A systematic review and meta-analysis of Long COVID symptoms

55 56 57

58 Ongoing symptoms or the development of new symptoms following a SARS-CoV-2 59 diagnosis has caused a complex clinical problem known as “Long COVID” (LC). This 60 has introduced further pressure on global healthcare systems as there appears to be 61 a need for ongoing clinical management of these patients. LC personifies 62 heterogeneous symptoms at varying frequencies. The most complex symptoms 63 appear to be driven by the neurology and neuropsychiatry spheres. 64 65

66 A systematic protocol was developed, peer reviewed and published in PROSPERO. 67 The systematic review included publications from the 1 st of December 2019-30 th 68 June 2021 published in English. Multiple electronic databases were used. The 69 dataset has been analysed using a random-effects model and a subgroup analysis 70 based on geographical location. Prevalence and 95% confidence intervals (CIs) 71 were established based on the data identified. 72 73

74 Of the 302 studies, 49 met the inclusion criteria, although 36 studies were included in 75 the meta-analysis. The 36 studies had a collective sample size of 11598 LC patients. 76 18 of the 36 studies were designed as cohorts and the remainder were cross-77 sectional. Symptoms of mental health, gastrointestinal, cardiopulmonary, 78 neurological, and pain were reported. 79 80

81 The quality that differentiates this meta-analysis is that they are cohort and cross-82 sectional studies with follow-up. It is evident that there is limited knowledge available 83 of LC and current clinical management strategies may be suboptimal as a result. 84 Clinical practice improvements will require more comprehensive clinical research, 85 enabling effective evidence-based approaches to better support patients. 86 87 88 89 Funding: None 90 91

Long Covid, Pain, Neuropsychiatry, Neurology, Autonomic Dysfunction, 92 Gastrointestinal 93 94 95 96 97 98 99 100 101 102 103 104 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 3

105 Global experience with a rapidly evolving and advanced strain of the coronavirus 106 have led to over a million deaths since January 2020. The first case of SARS-CoV-2 107 was reported in China around December 2019. Healthcare systems have been 108 under immense pressure to support both SARS-CoV-2 patients and survivors who 109 continue to demonstrate various symptomatologies which appear to impact the 110 overall quality of life and wellbeing. A report from the Center for Disease Control and 111 Prevention (CDC) in the United States reported that patients recovered from SARS-112 CoV-2 have continuous symptoms of shortness of breath, fatigue, brain fog, cough, 113 chest pain, stomach pain and headache. Bin Cao and colleagues reported that these 114 complications appear to last for at least 6 months thus far (1). Similarly, Carfi and 115 colleagues reported 87.4% of the survivors suffered from a variety of symptoms at 116 post-60 days since the original SARS-CoV-2 diagnosis (2). 117 118 As SARS-CoV-2 survivors continue to share their experiences, clinical researchers 119 hypothesize the continuation of complex symptomatologies for a longer period of 120 time than initially anticipated (3). As are a result, several independent authorities 121 have developed Long COVID guidelines, although the consensus continues to 122 change with the changing evidence base from data gathered from patients. 123 Therefore, a universally accepted Long COVID definition is yet to be elaborated, 124 although a general overview is available. One such important guideline set is from 125 the National Institute for Health and Care Excellence (NICE), which stipulates “Long 126 COVID’ (LC) is commonly used to describe symptoms that continue or develop after 127 acute SARS-CoV-2 diagnosis post-4 weeks” (4). The current research landscape 128 exploring LC is also limited due to the varying reports of symptoms identified in 129 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 4 clinical datasets that demonstrates it to be a ‘moving target’ and it is challenging 130 clinical researchers to guide clinicians on the most optimal steps to pursue when 131 managing the clinical care of these patients. The World Health Organization’s (WHO) 132 Novel Coronavirus Pneumonia Emergency Response Epidemiology Team describes 133 LC as a complex course of illness. Therefore, pandemic policymaking itself requires 134 evidence-based clinical research along with patient-reported outcomes and clinician 135 experiences to be reported in an effective manner to channel a more holistic 136 approach to optimise long-term clinical management. A key component appears to 137 be the difference in LC symptoms between men and women, as reported by Mathew 138 et al., who demonstrate that these observations are vital to understand, and that at 139 present this is based particularly on clinician experience with limited 140 pathophysiological and aetiology (5). 141 142 In this study, we conducted a meta-analysis of peer-reviewed and published data 143 using a systematic approach to better understand LC from a neurological and 144 neuropsychiatry perspective. 145 146 147

