Information on modifiable lifestyle factors associated with constipation is limited, especially among non-Western populations. We examined associations between dietary intake and self-reported constipation in young Japanese women.
A total of 1705 female Japanese dietetic students aged 18–20 years and free of current disease and current dietary counseling.
Dietary intake was estimated over a 1-month period with a validated, self-administered, diet history questionnaire, and lifestyle variables including self-reported constipation were assessed by a second questionnaire designed for this survey.
A total of 436 women (26%) reported themselves to be ‘constipated’. A multivariate odds ratio (OR) for women in the highest quartile of rice intake was 0.47 (95% confidence interval (CI): 0.33, 0.68) compared with the lowest. Additionally, women in the highest category of coffee intake had a multivariate OR of 0.67 (0.47, 0.94) compared with women in the lowest. Conversely, women in the highest quartile of confectionery intake had a multivariate OR of 1.54 (1.12, 2.13) compared with women in the lowest. Moreover, a multivariate OR for constipation for women in the highest quartile of Japanese and Chinese tea intake was 1.49 (1.09, 2.05) compared with women in the lowest. Neither total dietary fiber intake nor other lifestyle factors examined were associated with constipation.
The consumption of rice and coffee was inversely associated with and that of confectioneries and Japanese and Chinese tea was positively associated with a prevalence of self-reported constipation.
Constipation is a major health problem, although the criteria for constipation remain arbitrary (Thompson et al., 1999), and symptoms of constipation vary from a relatively mild bowel habit disturbance to rare serious sequelae (Talley et al., 2003). The reported prevalence of constipation ranges from 2 to 30% in Western countries, depending on the definition applied (Garrigues et al., 2004; Higgins and Johanson, 2004). In Japan, the prevalence of constipation, defined as 3 bowel movements weekly, also seems to be relatively high (6–25%) (Hirai and Takezoe, 1997; Hirai et al., 2001). As a result of its high prevalence, chronic nature and effect on quality of life (Talley, 2004), modifiable lifestyle factors associated with constipation need to be identified.
According to previous studies in the West, not only various factors including age (Everhart et al., 1989; Sandler et al., 1990; Campbell et al., 1993; Dukas et al., 2003), sex (Everhart et al., 1989; Sandler et al., 1990; Campbell et al., 1993), smoking status (Dukas et al., 2003), alcohol consumption (Dukas et al., 2003; Sanjoaquin et al., 2004), body mass index (BMI) (Sandler et al., 1990; Dukas et al., 2003; Sanjoaquin et al., 2004), and physical activity (Everhart et al., 1989; Sandler et al., 1990; Dukas et al., 2003; Sanjoaquin et al., 2004), but also several aspects of diet such as intakes of energy (Sandler et al., 1990; Towers et al., 1994), dietary fiber (Dukas et al., 2003; Sanjoaquin et al., 2004), and nonalcoholic beverages (Sandler et al., 1990; Sanjoaquin et al., 2004) have been associated with constipation. However, information on this issue is quite limited among people in Asian countries including Japan (Kunimoto et al., 1998; Wong et al., 1999; Nakaji et al., 2002; Fujiwara, 2003), where dietary habits and foods available differ considerably from those in Western countries. Moreover, quantitative assessment of diet was not performed in these Asian studies. Therefore, we investigated associations of dietary factors, which were assessed using a previously-validated self-administered diet history questionnaire (DHQ) (Sasaki et al., 1998a, 1998b, 2000b), as well as other lifestyle factors with self-reported constipation in young Japanese women.
Subjects and methods
Subjects and data collection
The subjects were students who entered dietetic courses at 22 colleges and technical schools in Japan in April 1997 (n = 2069) (Sasaki et al., 2002, 2000a, 2003a). A total of 2063 students (2017 women and 46 men) participated in the survey (response rate: 99.7%). The staff of each school checked the submitted questionnaires according to the survey protocol. When missing values and/or logical errors were detected, the subjects were asked to complete the questions again. The questionnaires were checked at least once by the staff at each school and by the staff at the survey center. Most surveys were completed by the end of May 1997.
