Could we really see another peak in Covid-19 hospitalisations as bad as January once society reopens in June? That was the story widely reported this morning, based on the latest modelling from SPI-M, the government’s advisory committee on modelling for scientific emergencies. The study caught attention not least because back in January very few people had received a vaccine: now, 56 per cent of the adult population has been vaccinated. By July, on current forecasts, every adult in Britain will have been offered at least a first vaccine dose.
How, if vaccines actually work — and there is a lot of evidence to suggest they do — could we end up in as bad a situation as we did before we had population-wide vaccination? As was explained in some reports, the prospect of a peak as bad as that in January is a worst-case scenario. But, as has happened throughout this crisis, worst-case modelling scenarios tend to be promoted very quickly as if they are scientific fact. A lot less attention has been given to the central projections — which show a much-less dramatic increase of additional cases in the summer or autumn. Nevertheless, SPI-M has fed this morning’s interpretation by picking it out in its summary, which reads:
“It is highly likely that there will be a further resurgence in hospitalisations and deaths after the later steps of the roadmap. The scale, shape, and timing of any resurgence remain highly uncertain; in most scenarios modelled, any peak is smaller than the wave seen in January 2021, however, scenarios with little transmission reduction after Step 4 or with pessimistic but plausible vaccine efficacy assumptions can result in resurgences in hospitalisations of a similar scale to January 2021.
What really catches the eye — and which has gone entirely unremarked so far this morning — is the assumptions on vaccine efficacy built into the SPI-M models. The models, produced by three different teams at Imperial College, Warwick and the London School of Hygiene and Tropical Medicine, assume that the Pfizer and AstraZeneca vaccines are markedly less effective than trials have shown them to be. Imperial, for example, has produced figures for two scenarios: one in which two doses of AstraZeneca is 63 per cent effective at preventing symptomatic infection and 80 per cent effective at preventing hospitalisation, and a ‘pessimistic’ scenario in which it is 50 per cent effective at preventing symptomatic infection and 70 per cent effective at preventing hospitalisation. By contrast, Phase 3 AstraZeneca trials found it to be 70 per cent effective at preventing symptomatic infection (falling to 62 per cent when two standard doses were administered, as opposed to a standard dose followed by a half dose).
Updated data from February suggested that efficacy against symptomatic infection rose to 82 per cent when the two doses were administered 12 weeks or more apart — as is the practice in Britain. A more recent US trial found AstraZeneca to be 76 per cent effective at preventing symptomatic infection and 100 per cent effective at preventing serious cases requiring hospitalisation.
The Warwick and LSHTM models also assume poor efficacy for AstraZeneca, assuming that it is only 90 per cent and 85 per cent effective at preventing hospitalisations respectively. The assumptions for the Pfizer vaccine are generally closer to the results of trials, but they are still lower in each case. In other words, these models seem to have ignored the real-world data from vaccine trials and assumed that the vaccines are not as good as they have proved to be in real life. No wonder, on those assumptions, we end up with teams modelling frightening spikes, even after most of the population has been vaccinated.