The protocol of this study was registered in PROSPERO (CRD42021239187) and the protocol required no ethical clearance. To ensure the robustness of the generated data, we followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines to search electronic databases and report our findings. ³⁹ The completed PRISMA checklist is presented in Figshare. ⁴⁰

Studies reporting at least one prolonged or persistent GI symptom such as nausea, vomiting, diarrhea, lack or loss of appetite, abdominal pain, and dysgeusia in patients with COVID-19 after being discharged from hospital were considered eligible. Editorials, commentaries, reviews, case reports, and case series were excluded. Diagnosis of COVID-19 must have been confirmed using RT-PCR of SARS-CoV-2 RNA from nasal or oropharyngeal swab samples. Studies that diagnosed patients with COVID-19 based on symptoms only (without nucleic acid testing) were excluded. COVID-19 survivors were defined as all patients with COVID-19 who met either the WHO or China National Health Commission discharge criteria. ^{41 , 42} Prolonged GI symptoms were defined as persistence of symptoms for at least two weeks after discharge in COVID-19 survivors.

The potential articles in three databases (PubMed, Scopus, and Web of Science) were searched as of January 30, 2021. The searches were limited to 2019-2021 and only articles written in English were considered eligible. The search strategies were as follows. PubMed ([Title](“SARS-CoV-2” OR “COVID-19” OR “Wuhan coronavirus” OR “Wuhan virus” OR “novel coronavirus” OR “nCoV” OR “severe acute respiratory syndrome coronavirus 2” OR “coronavirus disease 2019” OR “2019-nCoV” OR “2019 novel coronavirus” OR “SARS 2”) AND ([Title](“prolong*” OR “follow-up” OR “persistent” OR “sequelae” OR “consequen*” OR “prospective” OR “cohort” OR “long-term” OR “follow*” OR “longitudinal”). Web of Science ([Title](“SARS-CoV-2” OR “COVID-19” OR “Wuhan coronavirus” OR “Wuhan virus” OR “novel coronavirus” OR “nCoV” OR “severe acute respiratory syndrome coronavirus 2” OR “coronavirus disease 2019” OR “2019-nCoV” OR “2019 novel coronavirus” OR “SARS 2”) AND ([Title](“prolong*” OR “follow-up” OR “persistent” OR “sequelae” OR “consequen*” OR “prospective” OR “cohort” OR “long-term” OR “follow*” OR “longitudinal”). Scopus ([Title](“SARS-CoV-2” OR “COVID-19” OR “Wuhan coronavirus” OR “Wuhan virus” OR “novel coronavirus” OR “nCoV” OR “severe acute respiratory syndrome coronavirus 2” OR “coronavirus disease 2019” OR “2019-nCoV” OR “2019 novel coronavirus” OR “SARS 2”) AND ([Title](“prolong*” OR “follow-up” OR “persistent” OR “sequelae” OR “consequen*” OR “prospective” OR “cohort” OR “long-term” OR “follow*” OR “longitudinal”).

Essential information of all articles was imported to a reference manager (EndNote X9, Thompson Reuters, Philadelphia, PA, USA) and duplicated records among the three databases were removed. The titles and abstracts of all records were screened to identify eligible articles. The full texts of potentially eligible studies were downloaded and reviewed by two authors (MF and HH). The eligibility of each study was decided based on the eligibility criteria and the availability of the data. Data extraction was conducted as explained in previous studies. ^{23 , 24 , 43} Briefly, important data from the eligible articles were extracted and whenever required supplementary materials were extracted. The list of references was retrieved to search for additional relevant studies. The collected study characteristics of the eligible articles included author(s), year of study, study site and country, study design, extent of follow-up conducted after discharge, number of patients with COVID-19, number of patients with COVID-19 with prolonged specific GI symptoms, and severity of the COVID-19 infection during admission to the hospital.

