Empirical evaluation of grouping of lower urinary tract symptoms: principal component analysis of Tampere Ageing Male Urological Study data


Correspondence: Antti Pöyhönen, Department of Urology, Tampere University Hospital, Teiskontie 35/PO Box 2000, FIN-33521 Tampere, Finland.

e-mail: antti.j.poyhonen@gmail.com


What's known on the subject? and What does the study add?

  • The ICS has divided LUTS into three groups: storage, voiding and post-micturition symptoms. The classification is based on anatomical, physiological and urodynamic considerations of a theoretical nature. We used principal component analysis (PCA) to determine the inter-correlations of various LUTS, which is a novel approach to research and can strengthen existing knowledge of the phenomenology of LUTS. After we had completed our analyses, another study was published that used a similar approach and results were very similar to those of the present study.
  • We evaluated the constellation of LUTS using PCA of the data from a population-based study that included >4000 men. In our analysis, three components emerged from the 12 LUTS: voiding, storage and incontinence components. Our results indicated that incontinence may be separate from the other storage symptoms and post-micturition symptoms should perhaps be regarded as voiding symptoms.


  • To determine how lower urinary tract symptoms (LUTS) relate to each other and assess if the classification proposed by the International Continence Society (ICS) is consistent with empirical findings.

Materials And Methods

  • The information on urinary symptoms for this population-based study was collected using a self-administered postal questionnaire in 2004.
  • The questionnaire was sent to 7470 men, aged 30–80 years, from Pirkanmaa County (Finland), of whom 4384 (58.7%) returned the questionnaire.
  • The Danish Prostatic Symptom Score-1 questionnaire was used to evaluate urinary symptoms. Principal component analysis (PCA) was used to evaluate the inter-correlations among various urinary symptoms.


  • The PCA produced a grouping of 12 LUTS into three categories consisting of voiding, storage and incontinence symptoms. Post-micturition symptoms were related to voiding symptoms, but incontinence symptoms were separate from storage symptoms.
  • In the analyses by age group, similar categorization was found at ages 40, 50, 60 and 80 years, but only two groups of symptoms emerged among men aged 70 years. The prevalence among men aged 30 was too low for meaningful analysis.


  • This population-based study suggests that LUTS can be divided into three subgroups consisting of voiding, storage and incontinence symptoms based on their inter-correlations.
  • Our empirical findings suggest an alternative grouping of LUTS. The potential utility of such an approach requires careful consideration.

Danish Prostatic Symptom Score


International Continence Society


principal component analysis


Tampere Ageing Male Urological Study


Lower urinary tract symptoms include a wide range of complaints and physiological phenomena. Population-based studies have indicated that LUTS are very common and also affect younger men [1-4]. The ICS has divided LUTS into three groups: storage, voiding and post-micturition symptoms [5]. The classification is based on anatomical, physiological and urodynamic considerations of a theoretical nature. Storage symptoms are experienced during the storage of urine in the bladder, voiding symptoms while passing urine and post-micturition symptoms occur after voiding. Pain symptoms and those associated with sexual intercourse or pelvic organ prolapse are considered separately.

In the present study, we evaluated how well the current ICS classification is corroborated by empirical findings by exploring the inter-relationships among 12 LUTS to determine which symptoms occur together in a population-based study of men aged 30–80 years. We used principal component analysis (PCA) to determine the inter-corelations of various LUTS, which is a novel approach to research and could strengthen the existing knowledge about the phenomenology of LUTS. Recently, after we had completed our analyses, another study was published that used a similar approach and its results were similar to those of the present study [6].

Materials and Methods

The present report is a part of the population-based Tampere Ageing Male Urologic Study (TAMUS) cohort study [7, 8]. The first TAMUS survey was conducted in 1994 and the survey has been repeated every 5 years since then. Data for the present study was gathered in 2004.

The original TAMUS study population included men born in 1924, 1934 and 1944. They represented 80, 70 and 60-year-old men in the study. An additional group of men was recruited who were born in 1954, 1964 and 1974 to represent 50-, 40- and 30-year-old men.

The majority of the target population lived in Tampere, which was the third largest city in Finland with a population of 203 000 in 2004. About 20% of the target population lived in 11 small rural municipalities surrounding the city of Tampere in Pirkanmaa County. All men in this area, born in the aforementioned years, were included in the study. The men were identified from the Population Register Centre and invited to enroll in the study.

Self-administered questionnaires were posted in personally addressed letters with a prepaid return envelope enclosed. The first questionnaire was sent in April 2004 to 7464 men. A reminder questionnaire was sent 3 months later and for the younger age groups a second reminder questionnaire was also used. In all rounds, data collection finished and data were recorded 2 months after posting. Detailed participant numbers are shown in Fig. 1.

