Long COVID: A Narrative Review and Meta-Analysis of Individual Symptom Frequencies

Round 1

Reviewer 1 Report

The work presented here addresses a very relevant issue for the healthcare system. The prevalence rate for the various symptoms of Long-Covid. The literature research carried out and the mix of EHR, surveys and special Long-Covid care data make the analysis very valuable.

The centrepiece of the work is the different prevalence data for the various (21) symptoms investigated, depending on the basic scientific survey. Due to the heterogeneity of Long-Covid symptomatology, the authors implicitly assume very different pathogenetic causes and therefore ask (line 439 - 441): ‘What is the prevalence of symptoms in participants who have long-term symptoms after SARS-CoV-2 infection?’ and not ‘What is the prevalence in a population infected with SARS-CoV-2’.

This results in at least 2 very different ways of thinking:

1. the (21) different long-Covid symptoms are caused by different pathomechanisms and the question arises for each symptom individually as to their prevalence occurrence after SARS-CoV-2 infection for a population population.

2. long-Covid (more precisely, this applies to post-Covid) represents a nosological entity in which one or a few pathomechanisms can cause a variety of symptoms. (An example of this is the current hypothesis, where the measurement of the bioactivity of GPCR autoantibodies as an autoimmune process leads to the various agonistic adrenergic autoantibodies (e.g. alpha1-, beta 1 and 2 -, M2-, ACE2-, ET autoantibodies, etc.), which also affect the different organ and tissue systems differently depending on the receptor set. One pathomechanism leads to a great heterogeneity of symptoms, but represents a nosology.

This results in very different models for the research question, as well as for the consideration and calculation of bias.

In the assumptions and the entire discussion up to questions line 439-441, it becomes clear that this is set as a basic assumption and that the authors favour this assumption. 

This may and can of course be done, but then this should also be done critically in the justification of the assumptions in the text and the discussion.

If version 2 is favoured, this leads to different conclusions. The symptoms do not occur randomly in different people, but represent a heterogeneity of symptomatology, but are a nosology (autoimmune process with formation of different GPCR-AAK as a reaction to the spike protein with binding to the ACE2 receptor). Question 2 would now be relevant here. What is the prevalence of Long-Covid syndrome and what is the incidence of the various 21 symptoms? Mathematically, there are connected samples in model 2 instead of unconnected samples as in model 1.

As clinicians in post-Covid centres, symptom heterogeneity and the respective severity of symptoms is a secondary characteristic (and correlates to GPCR autoantibodies in bioassays). A further complicating factor is that the spike protein is the antigen for the formation of GPCR-AAK and therefore virus persistence and spike persistence are also associated variables and not unrelated variables.

As there is (still) no certainty about these pathomechanisms, these two basic variants should be separated in the bias assessment and discussion.

In the 2nd model, the question of the occurrence of Long-Covid/Post-Covid in the population would therefore be decisive for the clinician and, secondarily, which symptom heterogeneity and severity occur (Post-Covid).

This also implies the evaluation of bias assumptions and calculations. This also explains the result found in the paper that the highest prevalence of the various 21 symptoms occurred in long-Covid studies. Doctors with good knowledge are best advised to enquire about the various Long Covid symptoms and document them as nosology. Untrained physicians (EHR data) at best record the patient's symptoms, but do not inquire further about the typical symptoms that occur and there is a risk of underreporting.

These basic considerations should only serve to enable the comments made under Detailed comments to be processed accordingly. From the reviewer's point of view, the discussion should transparently explain the implied assumptions and justify in each case why the assumptions were made or deal critically with both models.

Line 2: Instead of ‘Post-Acute Sequelae of SARS-CoV2’, I recommend using the term ‘Long Covid’ in the title and text, even if one study formally examined patients after 3 weeks. The different designations of Long-Covid and Post-Covid already cause enough confusion, so that Post-Acute Sequelae of SARS-CoV2 creates further confusion.

