A study published in BMJ Evidence Based Medicine looks at the analysis of long COVID research and its potential flaws.
Dr Jeremy Rossman, School of Biosciences at the University of Kent and Research-Aid Networks, said:
“Unfortunately, this is a flawed analysis that generates more misconceptions about Long Covid than it addresses. Importantly, this is not an original research paper, nor is it a rigorous systematic review, instead the authors reference a limited number of studies (17 peer-reviewed publications) that are then broadly, and inappropriately, applied to Long Covid research as a whole. There are over 1500 clinical studies, meta-analyses or systematic reviews on Long Covid / PASC that have been published since 2020 (as listed on PubMed). In order to make broad conclusions about a body of research, a thorough investigation of the literature is required, such as would be done in a systematic review. For example, in discussing the “most well-designed studies”, the authors only reference two peer-reviewed publications, with no assessment of the limitations of these studies or why other studies that also include control groups and address sampling bias are excluded from this section. As such I am concerned with the authors selection of studies.
“COVID-19 and Long Covid are new conditions that we are still just beginning to understand. As such, the methodology, including definitions, used to study the disease evolve and are refined over time. This does not mean that the early research is flawed and should be discounted, but rather that our understanding is constantly improving. In addition, research into a new condition that is affecting millions of people during a pandemic will necessitate more expedient and agile methods. Again, this does not mean that the body of research is flawed but rather that each study will have its own limitations, which are usually acknowledged in the study itself. It is true that Long Covid lacks a highly precise clinical definition; however, this is not because previous studies have used flawed definitions, but because this is a new disease and we don’t yet have biomarkers or enough of an understanding of the disease process to develop a more precise causal definition. The current case definitions are the ones that have been the most studied, by the widest audiences and deemed to give the best identification of Long Covid patients. The authors also criticize the lack of a control group in many Long Covid studies. Unfortunately, the authors do not acknowledge the reality of investigating Long Covid in the context of a pandemic where the ideally-designed study may not be possible, as baseline data is often not present, COVID-19 testing was limited, not all people maintain seroconversion after infection, we have no biomarkers or precise knowledge of disease etiology and yet have a pressing need to understand more of this on-going disease that is affecting millions of people. Whilst we do not yet know the exact prevalence of Long Covid in different groups, we are getting closer thanks, in part to many new studies being published with control groups and systematic reviews / meta-analyses that have appropriately reviewed the body of clinical Long Covid research.
“The authors say “ultimately, biomedicine must seek to aid all people who are suffering”, this is true; however, this commentary could be damaging to both patients and Long Covid research and does so without providing any new data or rigorous analysis that would justify their claims. Many Long Covid patients already struggle to obtain care, with symptoms and illness often dismissed. This article risks further minimizing Long Covid which may worsen the ability of patients to obtain care. In addition, this article broadly paints Long Covid research as being flawed, potentially threatening funding into the disease at a time when we desperately need additional funding and researchers to continue leading high-quality, well-controlled research into Long Covid.”
Dr Adam Jacobs, Senior Director, Biostatistical Science, Premier Research, said:
“The paper by Hoeg et al on long covid makes some sensible points, although it then goes on to draw some odd conclusions. The paper provides a tour of some of the methodological flaws in the long COVID literature. It is of course inevitable that much of this literature is imperfect: long COVID simply didn’t exist 4 years ago, so researchers have had to get to grips with a new and challenging topic at top speed. It is therefore not surprising that different studies have different estimates of the prevalence of long COVID, as studies have used different case definitions, different populations, etc. We are still learning about the implications of those things. Many of the suggestions Hoeg et al make for how the literature could be improved are sensible.
