In this study, the dataset from the CHARLS was analyzed. CHARLS was a longitudinal project initiated in 2011 and followed up every 2 years in China (14). Adults aged ≥ 45 years in 28 provinces were sampled and investigated. To test the association between the CircS and LUTS/BPH and compare the predictive value between CircS and MetS, the 2011 baseline and 2015 follow-up data were employed to construct the retrospective cohort. In the baseline survey, 8,471 men participants were recruited. After cleansing the dataset, 4,390 men were enrolled into analyses in the baseline survey. Of them, 3,658 men were followed up until 2015 (Figure 1).

The concept of CircS was first proposed by Zimmet et al. in 2019 based on the concept of MetS and then defined with a cutoff value by Shi et al. (12, 15). In the definition of CircS, another two risk factors, which were depression and reduced sleep duration, were clustered with the components of MetS including abdominal obesity, hypertension, hyperglycemia, high triglycerides, and low high-density lipoprotein (HDL) cholesterol. In this study, abdominal obesity (waistline ≥ 85 cm), hypertension (systolic pressure ≥ 130 or diastolic pressure ≥ 85 mmHg or drug treatment for hypertension), elevated glucose (≥100 mg dl⁻¹ or drug treatment for elevated glucose), high triglycerides (≥150 dl⁻¹ or drug treatment for high triglycerides), and low HDL cholesterol (<40 mg dl⁻¹ or drug treatment for low HDL cholesterol) were defined according to the previous studies (10). The specific cutoff values of the seven components in CircS and MetS were presented in Supplementary Table 1. According to one previous study, the cutoff for MetS was set as ≥3 components, while for CircS, the cutoff was ≥4 components (15).

The participants were asked to remain fasting in the night, and the researchers collected their venous blood in the next morning. The venous blood was then centrifuged to separate the plasma and stored at −80°C for further determining the triglycerides, glucose, C-reactive protein, uric acid, creatinine, HDL cholesterol, and hemoglobin using enzymatic colorimetric tests. In the morning, when the participants remained seated and quiet for a while, the blood pressure was then measured for three times. These three measurements were recorded, and their mean was calculated to diagnose hypertension. Waist circumference was determined using a measuring tape at the navel level when the participants stood still.

Reduced sleep duration was defined as sleeping <6 h. Depression was assessed using the Epidemiological Studies Depression Scale-10 (CESD-10). Scores ≥ 10 were diagnosed as depression (16). Furthermore, all the relevant drug treatment histories for elevated glucose, low HDL cholesterol, high triglycerides, and hypertension were recorded through a personal face-to-face interview. The collection of blood samples and further tests for blood biomarkers were all performed by professional medical staffs, and the personal face-to-face interviews were finished by at least two staffs together.

Generally, BPH is a term based on histological alterations. Several men with BPH are asymptomatic who, therefore, do not undergo further examinations, diagnosis, and treatments. Enlarged prostate can compress the urethra and then lead to frequent voiding, nocturia, urine retention, etc., namely, LUTS. This condition severely affects men only when it correlates with LUTS. In light of this circumstance, “LUTS attributed to BPH” was abbreviated as LUTS/BPH and used in the previous studies (17). In CHARLS, participants were interviewed if they have ever been diagnosed with a prostate illness (excluding prostatic cancer). Related LUTS, including frequent voiding, nocturia, and urine retention, were interpreted to them. If “Yes,” the participants were defined as suffering LUTS/BPH. This method for diagnosing LUTS/BPH was also adopted in the previous studies (6, 18).

In this study, the following covariates were summarized: age, marital status [married/cohabitated, others (divorced, widowed, never married, and separated)], education, smoking (current smoker, non-smoker, and exsmoker), alcohol consumption (drinker or non-drinker), uric acid, anemia (<120 g/l), stroke (yes or no), and chronic kidney disease (CKD, yes or no). The estimated glomerular filtration rate (eGFR) was calculated to evaluate the kidney function using CKD-EPI creatinine equation according to one previous study (15). The cutoff for diagnosing CKD was set as <60 ml min⁻¹/1.73 m² (15).

