Trend and Risk Factors of Diverticulosis in Japan: Age, Gender, and Lifestyle/Metabolic-Related Factors May Cooperatively Affect on the Colorectal Diverticula Formation

Background Despite the marked increase of diverticulosis, its risk factors have not been adequately elucidated. We therefore aim to identify significantly associated factors with diverticulosis. We also aim to investigate the present state of diverticulosis in Japan. Methods We reviewed the medical records from 1990 to 2010 that included the data of consecutive 62,503 asymptomatic colonoscopy examinees from the general population in Japan. Most recent 3,327 examinees were analyzed with 16 background factors. Results Among the 62,503 subjects (47,325 men and 15,178 women; 52.1 ± 9.2 years old), diverticulosis was detected in 11,771 subjects (18.8%; 10,023 men and 1,748 women). The incidences of diverticulosis in 1990-2000 and 2001-2010 were respectively 13.0% (3,771 of 29,071) and 23.9% (8,000 of 33,432): the latter was much higher than the former in all age groups and for both genders. Considering the anatomical locations of colorectal diverticula, left-sided ones have markedly increased with age but not significantly changed with times. Univariate analyses of the 3,327 subjects showed significant association of diverticulosis with four basic factors (age, sex, body mass index, blood pressure), three life style-related factor (smoking, drinking, severe weight increase in adulthood), and two blood test values (triglyceride, HbA1c). The multiple logistic analysis calculating standardized coefficients (β) and odds ratio (OR) demonstrated that age (β = 0.217-0.674, OR = 1.24-1.96), male gender (β = 0.185, OR = 1.20), smoking (β = 0.142-0.200, OR = 1.15-1.22), severe weight increase in adulthood (β = 0.153, OR = 1.17), HbA1c (β = 0.136, OR = 1.15), drinking (β = 0.109, OR = 1.11), and serum triglyceride (β = 0.098, OR = 1.10) showed significantly positive association with diverticulosis whereas body mass index and blood pressure did not. Conclusions The large-scale data of asymptomatic colonoscopy examinees from the general population from 1990 to 2010 indicated that the prevalence of diverticulosis is still increasing in Japan. Age, male gender, smoking, severe weight increase in adulthood, serum HbA1c, drinking, and serum triglyceride showed significant positive association with diverticulosis.


Introduction
The exact prevalence of diverticulosis is difficult to define, because most subjects with diverticulosis might be asymptomatic and also because the large intestine of healthy individuals is not usually examined [1]. Nevertheless, it is broadly accepted that rates of diverticulosis have obviously increased over times worldwide [1][2][3][4][5]. In Japan, several early studies reported the prevalence of diverticulosis based on the barium enema examination: 1.6% in 1975 [6], 7.8% in 1983 [7], 13.3% in 1987 [8], 15.7% in 1995 [9], and 23.6% in 2000 [10]. These reports indicated that prevalence of Japan had increased like Western countries [2,3]. However, recent reports evaluating the prevalence of diverticulosis among the asymptomatic subjects from the general population are few: thence, one of the aims of this study is investigating the present state of diverticulosis in Japan, one of the typical developing countries in East Asia.
Despite the marked increase of disease rate in the last several decades [1], pathogenesis and risk factors of diverticulosis have not been adequately elucidated [4]. For the basic factors, advancing age is the definite risk factor of diverticulosis validated by many previous studies [6,7,9,[11][12][13]. On the contrary, an association of gender with diverticulosis is still controversial [6,14,15]. Low-fiber diet has been repeatedly reported as the risk factor of diverticulosis [11,[16][17][18], but a recent study achieved the opposite result [13]. As for drinking, results of some reports considered alcohol as a risk factor of diverticulosis whereas other reports did not [19][20][21][22]. Association of smoking with diverticulosis is also controversial [13,[21][22][23]. Recently, several studies indicated that body mass index and/or obesity may be associated with diverticulosis [24][25][26]. As a whole, definitively associated factors for diverticulosis have not been identified except for advancing age. Therefore, the main purpose of this study is to evaluate associations of several background factors with diverticulosis.

Methods
Asymptomatic colonoscopy examinees from the general Japanese population for 21 years Study subjects to evaluate the associated background factors of diverticulosis Of the 20,773 subjects who received medical checkup at our medical institute and approved participating in our study, 5,021 subjects underwent colonoscopy from 2008 to 2012 (Fig 1B). This study was approved by the ethics committee of the University of Tokyo, and written informed consents were obtained from all the study participants according to the Declaration of Helsinki.

