Association between lipid accumulation product index and lower urinary tract symptom—benign prostatic hyperplasia: a 7-year follow-up study

2.1 Study population

This study used datasets from CHARLS, which is an ongoing project in China [10]. This project was initiated in 2011, sampling participants in 28 provinces (150 counties and 450 communities) at the baseline survey. The sampled participants were followed up in 2013, 2015 and 2018. All interviews were completed by well-trained researchers to improve the quality of the collected data. Detailed information of CHARLS can be obtained from previous publications [1, 10].

Based on CHARLS, a retrospective cohort was constructed and analyzed from 2011–2018. In 2011, a total of 17,693 participants were surveyed. After excluding females, participants in non-fasting status or having missing values and outliers of age, LAPI and LUTS/BPH, 3967 males were finally included in the cross-sectional analysis (Fig. 1). Among them, 3072 males (85.93%) were followed up to 2013, 2976 males (83.24%) were followed up to 2015, and 2855 males (79.86%) were followed up to 2018. We mainly analyzed the 2011–2018 cohort, and the results from the 2011–2013 and 2011–2015 cohorts were used as sensitivity analyses.

Fig. 1.Overview of the study design and analysis strategy. In CHARLS, we first investigated the cross-sectional association between LAPI and LUTS/BPH in 2011 and the longitudinal association was explored in a seven-year follow-up survey from 2011–2018. The 2011–2013 and 2011–2015 cohorts were used as supplements. According to the histogram of LAPI, 91 participants with LAPI <−30 or >200 were excluded as outliers. CHARLS: China Health and Retirement Longitudinal Study; LAPI: lipid accumulation product index; LUTS/BPH: lower urinary tract symptoms suggestive of benign prostate hyperplasia; BMI: body mass index.

2.2 Measurements of LAPI and LUTS/BPH

LAPI was calculated as previously reported: LAPI = [waist circumference (cm) − 65] × triglycerides (mmol/L) [11]. To determine the concentration of blood lipids, the participants were asked to fast overnight. The next morning, venous blood was collected and immediately centrifuged to obtain plasma, which was then frozen at −20 °C for transport. The plasma samples were finally stored at −80 °C until determination. An enzymatic colorimetric test was employed to measure the concentration of lipids. The waist circumference was assessed as one previous study stated and were recorded in centimeters [12].

In histology, the enlarged prostate gland can compress the urethra, further triggering LUTS. In this study, LUTS/BPH was not defined based on pathological section and ultrasound examination due to the heavy workload in national cohorts. To simplify the process, the diagnosis of LUTS/BPH was based on self-reported doctor-diagnosed prostate illness. The male participants were asked “Have you ever been diagnosed with a prostate illness, such as prostate hyperplasia, excluding prostatic cancer?”. Related symptoms of LUTS, such as frequent voiding and urine retention, were explained to the respondents. Participants with “Yes” were defined as having LUTS/BPH. This method has also been used in previous epidemiological studies [13, 14].

2.3 Measurements of covariates

Some covariates were adjusted in the regression models. Demographic variables included age (years), educational levels (literate and illiterate) and marital status (married with spouse/cohabitating versus divorced/separated/widowed). Lifestyle factors included sleep duration (hours), afternoon naps (minutes), cigarette consumption (current, never or ex-smoker), and alcohol consumption (more than once a month, less than once a month, and never). Demographic and lifestyle variables were evaluated based on self-reports. Medical histories included hypertension (yes or no) and depression (yes or no). Hypertension was defined as systolic pressure ≥140 mmHg, diastolic pressure ≥90 mmHg or history of antihypertensive drugs [14]. Depression was defined as a score of the Center for Epidemiological Studies Depression Scale-10 ≥10 [14]. Blood biomarkers consisted of low-density lipoprotein (LDL), high-density lipoprotein (HDL), total cholesterol, and blood urea nitrogen (BUN).

2.4 Statistical analyses

LAPI was stratified into tertiles from lowest (first tertile, T1) to highest (third tertile, T3). We first summarized the clinical characteristics of the included participants in the baseline survey. Continuous variables with a normal distribution are displayed as the mean ± standard deviation, and categorical variables are shown as proportions (%). The differences in covariates across tertiles were tested by the one-way analysis of variance for continuous variables and the Chi-square test for categorical data. Five logistic regression models were constructed to assess the cross-sectional and longitudinal association between LAPI and LUTS/BPH.

In addition, some sensitivity analyses were added to verify the main findings. First, LAPI was included in the five regression models as a continuous variable, instead of tertiles. Second, the missing values in the baseline survey were interpolated using multivariate imputation by chained equations [14]. The interpolated dataset was used to reperform the above stated logistic regression. Third, subgroup and interactive analyses were adopted to explore potential interactive effects. Finally, the association between LAPI and LUTS/BPH was verified after excluding participants with BMI ≥28 kg/m². Moreover, we used restricted cubic spline (RCS) regression to investigate the dose-response association between LAPI and LUTS/BPH. To explore the longitudinal association between LAPI and LUTS/BPH, three cohorts (2011–2013, 2011–2015 and 2011–2018) were constructed. We mainly analyzed the 2011–2018 cohort and the results from the 2011–2013 and 2011–2015 cohorts were used as supplements.

