Definition

For the purpose of this guideline, IBS is defined using the Rome II criteria, used mainly in the context of research. The Rome group is a pan-European clinician group that have met for the last decade, seeking to provide both clarity and direction for clinicians and patients alike.

The Rome II criteria characterises IBS as:

• At least 12 weeks (which need not be consecutive), in the preceding 12 months, of abdominal discomfort or pain with two of the following three features:
  • Relief by defaecation
  • Onset associated with a change in frequency of stool
  • Onset associated with a change in stool appearance.

The IBS population

IBS most commonly affects people between the ages of 20 and 30 years and is twice as common in women as in men. The prevalence of the condition in the general population in the UK is estimated to lie somewhere between 10 and 20%. Recent trends indicate that there is also a significant prevalence of IBS in older people; therefore, IBS diagnosis should be a consideration when an older person presents with unexplained abdominal symptoms. Because incidences of other conditions with similar symptoms are higher in the elderly population, use of certain diagnostic tests is warranted. Co-morbid conditions and poly-pharmacy are common in this patient population. 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 rely on self-care. In England and Wales, the number of people consulting for IBS is extrapolated to between 1.6 and 3.9 million. Evidence suggests that age and race have no consistent effect on the incidence of symptoms. Healthcare professionals need to be sensitive to and take into consideration cultural, ethnic and communication needs of people for whom English is not a first language or who may have cognitive and/or behavioural disabilities. Appropriate action should be taken to facilitate effective consultation.

Investigations commonly requested by clinicians

Primary care investigations are likely to include: routine blood tests such as full blood count, urea and electrolytes, liver function tests; tests for thyroid function, tissue transglutaminase anti-endomysial antibodies (test for coeliac disease); inflammatory markers, stool microscopy; urinary screen for laxatives; and lactose tolerance testing. Other investigations such as gut transit studies (radiological tests to measure the time required for food to move through the digestive tract) and sigmoidoscopy (endoscopy of the lower part of the bowel) are routinely performed in secondary care. Determining the criteria for such requests and appropriate referral into secondary care will be addressed in the guideline.

The need for effective diagnosis – clarifying concepts

IBS is associated with a disproportionately high prevalence of abdominal and pelvic surgery, although the cause of this has not been established. Diagnostic test methodology has traditionally been applied when comparing a new or alternative test with the acknowledged gold standard reference.

Gold standard reference points aim to represent the ‘truth’, and when a test is carried out there are four possible outcomes. These are:

1. True positive (detects disease when present)
2. False positive (detects disease when it is absent)
3. True negative (can identify absence of disease)
4. False negative (can identify someone as being disease free when they have it).

It is widely acknowledged within the literature that there is no gold standard reference for the diagnosis of IBS, which means that comparison of definitive diagnostic tests remains difficult. Diagnostic criteria in themselves can be seen as having enormous value, and these could be directly compared in other disease areas to the gold standard reference. In this narrative review of 170 studies/papers, comparisons of criteria are made against a definitive diagnosis of IBS through clinician expertise, augmented by a whole battery in many cases of diagnostic investigations.

In measuring accuracy of diagnostic test/criteria, two measures are used. These are sensitivity and specificity.

Sensitivity

This is a measure (usually expressed as a % of the total population that the test is applied to) that indicates how good the test is in identifying people with the disease.

Specificity

This is a measure (usually expressed as a % of the total population that the test is applied to) that indicates how good the test is in identifying people without the disease.

Problems associated with using these as single measures are acknowledged, as they are difficult to interpret for individual patients. For example, if a test has a sensitivity of 85%, what if I am one of the 15% that the test has failed to identify. In real world situations, what patients and clinicians generally want to know is ‘If this test is positive, does it mean that I have a positive diagnosis?’ or; ‘If this test is negative, does this mean that I do not have the disease?’

What may be more useful is for these single measures to be expressed as a probability; a likelihood of accuracy. Again this is expressed as a %, with positive tests measured against a whole study population who had the test. For example, 37 positive results out of 100 would be a 37% prediction, expressed as a positive predictive value (PPV) of 37% (Dawes et al, 2005). This can be viewed from the reverse perspective; how many negative tests were recorded out of the total study population who had the test. For example 63 negative results out of 100 would be a 63% prediction, expressed as negative predictive value (NPV) of 63%. Using real data (Steurer 2002), of 1000 women who received a positive mammogram result, 90 actually had breast cancer, meaning that the PPV for mammography is 9%. Converted to probability, this means that women have a 1 in 11 chance of having breast cancer if they have a positive mammogram result. Of the 12,102 negative mammogram results, 12,090 did not have breast cancer (meaning that 12 did have breast cancer). This means that the NPV for mammography is 99.9%. Converted to a probability, this means that women have a 1 in 1000 chance of having breast cancer if they receive a negative mammogram result. This is extremely useful to clinicians in trying to establish risk and/or probability of a disease in a particular individual, enabling them to articulate this to the person seeking consultation.

Odds ratios

This is another way of measuring test accuracy (see Appendix 2 of this chapter). Its real value is in estimating test accuracy. This is calculated using test Likelihood Ratio’s (LR) by taking the positive LR and dividing this by the negative LR. Likelihood ratios are useful in estimating the value of diagnostic tests, and as a general principle, the higher the likelihood ratio the more useful that test will be. A high odds ratio is an indicator of a good diagnostic test.

A main aim of the guideline

One of the main aims of this guideline is to identify diagnostic criteria for people presenting with symptoms suggestive of IBS and to ensure that primary care clinicians and people who may have IBS have a reference tool that is both sensitive and specific, with high predictive value of the syndrome. This is an area of healthcare practice which is currently absent, and creates great uncertainty for both clinicians and people who may have IBS.

Types of studies

All published literature on IBS diagnosis was included. This resulted in a large search and, post-sifting, a large number papers being reviewed for potential inclusion in the review.

Pre-Rome

The first paper to address diagnostic criteria for IBS was a working team report published in 1989 for the 1988 International Congress of Gastroenterology in Rome, Italy. This is acknowledged as the Rome criteria.

Establishment of Rome Committee Process

Following the 1989 publication, a committee was set up the same year to develop for the first time a classification system for all the 21 functional gastro intestinal disorders (FGID). This report was published in 1990 heralding the beginning of the Rome Criteria process. The criteria for IBS in 1989 did not feature pain as a symptom, which is now a current ROME criteria requirement for the diagnosis of IBS.

Rome I

From 1990–1995, seven committees formed to elaborate on the 1990 classification system. Knowledge of this classification system was quite limited, since the journal had a small circulation and was not listed in MEDLINE. The committee however were able to publish a book which featured the updated Rome I criteria in 1992 and it was the first time that pain was required for the diagnosis.

Rome II

By 1995, interest had grown from both clinicians and the pharma industry. Funding was secured from industry to support the development of Rome II. The number of committees was expanded, with wider international contributions forming the basis of this updated set of criteria. Emerging from this process, the criteria were available from 1999 and first published in 2000.

Rome III

Because of the success of the Rome II process, funding support from industry was forthcoming to maintain this consensus process. A co-ordinating committee was formed in 2001 for Rome III (Drossman, Corazziari, Delvaux, Spiller, Talley, Thompson and Whitehead). Work began in May 2003, leading to publication of new criteria in 2007.

