Skip Navigation


Rheumatology Advance Access originally published online on June 11, 2007
Rheumatology 2007 46(8):1345-1354; doi:10.1093/rheumatology/kem115
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
46/8/1345    most recent
kem115v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (8)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Brennan, A.
Right arrow Articles by Symmons, D.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Brennan, A.
Right arrow Articles by Symmons, D.
Related Collections
Right arrow Health Economics
Right arrow Rheumatoid Arthritis
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© The Author 2007. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org

Modelling the cost effectiveness of TNF-{alpha} antagonists in the management of rheumatoid arthritis: results from the British Society for Rheumatology Biologics Registry

A. Brennan1,4, N. Bansback1,2,4, R. Nixon3,4, J. Madan1, M. Harrison4, K. Watson4 and D. Symmons4

1Health Economics and Decision Science, School of Health and Related Research (ScHARR), The University of Sheffield, UK, 2Centre for Health Evaluation and Outcome Sciences, St Paul's Hospital, Vancouver, BC, Canada, 3MRC Biostatistics Unit, Institute of Public Health, University Forvie Site, Robinson Way, Cambridge CB2 2SR and 4Arthritis Research Campaign Epidemiology Unit, The University of Manchester Medical School, Manchester M13 9PT, UK.

Correspondence to: Alan Brennan, Director of Health Economics and Decision Science, School of Health and Related Research (ScHARR), The University of Sheffield, UK. E-mail: a.brennan{at}sheffield.ac.uk


   Abstract
Top
 Abstract
Introduction
 Method
 Results
 Discussion
 Appendix
 Acknowledgements
 References
 
Objective. To evaluate the cost effectiveness of TNF-{alpha} antagonist therapies for rheumatoid arthritis (RA) in the United Kingdom using data from the British Society for Rheumatology Biologics Registry (BSRBR).

Methods. A simulation model is constructed to quantify the cost effectiveness of the TNF-{alpha} antagonist therapies (infliximab, etanercept and adalimumab) as a group versus traditional disease-modifying anti-rheumatic drugs, with a time horizon over the full patient lifetime. Participants are UK NHS patients in the BSRBR with RA who have failed at least two traditional disease-modifying anti-rheumatic drugs. The BSRBR aims to recruit all RA patients starting on a TNF-{alpha} antagonist agent and follows them 6 monthly via consultant and patient administered questionnaires. Data collected include disease activity scores (DAS28), the Health Assessment Questionnaire and the SF-36. Costs include drug, monitoring and hospitalisations. Benefits are measured in disability and quality of life improvements. The main outcome measure is the incremental cost per quality adjusted life-year gained (discounted).

Results. The basecase cost per quality adjusted life-year gained by using TNF-{alpha} antagonist therapies is estimated at £23 882, with probabilistic uncertainty analysis suggesting that the probability that treatments are below £30 000 per QALY is around 84%. The results are most sensitive to assumptions concerning long-term disability progression, discount rates and the validity or otherwise of SF6D derived utility measures. Subgroup analysis, monotherapy versus combination with methotrexate, and a limited analysis of sequential therapy with two TNF-{alpha} antagonist agents, suggest cost-effectiveness ratios around £20 000 to £30 000.

Conclusions. The BSRBR data provide valuable evidence for estimating cost-effectiveness. The analysis concludes that current policies and practice for the use of TNF-{alpha} antagonist therapies, after RA patients have failed at least two traditional disease-modifying anti-rheumatic drugs, appear cost-effective in the context of the NICE re-appraisal of 2006 for England and Wales, thus supporting their decision to continue their reimbursement. Decision-makers worldwide might adapt this analysis because differential costs, discount rates and other factors could affect results. There remains uncertainty, particularly on long-term disease progression. Further data collection using the BSRBR is recommended, together with a revision to this analysis when data become available.

KEY WORDS: Anti-TNF, cost, cost-effectiveness, rheumatoid arthritis


   Introduction
Top
 Abstract
 Introduction
Method
 Results
 Discussion
 Appendix
 Acknowledgements
 References
 
Rheumatoid arthritis (RA) is a chronic, progressive, inflammatory disease that affects ~0.8% of the adult population [1]. RA affects patients’ physical functioning, their psychological and social health and is associated with premature mortality [2–5]. The efficacy of tumour necrosis factor alpha (TNF-{alpha}) antagonists vs placebo regimens has been demonstrated in numerous clinical trials [6–16]. Costing between £8000 and £15 000 per year, a US study recently showed that the introduction of TNF-{alpha} antagonists has produced a 3-fold increase in annual direct costs per patient [17].

Policy makers, who wish to maximize their population of interest's health benefits within finite health care resources, are beginning to use cost-effectiveness analysis (CEA) to compare the efficiency of TNF-{alpha} antagonists with other technologies both within rheumatology and across other diseases. Existing CEAs of TNF-{alpha} antagonists for the UK [18–22], and elsewhere [18,23–28] give estimates that span the range from cost-effective to not cost-effective. Most recently, Welsing et al. [29] modelled the cost-effectiveness of several strategies including leflunomide and etanercept suggesting relatively high costs per quality adjusted life year (QALY). This led Maetzel [30] to question whether such analyses are out of touch with clinical reality. Wolfe et al. [31] also question the validity of these studies, because estimates of benefit are derived directly from ‘unrepresentative’ randomized controlled trials (RCTs), rather than real clinical practice. Only one study has directly used registry data, and this does not have a control arm to measure true effectiveness [19]. As yet no evaluation can be considered to be definitive [32,33].

In 2001, the policy makers for formulary decisions in the UK, the National Institute for Health and Clinical Excellence (NICE), recommended TNF-{alpha} antagonists for the treatment of patients with active RA unresponsive to conventional disease-modifying anti-rheumatic drugs (DMARDs) [34]. NICE mandated that all patients with RA exposed to TNF-{alpha} antagonists be followed-up to assess the long-term safety and effectiveness of these drugs. The British Society for Rheumatology Biologics Registry (BSRBR) was established in October 2001 for this purpose [35]. In 2004, NICE indicated that revised guidance on the use of TNF-{alpha} antagonists was to be developed.

Our study performs an independent analysis using a decision analytic model populated by BSRBR data to evaluate the cost-utility of TNF-{alpha} antagonists (infliximab, etanercept and adalimumab) as a class compared with conventional DMARD therapy (e.g. hydroxychloroquine, methotrexate, intramuscular gold, sulphasalazine and leflunomide). The two main research questions are: (i) can the current pattern of TNF-{alpha} antagonist use be considered cost-effective when compared with conventional DMARDs? (ii) If existing guidance from NICE and the BSR were strictly followed, would TNF-{alpha} antagonist therapy use be considered cost-effective? Evidence on subsidiary questions concerning cost-effectiveness in subgroups and for sequential anti-TNF-{alpha} therapy is also examined. This analysis provides the first evidence on cost-effectiveness from the BSRBR and, alongside the detailed project report [36], is intended as a source of evidence to be utilized in the UK and across the world.