All analyses were performed using R 4.0.2 software (R Foundation for Statistical Computing, Vienna, Austria). A two-tailed p-value < 0.05 was considered the significance threshold for statistical analysis.

3.1 Characteristics of the baseline participants

As displayed in Table 1, 3967 males were included. Participants in the T3 group were younger; literate; married with spouse/cohabitating; slept longer; smoked less; had higher blood pressure, BMI and waist circumference; and had no depression (all p < 0.01) in relation to those in the other groups. For blood biomarkers, the T3 group had significantly higher triglyceride, uric acid, LDL and total cholesterol and lower HDL and BUN (all p < 0.001) than the other groups.

3.2 Cross-sectional association between LAPI and prevalent LUTS/BPH

As shown in Fig. 2, a high LAPI was positively associated with prevalent LUTS/BPH. In the crude model, the ORs for prevalent LUTS/BPH were 1.08 (95% CI = 0.83–1.40, p = 0.584) and 1.36 (95% CI = 1.06–1.76, p = 0.017) for the T2 and T3 groups, respectively. In the full model (model 5), the ORs for the T2 and T3 groups were 1.21 (95% CI = 0.91–1.62, p = 0.197) and 1.56 (95% CI = 1.09–2.23, p = 0.014), respectively. All five regression models supported the increased risk of prevalent LUTS/BPH for individuals with high LAPI (all p for trend < 0.05). In sensitivity analyses, LAPI was used as a continuous variable in analysis. In the crude and full models, the ORs were 1.004 (95% CI = 1.001–1.008, p = 0.008) and 1.007 (95% CI = 1.001–1.013, p = 0.016), respectively, with an increment of 1 unit of LAPI (Supplementary Table 1).

In addition, we also interpolated the missing values and then reanalyzed the datasets. In Supplementary Table 2, LAPI was used as a continuous variable. In the crude and full models, the ORs were 1.004 (95% CI = 1.001–1.007, p = 0.008) and 1.009 (95% CI = 1.003–1.015, p = 0.004), respectively, with an increment of 1 unit of LAPI. In Supplementary Table 3, LAPI was used as a categorical variable. In the crude and full models, the ORs for the T3 group were 1.36 (95% CI = 1.06–1.76, p = 0.017) and 1.59 (95% CI = 1.12–2.25, p = 0.009), respectively.

Fig. 2.The cross-sectional association between LAPI and prevalent LUTS/BPH. In 2011, 3967 males were enrolled in the analysis. The T1 group was used as the reference group. Model 1—crude model; Model 2—adjusting for age, educational levels and marital status; Model 3—further adjusting for sleep duration, afternoon nap, cigarette and alcohol consumption; Model 4—further adjusting for hypertension and depression; Model 5—further adjusting for blood biomarkers including uric acid, LDL, BUN, HDL and total cholesterol. OR: odds ratio; CI: Confidence Interval.

3.3 Subgroup and interactive analyses

In the subgroup analysis (Table 2), the increased risk of prevalent LUTS/BPH was also verified in males aged >60 years (OR = 1.45 for the T2 group and 1.67 for the T3 group), being illiterate (OR = 1.70 for the T2 group), never smoking (OR = 2.03 for the T2 group), and having hypertension (OR = 1.99 for the T3 group) or depression (OR = 1.85 for the T3 group). However, in the interactive analysis, no significant interactive effects of age, educational levels, marital status, cigarette and alcohol consumption, hypertension and depression were detected (all p for interaction > 0.05).

3.4 Dose-response association between LAPI and prevalent LUTS/BPH

The non-linear relationship between LAPI and LUTS/BPH was summarized in Fig. 3. In Fig. 3A, a linear association was observed between LAPI and prevalent LUTS/BPH in the overall population (p for overall < 0.05). This linear association became more obvious in the interpolated dataset by random forest (p for overall = 0.005, Supplementary Fig. 1A). Given the age-specific prevalence of LUTS/BPH, we further verified the association in the two age groups (<60 years and ≥60 years). In men aged <60 years, a marginal significance was detected between LAPI and prevalent LUTS/BPH (p for overall = 0.062, Fig. 3B), which was significant in the interpolated dataset (p for overall = 0.018, Supplementary Fig. 1B). Notably, in males aged ≥60 years, an L-shaped association between LAPI and prevalent LUTS/BPH was observed (p for non-linear = 0.006, Fig. 3C), in line with the interpolated dataset (p for non-linear = 0.006, Supplementary Fig. 1C).