Two main outcomes were evaluated in this study: (a) global prevalence of prolonged GI symptoms including nausea, vomiting, diarrhea, lack of appetite, abdominal pain, and dysgeusia and (b) association of COVID-19 severity with the presence of prolonged GI symptoms (nausea, vomiting, diarrhea, lack of appetite, abdominal pain, and dysgeusia). In addition, the possible pathogenesis mechanisms of GI symptoms in COVID-19 including those with prolonged GI symptoms are discussed.

The prevalence of each prolonged GI symptom (nausea, vomiting, diarrhea, lack of appetite, abdominal pain, and dysgeusia) was calculated as the number of patients with a prolonged symptom divided by the total number of patients with COVID-19 with or without the specific GI symptom during the follow-up and expressed as frequency (%) with a 95% confidence interval (CI). The associations of COVID-19 severity and the risk of prolonged GI symptoms were also calculated. Forest plots were used to visualize the data.

The Newcastle-Ottawa scale (NOS) was used for critical assessment of the quality of each included study. ⁴⁴ The NOS evaluates nine characteristics of a study including four, one, and three items for sample selection, group comparison, and the outcome, respectively. The scores range between 0 to 9 and a study is classified into one of three groups based on the score: low (≤ 4), moderate (between 5–6), and high-quality (≥ 7) study.

The Q test was used to evaluate the heterogeneity of the pooled data and the data was analyzed using a random-effect or fixed-effect model as appropriate. Egger’s test was used to assess publication bias (a p < 0.05 is considered indicative of potential publication bias). The associations between severity of COVID-19 and the risk of GI symptoms were calculated using the Z test. Review Manager version 5.3 was used to analyze the data. ⁴⁵

The database searches yielded 4,050 eligible original articles, with 2,005 publications remaining after the duplicates were removed. Initial screening of the titles and abstract excluded 1,244 articles, leaving 761 studies ( Figure 1). After reviewing the full-texts of these studies, an additional 739 articles were excluded for several reasons including that they were reviews, case series, case reports, initial reports on COVID-19, letters or commentaries, studies on specific groups, recommendations, clinical trials, and studies with insufficient data. The final screening resulted in 22 articles which were included in this meta-analysis. Figure 1.

Among the 22 studies selected, the meta-analysis to calculate the prevalence of prolonged GI symptoms included four studies for nausea, ^{35 , 46 – 48} three studies for vomiting, ^{35 , 49 , 50} seven studies for diarrhea, ^{35 , 46 – 48 , 50 – 52} three studies for loss of or decrease in appetite, ^{35 , 49 , 50} three studies for abdominal pain, ^{46 , 47 , 50} and six studies for dysgeusia. ^{46 , 47 , 49 , 53 – 55} The studies included for the prevalence of prolonged GI symptoms are summarized in Table 1. Table 1.

