Diagnosis and treatment of female lower urinary tract symptoms (LUTS) has been a challenge due to overlapping symptoms. Identifying subtypes of women with LUTS would potentially improve diagnostic and treatment decisions. Our hypothesis is that symptom-based subtypes, identified previously [1] by using the LUTS Tool [2] and refined in this study by incorporating demographics, physical exam, and non-urologic patient reported outcomes (PRO), will be biochemically distinct, i.e. will demonstrate different serum-based biomarker signatures of over- and under-abundant proteins. This is the first large-scale targeted proteomics study of serum in women with LUTS.

A cohort of 545 treatment-seeking women with LUTS was subtyped by using resampling-based consensus clustering of the clinical data, which included demographic information, medical history, physical exam findings, bladder diaries, and self-reported questionnaires of urologic and non-urologic symptoms. Baseline serum samples were analyzed using the targeted proteomics approach Proximity Extended Assay (PEA) by Olink Proteomics (Uppsala, Sweden) for a subset of these women (n=230, 42%), representing all identified clusters. Serum samples were also analyzed from 30 controls that were not necessarily healthy but reported no LUTS and were age, race, and body mass index (BMI) frequency-matched to the above cases. Three Olink panels (cardio metabolic, inflammation, neurology) were used to quantify the abundances of 276 proteins. These data were not used for clustering; however, they served as an additional orthogonal approach for evaluation of the distinctiveness of the identified clusters. We compared the abundances of each serum protein in women with LUTS and in controls and tested for significant differences in each of the identified clusters versus controls. We used ANOVA, Wilcoxon rank sum tests, and chi square tests, where appropriate, to determine which of the variables used for clustering were significantly different in the pairwise comparison. ANOVA was used to determine which of the proteins were significantly differentially abundant in identified clusters relative to controls. All comparisons were adjusted using an FDR correction for multitesting. Pathway enrichment analysis was performed using MetaCore (Clarivate Analytics) software.

Five distinct clusters of women with LUTS were identified using 185 variables. Properties of the five clusters named W1-W5 are shown (Figure 1). Each column represents one of five clusters. Radar plots in the first row illustrate urinary symptoms measured by the LUTS Tool, the second row illustrates demographics, clinical measurements and non-urinary PROs, the third row presents categorical data on comorbidities and anomalies identified during the physical exam, and the fourth row shows intake and voiding pattern variables collected in bladder diaries. Radar plots represent the mean values of the variables across the members of each cluster. None of the clusters could be characterized by a single symptom, but rather by a combination of symptoms with various levels of severity. Women in all five clusters reported higher than normal frequency of voiding (with the highest frequency in W3 and W5). Women in all clusters except W1 reported urinary urgency and some level of incontinence. Women in cluster W1 (n=77) did not report incontinence, but had post-void dribbling, trickling, straining, and symptoms of incomplete bladder emptying. They were younger than average across the cohort, had lower than average weight, number of pregnancies, and vaginal births, and fewer comorbidities and abnormal physical exam findings. Women in cluster W2 (n=64) reported mild urinary symptoms including mild urinary incontinence. They were characterized by the presence of clinically significant pelvic organ prolapse (high values of POP-Q point Ba). They were on average older (66 vs. 53 years old), with more pregnancies (2.9 vs 1.8) and vaginal births (1.47 vs 0.92) than women in cluster W1. They also had the highest post void residual (75 mL) across the clusters. Women in cluster W3 (n=144) reported high urinary frequency, urgency, and urgency urinary incontinence. They had a greater weight, larger waist circumference, and higher functional comorbidity index (FCI) than women in W1, W2, and W4. They reported the most “urgency with fear of leakage” across the clusters, but did not report any substantial post voiding symptoms. Women in cluster W4 (n=95) reported stress incontinence, as well as urgency urinary incontinence, and some post void urinary incontinence. They were younger (mean 51 yo), healthier (FCI =1.24), and more physically active (PROMIS Physical Functioning T-score=53.2) than others in the cohort. Women in W5 (n=165) reported all LUTS at uniformly high levels. For 27 out of 30 urinary symptoms, they reported the highest levels across all five clusters. They demonstrated the greatest mean weight (87.6 Kg), were the least physically active (PROMIS Physical Functioning T-score=42.2), had more comorbidities (FCI=3.64) and pregnancies (3.07) than the rest of the cohort. Pairwise comparison of the clusters demonstrated that multiple variables were significantly different (52% on average across pairs of clusters; range from 27% for W1 vs W2 to 83% for W2 vs W5).
Figure 2 presents the volcano plots comparing abundances of 276 proteins in serum samples of women with LUTS versus controls. Fig 2a compares the abundances for all 230 women with LUTS to 30 controls, while Figs 2b-2f provide similar comparisons for the members of the identified clusters W1-W5 for whom proteomics assays were performed (n1=37, n2=38, n3=53, n4=42, n5=60).  Multiple differentially abundant proteins are observed in the serum samples of women with LUTS versus controls both overall and between each cluster and controls. While some of these proteins have been shown [3] to be associated with LUTS, e.g. TNF, IL-10, MCP, and TGF, the remainder are novel.

