Table 1. Study population characteristics, extracted outcomes and assessed limitations and Risks of Bias (ROBs).
| Study | Type | Population and setting† | Relevant outcome(s) | Number studies & participants in relevant outcome(s) | Limitations Described | Quality assessment of primary studies‡ | Top ROBs |
|---|---|---|---|---|---|---|---|
| Prevalence SRs | |||||||
| Di Gennaro et al. 2022 | SR/MA | Age: Adults and children, mean age 52.3 Sex: 48.8% female Exposure: PCR or lab confirmed Countries/continents: Australia (2), Austria (3), Bangladesh (2), Belgium (2), Brazil (3), Canada (2), China (17), Denmark (3), Ecuador (1), Egypt (3), Faroe Island (1), France (13), Germany (9), Hong Kong (1), India (3), Iraq (1), Iran (3), Israel (3), Italy (35), Japan (3), Korea (2), Latvia (1), Mexico (1), Netherland and Belgium (1), Netherlands (5), Norway (2), Poland (2), Russia (3), Saudi Arabia (1), Singapore (1), Spain (25), Switzerland (6), Turkey (3), UK (13), USA(20) |
Outcome 1) Cumulative incidence of any Long COVID signs and symptoms at follow-up from random-effects meta-analysis Outcome 2) Cumulative Incidence of any long COVID signs and symptoms at follow-up from random-effects meta-analysis for mean ages a) 18–60 or b) ≥60 Outcome 3) Follow-up at a) 3 month or b) 3–6 months Outcome 4) By hospitalization a) hospitalized, b) mixed, c) non-hospitalized |
Outcome 1) 196 (120,970) Outcome 2) NR Outcome 3) NR Outcome 4) NR |
1) Lack of objective and precise scales for self-report of long COVID symptoms 2) High heterogeneity only partially explained by meta-regression 3) Did not explore role of vaccination 4) Presence of publication bias |
Tool: NOS 67/196 moderate ROB (6–7 stars) 129/196 low ROB (8–9 stars) |
NR |
| Fernandez-de-las-Pena 2022 [19] | SR | Age: adults mean age 50.1 Sex: 56.6% female Exposure: PCR or serological test Countries/continents: Japan, India, Italy, Spain, USA, UK |
Outcome 1) Prevalence of at least 1 post-COVID symptom by variants Outcome 2) Risk of long COVID by variants |
Outcome 1) 3 (2,115) Outcome 2) 2 (98,104) |
1) High heterogeneity 2) Small number of studies 3) Lack of control for confounders (e.g., vaccination status, reinfections) 4) Many studies investigating long COVID with wild type or Alpha variants sampled hospitalized patients, so may overrepresent chronic fatigue or post-intensive care syndrome due to hospitalization 5) No inclusion of uninfected control groups in any study 6) Cannot exclude potential influence of pandemic-related factors such as social alarm, somatization, physical inactivity, etc. |
Tool: NOS 6/6 High quality (7, 7, 7, 8, 8, 9) |
1) Outcome of interest was not present at the start of the study (0/5 cohort/cross-sectional studies) 2) Assessment of outcome was through blind independent assessment or record linkage (2/5 cohort/cross-sectional studies) |
| Huang et al. 2022 | SR/MA | Age: Adults, 53.3 (mean) Sex: 44.8% female Exposure: NR Countries/continents: Europe, Western Pacific, America, Eastern Mediterranean, Asia and Africa |
Outcome 1) Pooled prevalence of any Long COVID symptoms at 1 month by random effects meta-analysis Outcome 2) Pooled prevalence of any Long COVID symptoms at 3 months by random effects meta-analysis Outcome 3) Pooled prevalence of any Long COVID symptoms at 6 months by random effects meta-analysis |
Outcome 1) 22 (NR) Outcome 2) 27 (NR) Outcome 3) 20 (NR) |
1) Substantial heterogeneity. Meta-regression found significant association with study region but no other variable (e.g., variable time zero) 2) Prevalence may be over-estimated as original studies included subjects in acute phase and relatively high proportion of severe/critical cases 3) Long COVID symptoms are nonspecific and lack of controls in original studies make it difficult to determine causality 4) Hard to explain temporal change |
Tool: NOS for cohort studies (Score interpretation not given) 4/76 3–4 out of 9 stars 53/76 5–7 out of 9 stars 19/76 8 out of 9 stars Tool: AHRQ for cross-sectional studies (Score interpretation not given) 1/5 7 stars 2/5 9 stars 2/5 10 stars |