Symptom	Year	Study design	City	Country	Days from discharge to follow-up	Prevalence of GI symptom over total followed survivors			Prevalence of GI symptom based on severity of COVID-19						NOS	Ref
						No patient	Total patient	Percentage	Mild- moderate	Total	Percentage	Severe	Total	Percentage
Nausea	2020	Prospective	Wuhan	China	14	2	131	1.53	1	62	1.61	1	69	1.45	7	35
	2020	Cohort	Georgia	USA	38 (21–49)	5	26	19.23	NA	NA		NA	NA		8	46
	2020	Prospective	Aachen	Germany	56 (48–71)	2	33	6.06	NA	NA		NA	NA		8	47
	2020	Prospective	Wuhan	China	14	1	337	0.30	NA	NA		NA	NA		7	48
	Total					10	527	1.90	1	62	1.61	1	69	1.45
Vomiting	2021	Cohort	Hubei	China	153 (146–160)	80	1655	4.83	75	1538	4.88	5	117	4.27	8	49
	2020	Prospective	Isfahan	Iran	28	12	452	2.65	11	400	2.75	1	52	1.92	9	50
	2020	Prospective	Wuhan	China	14	1	131	0.76	1	62	1.61				7	35
	Total					93	2238	4.16	87	2000	4.35	6	169	3.55
Diarrhea	2020	Prospective	Isfahan	Iran	28	1	452	0.22	NA	NA		NA	NA		9	50
	2020	Prospective	Wuhan	China	14	1	131	0.76	NA	NA		NA	NA		7	35
	2020	Prospective	Aachen	Germany	56 (48–71)	3	33	9.09	NA	NA		NA	NA		8	47
	2020	Prospective	Wuhan	China	14	4	337	1.19	NA	NA		NA	NA		7	48
	2020	Cohort	Georgia	USA	38 (21–49)	3	26	11.54	NA	NA		NA	NA		8	46
	2020	Prospective	Wuhan	China	??	2	18	11.11	NA	NA		NA	NA		9	51
	2020	Prospective	Wuhan	China	90	20	76	26.32	NA	NA		NA	NA		8	52
	Total					34	1073	3.17
Loss pf appetite	2021	Cohort	Hubei	China	153 (146–160)	138	1655	8.34	127	1538	8.26	11	117	9.40	8	49
	2020	Prospective	Wuhan	China	14	3	131	2.29	1	62	1.61	2	69	2.90	7	35
	2020	Prospective	Isfahan	Iran	28	15	452	3.32	13	400	3.25	2	52	3.85	9	50
	Total					156	2238	6.97	141	2000	7.05	15	238	6.30
Abdominal pain	2020	Prospective	Isfahan	Iran	28	6	452	1.33	NA	NA	NA	NA	NA		9	50
	2020	Cohort	Georgia	USA	38 (21–49)	1	26	3.85	NA	NA	NA	NA	NA		8	46
	2020	Prospective	Aachen	Germany	56 (48–71)	1	33	3.03	NA	NA	NA	NA	NA		8	47
	Total					8	511	1.57
Dysgeusia	2021	Cohort	Hubei	China	153 (146–160)	120	1655	7.25	112	1538	7.28	8	117	6.84	8	49
	2020	Prospective	Fuyang	China	90	1	60	1.67	NA	NA		NA	NA		7	53
	2020	Cohort	Georgia	USA	38 (21–49)	1	26	3.85	NA	NA		NA	NA		8	46
	2020	Prospective	Aachen	Germany	56 (48–71)	3	33	9.09	NA	NA		NA	NA		8	47
	2020	Prospective	Henan	China	90	2	55	3.64	NA	NA		NA	NA		7	54
	2020	Prospective	Tokyo	Japan	108 (±23)	3	58	5.17	NA	NA		NA	NA		7	55
	Total					130	1887	6.89	112	1538	7.28	8	117	6.84

The information about COVID-19 severity on admission and the occurrence of GI symptoms were provided in one article each for nausea ³⁵ and dysgeusia, ⁴⁹ and three articles each for vomiting ^{35 , 49 , 50} and loss of or decrease in appetite. ^{35 , 49 , 50}

Prolonged nausea was reported in 10 patients after recovery, with estimated prevalence of 3.23% (95% CI: 0.54%–16.53%) from a total of 527 patients with COVID-19 from four studies ( Figure 2). Persistent vomiting was identified in 93 of 2,238 patients with COVID-19 from three studies, which corresponded to a pooled prevalence of 3.19% (95% CI: 1.62%–6.17%). Seven articles reported the prevalence of prolonged diarrhea as 4.12% (34/1,073 patients with COVID-19) with 95% CI: 1.07%–14.64% . Loss of or decrease in appetite was reported in three studies that included 2,238 patients with COVID-19, among whom 156 patients were reported to have had the symptom (estimated prevalence of 4.41%, 95% CI: 1.91%–9.94%). Based on three studies, abdominal pain was reported in 8/511 patients with COVID-19 with an estimated prevalence of 1.68% (95% CI: 0.84%–3.32%). Six studies identified 130 cases of prolonged dysgeusia among a total of 1,887 patients with COVID-19 (7.04%, 95% CI: 5.96%–8.30%). Abdominal pain and dysgeusia were analyzed using a fixed-effect model of Egger’s test at p < 0.001. Figure 2.

