Study Participants
We extracted detailed data on respondents from the NHANES database of cross-sectional studies from 2005–2010 for analysis and research. The data for these three cycles included a bowel health questionnaire (BHQ). These data were collected using a hierarchical multilevel probability design, allowing for a weighted analysis of the population. The study was authorized by the National Center for Health Statistics’ Ethics Review Board. Each participant signed an informed consent form. The work on this database was organized by the National Center for Health Statistics of the Center for Disease Control and Prevention (Atlanta, GA, USA).
Data on 17 132 people aged 20 and older were collected.
17132 people aged 20 and older were collected from 2005-2010 in NHANES. Data for 2541 individuals that lacked information on stool consistency and frequency were removed. We also excluded 237 people with dietary phosphorus deficiency, as well as 379 pregnant women. Further 27 participants with phosphorus levels > 4000mg were eliminated. The detailed flow chart of population selection is presented in Figure 1.
Definition of Constipation
We defined chronic constipation using the description of stool frequency and consistency in the detailed bowel health questionnaire used in the following three cycles: 2005–2006, 2007–2008, and 2009–2010. The Bristol Stool Form Scale divides stool consistency into 7 types. The first two categories were regarded as constipation, while the remaining types were considered as non-constipation.
When collecting information on stool frequency, participants were asked the following: “How many times a week do you usually have a bowel movement?” Constipation was defined when stool frequency was not more than two bowel movements per week, and non--constipation when stool frequency was more than two bowel movements per week.
A sensitivity analysis was implemented using three constipation-related symptoms. We classified non-constipation and constipation according to self-reported constipation questionnaires. Laxatives were divided into used and not used. If people answered most days, use was defined as frequent. The following situations were considered infrequent: use of laxatives 2–3 times a month, once a month or 1–3 times a month. Because these three parameters were only used in the 2009–2010 cycle, we carried out the sensitivity analysis for this cycle only[11].
Phosphorus Intake
The collection of dietary information has been previously described[12-14]. The multi-pass approach was used to collect 24-hour recall of dietary phosphorus intake. Accurate information on participants’ food and drink intake was collected during the 24-hour period. Participants were each interviewed twice, and their dietary recalls were collected on both occasions. The first was a face-to-face interview with participants at the Mobile Examination Center, and the second was conducted by telephone 3–10 days later. We used the mean of the two diet recalls for the calculation, and if the participants had information missing from the second diet, we used the information from the first.
Covariates
After referring to previous reports on phosphorus and gut-related diseases, we included the following relevant covariates: milk consumption, energy, total fat, dietary fiber, selenium, magnesium, calcium, sodium, potassium, smoking, alcohol consumption, diabetes, hypertension, depression, poor oral health, income-poverty ratio, physical activity, body mass index (BMI), education, age, sex, and ethnicity [13-15]. We divided age into three groups (<45, ≥45<65, and ≥65 years old). Ethnicity included the following race categories: Non-Hispanic Black, Non-Hispanic White, Mexican American, Other Hispanic, and Other. Education level was divided into the following categories: grades 0-12, some college or above, and high school grad/GED. Income-poverty ratio (%) was classified into two types: <2 and ≥2. Based on the recommendations of the United States Departments of Health and Human Services, we used the weekly metabolic equivalent of task to classify physical activity as inactive (<500) or active (≥500). We defined participants who had smoked <100 cigarettes in their lifetime as never smoking, those who had smoked >100 cigarettes but did not smoke currently as former smokers, and those who smoked more than 100 cigarettes and still smoked on