Symptomatology of Coronavirus Disease 2019 (COVID-19) - Lessons from A Meta-Analysis Across 13 Countries

Background: COVID-19 pandemic has resulted in varying clinical manifestations and mortality rates. There is no consensus on the symptomatology that would guide researchers and clinicians. Objective: The objective of the study was to identify symptoms and their frequencies of coronavirus disease 2019 with a meta-analysis of studies from several countries. Data sources: A systematic review using PubMed and Google Scholar data sources and reference tracing were used to identify 7176 relevant articles. Eligibility criteria: Suitable articles were selected manually with selection criteria and 14 original articles included for meta-analysis. Data abstraction analysis: PRISMA guideline was used for abstracting data. Then a table was generated by feeding it with numbers and proportions of each symptom described in original studies. A meta-analysis was carried out using random effect models on each symptom separately across the studies and their prevalence rates and 95% condent intervals calculated. Results: We identied 14 relevant scientic papers, either cross-sectional or cohort studies and analyzed. There were 2,660 cases of COVID-19. he majority were from China (n=2,439, 91.7%) and remainder from the Netherlands, Italy, Korea and India and one article from Europe. There was a total of 32 symptoms (i.e. present in >50% of patients): fever (79.56%, 95% CI: 72.17-86.09%), malaise (63.3%, 95% CI: 53.1 – 73.0%), cough (56.7. %, 95% CI: 48.6 - 64.6 %) and cold (55.6%, 95% CI: 45.2 – 65.7%). Symptoms of intermediate incidence (5-49%) were; anosmia, sneezing, ocular pain, fatigue, sputum production, arthralgia, tachypnea, palpitation, headache, chest tightness, shortness of breath, chills, myalgia, sore throat, anorexia, weakness, diarrhea, rhinorrhea, dizziness, nausea, altered level of consciousness, vomiting and abdominal pain. Rare symptoms (<5% of patients) were: tonsil swelling, haemoptysis, conjunctival injection, lymphadenopathy and rash were uncommon symptoms of coronavirus disease (<5%). Conclusion and implications of key ndings: We found (25/32) symptoms to be present in =>5% of cases which could be considered as “typical” symptoms of COVID-19. The list of symptoms we identied are different from those documents released by the WHO, CDC, NHS, Chinese CDC, Institute Pasteur and Mayo Clinic. The compiled list would be useful for future researchers to document a comprehensive picture of the illness.

3. Studies with more than 90 cases for China and at least 20 cases for other countries 4. Studies conducted in any country

Articles published in English language
Exclusion criteria 1. Predominately paediatric studies 2. Those articles that had not stated the frequencies and/or percentages of incidence of symptoms of COVID-19 3. Studies with less than 90 cases for China and less than 20 cases for other countries 4. Articles published in languages other than English Please note that there were several studies from China; authors intentionally limited the number of studies from China when adequate numbers were included.
This was to enable the inclusion of a diverse population in order to improve the generalizability of ndings. We have limited the symptom analysis to mainly adult population, excluding primarily paediatric studies considering the potential variation in symptomatology, target groups and expertise of the authors. The authors' uency in languages is limited to English on published articles leaving those published in other languages, excluded.
Study population: we included data from 14 studies, collecting 2660 individuals in to the analysis. Ages ranged from 0 to 94 years. It was not possible to calculate means and modes due to differences in the data given in the studies. However, the age and sex parameters are given separately for each study in table 01.
The disease severity included mild, moderate, severe/ critical and fatal representing a wider spectrum of disease. Each symptom was taken separately across the studies, including only those tested for that symptom, to avoid confounding by the investigator 'not checking' for the particular symptom. Meta-analysis was carried out studying each symptom separately and their frequencies were calculated and ranked in order. Figure 1 demonstrates the prevalence of all the symptoms and further describes the results of meta-analysis for each symptom separately. These symptoms are illustrated with the relevant system involved in Figure 2.
Among the selected studies for symptom analysis, 9 were from China 7-13 , one study per each country included from Netherlands 14 , India 15 , Korea 16 and Italy 17 . Another article by Spiteri G et.al was included which represented the rst 38 cases in Europe 18 . A large study which included data from 5700 COVID-19 patients in New York, was not include in to the meta-analysis. This is because the clinical characteristics assessed in this study were only fever and tachypnea present at triage. Fever was present only at triage in 30.7% and tachypnea in 17.3% of cases 19 . Here the symptoms before and after the triage was not taken in to account thus limiting the feasibility in our analysis 19 . Several other similar studies were excluded from the meta-analysis concerning the selection criteria; doubts raised on accuracy of data, inadequate information on symptoms studied etc. 31-43 . Methodology and results of these articles were also studied prior to selection, to assure the quality of information. After assessing the suitability of the articles, we selected 14 original studies for the meta-analysis. Those articles were used to generate a table consisting of sample size, number, percentage and prevalence of each symptom. The ndings are presented in Table 1 as the Characteristics of studies. All the patients included were diagnosed to have COVID-19 by detection of nucleic acids (viral RNA detection by Reverse Transcription Polymerase Chain Reaction -RT-PCR).
Statistical analysis: All the symptoms encountered were considered for the analysis. Some symptoms were identi ed in all the studies (e.g. fever and cough), but certain symptoms were only described in one study (e.g. common cold, tonsil swelling, sneezing, palpitation, conjunctival congestion/ injection anosmia, rash, lymphadenopathy and malaise). Therefore, each symptom was considered separately and meta-analysis was carried out to obtain the prevalence of each symptom across all the studies. The sample size and number of events per each symptom in each study was considered in the analysis. Freeman-Tukey double arcsine transformation with inverse variance method was used to consider individual study weights. The overall prevalence of symptoms across studies along with 95% con dence intervals was calculated and symptoms were ranked in the ascending order as depicted in gure 1. R programming language version 3.6.3 22 and Meta package 23 were used in the analysis.