   Method
Top
 Abstract
 Introduction
 Method
Results
 Discussion
 Appendix
 Acknowledgements
 References
 
Overview
The computer simulation model is used to synthesize evidence from the BSRBR with other available sources from systematic reviews [20]. Even with a multitude of data, mathematical modelling is required to synthesize the data in order to be able to address specific questions [37]. The model describes clinical pathways for RA including initial and long-term response to a sequence of therapies, probabilities of withdrawal and the long-term disability and quality of life consequences. Benefits are measured using health utilities so that the cost effectiveness can be compared with a range of conditions and interventions [38]. QALYs are compared with drug, monitoring and hospitalization costs to give incremental cost-effectiveness ratios from the UK NHS perspective using a cost year of 2004. The model is structured as an individual sampling model [39]. Uncertainty is examined using probabilistic sensitivity analysis (PSA) [40]. Discount rates are 6% for costs, 1.5% for QALYS as per the NICE protocol [41] and varied to 3.5% for both in sensitivity analysis.

BSRBR
The BSRBR has up to 3 yrs follow-up on nearly 8000 RA patients with active disease treated with TNF-{alpha} antagonists and nearly 900 RA patients with active disease treated with conventional DMARDs. The BSRBR collects 6-monthly data on efficacy, safety, quality of life and resource use via the patient and the consultant rheumatologist [35]. In addition, all patients are flagged for mortality with the Office for National Statistics. The Disease Activity Score (DAS28) [42] and Health Assessment Questionnaire Disability Index (HAQ) [43] are the main effectiveness measures alongside the SF-36 [44] generic quality of life instrument. Treatment response was assessed using the EULAR response criteria [45], which are based on the absolute and relative values of the DAS28 scores. There are three categories of EULAR response (none, moderate and good). We analysed data from October 2001 to September 2004. The patients in the TNF-{alpha} antagonist group had a total of 8284 patient years’ follow-up (2971 etanercept, 4474 infliximab, 839 adalimumab). They were younger than those on conventional DMARDs (mean age 55 yrs vs 60 yrs) but had more severe disease at baseline (HAQ = 2.1 vs 1.6). Table 1 summarizes the baseline characteristics of BSRBR patients.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Clinical and demographic characteristics of the 7953 patientsa

 
Clinical pathways
The simulation model tracks a large number of individual patients’ event histories over 6-monthly intervals from the decision to prescribe a TNF-{alpha} antagonist to patients who have failed at least two conventional DMARDs until the end of their lifetime. The model runs the same patient through two arms i.e. TNF-{alpha} antagonist therapy vs ongoing conventional DMARDs. Figure 1 illustrates the five key events modelled:
  1. Initial response to treatment (in terms of EULAR response) and switching of therapy
  2. Impact of initial response on short-term (0–6 months) health utility
  3. Length of longer-term treatment if therapy is continued
  4. Impact of longer-term treatment on longer-term health utility.
  5. Utility worsening when treatment is withdrawn


Figure 1
View larger version (22K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		FIG. 1. Clinical pathways in the Sheffield BSRBR RA model

 
When a patient reaches the time to withdraw, because of either an adverse event or lack of response, the model moves on to a second treatment in the sequence, then a third and so on. The probabilities of moderate or good EULAR response, magnitude of improvement and time to withdrawal are all re-adjusted to the individual patient's updated characteristics. Rather than specifying particular conventional DMARDs at different positions in the sequence, we use a generalized DMARD in each position based on a weighted average of BSRBR patients’ DMARD use (see Table 1 for weightings). After the sixth treatment in each arm, we assume patients will no longer respond but will still receive some maintenance therapy on conventional DMARDs.

Quality of life measures
We analyse quality of life in two separate ways. The BSRBR collects 6-monthly SF-36 data which can be translated to a societal health utility through the SF-6D algorithm [46]. However, previous studies have found the SF-6D is unable to distinguish between states of severe health in general [47]. The EQ-5D questionnaire is an alternative instrument for which health utilities can be obtained [48], and has been proven to better distinguish between states of severe health specifically in RA [49]. Since the EQ5D is not collected directly by the BSRBR, a validated mapping which imputes the EQ5D from all 42 components of the HAQ disability questionnaire data is used [50]. We have used both the SF-6D and the imputed EQ-5D methods to derive health utility estimates and subsequent QALYs for BSRBR patients.

Statistical analyses
The probability of each event is quantified through separate statistical models now described in further detail. All models include the following covariates: treatment (TNF-{alpha} antagonist or not), EULAR response, HAQ score at baseline, utility at baseline, age, sex, disease duration, number of previous DMARDs and concomitant DMARD use. The probability of events is adjusted according to the characteristics of sampled patients using these covariates, in order to arrive at an unbiased estimate of mean cost and effect. The degree of multi-co-linearity between covariates is therefore not of concern, although caution is advised in the interpretation of some individual coefficients [51,52]. However, since the purpose of the inclusion of these parameters is to adjust for the characteristics of sampled patients, no bias will be introduced to overall cost-effectiveness estimates. In addition, as the modelling approach examines uncertainty in all parameters, including costs and effects, formal statistical significance is not of inherent interest.

Initial response to treatment (in terms of EULAR response) and switching of therapy
The probability of EULAR response (non/moderate/good) is examined in a proportional odds logit model using BSRBR data (Appendix part 1). The results (Table A1) showed a higher likelihood of successful response for patients who are younger, male, have had fewer previous DMARDs, have higher baseline health utility and those who are given a TNF-{alpha} antagonist therapy. Table 2 summarizes the probability of initial response for an average patient.


View this table:
[in this window]
[in a new window]

  		TABLE 2. Short-term results from BSRBR on probability of response, utility improvement at 6 months and drug costs over first 2 yrs

 
Impact of initial response on short-term (0–6 months) health utility
The magnitude of change in health utility in the first 6 months of a new treatment is estimated using multivariate regression (Appendix, Part 2). Higher health utility improvement is predicted for patients receiving TNF-{alpha} antagonist agents, those who are moderate/good responders, those who are younger, with shorter disease duration and fewer previous DMARDs. For each covariate, the EQ5D-based coefficients are approximately twice the size of those for SF-6D. In the absence of BSRBR data on 1-month or 3-month utility, we assumed that 80% of a patient's 6-month utility response will be achieved within the first month [53].

