Assessing the impact of nocturia on health-related quality-of-life and utility: results of an observational survey in adults

Abstract

Background and aim: The impact of nocturia (getting up at night to void) on health-related quality-of-life (HRQoL) is often under-estimated. This study investigated the relative burden in terms of HRQoL and utilities of nocturia in a real-world setting.

Methods: Patient data were collected from two surveys: a nocturia-specific, cross-sectional survey of physicians and their patients (DSP), and a general UK population health survey (HSFE). Utilities (EQ-5D-5L), productivity (Work Productivity and Activity Index), and the impact of nocturia symptoms (Nocturia Impact Diary and Overactive Bladder Questionnaires) were assessed against the number of voids. A robust linear regression model with propensity score weights was used to control for confounding factors in estimating utilities.

Results: Physician-recorded data were available from 8,738 patients across the US, Germany, Spain, France, and the UK; of these, 5,335 (61%) included patient-reported outcomes. In total, 6,302 controls were drawn from the two surveys and compared to 1,104 nocturia patients. Deterioration of HRQoL was associated with increasing number of night-time voids (p < 0.0001). In particular, significant differences were observed between 0–1 and ≥2 voids (p < 0.001). The regression model demonstrated that nocturia (≥2 per night) is associated with a modest but significant deterioration in utility of 0.0134 (p < 0.05).

Limitations: The cause of nocturia is multifactorial and the mostly elderly patients may have several concomitant diseases. The authors tried to adjust for the most common ones, but there may be diseases or unknown relationships not included.

Conclusions: Nocturia negatively affected HRQoL and patient utility. A clear effect is seen already at two voids per night. Every effort should, therefore, be made to reduce nocturia below the bother threshold of two voids per night.

Keywords:

• Nocturia
• Utility
• Quality-of-life (QoL)
• EQ-5D
• Nocturia Impact (NI) diary
• WPAI
• Observational survey

Introduction

The International Continence Society (ICS) defines nocturia as the condition when an individual has to wake at night to void one or more times, with each of the voids preceded and followed by sleep¹, and the American Urological Association (AUA) as the need to urinate at least twice during the night². The latter is the most frequently used, also used in this study.

Nocturia is a multifactorial disease that differs from other lower urinary tract symptoms (LUTS). While nocturia is mostly the result of excessive urine production during the night (nocturnal polyuria), it can also be the result of sleep disorders (e.g. sleep apnea) or benign prostatic hyperplasia (BPH) in men, but also chronic illnesses such as diabetes or cardiovascular diseases and/or as a side-effect to medication^3,4.

Nocturia is increasingly recognized as one of the most bothersome symptoms for those suffering from LUTS⁵, and the leading cause of sleep disturbance across all ages⁶. The repeated fragmentation of sleep that individuals affected with nocturia experience can have a profound effect on their sleep pattern and daily tiredness, and negatively impact their health-related quality-of-life (HRQoL)^7,8. The degree of bother and decrease in HRQoL has been shown to directly relate to the severity of nocturia as the HRQoL of individuals with two or more voids (2+ per night) worsens with increasing episodes of nocturia⁹.

While studies in various countries^7,10 and with different designs^8,11 reflect a causal association between nocturia and detrimental HRQoL, clear evidence of a direct relationship for nocturia and patient utilities in a clinical practice setting is lacking.

The importance of evidence-based systems in informing payer decision-making was a key conclusion in a 2013 report by the not-for-profit think tank and strategy company Meteos (Oxford, OX1 1JH, UK); it recognized the importance of real-world (i.e. clinical practice) observational studies to enable health systems to deliver increased productivity and better outcomes while managing unprecedented demand and intense cost pressure¹². Using real-world data from large multinational databases provides the best possible evidence for future health technology assessment calculations.

The objectives of this study were to assess the relative impact of nocturia on the HRQoL and utility of patients actively seeking help and to measure the absolute burden of two or more voids per night on a generic scale relative to a matched-control population (non-nocturics, i.e. 0–1 nocturnal voids).

