A systematic review and meta‐analysis of prevalence and clinical features of upper gastrointestinal (UGI) tract Crohn's disease in adults compared to non‐UGI types

Abstract Background and Aim Crohn's disease is an inflammatory condition that affects the gastrointestinal (GI) system. This study aimed to determine the prevalence of upper gastrointestinal Crohn's disease (UGICD) and compare its features to non‐UGICD types. Methods We conducted a systematic search in the databases PubMed, Web of Science, Scopus, and Google Scholar. The heterogeneity of prevalence estimates was examined, subgroup analyses were carried out, and meta‐analyses were conducted using random‐effects modeling. Prognostic data were qualitatively reviewed and combined. Results Two‐thousand nine‐hundred and forty studies were retrieved and 32 studies were included in the final analysis. Pooled prevalence of UGICD was 15% (CI: 11–18%) among 14 509 patients. UGICD prevalence did not show any significant increase with time (P = 0.45). The most prevalent (38%, CI: 30–47%) behavior of UGICD was B1 (nonstricturing‐nonpenetrating), while the most common concurrent location was L3 (ileocolon) with a prevalence of 47% (CI: 34–59%). UGICD patients had higher stricturing phenotype (B2) compared to non‐UGICD (0.38 vs 0.30; P = 0.03). There was no significant difference in the prevalence of UGICD between patients classified according to the Montreal or Vienna classification. Stricturing phenotype was more common among Asian patients compared to Western patients (0.44 vs 0.24; P < 0.001). UGICD was a risk factor for surgery and drug therapy and was associated with an aggressive course of the disease and more resections. Pooled prevalence of UGICD was 15%. Conclusion Nonstricturing‐nonpenetrating type was the most prevalent UGICD. UGICD patients had more complications and worse outcomes compared to non‐UGICD patients.


Introduction
Crohn's disease is an inflammatory bowel disease that commonly affects the distal ileum and colon. However, the distribution of lesions can be in various parts of the gastrointestinal (GI) tract, from mouth to anus. 1 Upper GI tract involvement is an infrequent disposition of the disease. 2 In recent decades, because of the advances in diagnostic tools and imaging technologies, the upper gastrointestinal Chron's disease (UGICD) has been more recognized. 3,4 According to Zallot et al. UGICD has a worse outcome and is associated with more hospitalizations and surgeries. 5 Several studies on the epidemiological characteristics of UGICD have been published in recent years. [6][7][8] Although Chin et al. 9 conducted a systematic review on the prevalence of UGICD, its risk factors, and outcomes, they failed to include almost all of the available documents. They searched for reports only in PubMed and Embase databases and ignored possible gray literature available and other less Westernized databases such as Scopus and Google Scholar. Accordingly, their study results may not be adequately robust. This systematic review aimed to determine the prevalence of UGICD and compare its features and outcomes with those of non-UGICD.

Methods
Inclusion and exclusion criteria. A systematic review and meta-analysis was designed and performed. We conducted a systematic search for available documents following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.
We searched for original documents on the prevalence, epidemiology, prognosis, clinical management, as well as its clinical features including location, behavior, and phenotype. Original studies with any observational design were included.
Information source and search strategy. Based on the search technique used in the study, searches were conducted systematically. Scopus, PubMed, and Web of Science were searched online for published and indexed documents. To find online gray literature, Google Scholar was used. We also manually searched the list of references cited in the selected documents. All online searches were conducted on November 22, 2021. An online search strategy was designed using the following keywords: "upper gastrointestinal tract" AND "Crohn's disease"; "epidemiology"; "clinical management"; outcomes; and "clinical features." The tools offered by each database were used to combine these keywords and their synonyms. The primary search strategy is described in the online Supplementary Material. Selection process. After retrieving the results of the searches and removing duplicate documents, the screening process was carried out independently by two authors according to prespecified criteria. If no consensus was available based on the document title and abstract, reviewers read the whole text, and in case of any disagreements about the inclusion of a document, the case was discussed by a methodologist.