Length of longer-term treatment if therapy is continued
Duration of TNF-{alpha} antagonist treatment in the BSRBR is modelled using a multivariate Weibull survival analysis (Appendix, Part 3). Duration of treatment is longer for patients with good EULAR response than for moderate or non-responders. For conventional DMARDs, the BSRBR data were limited mostly to 6 months follow-up, so we used estimates from the literature [20].

Impact of longer-term treatment on longer-term health utility
The model length of follow-up for disease progression (HAQ disability) on a TNF-{alpha} antagonist in the BSRBR was 18 months. In our basecase, longer-term progression after the first 6 months on TNF-{alpha} therapy is evidenced from observational data in Sweden, where trends over 3 yrs suggest that those patients continuing to respond to TNF-{alpha} antagonist therapy maintain a level HAQ disability score [53]. Our basecase assumes that the health utility achieved after 6 months on a TNF-{alpha} antagonist is maintained until patients’ withdrawal. For patients on conventional DMARDs, again the BSRBR follow-up is insufficient, and we have used literature evidence for a mean HAQ progression of 0.042 per annum [54], which is translated to health utility [50].

In a sensitivity analysis, we use a multivariate regression model based on the BSRBR TNF-{alpha} antagonist-treated patients (Appendix, Part 4) probabilistic sensitivity analysis (PSA). The BSRBR data over 6–18 months suggest a slight worsening of utility for patients who achieved good response, whilst moderate responders and poor responders have slight improvements in utility. This could be a slight regression to the mean effect whereby those who initially respond well are more likely to fall back a little, whilst those who had poor response due to random chance have a greater chance of improving slightly.

We make one further assumption concerning longer-term progression in the basecase. Evidence suggests that TNF-{alpha} antagonists prevent or even improve ongoing joint damage whereas conventional DMARDs are associated with worsening joint damage and long-term disability [55]. Information about radiographic damage is not collected in the BSRBR and so is not directly examined in our model. To assume that patients have no worsening in joint damage on TNF-{alpha} antagonist treatment, we use proxy variables—recognizing that disease duration is correlated with joint damage. We hold constant the patient's age and disease duration whilst they remain on a TNF-{alpha} antagonist. This means that when the patient is withdrawn from TNF-{alpha} antagonist therapy, they will follow precisely the same rate of disability progression as if they had gone directly to DMARDs instead of starting their TNF-{alpha} antagonist. This concept is well explained in Kirwan et al. [56]. Removing this assumption is tested in sensitivity analysis.

Utility worsening when treatment when treatment is withdrawn
Published data on health utility after withdrawal from TNF-{alpha} antagonists are almost non-existent. Initial analysis of BSRBR data suggested that utility worsening in the 6-month period after withdrawal appears much smaller than the initial improvements seen. However, the true worsening is not directly measurable for many patients in the BSRBR because of the difference in the timing of data collection, i.e. switching of treatment rarely occurs at exactly the time when the patient completes the quality of life questionnaire. By the time the patient completes the quality of life questionnaire subsequent treatments may have already taken action. In the absence of data from the BSRBR, we have made two assumptions. First, when a patient switches treatment, utility worsens temporarily until the new treatment becomes effective. Second, the worsening is equal to the initial improvement. There is some trial-based evidence that this may be slightly conservative that patients’ disability scores can quickly rebound back to near their baseline [21].

Mortality
We use standard UK life tables [57] adjusted for the standardized mortality ratios for patients with RA [58]. Although evidence is emerging on improved disability being linked to increased longevity, this has not been included, because it is an association between disability and relative risk of mortality [59,60]. There is not as yet a prospectively proven causative link showing that improvements in disability due to TNF-{alpha} antagonists cause a reduction in mortality risk. Mortality is assumed to be equivalent in the two arms.

Semi-annual resource use
Days of hospitalization per 6-month period in the BSRBR are modelled using a multivariate regression model (Appendix 1). Patients with higher baseline utility and those on TNF-{alpha} antagonists have lower levels of hospitalization, whilst those who are older and/or male and those with longer disease duration and higher numbers of previous DMARDs have higher levels of hospitalization. Unit cost per day of hospitalization is assumed the same for both groups.

Unit costs
Unit costs of drug treatment are taken from the British National Formulary [61]. The average cost of TNF-{alpha} antagonist treatment is based on the weighted average use of the different drugs in the BSRBR. The basecase assumes licensed doses are given to each patient. Dosage information is available from several fields in the BSRBR but there are interpretation issues including whether the recorded dose is that given directly to the patient or includes any unused component that may be discarded or used in other patients. We use the doses reported in the BSRBR in a sensitivity analysis including the increases in dosage seen in some patients over the 3 yrs. After 3 yrs, average dose is assumed to remain constant. Monitoring and administration costs are estimated from a previous costing exercise [18]. Sixty-nine per cent of patients treated with TNF-{alpha} antagonist therapy used a conventional DMARD in combination. The expected number of days of hospitalization for each patient is multiplied by the average cost per day in a rheumatology unit [62]. This methodology does not look separately at the type of adverse event or procedure. All costs were updated to financial year 2004 where necessary.

Finally, a number of important costs are excluded: costs of other medications, e.g. in primary care, costs of institutionalization due to disability, costs to or quality of life impact on carers and any lost work productivity due to disability among patients of working age. The likelihood that the excluded costs would differ between the TNF-{alpha} antagonist and conventional DMARDs groups differs for each type of cost. Primary care costs are possibly small and may not be differentially affected. There is little evidence of costs/quality of life on carers. If differentials in disability levels are maintained long term then it is possible that costs of institutionalization and the lost work productivity due to disability would be higher in the conventional DMARD group than in the TNF-{alpha} antagonist group.

sa PSA
PSA involves re-running the model many times, allowing all of the uncertain variables to vary across their range of existing uncertainty at the same time, and examining the effect on cost-effectiveness results. Each model parameter has been characterised with a probability distribution and, where measurable in our statistical modelling, correlation between parameters is also accounted for. This approach is particularly important since we are merging the results of a number of analyses and evidence sources. It should be noted that whilst some of the model parameters account fully for variation between individual patients using available data (e.g. baseline utility and individual good/moderate/poor response) other model parameters are modelled as cohort means (e.g. mean improvement in utility if good responder). PSA has previously been considered difficult or even impossible in such individual sampling models, but we have used recently developed methodology [63], to show that in this particular model, we need 100 Monte Carlo samples for analysing parameter uncertainty each with 50 hypothetical individual patients (total 5000 modelled patient runs) to provide robust estimates of uncertainty.

Guidance analysis model
Alongside the current practice analysis, the model also considers alternative guidance. In particular, we examine policy options of withdrawing patients at 3 months or 6 months if they have not achieved either moderate or good response. We also examine subgroups by age, sex, baseline disability, disease duration and number of previous DMARDs. Finally, we examine monotherapy, combination therapy and sequential therapy in which, given the absence of evidence on correlation, we assume that response to a second TNF-{alpha} antagonist is independent of response to the first.