Kruis criteria

The aim of the original study was to create a scoring system for IBS diagnosis incorporating history, physical examination and some basic investigations (ESR and blood count).

¹Validated criterion reference tool reviewed and acknowledged as used within practice over the last 3 decades.

Manning criteria

The patient should present with at least 2 of the following symptoms for an IBS diagnosis to be made:

• Onset of pain associated with more frequent bowel movements
• Onset of pain associated with more loose bowel movements
• Relief of pain with defaecation
• Abdominal distension
• Sensation of incomplete evacuation with defaecation
• Passage of mucus.

Rome Criteria

At the 13th International Congress of Gastroenterology in Rome in 1988 a group of physicians defined criteria to more accurately diagnose IBS. The Rome criteria are:

The patient should present with 3 months of continuous or recurring symptoms of abdominal pain or irritation that:

• May be relieved with a bowel movement
• May be coupled with a change in frequency, or
• May be related to a change in the consistency of stools.

Two or more of the following are present at least 25 percent (one quarter) of the time:

• A change in stool frequency (more than 3 bowel movement per day or fewer than 3 bowel movements per week)
• Noticeable difference in stool form (hard, loose and watery stools or poorly formed stools)
• Passage of mucous in stools
• Bloating or feeling of abdominal distention
• Altered stool passage (e.g. sensations of incomplete evacuation, straining, or urgency).

Rome I criteria (1992)

The patient should present with at least 3 months of continuous or recurrent symptoms for an IBS diagnosis to be made:

Abdominal pain or discomfort, which is:

• Relieved with defaecation
• and/or associated with altered bowel frequency
• and/or associated with altered stool consistency
• and/or two or more of the following, on at least 1/4 of days:
• Altered stool frequency
• Altered stool form
• Altered stool passage (straining, urgency or tenesmus)
• Passage of mucus
• Usually with bloating or a feeling of abdominal distension.

Rome II criteria

The Rome II Criteria, published in 2000, were developed by 10 multinational working teams that collaborated over 4 years to arrive at a consensus for symptom-based diagnostic standards.

Twelve weeks* or more in the past 12 months of abdominal discomfort or pain that has 2 out of 3 features:

• Relieved with defaecation
• Associated with a change in frequency of stool
• Associated with a change in consistency of stool.

*The twelve weeks need not be consecutive

The following are supportive, but not essential to the diagnosis. One or more are usually present. They add to the clinician’s confidence that the intestine is the origin of the abdominal pain. The more of these symptoms that are present, the greater the confidence with an IBS diagnosis:

• Abnormal stool frequency (> 3/day or < 3/week)
• Abnormal stool form (lumpy/hard or loose/watery) > 1/4 of defaecations
• Abnormal stool passage (straining, urgency or feeling of
• incomplete evacuation) > 1/4 of defaecations
• Passage of mucus > 1/4 of defaecations
• Bloating or feeling of abdominal distension > 1/4 of days.

Issues to consider

From this extensive review of the literature, a number of key observations have emerged which the GDG will need to consider in moving towards consensus opinion as to how IBS is diagnosed; which criteria to use in diagnosing IBS and how this potentially will move clinicians away from an exclusion diagnostic approach towards positive diagnosis of IBS, management and patient follow-up. Of significance is the potential cost saving to the NHS of tests that are routinely requested but prove to be of little added value.

Diagnostic thresholds

Clinicians need to be able to determine whether a person has IBS (or not). Balance between missing the diagnosis and over diagnosing is a possibility because criteria may be too vague. Thresholds can be set across different parameters. These include:

• Severity of bowel symptoms
• Symptom count – either all symptoms given equal weight (e.g. Manning – 2 or more symptoms being given equal weight) or identified symptoms being considered as essential with others considered as accessory symptoms (Rome I)
• Duration threshold in combination with symptom count (Rome II) and symptom frequency appears to be highly relevant (Mearin 2003)
• Rome II requires the presence of both abdominal pain and changes in bowel habit and duration of symptoms (at least 12 weeks in last 12 months)
• Rome considers abdominal pain and changes in bowel habit independently and no minimum duration of symptoms
• Kruis developed according to symptoms and evaluation of lab tests
• Following a positive diagnosis of IBS based on clinical criteria clinicians can be reassured that the diagnosis is durable. (Adeniji 2004, follow-up of 196 patients with a diagnosis of IBS between 1989–92 35/86 pts had had further diagnostic investigations without diagnosis changing)
• Prevalence of IBS measured across a New Zealand population cohort (N=980) using 2 of the manning criteria emerged as 18.1%. This decreased to 10.3% if 3 or more of the Manning criteria were used to identify an IBS diagnosis. This then fell to 3.3% when 4 or more of the criteria were used (Barbezat 2002).

A key question for the GDG was “Should a positive diagnostic approach include severity threshold or disabling threshold?”

What happens to patients who do not meet Rome II?

• How important is diagnostic precision for clinicians (this is a different priority for research)? Is the priority to exclude structural cause and/or red flags for symptoms?
• Would treatment be different if patients were diagnosed using Manning or Rome? Does this matter?
• Do they have alternative diagnosis (eg FBD) and/or go for lots of investigations and then turn out to be IBS?
• Manning: discriminates upper GI disease from IBS but not IBD (3 or more Manning criteria were frequent in patients with ulcerative colitis in remission (Isgar 1983).
• Kruis: was not able to discriminate IBS from organic GI Disease (Frigerio 1992).

