Methods
In the economic evaluation, three principal diagnostic pathways are compared. In the first, treatment is based on UDS and this represents the way treatment pathways are determined in current practice. In the second, treatment is based on the bladder ultrasonography test, which can be used to determine the appropriate treatment based on the measurement of BWT. In the third, treatment is based on clinical history and this represents the clinical pathway that a woman would follow if no diagnostic test was available and treatment relied on a patient’s history only. In a secondary analysis combinations of these strategies are also explored.
The primary symptoms of women with OABs are urgency and often urgency-related incontinence. These symptoms can be accompanied by another type of incontinence that occurs as a result of increased abdominal but not detrusor pressure and this is known as SUI. Where they coexist, the condition is referred to as MUI. Women who experience predominant symptoms of OAB are suspected to most likely have either DO or MUI. Sometimes, these women can have other syndromes – SUI, low compliance, VD – or, on the basis of diagnosis, are referred to as normal despite the existence of symptoms.
As far as treatment is concerned, all women are treated conservatively in the first instance. Conservative treatments typically comprise bladder training and oral treatment with antimuscarinic drugs. If these treatments are not effective, women undergo further investigation and tests before more invasive treatment is considered. The benefit of UDS compared with the other two diagnostic strategies is that it provides a more clear-cut diagnosis of the underlying syndrome. More specifically, UDS can identify (1) if the cause of the predominant symptoms of OAB is a DO, (2) if the symptoms observed are because of other syndromes (MUI, USI, low compliance, VD) or (3) if women are considered to have a normal bladder despite the existence of symptoms.
The hypothesis used in the primary study was that the bladder ultrasonography test predicts DO if BWT is above a predefined threshold. In our model we follow this using data on accuracy from the primary study and presume that bladder ultrasonography does not discriminate between MUI, USI, low compliance, VD and normal otherwise. For brevity, we will call this set of diagnoses MUI. Finally, clinical history according to whether women have only urinary urgency, with or without urgency incontinence, or had urgency-predominant MUI was used to indicate diagnoses either of OAB, most likely caused by DO, or of MUI. Despite the theoretical advantage of UDS, previous evidence has indicated – contrary to the findings reported in Chapter 6 – that an urodynamic observation rarely has an impact on the type of intervention undertaken or on outcomes.66 Therefore, a model-based economic evaluation was required to synthesise data on test accuracy and evaluate the costs and outcomes incurred by women following any particular treatment pathway based on the test result.
Model structure
In order to maintain patient history and given the short-term nature of the decision problem, the appropriate model structure to describe the options being compared and their treatment pathways is a decision tree. The model was developed in TreeAge Pro 2014 software (TreeAge Software, Inc., Williamstown, MA, USA) and the structure was informed by clinical input and NICE guidelines on the management of UI.36 Women with a mean age of 55 years (95% CIs 39 years to 71 years) enter the model if they present predominant symptoms of OAB and conservative treatments have not been effective. The term ‘predominant symptoms of OAB’ refers to symptoms of urgency or urgency-related incontinence, possibly accompanied by stress incontinence. Women who have not undergone conservative treatments and women for whom conservative treatments were effective in treating the urgency symptom but the stress symptom remained were not included in the analysis.
In the primary analysis of the model-based economic evaluation, women are assumed to follow one of three alternative strategies for the treatment of their symptoms: (1) based on UDS observation alone, (2) based on bladder ultrasonography test alone and (3) based on clinical history. These are presented in the branches to the right of the decision node (square symbol) in Appendix 9 (see ). For completeness, pathways for strategies that represent other combinations in which clinical history and diagnostic tests can in theory be used together are also considered. These are illustrated in Appendix 9 (see ).
Once women enter the model, it is assumed that they proceed from the least invasive (and more common) interventions to the most invasive ones. For the purpose of the model, it is assumed that further conservative treatment is not provided. However, in real life, conservative treatments are ongoing and can be complementary to the more invasive interventions.
As a first-line surgical intervention for the treatment of OAB caused by DO (diagnosed as DO by UDS, or a positive bladder ultrasonography test, or as urgency with or without urgency incontinence, or clinical history) women go through either BTX-A injections or PTNS depending on patient and physician preferences. If these interventions are not effective, the second-line intervention is determined through a peripheral nerve evaluation (PNE), which is the first stage of the SNS. The outcome then determines whether or not a permanent implantation of a stimulation device (stage II) will be given. If this is not indicated, women are assumed to undergo either BTX-A injections or PTNS depending on which of the two has not been applied earlier in the treatment pathway.
The treatment of SUI (USI diagnosed only by UDS) involves a sling surgery as a first-line intervention and Burch colposuspension as a second-line surgical treatment.
In cases of MUI (diagnosed as DO + USI by UDS, a negative bladder ultrasonography test, or urgency-dominant MUI clinical history) women can opt for BTX-A injections prior to the sling surgery and Burch colposuspension as a second-line surgical treatment.
Women who are diagnosed with a normal bladder, despite the existence of symptoms, can be identified by UDS only. These women are assumed to remain symptomatic without further interventions.
For the strategies bladder ultrasonography test and clinical history, when treatment either for OAB or MUI is initiated, women are assumed to receive invasive interventions as a result of a misdiagnosis. The same is assumed for women who have low compliance only or VD only despite the predominant symptoms of OAB. The treatment pathways for each syndrome in the UDS strategy are illustrated in Appendix 9 (see –). The pathways for the diagnostic strategies of bladder ultrasonography test and clinical history are shown in Appendix 9 (see –).