   Results
Top
 Abstract
 Introduction
 Method
 Results
Discussion
 Appendix
 Acknowledgements
 References
 
The results for analysing current UK practice with TNF-{alpha} antagonist therapies as described in the basecase show an estimated discounted mean lifetime cost of nearly £58 000 on TNF-{alpha} antagonist therapy vs around £21 000 on conventional DMARDs (top line of Table 3). The mean incremental cost of around £37 000 achieves an estimated mean discounted QALY gain of 1.5583 over a lifetime. The incremental cost per QALY gained is estimated at £23 882. This is around the range that has previously been considered acceptable by NICE. The PSA confirms this, showing an 84% probability of being cost-effective at a £30 000 threshold (Fig. 2).


View this table:
[in this window]
[in a new window]

  		TABLE 3. Lifetime cost-effectiveness of TNF-{alpha} antagonists vs traditional DMARDS using the health economic model

 

Figure 2
View larger version (29K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		FIG. 2. Probabilistic sensitivity analysis on the base case. (A) Cost-effectiveness plane. (B) Cost-effectiveness Acceptability Chart (CEAC).

 
Sensitivity analyses concerning each key variable in the model are shown in Table 3, with more details given in the independent project report [36]. The first important sensitivity analysis (S1a) concerns the guidance to withdraw patients not achieving moderate or good EULAR response at 3 months. If this were strictly applied, the model estimates a reduced incremental cost (11% lower than the basecase), a slightly smaller QALY gain (4% lower than the basecase) and a resulting cost per QALY gained of £22 203 (7% lower than the basecase) with a 97% probability of being cost-effective at a £30 000 threshold. Results for rules to withdraw patients unless they achieve ‘good’ EULAR response at 3 months and 6 months also show marginally improved cost-effectiveness (S1). If discount rates of 3.5% for costs and benefits (S2) were applied this would reduce the value of future health gains and the model estimates an incremental cost per QALY of around £32 000, on the borderline of cost-effective with a 36% probability of being below a £30 000 threshold. A more important effect would occur if decision makers were to choose to believe in the validity of SF-6D derived utility with its ‘floor-effect’ problems (S3)—the model estimates an incremental QALY that is almost halved, and the resulting cost per QALY is around £48 000 (double the basecase). The final important sensitivity analysis concerns the assumed disability progression on conventional DMARDs (basecase = 0.0418 HAQ points per annum from Scott et al.). An alternative estimate based on Early Rheumatoid Arthritis Study patients who have previously failed two DMARDs (S4a) would give a cost per QALY of around £18 500 (100% probability of cost-effectiveness below a £30 000 threshold). For patients progressing at a very high rate, e.g. functional class III/IV (S4b), the cost per QALY estimate is around £12 500 (almost 100% probability of cost-effectiveness at both £30 000 and £20 000 thresholds).

Most other variables have a lesser impact on the results. Assumptions concerning the relationship between HAQ disability and utility (S6), whether or not we assume no radiographic progression with the age/disease duration proxy variables (S7), a rule to withdraw any patient whose utility actually worsens (S8) and assuming the 18 months BSRBR trend data on utility levels achieved by successful responders followed by a level utility progression longer term (S10) make little difference to the results. There are important caveats around the reported BSRBR dose data, and if we make a ‘worst case’ assumption, then the cost per QALY would increase to around £30 000 (S9). A series of subgroup analyses (S10–S14) show marginally better cost-effectiveness for subgroups with higher baseline disability, younger starting age of treatment, shorter disease duration and fewer previous DMARDs.

Finally, the model has been used to compare mono-, combination with DMARD and sequential therapy. The results suggest that TNF-{alpha} antagonist plus combination DMARD therapy patients have a slightly higher cost per QALY (around £27 000) than those on monotherapy (around £23 000). When we analysed the use of two TNF-{alpha} antagonists in sequence, making the assumption that the probability of response and utility gained following a switch to a second TNF-{alpha} antagonist are able to be modelled using the same relationships as seen from the BSRBR data, then we find the results only marginally different from the basecase (cost per QALY increasing by just 2%).


   Discussion
Top
 Abstract
 Introduction
 Method
 Results
 Discussion
Appendix
 Acknowledgements
 References
 
This is the first study to assess the cost-effectiveness of TNF-{alpha} antagonist therapies using data from clinical practice using a control arm. The model has used data from the BSRBR, a large registry of patients with RA in the UK. Levels of response, duration of therapy, hospitalisations and ongoing HAQ disability scores to estimate lifetime cost-effectiveness are estimated. The basecase results suggest that current practice in the UK could be considered cost-effective when compared with use of conventional DMARDs only (around £24 000 per QALY gained). If the guidelines set out by NICE in their 2001 appraisal were strictly adhered to, and EULAR non-responders were withdrawn from therapy at 3 months, practice would be marginally more cost-effective (£22 000 per QALY gained), but with a reduction in mean lifetime costs in the TNF-{alpha} strategy of 7% (around £4000). Sensitivity analyses suggest that these conclusions are robust to changes in assumptions for many of the model variables, though some are more sensitive, including discount rates, long-term disease progression on conventional DMARDs and the use of SF-6D derived utility. Sequential therapy with two TNF-{alpha} antagonists appears to have the same order of cost-effectiveness as single therapy though our analysis assumes (in the absence of evidence on correlation) that response to a second TNF-{alpha} antagonist is independent of response to the first. The results of our analysis imply that the earlier decisions by NICE and other bodies to approve TNF-{alpha} antagonist therapies for use in patients who have failed on two previous DMARDs are resulting in cost-effective therapy in clinical practice. As these decisions are reviewed by many bodies across the world, the evidence from the BSRBR make an important contribution, and decision-makers worldwide might adapt this analysis using different costs, discount rates and other factors that might affect results.

In spite of having access to the best available evidence on long-term use of TNF-{alpha} antagonists via the BSRBR, several limitations apply to this analysis. We analyse TNF antagonist therapies as a class (rather than infliximab, etanercept and adalimumab separately) because there are treatment selection biases in the early data in the BSRBR due to differential availability of treatment over the first 4 yrs of use, making fair adjustment difficult. We have not looked at the differences in the cost-effectiveness between TNF-{alpha} antagonists due to more complex selection bias in the early years of BSRBR data that would not be sufficiently accounted for in the general case mix adjustment approach [64]. Recent meta-regression and mixed treatment comparison of clinical trial results suggests little difference in probability of initial response between TNF-{alpha} antagonists [65]. The differential cost of each drug and the cost of administration will directly affect the cost effectiveness results.

