Larger vegetable intake helps patients with constipation: socioeconomic analysis from United States–based matched cohorts

Abstract

Introduction

Constipation is one of the most common gastrointestinal complaints in the United States, and multiple interventions and behavioral changes are often required to alleviate it. Vegetables are often one of the diet recommendations for constipated patients, but the amount required for constipation impact is still limited.

Methods

We conducted a nationwide cross-sectional study with the 2006 to 2010 National Health and Nutrition Examination Survey (NHANES) database. Patients >20 years old were stratified into four quartiles of vegetable intake. We used multivariable logistic regression to determine the association between vegetable intake and constipation status as recorded in the database.

Results

A total of 13,832 patients were included in the study. The average age was 50.5 years; 49.6% of the population was Caucasian, and 26.43% were Hispanic. In the population, 9.93% had constipation, and 92.65% had vegetable consumption. After adjusting for multiple factors, patients with larger vegetable consumption had lower odds of constipation (adjusted odds ratio 0.60; 95% confidence interval 0.49, 0.73; P < 0.001) when compared to the first quartile. Postpropensity score matching revealed similar statistical significance.

Conclusion

A larger amount of vegetable intake is associated with lower odds of constipation. Additional investigations on vegetable subtype, as well as the longitudinal relationship, are required to understand this relationship.

Constipation is a common condition that affects patient quality of life. It is estimated that 10% to 25% of the US and worldwide population are affected with constipation.^1,2 Vegetables are one of the common food intakes, and current literature suggests the benefit of fruit in regard to constipation.³ Existing studies, however, focus on small sample sizes and specific vegetable intake, in which availability and generalizability may hinder the applicability of the results to the general public. Beyond that, various studies have established the association of socioeconomic status, especially in race,⁴ but comprehensive socioeconomic and comorbidity analyses of patients with vegetable consumption with constipation outcomes are still limited.

To the best of our knowledge, no studies have investigated the relationship between the quantity of vegetables and constipation outcomes in a large population. In this retrospective, observational, propensity-matched cohort study, we aimed to explore the differences in constipation outcomes between these two groups, as well as conduct a comprehensive review of factors predicting constipation outcome using multiple socioeconomic status and comorbidity variables available in this nationally representative sample dataset.

METHODS

We utilized the National Health and Nutrition Examination Survey (NHANES) from 2005 to 2010.⁵ NHANES, conducted by the National Center for Health Statistics of the Centers for Disease Control and Prevention, is a publicly available compilation of studies that examines the general health and nutritional status of adults and children in the United States. The data collects multiple dimensions of health, nutrition, socioeconomic status, and various exposures. The datasets used in this study were those in which constipation information was available, and no more recent dataset, as of this writing of the manuscript, contained constipation data. Since the NHANES database lacks patient- and hospital-specific identifiers, our study was exempt from institutional review board approval. However, we ensured that our study adhered to ethical standards for human subjects study.

The Food Patterns Equivalents Database (FPED)⁶ converts the foods and beverages in the Food and Nutrient Database for Dietary Studies to the 37 US Department of Agriculture (USDA) Food Patterns components. Combining NHANES and FPED data provided additional information to calculate each record’s dietary quality. We used the Healthy Eating Index 2010 to represent each record’s dietary quality.⁷ In it, nine adequacy components (total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood, plant proteins, and fatty acids) and four moderation components (refined grains, sodium, added sugars, and saturated fats) contribute to a score of 100, which we believe encompasses all dietary data that may confound our results.

We focused on patients aged 20 and older in the NHANES database during the period spanning January 2005 to December 2010. After selection, we excluded records with intestinal disease, colon cancer, pregnancy, celiac disease, and incomplete questionnaire responses. The age of 20 was used due to constipation records being collected for patients over 20 years old, as designed by the survey. Constipation was defined as either having stool type of 1 or 2, according to the Bristol Stool Score, or having a weekly frequency of bowel movement <3. We then stratified all included records into four groups, based on cups of vegetable intake. Vegetable consumption was defined in the FPED survey as total vegetable intake, without any stratification of fruit type, and included various types of vegetables. A similar study utilized similar definitions and obtained satisfactory results, and thus the approach was deemed usable in our study.⁸