Results
A total of 14 articles with original data describing the clinical manifestations of COVID-19 were retrieved. The largest study was done in China by Guan et.al 7 . Owing to the novelty of the current pandemic there were heterogeneity among the available data and not all symptoms were mentioned. The ages of patients varied from 0 to 94 years. All the studies were descriptive cross sectional or cohort studies and their characteristics are summarized in the Table 1 Clinical manifestations of COVID-19 We identi ed 32 symptoms mentioned in 14 studies. Table 2 summarizes clinical manifestation from those selected original articles with their frequencies.
There were symptoms noted only in one study (sneezing, tonsil swelling, cold, conjunctival injection/ congestion, ocular pain, rash, lymphadenopathy, anosmia and malaise) thus giving a poor statistical con dence on prevalence.

Discussion
This study was designed to identify the symptoms of COVID-19 and to rank them according to their frequencies of occurrence in a globally representative sample. This was di cult as the disease is novel and new symptoms and complications were frequently been reported. Our study recognizes 32 symptoms of Mayo Clinic 30 and in web-based trackers for self-assessment differ from that reported in our study. This comparison is shown in Table 3. We also found two studies which used such clinical criteria 55−5 .
We used original studies from different geographic locations having a range of severities to improve the generalizability of the information. However, nine studies representing about 90% of study population is from China, giving a publication bias for our statistical analysis. There is a signi cant variation in proportions of each symptom across countries and regions. This variability of presentations is likely to be due to the differences in demography of sample, virulence of strain of COVID-19, aggregation of severe cases in to certain centers with higher facilities and milder cases in to other care centers and variations in host response (both genetic and immunological) in different populations.
Wider spectrum of disease severity is covered by the included studies for the meta-analysis. Tostmann et al (14) has studied the COVID-19 in health care workers during a screening test while Chen T et al (8) included 113 fatal cases thus approaching the far severe aspect of the disease. However, the accessibility to health care facility and some factors that could modify clinical features (e.g pregnancy, co-morbidities of individuals) were not taken in to account. The studies primarily on paediatric and neonatal population were identi ed but not included in this analysis and suggest the need of a different study for that [45][46][47][48][49] .
Older age, male sex, presence of comorbidities and certain symptoms were associated with poor outcome. The median age in the Italian study was high (median of 67.5 years), which may at least partially explain the higher fatality rates observed in this population (17) . Sex-disaggregated data suggests a slight male predominance which was also observed in mortality rates (20) . Clinical data associated with disease evolution is critical knowledge especially in a new pandemic. Among the reported cases till February 2020, 14% of COVID-19 cases were severe, causing pneumonia and shortness of breath, and that of about 5% of patients had critical disease, including respiratory failure, septic shock, and multi-organ failure (21) . Host susceptibility is studied in detail by Shi Yu et al, including 487 patients outside Wuhan. They have developed a host risk score using 3 variables: age, sex and presence or absence of hypertension 44 . Further analyses of the symptoms indicate that, certain symptoms like dyspnoea/ shortness of breath (62% in diseased vs. 31% in recovered), chest pain (49% in diseased vs. 30% in recovered) and altered consciousness (22% in diseased vs. 1% in recovered), are associated with higher mortality (8) .
Asymptomatic cases in this study was 9/287 (prevalence − 2.71%, 95% CI 0.96%-5.12%) using data from articles by Chen J (9) and Spiteri G (18) . This might not reveal the true picture, because a large scale screening tests done on populations at risk needed to assess this. However, a study done in Japanese Diamond Princess Cruise ship by Mizumoto K et el. shows valuable results. Here 3,711 patients were kept quarantined after nding one patient with COVID-19.
Out of all, 634 cases became positive and 306 (48.3%) cases were symptomatic and 328 (51.7%) were asymptomatic (24) In this study, we did not concentrate on the chronology of development of symptoms and complications which is also very important for clinicians when assessing patients, and need to be addressed in detail separately. However, temporal clinical progression has been assessed by Chen J et al (N = 249) in their study 9 . There had been reports of possible 'reactivation' of COVID-19 after recovering from the rst infection 37 ; the symptoms of such cases are not taken in to this review.
The results of our systematic review highlight the common and uncommon clinical symptoms which will help clinicians across the globe in the diagnosis and management of suspected cases of COVID-19, especially during the early phase. This will help in de ning the disease presentation and improves diagnostic skills. These common and uncommon symptoms could be utilized in studying patients and designing future research.
There are a multitude of other uncommon or rare manifestations of COVID-19 not described in these studies presented from many countries that have been mainly the focus of case reports [50][51][52][53][54] .

Limitations
Some symptoms assessed were only present in one or two studies and other studies have not recorded them or not inquired about them making them statistically less reliable on their frequencies.
We have not focused on the chronology of symptom development and complications. Reports in languages other than English were not included

Conclusions
There are 32 symptoms of COVID-19 representing multiple organs and systemic features. Fever is the most common symptom followed by malaise, cough, cold and anosmia. Researchers and clinicians should be aware of a comprehensive list of symptoms to describe the illness and for research. Availability of data and materials: The pooled articles, data-sheets and analytic results are available with the authors for future references Competing interests: There are no competing interests.