We have excluded any potential benefit of TNF-{alpha} antagonist therapies on mortality and, since studies show a significant association between HAQ improvement and mortality risk reduction, evidence in the next few years from the BSRBR may allow us to quantify any such effect. On the cost side, we have excluded any cost differences which might be incurred in the future due to increased institutionalization due to disability, longer-term surgeries beyond the current 18-month BSRBR data horizon, costs to carers and any lost work productivity due to disability among patients of working age. In general, these exclusions might be expected to favour the comparator arm, that is, our analysis might be considered conservative. Were we to include these costs then the cost per QALY gained would probably reduce, showing TNF-{alpha} antagonists as more cost-effective than our quantified results so far.

Our modelling approach extends that of earlier analyses and differs in several important ways. Most importantly, we build upon a large database concerning current UK clinical practice. Because we have been able to analyse patient level data, the model parameters can all be adjusted for important covariates, resulting in greater confidence that any biases are accounted for, and at the same time providing the opportunity to undertake subgroup analyses. The model takes a UK NHS cost perspective and covers the whole lifetime (as required by NICE) and so our results are not comparable with other analyses, which use different cost perspectives and time horizons [19,23,26,27,29]. Welsing et al. argue for their 5-yr time horizon to avoid ‘too many assumptions’ though it effectively makes two strong assumptions of zero cost difference and zero benefit difference between the two arms after their chosen horizon. The other main differences between our approach and that of Barton and colleagues concern: the use of the concept of treatment withdrawal unless a particular level of response is achieved; the disability/utility improvement linked to the level of response achieved rather than using average improvement in HAQ scores, and the level of disability progression on conventional DMARDs, which we take as the weighted average of Scott et al. [54] as opposed to a simple average of the study means [66]. Under the assumptions represented by our basecase, we arrive at a central cost-effectiveness estimate of £23 882 per QALY. This is slightly higher than the figure of £16 330 per QALY produced by previous work modelling the cost-effectiveness of etanercept [21]. Previous work used one-way sensitivity analysis to generate a range of estimates from £7800 to £42 000 per QALY. Our work has also allowed us to examine the degree of certainty in cost-effectiveness. We can now say that there is an 84% chance that the true value is below £30 000 per QALY. It is for policy makers locally, nationally and worldwide to consider whether this is an appropriate use of resources in the context of other competing treatments and diseases.

The statistical modelling to adjust for covariates also has inbuilt assumptions. For the probability of response model, the proportional odds cumulative logit approach assumes that the odds ratio of ‘no DAS response’ to ‘no or moderate DAS response’ is the same for all values of the coefficients, e.g. the odds ratio is the same for men and women. This is similar for the ‘good DAS response’ to ‘good or moderate DAS response’ odds ratio too. This is the most common standard assumption made in regression models for ordered categorical responses, and we believe it is reasonable in this context. For the Weibull survival on therapy analysis, the proportional hazard assumption for the risk of withdrawal is, again, a common standard assumption. There are only 2 yrs of data collected every 6 months to estimate these curves so we are unable to check this assumption, and it should be noted that the inference from a Weibull analysis is quite robust to deviations from the proportional hazard assumption

There remain additional opportunities for further research using the BSRBR data. We accounted for selection bias between the TNF-{alpha} antagonist and control arms in the BSRBR data by adjusting for the main covariates, but this approach has limitations and alternative methods including propensity analysis might be worthwhile. Incidentally, the potential ‘bias’ in analyses using unrepresentative RCTs rather than registry information is not entirely clear cut in scale or direction. RCTs may, on the one hand, represent the best possible use of the drugs, in the sense of active preparation of the patients before treatment and more than usually intensive monitoring. On the other hand, RCTs select relatively healthy, younger patients with fewer co-morbidities and the improvements seen could be substantially different for older, less healthy patients in clinical practice. The BSRBR data on dose used requires some caution in interpretation (hence our basecase assuming standard recommended doses) and a further integrated analysis of the several fields collected regarding dosage would be useful. We have examined adverse events as a class in terms of hospitalisations but have not examined individual adverse events or their costs or the quality-of-life impact. More detailed analysis, particularly on chronic adverse events would be helpful in future evaluations. Finally, the comparison of hospitalisation costs between arms may be biased since patients about to receive joint surgery for example may be deemed unsuitable to start TNF-{alpha} antagonist therapy. Ongoing data collection on hospitalizations will be valuable. Once further data is available from the BSRBR, a revision to this analysis would be recommended.

In conclusion, our basecase analysis suggests that the probability the TNF-{alpha} antagonist therapies can be considered a cost-effective therapy (below £30 000 per QALY) is high in the UK context. The BSRBR has been a valuable source of evidence in deriving answers to the decision questions posed. There remain some uncertainties on the much longer-term costs and benefits and, as the BSRBR continues, it will be an increasingly useful source of such evidence.


   Appendix
Top
 Abstract
 Introduction
 Method
 Results
 Discussion
 Appendix
Acknowledgements
 References
 
Statistical models using BSRBR data
Table A1 shows the coefficients for the following statistical models.


View this table:
[in this window]
[in a new window]

  		TABLE A1 Statistical models

 
(Part 1) Predicted EULAR response
To predict the probability of a EULAR responce as none (DAS = 0), moderate (DAS = 1) or good (DAS = 2), we used a proportional odds cumulative logit model:


Formula



Formula



Formula

where {gamma} = ({gamma}1, ... , {gamma}n) are the coefficients for the covariates (i.e. the patient characteristics). A positive coefficient indicates greater likelihood of successful response. The negative coefficient for disease duration (Table A1) appears counter intuitive but as individuals progress through the clinical pathways model each year, the age and disease duration both increase and the negative disease duration effect is counter balanced by the positive age coefficient.

(Part 2) Short-term health utility from treatment
Utility is defined on a scale from –0.6 to 1, whereas the methods we wish to use require a variable to be in the range (0,1) i.e. not including 0 or 1. We define a function to transform the range –0.6 : 1 to 0.025 : 0.975


Formula

And hence a logit type function for the range –0.6 : 1


Formula

We modelled the short-term (0–6 months) impact of treatment on health utility. The predicted utility at 6 months u6 is given by


Formula

where {alpha} is an intercept, ß is the coefficient for the transformed baseline utility l(u0) and the {gamma} are the coefficients for the other covariates.

(Part 3) Predicted time to withdrawal: TNF-{alpha} antagonists
We fitted a Weibull survival curve to the BSRBR data, using a proportional hazard model to estimate the impact of patient characteristics.

This involves fitting a survival function of the form:


Formula

where t is the time in months, s is the probability of still being on the drug at time t, and {alpha} and ß are the shape and scale parameters of the distribution.

