Assessment group response to Wyeth commentary on assessment report Subject & Wyeth s comments related section/page Model patient The TAR economic model is a complex model that attempts to reflect the population multiple treatment options and pathways an RA patient can follow. Assessment group response Model distribution of baseline HAQ This model demonstrates that it would be cost effective to administer etanercept to a patient with high HAQ on presentation to a clinician, whether they had received prior DMARD therapy or not. The data as it is presented does not reflect the marketing authorisation of etanercept as it uses a patient population distribution that includes patients with milder disease. The clinical and cost effectiveness case put forward by Wyeth is that patients with a high HAQ, erosions identified by x-ray and are rheumatoid factor positive should receive etanercept therapy. Wyeth have undertaken further subgroup analysis to demonstrate this. The Birmingham TAR Group s model has a major weakness in the choice of baseline input data, which is based on NOAR. In this inception cohort, nearly 50% of patients did not receive treatment with either DMARDs or steroids over a 5-year period. Forty five percent of patients had baseline HAQ scores <1 and 65% <1.5. Patients with HAQ scores <1 would not be considered to have severe disease and therefore would not be candidates for therapy with etanercept. The scope for improvement and therefore the QALYs gained in these patients would be limited. It is not surprising therefore that the ICERs are much higher than those obtained in the 3 company models. Graph 1. below, demonstrates the differences in this distribution. The SPC for etanercept states: Enbrel can be used alone or in combination with methotrexate for the treatment of active rheumatoid arthritis in adults when the response to disease-modifying antirheumatic drugs, including methotrexate (unless contraindicated), has been inadequate. Disease may be severe in the absence of erosions and rheumatoid factor and their presence is not necessary in order to use etanercept. The baseline HAQ scores were chosen from Wiles et al A&R 2001 (Ref 169 in our report). This study provides HAQ scores for those starting DMARDs as well as those not given DMARDs. The median HAQ score for the population given DMARDs was 1.25 (IQR 0.625-1.75), as used in our model. DMARDs in the UK are virtually always initiated or recommended after assessment in hospital. This is therefore the appropriate data for a population of patients commencing DMARDs. Our model used the correct IQR but median 1; we have calculated revised results with the correct median and quartiles and the changes are very small. It is inappropriate to compare, in a chart, baseline HAQ scores for patients commencing DMARDs in a model of RA where patients may not be given a TNF inhibitor at all versus patients selected for inclusion in a clinical trial of etanercept.
Model MTX naïve patients Model Wyeth s additional analysis Trial data Executive summary, p17 Executive summary, p18, 1 st Executive summary, p.19, 2 nd bullet Executive summary, p.20 The approved indication for etanercept in patients who have not been treated with methotrexate specifically states severe, active and progressive rheumatoid arthritis. The BRAM therefore has not modelled within the etanercept licence. Appendix 1 outlines the cost-effectiveness, in terms of /QALY for etanercept used either as first, second or third line. The various incremental cost-effectiveness ratios (ICERs) are represented for each quartile of patients from the TEMPO study. Starting HAQ, initial HAQ change and medium term HAQ change have been incorporated for each quartile. Please be advised that the following data are soon to be published and are therefore no longer considered commercial in confidence: - Data from the 2-year analysis of the TEMPO study (protocol 881A- 308) have been accepted for publication - The 6-month results from the Codreanu study (protocol 881A-309) have been posted on htpp://www.clinicaltrials.gov and have been submitted for publication. Current recommendations and service provision The first bullet point should be...used in patients with clinically active disease that has not responded adequately to at least 2 DMARDs including MTX (unless contraindicated). Current recommendations and service provision It is stated that there is variable implementation of guidance. This needs to be expanded to give reasons (more details are provided in the body of the report). Direction of evidence and treatment effect In the bullet referring to adalimumab, reference is made to radiographic joint damage, which is a key endpoint. No reference is made to this in the bullet referring to etanercept, yet data from 2 studies (TEMPO and ERA) clearly demonstrate a significant effect when compared with MTX. Existing economic evaluations Responders and non-responders can be identified using clinical The SPC for etanercept does not define severe, active or progressive. HAQ scores for our population starting DMARDs cannot be regarded as indicating the HAQ of the population given TNF inhibitors. - Thank you. Accepted. We do not feel this is necessary, as it is made clear later. We accept the comments and the text will be revised to read: Adalimumab was marginally less effective and etanercept was marginally more effective than methotrexate in reducing signs and symptoms of RA. Etanercept was better tolerated than methotrexate. Both adalimumab and etanercept were more effective than methotrexate in slowing radiographic joint damage. DAS28 is not useful for decisions about individual patients, as described in the introduction.