Model assumptions
To carry out the model-based analysis some further pragmatic assumptions were required. These are presented below.
Diagnostic tests may be repeated once in each treatment pathway before the second-line interventions and intervention strategies will change if the second test contradicts the original test.
In the UDS and bladder ultrasonography strategies, treatment is assumed to be initiated 3 months after the diagnostic test, but instantly in the case of clinical assessment. Three months was selected as an arbitrary cut-off point below the maximum of 18 weeks used in the NHS
157 to reflect the waiting time for a subsequent visit for a first-line intervention.
Symptoms of VD and low compliance that can accompany DO, MUI or SUI are assumed to not have an impact on the clinical effectiveness of interventions.
Women diagnosed with VD require self-catheterisation training.
A woman can have up to three BTX-A injections at yearly intervals.
Subjective cure from BTX-A entails the possibility of acquiring voiding difficulties that will require self-catheterisation training.
PTNS is offered in 12 sessions, a week apart, in the final 3 months. If it is successful, a monthly session is required for the rest of the model period in order for its clinical effectiveness to be maintained.
In the case of SNS, a proportion of women will require a revision of the surgery and maintenance or removal of the implanted neurostimulator device, depending on the time between SNS surgery and the end of the model period. These were assumed to take place at the end of the model period.
De novo DO or urgency UI is a possible outcome of an antistress incontinence surgery. It is described as the development of a new OAB-related symptom in women who did not demonstrate DO in their pre-operative evaluation. Given that women entering the model have predominant symptoms of OAB, de novo DO or urgency UI is not modelled.
If a woman becomes subjectively cured, improvements are assumed to be maintained throughout the model period.
Clinical data
The prevalence and accuracy data used in the model are based on the results of the primary study using UDS as a reference standard. These are presented in –. The probability of each intervention being clinically effective is presented in . Data on the effectiveness of PTNS in women with DO,82 the clinical effectiveness of sling surgery in women with SUI160 and MUI,161 as well as the clinical effectiveness of colposuspension in women with SUI164 were drawn from meta-analyses. The probability of SNS being effective in women with DO was taken from a systematic review.89 Observational studies were used to inform the probability of three repetitive BTX-A injections being effective in women with DO158 and the cure rate of colposuspension in women with MUI.163 From the former observational study, the proportion of women that become subjectively cured after three BTX-A injections (56.8%), the disaggregation of the overall effectiveness across the three injections and the drop-out rate after each injection were used in the model (). In the absence of relevant literature, the disaggregation of the overall effectiveness across the three injections and the drop-out rate after each injection were assumed to be the same if BTX-A was offered in women with other syndromes.
Prevalence data used in the model
Accuracy data used in the model for clinical history
Probabilities for the different interventions in the model
Disaggregation of the effectiveness and drop-out rate in each BTX-A injection
Accuracy data used in the model for UDS
Accuracy data used in the model for the bladder ultrasonography test
A literature search was undertaken to identify the clinical effectiveness of interventions modelled in a situation of a misdiagnosis. Apart from the clinical effectiveness of a sling surgery in women with DO identified in a published economic evaluation,90 there was limited robust evidence to inform the model about the clinical effectiveness of BTX-A in women with MUI and SUI or the clinical effectiveness of PTNS and SNS in women with MUI or SUI. For this reason, these values were elicited from the anonymous expert opinion of eight study collaborators on 18 June 2014. Expert opinion is a legitimate source of information in decision modelling when other information is not available. Given that the parameters of interest apply to very rare circumstances, it was appropriate to decide in advance to use a beta distribution and use a simple approach to elicitation, asking for most likely value as well as lowest and highest, interpreting these as mean and lower and upper limits of a 95% CI. Use of more elaborate elicitation techniques165 was judged to be unlikely to make an appreciable difference to the modelling.
Other data used in the model are presented in . The proportion of women that proceed to an implantation of a neurostimulator device (SNS stage II) after a PNE was informed by a systematic review.88 SNS often requires a revision of the surgery and this was assumed to vary across time. More specifically, 9% of women undergoing a SNS surgery are assumed to require a revision of the surgery within 2 years from the surgery and 33% after 3–5 years. These values were taken from two systematic reviews.90,166 The probability of removal of the implanted device and the probability that a maintenance surgery will be required ≥ 2 years after surgery were also taken from these systematic reviews. In addition, VD is a possible adverse event after a subjective cure from BTX-A injections and the proportion of women acquiring VD was taken from a RCT.91 Finally, the proportion of women with a diagnosis of MUI that choose BTX-A injections prior to a sling surgery and women with a diagnosis of DO (or OAB) of that choose BTX-A injections instead of PTNS were informed by the anonymous collaborators of this study.
