SeHCAT [tauroselcholic (selenium-75) acid] for the investigation of bile acid malabsorption and measurement of bile acid pool loss: a systematic review and cost-effectiveness analysis

Bile acid malabsorption

The principal diagnosis/indication for this assessment is chronic diarrhoea due to bile acid malabsorption (BAM). Diarrhoea can be defined as the abnormal passage of loose or liquid stools more than three times daily and/or a daily stool weight greater than 200 g per day and is considered to be chronic if it persists for more than 4 weeks. 1 The cause of chronic diarrhoea in adults is often difficult to ascertain and patients may undergo several investigations without a definitive cause being identified. 2 Chronic diarrhoea is one of the most common reasons for referral to a gastrointestinal clinic,3 and could account for as many as 1 in 20 referrals. Estimates of the prevalence of chronic diarrhoea in a Western population are 4–5%. 4 Some of the causes of chronic diarrhoea are given in Box 1 .

Bile acid malabsorption is one of several causes of chronic diarrhoea (see Box 1 ) and results from failure to absorb bile acids (which are required for the absorption of dietary fats and sterols in the intestine) in the distal ileum. Normally, more than 90% of the acids are reabsorbed in the distal ileum. BAM results in excess bile acids in the colon where they cause diarrhoea by various mechanisms. These mechanisms include:

• inducing secretion of sodium and water, particularly at a concentration above 3 mmol/l

• increase colonic motility

• stimulating defecation

• inducing mucus secretion

• causing damage to mucosa, thereby increasing mucosal permeability.

Bile acid malabsorption has been divided into three types depending on aetiology:

1. type 1: following ileal resection, disease or bypass of the terminal ileum

2. type 2: primary idiopathic malabsorption

3. type 3: associated with cholecystectomy, peptic ulcer surgery, chronic pancreatitis, coeliac disease and diabetes mellitus.

Irritable bowel syndrome

People with chronic diarrhoea are often diagnosed as having diarrhoea-predominant irritable bowel syndrome (IBS-D) if a definitive cause has not been identified. There is evidence that suggests a high prevalence of BAM (up to one-third) in patients previously diagnosed with IBS-D. 3,5 On this basis, approximately half a million patients in the NHS who are currently treated for IBS-D actually have BAM, for which potential diagnosis and effective treatment are available. 6

Irritable bowel syndrome is one of the most common functional gastrointestinal disorders. It is a chronic, relapsing and often life-long disorder, characterised by the presence of abdominal pain/discomfort associated with defecation, a change in bowel habit together with disordered defecation (constipation or diarrhoea or both), the sensation of abdominal distension, and can include associated non-colonic symptoms. These morbidities can cause dehydration, lack of sleep, anxiety and lethargy, with consequences such as time taken off work, avoidance of stressful or social situations and significant reduction in quality of life. 7

Irritable bowel syndrome most commonly affects people between the ages of 20 and 30 years and is twice as common in women as in men. People with IBS are the largest group of patients seen in a general gastroenterology clinic (1 in 20 referrals). The prevalence of the condition in the general population is estimated at between 10% and 20%. 7 Recent trends indicate that there is also a significant prevalence of IBS in older people, and therefore IBS diagnosis should be a consideration when an older person presents with unexplained abdominal symptoms. The true prevalence of IBS in the whole population may be higher than estimated, because it is thought that many people with IBS symptoms do not seek medical advice; NHS Direct online data suggest that 75% of people using this service who have IBS symptoms rely on self-care. In England and Wales, the number of people consulting for IBS is extrapolated to between 1.6 million and 3.9 million. 7

Inflammatory bowel disease

Ulcerative colitis and Crohn’s disease are the two most common forms of inflammatory bowel disease (IBD). Together these long-term conditions are estimated to affect about 240,000 people in the UK: approximately 400 per 100,000 population. 8 Both ulcerative colitis and Crohn’s disease directly cause chronic diarrhoea.

Crohn’s disease is a chronic severe condition characterised by inflammation, ulcers and bleeding which may affect any part of the gastrointestinal tract but mostly the terminal ileum. There are approximately 60,000 people in the UK with this condition ( Figure 1 ). 9 Crohn’s disease is sometimes treated by ileal resection. In a study carried out by Smith et al. 3 BAM was found in 97% of people with Crohn’s disease with ileal resection who were in clinical remission and in 54% of people in clinical remission with unoperated Crohn’s disease.

FIGURE 1.

Venn diagram with approximate population sizes in the UK. 1, 3 million in the UK (prevalence of chronic diarrhoea in a Western population: 4–5%);1 2, unknown; 3, 1.3 million in the UK (there are up to 3.9 million adults in the UK being treated for IBS, with one-third of these having IBS-D);10 4, unknown (there are approximately 60,000 people in the UK with Crohn’s disease);9 5, unknown (BAM was found in 54% of people in clinical remission with unoperated Crohn’s disease);3 6, 500,000 in the UK. 10

Diagnosis

Patients with undiagnosed BAM are likely to present with chronic diarrhoea. The BSG guideline states that bile salt malabsorption occurs when normal active uptake from the ileum is disrupted by ileal inflammation or resection. It also states that the degree of malabsorption depends on the length of ileal involvement or resection. According to the BSG guidelines, diagnosis of BAM can be made via SeHCAT scanning.

During early scoping, two issues arose regarding the BSG pathway. First, all experts agreed that SeHCAT needs to be placed earlier in the pathway to help patients gain a firm diagnosis at an earlier stage. However, expert opinion varied as to where SeHCAT should be placed on the pathway. Some felt that it should be available to general practitioners (GPs) for use in all patients with chronic, erratic bowels with a tendency to diarrhoea, while others felt that it is more appropriate for use in secondary care. Second, the BSG guideline does not take into account the prevalence of BAM in people diagnosed with IBS. The BSG guideline place SeHCAT at the end of the diagnostic algorithm (position C in Figure 2 ). Possible alternatives are:

1. SeHCAT as part of the basic investigations for all patients presenting with chronic diarrhoea (position A in Figure 2 )

2. SeHCAT for all patients presenting with chronic diarrhoea and symptoms suggestive of functional disease (i.e. age < 45 and normal basic investigations) (position B1 in Figure 2 ); and also for patients with a history of findings suggestive of colonic or terminal ileal disease (position B2 in Figure 2 ).

FIGURE 2.

British Society of Gastroenterology’s guideline for the investigation of chronic diarrhoea (adapted from Thomas et al. 1). B12, vitamin B12; Ca, calcium; FBC, full blood count; LFT, liner function tests.

Selenium-75-homocholic acid taurine as part of the basic investigations (position A in Figure 2 ) means that all patients presenting with chronic diarrhoea will be tested with SeHCAT. However, during the scoping workshop clinical experts advised that a positive SeHCAT test at this stage does not rule out the possibility of organic disease. As no subsequent tests for organic disease are made redundant, it is unlikely that SeHCAT in position A will be more cost-effective than in position B1. Therefore, this assessment will focus on position B1.

The same applies to SeHCAT in position B2. A positive SeHCAT test in position B2 is not thought likely to stop clinicians from doing subsequent tests such as sigmoidoscopy, barium enema or colonoscopy. Therefore, in the assessment, using SeHCAT in position B2 and in position C will be considered as having the same effect on the care pathway.

This leaves two possible populations for investigation:

1. people presenting with chronic diarrhoea with unknown cause and symptoms suggestive of functional disease

2. people with Crohn’s disease and chronic diarrhoea with unknown cause (i.e. before resection of the terminal ileum).

Treatment

Following a definitive diagnosis, patients may be treated with bile acid sequestrants (BASs), which can cause significant reduction in bowel frequency and therefore a better quality of life. 13 There are currently two types of BAS available:

1. Bile binding resins (cholestyramine and colestipol): although they are tolerated by some, most people dislike these treatments because of side effects and the difficulty of administration, but take them in the long term because they can help in managing the condition. One in four patients cannot take more than a single dose.

2. Colesevelam (Cholestagel,^® Genzyme) (gel matrix): two-thirds of patients take this treatment for at least 4 years and most of those who stop do so because they found no benefit initially.

The response to BAS therapy varies among people with bile acid diarrhoea. For those with Crohn’s disease with ileal resection and BAM (assessed with SeHCAT), the response to BAS was 60%; response was 40% in those with Crohn’s disease without ileal resection and BAM and 70% in those with a diagnosis of IBS-D and BAM. 3

Participants

Study populations eligible for inclusion will be all patients (including children) referred to a gastrointestinal clinic for investigation and diagnosis of BAM which is a common underlying cause of chronic diarrhoea and the measurement of bile acid pool loss. 3,5

As explained above, this report will focus on two specific populations:

1. people presenting with chronic diarrhoea with unknown cause and symptoms suggestive of functional disease

2. people with Crohn’s disease and chronic diarrhoea with unknown cause (i.e. before resection of the terminal ileum).

Setting

The relevant setting is secondary care.

Interventions

The intervention is SeHCAT.

Comparators

The comparator will be no SeHCAT test (the current situation).

Outcomes

The following outcomes are considered:

• effect of testing on treatment plan (e.g. surgical or medical management), where information on the appropriateness of the final treatment plan is also reported

• effect of testing on clinical outcome (e.g. morbidity and adverse events)

• prognosis – the ability of test result to predict clinical outcome (e.g. response to treatment).

For included studies reporting any of the above outcome measures, the following outcomes will also be considered if reported:

• acceptability of tests to patients or surrogate measures of acceptability (e.g. waiting time and associated anxiety)

• adverse events associated with testing (e.g. pain/discomfort experienced during the procedure and waiting times before results).

Study design

The following study designs were eligible for inclusion:

• Randomised or non-randomised controlled trials, where participants are assigned to the intervention or comparator tests, for treatment planning, and outcomes are compared at follow-up.

• Observational studies which report the results of multivariable regression modelling with clinical outcome as the dependent variable and index test result as an independent variable. Included studies should control adequately for potential confounders (e.g. age, sex, disease, etc.).

The following study/publication types were excluded:

• pre-clinical and animal

• reviews, editorials, and opinion pieces

• case reports

• studies reporting only technical aspects of the test, or image quality

• studies with < 10 participants.

As no studies were found with either of the above-mentioned study designs, it was decided to broaden the inclusion criteria by allowing lower levels of evidence (change to protocol). Therefore, observational studies reporting data to calculate the accuracy of SeHCAT in predicting treatment response and studies reporting data on the clinical effectiveness of treatment given a positive and/or negative SeHCAT test will also be included.

Test accuracy

The results of DTA studies included in this review were summarised by clinical indication (chronic diarrhoea only). For all included studies, the absolute numbers of true-positive (TP), false-negative (FN), false-positive (FP) and true-negative (TN) test results of SeHCAT compared with the reference standard of treatment response, as well as sensitivity and specificity values, with 95% confidence intervals (CIs) were presented in results tables. The results of individual studies were plotted in the receiver operating characteristic (ROC) plane, with the diagnostic threshold used for the SeHCAT test indicated.

As no studies were found for the assessment of SeHCAT’s test accuracy and very few studies for the accuracy of SeHCAT to predict treatment response, it was decided to include studies reporting response to BAS given a positive test to estimate the probability of a positive BAS response at different SeHCAT cut-off points. Based on the data retrieved, a random-effects meta-analysis was performed to find a pooled estimate for each of the three cut-off values.

Dichotomous outcomes

Dichotomous data were analysed by calculating the relative risk (RR) for each trial using the random-effects DerSimonian and Laird method and the corresponding 95% CIs. 26

Continuous outcomes

Continuous data were analysed by calculating the standardised mean difference (SMD) between groups and the corresponding 95% CI, due to the different types of outcome measures. Where the standard deviations and means were not determinable, they were estimated from the data that were provided or from a representative value from other studies.

