Examining lower urinary tract symptom constellations using cluster analysis

Authors


Karin Coyne, Senior Research Scientist, Center for Health Outcomes Research, United BioSource Corporation, 7101 Wisconsin Ave Suite 600, Bethesda, MD 20814, USA.
e-mail: karin.coyne@unitedbiosource.com

Abstract

OBJECTIVE

To gain a better understanding of how patients experience lower urinary tract symptoms (LUTS) and to determine whether particular symptoms cluster together, as LUTS seldom occur alone.

SUBJECTS AND METHODS

A secondary analysis of a cross-sectional, population-based survey of adults in Sweden, Italy, Germany, UK and Canada was undertaken to examine the presence of LUTS groups. Of the 19 165 telephone surveys, 13 519 respondents reported at least one LUTS and were included in the analysis. All respondents were asked about the presence of 14 LUTS (International Prostate Symptom Score plus seven additional LUTS). K-means cluster analyses, a statistical method for sorting objects into groups so that similar objects are grouped together, was used to identify groups of people based on their symptoms. Men and women were analysed separately. A split-half random sample was selected from the dataset so that exploratory analyses could be conducted in one half and confirmed in the second. On model confirmation, the sample was analysed in its entirety.

RESULTS

Included in this analysis were 5014 men (mean age 49.8 years; 95% white) and 8505 women (mean age 50.4 years; 96% white). Among both men and women, six distinct symptom cluster groups were identified and the symptom patterns of each cluster were examined. For both, the largest cluster consisted of respondents with minimal symptoms (i.e. reporting essentially one symptom), 56% of men and 57% of women. The remaining five clusters for men and women were labelled based on their predominant symptoms. For men, the clusters were nocturia of twice or more per night (12%); terminal dribble (11%); urgency (10%); multiple symptoms (9%); and postvoid incontinence (5%). For women, the clusters were nocturia of twice or more per night (12%); terminal dribble (10%); urgency (8%); stress incontinence (8%); and multiple symptoms (5%). The multiple-symptom groups had several and varied LUTS, were older, and had more comorbidities. Clusters of terminal dribble and male postvoid incontinence had a lower prevalence of all other LUTS, but were fairly common (11% and 5% of men).

CONCLUSIONS

This analysis provides an empirical approach to examining the presentation of multiple LUTS and suggests it is possible to identify subgroups of patients with LUTS based on their symptom presentation. These analyses need to be replicated to evaluate the clinical relevance of these findings.

Abbreviation
(U)(S)UI

(urgency) (stress) urinary incontinence.

INTRODUCTION

Symptoms are subjective indicators of a condition or a change in a condition as perceived by an individual. Symptoms alone cannot be used to make a definitive diagnosis [1], but symptoms can guide diagnosis and treatment by suggesting appropriate diagnostic algorithms [2], which can be complicated when multiple or mixed symptoms are presented. Given the multitude of LUTS, there are endless possibilities of symptom combinations. Patients with LUTS might present with a combination of storage, voiding, or postmicturition symptoms, indicating possible multiple pathophysiologies, or with just one specific symptom. Regardless of pathophysiology, LUTS are bothersome symptoms [3–7] that are associated with a negative effect on daily living [8–11]. Given the high prevalence of LUTS noted by Irwin et al.[12], with 64.3% of respondents to a large European epidemiological study reporting at least one LUTS, with the increasing prevalence of LUTS with ageing [8,13], bothersomeness of LUTS, and patient experience of multiple LUTS, there is a need for a greater understanding of LUTS presentation and treatment algorithms.

Knowledge about constellations of LUTS might or might not alter the treatment offered to the patient, but it could assist the healthcare provider and patient in identifying outcomes and goals of treatment. Also, the use of cluster analysis might increase our understanding of the potential common causes of various LUTS that might group together. Targeting specific LUTS that might be more problematic, and focusing treatment on specific symptom targets, might provide the patient with reasonable treatment expectations and ultimately improved outcomes [14,15]. Thus, the objective of this research was to obtain a better understanding of how patients experience LUTS as symptom constellations, and to identify common combinations of LUTS using cluster analysis.