Clinician ignorance of IBS diagnostic criteria

• An Italian study (Bellini 2005) of 28 generalist GPs – 17 judged knowledge of IBS to be insufficient but only three thought that further education might be useful. Ten GPs were familiar with Rome II prior to the study; 19 agreed with Rome II criteria when they used them as part of the study. They reported satisfactory management in approximately 60% of patients.
• Important symptoms for this group – Primary symptoms: changes in bowel habit, abdominal pain relieved by evacuation, bloating. Secondary symptoms: difficult or incomplete evacuation, passing mucous with BM.
• Following clinical evaluation GPs ordered further investigations in large numbers of patients. GPs with more than 20 yrs experience requested less investigations than younger colleagues (p= 0.001). 168/229 patients had routine bloods, 30 – abdominal ultrasound, 87 FOB, 83 faecal analyses, 81 – Thyroid Function, 70 – lower GI endoscopy.
• All agreed that counselling, reassurance, information about natural course of condition and suggestions for coping strategies were first step in management. Patients with diarrhoea were prescribed wider ranging therapeutic interventions and perceived to be in more need of further investigations than those with constipation.
• In 145/229 cases referral to at least one specialist was made (GI specialist most common but gynaecological referral for 19% of women). Referral did not vary with clinical presentation and the most common reasons for referral were diagnostic confirmation, patient need for reassurance and patient request.
• Patients frequently attribute food intolerance (37%) and stress (43%) as the causes of IBS. GPs consider stress (71%), fibre deficiency (83%) and disturbed motility (62%) as most important factors (Bijkerk 2003).
• Helpful clues to aid diagnosis: symptoms chronic or recurrent; pain is variable in location and timing; D and C may alternate; onset sometimes follows GI tract infection; findings on physical examination are usually normal except for abdominal tenderness (Paterson 1999, Canadian IBS position statement).
• USA Primary Care practice based diagnostic evaluations differ significantly from speciality expert opinion based guidelines. Implementation of speciality guidelines in Primary Care would increase utilisation but with limited improvement of diagnostic outcomes (Yawn 2001).
• Patients under 50 yrs of age who meet Manning and have no red flags require no investigations (Paterson et al 1999 Canadian IBS position statement).
• Patient expectations: reassurance, counselling, pharmacotherapy, diagnostic tests and referral to specialist. Dietary interventions were considered less important. Most people with IBS in this study stated that improvement in worst symptom should be target of treatment. Global improvement and QoL were considered much less important as treatment goal.
• A British study in general practice (n=400) Gladman and Gorard (2003) Sent a questionnaire to a random selection of 200 GPs and 200 clinician members of British Society of Gastroenterologists asking about their knowledge of functional GI disorders and their knowledge and use of Manning and Rome criteria for diagnosis. 68/137 GPs believed functional GI disorders were psychosomatic compared to 36/167 of consultants (p=<0.001). Consultants believed that understanding had increased over last 20 years; 50%GPs believed it had not changed. Both believed diagnosis and management had not improved in past 20 years. Only 29/137 GPs had heard of Manning; 16/137 of Rome compared to 134/166 and 139/167 Consultants respectively (p=0.0001). Only 18 GPs used either Manning or Rome in practice and despite increased awareness only 40% consultants used one or other diagnostic criteria in their practice.

Many studies of IBS and developed guidelines to date have been produced by specialists who had seen patients with particularly severe or intractable symptoms. This clinical guideline was developed from a different starting point, with the essential focus being in primary care. All development and implementation must be framed with questions of applicability to General Practice.

The GDG noted that GPs consider Rome II too complicated and more suited for use in secondary care or for research purposes (Thompson 1997; Bellini 2005). The need for a pragmatic useful diagnostic tool was felt to be the most important aspect to the review. As the majority of IBS patients are treated by GPs, any recommendations for the use of diagnostic criteria should ensure that their ease of use by GPs in their practice is established, rather than expecting GPs to use criteria that do not apply in the reality of day-to-day practice (Corsetti and Tack 2004).

What is the role of red flag symptoms alongside diagnostic criteria IBS diagnosis?

• The Manning criteria do not consider red flag symptoms. The addition of red flag symptoms seems to enhance diagnostic accuracy (Paterson 1999, Canadian IBS position statement). The GDG considered this aspect of the review at length, recognising the need for recommendations supporting the IBS algorithm to ensure that red flag symptoms take the patient out of this guideline and into other related NICE guidance.
• The addition of red flags to the Manning criteria increases the PPV of Manning and Rome I and II (Vanner 1999; Hammer 2004).
• Red flag symptoms – these seem to enhance original criteria and importantly relate this guidance to other relevant NICE guidelines, in particular NICE Clinical Guideline 27 ‘Suspected Cancer Referral’ published in 2005.

Discussion

The need for clinicians to be guided in the diagnosis of IBS has emerged as a strong recurrent theme in this systematic narrative review. The seminal work of Manning laid a foundation to enable clinicians to be guided by such criteria in attempting to provide direct answers to patients presenting with a range of symptoms. This work has undoubtedly influenced the development of thinking within the Rome group, and the Rome criteria recently published as Rome III reflects the benefit of validation of the key aspects of the criteria and pragmatic decisions relating to the length of presenting symptom such as pain (6 months). The use of available diagnostic criteria summarised in this review (Table 1) have typically been augmented by further diagnostic investigations that have limited or no benefit and these add considerable costs to the NHS.

The use of consensus agreement regarding the recommendation of single diagnostic criteria serves three main purposes:

• Increased patient confidence through positive diagnosis
• Increased clinician confidence
• Potential for considerable NHS disinvestment in avoiding unnecessary investigations and referrals to multiple specialities.

When looking at combinations of possible criteria used in the available diagnostic tools reviewed, the emergence of Rome III has proven to be timely in relation to guideline development. It features strengths of previous diagnostic criteria, while minimising weaknesses of reviewed tools.

Key questions that emerge from the literature aim at identifying a gold standard or best reference tool. The challenge is that when thresholds differ, results are inconsistent. For clinicians, diagnostic precision of IBS is often of low priority when compared to excluding other structural cause. This is a conceptual misinterpretation which can be explained as under-confident application of clinical examination and clinical history interpretation. Perhaps of significant note, regardless of which criteria were used in included studies in the review, treatment rarely differed against symptom profiles.

Over a decade ago, Jeong and colleagues having identified that Manning had reasonable specificity, called for better diagnostic criteria with improved accuracy to be developed. The road to Rome and the subsequent development of international consensus over the last 15 years has been useful in predetermining consistency in research application. It however, may have distracted from refinement of a tool that is easy and straightforward to use for primary care clinicians.

What clearly emerges from the literature is that with careful history and physical examination, positive diagnosis of IBS is possible. This, augmented by simple laboratory investigations to rule out more serious underlying pathology in the absence of red flag symptoms, would seem to be a step forward for both clinicians in diagnostic practice and patients in receiving timely interventions.

Perhaps it is fitting to highlight within this review that clinicians have been seeking to change the way that they think about diagnostic approaches in relation to this chronic syndrome. The Manning (1978) paper titled “Towards a positive diagnosis of Irritable Bowel Syndrome’ clearly indicates a diagnostic aim, this over the last three decades has been lost, with negative diagnosis by ruling out other conditions being the predominant clinician approach. Returning to the original aim of Manning and colleagues by seeking to influence the behaviour of primary care clinicians in the way that they think and approach diagnosis of people presenting with IBS symptoms is an important objective of this guideline.

GDG interpretation of the review and application of the guideline

General practitioner (GP) training has focussed on the importance of what happens within a typical patient consultation. This is usually recorded and analysed to enable new GPs to reflect on the detail within the consultation, in particular the quality of verbal and non-verbal behaviour, the sequencing of questions and information gathered to enable diagnosis. This is based around simulation and objective structured clinical examination methodology and has effectively enabled GP trainees to experience and develop understanding related to the importance of clinical history prior to physical examination. Using this approach, the NCC-NSC planned an interactive session for the GDG to fully engage with relevant issues. This was felt to be important in demonstrating that in guideline development, the GDG had explored the utility of different criteria that would then inform any consensus recommendations and lay the foundation for a positive implementation experience.

In order to test the utility of different criteria, NCC-NSC staff ran an interactive diagnostic simulation with members of the GDG. A number of typical IBS patient profiles were written by the technical team, which were then shared with four subgroups of the GDG. Details of the patient profiles are listed in Appendix 3 to this chapter.

Sub group constituency:

• Primary or secondary care doctor and/or
• Primary or secondary care nurse and/or
• Allied health professional (eg. Dietician or pharmacist) and/or
• Patient representation allocated to comment and input across the four groups.

