Symptom Based Clustering of Men in the LURN Observational Cohort Study

Abstract

Purpose: Conventional classification of patients with lower urinary tract symptoms into diagnostic categories based on a predefined symptom complex or predominant symptom appears inadequate. This is due to the frequent presentation of patients with multiple urinary symptoms which could not be perfectly categorized into traditional diagnostic groups. We used a novel clustering method to identify subtypes of male patients with lower urinary tract symptoms based on detailed multisymptom information.

Materials and methods: We analyzed baseline data on 503 care seeking men in the LURN (Symptoms of Lower Urinary Tract Dysfunction Research Network) Observational Cohort Study. Symptoms and symptom severity were assessed using the LUTS (Lower Urinary Tract Symptoms) Tool and the AUA SI (American Urological Association Symptom Index), which include a total of 52 questions. We used a resampling based consensus clustering algorithm to identify patient subtypes with distinct symptom signatures.

Results: Four distinct symptom clusters were identified. The 166 patients in cluster M1 had predominant symptoms of frequency, nocturia, hesitancy, straining, weak stream, intermittency and incomplete bladder emptying suggestive of bladder outlet obstruction. The 93 patients in cluster M2 mainly endorsed post-micturition symptoms (eg post-void dribbling and post-void leakage) with some weak stream. The 114 patients in cluster M3 reported mostly urinary frequency without incontinence. The 130 patients in cluster M4 reported severe frequency, urgency and urgency incontinence. Most other urinary symptoms statistically differed between cluster pairs. Patient reported outcomes of bowel symptoms, mental health, sleep dysfunction, erectile function and urological pain significantly differed across the clusters.

Conclusions: We identified 4 data derived clusters among men seeking care for lower urinary tract symptoms. The clusters differed from traditional diagnostic categories. Further subtype refinement will be done to incorporate clinical data and nonurinary patient reported outcomes.

Keywords: cluster analysis; diagnosis; lower urinary tract symptoms; patient reported outcome measures; urinary bladder.

Figure 1.

Consensus matrix as a color map and dendrogram for hierarchical clustering. Each time we randomly selected 80% of the patients and performed k-means clustering. We repeated this step for 1000 times and calculated the probability that a pair of patients were clustered into the same cluster (i.e., the number of times this pair of patients being in the same cluster divided by the number of times this pair of patients both being selected). Probability is color-coded: bright yellow represents probability close to one, dark blue – probability close to zero. The index of patients is reordered so that the patients belonging to the same clusters are grouped together. The four bright-yellow blocks along the diagonal of the matrix as well as the dendrogram at the top of the figure both indicate four clusters.

Figure 2.

Upper row represents signatures based on the LUTS tool questions, where each “spoke” is an average of symptoms’ severity and bother within each cluster. Questions related to similar symptoms are grouped together and color-coded. LUTS tool questions are as follows: 1-frequency, 2-daytime frequency, 3-nocturia, 4-incomplete emptying, 5-trickle/dribble, 6-urgency, 7-hesitancy, 8-intermittency, 9-strain, 10-weak stream, 11-splitting/spraying, 12-urgency for fear of leaking urine, 13-pain, 14-burning, 15-leakage, 16a-leakage post voiding, 16b-leakage with urgency, 16c-leakage with laugh, 16d- leakage with exercise, 16e-leakage with sleeping, 16f-leakage with sex, 16g-leakage for no reason. Each radar plot in the lower row represents averaged response to the AUA-SI questions within the corresponding cluster. AUA questions are as follows: 1-nocturia, 2-incomplete emptying, 3-frequency, 4-intermittency, 5-urgency, 6-weak stream, 7-strain. Circles represent the highest possible severity level (8 for LUTS questions and 5 for AUA-SI questions).

Figure 3.. Mean symptom severity levels by cluster, pairwise symptom severity differences between clusters and statistical significance.

The bar charts along the diagonal present the average level of severity for each of the LUTS Tool questions for each cluster. Each bar chart in the lower triangle of the matrix shows the differences in the average severity levels of symptoms between corresponding clusters (i.e., the average severity levels of symptoms in column cluster subtracted from those of row cluster). Presence of each colored bar in the upper triangle indicates a statistically significant difference in the corresponding pairwise comparison. Absence of a colored bar indicates no significant difference could be detected in the corresponding pairwise comparison. For instance, the second rectangle in the first row demonstrates that 17 symptoms are significantly different in cluster M1 versus cluster M2: three related to frequency, one to post-micturition, two to urgency, five to voiding, two to pain, and four to incontinence.

Figure 4.. Non-urologic patient reported outcomes (PROs). Matrix of pairwise comparisons of the clusters based on adjusted regression models.

The bar plots along the diagonal illustrate the average score for each of the PROs. Similar to Figure 3, the bar charts in the lower triangle show the adjusted estimated differences in the average scores of the PROs between corresponding clusters. The colored bars in the upper triangle indicates that the corresponding pairs of clusters have significantly different scores on average after adjustment for relevant covariates. For instance, the second rectangle in the first row demonstrates that 9 scores of PROs are significantly different in cluster M1 versus cluster M2: four related to GUPI, one to bowel symptom, and four to psychological symptom. PROs are as follows: 1-GUPI Pain Subscale, 2-GUPI Urine Subscale, 3-GUPI QOL Subscale, 4-GUPI Total Score, 5-PROMIS Constipation, 6-PROMIS Diarrhea, 7-PROMIS Bowel Incontinence, 8-Perceived Stress Scale, 9-PROMIS Sleep Dysfunction, 10-PROMIS Depression, 11-PROMIS Anxiety, 12-IIEF.