Cluster analysis encompasses a wide range of empirical methods used across scientific disciplines to identify clusters or groups of objects (e.g. symptoms, patients, families) [16–18]. The basic principle of cluster analysis is that objects or individuals are grouped together based on specified variables so that members of each cluster are as similar as possible to others within their cluster, but as different as possible to those in adjacent clusters [16,17,19]. Cluster analysis is considered an exploratory process in which clusters are ‘discovered’ based on patterns within the data, without the influence of a priori hypotheses.

In healthcare research, cluster analysis is increasingly being used across a range of medical and psychiatric conditions to identify groups of patients with similar symptom or health status profiles [20–25]. By forming relatively homogeneous groups of respondents, cluster analysis can identify subgroups of patients that might benefit from targeted interventions [16,19].

SUBJECTS AND METHODS

A secondary analysis of data from the EPIC study, a cross-sectional, population-based survey of people aged ≥18 years in five countries (Canada, Germany, Italy, Sweden and the UK) was conducted. EPIC, described previously [12], used computer-assisted telephone interviews with a geographically stratified random sample of the population in each country. To provide representative population estimates, the sample was weighted by age, gender, household size and country size, but unweighted data were used for the present analysis.

Respondents were asked about the presence of specific LUTS using the seven-item IPSS and seven additional questions on the following symptoms: incomplete emptying, urinary frequency, intermittency, urinary urgency, weak stream, straining, nocturia, perceived frequency, urinary incontinence (UI, general), stress UI (SUI), urgency UI (UUI), terminal dribble, and postvoid UI. The ICS standard terminology for voiding, storage and postmicturition symptoms was followed. Specific questions are given in the Appendix.

In all, 19 165 telephone surveys were conducted, with 572 respondents excluded from the current analysis because they were either pregnant or had a UTI. Of the remaining 18 593 respondents, 5074 were excluded because they reported no LUTS. The remaining 13 519 respondents included in the analysis reported at least one LUTS. A split-half randomization of the 13 519 respondents was done so that half of the sample could be used for exploratory analyses while the second half could be used to confirm the exploratory findings. The split-halves were compared for all variables of interest (age, gender, country and LUTS) to assure an even distribution. Following model confirmation with the confirmatory half of the sample, the total sample of 13 519 respondents was re-analysed in its entirety.

Men and women were analysed separately. To begin the clustering method, all LUTS were standardized from 0 to 1 using the range method in PROC STDIZE (SAS v 9.1.3, Cary, NC, USA). The standardization process was necessary to accommodate the dichotomous, categorical and continuous question responses. Initial exploration as to the number of clusters was conducted using PROC CLUSTER, a hierarchical agglomerative approach with each subject beginning as his or her own cluster. Ward’s linkage, an anova approach that attempts to evaluate the distances between clusters by minimizing the sum of squares of any two clusters that can be formed at each step, was used to combine the initial clusters into larger clusters. Importantly, all cluster solutions are mutually exclusive, with no respondent being included in more than one cluster.

As the number of resulting clusters was not known, a series of exploratory cluster analysis models was run using the first half of respondents with the specification of three to seven clusters. Each cluster model was evaluated based on the clinical relevance and distinctiveness of each cluster, as well as the amount of variance accounted for by the cluster solution. As such, a six-cluster solution was considered to be optimal for both men and women.

The six-cluster solution from the exploratory cluster analysis was then tested in a confirmatory cluster analysis run with the second half of the respondents. The means of the standardized LUTS items from the exploratory analysis were used as seeds, or starting points, for the k-means cluster analysis procedure, PROC FASTCLUS. Upon successful replication of the six-cluster solution in the confirmatory half of the respondents, PROC FASTCLUS was executed on the entire sample, again using seeds from the exploratory half sample.