Additional model parameters
Cost data
Information from the NHS Reference Costs 2012–1330 and Curtis92 was used to parameterise the cost component of the decision model. Costs were calculated in 2012–13 UK Great British pounds (£). These costs of different interventions modelled are presented in . Costs selected from the NHS Reference Costs have been calculated based on the weighted average value of elective inpatient and day-case costs and the proportion of patients in each group. An exception is in the case of the evaluation stage of the SNS, which is offered as a day case. Unit costs from the urology category were selected instead of the average across different medical specialties, apart from the case of bladder ultrasonography for which only a total Healthcare Resource Group (HRG) cost was available. The cost of three BTX-A injections has been calculated by multiplying the unit cost of each injection (£912, 95% CI £704 to £1060) by the proportion of women undergoing each injection. This has been calculated based on the proportions of women becoming subjectively cured and dropping out after each BTX-A injection, which combines information from to . Information about the mean cost and 95% CIs for three BTX-A injections across the different syndromes is illustrated in .
Cost data used in the model (£, 2012/13 prices)
Undiscounted cost of three BTX-A injections for each syndrome
Outcomes
To examine whether or not a diagnostic test is a cost-effective component in the treatment pathway for women with predominant symptoms of OAB, three different outcomes were considered: (1) women successfully treated, as determined by subjective symptoms, (2) DO cases detected and (3) quality-adjusted life-years (QALYs) based on the best available QoL data in the literature. The third combines quantity with QoL, which is measured using utility weights. A review of the Cost-Effectiveness Analysis Registry167 using the terms ‘overactive bladder’, ‘urinary incontinence’ and ‘detrusor overactivity’ was performed to identify relevant utility weights for the outcomes experienced at a time point beyond the primary study end point. Eighteen studies168–186 were identified and utility weights from one study168 were selected to represent the QoL for subjective cure. This study was selected because the utility values were obtained with a sound theoretical approach (time trade-off) and were relevant to the symptoms of OAB and the type of interventions modelled.169 Utility weights from other studies were not considered, either because they were focusing on a population with OAB that was undergoing conservative treatments rather than invasive170–176 or because they were focusing on women with SUI.177–182 In four more studies,183–186 the values of 0.95 and 0.73 were used to represent the utility of continent and incontinent state for a relevant population and interventions, but these were not considered appropriate because the former value in reality represents the utility weight of people with no chronic condition187 and the latter is based on women with SUI.188 In the model, women who remain symptomatic were assumed to maintain their initial QoL, which was informed by the primary study. Thus, no QoL decrements were allowed to occur after diagnostic tests or after surgeries, even when they were the outcome of a misdiagnosis. The QoL scores used in the model are shown in . QALYs were estimated combining the utility weights with estimates of the duration of different health states.
Quality-of-life data used in the model
Analysis
The decision model was constructed to investigate the cost-effectiveness of UDS compared with the diagnostic strategies, the bladder ultrasonography test and clinical history alone. Two separate economic analyses are carried out. The primary analysis provides a comparative evaluation of the costs and benefits of the UDS, bladder ultrasonography and clinical history diagnostic strategies. In a secondary analysis, all the different ways in which clinical history and a diagnostic test (UDS, bladder ultrasonography test) can be used together are explored. This allows the exploration of whether or not a diagnostic test in selective subgroups of women with predominant symptoms of OAB is a more cost-effective strategy than the three strategies explored in the primary analysis.
In both primary and secondary analyses, subjective cure is used for the identification of women successfully treated and UDS has been used as a reference standard. In the UDS arm, women with ‘normal’ or ‘other’ observations are assumed to remain symptomatic throughout the model period. However, in the case of bladder ultrasonography and clinical history, these women receive treatment for OAB or MUI as a result of a misdiagnosis. In such a situation, women were assumed to have a probability of becoming asymptomatic mainly as a result of a placebo effect. In the absence of robust evidence to inform this probability, the smallest figure of (0.143, 95% CI 0.133 to 0.375) was selected.
The analyses were carried out from the perspective of the UK NHS and the primary outcome is in terms of cost per woman successfully treated. Other outcomes included in the analysis are cost per DO case detected and cost per QALY. The latter is the recommended outcome for economic evaluations in the UK.189 Results presented in terms of QALYs represent an additional analysis and were not part of the principal objective of the study. Results are presented in terms of incremental cost-effectiveness ratios (ICERs) and cost-effectiveness acceptability frontiers (CEAFs). A 5-year time horizon was considered appropriate to reflect all key differences, in terms of costs and benefits, for the options compared. Costs and QALYs accruing beyond 12 months were discounted at a rate of 3.5% per year. According to NICE,36 based on existing evidence,66 UDS should not be applied prior to the initiation of conservative treatments but only when these are not effective and women are expected to proceed to more invasive interventions. For this reason, the decision model is focusing on the subgroup population of the primary study that had been taking conservative treatment before enrolling into the study and had complete accuracy data on the three strategies modelled (n = 209).
Deterministic and probabilistic sensitivity analyses (PSAs) are performed to explore the effects of the inherent uncertainty in parameter estimates on model results. In deterministic sensitivity analysis, one or more parameters are varied while keeping the remaining at their baseline value. Although deterministic sensitivity analyses can be helpful to identify which model inputs are important in driving a decision or identify threshold values, comprehensive representation can be obtained by undertaking a PSA, in which the uncertainty around a parameter is represented with a probability distribution.190,191 In the PSA, using 10,000 repeated random draws from those distributions, a Monte Carlo simulation of the model provides an indication of how variation in the model parameters leads to a variation in the results generated. Beta and Dirichlet distributions were used for binomial and multinomial data, respectively (see – and ) and a Gamma distribution for costs (see ). The parameters of each distribution are shown in the corresponding tables.