Methods

Data sources, HRQoL instruments, and study population

Data for the study were collated from two sources: the 2013 Adelphi LUTS Disease Specific Programme (DSP), and the Health Survey for England 2011 (HSFE)¹³. Data from nocturia patients were extracted from the DSP survey whilst the control population (subjects without nocturia) were extracted from the HSFE and DSP surveys. These databases were chosen as both surveys use similar methodology for data collection and key variables required to assess HRQoL burden.

The DSP is a cross-sectional, real-world, multinational survey of physicians and their consulting LUTS patients, across the US and four EU countries (Table 1). The survey conducted in February–May 2013 involved 264 primary care physicians and 371 specialists (urologists, gynecologists, and urogynecologists). During the survey, physicians completed a patient record form (PRF) for the next 14 consulting patients whose diagnosis included BPH, and/or overactive bladder (OAB) and/or nocturia/nocturnal polyuria. The same patients were then invited to fill out a patient self-completion form (PSC) and several PROs such as the EQ-5D-5L and the OAB Questionnaire (OAB-q)^14,15. Patients were not tested or investigated prior to or during the survey.

Table 1. Characteristics of the two data sources.

Data source characteristic	DSP (patient data) (n = 8,738)	HSFE (control data) (n = 10,000)
Year of data collection	2013	2011
Survey type	Cross-sectional	Cross-sectional
Respondent/source	Patient-reported outcomes, with clinical data collected from physicians, self-reported completion method	Patient-reported outcomes, with clinical data collected from nurse, face-to-face interview method
Patient-reported outcome measures collected	EQ-5D-5L, WPAI, OAB-q, NI Diary	EQ-5D-5L
Confounding covariates collected	Demographics, comorbidities, drugs prescribed, health/lifestyle indicators, income, home circumstances, treatment history, physician-reported outcomes	Demographics, comorbidities, drugs prescribed, health/lifestyle indicators, income, home circumstances
Country coverage Patient sample	UK, France, Germany, Spain, USA LUTS	UK General population

DSP, Disease Specific Programme; HSFE, Health Survey for England; EQ-5D-5L, EuroQoL 5 Dimension 5 Level; WPAI, Work, Productivity and Activity Index; NI, Nocturia Impact; LUTS, lower urinary tract symptoms.

The DSP was conducted in accordance with the European Pharmaceutical Market Research Association code of conduct for international healthcare market research¹⁶ and the US Health Insurance Portability and Accountability Act 1996. Patients had to provide consent for de-identified and aggregated reporting of research findings as required. Data were collected by local fieldwork partners and fully de-identified. Ethical approval is not necessary for surveys of this nature, as the purpose of the research is to improve understanding and not to test any hypotheses nor treatments. Clinical practice should not be affected by the survey. The methodology has been published previously¹⁷.

HSFE 2011 is an independent cross-sectional data set, which, at the time of fieldwork and analysis, contained the most up-to-date health data from the general public in the UK (n = 10,000)¹³. To the best of our knowledge it is the only publically available general public dataset containing EQ-5D outcome data at the patient level.

For this study, two analyses were conducted: a regression model and a descriptive patient analysis. The regression analysis included nocturia patients from the DSP and controls comprising individuals without nocturia from both the DSP and HSFE samples. Inclusion criteria for these three groups were as follows:

• Nocturia patients from DSP—diagnosed with OAB and/or BPH and/or nocturia, experiencing two or more (≥ 2) nocturnal voids on average and providing complete data for all regression covariates. EQ-5D completed via unsupervised patient questionnaire.

• DSP controls—as above, but experiencing less than two (i.e. 0–1) nocturnal voids on average. EQ-5D completed via unsupervised patient questionnaire.

• HSFE controls—a general population sample excluding those with urinary conditions (assumes remaining individuals have less than two nocturnal voids; number of voids was not collected). EQ-5D completed by the patient in a supervised interview environment.

The descriptive analysis included all LUTS patients in the DSP stratified by number of nocturnal voids, but not restricted to those who provided complete data for all variables, as required for the regression analysis.