Descriptive statistics were used to present the demographic, comorbid and symptom characteristics of each cluster group. For ease of data presentation only, all Likert response options for LUTS items were re-coded dichotomously as having the symptom if the respondent reported that the symptom occurred more than half of the time or, in the case of nocturia, two or more nocturia episodes per night.

RESULTS

The 13 519 respondents who were included in the cluster analyses were white (95.6%), predominately female (62.9%) and currently married (57.4%), with a mean age of 50.2 years (Table 1). About a quarter (25.4%) of this group was from Italy, while somewhat fewer respondents were from Canada (23.9%), Germany (20.6%), the UK (19.1%) and Sweden (11.0%). The 5074 respondents who were excluded from the analyses because they reported no LUTS were significantly younger than the respondents with LUTS (44.0 vs 50.2 years; P < 0.001). There were also statistically significant differences between the included and excluded respondents in gender, race, marital status, and country, but the between-group differences appeared to be minimal.

Table 1. 
The demographics of the study population; the data presented are unweighted
VariableNo symptomsSymptoms
No. of respondents507413519
Mean (sd) age, years  44.0 (15.2)   50.2 (16.7)
n (%):
Gender, men2074 (40.9) 5014 (37.1)
Race
 White4815 (94.9)12926 (95.6)
 Black  49 (1.0)   131 (1.0)
 Asian 120 (2.4)  226 (1.7)
 Hispanic  13 (0.3)   28 (0.2)
 Other  56 (1.1)  172 (1.3)
 Missing  21 (0.4)   36 (0.7)
Education
 Primary school 832 (16.4) 2746 (20.3)
 Higher secondary school2512 (49.5) 6409 (47.4)
 University1319 (26.0) 3296 (24.4)
 Advanced university degree 229 (4.5)  526 (3.9)
 Other 140 (2.8)  456 (3.4)
 Missing  42 (0.8)   86 (0.6)
Marital status
 Single1706 (33.6) 3796 (28.1)
 Married2756 (54.3) 7755 (57.4)
 Living with partner 353 (7.0)  808 (6.0)
 Widow/widower 224 (4.4) 1098 (8.1)
 Missing  35 (0.7)   62 (0.5)
Country
 Canada 1152 (22.7) 3233 (23.9)
 Germany1372 (27.0) 2779 (20.6)
 Italy 1213 (23.9) 3436 (25.4)
 Sweden 397 (7.8) 1489 (11.0)
 UK 940 (18.5) 2582 (19.1)
Comorbid conditions
 Asthma 290 (5.7) 1244 (9.2)
 Chronic constipation  67 (1.3)  449 (3.3)
 Diabetes 136 (2.7)  849 (6.3)
 Depression 203 (4.0) 1186 (8.8)
 Hypertension 539 (10.6) 2972 (22.0)
 Neurological conditions  37 (0.7)  244 (1.8)
 Prostate/bladder cancer  13 (0.3)   141 (1.1)

Of the 7088 men responding, 2074 (29.3%) reported no LUTS. In the remaining 5014 men, six distinct clusters were identified (Table 2) and labelled based on their dominant symptom profile. Among the respondents with at least one LUTS, the largest identified cluster (minimal symptoms) which represented 52.9% of men (2654) reported a low prevalence of all LUTS with no predominant symptom (Table 2).