Executive summary, p.20 Background, p.23 3 rd Section 2.1.9, p.28 last Section 2.3.2, p-34-35 Section 2.3.4, p.36 assessments such as DAS28, as recommended by the BSR. Cost effectiveness Reference is made to the use of both early and late RA data giving different results. This is an expected result. Early RA data should not be used to model 3rd line as patients are very unlikely to be early (however defined the studies tended to require <3y, but average was <12m). This should be noted by the TAR group. Summary Guidance stated that etanercept and infliximab should be...used in patients with clinically active disease that has not responded adequately to at least 2 DMARDs including MTX (unless contraindicated). Data from the BSRBR shows that in practice, patients have on average failed 4 DMARDs before receiving these drugs. Current drug therapy The use of the word commonly for some of the drugs e.g. azathioprine is not appropriate. In the paper by Edwards et al (Rheumatol, 2005) based on the GPRD database, azathioprine was not mentioned, and in 2002 less than 3% of prescriptions were for gold or penicillamine. Etanercept (Enbrel ) No specific mention is made of use in early disease, only to patients not treated with MTX, whereas, the section on adalimumab refers to early disease. As per the emc, adalimumab is indicated for the treatment of severe, active and progressive rheumatoid arthritis in adults not previously treated with methotrexate which is the same as etanercept. Therefore the wording for the 2 drugs should be the same. Special precaution Data for adalimumab are limited but an increased risk has also been shown. Reference should be made to the oral presentation from Abbott at EULAR 05 (Perez et al OP0093. Impact of screening for latent TB prior to initiating anti-tnf therapy in North America and Europe). It should be noted that formal screening for TB was included in most of the adalimumab studies. Screening for TB has not been a requirement of We provided the results for 3rd line treatment with early RA data as a comparison. Note that the results Wyeth say should not be used are the results more favourable to TNF inhibitors. This is acknowledged in our report. GPRD data may not reflect the range of DMARDs used. In Birmingham, for example, many GPs do not initiate or prescribe azathioprine and gold. This is done by hospital rheumatology departments. Accepted and relevant text will be changed. The statement as it stands is accurate. We are not in a position to cite oral presentations. The BSR/NICE guidance of 2002 requires screening for TB in all patients treated with TNF inhibitors. This has been strengthened in recently
Section 2.3.4, p.36 Section 2.4, p.37, 1 st bullet Section 2.4, p.39, 1 st Section 2.4, p.39, 1 st bullet Section 3.1.1, p.42, last any study with etanercept. Special precaution Official summary of product characteristics (SPC) and guidance, including proposed guidance from the BSR, British Thoracic Society and the British Society for Gastroenterology recommends screening patients before treatment. The SmPC for etanercept does not recommend screening. Current Guidance The primary objective, for which the registry is powered, is to detect a 2- fold increase in lymphoma. Current Guidance..although the veracity of the recorded data is unclear seems overly critical of such a major initiative to assess these drugs in clinical practice. To institute audit etc would significantly impact on the study and increase costs. The report references the issues already experienced with collecting data for the registries. Current Guidance In the BSRBR 41% of patients classified as non-responders It should be made clear the number of patients who were nonresponders, and the % this was of the total, so that the 41% can be put into context. In fact the number is small. Therefore 41% could be misleading. Search Strategy: Clinical Effectiveness as the previous reported had should read as the previous report had published BSR and British Thoracic Society guidance. Accepted. The text will be revised to read: The summary of product characteristics (SPC) for adalimumab and infliximab and guidance including proposed guidance from the BSR, British Thoracic Society and the British Society for Gastroenterology recommend screening patients before treatment. Accepted. The original text: The BSRBR is a prospective cohort study designed to compare the risk, over 5 years, of developing malignancy, lymphoproliferative malignancy, infection requiring hospitalisation, serious comorbidity and death. will be revised to read: The aim of the BSRBR is to establish the long-term safety of a variety of biologic agents (including TNF inhibitors) in adult patients with RA and other rheumatic diseases. In particular the BSRBR is interested in mortality, malignancy and serious adverse events and its sample size was based on being able to detect a two-fold increase in risk of lymphoma over five years. We are concerned with data accuracy not with the perceived importance of this initiative. We believe there are serious questions about the accuracy of the data in the registry and that an audit would go some way to allaying our concerns. We believe this is sufficiently clear, when it is read in context. Accepted and text to be corrected.
