Is it Safe? Was the MHRA slow to detect the AZ vaccine's suspected CSVT?

Pharmacovigilance is more tricky than turning things up to 11

If you find these posts interesting and think others might do so, spread the word.

Share Is It Safe?

It’s a month after EMA and the MHRA simultaneously announced, after expert advisory group review in both cases, that despite a number of extremely rare reports of cerebro sinus vein thromboses (CSVT) that the benefits of the AZ vaccine still outweighed the harm.

Despite this, The Daily Telegraph has an exclusive : ‘Revealed: Why Britain’s regulator missed the link between the AstraZeneca jab and rare blood clots

In the article they argue that because the AZ vaccine started a rollout a full month earlier, meaning 11 million doses had been administered, the UK regulator should have found the CSVT issue.

They argue this was not found because:

  1. The MHRA’s signal detection methods were not sensitive enough, and that other countries had turned threat sensitivity gauge ‘up to 11’.

  2. The MHRA did not have an open and transparent approach to the CSVT cases and did not tell the public about the issue until the 18th of March when it announced 5 cases.

  3. That the MHRA had lost access to Eudravigilance, the European database of spontaneous reports.

The Telegraph does not say that the MHRA covered up the reporting of CVST, but merely says that the MHRA was not as quick to spot the issue as others.

They also note:

‘The issue is not that it has been left looking flatfooted or even that earlier detection would necessarily have altered its advice, but that the delay left it unable to shape international policy and confidence in what remains a vital vaccine in the fight against Covid 19 for the world.’

They also importantly note:

There remains no doubt the benefit of taking the AstraZeneca jab outweighs the risk now associated with the vaccine. This is the judgment not just of the MHRA but also the European Medicines Agency (EMA) and the World Health Organization (WHO).

So there is no complaint about the regulatory decisions of the MHRA, but a very narrow point about a failure to find the CSVT issue sooner, which they argue has undermined worldwide confidence in the vaccine.

I’m going to walk through the claims in the Telegraph, but first a brief timeline.

7th March: Denmark, Norway, and Iceland report cases of blood clots and deaths. It’s important to note these were not CSVT cases.

11th March: EMA and the MHRA both release similar press releases underlining the benefits of AZ vaccine outweigh the risks (since the blood clots weren’t out of line with expected thrombosis cases in a similar population).. No CSVT is noted.

15th March: Germany’s Paul Erlich Institute reports on seven cases of CSVT, including 3 deaths. Germany suspends the AZ vaccine.

18th of March: Both EMA and the MHRA release press releases and hold conferences on the CSVT issue, having both already convened and consulted with experts on the issue. Both organisations issue warnings to prescribers and patients, but back the benefits over the risks of the AZ vaccine.

7th of April: Both EMA and the MHRA again hold press releases at the same time, noting that the benefits continue to outright the risks of the AZ vaccine.

The Telegraph also has an analysis of a document published by the MHRA, which is just a simple count of cases carried by an academic:

From January 4 to March 14, a total of 532 “blood system events”, including 20 deaths, came through the UK’s Yellow Card system relating to the AstraZeneca jab, according to an analysis of published MHRA data by Dr Hamid Merchant, a pharmaceutical scientist at the University of Huddersfield. There were thousands of non-blood-related reports besides.

Of the thrombotic events recorded, four related to CVST (but no deaths were recorded), 55 were non-site specific and there were clusters of 64 and 66 cases in the lungs and deep veins respectively. There were then 267 general bleeding events and six deaths, three of which resulted from cerebral haemorrhage. Finally, there were 60 cases of thrombocytopenia, including 2 deaths.

This appears to be based on this document published by the MHRA.

The important thing to note here are the reports related to CVST. The first column is cases, the second is the number of those that were fatal. The term come from an adverse effect standardised set of terminology to enable sharing of regulatory information called MedDRA. Once a spontaneous report is received, in the UK a Yellow Card, it will be coded with this terminology.

So, on the 15th of March when the Paul Erlich Institute was reporting 7 cases with 3 fatal, the UK had 4 cases, none fatal. Germany had administered 1.6 million vaccines, the UK 11 million. Norway also had 5 cases, three of whom died, and had administered 130,000 vaccines.