some days or every day as now smokers. We referred to participants who drank 12 or more alcoholic beverages per year as drinkers, otherwise not. We divided the frequency of milk consumption into never drinking, <once a week as rarely drinking, ≥once a week but <once a day as sometimes drinking, and ≥once a day as frequent drinking. BMI (kg/m2) was divided into three categories: obese (≥30), overweight (25–29.9), and under/normal weight (<25). In all of the following cases, we diagnosed the participants as having diabetes: First, the doctor diagnosed the participant with diabetes mellitus. Second, the test result of the oral glucose tolerance test or random blood glucose was greater than or equal to 11.1 mmol/L. Third, the participant’s fasting blood glucose value was greater than or equal to 7.0 mmol/L. Fourth, the glycohemoglobin HbA1 of the respondents was greater than or equal to 6.5%. Fifth, participants received oral diabetes medication or intramuscular insulin. We used hypertension medication use, hypertension-related questionnaires, and systolic and diastolic blood pressure readings to determine whether or not the participant had hypertension. Participants with a PHQ-9 score of 10 or greater were diagnosed with depression. Dietary information on total fiber (T1<11.6; T2, 11.6–18.0; T3≥18.1 g/day), total fat (T1<55.4; T2, 55.4–85.0; T3 ≥ 85.1 g/day), protein (T1<61.0; T2, 61.1–88.1; T3≥88.2 g/day), calcium (T1<644.5; T2, 644.5–1004.0; T3≥1004.1 mg/day), sodium (T1<2523.5; T2, 2523.5–3689.0; T3≥3689.1 mg/day), potassium (T1<2060.0; T2, 2060–2910.0; T3≥2910.1 mg/day), and total energy (T1<1590.0; T2, 1590.0–2247.0; T3≥2247.1 kcal/day) intake was collected by trained interviewers.
Statistical Analyses
We explored the relationship between changes in dietary phosphorus intake and the prevalence of chronic constipation in the NHANES dataset using statistical methods for weighted sampling. As oversampling had occurred in some populations, the weighted statistical method was applied to the data, in order to avoid bias and ensure the accuracy of results. We set the continuous variables as categorical variables through the study of relevant literature. Categorical variables are expressed as weighted percentages and confidence intervals (95% CI). Furthermore, we investigated the relationship between the two using several multiple logistic regression models adjusted for relevant confounders. Model Ⅰ is the unadjusted model. In model Ⅱ the data were adjusted for participants’ age (<45; ≥45<65; ≥65 years old), sex, and ethnicity (Non-Hispanic White, Non-Hispanic Black, Mexican American, Other Hispanic, and Other). Model-Ⅲ continued to adjust for the following variables on the basis of model Ⅱ: income-poverty ratio (%) (<2, ≥2, or missing), physical activity (<500, ≥500, or missing), BMI (kg/m2) (<25, 25–29.9, and ≥30), poor oral health (yes, no), hypertension (yes, no), depression (yes, no), diabetes (yes, no), smoking (never, former, and now), alcohol consumption (yes, no), milk consumption (often, sometimes, rarely, never), and energy (T1<1590.0; T2, 1590.0–2247.0; T3≥2247.1 kcal/day).
In addition, we used smooth curve fitting after adjusting for confounding factors to show more intuitively the relationship between the two. In the curve fitting graph, the middle line represents the effect size, and the area on both sides of the line represents the 95% confidence interval. We further applied interaction, stratified analysis, and univariate analysis based on the variables, including age (<45, ≥ 45<65, and ≥ 65 years old), sex, and ethnicity (Non-Hispanic White, Non-Hispanic Black, Mexican American, Other Hispanic and Other), BMI (kg/m2) (<25, 25–29.9, and ≥ 30), physical activity (<500, ≥500, or missing), poor oral health (yes, no), hypertension (yes, no), depression (yes, no), diabetes (yes, no), smoking(never, former, and now), alcohol consumption (yes, no), milk consumption (often, sometimes, rarely, never), energy (T1<1590.0; T2, 1590.0-2247.0; T3≥2247.1 kcal/day), and income-poverty ratio (%) (<2, ≥2, or missing). Dummy variables were used to indicate missing covariate values[16].
P<0.05 was considered statistically significant. All statistical analyses were performed using R packages (The R Foundation; http://www.r-project.org; version 3.4.3) and Empower (R) (www.empowerstats. com, X&Y solutions, inc. Boston, Massachusetts).