Managing the long term care of inflammatory bowel disease patients: The cost to European health care providers

Transcription

1 Journal of Crohn's and Colitis (2011) 5, available at Managing the long term care of inflammatory bowel disease patients: The cost to European health care providers James Buchanan a,, Sarah Wordsworth a, Tariq Ahmad b, Angela Perrin c, Severine Vermeire d, Miquel Sans e, Jenny Taylor f, Derek Jewell c a Health Economics Research Centre, University of Oxford, Oxford, UK b Royal Devon and Exeter Hospital, Exeter, UK c Gastroenterology Unit, John Radcliffe Hospital, University of Oxford, Oxford, UK d University Hospital, Leuven, Belgium e Hospital Clinic de Barcelona, Barcelona, Spain f NIHR Oxford Biomedical Research Centre, Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, UK Received 5 November 2010; received in revised form 2 February 2011; accepted 2 February 2011 KEYWORDS Inflammatory bowel disease; Ulcerative colitis; Crohn's disease; Health economics; Costs Abstract Background and aims: Inflammatory Bowel Disease (which includes Crohn's Disease and Ulcerative Colitis), is a chronic condition characterised by substantial morbidity. Inflammatory Bowel Disease patients are considered expensive to manage, hence accurate estimates of care costs are crucial to help healthcare providers plan clinical management. The aim of this study is to estimate the cost of care for Crohn's Disease and Ulcerative Colitis patients in the United Kingdom and Western mainland Europe. Methods: Decision models were built to simulate the natural disease history of Crohn's Disease and Ulcerative Colitis, informed by United Kingdom and European clinical pathways. A healthcare provider perspective was adopted, model inputs were informed by published sources and expert opinion, and UK healthcare costs were used (2008 prices). Cohorts of 25 year old patients presenting with symptoms of varying severity were modelled over ten years, and annual treatment costs calculated per patient. Results: The average annual cost of care per Crohn's Disease/Ulcerative Colitis patient was 631/ 762 (United Kingdom) and 838/ 796 (Europe). Most costs were incurred immediately following diagnosis, particularly in European Crohn's patients, reflecting the earlier use of more aggressive treatments. Surgery, hospitalisation, and the use of biological therapies and Abbreviations CD, Crohn's Disease; IBD, Inflammatory Bowel Disease; NHS, National Health Service; PSA, Probabilistic Sensitivity Analysis; UC, Ulcerative Colitis Corresponding author at: Health Economics Research Centre, Department of Public Health, University of Oxford, Old Road Campus, Headington, Oxford, OX3 7LF, UK. Tel.: ; fax: address: james.buchanan@dphpc.ox.ac.uk (J. Buchanan) /$ - see front matter 2011 European Crohn's and Colitis Organisation. Published by Elsevier B.V. All rights reserved. doi: /j.crohns

2 302 J. Buchanan et al. mesalazine (in Ulcerative Colitis) were key cost drivers. The total annual cost to the United Kingdom National Health Service of caring for Inflammatory Bowel Disease patients was estimated to be 131 million. Conclusions: This study confirms that Inflammatory Bowel Disease patients are expensive to manage and illustrates the importance of differentiating between alternative clinical management scenarios European Crohn's and Colitis Organisation. Published by Elsevier B.V. All rights reserved. 1. Introduction Inflammatory Bowel Disease (IBD), which includes Ulcerative Colitis (UC) and Crohn's Disease (CD), is a chronic condition characterised by substantial morbidity. Prevalence in the United Kingdom (UK) is estimated to be between 160 and 240 cases per 100,000 for UC, and between 55 and 140 for CD. Around 150,000 people are estimated to have IBD in the UK, rising to 2.2 million people across Europe (including both EU and non-eu countries). 1,2 UC and CD are characterised by periods of remission and relapse, the incidence of the latter varying by disease severity. During relapse patients often suffer substantial morbidity and require intensive treatment, including invasive investigations, costly drugs (e.g. Infliximab) and surgery (e.g. intestinal resection). Patient response to these treatments varies considerably and is unpredictable, with many widely used treatments associated with a high incidence of relapse. 3 Tailoring disease management to individual patient characteristics can be time consuming, and neither medical treatment nor surgery offer a cure for IBD. Consequently, patients with IBD are considered to be expensive to manage. 4,5 Reliable estimates of resource use and treatment costs are crucial to allow healthcare commissioners and providers to plan and deliver a comprehensive IBD service for their population. However, few studies have attempted to estimate the total cost of IBD care, focusing instead on the cost-effectiveness of specific interventions and particular aspects of clinical management (e.g. Infliximab, management of perianal fistulae). 6 8 While these studies could help to guide clinical decision-making, a much broader analytical perspective is now needed, driven largely by the development of high-cost IBD-related interventions with the potential to impact on all aspects of clinical management (e.g. genetic diagnostic tools such as microarrays (DNA chips)). Alongside these novel genomic technologies, new biological therapies (e.g. certolizumab) are also likely to become available, coupled with p4 p1 PRESENT FOR TREATMENT p2 Symptomatic pre diagnosis workup p7 Symptomatic - severe attack p17 Steroid controlled remission p15 p3 Symptomatic mild/moderate* p11 5-asa non-responder* p6 p13 5-asa controlled remission p9 p10 5-asa controlled relapse p8 p5 Tapered treatment after severe attack* p16 p19 Steroid controlled relapse* p18 p14 Steroid non-responder p23 p20 p21 Steroid non-responder in remission p22 p26 p12 Immunosuppressant non-responder p27 p24 p25 Immunosuppressant nonresponder in remission p30 IV steroid nonresponder p31 p28 p29 IV steroid nonresponder in remission Surgery p36 p33 Postsurgical remission p34 p35 p32 Postsurgical complication Figure 1 Ulcerative Colitis Markov model. The patient cohort begins in Present For Treatment. All other boxes represent potential disease states, described in Table A1. Arrows indicate possible transitions between health states. Transition probabilities (p1 to p39) are outlined in Table 1. In health states denoted with (*), patients receive tapered treatment see Tables 3 and A1 for further details.