A further inclusion criterion was applied for the regression analysis but not the descriptive analysis. Nocturia patients and controls were only included if their propensity score fell inside a common support zone, a form of matching. The propensity score is the likelihood of an individual being assigned to the patient group based on their demographic and health-related characteristics¹⁸.

Based on the available nocturia literature, recorded variables in the databases, and the author’s experience, the model included: age, gender, body mass index (BMI), smoking status, comorbid cancer, arthritis, other joint condition, asthma, heart condition, digestive condition, other comorbid condition, educational attainment, employment status, and income. They, and some of the commonly used treatments, have all previously been shown to correlate with nocturia. Independent variables included in the model were also selected based on identifying appropriate variables included in both datasets, measured by either the same method or in a way that could feasibly be transformed to ensure comparability. The most problematic variables for comparison between datasets were those comorbidities which were defined differently. These were, therefore, slightly adjusted (e.g. sub-groups combined) to allow for meaningful comparison. As age has a non-linear confounding relationship with EQ-5D, with an increased effect beyond the age of 60 years, an additional dummy (binary) age variable was included for patients above this age. To investigate whether the data collection method (face-to-face interview vs patient self-reported questionnaire) was masking the disease effect (due to perfect co-linearity), controls were taken from both datasets and included in order to quantify this suspected collection bias.

The UK value set was applied to all patients (as the HSFE is UK-only), regardless of country of residence, in order to ensure all patients and controls used the same EQ-5D scoring method, thus providing comparable outcomes across respondents from different countries.

The listed variables were used to derive a propensity score for each patient and control. For the regression, a common support zone was identified using the propensity score range of nocturia patients and control subjects. Subjects with a propensity score outside the common range of patient and control scores (i.e. potential controls with a score higher than the highest patient score or lower than the lowest patient score) were excluded from the population (control) group. This ensured that unsuitable controls, such as children, were excluded from the analysis as their propensity score would fall outside the common support zone (the interval on the propensity score axis covered by both treatment and control observations).

EQ-5D-5L is a standardized measure of health status developed in order to provide a simple, generic measure of health for calculation of utilities, a measure for clinical and economic appraisal across different diseases¹⁴. It essentially consists of two measures—the EQ-5D-5L descriptive system and the EQ visual analog scale (EQ VAS). The former comprises the following five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The EQ VAS is an overall estimation of the present health status. An EQ-5D-5L estimation may be converted to a single summary index by applying a formula that essentially attaches weights to each of the levels in each dimension. This formula is based on the valuation of EQ-5D-5L health status from general population samples, in this case the UK.

The OAB-q is a 33-item, condition-specific measure designed to assess the impact of OAB symptoms on HRQoL¹⁵. The OAB-q consists of a Symptom Bother scale and four HRQoL sub-scales (Coping, Concern, Sleep, and Social Interaction). All scales scores are transformed to a 0–100-point scale with higher Symptom Bother scores indicating greater symptom severity. This instrument was included as previous research shows clinical overlap of OAB and nocturia, where nocturia until recently was regarded as a sub-component of OAB³.

Work productivity and activity impairment was assessed using the Work Productivity and Activity Impairment Questionnaire (WPAI). The WPAI is a specific tool for quantifying the effect of a specific health problem on absenteeism, presenteeism, work productivity loss, and activity impairment. It is summarized into percentage Overall Work Impairment and Activity Impairment, respectively¹⁹. Previous research shows that productivity loss is a very costly component of the total societal burden of nocturia¹¹.

The disease-specific impact of nocturia was investigated by the Nocturia Impact Diary, which is a 12-item instrument consisting of 11 core items and an overall impact question²⁰. Responses are scored from 0 (lowest) to 4 (highest). The Nocturia Impact total score is calculated by adding the 11 core items. The Nocturia Impact total score and the overall impact question are standardized from 0 (lowest) to 100 (highest). The NI Diary was included as it is the most recently developed HRQoL instrument for nocturia, in line with the recent methodological guidelines and regulatory requirements.