We fitted the model firstly to patients on their first TNF-{alpha} antagonist (Part 3a), and secondly to patients on continuous use of any TNF-{alpha} antagonist (Part 3b).

For duration on therapy for a DMARD, we used information from Barton et al.

(Part 4) Predicted improvement in utility over 6–18 months on TNF-{alpha} antagonist treatment
Predicted utility over 6–18 months after starting TNF-{alpha} antagonist treatment is a function of time t in months after 6 months of treatment, transformed utility at 6 months l(u6), and DAS response at 6 months.


Formula

where ß is the coefficient for time in months, and dDAS is a DAS response-time interaction effect. DAS = 1, 2, or 3 for none, moderate and good response. d3 = 0 for an identifiability constraint so is the reference case. Note that the coefficient on l(u6) is fixed at 1 so the utility progression over time in continuous between 0–6 months and 6–18 months. Although this model is fitted using data from 6–18 months, it is used to extrapolate utility beyond this time.

(Part 5) Predicted days of hospitalization per 6-month period
The number of hospital days (length of stay) in a 6-month period, s, is predicted as a function of patient characteristics.


Formula


   Acknowledgements
Top
 Abstract
 Introduction
 Method
 Results
 Discussion
 Appendix
 Acknowledgements
References
 
The British Society for Rheumatology (BSR) commissioned the Biologics Register (BSRBR) as an UK-wide national project to investigate the safety of biologic agents in routine medical practice. BSR receives restricted income financial support from UK pharmaceutical companies, presently Abbott Laboratories, Amgen, Schering Plough and Wyeth Pharmaceuticals. This income finances a wholly separate contract between the BSR and the University of Manchester who provide and run the BSRBR data collection, management and analysis services. The principal investigators and their team have full academic freedom and are able to work independently of pharmaceutical industry influence. All decisions concerning analyses, interpretation and publication are made autonomously of any industrial contribution. Members of the Manchester team, BSR trustees, committee members and staff complete an annual declaration in relation to conflicts of interest. Under the terms of the contract between the BSR and the pharmaceutical companies, copies of publications are provided prior to submission for review. The purpose of this review is to enable the companies to protect their internal confidential or proprietary information. With the sole and rare exception of bringing attention to matters of factual error, they have no role in the interpretation or preparation of material for publications. All papers are also reviewed and approved in advance of publication by the BSR Project Steering Committee. Full responsibility for all data, results, analyses, interpretation of findings rests solely with the authors and principal investigators. The BSRBR is also financially supported by the UK's Arthritis Research Campaign programme support to professors A.S. and D.S. The authors acknowledge the enthusiastic collaboration of all consultant rheumatologists and their specialist nurses in the UK in providing the data used in this report. The substantial contribution of Andy Tracey and Katie McGrother in database design and manipulation is acknowledged.

We also acknowledge the extensive support given to the project by members of the BSRBR project management and steering committees, its chairman, first Dr Ian Griffiths and now Professor David Isenberg plus the officers and staff of the British Society of Rheumatology.

This study was originally proposed by the authors for a research-funding bid to the Arthritis and Rheumatism Campaign. The bid was unsuccessful. The timely requirement for an analysis of the BSRBR for NICE) meant that the project team began work without funding. A small grant of £10 000 from the BSR has subsequently met a proportion of the costs.

The University of Sheffield received a research grant of £10 000 from the BSR to support this project. The authors have declared no other specific conflicts of interest.


   References
Top
 Abstract
 Introduction
 Method
 Results
 Discussion
 Appendix
 Acknowledgements
 References
 

