Faecal calprotectin diagnostic tests for inflammatory diseases of the bowel

5.10 Seven studies gave results that compared IBS and IBD, at 8 cut‑off levels ranging from 8–150 micrograms/g, all in adults in secondary care. All studies assessed enzyme‑linked immunosorbent assay (ELISA) tests, and one also assessed the performance of the point‑of‑care test (POCT) CalDetect. As expected, low cut‑offs gave high sensitivity for IBD but poor specificity. Sensitivity was consistently high (usually 100% at levels under 50 micrograms/g; ranging from 83–100% at a cut‑off of 50 micrograms/g), but specificity was more varied (51–100%).

5.11 Many of the studies had a small sample size. The largest study was by Li et al. (2006), which employed a sample of 240 people. Studies were of mixed quality. Schroder et al. (2007) and Schoepfer et al. (2008) were assessed as having the least risk of bias.

5.12 Five studies reported data for faecal calprotectin testing with ELISA with a cut‑off of 50 micrograms/g. This allowed for the meta‑analysis of the studies to provide an overall combined estimate of sensitivity and specificity. The combined estimates for ELISA tests, at a 50 micrograms/g cut‑off, were a sensitivity of 93% and a specificity of 94%. The meta‑analysis estimates were informed by a pool of 596 people, of which 40% were from the Li et al. (2006) study. The mean age of people in these studies, when reported, ranged from 40–52 years in people with IBS and 34–45 years in people with IBD. However, the age of people in the Schoepfer et al. (2008) study went as high as 78 years.

5.13 The only study using a POCT was Otten et al. (2008), which assessed the CalDetect test in a sample of 114 people. Otten et al. showed that the test performed well at a cut‑off of 15 micrograms/g, with a sensitivity of 100% and a specificity of 95%. At a cut‑off of 60 micrograms/g, although specificity improved slightly to 98%, sensitivity was only 61%, which the External Assessment Group considered to be unlikely to be acceptable in clinical practice given the importance of not missing people with IBD. The average age of people in the Otten et al. study was 52 years in people with IBS and 45 years in people with IBD.

5.14 The cost effectiveness of faecal calprotectin testing for distinguishing between IBD and IBS in an adult population in primary care was assessed in the economic evaluation conducted by the External Assessment Group.

5.15 Eleven studies reported IBD compared with non‑IBD, at 8 cut‑off levels. Eight studies were conducted in paediatrics and 3 in adults. All used ELISA tests, and one (Damms and Bischoff 2008) also assessed the Prevista POCT (not identified in the scope for the assessment).

5.16 The studies showed consistently high sensitivity at lower cut‑offs, nearly all over 90%, with most at the 50 micrograms/g cut‑off having sensitivities of 100%. Specificity was more varied, ranging from 44–93% at a 50 micrograms/g cut‑off. Most of these results were in paediatric groups. Most studies reported results at only 1 cut‑off, but 1 study reported 5 cut‑offs and another 4, both in paediatric populations. Studies were of mixed quality with Canini et al. (2006), Diamanti et al. (2010), Fagerberg et al. (2005), Henderson et al. (2012) and van de Vijver et al. (2012) assessed as having the least risk of bias compared with the other studies.

5.17 Six separate estimates of sensitivity and specificity were available at a cut‑off of 50 micrograms/g and another 6 estimates at 100 micrograms/g, which allowed the individual estimates to be meta‑analysed into combined overall estimates of sensitivity and specificity for ELISA tests. The overall pooled results for IBD compared with non‑IBD showed very high sensitivity of 99% but moderate specificity of 74% at a cut‑off of 50 micrograms/g. These estimates were informed by a pool of 531 people with most of these studies including people up to the age of 18 years. At a cut‑off of 100 micrograms/g, sensitivity was found to fall to 94% but specificity to improve to 82%. These estimates were informed by a pool of 656 people; however, the upper age limit varied in these studies. Two studies recruited people up to the age of approximately 15 years, 2 studies up to the age of 18 years and 1 study up to an age of 20 years. The age limit was not reported in the sixth study.

5.18 The cost effectiveness of faecal calprotectin testing for distinguishing between IBD and non‑IBD in a paediatric population in secondary care was assessed in the economic evaluation conducted by the External Assessment Group.

