Body Mass Index Is Associated with Inflammatory Bowel Disease: A Systematic Review and Meta-Analysis

Background Prior work suggested that patients with inflammatory bowel diseases (IBD) have lower body mass index (BMI) than controls and patients with lower BMI have more serious complications. Goal The study was aimed to find relationship between BMI in patients with and without IBD, investigate effects of medicine therapy and disease stages on patients’ BMI. Methods Potentially eligible studies were identified through searching PubMed, Cochrane and Embase databases. Outcome measurements of mean BMI and the number of patients from each study were pooled by a random-effect model. Publication bias test, sensitivity analysis and subgroup analysis were conducted. Results A total of 24 studies containing 1442 patients and 2059 controls were included. Main results were as follows: (1) BMI in Crohn’s disease (CD) patients was lower than that in health controls (-1.88, 95% CI -2.77 to -1.00, P< 0.001); (2) Medical therapy significantly improved BMI of CD patients (with therapy: -1.58, -3.33 to 0.16; without: -2.09, 95% CI -3.21 to -0.98) while on the contrary not significantly improving BMI of UC patients (with therapy: -0.24, 95% CI -3.68 to 3.20; without: -1.34, 95% CI -2.87 to 0.20, P = 0.57); (3) Both CD and UC patients in active phase showed significantly greater BMI difference compared with controls than those in remission (CD patients: remission: -2.25, 95% CI -3.38 to -1.11; active phase: -4.25, 95% CI -5.58 to -2.92, P = 0.03; UC patients: remission: 0.4, 95% CI -2.05 to 2.84; active phase: -5.38, -6.78 to -3.97, P = 0.001). Conclusions BMI is lower in CD patients; medical therapy couldn’t improve BMI of IBD patients; the state of disease affects BMI of CD patients and UC patients.


Introduction
Inflammatory bowel diseases (IBD), consisting of crohn's disease (CD) and ulcerative colitis (UC), is chronic inflammatory diseases with the gastrointestinal tract [1]. Pathophysiology of IBD is yet not fully apprehended, and it has been found related to overactive mucosal immune system within the bowel [2]. During recent decades, incidence of IBD in traditionally high incidence areas, such as the United States and Europe, has been relatively stable. While the incidence of IBD has been increasing in previously low incidence areas, including China [3], it highly advocates that a tremendous epidemic of IBD is coming and more contemplation should be paid to IBD prevention, diagnosis, treatment as well as prognosis.
Recently, another great health threat developed. Global epidemic of overweight and obesity -"globesity"-is rapidly becoming major public health problem and obesity has been leading to excess morbidity and mortality. Body mass index (BMI) is a simple index of weight-forheight commonly used to classify underweight, overweight and obesity in adults. BMI has been found related with numerous health conditions, such as cancers [4][5][6][7][8][9], osteoarthritis [10], obstructive sleep apnea [11], non-alcoholic fatty liver disease [12], gallbladder diseases [13], et al.
The relationship between body mass index (BMI) and factors of inflammatory bowel diseases (IBD), such as morbidity, complications, prognosis, the stage of the disease, medical therapy, has been reported in many studies. Some studies have discovered that in IBD patients, BMI were lower than that in non-IBD controls [14][15][16][17]. However, others indicated that UC patients have higher BMI than controls or have BMI in the normal range [18,19]. Some studies [15,17,[19][20][21][22][23] show that medical therapy could decrease lean mass, while others indicated that patients' BMI is lower than controls' before the use of therapy [24]. Decreased BMI was found in patients with active UC, but was not commonly observed in CD patients compared with healthy controls. Nevertheless, some studies pointed out that patients with inactive CD having a lower BMI than both UC and healthy controls [25]. Since a low BMI is associated with negative health outcomes and conclusions about BMI in IBD patients were inconsistent, we did a systematic review of all relevant literature, and pooled analysis of BMI in IBD patients versus controls in order to find out that: 1) whether there is a relationship between BMI and IBD, 2) whether medical therapies and the stage of disease have effect on patients' BMI.
All articles and relevant reviews included in this study were screened [26,27] for potential missing studies. After eliminating duplicate studies, titles and abstracts of all articles obtained were screened by Jie Dong and Yi Chen to exclude the irrelevant. The remaining articles were read thoroughly and those who met selection criteria were included. Divergence were resolved by consulting with the third author, Ning Dai.
Inclusion criteria were as follows: (1) studies published as epidemiology studies or clinical trials evaluating association between BMI and IBD, concerning risk of IBD in different BMI patients, effect of BMI on diagnosis or prognosis in IBD patients or merely reported BMI in both IBD patients and health controls; (2) BMI of patients were provided or could be calculated in form of Mean±SD; (3) BMI of controls and BMI of IBD patients were not matched.
Data extraction and quality assessment. The following information was extracted: the first author, year of publication, study aim, gender, medicine usage, assessment of disease activity, source of BMI and definition of controls, number, age and BMI of both patients and controls. Extracted data were looked over to consolidate by two authors Jie Dong and Yi Chen. Discrepancies were resolved by the third investigator, Ning Dai.
If a study did not incorporate enough data to be included in the meta-analysis (i.e., no risk estimates and/or 95% confidence intervals or no specific BMI in patients or controls), the corresponding author would be contacted via email and the missing data would be solicited at least twice.
The quality of each study was assessed according to Newcastle-Ottawa quality assessment scale [28]. This scale consists of three factors: patient selection, comparability of study groups, and assessment of outcomes. A score of 0-9 (labeled as stars) was used to indicate the quality of each study.