Another part of this story are some astute and vigilance physicians in Germany:

The VIPIT story began on 27 February, when Sabine Eichinger, a hematologist at the Medical University of Vienna, was confronted with an unusual patient. A 49-year-old nurse had sought help at a local hospital the day before, suffering from nausea and stomach discomfort, and was transferred to Eichinger’s hospital. She had a low platelet count and computed tomography scans found thromboses—blood clots—in the veins in her abdomen and later in arteries as well. “There was little we could do at this stage,” Eichinger says. The patient died the next day.

The combination of low platelet count, or thrombocytopenia, and clots kept Eichinger thinking, however. “It’s very striking,” she says. Platelets, also known as thrombocytes, help form blood clots, so low levels usually lead to bleeding, not clotting. “You would think that low platelets and thromboses are opposites really.” One condition where they occur together is called disseminated intravascular coagulation, when severe infection, injury, or cancer trigger clotting so widespread it uses up all the platelets, “but she had none of these things,” Eichinger says.

Vigilance is key.

The science of detecting adverse effects of medicines and vaccines is called pharmacovigilance. Vigilance is key. Without suspicious healthcare professionals and patients making a suspected link between a vaccine and drug or vaccine, we don’t get reports. And for that reason, along with other reasons, there is a level of under-reporting in pharmacovigilance systems. It can vary, with more severe reactions perhaps more likely to be reported, but if no link is made by individuals, then no report can be made. Under-reporting in one systematic review is estimated to be about 94%. That is, for every one report made, 19 may not be made.

So, whatever reports you have in your database, be it the Yellow Card scheme, or the Eudravigilance system in the EU, you have to be aware that not all the reports may be there. Indeed, if a type of reaction is very novel, then the a-ha moment may not arise in mind of the healthcare professional caring for the patient. It is also worth noting that the reports that are made as suspicions of harm, they are not proven to be be linked to the vaccine.

It is also worth noting that once a suspected adverse effect of a vaccine is highlighted by regulatory authorities and the media, more reports are stimulated. This can be termed a notoriety bias or stimulated reporting.

And now onto a few specifics in the Telegraph:

On March 11, the UK's newly sovereign Medicines and Healthcare products Regulatory Agency (MHRA) put out a statement saying it could see no evidence of a problem.

“We continually monitor the safety of vaccines to ensure that the benefits outweigh any potential risks,” said Dr Phil Bryan, the MHRA’s Vaccines Safety Lead.

“Reports of blood clots received so far are not greater than the number that would have occurred naturally in the vaccinated population. 

“The safety of the public will always come first.”

But the MHRA was, it appears, wrong. An investigation by The Telegraph has established that signals had been firing unnoticed in the UK’s Yellow Card database for at least a month, perhaps longer.

Remember in the timeline at this point, the issue of CSVT had not arisen, and the MHRA statement is dealing with reports of generalised blood clots in Denmark, Norway, and Iceland. There statement as far as it pertains to that issue is accurate. A similar statement was put out by EMA on the same day.

The Telegraph immediately follows the above quote with this:

In January, a patient suffered a brain clot following their first dose of the AstraZeneca jab, The Telegraph has confirmed. Then in early February, two similar cases followed, including a death and a life-changing CVST clot in a young adult. All had low platelets and all were reported into the Yellow Card system.

It is worth noting, that from the start of the use of the AZ vaccine to the 14th of March the MHRA had 4 cases with no fatalities. So, at least one of these cases, the fatal one, was definitely reported later than the 14th of March. The MHRA would not have known of it. So this segment does not prove the 11th of March press release ‘wrong’.

A psychologist argues that the MHRA was slow to respond to the emergence of the AZ vaccine CSVT issue.

Prof Stephan Lewandowsky, a psychologist at the University of Bristol studying the rollout of Covid-19 vaccines, told the Financial Times on Friday: “The MHRA was slow in responding to the emergence of a specific constellation of symptoms associated with the AstraZeneca vaccine and slow to communicate what they were finding — and I am not the only one who thinks so.”

