Appendix 1: Search strategy

Table 6: MEDLINE search (24-08-2012)

Limit 1 / Limit 2 / Limit 3 / Results
1# / (rheumatoid OR reumatoid OR rheumatic OR reumatic OR rheumat* OR reumat*) AND (arthrit* OR artrit* OR disease OR diseases OR condition* or nodule*) / English / ≥ 1 jan 2002 / 52881
2# / (cost OR costs OR costing OR economic* OR pharmacoeconomic* OR (cost AND (analys* OR benefit OR effectiveness OR utility))) AND (simulation OR model OR models OR modeling OR modelling OR (decision AND (analys* OR analytic))) / English / ≥ 1 jan 2002 / 60144
3# / (DMARD* OR antirheumatic* OR antireumatic* OR biologic* OR TNF-α OR tnf-alfa OR TNF-α OR TNF OR anti-tnf OR necrosis factor* OR *alpha OR *alfa OR gold OR auranofinOR methotrexate OR MTX OR cyclosporin OR ciclosporin OR *penicillamine OR leflunomide OR azathioprine OR sulfasalazine OR SSZ OR *chloroquine OR minocycline OR etanercept OR infliximab OR adalimumab OR golimumab OR certolizumab OR anakinra OR tocilizumab OR abatacept OR rituximab OR tofacitinib) / English / ≥ 1 jan 2002 / 1079797
4# / 1# AND 2# and 3# / English / ≥ 1 jan 2002 / Humans / 184
5# / "Arthritis, Juvenile Rheumatoid"[Mesh Major Topic] / English / ≥ 1 jan 2002 / Humans / 1851
6# / 4# NOT 5# / English / ≥ 1 jan 2002 / Humans / 183

Table 7: Embase search (24-08-2012)

Limit 1 / Limit 2 / Limit 3 / Results
1# / rheumatoid arthritis'/exp/mj AND [humans]/lim AND [english]/lim AND [1-1-2002]/sd NOT [31-8-2012]/sd AND [2002-2013]/py / English / ≥ 1 jan 2002 / Humans / 24305
2# / tumor necrosis factor alpha antibody'/exp OR'tumor necrosis factor alpha inhibitor'/exp ORdmard*ORantirheumatic*ORantireumatic*ORbiologic*OR'gold'/exp OR 'auranofin'/exp OR'methotrexate'/exp OR'mtx'/exp OR'cyclosporin'/exp OR'ciclosporin'/exp OR'penicillamine'/exp OR'leflunomide'/exp OR'azathioprine'/exp OR'sulfasalazine'/exp ORsszOR'chloroquine'/exp OR'minocycline'/exp OR'etanercept'/exp OR'infliximab'/exp OR'adalimumab'/exp OR'golimumab'/exp OR'certolizumab pegol'/exp OR'anakinra'/exp OR'tocilizumab'/exp OR'abatacept'/exp OR'rituximab'/exp OR'tofacitinib'/exp AND [humans]/lim AND [english]/lim AND [1-1-2002]/sd NOT [31-8-2012]/sd AND [2002-2013]/py / English / ≥ 1 jan 2002 / Humans / 4104832
3# / #1 AND #2 / English / ≥ 1 jan 2002 / Humans / 22472
4# / cost*OReconomic*ORpharmacoeconomic*OR'pharmacoeconomics'/exp OR'cost benefit analysis'/exp OR'cost effectiveness analysis'/exp OR'cost utility analysis'/exp AND [humans]/lim AND [english]/lim AND [1-1-2002]/sd NOT [31-8-2012]/sd AND [2002-2013]/py / English / ≥ 1 jan 2002 / Humans / 465002
5# / 'simulation'/exp OR'model'/exp ORmodelsOR'modeling'/exp OR'modelling'/exp OR (decisionAND (analys*ORanalytic)) AND [humans]/lim AND [english]/lim AND [1-1-2002]/sd NOT [31-8-2012]/sd AND [2002-2013]/py / English / ≥ 1 jan 2002 / Humans / 461653
6# / #4 AND #5 / English / ≥ 1 jan 2002 / Humans / 58050
7# / #3 AND #6 / 387
8# / 'nonhuman'/exp OR'juvenile rheumatoid arthritis'/exp AND [humans]/lim AND [english]/lim AND [1-1-2002]/sd NOT [31-8-2012]/sd AND [2002-2013]/py / 502257
9# / #7 NOT #8 / 338