“However, they then go on to draw the conclusion that there has been a “distortion of risk” and that well designed studies “have been reassuring”. The uncertainty itself in the long COVID literature is far from reassuring. The fact is there is still much we do not know about long COVID: in part because of many of the limitations of the existing literature that Hoeg et al themselves identify. If we could be sure that some of the most optimistic studies are correct, then perhaps we could be reassured. However, if some of the even moderately pessimistic, let alone the most pessimistic, studies are correct, then long COVID could represent a public health catastrophe the likes of which has not been seen in modern times.
“The ONS long covid survey, now sadly discontinued, counted 1.9 million people with long COVID in the UK as of March 2023, or 2.9% of the population. That is a staggeringly high number. Even if, as a result of limitations in definitions pointed to by Hoeg et al, it overestimated the real number by a factor of 10, we would still have more people living with long COVID than with Parkinson’s disease. It is also noteworthy that the number of people in the UK who are economically inactive because of sickness is currently at a record high. Although long covid is clearly not responsible for all the current long-term sickness, it would be naive to assume that it does not play a significant part.
“We don’t know how long those people will remain ill. If we are lucky, most of them will be healthy again within a year or two. If we are unlucky, a large number of them may have a permanent disability. And perhaps their numbers will be swelled still further as COVID continues to circulate in the population and people are repeatedly infected. Are we willing to just assume that we will be lucky in the face of such a catastrophic threat to population health? The precautionary principle requires that we take this threat very seriously indeed.
“New research is appearing all the time, and we are constantly learning more about the prevalence of long COVID, the various forms it can take, and the mechanisms behind it. In fact just last week another study was published showing that patients hospitalised with COVID were significantly more likely to have multiorgan abnormalities detected on MRI at a median of 5 months after hospital discharge, particularly in the lung, brain and kidneys. That does not strike me as “reassuring”.
“We continue to need more research, and if researchers can learn from the paper by Hoeg et al and make their research more methodologically robust, then that will be a good thing. But the message that we should be “reassured” seems to me to be completely inappropriate.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, Open University, said:
“This study doesn’t present new research. It interprets existing research on long COVID. That’s not in itself an issue, of course – as the authors point out, there’s a lot of inconsistency between research findings, and that leads to problems that need to be understood. I do agree with the main overall points that the authors make, particularly about existing definitions of long COVID and about the lack of control groups for comparison. I also agree with many, though not all, of their recommendations for future epidemiological research on long COVID.
“On the other hand, none of this is really new. Very similar points have been raised by several other epidemiologists* in the past and recently. There are some interpretations in the new paper that I just don’t agree with, and some of the implications are perhaps not obvious, as I’ll explain. In particular, it’s not at all clear to me that the changes that the authors propose will reduce the demand for health care from people who get ill, or whose symptoms get worse, some time after their COVID infection, even if in fact the new symptoms aren’t directly caused by the COVID infection. The people are, after all, still ill – though the proposed changes might (or might not) change how some people are treated medically.
“It’s true, as the authors say, that existing definitions of what’s generally referred to as ‘long COVID’ have been very broad. If you use a broad definition of a condition, then estimates of how many people have the condition will be larger than would be the case with a tighter definition. I’d say it’s inevitable that when a new condition emerges, it takes time to come up with an appropriate definition, and that’s particularly true of the many different symptoms that can arise after an acute COVID infection. Tying the details down does need to be done, and indeed that has been done to some extent for three of the four definitions from international organisations that are shown in Table 1 of the new paper. The definition from the American CDC (Centers for Disease Control and Prevention) stands out from the other three, in that it refers to conditions present 4 weeks or more after the initial infection, while the others use 3 months/12 weeks. It also gives more prominence to the term ‘long COVID’, whereas the other three use terms like ‘Post-COVID-19 condition’ or ‘Post-COVID-19 syndrome’ (and PASC, ‘post-acute sequelae of COVID-19’, is in common use also). But calling it long COVID isn’t going to stop. That term was essentially invented by patients and is well-established in our language now. It might well help doctors and epidemiologists to use terms that sound more scientific, but that won’t change common use.
