ALZHEIMER S DISEASE - MAIN ISSUES FROM ACD-2 CONSULTATION

Transcription

1 ALZHEIMER S DISEASE - MAIN ISSUES FROM ACD-2 CONSULTATION - The responses in the table below are from Consultees and Commentators to the appraisal. - The table is currently sorted on subject matter as follows: o Acetylcholinesterase inhibitors Impact of the guidance Evidence base Quality of life QALY Model structure Clinical effectiveness evidence Population responders Population - subgroup severity Carer benefits Carer costs Decision & neuroleptics Costs Cost-effectiveness subgroup severity MMSE & assessment Choice of AChE inhibitor Clinical effectiveness AD2000 Broad policy framework Model mortality o Memantine Impact of the guidance Clinical effectiveness Population - behavioural disturbances Cost-effectiveness behavioural subgroup Cost-effectiveness responders Carers National Institute for Health and Clinical Excellence Page 1 of 97

2 Costs Quality of life QALY Model MMSE Broad policy framework - The comments from Health Care Professionals that were received in writing or on the web are summarized in a separate document. - AG = Assessment Group / AR = Assessment Report / ACD = Appraisal Consultation Document / FAD = Final Appraisal Determination ACETYLCHOLINESTERASE INHIBITORS Issue Specific question Comment made by Impact of the guidance If approved, the decision would put professionals in the impossible position of withholding beneficial treatments because the patient is not ill enough. We view this as unethical. The current draft guidance goes against what is known to be best practice in dementia care (for example, guidance from the British Association of Psychopharmacology due to be published in the next few months) and would be extremely difficult to implement in clinical practice. Withdrawing anticholinesterase drugs from those people in the mild stages of Alzheimer s disease would also be very difficult guidance for clinicians to implement. Withholding a potentially effective treatment from a person who could benefit, especially when this contradicts best practice guidelines, would create a serious ethical dilemma for most doctors. This is clearly demonstrated by a recent survey of ART AS See FAD Noted. Noted. National Institute for Health and Clinical Excellence Page 2 of 97

3 consultants by the Royal College of Psychiatrists, which indicated that two-thirds of specialist Old Age Psychiatrists felt unable to withhold a trial of cholinesterase inhibitor treatment from someone with Alzheimer s disease who met the licensing requirements for these treatments. The Alzheimer's Society firmly believes that withdrawing access to the anticholinesterase drug treatments for people with an MMSE of 21 and above is the wrong decision and is not the most effective way to use the treatments. The evidence clearly demonstrates that, as used in clinical practice, the drugs are clinically and cost-effective in the mild and moderate stages of Alzheimer s. The overall outcome of the Committee s deliberations appears both perverse and parsimonious. Perverse in that it suggests not initiating treatment until sufferers have deteriorated to a stage when the quality of their actual life experience (not some dumb cost effect measure) has deteriorated to a stage when prolonging it seems cruel rather kind and then once they deteriorate further to a stage of real suffering to withdraw help again. It is unbelievably detached from the reality of suffering and symptomatic treatment. However our biggest concern regarding the restriction of ACHEIs to moderate AD is that it places clinicians in an, at best, unenviable position and, at worse, an unethical one. We will be faced with patients in the early stages of the decision who possess capacity to understand their disease and its consequences and we will be unable to prescribe medication that has been shown to be efficacious. Instead we will have to DCT IoP Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and Noted. National Institute for Health and Clinical Excellence Page 3 of 97

4 convey to them and their families that we are forced by the NICE decision to wait until they deteriorate, on essentially cost-benefit grounds. We would not be surprised to see such a decision challenged. I would ask all members of the committee to consider if they had dementia and could have treatment that would benefit them cognitively and effectively stabilise their condition for between one and two years, would they want that at the mild stage of dementia when they still had insight, significant retention of activities of daily living and reasonable social functioning, or would they rather wait until they had deteriorated to a point where they were becoming behaviourally disturbed, were losing insight and at a point where they were already needing substantial hands on care and supervision? The committee should also be mindful of the good trial evidence which shows that delaying treatment means that patients never catch up to the benefit gained by those whose treatment is started early. To deny mild patients such benefits clearly defies any logic as to how medicine should be practised in the 21st century. There is also an issue of lack of equity between those people who have already been prescribed these drugs and those in very similar or identical situations in the future who, if these recommendations are adopted, will have no access to these drugs until their cognitive abilities have become highly compromised. # 1.4. The wording of this # is going to lead to a lot of distress to carers and annoyance. If BOTH carer AND clinician concur for either continuance or cessation - no problem. But, if the Clinical specialist 2 MHF Patient expert See FAD See guidance 1.2 No change - see guidance 1.2 National Institute for Health and Clinical Excellence Page 4 of 97