5.24 The distribution of faecal calprotectin values is highly skewed and a wide range can be observed. Low levels may be seen in people with IBD and raised levels may be seen in people with IBS/non‑IBD (for example, people with infectious gastroenteritis or food poisoning).

5.25 In some studies, the ranges did not overlap, but in others they did. For example, in El‑Badry et al. (2010), the value of faecal calprotectin in people with IBD ranged from 98–637 micrograms/g, which did not overlap with the value of faecal calprotectin in people with IBS (14–65 micrograms/g). In all other studies, the range of faecal calprotectin in patients with IBD overlapped with the range of faecal calprotectin in patients with IBS. In some studies, such as Li et al. (2006) and Schroder et al. (2007), the range of faecal calprotectin levels in people with IBD was wide, with the lowest value being 15 micrograms/g and the highest being 2574 micrograms/g.

5.26 The range of results in studies comparing IBD and non‑IBD in children was similar to that found in studies comparing IBD and IBS in adults. In some studies (Canini et al. 2006; Diamanti et al. 2010; Sidler and Leach 2008), the ranges of faecal calprotectin levels overlapped in children and faecal calprotectin levels were high.

5.27 The External Assessment Group noted that faecal calprotectin levels were often raised in conditions other than IBD, such as larger colorectal adenomas and some colorectal cancers. The accuracy of faecal calprotectin testing is lower in these other conditions when compared with IBD.

5.28 The External Assessment Group summarised several studies that evaluated the comparative performance of faecal calprotectin tests in particular situations. For example, some studies assessed the performance of the tests for distinguishing IBS from IBD and others assessed the tests in distinguishing organic from non‑organic disease.

5.29 Overall, the External Assessment Group concluded that there are limited data comparing the performance of different faecal calprotectin tests. Of the studies conducted, they concluded that none suggested any considerable differences between the various faecal calprotectin tests.

5.34 The model uses a linked-evidence approach to combine the outcomes of diagnostic strategies with the management (induction therapy and remission patterns) of patients' conditions, to allow the estimation of clinical outcomes and QALYs. The model assesses multiple diagnostic strategies and allows for multiple test sequences to be considered (for example, an initial ELISA test followed by colonoscopy). The outcomes from the diagnostic pathway are linked to the care pathway following diagnosis. Patients with true positive results for Crohn's disease and ulcerative colitis are considered separately from one another because patients in these groups follow different and complicated induction and remission pathways post diagnosis. Both patients with true negative and with false negative results follow the care pathway for IBS, with those patients whose disease does not respond to dietary changes after advice and subsequent medical treatment for IBS being retested for IBD. Patients with false positive results (incorrectly diagnosed as having IBD) are eventually correctly diagnosed as having IBS, given that it is assumed all patients with false positive results are referred for specialist investigation and undergo a colonoscopy (assumed 100% specificity). The model employs a weekly cycle and adopts a 10‑year time horizon.

5.35 Although the scope allowed for the assessment of faecal calprotectin testing for both adults and children in both primary and secondary care, the External Assessment Group modelled 2 specific populations: an adult population in primary care, with faecal calprotectin test accuracies for IBD compared with IBS, and a paediatric population in secondary care, with faecal calprotectin test accuracies for IBD compared with non‑IBD. The External Assessment Group believed these populations reflect the most likely use of faecal calprotectin testing in clinical practice.

5.36 The main aim of the model was to assess the impact of faecal calprotectin testing when added to current clinical practice compared with current practice alone on the differentiation of IBD and IBS in primary care. This model was then adjusted to reflect the differing test performances and costs in the paediatric population to provide an approximation of the cost effectiveness of faecal calprotectin testing for distinguishing between IBD and non‑IBD. However, the External Assessment Group highlighted the limitation of this approach because the main model structure does not fully account for the non‑IBD case mix in the paediatric population (prevalence of IBS in the non‑IBD group is lower than that seen in adults).

5.37 The use of an ELISA faecal calprotectin testing service was evaluated in the base case for both of the primary and secondary care scenarios. The POCT CalDetect was evaluated in the base case primary care scenario.