“The authors recommend that long COVID should be ‘established as a diagnosis of exclusion’. Well, yes, but they don’t point out that, for three of the four definitions they quote, it already is a diagnosis of exclusion. The term means that there isn’t a test that can establish whether what someone is suffering from is long COVID, so that it has to be diagnosed by checking for other possible explanations. If the other possible explanations are ruled out, the person can be said to have long COVID. And all but the CDC definition say explicitly that a diagnosis of long COVID can arise only when the symptoms ‘cannot be explained by an alternative diagnosis’ (or other similar wording). The CDC definition doesn’t say that, and it is quite old now and does probably need to be retired. But to concentrate so much on one dubious definition without pointing out the better points of the other definitions (including the one from WHO and the one used in the UK) seems wrong.
“The issue with control or comparison groups is as follows. The symptoms described by people with long COVID are things that can also arise from many causes other than a COVID infection. Just because someone has ongoing fatigue or long-term ‘brain fog’ after they had COVID doesn’t mean that those symptoms were caused by the COVID infection. It’s possible that they might have had the symptoms even if they had not been infected with the COVID virus. So just counting how many people have symptoms like that after COVID can’t tell you how likely it is that a COVID infection will cause those symptoms.
“It’s difficult or impossible to tell this on an individual basis, since there’s no test for long COVID. But we can get estimates of the numbers involved in a population, by comparing the rate at which people, known to have had COVID, get long-term symptoms, with the rate at which people, known not to have COVID, get the symptoms. This obviously needs a ‘control group’ of people who are known not to have had COVID, and that group needs to be as alike as possible to the group who had COVID, apart from not having had COVID. In most research on long COVID, there has not been a control group. Even when there was one, it wasn’t always matched well enough to the group of people who had COVID, and/or not enough was done to make statistical adjustments to allow for differences between the groups.
“The authors of the new study are right to point out that this is a problem, and they are also right to point out that finding an appropriate control group isn’t at all straightforward. But one very important point, that I think they should have said more about, is as follows. If someone reports a symptom that is causing them problems, after having had COVID, and that symptom wasn’t actually caused by the COVID infection, that doesn’t mean that it needs no medical treatment. It might need a different medical treatment, and knowing more about what did or didn’t cause the symptoms might help in getting the right treatment. But even that would often require more knowledge about what can cause these symptoms, and how, than we currently have.
The most recent Office for National Statistics (ONS) survey results**, referred to in the new paper, estimated that 2.7% of the people living in private households in the UK in the 4 weeks to 5 March 2023 had self-reported long COVID, at least 12 weeks after they first had (or suspected) COVID. (That’s slightly less than the 2.9% mentioned in the new paper, which is for prevalence at least 4 weeks after the initial infection, but it seems more appropriate to use the 12 week figure, given that the NICE definition of post-COVID-19 syndrome refers to at least 12 weeks, not 4.) That’s an estimated 1.7 million people, and of those, about 1.4 million said that their symptoms reduced their ability to undertake day-to-day activities. Of the 1.4 million, 350,000 said that their symptoms reduced their ability to undertake day-to-day activities a lot. That figure of 350,000 is similar to the population of a biggish city like Cardiff.
“There’s no control group for those figures, so it remains quite possible that the symptoms of quite a few of those 1.4 million with reduced ability to do daily activities weren’t directly caused by COVID. We just can’t tell from this data source. But that doesn’t mean those people are well. If their symptoms can be distinguished from those of people whose symptoms were caused by COVID, which often won’t be possible on an individual basis, then maybe the treatment would be different. But they still need treatment. Changing the numbers attributed to long COVID won’t necessarily reduce the demand for treatment, or the cost of treatment, at all. That all depends on what the treatments are.