Both EMA and the MHRA communicated about the CSVT issue on the 18th of March and the 7th of April, so the UK was no slower than EMA. The Paul Erlich Institute did not report their cases until the 15th of March. At that point, the UK had only 4 cases reported and no fatalities in 11 million administered vaccines. One could make the argument that the UK should have found the issue sooner because they started vaccinating sooner, but the UK had only 4 reports submitted (none of which were fatal) and all of which fell within the realms of the natural occurrence of this rare event at that point.

The Telegraph also mentions the algorithms used to examine the Yellow Card Data. In simplistic terms, the system looks for a disproportionate number of reports of a particular adverse reaction compared to all the other drugs in the same database. There is a threshold above which it might be that a safety Signal arises.

To sift such data, regulators build algorithms that must balance “sensitivity” against leg-work. The more sensitive the algorithm, the more warning signals it will throw up to investigate - and many of those labour-intensive investigations will prove fruitless.

It is not known exactly what parameters the MHRA set but it is clear they were not as sensitive as those used by some regulators in Europe. 

The MHRA says it followed a principle of applying “statistical techniques which can tell us if we are seeing disproportionately more cases than we would expect to see based on what is known about background rates of illness in the absence of vaccination”. This is reflected in the regulator’s initial statement when it said clotting reports were not above normal.

But other countries turned the sensitivity gauge up to 11. “Our policy is if it is associated with a death, or very serious adverse drug reaction, we will look into it right away,” David Benee Olsen, senior advisor at the Norwegian Medicines Agency, told The Telegraph. 

There is no evidence that the MHRA signal detection systems were less sensitive than European regulators'. Certainly, I imagine they were more than likely set exactly the same as those at EMA. Such systems are not Marshall amps that one turns up to 11. The comment by Olsen at the Norwegian Medicines Agency about following up serious reports and patient deaths is as true of the MHRA as it is of the Norwegian Medicines Agency. But remember the UK had no fatal cases of CSVT up the 14th of March, and we have no evidence one way or the other than the MHRA was not looking at individual cases of CSVT in more depth.

The Telegraph also notes cultural differences on vaccine pauses and openness about vaccine harms, noting the J&J pause in the US, and the pauses in Europe on the 7th of March. However, the early pauses were not about CSVT, and arguably wrong because EMA did not later endorse them. They also note the warnings of anaphylaxis issues for the Pfizer vaccine by the MHRA, which were eventually withdrawn, and say:

Unfortunately, perhaps, the MHRA did not adopt the same approach with the problems surrounding the AstraZeneca jab - or at least it does not appear to have.

That sentence is heavily caveated for good reason. As already noted, by the 14th of March 2021 the MHRA did not have a good signal. The Telegraph says:

The MHRA says it got its first report of CVST with thrombocytopenia in the week of February 8, but the regulator did not tell the public about the issue until March 18 when it announced five cases. It did so just moments before the EMA opened a high profile press briefing on the issue, which by then was dominating world headlines. 

Remember the German cases were reported on the 15th of March. Before then the UK did not have a clear signal (4 cases), and no deaths. Three days after the German cases were reported the MHRA had issued warnings at the same time as EMA, had held expert committees, and had clearly worked with EMA. The Telegraph tries to make out that working with EMA is a bad thing, rather than good co-operation. Should the MHRA have released a competing press release the day before, without co-operating?

Since then, the MHRA has made three further announcements regarding new cases identified in the UK. Late on Good Friday, after most people had tuned out for the bank holiday weekend, it said it had recorded a total of 30 cases, including seven deaths. Then on 7 April, again during a high profile EMA press conference, it revealed the count had grown to 79 and 19 deaths. The latest figures, quietly released on Thursday evening, show the number of UK clotting cases with thrombocytopenia to stand at 100 with 22 deaths. 

The above suggests the MHRA were trying to sneak new cases our under cover of a bank holiday, undercover of an EMA press conferences, and ‘quietly’ on a Thursday night. That’s quite a way of looking at things.

