Family Practice Advance Access originally published online on September 12, 2006
Family Practice 2006 23(6):666-673; doi:10.1093/fampra/cml028
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
A decision aid for GPs for the treatment of elderly male patients with lower urinary tract symptoms (LUTS)
a Department of General Practice, Care and Public Health Research Institute, Universiteit Maastricht Maastricht, The Netherlands
b Department of Urology, University Hospital Maastricht Maastricht, The Netherlands
Correspondence to Roelf JC Norg, Department of General Practice, Care and Public Health Research Institute, Universiteit Maastricht, Peter Debyeplein 1, PO Box 616, 6200 MD Maastricht, The Netherlands. Email: roelf.norg{at}hag.unimaas.nl
Received 22 November 2005; Accepted 23 May 2006.
 |
Abstract |
|---|
 Top Abstract IntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Background. GPs have four main treatment options for lower urinary tract symptoms (LUTS): watchful waiting,
-blockers, 5--reductase inhibitors or (referral for) surgery. Guidelines do not provide clear cut-off values for (combinations of) symptoms and physical examination results to decide which treatment is best. Objective. (i) To develop a decision aid (‘checklist’) for GPs for the treatment of patients with LUTS. (ii) To assess its value for use in a primary care population.
Materials and methods. Population-based cross-sectional study. Included were subjects with uncomplicated LUTS for whom treatment in primary care may be appropriate. [International Prostate Symptom Score (IPSS) 
8, no prior prostate surgery, prostate-specific antigen (PSA) value <10 ng/ml]. For each subject the appropriateness of surgery and -blocker treatment was determined using a previously validated formalized international expert panel judgement. Regression models using data available in primary care were constructed to predict the panel judgement. Subsequently these models were transformed into simple checklists. Finally, the efficiency of these checklists was calculated.
Results. The best checklists consisted of age, symptoms severity, type of symptoms, a quality of life score and PSA value. Assuming one would like to provide at least 95% of the subjects for whom a certain treatment is appropriate with this treatment (i.e. ‘sensitivity’ of the checklist 95%), one can reach a positive predictive value of 50–60%.
Conclusion. Simple checklists based on the judgement of experts regarding the most appropriate therapy can help GPs to advise their patients of a rational treatment strategy.
Keywords. Benign prostatic hyperplasia, diagnosis, lower urinary tract symptoms, therapy.
|
Introduction |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Lower Urinary Tract Symptoms (LUTS) are becoming an increasing health problem in the elderly male. Ageing and more awareness and recognition among both patients and doctors have led to an increase in the demand and costs for diagnosis and treatment of LUTS.1,2 Almost half of the new patients can be treated in primary care.3 New patients tend to be younger and present themselves with less severe symptoms than before.4 Therefore, the involvement of GPs is expected to grow in the years to come.
Primary care guidelines for LUTS often start off with excluding conditions that can be easily diagnosed in primary care [e.g. urinary tract infections (UTI) or acute urinary retention (AUR)], or should be referred to secondary care for further diagnosis because of the seriousness of the disease (e.g. prostate cancer). The rest of the patients are regarded to have LUTS ‘suggestive’ for benign prostatic hyperplasia, also referred to as uncomplicated LUTS. These patients may be treated with watchful waiting (WW), -blockers (AB), 5--reductase inhibitors (5ARI) or surgery. Current guidelines offer little help to primary care physicians in deciding which of these treatments to choose. They leave much room for interpretation and shared decision making by the GP and the patient.2,3,5
In the process of shared decision making, the GP would be helped if he could fall back on a ‘reasonable’ treatment strategy. Such a strategy would be to anticipate the expected treatment strategy of the specialist. This means that the GP asks two questions. First, Should I refer? and the second, Should I myself treat? In case of LUTS these questions are specified to the following: Would surgery (probably) be appropriate? and Should I prescribe an -blocker? or, alternatively, Should I maintain a WW strategy? (5ARIs have limited value in primary care.5) The order of these questions may vary. A GP may first determine whether or not to refer and subsequently determine the appropriateness of -blocker treatment; or he may also first prescribe an -blocker if appropriate, and (if it should fail) refer for surgery if indicated. Some GPs may prefer a specific strategy; others may leave the choice to the patient, who may either have a preference for a quick and definitive solution if indicated, or may want a less invasive approach if possible. Both treatment strategies therefore are defendable, depending on the circumstances.
The question whether surgery or an -blocker would be appropriate was answered in an international study that used a formalized expert opinion procedure.6,7 In this study, 39 urologists considered vignettes of men aged 50 years, referred by their GP for the first time with—according to the first general diagnostic evaluation—uncomplicated LUTS, and no influencing co-morbidity or medication. Each case had a different combination of a total of eight diagnostic variables. For each case the experts judged a treatment as ‘appropriate’, ‘inappropriate’ or ‘uncertain’ compared to WW.