“A similar point applies about people who got post-ICU syndrome, or shortness of breath following pneumonia, after COVID. Yes, there’s perhaps some sense in saying this wasn’t directly caused by COVID, because these conditions can occur after someone has been in ICU or has had pneumonia for some other reason than COVID. And yes, the nature of their condition needs to be taken into account. But, in many cases at least, they wouldn’t have had pneumonia, or had to spend time in ICU, if they hadn’t had a COVID infection, so I’d argue that those conditions are caused by COVID in a real sense. They certainly need to be taken into account as possible long-term consequences of a COVID infection.
“My final general point is as follows. The authors of the new study seem to want to require that definitions of long COVID must require symptoms to be ‘persistent or continuous’. They say that three of the four definitions that they quote, all but the CDC definition, require this. I think that’s an odd reading of those definitions. The WHO definition does not mention anything about the symptoms being persistent or continuous. The definition from NICE and other groups says that the symptoms “can fluctuate and change over time”, and the Delphi definition says that symptoms “may fluctuate or relapse over time” (and, incidentally, this definition is not for children and young people, despite what Table 1 in the new paper says). None of this sounds like ‘persistent or continuous’ symptoms to me. And it’s hardly a new thing that symptoms caused by infections might not arise for some time after the initial infections. An extreme case is shingles, which is caused by the virus that causes chickenpox. That can stay inactive in the body for years or decades, before emerging again in the form of shingles.
“In any case, the main source for the authors’ claim that studies that don’t use a definition including persistent or continuous symptoms may exaggerate the numbers is what they refer to as a ‘UK study’ (their reference 7). But I suspect that this claim rather misses the point of that study. It was a UK study, in the sense that all but one of the authors are from the University of Oxford. But the data that they used is from the USA, from electronic health records. Nothing wrong with that, but it does impose some limitations and issues of interpretation.
“The authors of the new paper point out, correctly, that in this UK/US study, of all the patients whose health record showed any of the long COVID symptoms, that were included in the study, between 90 and 120 days after their initial infection, 40% did not have any symptom on their record during the first 90 days. The authors of the new paper say that this means these patients would not be included as having long COVID if the definition had insisted on persistent or contiguous symptoms. Well, yes, that’s a possibility, but it’s far from the only one. Because the data come from records of encounters with healthcare systems, it remains quite possible that there were persistent or contiguous symptoms in the first 90 days for these patients, but they did not seek health care intervention for them, so that nothing was recorded. There are several possible reasons for that. First, it seems quite a common reaction to a symptom that’s not too severe for the patient not to seek health care initially, because they feel that the symptom might go away of its own accord, but that they do seek health care later when it has become clear that the symptom is not going away. Second, all the records in question come from 2020, and during much of that year, there were issues in terms of access to health care. That was perhaps particularly true earlier in the year, during and shortly after the initial wave of the disease. So patients may have been unable to access health care in the first 90 days of their symptoms, or at least felt that they should not access it, even if they had continuous symptoms. And then the researchers on the UK/US study rightly list various limitations that generally arise in using data from electronic health records, including the fact that, if a person changes health care provider, only some of their records are available through the source of data that the researchers used. I’m certainly not saying that the UK/US research was a bad study, but its limitations do mean that it doesn’t really bear the weight that the new authors are attributing to it (in my view).”
“This new paper seems to me to show several signs of poor preparation, or perhaps inadequate checking. Some are trivial, but it all adds up, in my view. This does not mean that the overall thrust of the paper is wrong – I don’t think it is. But I’d say that it does mean we need to be careful in assessing the details.
“After referring to the ONS study on prevalence of long COVID, based on the ONS infection survey, that I have already mentioned, the researchers refer favourably to another ONS study. This one compared the symptom prevalence of any of 12 common symptoms of COVID at 12-16 weeks after the initial infection, in a group of people known to have had COVID and in a control group who had not had a positive test for the virus. (The reference given in the new paper for this study is wrong; see below.) What the new authors say about this study is true, but they miss out an important proviso. The 12 symptoms are relatively common, but particularly so in the acute stage of COVID, and they do not include some of the symptoms that are commonly reported for long COVID. So they do include weakness/tiredness, muscle ache, and shortness of breath, but they don’t include any of the aspects of mental health that are commonly reported, including those relating to ‘brain fog’ (difficulty concentrating, memory loss or confusion), worry and anxiety, low mood/not enjoying anything, or trouble sleeping. All of those were more commonly reported n the self-report ONS study than were any of those on the list of 12 in the study with a control group, apart from the three (weakness/tiredness, muscle ache, and shortness of breath) that I’ve already mentioned.