148 A systematic methodology was developed, peer reviewed, and published in 149 PROSPERO (CRD42021235351). The primary aim of this systematic review was to 150 determine the prevalence of LC symptomatologies pertaining to neuropsychiatry, 151 neurology, and pain. The secondary aim was to determine any other infrequently 152 reported symptoms that may influence neuropsychiatry and/or neurology and/or pain 153 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 5 diagnosis following LC. The Preferred Reporting Items for Systematic Reviews and 154 Meta-Analysis (PRISMA) was used to report this study. 155 156 Search strategy 157 Multiple databases of Embase, Pubmed, Science Direct, and ProQuest were used 158 with multiple MeSH terms such as nervous system diseases , autonomic central 159 nervous system diseases , autonomic diseases , autonomic nervous system 160 disorders, disorders of the autonomic nervous system , autonomic nervous system 161 diseases, peripheral autonomic nervous system diseases , autonomic peripheral 162 nervous system diseases , parasympathetic nervous system diseases , sympathetic 163 nervous system diseases , headaches, migraine, headache after mental exertion, 164 exertional headache , tension headache, cluster headache, intra cranial 165 hypertension, temporal headache , retro-orbital headache , cervicogenic headache , 166 chronic pain , fibromyalgia, back pain , erythromelalgia, endometriosis, intercostal 167 neuralgia, leg pain, neuropathic pain, chronic pelvic pain , sciatica, muscle fatigue, 168 metal fatigue, cognition, apathy, sleep arousal, sleep deprivation, sleep initiation and 169 maintenance, anxiety, depression emotional lability. 170 171 All studies and surveys were included in the Preliminary R1 round. The reviews and 172 metanalysis identified were scrutinized for references that can be included in our 173 metanalysis. A final set was arrived at looking at the possible relevance of the 174 studies comprising of 302 studies. This was analysed as per PRISMA diagram in 175 Figure 1 in the Results section. 176 177 178 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 6 Eligibility criteria 179 In this meta-analysis, we looked at persistent symptoms in COVID patients, including 180 cohort and cross-sectional studies. All studies included were reported in English. 181 182 Data extraction and synthesis 183 Screening and data extraction were performed by four independent reviewers. Any 184 disagreements were discussed and reached a consensus by two reviewers. To fully 185 investigate the impact of LC on the physical health of survivors, we grouped all 186 reported symptoms into five main categories: general symptoms (which includes 187 pain and other infrequently reported symptoms), neurological, mental disorders, 188 cardiopulmonary, and obstetric problems. 189 190 Data extractions were made via studies that included SARS-CoV-2 survivors that 191 had either been hospitalized or treated as outpatients. Therefore, these patients had 192 a confirmed positive test for SARS-CoV-2 in addition to relevant symptoms. All 193 studies that did not report on follow-up data were excluded. For studies that reported 194 on a control and patient group, only the patient data was extracted and used. A data 195 extraction sheet specific to the clinical question of this study was developed. This 196 Excel spreadsheet included study type, sample size, country, characteristics, 197 information, outcomes, duration of symptoms, and prevalence. 198 199 Risk of bias assessment 200 A quality assessment was performed using the Newcastle-Ottawa Scale (NOS) 201 (Table 1a) to critically appraise the literature included within the systematic review 202 using common variables. Methodological quality and risk of bias was assessed by 203 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 7 independent reviewers according to the NOS, which has validity for use in cohort 204 studies (6) and the adapted version (7) for cross-sectional studies. The scale 205 consists of eight items with three quality parameters: (i) selection, (ii) comparability, 206 and (iii) outcome. We scored the quality of the studies (poor, fair, and good) by 207 allocating stars to each domain as stated below: 208 • A Poor quality score was allocated 0 or 1 star(s) in selection, 0 stars in 209 comparability, and 0 or 1 star(s) in the outcomes domain 210 • A Fair quality score was awarded, 2 stars in selection, 1 or 2 stars in 211 comparability, and 2 or 3 stars in outcomes. 212 • A Good quality score was awarded, 3 or 4 stars in selection, 1 or 2 in 213 comparability, and 2 or 3 stars in outcomes. (6) 214 215 [Table 1a] 216 217 Data analysis 218 A random-effects model with an inverse variance method was used for the meta-219 analysis and the heterogeneity was assessed by I 2. A subgroup analysis was 220 conducted in terms of study geographical location on the symptoms that were 221 reported in more than 10 studies. Sensitivity analysis was used to test the 222 robustness of the results. Funnel plots and Egger’s tests for symptoms with more 223 than 10 studies would be analyzed to detect publication bias. All data analysis will be 224 carried out using R and STATA 15. 225 226 227 228 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 8