Total colonoscopy and diagnosis of diverticulosis
For bowel preparation for colonoscopy, each subject ingested 2 liters of polyethylene glycol electrolyte (PEG) solution. In case of inadequate preparation, the examinee additionally ingested 0.5-1 litters of PEG solution. After confirming that the bowel's output runs clear, total colonoscopy was performed by well-trained endoscopists both in the morning and afternoon. In case of failure to reach the cecum, the subject was omitted in the present analysis. During the procedure of total colonoscopy, the locations of diverticula were documented. The anatomical locations of colorectal diverticula were classified into cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

Categorization of blood pressure
According to the guideline of the Japanese society of hypertension [27], total subjects were classified into three classes as follows; i) optimal blood pressure: systolic blood pressure (SBP) lower than 120 mmHg and diastolic blood pressure (DBP) lower than 80 mmHg, ii) hypertension: SBP equal or more than 140 mmHg or DBP equal or more than 90 mmHg, and iii) normal range blood pressure: a combination of SBP and DBP not fallen under optimal blood pressure and hypertension.

Evaluation of serum anti-Helicobacter pylori IgG, Alcohol Drinking, and Smoking
Serum anti-Helicobacter pylori IgG was measured using a commercial kit (E-plate "EIKEN" Helicobacter pylori antibody, Eiken Chemical Co LTD., Tokyo, Japan) as we previously reported [28][29][30]. According to the manufacture's instruction, the titer of anti-Helicobacter pylori IgG 10 U/ml was considered as Helicobacter pylori-positive.
Based on our previous study [32,33], we used additional four questions to assess the lifestylerelated factors: 1) Has your body weight markedly increased in adulthood (more than 10 kg from age 20 years)? 2) Do you have a feeling of inadequate sleep? 3) Do you have a habit of frequent skipping of breakfast (more than three times a week)? 4) Do you have a habit of having dinner within two hours before going to bed?

Statistical Analyses
We used JMP 10 software or SAS 9.1.3 (SAS Institute Inc. Cray, NC, USA) for statistical analyses. In the univariate analyses, categorical data were analyzed by Pearson's chi-square test, and continuous data were analyzed by Welch's t-test. Covariates associated with the presence of diverticulosis were carefully selected based on known confounders and clinical knowledge as follows: age (categorical data), sex (categorical data), body mass index (categorical data), blood pressure (categorical data), habit of smoking (categorical data), habit of drinking (categorical data), severe weight increase in adulthood (categorical data), feeling of inadequate sleep (categorical data), habit of frequent skipping of breakfast (categorical data), habit of having dinner within two hours before going to bed (categorical data), anti-Helicobacter pylori IgG (categorical data), serum T-chol (total cholesterol, continuous data, [ To compare the incidence trend of anatomical locations of colorectal diverticula in 1990-2000 and that in 2001-2010, we constructed the interaction terms of diverticula's locations with advancing age or an examination period to apply "Zero-Inflated Poisson model" [34]. To evaluate relative strengths of predictors on diverticulosis, they were calculated by fitting a logistic regression model using all the relevant variables, which were derived from the univarite analyses. In all the analyses, p values <0.05 were considered as statistically significant.
As shown in Table 1 and Fig 2, the disease rates of diverticulosis in 2001-2010 were higher than those in 1990-2000 for both genders and in all the age groups. As described in Introduction, the prevalence of diverticulosis in Japan had increased from a few percents in 1970s to about 20% around 2000 on the basis of barium enema studies [6][7][8][9][10]. Our results indicated that the prevalence of diverticulosis has been still increasing in Japan, although it has not reached the still higher disease rates of diverticulosis in Western countries such as 39.9% (318 of 796) in France [12] or 55.5% (497 of 896) in the United States [35].
Left-sided colorectal diverticula increase markedly with age but not significantly with times Several previous studies reported that the incidence of diverticula in the left-sided colon increases with advancing age [7,9,10]. Our results based on the data of colonoscopy examinees over 21 years similarly showed that the rates of diverticula in sigmoid and descending colon obviously increase with age, and those of diverticula in cecum and ascending colon apparently reduce in parallel (Fig 3, Table 2, S1 Table and S2 Table).
On the contrary, distributions of colorectal diverticula in the first half 11 years and in the second half 10 years are rather similar (Fig 3). By applying the "Zero-Inflated Poisson model", we compared the incidence trend of the anatomical locations of colorectal diverticula in 1990-2000 with that in 2001-2010. This statistical evaluation revealed that there is no significant difference in the location of colorectal diverticula between the two periods (p = 0.4042).