Statistical analysis

HRQoL was assessed using different tools and across different symptom severities to identify a consistent pattern in nocturia patients that differs from the general population.

Absolute burden of nocturia on patients’ QoL was assessed with EQ-5D-5L. A robust linear regression model weighted by propensity scores was used to control for confounding factors and quantify the disease effect on HRQoL. Robust regression was used to allow for EQ-5D being a uniquely non-normal outcome variable^21,22. The regression included all of the previously listed clinical and demographic variables from which the propensity scores were derived as well as the propensity scores as weights. A sub-class propensity scoring method was used, using the MatchIt package in R version 3.0. For the general population controls only EQ-5D could be assessed in the regression model.

Variables collated from both data sources had to be similar and broad to be included in the analysis; therefore, allowing the inclusion of confounding factors influencing the outcome variables. Key differences between both data sources were country coverage and private vs interview methodology. Both were addressed in the methodology: the country coverage is only limited in the control sample and was adjusted for by using the UK value set of the EQ-5D²³ for both controls and patients, thus providing a like-for-like comparison of outcome measure (as recommended by the EQ-5D designers²⁴). The data collection method was controlled by using a dummy variable in the regression model to account for any potential bias in data collected.

For all patient-reported outcomes descriptive analysis was conducted for all DSP patients by frequency of nocturnal voids (0–1, 2, 3, 4, or ≥5). A Kruskal–Wallis test was performed to test for dependence between HRQoL measures and number of nocturnal voids. To test for differences between groups (0–1, 2, 3, 4, and 5 nocturnal voids), pairwise comparisons were conducted using a Mann–Whitney test with Bonferroni correction on adjusted p-values. An additional pairwise comparison was undertaken to compare patients with 0–1 vs 2+ nightly voids, as suggested by AUA guidelines to assess the impact of nocturia on HRQoL².

Results

Study population

Demographic characteristics and clinical variables for the nocturia patients (DSP, n = 1,104) and controls (HSFE + DSP, n = 6,302) included in the regression analysis are summarized in Table 2.

Table 2. Nocturia patient and control profiles (unadjusted) with complete data.

Mean or prevalence values	Nocturia patientsDSP (≥ 2 voids)(n = 1,104)	Controls (0–1 voids)DSP + HSFE(n = 1,663 + 4,639)	Significance(p)
Age, years	65.1	53.1	<0.01
Male gender, %	59.1	48.2	<0.01
BMI	27.6	27.6	NS
Current smoker, %	14.9	18.7	<0.01
Comorbid cancer, %	3.5	2.0	<0.01
Comorbid arthritis, %	12.6	9.8	<0.01
Comorbid asthma, %	3.6	5.1	<0.05
Comorbid heart condition, %	3.2	10.9	<0.01
Comorbid joint condition, %	7.6	9.5	<0.05
Comorbid digestive condition, %	7.0	4.3	<0.01
Other comorbid condition, %	24.9	11.9	<0.01
Degree-educated, %	31.0	31.1	NS
Unemployed, %	4.4	4.8	NS
Below median income, %	56.0	45.6	<0.01
Propensity score*	0.261	0.129	<0.01
EQ-5D-5L	0.781	0.836	<0.01

DSP, Disease Specific Programme; HSFE, Health Survey for England; n, number of patients; BMI, body mass index; EQ-5D-5L, EuroQoL 5 Dimension 5 Level; NS, not significant.

Note: Propensity score summarizes the differences in characteristics between patients and controls.

Of the 6,302 subjects in the comparison group, 4,639 (73.6%) were drawn from the HSFE database and 1663 (24.4%) from the DSP (< 2 voids) to adjust for differences in the methodology used for data collection. These subjects were inside the common support zone and had complete data.

Relative burden of nocturia on patients’ QoL—patient-reported outcomes and number of voids in the DSP population

PRFs were available from 8,738 DPS patients; of these, 5,335 (61.1%) completed a PSC. Table 3 presents the differences in EQ-5D, OAB-q, WPAI overall work impairment/activity impairment, and NI Diary scores as per disease severity.