Risk-of-bias assessment. Risk-of-bias assessment was done by two authors independently. The NOS scale (Newcastle-Ottawa scale) 10,11 designed specifically for observational studies was used for critical appraisal. Epidemiologic features were assessed for possible bias through seven different questions, and any disagreements were discussed by a third author.
Data items. Relevant data from the included documents were extracted into a piloted data form comprising the date of publication, first author name, geographical region (Western or Asian), country, period of the study, number of patients, family history of patients, female percentage, as well as disease location and phenotype according to Montreal 12 or Vienna 13 classification, concurrent location of upper gastrointestinal (L4) disease, diagnostic criteria for Crohn's disease, type of study, risk factors for UGICD, complications, and clinical or surgical management of the disease.
Statistical methods. Pooled prevalence values for L1 (terminal ileum), L2 (colon), L3 (ileocolon), P (perianal), L4 (UGICD), and only L4 location (only UGI involvement) and the following B1 (nonstricturing-nonpenetrating), B2 (stricturing), and B3 behavior (penetrating) were estimated by applying random-effects modeling and illustrated in forest plots. Heterogeneity was assessed using the I 2 and Cochrane Q tests. Heterogeneity was defined as low (I 2 < 25%Þ, moderate (I 2 : 25 -75%Þ, and significant (I 2 > 75%Þ. Subgroup analyses were carried out considering the region, classification, and diagnostic methodology applying meta-regression, and potential correlates of heterogeneity including female percent, smoker percent, and subgroup analyses were investigated. Forest plots were used to illustrate subgroup analysis based on region and involvement of the UGI tract (Supplementary Material-Forest plot). The Egger test was conducted and the funnel plots were used for the assessment of publication bias (Supplementary Material-Egger test and funnel plots). Data analyses were performed using Stata (StataCorp LP, College Station, TX, USA). Where meta-analysis of the extracted data was not possible, qualitative synthesis was performed to summarize the information.

Results
The flow diagram of the study was developed according to the PRISMA method 14 (Fig. 1).
According to the flow diagram of the study, from the 2940 initial citations that were retrieved from the databases, 170 studies were selected for full-text review and from those 170 papers, 32 essays with 35 Crohn's population were included in the final paper. The following three studies had two different Crohn's population bases: the Spekhorst study, 15 the Jess study, 16 and the Farkas study. 17 Thirty studies were included in the meta-analysis. Among the included papers, 21 (including 22 populations) studies were retrospective, 6 studies (including 8 populations) were prospective, and 1 study was cross-sectional (Table 1). Eight papers were not accessible due to the access policies of the journal, so we emailed the authors for the full text, but the papers were not made available, thus they were excluded from the study.
We had two overall populations of Crohn's patients in our investigation. The first population comprised all the Crohn's patients who were studied in the included papers, and the second population (the L4 population) consisted of Crohn's patients with upper GI involvement.
Comparison between the West and Asia. Fourteen studies including 16 populations (the Spephorst 15 and Jess 16 study had two populations) were from Western countries and 14 were from Asia. The pooled prevalence of UGICD in Asian patients was higher than in Western patients (0.18 [CI: 0.10-0.27] vs 0.12 [CI; 0.09-0.15]) though not significant (Supplementary Material-whole population data). The only significant difference between the pooled prevalence of measured parameters in Western and Asian patients with UGICD was the fact that the stricturing phenotype (B2) was more prevalent among Asian patients (0.44 vs 0.24; P < 0.001) ( Table 4). There was no significant difference in the characteristics of UGICD patients, namely the prevalence of the disease, patient's age, phenotype, and concurrent location of the disease, between studies classified according to Montreal and Vienna classifications (Table 5).
Comparison between L4 and non-L4 populations. The L4 population comprised Crohn's patients with upper GI involvement and the non-L4 population consisted of Crohn's patients without upper GI involvement. The stricturing phenotype (B2) was significantly more common in the L4 population compared to the non-L4 population (0.38 vs 0.30; P = 0.03); also, isolated colon involvement (L2) was significantly higher among the non-L4 papulation (0.25 vs 0.13; P = 0.01) ( Table 6).