1) Selection of the non-exposed cohort 2) Comparability of cohorts on the basis of the design or analysis 3) Adequacy of follow up of cohorts |
| Ma et al. 2023 | SR/MA | Age: Adults, 40–53 (range of means/medians)* Sex: 35–75%* Exposure: Serology or RT-PCR* Countries/continents: Italy, Germany, Brazil and Luxembourg* |
Pooled prevalence of at least 1 symptom among asymptomatic SARS-CoV-2 infected adults | 4(99) | 1) Absence of data on asymptomatic cases 2) Absence of some consequences that may not have been recorded |
Tool: NOS 4 low ROB (7, 8, 9, 9) |
Comparability of cohorts on the basis of design or analysis |
| Nasserie et al. 2021 | SR | Age: mean/median range 47–65.5* Sex: 54% male Exposure: PCR or lab confirmation in 14/16, NR in 2/16* Countries/continents: UK (4), Spain (4), China (3), Italy (2), France (1), Canada (1), Austria (1)* |
Outcome 1) Median and IQR of prevalence of 1 or more symptom at end of individual or study follow-up Outcome 2) Median and IQR of prevalence of 1 or more symptom at end of individual or study follow-up of a) <3 months or b) ≥ 3 months |
Outcome 1) 16 (4,695) Outcome 2) a) 8(1,386) b) 8(3,309) |
1) Design limitations among included studies prevented addressing symptom duration, resolution, and trajectory of global quality of life 2) Symptoms not captured using standardized definitions or instruments too difficult to compare frequency and severity 3) Studies measure same symptoms in different ways report different estimates within the same study 4) Few studies examined past history or baseline prevalence of similar symptoms assessed 5) Variable time-zero: diagnosis or symptom onset, hospital admission, hospital discharge, or recovery from illness 6) Variable follow-up duration 7) Heterogeneity of design features and quality |
Tool: NIH (overall summary not performed) a) Patients randomly selected or all eligible patients included 12/16 studies b) Baseline severity reported 11/16 studies c) Attrition 4/16 not reported or ≥30% 3/16 20–29% 6/16 10–19% 3/16 <10% d) Outcome repeatedly measured 2/16 studies e) Established outcome scales used 14/16 studies for some or most outcomes |
1) Repeated outcome measurement 2) Attrition 3) Baseline severity reporting |
| Nittas et al. 2022 | Umbrella Review and evidence synthesis | Age: Adults and children Sex: NR Exposure: clinical, serological or PCR testing Countries/continents: NR |
Outcome 1) Median and IQR of prevalence of at least one Long COVID symptom in adults and children Outcome 2) Median and IQR of prevalence of at least one Long COVID symptom in adult patient studies with population-based samples and adjusted prevalence for cohorts with negative comparators |
Outcome 1) 40 (46,144 cases and controls) Outcome 2) 10 (7097 cases, 11,050 controls) |
1) Much early research on SARS-CoV-2 was designed and implemented quickly with a focus on conveniently sampled hospital and outpatient participants so samples recruited early in the pandemic often not as widespread and captured more severe cases; only 4 population-based studies reported prevalence estimates 2) Prevalence of certain symptoms rarely placed in relation to their prevalence in persons without SARS-CoV-2 before or during the pandemic. Most studies fail to distinguish between COVID-related conditions and those linked to preexisting conditions; this is especially true for studies reporting vital organ impairment 3) Certain populations including the elderly, people with disabilities, children and asymptomatic SARS-CoV-2 patients are underrepresented 4) Little evidence on risk and protective factors |
Tool: Hoy et al. (overall summary not performed) a) Is the target population representative of the national population? 33/40 high risk 6/40 low risk 1/40 unclear b) Was some sort of random selection used to select the sample, OR was a census undertaken? 36/40 high risk 4/40 low risk c) Was the likelihood on non-response bias minimal? 24/40 high risk 16/40 low risk |