Section 3.1.5.2, p.46 Section 3.1.5.2, p.48 Section 3.2.3 Section 3.2.3.1 P.82 Approach for meta-analysis The methodology uses variable periods of follow-up so 3-month results or 2 year results are compared. The authors do accept this as a possible weakness (end of report). They have also looked at results available at fixed periods, 1, 3, 6, and 12 months, but only for ACR20. It could be argued that 40 mg ow of adalimumab and 25 mg ow of etanercept should have been included as they are both approved doses. Approach for meta-analysis 2nd para gives PREMIER as an example study but TEMPO was the 1st study with this design Etanercept General comments: The order in the tables is not the same as the text. Weinblatt 99 is not described in the text It would be more helpful if the studies were grouped in the way the metaanalysis was done. Versus conventional DMARD (parallel start) Versus placebo (with or without continued DMARD) Combination parallel start Description of included etanercept trials -ERA study compared with placebo. Should be MTX Unfortunately we are limited by the primary evidence available to us, however, we believe our approach is justified given the need to succinctly summarise a large volume of evidence and the fact that observed treatment effects in the trials do not appear to change significantly over time. The license indicates that the dose of adalimumab can be increased from 40 mg every other week to 40 mg every week if the patient experiences a decrease in response. This strategy was adopted in some of the adalimumab trials (such as PREMIER) and thus our analysis of patients starting 40 mg every other week does include such dose increase. The text in the method will be revised to read: Adalimumab: 40 mg every other week (may be increased to every week if response is inadequate) or 20 mg every week. Patients who started with 40 mg every week were excluded from the primary analysis but were included in the sensitivity analyses of above licensed doses and all doses. The results of sensitivity analyses are consistent with the primary analysis. BNF 50 (September 2005) does not indicate that etanercept 25 mg once weekly is an approved dose. We agree. PREMIER was used as an example of three-arm trials because the readers will encounter adalimumab trials first in our report, which presents the three TNF inhibitors in alphabetic order. Comments noted. We will edit the report so the studies will be grouped and ordered in a consistent way for all drugs. Accepted and text to be corrected.