3 Care costs for inflammatory bowel disease patients 303 greater intensity of use of existing therapies These two effects are likely to result in an increase in the cost of IBD care. Furthermore, existing studies of the cost of IBD care are likely to be of limited use when considering the impact of these new interventions. Differences in clinical practice ensure that many previously generated cost estimates have limited applicability outside of a particular study setting. Thus studies based on data collected from a single hospital or region (e.g. US surveys of health insurance databases) cannot be widely applied Other studies do not account for key cost drivers (e.g. drugs) and often exclude the most important patient subgroups (e.g. the most severely ill). 13 Many past estimates are also outdated, predating the widespread use of expensive biological therapies. 4,17 Our study aimed to develop an economic framework to calculate the cost of IBD care. The objectives of the framework were to: provide a total cost of caring for UK IBD patients; be applicable in different settings and be able to reflect variations in clinical practice, and be capable of considering the full impact of novel IBDrelated interventions. This paper reports the results of an analysis of the clinical management of IBD patients in the UK using the framework that was developed, and considers the cost impact of variations in clinical practice in Western mainland Europe. 2. Materials and methods 2.1. Clinical pathways and model design Data to construct a framework of IBD clinical management were derived from several sources. Interviews were conducted with gastroenterologists in two UK hospitals (John Radcliffe Hospital, Oxford (DPJ) and the Royal Devon and Exeter Hospital, Exeter (TA)) to map the typical clinical pathways of UC and CD patients in a UK setting from symptomatic presentation, through diagnosis and treatment, and finally to monitoring during remission (or further treatment due to relapse). Patient subgroups were considered (with mild, moderate or severe disease) and differences in clinical management noted. A distinction was made between CD patients with ileal/ileo-colonic disease and those with colonic disease. Resource use (staff, capital, consumables etc.) was estimated at all points in the pathways, along with the timing of patient progression through the pathways. A Stoma and Colorectal Nurse Specialist (AP) provided additional data on patient resource use post-surgery. The UK pathways were then reviewed by gastroenterologists in Barcelona, Spain (MS, Hospital Clinic de Barcelona) and Leuven, Belgium (SV, University Hospital), partners in an EU-funded IBD Consortium Project. Key differences in clinical management were identified, and a second set of clinical pathways specific to Western mainland Europe (the EU15 countries, except for the UK) were developed. Both pathways accounted for patients who were treated with biological therapies, and the pathway for Western mainland Europe accounted for the increased use of immunosuppressants as first-line therapy for CD patients presenting with moderate to severe disease. Infliximab was considered to be a treatment option in 20% of UC patients presenting with a severe attack and not responding to IV steroids in both settings (equal to 4% of all UC patients, with patients in Western mainland Europe receiving additional cycles of treatment). It was also considered to be a treatment option for a proportion of UC patients with no response to 5-asa drugs, steroids or immunosuppressants. Both Infliximab and adalimumab were considered to be treatment options for the small proportion of CD patients presenting with mild disease who do not respond to 5-asa drugs, steroids or immunosuppressants 18, and they were also considered p4 PRESENT FOR TREATMENT p1 p2 p8 Symptomatic mild pre diagnosis workup Symptomatic mild* p5 p3 p6 p9 p10 Symptomatic moderate/severe pre diagnosis workup p7 Symptomatic moderate/severe* p16 p11 Azathioprine controlled remission p12 p18 p17 First-line drug nonresponder mild* p14 Second-line drug nonresponder mild* p13 p20 Long-term remission p22 First-line drug nonresponder moderate/severe* p23 Azathioprine controlled relapse p19 p21 p28 Surgery p25 Postsurgical remission p26 p24 p27 Postsurgical complication p15 Figure 2 Crohn's Disease Markov model. The patient cohort begins in Present For Treatment. All other boxes represent potential disease states, described in Table A1. Arrows indicate possible transitions between health states. Transition probabilities (p1 to p31) are outlined in Table 2. In health states denoted with (*), patients receive tapered treatment see Tables 3 and A1 for further details.

4 Table 1 Transition probabilities ulcerative colitis. Prob a Transition Value b Notes Source c From To p Probability of presenting with mild/moderate E symptoms p Probability of presenting with severe symptoms E, 22,47 p3 Symptomatic pre-diagnosis workup Symptomatic mild/moderate 0.20 Proportion of starting population who make E transition each cycle. Increments by 0.2 each cycle (estimated to take 1 5 cycles for all patients to move (average 3)) p4 Symptomatic pre-diagnosis workup Symptomatic pre-diagnosis workup (1-p3) Proportion of starting population who remain in this stage after one cycle. p5 Symptomatic severe attack Tapered treatment after severe attack 0.70 E p6 Symptomatic mild/moderate 5-asa controlled remission Variable Transition probabilities p6 and p11 can take varying Various values depending on the treatment options used: (i) 5-asa patients (32.5% of all symptomatic mild/ moderate patients [E]) spend only one cycle in this state, after which 64% move into 5-asa controlled remission [E]; the remainder move into the 5-asa non-responder state. (ii) The remaining patients (67.5% of all patients [E]) are treated with prednisolone (tapered treatment). After 3 cycles 64% move into 5-asa controlled remission [E, 21,22,48,49 ]; the remainder move into the 5-asa non-responder state. p7 5-asa controlled remission 5-asa controlled remission (1-p9-p39) p8 Tapered treatment after severe attack 5-asa controlled remission (1-p39) p9 5-asa controlled remission 5-asa controlled relapse 0.05 E, 22 p10 5-asa controlled relapse 5-asa controlled remission p11 Symptomatic mild/moderate 5-asa non responder Variable See p6 See p6 p12 Symptomatic severe attack Surgery (1-p5-p39) p13 Symptomatic severe attack 5-asa non responder (1-p10-p39) p14 5-asa non responder Steroid non responder 0.31 E p15 5-asa non responder Steroid controlled remission (1-p14-p39)* p16 Steroid controlled remission Steroid controlled relapse 0.03 [E]* 0.01 in Western mainland Europe reflects Various azathioprine use instead of 5-asa drugs 51,52 p17 Steroid controlled remission Steroid controlled remission (1-p16-p39) 304 J. Buchanan et al.

