Reduced sleep duration increases the risk of lower urinary tract symptoms suggestive of benign prostatic hyperplasia in middle-aged and elderly males: a national cross-sectional study

Abstract Background The prevalence of lower urinary tract symptoms suggestive of benign prostatic hyperplasia (LUTS/BPH) remains high in men. However, whether reduced sleep duration enhances the risk of LUTS/BPH remains unknown. Materials and methods The 2015 China Health and Retirement Longitudinal Study was used in this study. Binary logistic regression was adopted to test the relationship between sleep duration and LUTS/BPH. Restricted cubic spline (RCS) regression was used to examine the non-linear association. In sensitivity analyses, propensity scores matching was performed to verify the robustness of the results. Results In this study, 8,920 males aged 40 years above were enrolled. In the fully adjusted logistic model, across the quartiles of sleep duration, the odds ratios of LUTS/BPH were 1.00 (reference), 0.94 (95% CI 0.77–1.15), 0.74 (95% CI 0.58–0.94), 0.54 (0.37–0.75), respectively. The results of RCS indicated a non-linear inverted U-shaped association between sleep duration and LUTS/BPH (p for non-linearity <0.05). In the subgroup analyses, no significant effects of settlements, alcohol and cigarette consumption, depression, and hypertension on the association between sleep duration and prevalent LUTS/BPH were observed (p for interaction >0.05). Conclusion Reduced sleep duration is significantly associated with the increases of the LUTS/BPH risk in Chinese middle-aged and elderly males.


Introduction
Benign prostatic hyperplasia (BPH) is prevalent in middle-aged and elderly men [1]. According to an epidemiological study, it affects 11.97% of elderly men in China [1]. The enlarged transitional zone of the prostate will compress the urethra and then lead to overactive bladder and lower urinary tract symptoms (LUTS) [2]. Mounting evidence shows that LUTS can be caused by various conditions. However, in males, it is well recognized that BPH is the leading cause [2]. LUTS attributed to BPH (LUTS/BPH) severely aggravates the life of patients. This condition is also closely correlated with other serious diseases, such as depression and stroke [3,4], which makes patients more painful and motivates researchers to explore its risk factors and develop effective treatments.
Existing literature has revealed numerous risk factors. One previous study reported that depression is an independent risk factor for LUTS/BPH [3]. Under a two-year follow-up period, depressive males displayed 2.1-folder risk of LUTS/BPH than the non-depression counterparts [3]. What's more, aging, obesity, inflammation, circadian syndrome, metabolic syndrome and dietary pattern were found to improve the risks of LUTS/BPH significantly [5][6][7][8]. Recently, the role of circadian rhythm in the occurrence of LUTS/BPH has been noted [6]. It is postulated that dysregulated circadian clock genes are involved with the pathological alterations of BPH. However, little evidence is observed in clinical and laboratory settings as of now. As a pivotal part of circadian rhythm, the role of sleep duration in the occurrence of LUTS/BPH remains requisite to be explored.
Sleep is a highly conservative behavior among animals. Generally, adults are recommended to sleep 7-8 h per day. But currently, it is frequently violated in modern society due to numerous reasons such as shift work and entertainment. In the US, only 65.2% adults report a healthy sleep duration [9]. Insufficient sleep duration places heavy adverse effects on human health. Previous studies have revealed that reduced sleep duration increased the risk of subclinical multiterritory atherosclerosis, depression, metabolic syndrome and dementia [10][11][12][13]. But to date, no study confirms the risky role of reduced sleep duration in the suffering of LUTS/BPH, which is detrimental to suppress the occurrence of LUTS/BPH. To explore the association between sleep duration and LUTS/BPH, we used a dataset from a national longitudinal survey, the China Health and Retirement Longitudinal Study (CHARLS), and performed this study.

Data source and analytical sample
To explore the relationship between sleep duration and LUTS/BPH, CHARLS 2015 survey dataset was used. CHARLS is an ongoing project designed to assess the current socioeconomic, physical and psychological status of the Chinese aged population [14]. The middle-aged and elderly population was sampled, as the previous study revealed [14]. Currently, four waves (2011, 2013, 2015, and 2018) are issued, and only the 2011 and 2015 survey collected blood samples for determining blood biomarkers, such as C reactive protein, blood urine nitrogen. Since blood biomarkers were not measured in the 2018 survey, we used the CHARLS 2015 survey dataset to perform this study.
CHARLS 2015 survey was performed in 28 Chinese provinces. This survey included 21,100 participants with respondents, non-response samples in previous waves, and fresh samples in this wave. After removing samples who did not meet the data cleansing principles in Supplemental Figure 1, a total of 8,920 males were remained and further analyzed in this study.

