INTRODUCTION
Benign prostatic hyperplasia (BPH) is one of the most prevalent conditions affecting older men, a population that already faces an increased risk of cognitive decline and dementia. Its prevalence rises considerably after age 40, affecting approxi-mately 8% to 60% of individuals by age 90 [1]. Because age is the primary risk factor for dementia, understanding the potential cognitive effects of BPH medications is essential, particularly for older patients who are already vulnerable to neurodegenerative disorders [2,3].
According to American Urological Association guidelines, BPH treatments mainly include alpha-blockers and 5-alpha reductase inhibitors (5ARIs), both of which may influence cognitive outcomes through several mechanisms [4]. Multiple biological pathways have been proposed to explain how these medications may contribute to dementia risk. Alpha-blockers may alter norepinephrine signaling, impairing attention, memory, and executive function and thereby promoting cognitive dysfunction [5]. 5ARIs inhibit dihydrotestosterone (DHT) synthesis, a process that may reduce hippocampal neurogenesis and facilitate β-amyloid accumulation, both central to Alzheimer disease pathology [6]. Additionally, both drug classes have been linked to vascular and autonomic dysregulation, which may result in hypotension, reduced cerebral perfusion, and a heightened risk of vascular dementia [7].
Real-world clinical studies have reported higher dementia incidence among BPH patients receiving these medications, with retrospective cohort analyses demonstrating potential dose- and duration-dependent relationships, particularly for 5ARIs [8,9]. Despite these observations, the literature remains fragmented, and no meta-analysis has comprehensively evaluated dementia risk associated with BPH medications. A systematic and integrative assessment is therefore needed to offer clinicians and policymakers clearer evidence regarding potential cognitive risks. Network meta-analysis (NMA), especially within a Bayesian framework, enables comparison across multiple interventions by integrating both direct and indirect evidence. In contrast to pairwise meta-analysis, which compares treatments only in head-to-head studies, NMA simultaneously evaluates all available treatments within a unified analytic structure. The additional use of SUCRA allows quantitative ranking of treatment-related risks. This study provides the first NMA to systematically examine whether alpha-blockers and 5ARIs increase dementia risk and aims to offer clinically meaningful interpretations. To our knowledge, this is the first NMA to apply these advanced methods in the context of BPH-related dementia, representing the principal novelty of this work and expanding the evidence base beyond prior reviews.
MATERIALS AND METHODS
Protocol Registration and Reporting Guidelines
This study followed a predefined protocol registered in the International Prospective Register of Systematic Reviews (PROSPERO; CRD42025643431) and adhered to the PRISMA-NMA (Preferred Reporting Items for Systematic Reviews and Metaanalyses for Network Meta-Analyses) guidelines [10].
Eligibility Criteria
Eligible primary studies were selected based on their PICO-TS (population, intervention, comparator, outcome, timing, and study design) framework. We included observational studies enrolling male patients aged 50 years or older who were diagnosed with BPH using prostate-specific antigen testing or imaging and were indicated to receive BPH medications, specifically alpha-blockers (alfuzosin, doxazosin, terazosin, and tamsulosin) or 5ARIs (dutasteride and finasteride). Comparator groups included other BPH medications or no treatment. Eligible studies were required to report dementia as a study outcome. We included observational studies. Studies were excluded if they involved patients under 50 years of age, used BPH treatment beyond the specified medications, or did not assess dementia risk. Nonoriginal research articles (e.g., reviews, editorials, case reports) and animal studies were excluded. The exclusion criteria also included inaccessible full texts and nonoriginal research articles (including reviews, editorials, and case reports). Studies were additionally excluded if they lacked sufficient data for effect size estimation, had overlapping populations, or did not provide appropriate comparator groups. We applied these stringent inclusion and exclusion criteria to ensure methodological rigor and clinical relevance. Only highquality observational studies that reported dementia outcomes in patients with BPH were included. This approach was necessary to minimize bias and maintain the validity of the NMA. No publication date restrictions were applied.
Data Sources and Search Strategy
A comprehensive search was conducted across 7 electronic databases (EBSCOHost, Google Scholar, ProQuest, PubMed, SAGE Journals, ScienceDirect, and Wiley Online Library) from their inception through February 2025. The search strategy used a combination of pertinent keywords and Medical Subject Headings terms, including “alpha-blocker,” “alfuzosin,” “doxazosin,” “terazosin,” “tamsulosin,” “5-alpha reductase inhibitors,” “dutasteride,” “finasteride,” “dementia,” “Alzheimer’s disease,” and “cognitive.” Boolean operators (AND/OR) were applied to refine the search results. The reference lists of relevant review articles were also systematically examined to identify additional eligible studies.
Study Selection and Data Extraction
EndNote X9 (Clarivate Analytics, USA) was used to import all retrieved citations and remove duplicates. Three authors (FXR, KJ, and RY) independently screened titles and abstracts, followed by full-text assessments to determine final eligibility. Data extraction was performed by 4 authors (FXR, KJ, RY, and PAS) using a standardized form to document essential study details, including author and publication year, study design, study period, dementia diagnostic criteria, participant demographics (mean age and race), pertinent comorbidities (hypertension and diabetes mellitus), and duration of drug use. Discrepancies were resolved through discussion or consultation with senior authors (PYS and AW).