229 Of the 302 studies identified, 49 met the inclusion criteria. 36 studies were included 230 in the final meta-analysis. This was reported within the PRISMA document as 231 demonstrated in Figure 1. 232 233 The 36 studies included comprised of a total sample size of 11,598 people. Of the 36, 234 50% were cohort studies and the remainder cross-sectional. The longest follow-up 235 time among the 36 studies was 8 months, although the most common follow-up time 236 was 4 months. The 36 studies covered multiple geographical locations, where 19 237 countries reported five primary classifications of symptomatologies of general clinical, 238 neurological, neuropsychiatry, and cardiopulmonary. Primary clinical features within 239 these categories included fatigue, cognitive impairment, joint pain, anxiety, and 240 depression. These appear to align with the present understanding of LC 241 symptomatologies. Study-based characteristics and outcomes are demonstrated in 242 Table 1b. 243 244 [Table 1b] 245 [Figure 1] 246 247 Meta-analysis 248 The meta-analysis included 36 studies, which are summarized in Figure 2. 249 250 [Figure 2] 251 252 253 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 9 Categorization: 254 General symptoms 255 General symptoms included those associated with pain (such as general pain, 256 muscle or joint pain, and mobility dysfunction), fatigue, fever, hair fall, skin rash, and 257 weight loss. The pooled prevalence of the general problem was 14.4% with a 95%CI 258 of 11.63% to 17.81%. A forest plot for general symptoms is shown in Figure 2.1. 259 260 [Figure 2.1] 261 262 Fatigue was the most frequently reported symptom within the general problem 263 category. Twenty-one of the thirty-six studies reported fatigue symptoms and the 264 pooled prevalence of fatigue was 29.2% with a 95%CI of 21.59% to 39.45%. Muscle 265 pain was the second most prevalent symptom reported among the 13 studies, which 266 led to a pooled prevalence of 13.30% with a 95%CI of 7.48% to 23.67%. However, 267 the prevalence rate of muscle pain is not as high as some of the other symptoms 268 associated within the generalized category. 269 270 The pooled prevalence of joint pain and hair fall were 28.25% (95%CI 14.76% to 271 54.05%) and 20.29% (95%CI 10.56% to 38.98%) respectively. It appears that the 272 prevalence of these two symptoms were high, but only a few studies mentioned 273 these in comparison to those reporting fatigue and muscle pain. Therefore, it is worth 274 standardizing these variables across all studies to manage a better understanding of 275 the clinical relevance. 276 277 278 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 10 Neurological symptoms 279 The neurological symptoms included headache, cognitive impairment, and loss of 280 smell, taste, and hearing. As shown in Figure 2.2, the most frequently reported 281 neurological problems were loss of smell and taste, and headache. The pooled 282 prevalence for loss of smell or taste and both taste and smell as well as headaches 283 were 14.76%, 11.98%, 18.05%, and 10.45% respectively. However, the most 284 prevalent neurological symptom reported appears to be cognitive impairment with a 285 pooled prevalence of 28.85% with a 95%CI of 9.99% to 83.18%. The 95% CI is wide, 286 and the identified heterogeneity based on I 2 was 91%. Despite the high 287 heterogeneity, only 3 studies mentioned the symptoms of cognitive impairment. 288 Further studies and improved sampling would be required to demonstrate a more 289 precise statistical conclusion in regard to cognitive impairment and LC. 290 291 [Figure 2.2] 292 293 Mental health symptoms 294 Four symptoms, and mental health (MH) symptoms including depression, anxiety, 295 PTSD, and sleep disturbances, were reported within the neuropsychiatry category. 296 The pooled results can be found in Figure 2.3. The collective prevalence of MH 297 symptoms was 21.26% (95%CI 16.81% to 26.9%), while each symptom 298 independently also demonstrated a high prevalence. Anxiety prevalence was 299 identified to be 27.77% with a 95%CI of 16.56% to 46.53%, while the prevalence of 300 depression was 22.44% (95%CI 10.22% to 49.35%). The pooled prevalence of 301 studies reporting patients with both anxiety and depression was 23.45% (95%CI 302 19.79% to 27.8%). The prevalence of sleep disturbance was identified to be 19.13% 303 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 11 with a 95%CI of 12.44% to 29.43%. This is an important facet to demonstrate given 304 that there is a large number of studies demonstrating depression and anxiety to be 305 the most commonly reported MH outcomes among SARS-CoV-2 patients. 306 307 [Figure 2.3] 308 309 Cardiopulmonary symptoms 310 LC patients demonstrated cardiopulmonary symptoms with 5 commonly reported 311 issues of chest pain, sore throat, dyspnea, palpitations, and cough. As can be seen 312 from Figure 2.4, dyspnea appeared to have the highest prevalence with 17 of 36 313 studies reporting it as a primary end point. The pooled prevalence was therefore 314 21.48% with a 95%CI of 14.37% to 21.2%. Cough was the second most commonly 315 reported symptom across 14 of 36 studies. The pooled prevalence was 17.83% with 316 a 95%CI of 13.34% to 23.86%. 317 318 [Figure 2.4] 319 320 Gastrointestinal symptoms 321 The overall prevalence of gastrointestinal problems, as shown in Figure 2.5, was 322 6.22% with a 95%CI of 4.61% to 8.39% and is comparatively minimal to the other 323 categorical symptoms identified. Commonly reported symptoms reported in this 324 category were poor appetite, diarrhea and emesis, diarrhea or emesis, nausea, and 325 abdominal pain. Diarrhea and emesis had the highest prevalence of 14.64% with a 326 95%CI of 1.72% to 124.46%. Studies about diarrhea/emesis were too small. Only 327 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 12 two studies mentioned diarrhea and emesis, which also indicated a high 328 heterogeneity with an I2 =97.9%. 329 330 [Figure 2.5] 331 332 Subgroup analysis 333 A subgroup analysis was conducted based on geographical regions correlated with 334 the 8 symptoms of fatigue, headache, cough, loss of smell and taste, dyspnoea, 335 chest, and muscle pain (see Figure 3). 336 337 [Figure 3] 338 339 High prevalence of each symptom was reported by the studies from North America 340 (mainly USA), followed by the Middle East and Australia. Most of the symptoms had 341 a lower prevalence in Africa and Asia. Due to the small number of studies in the 342 subgroup, the conclusions may have bias; for this reason, data from one subgroup 343 was of concern to us. 10 studies from Europe reported dyspnoea in this subgroup 344 and the pooled prevalence of this subgroup was 30.87% with 95%CI of 20.18% to 345 41.55%, which was the second highest prevalence among different regions. This 346 suggested that dyspnoea was a highly prevalent symptom in European countries 347 and should be addressed by the healthcare system to improve post-discharge care. 348 349 Funnel plots of the eight symptoms identified were reported in Figure 4. It is 350 apparent, based on the funnel plots, there is the presence of high heterogeneity. 351 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 13 Many studies were outside the scope of 95% CI, so it was difficult to intuitively detect 352 the bias. Therefore, Egger’s test was used to determine publication bias. 353 354 [Figure 4] 355 356 Sensitivity analysis 357 Many studies were outside the 95% confidence interval as demonstrated within the 358 funnel plots, which could impact the overall conclusions of this study. Therefore, a 359 sensitivity analysis was conducted to determine the consensus of the overall 360

