Systematic Review and Meta-Analysis: What is the Prevalence of Non-Malignant, Organic Gastrointestinal Disorders Misdiagnosed as Lrritable Bowel Syndrome?

may be overlooked. The prevalence of ve gastrointestinal conditions - bile acid diarrhoea (BAD), carbohydrate malabsorption (CM), microscopic colitis (MC), pancreatic exocrine insuciency (PEI) and small intestinal bacterial overgrowth (SIBO) was systematically assessed from studies including consecutive patients meeting diagnostic criteria for IBS. Methods 4 databases were searched from 1978-2020. Studies were included if they evaluated the prevalence of these conditions in secondary healthcare setting. Estimated pooled rates were calculated and statistical heterogeneity between studies was evaluated using Q and I 2 statistics. 17 studies (n=5,068) estimated that of MC as 3% (95% CI 2-4%). Two studies (n=478) suggested a rate of 4·6% (range: 1·8-6·1%) for PEI. Using breath testing, 26 studies (n=6,700) and 13 studies (n=3,415) estimated the prevalence of lactose and fructose malabsorption as 54% (95% CI 44-64%) and 43% (95% CI 23-62%); 36 studies (n=4,630) and 22 studies (n=2,149) estimated that of SIBO as 49% (95% CI 40-57%) with lactulose and 19% (95% CI 13-27%) with glucose. Rates of all conditions were signicantly higher than in healthy controls.


Introduction
Irritable bowel syndrome (IBS) is de ned by a symptom cluster. Cluster content has changed over time but includes recurrent, defecation-associated abdominal pain and change in bowel habit. There is no biological marker for IBS [1]. Patients with IBS-like symptoms are the single largest group of patients presenting with gastrointestinal (GI) complaints to both primary and secondary healthcare [2,3]. IBS has been estimated to affect at least 7-21% of the global adult population [4].
Terms such as 'irritable colon syndrome', 'spastic colon', 'painless diarrhoea' and 'mucus colitis' were used until the mid-20th century, with classi cation de ned subjectively by clinical experience [5][6][7][8]. The Manning and Kruis criteria were developed, in 1978 and 1984 respectively, to rationalise the use of investigations and minimise surgery prior to diagnosis of IBS [7,8]. Neither of these criteria was adequate for clinical practice and they were superseded by the consensus-based Rome criteria [9]. One aim of such criteria was to promote a positive diagnosis of functional GI disorder where no biological or structural cause could be identi ed [10]. This standardised de nition has facilitated research into the underlying pathophysiology of IBS but successful translation into therapy has thus far been limited. The Rome criteria have been developed using community populations rather than patients selected for secondary care referral and assessment.
The current guidelines recommend that patients can be diagnosed with IBS, in the absence of 'red ags', when: their symptoms ful l the Rome IV criteria for IBS; their full blood count, coeliac serology, C reactive protein and erythrocyte sedimentation rate are normal; and they have a faecal calprotectin level of < 50 µg/g [11,12]. However, many physicians nd these diagnostic criteria too restrictive and unclear [13].
Additionally, this diagnostic approach has only modest ability in differentiating IBS from other diseases [14,15]. This may explain the low proportion of patients responding to treatments for IBS [16].
Gastroenterologists are aware of coeliac disease, in ammatory bowel disease and colorectal cancer as possible confounding diagnoses in up to 8·6% of patients who meet clinically de ned criteria of IBS [17][18][19][20][21][22]. These conditions are often associated with alarming symptoms which prompt further investigations, allowing them to be diagnosed. In this systematic review we aimed to clarify the prevalence of ve other, often overlooked, conditions -bile acid diarrhoea (BAD), carbohydrate malabsorption (CM), microscopic colitis (MC), pancreatic exocrine insu ciency (PEI) and small intestinal bacterial overgrowth (SIBO) in adults presenting to secondary care with IBS-like symptoms.