Section 3.2.3.1 p.83 last Section 3.2.3.1 p.84 1st Section 3.2.3.2 p.86 1st Section 3.2.3.2 p.98 Table 10 Section 3.1.3 p.138 Description of included etanercept trials -TEMPO The methodology used was the Van der Heidje modification of TSS Description of included etanercept trials -TEMPO However around 42% of patients in each arm of this trial had previously tried methotrexate. It is not at all clear why these individuals discontinued methotrexate in the face of active disease if, as stated in the entry criteria, the drug was not ineffective nor toxic. Patients who had previously received MTX in this study took the drug a mean of 2 years previously. The doses were generally low by current standards and MTX was usually discontinued due minor adverse events or lack of efficacy at that low dose. Such patients were considered eligible provided investigator considered the adverse event did not represent significant toxicity or the assessment of lack of efficacy was at a dose that would now be considered a target dose for treatment. Other reasons for discontinuation included personal reasons such as wishing to become pregnant, not wanting regular blood tests and travel. Meta-analysis of etanercept trials -Etanercept v conventional DMARD The inclusion of the Codreanu study (protocol 881A-309) in this section is questionable. In Codreanu patients were all receiving SSZ at baseline but had active disease. Patients were randomised to one of 3 groups. Two groups were treated with etanercept, in one of which SSZ was discontinued and in the other it was continued. The third group received placebo to etanercept and SSZ was continued. This study was therefore of a step-up design. In contrast ERA and TEMPO were parallel start studies of etanercept and MTX Meta-analysis of etanercept trials -Etanercept v placebo Codreanu is correctly included in this section TNF inhibitors v DMARDs Direction of effect Accepted and text to be revised. Thank you for this additional information. Unfortunately this was not available in the original report and finding such a population today would be difficult. Only one of the 3 groups can be regarded as step-up. One of the remaining two groups would be regarded as saw-tooth strategy i.e. switching agents in the face of disease activity. However we accept that this study does not represent a direct comparison of a TNF inhibitor with a DMARD and have already emphasised in the text this trial should not be regarded as a head-to-head comparison. - Notwithstanding accepted dogma until a clear relationship has been shown between radiographic change and other clinical outcomes in longitudinal
Section 3.3.3 p.142 1 st Section 4.1.5 p.160 2nd P.163 1st but the clinical relevance is not clear. If this significant, and the case of etanercept + MTX (TEMPO) substantial, effect on radiographic progression is not clear, then it would be also be reasonable to conclude that the clinical relevance of such differences in studies of any DMARD v placebo or a second DMARD are also unclear. It is accepted dogma that slowing of radiographic progression has important long-term benefits on prevention of disability. Combination of TNF inhibitor + MTX versus MTX All three TNF inhibitors, when combined with MTX, showed a trend for increased serious adverse events but again this was not statistically significant. The 1-year TEMPO data shows 10 serous infections in each treatment arm and in year 1 plus year 2 the numbers were ETN; 14 (6%), MTX; 15 (7%) and ETN + MTX; 13 (6%). For serious adverse events excluding infections, the corresponding figures were 35 (16%), 35 (15%), and 40 (17%) respectively. It is incorrect therefore to state that with etanercept there was a trend for increased serious adverse events when combined with MTX. Health economics- Abbott submission This refers to infliximab not adalimumab The Wyeth submission carefully presented the general argument that treatment of patients with severe active disease should be with the most effective available options. Evidence from studies with DMARDs have clearly established that use of combination therapy early in the course of RA has long term benefits on both progression of structural damage and disability, even when the combination treatment has been discontinued some time before the end of follow-up. Two such examples are the COBRA and FinRACo studies. It is perfectly reasonable and appropriate therefore to propose that the combination of etanercept and MTX be the studies, accepted dogma remains just that. In the clinical study report of TEMPO, the numbers for patients with serious infections and for those with serious adverse events excluding infections were listed separately, as shown in Wyeth s comment here. However, the number of patients who experienced at least one serious adverse event is not simply the addition of the two as some patients would have experienced both serious infections and other serious adverse events that were not infection. The data that we have used in our meta-analysis were extracted from a listing of all patients with serious adverse events (both infection and non-infection) in the clinical study report. We believe the data and our analysis, and hence the statement, are correct. They are correct this belongs in section 4.1.7. We are concerned with current practice not with perceptions of what is regarded as most optimal in severe, active disease. It is by no means universally accepted that DMARD combinations are better, indeed much of benefit seen in COBRA, for example, is attributed to the early use of corticosteroids.