Figure 1.

Flow chart showing responders for analysis.

Background information about men was obtained through the questionnaire and has been published previously in more detail [9, 10]. LUTS were evaluated using the Danish Prostatic Symptom Score (DAN-PSS)-1 [11, 12]. The score system is based on the severity of 12 LUTS related to storage, voiding and post-micturition symptoms. Three questions related to sexual function were not used in the present study. The severity of each symptom was graded by the respondent on a scale from 0 (no symptoms) to 3 (severe symptoms). The preceding 4-week period was used as the time frame for the questionnaire.

Background information was analysed using a two-sided chi-squared test and a linear-by-linear test. Correlations between LUTS were evaluated using Spearman's rank correlation coefficient (r). A value of zero indicates the absence of association and value 1 corresponds to perfect association. LUTS were evaluated by using PCA. PCA provides a means of reducing the dimensionality of a dataset consisting of a large number of inter-related variables, while retaining as much as possible of the variation present in the dataset. In other words, it can be used to combine nearly the same information as provided by the original variables by combining them into a small number of new summary vectors. This is achieved by transforming to a new set of variables, the principal components, which are uncorrelated, and ordered so that the first few retain most of the variation present in all of the original variables [13]. We used PCA to divide the 12 LUTS into smaller groups and evaluated which symptoms seemed to relate to each other. For this purpose a covariance or correlation matrix is created and principal components selected that provide both maximum correlation with the original variables and minimum correlation with each other. Various rotation methods are used to simplify data structure. Unlike confirmatory factor analysis, the PCA was entirely data-driven (exploratory), i.e. we imposed no restrictions on the number of summary variables (principal components), nor was any previous hypothesis incorporated into the analysis about which symptoms should be combined together (such as symptoms should represent a meaningful joint entity or set of characteristics i.e. comprehensibility criterion). An eigenvalue >1.0 was used as the criterion to include a summary variable in the analysis, in accordance with conventional procedures in PCA.

The PCA was performed for all men and then for each age group separately. Data were checked and a Kaiser–Meyer–Olkin index of sampling adequacy and a Bartlett test of sphericity were used to evaluate the factorability of the correlation matrix. In all analyses, Bartlett tests gave a P value <0.001, but it must be borne in mind that the Bartlett test can also give statistically significant results for minor correlations in large samples. Kaiser–Meyer–Olkin test scores were also high (0.876 for the entire dataset and, although they decreased with age, the score was still 0.809 even in the age group of 40 years). Generally it is required that the Kaiser–Meyer–Olkin score should be ≥0.60 for the data to be suitable for PCA. Oblimin rotation was used for the analysis. SPSS (Statistical Package for the Social Sciences) version 17.0 was used in the data analysis.

The study protocol was approved by the Tampere University Hospital Committee of Research Ethics (tracking number 99050).


The overall participation rate was 58.7% (4384 returned questionnaires out of 7470). More detailed information about participation is shown in Fig. 1. A total of 226 men did not answer all the 12 questions in DAN-PSS-1 and they were excluded from analysis. Incomplete responses were more common in the older age groups. In total, 4158 men were included in the PCA.

For all individual LUTS, mild symptoms were much more common than moderate or severe symptoms (Fig. 2). Correlations between LUTS are shown in Table 1. Correlations among different voiding symptoms, including feeling of incomplete emptying, were stronger than among other LUTS, and high correlations were also found among different types of incontinence. Overall correlations among symptoms were weak.

Figure 2.

Prevalence (%) of individual LUTS in the TAMUS 2004 cohort.

Table 1. Relationships among LUTS using Spearman's rank correlation coefficient in men aged 30–80 years.
Slow stream0.3680.3941         
Feeling of incomplete emptying0.3930.4470.4720.2761       
Post-micturition dribble0.2730.3200.2790.2160.3781      
Increased daytime frequency0.1590.1960.2380.1380.2650.2161     
Urgency incontinence0.1450.1870.2840.1440.3050.2930.3060.3270.4231  
Stress incontinence0.1520.1750.1990.1610.2100.2270.1930.2050.2240.3541 
Other incontinence0.1180.1540.2160.1670.2130.2330.2090.2170.2420.4300.4761
 HesitancyStrainingSlow streamDysuriaFeeling of incomplete emptyingPost-micturition dribbleIncreased daytime frequencyNocturiaUrgencyUrgency incontinenceStress incontinenceOther incontinence