Deterministic sensitivity analyses
A number of deterministic sensitivity analyses were conducted in both the primary and secondary analyses. These deterministic analyses include both univariate and multivariate analyses to assess the impact of any uncertainty in model parameters on the final results. Three univariate analyses and four multivariate analyses were conducted based on the following justifications:
Univariate analyses
Reducing the cost of UDS from £401 (95% CI £216 to £462) to £173. The base-case value was calculated based on the weighted average value of elective inpatient and day-care costs for a urological intervention and the proportion of patients in each group. The value used in this sensitivity analysis represents the weighted average value of total HRGs.
30Increasing the cost of sling surgery by 50% to account for possible adverse events such as bladder injury, vaginal erosion or groin pain.
160Lowering the utility weight of women subjectively cured from 0.92 to 0.84. This value is drawn from the BUS and represents the EQ-5D QoL score of those women that scored below four in the ICIQ (low bother).
Multivariate analyses
Changing the accuracy data. The main analyses use UDS as a reference standard, which possibly leads to an overestimation of its diagnostic accuracy. In addition, urinary diaries can complement clinical history and provide a more accurate clinical assessment. The accuracy of UDS was reduced using evidence from an economic evaluation,
64 which reports 14% probability of false-positive diagnoses for MUI when DO or SUI are the true conditions and 25% probability of false-positive diagnoses for SUI when MUI is the true condition. Similarly, a systematic review
150 has reported that clinical history and information from diaries detect 53.4% of women with DO, increasing the accuracy of clinical history by five percentage points.
Changing the rates of clinical effectiveness elicited from expert opinion to their lowest value.
Changing the rates of clinical effectiveness elicited from expert opinion to their highest value.
Assuming one diagnostic test is offered prior to the initiation of the first-line interventions in line with NICE guidelines.
36
Results
Primary analysis
The costs and outcomes of each strategy in the primary analysis are shown in . In terms of cost, UDS is the least expensive strategy (£4524) with clinical history and bladder ultrasonography costing an additional £1278 (£5801) and £1424 (£5947), respectively. In terms of outcomes, treatment based on clinical history appeared to be the most effective strategy apart from the number of DO cases detected, for which UDS was more effective (as DO is an urodynamic observation rather than a clinical symptom, a DO diagnosis is never made based on clinical history).
Deterministic analyses results for primary and secondary analysis across the three outcome measures
It is important to note that the total cost per patient (from £4524 to £5947) greatly exceeds the cost of the initial tests (£401 for UDS or £51 for bladder ultrasonography) and thus the major determinants of the costs are the treatment selections made based on the results of the tests.
In the deterministic analysis, the bladder ultrasonography test strategy was dominated by the other two strategies as it was the most expensive and least effective strategy across all outcomes explored in the analysis. In the remainder of this subsection, the results of the incremental analysis are described with more information provided about the cost for an additional unit of outcome between the strategies compared. Dominated strategies, such as bladder ultrasonography, are not described as a part of the incremental analysis.
Results of primary analysis for the outcome of women successfully treated
As shown in , the diagnostic strategy of clinical history leads to an additional 26 cases per 10,000 women successfully treated compared with UDS at an additional cost of £1278 per woman. This results in an ICER of £491,100 per woman successfully treated, which means that an additional £491,100 is required for each additional woman to be successfully treated as a result of a diagnosis by clinical history compared with a diagnosis using UDS. The scatterplot in Appendix 10 (see ) shows the modelled uncertainty in the cost and clinical effectiveness between UDS and clinical history from 10,000 Monte Carlo simulations. The result of each simulation is plotted on the cost-effectiveness plane providing information about the joint density of the differences in cost and clinical effectiveness between the two strategies. It is evident that even though UDS is almost certainly a less costly diagnostic strategy, it is uncertain whether or not it is more effective than clinical history. It is thus uncertain whether or not the strategy based on UDS dominates clinical history or whether it provides savings for one woman less successfully treated.
Deterministic results for non-dominated strategies of the primary analysis for women successfully treated
The results of the PSA based on Monte Carlo simulation can also be used to illustrate, with a CEAF, the probability that the optimal strategy, in terms of maximising the net benefit, is cost-effective under current uncertainty at different levels of decision-makers’ willingness to pay (WTP) per additional woman successfully treated. As illustrated in the CEAF for the comparison between UDS and clinical history (see Appendix 10, ) for any value of WTP per woman successfully treated below £100,000, the probability that UDS is cost-effective exceeds 61%.
Given that there is not a pre-specified threshold of WTP for an additional woman successfully treated, as in the case of QALYs for which £20,000 to £30,000 are the recommended cut-off points by NICE,192 the identification of the probability of UDS being cost-effective is less straightforward and subject to uncertainty owing to the number of assumptions required. Evidence suggests that the QoL of a woman successfully treated is approximately 0.92173 and that improvements could be maintained for 5 years,161,166,193 which results in 4.6 QALYs. If a woman remained symptomatic for these years then she would have 3.2 QALYs. This is a multiplication of the weighted average of the initial QoL of each syndrome (see ) and the number of women with each syndrome (see ), which leads to a value of 0.69, with the number of years. The QALY gain from becoming subjectively cured is thus 1.4, and if the WTP for a QALY is £20,000, £28,000 for a woman subjectively cured could be an acceptable WTP threshold. At this value, there is a probability of 86% that UDS is the optimal strategy.