Data were collected using the following two questionnaires: DHQ and a questionnaire on general lifestyle. The DHQ is a previously validated, structured 16-page questionnaire for assessing dietary habits in the previous month, consisting of the following seven sections: overall dietary behaviors; major cooking methods; consumption frequency and amount of six alcoholic beverages; consumption frequency and semi-quantitative portion size of selected 121 food and nonalcoholic beverage items; dietary supplement; consumption frequency and amount of 19 staple foods (rice, bread and noodles) and miso-soup; and open-ended sections for foods consumed regularly (once/week) but not appearing in the DHQ (Sasaki et al., 1998a, 1998b, 2000b). The food and beverage items and their portion sizes in the DHQ were derived mainly from the data of the National Nutrition Survey of Japan (Ministry of Health and Welfare, 1994). Dietary intake, including 147 food and beverage items, energy and dietary fiber, was calculated using an ad hoc algorithm for the DHQ, which was based on the food composition table in Japan (Science and Technology Agency, 2000); information on dietary supplement and from the open-ended section is not used in the calculation. Dietary fiber intake was estimated by the modified Prosky method (Science and Technology Agency, 2000) from the intake of 86 fiber-containing foods in the DHQ. The food and nonalcoholic beverage items were grouped into the following 18 food groups: rice; bread; noodles; potatoes; confectioneries (including sugar and sweeteners); fat and oil; pulses (including nuts); fish and shellfish; meat; eggs; dairy products; vegetables (including mushrooms and sea vegetables); fruits; water; Japanese and Chinese tea (nonfermented type of tea (green tea) and semi-fermented type of tea (oolong tea)); black tea (fermented type of tea); coffee; other nonalcoholic beverages. A detailed description and methods of calculating dietary intake and the validity of the DHQ have been published elsewhere (Sasaki et al., 1998a). The Pearson correlation coefficient between the DHQ and 3-d dietary records was 0.48 for energy intake among 47 women (Sasaki et al., 1998a). For dietary fiber intake (g/1000 kcal), the Pearson correlation coefficient between DHQ and 16-d dietary records was 0.69 among 92 women; the mean value of the Spearman correlation coefficients for intakes of 16 food groups (g/1000 kcal) was 0.35 (range: 0.05–0.59) (unpublished observations, Sasaki, 2004).
Body weight and height were self-reported as part of the DHQ. BMI was computed as weight (kg) divided by square of height (m). We classified BMI into three categories (<18.5, 18.5–24.9, and 25) according to the Japan Society for the Study of Obesity (Matsuzawa et al., 2000). The subjects were also asked in the DHQ whether they currently received dietary counseling.
The questionnaire on general lifestyle during the previous month is a 4-page questionnaire designed for this survey. In this questionnaire, subjects reported residential area (a place where the subject mainly lived during the previous month), participation in sports club activities (times/months), without inquiring into the types of sports, their intensity or duration, and smoking status (‘never,’ ‘past’ or ‘current’). They were also asked whether or not they were currently suffering from some diseases. Residential areas were categorized into 12 blocks according to the National Nutrition Survey in Japan (Ministry of Health and Welfare, 2004). Since relatively few subjects were categorized into three of these blocks, they were included in their neighboring blocks. The residential areas were also divided into three categories according to population size (cities with population 1 million, cities with population <1 million, and towns and villages). The subjects who participated in sport club activities at least once per week were regarded as ‘active’ and all others as ‘sedentary’.
Constipation was assessed by the following question in the questionnaire: do you often have constipation? The possible answers were ‘yes’, ‘sometimes’, or ‘no’. The subjects with an answer of yes to the question were considered to be ‘constipated’. We examined the validity of this question in 145 female Japanese dietetic students (mean age: 21.2 years) using 14-d bowel movement diaries as the standard; 33 subjects with an answer of yes had significantly (P<0.001) fewer bowel movements (mean±s.d.: 3.4±1.1 day/week) than did 60 subjects with an answer of sometimes (4.5±1.3 day/week) or 52 subjects with an answer of no (6.2±1.0 day/week).
For statistical analysis, we selected female subjects aged 18–20 years (n=1960). We excluded one woman whose residential area was not in Japan, 154 women currently having some diseases, and 33 women currently receiving dietary counseling. Also excluded were 43 women with a reported energy intake less than half the energy requirement for the lowest physical activity category (<775 (1550 × 0.5) kcal/day) or a reported energy intake more than 1.5 times the energy requirement of the highest physical activity category (>3450 (2300 × 1.5) kcal/day) according to the Recommended Dietary Allowance for Japanese (Ministry of Health and Welfare, 1999). We further excluded 47 women with missing values in the variables used. A total of 1705 women remained for the present analysis; some women were in more than one exclusion category.
