Question formulation

Two clinical questions were formulated: in adults with a history of SARS-CoV-2 infection, what is the prevalence of long COVID-19 and the frequency of long COVID-19 symptoms? And, in adults with a history of SARS-CoV-2 infection, what are the factors associated with the presentation of long COVID-19?

Search and selection of evidence

The bibliographic search was carried out in MEDLINE (via PubMed), through a search strategy that included free terms and controlled language descriptors for long COVID-19 (Supplementary material 1: search strategy), the search was carried out until April 7th, 2022. The selection of the articles was carried out individually by the authors, considering an initial phase of reading the titles and abstracts through the Rayyan platform (www.rayyan.ai) and a phase of reading the full text of the potentially relevant publications to determine their eligibility. The inclusion criteria were:

1) systematic reviews and meta-analyses of cohort, case-control, or cross-sectional studies that report results for outcomes of interest, evaluated at least three months from the onset of COVID-19 in adults; 2) reviews published in English and Spanish. If more than one systematic review was identified, the one with the best methodological quality was chosen. The following were excluded: 1) systematic reviews that had not assessed the risk of bias or the methodologic quality of the included studies; 2) systematic reviews focused on determining the prevalence of symptoms of a single organ or system and 3) letters to the editor, narrative reviews, preclinical studies (studies in vitro or animal models), and opinion articles.

Data extraction

The data was extracted in a standardized form that included the following information: author, year of publication, number of studies included in the review, number of participants, design, place, population characteristics, prevalence outcomes, associated factors, and assessment tool. risk of bias assessment and AMSTAR 2 score.

Assessment of methodological quality and risk of bias

The AMSTAR 2 tool ¹⁶⁾ was used to assess the methodological quality of identified systematic reviews. The evaluation was carried out by four authors (FHR, DG, DR, MCR) in a paired and independent manner and discrepancies were resolved by consensus. The risk of bias in the included studies was considered from the reviews with the best AMSTAR 2 score.

Characteristics of the studies

The study by Chen et al. ⁹ was a systematic review and meta-analysis that aimed to examine the prevalence of post-acute sequelae of COVID-19 around the world⁽⁹⁾. The databases consulted were PubMed, Embase, and iSearch for studies without peer review (preprints) of medRxiv, bioRxiv, Social Science Research Network (SSRN), and others (search performed on July 5, 2021, with an extension to March 13 of 2022). Studies in English that evaluated long COVID-19, defined as conditions that persist for at least 28 days after diagnosis or recovery from SARS-CoV-2 infection, were selected. The outcomes evaluated were prevalence, risk factors, duration, or associated symptoms.

The search identified 5125 studies, from which 40 were included for qualitative synthesis, and 33 for meta-analysis (number of participants: 886,388). The selected studies were of the prospective cohort type (n = 23), retrospective cohort (n = 6), bidirectional cohort (n = 3), and cross-sectional studies (n = 8).

The risk of bias assessment was carried out using the Joanna Briggs Institute tool for studies with prevalence results, for which they added the total responses classified as "Yes" concerning 9 questions of the tool (score from 0 to 9) where Aspects such as the representativeness of the population included (sampling framework, sampling), the size of the sample, the description of the population and the environment, the methods to evaluate the study condition, the statistical analysis and the percentage of response of the participants¹⁷. After updating the search to March 13, 2022, an additional 10 studies were included for qualitative synthesis, of which 8 were included in the meta-analysis. The selected studies were of the prospective cohort type (n = 5), retrospective cohort (n = 1), bidirectional cohort (n = 3), and cross-sectional studies (n = 1).

The total number of participants after the new systematic search was 1,680,003. The characteristics of the participants were: non-hospitalized participants (4,165 of 5 studies), hospitalized participants (67,161 of 22 studies), and any COVID-19 positive patient, regardless of their hospitalization status (1,608,677 of 23 studies). According to the risk of bias assessment with the Joanna Briggs Institute tool, the most frequent methodological limitations were: sampling was not adequate (16 of 50 studies), a valid method was not used to identify the study condition (15 of 50 studies), the sampling frame was not appropriate to address the target population (7 of 50 studies) and the data analysis was not performed with sufficient coverage of the identified sample (7 of 50 studies).

The study by Michelen et al. ⁸ aimed to synthesize the evidence on the characteristics of long-lasting COVID-19⁽⁸⁾. The study design was of the living systematic review type. The authors performed the systematic search in the MEDLINE, CINAHL, Global Health (Ovid), WHO Global Research on COVID-19, LitCOVID, and Google Scholar databases (search period from January 1, 2002, to March 17, 2021). The outcomes evaluated were the prevalence of signs and symptoms and associated factors. A total of 39 studies were selected (cohort studies: 32; cross-sectional studies: 6; case-control studies: one).