Data were then categorized based on socioeconomic status and demographic data. From the NHANES dataset, we obtained age, gender, race, education level, marital status, smoking, alcohol, poverty ratio, physical activity/exercise intensity, supplementary medications, medications that cause constipation, and healthy eating index. Age groups were categorized into young (20 to 29), middle age (40 to 64), and elderly (≥65), to understand the impact of different population ages. The poverty-to-income ratio served as a measure of socioeconomic status in our study. Due to limitations in sample size, the poverty ratio was grouped into those <1, between 1 and 2, and >2, whereas the Healthy Eating Index was grouped into good (>70), average (60 to 70), and unhealthy (<60). Common medications deemed to cause constipation were pulled from current literature,⁹ and any patients who used the medications in the list within 30 days before the survey were assigned to the “constipation drug usage” group. Due to the limitations of the dataset, in which only common names were used, the following drugs were included as medications associated with constipation: tricyclic antidepressants, anti-epileptic drugs, antihistamines, antacids, antispasmodics, calcium channel blockers, diuretics, bismuth, lithium, and opioids (morphine, oxycodone, tramadol). The list is similar to that used in a national survey of constipation drug use.¹⁰ Dietary supplements included vitamins, minerals, or other dietary supplements in the past month, regardless of whether they were prescribed or were bought over the counter. Due to the limitations of understanding the components of each supplement, patients were stratified as using or not using supplements within 30 days of the survey.

Socioeconomic variables were matched to generate a propensity-matched cohort¹¹ to ensure that there were minimal differences in baseline characteristics. The process of using propensity score analysis is well described.¹² A combination of various balancing, matching, weight readjustment, and interpretation strategies was utilized, with nearest-neighbor 1-to-10 matching to minimize confounding. A standardized difference of <10% suggested adequacy of the match between two groups among the measured covariates, with less potential for bias.¹³ Figure 1 demonstrates the standardized difference before and after matching.

Figure 1.

Comparison before and after propensity matching.

In our study, the primary outcome was the presence of constipation between each quartile of vegetable consumption with respect to the lowest quartile, as well as other socioeconomic statuses predicting constipation in both pre- and post-propensity score cohorts. Data analyses were conducted using StataBE 17.0 software (StataCorp, College Station, TX). Continuous variables were presented as mean, while categorical variables were presented as percentages. Proportions were compared using Fisher’s exact test for categorical variables, and the Student t test was employed for continuous variables. Covariates for our analyses were selected based on a literature review, previous findings, and commonly known confounding factors. Multivariable survey logistic regression analyses were used to calculate adjusted odds ratios for primary and secondary outcomes. Additionally, multivariable linear regression analyses were employed for continuous outcomes. Outcomes were adjusted for potential patient and region-level confounders, including age, gender, ethnicity, education, marital status, poverty ratio, body mass index, healthy food index, physical activity, supplemental usage, medications, smoking, and alcohol usage. A P value of <0.05 was considered statistically significant.

RESULTS

A total of 13,832 patients were identified per our inclusion criteria. Among them, 1374 (9.93%) patients had the presence of constipation. Supplemental Table 1 shows the baseline characteristics of the study population, stratified by the presence of constipation. Stratifying all of the patients included in our study by vegetable consumption revealed that 12,333 patients (92.65%) had vegetable consumption. We found that between the groups, gender, race, and most elements of socioeconomic status differed statistically significantly before propensity score matching. Table 1 shows the baseline characteristics of the study population, stratified by the presence of vegetable consumption. We also stratified vegetable consumption into four quartiles (Table 2). At the 50th percentile, patients were consuming the equivalent of 25.8 cups of vegetables.

Table 1.

Baseline characteristics of the study population, stratified by total vegetable consumption

Variable	No vegetable consumption	Vegetable consumption	P value
Total	1,499 (7.35%)	12,333 (92.65%)
Constipation	12.28%	9.65%	0.01
Sex, N (%)			<0.01
Male	55.69%	51.02%
Female	44.31%	48.98%
Age group			0.09
20–29	14.96%	15.75%
40–64	57.59%	60.08%
≥65	27.46%	24.17%
Ethnicity			<0.01
Non-Hispanic White	41.18%	50.22%
Non-Hispanic Black	27.90%	19.42%
Mexican American	27.79%	26.32%
Other Hispanic/race	3.13%	4.04%
Education			<0.01
Less than high school	41.07%	27.23%
High school/equivalent	23.21%	23.85%
Above high school	35.71%	48.92%
Marital status			<0.01
Solitary status	48.10%	37.32%
Partner status	51.90%	62.68%
Poverty Index Ratio			<0.01
<1.00	24.78%	16.86%
1.00–2.00	28.01%	24.21%
>2.00	47.21%	58.93%
Body mass index (kg/m2)			0.53
<25 (lean/normal)	28.86%	27.87%
≥25 (overweight)	71.14%	72.13%
Healthy Eating Index–2010			<0.01
70–100 (good)	23.77%	0.18%
60–69 (average)	0.22%	6.12%
0–59 (poor)	76.00%	93.71%
Physical activity			<0.01
Mild	52.57%	42.94%
Moderate	20.42%	23.64%
Extreme	27.01%	33.43%
Smoking			<0.01
Never	46.93%	52.97%
Former	23.02%	25.81%
Active	30.06%	21.22%
Alcohol consumption			0.21
Never	14.86%	13.00%
Former	15.64%	15.05%
Active	69.50%	71.95%
Supplemental medication usage			<0.01
Absent	62.23%	51.19%
Present (within 30 days)	37.77%	48.81%
Constipation drug usage			0.84
Absent	94.20%	94.01%
Present (within 30 days)	5.80%	5.99%