 

Line 15: Explain the abbreviation EHR at the beginning. EHR (= electronic healthcare record)

Line 25/26: Not only laypeople speak of Long Covid syndrome. The British National Institute for Health and Care Excellence (NICE) definition of ‘long COVID’ as health complaints that persist beyond the acute phase of a SARS-CoV-2 infection of 4 weeks or even occur again could be used here. According to NICE, ‘post-COVID syndrome’ refers to symptoms that are still present more than 12 weeks after the onset of SARS-CoV-2 infection and cannot be explained in any other way. ‘Long COVID’ therefore includes both symptoms that persist 4 to 12 weeks after the onset of symptoms following an acute COVID-19 illness and “post-COVID-19 syndrome”.

Line 27: The number of 704 million SARS-CoV-2 infections worldwide by April 2024 according to (Our world in Data) is viewed very differently in terms of its value, as many child infections and countries have not collected reliable data. It can therefore be omitted at this point.

Line 36: Here, multisystem disease is explicitly mentioned as possibly having only one underlying aetiology, which is not taken into account later in the discussion. See the comment under Major comments.

Line 89: It is written here that studies with experimental therapy were excluded. This needs to be explained or at least specified, as all therapies were experimental at the beginning of the pandemic.

Line 124 and 125: Hair loss is listed twice here.

Line 289-291: Here it must be explained why the long-covid studies are categorised as high-risk with regard to bias and, above all, with regard to which bias. Since the Hoy et al criteria are applied, the list of Hoy criteria should be given from a to i for the study types. Unfortunately, the supplement figures A - D were not available for assessment for the review. 

Line 342: 'with suspected Long COVID that do not reflect general population 342 prevalence ' 

What is the assumption that Long-Covid Population does not reflect the general population in its prevalence? 

If Long-Covid is a nosology, it is the primary number of interest as prevalence and symptom heterogeneity is secondary. If the symptoms relate to completely different aetiologies or pathomechanisms and are independent variables, then the statement is correct.

Line 395: In addition to virus persistence, please also list spike protein persistence alone.

Line 439 - 441: See comment under Major comments.

Line 450 ff: The authors' contributions still need to be filled in.

 

 

 

 

 

 

Author Response

1. The authors write ‘ Post-Acute Sequelae of SARS-CoV2: “. Medical terminology uses the terms Long-Covid and Post Covid, with the UK National Institute for Health and Care Excellence (NICE) guideline recommendations defining ”Long-COVID’ as health complaints that persist or reappear beyond the acute disease phase of a SARS-CoV-2 infection of 4 weeks. Post-Covid are symptoms lasting longer than 12 weeks after infection. Even if the work in a publication includes symptoms as early as 3 weeks after infection, it should be written as Long-Covid.

Response 1: We understand the need for clarity and consistency and have changed the term to Long COVID.

2. A substantial part of the work analyses the prevalence frequency using different scientific sources (EHR, surveys and long-covid studies) and attempts to discuss the different prevalence results by reducing systematic bias. A more detailed discussion of the assumptions for bias reduction and, in some cases, further justification is required. This is explained in detail in the Detail Comments.

Response 2: Please see the comments below.

3. The manuscript refers to Supplementary Figures (A to D), which are not attached to the manuscript and could not be assessed by additional download. These should be made available for the review process. The illustrations in the manuscript are very good.

Response 3: We apologize as we were unaware there was an issue uploading supplementary figures. They are now ready for your review (Supplementary Figures A to D).

Major Comments

4. The work presented here addresses a very relevant issue for the healthcare system. The prevalence rate for the various symptoms of Long-Covid. The literature research carried out and the mix of EHR, surveys and special Long-Covid care data make the analysis very valuable. The centrepiece of the work is the different prevalence data for the various (21) symptoms investigated, depending on the basic scientific survey. Due to the heterogeneity of Long-Covid symptomatology, the authors implicitly assume very different pathogenetic causes and therefore ask (line 439 - 441): ‘What is the prevalence of symptoms in participants who have long-term symptoms after SARS-CoV-2 infection?’ and not ‘What is the prevalence in a population infected with SARS-CoV-2’. This results in at least 2 very different ways of thinking:
   1. the (21) different long-Covid symptoms are caused by different pathomechanisms and the question arises for each symptom individually as to their prevalence occurrence after SARS-CoV-2 infection for a population population.
   2. long-Covid (more precisely, this applies to post-Covid) represents a nosological entity in which one or a few pathomechanisms can cause a variety of symptoms. (An example of this is the current hypothesis, where the measurement of the bioactivity of GPCR autoantibodies as an autoimmune process leads to the various agonistic adrenergic autoantibodies (e.g. alpha1-, beta 1 and 2 -, M2-, ACE2-, ET autoantibodies, etc.), which also affect the different organ and tissue systems differently depending on the receptor set. One pathomechanism leads to a great heterogeneity of symptoms, but represents a nosology.