One patient profile randomly selected from the total number of prepared patient profiles were randomly allocated to each of the GDG subgroups. Each group then had to discuss the use of allocated diagnostic criteria and elect one member of the group to role-play a GP consultation, responding to the simulated patient profile. Group A were asked to use Kruis Criteria; Group B were asked to use Manning criteria; Group C were asked to use Rome criteria;Group D were asked to use Rome II criteria.

Each group selected a physician to role-play the consultation, timed at a typical 8 minute general practitioner consultation. Two groups selected their GP member, with the other group selecting the GDG clinical lead who is a Gastroenterologist. Members of the NCC-NSC team role-played the four different patients. ROME III at the time of the patient simulation was unpublished, and therefore was unable to be used in this exercise. During each of the four role-plays, GDG members were asked to observe the consultation and record their observations. These typically related to the ease and logical progression of the consultation, shaped by the diagnostic criteria. This simulation enabled the GDG to both interpret the evidence and evaluate how easily the criterion reference based tool could work within a busy primary care environment.

The NCC-NSC technical team transcribed the detail within each of the GP consultations and recorded the information gathered from the patient using each of the three criteria referenced tools. The content was analysed and grouped in emerging themes (see Table 1) to enable the group to fully understand what was possible in recreating the primary care consultation. Typically patients are reticent to come and see a primary care clinician with issues relating to bowel habit/function, and this reticence was simulated with behaviours that demonstrated both hesitancy and embarrassment.

Observations were recorded from three main areas for feedback:

1. How the GDG clinician felt using the diagnostic criteria allocated to them.
2. How GDG members felt each of the diagnostic criteria worked in this simulated patient consultation.
3. How the NCC team member felt when role playing the IBS patient, in relation to the sequencing of ideas and extracting of important patient information, facilitating an effective diagnosis.

This was a powerful exercise in embedding the evidence review into a simulated patient-clinician exchange. The importance of ensuring that the guideline recommendations are able to be effectively implemented into routine primary care is clearly important in ensuring that current variations in diagnosing IBS are addressed.

Outcomes from the evidence review and diagnostic criteria simulated exercise were:

• A strong evidence base for the use of diagnostic criteria with good predictive value.
• Expert (GDG) evaluation of how potential tools could enable primary care clinicians to make a positive diagnosis of IBS, supported by a limited number of investigations that may augment an IBS positive diagnosis.
• Agreement of evidence based positive diagnostic criteria for use in primary care which reflects current evidence.
• A contemporised Manning criteria which are consistent with ROME III criteria.
• The decision to refer to agreed criteria as ‘Positive Diagnostic Criteria’.

KRUIS (A SCORE OF >44 = IBS)

MANNING (> 3 CRITERIA = IBS)

ROME I

ROME II

Themes emerging from the exercise and focussed GDG discussion

• IBS is a lifelong condition that needs to be managed effectively.
• Symptoms that are most crucial in diagnosis are pain/discomfort relieved by bowel movement, bloating (more common in women; men describe it as abdominal tension) and disordered bowel habit. It was noted that language in the Manning criteria needed to be contemporised; this was discussed and agreed, e.g. pain was contemporised to pain/discomfort.
• Pain/discomfort induced by eating is also common symptom.
• Extra-colonic symptoms were commonly reported in secondary care – with good discussion around their prevalence in primary care.
• The severity of the condition may or may not be useful as a threshold. Whilst high sensitivity maybe attractive, it is important not to miss patients by having too high an exclusion criteria, as reported in the evidence for Rome II, and supported by the simulated consultation feedback and analysis.
• IBS co-exists with other conditions. The possibility of missing inflammatory bowel disease initially would not be perceived by the GDG as problematic
• There is clear evidence supporting diagnostic criterion based reference tools, but their use in practical clinical settings has been reported to be difficult, this was noted by the group and it was felt that published criteria in this guideline should reflect the validated tools, but ordered in such a way that ensures that the tool is intuitive for clinicians to use. It should also facilitate the type of discussion that enables a full history to emerge.
• The individual patient story is very important, emphasising the need for the primary care clinician to focus on the most severe symptom while also establishing other related symptoms.

Published evidence from the diagnostic tools has shaped recent diagnostic approaches for IBS. Whorwell (2006) refers to this as a diagnostic triad, seen below:

• Pain/discomfort – quality and quantity
  Site of pain: in IBS it can be anywhere in the gut. If the site of pain varies it is unlikely to be cancer (tumour fixed). Need to distinguish this IBS pain discomfort from that caused by gall bladder disease. IBS patients do not tolerate abdominal surgery well.
• Bowel habit – quality and quantity
  Giving patients’ descriptive examples (e.g. like porridge, rabbit pellets) and using the Bristol Stool Form Scale helps. Incomplete evacuation is reported, creating rectal hypersensitivity. Urgency is increased in Diarrhoea; prevalence for those incontinent is 20% (patients often do not disclose unless asked directly).
• Bloating in women
  (Absence of bloating in women = red flag) Less common in men, although they may report that the abdomen is tight/hard.

The diagnostic triad clearly reflects the valuable work published by Manning and the Rome group. It also highlights the importance of the extra-colonic features that maybe reported by people presenting with IBS symptoms, typically these include nausea, low back pain, bladder symptoms and thigh pain.

The GDG agreed that primary care clinicians should ask open questions to establish the multiple features of the syndrome, recognising that a potential conflict may exist within Primary Care in terms of the time available to the clinician in exploring the whole range of presenting symptoms.

Diagnostic certainty

In establishing the sensitivity and specificity of different diagnostic criteria by using a clinical reference standard (expert gastroenterologist diagnosis), looking at a pragmatic diagnostic reference tool appears to be of great value to the primary care clinician. The advent of ROME III during the development of this guideline was both timely and beneficial in shaping the and further strengthening the diagnostic criteria agreed within final recommendations. Of equal value, is the provision of clear economic evidence relating to supplementary diagnostic tests.

Mein (2004)

The primary aim of this study was to assess the cost-effectiveness of coeliac disease testing in patients with suspected irritable bowel syndrome in the US health care system. This was done by using a decision tree to estimate the number of coeliac disease cases detected, QALYs gained and costs resulting from three testing strategies and comparing these to no testing. The three strategies were; tissue transglutaminase antibody (TTG), antibody panel, or upfront endoscopy with biopsy. All positive serological tests were followed by endoscopy with small bowel biopsy and the potential complications of this procedure were accounted for. All positive upfront endoscopies were assumed to be confirmed by an antibody panel. Long-term treatment costs were assumed similar between patients with IBS and those diagnosed with coeliac disease. The increase in health-related quality of life associated with correctly diagnosing coeliac disease in patients with suspected IBS and initiating a gluten-free diet was estimated indirectly by comparing utility estimates for treated and untreated coeliac disease and IBS measured in different populations. This is less reliable than direct utility measurement as it combines estimates from different populations. However, the uncertainty surrounding this parameter has been adequately examined in a sensitivity analysis. The authors stated that they used conservative assumptions to deliberately bias the model against testing for coeliac disease. These included assuming no reduction in resource use or increase in life-expectancy following correct diagnosis of coeliac disease and initiation of treatment. The base case prevalence for coeliac disease in patients with suspected IBS was 3% and was varied from 1 to 5% in a sensitivity analysis.