Search strategy
This systematic review and meta-analysis was performed in accordance with PRISMA recommendations.
The PRISMA checklist for this study can be found in the Supplementary Material. The literature search was conducted using CINAHL, Cochrane, Embase and Pubmed from 1978, the year of publication of the Manning criteria, to November 2020. Literature searches were conducted individually for 5 conditions. The terms 'irritable bowel syndrome' OR 'irritable colon' OR 'IBS' OR 'functional bowel' OR 'diarrhoea' OR 'diarrhea' were used as medical subject heading (MeSH) and free text terms for each condition.
These were combined using the set operator AND with following terms:

Study selection
Studies were included if they assessed prevalence of any of these conditions in consecutive patients meeting the Manning, Kruis or Rome I-IV criteria system in a secondary healthcare setting. Studies whose primary aim was not evaluating prevalence of the conditions, but presented these data, were also included. Studies with zero cases were also included in the analysis. Important reasons to exclude studies were: primary care studies; lack of prevalence data; no clear de nitions of IBS; inclusion of individuals with any other known organic GI diseases (coeliac disease, in ammatory disease, etc) or previous abdominal surgery; only reported as conference abstracts with no full text available.
The titles of all papers identi ed from the literature searches were scanned and abstracts of potentially relevant papers were reviewed. The full text of any papers meeting the inclusion criteria were obtained. Where data were missing within the eligible studies, the authors were contacted. Studies were assessed using the pre-de ned eligibility criteria independently by two reviewers (DP and HJNA). Any discrepancies were resolved by consensus.
The protocol used for this systematic review was registered with PROSPERO, number CRD 42019145486.
Quality assessment of the individual studies Quality and risk of bias of the included studies were analysed independently by two reviewers (DP and HJNA) using Critical Appraisal Skills Programme (CASP) checklist for cohort studies [23].
We were unable to identify a single appraisal tool which was applicable to all studies. Some items of this checklist were not applicable to the non-cohort studies that presented prevalence data. We modi ed the checklist by removing items 6, 7 and 8 regarding length of follow-up and precision of study results. This led to a maximum score of ten points, one for each item on the checklist. A score ≤ 7 points indicated high risk of bias. Appraisal for the included studies can be found in Supplementary Tables 1-5. BAD -75-selenium homocholic acid taurine (SeHCAT) scan with 7-day retention < 15% (severe BAD is de ned as SeHCAT < 5%; moderate as SeHCAT < 10%; mild as SeHCAT < 15%) [24], or -total 48-hour faecal bile acids > 2,337 µmol [25], or -an elevated fasting 7α-hydroxy-4-cholesten-3-one (7α-C4) level CM -a positive lactose, fructose, sucrose, sorbitol or mannitol hydrogen breath test, or -genotyping for adult-type hypolactasia leading to lactose intolerance MC -abnormal histological ndings on colonic biopsies meeting criteria for lymphocytic or collagenous colitis PEI -faecal elastase-1 level < 200 µg/g [26] SIBO Page 6/26 -a positive lactulose or glucose hydrogen breath test, or -a positive bacterial culture or a bacterial count of > 10 3 colony-forming units (cfu) per millilitre in the small intestinal aspirate [27] The total number of patients with IBS were extracted from each study (denominator), followed by the number of those patients diagnosed with an organic condition using one of the recognised diagnostic tests above (numerator). The prevalence was expressed as the proportion of patients meeting IBS criteria, who underwent one of the above tests, had an organic condition.
Additionally, subgroup analyses for each condition were performed incorporating: only studies with a sample size ≥ 100; studies utilising Rome as the diagnostic criteria for IBS; and prospective studies. In addition, we included the study location and speci c aspects of the tests or criteria used for diagnosing the organic condition, e.g. dose of test substrate used in breath testing for CM and SIBO; number of lymphocytes per 100 epithelial cells to de ne lymphocytic colitis; or thickness of sub-epithelial collagen to de ne collagenous colitis on histological examination.
All subgroup analyses can be found in the Supplementary Material.