Such a study revealing the general opinion of an international expert panel could be helpful to GPs to determine their treatment strategy. However, application in the primary care setting is problematic. Usually only five of these diagnostic variables are available in primary care (i.e. age, International Prostate Symptom Score, type of symptoms, bother and history of UTIs); two require specialized diagnostic procedures that are usually not available in primary care [maximum urinary flow (Qmax) and post-void residual volume (PVR)]; one is measured differently in primary care (prostate size via digital rectal examination versus ultrasound). Furthermore, the case-mix in primary care differs from that in secondary care. Most patients in primary care have only mild to moderate LUTS. Consequently, a modification and validation of this expert opinion for the primary care setting is needed, preferably into a simple checklist with (weighted) key symptoms and investigations, and clear cut-off values determining the best treatment option available.
We developed such a checklist in a population-based sample of elderly males with LUTS. In this article we report on its development, and provide an estimation of its value for the primary care situation.
|
Methods and materials |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Study population
Our source population consisted of all men aged 55 or older registered with 14 general practices. They were invited by mail to participate in a cohort study. During a qualification period the participants repeatedly filled in questionnaires [IPSS, Danish Prostate Symptom Score (Dan-PSS), micturition behaviour], underwent urinary flow measurement and physical examination. Also, prostate-specific antigen (PSA) and creatinin were measured, and the subjects kept a 3 day micturition diary and underwent ultrasound measurement of PVR and prostate size. Included in this analysis were subjects with untreated uncomplicated LUTS, i.e. an IPSS
8 points, no prior prostate surgery or current use of medication for LUTS, and a PSA value <10 ng/ml. Consulting for LUTS was not an inclusion criterion.
Analysis
We used the combined opinion of the expert panel as published in the studies of Stoevelaar6 and McDonnell7 to determine the appropriateness of surgery and -blockers for each member of our study sample. Subsequently, we constructed logistic regression models that predicted whether or not a certain treatment would be appropriate or not (i.e. ‘appropriate’ versus ‘inappropriate’ or ‘uncertain’). We used a backward stepwise approach with –2 log likelihood (P < 0.05) as criterion. Only data available in primary care were used in the primary analyses, although we also constructed similar models including Qmax to investigate the added value of urinary flow measurement. We checked for collinearity between the included variables. We also compared the receiver operating characteristic (ROC) curves of the models, and calculated the area under these curves (AUC). ROC curves indicate the sensitivity and the corresponding specificity at various cut-off points of a test (i.e. our regression models and our checklist). The AUC indicates the diagnostic accuracy of the checklists. An AUC of 0.50 indicates that the checklist does not discriminate between appropriate and inappropriate treatment at all. An AUC of 1.0 would imply perfect discriminative power. In a second step, the final models were transformed into simpler checklists, preferably based only on simple dichotomous items.
During the analysis we followed the two possible strategies described in the Introduction. According to strategy 1 (referral first strategy) we first constructed regression models to predict the appropriateness of surgery. Subsequently, for those for whom surgery (and therefore referral) was not indicated, we constructed regression models to predict the appropriateness of -blocker treatment.
In strategy 2 the sequence was reversed (medication first strategy). First we constructed regression models to select those for whom an -blocker would be appropriate. Within this group we constructed models to predict whether surgery would be appropriate in case of failure of -blocker treatment.
Having translated the regression models into simpler checklists we—as a preliminary investigation—calculated the sensitivity, specificity and positive predictive value. In this specific setting the sensitivity represents the percentage of subjects for whom a referral (or prescription of medication) would be appropriate, and who—using the checklist—would indeed be referred (or be prescribed medication). The specificity represents the percentage of subjects that would not receive a treatment not considered appropriate. The positive predictive value represents the percentage of referred subjects for whom surgery indeed would be appropriate.
Our aim was to create checklists that would have a high sensitivity, whereas the specificity and positive predictive value of the models were less important. Indeed, a GP must select and refer those for whom surgery is appropriate, accepting that a certain percentage of those referred will not be operated at all. This is not problematic, because the urologist makes a second selection using more extensive diagnostic tools (ultrasound, urinary flow measurement, urodynamics). This second selection will guarantee that no unnecessary operations are executed. The GP must however not be too selective. A too selective GP might withhold patients for whom surgery would be appropriate possibly beneficiary therapy. A similar reasoning holds for -blocker treatment. Current guidelines state that—when in doubt—a trial of treatment is appropriate. Consequently, the GP should aim to select those for whom -blocker treatment will be useful, accepting that some of those for whom these are less appropriate will also receive such a prescription.