The probabilistic model estimated that testing with TTG would detect 28 out of 30 cases present in a population of 1000 individuals but testing with a full antibody panel would only detect one further case. The median incremental cost per case detected was $6,700 (interquartile range $4,800 – $9,700) for TTG vs no testing and $12,300 ($8,900 – $17,700) for antibody panel vs no testing. The incremental cost per case detected for antibody panel vs TTG was $167,000 ($110,000 – $279,000). The incremental cost per QALY was $11,200 ($7,200 – $17,900) for TTG vs no testing and $20,900 ($13,500 – $34,300) for antibody panel vs no testing. The incremental cost per QALY for antibody panel vs TTG was $287,000 ($99,400 – $675,000). The upfront biopsy strategy resulted in a lower QALY gain and higher costs than TTG testing and was therefore dominated by TTG testing. In the one-way deterministic sensitivity analysis, reducing the prevalence to 1% increased the cost per QALY of TTG vs no testing from $7,400 to $19,900 and decreasing the utility gain associated with treatment from 0.024 to 0.01 increased the cost per QALY to $17,900. This demonstrates that whilst the cost-effectiveness results are sensitive to changes in these parameters, the TTG testing strategy is still cost-effective compared to no testing at the thresholds considered ($50,000 to $100,000 per QALY).

This study provided evidence that TTG testing followed by confirmatory endoscopy with biopsy would be cost-effective in patients with suspected IBS in the US health-care system. We converted the cost per QALY directly from 2003 US$ to 2006 UK£ using Health Care Purchasing Power Parity rates (2003 PPP rates UK/US = 2317/5711, OECD 2006) and Hospital and Community Health Services Pay and Pricing Index (2006/2003 = 241.3/224.8 (Netten 2006) and this gave a cost per QALY for TTG vs no testing of £4,900. This is a crude estimate as it assumes that each component of the total cost has an equal weighting in both counties, which may not be true due to differences in the health care systems between the US and UK. However, the relatively low value of this estimate compared to typical UK thresholds of £20,000 to £30,000 per QALY, suggests that this intervention may also be cost-effective from a UK NHS and PSS perspective.

Spiegel (2004)

This aim of this study was to assess the cost-effectiveness of screening for coeliac disease in patients fulfilling the Rome II criteria for diarrhoea predominant IBS (IBS-D) in the US health care system. A strategy of screening for coeliac disease using serological tests followed by confirmatory endoscopy with biopsy was compared with a strategy of initiating IBS therapy without screening for coeliac disease. This was done using a decision tree to estimate the number of patients receiving appropriate therapy for either IBS or coeliac disease, the number of missed coeliac disease diagnoses and the number of patients for whom IBS treatment was delayed due to coeliac disease testing. It was assumed that 1 in 4 clinicians eventually test for coeliac disease in patients who do not respond to empiric IBS treatment, resulting in an average diagnostic delay of 6 months. A Markov model was then used to estimate transitions between states of symptomatic improvement and remission once patients have begun treatment for either IBS or coeliac disease. The analysis was based on a generic serological test using data which reflected the range of serological tests available (anti-EMA and anti-TTG IgA antibodies). The results are presented in terms of the cost per additional patient with symptomatic improvement after 10 years. The authors state that the model was deliberately biased in favour of IBS treatment without testing for coeliac disease in order to place the burden of proof for cost-effectiveness on coeliac testing. This was done by using estimates from the unfavourable end of the range presented in the literature for the following parameters; coeliac disease prevalence, sensitivity and specificity of tests for coeliac disease, rate of coeliac disease testing in patients not responding to empiric IBS therapy, IBS treatment effectiveness and cost. For example, IBS treatment was assumed to be effective in 75% of patients based on the effectiveness of alosetron but the cost of therapy was assumed to be $45 per month which is similar to the cost of loperamide. The model also assumed that 30% of the population with coeliac disease had “latent” or “potential” coeliac disease which would not be detected by small bowel biopsy but would have the potential to benefit from a gluten free diet. It also assumed that 5% of the population with coeliac disease would have concurrent IgA deficiency which would render serological screening for IgA antibodies ineffective. These assumptions were based on limited data but were included to bias the model against testing for coeliac disease and their impact was explored through sensitivity analyses.

The deterministic base case model estimated that testing for coeliac disease resulted in 51.6% of the cohort achieving symptomatic improvement at 10 years, whilst initiating IBS therapy without testing for coeliac disease resulted in 50.9% of the population achieving symptomatic improvement. The incremental cost was $77 per patient resulting in a cost per additional symptomatic improvement after 10 years of $11,000. The probabilistic model resulted in a median cost per symptomatic improvement of $12,983 (95% CI: Dominating to $41,031). The results were sensitive to the prevalence of coeliac disease in the population considered. The cost-effectiveness ratio was under $50,000 when the prevalence was >1% and screening dominated no screening (resulted in more health gain at reduced cost) when the prevalence was over 8%.

These results were difficult to interpret as they were presented for the US health care system and did not provide benefits measured in QALYs. The aim of a Markov model is usually to determine the proportion of time a patient spends in each health state over the duration of the model and to use this to estimate their aggregate health gain over the time-horizon considered. This analysis did not present results in terms of the time spent in the symptom remission state, but instead presented the results in terms of the number of patients in this state at the end of the model, which may not accurately reflect the amount of health benefit accrued over the duration of the model. It was therefore less useful in determining whether testing for coeliac disease is cost-effective compared to no testing than the evidence provided by Mein (2004).

Dubinsky (2002)

This study examined the cost-effectiveness of initial serodiagnostic screening followed by standard invasive testing, compared to standard testing alone in patients presenting with symptoms suggestive of IBD from a third-party payer perspective in the US health care system. The authors state that the population considered by this analysis was patients presenting with symptoms which did not meet the Rome I criteria for IBS. As the aim of this review is to consider the cost-effectiveness of testing for alternative diagnoses in patients meeting the diagnostic criteria for IBS, following the application of a criterion based reference tool, this study was not directly relevant to the target population. We would expect the patient population meeting the diagnostic criterion for IBS to have a lower prevalence of IBD. As the analysis considered a wide range of prevalence values (5% to 75%) in a sensitivity analysis, the results for the lower end of this prevalence range were considered to have some relevance to the target population.

The decision analytic model considered six alternative diagnostic strategies. Two levels of serodiagnostic screening were evaluated. In the primary screening strategy (PR 1) patients received a primary assay followed by a gold standard invasive diagnostic test if the primary screen was positive or if a negative primary screen was followed by persistent symptoms. In the sequential screening strategy (SS 1) a positive primary assay was followed by a confirmatory assay and if this was positive it was followed by a gold standard test. Negative results followed by persistent symptoms were investigated using the gold standard test as in the primary screen. These were compared to gold standard testing upfront (GS 1) which consisted of colonoscopy with biopsies and histological examination as well as a barium upper GI series and small bowel follow-through. Three additional strategies were also considered in which the first three strategies were extended to include a second gold standard test in patients with persistent symptoms following the first gold standard test (PS 2, SS 2 and GS 2). The proportion of patients returning with persistent symptoms due to IBS not meeting the Rome I criteria or other causes of symptoms was assumed to be 50% based on expert opinion and varied from 0 to 100% in a sensitivity analysis. A decision tree model was used to estimate the accuracy and cost of each of the six strategies. No costs or health benefits following diagnosis were estimated. An incremental cost-effectiveness analysis was also presented which compared the relative cost-effectiveness of the six competing strategies.

