I came across this in an article in The Daily Sceptic by Dr Richard Ennos a retired evolutionary biologist. He used a method called proportional reporting rate to try and show that the UK Yellow Card system signals the mRNA vaccines are unsafe and should be withdrawn with immediate effect. The idea of PRR is to see if there is a significantly different reporting rate of some incident of interest as a proportion of all reports compared to the same proportion in other interventions. For example, are there more heart attacks reported for vaccine X as a proportion of all reports about vaccine X compared to vaccines in general. If so, this might be a safety signal. In this case Dr. Ennos found that there are statistically significant more incidents of several types among certain age/sex groups when comparing the mRNA vaccines to the Astrazeneca vaccine. He then goes on to claim that this is a strong safety signal and There can be no question that the mRNA vaccines should be withdrawn with immediate effect.
This procedure is littered with conceptual issues.
Dr Ennos justified comparing the mRNA vaccines to the AZ vaccine on the basis that we make the very conservative assumption that the AZ vaccine does not increase the frequency of the particular adverse reactions that we are investigating. This a complete muddle. AZ might increase the frequency of particular adverse reactions compared to what? All vaccines? All interventions? A placebo? Whatever the the comparison, why should AZ not decrease the frequency? Note that this is compatible with it being unsafe. There might be a particularly large number of non-serious adverse reactions in the case of AZ which would lower the frequency. The mRNA vaccines may not be exceptional compared to other vaccines. It might be the AZ vaccine.
In any case, the proportion of some type of reported incident tells us nothing about the frequency of that type of incident. For example, Dr Ennos points out that the rate of reporting of serious and fatal adverse events is nearly three times higher for the adenovirus AZ vaccine (3.912 serious or fatal reaction reports per 1,000 doses) than for either of the mRNA vaccines PF or MO (1.341 and 1.344 serious or fatal reaction reports per 1,000 doses respectively). This means that even if the proportion some incident of interest, say heart attacks, is higher among the mRNA vaccines than the AZ vaccine, the frequency of reported heart attacks may well be lower than for the AZ vaccine. More significantly, it tells us nothing about how the frequency of heart attacks among those who received an mRNA vaccine compares to those of the same age and gender who were not vaccinated.
These are reported incidents, not proven to be actual incidents. As such, they are prone to confounding factors, particularly as the AZ vaccine and the mRNA vaccines were mostly administered at different stages of the epidemic. There were differences in the publicity around covid vaccines in general and around the specific vaccines, in the ease of reporting of incidents, in public concern with health and in the background rate of other health concerns which might mistakenly be reported as vaccine incidents. For example, in the early stages of the vaccination programme the vast majority of vaccines were AZ. This was a brand new vaccine at a time when many people were very aware of health issues. It would not be surprising if an exceptional number of people reported minor side effects or minor symptoms that were not in fact connected to the vaccine, thus lowering the ratio of serious reported incidents to minor reported incidents. By the time the mRNA vaccines were administered the population had grown more familiar with the idea of a covid vaccine, vaccination had become routine, and there would be less motivation to report a minor symptom.
In Moderation 