5 clinician wishes to cease and carer wishes to continue then I foresee a lot of unwarranted stress for carers. I would propose the last phrase be reworded to indicate that the carer is the final arbiter. I am also concerned that this will lead to a two-tier healthcare system within the UK. Those who are able to afford these drugs will purchase them privately, in conjunction with their General Practitioner and hospital specialist, as happened before the drugs were originally supported by NICE. The poorest in society will be deprived of these treatments, that have been shown to benefit not only the person with the dementia, but also those caring for them, who are often frail and elderly, and trying to cope with other chronic conditions associated with ageing. We believe this piece of work has been extremely thorough and painstaking, and we do not wish to make any specific comment on the methodologies used. We believe the conclusions should be widely adopted. We would only make three general comments and one specific comment. In the wider context of treating patients with Alzheimer s disease, we would stress and ask the committee to consider how ethically difficult it will be for clinicians to deny patients they see with mild Alzheimer s disease effective treatment. If mild AD patients are informed following initial diagnosis of this chronic debilitating disease that they should await treatment until their disease deteriorates to moderate, a management vacuum will be created during which the patient and caregiver will be unsure when this degree of severity is likely to be reached and treatment may commence. Clinical specialist 3 RCPhysicians RCPsychiatrists SHIRE Prescribing AChEinhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and Noted. Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and National Institute for Health and Clinical Excellence Page 5 of 97

6 The current proposed guidance which requires that patients have reached moderate severity is also likely to have a major impact on use of specialist services for assessment (with resultant additional costs) with patients asking for repeat assessments leading to a very likely significant impact on NHS waiting lists for neurology, geriatrics and old age psychiatry services. Delayed initiation of treatment is associated with a loss of benefit for patients and their carers. For example analyses of galantamine trial in the US found that patients treated continuously with galantamine 24mg for twelve months had a better cognitive profile compared to patients initiated on placebo for 6 months followed by treatment with galantamine for 6 months(12). It is likely that in many cases undue anxiety will develop during this period, with possible over-use of NHS services - including alternative symptomatic AD medication while AChEI medication is denied Evidence base As stated in previous submissions (response to the first ACD dated 22 March 2005 at Appendix 1 [Section 1.0]). We believe that the exclusion of most new trial data on the basis that the studies were not randomised or placebo controlled is inappropriate. As indicated previously, use of placebos are unethical following the demonstration of benefit in these patients. The exclusion of these studies unnecessarily limits the data available for consideration and underestimates the true effectiveness of donepezil. Has ALL relevant evidence been taken into account? In my Eisai Patient expert See FAD See response table to comments on ACD (published January 2006) See FAD to 8 and National Institute for Health and Clinical Excellence Page 6 of 97

7 view NO. The sections on carers quality of life and NHS costs have been poorly addressed. By example, I offered my detailed Day Diary for detailed consideration by the Committee; this offer was NOT taken up. Alzheimer Scotland is impressed with the number of studies that have been taken into account by NICE. Our concern is that, in its interpretation of the evidence, NICE appears neither to have understood nor taken into account fully the complexity of Alzheimer s disease and how it affects people with the illness. For example, in the discussion about the cost effectiveness of memantine (ACD p 23, p 46) it is recognised that memantine versus placebo showed statistically significant advantages on a number of outcomes but this was not the case on all scales and the absolute magnitudes of differences on all outcomes were modest. For a complex illness which manifests itself in different ways for different patients it is not surprising that some of the measures of change have failed to capture the benefits of treatment. It is therefore wrong to underplay the value of the scales which recorded statistically significant benefits. This is of particular importance because MMSE scores are neither meaningful nor reliable at this stage of the illness.1 NICE state that they have considered evidence from a wide variety of sources. However, Schneider (Int J Geriatric Psychiatry, Jan 2006) provides a controversial critique identifying problems with the systematic review underpinning the NICE decision which cannot be matched here but which NHS-QIS NHS-QIS See FAD See inclusion criteria Assessment Report. 1 Health Technology Assessment 2005; Vol 9; No 10. National Institute for Health and Clinical Excellence Page 7 of 97