Informed consent
According to Dutch law the study was approved by the Ethics Committee of the Academic Hospital Maastricht/Maastricht University. All participants gave written informed consent to participate.
|
Results |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Included in the analysis were 512 subjects. Table 1 shows the baseline characteristics and their association with the appropriateness of surgery and
-blocker treatment. A total of 432 subjects (86%) had moderate symptoms. The median age was 66 years. -Blockers would be appropriate in 86 subjects (17%); surgery in 39 subjects (8%). There was much overlap between these groups: for 37 subjects both surgery and -blockers would be considered appropriate.
|
Analysis of the treatment strategies
The results of the analysis of the treatment strategies are presented in Table 2.
|
(i) Referral first strategy. The regression model that indicates for whom surgery would be appropriate consists of age, IPSS score
20, IPSS-Qol score 4, irritative symptoms and PSA value. Inclusion of Qmax resulted in a significantly better regression model (P < 0.001), which, however, was hardly visible in the ROC curve. For those for whom surgery was not indicated (N = 473), a similar model without PSA value fitted best to predict the appropriateness of -blocker treatment. Again, inclusion of Qmax resulted in a significantly better regression model (P < 0.001) with a comparable AUC.
(ii) Medication first strategy.
The best fitting models included age, IPSS score 20, IPSS-Qol score 4 and irritative symptoms for -blocker treatment, and the same variables plus PSA value and prostate size for surgery. Inclusion of Qmax resulted in significantly better regression models (P < 0.001), with comparable AUCs.
Evaluation of the checklists
The most feasible situation for daily practice would be if all regression equations would consist of the same variables, making it possible to transform them into one single checklist with different cut-off values for the different situations. However, this was not possible without considerable loss of predictive value of the regression equations. Our trade-off between feasibility and predictive power is shown in Table 3. The corresponding ROC curves are shown in Figure 1.
|
|
(i) Checklists of referral first strategy. Within the referral first strategy, checklist 1 can be used to select patients for referral. In our population referral of anyone with
2 points would have a sensitivity of 95% and a specificity of 78% (Fig. 1). This means that in 26% (37/141) of the patients referred for (determination of the eligibility and necessity of) surgery an operation would indeed be appropriate. Using 3 points as the cut-off these values would be 92, 92 and 49% (36/73), respectively.
Checklist 2 can be used to decide whom to prescribe medication. In our population prescription of an -blocker for anyone with 3 points would have a sensitivity of 92% and a specificity of 90%. In 52% (45/87) of the patients who received a prescription of an -blocker, this would be appropriate; the others (48%, 42/87) were classified by the expert panel as ‘uncertain’; none of them was classified as inappropriate (according to the expert panel). The cut-off of 1 point would almost be unselective: all subjects for whom an -blocker was considered inappropriate by the expert panel would receive this medication.
(ii) Checklists of medication first strategy.
Checklist 3 can be used to select those for whom an -blocker would be appropriate. It is identical to the situation in which already a selection of those for whom surgery would be indicated had taken place. In our population prescription of an -blocker for anyone with 3 points would have a sensitivity of 95% and a specificity of 90%, and a positive predictive value of 66% (82/124). Again, the others (34%, 42/124) consisted only of people for whom the appropriateness was uncertain.
Since surgery would only be considered in those situations that an -blocker has failed, we considered the sensitivity and specificity in the 124 subjects that according to the checklist 3 would have received an -blocker. Using 7 points on checklist 4 as the cut-off, there would be a calculated sensitivity of 95% and a corresponding specificity of 90%.
|
Discussion |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Main results
We developed a decision aid using an approach in which GPs anticipate on the probable treatment of the urologist (Fig. 2). Using simple checklists GPs can follow a rational treatment strategy without gross overprescription or undertreatment of patients.
|
Strengths and limitations
Any research using consensus procedures or panel judgements to define the presence or absence of a health status or—like in this study—the appropriateness of a certain treatment, is subject to changes due to new insights and developments in medicine. However, there has not been a major breakthrough recently.
Our study population was based on the general population. This is very similar to actual everyday patients in primary care where new patients with LUTS tend to be less symptomatic.
Our study was based on the results of a study with a Dutch expert panel, which was subsequently validated with an international panel of urologists. Our results can therefore be generalized beyond the Dutch situation. Of course, a local urologist may have different opinions than the expert panel. Opinions and daily practice have been shown to differ sometimes substantially. Even then, however, the strategies proposed here are defendable due to the multinational composition of the expert panel.
We developed the checklists and subsequently calculated the effects of their use with regard to the number of adequate referrals and prescriptions. These figures were not calculated in a separate ‘test’ sample, but in the same sample in which the models were constructed. They therefore must be regarded as—somewhat optimistic—indications of the real sensitivity, specificity and predictive value. The checklists should be validated in other primary care populations, ideally with real decisions of an expert panel as outcome parameter. Such studies, however, are difficult to perform and probably expensive.