Systematic differences between studies (heterogeneity) are likely; therefore, the random-effects model was used for the calculation of RRs or SMDs. Heterogeneity was initially to be assessed by measuring the degree of inconsistency in the studies’ results (I ²). This measure (I ²) describes the percentage of total variation across studies that were due to heterogeneity rather than the play of chance. The value of I ² can lie between 0% and 100%. Low, moderate and high I ² values correspond to 25%, 50% and 75%.

Where important heterogeneity was identified, we planned to formally investigate this using meta-regression. In addition, a funnel plot (plots of logarithm of the RR for efficacy against the precision of the logarithm of the RR) was planned to be generated in order to estimate potential asymmetry, which is indicative of small study effects. In addition, we wanted to use the Egger regression asymmetry test in order to facilitate the prediction of potential publication biases. This test detects funnel plot asymmetry by determining whether or not the intercept deviates significantly from zero in a regression of the standardised effect estimates against their precision. However, due to the lack of data this was not possible.

Statistical analyses were performed using the following software: RevMan (version 5, The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark), Comprehensive Meta-Analyses (CMA version 2, Biostat, Englewood, NJ, USA: www.meta-analysis.com/pages/about_us.html) and Stata (version 10, StataCorp LP, College Station, TX, USA).

A detailed commentary on the major methodological problems or biases that affected the studies was also included, together with a description of how this may have affected the individual study results. Recommendations for further research were made based on any gaps in the evidence or methodological flaws.

Accuracy of selenium-75-homocholic acid taurine for the detection of bile acid malabsorption

As mentioned in the NICE scope, there is no direct comparator for SeHCAT. Current diagnostic options include analysis of a patient’s history, investigations to exclude ‘red flag’ symptoms and a variety of other diagnostic tests such as blood tests and lactose tolerance tests. Trial of treatment and measurement of faecal bile acids are two methods used, with mixed results, to diagnose BAM. They are, however, not widely used in current practice.

In addition, the clinical effectiveness of BASs in people with chronic diarrhoea with unknown cause and in people with Crohn’s disease and chronic diarrhoea of unknown cause is not known. Therefore, this assessment will focus on the value of a SEHCAT test in predicting the response to treatment with BASs.

Any studies described in the literature as accuracy studies (Sciaretta42 in Johnston et al. ;52 Merrick39 and Sciaretta42 in Kurien et al. ;53 and Sciaretta42 in Wedlake et al. 6), will be included in this review under studies predicting the response to BAS, if treatment response was assessed.

None of the studies evaluated in this report, including the studies described in the literature as accuracy studies, was included in this review as diagnostic test accuracy (DTA) studies, because they either do not use an acceptable reference standard or they include a population not in line with the scope (i.e. healthy volunteers or people with ileal resection).

Accuracy of selenium-75-homocholic acid taurine for the assessment of response to treatment in people with chronic diarrhoea

Nineteen of the 20 studies with information on the relationship between SeHCAT and response to treatment were in patients with chronic diarrhoea. These can be divided in different groups, depending on the reliability of the results.

Most studies (12 out of 19) have no information about respondents with a negative SeHCAT test. 2,3,5,31–33,37,38,40,44,45,49 Therefore, these studies should be regarded as flawed design for the purposes of this review and cannot be used to reliably asesess the relationship between SeHCAT and response to treatment. Another three studies34,35,41 have very limited data for respondents with a negative SeHCAT test.

Four studies36,39,42,43 have data for all respondents with a negative SeHCAT test. However, in two of these studies36,39 it is unclear why certain patients are treated and others not. In Fellous et al. ,36 16 out of 53 patients with functional diarrhoea were treated with cholestyramine; 11 of these had a positive SeHCAT test and 5 a negative SeHCAT test. It is not clear why the other 37 patients with functional diarrhoea did not receive cholestyramine.

This leaves three studies to assess the relationship between the SeHCAT test and treatment with cholestyramine:

• Merrick et al. 39 estimated the sensitivity of SeHCAT in predicting a positive response as 0.667 (95% CI 0.223 to 0.957), and the specificity as 0.971 (95% CI 0.847 to 0.999) using a cut-off of 8% for the test. Using a cut-off of 15%, the sensitivity is 1.000 (95% CI 0.541 to 1.000) and the specificity is 0.912 (95% CI 0.763 to 0.981).

• Sciaretta et al. 42 estimated the sensitivity of SeHCAT in predicting a positive response as 0.857 (95% CI 0.421 to 0.996), and the specificity as 1.000 (95% CI 0.541 to 1.000) using a cut-off of 5% for the test. However, only 13 patients were included in this analysis.

• Sciaretta et al. 43 estimated the sensitivity of SeHCAT in predicting a positive response as 0.950 (95% CI 0.751 to 0.999), and the specificity as 0.962 (95% CI 0.804 to 0.999) using a cut-off of 8% for the test.

It should be noted here that in the study by Merrick et al. ,39 31 patients were considered TNs. This assessment was based on long-term follow-up: ‘None of the 31 patients with irritable bowel disease who retained more than 15% at seven days showed any evidence of small bowel disease, and none appeared during a follow-up of at least 12 months, and in some up to 24 months. Simple conservative treatment resolved or eased most symptoms’ (p. 266). The remaining nine patients were treated with cholestyramine; five of these had a SeHCAT test result < 8%, one of whom did not respond to treatment and four had an equivocal (between 8% and 15%) SeHCAT test result; two of these patients responded to cholestyramine and two did not.

In the study by Sciaretta et al. ,43 the cut-off for a positive SeHCAT was 8% based on a 7-day SeHCAT retention measurement. However, in the study by Sciaretta et al. ,42 the cut-off for a positive SeHCAT was most likely based on a 3-day retention measurement, which was according to the authors equivalent to 5% at 7 days. These issues and possible overlap in populations in both Sciaretta papers are discussed in more detail in the quality assessment (QUADAS-2, see Appendix 3 ) of these studies.

As can be seen in Figure 5 , the sensitivity is highest with a cut-off of 15% and decreases with lower cut-offs; while the specificity is less clearly related to the cut-off used.

FIGURE 5.

‘Accuracy’ of SeHCAT to predict a response to treatment at different cut-offs. The centre dots represent the point estimates for sensitivity and specificity of SeHCAT in predicting response to treatment in the three studies at different cut offs (5%, 8% and 15%). The vertical and horizontal lines represent the 95% CIs for sensitivity and specificity respectively; see data in Table 4 .

The between-study heterogeneity in these three studies is considerable. The principal diagnosis, treatment dose, the definition of response, follow-up period and SeHCAT administration was different between trials. Most of these differences were between Merrick and the two Italian studies, but even the two Italian studies were not completely similar (see Appendices 3 and 4 for details).

Three studies were included in the QUADAS-2 assessment; all three studies were rated as ‘unclear’ risk of bias for the patient selection domain. Merrick et al. 39 was rated ‘high’ risk of bias for the ‘flow and timing’ domain of QUADAS-2 because only test positive and equivocal patients received the reference treatment. The test-negative patients were followed up and hence all patients did not receive the same reference standard. Sciaretta et al. 42 was rated ‘high’ risk of bias for the ‘index test’ domain of QUADAS-2 because a threshold was not prespecified. Sciaretta et al. 43 was rated as ‘low’ risk of bias for all the QUADAS-2 domains in this assessment except for ‘patient selection’ domain, where it was rated ‘unclear’ risk of bias because the study did not clearly state whether the patient enrolment was prospective or retrospective. The applicability concerns were high in two studies for the ‘patient selection’ domain because some patients had cholecystectomy and < 90% of patients had unknown-cause diarrhoea. 42,43 Merrick et al. 39 had low applicability concerns for the ‘patient selection’ domain.

Table 3 provides a summary of the quality assessments for studies in this section and Table 4 summarises individual study results.

Accuracy of selenium-75-homocholic acid taurine for the assessment of response to treatment in people with Crohn’s disease

None of the studies looking specifically at people with Crohn’s disease presented reliable data for the prediction of response to treatment with BAS. Only two studies looked at this population. 3,46 Neither of these studies presented data for people with a negative SeHCAT test. In addition, it was not clear why certain people were treated and others not. In the study by Nyhlin et al. ,46 34 out of 51 patients had a positive SeHCAT test at a 10% cut-off, while 22 patients were treated, of whom one had a negative SeHCAT (at 10%). In the study by Smith et al. ,3 24 out of 44 patients had a positive SeHCAT test at a 10% cut-off; 11 of these were successfully treated with conventional treatment (prednisolone with or without 5-ASA) and a further nine patients were treated with BAS after conventional treatment had failed. Finally, Nyhlin et al. 46 included patients with or without resection.

Effectiveness of bile acid sequestrants for the treatment of bile acid malabsorption in patients with chronic diarrhoea

One controlled clinical trial compared colesevelam with placebo for patients with IBS-D. 47,50,51 All participants had fasting plasma 7α-C4 (C4) measured to assess for underlying BAM and had serum FGF-19 measured. However, it is not certain whether or not this was used as an inclusion criterion. SeHCAT was not used in this trial.

No controlled trials were found to assess the clinical effectiveness of cholestyramine in terms of bowel function in patients with chronic diarrhoea of unknown cause, nor were any such trials found for other BASs.

This randomised trial was considered to have ‘risk of bias’ in a number of areas: sequence generation was described as independently generated by a statistician at the Mayo Clinic, no further details were reported, allocation concealment was not described, and results for some of the outcomes were not reported.

According to the authors, ‘colesevelam modestly affected overall colonic transit (24 hours, p = 0.22). Emptying of the ascending colon took an average four hours longer in patients given colesevelam compared with placebo. Colesevelam was associated with greater ease of stool passage (p = 0.048) and somewhat firmer stool consistency (p = 0.12). No effects on mucosal permeability or safety were identified’ (p. 160). 47 Our analyses using Cochrane’s Review Manager software identified no significant differences between colesevelam and placebo for all outcomes.

Table 5 provides a summary of the risk of bias assessment for this study and Table 6 summarises results.

In addition to the RCT reported above, information about the clinical effectiveness of BAS for the treatment of BAM can be derived from the 19 studies reported in Chapter 4 (see Accuracy of SeHCAT for the assessment of response to treatment in people with chronic diarrhoea). All 19 studies provide data on the clinical effectiveness of BAS given a positive SeHCAT test; three studies39,42,43 also provide data on the effectiveness of BAS given a negative SeHCAT test. Table 7 provides a summary of the risk of bias assessment for these studies; the results are reported in Tables 8 and 9 . In these tables we have used data for the population defined in the scope where we could find them and we have included only patients who were actually treated, ignoring those with a positive SeHCAT test who were not treated.

Meta-analysis of test accuracy studies was considered inappropriate, owing to the small number of studies with varying diagnostic thresholds and between-study heterogeneity in other study design categories (principal diagnosis, treatment dose, definition of response, follow-up period and SeHCAT administration); therefore, we employed a narrative synthesis in Chapter 4 (see Accuracy of selenium-75-homocholic acid taurine for the assessment of response to treatment in people with Crohn’s disease). As mentioned previously, the between-study heterogeneity in the 19 studies included in this section is also considerable. The principal diagnosis, treatment dose, the definition of response, follow-up period and SeHCAT administration were different between trials (see Appendix 4 for details). However, for the economic model we need estimates for the probability of a positive SeHCAT test in people with IBS-D or Crohn’s disease and the probability of a positive BAS response given a positive test result (see Chapter 5 , Model parameters). These probabilities can be derived from the data in these studies. Despite the heterogeneity between studies, we decided to combine the results from different studies to derive an estimate for these probabilities. As no studies could be classified as superior based on risk of bias, we decided to include all studies in the meta-analysis. We chose a random-effects model, as this would ensure that the CI would be wide enough to capture most uncertainties. However, given the large heterogeneity, these results should be treated with appropriate caution.

For those with a positive SeHCAT test response rates ranged from 74% to 100% at a cut-off of 5% and from 62% to 86% at a cut-off of 15%. For those with a negative SeHCAT test the response rate was 14% at a cut-off of 5% and 0% at a cut-off of 15%.