The total number of participants was 10,951 (48% women) from 12 countries. The main finding was the estimation of the prevalence of symptoms of long COVID, in addition, a qualitative synthesis was carried out on diagnostic imaging (13 of 39; diagnostic methods included tomography, ultrasound, and artificial intelligence) and functional tests (10 of 39; methods included spirometry, diffusing capacity, lung volume, and exercise tests). The risk of bias was evaluated with the instrument developed by Hoy et al.¹⁸, which is a validated tool for bias assessment in prevalence studies. The studies had a low risk of bias (4/39), moderate risk (23/39), and high risk (12/39). The domains that presented more studies with a high risk of bias were: 1) representation of the national population (21 of 39 studies), 2) true sampling frame or close representation of the target population (24 of 39) and 3) random selection used for sample selection (32 of 39 studies). The study by Maglietta et al. ⁽¹² aimed to identify, in patients who had been hospitalized for COVID-19, which factors were already present or emerging during hospitalization, It was associated with a higher risk of presenting new or persistent symptoms¹². The study design was a systematic review and a bibliographic search was carried out in 2 databases (MEDLINE and Web of Science) until September 20, 2021, including observational studies in English, with 12 weeks or more of prospective follow-up. Odds ratios were estimated for each assessed factor using unadjusted data. They also performed a random effects meta-analysis, using the Paule and Mandel method for estimating variance between studies^19,20. The confidence intervals for the global effect of the factors of interest were adjusted by applying the Hartung-Knapp-Sidik-Jonkman (HKSJ) approach, which takes into account the uncertainty in the variance of the estimates²¹.

Risk of bias assessment was performed using the QUIPS tool²². This review provides evidence based on 20 observational studies and association measures for the factors of gender and severity of the initial condition concerning outcomes such as any symptoms, respiratory symptoms, mental health symptoms, and fatigue. Most studies (11 of 20) were rated as high risk of bias in at least 1 domain of the QUIPS tool and included: loss of participants to follow up (10 studies), the study sample was not representative of the population of interest (4 studies), limitations in statistical analysis and reporting of results (3 studies), and potential confounders not adequately addressed (1 study). The remaining studies (9 of 20) were at moderate risk of bias.

Prevalence of Long COVID-19

According to Chen et al. (2022), the overall prevalence of post-acute COVID-19 syndrome was 0.43 (95% CI: 0.39 - 0.46) (I2 = 100%; p < 0.001). The authors stratied the analysis according to sex, region, hospitalization, and follow-up time. According to sex, the prevalence of PACS in men was 0.37 (95% CI: 0.24 - 0.51) and in women, 0.49 (95% CI: 0.35 - 0.63); while, by region, the highest prevalence was found in Asia (0.51 [95% CI: 0.37 - 0.65]), followed by Europe (0.44 [95% CI: 0.32 - 0.56]) andtheUnitedStates(0.31[95%CI:0.21-0.43]); Accordingtothehistoryofhospitalization,the prevalence in hospitalized the COVID-19 patients was 0.54 (95% CI: 0.44 - 0.63), in the non-hospitalized group it was 0.34 (95% CI: 0.25 - 0.36), and in the mixed group between hospitalized and non-hospitalized was 0.33 (95% CI: 0.29 - 0.37).

According to the follow-up time, the prevalence of PACS after 90 days was 0.32 (95% CI: 0.14 - 0.57), while the prevalence after 120 days was 0.49 (95% CI: 0.40 - 0.59).

Prevalence of long-term clinical manifestations of COVID-19

According to Chen et al. ⁹ , the most frequent clinical manifestations of PACS were: fatigue (prevalence: 0.23 [95% CI: 0.17 - 0.30]), memory problems (0.14 [95% CI: 0.10 - 0.19]), dyspnea (0.13 [95% CI: 0.11 - 0.15]), insomnia (0.11 [95% CI: 0.05 - 0.23]), and joint pain (0.10 [95% CI: 0.04 - 0.22]). On the other hand, according to Michelen et al. ⁸, the most frequent clinical manifestations were weakness (prevalence: 41% [95% CI: 25.43 - 59.01]), malaise (33% [95% CI: 14.91 - 57.36]), fatigue (31% [95% CI: 23.91 - 39.03]), altered concentration (26% [95% CI: 20.96 - 31.73]), and shortness of breath (25% [95% CI: 17.86 - 33.97]). The list of signs and symptoms is described inTable 2.