This results in very different models for the research question, as well as for the consideration and calculation of bias. In the assumptions and the entire discussion up to questions line 439-441, it becomes clear that this is set as a basic assumption and that the authors favour this assumption. This may and can of course be done, but then this should also be done critically in the justification of the assumptions in the text and the discussion. If version 2 is favoured, this leads to different conclusions. The symptoms do not occur randomly in different people, but represent a heterogeneity of symptomatology, but are a nosology (autoimmune process with formation of different GPCR-AAK as a reaction to the spike protein with binding to the ACE2 receptor). Question 2 would now be relevant here. What is the prevalence of Long-Covid syndrome and what is the incidence of the various 21 symptoms? Mathematically, there are connected samples in model 2 instead of unconnected samples as in model 1. As clinicians in post-Covid centres, symptom heterogeneity and the respective severity of symptoms is a secondary characteristic (and correlates to GPCR autoantibodies in bioassays). A further complicating factor is that the spike protein is the antigen for the formation of GPCR-AAK and therefore virus persistence and spike persistence are also associated variables and not unrelated variables. As there is (still) no certainty about these pathomechanisms, these two basic variants should be separated in the bias assessment and discussion. In the 2nd model, the question of the occurrence of Long-Covid/Post-Covid in the population would therefore be decisive for the clinician and, secondarily, which symptom heterogeneity and severity occur (Post-Covid). This also implies the evaluation of bias assumptions and calculations. This also explains the result found in the paper that the highest prevalence of the various 21 symptoms occurred in long-Covid studies. Doctors with good knowledge are best advised to enquire about the various Long Covid symptoms and document them as nosology. Untrained physicians (EHR data) at best record the patient's symptoms, but do not inquire further about the typical symptoms that occur and there is a risk of underreporting. These basic considerations should only serve to enable the comments made under Detailed comments to be processed accordingly. From the reviewer's point of view, the discussion should transparently explain the implied assumptions and justify in each case why the assumptions were made or deal critically with both models.

Response 4: This is a very interesting and important comment from the reviewer. However, it is beyond the scope and goals of this current review article. We did not intend to demonstrate pathobiology of Long COVID or comment on whether studies support any one model of pathogenesis, or whether studies support a nosology of clusters of symptoms. In the introduction, we simply state: 

It is likely that Long COVID is a heterogeneous set of syndromes, potentially with variable etiopathogenesis(e.g., initial organ injury from acute infection, viral persistence, immune dysregulation, autoimmunity, unrepaired tissue damage, dysbiosis of microbiome or virome).^6,7

The primary goal of this review was to summarize data regarding the prevalence of Long COVID symptoms reported from studies using varied study design, and to demonstrate differences in prevalence rates according to study design. 

Please Note: Line numbers refer to revised manuscript WITH tracked changes.

The introduction has been modified to make this goal clear (Lines 29-33):

‘COVID has been reported in multiple studies from throughout the world, most of which have focused on individual symptom frequency; the primary goal of this narrative review is to summarize current literature on this topic focusing on reported prevalence of individual symptoms among the general population infected with SARS-CoV2 prior to the omicron wave.’

We would like to clarify the intent of the text in Lines 439 – 441 as it seems to have caused confusion. This text refers to the study design differences between studies of Long COVID participants 1: ‘What is the prevalence of symptoms in participants who have long-term symptoms after SARS-CoV-2 infection?’ vs. studies of 2: ‘What is the prevalence in a population infected with SARS-CoV-2’.  We have edited the manuscript to reflect the main study goal and the limitations of studies that included only participants with Long COVID due to selection bias.