5.38 The base case applied the quality‑of‑life decrements from remission to active disease of 0.280 for Crohn's disease and 0.200 for ulcerative colitis from Stark et al. (2010). But sensitivity analyses applying the quality‑of‑life decrements from mild‑to‑moderate disease of 0.075 for Crohn's disease, as drawn from Gregor et al. (1997), and of 0.165 for ulcerative colitis, as drawn from Poole et al. (2010), were also explored. The utility decrements for IBS were less important for modelling purposes, given that the 100% specificity assumed for colonoscopy meant that there were no patients with false positive results by the end of the test sequence. For the base case, the 0.071 increment for response to treatment estimated in the NICE guideline on irritable bowel syndrome in adults was applied. The 0.662 baseline HRQoL that this increment was applied to was taken from Brazier et al. (2004). A sensitivity analysis using values from Spiegel et al. (2009) was also considered: 0.780 for response to treatment and 0.730 for no response to treatment, but the algorithm used to construct the EQ‑5D utilities was not clear. The baseline HRQoL value for IBS has an impact because of the small mortality rate associated with colonoscopy, with this impact enduring for the 10‑year time horizon of the model.

5.39 Because of data constraints, the cost impacts were limited to modelling the relatively rare (less than 0.5%) serious adverse events of bleeds and perforations. The quality‑of‑life impacts were limited to the mortality associated with perforations. While perforations are rare, resulting in a very low mortality rate, the QALY impact of this persisted for the duration of the model.

5.40 There is evidence from the literature that colonoscopies result in minor adverse events among a reasonable proportion of patients; for example, de Jonge et al. (2012) suggested that around 40% of those investigated with colonoscopy have some effects persisting for 30 days after the colonoscopy. In common with the NICE guideline on colonoscopic surveillance for the prevention of colorectal cancer, these minor adverse events were not taken into account in the modelling principally because of a lack of quality‑of‑life data.

5.41 The costs included in the model were the costs of the different tests, treatment costs (including induction therapy and maintenance therapy costs for people in remission), NHS resource costs (for example, staff time) and costs of adverse effects associated with colonoscopy.

5.42 The per person costs of an ELISA test and POCT CalDetect were estimated to be £22.79 (based on an assumption of 40 patient samples per 96 well‑plate, costed at the list price, plus an average 11–12 minutes of staff time at grade 6/7) and £24.03 (test list price plus cost of 15 minutes of GP practice‑nurse time) respectively.

5.43 Colonoscopy was estimated to cost £741.68 per person. This estimate was based on a weighted average of the NHS reference cost for outpatient and day cases without biopsy (procedures payment by results code FZ51Z/FZ54Z), or with biopsy (procedures payment by results code FZ52Z/FZ55Z) for colonoscopy or, when used, sigmoidoscopy. The cost included an outpatient gastroenterology appointment (£164) and costs of adverse effects (an average of £12 per colonoscopy).

5.45 For the primary care adult population, the model adopted a baseline age of 25 years for those presenting with symptoms, as used in the NICE guideline on Crohn's disease. Consistent with the modelling in this guideline, the proportion of females was taken to be 50% for both Crohn's disease and ulcerative colitis. It appears that a higher proportion of people with IBS are female; in the Brazier et al. (2004) sample, 86% were female, although the External Assessment Group suggested this estimate may be towards the upper end. For IBS, the base case adopted a proportion of females of 75%. These estimates only affect the all‑population mortality risks. Because these risks are low during mid-adulthood, for both women and men, the average age and proportion of model inputs for women will have had a minimal impact on the results.

5.46 The base-case prevalence of IBD (6.3%) was drawn from the Durham Dales data, while the prevalence of ulcerative colitis among patients with IBD (a ratio of ulcerative colitis to Crohn's disease of 3:2) was drawn from Shivananda et al. (1996).

5.47 The strategies assessed were:

5.48 The External Assessment Group assumed, given lack of evidence to the contrary, that the accuracy of ELISA testing is the same as would be obtained from ELISA testing in conjunction with GP current practice.

5.49 The delay between referral and colonoscopy was assumed to be 4 weeks and the time to retesting among those with negative tests but not responding to IBS therapy was assumed to be 12 weeks, both estimates being based on expert opinion. The latter estimate may have been optimistic because a sequence of unsuccessful treatments might be tried for people with IBS. This was explored in sensitivity analyses.