Results of primary analysis for the outcome of quality-adjusted life-years
In terms of QALYs, the diagnostic strategy of clinical history results in an additional 0.0212 QALYs gained per woman compared with UDS. Given the additional cost of £1278 per woman, the mean ICER for clinical history compared with UDS was estimated at £60,200 per QALY (). The results of the PSA show that UDS is likely to be the most cost-effective diagnostic strategy for the commonly used £20,000–30,000 threshold of WTP for a QALY192 with 72% and 61% probability, respectively. If decision-makers are willing to invest more than £60,200 for an additional QALY, clinical history becomes the optimal strategy with a probability ranging from 52% to 57% as the WTP per QALY increases from £60,200 to £100,000. The incremental cost-effectiveness scatterplot and the CEAF are shown in Appendix 10 (see and ).
Deterministic results for non-dominated strategies of the primary analysis for QALYs
Results of primary analysis for the outcome of detrusor overactivity cases detected
As evident in , UDS dominates clinical history as it is less expensive and detects a greater number of cases of DO. However, the outcome of DO cases detected should be used with caution given that DO constitutes a UDS observation rather than a clinical symptom, which consequently favours UDS. The dominance of UDS is also depicted in a scatterplot and CEAF (see Appendix 10, and ).
Secondary analysis
In this analysis, apart from the main strategies explored in the primary analysis, four further strategies that use a diagnostic test as an adjunct to the clinical history have been included. In these strategies, a diagnostic test (UDS, bladder ultrasonography) is conducted when clinical history has indicated either OAB or MUI. This analysis aims to provide a further insight into the cost-effectiveness of performing a diagnostic test only in selective subgroups of the population modelled. When UDS is used as an adjunct to the clinical history if OAB or MUI has been indicated from patient history, the strategies have been termed as clinical history and UDS in OAB and clinical history and UDS in MUI, respectively. Similarly, when the bladder ultrasonography test complements clinical history, the terms clinical history and BUS in OAB as well as clinical history and BUS in MUI are used.
The mean estimated costs and outcomes for each of the seven strategies compared in the secondary analysis are presented in . In terms of cost, UDS is again the least expensive strategy with a mean cost of £4524. In terms of outcomes, conducting a diagnostic test when clinical history suggests MUI appears to be more effective in all three outcomes considered in the analysis. More specifically, performing bladder ultrasonography if the patient’s history indicates MUI is the most effective strategy when the outcome is measured in women successfully treated, while conducting UDS if MUI is indicated by the clinical history is the most effective strategy when the outcome is measured in terms of QALYs and DO cases detected.
Results of secondary analysis for the outcome of women successfully treated
Appendix 10, , illustrates the mean cost and clinical effectiveness in terms of women successfully treated for all strategies included in the secondary analysis. The line connecting the estimates for UDS, to that for the strategy of clinical history and UDS in MUI, then to that for clinical history and BUS in MUI, creates a cost-effectiveness frontier. These strategies, as indicated in , have been dominated. The deterministic results of secondary analysis for the non-dominated strategies are presented in . The results show that performing UDS when clinical history indicates MUI means there will be an additional 309 successfully treated women per 10,000 and an additional cost of £603 per woman. This leads to an ICER of £19,500 per woman successfully treated. Performing a bladder ultrasonography test when clinical history indicates MUI requires an additional £641 per woman and leads to an additional 82 cases per 10,000 women successfully treated compared with UDS, which gives an ICER of £78,600 for an additional woman successfully treated.
Deterministic results for non-dominated strategies of the secondary analysis for women successfully treated
The scatterplot in Appendix 10, , depicts the overall uncertainty in the cost and clinical effectiveness (women successfully treated) when 10,000 random draws from parameter distributions in a Monte Carlo simulation process are used instead of mean estimates. Given the degree of overlap in the results obtained from the different strategies compared, it is evident that there is uncertainty as to which diagnostic strategy is optimal. This uncertainty has been graphed in the CEAF in Appendix 10, . As shown in , if the WTP for a woman successfully treated is below £19,500, UDS is the optimal strategy under current uncertainty with a probability of being cost-effective above 52%. Above this WTP threshold, clinical history with UDS in MUI becomes the optimal strategy. At WTP values between £19,500 and £78,600 per woman successfully treated, there is a probability of 37–48% that clinical history and UDS in MUI is the strategy that maximises the net benefit. If a decision-maker is willing to invest more than £78,600 per woman successfully treated, clinical history with bladder ultrasonography in MUI becomes the optimal choice. At the WTP threshold of £28,000 per woman successfully treated, which, according to the corresponding section of the primary analysis, was considered to be an acceptable WTP threshold for this outcome, there is a probability of almost 50% that clinical history with UDS in MUI is the optimal strategy.