In the PCA for all 4158 men, three summary variables (principal components) were formed as a result of the analysis of inter-correlations among symptoms, and they largely corresponded to entities classifiable as voiding, incontinence and storage symptoms (Table 2). No main component corresponding to post-micturition symptoms emerged, but post-micturition dribble showed a weak correlation with voiding symptoms. The main component corresponding to voiding symptoms showed a strong correlation (r > 0.8) with hesitancy and straining, but only a weak (<0.5) correlation to post-micturition dribble. Similarly, the incontinence component had strong correlations with stress and other incontinence symptoms. Urge incontinence was classified with the storage symptoms, but it also had a strong association with incontinence symptoms. Likewise, weak stream and feeling of incomplete emptying were allocated among voiding symptoms, but were also associated with storage symptoms. Overall, the main component related to storage symptoms had the least prominent profile, with moderate to weak correlations with individual symptoms, except in the oldest age group (80 years), among whom the main component that was related to storage symptoms showed good correlations to its components.

Table 2. Principal component analysis in men aged 30–80 years.
Weak stream0.552−0.017−0.303
Feeling of incomplete emptying0.517−0.055−0.400
Post-micturition dribble0.3280.212−0.219
Stress incontinence0.0550.8490.054
Other incontinence−0.0160.841−0.046
Daytime frequency−0.072−0.002−0.749
Urge incontinence−0.0770.479−0.514

In the analyses by age group, the youngest men (30-year-olds) could not be analysed separately because of the failure of the Oblimin rotation to converge owing to low prevalence of LUTS. In the 70-year-old men, only two components were found (voiding component separately and incontinence/storage symptoms combined), while in the other age groups three components emerged similar to the main analysis of all men, containing mainly voiding, storage and incontinence symptoms, respectively (Table 3). Urge incontinence showed loadings with both incontinence and storage components in the age groups 50, 60 and 80 years, similar to the overall analysis. In addition, hesitancy and straining showed the strongest correlations with the voiding component in the age-specific analyses. Likewise, the individual symptoms had only weak-to-moderate correlations with the summary component for storage symptoms in each age group. Post-micturition dribble showed only a weak correlation (r = 0.3–0.4) with the voiding component (except in the age group 70 years, where there was a moderate correlation with the combined incontinence/storage component).

Table 3. Correlations among symptoms (ICS classification) and principal components in different age groups.
SymptomComponent40-year-old men50-year-old men60-year-old men70-year-old men80-year-old men
  1. aNot included in ICS classification.
Voiding symptoms               
Hesitancy 0.807−0.044−0.0950.777−0.0800.0010.904−0.0290.1800.787−0.125−0.1870.917−0.075
Straining 0.777−0.1100.0630.810−0.0850.0300.8300.0230.0380.851−0.089−0.0070.780−0.201
Slow stream 0.6840.128−0.2200.6320.0860.0940.730−0.126−0.1370.6200.2100.2500.6190.081
Dysuriaa 0.4670.0580.1010.5860.297−0.2810.5380.087−0.0320.3650.3340.1880.5800.123
Post-micturition symptoms               
Feeling of incomplete emptying 0.683−0.0220.1720.644−0.0480.2140.6140.043−0.1930.6910.1130.7070.068−0.031
Post-micturition dribble 0.4090.0820.1240.4020.1240.2730.2960.303−0.1660.2030.5570.3800.143−0.351
Storage symptoms               
Increased daytime frequency −0.0870.0730.748−0.004−0.0500.737−0.040−0.028−0.7940.1270.4870.669−0.019−0.141
Nocturia 0.087−0.0780.6470.225−0.0690.4950.095−0.095−0.6710.1920.4450.7610.0230.133
Urgency 0.2090.0670.5210.1520.1220.6060.0560.108−0.6750.2180.5900.6730.084−0.102
Urgency incontinence −0.0400.6000.334−0.1370.4760.573−0.0680.475−0.461−0.0740.8030.498−0.002−0.464
Stress incontinence 0.1330.771−0.1640.0940.829−0.0730.0010.8500.084−0.1690.775−0.0890.147−0.870
Other incontinence −0.0350.816−0.003−0.0220.8150.0890.0280.8280.008−0.1880.7760.171−0.068−0.818


We evaluated the constellation of LUTS using PCA of the data from a population-based study of >4000 men. In our analysis, three components emerged from the 12 LUTS, namely, voiding, storage and incontinence components. Overall, our findings pointed to a need for some revision of the ICS classification, as the grouping into voiding, storage and post-micturition symptoms were partly, but not entirely consistent with the current taxonomy. Our results indicated that incontinence may be separate from the other storage symptoms and post-micturition symptoms should perhaps be regarded as voiding symptoms.