Bold indicates statistical significance at p < 0.005.

Table 2.

Statistical data for vegetable consumption, stratified by quartiles

Quartile	Mean (cups equivalent) per day
	Prepropensity match	Postpropensity match
First	0.43	4.94
Second	6.62	14.83
Third	17.43	28.79
Fourth	51.87	68.10

Crude constipation rates in both groups were reported. After adjusting for socioeconomic status and potential cofounders, we found that patients with larger vegetable consumption had statistically lower odds of constipation. Table 3 shows factors for patients with constipation, stratified by each quartile of consumption of vegetables, with multivariate analysis. Utilizing propensity score matching, with matched socioeconomic status, patients with higher consumption of vegetables still had statistically lower odds of constipation (Table 4 and Figure 2).

Table 3.

Factors predicting constipation, within the subpopulation with vegetable consumption, before propensity score matching

Variable	Constipation (%)	Odds ratio	P value
1st quartile	13.20%	Reference
2nd quartile	12.02%	0.82 (0.67, 1.01)	0.06
3rd quartile	9.95%	0.72 (0.59, 0.88)	<0.01
4th quartile	7.47%	0.60 (0.49, 0.73)	<0.01
Gender
Male	6.20%	Reference
Female	13.69%	2.26 (1.96, 2.59)	<0.01
Age group
20–29	13.07%	Reference
40–64	9.27%	0.75 (0.63, 0.88)	<0.01
≥65	9.17%	0.70 (0.56, 0.86)	<0.01
Ethnicity
Non-Hispanic White	8.87%	Reference
Non-Hispanic Black	12.77%	1.31 (1.11, 1.53)	0.01
Mexican American	9.68%	0.89 (0.75, 1.05)	0.18
Other Hispanic/race	8.26%	0.81 (0.57, 1.14)	0.23
Education
Less than high school	11.53%	Reference
High school/equivalent	11.34%	1.00 (0.84, 1.18)	0.98
Above high school	8.11%	0.71 (0.60, 0.84)	<0.01
Marital status
Solitary status	11.17%	Reference
Partner status	9.03%	1.01 (0.89, 1.15)	0.79
Poverty Index Ratio
<1.00	12.84%	Reference
1.00–2.00	11.19%	0.93 (0.78, 1.11)	0.45
>2.00	8.38%	0.78 (0.66, 0.92)	<0.01
Body mass index (kg/m2)
<25 (lean/normal)	12.36%	Reference
≥25 (overweight)	8.85%	0.66 (0.58, 0.75)	<0.01
Healthy Eating Index–2010
70–100 (good)	7.30%	Reference
60–69 (average)	6.06%	1.39 (0.74, 2.60)	0.30
0–59 (poor)	10.13%	2.16 (1.26, 3.69)	<0.01
Physical activity
Mild	11.18%	Reference
Moderate	9.22%	0.94 (0.80, 1.10)	0.48
Extreme	8.51%	0.95 (0.81, 1.11)	0.53
Smoking
Never	10.84%	Reference
Former	7.91%	0.91 (0.78.1.08)	0.31
Active	9.72%	0.80 (0.68, 0.95)	0.01
Alcohol consumption
Never	12.94%	Reference
Former	13.65%	1.23 (1.00, 1.51)	0.05
Active	8.46%	0.90 (0.75, 1.08)	0.27
Supplemental medication usage
Absent	10.86%	Reference
Present (within 30 days)	8.74%	0.84 (0.73, 0.96)	0.01
Constipation drug usage
Absent	9.71%	Reference
Present (within 30 days)	11.93%	1.31 (1.03, 1.66)	0.02

Bold indicates statistical significance at p < 0.005.

Table 4.