Considering the significant results in the subgroup analysis, we also investigated the dose-response relationship in patients with hypertension. In participants without hypertension, RCS regression identified no association between LAPI and prevalent LUTS/BPH (both p for overall and non-linear > 0.05, Supplementary Fig. 2A), in line with the subgroup analysis However, in participants with hypertension, a significant linear upward trend was observed between LAPI and the risk of prevalent LUTS/BPH (p for overall = 0.036, Supplementary Fig. 2B). After interpolating the dataset, these associations in participants without (Supplementary Fig. 3A) and with hypertension (p for overall = 0.021, Supplementary Fig. 3B) remained consistent.

Fig. 3.The dose-response association between LAPI and prevalent LUTS/BPH. Restricted cubic spline regression was used to evaluate the dose-response relationship between LAPI and prevalent LUTS/BPH. In the overall population (A), age, educational levels, marital status, sleep duration, afternoon nap, cigarette and alcohol consumption, hypertension, depression, uric acid, LDL, BUN, HDL and total cholesterol were adjusted. In males aged <60 years (B) and ≥60 years (C), age was not adjusted. The red line shows the odds ratio and the pink area shows the 95% confidence interval. OR: odds ratio; CI: Confidence Interval; LAPI: lipid accumulation product index.

3.5 Association between LAPI and prevalent LUTS/BPH excluding participants with obesity

To verify the association between LAPI and LUTS/BPH in the non-obese population, we conducted an additional sensitivity analysis. Participants who were obese (BMI ≥28.0 kg/m²) were excluded in the overall population. In the crude model, the ORs for prevalent LUTS/BPH were 1.06 (95% CI = 0.80–1.40, p = 0.678) and 1.29 (95% CI = 0.99–1.69, p = 0.063) for the T2 and T3 groups, respectively (Table 3). In the full model (model 5), the ORs for the T2 and T3 groups were 1.17 (95% CI = 0.86–1.59, p = 0.312) and 1.58 (95% CI = 1.09–2.27, p = 0.015), respectively. In the non-obese population, high LAPI was also positively associated with prevalent LUTS/BPH.

3.6 Longitudinal association between LAPI and incident LUTS/BPH

To investigate the longitudinal association between LAPI and incident LUTS/BPH, we constructed three cohorts (2011–2013, 2011–2015 and 2011–2018). In the 2011–2018 cohort, the T2 group had a 1.62-fold risk of incident LUTS/BPH (95% CI = 1.22–2.14, p < 0.001), and the T3 group had a 2.07-fold risk of incident LUTS/BPH (95% CI = 1.58–2.71, p < 0.001) in the crude model (Fig. 4). In the full model, the ORs for the T2 and T3 groups were 1.51 (95% CI = 1.11–2.04, p = 0.008) and 1.74 (95% CI = 1.21–2.50, p = 0.003), respectively. All five regression models supported the increased risk of incident LUTS/BPH for participants with high LAPI (all p for trend < 0.01), consistent with the results using LAPI as a continuous variable (all p < 0.05, Supplementary Table 4).

Fig. 4.The longitudinal association between LAPI and incident LUTS/BPH. In 2011, 3967 males were enrolled in the analysis. Of them, 2855 males were followed up to 2018. The T1 group was used as the reference group. Model 1—crude model; Model 2—adjusting for age, educational levels and marital status; Model 3—further adjusting for sleep duration, afternoon nap, cigarette and alcohol consumption; Model 4—further adjusting for hypertension and depression; Model 5—further adjusting for blood biomarkers including uric acid, LDL, BUN, HDL and total cholesterol. OR: odds ratio; CI: Confidence Interval.

For sensitivity analysis, we also investigated the longitudinal association in the 2011–2013 and 2011–2015 cohorts. In the 2011–2013 cohort, the longitudinal association was significant for the T2 group (OR = 1.46, 95% CI = 1.04–2.04, p < 0.05) and the T3 group (OR = 1.61, 95% CI = 1.15–2.26, p < 0.01) in the crude model (Supplementary Table 5). However, in the full model, this association did not reach the significance threshold (OR = 1.40 for the T2 group, p = 0.071; OR = 1.54 for the T3 group, p = 0.063). As a continuous variable (Supplementary Table 6), a significant longitudinal association was found between LAPI and incident LUTS/BPH in the full model (OR = 1.011, 95% CI = 1.003–1.018, p = 0.006). In the 2011–2015 cohort, all five regression models supported the association between tertiles of LAPI and LUTS/BPH (all p < 0.05, Supplementary Table 7). In the full model, the ORs were 1.39 (95% CI = 1.02–1.91, p < 0.05) for the T2 group and 1.57 (95% CI = 1.06–2.31, p < 0.05) for the T3 group. As a continuous variable (Supplementary Table 8), we identified a longitudinal association between LAPI and incident LUTS/BPH in the full model (OR = 1.012, 95% CI = 1.005–1.018, p = 0.001).