Risk of Bias Assessment
Three reviewers (FXR, PAS, and NN) independently assessed methodological quality using the Risk of Bias in Non-randomized Studies of Interventions, Version 2 (ROBINS-I V2) tool. This tool evaluates 5 domains: the randomization process, deviations from the intended intervention, missing outcome data, outcome measurement, and selection of the reported result [11]. Each domain was categorized as having low risk, some concerns, or high risk of bias. Disagreements were resolved by consensus or with input from senior authors (PYS and AW).
Data Synthesis and Statistical Analysis
A meta-analysis was performed using Review Manager 5.4.1, provided by the Cochrane Collaboration Network. Results were evaluated using risk ratios (RRs) with corresponding 95% confidence intervals (CIs). An inverse variance random-effects model was used to assess heterogeneity across the included studies. A P-value less than 0.05 and an I² value greater than 50% indicated statistically significant heterogeneity. Statistical significance was defined as a P-value less than 0.05. A Bayesian NMA using random-effects models was conducted in MetaInsight software v6.3.0 for binary outcomes, utilizing the “gemtc” and “BUGSnet” packages [12-14]. RR with 95% credible intervals (CrI) was calculated to assess dementia risk. Markov Chain Monte Carlo (MCMC) simulations were performed with a minimum of 4 parallel chains, with each chain comprising 20,000 iterations and a burn-in period of 5,000 iterations. Default diffuse priors were applied. Convergence was evaluated using trace plots and Gelman-Rubin statistics, while consistency was assessed using node-splitting methods [15]. The surface under the cumulative ranking curve was used to rank dipping categories for each outcome. A network meta-regression would have been conducted if the number of included studies reached 10 or more [16].
RESULTS
Study Selection and Characteristics
A cumulative total of 769 records was obtained from 9 databases: EBSCOHost (n=232), Google Scholar (n=37), ProQuest (n=90), PubMed (n=317), SAGE Journal (n=17), ScienceDirect (n=37), and Wiley Online Library (n=39). After removing duplicates and screening the titles and abstracts, 34 studies were selected for full-text evaluation. Of these, 19 were excluded due to irrelevant study outcomes, and 10 were excluded because they contained insufficient data, resulting in 5 studies that met all inclusion criteria (Fig. 1). These included 4 retrospective cohort studies and 1 case-control study, encompassing 3,650,434 participants with a mean age of 71.1 years. Table 1 summarizes the characteristics of the included studies [17-21].
Risk of Bias Assessment
The risk of bias assessment is presented in Fig. 2. All included studies were evaluated using the ROBINS-I V2 tool and were predominantly categorized as having a low risk of bias. Most domains, including selection of participants, classification of interventions, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting, were consistently rated as low risk across all studies. This pattern indicates adequate methodological rigor and reliability in data collection and reporting. Two studies by Garcia-Argibay et al. [20] and Fung et al. [21] demonstrated a moderate risk of bias related to confounding factors. Nonetheless, a study with only a single moderate-risk domain related to confounding is still considered to have an overall low risk of bias.
Pairwise Meta-analysis for Dementia Risk Associated With Alpha-Blockers and 5ARIs
The meta-analysis included 18 pairwise comparisons across all treatment groups. Statistically significant results were observed for tamsulosin versus dutasteride (RR, 1.27; 95% CI, 1.07–1.51; P=0.005; I²=75%), tamsulosin versus finasteride (RR, 1.16; 95% CI, 1.08–1.26; P=0.0001; I²=68%), and terazosin vs. alfuzosin (RR, 1.17; 95% CI, 1.09–1.24; P<0.00001; I²=0%). All other comparisons did not show statistically significant results (Table 2).
Bayesian NMA for Dementia Risk Associated With Alpha-Blockers and 5ARIs
The network examining dementia risk consisted of 5 alphablocker drugs (alfuzosin, doxazosin, terazosin, and tamsulosin) and 2 5ARIs (dutasteride and finasteride) across 5 studies, generating 21 direct comparisons (Fig. 3).
Overall, none of the medications demonstrated statistically significant differences in dementia risk when compared with each other (Table 3). However, SUCRA values indicated that the no-treatment group had the lowest estimated dementia risk (89.11), followed by alfuzosin (59.82), terazosin (53.39), doxazosin (52.55), tamsulosin (46.07), dutasteride (25.24), and finasteride (23.81), as illustrated in the litmus rank-o-gram (Fig. 4). Because fewer than 10 studies were included, meta-regression could not be conducted.
Node-Splitting and Convergence Diagnostics
Node-splitting analyses showed no statistically significant inconsistency between direct and indirect estimates, as the 95% CIs for all comparisons crossed zero and P-values exceeded 0.05, indicating no evidence of inconsistency. Convergence was confirmed through trace plots and Gelman-Rubin diagnostics, with shrink factors approaching 1.00 across all parameters. These findings indicate that the MCMC simulations exhibited adequate mixing across chains, supporting the reliability of the Bayesian model estimates and the validity of the reported 95% CrI (Fig. 5).