L4 trend. There was no significant relationship between L4 pooled prevalence and time (P = 0.45).
Meta-regression L1. Although L1 prevalence did not differ significantly between L4 and non-L4 in simple subgroup analysis, according to data in meta-regression, all the three variables, namely female percent, smoker percent, and subgroup (analysis between L4 and non-L4 group), significantly affect the heterogeneity (differences) of L1 prevalence between L4 and non-L4 population (subgroup P = 0.02; smoker percent P = 0.01; female percent P <0.001).
L2. The effect of the subgroup was not significant on the heterogeneity of L2 prevalence (P = 0.45) but the smoker percent (P = 0.02) and female percent (P < 0.001) were both significantly correlated with the heterogeneity of measured L2 prevalence. (In subgroup analysis, the subgroup effect was significant, but when these three variables were analyzed together, the significance of the subgroup effect disappeared).
L3. When the variables smoker percent, female percent, and subgroup were analyzed simultaneously, all three of them (smoker P < 0.001, female percent P < 0.001, subgroup P < 0.001) were  significantly effective on the heterogeneity of L3 prevalence between L4 and non-L4 population.
B1. Also, all the mentioned variables were significantly effective in the heterogeneity of B1 prevalence (smoker percent P = 0.00; female percent P < 0.001; subgroup P = 0.05).
B2. In B2, the only effective variable on the heterogeneity of prevalence was female percent (female percent P < 0.001, subgroup P = 0. 93, smoker P = 0.59).
B3. Subgroup was the only variable significantly responsible for making a difference in the heterogeneity of B3 prevalence between the L4 and non-L4 populations. (In simple subgroup analysis, the subgroup was not effective in the heterogeneity of B3 prevalence, but after controlling the confounder variables, the subgroup effect became significant: subgroup prevalence = 0.05; female prevalence = 0.72; smoking prevalence = 0.35.) P. In meta-regression, none of the variables, namely female percent, smoker percent, and subgroup, was correlated with the heterogeneity of prevalence between L4 and non-L4 (see Table 7).
Outcome. All the relevant qualitative data on UGICD outcome and its comparison to non-UGICD are given in Table 8.
Risk factors. All the relevant qualitative data on UGICD risk factors are given in Table 9.

Discussion
This study is a systematic review and meta-analysis conducted on UGICD prevalence and its risk factors and outcome. Thirty studies comprising 33 Crohn's populations with 17 071 patients were included in the meta-analysis. The pooled prevalence of UGICD was 15% (CI: 11-18%) in 14 509 patients across 32 populations, consistent with a 13% pooled prevalence of UGICD in the Chin and colleagues 0 study, 9 which is a previous systematic review conducted on this subject in 2021.
Although the L4 (upper GI involvement) prevalence trend did not show any significant evidence of increase with time (P = 0.45), studies after 2019 had a nonsignificantly higher L4 prevalence (15% in studies between 2015 and 2019 vs 24% in studies after 2019) and also pooled prevalence of UGICD based on the Montreal classification; the newer classification was nonsignificantly higher (P = 0.32) (16 vs 12%) in contrast to the Vienna classification. This late increase in L4 prevalence can be due to the limitations of our review process. Another hypothesis is that recent advances in diagnostic and imaging technologies are leading to more accurate detection of UGI involvement in Crohn's patients; however, to the best of our knowledge, there is no convincing study on the changes in UGICD prevalence rate in  Comparison of upper gastrointestinal tract Crohn's disease (UGICD) and Non-UGICD in adults B Tamizifar et al.  recent years, and our findings need to be verified in future studies. There was no significant difference in measured parameters such as age, phenotype, and location of the disease between UGICD patients in studies classified according to the Montreal or Vienna classification. The pooled prevalence of UGICD in Asian and Western countries did not have any significant difference; however, the prevalence of upper GI involvement in the Asian population was higher (18 vs 12%; P = 0.17). Similar findings regarding the higher prevalence of UGICD in Asia have also been reported in Chin and colleagues' review 9 and they are in accordance with the rapid increase in the incidence and prevalence of inflammatory bowel disease (IBD) in Asian countries over the recent decades. 45,46 The only significant difference between the pooled prevalence of measured parameters between Western and Asian patients with UGICD was the fact that stricturing phenotype (B2) was more common among Asian patients (0.44 vs 0.24; P < 0.001). This finding is congruent with Ng and colleague's large-scale population-based study, which found that complicated CD (stricturing, penetrating, or perianal disease) was more common in Asia than Australia (52 vs 24%; P = 0.001). According to their result, IBD can be as severe or more severe in Asia than in the West. 47 Based on the results, UGICD patients compared to non-UGICD patients had more stricturing behavior (B1) (0.38 vs 0.30; P = 0.03), which is in agreement with our overall prognostic data in which UGI involvement in Crohn's patients is a risk factor for more surgeries and resection therapies. 31,37,41 The pooled prevalence of isolated colon involvement (L2) was  In subgroup analysis, comparisons between L4 and non-L4 populations based on the mentioned variables in Table 7 were carried out. The L4 population was defined as Crohn's patients with upper GI involvement and non-L4 was defined as Crohn's patients without upper GI involvement.