5 p18 Steroid controlled relapse Steroid non responder 0.06 E p19 Steroid controlled relapse Steroid controlled remission (1-p18-p39) p20 Steroid non responder Steroid non responder in remission 0.90 Assumed to be equivalent to the probability of 50 moving into remission when being treated with IV steroids p21 Steroid non responder in remission Immunosuppressant non responder 0.01 Assumed to be equivalent to the probability of 51,52 relapse in CD patients treated with maintenance azathioprine p22 Steroid non responder in remission Steroid non responder in remission (1-p21-p39) p23 Steroid non responder Immunosuppressant non responder (1-p20-p39) p24 Immunosuppressant non responder Immunosuppressant non responder in remission p20 Assumed to be equivalent to the probability of moving into remission when being treated with IV steroids 50 p25 Immunosuppressant non responder in remission IV steroid non responder p21 Assumed to be equivalent to the probability of relapse in CD patients treated with maintenance azathioprine 51,52 Immunosuppressant non responder (1-p25-p39) in remission p26 Immunosuppressant non responder in remission p27 Immunosuppressant non responder IV steroid non responder (1-p24-p39) p28 IV steroid non responder IV steroid non responder in remission p29 IV steroid non responder in remission Surgery p21 Assumed to be equivalent to the probability of 51,52 relapse in CD patients treated with maintenance azathioprine p30 IV steroid non responder in remission IV steroid non responder in remission (1-p29-p39) p31 IV steroid non responder Surgery (1-p28-p39) p32 Surgery Postsurgical complication 0.02 E p33 Surgery Postsurgical remission (1-p32-p38) p34 Postsurgical remission Postsurgical complication p32 Assumed to be equivalent to suffering a E complication following surgery p35 Postsurgical complication Postsurgical remission (1-p39) p36 Postsurgical remission Postsurgical remission (1-p34-p39) p37 Various states Dead from disease 0 No change in mortality in UC patients. 25 Varied in sensitivity analysis p38 Various states Dead from surgery p39 All states Dead from other causes Variable Age specific 25 Parameter values that differ in a Western mainland Europe setting are starred. a Prob=Probability. b Unless otherwise noted, all parameter values apply in both a UK and Western mainland Europe setting. Parameter values that differ in a Western mainland Europe setting are starred. c All references that have informed parameter estimates are given. E=Values based on expert opinion estimates obtained through interviews with gastroenterologists. For parameters requiring more detailed calculation using multiple references, further details are provided within the Notes section. Care costs for inflammatory bowel disease patients 305

6 Table 2 Transition probabilities Crohn's disease. Prob a Transition Value b Notes Source c From To p Probability of presenting with mild symptoms E p Probability of presenting with moderate/ E severe symptoms p3 Symptomatic mild-pre-diagnosis Symptomatic mild 0.05 Proportion of mild starting population who E workup make transition each cycle, from 3 months onwards. Increments by 0.05 each cycle (estimated to take 3 24 cycles for all patients to move (average 13.5)) p4 Symptomatic mild-pre-diagnosis Symptomatic mild-pre-diagnosis (1-p3) Proportion of mild starting population who workup workup remain in this stage each cycle. p5 Symptomatic mild First-line drug non-responder mild Variable* Transition probabilities p5, p6 and p19 take Various varying values depending on the treatment options used: (i) 5-asa patients (1/3 of all symptomatic mild patients) spend only one cycle in this state. After one cycle, 17% move into long-term remission 54, 10% require surgery [E], and the remainder require more aggressive treatment [E]. 5-asa drugs are assumed not to be used in this state in Western mainland Europe, hence these probabilities do not apply in that location. (ii) The remaining patients (2/3 of all patients) are treated with prednisolone or budesonide (tapered treatment). After 3 cycles, 54% move into long-term remission 21,23,54 56, 15% require surgery [E], and the remainder require more aggressive treatment [E]. p6 Symptomatic mild Long-term remission Variable See p5 See p5 p7 Symptomatic moderate/severepre-diagnosis workup Symptomatic moderate/severe 0.25 Proportion of moderate/severe starting population who make transition each cycle, from 3 months onwards. Increments by 0.25 each cycle (estimated to take 3 6 cycles for all patients to move (average 4.5)) E Symptomatic moderate/severepre-diagnosis (1-p7) workup p8 Symptomatic moderate/severepre-diagnosis workup p9 Long-term remission Long-term remission (1-p31) p10 Symptomatic moderate/severe Long-term remission 0.10* 0.05 in Western mainland Europe, E reflecting more aggressive treatment. 306 J. Buchanan et al.

7 p11 Symptomatic moderate/severe Azathioprine controlled remission 0.60* 0.40 in Western mainland Europe, E reflecting more aggressive treatment. p12 Azathioprine controlled remission Azathioprine controlled relapse ,52 p13 First-line drug non-responder mild Long-term remission Variable Transition probabilities p13 and p14 Various take varying values depending on the treatment options used: (i) Patients treated with azathioprine, methotrexate or mercaptopurine (3/4 of all first-line drug non-responder (mild) patients) spend only one cycle in this state. After one cycle, 52% move into long-term remission 21,23,52,57 59, and the remainder require more aggressive treatment [E]. (ii) The remaining patients (1/4 of all patients) are treated with a higher steroid dose (tapered treatment). After 2 cycles, 31% require more aggressive therapy [E], and the remainder enter long-term remission [E]. p14 First-line drug non-responder mild Second-line drug non-responder mild Variable See p13 See p13 p15 Symptomatic moderate/severe Surgery 0.10* 0.30 in Western mainland Europe, E reflecting more aggressive treatment. p16 Symptomatic moderate/severe First-line drug non-responder (1-p10-p11-p15) moderate/severe p17 Azathioprine controlled remission Azathioprine controlled remission (1-p12) p18 Azathioprine controlled relapse Azathioprine controlled remission (1-p29-p31) p19 Symptomatic mild Surgery Variable See p5 See p5 p20 Second-line drug non-responder mild Long-term remission 0.58 Average based on three sources p21 Second-line drug non-responder mild Surgery (1-p20-p29-p31) p22 First-line drug non-responder moderate/severe Long-term remission Variable Transition probabilities p22 and p23 take varying values depending on the treatment options used: (i) Patients treated with azathioprine, methotrexate or mercaptopurine (3/5 of all first-line drug non-responder (moderate/severe) patients) spend onlyonecycleinthisstate:52% then enter long-term remission 52,57,58, and the remainder require more aggressive treatment [E]. (ii) The remaining patients (2/5 of all patients) are treated with infliximab or adalimumab (tapered treatment). After 2 cycles, 58% move into long-term remission ; the remainder require surgery [E]. Various (continued on next page) Care costs for inflammatory bowel disease patients 307