Table 10 shows the average response rates given a positive or negative SeHCAT test at all different cut-offs using all available data for patients with unknown cause chronic diarrhoea. This analysis combines results from different studies weighted by population size.

Using the random-effects analysis,26 the results are very similar ( Table 11 ). Data are now grouped in cut-off bands of 5%, 10%, and 15% plus or minus 2% points. These are the data used in the economic model.

Effectiveness of bile acid sequestrant for the treatment of bile acid malabsorption inpatients with Crohn’s disease

As reported earlier in this chapter (see Accuracy of selenium-75-homocholic acid taurine for the assessment of response to treatment in people with Crohn’s disease), none of the studies looking specifically at people with Crohn’s disease presented reliable data for the prediction of response to treatment with BAS. Only two studies looked at this population. 3,46 Neither of these studies presented data for people with a negative SeHCAT test. In addition, it was not clear why certain people were treated and others not. In Nyhlin et al. ,46 34 out of 51 patients had a positive SeHCAT test at a 10% cut-off, while 22 patients were treated, of whom one had a negative SeHCAT (at 10%). In Smith et al. ,3 24 out of 44 patients had a positive SeHCAT test at a 10% cut-off; 11 of these were successfully treated with conventional treatment (prednisolone with or without 5-ASA) and a further nine patients were treated with BAS after conventional treatment had failed. Finally, Nyhlin et al. 46 included patients with or without resection. Nevertheless, information about the effectiveness of BAS for the treatment of BAM given a positive SeHCAT test can be obtained from the two studies.

Table 12 provides a summary of the quality assessments for studies in this section and Table 13 summarises individual study results.

For those with a positive SeHCAT test the response rate was 95% at a cut-off of 5% and 86% or 89% at a cut-off of 15%.

Table 14 shows the average response rates given a positive SeHCAT test at two different cut-offs using all available data.

Diagnostic and initial treatment model for diarrhoea-predominant irritable bowel syndrome

Given the fact that no data are available on the accuracy of SeHCAT, the model uses response to treatment, dependent on the test outcome (see also Chapter 3 , Accuracy of selenium-75-homocholic acid taurine for the detection of bile acid malabsorption). We compare two no SeHCAT strategies with three SeHCAT strategies. Figure 7 presents the decision tree used.

FIGURE 7.

Decision tree for IBS-D population.

In the SeHCAT strategies, patients may have a positive or negative test result. If the test is positive, that is to say the percentage resorption of bile acids is below a certain cut-off, patients are treated with cholestyramine (a BAS) and they may or may not respond to that treatment. If the test is negative, the patient is diagnosed as IBS-D and treated accordingly. Again, patients may or may not respond to that treatment. The first no SeHCAT strategy assumes that all patients are treated with IBS-D treatment and patients may or may not respond to that treatment. Finally, in the trial of treatment, all patients receive cholestyramine. If patients do not respond to this, they receive IBS-D treatment and again, patients may respond or not respond to this treatment.

Diagnostic and initial treatment model for Crohn’s disease without ileal resection

For the Crohn’s disease population, the outline of the model is in essence the same as for the IBS-D population ( Figure 8 ). However, here, if no SeHCAT test is used, if the test result is negative or if patients do not respond to a trial of cholestyramine, patients will receive treatment for diarrhoea. This treatment can vary between patients, as the diarrhoea may be a result/symptom of a relapse of the disease, and thus focus could be on treating the relapse, but it may also occur in patients who are in remission. 3 Chronic diarrhoea in patients in remission has been suggested to be the result of repeated relapses, which have led to fibrosis. 46 For these patients, treatment is targeted at controlling the diarrhoea.

FIGURE 8.

Decision tree for Crohn’s disease population.

This model assesses only two SeHCAT strategies; no data on the probability of a positive test result were available for a cut-off absorption of < 5%, and so this strategy was not evaluated in this assessment.

Markov model

To assess the long-term costs and effects of the various strategies, patients enter into a three-state Markov model ( Figure 9 ). Patients who had a treatment response enter the Markov model into the ‘no diarrhoea’ state and patients who did not respond enter into the ‘diarrhoea’ state. As the model has a lifetime time horizon, the third state included is death.

FIGURE 9.

Outline Markov model.

In theory, patients can either move between the ‘no diarrhoea’ state and the ‘diarrhoea’ state from one cycle to another, stay in the same state or die. The cycle length used is 6 months. The general Markov model is parameterised according to treatment; that is to say, in the IBS-D population, patients who receive(d) BAS enter the BAS Markov model and patients who receive(d) IBS-D treatment enter the IBS-D Markov model. Likewise, in the Crohn’s disease population, patients who receive(d) BAS enter the BAS Markov model and patients who receive(d) treatment for Crohn’s disease with chronic diarrhoea (non-BAM diarrhoea) enter the Crohn’s disease Markov model.

Diagnostic and initial treatment model for diarrhoea-predominant irritable bowel syndrome

The diagnostic model estimates the initial costs of diagnosis and initial treatment. We assume that the diagnostic model reflects the first 6 months after the moment the patient is seen by a specialist.

Markov model for diarrhoea-predominant irritable bowel syndrome population

Diagnostic and initial treatment model for Crohn’s disease without ileal resection

Markov model for Crohn’s disease without ileal resection

Uncertainty analysis

In the model many uncertainties exist with regard to the input data. The impact of these uncertainties was explored through probabilistic sensitivity analysis. Cost-effectiveness acceptability curves (CEACs) have been used to describe which strategy has the highest probability of being considered cost-effective given a threshold incremental cost-effectiveness ratio (ICER). All values used are listed in Tables 31 and 32 .

We also explore the value of information associated with the model uncertainty by estimating the expected value of perfect information (EVPI), which is the amount the decision-maker should be willing to pay to eliminate all uncertainty in the decision. For the IBS-D model we assumed a potential population of 1.3 million in the UK (see Figure 1 ). For the Crohn’s disease model, we do not know the size of the population and present, therefore, the EVPI per patient.

In addition, extensive scenario analyses were performed for all those input parameters where, due to lack of published data, assumptions had to be made.

Model assumptions

Costs, effects and cost-effectiveness for diarrhoea-predominant irritable bowel syndrome

Short-term results for diarrhoea-predominant irritable bowel syndrome (decision tree)

Table 33 presents the probabilistic outcomes of the base-case decision tree for the IBS-D population. Results are presented in order of increasing number of responders. From this, it becomes clear that the SeHCAT 15% strategy dominates the other SeHCAT strategies. Also, SeHCAT 15% is more expensive while also generating more responders.

TABLE 33 - Short-term results for IBS-D population, base case (scenario A1)

Strategy	Percentage of responders	Costs (per patient)	Incremental responders	Incremental costs	ICER
No SeHCAT test – treat IBS-D	0.5145	208
SeHCAT test – 5% cut-off	0.632	569	Dominated by SeHCAT 15%
SeHCAT test – 10% cut-off	0.654	555	Dominated by SeHCAT 15%
SeHCAT test – 15% cut-off	0.6596	553	0.1451	345	2378

To investigate what the impact is of the various (statistical) uncertainties in the model, we also present the CEAC ( Figure 10 ). It should be noted that the outcome here is cost per responder; we cannot judge the curve using the usual threshold, which relates to cost per QALY.

FIGURE 10.

Cost-effectiveness acceptability curve short-term (cost per responder) results for IBS-D population, base case (scenario A1).

As mentioned in the methods section, various assumptions regarding probabilities had to be made where neither literature nor expert opinion was available. Additionally, the costs of IBS-D treatment are very uncertain, as the experts had very differing responses, and some uncertainty exists about the total costs of SeHCAT.

Thus, we formulated several scenarios, which are listed in Table 34 . We varied the percentage point change in the response rate to IBS-D treatment for SeHCAT-negative and BAS trial-negative patients from 10% (i.e. response rate 62%) to both 0% and 25%; we increased IBS-D treatment costs by 100% and decreased these costs by 50%; and finally we increased SeHCAT cost by £55 and £105 and decreased it by £95.

TABLE 34 - Overview of scenarios explored for short-term IBS-D

Scenarios	SeHCAT service cost (£)	Probability response to IBS-D treatment in SeHCAT-negative patients	Costs IBS-D treatment	Cost SeHCAT capsule (£)	Results presented?
A1	0	0.62		195	Yes
A2	55	0.62		195	No
A3	105	0.62		195	No
A4	0	0.52		195	Yes
A5	0	0.87		195	No
A6	0	0.62	× 2	195	No
A7	0	0.62	/2	195	No
A8	0	0.52	/2	195	Yes
A9	0	0.62		100	Yes
A10	0	0.52		100	No
A11	0	0.87		100	No

We found that increasing SeHCAT costs has no impact on the conclusions from the base case. This is intuitively clear, as in the base case, the SeHCAT strategies are already the most expensive, and these only become slightly more expensive. When we change the probability of IBS-D treatment response in SeHCAT-negative and BAS trial-negative patients to the same value as when all patients receive IBS-D treatment (52% response rate), we see that the number of responders in the SeHCAT strategies decreases ( Table 35 ). The CEAC ( Figure 11 ) shows that now SeHCAT 10% dominates the other SeHCAT strategies for all larger threshold ICERs.

TABLE 35 - Short-term results for IBS-D population, response to IBS-D treatment in SeHCAT-negative patients 52% (scenario A4)

Strategy	Percentage of responders	Costs (per patient)	Incremental responders	Incremental costs	ICER
No SeHCAT test	0.5202	207
SeHCAT test – 5% cut-off	0.5971	568	Dominated by SeHCAT 10%
SeHCAT test – 15% cut-off	0.6014	553	Extendedly dominated by SeHCAT 10%
SeHCAT test – 10% cut-off	0.6068	555	0.0865	348	4019

FIGURE 11.

Cost-effectiveness acceptability curve short-term (cost per responder) results for IBS-D population, response to IBS-D treatment in SeHCAT-negative patients 52% (scenario A4).

When we also decrease the cost of IBS-D treatment to 50% of the base case we see costs per strategy decrease while the incremental costs stay approximately the same ( Table 36 ). Thus, the CEAC ( Figure 12 ) is similar to the previous. Clearly, the impact of the treatment costs for IBS-D for the short term is limited.

TABLE 36 - Short-term results for IBS-D population, response to IBS-D treatment in SeHCAT-negative patients 52% and cost of IBS-D treatment 50% of base case (scenario A8)

Strategy	Percentage of responders	Costs (per patient)	Incremental responders	Incremental costs	ICER
No SeHCAT test – treat IBS-D	0.5272	104
SeHCAT test – 5% cut-off	0.6024	487	Extendedly dominated by SeHCAT 10%
SeHCAT test – 15% cut-off	0.6058	489	Dominated by SeHCAT 10%
SeHCAT test – 10% cut-off	0.6111	489	0.0839	385	4591

FIGURE 12.

Cost-effectiveness acceptability curve short-term (cost per responder) results for IBS-D population, response to IBS-D treatment in SeHCAT-negative patients 52% and cost of IBS-D treatment 50% of base case (scenario A8).

When we reduce the cost of the SeHCAT test we observe that now SeHCAT 15% again dominates the other SeHCAT strategies ( Table 37 ), but overall the impact is limited. This is confirmed by the CEAC where we still find that trial of treatment is dominant. The CEAC ( Figure 13 ) is very similar to the base-case curve, indicating that the impact of lowering the cost of SeHCAT has little impact on the results.

TABLE 37 - Short-term results for IBS-D population, costs of SeHCAT reduced to £100 (scenario A9)

Strategy	Percentage of responders	Costs (per patient)	Incremental responders	Incremental costs	ICER
No SeHCAT test – treat IBS-D	0.5275	208
SeHCAT test – 5% cut-off	0.6421	474	Dominated by SeHCAT 15%
SeHCAT test – 10% cut-off	0.6627	460	Dominated by SeHCAT 15%
SeHCAT test – 15% cut-off	0.668	459	0.1405	251	1784

FIGURE 13.