  1. Symmons D, Turner G, Webb R, et al. The prevalence of rheumatoid arthritis in the United Kingdom: new estimates for a new century. Rheumatology (2002) 41:793.[Abstract/Free Full Text]
  2. Yelin E, Callahan LF, for the National Arthritis Data Work Group. The economic cost and social psychological impact of musculoskeletal conditions. Arthritis Rheum (1995) 38:1351–6.[Web of Science][Medline]
  3. Pincus T, Callahan LF. The ‘side effects’ of rheumatoid arthritis: joint destruction, disability, and early mortality. Br J Rheumatol (1993) 32:28–37.
  4. Wolfe F, Michaud K, Gefeller O, et al. Predicting mortality in patients with rheumatoid arthritis. Arthritis Rheum (2003) 48:1530–42.[CrossRef][Web of Science][Medline]
  5. Wong JB, Ramey DR, Singh G. Long-term morbidity, mortality, and economics of rheumatoid arthritis. Arthritis Rheum (2001) 44:2746–9.[CrossRef][Web of Science][Medline]
  6. Olsen NJ, Stein CM. New drugs for rheumatoid arthritis. N Engl J Med (2004) 350:2167–79.[Free Full Text]
  7. Moreland LW, Schiff MH, Baumgartner SW, et al. Etanercept therapy in rheumatoid arthritis - A randomized, controlled trial. Ann Intern Med (1999) 130:478–86.[Abstract/Free Full Text]
  8. Weinblatt ME, Kremer JM, Bankhurst AD, et al. A trial of etanercept, a recombinant tumor necrosis factor receptor: Fc fusion protein, in patients with rheumatoid arthritis receiving methotrexate. N Engl J Med (1999) 340:253–9.[Abstract/Free Full Text]
  9. Bathon JM, Martin RW, Fleischmann RM, et al. A comparison of etanercept and methotrexate in patients with early rheumatoid arthritis. N Engl J Med (2000) 343:1586–93.[Abstract/Free Full Text]
  10. Klareskog L, van der Heijde D, de Jager JP, et al. Therapeutic effect of the combination of etanercept and methotrexate compared with each treatment alone in patients with rheumatoid arthritis: double-blind randomised controlled trial. Lancet (2004) 363:675–81.[CrossRef][Web of Science][Medline]
  11. Maini R, St Clair EW, Breedveld F, et al. Infliximab (chimeric anti-tumour necrosis factor alpha monoclonal antibody) versus placebo in rheumatoid arthritis patients receiving concomitant methotrexate: a randomised phase III trial. Lancet (1999) 354:1932–9.[CrossRef][Web of Science][Medline]
  12. St Clair EWS, van der Heijde DMFM, Smolen JS, et al. Combination of infliximab and methotrexate therapy for early rheumatoid arthritis - A randomized, controlled trial. Arthritis Rheum (2004) 50:3432–43.[CrossRef][Web of Science][Medline]
  13. Weinblatt ME, Keystone EC, Furst DE, et al. Adalimumab, a fully human anti-tumor necrosis factor alpha monoclonal antibody, for the treatment of rheumatoid arthritis in patients taking concomitant methotrexate: the ARMADA trial. Arthritis Rheum (2003) 48:35–45.[CrossRef][Web of Science][Medline]
  14. Keystone EC, Kavanaugh AF, Sharp JT, et al. Radiographic, clinical, and functional outcomes of treatment with adalimumab (a human anti-tumor necrosis factor monoclonal antibody) in patients with active rheumatoid arthritis receiving concomitant methotrexate therapy: a randomized, placebo-controlled, 52-week trial. Arthritis Rheum (2004) 50:1400–11.[CrossRef][Web of Science][Medline]
  15. van de Putte LB, Atkins C, Malaise M, et al. Efficacy and safety of adalimumab as monotherapy in patients with rheumatoid arthritis for whom previous disease modifying antirheumatic drug treatment has failed. Ann Rheum Dis (2004) 63:508–16.[Abstract/Free Full Text]
  16. Breedveld FC, Weisman MH, Kavanaugh AF, et al. The PREMIER study - A multicenter, randomized, double-blind clinical trial of combination therapy with adalimumab plus methotrexate versus methotrexate alone or adalimumab alone in patients with early, aggressive rheumatoid arthritis who had not had previous methotrexate treatment. Arthritis Rheum (2006) 54:26–37.[CrossRef][Web of Science][Medline]
  17. Michaud K, Messer J, Choi HK, et al. Direct medical costs and their predictors in patients with rheumatoid arthritis: a three-year study of 7527 patients. Arthritis Rheum (2003) 48:2750–62.[CrossRef][Web of Science][Medline]
  18. Jobanputra P, Barton P, Bryan S, Burls A. The effectiveness of infliximab and etanercept for the treatment of rheumatoid arthritis: a systematic review and economic evaluation. Health Technol Assess (2002) 6:1–110.[Medline]
  19. Kobelt G, Jonsson L, Young A, Eberhardt K. The cost-effectiveness of infliximab (Remicade) in the treatment of rheumatoid arthritis in Sweden and the United Kingdom based on the ATTRACT study. Rheumatology (2003) 42:326–35.[Abstract/Free Full Text]
  20. Barton P, Jobanputra P, Wilson J, Bryan S, Burls A. The use of modelling to evaluate new drugs for patients with a chronic condition: the case of antibodies against tumour necrosis factor in rheumatoid arthritis. Health Technol Assess (2004) 8:iii. 1–91.[Medline]
  21. Brennan A, Bansback N, Reynolds A, Conway P. Modelling the cost-effectiveness of etanercept in adults with rheumatoid arthritis in the UK. Rheumatology (2004) 43:62–72.[Abstract/Free Full Text]
  22. Clark W, Jobanputra P, Barton P, Burls A. The clinical and cost-effectiveness of anakinra for the treatment of rheumatoid arthritis in adults: a systematic review and economic analysis. Health Technol Assess (2004) 8:iii–iv. ix–x, 1–105.[Medline]
  23. Choi HK, Seeger JD, Kuntz KM. A cost-effectiveness analysis of treatment options for patients with methotrexate-resistant rheumatoid arthritis. Arthritis Rheum (2000) 43:2316–27.[CrossRef][Web of Science][Medline]
  24. Rubio-Terres C, Dominguez-Gil A. Pharmacoeconomic analysis of the treatment of rheumatoid arthritis with leflunomide in comparison with the combination of infliximab and methotrexate. J Drug Assess (2001) 4:39–54.
  25. Choi HK, Seeger JD, Kuntz KM. A cost effectiveness analysis of treatment options for methotrexate-naive rheumatoid arthritis. J Rheumatol (2002) 29:1156–65.[Abstract/Free Full Text]
  26. Wong JB, Singh G, Kavanaugh A. Estimating the cost-effectiveness of 54 weeks of infliximab for rheumatoid arthritis. Am J Med (2002) 113:400–8.[CrossRef][Web of Science][Medline]
  27. Kobelt G, Eberhardt K, Geborek P. TNF antagonists in the treatment of rheumatoid arthritis in clinical practice: costs and outcomes in a follow up study of patients with RA treated with etanercept or infliximab in southern Sweden. Ann Rheum Dis (2004) 63:4–10.[Abstract/Free Full Text]
  28. Bansback NJ, Brennan A, Ghatnekar O. Cost effectiveness of adalimumab in the treatment of patients with moderate to severe rheumatoid arthritis in Sweden. Ann Rheum Dis (2005) 64:995–1002.[Abstract/Free Full Text]
  29. Welsing PM, Severens JL, Hartman M, van Riel PL, Laan RF. Modeling the 5-year cost effectiveness of treatment strategies including tumor necrosis factor-blocking agents and leflunomide for treating rheumatoid arthritis in the Netherlands. Arthritis Rheum (2004) 51:964–73.[CrossRef][Web of Science][Medline]
  30. Maetzel A. Cost-effectiveness analysis: Out of touch with clinical reality? Arthritis Care Res (2005) 53:3–4.[CrossRef][Web of Science]
  31. Wolfe F, Michaud K, Pincus T. Do rheumatology cost-effectiveness analyses make sense? Rheumatology (2004) 43:4–6.[Free Full Text]
  32. Bansback NJ, Young A, Brennan A. The NICE reappraisal of biologics in 2005: what rheumatologists need to know. Rheumatology (2005) 44:3–4.[Free Full Text]