8 Quality of life QALY needs to be echoed in this document. There is a strong likelihood that many of the testimonials received from users and carers will be from those who have shown substantial benefit from being treated while their MMSE is above, or even well above, the value of 20 which NICE select as their arbitrary cut-off point for future treatment. One would have to assume that these testimonials have been ignored. This would seem inappropriate. In addition, the current document appears on the one hand to acknowledge the significant problems of selection bias associated with post hoc subgroup analyses, yet they rely on this very type of analysis in their cost effectiveness model which they use to justify their conclusions. Indeed, the number of prospective randomised controlled trials of cholinesterase inhibitors with a baseline MMSE towards the midpoint of their suggested range for future treatment is extremely small by comparison to the overall number of RCTs with this group of drugs. Therefore, the derivation of their models on which cost effectiveness calculations are undertaken are likely to be prone to bias and flawed. The current scores used in the appraisal are based upon Neumann et al in which carer proxies responded on behalf of people with dementia using the Health Utilities Index 2 (HUI2).i, iias the authors point out, this is a scale which has not been validated for use in Alzheimer s disease and certainly not for proxy use. Results should therefore be used with extreme caution. In the response to the first ACD, we also noted that four of the six subscales in the index do not change between mild and severe dementia, so it is AS See FAD See response table to comments on ACD (published January 2006) National Institute for Health and Clinical Excellence Page 8 of 97

9 completely counter-intuitive to base a model for change between mild and moderate dementia on these parameters. Evidence suggests that self-care, but not cognition, is associated with quality of life. iii Therefore, it would appear valid to use only scores from the self-care subscale. This results in scores of 0.88 for mild dementia and 0.14 for severe dementia rather than 0.60 and The fact is that in this condition the greatest quality of life gains will occur by preserving time in the mild stages, even though illness costs are at their least, rather than prolonging time in the moderate and severe stages and this guidance will do just the opposite. At a number of points in the ACD consulted on in 2005, the Committee had acknowledged that there were some concerns around the use of the cost per QALY approach in this appraisal. This does not appear to feature in the latest appraisal document. To what extent have the Committee s previous concerns now been addressed? The economic analyses do not reflect the real costeffectiveness of AChEI treatment of patients with mild AD. We refer again to our fundamental concern that reliance of QALYs is inappropriate in this therapeutic area because of the nature of the disease process and the difficulty measuring quality of life in those patients. The utility scores are based on cognitive endpoints only. The disutility associated with side effects is not incorporated, although this may affect overall utility gain with the various treatments, and hence influence cost-effectiveness ratios. This is not addressed in the model, yet differences in Clinical specialist (see also RCP) DH/WAG Eisai Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and See response table to comments on ACD (published January 2006) See response table to comments on ACD (published January 2006) National Institute for Health and Clinical Excellence Page 9 of 97

10 adverse events by product are noted in Section Factors that are taken into account in the descriptions of study data should be reflected in the model. Section : We maintain that the utility estimates and hence cost-effectiveness may be underestimated. As described in our response to Technical Report no. 2, the application of these health state utilities has recently been questioned by Neumann (2005). Recognising the conservative nature of the utilities, Neumann acknowledges the limitation of implementing the utilities in the manner employed in the economic model, and suggests that the difference in utilities between dependent and nondependent AD patients may be underestimated. o An underestimation of the value would affect costeffectiveness estimates, and hence bias subsequent conclusions. While it is the accepted practice to use QALY's as a measure, this disadvantages this group as the vast majority are elderly people and therefore have less life years ahead of them. This brings into question the accuracy of the cost effectiveness. We continue to feel that the Appraisal Committee is using as a base for its decisions a cost effectiveness model and QALY measure that is both inappropriate and totally inconsistent with the previous (2001) Appraisal. The central issue of using utility scores obtained with an instrument designed for use in children, and never validated for proxy use for people with dementia, undermines the whole basis of the NICE model which the Appraisal Committee uses for its decisions. RCN RCPsychiatrists See response table to comments on ACD (published January 2006) See response table to comments on ACD (published January 2006) National Institute for Health and Clinical Excellence Page 10 of 97