Data Synthesis and Statistic Analysis
All analyses were conducted using the STATA software (version 12, STATA Corporation, College Station, TX, U.S.). Publication bias was conducted through Begg's adjusted rank correlation test. [29] Random effect meta-analysis was employed to compare difference in BMI between IBD patients and healthy controls in the form of weighted mean difference (WMD). Statistical heterogeneity among studies was evaluated though Cochran's Q test and I 2 statistic [30]. The extent of heterogeneity across studies was checked using the chi-square test and I 2 test; P0.10 and/or I 2 >50% indicates significant heterogeneity. If so, subgroup analysis will be conducted. Sensitivity analysis was performed to investigate the contribution of each study to the heterogeneity by comparing results before and after sequentially removing one study and reanalyzing the pooled estimate for remaining studies.

Results
Search results and Study characteristics 9,493 potential relevant articles were identified by searching the databases. An additional five studies were included from reference retrieve. After eliminating duplicates, 8,926 papers were screened. With reading titles and abstracts, a total of 220 articles were considered potentially eligible and full texts were carefully reviewed for inclusion. Among these articles, 196 were subsequently excluded for the following reasons: 173 did not investigate the relationship between BMI and IBD; 13 did not provide sufficient data and did not reply to us; in five studies, BMI couldn't be calculated in form of Mean±SD; in another five articles, BMI of controls were matched with patients' BMI. Thus, a total of 24 articles were finally included (Fig 1). 19 articles focus on Europeans, one on Americans, two on Asians, one on Australians and one on Africans. Study designs included three prospective cohort studies, four cross-sectional studies and 17 case-control studies. 10 assessed BMI in UC patients [15,16,18,19,21,25,[31][32][33][34], 21 enrolled CD patients [15, 17-25, 31-33, 35-43] and three included patients of the undefined IBD type [14,15,33]. A total of seven studies investigated the effect of medical therapy on BMI improvement [15,17,[19][20][21][22][23]. For IBD state, four studies assessed patients in active phase [15,16,37,43], while 12 evaluated patients in clinical remission and 9 reported both [16,17,23,25,31,35,36,38,[40][41][42][43].
The aims and quality assessments of these studies and baseline characteristics of patients and controls are specified in Table 1.

Primary Meta-analysis
A total of 24 studies containing 1442 patients and 2059 controls were included. Pooled data showed significant BMI difference of -1.88 for CD patients compared to controls (95% CI -2.77 to -1.00, P< 0.001) (Fig 2), while no obvious difference was observed for -0.94 in UC patients (95% CI -2.54 to 0.66, P< 0.001) (Fig 3) or undefined IBD patients (WMD of -2.64,  95% CI -5.43 to 0.16, P = 0.064) (Fig 4). Great heterogeneity was found in primary meta-analysis (P< 0.001 for CD, UC as well as non-identified group). There was no evidence of publication bias in primary meta-analysis (P value of Begg's test: 0.735 for CD, 0.276 for UC and 1.000 for non-identified group). Sensitivity analyses indicated that pooled estimate was not excessively changed by any individual studies (Tables 2-4, S1 Fig, S2 Fig).