Comparison of upper gastrointestinal tract Crohn's disease (UGICD) and Non-UGICD in adults B Tamizifar et al.
significantly higher among patients without UGI involvement (0.25 vs 0.13; P = 0.01). This could be due to the presence of the more extensive ileocolonic form of the disease with small bowel involvement in UGICD, as suggested in the Park and Kim study, which was included our studies in the qualitative data. 43,44 Male sex was a common risk factor in the two included studies on the risk factors for the occurrence of UGI involvement, 26,40 Yet, since there is not adequate data on the subject, the interpretation is inconclusive.
According to the result of meta-regression, based on the female percent, smoker percent, and L4 and non-L4 subgroup analysis, female percent was significantly effective in the heterogeneity of pooled prevalence of L1, L2, L3, B1, and B2 between the two population of L4 and non-L4. The smoker percent and the subgroup analysis significantly affect the heterogeneity of pooled prevalence of L1, L2, L3, B1, and L1, L3, respectively. To the best of our knowledge, there is no data available in this regard and thus our findings need to be verified in future studies.
In the qualitative segment of the study (Tables 8 and 9), the studies of Kim et al. 44 Lazarev et al. 41 and Park et al. 43 compared Crohn's disease with and without small bowel involvement. Lazarev and colleagues 41 in a cross-sectional study of the  31 2019 L4 disease was an independent risk factor for abdominal surgery within 1 year post diagnosis (OR 6.335; 95% CI 3.862-10.390) and was associated with higher rates of abdominal surgery (41.3% vs 11.4%, P < 0.001) but similar rates of hospitalization within 1 year post-diagnosis compared to non-L4 disease. Jejunal and proximal ileal involvement was associated with stricturing behavior (P = 0.034, P < 0.001) and a higher abdominal surgery rate (both P < 0.001). Early outcomes are worse for L4 than for non-L4 disease thus more aggressive initial therapy is needed to improve L4-disease prognosis. Mao et al. 33 2018 Compared to non-L4 patients, L4 patients were more likely to have intestinal surgeries during follow-up (31% vs 16%, P < 0.001). L4-jejunal (HR 3.08; 95% CI 1.30-7.31) and L4-proximal ileal disease (HR 1.83; 95% CI 1.07-3.15) were independent predictors for intestinal resection in multi-variable analysis. Kim et al. 44 2018 compared to patients without proximal small bowel involvement, those with small bowel involvement: Were more likely to display stricturing behavior (19.8% vs 12.7%, P = 0.020).