11 Appraisal Committee is basing its decisions on a costeffectiveness model and QALY measure that is flawed and inappropriate for people with AD and their families (and does not take into account the full social, personal and healthrelated costs). We have previously referred to the article by Rawlins and Culyer in the BMJ (2004, 239: 224-6) that describes circumstances where NICE s preferred measure of the cost/qaly is inappropriate. This includes where appropriate data on quality of life are not available and this is undoubtedly still the situation for AD. It was accepted as such by the Appraisal Committee in 2001 and should be accepted again now. If not, the Appraisal Committee should explicitly state the basis for these changes particularly as the data on cost-effectiveness are based on models that use data prior to the 2001 report. Model - structure We do not believe that NICE has given appropriate weight to the views of people with dementia and their carers about the benefits of the drugs. This is especially the case when the model of clinical and cost effectiveness used is regarded by ourselves and most dementia research experts to be inaccurate and unreliable in capturing the complex nature of dementia. Nice have responded to the points raised by the Alzheimer s Society and other consultees about the reliability of the model. However, we do not believe many of these responses adequately address the flaws raised. Nice have dismissed some key points with inadequate explanation. We feel strongly that the model remains flawed and any resulting estimates of cost-effectiveness are unreliable and inaccurate. RICE ART AS See response table to comments on ACD (published January 2006) See response table to comments on ACD (published January 2006) See response table to comments on ACD (published January 2006) National Institute for Health and Clinical Excellence Page 11 of 97

12 Limitations concerning the modelling structure include the simplistic binary health states, the use of only 1 parameter to assess treatment efficacy, and the relatively short modelling period. Limitations concerning the sources of data include the risk prediction equation, the health state utilities, the mortality rate and the sources of cost. Many of these are derived from studies published years ago using patient data from the United States. The relevance of these data to the treatment of, and patterns of utilisation of healthcare services by, patients in England and Wales in 2005 is open to question. Furthermore it appears inconsistent to reject published economic evaluations for donepezil on the basis that these relate to treatment outside the UK and then to rely on a model using data from the United States. In some cases it is clear that there is a lack of good quality up-to-date UK data, which in turn identifies the need for more research. However in other cases, for example in costs where there are recent data, these have not been used in the model. The inappropriateness of some of these input values is compounded because the model is very sensitive to small changes in cost and utility estimates. Sensitivity analyses around these variables cause large changes to the ICER and therefore place the robustness of the model in doubt. As described in our response to Technical Report no. 2, use of the placebo group as the comparator in the model is conservative. Patients who receive no treatment are likely to deteriorate faster than placebo-treated patients. Whatever one s opinion about the use of the QALY measure in Alzheimer s disease, the committee has to acknowledge Eisai Clinical specialist 2 See response table to comments on ACD (published January 2006) and FAD See Technical Report 2 and see FAD and See response table to comments on ACD National Institute for Health and Clinical Excellence Page 12 of 97

13 that the assumptions behind the economic model are extremely questionable and the resulting uncertainties huge. The health utility scores were derived from an unvalidated health utility index which was designed for use in children, yet administered to proxy carers of patients with dementia. The economic model fails to take into account many significant benefits of the drugs, for example reduced carer time spent supervising patients with dementia. They have to realise that to use this as the sole basis for decision making in this case remains a fundamental flaw. It can be argued that cost-benefit analysis is going to be stacked against older people and that a similar analysis relating to younger people would be less likely to lead to the conclusion reached by NICE. This ACD is flawed on the concept of delay in time in admission to Full Time Care [FTC]. Admission to F/T Care may [as was the case of Beryl] NOT the mental state of Beryl, but, my inability to care for Beryl especially at nights, and due to my declining health, for reasons other than caring. If this criterion is to be applied to AS patients then it is not logical to deduce that ALL patients who cannot be cared for by a [sole] carer cannot have other [expensive] drugs, e.g. cancer, cardiac drugs, or expensive technologies, e.g. stents? Moving on to the cost effectiveness assessments, in earlier submissions Alzheimer Scotland has expressed its major reservations about the way in which NICE has used QALYs to assess these treatments. Although the Assessment Group noted that the use of cognitive function alone to model MHF Patient expert NHS-QIS (published January 2006) See response table to comments on ACD (published January 2006) See response table to comments on ACD (published January 2006) See response table to comments on ACD (published January 2006) National Institute for Health and Clinical Excellence Page 13 of 97