  33. Bansback NJ, Regier DA, Ara R, et al. An Overview of Economic Evaluations for Drugs Used in Rheumatoid Arthritis Focus on Tumour Necrosis Factor-{alpha} Antagonists. Drugs (2005) 65:473–96.[CrossRef][Web of Science][Medline]
  34. National Institute for Clinical Excellence. Guidance on the use etanercept and infliximab for the treatment of rheumatoid arthritis. (2002) March. http://www.nice.org.uk/pdf/RA-PDF.pdf.
  35. Watson K, Symmons D, Griffiths I, Silman A. The British Society for Rheumatology biologics register. Ann Rheum Dis. (2005) 64:iv42–3.[Abstract/Free Full Text]
  36. Brennan A, Bansback N, Nixon R. Modelling the cost effectiveness of TNF inhibitors in the management of rheumatoid arthritis: Results from the British Society for Rheumatology Biologics Registry, An Independent Report. (2005) May. Available at: http://www.shef.ac.uk/scharr/sections/heds/discussion.html.
  37. Maetzel A, Tugwell P, Boers M, et al. OMERACT 6 Economics Research Group. Economic evaluation of programs or interventions in the management of rheumatoid arthritis: defining a consensus-based reference case. J Rheumatol (2003) 30:891–6.[Abstract/Free Full Text]
  38. Torrance GW. Measurement of health state utilities for economic appraisal. J Health Econ (1986) 5:1–30.[CrossRef][Web of Science][Medline]
  39. Barton P, Bryan S, Robinson S. Modelling in the economic evaluation of health care: selecting the appropriate approach. J Health Serv Res Policy (2004) 9:110–18.[Abstract/Free Full Text]
  40. Claxton K, Sculpher M, McCabe C, et al. Probabilistic sensitivity analysis for NICE technology assessment: not an optional extra. Health Econ (2005Apr) 14:339–47.[CrossRef][Web of Science][Medline]
  41. NICE. Retrieved from: http://www.nice.org.uk/pdf/Final_protocol_Anti_TNF.pdf.
  42. Prevoo ML, van't Hof MA, Kuper HH, van Leeuwen MA, van de Putte LB, van Riel PL. Modified disease activity scores that include twenty-eight-joint counts. Development and validation in a prospective longitudinal study of patients with rheumatoid arthritis. Arthritis Rheum (1995) 38:44–8.[Web of Science][Medline]
  43. Fries JF, Spitz P, Kraines G, Holman H. Measurement of Patient Outcome in Arthritis. Arthritis Rheum (1980) 23:137–45.[Web of Science][Medline]
  44. Ware JE Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care (1992) 30:473–83.[Web of Science][Medline]
  45. van Gestel AM, Prevoo ML, van 't Hof MA, van Rijswijk MH, van de Putte LB, van Riel PL. Development and validation of the European League Against Rheumatism response criteria for rheumatoid arthritis. Comparison with the preliminary American College of Rheumatology and the World Health Organization/International League Against Rheumatism Criteria. Arthritis Rheum (1996) 39:34–40.[Medline]
  46. Brazier J, Roberts J, Deverill M. The estimation of a preference-based measure of health from the SF-36. J Health Econ (2002) 21:271–92.[CrossRef][Web of Science][Medline]
  47. Brazier J, Roberts J, Tsuchiya A, Busschbach J. A comparison of the EQ-5D and SF-6D across seven patient groups. Health Econ (2004) 13:873–84.[CrossRef][Web of Science][Medline]
  48. Dolan P. Modeling valuations for EuroQol health states. Med Care (1997) 35:95–108.[CrossRef][Web of Science][Medline]
  49. Marra CA, Woolcott JC, Kopec JA, et al. A comparison of generic, indirect utility measures (the HUI2, HUI3, SF-6D, and the EQ-5D) and disease-specific instruments (the RAQoL and the HAQ) in rheumatoid arthritis. Soc Sci Med (2005) 60:1571–82.[CrossRef][Web of Science][Medline]
  50. Bansback N, Marra C, Tsuchiya A, Anis A, Guh D, Hammond A, Brazier JE. Using the Health Assessment Questionnaire to estimate preference-based single indices in patients with Rheumatoid Arthritis. In: [2006] Arthritis Care & Research. In Press.
  51. Murray MP. Econometrics. In: A modern introduction (2006) Boston: Pearson Education Inc.
  52. Greene WH. edn. In: Econometric analysis (2000) 4th. New Jersey: Prentice Hall Inc.
  53. Kobelt G, Eberhardt K, Geborek P. TNF inhibitors in the treatment of rheumatoid arthritis in clinical practice: costs and outcomes in a follow up study of patients with RA treated with etanercept or infliximab in southern Sweden. Ann Rheum Dis (2004) 63:4–10.[Abstract/Free Full Text]
  54. Scott DL, Garrood T. Quality of life measures: use and abuse. Baillieres Clin Rheumatol (2000) 14:663–87.[CrossRef]
  55. Scott DL. Prognostic factors in early rheumatoid arthritis. Rheumatology (2000) 39:24–9.[Web of Science][Medline]
  56. Kirwan JR. Conceptual issues in scoring radiographic progression in rheumatoid arthritis. J Rheumatol (1999) 26:720–5.[Web of Science][Medline]
  57. Standard UK life table (Table 16). Available at: www.gad.gov.uk.
  58. Standard UK life table (Table 17). [WHO Global Burden of Disease programme].
  59. Wolfe F, Mitchell DM, Sibley JT, Fries JF, Bloch DA, Williams CA. The mortality of rheumatoid arthritis. Arthritis Rheum (1994) 37:481–94.[Web of Science][Medline]
  60. Sokka T, Hakkinen A, Krishnan E, Hannonen P. Similar prediction of mortality by the health assessment questionnaire in patients with rheumatoid arthritis and the general population. Ann Rheum Dis (2004) 63:494–7.[Abstract/Free Full Text]
  61. British National Formulary. Accessed February 2005. Available at: www.bnf.org/bnf/.
  62. Netten A, Curtis L. Unit Costs of Health and Social Care. (2002) Personal Social Services Research Unit, University of Kent.
  63. O'Hagan A, Stevenson M, Madan J. Monte Carlo probabilistic sensitivity analysis for patient level simulation model, (2005). Health Economics, In Press.
  64. Deeks JJ, Dinnes J, D'Amico R, et al. International Stroke Trial Collaborative Group; European Carotid Surgery Trial Collaborative Group. Evaluating non-randomised intervention studies. Health Technol Assess (2003) 7:iii–x. 1–173.[Medline]
  65. Nixon RM, Bansback N, Brennan A. Using mixed treatment comparisons and meta-regression to perform indirect comparisons to estimate the efficacy of biologic treatments in rheumatoid arthritis. Statistics in Medicine (2007) 26((6):):1237–54.[CrossRef][Web of Science][Medline]
  66. Bansback NJ, Young A, Brennan A. The NICE reappraisal of biologics in 2005: what rheumatologists need to know. Rheumatology (2005) 44:3–4.[Free Full Text]
Submitted 18 September 2006; revised version accepted 26 March 2007.
Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Rheumatology (Oxford)Home page
R. F. van Vollenhoven
Comment on: Modelling the cost effectiveness of TNF antagonists in the management of rheumatoid arthritis: results from the British Society for Rheumatology Biologics Registry
Rheumatology, July 1, 2008; 47(7): 1106 - 1107.
[Full Text] [PDF]

Home page
 Rheumatology (Oxford)Home page
N. Bansback, A. Brennan, D. Symmons, R. Nixon, J. Madan, M. Harrison, and K. Watson
Comment on: Modelling the cost effectiveness of TNF-{alpha} antagonists in the management of rheumatoid arthritis: results from the British Society for Rheumatology Biologics Registry: reply
Rheumatology, July 1, 2008; 47(7): 1107 - 1107.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
46/8/1345    most recent
kem115v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (8)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Brennan, A.
Right arrow Articles by Symmons, D.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Brennan, A.
Right arrow Articles by Symmons, D.
Related Collections
Right arrow Health Economics
Right arrow Rheumatoid Arthritis
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?