Measurements of variables: sleep duration and LUTS/BPH
All the participants were interviewed face to face by two well-trained researchers. The evaluation of sleep duration was mainly based on a questionnaire. The participants were asked, "During the past month, how many hours of actual sleep did you get at night (average hours for one night)? (This may be shorter than the number of hours you spend in bed)." Answers from respondents were recorded by the researchers.
BPH is the leading cause of LUTS, and exactly, this term is based on pathological evidence. However, parts of BPH cases are asymptomatic and therefore not self-recognized till medical examinations or the emergence of LUTS. In this survey, the asymptomatic cases cannot be evaluated. Therefore, "LUTS attributed to BPH" was abbreviated as LUTS/BPH and used in this study and previous literature [1,6,15].
In CHARLS, male participants were further interviewed that "Have you ever been diagnosed with a prostate illness, such as prostate hyperplasia (excluding prostatic cancer)?" The symptoms of BPH were interpreted to the participants as follows: "The main symptoms of benign prostatic hyperplasia are difficulty in urination. For mild cases, patients get up at night for urinating more frequently than their healthy counterparts. Moreover, mild cases also have increased frequency and urgency of voiding. In severe cases, it is noted that patients have a slow urinary stream, even urine retention, generally accompanied by backache, limb weakness, spermatorrhea, etc." If positive, the respondents were noted as LUTS/BPH cases. The same approaches were also used in previous studies [1,6,15].

Covariates
In this study, demographic features, physical and psychological health status, blood biomarkers, and medical histories were recorded. Age, settlements (rural/ urban), marital status, cigarette consumption (current/ none/ever-smoker) and alcohol consumption (drink more/less than once a month or non-drinker), waist circumference, body mass index (BMI), time of afternoon nap, depression (yes or no), stroke (had or had not), hypertension (yes or no), arthritis (yes or no), kidney diseases excluding cancers (yes or no), liver diseases excluding cancers and fatty liver (yes or no), stomach diseases excluding cancers (yes or no), chronic kidney disease (CKD), anemia (<120 g.L À 1 ), high-sensitivity C reactive protein (CRP), uric acid, lowdensity lipoprotein (LDL), triglyceride, blood urea nitrogen (BUN) were investigated or determined.
Among the variables, depression was assessed using the Center for Epidemiological Studies Depression Scale-10 (CESD-10) questionnaire [3]. The cutoff for diagnosing depression was set as �10 [3]. CKD was diagnosed according to estimated glomerular filtration rate (eGFR), calculated by the CKD-EPI creatinine equation. The cutoff was set as <60 ml.min À 1 per 1.73 m 2 [6]. All the venous blood samples were collected by professional medical staff, and the blood biomarkers were further determined using enzymatic colorimetric tests.

Statistical analyses
In this study, variables were described as mean ± standard deviation (SD), proportions (%) and median (25-75% quartile) according to data types. To test the differences of covariates among different sleep duration groups, the Chi-square test, ANOVA, and Kruskal-Wallis tests were used in categorical and continuous variables, respectively. Further, correlation coefficients were calculated using Pearson or Spearman tests to avoid collinearity before performing regressions, which were also confirmed by variance inflation factors (Supplemental Figure 2 & 3). To evaluate the association between sleep duration and prevalent LUTS/BPH, the multivariate binary logistic regression was performed. The first quartile of sleep duration (Q1) was set as the reference group. A set of logistic regression models were built: Model 1adjusting for age and settlements; Model 2 -further adjusting for cigarette and alcohol consumption; Model 3 -further adjusting for waist circumference, BMI, and afternoon nap based on Model 2; Model 4further adjusting for self-reported hypertension, stroke, arthritis, stomach diseases, kidney diseases, liver diseases, CKD and depression based on Model 3; Model 5 -further adjusting for uric acid, low-density lipoprotein and CRP. It should be noted that anemia and marital status were not enrolled into regression models for insignificance in univariate logistic regression (p > 0.1).
The relative importance of variables included in Model 5 was measured as the partial chi-square statistic minus the predicted degrees of freedom (v 2 -df). Additionally, subgroup analyses were performed. The multiplicative interaction between sleep duration and covariates was tested by adding a cross-product term in the full multivariate logistic regression model, such as age, settlements, depression, and CKD. To further flexibly model and visualize the relationship between sleep duration and LUTS/BPH, restricted cubic spline (RCS) regression was used with three knots (at 10, 50, 90 quartiles). According to the results of RCS, we recoded the variable, sleep duration, as a binary variable (<6 h & �6 h) to explore the increased risks of insufficient sleep duration. Finally, propensity scores matching (PSM) was adopted to balance the disparities between covariates and confirm the results. To perform the sensitivity analyses and confirmed the results again, we also interpolated all the missing values of covariates using multivariate imputation by chained equations based on random forest methods. The interpolated dataset was reanalyzed again as we did before. All the analyses and figures were made using R3.6.3 (R Foundation for Statistical Computing, Vienna, Austria). p < 0.05 (two-sided) was seen as significant in statistics.