option modelled for patients with severe, active and progressive disease. p.164 1st p.164 1st p.164 2nd In addition Wyeth do not accept that this weights the results in favour of etanercept as the 1 st drug in the comparator sequence is MTX, and therefore the model compares the incremental benefit of the combination over that of MTX, at least in the situation where etanercept is considered 1 st line. However for the purpose of switching thresholds a relationship between HAQ and DAS28 was required and changes in HAQ score were used as a proxy for changes in the DAS28. Perhaps here it would have been more appropriate to use actual switching rates from clinical observation rather than this conversion which potentially introduces more uncertainty into the model. Wyeth are not aware of any source currently available, and therefore took a reasonable assumption based on consultation with our clinical experts. Using a fixed HAQ at start of treatment has limitations and the heterogeneity of response is not taken into account Wyeth does not believe that this would have a large influence on the outcome of the cost effectiveness analysis. If this is deemed to have a very large effect our model can be modified to take this into account. The HAQ change for unspecified DMARD was based on the TICORA trial. This is inappropriate since data for individual drugs is available. Specific data were used for drugs specified in the model. The specified drugs were those commonly used in the UK. After these particular DMARDs have been used there is no common pattern of use. Hence it was decided to use non-specific data at this late stage in the model. The standard treatment arm of TICORA represents an average response as patients were treated with a range of treatments at the discretion of the We disagree. The BSRBR provides data on actual switching rates for etanercept. It is unclear why these were not used. We would be interested to see the results of such an analysis. Thank you for this clarification.
p.164 2nd rheumatologist. However, the definition of remission is problematic, and the change in HAQ may have been sufficient to represent remission without assuming further treatment benefits in modelling. It is not explained why the definition of remission is problematic. The clinical remission state has been defined using the criteria developed for DAS that have been widely accepted. The definition of remission is subject to a great deal of debate and we do not accept that there is a widely accepted gold standard. Additional analyses of the data from the TEMPO study have provided further evidence that the combination of etanercept and MTX has important effects on the relationships between disease activity, disability and joint damage. These analyses support the additional benefits that the remission state has on long term outcome. Thank you. p.165 1st p.165 Table 26 This assumption would be unnecessary if actual data on switching were used and the probability of switching may actually be much higher than a third Wyeth would welcome the use of the BSRBR to provide further information on switching. Data on IFN and ADL are included in the Table and yet there is no previous reference in the text to Wyeth including these drugs in the model. See above. We do not understand this comment
p.166 2nd p.166 2nd Section 4.2 p.173 Last sentence Section 4.2.2 p.185 Table 36 Inflating the already increased mortality on the basis of HAQ appears to introduce double-counting and is therefore inappropriate There is a clear relationship between HAQ and mortality, which is why mortality was adjusted to take this into account. In any event mortality effects are limited. In the BRAM model, risk of death is related to current HAQ score, age and sex. This was further adjusted to consider SAEs with a loss of 0.05 for each SAE experienced, but this assumption for a six-month period for someone experiencing a SAE appears to be an underestimate. Wyeth would like further clarification of why the TAR group viewed this as an underestimate. Clinical expert opinion sort by Wyeth deemed this a reasonable assumption. Economic analysis in this report Reflecting observed data, while on any treatment, a patient s condition is assumed to decline slowly over time; this is modelled as periodic increases of 0.125 in HAQ score. This reflects data available for traditional DMARDs and is the same as used in the Wyeth model. However data that now extends to 8 years clearly demonstrate that for etanercept there is no increase in HAQ over time. Data from TEMPO over a 3 year period show that this is also the case for the ETN+MTX combination. The BRAM does not therefore reflect the observed data for etanercept. Data used in the BRAM The distribution of HAQ scores comes from the NOAR cohort. This is a population based inception cohort of polyarticular inflammatory arthritis. The data used is derived from the population described by Wiles et al. The patients were from a group of 761 patients from which included In Wyeth s submission, it appears that the baseline mortality ratio of 1.63 has been applied to patients with HAQ 0, and higher ratios for higher HAQ scores. This would be double counting because the ratio 1.63 is taken from a source where it is applied to all RA patients regardless of HAQ score. Our opinion differs and we do not feel that expert opinion is an acceptable source for such data. We feel that selection bias in such observational cohorts necessarily limits the validity of the assumption the HAQ remains stable over 8 years. Population based studies indicate that even those without RA experience a decline in HAQ with ageing, unsurprisingly. We have commented above on our initial distribution of HAQ scores.