Table 2.  Frequency of LUTS in the clusters in men
VariableMinimal symptomsPostvoid UIUrgency (ICS)Terminal dribbleNocturia ≥2Multiple symptoms
N (%)2654 (52.9)255 (5.1)500 (10.0)558 (11.1)581 (11.6)466 (9.3)
Mean (sd) age  48.1 (16.7) 48.4 (15.6) 53.4 (17.7) 48.4 (17.0) 53.4 (16.4) 53.6 (17.3)
Age >50, n (%) 1150 (43.6) 113 (44.8)285 (57.6)259 (46.4)324 (55.8)270 (57.9)
Mean (sd) frequency   6.0 (3.1)  6.1 (2.7)  6.2 (2.7)  5.7 (2.4)  6.3 (3.1)  8.7 (5.2)
Mean (sd) nocturia   0.8 (0.7)  1.5 (1.0)  1.4 (1.5)  1.0 (1.3)  4.2 (0.3)  1.9 (1.6)
LUTS frequency, n (%)
 Not empty ( yes)  59 (2.2) 26 (10.2) 54 (10.8) 24 (4.3) 30 (5.2)125 (26.8)
 Frequency (IPSS) 127 (4.8) 42 (16.5) 90 (18.0) 41 (7.3) 66 (11.4)199 (42.7)
 Intermittency  51 (1.9) 27 (10.6) 24 (4.8) 18 (3.2) 24 (4.1) 69 (14.8)
 Urgency (IPSS)  32 (1.2) 12 (4.7) 62 (12.4)  7 (1.3) 23 (4.0)108 (23.2)
 Weak stream  70 (2.6) 22 (8.6) 55 (11.0) 33 (5.9) 18 (3.1) 86 (18.5)
 Strain  31 (1.2) 10 (3.9) 20 (4.0) 18 (3.2) 18 (3.1) 43 (9.2)
 Nocturia ≥ 2 382 (14.4) 77 (30.2)168 (33.6)124 (22.2)575 (99.0)225 (48.3)
 Perceived frequency   0 14 (5.5)  0  1 (0.2)  0461 (98.9)
 Urgency (ICS)   0  0499 (99.8)  0  0190 (40.8)
 General UI (yes)  88 (3.3) 67 (26.3)105 (21.0) 24 (4.3) 18 (3.1) 94 (20.2)
 UUI (yes)   11 (0.4) 14 (5.5) 53 (10.6)  6 (1.1)  3 (0.5) 43 (9.2)
 SUI  18 (0.7)  9 (3.5) 25 (5.0)  4 (0.7)  2 (0.3) 23 (4.9)
 Terminal dribble   0126 (49.4)167 (33.4)542 (97.1)  0134 (28.8)
 Postvoid UI   0255 (100) 83 (16.6)  0  0 74 (15.9)
 Sweden 306 (11.5) 46 (7.6) 80 (16.0) 63 (11.3) 26 (4.5) 88 (18.9)
 Italy 480 (18.1) 42 (16.5) 119 (23.8)272 (48.8) 119 (20.5) 58 (12.5)
 Canada 661 (24.9) 69 (27.1) 98 (19.6) 76 (13.6)234 (40.3)128 (27.5)
 Germany 636 (24.0) 33 (12.9)155 (23.0) 92 (16.5) 63 (10.8) 88 (18.9)
 UK 571 (21.5) 65 (25.5) 88 (17.6) 55 (9.9)139 (23.9)104 (22.3)
Comorbid conditions (yes)
 Asthma 187 (7.1) 23 (9.0) 52 (10.4) 47 (8.4) 46 (7.9) 65 (14.0)
 Bladder/prostate cancer  33 (1.3)  4 (1.6) 23 (4.6) 13 (2.4) 23 (2.3) 20 (4.4)
 Chronic constipation  39 (1.5)  7 (2.8) 18 (3.6) 16 (2.2) 13 (2.2) 14 (3.0)
 Depression 144 (5.4) 25 (9.8) 45 (9.0) 35 (6.3) 32 (5.5) 65 (14.0)
 Diabetes 155 (5.9) 20 (7.8) 38 (7.6) 34 (6.1) 38 (6.5) 64 (13.8)
 Hypertension 520 (19.7) 60 (23.6)135 (27.1) 117 (21.0)124 (21.4)159 (34.1)
 Neurological  37 (1.4)  4 (1.6) 16 (3.2)  7 (1.3)  8 (1.4) 22 (4.7)