5.50 The modelling assumed that all people who test positive or have an indeterminate result are referred to secondary care and all of these people receive a colonoscopy (indeterminate results are treated as if the results are determinate). Because of a lack of data, the External Assessment Group was not able to incorporate the impact of a gastroenterologist's assessment on the number of people who will go on to receive a colonoscopy (which may also include the use of faecal calprotectin testing) in the base case; however, this is explored in sensitivity analysis.

Abbreviations: CI, confidence interval; ELISA, enzyme-linked immunosorbent assay; POCT, point-of-care test.

5.51 Without faecal calprotectin testing, GP current practice is highly sensitive in terms of referring people with IBD and is as good as, if not better, than faecal calprotectin testing. Of the 6.3% of people with IBD in the total population, all were identified by the GP current practice strategy and the POCT CalDetect strategy. Colonoscopy would correctly identify 6.0% of the 6.3% referred as patients with true positive results (because of its 95% sensitivity), resulting in a total of 0.3% of patients with false negative results. ELISA testing is slightly worse, identifying 5.9% of the 6.3% (because of its lower sensitivity when compared with current practice and the POCT), with 0.4% of patients being classified as having false negative results. Of the 5.9% referred for colonoscopy, 5.6% of patients would be identified as having true positive results, with 0.3% being classified as having false negative results, resulting in a total of 0.7% of patients with false negative results. Therefore, a slightly larger number of people will have IBD but will be incorrectly diagnosed as having IBS when using an ELISA testing strategy when compared with current practice strategy and a POCT CalDetect strategy (0.7% compared with 0.3%).

5.52 Within the total patient population, GP current practice incorrectly identified 19.8% of patients as having false positive results (people thought to have IBD but who actually have IBS) and requiring referral for colonoscopy. The rates of patients with false positive results incorrectly referred for colonoscopy for POCT CalDetect and ELISA were much lower, at 5.1% and 5.6% respectively. Therefore, without faecal calprotectin testing, many of the patients with false positive results would go on to have a colonoscopy, which has a risk (although low) of serious complications such as perforation. Such events are too rare to significantly affect costs, but they do have some QALY impact. This is also true for the more common minor adverse effects of colonoscopy (which were not explicitly considered in the model because of a lack of data).

5.53 Taking the diagnostic performance of the different testing strategies summarised above into account, the resulting total per patient costs and QALYs are provided in tables 3 to 5.

Abbreviations: IBS, irritable bowel syndrome; QALY, quality-adjusted life year.

5.54 The faecal calprotectin tests were estimated to result in similar average cost savings compared with GP current practice: £83 for the POCT CalDetect and £82 for ELISA per patient. This was mainly because of the lower number of referrals and colonoscopies for false positive results. Average QALY gains of around 0.0007 QALYs were also accrued, although these were limited because the low prevalence of IBD and the similar high sensitivities of the tests resulted in relatively few false negative results. Therefore, the faecal calprotectin testing strategies dominated current practice (provided greater benefit at reduced cost). Some of the QALY differences accrued were from the very slightly lower mortality associated with the lower number of colonoscopies. The POCT CalDetect and ELISA strategies were estimated to be broadly equivalent in terms of costs and QALYs, with only minor differences between them.

5.55 A range of sensitivity analyses were conducted to explore the impact of varying the main model parameters. These included: varying the prevalence of IBD between 5–25% (6.3% used in the base case); changing the source of utility values; adjusting the costs of colonoscopy (no outpatient appointment cost) and removing any associated mortality; varying the number of patients whose condition did not respond to IBS medication; varying the time it takes for patients with false negative results to re‑present to the clinician (8, 16 and 24 weeks; 12 weeks was used in the base case); and exploring the impact of varying the specificity of the consultant's diagnosis at an outpatient clinic assessment in people referred with a false positive diagnosis. Scenario analyses were also undertaken using different sources of test accuracy for faecal calprotectin and alternative assumptions surrounding the uptake of faecal calprotectin testing (assuming 50% of patients are tested as opposed to the 25% used in the base case) in primary care.