Data synthesis and statistical analysis
We performed statistical analysis on the extracted data using Stata SE 16 (StataCorp LLC, USA). Crude pooled prevalence for each condition was obtained by combining, from all studies, the prevalence proportions as described above. Each condition is diagnosed using different tests so prevalence was presented individually, based on the diagnostic modality used. Estimated pooled rates with 95% con dence intervals calculated using random effects with DerSimonian-Laird method [28] were presented in forest plots. Statistical heterogeneity between studies was evaluated using Q and I 2 statistics [29], with I 2 > 50% indicating signi cant heterogeneity [30]. Subgroup analysis by study characteristics was performed to explore any heterogeneity observed. When there were ten or more studies in the metaanalysis, we undertook a general inspection of funnel plots to assess for small-study effects and publication bias [31,32].

Search results
Bile Acid Diarrhoea (Bad) 3,173 citations were identi ed. Of these, 94 were potentially relevant and the full text of these citations were retrieved for full evaluation. 15 papers were included, 79 papers did not satisfy the eligibility criteria ( Supplementary Fig. 1).
Nine studies reported the prevalence of BAD based upon an abnormal SeHCAT scan. Two studies used elevated faecal bile acids over 48 hours and one study used an elevated serum 7α-C4 level. The remaining three studies used two diagnostic tests (SeHCAT scan and 7α-C4 level, total 48-hour faecal bile acids and 7α-C4 level).

Prevalence of a positive SeHCAT scan in patients ful lling IBS criteria
The ten studies that reported the prevalence of BAD based upon an abnormal SeHCAT study are shown in Supplementary Table 1. Across the ten studies, seven different 7-day SeHCAT retention cut-off values (i.e. <5%, < 10%, < 11%, < 11·7%, < 15%,) were used. The two studies which had reported values within 2% of the < 10% cut-off (i.e. <8% to < 12%) were included in the prevalence calculation of the 7-day SeHCAT < 10% subgroup. Six papers [33][34][35][36][37][38] demonstrated the change in their rates of positive scans by using different cut-off values but only one [36] assessed treatment response to cholestyramine by comparing with placebo. Nevertheless, a previous systematic review showed that the response rate to bile acid sequestrants was higher in patients with severe BAD than those with mild or moderate disease [24]. One study [38]evaluated and compared the prevalence of BAD with different SeHCAT retention values as a result of using the Rome III and Rome IV criteria, respectively, to de ne their patients.
Carbohydrate Malabsorption (Cm) 875 citations were identi ed. Of these, 97 appeared to be potentially relevant and the full texts of these citations were retrieved for full evaluation. 39 were included and 58 papers did not satisfy the eligibility criteria ( Supplementary Fig. 7).
The carbohydrates studied included lactose, fructose, sorbitol and mannitol. Using hydrogen breath testing, 28 studies evaluated lactose malabsorption and/or intolerance; 12 examined fructose malabsorption and/or intolerance and six examined alterative or mixed forms of CM. Five groups utilised genotyping studies to identify lactase de ciency. Several of these studies reported prevalence of more than one form of CM.
Five studies including 970 patients applied genotyping studies to identify the homozygous state of a genetic variant, C/C-13910, for adult-type hypolactasia, also known as lactose intolerance (Supplementary Table 10). Two of these studies also evaluated another genetic variant, G/G-22018, which has also been shown to be associated with lactase de ciency, leading to lactose intolerance. The crude pooled rate for lactose intolerance as a result of the C/C-13910 gene was 37·9% (range 15·1-100·0%). The estimated pooled rate was 62% (95% CI 27-91%, Supplementary Fig. 11) by the random effects model, with signi cant heterogeneity in the effect sizes (Q-test X 2 = 412·2, P < 0·0001; I 2 = 99·0%).

Prevalence of alternative or mixed forms of CM in patients ful lling IBS criteria
Six studies shown (Supplementary Table 12) reported the rates of alternative or mixed forms of CM.
Two groups reported that 31·4% and 70·2% of their patients ful lling IBS criteria had a positive breath test for combined fructose-sorbitol malabsorption. Another group concluded that 9·4% of patients had either lactose or combined fructose-sorbitol malabsorption. One of the groups also reported that 20% of patients also had mannitol malabsorption.