Results of secondary analysis for the outcome quality-adjusted life-years
In terms of QALYs, the only non-dominated strategies are the UDS and the UDS as an adjunct to the clinical history in women with patient history of MUI (see Appendix 10, ). As shown in , the latter strategy leads to an additional 0.476 QALYs gained per woman at an additional cost of £603. This means that the strategy of performing UDS in women with a clinical history of MUI requires an additional £12,700 for an additional QALY compared with UDS, which is considered to be cost-effective. The results of the PSA indicate that at commonly cited ceiling ratios between £20,000 and £30,000 per QALY, clinical history and UDS in MUI is likely to be the optimal strategy with a probability of being cost-effective at 72% and 78%, respectively. The graphs representing the uncertainty in the estimated costs and QALYs and the uncertainty around the optimal strategy across a range of possible values of WTP for an additional QALY are shown in Appendix 10 (see and ).
Deterministic results for non-dominated strategies of the secondary analysis for QALYs
Results of secondary analysis for the outcome of detrusor overactivity cases detected
According to the evidence of the deterministic analysis presented in , having UDS as an adjunct to the clinical history in women with a patient history of MUI is £603 (£5126) more expensive and more effective, leading to an ICER of £62,100 for an additional DO case detected compared with UDS. According to the results of the PSA, if decision-makers’ WTP for a DO case detected is below £62,100, UDS is the optimal strategy. As the WTP for this outcome increases from £62,100 to £100,000, the strategy of clinical history and UDS in women with a clinical history of MUI becomes more cost-effective with a probability ranging from almost 50% to 75%. As in the outcome of women successfully treated, the identification of the optimal strategy and the probability of this strategy to be cost-effective for an additional DO case detected is not straightforward as there is not a predefined threshold of WTP. However, it was shown that the probability of a DO case to become subjectively cured is approximately 60%. In this situation, an increase in QoL from 0.6, which is the QoL of a woman with DO (see ), to 0.92 would be expected to occur. That means that the weighted average of the QoL for a DO case detected is 0.79. A DO case detected could thus lead to a 0.95 QALY gain. As a consequence, a plausible WTP threshold for a DO case detected could be £19,000. For this value, there is a 99% probability that UDS is the optimal choice. The cost-effectiveness plane, the scatterplot and the CEAF for this outcome are shown in Appendix 10 (see –).
Deterministic results for non-dominated strategies of the secondary analysis for DO cases detected
Sensitivity analyses for the primary and secondary analysis: univariate sensitivity analyses
Reducing the cost of urodynamics
The cost of UDS used in main analyses (£401) represents the weighted average value between elective inpatient and day-case costs and the proportion of patients in each of the two groups for the urology service description. In this sensitivity analysis, the average weighted value across all HRGs (£173) is used.30 Lowering the value of UDS would lead to an increase in the ICER of any strategy compared with this diagnostic test. As evident in , the ICER of all strategies increased in all three outcomes compared with the deterministic results of the main analysis without any impact on the conclusions drawn.
Results of univariate sensitivity analyses for primary and secondary analysis
Increasing the cost of sling surgery by 50%
In this sensitivity analysis, the cost of sling surgery was increased from £3917 to £5876 with the purpose of capturing other possible adverse events, such as bladder injury, vaginal erosion or groin pain,160 that were not modelled. An increase of the cost of sling surgery led to an increase of the ICER of clinical history because of the larger number of women with DO (or OAB) undergoing surgeries for MUI or SUI as a result of a misdiagnosis. Having UDS as an adjunct to the clinical history when MUI is indicated by the patient does, in theory, mean fewer women with DO (or OAB) undergoing sling surgery owing to the larger sample size on which diagnostic accuracy is applied. This explains the decrease of the ICER of clinical history and UDS in MUI strategy. The results obtained from the main analyses remained unchanged.
Lowering the utility weight of women subjectively cured
This sensitivity analysis explores the impact that a lower QoL score for women becoming subjectively cured would have on model results. Given that more women become subjectively cured from any strategy containing clinical history than with UDS and bladder ultrasonography test alone, lowering the utility weight would result in an increase of the ICER for the different strategies of clinical history. As shown in , lowering the utility weight of women becoming subjectively cured from 0.92 to 0.84 makes clinical history dominated by UDS in the primary analysis and leads to an increase in the ICER of clinical history and UDS in MUI from £12,700 to £22,500 in secondary analysis. Again, the conclusions drawn from the main analyses remained unchanged.
Sensitivity analyses for the primary and secondary analysis: multivariate sensitivity analyses
Changing the accuracy data
In the main analyses, UDS was used as a reference standard. This possibly led to an overestimation of its diagnostic accuracy and subsequently an overestimation of the results obtained from the economic evaluation. In addition, clinical history can be complemented by urinary diaries (clinical assessment) and possibly provide a more accurate diagnosis of the underlying syndrome. For this reason, this sensitivity analysis explores the impact on the results generated from the main analysis if lower accuracy is assumed for UDS and higher for clinical history. A 14% probability of false-positive diagnoses for MUI when DO or SUI are the true conditions and 25% probability of false-positive diagnoses for SUI when MUI is the true condition were assumed for UDS.64 For clinical history, a 53.4% accuracy of detecting DO from OAB predominant symptoms was assumed.150 As seen in , this has significantly improved the ICERs but without a major impact on the conclusions from the primary analysis. The ICER of clinical history from the primary analysis was reduced from £60,204 to £31,000, which is slightly over the upper WTP threshold for a QALY. The only change in the decision was in the case of DO cases detected, in which the clinical history and UDS in MUI strategy became cost-effective, as opposed to dominated, considering the ICER found from the sensitivity analysis (£14,100) is below £19,000, which was considered a plausible value of WTP for a DO case detected. With this change, clinical history and UDS in MUI is the optimal choice across the three outcomes used in the analysis.