8 Table 2 (continued) Prob a Transition Value b Notes Source c From To p23 First-line drug non-responder Surgery Variable See p22 See p22 moderate/severe p24 Surgery Postsurgical complication 0.01 E p25 Surgery Postsurgical remission (1-p24-p30-p31) p26 Postsurgical remission Postsurgical complication p24 E p27 Postsurgical complication Postsurgical remission (1-p29-p31) p28 Postsurgical remission Postsurgical remission (1-p26-p31) p29 Various states Dead from disease 0 No change in mortality in CD patients. 25 Varied in sensitivity analysis p30 Various states Dead from surgery p31 All states Dead from other causes Variable Age specific 25 Parameter values that differ in a Western mainland Europe setting are starred. a Prob=probability. b Unless otherwise noted, all parameter values apply in both a UK and Western mainland Europe setting. Parameter values that differ in a Western mainland Europe setting are starred. c All references that have informed parameter estimates are given. E=Values based on expert opinion estimates obtained through interviews with gastroenterologists. For parameters requiring more detailed calculation using multiple references, further details are provided within the Notes section. 308 J. Buchanan et al.

9 Care costs for inflammatory bowel disease patients 309 Table 3 Unit costs. Parameter Cost ( ) Notes Source Drug treatments a Mesalazine month of treatment (4 g daily) month of maintenance treatment (2 g daily) 63 Prednisolone 5.90 Mild tapered dose average monthly cost 63, Moderate tapered dose average monthly cost 63,64 Budesonide Monthly cost for 3 month total course (9 mg daily for 63 8 weeks, 6 mg daily for 4 weeks) month of maintenance treatment (6 mg/daily) 63 Hydrocortisone day of treatment (300 mg/24 h) 63 Metronidazole days of treatment (800 mg daily) 63 Ciprofloxacin days of oral treatment (500 mg b.d.) days of IV treatment (500 mg b.d.) 63 Azathioprine month of maintenance or active treatment (2.5 mg/kg daily) 63 Mercaptopurine month of treatment (1 mg/kg daily) 63 Methotrexate month of treatment (15 mg sc injection, weekly) 63 Ciclosporin days of IV treatment (2 mg/kg daily) month of oral treatment (5 mg/kg daily) 63 Infliximab Cost of a single infusion (5 mg/kg) 63 Adalimumab First month of treatment (week mg; 63 week 2 80 mg; week 4 40 mg) Subsequent months of treatment (2 40 mg doses) 63 Surgeries, tests, and procedures b Complex surgical procedure Includes proctocolectomy with ileostomy, subtotal 65 proctocolectomy plus creation of pouch, ileal resection and strictureplasty, colectomy with ileo-colonic anastomosis Standard surgical procedure Includes closure of ileostomy, pouch removal, ileal resection, 65 strictureplasty, drainage of sepsis and insertion of setons Examination under anaesthesia Sigmoidoscopy Cost for day case procedure 65 Colonoscopy Cost for day case procedure 65 MRI scan Magnetic resonance imaging scan, one area, no contrast 65 Barium enema Contrast fluoroscopy procedures min 65 CT scan Computerised tomography scan, one area, no contrast 65 X-ray X-ray costs no longer unbundled in NHS Reference 67 Costs; cost inflated from 2006 to 2007 plain film, one area Stool culture 7.53 Microbiology/virology test 65 C-reactive protein test (CRP) Erythrocyte sedimentation rate test (ESR) Full blood count (FBC) Liver function test (LFT) 1.34 Biochemistry test 65 Staff costs (per hour) c General practitioner Surgeon Gastroenterologist Stoma & colorectal nurse specialist Agenda for change, Band 7 68 Miscellaneous Inpatient day in hospital Cost for single day in gastroenterology ward 65 Stoma related consumables Average cost for stoma supplies for one month 69 Stoma malfunction Average cost for treatment of a stoma malfunction (Various sources see Table A3 for further details and calculations) a Dosages based on an adult weighing 75 kg 70 informed by the British National Formulary 63 and expert opinion estimates obtained through interviews with gastroenterologists at UK hospitals. Patients assumed to complete a full treatment course. b Surgery, test and procedure costs informed by NHS Reference Costs 65 are based on NHS Trust data for patients aged over 18 years. c All staff costs inflated by 20% to account for overheads (Superannuation/National Insurance). A standard working week of 37.5 h, and a working year of 46 weeks were assumed.

