The OpenSAFELY paper gets Professor Graeme Jones's gong for the most impactful paper of 2020. What does it hold for rheumatologists?
To say the least, 2020 was an unusual year. Dealing with anxious patients was made much more difficult by a certain level of anxiety among us carers as well as a virtually total lack of useful information about the intersection between Covid-19 and rheumatology in the first part of 2020.
Tasmania had a short-lived lockdown and the southern part of the state still hasn’t had any community transmission! Having a moat and a drawbridge may have contributed to this, although I remain unconvinced that closing borders has achieved anything apart from re-election of the premiers involved and chaos in the hospitality/tourism sector.
This relative lack of disease didn’t stop all sorts of strange happenings. The patient who cancelled a phone consultation because she was worried about catching Covid. Many patients falling through the cracks by staying firmly in their isolated shacks for most of last year. My institution still insisting on VC approval for travel in Australia and wearing PPE gear while screening clinical trial patients (justified based on reputational risk but which only makes people think we must be doing something dangerous).
All of the immunosuppressed patients were listed as at-risk and many stopped face-to-face work during this period. The SLE patient delivering meals to the index Covid case in Burnie hospital without PPE was a notable exception. When I asked her to query this with management, she was told Covid only happens in big cities. Following federal intervention, one month’s isolation for 5000 people and 128 cases later the Burnie outbreak was finally quelled, so hopefully lesson learnt. The official report did not mention any of this of course.
From February to June most of the advice was based on first principles and we all know how right that approach often is! I was stopping and reversing MTX in anyone who had Covid. Indeed, I was giving folinic acid scripts to everyone on MTX just in case. Of the 6 patients I had, two were on MTX, many were older and most had other ‘at risk’ illnesses. All survived but two have had a CFS-like illness after. By June Tasmania was starting to open up and practice has been busy ever since. The patients were a bit ahead of the government on this one.
It was around this time that the OpenSAFELY paper was published (1). While there were many other notable papers, this gets my gong for the most impactful paper of 2020. The name is a bit of an oxymoron as England is still in lockdown, but the results really gave a lot of clarity to the advice we could give our patients.
In summary, this paper involved record linkage in the UK organised through the University of Oxford. It was incredibly large with over 17 million people included and over 10,000 Covid related deaths. This was around 40% of all patients in England.
There were the usual caveats with large, matched data: lots of missing data (which was imputed, which generally weakens associations), no PCR confirmation of cases (as many older people did not have confirmation) and some difficulties with telling whether deaths were directly due to Covid or associated. Somewhat surprisingly, there were fewer subjects from London and more rural subjects. This may have led to some lower estimates of death rates.
These caveats were balanced against the speed of data collection, a prespecified analysis plan (which avoids spurious results obtained by massaging the data) and early data sharing.
There are some key take-home big picture messages that are relevant for all of us.
- The model was fantastic as it explained 93% of the variation in risk of death. I am generally ecstatic if I get 20% or better in my epidemiologic studies. This means there is little unexplained risk – for example, genetic factors – or error in the variables don’t have much impact.
- The overall mortality rate was quite low, peaking at under 1% even in those aged above 80. This is a lot better than the 10% that was reported in Northern Italy in the early days of the pandemic.
- Males were more likely to die than females.
- Whites were less likely to die than all other races.
- Deprivation had an important independent association, which is why developing countries are really struggling.
- Heart disease, diabetes (doesn’t specify type 1 or type 2), cancer and kidney failure all had epidemiologically (as well as statistically) strong associations with Covid death. Corticosteroid users with asthma had a slightly higher death rate. This data is broadly in agreement with other studies and this is the group to protect. My stepson labelled the virus the boomer remover but, really, it is the sick boomer remover (and I am not including him in my will).
- Smoking and hypertension were protective in adjusted models despite being deleterious in unadjusted models. This is an excellent example of confounding, as this apparent increase was due to higher prevalence of the other risk factors.
The small-picture finding relevant directly to our patients was the report that rheumatoid arthritis, lupus or psoriasis were associated with a very small increase in risk (1.19, 95% CI 1.11-1.27). This is statistically significant but not clinically significant, so I have been reassuring my patients successfully ever since. This compares to 3.5 for an organ transplant.
I was initially excited by other data suggesting some of our therapies may work but most of this has bitten the dust once the well-designed trials were completed. As an aside, there were many very poor papers published in haste last year. This stimulated a lot of work which wasn’t well justified, so there was a lot of wastage among some glittering vaccine triumphs.
Come 2021, it is time to roll out the vaccine to everyone who wants it. It will be interesting to see if the ACR recommendations for biologics and vaccination are accurate when subject to the retrospectoscope.
- Elizabeth J Williamson et al. Factors associated with COVID-19-related death using OpenSAFELY. Nature 2020 Aug;584(7821):430-436.
Graeme Jones is Professor of Rheumatology and Epidemiology and Head of the Musculoskeletal Unit at the Menzies Institute for Medical Research.