Microscopic Colitis (Mc)
The search identi ed 1,484 citations. Of these, 91 were potentially relevant and full texts of these citations were retrieved. 17 papers were included, 74 did not satisfy our eligibility criteria (Supplementary Fig. 18 and Table 13). Four of these 17 studies reported the prevalence of MC but did not specify the subtype, one study only reported patients with lymphocytic colitis. The remaining 12 studies reported the prevalence of both lymphocytic and collagenous colitis, de ned by a con rmed histological diagnosis (> 15 or > 20 lymphocytes in 100 epithelial cells for lymphocytic colitis and > 10 µm or > 15 µm thickened sub-epithelial collagen band for collagenous colitis).

Pancreatic Exocrine Insu ciency (Pei)
The search strategy identi ed 314 citations. Of these, 11 appeared to be potentially relevant and the full text of these citations were retrieved. Two papers were included, nine did not satisfy our eligibility criteria ( Supplementary Fig. 24). Both papers utilised faecal elastase-1 to diagnose PEI.

Prevalence Of Pei In Patients Ful lling Ibs Criteria
A meta-analysis was not performed due to the number of papers identi ed. The two papers gave a crude rate of 4·6% (22/478), ranging between 1·8% and 6·1% (Supplementary Table 14).

Small Intestinal Bacteria Overgrowth (Sibo)
The search strategy identi ed 1,674 citations. Of these, 114 appeared to be potentially relevant and the full text of these citations were retrieved. 59 papers were included, 55 did not satisfy our eligibility criteria ( Supplementary Fig. 25). 55 studies evaluated the prevalence of SIBO within their IBS patient cohorts utilising breath testing with glucose and lactulose. Three of these studies, alongside an additional four studies, examined small bowel uid from their patients to identify those with SIBO (Supplementary Table 15).

