AUA2023 BEST POSTERS Benign Prostatic Hyperplasia/Lower Urinary Tract Symptoms Phenotypes Identified by Cluster Analysis

By: Matthew N. Simmons, MD, PhD, FACS, Medical College of Georgia, Augusta University; Nathaniel S. Taylor, MD, Medical College of Georgia, Augusta University; W. Carter Reed, MD, Medical College of Georgia, Augusta University; Pablo Santamaria, MD, Medical College of Georgia, Augusta University; Zachary Klaassen, MD, Medical College of Georgia, Augusta University; Sherita A. King, MD, Medical College of Georgia, Augusta University; Martha K. Terris, MD, Medical College of Georgia, Augusta University | Posted on: 30 Aug 2023

One of the more enigmatic aspects of benign prostatic hyperplasia/lower urinary tract symptoms (BPH/LUTS) is the highly variable nature of the disease. For example, a patient with a 40-mL prostate may have bothersome LUTS, while another with a 150-mL prostate has none. Data support that BPH/LUTS are impacted by several different pathological processes.1 Because of this heterogeneity, traditional statistical methods such as regression analysis fail to identify consistent correlates of disease severity. Regression analysis aims to find predictors of disease severity or response to treatment. It assumes that BPH patients have a uniform disease process. It also requires that the variables are independent of each other. In the case of BPH, the disease is not uniform. Also the variables are not independent (such as the relationship between age and prostate volume [PV]).

The study we conducted used hierarchical cluster analysis (HCA) to study BPH/LUTS patients. HCA is useful to identify patterns in complex data sets. In essence, HCA clusters individuals based on their similarity. In contrast to regression analysis, HCA makes no assumptions about relationships between variables. We hypothesized that it could be used to cluster BPH/LUTS patients based on similarity of clinical data.

Out of several thousand patients treated from 2017-2022 for BPH/LUTS in our clinics we identified 111 who underwent transurethral resection of the prostate (TURP) and who had pelvic CT imaging conducted within 12 months of surgery. CT scans were done to assess other issues (eg, hematuria, abdominal pain), and were not part of the BPH evaluation. All patients had International Prostate Symptom Scores >20 and either medically refractory disease or absolute indications for TURP, such as acute retention or obstructive uropathy. There was no precedent for the analysis, so we selected data points based on several criteria. First was that data points were clinically relevant and readily available in the electronic medical record. Second was that data were quantifiable and unambiguous. Lastly, we made an effort to include data that represented putative BPH/LUTS disease drivers—innate/genetic factors, systemic inflammation and localized/reactive inflammation. Innate/genetic factors were represented by age, PV, and presence of a median lobe. Systemic inflammation severity was represented by a metabolic syndrome score termed “MetX” (based on presence of obesity, diabetes, hypertension, and hyperlipidemia). We used presence and severity of intraprostatic calcifications (IPCs, as seen on CT) as a surrogate for localized/reactive inflammation.

The Table summarizes clinical data for the cohort. Mean age of the cohort was 67.6 years (SD = 8.5). Of the cohort 41% were men of African descent. Mean BMI was 27.8 kg/m2 (SD = 6.2). Mean PV was 80.5 mL (SD = 54). Hypertension, hyperlipidemia, and type 2 diabetes mellitus were present in 73%, 58%, and 32% of patients, respectively. Median lobes were present in 22% of men. IPCs were present in 41% and were graded as scant (n = 29) or extensive (n = 17).

Table. Clinical Features of Cohort and Cluster Analysis Groups

Cohort Group 1 Group 2 Group 3 ANOVA P value
No. 111 36 24 51
Mean age (SD), y 67.6 (8.5) 63.4 (6.5) 63.4 (6.5) 70.5 (9.2) .001
No. AA (%) 45 (41) 10 (28) 14 (58) 21 (41) .079
Mean BMI (SD), kg/m2 27.8 (6.2) 27.1 (4.4) 32.6 (7.9) 26 (5.3) < .001
Mean MetX (SD) 3.6 (1.5) 3.4 (1.4) 4.96 (1.3) 3.06 (1.4) < .001
Mean PV (SD) 80.5 (54) 53.1 (30.5) 84.1 (52.8) 98.1 (60.1) < .001
No. with MLs (%) 24 (22) 0 1 (4) 23 (45) < .001
No. with IPCs (%) 45 (41) 36 (100) 2 (8) 7 (14) < .001
Abbreviations: AA, African American; ANOVA, analysis of variance; BMI, body mass index; IPC, intraprostatic calcification; MeTX, metabolic syndrome score; ML, median lobe; PV, prostate volume; SD, standard deviation.

HCA revealed 3 distinct patterns of BPH/LUTS patients (see Figure). The common factor in group 1 patients was presence of IPCs. The patients had lower than average PV, lower age, and low incidence of metabolic syndrome–related diseases. Group 2 consisted of patients with metabolic syndrome. They had lower than average age and average PV. Group 3 consisted of patients with innate factors such as advanced age, presence of a median lobe, or isolated high PV.

Figure. Hierarchical cluster analysis heat map and dendrogram. IPC indicates intraprostatic calcification; ML, median lobe; MtX, metabolic syndrome score; PV, prostate volume.

The analysis was not meant to definitively characterize BPH/LUTS causality. Rather, it was meant to allow for identification of patterns of presentation of BPH/LUTS patients. It appeared successful in that regard, and based on these patterns we can make some speculations. Group 1 patients tend to be men with small fibrotic glands. Notably, 60% of men in group 1 had a prior history of acute or chronic prostatitis. The correlation between IPCs and a history of prior inflammation or infection is established but limited.2 Better markers are needed. Many of these men respond poorly to conventional medical therapy and favorably to TURP. Further study may support an increased role for procedural intervention in this group. Group 2 patients had high MetX scores. This is consistent with an emerging body of data supporting presence of a metabolic syndrome–specific pathological pathway to BPH/LUTS.3,4 It is unclear if metabolic syndrome affects prostate growth, affects neurological function of the bladder, or both. Response to medical therapy is variable in this group. Also, anecdotally, there seems to be an increased incidence of LUTS persistence after TURP in these patients. It may be possible that treatment of metabolic syndrome conditions is an integral part of management in this group. Group 3 patients tended to have a more straightforward cause for their LUTS, namely resistance to flow due to obstruction by either a voluminous gland or a median lobe. Further study of responses to medical and surgical treatment may allow for improved treatment.

The study demonstrated that HCA analysis provides a powerful tool for the study of patients with BPH/LUTS. Future HCA analyses will integrate molecular data, functional data, and symptom scores. Use of this approach may allow for refinement of diagnosis and targeted treatment of BPH/LUTS patients.

  1. Devlin CM, Simms MS, Maitland NJ. Benign prostatic hyperplasia—what do we know?. BJU Int. 2021;127(4):389-399.
  2. Wang H, Ma M, Qin F, Yuan J. The influence of prostatic calculi on lower urinary tract symptoms and sexual dysfunction: a narrative review. Transl Androl Urol. 2021;10(2):929-938.
  3. Liu D, Shoag JE, Poliak D, et al. Integrative multiplatform molecular profiling of benign prostatic hyperplasia identifies distinct subtypes. Nat Commun. 2020;11(1):1987.
  4. Xia B, Zhao S, Chen Z, et al. The underlying mechanism of metabolic syndrome on benign prostatic hyperplasia and prostate volume. Prostate. 2020;80(6):481-490.