The second largest cluster for men was the ‘nocturia twice or more’ cluster (581, 11.6%). Respondents in this cluster reported a mean of 4.2 nocturia episodes per night, which was more than double the rate of nocturia in any of the other clusters. Respondents in the nocturia twice or more cluster had a relatively low prevalence of other LUTS. The third cluster (558, 11.1%) was characterized by terminal dribble, but with a minimal report of other LUTS, notably postvoid UI. The fourth cluster, labelled as the urgency cluster (500, 10.0%), was characterized by the presence of urinary urgency more than any other cluster. Nocturia and terminal dribble were also somewhat commonly reported among respondents in the urgency cluster, with each of these two symptoms reported by about a third of respondents in the cluster. The fifth cluster, termed the ‘multiple symptoms’ cluster (466, 9.3%), was the most symptomatic group, reporting greater urinary frequency than any other group (8.7 micturitions/day), in addition to relatively high rates of most other LUTS (i.e. nocturia twice or more 48.3%; urgency 40.8%; terminal dribble 28.8%; incomplete emptying 26.8%; general UI 20.2%; postvoid UI 15.9%; weak stream 18.5%; straining 9.2%; and UUI 9.2%). The sixth and smallest cluster among men was the postvoid UI cluster (255, 5.1%), in which all respondents reported the predominant symptom of postvoid UI, and a high prevalence of general UI (26.3%) and terminal dribble (49.4%).

There were 11 505 women respondents, of whom 3000 (26.1%) reported no LUTS. Among the remaining 8505 women, six clusters were identified (Table 2) and labelled according to the dominant symptom. As with men, the largest identified cluster for women was the minimal symptom cluster (4757, 55.9%) which had a low prevalence of all LUTS (Table 3). Nocturia of twice or more was the second largest cluster among women (1057, 12.4%), with the predominant symptom of nocturia (mean nocturia episodes per night, 4.3) and a relatively low prevalence of all other LUTS. Following the same pattern as men, terminal dribble was the third largest cluster for women (883, 10.4%). A fourth cluster that was unique to women was characterized by SUI (692, 8.1%) with few other LUTS (other than nocturia, at 30.9%). Urinary urgency was the dominant symptom of a fifth cluster (651, 7.7%) but respondents in this cluster also reported a moderate prevalence of nocturia (33.2%) and UUI (7.0%). The sixth and smallest cluster among women (465, 5.5%) was characterized predominantly by UI (95.1%), but respondents in this cluster also reported a multitude of other LUTS, including urinary urgency (87.3%), UUI (67.5%), SUI (76.6%), terminal dribble (42.8%), and postvoid UI (24.9%), incomplete emptying (31.0%), intermittency (18.5%) and weak stream (17.8%).