“These omissions aren’t due to ONS incompetence – the list of 12 symptoms was used right from the start of the ONS infection survey (the source for these results) in April 2020, and are based on the symptoms of the acute stage of COVID, because the list was drawn up before long COVID had emerged as an important issue. But that was the only list available from the ONS infection survey, that could be used with a control group, up to the time the most recent version of this study was published (September 2021). A longer list of symptoms had been added to the ONS survey questionnaire by then, with 21 symptoms rather than 12, and later another two symptoms were added. Several commentators*** at the time commented on this, and there were some other methodological issues – this ONS analysis was always said to be ‘experimental’ – so it was not updated after September 2021. Relying on it as a source of good numerical estimates of the difference in prevalence of symptoms, between people known to be infected and known not to be infected, as the new authors do, is simply not justified.
“The ’well-designed Swiss study’ of infection in children, referred to in the section headed “The most well-designed studies provide reassuring estimates”, was indeed well designed, for the reasons stated. But (as the new authors point out) it was very small, and so wasn’t really able to answer the question of interest for the new study. (The reference given in the new paper for this study is again wrong; see below.) The numbers given in the new paper from this study are correct, but it’s odd that (again) the new authors refer to lasting symptoms 4 weeks after initial infection, rather than the 12 weeks given in three of the four definitions in Table 1. The Swiss study did also look at symptoms after 12 weeks, and found that the rate of such symptoms in children who were antibody positive was double the rate in those who were antibody negative (4% compared to 2%). The numbers are so small that this effectively means nothing statistically, but the same really goes for the 4 week figures, which give the opposite impression. One wonders why the new authors chose to give the 4 week figure, and not the 12 week figure as well or instead.
“There are some more trivial points too. The affiliation of the author Shamez Ladhani is given as Public Health England, even though that is an organisation that ceased to exist two years ago (1 October 2021), when its functions were divided between the UK Health Security Agency (UKHSA) and some other bodies. I believe Dr Ladhani currently works for UKHSA (and also at St George’s, University of London). The long COVID definition said to be a “Delphi definition for children and young people” in Table 1 of the paper is actually not restricted to children and young people, according to the reference given (Reference 29). That paper does say that “A separate definition might be applicable for children,”, but it does not give one.
“And finally (and I mention this really only to avoid anyone else having to do what I did and dig around to find what’s really where), the reference numbers given in the section headed “The most well-designed studies provide reassuring estimates” are wrong. The ONS study listed in the text as having reference 21 is really reference 23. The one listed in the text as reference 23 is really 24, and the one listed as reference 24 is really 25. And, in the next paragraph, the Swiss study, said to have reference 25 is in fact reference 22, and that reference is to an overall website for a linked series of projects in Switzerland, where one has to dig out the relevant paper from a list of over 60 publications. The specific paper is actually at https://jamanetwork.com/journals/jama/fullarticle/2782164.”
* Among them the Australian epidemiologist and writer Gideon Meyerowitz-Katz, who blogged about it a year ago at https://gidmk.medium.com/how-many-people-get-long-covid-15d2e433a846 and, coincidentally perhaps, returned to the subject just a couple of days ago, at https://gidmk.medium.com/the-ongoing-risk-of-long-covid-cf2eb8a39a78 and https://gidmk.substack.com/p/the-ongoing-risk-of-long-covid.