Lack of random selection or population-based sampling |
| O’Mahoney 2023 | SR/MA | Age: 5–73* Sex: 14–94% male (range by studies) Exposure: self-diagnosed or confirmed by a PCR, antigen or antibody test Countries/continents*: Africa (3), Asia(15), Oceania (1), Europe (53), North America (21), South America(1), multiple (1) |
Outcome 1) Pooled prevalence of at least 1 symptom at follow-up by random effects meta-analysis for all studies Outcome 2) Pooled prevalence of at least 1 symptom at follow-up by random effects meta-analysis for studies that included a) hospitalized patients, b) mixed (both hospitalized and non-hospitalized), or c) non-hospitalized Outcome 3) Median prevalence and interquartile range of at least 1 symptom at follow-up time for studies that included only adult participants Outcome 4) Median prevalence and interquartile range of at least 1 symptom at follow-up time for studies that included only adult participants at a) <12 weeks of follow-up or b) at ≥12 weeks of follow-up |
Outcome 1) 95 (411,630) Outcome 2) a) 48 (74,422) b) 36 (133,321) c) 11 (203,887) Outcome 3) 83(406,394) Outcome 4) a) 19 (5171) b) 64 (401,223) |
1) 144/194 studies did not report race/ethnicity 2) Limited standardization in using self-report tools 3) Lack of unified consensus definition on long COVID 4) Wide-ranging follow-up period 5) Small number of studies with control/comparator groups 6) Exclusion of studies that recruited from long COVID clinics to avoid selection bias 7) Geographic homogeneity—most studies derive from Europe 8) Did not assess impact of vaccination status and variants |
Tool: NIH# Cross-sectional and cohort#: 114 good quality 57 fair quality 14 poor quality Case-control#: 5 good quality 4 fair quality Case series#: 1 good 1 fair |
1) Sample size justification, power description, or variance and effect estimates was not provided (170/184 cohort studies) 2) Exposure was not assessed more than once over time (170/184 184 cohort studies) 3) Key potential confounding variables were not measured or adjusted on the impact between exposure and outcomes (95/184 184 cohort studies) |
| Rahmati et al. 2023 | SR/MA | Age: Adults, median age range 40–61* Sex: 28–68% female* Exposure: NR Countries/continents: China (7), USA (1), Spain (1), France (1)* |
Pooled event rate of any Long COVID symptom from random-effect meta-analysis | 10(4,589) | 1) Substantial heterogeneity possibly due to small study size, different assessment scales and wide variation in reported prevalence data (e.g., prevalence of at least one unresolved symptom ranged from 16% to 76%) 2) Paucity of data up to 2 years of COVID-19 follow-up 3) Only 1/3 of included studies conducted in-person assessment during follow-up, the rest was mixed of in-person and phone, only-phone or via electronic health record, so risk of recall bias 4) Most studies conducted in China so may not generalize to Europe, USA and resource-poor nations 5) Most study participants were from hospitalized patients during early waves of the pandemic so cannot generalize to new variants or non-hospitalized patients 6) Only a slight majority of studies had control/comparator groups |
Tool: NOS 9/10 moderate ROB (7 stars) 1/10 low ROB (8 stars) |
1) Selection of unexposed cohort was from the same community as exposed (1/10 studies) 2) Assessment of outcome was through blind independent assessment or record linkage (0/10 studies) |
| Zeng et al. 2022 | SR/MA | Age: Adults, 58–68.8 (range of means/medians)* Sex: 54.8% (individual-level)* Exposure: NR* Countries/continents: Spain (3), France (2), UK, USA, China* |
Pooled prevalence of at least 1 symptom among SARS-CoV-2 infected adults from cohorts with median/mean age ≥60 | 8(3658) | 1) High heterogeneity between studies with most I-squared >50%; heterogeneity may be due to different case definitions, diagnostic criteria, and follow-up durations 2) Outcome based mainly on self-report and limited objective evidence 3) Cannot ascertain causality 4) Only reviewed work done with Alpha variant 5) Did not involve vaccination because of limited reporting in primary studies |