Subgroup analysis
Medical therapy. Among the 21 articles enrolled CD patients, 9 articles containing 429 patients and 565 controls mentioned the current use of therapy, like corticosteroid, azathioprine, mesalamine, or TNF-alpha antagonist. 468 patients and 637 controls in 12 studies hadn't taken any medicine. Pooled data showed significantly different BMI in patients without therapy compared to controls (WMD = -2.09, 95% CI -3.21 to -0.98, P < 0.001), difference of -1.58 in BMI in patients with therapy (95% CI -3.33 to 0.16, P < 0.001) (Fig 5). There was no significant difference between the two subgroups (P = 0.63).
There were 10 studies concerning UC patients. 178 patients from four studies had received therapy, like corticosteroid, azathioprine or mesalamine, during the studies and a total of 169 people were included in the control groups. 275 patients and 820 controls in six studies hadn't    taken any medicine. We observed no significant difference of BMI in patients without medical therapy (pooled WMD = -1.34, 95% CI -2.87 to 0.20, P = 0.089), and in BMI in patients with medical therapy (pooled WMD = -0.24, 95% CI -3.68 to 3.20, P = 0.891), compared with controls. There was no significant difference between the two subgroups (P = 0.57) (Fig 6). Disease state. Crohn's Disease Activity Index (CDAI), Simplified Crohn's Disease Activity Index, Pediatric Crohn's Disease Activity Index (PCDAI) or Harvey-Bradshaw Index (HBI) were used in different studies to assess CD patients' disease activity. Active phase was defined as CDAI score over 150, below was categorized as in remission. 344 patients in remission and 303 controls were enrolled in 11 studies. A significant difference of BMI was suggested (pooled WMD = -2.25, 95% CI -3.38 to -1.11, P < 0.001). Three studies focused on 51 CD patients in active phase and 65 healthy controls. Pooled data showed a significant different BMI in patients  in active phase (pooled WMD = -4.25, 95% CI -5.58 to -2.92, P < 0.001). There was a significant difference between the two subgroups (P = 0.03) (Fig 7). Truelove-Witts' index, the Mayo score, Powell-Tuck index or Colitis activity index were used in different studies to assess UC patients' disease activity. Two articles contained 73 UC patients in active phase and 565 controls. A significant difference of BMI was found in patients in active phase (pooled WMD = -5.38, 95% CI -6.78 to -3.97, P < 0.001). Pooled data showed no significant difference of 0.40 in BMI in 51 patients in remission and 86 controls (95% CI -2.05 to 2.84, P = 0.752) from four studies. There was a significant difference between two subgroups (P = 0.001) (Fig 8).