Had more upper gastrointestinal involvement (OR, 1.643; 95% CI: 1.008-2.677) and lower surgery-free survival (10-year surgery-free survival: 58.4% vs 67.7%, respectively, P < 0.001) Park 43 2013 Jejunal group in comparison to non-jejunal group: Had more ileal location (28.3 vs 20.6%, P < 0.001) and stricturing behavior (16.7 vs 9.4%, P = 0.001). In univariate analyses, had significantly higher cumulative probabilities of treatment with corticosteroids (P = 0.014) and thiopurines (P = 0.008), the first major surgery (P = 0.021), and the first hospitalization (P = 0.015) In multivariate analyses, had greater use of corticosteroids (hazard ratio, IBD genetic consortium compared jejunal with non-jejunal Crohn's disease, and according to their results, jejunal involvement is a significantly greater risk factor for stricturing behavior and multiple abdominal surgeries. Park and coworkers 43 in a retrospective study in Korea compared the prognosis of two groups of jejunal and non-jejunal Crohn's patients. According to their report, jejunal patients had more ileal location and stricturing behavior than the non-jejunal patients. In their study, on both univariate and multivariate analysis, jejunal involvement was independently associated with more corticosteroid and thiopurine use, more surgery, and more hospitalizations. Kim and their colleagues 44 in the connect cohort study compared Crohn's patients with and without proximal small bowel involvement. According to their results, patients with small bowel involvement had poorer surgery-free survival and more upper GI involvement. All three mentioned studies maintain the fact that the involvement of small bowel disease, and particularly jejunum, had a poor prognosis and worse outcome in contrast to Crohn's patients without small bowel involvement. Regarding the correspondence of clinical features including prognosis and outcome between UGICD and non-UGICD, the studies of Louis et al. 38 Chow et al. 19 Mao et al. 33 and Sun et al. 31 demonstrate that UGICD in contrast to non-UGICD was associated with significantly more surgeries; besides, the studies of de Barros et al. 37 Lazarev et al. 41 and Sun et al. 31 suggest that UGICD patients had more stricturing behavior in contrast to non-UGICD patients. Furthermore, studies of Chow et al. 19 and Park et al. 43 found that UGICD patients had a longer duration of hospitalization than non-UGICD patients. However, according to Sun et al. 31 and Greteur et al. 40 there were no significant differences between UGICD and non-UGICD disease courses. According to Sun and colleagues, 31 UGICD patients had similar rates of hospitalization compared to non-UGICD patients, and according to the Greteur et al. 40 study, PCD (proximal Crohn's disease) group had a disease course similar to that of the non-PCD group of patients. However, the findings of these last two works are in contrast with those of the rest of the included studies in the review; nevertheless, seven of nine papers included in our study maintain that UGICD was associated with more surgeries, complications, and worse prognosis. Thus the overall result of the included studies favor the conclusion that UGICD is a risk factor for surgery, resection, and drug therapy and has a more aggressive disease course.
In contrast to Chin's study, this paper has conducted the first review in three databases, namely the Web of Science, Scopus, and Google Scholar, and included six (23%) more reports with nine (34%) more population bases of UGICD, more comprehensive literature search, and a convincing number of eligible reports, thus increasing the validity of our results and making it more principled; however, the overall findings of our study are supportive of their results. Besides that, this review has three distinct populations of Crohn's patients with their measured demographic characteristics according to Montreal and Vienna classifications: The first and the main population is the L4 or UGICD population, and the second and the last populations are the non-L4 and the whole populations, respectively. We have carried out two comparisons: comparison of demographic characteristics between the L4 and non-L4 populations, and a meta-regression analysis between the two populations of L4 and non-L4 based on the variables female percent, smoking, and subgroup analysis; however, due to the originality of this data, the conclusions of this section need to be verified in future original studies. And our qualitative findings in Tables 8 and 9 regarding the outcome and risk factor of UGICD patients due to limitations of the qualitative review process have to be appraised more in future studies.

Conclusions
In light of the 15% prevalence of UGICD according to our analysis, clinicians should be aware of possible upper GI involvement and conduct required investigations in the early phase of the disease to prevent complications, if necessary.

Ethics approval
Committee of Medical Ethics, Isfahan Medical Institute, approved our study. Our ethical code is IR.ARI.MUI.REC.1400.095.
Data availability statement. Data and materials of our cases are available in Supplementary Material.