The highest number of the differentially abundant proteins of 70 (29 after FDR correction) is observed for cluster W5, which demonstrated the highest level of all urinary symptoms. The overlap between the lists of differentially abundant proteins is quite low meaning that clusters W1-W5 are “biochemically” different. The highest overlap of 18 differentially abundant proteins is observed for cluster W5 where all urinary symptoms are present at high level and cluster W3 defined mainly by high urinary frequency, urgency, and urgency urinary incontinence. Interestingly, the lowest number of differentially abundant proteins (n=10) are observed in clusters W2 (characterized by the presence of prolapse) and W4 (characterized by the presence of stress urinary incontinence), which are presumably driven by mechanistic or anatomic rather than biochemical factors. Mapping of the differentially abundant proteins on the known biochemical pathways and performing enrichment analysis with MetaCore demonstrated that affected pathways are different across the clusters as well, identifying that the clusters are different “biochemically” and have different etiologies.

Subtypes of women with LUTS are refined into five distinct, clinically recognizable clusters by incorporating demographics, physical exam, urologic and non-urologic PROs. The first large scale proteomics analysis of serum samples from women with LUTS identified multiple differentially abundant proteins relative to controls. Furthermore, differentially abundant proteins and affected biochemical pathways are different across five identified subtypes of women with LUTS. Since proteomics data were not used for clustering, this result independently corroborates the distinctiveness of the identified subtypes. Further investigation of the etiologies and clinical significance of the identified subtypes is necessary.

Andreev VP, Liu G, Yang CC, Smith AR, Helmuth ME, Wiseman JB, Merion RM, Weinfurt KP, Cameron AP, Lai HH, Cella D, Gillespie BW, Helfand BT, Griffith JW, DeLancey JOL, Fraser MO, Clemens JQ, Kirkali Z, and the LURN Study Group. Symptom-based Clustering of Women in the Symptoms of Lower Urinary Tract Dysfunction Research Network (LURN) Observational Cohort Study. J Urol 2018; 200(6) 1323-1331.
Coyne KS, Barsdorf AI, Thompson C, et al. Moving towards a comprehensive assessment of lower urinary tract symptoms (LUTS). Neurourol. Urodyn. 31:448–454, 2012.
NY Siddiqui, BT Helfand, VP Andreev, JT Kowalski, MS Bradley, HH Lai, MB Berger, MG Mueller, JA Bickhaus, VT Packiam, D Fenner, BW Gillispie, Z Kirkali. Biomarkers implicated in lower urinary tract symptoms: systematic review and pathway analyses. J Urol 2019, 202 (5) 880-889.