Cost-effectiveness acceptability curve short-term (cost per responder) results for IBS-D population, costs of SeHCAT reduced to £100 (scenario A9).

Lifetime results for diarrhoea-predominant irritable bowel syndrome

As mentioned in the methods section, no information was available to estimate the transition probabilities in the Markov models other than the all-cause mortality rates. Thus, we formulated various scenarios with different transition probabilities for ND-BAM → D-BAM, D-BAM → ND-BAM, ND-IBS → D-IBS and D-IBS → ND-IBS. Additionally, we have two scenarios regarding the utility in the no diarrhoea health state; either this is the same for BAS and IBS-D or utility of BAS responders is lower than for IBS-D responders. Table 38 presents an overview of all scenarios we explored. As many yielded the same results, we only present a selection of the scenarios, as indicated in the table.

TABLE 38 - Overview of scenarios explored for the long-term analysis, no trial of treatment

Scenario	IBS-D P(ND→D)	IBS-D P(D→ND)	BAM P(ND→D)	BAM P(D→ND)	Utility BAS no diarrhoea	Results presented?
1	0.05	0.05	0.05	0.05	BAS < IBS	Yes
2	0	0	0	0	BAS < IBS	Yes
3	0.05	0	0.05	0	BAS < IBS	Yes
4	0.05	0	0	0	BAS < IBS	Yes
5	0	0.05	0	0	BAS < IBS	No, same as 1
6	0	0	0.05	0	BAS < IBS	No, same as 1
7	0	0	0	0.05	BAS < IBS	No, same as 4
8	0.05	0.05	0.05	0.05	BAS = IBS	Yes
9	0	0	0	0	BAS = IBS	No, same as 2
10	0.05	0	0.05	0	BAS = IBS	No, same as 3
11	0.05	0	0	0	BAS = IBS	No, same as 4
12	0	0.05	0	0	BAS = IBS	Yes
13	0	0	0.05	0	BAS = IBS	No, same as 6
14	0	0	0	0.05	BAS = IBS	No, same as 7

Below are the results for scenario 1. Scenario 1 assumes that all four transition probabilities are 5% per cycle. We observe that all SeHCAT strategies are dominated by no SeHCAT ( Table 39 ). When we look at the CEACs ( Figure 14 ), we notice that no SeHCAT has the highest probability of being cost-effective for the whole range of thresholds.

TABLE 39 - Lifetime results for IBS-D population, scenario 1

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	15.273	1624	Dominated by no SeHCAT
SeHCAT test – 10% cut-off	15.2744	1611	Dominated by no SeHCAT
SeHCAT test – 5% cut-off	15.2925	1462	Dominated by no SeHCAT
No SeHCAT test	15.3006	818

FIGURE 14.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 1.

Figure 15 shows the population EVPI; from this we see that at a threshold of £30,000, the value of perfect information amounts to £10,000M.

FIGURE 15.

Population EVPI for lifetime IBS-D population, scenario 1.

Scenario 2 assumes that all four transition probabilities are 0% per cycle. We now observe ( Table 40 ) that the ICER of SeHCAT 5% compared with no SeHCAT falls well below the commonly used £30,000 threshold. The ICER of SeHCAT 15% compared with SeHCAT 5% is approximately equivalent to this threshold. When we look at the CEAC ( Figure 16 ), we notice that, for small thresholds, no SeHCAT – treat IBS-D has the highest probability of being cost-effective while for larger thresholds (around £10,000) SeHCAT 5% has the highest probability of being cost-effective.

TABLE 40 - Lifetime results for IBS-D population, scenario 2

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	15.3244	823
SeHCAT test – 10% cut-off	15.4127	1783	Dominated by SeHCAT 5%
SeHCAT test – 5% cut-off	15.4133	1613	0.0889	790	8886
SeHCAT test – 15% cut-off	15.4191	1793	0.0058	180	31031

FIGURE 16.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 2.

Scenario 3 assumed that the two transition probabilities from ND to D are 5% per cycle while the other two are 0. As can be seen in Table 41 , SeHCAT 5% has a favourable ICER compared with no SeHCAT, and the subsequent ICER of SeHCAT 15% versus SeHCAT 5% may also be considered acceptable given a threshold of £30,000. The CEAC presented in Figure 17 shows that for lower threshold values (< £20,000) no SeHCAT has the highest probability of being cost-effective while for higher threshold values (> £30,000), all three SeHCAT strategies have a higher probability of being cost-effective than no SeHCAT.

TABLE 41 - Lifetime results for IBS-D population, scenario 3

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	14.8964	751
SeHCAT test – 5% cut-off	14.9272	1307	0.0308	557	18077
SeHCAT test – 10% cut-off	14.9275	1379	Extendedly dominated by SeHCAT 15%
SeHCAT test – 15% cut-off	14.93	1382	0.0028	75	26786

FIGURE 17.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 3.

Figure 18 shows the population EVPI; from this we see that at a threshold of £30,000 the value of perfect information amounts to £30,000M. This is three times higher than in scenario 1, which is explained by the fact that at that threshold value all four strategies have approximately the same probability of being the most cost-effective.

FIGURE 18.

Population EVPI for lifetime IBS-D population, scenario 3.

Scenario 4 assumes that the transition probability from ND to D in the IBS model is 5% per cycle while all others are 0. We observe that SeHCAT 15% compared with no SeHCAT yields an ICER well below the common threshold of £30,000, while the other two SeHCAT strategies are extendedly dominated by SeHCAT 15% ( Table 42 ). When we look at the CEAC ( Figure 19 ), we again observe that, for small thresholds, no SeHCAT – treat IBS-D has the highest probability of being cost-effective while for larger thresholds (around £5000) SeHCAT 15% has the highest probability of being cost-effective.

TABLE 42 - Lifetime results for IBS-D population, scenario 4

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	14.8939	749
SeHCAT test – 5% cut-off	15.0432	1548	Extendedly dominated by SeHCAT 15%
SeHCAT test – 10% cut-off	15.0965	1727	Extendedly dominated by SeHCAT 15%
SeHCAT test – 15% cut-off	15.1003	1736	0.2064	987	4782

FIGURE 19.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 4.

Scenario 8 is the same as scenario 1, but now with utilities for IBS-D and BAM the same. As can be seen in Table 43 , SeHCAT 5% has a favourable ICER compared with no SeHCAT, and the ICER of SeHCAT 15% versus SeHCAT 5% may also be considered acceptable given a threshold of £30,000. The CEAC presented in Figure 20 shows that for lower threshold values no SeHCAT has the highest probability of being optimal while at a threshold of between £20,000 and £30,000, the three SeHCAT strategies have approximately the same probability of being optimal.

TABLE 43 - Lifetime results for IBS-D population, scenario 8

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	15.3034	819
SeHCAT test – 5% cut-off	15.334	1460	0.0306	642	20,976
SeHCAT test – 10% cut-off	15.3388	1610	Extendedly dominated by SeHCAT 15%
SeHCAT test – 15% cut-off	15.3408	1625	0.0068	165	24,265

FIGURE 20.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 8.

Figure 21 shows the population EVPI; from this we see that at a threshold of £30,000, the value of perfect information amounts to £20,000M.

FIGURE 21.

Population EVPI for lifetime IBS-D population, scenario 8.

Finally, scenario 12 is the same as scenario 5, but now with utilities for IBS-D and BAM the same. As can be seen in Table 44 , due to the higher utility for BAM responders, now SeHCAT 5% has a favourable ICER compared with no SeHCAT given a threshold of £30,000. The CEAC presented in Figure 22 shows that for lower threshold values no SeHCAT has the highest probability of being optimal while at a threshold of between £20,000 and £30,000, SeHCAT 5% has the highest probability of being optimal.

TABLE 44 - Lifetime results for IBS-D population, scenario 12

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	15.7042	1823	Dominated by no SeHCAT
SeHCAT test – 10% cut-off	15.7136	1820	Dominated by no SeHCAT
No SeHCAT test	15.7239	893
SeHCAT test – 5% cut-off	15.7539	1660	0.03	767	25,567

FIGURE 22.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 12.

Costs, effects and cost-effectiveness diarrhoea-predominant irritable bowel syndrome with trial of treatment included

Short-term results (decision tree) for diarrhoea-predominant irritable bowel syndrome with trial of treatment

Table 45 presents the probabilistic outcomes of the base-case decision tree for the IBS-D population when trial of treatment is also considered as comparator. Results are presented in order of increasing number of responders. From this, it becomes clear that the trial of treatment strategy dominates all other strategies: it leads to the highest number of responders for the lowest costs. Thus, no ICERs should be calculated. From the table we also notice that of the three SeHCAT scenarios, the 15% cut-off dominates the other two cut-offs.

TABLE 45 - Short-term results for IBS-D population: base case with trial of treatment

Strategy	Percentage of responders	Costs (per patient)	Incremental responders	Incremental costs	ICER
No SeHCAT test	0.5145	208	Dominated by trial of treatment
SeHCAT test – 5% cut-off	0.632	569	Dominated by trial of treatment
SeHCAT test – 10% cut-off	0.654	555	Dominated by trial of treatment
SeHCAT test – 15% cut-off	0.6596	553	Dominated by trial of treatment
No SeHCAT test – trial of treatment	0.6862	148

To investigate what the impact is of the various (statistical) uncertainties in the model, we also present the CEAC ( Figure 23 ). We notice that the trial of treatment strategy has the highest probability of being considered cost-effective over the whole range of possible threshold ICERs.

FIGURE 23.

Cost-effectiveness acceptability curve for short-term (cost per responder) results for IBS-D population, base case with trial of treatment.

As mentioned in the methods section, various assumptions regarding probabilities had to be made where neither literature nor expert opinion was available. Additionally, the costs of IBS-D treatment are very uncertain, as the experts had very differing responses and some uncertainty exists about the total costs of SeHCAT. Thus, we formulated several scenarios, which are listed in Table 46 . However, none of these scenarios led to any results different from the base case; in all cases trial of treatment was the dominant strategy. For all scenarios trial of treatment had the highest probability of being cost-effective; this probability was for most scenarios around 90%, decreasing for some scenarios to 50%.

TABLE 46 - Overview of scenarios explored for short-term IBS-D with trial of treatment

Scenario	SeHCAT service cost	Probability response to BAS with trial of treatment	Probability response to IBS-D treatment in SeHCAT-negative patients	Costs IBS-D treatment	Cost SeHCAT capsule	Results presented?
B1	0	0.28	0.62		195	Yes
B2	55	0.28	0.62		195	No, same as 1
B3	105	0.28	0.62		195	No, same as 1
B4	0	0.28	0.52		195	No, same as 1
B5	0	0.28	0.87		195	No, same as 1
B6	0	0.21	0.62		195	No, same as 1
B7	0	0.21	0.52		195	No, same as 1
B8	0	0.21	0.87		195	No, same as 1
B9	0	0.28	0.62	× 2	195	No, same as 1
B10	0	0.28	0.62	/2	195	No, same as 1
B11	0	0.28	0.52	/2	195	No
B12	0	0.21	0.52	/2	195	No, same as 1
B13	0	0.28	0.62		100	No, same as 1
B14	0	0.28	0.52		100	No, same as 1
B15	0	0.28	0.87		100	No, same as 1
B16	0	0.21	0.62		100	No, same as 1
B17	0	0.21	0.52		100	No, same as 1
B18	0	0.21	0.87		100	No, same as 1

Lifetime results for diarrhoea-predominant irritable bowel syndrome with trial of treatment

As in Lifetime results for diarrhoea-predominant irritable bowel syndrome, above, we have again explored a list of 14 scenarios regarding transition probabilities of the Markov models and the utility of the no diarrhoea health state. Again, we present only a limited number of these scenarios ( Table 47 ).