10 310 J. Buchanan et al. Table 4 Stage costs. Stage Cost ( ) UK Europe a Ulcerative colitis b Symptomatic pre diagnosis workup Symptomatic mild/moderate Symptomatic mild/moderate Symptomatic mild/moderate asa controlled remission asa controlled relapse asa non-responder asa non-responder asa non-responder Steroid controlled remission Steroid controlled relapse Steroid controlled relapse Steroid controlled relapse Steroid non-responder Steroid non-responder in remission Immunosuppressant non-responder Immunosuppressant non-responder in remission IV steroid non-responder IV steroid non-responder in remission Symptomatic severe attack Tapered treatment after severe attack Tapered treatment after severe attack Surgery 11, , Postsurgical remission Postsurgical complication Dead from disease 0.00 Dead from other causes 0.00 Crohn's disease c Symptomatic mild-pre diagnosis 3.08 workup Symptomatic mild Symptomatic mild Symptomatic mild First-line drug non responder mild First-line drug non responder mild 2 Second-line drug non responder mild Second-line drug non responder mild 2 Long-term remission 1.70 Symptomatic moderate/ 6.15 severe-pre-diagnosis workup Symptomatic moderate/severe Symptomatic moderate/severe Symptomatic moderate/severe First-line drug non responder moderate/severe First-line drug non responder moderate/severe Table 4 (continued) Stage Cost ( ) UK Europe a Crohn's disease c Azathioprine controlled remission 3.40 Azathioprine controlled relapse Surgery Postsurgical remission Postsurgical complication Perianal disease d Dead from disease 0.00 Dead from other causes 0.00 a Where stage costs are not indicated for Europe, these are equivalent to UK costs. All European stage costs were based on UK sources, to ensure that cost differences between locations reflected differences in clinical management. b Detailed costs described in Table A3(i ii). c Detailed costs described in Table A4(i ii). d A proportion of CD patients (assumed to be 20%) suffer from perianal disease, regardless of disease severity This is not modelled as a separate health state: rather the costs associated with perianal disease are incorporated into the model at each stage as being accrued by 20% of the patients in the active disease states of the CD model. (alongside immunomodulators such as methotrexate, azathioprine and 6-mercaptopurine) as treatment options in CD patients presenting with moderate to severe disease with no response to steroid treatment. In both settings, data obtained via interview were supplemented by data from published sources. Economic decision models simulating the natural disease history of IBD were constructed based on these clinical pathways. Specifically, Markov models were developed for UC and CD (Figs. 1 and 2). Markov modelling is commonly used to study diseases with health states that recur over time. Therefore, it is particularly well suited to the study of chronic diseases such as IBD. 19 Health states within the two models were defined by both disease activity and the type of medical or surgical therapy required, an approach taken in previous studies. 20 The same model structure was used in both settings, incorporating variations to reflect differences in clinical management. All model states are described in Table A1. Starting and transition probabilities are described in Tables 1 and 2, and differences in these probabilities in the UK and Western mainland Europe are highlighted. Probability estimates from published sources such as international guidelines were used where possible In a number of cases, no published data sources were identified, so model parameters were based on the clinical pathways and informed by the gastroenterologist interviews. A cycle length of one month was assumed, reflecting the likely timeframe during which potential changes in patient clinical status may occur. Transition probabilities which were reported in studies for periods other than one month were converted into monthly probabilities as required. Tunnel states (that patients progress through in a fixed sequence over a specified time period) were used to represent health states in which patients received a tapered steroid dose. Microsoft Excel 2007 was used to construct the models and conduct all analyses. In both models patients were grouped according to disease severity. The length of time taken to move from a symptomatic state to treatment initiation was variable (except for UC patients with severe symptoms). However, it was assumed that all patients ultimately received a definitive, correct, diagnosis.

11 Care costs for inflammatory bowel disease patients 311 Given mixed evidence in the literature 24, no excess mortality (exceeding that observed in the general population) from UC or CD was assumed (varied in a sensitivity analysis). Age-specific all cause mortality rates were informed by data from UK life tables. 25 Transition to death was possible in all health states, but is not indicated in Figs. 1 and 2 for brevity. No gender differences in disease progression were assumed: previous studies have reported that gender is not a significant predictor of total treatment costs Cost data The costing perspective was the UK NHS. Costs were extracted from UK sources and are reported in pounds sterling in 2008 prices (Table 3). Cost data for the European models were taken from the same UK sources. Applying UK costs throughout ensured that cost differences between locations reflected differences in clinical management, facilitating comparisons: using costs specific to each location may obscure these differences. Table A2 reports additional details concerning the cost of treating stoma malfunctions. Table 4 reports the costs for each model stage. Where differences in clinical management were noted by location, stage costs differ (see Table A3(i ii) and Table A4(i ii) for further details). All costs were discounted at a rate of 3.5%. 26,27 Results are presented in both UK pounds sterling and Euros, converted by applying an exchange rate of 1: 1.26, based on average rates in Analytical methods Hypothetical cohorts of 10,000 patients presenting with symptoms of varying severity were assigned to several initial health states in each model, and their transitions between health states over time were modelled. Analyses were undertaken in UK and Western mainland Europe settings. A ten year time horizon was used, as the clinical course of disease remains settled for most patients after this time. 29 The baseline starting age for all patients in each cohort was 25, reflecting the average age at diagnosis. 1,21 The number of patients in each health state at the end of every cycle was observed and the costs accrued over time were calculated. The total cost of caring for the whole cohort and the cost of care per patient year were calculated. The number of patients in each health state after ten years was observed, along with the total number of months spent in each health state. The estimates of cost per patient year for 25 year old patients were used to calculate the annual cost of caring for all UC and CD patients in the UK and in Western mainland Europe. Prevalence estimates were taken from the literature 2 (higher estimates have been reported; 21,30 our results therefore reflect a conservative approach). Mid-point values were used in the cost calculations; high and low estimates of the total annual cost of care were based on the high and low prevalence estimates. The UK and Western mainland Europe populations in 2008 were estimated to be 61.4 million people 31 and million people, 32 respectively Sensitivity analysis Key model parameters were varied in a one-way sensitivity analysis. Parameter ranges reflected evidence from both the literature and expert opinion, elicited through interviews with gastroenterologists. Parameters varied included the cost of treatment, procedures and staff time, the proportion of patients with severe forms of UC and CD, time to diagnosis, length of inpatient stay and several transition probabilities. For full details of the parameter variations, see Table A5. Probabilistic sensitivity analysis (PSA) was used to further consider the impact of parameter uncertainty within the model. PSA allows the uncertainty across all model parameters (costs, probabilities etc.) to be assessed simultaneously by assigning probability distributions to parameters based on measures of variance reported in the original studies. 33 Where parameters were informed by expert opinion, or no information on variance was available, a coefficient of variation equal to 0.25 was assumed to reflect this uncertainty. 34 Values from these distributions were randomly selected for each parameter and the model was run 1000 times for different parameter combinations, allowing confidence intervals to be calculated for annual treatment costs. 3. Results The average cost of treatment per patient year for a cohort of 25 year old UC patients was 762/ 959 (95% CI: / ) in the UK and 796/ 1002 (95% CI: / ) in Europe (Table 5). The average cost per patient year for a cohort of 25 year old CD patients was 631/ 795 (95% CI: / ) in the UK and 838/ 1055 (95% CI: / ) in Europe. The costs of UC treatment were similar in both locations, while the average cost per patient year for CD patients was 207/ 261 higher in Europe. The relative costs of UC and CD also differed within locations. In the UK, the average annual cost of treatment was generally higher for UC patients as compared to CD patients. In Europe this pattern of results was reversed. Most patients accrued the bulk of their treatment costs in active disease states in year one. However, a number of patients were still being treated for active disease, and therefore incurring costs, at ten years. This result was repeated in other age groups and settings for both UC and CD. The distribution of costs across the 10 year time horizon was similar for both UC cohorts (Table 6), with costs much higher in the first year of treatment. For CD patients, costs in year 1 were 1645/ 2070 higher in Europe compared to the UK. This reflects the earlier use of more aggressive (often more expensive) therapies and procedures in Europe (e.g. bowel resections). Again, costs in the first year of treatment were higher than costs in years 2 9 in both locations. Table 5 Results, by disease and location. Disease Setting Average cost per patient year (95% CI) Cohort health states at 10 years Months spent by entire cohort in health states at 10 years Remission Active disease Dead Remission Active disease Dead UC UK 762 ( ) 959 ( ) ,068, , Europe 796 ( ) 1002 ( ) ,088, , CD UK 631 ( ) 795 ( ) ,097,140 99, Europe 838 ( ) 1055 ( ) ,093, ,