In the basecase model all of the serodiagnostic strategies had lower costs and higher diagnostic accuracy than the gold standard strategies. The SS 1 strategy had the lowest cost and a diagnostic accuracy of 96.95%. The SS 2 strategy cost $20.30 more per patient but had a slightly higher diagnostic accuracy of 97.90% resulting in a cost per % increase in accuracy of $2,137. The SS 1 strategy dominated all other strategies by having higher accuracy and lower cost and also resulted in the lowest number of invasive procedures out of all six strategies (610 for SS 1 vs 1000 for GS 1 and 1010 for GS 2). In the cost sensitivity analysis, standard invasive testing was more cost-effective when the costs of testing were varied outside of the plausible range considered by the sensitivity analysis.

Standard invasive testing was more cost-effective when the prevalence of IBD was varied to >76%, or when the proportion of patients with persistent symptoms was varied to over 89%. These results suggest that serodiagnostic screening for IBD in patients with “atypical” IBS symptoms would be less costly and more effective than immediate gold standard invasive testing. These results are based on cost data from the US and the conclusions may be different in a UK analysis if the relative costs of invasive and non-invasive testing are significantly different. Whilst these results apply to patients with “atypical” IBS who do not meet the Rome I criteria, the sensitivity analyses carried out demonstrate that they will apply equally to groups with lower prevalence rates of IBD, provided that less than 89% of patients are given the gold standard test after returning with persistent symptoms following a negative serodiagnostic test. This study does not address whether further testing in patients returning with persistent symptoms is beneficial but assumes that this occurs in practice regardless. This number may be higher or lower in the group meeting the diagnostic criteria for IBS depending on the confidence placed on the positive diagnosis. This study did not address whether these strategies for diagnosing IBD are cost-effective compared to a strategy of initiating IBS treatment, following a positive IBS diagnosis, without excluding IBD. It therefore did not demonstrate the cost-effectiveness of serological testing for IBD in patients meeting diagnostic criteria for IBS.

Suleiman (2001)

The aim of this study was to assess the incremental cost-effectiveness of endoscopic procedures in the work-up for IBS. It did not consider the incremental cost-effectiveness of a specific test for an alternative diagnosis in patients with IBS, but considered the increase in diagnostic probability achieved by using various sequences of tests to exclude alternative diagnoses. These tests included; hydrogen breath test to exclude lactase deficiency or bacterial overgrowth, flexible sigmoidoscopy and colonoscopy to exclude inflammatory colitis, diverticular disease and colon cancer, and small bowel follow-through to exclude small bowel cancer. A decision tree was used to estimate the probability of IBS in the remaining population following each diagnostic test. Various sequences of tests were considered but each began with a clinical history, physical examination and laboratory tests. The costs of further testing following false positive tests and the costs and health impact of delayed diagnosis of IBS or alternative diagnoses were not considered. The authors presented incremental cost-effectiveness ratios (ICERs) for flexible sigmoidoscopy and colonoscopy in terms of the incremental cost per 1% increase in IBS probability, but these figures considered the cost in the individual and did not take into account the number of patients who would be tested at each stage of the diagnostic sequence. The authors also presented average cost-effectiveness ratios, “ACERs” which gave the cost of the whole diagnostic sequence in a cohort of patients divided by the number of correct diagnoses.

The model demonstrated that lower ACERs are achieved by using flexible sigmoidoscopy after rather than before hydrogen breath testing and small bowel follow-through. The same was found for colonoscopy in the absence of a previous flexible sigmoidoscopy. The results demonstrated that carrying out colonoscopoy without flexible sigmoidoscopy at the end of the diagnostic sequence would result in a lower ACER than carrying out colonosocopy following flexible sigmoidoscopy at the end of the diagnostic sequence.

The relevance of these results to this review was limited as the study did not consider the incremental cost-effectiveness of testing for a specific alternative diagnosis in patients meeting IBS diagnostic criterion. However, it did demonstrate that the cost of diagnostic testing can be reduced by using more costly interventions at the end of a diagnostic sequence without changing the number of correct diagnoses. The clinical outcomes did not vary when the ordering of tests was varied due to the assumption that each test is independent of the next. In practice this may not be strictly true and there may be some dependence resulting in slightly different clinical outcomes depending on the test sequencing. However, it is still likely that lower costs would be achieved in practice by using more costly invasive investigations at the end of the diagnostic sequence as this would reduce the number of people who require these invasive tests. This would also minimise adverse health outcomes due to complications.

Summary

There was some relevant published literature concerning the cost-effectiveness of screening for coeliac disease in patients with suspected IBS. The study by Mein (2004) provided a cost per QALY for coeliac testing vs no testing from a US perspective. Whilst this could not be applied directly to the population under consideration, due to differences in the health care systems between the US and the UK, the low cost per QALY suggested that this intervention may also be cost-effective from a NHS and PSS perspective. The studies by Dubinsky (2002) and Suleiman (2001) did not consider directly whether further diagnostic testing would be cost-effective in patients meeting diagnostic criteria for IBS compared to no further diagnostic testing. They did provide some evidence that where diagnostic testing does take place, it is cost-effective to use less costly and less invasive tests first in the diagnostic sequence with positive results confirmed by standard invasive testing compared to invasive testing early in the diagnostic sequence.

Having considered the evidence on the clinical utility of diagnostic tests in patients meeting IBS diagnostic criteria, the GDG decided that there was insufficient evidence of clinical utility to warrant further economic analysis on the cost-effectiveness of diagnostic testing, except for serological testing for coeliac disease. It was not considered necessary to carry out further economic analysis to estimate the cost-effectiveness of routine laboratory investigations, such as FBC, ESR and CRP. They are unlikely to result in a significant cost burden for the NHS and it would be difficult to estimate their cost-effectiveness due to their non-specific nature which means that an abnormal result may be due to a variety of causes. Therefore, further economic modelling focused on the cost-effectiveness of serological testing for coeliac disease.

Cost-effectiveness of screening for coeliac disease in patients meeting IBS diagnostic criteria – adaptation of a published economic evaluation

Further analysis was carried out to adapt the cost-effectiveness estimate provided by Mein (2004) to make it more applicable to the NHS in England and Wales. UK specific data was obtained for the prevalence of undiagnosed coeliac disease, diagnosis costs, and HRQoL and ongoing resource use for individuals with IBS. A discounting rate of 3.5% was applied to both costs and QALYs in line with the NICE reference case for cost-effectiveness analysis (NICE 2007). Mein (2004) did not consider the additional cost of gluten-free foods in their analysis, but as gluten-free foods can be prescribed through the NHS this cost was also considered in our analysis. Mein (2004) did not allow for any increased life-expectancy that may result from adherence to a gluten-free diet in patients diagnosed with coeliac disease. This was considered to be overly conservative as one of the main aims of adherence to a gluten-free diet in coeliac disease is to reduce the risk of malignant diseases associated with coeliac disease such as Non-Hodgkin Lymphoma (West 2004). The model was therefore adapted to include an estimated survival difference between patients with diagnosed and undiagnosed coeliac disease. The economic model reported by Mein (2004) compared serological testing for IgA tissue transglutaminase (TTG) antibodies against a strategy of no testing. However IgA EMA testing is more commonly used in the UK than IgA TTG so this was used in the UK adaptation and TTG was considered in a sensitivity analysis.