TABLE 47 - Overview of scenarios explored for the long-term analysis with trial of treatment

Scenario	IBS-D P(ND→D)	IBS-D P(D→ND)	BAM P(ND→D)	BAM P(D→ND)	Utility BAS no diarrhoea	Results presented?
1	0.05	0.05	0.05	0.05	BAS < IBS	Yes
2	0	0	0	0	BAS < IBS	Yes
3	0.05	0	0.05	0	BAS < IBS	No, same as 2
4	0.05	0	0	0	BAS < IBS	No, same as 2
5	0	0.05	0	0	BAS < IBS	No, same as 1
6	0	0	0.05	0	BAS < IBS	No, same as 1
7	0	0	0	0.05	BAS < IBS	Yes
8	0.05	0.05	0.05	0.05	BAS = IBS	Yes
9	0	0	0	0	BAS = IBS	No, same as 2
10	0.05	0	0.05	0	BAS = IBS	No, same as 3
11	0.05	0	0	0	BAS = IBS	No, same as 4
12	0	0.05	0	0	BAS = IBS	Yes
13	0	0	0.05	0	BAS = IBS	No, same as 6
14	0	0	0	0.05	BAS = IBS	No, same as 7

Below are the results for scenario 1. Scenario 1 assumes that all four transition probabilities are 5% per cycle. We observe that all SeHCAT strategies are dominated by no SeHCAT ( Table 48 ). When we look at the CEACs ( Figure 24 ), we notice that no SeHCAT has the highest probability of being cost-effective for the whole range of thresholds.

TABLE 48 - Lifetime results for IBS-D population, scenario 1 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	15.273	1624	Dominated by no SeHCAT
SeHCAT test – 10% cut-off	15.2744	1611	Dominated by no SeHCAT
No SeHCAT test – trial of treatment	15.2863	1211	Dominated by no SeHCAT
SeHCAT test – 5% cut-off	15.2925	1462	Dominated by no SeHCAT
No SeHCAT test	15.3006	818

FIGURE 24.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 1 with trial of treatment.

Figure 25 shows the population EVPI; from this we see that at a threshold of £30,000, the value of perfect information amounts to £10,000M.

FIGURE 25.

Population EVPI for lifetime IBS-D population, scenario 1 with trial of treatment.

Scenario 2 assumes that all four transition probabilities are 0% per cycle. We now observe ( Table 49 ) that the ICER of trial of treatment compared with no SeHCAT falls well below the commonly used £30,000 threshold, while all SeHCAT strategies are dominated by trial of treatment. When we look at the CEAC ( Figure 26 ), we notice that for small thresholds no SeHCAT has the highest probability of being cost-effective while for larger thresholds (around £5000) trial of treatment has the highest probability of being cost-effective.

TABLE 49 - Lifetime results for IBS-D population, scenario 2 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	15.3244	823
SeHCAT test – 10% cut-off	15.4127	1783	Dominated by trial of treatment
SeHCAT test – 5% cut-off	15.4133	1613	Dominated by trial of treatment
SeHCAT test – 15% cut-off	15.4191	1793	Dominated by trial of treatment
No SeHCAT test – trial of treatment	15.45	1459	0.1256	636	5064

FIGURE 26.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 2 with trial of treatment.

Scenario 7 assumes that the transition probability from D to ND in the BAM model is 5% per cycle while all others are 0. We observe that trial of treatment compared with no SeHCAT yields an ICER well below the common threshold of £30,000, as is the ICER of SeHCAT 15% compared with trial of treatment ( Table 50 ). When we look at the CEAC ( Figure 27 ), we again observe that for small thresholds no SeHCAT has the highest probability of being cost-effective, for thresholds between £5000 and £15,000 trial of treatment is optimal and for thresholds higher than £15,000 SeHCAT 15% has the highest probability of being cost-effective.

TABLE 50 - Lifetime results for IBS-D population, scenario 7 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	15.3192	822
SeHCAT test – 5% cut-off	15.4264	1643	Dominated by trial of treatment
No SeHCAT test – trial of treatment	15.4451	1459	0.1259	637	5060
SeHCAT test – 10% cut-off	15.4653	1894	Extendedly dominated by SeHCAT 15%
SeHCAT test – 15% cut-off	15.4797	1919	0.0346	460	13,295

FIGURE 27.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 7 with trial of treatment.

Figure 28 shows the population EVPI; from this we see that at a threshold of £30,000, the value of perfect information amounts to £15,000M.

FIGURE 28.

Population EVPI for lifetime IBS-D population, scenario 7 with trial of treatment.

Scenario 8 is the same as scenario 1, but now with utilities for IBS-D and BAM the same. As can be seen in Table 51 , now trial of treatment has a favourable ICER compared with no SeHCAT, while all the SeHCAT strategies are dominated by trial of treatment. The CEAC presented in Figure 29 shows that for lower threshold values no SeHCAT has the highest probability of being optimal while for thresholds above £10,000 trial of treatment has the highest probability of being optimal.

TABLE 51 - Lifetime results for IBS-D population, scenario 8 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	15.3034	819
SeHCAT test – 5% cut-off	15.334	1460	Dominated by trial of treatment
SeHCAT test – 10% cut-off	15.3388	1610	Dominated by trial of treatment
SeHCAT test – 15% cut-off	15.3408	1625	Dominated by trial of treatment
No SeHCAT test – trial of treatment	15.3412	1210	0.0378	391	10,344

FIGURE 29.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 8 with trial of treatment.

Figure 30 shows the population EVPI; from this we see that at a threshold of £30,000, the value of perfect information amounts to £15,000M.

FIGURE 30.

Population EVPI for lifetime IBS-D population, scenario 8 with trial of treatment.

Finally, scenario 12 is the same as scenario 5, but now with utilities for IBS-D and BAM the same. As can be seen in Table 52 , due to the higher utility for BAM responders, now trial of treatment has a favourable ICER compared with no SeHCAT given a threshold of £30,000. The CEAC presented in Figure 31 shows that for lower threshold values no SeHCAT has the highest probability of being optimal while for thresholds above £10,000 trial of treatment has the highest probability of being optimal.

TABLE 52 - Lifetime results for IBS-D population, scenario 12 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	15.7042	1823	Dominated by no SeHCAT
SeHCAT test – 10% cut-off	15.7136	1820	Dominated by no SeHCAT
No SeHCAT test	15.7239	893
SeHCAT test – 5% cut-off	15.7539	1660	Dominated by trial of treatment
No SeHCAT test – trial of treatment	15.7956	1508	0.0717	615	8577

FIGURE 31.

Cost-effectiveness acceptability curve for lifetime IBS-D population, scenario 12 with trial of treatment.

Costs, effects and cost-effectiveness Crohn’s without ileal resection

Short-term results for Crohn’s disease (decision tree)

Table 53 presents the probabilistic outcomes of the decision tree for the Crohn’s disease population. Results are presented in order of increasing number of responders. From this, it becomes clear that testing with SeHCAT leads to more costs and more responders. Note that the ICER is in costs per additional responder and, thus, we cannot compare it with the usual threshold.

TABLE 53 - Short-term results for Crohn’s population, base case (scenario C1)

Strategy	Percentageof responders	Expected costs	Incremental responders	Incremental costs	ICER
No SeHCAT test – treat Crohn’s	0.6226	327
SeHCAT test – 15% cut-off	0.7296	573	0.1069	245	2293
SeHCAT test – 10% cut-off	0.7343	592	0.0046	19	4134

Figure 32 presents the CEACs. We notice that for smaller thresholds no SeHCAT is the strategy with the highest probability of being cost-effective whereas for larger thresholds, both SeHCAT strategies are equally likely to have the highest probability of being cost-effective.

FIGURE 32.

Cost-effectiveness acceptability curve for short-term (cost per responder) results for Crohn’s population, base case (scenario C1).

As for the IBS-D population we had to make various assumptions regarding probabilities where neither literature nor expert opinion was available. Additionally, the costs of treating Crohn’s patients with chronic diarrhoea are very uncertain, as the experts had very differing responses and some uncertainty exists about the total costs of SeHCAT.

These assumptions were explored through scenario analysis. The list of scenarios analysed is presented in Table 54 .

TABLE 54 - Overview of scenarios explored for short-term Crohn’s

Scenario	SeHCAT service cost	Probability response to Crohn’s treatment in SeHCAT-negative patients	Costs Crohn’s treatment	Cost SeHCAT capsule	Results presented?
C1	0	0.72		195	Yes
C2	55	0.72		195	No
C3	105	0.72		195	No
C4	0	0.62		195	No
C5	0	0.97		195	Yes
C6	0	0.72	× 2	195	No
C7	0	0.72	/2	195	No
C8	0	0.62	/2	195	No
C9	0	0.72		100	No
C10	0	0.62		100	No
C11	0	0.97		100	No

We found that increasing or decreasing costs related to SeHCAT did not alter the results. For most scenarios, the CEACs were close to the base-case CEAC, except for scenarios C5 and C11. These scenarios have a near-perfect response to Crohn’s treatment in SeHCAT-negative patients. The main difference that is observed in the CEAC for that scenario ( Figure 33 ) is that SeHCAT 15% and SeHCAT 10% are no longer equally likely to be the most cost-effective strategy. This is explained by the fact that in SeHCAT 10% more patients have a negative test result and then receive Crohn’s treatment, with a 97% response rate; thus, this strategy becomes more favourable ( Table 55 ).

TABLE 55 - Short-term (cost per responder) results for Crohn’s population, response to Crohn’s treatment in SeHCAT-negative patients 97% (scenario C5)

Strategy	Percentage of responders	Expected costs	Incremental responders	Incremental costs	ICER
No SeHCAT test – treat Crohn’s	0.6144	324
SeHCAT test – 15% cut-off	0.8108	574	Extendedly dominated by SeHCAT 10%
SeHCAT test – 10% cut-off	0.8366	593	0.2222	268	1208

FIGURE 33.

Cost-effectiveness acceptability curve for short-term (cost per responder) results for Crohn’s population, response to Crohn’s treatment in SeHCAT-negative patients 97% (scenario C5).

Lifetime results for Crohn’s population

As mentioned in the methods section, no information was available to estimate the transition probabilities in the Markov models other than the all-cause mortality rates. Thus, we formulated various scenarios with different transition probabilities for ND-BAM → D-BAM, D-BAM → ND-BAM, ND-Crohn → D-Crohn, D-Crohn → ND-Crohn. Additionally, we have two scenarios regarding the utility in the no diarrhoea health state; either this is the same for BAS and Crohn’s or utility of BAS responders is lower than for Crohn’s responders. Table 56 presents an overview of all scenarios we explored. As many yielded the same results, we only present a selection of the scenarios, as indicated in the table.

TABLE 56 - Overview of scenarios explored for the long-term analysis, no trial of treatment

Scenario	Crohn P(ND → D)	Crohn P(D → ND)	BAM P(ND → D)	BAM P(D → ND)	Utility BAS no diarrhoea	Results presented?
1	0.05	0.05	0.05	0.05	BAS < Crohn	Yes
2	0	0	0	0	BAS < Crohn	Yes
3	0.05	0	0.05	0	BAS < Crohn	No, same as 2
4	0.05	0	0	0	BAS < Crohn	Yes
5	0	0.05	0	0	BAS < Crohn	No, same as 1
6	0	0	0.05	0	BAS < Crohn	No, same as 1
7	0	0	0	0.05	BAS < Crohn	No, same as 4
8	0.05	0.05	0.05	0.05	BAS = Crohn	Yes
9	0	0	0	0	BAS = Crohn	No, same as 8
10	0.05	0	0.05	0	BAS = Crohn	No, same as 8
11	0.05	0	0	0	BAS = Crohn	No, same as 4
12	0	0.05	0	0	BAS = Crohn	Yes
13	0	0	0.05	0	BAS = Crohn	No, same as 6
14	0	0	0	0.05	BAS = Crohn	No, same as 4

Below are the results for scenario 1. Scenario 1 assumes that all four transition probabilities are 5% per cycle. We observe that no SeHCAT is the most favourable strategy, with not only the highest costs but also the highest effect ( Table 57 ). When we look at the CEACs ( Figure 34 ), we notice that no SeHCAT has the highest probability of being cost-effective for thresholds above £5000.