12 312 J. Buchanan et al. Table 6 Distribution of average costs by year for cohorts of 25 year old patients. Year Average cost per patient UC CD UK Europe UK Europe The total number of people with UC and CD in the UK in 2008 was estimated to be 122,823 (range: 98, ,388) and 59,876 (33,776 85,976) respectively (Table 7). The total number of people with UC and CD in Western mainland Europe in 2008 was estimated to be 666,383 (range: 533, ,659) and 324,861 (183, ,468) respectively. Around 1/336 individuals were estimated to have either UC or CD across both locations. The total annual cost of caring for these patients in the UK was estimated to be 131 million/ 165 million ( million/ million): 94 million/ 118 million ( million/ million) for UC patients and 38/ 48 million ( million/ million) for CD patients. The total annual cost of caring for these patients in Western mainland Europe was estimated to be 803 million/ 1010 million ( million/ million): 531 million/ 668 million ( million/ million) for UC patients and 272/ 343 million ( million/ million) for CD patients Sensitivity analysis Sensitivity analysis results are reported in Table A6 for the UK. Several parameters impact on model results. For CD patients, doubling the cost of a hospital inpatient day increases the average cost per patient year by 178 (from 631 to 809). In several cases (the cost of surgery, or infliximab treatment), parameter changes have a greater effect on the cost of treating CD rather than UC patients. A possible explanation is that a greater proportion of these costs occur earlier in the disease pathway for CD patients, diminishing the impact of discounting. The exception to this pattern is mesalazine treatment for UC patients, doubling this cost increases the average cost per patient year by 278 (from 762 to 1040). Variations in the proportion of patients who present initially with a severe attack are also important for UC (the base case probability of severe attack is lower than that for CD). For UC patients, an increase of 15% in the likelihood of presentation with a severe attack increases the average cost per patient year by 181 to 943. In addition, the proportion of CD patients with moderate to severe disease who achieve azathioprine controlled remission is an important cost driver. Finally, it is possible that older patients may develop more serious infections due to either the disease or the treatments than younger patients, leading to an increase in the length of inpatient stay. We considered the impact of doubling the length of inpatient stay and found that this increased the average cost per patient year by 105 to 867 for UC patients, and by 178 to 809 for CD patients. Table 7 Annual cost of caring for UC and CD patients in the UK and Western mainland Europe. Total cost of caring for all patients for 1 year (Range) Number of people with disease (Range) Average annual cost of care per person UK Western Mainland Europe Prevalence per 100,000 population (range) 1,2 UK Western Mainland UK Western Europe a Mainland Europe million million million million ( ) ( ) ( ) ( ) 666,383 (533, ,659) UC 200 ( ) 122,823 (98, ,388) ( ) 47.6 ( ) ( ) ( ) 324,861 (183, ,468) CD 98 (55 140) 59,876 (33,776 85,976) ( ) ( ) ( ) ( ) 991,244 (716,361 1,266,127) Combined 298 ( ) 182,700 (132, ,364) Comprising the EU15 countries, except for the UK: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain and Sweden. a