14 disease progression is likely to misrepresent disease progression it still plays a prominent part in the AHEAD modelling of the progression of Alzheimer s disease. Hence, because there is not a simple correlation between cognitive function score and quality of life, the figures for the cost of QALYs gained are neither reliable nor valid measures to use in this instance. The key issue with the original AHEAD model framework (retained by NICE) is that it assumes no additional treatment benefit beyond the original short term trial module. After the initial treatment effect patients are assumed to follow a natural untreated course of decline in AD from the cognitive level obtained after 6 months of treatment. This assumption is overly restrictive to mild patients where data from four separate open label extension studies from galantamine show that the treatment effect between treated and expected untreated decline opens up further over time for mild patients (7;8). Figure 1 illustrates the effect of the assumption compared to data for actual patients followed long-term. In addition the likely impact of clinical trial and placebo effects on reducing likely untreated decline are also shown. Analyses of the 6 month clinical trials for galantamine are consistent with this observation with untreated decline being greater in more advanced patients(4;5). Data from the Nordic donepezil 12 month placebo controlled study also appear consistent (9). SHIRE See response table to comments on ACD (published January 2006) and see FAD National Institute for Health and Clinical Excellence Page 14 of 97

15 Clinical effectiveness - evidence Importantly, research suggests that people who begin drug treatment at a later stage never catch up with those who began earlier, strongly indicating that earlier treatment leads to an improved long term prognosis. Open label continuation trials support this conclusion. For example, Farlow et al reported that patients who received placebo for the first 26 weeks and were then commenced on rivastigmine (Exelon) for weeks had a less favourable outcome than people prescribed rivastigmine from the beginning of the trial (1.4 difference on the ADAS-cog).iv An open label extension study of donepezil (Aricept) supports this conclusion. Patients who received placebo in the original 26 week randomized controlled trial (RCT), and were then commenced on donepezil declined by points on the ADAS-cog over 12 AS See FAD National Institute for Health and Clinical Excellence Page 15 of 97

16 months, compared to a point decline in the group that had received donepezil from the beginning of the RCT.v Similar findings were also evident in an open label extension study of galantamine (Reminyl), where the group who had received 24 mg/day of the active treatment from the beginning of the RCT were more likely to maintain ADAS-cog scores near their baseline values than people only prescribed galantamine in the open label extension phase.vi Trying to make judgements about the management of AD whilst ignoring the data on amnestic minimal cognitive impairment (MCI), seen by many as the very early, or prodromal, stages of AD, is to be doing the job with one hand behind ones back. Two large scale well designed and run studies of donepezil in MCI have shown that there is very little advantage in treating these patients. (Salloway et al Neurology.2004 Peterson et al NEJM 2005) It is true that this is not the case for clearly diagnosed very mild AD where the drug is effective but this gives a clear guide as to when to start treatment when cost minimisation and rational use of resources is concerned. Clearly patients who have cognitive impairment and no dementia will not benefit and so one should wait to see if dementia develops before starting. If you wish you can advise that in most cases dementia in AD, meaning a cognitive impairment affecting work or social functioning, will not be apparent, in those of average premorbid IQ until MMSE <26 has been reached. MMSE 20 is far too low for the reasons outlined above. But dementia is a clinical diagnosis and is not determined by an MMSE score. The clinical meaningfulness and cumulative benefits of AChEI Clinical specialist (see also RCP) Eisai Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and Prescribing AChE-inhibitors National Institute for Health and Clinical Excellence Page 16 of 97

17 therapy in mild AD have not been adequately recognised. AChEI treatment of patients with mild AD has been shown to be effective (Wolfson, 2002; Birks and Harvey, 2006 [Cochrane review]). Donepezil efficacy in a group composed wholly of patients with mild AD has been demonstrated (Seltzer, 2004). The benefit of AChEI therapy is cumulative, ie, early and sustained treatment maintains patients at higher levels of function for longer. Treatment with donepezil delays functional decline compared with placebo (Mohs, 2001), and delays time to nursing home placement (Lopez, 2002; Geldmacher, 2003). Delaying start of treatment by as little as 15 weeks can mean patients might not have the opportunity to attain maximal benefits because of more advanced disease (Doody, 2001). To investigate whether patients with mild AD derive long-term cumulative benefits from AChEI therapy, we conducted an exploratory analysis using data disclosed to NICE (Nordic study) comparing the change in cognition from baseline to Months 6, 9, and 12 between donepezil- and placebo-treated patients with mild AD (see Appendix). This analysis suggests that patients with mild AD treated with donepezil for more than 6 months do gain cumulative benefits from early and continued therapy. Since these benefits are not taken into account in the SHTAC/AHEAD model, the long-term clinical impact of treating patients with mild AD is therefore being underestimated. The ACD states that the evidence suggests that the NHS-QIS for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and Prescribing AChE-inhibitors National Institute for Health and Clinical Excellence Page 17 of 97