of the study. A Copas selection model was used (8,9) to adjust the 361 pooled prevalence, as demonstrated in Table 2. 362 363 364 Table 2: Summarized results of sensitivity analysis 365 Phenotype N of study Model Probability of publishing study with largest standard error Proportion(%) lower(%) upper(%) p-value for differences between two conclusions headache 14 copas selection model 49.55% 13.4637 10.0332 18.0672 0.1135 random effects model 9.4972 4.8842 18.4673 smell dysfunction 18 copas selection model 100.00% 14.1682 10.1188 19.8380 0.5006 random effects model 14.3823 11.3816 18.1741 taste dysfunction 12 copas selection model 100.00% 12.2338 8.2962 18.0402 0.5495 random effects model 12.3296 9.0703 16.7601 chest pain 11 copas selection model 100.00% 12.4236 7.2232 21.3681 0.103 random effects model 12.1217 6.1288 23.9508 dyspnea 17 copas selection model 100.00% 21.5591 15.1515 30.6797 0.1819 random effects model 21.4774 14.368 32.1046 cough 14 copas selection model 91.11% 18.176 12.8109 25.7852 0.1238 random effects model 17.8321 13.3431 23.8551 fatigue 21 copas selection model 92.46% 29.1951 21.5850 39.4487 0.1478 random effects model 28.9161 20.3158 41.1531 muscle pain 13 copas selection model 67.65% 15.6426 9.1963 26.6103 0.1038 random effects model 13.3031 7.4783 23.6651 366 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 14 367 In Table 2, the proportion of selected studies varied, and the changes in the P-value 368 of the residual selection bias are depicted in Figures 5 (1)-(8). The Copas model 369 (CSM) was used to determine bias within studies based on P-values exceeding 0.1. 370 The proportion of studies used within the CSM are listed within Table 2. It is evident 371 the CSM selected 49.55% studies with headache as a symptom, while the remaining 372 50.45% indicated a significant standard error, demonstrating poor quality and high 373 heterogeneity, thus were excluded. 374 375 The result from the CSM was compared to a random effects model (REM), indicating 376 P-values exceeding 0.05, which demonstrates a lack of statistical significance. 377 Therefore, the results of this study are consistent and provide robust conclusions. 378 379 [Figure 5 (1)-(8)] 380 381 Publication bias 382 Egger’s test was used to determine publication bias. The P-values were calculated 383 based on Egger’s test. 384 385 Table 3: Summarized P-values of Egger’s tests for symptoms with more than 386 10 studies 387 388 Phenotype Number of studies P-value of Egger’s test General problems fatigue 21 0.14 muscle pain 13 0.093 Neurological problems Smell loss 18 0.616 Taste loss 12 0.517 headache 14 0.022* Cardiopulmonary problems . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 15 Chest pain 11 0.048 dyspnea 17 0.007* cough 14 0.085 389 Note: ( * ) : p<0.05 indicates significance 390 391 As shown in Table 3, studies reporting symptoms of headache and dyspnea have 392 significant bias, with P-values of 0.022 and 0.007 respectively. Therefore, the pooled 393 prevalence of headache and dyspnea were 9.5% and 21.48%. In Figures 4(2) and 394 (3), the prevalence of headache and dyspnea may have been underestimated, and 395 more studies should be found to further confirm the conclusions. 396 397 [Figure 4(2) and 4(3)] 398 399

400 There are strengths and weaknesses to our study given that comparing patients with 401 maximum symptoms risks bias reporting. Studies that reported on neuropsychiatry 402 symptoms of depression and anxiety, for example, did not demonstrate a clinical 403 diagnosis. The identified and reported features cannot be deemed to be LC as these 404 patients could have underlying conditions that may not have been reported. Patients 405 who may have had critical respiratory illness, for example, may have been part of 406 these studies, but this data was not captured within the original peer-reviewed 407 publications. This would influence the analysis conducted within our study; therefore, 408 an underrepresentation and/or overrepresentation of some of these symptoms is a 409 point to consider. 410 411 412 413 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 16 414 415

416 This meta-analysis demonstrates the most recent studies identified with possible 417 long COVID symptoms. The pooled data indicate both self-reported and clinically 418 reported symptoms. This initial step is vital to design and develop comprehensive 419 research in the future, especially since SARS-CoV-2 appears to be reporting a 420 varying degree of symptoms. 421 422 The evidence identified demonstrates that long COVID appears to have multiple 423 symptoms, without clear aetiology similar to fibromyalgia and chronic fatigue 424 syndrome. The forementioned conditions also have an association with postviral 425 illness which appear to last longer than previously anticipated. As a result, 426 healthcare systems endeavour challenges with draining resources and souring costs 427 in addition to wellbeing concerns for staff. Another direct result of long COVID 428 disease will be the added burden on waiting times for patients requiring other clinical 429 care and elective procedures created by the pandemic. 430 431 The population prevalence of long COVID identified here could be used to determine 432 symptom-based models to evaluate the requirement for healthcare system 433 resources, and possible disease sequalae which may require care. Presently, 434 instituting effective therapies is based upon present clinical knowledge than 435 evidence-based practices. Repurposing drugs is another common theme among 436 clinicians, and these raise concerns around long COVID potentially becoming a 437 chronic condition in the near future, especially for patients who had significant issues 438 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 17 with COVID. With a growing number of variants of SARS-Cov-2 virus, this further 439 exacerbates the present unknowns of managing these patients in an optimal 440 manner. However, this meta-analysis does provide an opportunity to plan early 441 intervention strategies and target therapies in the future. 442 443 As it is an evolving pathology, further studies are being reported and published 444 swiftly, which has its own challenges. Therefore, to consistently report the latest 445 evidence, there is a requirement for a living systematic review and meta-analysis as 446 well as better methodologies should be developed. It is interesting to note that 447 developed countries appear to have a higher incidence of long COVID based on the 448 geographical data identified within this study. There is a possibility of over- and 449 under reporting, as well as validation of self-reported data. The lack of accurate 450 validated measurements for reported long COVID symptoms similar to other 451 fundamental clinical measures such as blood pressure and temperature causes 452 further problems. 453 454 In addition to these factors, ethical and moral implications to patients, the public, and 455 healthcare professionals continue to augment debates as the pandemic has forced 456 all stakeholders to rethink access to healthcare and treatment. 457 458 It is evident from this study that there are post-COVID symptoms that patients 459 continue to report. It might be beneficial to reduce the severity of the disease. A 460 useful method to reduce these of course would be to increase the vaccination 461 program outputs globally. With mass migration also attributed to the spread of Covid-462 19, an important facet to consider would be to understand the barriers and potential 463 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 18 issues around vaccine acceptability, especially for those returning to work or their 464 education in countries of residence. 465 The COVAX Facility is an international collaborative effort shared between the 466 Coalition for Epidemic Preparedness Innovations (CEPI), the Global Alliance for 467 Vaccines and immunizations (GAVI), the World Health Organization, supporting 468 governments and international organizations. (10) The COVAX Facility is meant to 469 facilitate the development and production of diagnostics, therapeutics, and vaccines 470 to combat the COVID-19 pandemic and to make them accessible to LMIC 471 governments. (11,12) The COVAX Facility does not have a legal 472 entity, therefore cannot enter into binding agreements, and relies upon agreements 473 between its constituent partners (e.g. GAVI, WHO) procuring government, and the 474 vaccine manufacturers. (13) Therefore, the law of contract governs access to 475 vaccines, data related to vaccine procurement and distribution, and related matters. 476 Similarly, sharing of data to better assess the mental and physical health sequalae 477 has been hampered by the lack of an international coordinating mechanism to do so 478 or a uniform set of guidelines that governments, public health officials, private 479 companies, and others may use to share such data. (14) As a result, COVID data 480 related to incidence, disease burden, and long COVID as well as a potential disease 481 sequalae may not be fully understood by the global healthcare community. This is a 482 particular a problem for assessing both COVID and long COVID syndrome impact on 483 differing ethnicities, age groups, and overall health status. Even within academic and 484 clinical research, only open access publications provide insight into evidence. 485 486 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 19 The justification of resources being reallocated to non-life-threatening sequelae of 487 long COVID could be a contentious topic. This further raises legal implications for 488 policymakers. 489 490 The research on nociplastic and immunological explanations for pain symptoms 491 could throw more light in future on the development of pain with long 492 COVID. Genetic studies may also throw some light into the development of long-493 standing chronic pain or even long COVID, although this requires bio-sampling at a 494 high frequency. The role of nutritional status and activity levels also needs to be 495 established and its association with long COVID needs further study. 496 497 498