Table 3.  Frequency of LUTS in the clusters in women
VariableMinimal symptomsSUIUrgency (ICS)Terminal dribbleNocturia ≥2Multiple symptoms
N (%)4757 (55.9)692 (8.1)651 (7.7)883 (10.4)1057 (12.4)465 (5.5)
Mean (sd) age, years  48.8 (16.4) 53.3 (14.3) 51.9 (17.0) 47.8 (16.6)  53.4 (16.9) 58.0 (15.4)
Age >50 years, n (%)2099 (44.1)387 (55.9)325 (49.5)375 (42.5) 585 (55.3)320 (68.8)
Mean (sd) frequency   6.6 (3.0)  6.5 (2.8)  7.0 (3.7)  6.5 (3.8)   6.7 (2.5)  8.6 (4.1)
Mean (sd) nocturia   0.9 (0.7)  1.3 (1.5)  1.4 (1.5)  1.1 (0.4)   4.3 (0.8)  2.4 (1.8)
LUTS frequency, n (%)
 Not empty (yes) 132 (2.8) 45 (6.5) 55 (8.4) 55 (6.2)  46 (4.4)144 (31.0)
 Frequency (IPSS) 356 (7.5) 110 (15.9) 117 (17.8)103 (11.7) 142 (13.4)215 (46.2)
 Intermittency  90 (1.9) 37 (5.3) 23 (3.5) 35 (4.0)  31 (2.9) 86 (18.5)
 Urgency (IPSS)  91 (1.9) 47 (6.8) 87 (13.2) 29 (3.3)  24 (2.3)207 (44.5)
 Weak stream  55 (1.2) 21 (3.0) 27 (4.1) 31 (3.5)  19 (1.8) 83 (17.8)
 Strain  45 (0.9)  8 (1.2)  9 (1.4) 17 (1.9)  16 (1.5) 40 (8.6)
 Nocturia ≥ 2 756 (15.9)214 (30.9)218 (33.2)224 (25.4)1046 (99.0)278 (59.8)
 Perceived frequency 253 (5.3) 43 (6.2) 110 (16.7) 93 (10.5)  56 (5.3)196 (42.2)
 Urgency (ICS)   0 71 (10.3)647 (99.4)144 (16.3)   0406 (87.3)
 General UI 206 (4.3)674 (97.4) 90 (13.7) 57 (6.5)  47 (4.4)442 (95.1)
 UUI (yes)  30 (0.6) 48 (6.9) 46 (7.0) 17 (1.9)   7 (0.7)314 (67.5)
 SUI   0690 (99.7)  0  0   0356 (76.6)
 Terminal dribble   0 79 (11.4)  0868 (98.3)   0199 (42.8)
 Postvoid UI  72 (1.5) 34 (4.9) 25 (3.8) 67 (7.6)   9 (0.9) 116 (24.9)
 Sweden 450 (9.5)146 (21.1) 91 (14.0) 66 (7.5)  38 (3.6) 89 (19.1)
 Italy 1164 (24.5)149 (21.5)149 (22.9)516 (58.4) 281 (26.6) 87 (18.7)
 Canada 1102 (23.2) 131 (18.9) 118 (18.1) 84 (9.5) 419 (39.6) 113 (24.3)
 Germany 1166 (24.5)108 (15.6)189 (29.0)111 (12.6)  101 (10.0) 77 (16.6)
 UK 875 (18.4)158 (22.8)104 (16.0)106 (12.0) 218 (20.6) 99 (21.3)
Comorbid conditions (yes)
 Asthma 419 (8.8) 93 (13.4) 66 (10.2) 79 (9.0)  94 (8.9) 73 (15.7)
 Bladder cancer  16 (0.3)  4 (0.6)  2 (0.3)  3 (0.3)   3 (0.3)  7 (1.5)
 Chronic constipation  131 (2.8) 37 (5.4) 41 (6.3) 48 (5.5)  37 (3.5) 48 (10.4)
 Depression 374 (7.9)106 (15.3) 80 (12.3)104 (11.8)  68 (6.4)108 (23.3)
 Diabetes 228 (4.8) 44 (6.4) 50 (7.7) 51 (5.8)  84 (8.0) 43 (9.3)
 Hypertension 921 (19.4) 171 (24.7)169 (26.0)184 (20.9) 237 (22.4)175 (37.7)
 Neurological  69 (1.5) 12 (1.7) 23 (3.5) 16 (1.8)   4 (0.4) 26 (5.6)

The star and radar plots in Fig. 1a,b show the symptom variability within each cluster, to highlight the unique symptom presentations within each and the dominant symptom presentation.

Figure 1.

Cluster analysis: radar plots of a, men and b, women. Star and radar plot spokes represent the frequency of each LUTS variable on a standardized scale of 0–1.

For both men and women, the minimal symptom clusters consisted of the youngest respondents while the most symptomatic clusters consisted of the oldest respondents. Similarly, the most symptomatic clusters had the greatest prevalence of comorbid conditions; however, the minimal symptom clusters did not always have the lowest prevalence of comorbid conditions among the remaining clusters. Country appeared to play some role in the clustering of respondents, as Italy consisted of 48.8% of men and 58.4% of women in the terminal dribble cluster, but when Italy was removed from the sample, the same six-cluster solutions were identified.