18 cholinesterase inhibitor drug treatments are beneficial in treating Alzheimer s disease (see pp 14, 17, 21). However, it does so without enthusiasm and anticipates the limited scope of its recommendations. Alzheimer Scotland believes that NICE has been unduly negative in interpreting the available evidence. It is well known that these treatments do not benefit all patients with mild to moderately severe Alzheimer s disease but the evidence base does provide a convincing case for a consistent response rate and statistically significant treatment effects. This should have been summarised in a more balanced way. In the discussions of adverse side effects of the cholinesterase inhibitors (ACD, pp 11, 16, 19) it is reported that higher numbers of participants in the higher dose group withdrew because of adverse events (donepezil), this number increased with higher doses of galantamine and the percentage of participants reporting adverse events, namely nausea and vomiting, resulting from treatment with rivastigmine was particularly high in those treated at a higher dose. The impression is deliberately being created that the side effects of these treatments make them unsuitable for use. In clinical practice these problems are minimised and made manageable by introducing treatment at lower dose rates. This would have been mentioned in a more balanced interpretation of the evidence. Clinical effectiveness of galantamine vs. placebo has been demonstrated for mild patients (MMSE 21-24) across multiple efficacy domains using validated instruments - see galantamine submissions of June 2004 and October 2005 SHIRE for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and See FAD 3.5 Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost National Institute for Health and Clinical Excellence Page 18 of 97

19 (4;5) -. o Significant benefits were seen for cognition (ADAS-cog change vs. baseline galantamine 2.0 points vs. 0.3 points for placebo; overall treatment difference 2.4 points, observed cases [OC] analyses, p<0.001); o The actual cognitive improvement (2.0 points OC analyses) in ADAS-cog score vs. baseline after 5/6 months is greatest for mild AD patients treated with galantamine compared to moderate patients where the improvement vs. baseline is lower (1.1 points OC analyses). o Activities of daily living results including treatment difference on ADCS-ADL, (1.7 points OC analyses p=0.02): pooled analyses all trials using ADCS-ADL and DAD showed a standardised mean difference for galantamine vs. placebo of 0.23 (observed cases analyses, p<0.01); o Clinical global impression showed a greater proportion of patients improving with fewer patients declining when treated with galantamine vs. placebo (CIBICplus, observed cases analyses p<0.001) in the mild AD patient group. Placebo effects in mild AD patient observed in double-blind clinical trials are not replicable in actual clinical practice. The proposed guidance which excludes treatment of mild AD is based on an very small untreated decline observed on placebo in the trials which reflects both clinical trial (improvement of other aspects of health influencing cognition & learning effects) and placebo effects on patients cognitive effective use of NHS resources. See FAD and National Institute for Health and Clinical Excellence Page 19 of 97

20 Population - responders performance observed over 5-6 months. Patients who remain transparently untreated at home are likely to display a much greater decline than their placebo-treated counterparts. Consequently the positive impact of treating mild patients in the NHS in England and Wales is underestimated. The results of the responder analysis demonstrated that the cholinesterase inhibitor drugs are both clinically effective and cost effective across the full range of mild to moderate dementia. We are therefore very disappointed that the appraisal committee decided to disregard the results of this analysis. The ACD states that the committee had been advised that the analysis was subject to bias and therefore any results were unreliable. This is a very vague explanation of a decision with serious implications and more detail must be provided to justify the decision, especially as any other form of analysis grossly over-estimates the drug costs over the treatment period. It is disappointing that it chooses to ignore the subgroup responder analysis which it sought despite the compelling evidence of additional cost benefit. Indeed it is difficult not to think that all the additional evidence has been sought for appearances sake rather than actually considered in detail by every Committee member the mere volume of the data would make that an unlikely achievement. Section : The relationship between the moderate subgroup CQGs and the recommendation to treat responders within the moderate subgroup is not clear. The minimum value for the CQG ranges for the individual drugs and the meta-analyses appear to be taken from the AS DCT Eisai See FAD and See FAD The guidance in the FAD does not specify a provision for stopping the use of AChE inhibitors based on a specific responder definition National Institute for Health and Clinical Excellence Page 20 of 97