The expert panel classified the cases into three categories: appropriate, uncertain and inappropriate. In a primary care situation, and for this ailment, it is reasonable to treat the ‘uncertain’ group as (for the time being) ‘not appropriate’. In case of referred patients (secondary care) the primary care physician has already decided not to follow a WW strategy. It is therefore logical that urologists who take their primary care colleagues seriously want to further investigate whether there are specific circumstances that may justify a certain treatment in an individual patient. In a primary care situation however, it is often justified to start with a WW strategy, provided this includes a proper follow-up policy.
The inclusion and exclusion criteria of our study signify that our study does not apply to patients with mild symptoms (IPSS <8), nor to situations of AUR or active UTIs, which require different therapies. It furthermore presupposes that the GP has checked for (a high risk for) prostate cancer first (using digital rectal examination and PSA measurement). It is however applicable in the situation in which an elevated PSA (between 2.5 and 10 ng/ml) requires follow-up (watchful waiting) for prostate cancer (but not necessarily also for LUTS).
What this study adds
In a process of shared decision making, the patient should make a personal decision, weighing the information provided by the health professional. Preferably, this information is based on evidence from randomized controlled trials with hard clinical endpoints performed in a similar population and under similar circumstances. This means effectiveness studies in primary care. Lacking such trials we think our study is the next best option to create a rational and transparent treatment strategy. Rationality and transparency are important conditions for adequate patient information. Our study shows that such an opinion-based guideline can be translated into a simple and practical tool for primary care. It can be used in the overwhelming majority of patients consulting with LUTS. We consider our checklist feasible for daily practice since the IPSS and IPSS-Qol consist of only a total of eight questions. These are designed to be filled out by patients themselves. The regular use of the IPSS/IPSS-Qol has been recommended in primary care guidelines.3,4,8 The time required is <1 minute. The length of a regular consultation will therefore not increase considerably.
Our checklists predict the treatment policy of urologists well. The added value of urinary flow measurements is too small to promote its use for improving decision making in general practice.
We analysed two different strategies, which in principle are equally defendable. However, our results suggest that the medication first strategy is the most sensible of the two. This strategy guarantees the most efficient trial of -blockers as recommended by the current guidelines.3
|
Conclusion |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Using short checklists GPs can follow a reasonable strategy for the treatment of LUTS. However, our decision aid should be challenged and validated in other primary care settings and populations.
|
Acknowledgments |
|---|
Funding to pay the Open Access Publication charges for this article was provided by the Care and Public Health Research Institute (CAPHRI) of the University of Maastricht.
|
Notes |
|---|
Norg RJC, Portegijs PJM, van de Beek C, van Schayck CP and Knottnerus JA. A decision aid for GPs for the treatment of elderly male patients with lower urinary tract symptoms (LUTS). Family Practice 2006; 23: 666–673.
|
References |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
1 Prescription Cost Analysis: Department of Health. (2004–12) Bulletin 2004-12; ISBN 1 84182 866 1 and 1 84182 710 X; Publication date: 30 June 2004. Available at: http://www.dh.gov.uk/healthcarestatistics.
2 De la Rosette JJ, Madersbacher S, Alizivatos G, Rioja Sanz C, Emberton M, Nordling J. (2004) Guideline on benign prostatic hyperplasia: European Association of Urology. December 2004. Available at: http://www.uroweb.org/files/uploaded_files/BPHDec2004Webversion.pdf.
3 Speakman MJ, Kirby RS, Joyce A, Abrams P, Pocock R. (2004) Guideline for the primary care management of male lower urinary tract symptoms. BJU Int 93:985–990.[CrossRef][Web of Science][Medline]
4 Bruskewitz R. (1999) Management of symptomatic BPH in the US:who is treated and how? Eur Urol 36:Suppl. 3, 7–13.
5 Wolters RJ, Spigt MG, van Reedt Dorland PFH, et al. (2004) NHG-Standaard Bemoeilijkte mictie bij oudere mannen. Huisarts en Wetenschap 47:571–586.
6 Stoevelaar HJ, Van de Beek C, Casparie AF, McDonnell J, Nijs HG. (1999) Treatment choice for benign prostatic hyperplasia: a matter of urologist preference? J Urol 161:133–138.[CrossRef][Web of Science][Medline]
7 McDonnell J, Stoevelaar HJ, Bosch JL, Kahan JP. (2001) The appropriateness of treatment of benign prostatic hyperplasia: a comparison of Dutch and multinational criteria. Health Policy 57:45–56.[CrossRef][Web of Science][Medline]
8 American Urologic Association. (2003) Guideline on the management of Benign Prostatic Hyperplasia (BPH). Available at: http://www.auanet.org/guidelines/bph.cfm.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||