Clinical profiles of respondents attending the CHARLS 2015 survey
In the survey, 8,920 respondents were enrolled in the final analyses. In Table 1, the clinical profiles are shown. The mean age of the respondents was 59.77 (SD ¼ 10.03). Across the quartiles of sleep duration, the mean (SD) of sleep duration (hours) was 4.27 ± 1.02, 6.45 ± 0.49, 7.97 ± 0.12, and 9.71 ± 0.95, respectively. Table 1 indicates that males in the first quartile are more likely to have a higher prevalence of CKD, anemia, depression, arthritis, stomach diseases, and kidney diseases (p < 0.05).

The association between sleep duration and prevalent LUTS/BPH
A total of 1,071 LUTS/BPH cases were recorded in this survey. The relative importance of covariates in the fully adjusted model 5 was displayed in Supplemental 4. As shown in Supplemental Figure 4, age was the  Table 2). The negative association between sleep duration and LUTS/BPH was observed in the third and fourth quartiles of sleep duration in all the five models (all the p values <0. 05). Compared with the first quartile of sleep duration, the ORs were 0.82 (95% CI 0.70-0.96) in the second quartile, 0.62 (95% CI 0.51-0.75) in the third quartile, and 0.51 (95% CI 0.39-0.67) in the fourth quartile, when adjusted for age and settlements (p for trend <0.001). Compared with model 1, the association became slightly stronger in model 2 and model 3, adjusting for cigarette and alcohol consumption, waist circumference, BMI, and afternoon nap (all the p values <0.05). In model 4 and model 5, the decreased risk remained consistent in the third and fourth quartiles (ORs: 0.74 in Q3, p < 0.05; 0.54 in Q4, p < 0.01), but not significant in the second quartile in statistics (OR: 0.94, p > 0.05). However, the declining trend of ORs remained consistent.

The subgroup analyses in settlements, cigarette and alcohol consumption, depression and, hypertension
In the subgroup analyses (Table 3), we did not observe significant effects of settlements, alcohol and cigarette consumption, depression, and hypertension on the association between sleep duration and prevalent LUTS/BPH (p for interaction >0.05). The decreased ORs in the fourth quartile were significant in urban and rural areas, non-drinkers, non-smokers, and respondents without depression, with or without hypertension (p < 0.05).

Restricted cubic spline regression
Considering the significance of p-trend, we further used RCS regression to examine whether the non-linear relationship between sleep duration and prevalent LUTS/BPH existed or not. The results of RCS ( Figure  1(A)) indicated that the ORs of LUTS/BPH increased when the respondents slept less than six hours and declined when participants slept more than six hours, which did not support the linear association (p for non-linearity ¼ 0.01). An inverted U-shaped relationship between sleep duration and LUTS/BPH was noted according to Figure 1(A).
Since the trend of ORs altered in six hours, we recoded the variable, sleep duration, as a binary variable (<6 h & �6 h) and performed the multivariate logistic regression (Figure 1(B)). Respondents with sleep duration �6 h had lower OR than those <6 h (OR: 0.75, 95% CI ¼ 0.63-0.89, p ¼ 0.001). To confirm the efficacy of balancing the disparities among covariates, a love plot was drawn and displayed in Supplemental Figure 5. It showed that all the standardized biases were <10% after PSM. The differences between the groups were well-balanced. After PSM, this declining trend remained consistent (OR: 0.76, 95% CI ¼ 0.64-0.96, p < 0.01).

Sensitivity analyses
In this part, missing values of covariates were interpolated using multivariate imputation by chained equations based on random forest methods. The interpolated dataset was reanalyzed as we did in the "3.2" section. The results are shown in Table 4. As indicated in Table 2, ORs in the third and fourth quartiles decreased significantly in all the five models (p < 0.05), which confirmed the robustness of the results in Table 2.