21 Population subgroup severity Scenario D analysis (Table 13, Technical Report no. 2; Tables A.5 and A.6, Addendum to Technical Report no. 2). This scenario models the moderate subgroup regardless of response. However, the ACD recommendations are to treat the responders within the moderate subgroup of patients. This would reflect Scenario E, which has a lower CQG for donepezil ( 22,000) than that from Scenario D (although the same definition for response is not used in the ACD). The differences in the CQGs between the total group responder analysis and the moderate subgroup responder analysis are small (Scenario A/C vs Scenario E, Table 13, Technical Report no. 2). Section states, Finally the third method focused on the extra effect of responders on the drug over and above responders on placebo. Novartis would submit that this is not a valid comparison as it has no relevance to clinical practice. Stabilisation and a delay in progression in the early stages gives people with dementia longer in the period when symptoms are mild. At this stage people are likely to be relatively independent, to be better able to plan together with their family for the future, and to enjoy hobbies and interaction with others. They usually experience a very good quality of life with the right support, care and treatment. People with dementia and their carers talk about the value of this borrowed time when their relationships are as similar to how they were before the diagnosis as they will ever be. The natural progression of Alzheimer s means that a drug treatment that slows progression during the early stages will Novartis for people with moderate Alzheimer s disease as was included in the 2001 guidance. Noted. AS See FAD See FAD National Institute for Health and Clinical Excellence Page 21 of 97

22 show a smaller effect size using an assessment scale such as ADAS-cog. This is because progression of symptoms is slower in the mild stages than in the moderate stages. Therefore a drug treatment that delays progression for six months (for example) in the early stages will have a smaller magnitude of change than a drug that delays progression for the same amount of time in the moderate stages. This will clearly impact on the subgroup analysis and should be taken into consideration by the committee. You will now be aware that the metrics of the scales make it statistically inevitable that there will be greater changes in the scale scores in the more severe patients but there is no evidence that there is less benefit in the more mild patients. A MMSE change from brings a patient back into the normal range but a change from whilst correlating to an improvement in the patient does not. It is sadly an unscientific appraisal of the data and fails to take into account the vast amount of data beyond the cognitive scores, and the personal experience from those involved daily with the management of AD - leaving some of the most needy patients in our purview denied the widely recognised standard of care unless they can afford to access it privately. The clinical significance of AChEI therapy in the population with mild AD has not been adequately recognised, and the cumulative benefit of therapy is not properly appreciated. The committee s recommendation is to treat patients with moderate, but not mild, AD, despite formal analyses demonstrating that the magnitude of patient response was not affected by disease severity at treatment initiation. In Clinical specialist (see also RCP) Eisai See FAD Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and National Institute for Health and Clinical Excellence Page 22 of 97

23 addition, we disagree with the interpretation that clinical evidence for the beneficial effect of donepezil therapy on behaviour is unclear, and are concerned that the flawed AD2000 study has again been relied upon too much in the body of evidence. No significant interaction of baseline severity and magnitude of effect has been shown. The ACD states (Section ): subgroup analyses based on severity of cognitive impairment suggest some differential advantage for more severely cognitivelyimpaired subgroups. Caution should be exercised when interpreting data on treatment effect by disease severity, as sample sizes were relatively small (mild, n=287; moderate, n=111; moderate to severe, n=55), and only 2 studies were used in the analyses (see Table 6, p20, MRC report, November 2005). The relatively smaller effect sizes observed in the mild AD than the moderate AD treatment groups are largely driven by less decline in the comparator placebo groups, rather than by differences in the donepezil treatment responses from baseline. It should also be borne in mind that the ADAS-cog may not be uniformly sensitive to cognitive change over the course of AD, such that it is possibly less sensitive to change in patients with mild AD compared to moderate AD, and scores can be affected by the ceiling effect of the scale (Stern, 1994). Subgroup estimates have only been calculated for moderate AD. Larger numbers were included in the metaanalysis for all three AChE inhibitors. See Technical Report no. table 4 page 11. See FAD See FAD Prescribing AChE-inhibitors for people with mild National Institute for Health and Clinical Excellence Page 23 of 97