patients with IP, RA or PsA (ie not just RA). All patients however did attend hospital within 12 months of registration. It is important to note that only 48% of the patients were treated with DMARDs or steroids over the 5 year follow-up. Therefore over half the patients were considered to have disease that did not require aggressive treatment. It should be noted that patients with higher HAQ scores were more likely to be treated with DMARDs (Table 4) The HAQ distribution shows that 45.2% of patients had scores <1 and 65.2% <1.5. It is unlikely that patients with scores <1 would be considered for biologics. The model would be greatly improved if only data on patients who received DMARDs or steroids was used. In addition decisions on selection of a biologic should only be made if a HAQ threshold, say >1.5, was reached. This would more closely reflect clinical decision making. Patients on the BSR Biologics Register who were treated with etanercept had a baseline mean HAQ score of 2.1 (SD 0.5). Section 4.2.2 p.185 Table 36 Data used in the BRAM Given the HAQ distribution it is not surprising that the ICERs are higher than those obtained in the Wyeth model as patients with mild disease with limited room for utility gain would be treated with expensive drugs. This would be particularly true 1 st line. Section 4.2.2 p.195 It can be concluded that BRAM did not model the licensed indication for RA patients prior to MTX use, that is patients with severe, active and progressive disease. In addition it should be noted that a standard HAQ progression rate of 0.033 pa has been used but in the NOAR data set, the untreated group did not change over 5 years, and those treated within 6 months had a decline in median HAQ from 1.25 to 0.875. Data used in the BRAM -HAQ changes on treatment Data on etanercept that now extends to 8 years show that in the long term there is no increase in HAQ, unlike treatment with traditional DMARDs This is the same assumption as we have always made. We have been given no reason to change it. See above.
Section 7.1 p.219 last bullet Section 7.1 p.220 TEMPO bullet Section 7.1 p.221 Principal Findings Etanercept alone (at licensed dose) was as effective or slightly more effective than MTX in controlling RA symptoms and retarding joint damage in patients who were naïve to or who had no treatment failure with MTX in the ERA 119 and TEMPO 123 trials. The joint damage data for Etanercept is more than slightly more effective, the 2 year change for Etanercept being 1.3 compared with 3.2 for MTX (p = 0.001) Principal Findings It is commented that it is not clear that the results are generalisable to the UK. This was a multinational, multicentre study involving 19 countries and 92 centres. Three UK centres enrolled 19 patients. This is typical of large studies in RA. TNF inhibitors versus placebo Whilst evidence from Codreanu and ADORE show no benefit of adding etanercept to current DMARDs in patients with active disease, evidence from both the BSRBR and Swedish registries have shown improved outcomes in such circumstances compared with patients receiving Etanercept alone. We stand by our belief that clinically this represents a questionable difference notwithstanding a p value of 0.001. For instance, in the COBRA study the smallest detectable difference, which assesses measurement error for radiographic scores, was 8.7 Sharp units (95% agreement threshold). We refer Wyeth to Landewe and colleagues thoughtful commentary debates the importance of measured changes. 1 We are not questioning the generalisability of TEMPO because of it multinational nature. Our statement refers to the fact that half of the patients in the trials were reported as having previously received methotrexate without toxicity or lack of efficacy and yet these patients had not been treated with methotrexate for at least 6 months prior to the study. Noted. 1 Landewe RBM, Boers M, van der Heijde D. How to interpret radiographic progression in randomised clinical trials. Rheumatology 2003;42:2-5.