13 Care costs for inflammatory bowel disease patients Discussion IBD is a widespread chronic condition costing between 631 and 762 ( ) per patient for treatment annually in the UK. An estimated 182,700 people in the UK have a form of IBD, and the total annual cost to the UK NHS of caring for these patients is estimated to be 131 million ( 165 million). Many patients accrue a large proportion of their total treatment costs at the start of their disease course. However IBD is a chronic condition in which neither medical treatment nor surgery are curative, and a significant proportion of patients are still undergoing treatment (and incurring costs) up to ten years after diagnosis. The cost of caring for IBD patients in Western mainland Europe was also estimated. The annual cost of treating CD patients was higher in Western mainland Europe, while the treatment costs of UC patients weresimilar inbothlocations. Inmost model states, monthly treatment costs in Western mainland Europe were equal to or greater than those estimated for the UK. UK costs were used in both locations, therefore cost variations between locations reflect differences in clinical management. The higher cost of treating European CD patients primarily reflects variations in the drugs given as initial therapy, and more frequent and earlier use of surgical procedures. Since these differences in clinical management did not significantly affect the number of months spent by both cohorts in remission, the combined effect was a higher annual treatment cost for European CD patients. Conversely, the annual treatment costs for UC were similar in both locations. Again, differences in clinical management resulted in higher stage costs in Europe. However, this effect was combined with an increase in the number of months spent in (less costly) remission states, with the two effects acting on the overall treatment cost in opposite directions. The spread of treatment costs over the time horizon also differed between locations. In Europe, a greater proportion of total CD costs occurred in year one (compared to UC). In the UK, where more aggressive treatments are only used if less aggressive treatments have not alleviated a patient's symptoms, treatment costs were spread more evenly over the timeframe, allowing discounting to have a greater impact on the results. UC treatment costs were also higher in years 2 10 in the UK compared to Europe. Several other studies have estimated the total cost of care for IBD patients (Table A8 summarises a selection of these). Many of these estimates are higher than those reported in our study, although differences in study design and setting limit comparability. The only study reporting results of a similar magnitude also took a modelling approach, reporting a median cost. 35 Modelling studies by nature focus on average cohort costs rather than explicitly accounting for patients with very high treatment costs. Hence cost estimates based on modelling studies are commonly lower than those reported by single centre surveys or studies based on insurance claims databases. 14 Furthermore, median costs may more accurately reflect the true cost of treating an average IBD patient as mean healthcare costs can be affected by skewed cost distributions. 5 Our study therefore generates cost estimates which are particularly useful for planning IBD clinical management. A number of our findings replicate those from previous reports. Surgery and hospitalisations are commonly important cost drivers 12,15,17,36 although rates of hospitalisation differ widely between settings. 17 Treatment costs for CD are also often higher than those for UC (Table A8): studies suggest that CD patients are characterised by greater frequency of severe disease, requiring more frequent and intensive treatment. 15 This study only supports that conclusion in Western mainland Europe, suggesting that this pattern of results is sensitive to variations in clinical management by location. Our finding that mesalazine treatment is an important cost driver for UC is also supported by other studies. 17 Other studies also report that treatment costs are highest in the first year after diagnosis (to initially induce remission), decreasing sharply and then stabilising. 17,37 Hence studies measuring costs over long time horizons may produce lower estimates of average annual treatment costs than those covering shorter time periods. A comparison between our results and other studies supports this conclusion. Finally, previous studies suggested that treatment patterns will change in the future, with increased use of diagnostics and biological therapies, and a reduction in the frequency of surgery and hospitalisation. 17,38,39 The latter cost savings could exceed the additional treatment costs, reducing spending on IBD care overall. Some limited evidence exists to support the existence of this effect. 40 The cost estimates reported here may also reflect such changes in clinical practice. The costs of caring for IBD patients can be compared with those for patients with other chronic diseases. The annual cost of treating asthma patients was reported to be 754 million in the UK in Approximately 2.6 million people in the UK experience severe asthma symptoms, suggesting an annual cost of treatment of 290 per patient, lower than our estimate for IBD. 42 The annual cost of treating Type II diabetes mellitus was reported to be 2214 ( 1574) per UK patient in 2002, greater than our estimate for IBD. 43 Other studies compared the costs of caring for IBD and non-ibd patients, estimating that annual per capita healthcare costs were more than twice as high for IBD patients. 37 This study adds to the literature on the cost of IBD care by estimating current costs which are applicable across whole health care systems, rather than in specific study settings within healthcare systems. Key cost drivers such as expensive biological therapies are included and important patient subgroups, such as the most severely ill, are accounted for. A key strength of our study is that it illustrates the importance of differentiating between clinical management settings: the costs of care and the length of time spent in different disease states can vary considerably by location. However, these results do not necessarily reflect more or less effective clinical management strategies: they may instead reflect patient preferences for earlier aggressive treatment in different locations. Information on the value to patients of achieving these preferences is required prior to drawing conclusions about the relative effectiveness of different management strategies, as patient satisfaction may not necessarily be correlated with expenditure on treatment. 3,44 Furthermore, the estimates generated in this study for both settings are based on UK costs. Applying costs from Western mainland Europe instead may lead to a different pattern of results. Our analytical framework allows healthcare providers to predict the resources required and costs incurred from treating current IBD patients. It can also be adapted to consider the impact on IBD care costs of introducing new interventions. This is particularly useful if such interventions are expensive and other forms of assessment are impractical (e.g. randomised controlled trials). Several novel genomic technologies with diagnostic and prognostic applications in IBD (e.g. microarrays) meet both of these conditions The use of these technologies in clinical practice has been shown to be costeffective in other disease areas, 45 but given that these tools could impact on all aspects of IBD clinical management, their positioning within clinical pathways must be carefully planned. In terms of study weaknesses, our results are based on an analysis of hypothetical patient cohorts. However, it could be important to focus on particular patient subgroups. For example, the most costly patients in a particular cohort can account for a significant proportion of total costs. 12,15,20 Cohort composition may differ by location: larger tertiary hospitals may treat more severely ill patients. 12 Furthermore, this study adopts a national healthcare provider perspective when calculating the cost of IBD care. However costs accruing to patients or society (e.g. lost work productivity) are also important, and can in some cases surpass treatment costs. 20 Modelling studies are by design simplifications of clinical practice. This study was restricted to a ten year time horizon as many IBD treatment costs are incurred at the beginning of the disease pathway. However, patients continue to incur costs beyond ten years. It is possible that the type and magnitude of these additional costs differ from those captured by this study. Many CD patients present with an inflammatory phenotype, which can develop into a stenosing or penetrating phenotype over time,