Results of multivariate sensitivity analyses for primary and secondary analysis
Using the lowest rates of clinical effectiveness elicited from expert opinion
In this sensitivity analysis, the lowest rates of clinical effectiveness elicited from expert opinion are used to examine the impact of lower effectiveness in cases of a misdiagnosis on the model results (see Clinical data). For this analysis, the clinical effectiveness of BTX-A, when the true conditions are MUI and SUI, were decreased from 0.329 and 0.143 to 0.200 and 0.133, respectively. The clinical effectiveness of PTNS and SNS in women without DO was also reduced from 0.314 and 0.271 to 0.283 and 0.233, respectively. Finally, the clinical effectiveness of colposuspension in women with DO was reduced from 0.163 to 0.153. Reducing the effectiveness of interventions in situations of a misdiagnosis was expected to lead to an increase in the ICER of the strategies compared with UDS. This is evident in the results of . The results obtained from the sensitivity analysis did not impact on the conclusions drawn from the main analyses.
Using the highest rates of clinical effectiveness elicited from expert opinion
In this sensitivity analysis, effectiveness rates elicited from expert opinion were taken to their highest value (see Clinical data). More specifically, the effectiveness of BTX-A, when the true conditions are MUI and SUI, were increased to 0.483 and 0.375, respectively. For the effectiveness of PTNS and SNS in women without DO the values 0.567 and 0.555 were used. Finally, the effectiveness of colposuspension in women with DO was increased to 0.386. According to the results of Chapter 5, Table 16, this sensitivity analysis affected the initial decision in one outcome of the base-case and secondary analysis. More specifically, in the base-case analysis, clinical history became cost-effective in terms of QALYs with an ICER of £21,000 compared with UDS. In the secondary analysis, clinical history and bladder ultrasonography in MUI became cost-effective with an ICER of £14,000 for a woman successfully treated compared with UDS.
Assuming one diagnostic test is given in the treatment pathway
This sensitivity analysis explores the impact on the results of the main analyses if one diagnostic test was performed in the treatment pathway prior to the initiation of more invasive interventions, as indicated in the 2006 NICE report.36 As shown in , the ICER of all strategies compared with UDS was increased without any impact on the conclusions from the main analyses. This finding implies that at the level of accuracy data used in the model, having one UDS is more cost-effective.
Discussion
Principal findings
The results of the deterministic analysis suggest that a patient management strategy based on using bladder ultrasonography to select treatment options is more costly and less effective than the strategies based on UDS and clinical history. Given that the results of the main study show that bladder ultrasonography is of limited diagnostic value, the uncertainty around the deterministic results of this strategy was not explored in a PSA.
With regard to the other two strategies, treating according to the primary symptoms in clinical history is more effective than UDS, leading to an additional 26 cases per 10,000 women successfully treated and 0.02 QALYs gained per woman. This comes at an additional cost of approximately £1300 per woman arising from the costs of the treatments used. Given current acceptable thresholds, this additional cost would not be justified by the clinical gain. In terms of DO cases detected, UDS detects more cases than clinical history at a lower cost. Thus, the results of the primary analysis suggest that UDS would be the preferred strategy on cost-effectiveness grounds for women with predominant symptoms of OAB. This result holds for all three outcome measures used in the primary analysis. This finding has arisen because, although the UDS test itself adds to the cost of clinical history, basing treatment decisions on the UDS leads to more efficient use of expensive interventions than basing decisions on clinical history alone.
In the secondary analysis, the strategy of giving UDS to women with a clinical history of MUI is more effective than universal UDS. The selective use was shown to result in an additional 309 cases per 10,000 women successfully treated and the QALY gain per woman was 0.05. The additional cost per woman was £600. This strategy would be considered value for money according to acceptable thresholds in the UK. Based on cost per DO case detected, UDS was not deemed cost-effective for these women given the ICER of £62,000. This is not a surprising finding as DO is an urodynamic observation and not a clinical symptom.
Univariate and multivariate sensitivity analyses were performed to explore the robustness of the model results to changes in the assumptions made. Conclusions drawn from the main analysis remained robust to all sensitivity analyses apart from the scenario in which the highest effectiveness rates were assumed in cases of misdiagnosis.
Strengths and limitations of the study
The prevalence and accuracy data were estimated in the primary study and the main clinical pathways were parameterised with clinical effectiveness data mostly from systematic reviews and meta-analyses. In addition, costs were drawn from national sources and the development of the model relied on national guidelines. These are likely to enhance the generalisability of the study’s findings. Finally, all assumptions used in the model were agreed a priori and key assumptions have been tested in sensitivity analyses.
However, there are a few weaknesses. Instead of clinical history, the best alternative strategy to compare the cost-effectiveness of UDS with could have been clinical assessment. Urinary diaries, for example, are used to complement clinical history and provide a joint decision based on clinicians’ assessment rather than on patient’s history alone. Furthermore, ambulatory UDS has been shown to be more sensitive in detecting DO,194 or generally the underlying syndrome of UI,195,196 than conventional UDS and thus the results of the economic evaluation may overstate the cost-effectiveness of UDS, as it has been used as a reference standard. Ambulatory UDS was used as part of the primary study in situations of a negative UDS, but in the subgroup population used for the model, the number of observations were not enough to be tested in a sensitivity analysis (n = 7).