The association between self-reported constipation (the dependent variable) and a number of variables was examined. The variables examined were six nondietary variables, that is, residential blocks (nine categories), size of residential area (three categories), physical activity (two categories), smoking status (three categories), alcohol drinking habits (two categories (‘yes’ or ‘no’) because of extremely low alcohol intake (mean: 0.7 g/day)), and BMI (three categories) and 22 dietary variables, that is, intakes of energy (kcal/day), 18 food groups mentioned above (g/1000 kcal), and total, soluble, and insoluble dietary fiber (g/1000 kcal) (quartiles except for water (four categories), black tea (four categories), and coffee (three categories) because of more than one quarter nonconsumers). We calculated both crude and multivariate odds ratios (ORs) and 95% CIs for self-reported constipation for each category of variables included using the logistic regression analysis; multivariate ORs were calculated by adjusting for six nondietary variables and energy intake. As results for the crude and multivariate analyses were similar for all variables considered, we presented only the results derived from the multivariate models. Trend of association (for only dietary variables) was assessed by a logistic regression model assigning scores to the levels of the independent variable. All statistical analyses were performed using the SPSS for Windows software program, version 11.5, (SPSS Japan Inc.) and the SAS statistical software, version 8.2 (SAS Institute Inc.). A two-sided P value of <0.05 was considered statistically significant.
The mean (±s.d.) of selected physical characteristics was as follows: 18.1±0.4 years for age, 157.9±5.2 cm for height, 51.8±7.3 kg for weight, and 20.8±2.6 kg/m2 for BMI. A total of 436 (26%) out of 1705 women reported themselves to be ‘constipated’. Table 1 presents the multivariate ORs (95% CIs) for constipation in each category of selected demographic and lifestyle factors. Living in town or village was associated with a decreased prevalence of constipation compared with living in city with population 1 million (OR: 0.64; 95% CI: 0.43, 0.97). Residential block, physical activity, smoking status, alcohol drinking habits, and BMI were not significantly associated with constipation.
Table 2 shows the associations between dietary intake and constipation. Energy intake was not associated with a prevalence of constipation. There was a clear dose-response relationship between increased intake of rice and a decreased prevalence of constipation (P for trend <0.0001). Women in the highest quartile had a multivariate OR of 0.46 (95% CI: 0.32, 0.66) compared with women in the lowest. Other staple foods including bread and noodles were not associated with prevalence of constipation. Because only staple foods were assessed for each meal separately in DHQ, we further assessed the relationships of intakes of rice from each meal with constipation. Increased intakes of rice at breakfast, lunch, and dinner were all associated with a decreased prevalence of constipation (multivariate OR (95% CI) in the highest quartile compared with the lowest: 0.62 (0.44, 0.86) for breakfast (P for trend=0.002); 0.65 (0.46, 0.91) for lunch (P for trend=0.001); 0.55 (0.39, 0.78) for dinner (P for trend=0.001)).
The prevalence of constipation increased with increasing intake of confectioneries (P for trend <0.001). In comparison with women in the lowest quartile, the multivariate OR for women in the highest was 1.56 (95% CI: 1.13, 2.14). There was also a positive association between intake of Japanese and Chinese tea and a prevalence of constipation (P for trend=0.004). Women in the highest quartile of the intake had a multivariate OR of 1.54 (95% CI: 1.12, 2.11) compared with those in the lowest. On the other hand, there was an inverse association between coffee intake and a prevalence of constipation (P for trend=0.045). Women in the highest category of the intake had a multivariate OR of 0.66 (95% CI: 0.47, 0.94) compared with those in the lowest. No clear associations were observed between constipation and the intake of other food groups examined. As shown in Table 3, further adjustment for total dietary fiber, as well as soluble and insoluble dietary fiber (data not shown), did not change the results of rice (excluding dietary fiber content) (P for trend <0.0001), confectioneries (excluding dietary fiber content) (P for trend=0.0005), Japanese and Chinese tea (P for trend=0.0067), and coffee (P for trend=0.0563) materially, indicating that these observed associations are independent of dietary fiber intake.
There was a positive association of intake of total and soluble dietary fiber with a prevalence of constipation (P for trend=0.07 and 0.01, respectively). The association between total dietary fiber and constipation, however, disappeared when further adjusted for rice (excluding dietary fiber content) (P for trend=0.41; Table 3), confectioneries (excluding dietary fiber content) (P for trend=0.16), Japanese and Chinese tea (P for trend=0.09), or coffee (P for trend=0.09). Additionally, although the positive association between soluble dietary fiber and constipation remained when further adjusted for Japanese and Chinese tea (P for trend=0.01) or coffee (P for trend=0.02), the association disappeared when further adjusted for rice (excluding dietary fiber content) (P for trend=0.37) or confectioneries (excluding dietary fiber content) (P for trend=0.08). Thus, the positive association between dietary fiber and constipation seemed to be largely dependent on rice intake.