Table 8: NHS EED search(24-08-2012)

Limit 1 / Results
1# / Rheumatoid AND Arthritis / > 1 jan 2002 / 115

Table 9: Merging databases

Databases / Duplicates / Total
MEDLINE + Embase / 83 / 438
(MEDLINE Embase) + NHS EED / 55 / 498

Appendix 2: Transferability factors

Table 10: Explanation on the evaluation of Welte's transferability factors

Transferability factors
Perspective / The perspective of the analysis affects the choice which costs and effects to account for in the model. As costs and effects can be easily adjusted, this is regarded parameter uncertainty.
Discount rate / Discount rates are usually included as variables in the model. These are thus easily adjustable and therefore this is regarded parameter uncertainty.
Medical cost approach / These factors all influence the cost of the different technologies under evaluation. As these can be easily adjusted, this is regarded parameter uncertainty.
Productivity cost approach
Absolute and relative prices
Practice variation / Practice variation may affect the costs and effectiveness of a technology (e.g. if patients receive more or less instructions or guidance in complying withtheir treatment, this may affect the effectiveness and costs of care). If practice variationis only expected to affect the costs and/or effects, parameter uncertainty is presentand simple adjustments can be made. However, practice variation may also account for structural uncertainty in case practice patterns are somehow embedded in the model structure. For example, a higher frequency of follow-up visits could result in shorter treatment durations, as non-effectiveness would be noticed sooner. If treatment discontinuation is hard coded in the model, structural bias will be present. Also, if treatment sequences are hard coded in the model structure, practice variation in such sequences will result in structural bias.
Technology availability / This could result in parameter bias, for example higher distances to the hospital will result in more travelling costs for the patient. Technology availability could also induce structural bias, for example if a lack of diagnostic/monitoring tools results in different treatment patterns. In this regard, technology availability is closely related to practice variation.
Disease incidence/prevalence / Disease incidence/prevalence is important especially when looking at infectious diseases, as this may affect the spread of the disease. In such models, however, the incidence/prevalence should be adaptable (if not, this will induce structural uncertainty!). In models for screening programs, incidence/prevalence may affect the effectiveness of the treatment, and it may affect the unit cost of a treatment, as unit costs will generally be lower if the utilization frequency is higher. As costs and effectiveness are easily adaptable, this is regarded parameter uncertainty.
Case-mix / Differences in case-mix may affect the effectiveness of a treatment. For example, some drugs show higher response rates in patients of Asian or Afro-American origin, which can impact the cost-effectiveness of a treatment in a certain population. Using simple adaptation (i.e. adapting the effectiveness inputs), parameter bias due to case-mix can be limited. Case-mix may, however, induce structural uncertainty when specific patient characteristics affect treatment patterns or disease progression which are hard coded into the model. For example, the presence of certain (additional) risk factors may accelerate disease progression, when compared to a population that is not exposed to such risk factors.
Life expectancy / Life expectancy tables should be adjusted per country, which is usually possible through simple adaptation (if not, this will induce structural bias). This is especially important in case of chronic diseases, or prevention of fatal diseases. In case of fatal diseases such as cancer, life expectancy is inherent to the effectiveness inputs as survival is usually a primary outcome measure.
Health status preferences / These are usually included as variables in the model. These are thus easily adjustable and therefore this is regarded parameter uncertainty.
Acceptance, compliance and incentives to patients / These factors all influence the effects of the different technologies under evaluation. As these can be easily adjusted, this is regarded parameter uncertainty.
Productivity and absenteeism / This all influences the cost of the different technologies under evaluation. As this can be easily adjusted, this is regarded parameter uncertainty.
Disease spread / This is only applicable when looking at infectious diseases. In such cases, disease spread may cause parameter uncertainty and structural uncertainty if factors like incidence/prevalence, population density, population susceptibility (case-mix), and practice variations are different between countries. Some of these factors may be adjustable, but others may not. This can differ between models.

Page 1 of 5