Factors predicting constipation, within the subpopulation with vegetable consumption, after propensity score matching

Variable	Constipation (%)	Odds ratio	P value
1st quartile	12.69%	Reference
2nd quartile	10.79%	0.88 (0.75, 1.05)	0.16
3rd quartile	7.61%	0.65 (0.54, 0.79)	<0.01
4th quartile	7.50%	0.73 (0.60, 0.88)	<0.01
Gender
Male	5.95%	Reference
Female	13.51%	2.31 (2.00, 2.67)	<0.01
Age group
20–29	13.34%	Reference
40–64	8.94%	0.69 (0.58, 0.82)	<0.01
≥65	9.02%	0.66 (0.53, 0.83)	<0.01
Ethnicity
Non-Hispanic White	8.57%	Reference
Non-Hispanic Black	12.44%	1.34 (1.13, 1.59)	<0.01
Mexican American	9.78%	0.90 (0.75, 1.07)	0.24
Other Hispanic/race	8.65%	0.87 (0.61, 1.24)	0.45
Education
Less than high school	11.76%	Reference
High school/equivalent	10.65%	0.90 (0.75, 1.08)	0.27
Above high school	7.98%	0.67 (0.56, 0.80)	<0.01
Marital status
Solitary status	10.78%	Reference
Partner status	8.97%	1.05 (0.91, 1.20)	0.44
Poverty Index Ratio
<1.00	12.62%	Reference
1.00–2.00	11.12%	0.94 (0.78, 1.14)	0.57
>2.00	8.20%	0.78 (0.65, 0.93)	<0.01
Body mass index (kg/m2)
<25 (lean/normal)	12.10%	Reference
≥25 (overweight)	8.69%	0.67 (0.58, 0.77)	<0.01
Healthy Eating Index–2010
70–100 (good)	6.45%	Reference
60–69 (average)	6.08%	1.36 (0.30, 6.13)	0.68
0–59 (poor)	9.89%	2.13 (0.49, 9.23)	0.31
Physical activity
Mild	10.97%	Reference
Moderate	9.05%	0.95 (0.80, 1.12)	0.57
Extreme	8.35%	0.95 (0.80, 1.11)	0.54
Smoking
Never	10.83%	Reference
Former	7.61%	0.88 (0.74, 1.04)	0.15
Active	9.17%	0.75 (0.63, 0.90)	<0.01
Alcohol consumption
Never	12.70%	Reference
Former	13.18%	1.22 (0.98, 1.51)	0.06
Active	8.34%	0.94 (0.77, 1.14)	0.55
Supplemental medication usage
Absent	10.65%	Reference
Present (within 30 days)	8.59%	0.84 (0.73, 0.97)	0.02
Constipation drug usage
Absent	9.53%	Reference
Present (within 30 days)	11.42%	1.28 (1.01, 1.66)	0.05

Bold indicates statistical significance at p < 0.005.

Figure 2.

Factors predicting constipation, by quartiles of vegetables intake, after propensity matching.

Table 4 and Figure 2 also show constipation odds for the socioeconomic and comorbidity variables used in this study. After adjusting bias by matching both cohorts, we found that sex type, age group, ethnicity, education, poverty ratio, body mass index, smoking, supplemental medication usage, and constipation drug usage had statistically significant levels. Although diet quality (by Healthy Eating Index) and alcohol consumption were statistically significant in the original cohort, they were not statistically significant in the matched cohort.

DISCUSSION

To our knowledge, this is the first study to explore the association of quantity of vegetable consumption with constipation on a large, nationwide scale. Furthermore, this is also the first study to understand factors predicting the odds of constipation, utilizing the lens of socioeconomic status. We report several important findings from the study. First, larger vegetable consumption decreases the risk of constipation, even after balancing socioeconomic status. Second, some socioeconomic status and comorbidity factors contribute to the risk of constipation. However, after balancing both cohorts, those factors may not impact the odds of constipation. Third, we report a comprehensive review of all socioeconomic and major comorbidity factors, in regards to the odds of constipation, which has not been reported elsewhere as extensively.