Tool: NOS <5 stars = low quality 5–7 stars = moderate quality >7 stars = high quality Full sample: 12/151 high quality 124/151 medium quality 15/151 (9.9%) low quality Relevant outcome: 1/8 low quality (4) 7/8 moderate quality (5,5,5,6,6,6,7) |
NR |
| Risk Factors SRs | |||||||
| Byambasuren et al. 2023 | SR | Age: Individuals eligible to receive any COVID-19 vaccine during study period (only adults in at least 15/16 studies) Sex: NR Exposure: NR Countries/continents: USA (8), UK (4), Netherlands (2), France and Italy |
HR or OR of at least 1 long COVID symptoms, most common symptoms, long COVID of any severity, receiving care >3 months after infection, confusion/difficulty concentrating, risk of fatigue after: 1) 1 dose of pre-infection vaccine vs. 0 dose 2) 2 doses of pre-infection vaccine vs. 0 dose 3) 3 doses of pre-infection vaccine vs. 0 dose 4) Any dose of pre-infection vaccine vs. 0 dose Median and IQR given if there are 5 or more studies; OR and 95% CI given if there is only 1 study |
1) 1 dose: 5(NR) 2) 2 doses: 5(NR) 3) 3 doses: 1(318 vaccinated; 421 unvaccinated) 4) Any dose: 5(NR) |
1) Lack of consistent definition of Long COVID 2) Could not recalculate common ratio so used HR/OR/RR depending on study. 3) Could not conduct meta-analysis due to high heterogeneity and lack of data on specific vaccine type, time between exposure and disease and viral variant. 4) Could not determine prevalence of individual symptoms as not reported in most studies. 5) Lack of high quality primary literature, in particular RCTs 6) Less than half of studies used PSM to form comparison group. |
Tool: ROBINS-I Assessed adjustment for predetermined confounders: age, sex, BMI, initial disease severity, comorbidity, vaccine hesitancy 3 critical ROB 5 serious ROB 3 moderate ROB |
1) Selection of reported result (serious or critical in 11/16 studies) 2) Confounders not adjusted for (serious or critical in 9/16 studies): vaccine hesitancy, initial disease severity, sex 4) Dealing with missing data (not described or serious ROB in 7/16 studies) 4) Measurement of outcome (serious or critical in 5/16 studies): ICD-10 codes with high detection bias; unclear definition and self-reported outcomes |
| Notarte et al. 2022 | SR/MA | Age*: 40–65 (range of means/medians) Sex*: 51.9% female (NR in 2/7 studies) Exposure*: RT-PCR Countries/continents*: UK(3), Switzerland (1), Italy (1), Faroe Islands (1), Spain (1) |
Pooled OR and 95% CI for the association between sex and presence of any long COVID-19 symptom | 7(386,237 COVID survivors; 1,944,580 COVID-negative controls) | 1) Lack of consistent definition of long COVID 2) Only studies that used WHO definition included in meta-analysis so small number of studies 3) Lack of differentiation in risk factors (i.e., hospitalization status, variants of concern) 4) Did not investigate COVID-19 associated risk factors (e.g., severity of acute infection) |
Tool: QUIPS 6 domains: Low ROB if ≤1 domain has moderate ROB; High ROB if ≥1 domain high ROB or ≥3 domains moderate ROB; all papers in between classified as moderate ROB 6/16 high ROB 5/16 moderate ROB 5/16 low ROB |
1) Adjustment for other prognostic factors (2H, 10M) 2) Attrition (4H, 5M) 3) Participation (2H, 2M) 3) Prognostic factor measurement (2H, 2M) |
| Pillay et al. 2022 | SR/MA | Age: 42.7–69 (median range)* Sex: 49.0% (median)* Exposure: ≥90% lab confirmed in 12/17 studies* Countries/continents: China (5), Italy (2), Norway (2), Russia (2), Switzerland (2), USA (1), UK (1), Sweden(1) and Turkey (1)* |
Pooled OR and 95% CI for non-recovery/persistent systems from random-effects meta-analysis for: 1) Age (continuous) 2) Age (categorical) 3) Sex (female vs. male) 4) Comorbidities (≥1 vs. 0) 5) Acute COVID-19 Severity (Critical/ICU vs. not) 6) Acute COVID-19 Severity (Severe/Critical vs. not) 7) Need for hospitalization |
Total: 9(7170) 1) 2(3296) 2) 40–60 vs 18–40: 4 (2867) >60 vs 18-40yrs: 3(1440) 3) 8(6163) 4) 4(2069) 5) 3(1722) 6) 2(1438) 7) 2(1030) |