The PCA results were very similar in all age groups. In the youngest age group (30-year-old men) symptoms were so mild and uncommon that PCA could not be performed. In the 70-year-old men, the analysis resulted in only two components, the incontinence group was linked to the storage component, which in older men is to be expected and in fact incontinence symptoms are storage symptoms in the ICS classification. The oldest group (80-year-old men) was substantially smaller than the other groups, but the classification of symptoms was consistent with the younger men (40- and 50-year-old men). Storage symptoms had a prominent profile in the oldest age group, as expected. Overall, the correlations among LUTS were similar across most age groups, which indicated the validity of our analysis and the uniform nature of LUTS over a wide age range. Post-micturition symptoms (post-micturition dribble and incomplete emptying) were fused to the voiding component. By contrast, incontinence symptoms (urgency incontinence, stress incontinence and other incontinence) formed a component of their own. Urgency and urgency incontinence were very closely correlated. Urgency incontinence was related to two components (cross-loading) in almost every age group. This is probably because urgency incontinence is firstly an incontinence symptom but also closely associated with urgency and it can be difficult even for patients to distinguish between these two.

Voiding symptoms seemed to correlate with each other and the so-called post-micturition symptoms were included in the voiding symptoms. Storage symptoms formed another main group, but it seemed that incontinence symptoms should be a third class, instead of post-micturition symptoms, if we base the categorization on how these symptoms are related to each other. The ICS symptom taxonomy was conceived on theoretical grounds and lacked an empirical foundation. The present study provides a new perspective on LUTS.

Principal component analysis is a mathematical procedure (based on linear orthogonal transformation) that can be regarded as an exploratory analysis for revealing interdependencies within complex, multidimensional data by reducing redundancies between variables. It can be used to identify unobserved (latent) variables (principal components) that can be used for expressing the observations as linear combinations of such components. PCA has been used previously in only one study of LUTS which was limited to nocturia [14]. Studies evaluating the composition of various individual LUTS to divide them into smaller sub-groups were also found, but these used a cluster analysis [15, 16]. These studies aimed to group individual patients according to their presenting LUTS, while our goal was to determine the inter-correlations of various LUTS in order to assess the current taxonomy. An alternative approach could have been confirmatory factor analysis, using the ICS classification as an a priori structure, but we wanted to introduce no such constraints to the analysis. Our data covered well over 4000 men and the ratio of data items (symptoms) to observations was well above the suggested 10:1 ratio, although the binary data in symptom occurrence reduce the variability and hence statistical efficiency. After we had completed our analyses, a study evaluating the grouping of LUTS using exploratory factor analysis was published [6]. Exploratory factor analysis was used to reclassify various LUTS for a secondary analysis, not performed as the main goal of the paper. The results were similar to those of the present study, however, with three categories: voiding symptoms, predominantly storage symptoms and storage items related to urinary incontinence. In the present analysis, dysuria was included among voiding symptoms and both urgency incontinence and stress incontinence belonged to the main group of incontinence, while Coyne et al. [6] were unable to classify pain symptoms, and urgency incontinence and stress incontinence were classified into different main groups.

The DAN-PSS-1 questionnaire used in the present study was slightly different from that used in the EPIC and BACH studies, as the items on terminal dribble and post-micturition dribble were combined into a single question. Our questionnaire also included assessment of dysuria among the voiding symptoms, unlike the ICS symptom categorization [5]. Correlations among individual symptoms were relatively low indicating that none of the 12 items could be replaced.

Finland has a strong history of population-based epidemiological studies, and we not only achieved a high response proportion, but also maintained it over a follow-up period spanning 15 years. A full description of the TAMUS study protocol has been published earlier [7, 8]. We used a postal questionnaire, which is less prone to bias in symptom reporting than other methods of data collection [17]. There were no major differences in the symptom prevalence between early and late responders, which indicates that major selection bias is unlikely, but no information was available on non-responders. Non-response would, however, affect our findings only if the pattern between LUTS was strongly different among non-responders than participants.

Our findings suggest that LUTS could be coherently divided into subgroups consisting of voiding, storage and incontinence symptoms. Post-micturition symptoms could be included in the voiding symptoms. Combining various incontinence symptoms into a separate category might draw attention to therapeutic approaches tailored for the group of patients. The findings are consistent with clinical practice, with different diagnostic and treatment options for incontinence symptoms than other LUTS.

In conclusion, in this population-based study, PCA based on correlations among different LUTS was used to construct an empirical classification of urinary symptoms. Three components emerged that largely correspond to voiding, storage and incontinence symptoms. As compared with the ICS classification, post-micturition symptoms were combined with voiding symptoms. Storage symptoms formed one entity, but incontinence symptoms constituted a separate component. The constellation of symptoms was consistent for men aged 40, 50, 60 and 80 years.


The work of Antti Pöyhönen was supported by a grant from the Finnish Association of Urology.

Conflict of Interest

None declared.