The prevalence of undiagnosed coeliac disease in patients meeting IBS diagnostic criteria was taken from a cross-sectional study conducted in a UK primary care setting (Saunders 2003). The study population was randomly sampled from all adults entering the GP premises on study days. A subgroup of individuals meeting the ROME II diagnostic criteria for IBS was identified. The prevalence of undiagnosed coeliac disease was 3.3% (4/123) in individuals who fulfilled the ROME II criteria for IBS. This estimate was used in the model as the prevalence of undiagnosed coeliac disease in patients meeting IBS diagnostic criteria. The prevalence from the primary care sample as a whole (1%) was used in a sensitivity analysis as the expected lower limit for the prevalence in the IBS population.

Sensitivity and specificity values for IgA EMA were taken from a published health technology assessment which included a systematic review of autoantibody testing in children with type I diabetes (Dretzke 2004). This systematic review included studies carried out in symptomatic populations or populations at a higher risk of developing coeliac disease but not exclusively type I diabetes. The sensitivity and specificity estimates used in the model were the Q values (overall best test performance with equal sensitivity and specificity) from the well-described studies as given in Table 17 of Dretzke (2004).

The NHS cost of an IgA EMA antibody test was also taken from Dretzke (2004). The cost of esophagogastroduodenoscopy (EGD) with biopsy to confirm coeliac disease was taken from the NHS references costs (2005–06) for day case endoscopic procedures on the stomach or duodenum (Department of Health 2006). The cost for an EGD with complications was assumed to be equal to the NHS reference cost for the same procedure as a non-elective in patient (average length of stay of 1 day). The cost of care for IBS was taken from a study by Akehurst (2002) which estimated the NHS costs for IBS patients and matched controls. As in the cost-effectiveness analysis by Mein (2004), it was assumed that the NHS costs of managing coeliac disease are equal to the costs of managing IBS except that there is the additional cost of gluten free foods on prescriptions. This may be an overestimate if IBS-like symptoms are reduced when patients with coeliac disease are established on a gluten-free diet.

The NHS cost of supporting a gluten-free diet by providing foods on prescription was calculated by estimating the total cost of gluten-free foods prescribed by the NHS in England in 2005 (£21.2million) from the Prescription Cost Analysis (NHS Health and Social Care Information Centre 2006) and the number of people diagnosed with coeliac disease based on a population for England of 50.4million and a prevalence for diagnosed coeliac disease of 0.26%.(Fowell 2006). This gave an annual cost of £162 per diagnosed case of coeliac disease.

The health utility of IBS was taken from the study by Akehurst (2002) which estimated health utility for IBS patients using the EQ-5D. Mein (2004) attempted to estimate the utility gain associated with diagnosing coeliac disease in patients with IBS-like symptoms. However, this estimate was considered to be unreliable as it was calculated by comparing utility values for health states estimated in different populations. No direct evidence was available on the utility gain achieved by diagnosing and treating coeliac disease in patients with IBS-like symptoms. O’Leary (2002) found that coeliac patients with IBS-like symptoms had a lower HRQoL than those without symptoms, but these symptoms were equally common in coeliac patients who did and didn’t adhere to a gluten-free diet. Casellas (2005) found that recently diagnosed patients who had not started a gluten-free diet had a lower quality of life and a higher prevalence of IBS-like symptoms compared to patients who had been established on a gluten-free diet, but the study design was cross-sectional, so it was not possible to say from this whether the diet itself provided an improvement in quality of life. In the basecase analysis it was assumed that the gluten-free diet did not provide any gain in health utility, so the only benefit was from improved survival. A threshold analysis was carried out to assess the size of health utility gain that would need to be achieved by adherence to a gluten-free diet, in order to give a cost per QALY under £20,000, when assuming that the gluten-free diet does not provide a survival gain.

There is evidence that patients with coeliac disease have a significantly higher than expected mortality (SMR = 2.0, p<0.0001) (Corrao 2001) and that mortality risk is significantly increased for patients with a diagnostic delay of over 12 months but is not significantly increased when it is less than this. Mortality is also significantly higher than expected (SMR 2.5, p<0.0001) in patients with severe symptoms of malabsorption such as diarrhoea or weight loss but not significantly increased in patients with milder symptoms which may be seemingly unrelated to coeliac disease. There is evidence that survival is significantly reduced in the first 3 years after diagnosis but not beyond. The timing of the observed excess mortality may be due to excess deaths in patients who had extended diagnostic delay and who were not diagnosed until after symptoms had become severe. It may be possible to prevent the excess mortality in patients with IBS–like symptoms by prompt diagnosis through serological screening. In order to estimate the survival gain associated with prompt diagnosis we have assumed that undiagnosed cases have a reduced survival compared to diagnosed cases. We have taken the survival ratio for coeliac patients compared to the general public and used this to estimate the reduction in mortality avoided by prompt diagnosis. This is equivalent to a relative reduction in cumulative survival of around 2% over the first 3 years of the model for patients with coeliac disease presenting with IBS-like symptoms whose coeliac disease remains undiagnosed. This may have underestimated the survival gain associated with diagnostic testing, as the SMR in the whole coeliac population is lower than in those patients with extended diagnostic delay, but it may also have overestimated the survival gain as prompt diagnosis may not result in a complete reduction of mortality to general population levels. These survival ratios were applied to UK life-tables (Office for National Statistics 2006), assuming a male to female ratio of 1:2, and gave an estimated difference in expected life-years of 1.4 LYs for patients with diagnosed and undiagnosed coeliac disease. Once discounting was applied to the expected survival this difference was reduced to 0.54 discounted LYs. A sensitivity analysis was also carried out using the upper 95% CI of the survival ratio which resulted in a lower estimated survival gain of 0.31 discounted LYs.

The parameter values used in the UK adaptation are summarised in Table 11 alongside those used by Mein (2004) in the US basecase. Univariate sensitivity analysis was used to determine whether the deterministic estimate of cost-effectiveness was sensitive to changes in the UK specific parameters. This included a threshold analysis on the utility gain associated with establishing a gluten-free diet and the cost of prescribing gluten-free foods on the NHS. The parameter values used in the sensitivity analysis are also given in Table 11. All costs were uplifted to 2005/06 values where applicable and the uplifted values are included in Table 11 in italics.

Table 11

Parameters used in Mein (2004) analysis and in the UK adaptation.