Table 3. Four measures of patient-reported outcomes in relation to number of voids in the DSP population.

Number of nightly voids	EQ-5D-5L* (n = 4,931†)	EQ-5D VAS* (n = 4,875†)	OAB-q total HRQoL score‡ (n = 4,907†)	WPAI % overall work impairment‡ (n = 1,246†)	WPAI % activity impairment‡ (n = 4,768†)	NI Diary overall score‡ (n = 4,592†)
0–1	0.88 (n = 1388)	79 (n = 1366)	19.9 (n = 1373)	17 (n = 442)	25 (n = 1341)	16.6 (n = 1253)
2	0.84 (n = 1388)	74 (n = 1373)	26.7 (n = 1389)	23 (n = 333)	32 (n = 1337)	25.4 (n = 1314)
3	0.83 (n = 1082)	71 (n = 1070)	33.9 (n = 1084)	30 (n = 246)	38 (n = 1051)	34.6 (n = 1027)
4	0.83 (n = 592)	68 (n = 588)	39.7 (n = 584)	35 (n = 129)	42 (n = 576)	40.9 (n = 557)
≥5	0.79 (n = 481)	66 (n = 478)	44.6 (n = 477)	34 (n = 96)	48 (n = 463)	44.0 (n = 441)

* higher = better health;

† number who completed question;

‡ lower = better health.

DSP, disease specific program; NI, nocturia impact; OAB-q, overactive bladder questionnaire (a majority of the sample has some form of OAB diagnosis, either alone or in combination with BPH/nocturnal polyuria/nocturia); VAS, visual analog scale, WPAI, work productivity and activity index.

Note: For all measures, lower outcome scores were observed to be associated with increasing number of night time voids (p < 0.000,1). Significant difference was also observed between 0–1 and 2+ nightly voids (all measures p < 0.001).

For all HRQoL and symptom measures, there was a clear pattern showing deterioration in outcome with increasing number of night-time voids (p < 0.000,1).

Absolute burden of nocturia on patients’ QoL—utility of nocturia compared to the group of comparison

The overall EQ-5D score was significantly lower in nocturia patients than in the control population (0.781 vs 0.836; p < 0.01) (Table 2), as was also evident in Table 3. This difference may be affected by several factors other than nocturia, such as method of data collection, comorbidities, income, and age.

The purpose of the regression model (Table 4) was to take into account both patient clinical and demographic characteristics (from which the propensity score is derived) and bias due to the data collection method. The model showed that nocturia (2+ voids per night) is associated with a modest but significant deterioration in HRQoL of 0.013,4 (p < 0.05) as measured by EQ-5D-5L.

Table 4. Results of multivariate regression predicting nocturia vs control group status.

Variable	Variable type	Coefficient estimate	Standard error	Significance (p)
Intercept	Continuous, model-derived	1.0047249	0.0149731	<0.001
Interviewed patient	Binary	0.0184052	0.0046686	<0.001
2+ voids	Binary	−0.0133981	0.0063830	<0.05
Age	Integer	−0.0003849	0.0002664	NS
Over 60 years old	Binary	0.0151501	0.0064296	<0.05
Gender (male)	Binary	0.0113263	0.0033314	<0.001
BMI	Continuous	−0.0032289	0.0003746	<0.001
Current smoker	Binary	−0.0087762	0.0044871	NS
Comorbid cancer	Binary	−0.0618102	0.0146990	<0.001
Comorbid arthritis	Binary	−0.1100500	0.0068219	<0.001
Comorbid asthma	Binary	−0.0366012	0.0084888	<0.001
Comorbid heart condition	Binary	−0.0009191	0.0086220	NS
Comorbid digestive condition	Binary	−0.0147439	0.0093484	NS
Comorbid joint condition	Binary	−0.1302524	0.0068460	<0.001
Comorbid other	Binary	−0.0371602	0.0072231	<0.001
Education: Less than high school	Binary	−0.0375601	0.0056467	<0.001
Education: College or more	Binary	−0.0072142	0.0040172	NS
Unemployed	Binary	0.0129933	0.0097849	NS
Employed	Binary	0.0223549	0.0051461	<0.001
Cohabiting	Binary	0.0297289	0.0044793	<0.001
Below median income	Binary	−0.0063736	0.0044570	NS
Propensity score	Continuous	−0.2323123	0.0428223	<0.001

NS, Not significant; BMI, Body Mass Index.