Significantly associated factors of diverticulosis
From the 5,021 general asymptomatic colonoscopy examinees who approved participating in our study, 1,694 subjects were excluded due to insufficient data for analyses ( Fig 1B). Of the 3,327 eligible subjects (2,485 men and 842 women; mean age 55.0 ± 9.1 years; range 20 to 86 years), colorectal diverticula were detected in 858 subjects (25.8%, Table 3), which comprised 734 men (29.5% of 2,485 men) and 124 women (14.7% of 842 women). To identify predictive factors of diverticulosis, we statistically evaluated association of diverticulosis with four basic factors (age, sex, body mass index, and blood pressure), six lifestyle-related factors (including smoking and drinking), and six blood test values (Table 3).
Of the 16 analyzed variables, univariate analyses showed the statistically significant association between diverticulosis and all the four basic factors: namely, older age, male gender, higher  body mass index, and hypertension were significantly associated with the presence of colorectal diverticula (Table 3). In addition, three lifestyle-related factors (smoking, alcohol drinking, and severe weight increase in adulthood) and two blood test values (serum triglyceride and HbA1c) denoted strong association with diverticulosis (Table 3). Based on the univariate analyses, we further performed the multivariate logistic analysis evaluating associations of diverticulosis with selected nine variables (Table 4). Judging from the p values and standardized coefficients, diverticulosis showed significantly positive association with older age, male gender, smoking, severe weight increase in adulthood, serum HbA1c value, alcohol drinking, and serum triglyceride value in this order (Table 4). On the contrary, body mass index and blood pressure did not denote meaningful association with diverticulosis.

Discussion
The present analysis of the 3,327 colonoscopy examinees from 2008 to 2012 showed that incidence of diverticulosis is 25.8% (858 subjects, Fig 1B). Past studies focusing on diverticulosis in Japan were mostly based on the barium enema examination, in which the prevalence of colorectal diverticula has been reported to increase from 1.6% to 23.6% in the quarter of century [6][7][8][9][10]. More recently, two studies analyzing the data from colonoscopy examinees were reported, in which the incidence of diverticulosis was 24.5% (165 of 672 subjects) in 2008 [15] or 25.0% (542 of 2164 subjects) in 2013 [22]. These two studies evaluated the patients who consulted to a doctor for some physical problems, whereas our study subjects were asymptomatic general adults who underwent total colonoscopy for medical checkup. Nevertheless, the disease rates in the three studies were quite similar: at present, the prevalence of diverticulosis would be about 25% in Japan. Recently, Niikura et al compared the diagnostic ability of barium enema and colonoscopy for diverticulosis: they reported that only half of diverticula were detected by colonoscopy compared with barium enema [36]. Niikura's study suggested that present true prevalence of diverticulosis in Japan may be even higher than 25%, which is not contradictory to the above-mentioned colonoscopy-based studies including the present one [15,22]. Of the 16 background factors we examined, the univarite and multivariate analyses demonstrated that diverticulosis has significantly positive associations with two basic factors (age and male gender), two blood test values (serum triglyceride and HbA1c), and three life style-related factors (smoking, alcohol drinking, and severe weight increase in adulthood). Judging from the values of standardized coefficients and odds ratios (Table 4), we could not identify the overwhelmingly associated factors for diverticulosis. These results suggest that multifactorial mechanisms should work on the formation of colorectal diverticula.
Our results also revealed positive associations of diverticulosis with smoking and drinking, both of which were controversial in the previous studies [13,[19][20][21]23]. In Japan, the number of smokers has been obviously decreased in the recent 50 years, which is opposite to the increasing incidence of diverticulosis (Fig 2). This suggests that other risk and/or preventive factors have overcome the possible influence of decreasing rate of smoking in Japan. On the other hand, alcohol consumption has gradually increased in Japan since the end of World War II (1945) [37,38], which may be one of the important causes of remarkably increased diverticulosis.
When thought of a noticeable increase of diverticulosis in the recent several decades [1,2], we speculate that various metabolic-related factors should have considerable roles on the formation of diverticula. Though the p value was not significant, higher body mass index tends to accompany with higher standardized coefficient and higher odds ratio of diverticulosis (Table 4). Among our identified significant variables, severe weight increase in adulthood, serum HbA1c, and serum triglyceride are metabolic-related factors, which may work on the increase of colorectal diverticula. Among them, serum HbA1c is one of the most important markers for diabetes mellitus, the prevalence of which has regrettably increased in the recent 30 years in Japan [39][40][41]. We speculate the increased prevalence of diabetes not only in Japan but worldwide [42,43] may be a great risk of diverticulosis in the past, present, and future.
There are some limitations in the present study. First, this is the single center study with cross-sectional or retrospective design: a future prospective study is necessary to confirm the results of the present study. Second, detectability of the colorectal diverticula may change during the 21 years (between 1990 and 2010). As preparation and procedure of total colonoscopy have not been changed in the 21 years, and also as the skill level of the endoscopists has been kept high in our institute, we think the detectability of colorectal diverticula has not been altered in the 21 consecutive years. Nevertheless, we cannot completely deny the possibility that improved colonoscopes over 20 years may increase the detection rate of diverticula. Third, we did not have detailed information for smoking and drinking. Data of quantified alcohol intake or duration of smoking might improve our analyses. However, we have already used the 5-grade scales of alcohol intake and 3-group classification of smoking in our six previous studies [28][29][30][31][32][33]: therefore, we think our categorization for drinking and smoking has been validated to a certain extent. Fourth, we had no available information about the diet and physical activity of the study subjects. In particular, it was a matter of regret that we had no data about low/high fiber diet, influence of those upon diverticulosis is still controversial [11,13,[16][17][18].