The estimated prevalence of constipation is around 10% worldwide,¹⁴ and we observed similar results in our study. We found that women had a higher risk of constipation. This is consistent with the study of McCrea et al¹⁵ who found a similar female-to-male ratio. In addition, we hypothesized that sex hormones may affect gastrointestinal mobility¹⁶ and that women might be more likely to respond to surveys. We also found disparities in non-Hispanic Black patients, in which they had higher odds of constipation, as found in other studies.¹⁷ Positive provider interaction and prioritization were cited as possible contributors to racial disparities.¹⁸ Our results revealed that higher education and a higher poverty-income ratio (corresponding to higher income status) lower the risk of constipation. As seen in the studies of Ozturk et al¹⁹ and Potter et al,¹⁸ a lack of knowledge and income can disrupt digestive health significantly. Unexpectedly, we found that older respondents had a lower chance of constipation. Although Stewart et al reported a similar decreasing prevalence in a more elderly group,²⁰ we believe that stool types may be harder to recall for the elderly population, which may contribute to an underreporting of constipation. We also found that active smokers had lower odds of constipation, which may be explained by a lower colonic transit time.²¹ Although we found lower odds of constipation in patients who are overweight, Sadik et al suggested that overweight patients may have faster colonic and rectosigmoid transit, which may explain our results. Lastly, supplements and constipation drug usage were associated with lower and higher odds of constipation, respectively. Medications are a well-known cause of constipation, and our results align with the current literature.⁹ Due to the limitations of the supplement component, we hypothesized that supplements may contain digestive enzyme supplements, which are widely advertised to the general public,²² thus contributing to the lower odds of constipation.

We found a reduction of constipation when comparing vegetable consumption in each quartile, even after adjusting using propensity score matching. Although prior research supported the view that the consumption of vegetables decreases the prevalence of constipation,^23,24 this is the first study to understand the amount of vegetable intake and the association with constipation. As seen in patients with fruit intake, a similar association of decreasing constipation odds was observed.^25,26 Several factors may contribute to this result. First, vegetables are rich in fiber and can hold large amounts of water and soluble ingredients, which can lead to stool softening and higher stool frequency.²⁷ Therefore, we hypothesized that fiber in vegetables might be one explanation. Second, vegetables can modulate various gut microbiota such as Ruminococcus and Lachnobacterium,²⁸ and studies have reported the roles of these microbiota in alleviating constipation.^29,30 Finally, vegetables may cause changes in colonic transit time. Kelsay et al³¹ showed that vegetables decreased colonic transit time and increased the number of defecations. Although current studies are limited and focus on only some fruit types,³² additional studies on colonic transit time with regard to fruit intake will be beneficial to understanding this association.

We also explored the association between the amount of vegetable intake and the odds of constipation. Current USDA guidelines recommend 2 to 3 cups of daily fruit intake (equivalent to 4 to 5 servings),³³ as well as 2 to 3 cups of vegetables. We found that most participants exceeded this recommended amount. Our study revealed that a larger amount of vegetable consumption was associated with lower odds of constipation. One possible reason is that the fiber within these vegetables has a dose-response relationship, as seen in breast cancer risk³⁴ and chronic pulmonary obstructive disease.³⁵

A significant strength of this study lies in its large, nationally representative sample. This approach mitigates the referral bias often encountered in single-center cohort studies. The diversity of the patient population accurately portrays the diverse socioeconomic, dietary intake, and racial differences throughout the United States. Moreover, the sizeable dataset bolsters the study’s statistical power, facilitating the identification of even subtle differences between groups.³⁶ Finally, rigorous adjustments for patient demographics, socioeconomic characteristics, and medication and supplement usage, as well as propensity score matching, help to reduce the influence of potential confounding factors.

This study is not without limitations. The administrative and cross-sectional nature of the NHANES database restricted the acquisition of other patient-level data, such as radiographic, echocardiographic, and colonic investigative results, which are essential for stratifying patient severity and accurately characterizing their health status. In addition, the lack of detail on vegetable intake, such as duration and rationale of eating, may reduce the level of granularity in applying these findings to practice. Furthermore, the database’s primary focus as a questionnaire introduces the possibility of neglecting important postquestionnaire outcomes, such as long-term mortality and the development of complications, as well as the issue of recall bias. Finally, the limitations of observational studies hinder the establishment of causal relationships between vegetable intake and constipation due to the inability to definitively determine the temporal sequence between the disease and the observed outcomes.

In conclusion, our findings underscore the relationship between constipation and vegetable consumption, highlighting lower constipation odds and the possibility of a dose-response relationship between vegetable consumption and constipation. Further epidemiological investigations are necessary to elucidate the causal mechanisms underlying the observed associations between vegetable consumption and constipation. Additional studies exploring the quantitative amount of vegetable intake and constipation should be conducted to understand this relationship. Encouraging the consumption of vegetables, in larger than recommended amounts, may be one possible intervention to alleviate constipation in the future.

CONFLICT OF INTEREST

The authors report no funding or conflicts of interest.