14 314 J. Buchanan et al. impacting on treatment costs. Some subgroups of IBD patients may also be at an increased risk of cancer (although this increased risk may only apply after ten years of disease 46 ). Finally, the costs incurred as a result of azathioprine-related complications (e.g. leukopenia) are not considered. This paper describes a comprehensive analytical framework which can be used to precisely identify where IBD treatment costs are incurred and to quantify the magnitude of these costs. This framework also illustrates the importance of differentiating between alternative clinical management scenarios. With the emergence of expensive but effective targeted biological therapies and novel diagnostic tools, this framework will provide a vital tool for healthcare providers to consider the potential economic impact of new IBD-related interventions, and to appropriately allocate limited healthcare resources. Our own future work will focus on applying the framework to such novel interventions. Conference presentations Preliminary results from this study were presented as posters at two conferences: British Society of Gastroenterology Annual Meeting, Liverpool, UK, March 2010 (abstract published in Gastroenterology 2010;138:5, S-317). Digestive Disease Week, New Orleans, USA, May 2010 (abstract published in Gut 2010; 59:A60). Awarded Poster of Distinction. Acknowledgements The authors acknowledge funding from the European Commission's Sixth Framework Programme and the Oxford Biomedical Research Centre. The study funders had no involvement in the study design, the collection, analysis and interpretation of data, the writing of the manuscript, or the decision to submit the manuscript for publication. We thank the IBD Consortium members for assistance. We also thank Dr John Ramage, Professor John Williams, Robert Peacock, Alastair Windsor, Dr Stirling Pugh and Jill Cockrill for helpful discussions concerning NHS Reference Costs. Any views expressed in this paper are solely those of the authors. Statement of authorship JB designed and carried out the economic analyses, interpreted the results and drafted the manuscript. SW advised on the study design and analysis and helped to draft the manuscript. TA and AP contributed to the collection of data on IBD clinical management in the UK and helped to draft the manuscript. SV contributed to the collection of data on IBD clinical management in Western mainland Europe and helped to draft the manuscript. MS led the funding application, managed the project, contributed to the collection of data on IBD clinical management in Western mainland Europe and helped to draft the manuscript. JT managed the project and helped to draft the manuscript. DJ contributed to the collection of data on IBD clinical management in the UK, advised on the study design and analysis, managed the project, and helped to draft the manuscript. Thus, all authors made substantial contributions to all of the following: (1) the conception and design of the study, acquisition of data or data analysis, and interpretation of data, (2) drafting the article or revising it critically for important intellectual content, and (3) read and approved the final manuscript to be submitted. None of the authors have any competing interests to declare. Appendix A. Supplementary data Supplementary data to this article can be found online at doi: /j.crohns References 1. Loftus EV. Clinical epidemiology of inflammatory bowel disease: incidence, prevalence, and environmental influences. Gastroenterology 2004;126(6): Williams JG, Roberts SE, Ali MF, Cheung WY, Cohen DR, Demery G, Edwards A, Greer M, Hellier MD, Hutchings HA, Ip B, Longo MF, Russell IT, Snooks HA, Williams JC. Gastroenterology services in the UK. The burden of disease, and the organisation and delivery of services for gastrointestinal and liver disorders: a review of the evidence. Gut February 1, 2007;56: Feagan BG. Review article: economic issues in Crohn's disease assessing the effects of new treatments on health-related quality of life. Aliment Pharmacol Ther 1999;13(s4): Kappelman MD, Rifas-Shiman SL, Porter CQ, Ollendorf DA, Sandler RS, Galanko JA, Finkelstein JA. Direct health care costs of Crohn's disease and ulcerative colitis in US children and adults. Gastroenterology 2008 Dec;135(6): Odes S. How expensive is inflammatory bowel disease? A critical analysis. World J Gastroenterol 2008;14(43): Arseneau KO, Cohn SM, Cominelli F, Connors Jr. Cost-utility of initial medical management for Crohn's disease perianal fistulae. Gastroenterology 2001;120(7): Bodger K. Economic implications of biological therapies for Crohn's disease: review of infliximab. Pharmacoeconomics 2005;23(9): Cohen RD. Effects of fistula on healthcare costs and utilization for patients with Crohn's disease treated in a managed care environment. Inflamm Bowel Dis 2008;14(12): Dryden Jr GW. Overview of biologic therapy for Crohn's disease. Expert Opin Biol Ther 2009 Aug;9(8): Reenaers C, Louis E, Belaiche J. Review: current directions of biologic therapies in inflammatory bowel disease. Ther Adv Gastroenterol March 1, 2010;3: Sandborn WJ, Feagan BG, Stoinov S, Honiball PJ, Rutgeerts P, Mason D, Bloomfield R, Schreiber S. Certolizumab pegol for the treatment of Crohn's disease. N Engl J Med 2007 Jul 19;357(3): Bassi A, Dodd S, Williamson P, Bodger K. Cost of illness of inflammatory bowel disease in the UK: a single centre retrospective study. Gut 2004;53(10): Bodger K. Cost of illness of Crohn's disease. Pharmacoeconomics 2002;20: Cohen RD, Yu AP, Wu EQ, Xie J, Mulani PM, Chao J. Systematic review: the costs of ulcerative colitis in Western countries. Aliment Pharmacol Ther 2010;31(7): Hay J, Hay A. Inflammatory bowel disease: costs-of-illness. J Clin Gastroenterol 1992;14(4): Feagan BG, Vreeland MG, Larson LR, Bala MV. Annual cost of care for Crohn's disease: a payor perspective. Am J Gastroenterol Aug 2000;95(8): Odes S, Vardi H, Friger M, Wolters F, Russel MG, Riis L, Munkholm P, Politi P, Tsianos E, Clofent J. Cost analysis and cost determinants in a European inflammatory bowel disease inception cohort with 10 years of follow-up evaluation. Gastroenterology 2006;131(3): D'Haens G, Baert F, van Assche G, Caenepeel P, Vergauwe P, Tuynman H, De Vos M, van Deventer S, Stitt L, Donner A, Vermeire S, Van De Mierop FJ, Coche J-CR, van der Woude J, Ochsenkühn T, van Bodegraven AA, Van Hootegem PP, Lambrecht GL, Mana F, Rutgeerts P, Feagan BG, Hommes D. Early combined immunosuppression or conventional management in patients with newly diagnosed Crohn's disease: an open randomised trial. Lancet 2008;371(9613): Briggs A, Sculpher M. An introduction to Markov modelling for economic evaluation. Pharmacoeconomics 1998 Apr;13(4):