** Reference 23 in the new paper, https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/prevalenceofongoingsymptomsfollowingcoronaviruscovid19infectionintheuk/30march2023. These results refer to the position in the four weeks to 5 March 2023. Things will have changed since, but the ONS Infection Survey that provided these estimates has been discontinued.
*** Including me, https://www.sciencemediacentre.org/expert-reaction-to-updated-estimates-of-the-prevalence-of-post-acute-symptoms-among-people-with-coronavirus-covid-19-in-the-uk-26-april-2020-to-1-august-2021/, though that’s hardly the most exciting read.
Prof Paul Garner, Professor and Director of the Centre for Evidence Synthesis in Global Health, Liverpool School of Tropical Medicine (LSTM), said:
“An explosion of studies and meta-analyses estimating prevalence of the post COVID condition based on symptoms has led to a widely held belief that the condition is extremely common, with the CDC stating, “one in five American adults who have had COVID-19 still have long Covid”. A team of excellent epidemiologists have prepared this narrative commentary that takes a broad look at the body of research and has found it wanting.
“The researchers report that most studies are flawed, as studies use definitions that are too broad, lack control groups and have a range of other biases. On top of this, the systematic reviews arising from these studies inherit these flaws and compound these inflated estimates. These biases are leading to overestimates of the condition generally. The authors point out this increases social anxiety about the condition and distorts healthcare spending.
“This overestimate has policy significance as there is an emerging hypothesis that societal anxiety and fear around COVID-19 may contribute to the persistence of symptoms associated with the post-covid condition. Thus, the spectre of so many people with a long term illness that is unexplained could itself create fear, generate catastrophic thought and unwittingly may potentially influence the meaning patients attribute to symptoms they experience.
“The authors highlight that one recent systematic review showing only 11% of the studies had a control group. Indeed, some of the smaller more carefully conducted studies with proper controls reveal important paradoxes: for example, a study from Norway showed that about 49% of adolescents in the cohort who had been infected with COVID-19 met the criteria for WHO’s definition for the post covid condition at six months, but 47% in the control group also met the WHO definition-yet had clear serological evidence that they had never been infected. Such findings need explanation, and may demonstrate that the cause of the symptoms that make up the post covid condition may result from more complex human biological responses that include both the virus and the broader impact of the pandemic on our society.
“Rapidly conducted scientific research in COVID-19 was needed and has had some real benefits. This paper highlights the importance of avoiding “COVID-19 exceptionalism”: we must apply the same critical appraisal to the research done before reaching conclusions on estimates and on causality. This is particularly important now in unpacking the symptoms associated with the post covid condition.”
Dr Gavin Stewart, Senior lecturer in Evidence Synthesis, Newcastle University, said:
“This opinion piece highlights many problems and potential biases with the evidence-base underpinning long Covid. However, in the absence of evidence synthesis, the magnitude of the biases remains largely unknown. The authors are right to highlight uncertainty in estimates of long Covid prevalence but the conclusion that effects are over-estimated may prove premature.”
Prof Danny Altmann, Professor of Immunology, Imperial College London, said:
“This paper would have benefited from a more comprehensive analysis of the extensive Long Covid literature with respect to epidemiology and mechanism. The current consensus dataset makes a robust case that, in all parts of the world where reporting is available, infection with SARS-CoV-2 is associated in around 10% of cases with variably disabling, persistent symptoms (see, for example Altmann et al., Nature Rev Immunol 2023; Davis et al Nature Microbiol 2023; Marjenberg et al., Sci Rep 2023). This is unsurprising to those with a track record in the field as it is reminiscent of related post-viral sequelae, including from SARS, MERS, Chikungunya and Ebola (Choutka et al., Nature Med 2022). Far from being derived from uncontrolled or self-reported studies, much of the strongest data on Long Covid come from well-controlled analysis of huge electronic healthcare record datasets, including the diverse consequences for persistent disease evidenced by the US Veteran Affairs datasets (see refs 1-7 below), comparing records of individuals with and without Covid-19 infection. Also important in this context are the strong serum biomarker correlations with long term symptoms in large cohort studies such as PHOSP-COVID (F Liew et al., (Imperial College) In press, Nature Immunology: ‘Large scale phenotyping of long COVID inflammation reveals mechanistic subtypes of disease’); M Taquet et al., Nature Medicine) or indeed, the correlations between COVID-19, CNS MRI changes and before-and-after infection neurocognitive scoring changes in the large UK Biobank dataset (Douaud et al., Nature 2022). Note that all of these studies have the common feature of publication in the very highest ranked, peer-reviewed journals. There is a massive global need to supply clinical answers for the many millions affected by Long Covid. Long Covid, like all aspects of COVID-19, has disproportionately affected the most socioeconomically deprived. As one might expect, there has been a disproportionate impact also on those frontline workers who gave the most to see us through the pandemic”.