Discussion
The present meta-analysis is the first to assess the association between BMI and IBD patients. The inclusive finding of our systematic review is that both UC and CD patients had consequential lower BMI than controls. This might be illustrated by reduced dietary intake, malabsorption. Metabolic disturbances, such as increased energy expenditure, decreased respiratory quotient, and increased substrate oxidation rates, may also be a reason causing patients' loss of glucose, protein and fat [40,42]. In prior works [15,33], researchers have proven that BMI was one of the most predictive parameters to assess the nutritional status in IBD patients. Meanwhile BMI was quite simple and convenient to acquire. Since malnutrition is frequently found in patients with IBD, physicians should be more aware of this concern in patients with low BMI. Consequently, IBD patients are most likely at an increased risk to develop osteopenia and osteoporosis. More studies are needed to elucidate whether nutritional supplementation in IBD patients may improve the clinical course of the disease or not [21].
To some degree, medical therapy, including corticosteroid, azathioprine, mesalamine and TNF-alpha antagonists, could improve CD patients' BMI. Without medicine, patients' BMI was significantly lower than non-IBD controls' while with medical management, the difference was not significant. Medical therapy could help improve patients' health condition by relieving nutrition loss or delay disease progression [19]. In a preceding study, decreased bone mass was already present at the time of diagnosis inpatients with CD. There was no significant deference of BMI in pooled data among the subgroups of CD patients with and without therapy versus controls shown in Geltrude Mingrone's study [24], BMI of treated Crohn disease patients is significantly higher when compared with untreated Crohn disease patients (P = 0.004). In a recent study of postmenopausal women in the USA, a significant linear trend across BMI categories with bone mineral density was observed [44]. In a meta-analysis of 60 000 patients from 11 prospective studies, relative risk of fracture rose from 1.4 in females with BMI of 20 to 2.2 in females with BMI of 15 [45]. Patients' diet also affects on bone metabolism since calcium, vitamin D, and vitamin K is necessary for bone metabolism [46]. Chronic inflammation, prolonged use of steroids, as well as low intake of calcium and vitamin D is other most frequently implicated factors [47].
However, we did not observe similar result in UC patients who might be due to the relative limited number of studies included. CD patients suffered more from subordinate BMI both in active phase and in remission than controls. UC patients had a lower BMI in active phase, while in remission, there was no difference compared with controls. It indicated that BMI could be a marker of disease state which could further develop into a prognosis predictor and treatment efficacy parameter. Discrepant conclusions were achieved from previous studies. Denia Stabroth-Akil' team found that high BMI had a positive effect on the prognosis, whereas low BMI pointed to a more severe course of the disease after a retrospective analysis of data from 202 UC patients [48]. However, the outcome from Millie D. Long's research was quite opposite. Obese IBD patients may have a more severe disease course and increased need for surgery [49]. In a retrospective cohort of 124 IBD patients treating with infliximab, researchers performed a multivariable logistic regression between BMI category and response to infliximab. Obesity was associated with an earlier time to loss of response to infliximab [50]. In Avegail Flores' study, IBD patients with low BMI were more likely to receive anti-TNF treatment, undergo surgery, or experience a hospitalization than patients with high BMI [51]. Continued observation of BMI might help appraise medicine efficacy and contribute instruction in medical therapy readjustment.
Obesity has been discovered to be associated with excess adipocyte hypertrophy generating a proinflammatory state through secretion of inflammatory cytokines and chemokines, including interleukin (IL)-1β, IL-6, IL-8, monocyte chemoattractant factor, [52] tumor necrosis factor-α [53], and C-reactive protein [54]. These bio-factors might be closely related to pathogenesis of IBD. In the existing analysis, there is one prospective study observing IBD incidence among patients with various BMI. After following 300,724 people for an average of 4.5 years (range from 1.6 to 15.6 years), researchers have revealed that BMI is not associated with IBD morbidity, neither CD nor UC [18], which might be explained by that the results of existent study may be modified by certain factors, such as population, follow-up time, age and geographic distribution of study population, so conclusions of whether BMI contributes to IBD development can yet be drawn until more researches have been done.
In addition to being a potential prognosis marker or risk factor for IBD, BMI may also affect severity of complication in IBD. Increasing data are emerging both in patients with IBD and in the healthy population designating that low BMI is an indicative independent risk factor for osteoporosis.
Our study has a couple of strengths. First of all, this is the first systematic review and metaanalysis with regard to BMI in IBD patients. Second, most of the incorporated studies were of high methodological quality. Third, no publication bias was ascertained, and subgroup analysis was applied to assess the role of therapy and disease stage.
However, inevitable limitations existed in our analysis. Proportionately large heterogeneity may influence the combined result and thus the conclusion. Besides, small sample size may affect the outcome of the analysis. In addition, not enough studies were included to conduct quantity's analysis in certain concerns. Furthermore, investigation is needed to figure out the complicated role of BMI in IBD patients.
In conclusion, our meta-analysis revealed that IBD patients had lower BMI than normal controls, and this dissimilarity might be rationalized by disease remission and medical therapy, indicating that BMI may serve as an easily accessible factor in IBD prognosis and treatment effectiveness.