1) Findings applicable mainly to long-term consequences ≥22 weeks after acute illness 2) Large proportion of hospitalized population 3) Several potential risk factors but none identified as strong association with long COVID outcomes 4) Evidence sparse on pre-existing socioeconomic variables (e.g., race/ethnicity, income, education, employment) 5) Self-reported outcome/exposure data so subject to recall and misclassification bias |
Tool: JBI Checklist for Cohort Studies 5/9 studies high ROB (≥2 domains high ROB) 4/9 some concern for ROB (<2 domains high ROB) |
1) Use of appropriate statistical analysis (2/9) 2) Risk factor measurement in valid/reliable way (4/9) 3) Outcome measured in valid/reliable way (5/9) 3)Follow up complete, and if not, the reasons to loss to follow up were described and explored (5//9) |
| Tsampasian et al. 2023 | SR/MA | Age: adults (≥18) Sex: NR Exposure: RT-PCR, serology or other laboratory confirmation in 38/41 studies; self-report admitted in 3/41 studies Countries/continents: Europe (30), Americas [Brazil, Canada, US] (6), Asia (5), Africa (1) |
Pooled OR and 95% prediction intervals (PI) for developing post-COVID condition from random-effects meta-analysis for: 1) Sex (female vs. male) 2) Age (≥40 vs. 18–40) 3) BMI (≥30 vs. <30) 4) Smoking status (current smoker vs. nonsmokers) 5) Comorbidities a) Anxiety/depression b) Asthma c) CKD d) COPD e) Diabetes f) Immunosuppression g) Ischemic heart disease 6a) Hospitalization (hospitalized vs. not) 6b) ICU admission (admitted to ICU vs not) 7) Vaccination status (2 doses vs. unclear) |
1) 38 (727,630) 2) 9 (324,950) 3) 16 (701,807) 4) 20 (455,204) 5a) 4(634,734) 5b) 13(639,397) 5c) 8(255,791) 5d) 10 (257,340) 5e) 18(259,978) 5f) 3(967) 5g) 5(201,906) 6a) 8(265,466) 6b) 10(213,441) 7) 4(249,788) |
1) High heterogeneity in many outcomes 2) Limitation of NOS scale itself even though all studies were rated as moderate or high quality 3) High ROB associated with observational studies 4) Different definitions of symptoms included among different studies 5) Pooled results independent of variants |
Tool: NOS 11/41 moderate quality (6/9 stars) 30/41 high quality (7-9/9 stars) |
NR |
| Watanabe et al. 2023 | SR/MA | Age: 45–58 years (range of medians/means)* Sex: 9.0–63% female* Exposure: PCR, serology or symptoms Countries/continents: UK (2), USA (2), Turkey (1) and Italy (1)* |
Pooled OR and 95% CI for the incidence of long COVID after: 1) 2 pre-infection vaccination doses vs. 0 dose 2) 2 pre-infection vaccination dose vs. 1 dose; 3) 1 pre-infection vaccination dose vs. 0 dose) by random effects meta-analysis |
2 doses vs. 0 dose: 4(2 doses: 60,099; 0 dose: 536,291) 2 doses vs. 1 dose: 3(2 doses: 3142; 1 dose: 21,872) 1 dose vs. 0 dose: 2(1 dose: 15,842; 0 dose: 392,745) |
1) Only observational studies included 2) Varied proportion of ICU-admitted patients 3) Definition of long COVID varied; 4) Could not evaluate variants of concern as observational period varied widely 5) Could not evaluate effect 3–4 vaccine doses |
Tool: NOS 1/5 moderate risk of bias (7) 4/5 low risk of bias (8, 8, 9, 9) |
Representativeness of the exposed cohort |
AHRQ, Agency for Healthcare Research and Quality; BMI, body mass index; CI, confidence interval; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; H, high; HR, hazard ratio; ICU, intensive care unit; IQR, interquartile range; JBI, Joanna Briggs Institute; M, moderate; MA, meta-analysis; NIH, National Institutes of Health; NOS, Newcastle-Ottawa Scale; NR, not reported; OR, odds ratio; ROB, risk of bias; ROBINS-I, Risk of Bias in Non-randomized Studies–of Interventions; RT-PCR, reverse transcriptase polymerase chain reaction; QUIPS, Quality in Prognosis Studies; SR, systematic review; WHO, World Health Organization
†: data in column extracted for all studies included in SR unless otherwise specified by asterisk (*); ‡: data in column extracted only for studies included in relevant outcomes unless otherwise stratified by number sign (#)