A further limitation relates to the QoL estimates used for the outcomes experienced beyond the primary study end point. In the absence of robust utility weights, it was assumed that women maintain their QoL if an invasive intervention is not successful and that all women have a utility score of 0.92 once they become subjectively cured. In reality, however, utility decrements can occur after invasive and poorly tolerated diagnostic strategies, such as UDS or surgeries. In addition, utility gains tend to differ between different syndromes and severity of symptoms. Despite the low accuracy of the QALY estimates used in the model, there was a significant consistency between the results obtained from the primary outcome (cost per woman successfully treated) and the results obtained using QALYs.
Furthermore, in the absence of available literature on the clinical effectiveness of the different interventions modelled in situations of misdiagnosis, values had to be elicited from expert opinion (see Clinical data). Probabilistic and deterministic sensitivity analyses were undertaken to reduce the uncertainty around these parameter estimates and explore their impact on conclusions drawn. Another limitation relates to the fact that the order of the interventions modelled for the treatment of women with predominant symptoms of OAB might be slightly different in practice. For example, after an unsuccessful sling surgery, instead of going through a Burch colposuspension, women may repeat the sling surgery. However, such differences are not expected to significantly impact on the results of the model. Finally, the syndromes of OAB, SUI and MUI are often associated with substantial personal costs. These costs are not included in economic analyses conducted from a health-care perspective and thus economic evaluations from a societal perspective could shed further light onto the cost-effectiveness of the alternative diagnostic strategies for women with predominant symptoms of OAB.
Strengths and limitations in relation to other studies
Economic evaluations that have explored the role of UDS in the treatment of women with predominant symptoms of SUI65 or genuine SUI64,89 have concluded that UDS is not a cost-effective diagnostic test. However, for women with predominant symptoms of OAB, relevant economic evidence on the role of UDS is lacking, while clinical evidence remains inconclusive.17,29,66
The results of the economic evaluation of this chapter suggest that UDS has a role to play in the treatment pathway of women with OAB. An investigation into whether or not a diagnostic test is more cost-effective when performed in only specific subgroups of women concluded that the most cost-effective diagnostic strategy is to perform UDS in women with a clinical history of mixed incontinence. This finding seems to be further supported by evidence from clinical studies indicating that in women with OAB only an urodynamic observation of DO does not have an impact on the outcome of more invasive interventions,84,86,197 but also from evidence indicating that mixed urinary symptoms tend to be more common than mixed UI on UDS.13,198 The magnitude of DO appears to be close to 50%,198 as it was also found in the primary study (55%).
Meaning of the study
The results of the economic evaluation showed that UDS is a cost-effective diagnostic strategy for women with predominant symptoms of OAB for whom conservative treatments were not effective. The primary analysis suggested that UDS is the most cost-effective strategy across all outcomes providing significant cost-savings for a small reduction in outcome. The secondary analysis suggested that selective UDS, restricting its use to women with a clinical history of MUI, is a cost-effective strategy.
Unanswered questions and future research
In this study, UDS was used as the reference standard, which may have led to an overestimation of its diagnostic accuracy. Evidence suggests that ambulatory UDS is more sensitive in detecting DO194 and other syndromes of UI.195,196 Furthermore, a more appropriate comparator for UDS would have been the clinical assessment, which would provide an overall clinical diagnosis based on the clinical history (validated questionnaires) and urinary diaries. Thus, a comparative evaluation of the costs and benefits of UDS and clinical assessment using ambulatory UDS as a reference standard would help gain further insight into the cost-effectiveness of UDS in women with predominant symptoms of OAB.
Furthermore, there are significant inconsistencies in the instruments used in the literature to capture QoL in women with urinary symptoms. Disease-specific instruments, such as the incontinence-specific QoL questionnaire and other non-preference-based measures that are often used in women with OAB symptoms, offer limited usefulness as outcome measures in cost-effectiveness analyses;199 although, mapping algorithms have now started to be developed.200 Nevertheless, for conditions such as OAB, in which symptoms can also be associated with non-health impacts (e.g. embarrassment), instruments that go beyond health and capture wider changes in individual’s QoL may be deemed more pertinent. An example is the ICECAP-A measure,201 which incorporates capability attributes on Attachment, Security, Role, Enjoyment and Control. Further research is required in the use of such instruments in women with urinary symptoms.
Finally, it appears that further research is needed to resolve current decision uncertainty around the cost-effectiveness of UDS in all women with predominant symptoms of OAB or in only those with a clinical history of MUI. A measure of the maximum possible value of any such research is the expected value of perfect information (EVPI). EVPI is estimated from the difference between the expected net benefit with perfect and current information, and multiplied by the size of the population that could benefit from additional information. Based on estimates of prevalence of women with predominant symptoms of OAB in the UK202–204 and an incidence rate of 6% (54,000 women),205 while assuming a time horizon of 10 years and an annual discount rate of 3.5%, the EVPI was estimated at almost £36M and £77M at a WTP threshold of £28,000 per additional woman successfully treated for the primary and secondary analyses, respectively (see Appendix 10, and ). These figures reflect the large expected opportunity loss associated with current decision uncertainty and the need for further primary research.