499 A key finding of this study is that the speed at which SARS-CoV-2 research is being 500 conducted has meant epistemic authority consolidates around particular clinical 501 areas. Therefore, it is vital to synthesise the evidence without any background noise. 502 However, as demonstrated in this study, the gathering of LC data has been limited. 503 The identified data could be associated with autonomic dysfunction, although to 504 confirm this, further investigations would be required. Mapping LC outcomes would 505 be a long-term commitment; therefore, future systematic reviews and meta-analysis 506 should be reported in a living format, combining both clinical and research data to 507 allow a more comprehensive synthesis of evidence with a view to using surveillance 508 data. 509 510 511 512 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 20 List of abbreviations 513 514 CI: confidence interval 515 CSM: Copas selection model 516 GAVI: Global Alliance for Vaccines and Immunizations 517 LC: Long-Covid 518 MH: Mental health 519 NOS: Newcastle-Ottawa Scale 520 PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analysis 521 REM: Random effects model 522 WHO: World Health Organization 523 524 525 526 527 Declarations 528 529 Ethics approval and consent to participate 530 Not applicable 531 532 Consent for publication 533 Not applicable 534 Availability of data and materials 535 All data used within this study have been publicly available. The authors will consider 536 sharing the dataset gathered upon request. 537 538 Competing interests 539 PP has received a research grant from Novo Nordisk, and the other, educational 540 from Queen Mary University of London, from John Wiley & Sons, and other from 541 Otsuka, outside the submitted work. SR reports research funding associated with 542 other studies from Janssen, Otsuka, and Lundbeck. AS reports funding associated 543 with other projects from Medtronic. GD reports research funding associated with 544 NIHR RCF. VR reports funding associated with other studies from the Medical 545 Research Council. 546 547 This research is based on evidence gathered systematically and has not been 548 influenced unduly by expertise. 549 550 All other authors report no conflict of interest. 551 552 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 21 The views expressed are those of the authors and not necessarily those of the NHS, 553 the National Institute for Health Research, Department of Health and Social Care 554 or Academic institutions. 555 556 Funding 557 KE and GD are supported by National Institute for Health Research (NIHR) 558 Research Capability Funding (RCF) and by Southern Health NHS Foundation Trust. 559 All study sponsors had no further role in the study design, data collection, analysis, 560 and interpretation of data; in the writing of the report and in the decision to submit the 561 paper for publication 562 563 Authors’ contributions 564 PP and GD developed study design and GD wrote the first draft of the manuscript. 565 AN conducted database searches and study selection and data extraction. YZ, JQS, 566 GD performed statistical analyses and contributed to the results’ section. AN, AS, 567 GD, YE, YZ, DK, VR, SR, SH, KE, JQS and PP critically reviewed and revised the 568 manuscript. All authors approved the final version of the manuscript. 569 570

571 The authors acknowledge initial contributions made by Dr M Sam Chong, Dr Robert 572 Shane Delamont and Dr Mayur Bodani. We would like to acknowledge Professor 573 Balakrishna Shetty for providing us his thoughts on Covid and Long-covid as a 574 practicing physician managing the ongoing care of patients. 575 576 This paper is part of the multifaceted EPIC project, sponsored by Southern Health 577 NHS Foundation Trust and in collaboration with the University of Oxford, University 578 College London, University College London NHS Foundation Trust and Southern 579 University of Science and Technology (China). 580 581 582 583 584