We found that increased intakes of rice and coffee were associated with a decreased risk of constipation in young Japanese women. We also found that lower intakes of confectioneries and Japanese and Chinese tea were associated with a decreased risk of constipation. While a limited number of studies on this issue conducted in Asian countries used non-validated, relatively simple questionnaires for the assessment of dietary factors (Kunimoto et al., 1998; Wong et al., 1999; Nakaji et al., 2002; Fujiwara, 2003), we used a previously validated DHQ for quantitative assessment of dietary intake.
We found dose–response relationships of increased intake of rice with a decreased risk of constipation. Furthermore, increased intakes of rice from breakfast, lunch, and dinner were all associated with decreased risk of constipation. The protective effect of rice on constipation has also been indicated in two previous studies conducted in Asian communities (Wong et al., 1999; Nakaji et al., 2002) where rice is the main staple food. The reason for the association is not well known. Nakaji et al. (2002) hypothesized that the effect of rice is due to dietary fiber in rice because rice is the largest source of dietary fiber in Japanese people (Sasaki et al., 2003b). Conversely, Wong et al. (1999) hypothesized that the effect of rice is explained by the increased energy intake because rice is high in energy but low in fiber. In these studies, however, quantitative assessment of dietary intake was not available because of the use of relatively simple questionnaire. Our data do not support their hypotheses since the association between rice and constipation was independent of both energy and dietary fiber intake. Rice is a staple food in Japan and a major contributor of many vitamins and minerals; some of constituents in rice and/or combinations of these constituents might exert a preventive effect on constipation. Alternatively, rice intake might merely reflect an overall healthier lifestyle that may not have been accurately captured and controlled in our analysis.
Several studies have suggested the association of breakfast-skipping and constipation (Kunimoto et al., 1998; Fujiwara, 2003), but we did not assess this association because of a quite small number of women with the habit of breakfast-skipping (n=30). In the present study, however, 65% of the staple food intake at breakfast was derived from rice, while a decreased intake of rice at breakfast was associated with increased risk of constipation. This might suggest breakfast-skipping as a risk factor of constipation.
A positive association between confectionery intake and constipation was observed, although we are not aware of any research reporting this association. We also found an adverse effect of Japanese and Chinese tea, which is in agreement with a study of Singapore (Wong et al., 1999), and a preventive effect of coffee, generally consistent with a study of the US (Dukas et al., 2003). It is unclear why these foods had such effects on constipation. Although our finding regarding these foods may have been due to chance alone given the large number of statistical analyses conducted in the present study and intake of these foods may be a marker of other lifestyle factors that were not addressed, further studies examining the association between constipation and these foods would be some of interest.
Constipation seemed to be associated with intake of energy (Sandler et al., 1990; Towers et al., 1994), fluids (water and pure fruit juices) (Sanjoaquin et al., 2004), beverages (sweetened, carbonated, and noncarbonated) (Sandler et al., 1990), tea (Sandler et al., 1990), meats (Sandler et al., 1990; Sanjoaquin et al., 2004), eggs (Nakaji et al., 2002), dairy products (Sandler et al., 1990), and fish (Sandler et al., 1990; Sanjoaquin et al., 2004) in previous studies. We, however, did not find any association of constipation with these dietary factors in the present study. These discrepancies may be, at least partially, explained by the differences in the characteristics, dietary habits, and lifestyle of the subjects examined, dietary assessment methods used, and definitions of constipation applied among studies.
The effect of dietary fiber on constipation is widely accepted, but only a few studies have found an inverse association between dietary fiber and constipation (Dukas et al., 2003; Sanjoaquin et al., 2004), and many other studies have failed to find this association (Everhart et al., 1989; Whitehead et al., 1989; Campbell et al., 1993; Towers et al., 1994). Unexpectedly, there seemed to be a positive association between dietary fiber intake and constipation in the present study, although the association disappeared after further adjustment for rice intake, suggesting that the association is largely due to an inverse association between rice intake and constipation. The positive association between dietary fiber and constipation may be because subjects suffering from constipation might increase their dietary fiber intake. This is particularly prevalent in the present study because the subjects are dietetic students and therefore may be highly health conscious. However, women with current dietary counseling were excluded from the present analysis. Additionally, not only was intentional dietary change, self-reported in DHQ, not significantly associated with constipation, but also the analyses, further adjusted for intentional dietary changes or after excluding the subjects who reported intentional dietary change within one year, provided identical results (data not shown). Another explanation of the positive association between dietary fiber and constipation is that dietary fiber intake was too low to have a protective effect for constipation for most women. The amount of dietary fiber estimated in the present study (mean: 12.0 g/day), however, was comparable with that observed in women aged 18–29 years in the Japanese National Nutrition Survey in 2001 (12.8 g/day) (Ministry of Health and Welfare, 2003) and 2002 (12.0 g/day) (Ministry of Health and Welfare, 2004), which has been available since 2001.