In this study, continuous variables with normal distribution were displayed as mean ± SD and those with non-normal distribution were showed as median (25–75% quantiles). Furthermore, categorical data were summarized as proportions (%). Differences within the groups were tested using one-way ANOVA, Kruskal–Wallis, or chi-square test according to the data type and distribution. Furthermore, multiple comparisons between groups were adjusted by the Bonferroni method.

To test the cross-sectional relationship between CircS, MetS, and LUTS/BPH, logistic regression was used. Moreover, in this part, propensity score matching (PSM) using MatchIt package was employed for sensitivity analyses. The 1:1 nearest neighbor matching was used to balance the differences between covariates, and then, ORs were calculated. Furthermore, we excluded the participants who had LUTS/BPH at the baseline to test the longitudinal relationship between CircS, MetS, and LUTS/BPH. P < 0.05 (two-sided) were seen as significant in statistics.

To compare the predictive power using the number of CircS and MetS components, receiver operating characteristic (ROC) curve analysis was used. PROC package was adopted to calculate the AUCs and draw the ROC curve. The DeLong test (one-sided) was employed to compare the difference of area under curves (AUCs). All the statistical analyses were performed using R 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).

A total of 4,390 participants attending the CHARLS baseline survey in 2011 were enrolled (Table 1). The mean age of the enrolled participants was 60.23 (SD = 9.33). The participants were divided into the following four groups: normal, CircS alone, MetS alone, and both CircS and MetS. It was noted that CircS alone group had higher prevalence of stroke, anemia, and CKD than the MetS alone group (3.77 vs. 1.94%; 7.05 vs. 3.21%; and 5.03 vs. 3.87%, Figure 2). Moreover, all the CircS alone group had reduced sleep duration and depression. None of the participants in MetS alone group had depression, and of them, only 3.16% had reduced sleep duration.

It was indicated that both CircS and MetS correlated positively with prevalent LUTS/BPH. In Figure 3A, the numbers for CircS and MetS components were set as independent variables. All the ORs for CircS were higher than those for MetS in all the four models, suggesting that patients with CircS had higher risks for prevalent LUTS/BPH than patients with MetS. In Figure 3B, after adjusted for all the covariates (Model 4), the OR value for CircS was 1.61 (95% CI 1.29–2.00, P < 0.001), while that for MetS was 1.34 (1.08–1.66, P < 0.01). To balance the differences between covariates, PSM analysis was performed. After PSM, all the mean differences were <10%, indicating a good attenuation of the disparities between the groups. It was revealed that the CircS group had 1.59 times higher risks for prevalent LUTS/BPH (95% CI 1.22–2.06, P < 0.001) than their counterparts and that the ORs for the MetS group were 1.33 (1.04–1.68, P < 0.05). To detect the risks for patients who had both CircS and MetS, this subgroup was separated and subjected to further analyses (Figure 3C). After adjusted for all the covariates (Model 4), it was revealed that patients having both CircS and MetS had 1.59 times higher risks for prevalent LUTS/BPH than the normal participants (95% CI 1.25–2.02, P < 0.001), which was lower than that for the CircS alone group (ORs: 1.59 vs. 1.78), but higher than the MetS alone group (ORs: 1.59 vs. 1.05).

The patients with LUTS/BPH in 2011 were excluded, and the rest of the participants were followed to 2015. ORs for CircS were higher than those for MetS in all four models. In Model 4, it was disclosed that having CircS was linked to 150% increased risks of developing LUTS/BPH, while the increased risks for MetS were 132% (Table 2).

To compare the predictive power of CircS and MetS, the ROCs were drawn (Figure 4). For the prevalent LUTS/BPH, the AUCs for CircS and MetS were 0.582 (0.553–0.611) and 0.556 (0.526–0.585), respectively. For the incident LUTS/BPH, the AUCs for CircS were 0.574 (0.540–0.609), which was higher than that for MetS [AUC = 0.561 (0.526–0.597), P < 0.05]. It indicated that CircS had better predictive powers than MetS both for prevalent and incident LUTS/BPH.