1: Bowe B, Xie Y, Al-Aly Z. Postacute sequelae of COVID-19 at 2 years. Nat Med.
2023 Sep;29(9):2347-2357. doi: 10.1038/s41591-023-02521-2. Epub 2023 Aug 21.
PMID: 37605079; PMCID: PMC10504070.
2: Bowe B, Xie Y, Al-Aly Z. Acute and postacute sequelae associated with SARS-
CoV-2 reinfection. Nat Med. 2022 Nov;28(11):2398-2405. doi:
10.1038/s41591-022-02051-3. Epub 2022 Nov 10. PMID: 36357676; PMCID: PMC9671810.
3: Al-Aly Z, Bowe B, Xie Y. Long COVID after breakthrough SARS-CoV-2 infection.
Nat Med. 2022 Jul;28(7):1461-1467. doi: 10.1038/s41591-022-01840-0. Epub 2022
May 25. PMID: 35614233; PMCID: PMC9307472.
4: Xie Y, Xu E, Bowe B, Al-Aly Z. Long-term cardiovascular outcomes of COVID-19.
Nat Med. 2022 Mar;28(3):583-590. doi: 10.1038/s41591-022-01689-3. Epub 2022 Feb
5: Xie Y, Bowe B, Al-Aly Z. Burdens of post-acute sequelae of COVID-19 by
severity of acute infection, demographics and health status. Nat Commun. 2021
Nov 12;12(1):6571. doi: 10.1038/s41467-021-26513-3. PMID: 34772922; PMCID:
6: Bowe B, Xie Y, Xu E, Al-Aly Z. Kidney Outcomes in Long COVID. J Am Soc
Nephrol. 2021 Nov;32(11):2851-2862. doi: 10.1681/ASN.2021060734. Epub 2021 Sep
7: Al-Aly Z, Xie Y, Bowe B. High-dimensional characterization of post-acute
sequelae of COVID-19. Nature. 2021 Jun;594(7862):259-264. doi:
10.1038/s41586-021-03553-9. Epub 2021 Apr 22. PMID: 33887749.”
‘How methodological pitfalls have created widespread misunderstanding about long COVID’ by name of first author et al. was published in BMJ Evidence Based Medicine at 23:30 hours UK time Monday 25 September 2023.
Dr Gavin Stewart: No known conflicts of interest
Prof Danny Altmann: DMA is co-author of the Penguin Long Covid Handbook and Lead Investigator of the NIHR Long Covid WILCO Study and has received honoraria for consultancies with Oxford Immunotec, Pfizer, Novavax, AstraZeneca and Shionogi.
Prof Paul Garner: PG is a medical epidemiologist who has worked in evidence synthesis in infectious disease for thirty years, with personal experience of the post-covid syndrome, and recovered.
Dr Adam Jacobs: I have no relevant competing interests.
Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee. My quote above is in my capacity as an independent professional statistician.”
Dr Jeremy Rossman: Dr Rossman declares that he is the Chair of the Board of Trustees of the charity Long Covid Kids.