The model demonstrates a reliable goodness of fit, with a low robust residual standard error (0.115,7), and approximately normal residuals (median =0.015, quartile one = −0.115, quartile three =0.084). The crude R-squared was 0.242. However, the model fit was less predictively reliable for values of EQ-5D below 0.6, where the mean difference between observed and predicted values increases substantially. Above EQ-5D values of 0.6, which makes up over 90% of the observations in the model, the model is reliable, providing mean differences ranging from −0.129–0.101. Overall observed vs predicted value differences were −0.03.

Discussion

Several previous studies on nocturia have considered only patients with two or more voids per night^8,25 on the basis that a nocturnal frequency of one void per night does not seem to be harmful or bothersome²⁶. This HRQoL per number of void analysis showed an increasing burden in all aspects of QoL/symptom and activity impairment. In particular, significant differences were observed between <2 and 2 or more nightly voids (for all measures p < 0.001), therefore supporting a more restrictive definition of nocturia of 2+ voids per night, as recommended by AUA². We applied a broad set of PRO (patient-reported outcome) instruments, covering such different aspects as nocturia and the relation to OAB, impact on work and free-time activities, and nocturia-specific QoL. These instruments have been widely used before, comprehensively validated, and they all clearly point to the threshold of two voids or more. This data also confirms the comprehensive Tikkinen et al.²⁷ study, which showed that the degree of bother starts having a significant impact at two or more voids per night and increases further with each additional void. To maximize the beneficial impact of nocturia-specific therapies, physicians should be encouraged to consider the more restrictive definition suggested by the AUA, along with the use of voiding diaries and measurement of patient bother.

The multivariate analysis focused on the EQ-5D has four key advantages. First, EQ-5D is well known and widely used by healthcare decision-makers and payers. Second, it provides a cross-disease standardized measure for assessing patient burden relative to other conditions, therefore enabling efficient allocation of healthcare resources. Third, by applying the propensity score matching we reduce the potential bias due to confounding variables, in an attempt to mimic randomization. Finally, it enables EQ-5D-based QALY calculation, allowing economic evaluations that are credible to decision-makers and widely understood.

The disadvantage of the EQ-5D is that it has been shown to be insensitive to urinary diseases^28,29, and thus the modest (although significant) findings observed in this study may be expected. It is reasonable to assume that these findings are an under-estimate given that nocturia severity and its treatment have been demonstrated in previous research to be sensitive to disease-specific HRQoL measurement^20,30. This is supported by our descriptive findings suggesting that EQ-5D VAS, NI Diary, and OAB-q all are more sensitive to nocturia severity than the EQ-5D utility score.

Furthermore, whilst significant, it is unknown whether or not the EQ-5D impact demonstrated by this model meets the minimally important difference (MID) threshold required for urinary conditions, but an indicator is Walters and Brazier³¹, who estimated a mean MID of ∼0.07, from a review of 11 different studies in various diseases. As a HRQoL reference, the unadjusted NI Diary scores in Table 2 show a difference of 8.8 points for zero-to-one voids as compared to two voids, whereas Holm-Larsen et al.²⁰ have proposed a MID range for NI Diary of 5–10 points.