Would we prefer the MHRA waited until after the Bank Holiday? I doubt it. The release on the 7th of April was during an EMA press conference, but again the MHRA was holding a press conference itself at the same time. Yet another example of the MHRA and EMA working together. The MHRA/JCVI panel was widely broadcast on TV, and highly praised, and was hardly hidden. The data released last Thursday evening was released when expected and changed nothing.

It is not known how many of those 22 deaths are covered in the Yellow Card data Dr Merchant analysed between January 4 and 14 March but it seems likely that a large number of them - perhaps the great bulk - will date back to that period. 

I find this quite confusing. The data Mr Merchant analysed between January 4th and 14th of March contained 4 cases of CSVT and no deaths. So none of those 22 deaths are covered in the Yellow Card data Dr Merchant looked at, because they did not exist at that point.

As noted before, once a rare adverse effect is drawn attention to by regulatory authorities and the media, stimulated reports are created as doctors, particularly haematologists dealing with rare events like this would go looking for them retrospectively. Some of those cases might have happened before the 14th of March, but none of them were reported to the MHRA until after wider reporting of the issue.

Another issue raised in the article is the UK’s access to Eudravigilance:

Another reason for the MHRA’s slower reaction, suggest observers, could be that it lost access to Eudravigilance, the vast European database into which all adverse drug reactions are reported, when the UK left the orbit EMA regulation on 31 December last year.

I’m extremely skeptical that this is major issue. EMA would want their hands on any data the UK obtained about vaccines given our early roll-out. Since it is clear that EMA and the MHRA have been working closely, look at the co-ordination in three press releases and statements from each body which shows they have been working closely together. It’s also worth noting that on the 11th of March 2021 EMA said ‘As of 10 March 2021, 30 cases of thromboembolic events had been reported among close to 5 million people vaccinated with COVID-19 Vaccine AstraZeneca in the European Economic Area.’

This data would not have strengthened any signal of CSVT the MHRA would have had, so I think it is worth discounting that as an issue.

The key moment was the release of the Paul Erlich Institute data on the 15th of March which changed everything. Within 3 days the UK’s MHRA had acted as quickly as the EU’s EMA.

So, it appears that the MHRA did not have a clear signal in the UK before the astute and vigilant physicians in Germany spotted unusual cases which were reported to EMA by the Paul Erlich Institute. That lead to an increased vigilance worldwide with further cases reported. The Telegraph piece itself notes that a UK group of leading haematologists started discussing a case in the UK on the 16th of March, the data after the Paul Erlich Institute cases had been reported, and reported them on the 17th of March.

If there is anything to be learnt here, it is that one should be vigilant to the potential adverse effects of any new treatment, and if more cases had been reported perhaps a signal may have been detected sooner. But is far from clear that the MHRA had a legitimate signal with 4 cases in 11 million vaccines, or that they have responded slowly to the safety signal raised by the Paul Erlich Institute, or failed to work closely with European colleagues.

Of course, you can construct a retrospective narrative, as the Telegraph has, but hindsight is a wonderful tool for doing so.

I should also make my conflict of interest clear, that I have worked at a MHRA Yellow Card centre, and hold an ongoing honorary position (although I have no committee positions at the MHRA, income from the MHRA, or insider information from the MHRA). I neither speak for the MHRA, not do the MHRA have any hold over me. So what I write is based on publicly available material, and no doubt subject to the biases that I have, some of which I am aware of, and others of which I am no doubt not.

I freely admit to a bias based on knowledge which is in the public domain, that people at the MHRA have been planning for months to address safety monitoring in the COVID-19 vaccines to a high degree of professionalism. I believe people at the MHRA believe passionately in the job they are doing, and find some of the suggestions that things were ‘quietly’ released pretty disgraceful without evidence of such intent. I dislike the tone of the Telegraph article.

I also think that the Telegraph is playing with fire here, during the ongoing vaccination campaign, and I don’t think their story warrants the title ‘Why Britain’s regulator missed the link between the AstraZeneca jab and rare blood clots’.

Because they didn't based on the data they had, from what I can see.

That’s all and stay safe.


Don’t forget to report your suspected adverse effects from medicines and vaccines. In the UK, this means using the Yellow Card Scheme.