufficient information, dramatically changes the way we act. It teaches us that we cannot just “follow the science” because that pathway is way too conservative (at least if that means only do things for which there is clear, direct evidence). But at the same time, it does not promote “guessing” – you can clearly distinguish a decision, such as the approval of aducanumab, which is a bad idea from the adoption of single-dose COVID vaccinations, which is a good idea, despite the lack of definitive proof in both cases.
This why we should have scientific advisors, but not rely on scientists to make decisions. It is why scientists rarely make good investors or businessmen – too many decisions need to be made in the absence of much information.
So next time you hear a scientist arguing against the UK’s enlightened single dose COVID vaccination policy, point them to DrugBaron.
RxCelerate Ltd is an outsourced drug development platform based near Cambridge, UK. We specialize in delivering an entire road map of drug development services from discovery and medicinal chemistry through to formal preclinical development and clinical up to Phase IIa. In the last five years, we have witnessed dramatic changes in the drug development …