A stratification was performed according to hospitalization status for COVID-19 and it was found that the prevalence of clinical manifestations was significantly higher in hospitalized patients, compared to non-hospitalized patients with fatigue (hospitalized: 37.10% [95% CI: 26.54 - 49.06]; not hospitalized: 24.60% [95% CI: 20.11 - 29.72]; p = 0.012), shortness of breath (hospitalized: 28.68% [95% CI: 18.48 - 41.64]; no hospitalized: 13.72% [95% CI: 8.51 - 21.37]; p = 0.003), weight loss (hospitalized: 37.31% [95% CI: 29.55 - 45.79]; non-hospitalized: 10.83% [95% CI: 8.23 - 14.12]; p

In both studies, limitations focused on the heterogeneity of the selected investigations in aspects such as design, population, measurement of the disease (heterogeneity in access to diagnostic tests), measurement of outcomes (self-diagnosis and differences in access to health), and follow-up period. In addition, there were inconsistencies in the terms used to describe symptoms, as well as limitations in the details and stratification of pre-existing comorbidities, the severity of COVID-19, and treatment methods. Also, the geographical distribution of the participants was another limitation. For example, few studies on long COVID-19 in low- or middle-income countries were identified, no studies were found in the pediatric population, and analyses stratified by ethnicity were also not performed. Other factors that can affect the measurement of prevalence are the type of predominant variant; thus, the Omicron variant (B.1.1.529) is related to mild acute symptoms in the vaccinated population; furthermore, the selection of articles in the English language excludes other important studies published in different languages.

Associated factors

The evidence came mainly from studies on people with a history of hospitalization for COVID-19, included in the systematic review by Michelen et al. (2021) and Maglietta et al. (2022) (8.12). There was high heterogeneity between the studies included in the reviews, given the different operational definitions for the prognostic factors evaluated and the outcomes of interest (Table 3).

Female gender was associated with the presence of any long-lasting COVID-19 symptoms (8 studies; OR: 1.52 [95% CI: 1.27 - 1.82]; I2 = 68%), with the presence of mental health symptoms, such as anxiety, post-traumatic stress disorder, insomnia, among others. (7 studies; OR: 1.67 [95% CI: 1.21 - 2.29]; I2 = 58%), and with fatigue (7 studies; OR: 1.54 [95% CI: 1.32 - 1.79]; I2 = 49%). However, no association was found between the female gender and respiratory symptoms (12 studies; OR: 1.20 [95% CI: 1.00 - 1.45]; I2 = 65%).

The severity of the initial symptoms of COVID-19 was associated with the persistence of respiratory symptoms (9 studies; OR: 1.66 [95% CI: 1.03 - 2.68]; I2 = 71%). In the analysis by symptom, the severity of the initial symptoms was associated with the persistence of cough (2 studies; OR: 1.78 [CI95%: 1.05 - 3.03]; I2 = 0%), and with the diffusing capacity of carbon monoxide ( DLCO) < 80% (6 studies; OR: 2.05 [95% CI: 1.06 - 3.96]; I2 = 49%). No statistically significant association was found with the presence of fatigue (5 studies; OR: 1.23 [95% CI: 0.73 - 2.07]; I2 = 71%).

In the review by Michelen et al. (2021), age was not included in the meta-analysis due to the high variability of the definitions for this variable and the different outcomes. Age ≥ 50-60 years was associated with a higher frequency of low quality of life scores (1 study), the persistence of symptoms assessed at 125 days of follow-up (1 study), and a lower frequency of olfactory dysfunction (1 study). Likewise, as age increased, an increased risk of deterioration in functional status measured by the post-COVID-19 functional status scale was observed (1 study)

Michelen et al. ⁸ narratively synthesized the assessment of the presence of comorbidities and their association with the long-term persistence of COVID-19 symptoms. Having a previous diagnosis of psychiatric illness was significantly associated with the persistence of depressive symptoms (1 study). Likewise, having 2 or more comorbidities were risk factors for the persistence of symptoms during follow-up (1 study). An additional study identified that the presence of spirometric abnormalities 3 months after hospital discharge was more frequent among those with a history of cardiovascular disease and diabetes.

Methodological quality of the included systematic reviews

The quality assessment of the reviews included with AMSTAR 2 is presented in Table 4. Two of the reviews^8,12had low confidence due to the presence of a critical weakness (item 7: list of items not provided). excluded studies and justification for exclusions), and confidence was very low in a further review due to two critical weaknesses (item 2: lack of an explicit statement of the existence of a protocol and item 7 already noted).