TABLE 57 - Lifetime results for Crohn’s population, scenario 1

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	14.0847	2710
SeHCAT test – 10% cut-off	14.0994	2828	Extendedly dominated by no SeHCAT test
No SeHCAT test	14.1571	3023	0.0724	313	4323

FIGURE 34.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 1.

Figure 35 shows the patient EVPI; from this we see that at a threshold of £30,000, the value of perfect information amounts to £100.

FIGURE 35.

Patient EVPI for Crohn’s population, scenario 1.

Scenario 2 assumes that all four transition probabilities are 0% per cycle. We now observe ( Table 58 ) that the ICER of SeHCAT 10% compared with SeHCAT 15% falls well below the commonly used £30,000 threshold and that SeHCAT 10% dominates no SeHCAT. When we look at the CEAC ( Figure 36 ), we notice that for small thresholds SeHCAT 15% has the highest probability of being cost-effective while for larger thresholds (above £8000) SeHCAT 10% has the highest probability of being cost-effective. However, for no SeHCAT this probability is almost the same as for SeHCAT 10%.

TABLE 58 - Lifetime results for Crohn’s population, scenario 2

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	14.2517	3196
No SeHCAT test	14.2635	3400	Dominated by SeHCAT 10%
SeHCAT test – 10% cut-off	14.2712	3336	0.0195	140	7179

FIGURE 36.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 2.

Scenario 4 assumes that the transition probability from ND to D in the Crohn’s model is 5% per cycle while all others are 0. We observe that SeHCAT 15% compared with no SeHCAT yields an ICER well below the common threshold of £30,000, while SeHCAT 10% is dominated by SeHCAT 15% ( Table 59 ). When we look at the CEAC ( Figure 37 ), we observe that for small thresholds no SeHCAT has the highest probability of being cost-effective while for larger thresholds (around £5000) SeHCAT 15% has the highest probability of being cost-effective.

TABLE 59 - Lifetime results for Crohn’s population, scenario 4

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	13.8169	1972
SeHCAT test – 10% cut-off	14.042	2596	Dominated by SeHCAT 15%
SeHCAT test – 15% cut-off	14.0657	2595	0.2488	623	2504

FIGURE 37.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 4.

Scenario 8 is the same as scenario 1, but now with utilities for Crohn’s and BAM the same. As can be seen in Table 60 , SeHCAT 15% dominates no SeHCAT, and the ICER of SeHCAT 10% versus SeHCAT 15% would be considered too high given a threshold of £30,000. The CEAC presented in Figure 38 shows that SeHCAT 15% has the highest probability of being optimal over the whole range of thresholds. However, the curves of SeHCAT 15% and SeHCAT 10% are quite close, thus leading to more decision uncertainty. This is reflected in the EVPI ( Figure 39 ), which at a threshold of £30,000 amounts to £400 per patient.

TABLE 60 - Lifetime results for Crohn’s population, scenario 8

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	14.1592	3031	Dominated by SeHCAT 15%
SeHCAT test – 15% cut-off	14.1865	2708
SeHCAT test – 10% cut-off	14.1876	2828	0.0011	119	108,412

FIGURE 38.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 8.

FIGURE 39.

Patient EVPI for Crohn’s population, scenario 8.

Finally, scenario 12 assumes that the transition probability from D to ND in the Crohn’s model is 5% per cycle while all others are 0 and that the utilities for IBS-D and BAM are the same. As can be seen in Table 61 , here no SeHCAT can be considered the most favourable strategy given a threshold of £30,000. The CEAC presented in Figure 40 shows that for lower threshold values (below £7000) SeHCAT 15% has the highest probability of being optimal, for thresholds between £7000 and £14,000 SeHCAT 10% has the highest probability while for thresholds above £14,000 no SeHCAT has the highest probability of being optimal.

TABLE 61 - Lifetime results for Crohn’s population, scenario 12

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	14.4624	3426
SeHCAT test – 10% cut-off	14.4917	3639	0.0293	213	7282
No SeHCAT test	14.5362	4246	0.0445	606	13,621

FIGURE 40.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 12.

Figure 41 shows the patient EVPI; from this we see that at a threshold of £30,000, the value of perfect information amounts to £400.

FIGURE 41.

Patient EVPI for Crohn’s population, scenario 12.

Costs, effects and cost-effectiveness for Crohn’s disease without ileal resection with trial of treatment included

Short-term results for Crohn’s disease (decision tree) with trial of treatment

Table 62 presents the probabilistic outcomes of the base-case decision tree for the Crohn’s population when trial of treatment is also considered as comparator. Results are presented in order of increasing number of responders. From this, it becomes clear that the trial of treatment strategy dominates all other strategies: it leads to the highest number of responders for the lowest costs. Thus, no ICERs should be calculated.

TABLE 62 - Short-term results for Crohn’s population, base case with trial of treatment (scenario D1)

Strategy	Percentage of responders	Expected costs	Incremental responders	Incremental costs	ICER
No SeHCAT test – treat Crohn’s	0.6226	327	Dominated by trial of treatment
SeHCAT test – 15% cut-off	0.7296	573	Dominated by trial of treatment
SeHCAT test – 10% cut-off	0.7343	592	Dominated by trial of treatment
No SeHCAT test – trial of treatment	0.8232	171

To investigate what the impact is of the various (statistical) uncertainties in the model, we also present the CEAC ( Figure 42 ). We notice that the trial of treatment strategy has the highest probability of being considered cost-effective over the whole range of possible threshold ICERs.

FIGURE 42.

Cost-effectiveness acceptability curve for short-term (cost per responder) results for Crohn’s population, base case with trial of treatment (scenario D1).

As for the IBS-D population, we formulated several scenarios to explore the sensitivity of the model for various major assumptions we made. The scenarios are listed in Table 63 . None of these scenarios led to any result other than the base case; in all cases trial of treatment was the dominant strategy.

TABLE 63 - Overview of scenarios explored for short-term Crohn’s with trial of treatment

Strategy	SeHCAT service cost	Probability response to BAS with trial of treatment	Probability response to Crohn’s treatment in SeHCAT-negative patients	Costs Crohn’s treatment	Cost SeHCAT capsule	Results presented?
D1	0	0.46	0.72		195	Yes
D2	55	0.46	0.72		195	No, same as 1
D3	105	0.46	0.72		195	No, same as 1
D4	0	0.46	0.62		195	No, same as 1
D5	0	0.46	0.87		195	No, same as 1
D6	0	0.28	0.72		195	No, same as 1
D7	0	0.28	0.62		195	No, same as 1
D8	0	0.28	0.87		195	No, same as 1
D9	0	0.46	0.72	× 2	195	No, same as 1
D10	0	0.46	0.72	/2	195	No, same as 1
D11	0	0.46	0.62	/2	195	No, same as 1
D12	0	0.28	0.62	/2	195	No, same as 1
D13	0	0.46	0.72		100	No, same as 1
D14	0	0.46	0.62		100	No, same as 1
D15	0	0.46	0.87		100	No, same as 1
D16	0	0.28	0.72		100	No, same as 1
D17	0	0.28	0.62		100	No, same as 1
D18	0	0.28	0.87		100	No, same as 1

Lifetime results for Crohn’s with trial of treatment

As above (see Lifetime results for Crohn’s), we have again explored a list of 14 scenarios regarding transition probabilities of the Markov models and the utility of the no diarrhoea health state. Again, we present only a limited number of these scenarios ( Table 64 ).

TABLE 64 - Overview of scenarios explored for the long-term analysis with trial of treatment

Scenario	Crohn P(ND→D)	Crohn P(D→ND)	BAM P(ND→D)	BAM P(D→ND)	Utility BAS no diarrhoea	Results presented?
1	0.05	0.05	0.05	0.05	BAS < Crohn	Yes
2	0	0	0	0	BAS < Crohn	Yes
3	0.05	0	0.05	0	BAS < Crohn	No, same as 2
4	0.05	0	0	0	BAS < Crohn	No, same as 2
5	0	0.05	0	0	BAS < Crohn	Yes
6	0	0	0.05	0	BAS < Crohn	No, same as 5
7	0	0	0	0.05	BAS < Crohn	No, same as 3
8	0.05	0.05	0.05	0.05	BAS = Crohn	No, same as 2
9	0	0	0	0	BAS = Crohn	No, same as 2
10	0.05	0	0.05	0	BAS = Crohn	No, same as 3
11	0.05	0	0	0	BAS = Crohn	No, same as 4
12	0	0.05	0	0	BAS = Crohn	No, same as 2
13	0	0	0.05	0	BAS = Crohn	No, same as 6
14	0	0	0	0.05	BAS = Crohn	Yes

Below are the results for scenario 1. Scenario 1 assumes that all four transition probabilities are 5% per cycle. We observe that both SeHCAT strategies are dominated by trial of treatment, while no SeHCAT can be considered cost-effective compared with trial of treatment for thresholds above £14,000 ( Table 65 ). The CEAC confirms this assessment ( Figure 43 ).

TABLE 65 - Lifetime results for Crohn’s population, scenario 1 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	14.0847	2710	Dominated by trial of treatment
SeHCAT test – 10% cut-off	14.0994	2828	Dominated by trial of treatment
No SeHCAT test – trial of treatment	14.1227	2545
No SeHCAT test	14.1571	3023	0.0343	477	13,889

FIGURE 43.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 1 with trial of treatment.

Figure 44 shows the patient EVPI; from this we see that at a threshold of £30,000, the value of perfect information amounts to £300.

FIGURE 44.

Patient EVPI for Crohn’s population, scenario 1 with trial of treatment.

Scenario 2 assumes that all four transition probabilities are 0% per cycle. We now observe ( Table 66 ) that no SeHCAT and SeHCAT 10% are dominated by trial of treatment and the ICER of trial of treatment compared with SeHCAT 15% falls well below the commonly used £30,000 threshold. When we look at the CEAC ( Figure 45 ), we notice that for almost the whole range of thresholds trial of treatment has the highest probability of being cost-effective, up to even 90% for thresholds above £10,000.

TABLE 66 - Lifetime results for Crohn’s population, scenario 2 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	14.2517	3196
No SeHCAT test	14.2635	3400	Dominated by trial of treatment
SeHCAT test – 10% cut-off	14.2712	3336	Dominated by trial of treatment
No SeHCAT test – trial of treatment	14.3572	3213	0.1055	17	161

FIGURE 45.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 2 with trial of treatment.

Scenario 5 assumes that the transition probability from D to ND in the Crohn’s model is 5% per cycle while all others are 0. We observe that SeHCAT 15% compared with no SeHCAT yields an ICER well below the common threshold of £30,000, while SeHCAT 10% is dominated by SeHCAT 15% ( Table 67 ). When we look at the CEAC ( Figure 46 ), we observe that for lower threshold values (below approximately £1000) SeHCAT 15% has the highest probability of being optimal, for thresholds between £1000 and £13,000 trial of treatment has the highest probability while for thresholds above £13,000 no SeHCAT has the highest probability of being optimal.

TABLE 67 - Lifetime results for Crohn’s population, scenario 5 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
SeHCAT test – 15% cut-off	14.326	3434
SeHCAT test – 10% cut-off	14.3689	3646	Dominated by trial of treatment
No SeHCAT test – trial of treatment	14.4834	3612	0.1574	178	1131
No SeHCAT test	14.5341	4250	0.0506	637	12,581

FIGURE 46.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 5 with trial of treatment.