 The key difference between studies that include only participants who had reported Long COVID, is in the denominator for the prevalence calculation.  When all participants already have symptoms, the prevalence is calculated among a biased group.  In studies that we classified as “survey” studies, the symptoms were ascertained among participants with and without Long COVID symptoms, hence the denominator would be a group of participants that includes participants without Long COVID symptoms. We thus interpret the survey studies as those amongst a general population, and the Long COVID studies, as potentially biased by selection bias (selecting participants who have symptoms).

We revised the Discussion to make this comparison clearer (Line 222-229):

‘Studies that focused on populations of reporting symptoms after SARS-CoV2 infection rather than focused on anyone infected in a population included: four studies that conducted surveys using social media and online outreach to support groups for individuals with post-COVID symptoms^12,24,30,33; two studies that surveyed patients visiting local clinics for Long COVID symptoms.^42,47; and one study that relied on EHR data and used the Long COVID International Classifications of Disease, 10^th Edition (ICD-10) code to identify the cohort.⁴⁸

Minor Comments

5. Line 2: Instead of ‘Post-Acute Sequelae of SARS-CoV2’, I recommend using the term ‘Long Covid’ in the title and text, even if one study formally examined patients after 3 weeks. The different designations of Long-Covid and Post-Covid already cause enough confusion, so that Post-Acute Sequelae of SARS-CoV2 creates further confusion.

Response 5: Thank you for this feedback; your point is well taken. The title and several instances of Post-Acute Sequelae of SARS-CoV2 have been changed to “Long COVID”. The Post-Acute Sequelae description remains only when describing the RECOVER initiative, as it is the phrase the study uses in publications (Thaweethai T, Jolley SE, Karlson EW, et al. Development of a Definition of Postacute Sequelae of SARS-CoV-2 Infection. JAMA 2023;329(22):1934-1946. DOI: 10.1001/jama.2023.8823).

6. Line 15: Explain the abbreviation EHR at the beginning. EHR (= electronic healthcare record)

Response 6: Done.

7. Line 25/26: Not only laypeople speak of Long Covid syndrome. The British National Institute for Health and Care Excellence (NICE) definition of ‘long COVID’ as health complaints that persist beyond the acute phase of a SARS-CoV-2 infection of 4 weeks or even occur again could be used here. According to NICE, ‘post-COVID syndrome’ refers to symptoms that are still present more than 12 weeks after the onset of SARS-CoV-2 infection and cannot be explained in any other way. ‘Long COVID’ therefore includes both symptoms that persist 4 to 12 weeks after the onset of symptoms following an acute COVID-19 illness and “post-COVID-19 syndrome”.

Response 7: Thank you for pointing this out; this line in the introduction and the manuscript throughout have been amended. We agree “Long COVID” is more appropriate as per the recent NASEM recommendations for terminology (National Academies of Sciences, Engineering, and Medicine. 2024. A Long COVID Definition: A Chronic, Systemic Disease State with Profound Consequences. Washington, DC: The National Academies Press. https://doi.org/10.17226/27768).

8. Line 27: The number of 704 million SARS-CoV-2 infections worldwide by April 2024 according to (Our world in Data) is viewed very differently in terms of its value, as many child infections and countries have not collected reliable data. It can therefore be omitted at this point.

Response 8: Done.

9. Line 36: Here, multisystem disease is explicitly mentioned as possibly having only one underlying aetiology, which is not taken into account later in the discussion. See the comment under Major comments.

Response 9:  To reduce confusion, we have removed this text (Lines 38-39): ‘However, PASC could be a multi-system disease with a single underlying etiology.’

10. Line 89: It is written here that studies with experimental therapy were excluded. This needs to be explained or at least specified, as all therapies were experimental at the beginning of the pandemic.

Response 10: This is a good point, and the description has been re-written for clarity. Specifically, any study for which treatment was the focus was excluded (Line 95).

11. Line 124 and 125: Hair loss is listed twice here.

Response 11: Corrected, thank you.