(A) Estimated prevalence of prolonged nausea in COVID-19 (event rate 3.23%, 95%CI: 0.54%–16.53%, p < 0.001, p Egger 1.683, and p heterogeneity <0.001). (B) Estimated prevalence of prolonged vomiting in COVID-19 patients (event rate 3.19%, 95%CI: 1.62%–6.17%, p = 0.028, p Egger 0.482, and p heterogeneity 0.028. (C) Estimated prevalence of prolonged diarrhea in COVID-19 patients (event rate 4.12%, 95%CI: 1.07%–14.64%, p < 0.001, p Egger 1.726, and p heterogeneity <0.001). (D) Estimated prevalence of prolonged of loss of appetite in COVID-10 patients (event rate 4.41%, 95%CI: 1.91%–9.94%, p < 0.001, p Egger 0.690, and p heterogeneity <0.001. (E) Estimated prevalence of prolonged of abdominal pain in COVID-19 patients (event rate 1.68%, 95%CI: 0.84%–3.32%, p = 0.499, p Egger <0.001, and p heterogeneity 0.499). (F) Estimated prevalence of prolonged of dysgeusia in COVID-19 patients (event rate 7.04%, 95%CI: 5.96%–8.30%, p = 0.526, p Egger <0.0001, and p heterogeneity 0.052).

Owing to the lack of studies presenting prolonged GI symptoms among patients with mild-moderate and severe COVID-19, the associations were calculated only for vomiting and loss of appetite. The severity of COVID-19 was not associated with the presence of either vomiting or loss of appetite in patients with COVID-19 (odds ratio (OR): 1.19, 95% CI: 0.51–2.78 and OR: 0.84, 95% CI: 0.47–1.5, respectively). Both symptoms were analyzed using the fixed-effect model of Egger’s test at p < 0.001.

Studies have confirmed that the existence of SARS-CoV-2 in the GI tract can last for several weeks after a throat swab shows a negative result. ^{56 , 57} One study found that the elimination of SARS-CoV-2 from fecal samples was completed more than a month after samples collected from the respiratory tract turned out negative. ⁵⁸ This may explain the prolonged GI symptoms observed in patients with COVID-19 in the present study.

To date, no satisfying explanation is available as to why the virus lasts longer in the gut than in the other systems. Although the exact mechanisms are not fully elucidated, some putative pathophysiological mechanisms underlying the occurrence of COVID-19-induced GI symptoms have been put forward. These potential mechanisms encompass the direct invasion of SARS-CoV-2 in GI cells, secondary effects after other organs are infected, and drug treatment-induced digestive symptoms. ^{59 , 60}

It has been shown that angiotensin-converting enzyme 2 (ACE2), the entry receptor of SARS-CoV-2, is also highly expressed in the digestive system organs such as the esophagus, small intestine, and colon. ^{61 , 62} Therefore, it is quite plausible to observe the advent of several GI symptoms induced by SARS-CoV-2 infection in patients ranging from nausea, vomiting, and diarrhea to loss of appetite and abdominal pain. ^{59 , 60 , 63} The expression of the ACE2 receptor in the gut results in the digestive system also being vulnerable to attack by SARS-CoV-2. After being occupied by the virus, ACE2 becomes dysfunctional, resulting in the impairment of the protective activity of the ACE2/Ang-(1–7)/Mas axis, whereas the activity of the ACE/Ang II/AT1R axis is elevated. ⁶⁴ Following this condition, nicotinamide adenine dinucleotide phosphate oxidases are excessively activated leading to the occurrence of oxidative stress-induced inflammation and this finally causes tissue damage as described in Figure 3. ^{65 , 66} Figure 3.