A probabilistic sensitivity analysis (PSA) was carried out to estimate the uncertainty in the cost-effectiveness estimate due to the uncertainty in the model input parameters. We characterised the parameter uncertainty by using a probability distribution to describe each of the parameters, details of which can be found in Appendix H. We sampled randomly from these distributions 1000 times and estimated the model outputs (incremental costs and incremental QALYs) for each set of sampled parameters and used these to estimate the uncertainty surrounding the cost per QALY estimate. We based our PSA on 1000 samples of the parameter distributions. The results are presented as a cost-effectiveness acceptability curve (CEAC) which shows the proportion of samples that resulted in a cost per QALY value below various thresholds. It should be noted that the PSA does not account for uncertainty around the model assumptions and these have been explored separately using univariate sensitivity analysis.

The deterministic results for the UK basecase estimate are given in Table 12. The serological testing strategy identified 32 out of 33 prevalent cases of coeliac disease in the cohort of 1000 patients with IBS symptoms for a diagnosis cost of £36,300, giving a cost per correctly diagnosed case of coeliac disease of £1,122. These diagnoses resulted in an additional 43.4 LYs (undiscounted) over the lifetime of the cohort which is equivalent to 11.69 QALYs (discounted). This was associated with a further £122,300 (discounted) of treatment costs, including gluten-free products for patients diagnosed with coeliac disease, over the lifetime of the cohort. The overall cost per QALY for serological testing compared to no testing was £13,560 for a life-time horizon.

Table 12

Deterministic basecase results for 1000 patients meeting IBS diagnostic criteria. Costs, LYs and QALYs are discounted at 3.5% per annum.

The mean cost per QALY over the 1000 samples carried out for the probabilistic analysis was £14,300. The CEAC in Figure 1 shows the probability that the cost per QALY is under various cost per QALY thresholds given the uncertainty in the parameters used to estimate cost-effectiveness. It shows that the cost per QALY had an 80% probability of being under £20K per QALY and a 96% probability of being under £30K per QALY under the basecase assumptions.

Figure 1

Cost-effectiveness acceptability curve (CEAC) for coeliac disease testing in patients with IBS-like symptoms compared to no testing.

The univariate sensitivity analysis in Figure 2, shows that the cost per QALY estimate was not particularly sensitive to the age of the patient at presentation. This may be because younger patients have a longer life-expectancy, but this increases both their lifetime cost of care and their survival gain from preventing excess mortality. Cost-effectiveness was not significantly impacted by higher testing costs, higher costs for ongoing IBS / coeliac disease management, lower health state utility values for patients with IBS / coeliac disease or lower sensitivity and specificity values for serological testing. Using a zero discounting rate lowered the cost per QALY as the majority of the survival benefit was gained over the long-term whilst the upfront diagnosis costs occurred early in the model.

Figure 2

Univariate sensitivity analysis results for coeliac disease testing in patients with IBS-like symptoms compared to no testing.

The cost-effectiveness estimate was sensitive to the survival gain attributed to identifying patients with coeliac disease and establishing them on a gluten-free diet, as this was the only benefit included in the basecase model. Using the lower estimate of survival benefit increased the cost per QALY to £23,000. The threshold analysis on utility gain demonstrated that establishing patients with coeliac disease on a gluten-free diet would need to produce a utility gain of 0.011 in order for the cost per QALY to remain under £20,000, when assuming that there is no survival gain. This is a small utility gain compared to the difference in health utility between IBS patients and matched controls (0.135, Akehurst 2002). These two sensitivity analyses on the survival and QALY gain demonstrated that whilst there is some uncertainty surrounding the expected benefits of identifying individuals with coeliac disease and initiating a gluten-free diet, testing is likely to be cost-effective in patients with IBS-like symptoms, so long as there is a small improvement in quality of life or a small reduction in mortality risk as a result of a correct diagnosis of coeliac disease.

The threshold analysis on the cost of prescribing gluten-free foods shows that up to £263 per patient per annum could be spent on gluten-free foods before the cost per QALY reached the threshold of £20,000. The estimated cost for providing gluten-free foods on prescription is based on the total costs of prescriptions for gluten-free foods in 2005 and an estimate of the prevalence of diagnosed coeliac disease. Using the lower estimate of prevalence from Fowell (2006) gave a higher cost of £234 per patient per annum, which based on our threshold analysis, would still provide a cost per QALY under £20,000.

We have estimated the cost-effectiveness of testing with EMA compared to no testing as this is the test most commonly available to primary care clinicians in the UK. However, TTG is also available in some areas of the NHS. Sensitivity and specificity values were available for TTG (96%, 95% CI of 92%–98%) from the Dretzke (2004) HTA, but a direct cost estimate for TTG was not available. In the economic analysis conducted as part of the HTA, the cost of TTG was assumed equal to the cost of testing for anti-gliadin antibodies (AGA) as these tests use similar techniques. The cost for these tests was estimated to be slightly higher than the cost of EMA. We carried out a sensitivity analysis to see whether testing using TTG would also be cost-effective when using the evidence on test cost and accuracy from the HTA (Dretzke 2004). The slightly lower accuracy for TTG resulted in a slightly lower QALY gain of 11.53 per 1000 people tested, for testing compared to no testing. The slightly higher test cost (£14 compared to £12) resulted in a slightly higher total cost £164,683 per 1000 people tested. The overall cost per QALY for TTG compared to no testing was therefore, £14,283. This suggests that testing with TTG would also be cost-effective compared to no testing. We have not carried out an analysis to consider which antibody test is the more cost-effective test to use as we did not feel that the cost data was sufficiently robust to allow a reliable comparison. In addition, TTG is a relatively new technology and the evidence base may have improved since the searches carried out by Dretzke (2004). There are also other factors that must be taken into account when deciding which tests should be available to primary care clinicians in the NHS, which we have not considered. Therefore we did not feel that our cost-effectiveness analysis was sufficiently robust to recommend the use of either test in preference to the other. However, there is good evidence that using either of these tests is cost-effective compared to no testing in people with IBS.

Duration of symptom profile

Having reviewed the evidence and analysed application of the criterion referenced diagnostic tools, duration of symptom profile was recognised to be an important aspect to consider in making recommendations for practice. Three, six and twelve month durations were all discussed and the consensus of the group was that a duration of 6 months was the most appropriate.

Entry filter for use of the diagnostic tool

Primary care clinicians should consider assessment for IBS if the patent reports any of the following symptoms for at least 6 months:

• Change in bowel habit
• Abdominal pain/discomfort and or bloating.

Positive Diagnostic Criteria

For a positive diagnosis to be made, the patient must present with at least 3 of the agreed diagnostic criteria. Language used in the tool was contemporised as follows:

• Pain was modified to include discomfort
• Pain/discomfort was changed to recurrent/episodic, experienced for at least 6 months duration
• Abdominal distension/bloating/abdominal tension was added.

Supportive investigations

Appropriate investigations identified were:

• FBC (full blood count)
• ESR (erythrocyte sedimentation rate)
• CRP (inflammatory marker)
• Antibody testing for Coeliac disease (EMA or TTG)

Follow-up

Once a positive diagnosis has been made, patient follow up is a key aspect of longer term management in managing and evaluating the response to first line therapy interventions. Giving patients the opportunity either to re-attend as required or possibly making regular appointments was discussed. It was agreed by the GDG that follow up should be explicitly stated within the recommendations, and in the absence of any evidence supporting this, consensus agreement would be used.

IBS GDG MEETING 30^th November- 1^st December 2006