Scenario 14 assumes that the transition probability from D to ND in the BAM model is 5% per cycle while all others are 0 and that utilities per health state are the same in the Crohn’s model and the BAM model. In this scenario we observe ( Table 68 ) that no SeHCAT and SeHCAT 10% are dominated by trial of treatment and the ICER of SeHCAT 15% compared with trial of treatment falls well below the commonly used £30,000 threshold. When we look at the CEAC ( Figure 47 ), we notice that for thresholds up to £10,000 trial of treatment has the highest probability of being cost-effective, whereas for higher thresholds SeHCAT 15% has the highest probability of being cost-effective, up to 50% for thresholds above £20,000.

TABLE 68 - Lifetime results for Crohn’s population, scenario 14 with trial of treatment

Strategy	Expected QALYs	Expected costs	Incremental QALYs	Incremental costs	ICER
No SeHCAT test	14.2627	3379	Dominated by trial of treatment
SeHCAT test – 10% cut-off	14.4892	3490	Dominated by trial of treatment
No SeHCAT test – trial of treatment	14.4943	3197
SeHCAT test – 15% cut-off	14.5145	3399	0.0201	202	10,020

FIGURE 47.

Cost-effectiveness acceptability curve for lifetime Crohn’s population, scenario 14 with trial of treatment.

Clinical effectiveness

Three studies were reasonably reliable in assessing the relationship between the SeHCAT test and treatment with cholestyramine. 39,42,43 However, the studies had small numbers of patients with unknown cause chronic diarrhoea, they used different cut-offs for the assessment of BAM and between-study heterogeneity was considerable.

Sensitivity ranged from 0.67 (at a cut-off of 8%) to 1.00 (at a cut-off of 15%) and specificity ranged from 0.91 (at a cut-off of 15%) to 1.00 (at a cut-off of 5%) ( Table 69 ).

None of the studies looking specifically at people with Crohn’s disease presented reliable data for the prediction of response to treatment with BAS because no data were presented for people with a negative SeHCAT test in the two studies.

One RCT in patients with IBS-D which compared treatment with BAS (colesevelam) with placebo showed no significant differences in terms of colonic transit [e.g. geometric centre at 48 hours: MD = 0.18 (95% CI –0.29 to 0.65)] bowel function [e.g. stool frequency per day: MD = –0.11 (95% CI –1.01 to 0.79)] or adverse events [e.g. uterine cramps: OR = 0.45 (95% CI 0.04 to 5.81)]. However, randomisation (sequence generation and allocation concealment) was not adequately reported and groups were small (n = 12 in both arms).

For people with chronic diarrhoea, 19 studies provide data on the clinical effectiveness of BAS given a positive SeHCAT test; three studies also provide data on the effectiveness of BAS given a negative SeHCAT test. For those with a positive SeHCAT test response rates were on average 85%, 73% and 72% for cut-offs at 5%, 10% and 15%, respectively. For those with a negative SeHCAT test the response rate was 14% at a cut-off of 5% and 0% at a cut-off of 15%. For people with Crohn’s disease and a positive SeHCAT test the response rate was 95% at a cut-off of 5% and 86% or 89% at a cut-off of 15%.

Several studies reported data on the accuracy of SeHCAT for the detection of BAM in people with chronic diarrhoea ( Table 70 ). However, all of these studies are flawed because they either do not use an acceptable reference standard or include a population not in line with the scope (healthy volunteers or people with ileal resection). Therefore, these data are not used in this review.

Cost-effectiveness

In Chapter 4 we assessed the cost-effectiveness of SeHCAT in two different populations. The first is the population of patients with chronic diarrhoea with unknown cause and symptoms suggestive of IBS-D and the second population concerns patients with Crohn’s disease without ileal resection with chronic diarrhoea.

For both populations the cost-effectiveness of SeHCAT compared with no SeHCAT was assessed. For the SeHCAT option we defined various strategies based on the test cut-off points used to classify patients. For the IBS-D patient population data were available to be able to distinguish between cut-off points of 5%, 10% and 15%. For the Crohn’s patient population, only data on a 10% and 15% SeHCAT test cut-off were available. For the no SeHCAT strategy all patients received regular treatment for either IBS-D or chronic diarrhoea in Crohn’s. Additionally, in the scoping document, trial of treatment with BAS was mentioned as another possible strategy without specifically including it as a comparator. According to the clinical experts at the scoping meeting, trial of treatment is rarely used as a treatment strategy and was thus not considered relevant. However, trial of treatment could also not be completely excluded as an option. Thus, in this report, for both populations we presented two sets of results: one where trial of treatment is not considered as a comparator and one where it is. In the trial of treatment strategy, patients first receive a BAS and when patients do not respond they receive regular treatment for either IBS-D or chronic diarrhoea in Crohn’s.

For each population, three models were combined:

1. a short-term decision tree that models the diagnostic pathway and initial response to treatment (first 6 months)

2. a long-term Markov model that estimates the life time costs and effects for patients initially receiving BAS

3. a long-term Markov model that estimates the life time costs and effects for patients initially receiving regular treatment (IBS-D treatment in the first population and Crohn’s treatment in the second population).

In the decision tree the 6-month number of responders and the expected costs were calculated for each comparator while for the Markov models lifetime expected QALYs and expected costs per patient were calculated for each comparator.

Where possible, input for the model was based on our SeHCAT systematic review, other published literature and UK databases. When such evidence was not available, expert opinion was used. The impact of uncertainty about the various input parameters on the outcomes was explored through probabilistic sensitivity analyses and scenario analyses.

Incremental cost-effectiveness ratios were estimated as additional cost per additional responder in the short term (first 6 months) and per additional QALY in the long term (lifetime).

When trial of treatment is not considered as a comparator, the evaluation for the IBS-D population showed, for the short term, that the optimal choice depends on the willingness to pay for an additional responder. For lower values (base case £2400, scenarios between £1800 and £4600) the choice will be no SeHCAT in all scenarios; for higher values either SeHCAT 10% or SeHCAT 15% becomes cost-effective.

For the lifetime perspective, we did not define a base case, as we had no information of any kind to inform the transition probabilities between the health states diarrhoea and no diarrhoea and vice versa. Thus, only scenario analysis was performed. The various scenarios showed widely differing results. Depending on the scenario, in the threshold range of £20,000–30,000 per QALY gained, we found as optimal choice either no SeHCAT, SeHCAT 5% or SeHCAT 15%; only SeHCAT 10% never had the highest probability of being cost-effective.

When trial of treatment is considered a comparator, the analysis showed for the IBS-D population that for the short term, trial of treatment is the optimal choice across a range of scenarios. In the base-case scenario, trial of treatment dominated all other strategies and had a 95% probability of being the most cost-effective option. In the various scenarios, trial of treatment was dominant compared with all strategies. For all scenarios trial of treatment had the highest probability of being cost-effective; this probability was for most scenarios around 90%, decreasing for some scenarios to 50%.

For the lifetime perspective, for all but two scenarios, trial of treatment was the strategy with the highest probability of being cost-effective for thresholds above £5000 to £15,000, with no SeHCAT the most favourable strategy for lower threshold ICERs. In the two scenarios where the transition probability from diarrhoea to no diarrhoea in the BAM model was 5% per cycle while all others were 0, we observed that for thresholds higher than £15,000 SeHCAT 15% had the highest probability of being cost-effective.

For the Crohn’s population, the short-term evaluation without trial of treatment as comparator showed for the short term that the optimal choice depends on the willingness to pay for an additional responder. For lower values (base case £2300, scenarios between £1500 and £4000) the choice will be no SeHCAT in all scenarios, for higher values SeHCAT 10% becomes cost-effective.

For the long term, the various scenarios show widely differing results. Depending on the scenario, in the threshold range of £20,000–30,000 per QALY gained, we found as optimal choice either no SeHCAT or SeHCAT 15% while some scenarios found that all three strategies had the same probability of being cost-effective.

When trial of treatment is considered a comparator, the analysis showed for the Crohn’s population that for the short term, trial of treatment dominated all other strategies (in terms of number of responders) and had an almost 100% probability of being the most cost-effective option. In the various scenarios, trial of treatment was dominant compared with all strategies. For all scenarios trial of treatment had the highest probability of being cost-effective; for most scenarios this probability was around 90%.

For the lifetime perspective in Crohn’s with trial of treatment, again the various scenarios show widely differing results. Depending on the scenario, in the threshold range of £20,000–30,000 per QALY gained, we found as optimal choice either trial of treatment, no SeHCAT or SeHCAT 15%.

In conclusion, the various analyses have shown that for both populations considerable decision uncertainty exists, and that no firm conclusions can be formulated as to which strategy is optimal.

Clinical effectiveness

A main strength for this assessment is the fact that the SeHCAT test has been long established; it has been used for over 25 years in clinical practice. The adverse events profile of the test seems favourable. Unfortunately, there are no published data to support this statement.

Extensive literature searches were conducted in an attempt to maximise retrieval of relevant studies. These included electronic searches of a variety of bibliographic databases, as well as screening of clinical trials registers and conference abstracts to identify unpublished studies. Because of the known difficulties in identifying test accuracy studies using study design-related search terms,87 no study design filters were used in order to maximise sensitivity at the expense of reduced specificity. Thus, large numbers of citations were identified and screened, many of which did not meet the inclusion criteria of the review.

The possibility of publication bias remains a potential problem for all systematic reviews. Considerations may differ for systematic reviews of test accuracy studies. It is relatively simple to define a positive result for studies of treatment, e.g. a significant difference between the treatment and control groups that favours treatment. This is not the case for test accuracy studies, which measure agreement between index test and reference standard. It would seem likely that studies finding greater agreement (high estimates of sensitivity and specificity) will be published more often. In addition, test accuracy data are often collected as part of routine clinical practice, or by retrospective review of records; test accuracy studies are not subject to the formal registration procedures applied to RCTs and are therefore more easily discarded when results appear unfavourable. The extent to which publication bias occurs in studies of test accuracy remains unclear; however, simulation studies have indicated that the effect of publication bias on meta-analytic estimates of test accuracy is minimal. 88 Formal assessment of publication bias in systematic reviews of test accuracy studies remains problematic and reliability is limited. 88 We did not undertake a statistical assessment of publication bias in this review. However, our search strategy included a variety of routes to identify unpublished studies and resulted in the inclusion of a number of conference abstracts.

Clear inclusion criteria were specified in the protocol for this review and the one protocol modification that occurred during the assessment has been documented in the methods section (see Chapter 3, Assessment of clinical effectiveness, Methods) of this report. The eligibility of studies for inclusion is therefore transparent. In addition, we have provided specific reasons for excluding all of the studies considered potentially relevant at initial citation screening (see Appendix 5 ). The review process followed recommended methods to minimise the potential for error and/or bias;14 studies were independently screened for inclusion by two reviewers and data extraction and quality assessment were done by one reviewer and checked by a second (RR and SD). Any disagreements were resolved by consensus. The only French-language study was extracted by one reviewer and checked by a second (Shona Lang and RR).

Three studies included in the review were test accuracy studies. The methodological quality of these studies was assessed using a modification of the QUADAS-2 tool. 25 The QUADAS tool has been recommended for assessing the methodological quality of test accuracy studies,14,17 and has been widely adopted by researchers and key organisations such as the Cochrane Collaboration, NICE in the UK, and Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG) in Germany. It has been mentioned in more than 200 abstracts on the DARE database and has been cited more than 500 times. The revised version of QUADAS (QUADAS-2) has recently been published. 25 QUADAS-2 more closely resembles the approach and structure of the Cochrane risk of bias tool. It is structured into four key domains covering participant selection, index test, reference standard, and the flow of patients through the study (including timing of tests). Each domain is rated for risk of bias (low, high or unclear) and the tool provides signalling questions, in each domain, to help reviewers in reaching a judgement. The participant selection, index test and reference standard domain are also, separately, rated for concerns regarding the applicability of the study to the review question (low, high or unclear). However, the QUADAS-2 tool does not currently include domains specific to the assessment of studies comparing multiple index tests; further development of QUADAS-2 in this area is planned. This assessment used a modified version of the QUADAS-2 tool, which includes an additional domain for the comparator test and additional signalling questions in the ‘flow and timing’ domain. It should be noted, however, that these components of the tool were not developed using the same rigorous evidence-based approach as the core QUADAS-2 tool. The inclusion criteria for this review were considered to largely match the review question and questions of applicability were, therefore, only relevant to the ‘patient selection’ domain. The review-specific guidance used in our QUADAS-2 assessment is reported in Appendix 2 . The results of the risk of bias assessment are reported, in full, for all included studies (see Appendix 3 ) and in summary in the results section (see Chapter 3, Assessment of clinical effectiveness, Results). However, the usefulness of this assessment was limited by poor reporting of primary study methods, particularly with respect to how the index and comparator tests and the reference standard were applied.