The inactive Ang I is converted into Ang II which produces its biological activities via its binding to AT1R. To maintain the homeostasis status, the catalytic activity of ACE2 converts Ang II to Ang 1-7 which has the opposite action of Ang II through its binding to Mas receptor. The invasion of SARS-CoV-2 to ACE2 causes an accumulation of Ang II and decreased level of Ang 1-7. This condition is linked to the increased interaction between Ang II and its receptor AT1R resulting in the hyperactivity of NADPH oxidase which is related to oxidative stress, massive production of proinflammatory cytokines, increased activity of thrombogenic factors and vasoconstriction. These events are eventually associated with the emergence of the COVID-19 GIT symptoms.

The interrelation between the digestive system and other systems or organs inspires another proposed mechanism underlying the emergence of GI symptoms during COVID-19 infection ( Figure 5). As SARS-CoV-2 is not always detected in the feces of patients with COVID-19 who are also displaying GI symptoms, a study speculated that the symptoms are not usually linked to the direct invasion of the virus into the intestines. ⁵⁹ For example, through the gut-lung axis, dysbiosis in the gut is putatively linked to the disturbances in the respiratory flora and vice versa. ^{59 , 82 , 83} It has been reported that the level of lung-derived C-C chemokine receptor type 9 (CCR9), a chemokine receptor required by CD4 ⁺ to move to the small intestine, increases during respiratory influenza virus infection. ⁸⁴ The movement of CCR9-CD4 ⁺ T cells to the intestine is promoted by the high abundant expression of CCL25 in the small intestine. ^{59 , 85} CCL25/CCR9 is found to have a critical responsibility in directing the recruitment of lymphocytes to the small intestine which is subsequently followed by the disruption of intestinal flora homeostasis. ^{59 , 86} Figure 5.

GIT symptoms during the COVID-19 course can be influenced by the virus infections in several sites such as in the CNS, lung and liver. The infections occurring in those organs can be sensed by the GIT leading to the emergence of the symptoms. Several axes are involved in this interrelationship such as the gut-brain axis which is mainly mediated by the enteric nervous system, gut-lung axis involving the movement of CCR9-CD4+ from the lung to the intestine driven by the CCL25, and gut-liver axis which is connected through the portal vein and biliary tract.

The various drugs administered during the COVID-19 course could be linked to the emergence of GI symptoms as adverse reactions. As shown in antimicrobial agents, antiviral agents can also change the steady-state level of the gut microbiota which can cause diarrhea. ⁵⁹ Some antibiotics, such as cephalosporins, penicillins, quinolones, and macrolides are known to induce diarrhea when they are used to treat infections. ^{59 , 98} During COVID-19 infection, the use of these antibiotics is common, and has been correlated with the increased number of drug-induced diarrhea cases. ^{59 , 99 , 100} In general, pharmacological agents can cause diarrhea via a number of mechanisms such as by disturbing the normal flora that reside in the gut, promoting the growth of pathogenic microbes, inducing allergic or toxic reactions in the intestinal mucosa, or by stimulating the motility of the gut. ¹⁰¹ In particular, the use of broad-spectrum antibiotics such as penicillins and cephalosporins, is found to be one of the causes of Clostridium difficile hegemony over the normal intestinal microbiota. ^{102 , 103} This occurs as the antimicrobial agents could kill the flora while leaving the pathogenic microorganisms without control from the normal flora. ¹⁰¹

Several antiviral agents are also reported to have GI symptoms as their adverse effects when administered to patients with COVID-19. The use of remdesivir, lopinavir, and ritonavir was found to induce nausea and vomiting. ¹⁰⁴ The increased level of noxious chemicals, including drugs, in the GI tract could send a signal to the vomiting center in the CNS through afferent fibers of the glossopharyngeal and vagal nerves to induce emesis. ¹⁰⁵

All data underlying the results are available as part of the article and no additional source data are required.

Figshare: PRISMA checklist for ‘Global prevalence of prolonged gastrointestinal symptoms in COVID-19 survivors and potential pathogenesis: A systematic review and meta-analysis’, https://doi.org/10.6084/m9.figshare.14083613. ¹⁰⁶

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC BY 4.0).