This cross-sectional study enrolled 5557 adults with age above 20 years old who underwent health examination at the TSGH from 2000–2016. The health examination included a laboratory data, body composition and detailed self-reported questionnaires. In addition, a diagnostic colonoscopy as a part of health examination was performed to identify any malignant polyps or other disease of colon such as diverticulosis. Before a diagnostic colonoscopy, all the participants were asked to complete detailed self-reported questionnaires. Those who reported any related abdominal symptom, such as abdominal tenderness or cramps, active rectal bleeding, fever, and nausea were not be able to have colonoscopy. All the colonoscopy was performed by experienced specialists after the participants had proper bowel preparation and fasted overnight. Any findings of diverticular bleeding, diverticulitis, familial polyposis syndrome or malignant tumors were excluded. All protocols in this cross-section study were approved by the Institutional Review Board (IRB) of TSGH. The TSGH IRB waived the need to obtain individual informed consent because these data were analyzed anonymously.

Bioelectrical impedance analysis (BIA) was a validated for assessing body composition analysis [16]. In comparison with other medical equipment, such as dual-energy X-ray absorptiometry, computerized tomography or magnetic resonance imaging, BIA was more maneuverable and less expensive and radiation exposure. We measured PBF by using BIA (InBody720, Biospace, Inc., Cerritos, CA, USA).

The information of participants such as age, gender, smoking status and alcohol consumption were obtained by a self-completed questionnaire. We acquired smoking status by asking the question “Do you have a smoking habit”. For the alcohol consumption, we divided participants into two groups which were alcohol consumption and those who never consumed alcohol. All the participants had been fasting for at least 8 hours prior to blood draw. Blood samples such as total cholesterol, triglycerides, creatinine, and glucose were analyzed by standard methods.

Student’s t test was used to assess the differences in continuous variables, such as age, percentage of total body fat, serum cholesterol, triglycerides, creatinine, and fasting glucose. Smoking habit and alcohol consumption were regarded as categorical variables which were assessed by using Pearson’s chi-square tests. Three extended models were established for multivariable adjustment. The respective models were adjusted as follows: Model 1 = unadjusted; Model 2 = age, sex, total cholesterol, creatinine, triglyceride, fasting glucose; Model 3 = Model 2 + smoking, alcohol consumption. Logistic regression was applied to analyze the association between PBF and the risk of diverticulosis. Furthermore, we used receiving operator characteristic (ROC) curve analysis to determine the ideal gender-specific cut-off point for PBF to predict the risk of diverticulosis. We used the Youden’s index in order to find the ideal cut-off point which was regarded as the shortest distance between the upper left corner of the graph and the ROC curve. Two-side p values of less than or equal to 0.05 were considered as significant difference. All statistical analyses of this study were analyzed between June and August 2019 by using the Statistical Package for the Social Sciences, version18.0 (SPSS Inc., Chicago, IL, USA).

The present study showed a positive correlation between obesity and diverticulosis. In our analysis, participants with higher PBF were more likely to have diverticulosis. In addition, potential confounders were adjusted, such as age, gender, total cholesterol, triglyceride, creatinine, fasting glucose, alcohol consumption and smoking status. We found that females with higher PBF tended to have higher risk of diverticulosis in univariable analysis and the positive association was not displayed in males. Furthermore, we found that there was a stronger predictive ability for diverticulosis with a use of ROC curve analysis to decide an optimal gender-specific cut-off point for PBF.

Few previous studies examined the impacts of BMI-defined obesity on diverticulosis and none of them revealed positive associations [11,17,18]. In contrast, a meta-analysis included 9 cross-sectional studies and 1 prospective cohort study through January 2017 reported an association between BMI-defined obesity and the risk of diverticulosis [19]. In spite of the positive relationship, the statistical heterogeneity of the meta-analysis was high and each of the studies from eastern countries revealed no association between obesity and diverticulosis which were opposite to the results of western countries. Two recent studies focused on the association with body fat, especially visceral body fat, on the risk of diverticulosis [12,15]. One study by Lee showed that total abdominal fat, abdominal visceral fat, and abdominal subcutaneous fat were significantly higher in the diverticulosis subgroup than in the controls. Besides, the other study by Nagata showed that diverticulosis was significantly associated with visceral adipose tissue area, even in people with normal weight [12]. The body fat of the two studies was both measured by CT scan.

In our studies, bioelectrical impedance analysis was used as the measurement of PBF which created a non-invasive, maneuverable and radiation-free method to assess the risk of diverticulosis. Although the mechanism that linked obesity to diverticulosis still remained unclear, several connections behind the association had been promoted by previous studies. It was suggested that increasing intraluminal pressure was a plausible cause of diverticulosis. The speculated mechanism was intraluminal pressure caused diverticula protruding from mucosa and submucosa of colon. Besides, several studies identified obese individuals produced higher concentration of methane gas by gut bacterial flora than nonobese individuals. A higher concentration of methane gas in gut might lead to delaying transition of small bowel and decreasing motility of colon which was conjectured to increase intraluminal pressure and resulted in diverticulosis.

Besides, the correlation between gender and diverticulosis remained controversial in previous studies [17,20,21]. In a retrospective study of Japanese patients, Yamamichi et al. identified there was a significant positive correlation between diverticulosis and male subgroups in multivariate analysis [10]. However, in a prospective study of Korean patients, Song et al. reported no association between gender and diverticulosis [11]. We also found that the participants with older age were more likely to have diverticulosis than the younger participants which was in line with previous studies in Taiwan [20,21].

Our study had four limitations needed to be considered. First, a few well-known lifestyles factors were not taken into account, such as dietary fiber intake and physical activity. However, the correlation between fiber intake and diverticulosis was still unclear. Several studies reported that low dietary fiber was associated with diverticulosis, but others revealed conflicting data [22–24]. Physical activity was shown to have an inversely correlation with the risk of symptomatic diverticular disease in a large prospective study [9]. Despite that, to our knowledge, the association between dietary, physical activity and PBF should be existed. Thus, investigations of dietary and physical activity should be involved in the questionnaire in future studies. Second, it was well known that multiple risk factors were associated with diverticulosis. In our study, age, gender and PBF might be important risk factors in diverticulosis. However, the other risk factors needed to be considered as predictors and might have varying degrees of effects on diverticulosis. Third, self-election bias of the participants in the study might exist because all enrolments received health examination at their own expense, which meant that they intended to pay more attention to their health condition and higher economic status among population in Taiwan. Finally, we acquired smoking status by asking the question “Do you have a smoking habit”. For the alcohol consumption, we divided participants into two groups which were alcohol consumption and those who never consumed alcohol. However, without having information about duration and daily amount of alcohol consumption as well as smoking status might lower the power of these factors. And forth, the proportion of diverticulosis in our research was relatively lower than other researches. This might be due to the selection bias we mentioned above, and moreover, the sensitivity of colonoscopy to detect diverticulosis was less than radiological techniques [25].