585 586 (1) Centers for Disease Control and Prevention. Post-COVID Conditions. 2021. 587 https://www.cdc.gov/coronavirus/2019-ncov/long-term-effects.html. 588 589 (2) The Lancet. Facing up to long COVID. Lancet. 2020;396:1861. 590 591 (3) Carfì A, Bernabei R, Landi F, for the Gemelli Against COVID-19 Post-Acute Care 592 Study Group. Persistent symptoms in patients after acute COVID-19. JAMA. 593 2020;324:603-605. 594 595 (4) National Institute for Health and Care Excellence (NICE), COVID-19 rapid 596 guideline: managing the long-term effects of COVID-19. 2020. 597 www.nice.org.uk/guidance/ng188. 598 599 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 22 (5) Darley DR, Dore G, Byrne A, Plit M, Brew BJ, Kelleher AD, Matthews GV. Limited 600 recovery from post-acute sequelae of SARS-CoV-2 (PASC) at eight months of a 601 prospective cohort. medRxiv [Preprint] 2021. 602 603 (6) Wells G, Shea B, O’Connel D, Peterson J, Welch V, Loso M, Tugwell P. The 604 Newcastle-Ottawa Scale (Nos) for Assessing the Quality of Nonrandomised Studies 605 in Meta-Analyses. Ottawa, Canada: Ottawa Hospital Research Institute; 2000. 606 607 (7) Modesti PA, Reboldi G, Cappuccio FP, Agyemang C, Remuzzi G, Rapi S, 608 Perruolo E, Parati G, ESH Working Group on CV Risk in Low Resource Settings. 609 Panethnic Differences in Bloody Pressure in Europe: A Systematic Review and 610 Meta-Analysis. PloS One. 2016;11:e0147601. doi:10.1371/journal.pone.0147601. 611 612 (8) Copas J, Shi JQ. Meta-analysis, funnel plots and sensitivity 613 analysis. Biostat. 2000;1:247-262. 614 615 (9) Copas JB, Shi JQ. A sensitivity analysis for publication bias in systematic 616 reviews. Statistical Methods in Med Res. 2001;10:251-265. 617 618 (10) Berkley S. COVAX explained. 2020. www.gavi.org/vaccineswork/covax-619 explained. 620 621 (11) Lurie N, Gilbert N, Hamburg M, Hoyt K, Morrison JS. The Scramble for Vaccines 622 and the COVAX Facility. [Online event] Center for Strategic and International 623 Studies. 11th August 2020. 624 625 (12) Gavi. Donor profiles. 2020. www.gavi.org/investing-gavi/funding/donor-profiles. 626 627 (13) Halabi S. Solving the Pandemic Vaccine Product Liability Problem. UC Irvine 628 Law Rev. 2021. 629 630 (14) Fegan G, Cheah PY, Data Sharing Working Group. Solutions to COVID-19 data 631 sharing. Lancet Digital Health. 2021;3:e6. 632 633 634 635 636 Figure titles & legends 637 638 Figure 1: PRISMA Flow Diagram 639 Figure 2: Summary of studies included in meta-analysis 640 Figure 2.1: Forest plots for general symptoms 641 Figure 2.2: Forest plots for neurological symptoms 642 Figure 2.3: Forest plots for mental health symptoms 643 Figure 2.4: Forest plots for cardiopulmonary symptoms 644 Figure 2.5: Forest plots for gastrointestinal symptoms 645 Figure 3: Forest plots of subgroup analysis 646 Figure 4: Funnel plots of eight symptoms (reported in more than 10 studies) 647 Figure 5: P-values for residual selection bias 648 649 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 23 650 651 652 653 654 655 656 657 Figure 1: PRISMA Flow Diagram 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 Records identified through database searching (n = 293) Sc re eni ng Inc lud ed Eli gib ilit y Id en tifi cat ion Additional records identified through other sources (n = 9) Records after duplicates removed (n = 302) Records screened (n = 263) Records excluded (n = 3) (exclusion reason is no full text) Full-text articles assessed for eligibility (n = 260) Full-text articles excluded, with reasons (n = 211) - Not COVID-19 survivors (n = 102) - Not related with long COVID (n = 44) - Not long COVID related follow-up symptoms (n = 46) - Not clinical studies (n = 11) - Studies with unrelated condition or intervention (n = 8) Studies included in qualitative synthesis (n = 49) Studies included in meta-analysis of all kinds of symptoms of long COVID (n = 36) Duplicate Records removed (n = 39) . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 699 700 701 702 703 704 705 706 707 708 Figure 2: Summary of studies included in meta-analysis 709 710 711 Figure 2.1: Forest plots for general symptoms 712 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 713 714 Figure 2.2: Forest plots for neurological symptoms 715 716 717 718 719 720 721 Figure 2.3: Forest plots for mental health symptoms 722 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 723 724 Figure 2.4: Forest plots for cardiopulmonary symptoms 725 726 727 728 729 730 731 Figure 2.5: Forest plots for gastrointestinal symptoms 732 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 733 734 735 736 737 738 739 Figure 3: Forest plots of subgroup analysis 740 (1)Fatigue (2)Headache 741 742 743 744 745 746 747 748 749 750 751 (3)Dyspnea (4) Cough 752 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 28 753 (5)Smell dysfunction (6) Taste dysfunction 754 755 (7)Chest pain (8) Muscle pain 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 Figure 4: Funnel plots of eight symptoms (reported in more than 10 studies) 771 (1)Fatigue (2)Headache 772 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 29 773 (3)Dyspnea (4) Cough 774 775 (5)Smell dysfunction (6) Taste dysfunction 776 777 (7)Chest pain (8) Muscle pain 778 779 780 781 782 783 784 785 786 787 788 789 Figure 5: P-values for residual selection bias 790 (1) Fatigue (2) Headache 791 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 792 793 794 (3) Dyspnea (4) Cough 795 796 797 798 (5) Smell dysfunction (6) Taste dysfunction 799 800 801 802 803 (7) Chest pain (8) Muscle pain 804 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 32 Tables 843 844 Table 1a [line 216]: Risk of bias quality assessment 845 846 Selection (S) Comparability (C) Exposure/Outcome (E/O) Sub Total assessment Quality Assessment 1 2 3 4 1a 1b 1 2 3 S + C & E/O & Conclusion NOS Akter et al * NO * * * * * * * Good Good Good Good 7 Huang et al * NO * * NO * * * * Good Good Good Good 6 Humphreys et al * NO * NO * NO NO * * Fair Good Good Fair 5 Simani et al * NO * * NO * * * * Good Good Good Good 6 Taylor et al * NO * * NO * * * * Fair Good Fair Fair 6 Felipe et al NO NO * * * * * * * Fair Good Good Good 6 Hopkins et al * NO * * * * * * * Good Good Good Good 7 Petersen et al * NO * * * * * * * Good Good Good Good 7 Iqbal et al * * * * * * * * * Good Good Good Good 8 Poncet- Megemont et al * * * * * * * * * Good Good Good Good 8 Trevisan et al * * * * * * * * * Good Good Good Good 7 Klein et al * NO * NO * NO * NO NO Fair Good Poor Poor 4 Munro et al * NO NO * NO NO * NO NO Fair Poor Fair Poor 4 Chopra et al * NO * * NO NO * NO NO Good Poor Fair Fair 5 Putri et al * NO * * * * * * * Good Good Good Good 7 Liu et al * NO * * * * * NO NO Good Good Fair Good 6 Tenforde et al * * * * * * * * * Good Good Good Good 8 Sykes et al * NO * * * * * * * Good Good Good Good 6 Townsend et al * NO * * NO * * * * Good Fair Good Good 7 Writing Committee for the COMEBAC study group, 2021 * NO * * * * * NO NO Good Good Fair Fair 5 Augustin et al * NO * * * * * * * Good Good Good Good 6 Duncan et al * NO * NO * * * NO NO Fair Good Fair Fair 4 Osikomaiya et al * NO * * * * * * * Good Good Good Good 6 Orrù et al * * * * * * * * * Good Good Good Good 7 Pujari et al * NO * NO * * * * * Fair Good Good Good 6 Frontera et al * * * * * * * * * Good Good Good Good 7 Holmes et al * * * * * * * * * Good Good Good Good 7 Townsend et al * NO * NO * * * NO NO Fair Good Fair Fair 4 Estiri et al * NO * * * * * NO NO Good Good Fair Fair 5 Chevinsky et al * NO * * * * * NO * Good Good Good Good 6 Pereira et al * NO * * * * * NO NO Good Good Fair Fair 6 Romero- Duarte et al * NO * * * NO * NO NO Good Fair Fair Fair 5 Graham et al * NO * * * * * * * Good Good Good Good 7 Trinkmann et al * * * * * * * * * Good Good Good Good 7 Nguyen et al * NO * * * NO * NO NO Good Fair Fair Fair 5 Vrillon et al * NO * * * * * * * Good Good Good Good 7 Modi et al * NO * * * NO * NO NO Good Fair Fair Fair 4 Pasquini et * NO * * * * * NO NO Good Good Fair Good 5 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 33 al Boscolo- Rizzo et al * NO NO * * * * NO NO Fair Good Fair Fair 5 Capelli et al * NO * * NO * * NO NO Good Fair Fair Fair 4 Yvonne et al * * * * * * * * * Good Good Good Good 7 Raman et al * * * * * * * * * Good Good Good Good 7 Swapna Mandal et al * * * * * * * * * Good Good Good Good 6 Woo et al * NO * * * * * * * Good Good Good Good 6 Puntmann et al * NO * * * NO * * * Good Fair Good Good 5 Bellan et al * * * * * * * * * Good Good Good Good 7 Stavem et al * NO * * * * * * * Good Good Good Good 6 Malek et al * NO * * NO * * NO NO Good Fair Fair Fair 5 Printza et al * NO * NO * * * NO NO Fair Good Fair Fair 4 847 848 849 Table 1b [line 245]: Characteristics of studies included in meta-analysis 850 851 First Author Publication year Study type Sample size Country Percent of Women Ethnicity follow-up time, months p-value Akter 2020 cross-sectional study 734 Bangladesh 24% / 1 / Huang 2021 cohort study 1733 China 48% / 5 / Humphreys 2021 qualitative study 18 UK 50% 55.6% white 16.7% white other 16.7% Asian 5.6% black 5.6% mixed 1 / Simani 2021 cross-sectional study 120 Iran 33.3% / 6 / Taylor 2021 qualitative study 13 UK 84.6% 84.6% white British / / Felipe 2020 cross-sectional study 46 Brazil 54.3% / 4 / Hopkins 2020 cohort study 382 UK 74.6% / 1 loss of smell p<0.001 Petersen 2020 cohort study 180 Faroe Islands 54.4% / 4 / Iqbal 2021 cross-sectional study 158 Pakistan 55.1% / 1 / Poncet- Megemont 2020 cohort study 139 France 62.6% / 1 / Trevisan 2021 observational study 1618 Italy, Spain and Norway 55% / 6 / Klein 2021 cohort study 103 Israel 37.9% / 6 / Munro 2020 cross-sectional study 138 UK 12.5% / / / Chopra 2020 cohort study 488 USA / 51.6% Black 37.3% White 11.1%other/unknown 0r 4.4% Hispanic 86.7% Non-Hispanic 9.3% Unknown 2 / Putri 2021 survey 109 Taiwan 44.95% 100% Asian 0.25 / Liu 2020 cross-sectional study 675 China 53% / 1 / Tenforde 2020 cross-sectional study 292 USA 52% 34.8% White, non- Hispanic 17% Black, non- Hispanic 36.3% Hispanic 11.9% other 0.5 p=0.01 Sykes 2021 cross-sectional study 134 UK 34.3% 91% White 1.5% Black 4 / . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 34 6% Asian 1.5% Mixed/other Townsend 2021 cohort study 40 Ireland 90% / 5 / Writing Committee for the COMEBAC Study Group, 2021 2021 cohort study 478 France 42.1% / 4 / Augustin 2021 cohort study 353 Germany 53.5% / 7 / Duncan 2021 survey NA Scotland / / / / Osikomaiya 2021 cohort study 274 Nigeria 33.9% / 0.5 / Orrù 2021 cross-sectional study 152 Italy / / 3 + insomnia p<0.05 quality of life p<0.05 Pujari 2021 cross-sectional study 94 India 26.6% / 0.5 / Frontera 2021 prospective study 382 US 35% Hispanic 15%/22% Non-Hispanic 62%/59% Prefer not to answer 23%/19% 6 / Holmes 2021 cohort study 27 Australia / / 6 / Townsend 2020 longitudinal study 111 Ireland 63% / 3 / Estiri 2021 cohort study 57622 US / / 3-6, 6-9 / Chevinsky 2021 cohort study 148892 US 57% 40.9% white 25.2% Black 2.4% Asian 21% Hispanic 10.6% others 1-4 / Pereira 2021 cohort study 38 UK 84% BAME group 37% 7 / Romero- Duarte Á 2021 cross-sectional study 797 Spain 46.3% / 6 / Graham 2021 cohort study 50 USA 66% 88% White, 4% Black or African American, 4% Asian, 0% American Indian or Alaskan Native, 4% other Or Hispanic or Latino 12% Not Hispanic or Latino 88% 4 / Trinkmann 2021 cross-sectional study 246 Germany 56.1% / 2 p<0.01 Nguyen 2021 cohort study 125 France 55.2% / 7 / Vrillon 2021 cohort study 125 France 58.4% / 0.7 / Modi 2021 qualitative study 131 US 47% 71% white(non- Hispanic) 7% white(Hispanic) 8% black 2% Asian 1% American Indian 8% Multiracial 4% other (Hispanic) 6 / Pasquini 2021 cross-sectional study 26 Italy 65.4% / 4 / Boscolo- Rizzo 2021 cohort study 183 Italy 54.6% / 6 / Capelli 2021 cohort study 55 Italy 64% / 8 / Yvonne 2020 cross-sectional study 2113 Netherlands and Belgium 85% / 2 p<0.001 Raman 2021 cohort study 58 UK 41.4% BAME group 22.4% 2 p<0.0001 to 0.044 Swapna Mandal 2020 cross-sectional study 384 UK 38% 38.8% British Caucasian 17.1% Other Caucasian 6.5% British Asian 2 p<0.0001 for all symptoms . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint 35 10.3% Other Asian 6.8% Black British 7.6% Other black 13.9% Other ethnicity Marcel S. Woo 2020 cross-sectional study 18 Germany 57.9% / 3 / Puntmann 2020 cohort study 100 France 47% / 2.5 / Bellan 2021 cohort study 238 Italy 59.7% / 4 / Stavem 2020 cross-sectional study 451 Norway 56% / 3 p<0.001 Małek 2021 cohort study 26 Poland 81% / 1.5 / Printza 2020 cross-sectional study 90 Greece 41.1% / 1 / 852 P-value ( * ): P-value <0.05 represents a significant improvement in symptoms at follow-up 853 time compared to onset. 854 855 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint Figure 1: PRISMA Flow Diagram Records identified through database searching (n = 293) Sc re eni ng Inc lud ed Eli gib ilit y Id en tifi cat ion Additional records identified through other sources (n = 9) Records after duplicates removed (n = 302) Records screened (n = 263) Records excluded (n = 3) (exclusion reason is no full text) Full-text articles assessed for eligibility (n = 260) Full-text articles excluded, with reasons (n = 211) - Not COVID-19 survivors (n = 102) - Not related with long COVID (n = 44) - Not long COVID related follow-up symptoms (n = 46) - Not clinical studies (n = 11) - Studies with unrelated condition or intervention (n = 8) Studies included in qualitative synthesis (n = 49) Studies included in meta-analysis of all kinds of symptoms of long COVID (n = 36) Duplicate Records removed (n = 39) . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint Figure 2: Summary of studies included in meta-analysis Figure 2.1: Forest plots for general symptoms . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint Figure 2.2: Forest plots for neurological symptoms Figure 2.3: Forest plots for mental health symptoms . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint Figure 2.4: Forest plots for cardiopulmonary symptoms Figure 2.5: Forest plots for gastrointestinal symptoms . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint Figure 3: Forest plots of subgroup analysis (1)Fatigue (2)Headache (3)Dyspnea (4) Cough . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint (5)Smell dysfunction (6) Taste dysfunction (7)Chest pain (8) Muscle pain Figure 4: Funnel plots of eight symptoms (reported in more than 10 studies) (1)Fatigue (2)Headache . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint (3)Dyspnea (4) Cough (5)Smell dysfunction (6) Taste dysfunction (7)Chest pain (8) Muscle pain Figure 5: P-values for residual selection bias (1) Fatigue (2) Headache . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint (3) Dyspnea (4) Cough (5) Smell dysfunction (6) Taste dysfunction (7) Chest pain (8) Muscle pain . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 9, 2022. ; https://doi.org/10.1101/2022.03.08.22272091doi: medRxiv preprint