5.56 The sensitivity and scenario analyses appeared to broadly affect the 3 strategies in a similar way and suggested that the results of the base case were reasonably robust. The most notable impact was from assuming 50% of patients are tested in primary care, as opposed to 25%, which reduced the cost savings with faecal calprotectin testing.

5.58 For the secondary care paediatric population, the proportions of females included were 38% (35/91) for patients with IBD and 44% (44/99) for patients without IBD; these were drawn from Henderson et al. (2012). An average age of 16 years was assumed because, for the adult modelling, this had minimal impact on results.

5.59 In the base case, a 48% (91/190) prevalence of IBD and a 75% (62/83) prevalence of Crohn's disease among patients with IBD were drawn from Henderson et al. (2012).

5.60 The strategies assessed were:

Abbreviations: CI, confidence interval; ELISA, enzyme-linked immunosorbent assay.

5.61 The base-case prevalence of IBD of 47.9% increased the importance of test sensitivities compared with the primary care setting, and so the effect of false negative results on the modelling outputs. Within the total patient population, ELISA with the 50 micrograms/g cut‑off led to 47.4% of patients with true positive results being referred for colonoscopy, while ELISA with the 100 micrograms/g cut‑off led to 45.0% of patients with true positive results being referred for colonoscopy. Colonoscopy was assumed to have a sensitivity of 95%. So, if all (47.9%) patients were referred immediately for colonoscopy, 45.5% would be diagnosed with IBD. With ELISA with the 50 micrograms/g cut‑off, 45.0% of the 47.4% of patients referred for colonoscopy were diagnosed as having IBD, while 42.8% of the 45.0% of patients referred for colonoscopy after ELISA with the 100 micrograms/g cut‑off were diagnosed as having IBD; a net difference between the cut‑offs of 2.2%.

5.62 Despite the higher IBD prevalence in the secondary care population, the main test differences still lay in the number of unnecessary colonoscopies. Without faecal calprotectin testing, all 52.1% of patients without IBD received a colonoscopy, compared with 13.5% for ELISA with the 50 micrograms/g cut‑off and only 9.4% for ELISA with the 100 micrograms/g cut‑off.

5.63 Taking the diagnostic performance of the different testing strategies summarised above into account, the resulting total per patient costs and QALYs are provided in tables 7 to 9.

Abbreviations: IBD, inflammatory bowel disease; QALY, quality-adjusted life year.

5.64 Prior testing using ELISA was estimated to dominate (provided greater benefit at reduced cost) the strategy of sending all patients directly for colonoscopy. Compared with referring all patients directly for colonoscopy, ELISA used at the 50 micrograms/g cut‑off was estimated to save £205 per patient, while ELISA used at the 100 micrograms/g cut‑off was estimated to save £240 per patient. QALY gains of around 0.001 QALYs were estimated for ELISA compared with direct referral for colonoscopy, these being slightly larger for ELISA with the 50 micrograms/g cut‑off because of its better sensitivity. The additional average cost of £35 and additional average QALYs of 0.0001 for ELISA with the 50 micrograms/g cut‑off compared with ELISA with the 100 micrograms/g cut‑off resulted in an incremental cost‑effectiveness ratio (ICER) of £35,000 per QALY gained.

5.65 A range of sensitivity analyses were conducted to explore the impact of varying the main model parameters. These included: varying the prevalence of IBD to 40% and 60% (48% used in the base case); changing the source of utility values; removing any associated mortality of colonoscopy; varying the time it takes for patients with false negative results to re‑present to the clinician (8, 16 and 24 weeks; 12 weeks was used in the base case); and changing the annualised net cost of false negative results to £376 (£188 was used in the base case).

5.66 As for primary care, most of the changes appeared to broadly affect the 3 strategies in a similar manner. The main difference arose from varying the prevalence of IBD, which tended to reduce the cost savings from faecal calprotectin testing because of the rise in prevalence, as would be anticipated. The source of utilities also had an impact on the anticipated net gain from ELISA with the 50 micrograms/g cut‑off compared with ELISA with the 100 micrograms/g cut‑off, the ICER for which increased to £117,000 per QALY gained. However, the External Assessment Group thought that this may have overstated the effect, given the prevalence of Crohn's disease within the presenting population and the perhaps rather small quality‑of‑life decrement sourced from Gregor et al. (1997).