Discussion in 'Long Covid research' started by Cheshire, Sep 22, 2022.
Twitter thread about this study
Abstract as open text:
"The neurologic manifestations of acute COVID-19 are well characterized, but a comprehensive evaluation of postacute neurologic sequelae at 1 year has not been undertaken.
Here we use the national healthcare databases of the US Department of Veterans Affairs to build a cohort of 154,068 individuals with COVID-19, 5,638,795 contemporary controls and 5,859,621 historical controls; we use inverse probability weighting to balance the cohorts, and estimate risks and burdens of incident neurologic disorders at 12 months following acute SARS-CoV-2 infection.
Our results show that in the postacute phase of COVID-19, there was increased risk of an array of incident neurologic sequelae including ischemic and hemorrhagic stroke, cognition and memory disorders, peripheral nervous system disorders, episodic disorders (for example, migraine and seizures), extrapyramidal and movement disorders, mental health disorders, musculoskeletal disorders, sensory disorders, Guillain–Barré syndrome, and encephalitis or encephalopathy.
We estimated that the hazard ratio of any neurologic sequela was 1.42 (95% confidence intervals 1.38, 1.47) and burden 70.69 (95% confidence intervals 63.54, 78.01) per 1,000 persons at 12 months. The risks and burdens were elevated even in people who did not require hospitalization during acute COVID-19. Limitations include a cohort comprising mostly White males.
Taken together, our results provide evidence of increased risk of long-term neurologic disorders in people who had COVID-19."
Lots in this. My quick view:
1.Hospitalisation and ICU have cumulative impact on risk of serious disease across the board, orders of magnitude over non hospitalised in most disease categories.
2. Being male is not protective (but note the study limitations)
3. Weird stat of the day - having diabetes seems 'protective' in a large number of disease categories !!
Main take - Covid 19 messes people up in a wide number of ways. Neurological sequalae are strongly related to hospitalisation, with only Dystonia, Loss of Taste, Loss of Smell, Somnolence, Guillain Barre syndrome, Encephalitis/Encephalopathy and Transverse Myelitis in non hospitalised patients showing overlaps with hospitalised and ICU patients - and that at a low hazard numbers.
Supplementary Table 15 - hazard ratios for fatigue
(they seemed to do this as a test of the approach, suggesting that it is already known that fatigue is more common after Covid-19. They don't seem to be suggesting that fatigue is a neurological outcome, quite the opposite)
Supplementary table 16 is interesting to look at (indicative excerpt below). They compared outcomes for people vaccinated against the flu on even versus odd days. Obviously, that's a grouping that should show no differences. Despite that, some of the hazard ratios are fairly large e.g. for stroke 1.21; thrombosis 0.88.
But, you have to look at the 95% confidence intervals. They all straddle zero. This is an issue the authors of that recent Wyller paper looking at the connection between various risk factors and post-Covid symptoms in young people seemed to not have come to grips with. If you take enough possible risk factors, even with a very large sample as is the case with this Al-Aly study (>150,000 patients, >5 million controls), you will find some hazard ratios that, (without looking at the confidence interval) seem interesting. When you are messing about with around just 85 controls, as is the case with the Wyller study that was so quick to point to personality issues as the cause of Long Covid, there's a big risk of sampling error.
Warning:the above table is not representing the results the actual study. It is of a nonsense study done to look at the chance of positive findings that are just due to statistical noise
I'm very skeptical of the dysautonomia numbers, though. It's definitely one of the most common outcomes out there and should be higher. Likely most of the "anxiety" out there is actually dysautonomia.
A lot of this is temporary, too, so likely a vast undercount as most of this happens at home. Same for movement disorders. Neurology has so much work ahead, especially in the area currently squatted by psychosomatics/functional/conversion stuff. Seizures, too. Smack in the field where "functional disorders" have taken root.
I wonder whether that relates to patients being on metformin - which is an autophagy promoter.
From Cleaning Crisis? Is Defective Mitochondrial Cleanup Impairing Energy Production in ME/CFS?
Sorry, I may have created confusion, not in@Rvallee, but perhaps in others, by posting that table above. To be clear, that is sort of an associated nonsense study the authors did, just to give an indication of the risk of finding incorrect associations.
Here's an excerpt of the results from the actual study. (Supplementary Table 3)
The hazard ratio for dysautonomia is, for example 1.3 and the confidence interval is pretty tight. So we can be fairly confident that it's a real result, that having had Covid-19 really does increase the risk of "dysautonomia". So I assume the chance of a person who had Covid-19 suffering from diagnosed dysautonomia during the study period is 1.3 times the chances of a control. If the chance of the control suffering from dysautonomia is small, then the chance of a person who had Covid-19 suffering from dysautonomia is also small. In this case the burden of dysautonomia in controls is 5.36 per 1000 people, and the burden of dysautonomia in people who had Covid-19 is 6.96 per 1000 people (roughly 5.36 *1.3).
The suggested incidence of dysautonomia in people in the year following having Covid-19 of 6.96 per 1000 people does seem quite low given what we know about orthostatic intolerance type symptoms, but I don't know how dysautonomia was defined, and it might not be diagnosed very often or quickly, compared to some other conditions.
It's not just diabetes - some other otherwise undesirable health conditions have lower hazard ratios for neurological sequelae of Covid-19. Hypertension, hyperlipidemia, immune dysfunction. In particular look at the risk of cerebro-vascular disorders. People with high cholesterol having a markedly lower risk of cerebro-vascular diseases than people who don't have high cholesterol?
One possibility is that something mathematical is happening in the comparisons. The figures are the risks compared to controls, I think with the specific disease. So, the risk of CVDs in a person without diabetes and no Covid is taken to be 1, and the risk of CVDs in a person without diabetes who has had Covid is 1.74. The risk of CVDs in a person with diabetes and no Covid is also taken to be 1, and the risk of CVDs in a person with diabetes and Covid is 1.43.
But say, (and I'm just making these numbers up) the risk of CVDs in a person without diabetes and no Covid is 2 per 1000.
The risk of CVDs in a person without diabetes and Covid would be 2*1.74= 3.48 per 1000.
If the risk of CVDs in a person with diabetes is much higher, say 20 per 1000,
then the risk of CVDs in person with diabetes and Covid would be 20 *1.43 =28.6 per 1000.
So, having diabetes isn't really protective for CVDs following Covid, it's just that having diabetes is a major risk factor that potentially sort of swamps other factors.
Does that make sense?
Supplementary Table 4 - part
(the risks are compared to the risk of controls)
After the last Al-Aly study, I'm not too inclined to take this one seriously. The patient group is quite unrepresentative (older, unvaccinated men with quite a few health problems, many ending up hospitalized), the method is weak (electronic health records aren't going to tell you much about fatigue or other non-specific issues or even about underdiagnosed problems like dysautonomia), and the stats are perhaps not great (lots of criticisms of this and the other Al-Aly studies, especially with respect to residual confounding, which don't seem to have been addresses at all).
I think we can certainly take this to confirm what we already know: that having had Covid increases the risk of all kinds of negative outcomes. But should we trust the specifics here or apply these numbers to the general population in 2022? Definitely not.
I'm sure that is the right (and very clearly made !) explanation - my comment re: diabetes was just a lazy aside, I intended it as a cautionary: "here's a weird statistical effect that probably doesn't mean anything at all." Apologies all if it looked like I was suggesting there was significance. Anyway it is a good example of how the numbers don't always tell a sound story.
I still think this is a useful study, just with a lot of limitations. But that's how science often is, you have to go with the data that's available not the data you wished you had.
Separate names with a comma.