The main limitations for this assessment are the lack of data, differences between studies included in the review, and the generally poor quality of included papers. As reported in Chapter 3 (see Assessment of clinical effectiveness,Results), most studies did not report information on the probability of a response to treatment with BAS for people with a negative SeHCAT test. Therefore, information on the accuracy of the SeHCAT test to predict a treatment response is largely missing. In addition, there were considerable differences between studies; the principal diagnosis, treatment dose, the definition of response, follow-up period and SeHCAT administration were different between trials (see Appendix 4 for details).

Three studies were included for the assessment of the accuracy of a SeHCAT test to predict a response to treatment. All three studies were published between 1985 and 1987; two were Italian studies and one was a UK study. It is unclear whether or not there was any overlap in study populations included in the two Italian studies, which were performed by the same group.

Cost-effectiveness

Searching the literature, no economic evaluation studies were found regarding the use of SeHCAT-testing in chronic diarrhoea patients. This report therefore presents, as far as we are aware, the first full economic evaluation study, both in (a) patients presenting with chronic diarrhoea with unknown cause and symptoms suggestive of functional disease (IBS-D) and (b) people with Crohn’s disease and chronic diarrhoea with unknown cause (i.e. without resection of the terminal ileum) assessing the short- and long-term consequences both in costs and effects of using SeHCAT. For this purpose, for each of these patient populations, a linked evidence approach was used in modelling cost and consequences, combining outcomes of the diagnostic test and the related changes in treatment decisions and final health outcomes. A clear distinction was made between the initial diagnostic phase (treatment responder vs. non-responder at 6 months) and the long-term projection of this intermediate outcome of the diagnostic phase into final health outcomes (lifetime costs and consequences, the latter presented in QALYs).

The available evidence regarding the cut-off values defining a positive and negative SeHCAT test shows that various cut-off values are used, which influences test-accuracy estimates expressed in BAS treatment response. In our approach we assessed the impact of different cut-off values explicitly by making a distinction between SeHCAT test strategies based on three levels: 5%; 10% and 15%. Furthermore, current clinical practice shows that no-SeHCAT testing implies different clinical approaches. Two major approaches were explicitly modelled: first, IBS-D treatment in all patients involved, and secondly, the trial of treatment approach, in which all patients initially receive BAS and within the first period of 6 months (the time horizon of the diagnostic decision tree) non-responders will switch from BAS to IBS-D treatment. Our methods made it possible to analyse and present the cost-effectiveness of all competing five strategies simultaneously.

We included the impact of using SeHCAT in terms of BAS-treatment response as reported in peer reviewed papers and selected for this purpose only those papers that fulfilled our quality criteria as presented in Chapter 3 . In all models presented (both the decision trees that represented the diagnostic phase and the Markov models represented the subsequent life-years of patients) we have used evidence to inform parameters that was UK relevant and as up to date and high quality as possible. Where evidence was not available through published studies or databases, we used the most likely and plausible values and ranges as reported by clinical experts. However, given the time constraints, no formal expert solicitation process could be conducted and instead a questionnaire was sent out. The seven experts who filled out the questionnaire showed a wide variation in their responses, highlighting the uncertainty about important input parameters.

The remaining lack of evidence of certain parameters was handled by performing extensive uncertainty analyses, both using probabilistic sensitivity analyses and running a wide range of scenario analyses. For the latter, because the plausibility of one of the scenarios was not larger over any other, we did not judge about the likelihood of any of them. Additionally, given the discussion on the acceptability of trial of treatment with BAS as a comparator, we also ran all analyses once with trial of treatment as comparator and once without.

One of the main limitations of the study is that the studies used to estimate the probability of a positive SeHCAT test and the probability of a BAS response were based on other populations than the ones defined in this evaluation. Most, if not all, IBS-D studies included patients in whom various tests had been performed and where no organic cause of the diarrhoea could be found. This is in contrast with the population defined here, which is patients with symptoms suggestive of functional disease in whom only basic blood tests have been performed. It is not unlikely that in our population the prevalence of BAM is lower than that observed in the published studies.

For the Crohn’s population, we combined the estimates from Tunney et al. 32 and Smith et al. 3 The Tunney paper does not provide much detail on the population of surgically naïve Crohn’s patients, so it is unclear whether or not this group would be representative of the population under investigation here. The study by Smith was specifically in patients who were in clinical remission, thus excluding patients having a relapse. It is unclear how this would impact the results.

Another limitation concerns the modelling of the IBS-D population. It is assumed in the model that patients not responding to IBS-D treatment will only use loperamide for some symptomatic relief. It is, however, not unlikely that in our IBS-D population (i.e. patients in whom no diagnostic testing other than initial blood work has been performed) some non-responders will be referred for diagnostic testing to check for organic causes of the chronic diarrhoea. However, the discussion among the experts during the scoping meeting suggested that SeHCAT testing would not make subsequent tests for organic disease redundant in which case these costs would be the same in all strategies.

We also did not incorporate patients switching from treatment with cholestyramine to colesevelam because of non-response or intolerability, as no data on response rates for colesevelam were available. Including that option would most likely lead to a higher response rate, a smaller percentage of patients moving from the ‘no diarrhoea’ health state to the ‘diarrhoea’ health state over time (due to increased persistence) and to higher treatment costs since colesevelam is more expensive than cholestyramine.

However, the most important limitation is the lack of data on various important parameters, and thus the necessity to rely on expert opinion for the short term and to rely on scenario analysis for the long term.

Clinical effectiveness

The main uncertainties regarding clinical effectiveness are both the accuracy of the SeHCAT test in predicting either BAM or response to treatment and the effectiveness of BAS for the treatment of chronic diarrhoea caused by BAM.

For the assessment of the accuracy of the SeHCAT test in predicting BAM there was no evidence, mainly because there is not an appropriate reference standard for this assessment. For the assessment of the accuracy of the SeHCAT test in predicting a response to treatment, we found three studies. However, the studies had small numbers of patients with unknown cause chronic diarrhoea, they used different cut-offs for the assessment of BAM and between-study heterogeneity was considerable. One study was included for the assessment of the clinical effectiveness of BAS for the treatment of chronic diarrhoea caused by BAM. This study looked at colesevelam versus placebo and included 24 patients: 12 in each arm. All other included studies used cholestyramine.

We found no data on the accuracy of the SeHCAT test in predicting either BAM or response to treatment, or the clinical effectiveness of BAS for the treatment of chronic diarrhoea caused by BAM in patients with Crohn’s disease.

Cost-effectiveness

The main uncertainties in the cost-effectiveness analyses are caused by lack of essential data. Several assumptions had to be made to make it possible to perform the cost-effectiveness analyses.

The lack of evidence of the accuracy of SeHCAT based on a reference test meant that in the diagnostic decision trees the conventional way of modelling test accuracy using sensitivity and specificity of testing was not feasible. Therefore, it remained impossible to indicate FP and FN probabilities of testing. The accuracy of SeHCAT testing was necessarily based on test result in combination with response to BAS treatment. Logically, this response after testing is an unknown function of actual test accuracy and BAS efficacy. In other words, patients responding to BAS may be TP patients with a true response but may also be FP patients with a placebo response.

The limited number of papers was heterogeneous regarding cut-off value of SeHCAT test results. Given the variety of cut-off values between 8% and 12% and the small number of studies, these studies were considered one group. Of course it is not possible to make a distinction between these grouped cut-off values. The 5% and 15% cut-off values based date are more homogenous in this respect.

The parameter value for BAS treatment response given a positive SeHCAT test was derived from the literature. However, for the SeHCAT-negative patients the model assumed IBS-D treatment and for this specific response parameter only a subjective assumption could be made with what was only indirectly related expert opinion: the expert opinion on IBS-D treatment response without testing (52%) was increased by 10 percentage points for this purpose. This was motivated by the idea that a positive SeHCAT test would have selected potential IBS-D non-responders to be treated by BAS.

Important uncertainties exist about the trial of treatment strategy. No data were available to inform the response rate to BAS, and so we made assumptions based on the total number of BAS responders in the SeHCAT 15% strategy. One question is to what extent it is reasonable to choose this strategy rather than SeHCAT 10% to base the response rate on, as resorption values between 10% and 15% are often seen as subnormal. However, we found in our model that the combination of first a positive SeHCAT test and then a positive response to BAS occurs almost as often in the SeHCAT 10% strategy as in the SeHCAT 15% strategy. Thus, this choice does not influence the results. Other issues regarding the trial of treatment strategy relate to the placebo response that may be expected in the true IBS-D patients receiving BAS. It is a well-known fact that patients with IBS-D tend to show high placebo responses to treatment. 68 One of the clinical experts pointed out that long-term inappropriate treatment with BAS may have implications for instance absorption of other drugs and vitamins. Ideally, such long-term consequences would need to be included when modelling the trial of treatment strategy. Additionally, long-term transitions between ‘diarrhoea’ and ‘no diarrhoea’ may not be the same for BAS patients having a positive SeHCAT result and patients responding to a trial of treatment, as patients without a positive diagnosis may be less inclined to accept the side effects of BAS.

In the long-term Markov model, the state of having diarrhoea was valued by cost and utilities irrespective of the cause of the symptom, being either BAM or IBS-D. It remains to be seen whether or not this is fact. For the increase in utility when patients become responders (i.e. are in the ‘no diarrhoea’ health state) we made two assumptions; one where this increase is the same for BAM and IBS-D patients and one where the increase in BAM patients is only 75% of the increase seen for IBS-D patients. This latter scenario reflects the idea that BAS is usually not well tolerated by patients with side effects leading to a smaller utility gain. To what extent this assumption of 75% is realistic is unknown.

The lifetime transition probabilities between the Markov states diarrhoea and non-diarrhoea could not be based on reliable evidence. For this reason, several equally plausible scenarios were used. Furthermore, it turned out to be impossible to include a health state constipation in the long-term Markov model because data on this issue were lacking and the impact on the results was estimated to be negligible.

In the protocol of the study, it was suggested that in patients who tested positive for BAM despite in fact not having BAM (FPs) an important question would be whether or not (some of) these patients will be detected at some point as having IBS-D. Likewise, patients who tested negative for BAM despite in fact having BAM (FNs) are assumed to receive treatment for IBS-D. Again, an important question is whether or not (some of) these patients will be detected at some point as having BAM. Experts indicated that in the latter situation, patients most likely would not be diagnosed as BAM with some delay. For the first situation the lack of essential data made it impossible to incorporate the impact of a possible delay in efficacious treatment in the cost-effectiveness analyses.

As stated, for the Crohn’s disease population no evidence regarding essential parameters could be found to be used in the model calculations. This raises the question if modelling should have even been attempted. However, as the NICE diagnostic assessments may also be used to formulate which, if any, additional research is required, it was felt useful to make all the uncertainties for the Crohn’s population explicit. Thus, for various parameters the assumption was made that the data defined for the population of patients presenting with chronic diarrhoea with unknown cause and symptoms suggestive of functional disease were considered to be generalisable to the Crohn’s population. Expert inputs did not state to do otherwise.

The cost-effectiveness analyses are based on adult patient data only. Whether or not the findings hold for children and adolescents is unknown.