DISCUSSION

Cluster analysis provides an empirical approach to examine the presentation of multiple LUTS, and in the present study is an expansion of a cluster analysis conducted by Norman et al.[24] in both the number of LUTS collected and sample size, and to gain insights on the potential underlying and overlapping pathophysiology. In the present analysis, six unique symptom constellations or clusters appeared for both men and women. As expected among this community sample, the largest cluster for both men and women reported minimal symptoms. Respondents in this cluster had few and varied LUTS, reflective of the general population and a group of people who are unlikely to seek care. Most of the remaining clusters were characterized by predominant symptoms, i.e. postvoid UI, urinary urgency, terminal dribble, and nocturia twice or more for men, and SUI, urinary urgency, terminal dribble and nocturia twice or more for women. The predominant single symptom cluster groups are interesting in the simplicity of diagnosis and treatment options. Predominant SUI among women with no additional LUTS is relatively easily treated by the clinician (depending upon the severity of the SUI) whereas the clusters that are defined primarily by terminal dribble and postvoid UI might not be as easily addressed in terms of successful treatment options. Among men, terminal dribble is frequently thought to be suggestive of prostatic enlargement and benign prostatic obstruction, although other symptoms that often appear with prostatic enlargement and benign prostatic obstruction, such as intermittency or urgency, were not particularly common in the male terminal dribble cluster [3]. Furthermore, little is known about terminal dribble in women, despite this having emerged as a somewhat prevalent cluster among women in the current analysis. Given that a cluster was defined by this symptom in both men and women, more research is needed to better understand the role of terminal dribble in diagnosis and treatment of LUTS.

The ‘nocturia twice or more’ cluster reflects an interesting group of respondents, in that while the respondents in this cluster reported low rates of all other LUTS, they reported an average of over four nocturia episodes per night. Given that this symptom is quite bothersome, with a substantial impact on health-related quality of life, [4,26], respondents within this group might be likely to seek treatment. Given the lack of additional LUTS in this cluster, the reason for nocturia might be not be related to prostate or bladder physiology, but to other clinical or lifestyle factors, such as nocturnal polyuria, sleep apnoea, or drinking too much fluid at night.

Urinary urgency, defined by the ICS as a sudden compelling desire to pass urine which is difficult to defer [1], was the predominant symptom in the second most severe LUTS cluster for both men and women. Some respondents in the urgency cluster also reported other symptoms of overactive bladder, e.g. UI, nocturia twice or more, and increased frequency, but none of these other symptoms were reported by most respondents within the urgency cluster, suggesting that urinary urgency occurs without multiple other LUTS in a portion of the population.

In both men and women, there was a single cluster that represented respondents with multiple and varied LUTS or the ‘multisymptom’ groups. These clusters would most likely benefit from a complete assessment of LUTS to differentiate treatment regimens and thus maximize treatment benefits. As noted by Chapple and Roehrborn [27], pharmacotherapies that target only the prostate do not alleviate symptoms that are attributed to overactive bladder and therefore combined treatments might be more effective. The targeting of interventions within the highly symptomatic groups to address the multitude of symptoms is an area in need of further research in terms of respondent expectations and outcomes.

While this cluster analysis yields interesting results in terms of considering multiple LUTS, the clinical relevance of these findings is yet to be determined. Important questions which cannot be addressed within the current dataset are the ‘bother’ impact and treatment-seeking behaviours of each of the symptom clusters. Also, the current cluster analysis is limited by the symptoms that were asked about and how they were elicited within the EPIC survey. The inclusion of additional LUTS that were not captured in EPIC (e.g. hesitancy, split stream, bladder pain) might or might not alter the cluster results found in the present analyses. Cluster analyses are needed in other large epidemiological databases to confirm or refute the clusters detected here, to further enhance the general applicability of these findings.