As a result of this unexpected association between dietary fiber and constipation and the possibility that subjects suffering from constipation might increase their dietary fiber intake and hence change their diet, the findings regarding foods, particularly those significantly associated with constipation in the present study (rice, confectioneries, Japanese and Chinese tea, and coffee), should be interpreted with great caution. We cannot deny that the association between these foods and constipation merely reflects dietary behaviors changed after, not before, suffering from constipation, although the findings on these foods were independent of dietary fiber intake and these foods are generally unlikely to be recognized as those having an influence of constipation. As mentioned above, however, previous studies have shown similar findings on rice (Wong et al., 1999; Nakaji et al., 2002), Japanese and Chinese tea (Wong et al., 1999), and coffee (Dukas et al., 2003). Unfortunately, these are all cross-sectional findings; prospective research on this area is required.
Findings regarding dietary factors also need to be cautiously interpreted in terms of dietary assessment methodology. First, the DHQ measures only the memory and perception of usual diet, although we used a previously-validated questionnaire (Sasaki et al., 1998a, 1998b, 2000b). Second, selective under- and/or overestimation of dietary intake, which may affect the energy-adjusted intake in a biased way, is a serious problem in many populations (Livingstone and Black, 2003) as well as the women examined here (Okubo and Sasaki, 2004). However, a repeated analysis presented in Tables 2 and 3 after excluding 400 subjects with implausible reported energy intake (women with the ratio of reported energy intake to basal metabolic rate, estimated using the FAO/WHO/UNU equation (FAO/WHO/UNU, 1985), of <1.2 or >2.5 (Black et al., 1996)) provided the similar results. We thus believe that the associations in the present study are not spurious associations created by inaccurate dietary data.
Constipation has been associated with smoking status (Dukas et al., 2003), alcohol drinking (Nakaji et al., 2002; Dukas et al., 2003; Sanjoaquin et al., 2004), and BMI (Sandler et al., 1990; Dukas et al., 2003; Sanjoaquin et al., 2004) in previous studies, but we observed no significant association of constipation with these variables. These null associations may be due to the large proportions of women without habits of smoking (94%) or alcohol drinking (78%) and with a normal BMI (78%). Physical activity has also been associated with constipation in several studies (Everhart et al., 1989; Sandler et al., 1990; Dukas et al., 2003; Sanjoaquin et al., 2004); we, however, did not find the association. This may be because of our relatively rough assessment of physical activity because we classified the subjects only into two groups according to the frequency of participating sports club activities without consideration of other kinds of activities.
In the present study, the assessment of constipation was based strictly on self-reporting, although subjects who were considered to be ‘constipated’ had significantly fewer bowel movements than did other subjects. The proportion of the subjects who were considered ‘constipated’ in the present study seemed to be relatively high (26%); some of those may not be classified as ‘constipated’ according to symptom-based criteria such as Rome I and Rome II criteria (Thompson et al., 1999). In fact, the prevalence of self-reported constipation was much higher compared with the prevalence based on Rome I and Rome II criteria in a study of Spain (30 vs 19 and 14%) (Garrigues et al., 2004). Thus, whether the same associations we observed would hold for constipation according to symptom-based criteria is not known, which should be addressed in future studies.
Although the use of medications may be associated with constipation (Wong et al., 1999; Dukas et al., 2003; Talley et al., 2003), this variable was not assessed in the present study. We, however, analyzed only the data of apparently healthy women without any disease at the time of study to minimize the confounding by medication usage. Additionally, our results might not be representative because the subjects were selected female dietetic students.
In conclusion, intake of rice and coffee was inversely and intake of confectioneries and Japanese and Chinese tea was positively associated with self-reported constipation in a group of young Japanese women. As a result of the cross-sectional nature of the present study, which precludes any causal inferences, several limitations, particularly possibility that subjects suffering from constipation might increase their dietary fiber intake and hence change their diet and the use of self-reported constipation, and the lack of biological explanation for the associations we observed, however, further observational (favorably, prospective) and experimental studies are required to clarify these relationships.
We would like to thank Ms Yukari Takemi, RD, PhD and Ms Ayako Miura, RD for data collection regarding the validation of the question on constipation.