DISCUSSION
This NMA demonstrated that neither the use of alpha-blockers nor 5ARIs significantly increased the risk of dementia compared with the no-treatment group, although the pairwise comparisons revealed significant differences between tamsulosin and dutasteride, tamsulosin and finasteride, and terazosin and alfuzosin. SUCRA values also suggested a higher probability of dementia risk with 5ARIs (dutasteride and finasteride). Nevertheless, at the network level, no significant differences in dementia risk were observed among the drugs. The pairwise estimates were constrained by the small number of available studies, whereas the Bayesian NMA, which integrates both direct and indirect evidence within a unified framework, provided more precise and comprehensive estimates across the network.
A recent systematic review highlighted the variability in outcomes regarding the association between alpha-blocker use and dementia risk [22]. For example, although 2 randomized con-trolled trials assessing tamsulosin reported no significant cognitive changes, observational studies suggested a potential increase in dementia risk [17,19,23,24]. Similarly, evidence for alfuzosin and terazosin has been inconsistent, with conflicting findings across different study designs [17-19]. Regarding 5ARIs, cohort studies have also produced variable outcomes. While no significant association was observed during the first and second years of 5ARI use, prolonged exposure was linked to a significant increase in dementia risk [9]. Despite these contradictory findings, no recent meta-analysis has comprehensively clarified this association. In our NMA, SUCRA rankings indicated that 5ARIs, including dutasteride and finasteride, have a higher probability of contributing to cognitive impairment and an increased risk of dementia compared with alphablockers and the no-treatment group. This potential mechanism is thought to involve the suppression of DHT, which plays an important role in hippocampal function. A recent in vivo study suggested that DHT contributes to improved synaptic plasticity in the hippocampus by upregulating synaptic plasticity markers, including CREB, PSD95, SYN, and drebrin [25]. This finding is further supported by an observational study demonstrating a significantly higher risk of cognitive decline in individuals with lower DHT levels [26].
Clinicians should carefully balance the risk of cognitive decline relative to the therapeutic benefits of alpha-blockers and 5ARIs in managing BPH, particularly in patients who may have a predisposition to dementia [7]. Given these concerns, treatment options with no established association with cognitive decline should be prioritized. Muscarinic receptor antagonists, a class of anticholinergics, have been linked to an increased risk of cognitive impairment, especially in older patients or those at risk for dementia [27]. Therefore, the use of these agents should be approached with caution and avoided when possible, in favor of safer alternatives such as phosphodiesterase type 5 inhibitors or surgical interventions [28]. A patient-centered approach incorporating shared decision-making can help balance symptom relief with cognitive safety. Even when combination therapy is considered clinically appropriate, clinicians should remain attentive to potential cognitive consequences. Regular cognitive assessments should be considered for patients receiving longterm therapy to facilitate the early detection of cognitive decline [29]. Integrating cognitive assessments into routine follow-up visits for BPH management may improve early intervention and overall patient outcomes [30]. This proactive approach ensures that treatment decisions prioritize not only urological health but also long-term cognitive well-being, particularly in vulnerable patient populations.
While our findings provide important insights into the potential cognitive effects of BPH pharmacotherapies, they should be interpreted in light of several limitations. The relatively small number of included studies reduces the statistical power of our analyses. Although Bayesian NMA can partially account for variability, the substantial heterogeneity observed in our results warrants careful interpretation. This heterogeneity likely reflects differences in study populations and outcome measures across included studies, which may increase uncertainty in the findings. Because the included studies were observational, methodological limitations such as the absence of randomization, the inability to establish causality, and restricted control over key variables could not be mitigated. In addition, evaluation of the roles and dosing schedules of the various classes of BPH medications was not feasible. These deficiencies in the data restrict generalizability to broader populations and limit the ability to draw definitive conclusions regarding long-term efficacy and safety. Future research should prioritize large-scale, longitudinal studies with standardized outcome definitions. Such studies are needed to clarify causal pathways, including those involving hormonal regulation and vascular effects, and to investigate high-risk subgroups to better understand underlying mechanisms. In addition, comparative research involving all classes of BPH medications with extended follow-up periods is needed to clarify the potential link between alpha-blockers or 5ARIs and dementia risk, ultimately guiding safer and more effective therapies within precision medicine.
This NMA found no significant overall association between BPH medications and an increased risk of dementia. However, pairwise comparisons and SUCRA rankings indicated a relatively higher risk of dementia associated with 5ARIs. Given the uncertain impact of 5ARIs on disease progression and mental health outcomes, further research is needed. Clinicians should carefully evaluate the long-term use of 5ARIs or surgical intervention in older BPH patients, particularly those with dementia. Patient counseling should include discussion of potential adverse effects, and management strategies should be individualized. Routine cognitive assessments are recommended to facilitate early detection and intervention for cognitive decline, thereby improving long-term outcomes.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