12. Line 289-291: Here it must be explained why the long-covid studies are categorised as high-risk with regard to bias and, above all, with regard to which bias. Since the et al criteria are applied, the list of Hoy criteria should be given from a to i for the study types. Unfortunately, the supplement figures A - D were not available for assessment for the review. 

Response 12: The Hoy criteria were used to rate studies based on the following criteria (Lines 151-159):

 ‘The included studies were assessed for risk of bias using the Hoy et al.²⁸ critical appraisal tool for prevalence studies. The critical appraisal checklist for studies reporting prevalence consists of nine topics: (a) target population representativeness (b) sample frame suitability, (c) sampling method appropriateness, (d) likelihood of non-response bias (e) direct data collection from subjects, (f) usage of valid methods for identification of the condition, (g) use of valid, reliable study instrument to measure parameter of interest (h) same mode of data collection for all subjects, and (i) adequate response rate. Each study was assessed across each of these areas, with results reported as Yes, No, or Unclear. Studies were assigned an overall score, reflecting the number of questions with a Yes response.’

Long COVID studies were rated as not representative of a population infected with COVID due to selection bias from limiting analyses to only participants with Long COVID symptoms (low scores for items a and c). Survey studies that included studying a large group of infected participants from a large geographic region were rated highly on items a and c, because the studies had less selection bias.

We added a clearer explanation (Lines 300-303):

‘Since the primary goal of this meta-analysis was to summarize the prevalence of Long COVID symptoms among the general population, all Long COVID studies were classified as high risk of bias due to limiting data collection to those with symptoms and not the general population’; 

And in the Discussion (Lines 353-356):

‘The challenges with varied study design include potential biases resulting in under-ascertainment of symptoms in EHR studies to including symptom frequencies only among individuals with suspected Long COVID that do not reflect general population prevalence’;

And (Lines 380-384): ‘Surveys among individuals seeking resources for Long COVID, such as in specialty clinics or social media support groups, are most likely to exhibit selection bias.’

Supplemental Figure D is also ready for your review.

13. Line 342: 'with suspected Long COVID that do not reflect general population prevalence' What is the assumption that Long-Covid Population does not reflect the general population in its prevalence? 

Response 13: Similarly to the above, Long COVID clinics have a selected group of patients who are seeking care for post-COVID symptoms and perhaps greater severity of symptoms than the general population of people infected with COVID.

14. If Long-Covid is a nosology, it is the primary number of interest as prevalence and symptom heterogeneity is secondary. If the symptoms relate to completely different aetiologies or pathomechanisms and are independent variables, then the statement is correct.

Response 14: As above, the primary goal of this study was to assess the prevalence of each symptom among populations of infected participants. As few studies performed cluster analyses, we did not include clusters of symptoms in this review. Therefore, we cannot comment on a nosology for Long COVID.

15. Line 395: In addition to virus persistence, please also list spike protein persistence alone.

Response 15: Done, see Line 409.

16. Line 450: The authors' contributions still need to be filled in.

Response 16: Done.

Author Response File: Author Response.docx

Reviewer 2 Report

The main question of the “Safety of Nintedanib in a patient with chronic pulmonary fibrosis and kidney disease” paper by R.Atchley-Challenner, Z.Strasser, D.Pant, A.Krishnamoorthy, L.B.Chibnik and E.W.Karlson was to identify and synthesize symptoms of post-acute sequela of SARS-CoV-2 (PASC) according to published scientific articles and to define PASC phenotypes.

The importance of this topic is due to the fact that about 10% of patients survived the acute COVID-19 are suffering from long COVID symptoms for several months and currently, there are no definite understanding of managing such the patients.

The present review provides the basis for future research to more effectively cluster symptoms and better recognition of its pathophysiology.

No additional remarks. 

The manuscript can be published in its current view.

Comments for author File: Comments.pdf

Author Response

Thank you to this reviewer for their kind endorsement. There were no suggestions to review.

Round 2

Reviewer 1 Report

The comments of the initial review were largely implemented by the authors, so I am in favour of accepting the article.

No further comments.