Discussion
This systematic review brings together data on multiple conditions other than coeliac disease, in ammatory bowel disease or colorectal cancer that may be present in adults ful lling clinical criteria for IBS. These data show a high frequency of treatable gastrointestinal diseases in this patient group. Approximately 40% had BAD; CM was often detected − 54% with lactose malabsorption and 44% with fructose malabsorption; 5% had PEI; 3% had MC; and 13-49% had SIBO. Importantly, pooled data from the studies which used healthy controls, the rates of BAD, lactose and fructose malabsorption, MC and SIBO were also signi cantly higher in patients ful lling IBS criteria (Supplementary Tables 16-24).
Some may argue that abnormal test results do not correlate perfectly with treatment responses, so the clinical relevance of such results is not de ned. Although this study did not assess treatment responses to these possible diagnoses, our ndings are consistent with the substantial published data which report at least good short term symptom response rates in large numbers of patients when appropriate treatment is given once these diagnoses are made [34,[48][49][50][51][52][53][54][55][56][57][58][59][60][61][62][63]. However, more studies are required to assess whether prediction of long-term treatment response can be improved.
Our estimated rate for MC is slightly lower than 9.8% reported previously in patients with diarrhoeapredominant IBS and functional diarrhoea [64]. This could be explained by the fact that we have included studies that selected patients with symptoms ful lling diagnostic criteria for IBS and excluded studies recruiting subjects with functional diarrhoea. Also, patients with constipation predominant IBS were included in some of these studies. Two of the studies we included only examined biopsies from the left side of colon [45,65] which may also underestimate prevalence rates. Our calculated crude pooled rate may also underestimate PEI as Leeds et al. used a faecal elastase-1 cut-off of < 100 µg/g, re ecting only the presence of severe PEI [66].
We found similar diagnostic yields for BAD in studies using total faecal bile acids and fasting serum 7α-C4. Glucose hydrogen breath testing and small bowel aspirate for SIBO give relatively comparable yields.
However, the prevalence rate of SIBO using small bowel aspirate is likely to be over-estimated as one study [67] used a low bacterial count > 10 3 cfu/mL to diagnose SIBO, which has been shown to result in a 2-fold increase in diagnosis [68]. Interestingly, the two studies, both by Ghoshal et al. [69,70] comparing the two diagnostic tests reported very different rates for SIBO. The small sample size used in the earlier study was likely the cause of such results. It is not as clear why in some studies the rates of SIBO using glucose breath testing were much lower than most of the other studies, but one of the possible explanations was that a more stringent criteria was used to de ne SIBO.
Although our data show that both lactose and fructose malabsorption were signi cantly more prevalent in IBS patients than in healthy individuals, it is worth noting that CM may be physiological (Supplementary Tables 17 and 24). It is often the sensory response following ingestion of carbohydrates -intolerance, that is pathological. For lactose intolerance, similar observation to lactose malabsorption can be made among the IBS patients and healthy controls (Supplementary Tables 17 and 24). Reduction of exposure to such carbohydrates may be therapeutic and should be considered in managing CM [71][72][73].
Most systematic reviews restricted their study inclusion to one type of study design (i.e. case-control studies) and those with a large sample size (i.e. >100 patients). We have included studies of all types and sample sizes recruiting consecutive patients to minimise potential selection bias [74]. We included only studies that investigated patients with IBS de ned by strict diagnostic criteria, and using recognised diagnostic tests for an alternative disorder.
There are several limitations to this study. First, the inclusion of smaller studies may have biased our estimated rates, particularly in conditions where only a small number of studies were included in the overall analyses. Modestly different rates can be seen after excluding studies with a sample size < 100.
Such disparity is not as evident in conditions with a greater number of large studies. Also, the lack of an appraisal tool that accurately assesses all the types of study included limits con dent identi cation and reporting of bias. The exclusion of non-English publications and those in only abstract form might have introduced bias; however, in comparison to the number of literatures included in this review, the number of these excluded publications was very small so they are unlikely to affect our estimates.
Secondly, substantial statistical heterogeneity was present in all of the pooled estimates and usually remained following subgroup analysis. Signi cant clinical heterogeneity among studies including difference in sample size, diagnostic criteria, diagnostic tests used and interpretation of their results could all have contributed. It is important to point out that whilst some studies examined only patients with diarrhoea predominant IBS, others included those with constipation predominant and mixed type of IBS and this can impact on the reported rates. It is also clear that the accuracy of some tests used to diagnose these conditions is suboptimal (Supplementary Table 25), and a better diagnostic test is urgently needed to inform us a more accurate diagnosis of CM and SIBO. Further work is needed to characterise test performance and how it relates to treatment response.
Thirdly, while pooled rates for the ve conditions were calculated, we did not compare them with those in controls to calculate odds ratios. However, these conditions have already been shown to be more prevalent in patients diagnosed with IBS [34,54,72,[75][76][77] which is consistent with our ndings. Lastly, all of the included studies were conducted in a secondary or tertiary healthcare settings so our data here may not be relevant to primary care. Current management strategies for patients with IBS-like symptoms work poorly with up to 50% of patients remaining signi cantly symptomatic six years after diagnosis [78]. This systematic review strongly suggests that one of the reasons for this is may be because organic disorders, which do not respond to conventional treatments for IBS, will be missed if patients are not investigated adequately. However, our ndings are limited by the signi cant clinical and statistical heterogeneity between selected studies and the shortcomings of current diagnostic tests, in particular breath testing for CM and SIBO, and hence they should be interpretated with a degree of caution. Appropriate studies measuring long term outcomes and the bene ts for patients, society and healthcare from accurate diagnosis and optimal treatment are urgently needed to examine the true signi cance of these organic conditions in patients with IBS-like symptoms.

Conclusion
Our study has shown that treatable GI conditions can be found in a proportion of patients with IBS-like symptoms referred to secondary care. Specialist clinicians should be aware of such conditions and further tests should be considered in patients who do not respond to conventional treatments for IBS.
Further research is needed to evaluate the clinical and cost-effectiveness of such an approach compared to symptom-based management alone.