The use of EQ-5D as a dependent variable allows comparison with previous research by Kobelt et al.³² which focused explicitly on this issue of utilities. There were a number of limitations in the Kobelt et al. study that we set out to address in the present study, including: (1) a small sample size (n = 203 patients, n = 80 controls), (2) an analysis conducted in one single country (Sweden), and (3) the control group, which was not representative of a healthy, working population. In our study, the sample size was much greater, patients were drawn from a number of countries (DSP), and the control group was representative of a general population (HSFE). Through the multivariate regression model, including propensity score weights, we were able to control for clinical and demographic patient differences and differences in data collection (interview vs questionnaire), thus enabling the incremental utility burden of nocturia to be more fairly assessed. The findings of this study also highlighted the importance of controlling for the data collection methodology to correct for potential bias when making HRQoL comparisons across samples, even when using the same validated patient-reported outcome measures across all observations.

This study has several limitations that should be considered when making inferences about the HRQoL burden of nocturia. A recent study has shown the importance of both number of voids and time to first awakening to void (also denoted first uninterrupted sleep period—FUSP) when estimating utilities³³. Combining these two variables in a utility mapping model can cause changes of up to 0.1 in utility score. However, time to first awakening was not measured in the HSFE survey, nor was number of voids.

Specific limitations typical of observational studies include a reliance on participating physicians to select patients and their judgment in the assessment of disease characteristics, and the possibility of unmeasured or unrecognized confounding factors. With regards to the multivariate analysis, the use of regression modeling will have reduced any bias associated with the DSP data collection methodology.

The distribution of EQ-5D is problematic when fitting a model to predict its value^21,22. Whilst a robust linear model mitigates this problem, it does not provide an ideal solution, as demonstrated by the unpredictability of EQ-5D below 0.6. A robust linear regression was chosen due to the non-parametric distribution of the dependent variable, whereby outliers are highly influential leading to a bimodal distribution. It allows for regression estimates to be made with heteroskedastic residuals, and the presence of many influential outliers, as is the case with this data, with robust standard errors providing conservative hypothesis testing. This method has previously been demonstrated to be appropriate for EQ-5D-measured disease burden regression models²². For a more comprehensive discussion on the different alternative models to estimate utilities from EQ-5D, please see Hernández Alava et al.²¹.

Finally, the cause of nocturia is multifactorial and patients (mostly elderly) may have several concomitant diseases³⁴. While we in the regression model have tried to adjust for the most common and relevant ones, there may of course be diseases or relationships that were either not recorded in these surveys or were incorrectly excluded. For instance, mental health disorders were not considered as a confounding variable, although some previous studies^35,36 have found a relationship between depression and nocturia. However, the causal relationship can be questioned. A recent, comprehensive, systematic review of depression³⁷ found this disease to be a condition that is suggested to have a bidirectional association with nocturia. This finding supports our exclusion of this variable as causality cannot be proven. A limitation of the DSP is also that the survey asks for comorbidities without stating whether they are pre-existing or not. An interesting point, which requires further study, relates also to the potential impact of comorbidities on utilities. For instance, Gaujoux-Viala et al.³⁸ found that the number of comorbidities has a negative but relatively marginal impact on indirect utility scores in rheumatoid arthritis.

Conclusion

Nocturia negatively affects HRQoL. This impact is seen in patients with two voids or more per night, and the univariate analysis shows that it may increase per each additional void for NI Diary, WPAI, EQ-5D VAS, and OAB-Q. When adjusting for potential confounding variables, we demonstrated a significantly higher EQ-5D utility score of 0.0134 if patients have less than two voids compared to two or more nightly voids. This is the first demonstration of this relationship in a robust, sizable study using a general rather than disease-specific HRQoL outcome measure, which is of value to health technology assessors and payers. Every effort should, therefore, be made to reduce the number of nightly voids below the bother threshold of two voids per night.

Transparency

Declaration of funding

This study was funded by Ferring Pharmaceuticals.

Declaration of financial/other interests

FA is employed by Ferring. THL is a consultant to Ferring. KE has received research and travel grants, and is also a speaker and consultant for Ferring. PA, JP, and THL are employees of Adelphi Group Ltd. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Previous presentations

Parts of this manuscript were presented at the European Association of Urology (EAU) congress in 2014.