Although cluster analysis is not a new empirical technique, it is a novel way of examining LUTS that might provide insights on the potential underlying and overlapping pathophysiologies. In this analysis of EPIC data, three types of clusters were identified: a cluster with minimal symptoms, clusters characterized by one predominant symptom, and a cluster characterized by multiple LUTS. Future research with other epidemiological data is needed to replicate the current findings and provide greater confidence that these clusters truly represent distinct groups of people with LUTS. Cluster analysis has the potential to provide a new insight into the diagnosis and treatment of LUTS.

ACKNOWLEDGEMENTS

The authors acknowledge the Epic Study Group: Paul Abrams, Bristol Urological Institute, Southmead Hospital, Bristol, UK; Walter Artibani, Urology Unit, University of Padova, Padova, Italy; Christian Hampel, Urologische Klinik und Polyklinik, Johannes Gutenberg-Universität, Mainz, Germany; Sender Herschorn, Sunnybrook Health Science Centre, University of Toronto, Toronto, ON, Canada; Steinar Hunskaar, Department of Public Health and Primary Health Care, University of Bergen, Bergen, Norway; Con Kelleher, St. Thomas’ Hospital, London, UK; and Kate Reilly, Pfizer Inc, New York, NY, USA.

CONFLICT OF INTEREST

Karin Coyne, Louis Matza, and Christine Thompson are employees of United BioSource Corporation, who were paid consultants to Pfizer in connection with the development of the manuscript. Zoe Kopp is an employee of Pfizer. David Henry received an honorarium from United BioSource Corporation in connection with the development of this manuscript. S. Herschorn: Consultant – Pfizer, Astellas, Oryx, Triton, Lilly; Investigator – Pfizer, Sanofi, Astellas, Allergan; Grant Support – Pfizer, Astellas, Sanofi, Allergan; Lecturer – Pfizer, Astellas, Allergan, Triton. W. Artibani: Advisory Board – Astellas, Pfizer, Pierre-Fabre. I. Milsom: Consultant – Pfizer, United BioSource; Investigator– Pfizer, Astellas; Grant Support – Pfizer, Astellas; Lecturer – Pfizer, Astellas, Novartis. D. Irwin: Consultant – Pfizer; Meeting Participant – Pfizer; This study was sponsored by Pfizer Inc.

Appendix

The EPIC LUTS Questions:

Symptom/condition; defining question

Urinary urgency; Do you experience a sudden compelling desire to urinate which is difficult to put off? What I mean is a sudden intense feeling of urgency where you feel you must urinate immediately. (ICS definition)

Perceived frequency; In your opinion, do you feel that you urinate too often during the day? (ICS definition)

Frequency; Over the past month, how often have you had to urinate again less than two hours after you finished urinating? (IPSS)

Intermittency; Over the past month, how often have you found you stopped and started again several times when you urinated? (IPSS)

Urgency; Over the last month, how difficult have you found it to postpone urination? (IPSS)

Slow stream; Over the past month, how often have you had a weak urinary stream? (IPSS)

Straining; Over the past month, how often have you had to push or strain to begin urination? (IPSS)

Terminal dribble; Do you experience prolonged trickle or dribble at the end of your urine flow? (IPSS)

Incomplete emptying; Over the past month, how often have you had a sensation of not emptying your bladder completely after you finish urinating? (IPSS)

Nocturia; Over the last week, how many times did you typically get up to urinate from the time you went to bed at night until the time you got up in the morning? (IPSS)

Postmicturition UI; Do you experience urine leakage almost immediately after you have finished urinating and walked away from the toilet? (ICS definition)

UI; How often do you experience urinary leakage?

UUI; Do you leak urine in connection with a sudden compelling desire to urinate? By that, I mean in connection with a sudden intense feeling of urgency?

SUI; Do you leak urine in connection with sneezing, coughing, or when doing physical activities such as exercising or lifting a heavy object?

Ancillary