Discussion
In recent years, aging has been noted globally, especially in China. As a prevalent disease in middle-aged and elderly men, LUTS/BPH severely affects their daily life. Mounting studies explore its risky factors from different perspectives, but few studies notice the role of sleep in the occurrence of LUTS/BPH, which we explore in this study. In this national cross-sectional study, respondents with reduced sleep duration are found to have higher odds of LUTS/BPH. Previous studies mainly reported that males with LUTS/BPH have bad sleep quality due to nocturia [16,17]. These findings highlight the poor sleep quality of LUTS/BPH patients, but the causal direction is opposite to what we report in this paper. One previous study by Chou PS et al. found that males with sleep apnea (SA) had 2.35-folder higher risks of BPH than the non-SA participants [18]. Especially for males aged 51-65, the hazard ratio was 5.59. However, SA patients are reported to suffer hypoxia and following systemic inflammation, such as tumor necrosis factor-a and interleukin 6, which induce the development of LUTS/BPH [19]. It mainly reminds the possible risky role of reduced sleep duration in the occurrence of LUTS/BPH. In our study, an inverted U-shaped relationship between sleep duration and LUTS/BPH is found. Enough sleep duration may protect males from suffering LUTS/BPH.
Although a close association between sleep duration and LUTS/BPH is identified, the specific molecular mechanisms remain unclear. There are several possible explanations. First, the circadian clock may be one possible mediator linking sleep to LUTS/BPH. Generally, sleep is considered a core part of the circadian rhyme [20]. Reduced sleep duration will lead to circadian dysfunction, followed by dysregulated circadian clock genes, such as Per 2, Per 3 [21]. As revealed by Li et al., decreased expression of Per 2 could inhibit apoptosis, leading to BPH [22]. Hence, good sleep quality may help maintain the circadian rhyme. Results of logistic regression before and after PSM. "rms" and "MatchIt" package in R 3.6.3 were used to perform RCS and PSM analyses. In Figure 1(A), the RCS regression model was adjusted as model 4 in Table 2. In Figure 1(B), the logistic regression model built before PSM was adjusted as model 4 in Table 2. RCS: restricted cubic spline; PSM: propensity scores matching; LUTS/ BPH: lower urinary tract symptoms suggestive of benign prostate hyperplasia; CI: confidence interval. Medications conducive to sleep and LUTS may benefit patients with BPH, but may further induce erectile dysfunction [23]. Moreover, patients with sleep disturbance usually have metabolic problems [12]. One previous study reported that men with metabolic syndrome have 1.47-folder risk of LUTS/BPH [12]. And as indicated by Russo G I et al., patients with insulin resistance have higher odds of severe LUTS [24]. Therefore, metabolism may be involved in the association between reduced sleep duration and LUTS/BPH. The mechanisms linking sleep duration and LUTS/BPH may be various pathway and need further confirmation in future studies. The subgroup analyses show that the protective effects are evident in non-drinkers, non-smokers, or males without depression, but not significant in their counterparts. Participants exposed to cigarettes and alcohol have higher odds of metabolic disorder [25,26]. Moreover, cigarette consumption can increase alcohol consumption [27]. Harms from the metabolic disorders may offset the protective effects of enough sleep duration. Additionally, according to one previous study in Chinese aging males, depression will increase the risk of LUTS/BPH [3]. The adverse effects of depression may lead to insignificance in our results.
Although a national survey is performed in this study, some limitations still exist. The major demerit is the simplified method to define LUTS/BPH, which was used according to Zhang's approach [15]. The definition of LUTS/BPH was mainly based on self-report as opposed to more objective tests such as prostatic ultrasonography, which may bias the diagnoses. Additionally, the cross-sectional design limits the confirmation of causal effects. Future cohort studies should be performed to verify the conclusion. Moreover, this study is mainly conducted in Chinese. The conclusion still needs further validation in other countries under different socioeconomic status.
To conclude, this study shows that reduced sleep duration enhances the risks of LUTS/BPH. A non-linear inverted U-shaped association exists between sleep duration and LUTS/BPH. For aging males, enough sleep duration may alleviate LUTS/BPH.

Conclusions
This national cross-sectional study in China shows that reduced sleep duration is significantly associated with the increases of the LUTS/BPH risk in the middle-aged and elderly males.