24 As such, CQG estimates can only be compared between the moderate subgroup and the total patient group. It is therefore not known to what degree cost-effectiveness may overlap between patients with mild and moderate AD. When analysing the cost-effectiveness of treating the moderate subgroup, Scenarios D and E modelled treatment of only the cohort of patients with AD and moderate cognitive impairment at baseline with an AChE-inhibitor (Section 2.1, Technical Report no. 2). As a result, the potentially large number of patients likely to transition to moderate AD during the time course of the model, and the different rates of transition that will occur if patients with mild AD are treated or not treated with AChEIs, has not been taken into account. The apparently greater magnitude of response for cognition in moderately impaired patients is because the scales used (the MMSE and the ADAS-Cog) are not linear and show both floor and ceiling effects that limit the interpretation of this observation. It is inappropriate to restrict therapy to more severely impaired people and illogical to wait for patients to deteriorate. The drugs buy people time in the course of a relentless and progressive condition and to wait until a person with AD deteriorates before offering therapy is unkind especially when the drugs clearly have benefit in a significant number of people with mild AD. The decision on restricting ACHEIs to moderate AD does not make sense from a perspective informed by the mode of action of these drugs. The more advanced the condition the more cholinergic function will have been lost and so this class GDG IoP Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and The economic model follows a cohort of people with moderate Alzheimer s disease over time and does not model the prognosis of all people with moderate Alzheimer s disease that are prescribed an AChE inhibitor. See also the Assessment Report for a description of the AHEAD model. See response table to comments on ACD (published January 2006) Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost National Institute for Health and Clinical Excellence Page 24 of 97

25 of compound will have greater biological effects earlier in the condition when there is a greater capacity for the system to respond. If the drugs appear in clinical trials to show more benefits in advanced AD then this is inevitably an artefact of clinical trials data. Most obviously the slope of deterioration is steeper in moderate AD than in mild AD and so it becomes easier to show an effect. A small effect in moderate dementia is more visible on trials data than a large effect in mild AD. If the measures used in clinical trials accurately and comprehensively captured end-points important to patients then differential biological events in mild vs. moderate AD would be of little or no interest. However this is not the case - the wealth of evidence presented to NICE by the Alzheimer s Society and others attest to this. Our understanding of the biology of the condition and the mode of action of the drugs suggests it is likely that the effects on the brain will be greater in mild AD and our knowledge of clinical trials methodology in this area suggests that measuring outcome in mild AD will be more challenging than in moderate AD. Therefore we think that NICE should be extremely cautious about overinterpreting sub-analyses of data for mild vs. moderate dementia. The focus on moderate dementia, because of a greater cognitive response in that group, is purely artefactual because of the non linearity of the cognitive assessment instrument used. This has been clearly demonstrated statistically and, as objective evidence of this, if one compares global assessments of outcome such as the CIBIC there is no difference in response between mild, moderate Clinical specialist 2 effective use of NHS resources. See FAD and See FAD National Institute for Health and Clinical Excellence Page 25 of 97

26 and severe disease. To exclude mild patients from benefits because of an artefact of an assessment tool would be quite wrong. It is NOT reasonable to EXCLUDE people with Alzheimer s who have an MMSE of 20 or above. Such persons can benefit from the AChE-I drugs considerably. During the period of MMSE ~ 25 to 20 the strain on their carers can be very considerable and in all likelihood would cost the NHS > 1,000 just to maintain carers well being. However, the reduction of MMSE from 12 to 10 is very welcome. Perhaps the greatest example of NICE not understanding the significance of the information it has analysed is the use of the analyses of the MRC Biostatistics Unit on sub groups by severity of cognitive impairment. These show (ACD p 13, p 17, p 21) that the magnitude of response on ADAS cog scores is less for people with mild Alzheimer s disease than for those with moderate and moderately severe Alzheimer s disease. NICE uses this finding to justify its exclusion of people with mild Alzheimer s disease from the use of the three cholinesterase inhibitor treatments. Yet, it is this group of people with Alzheimer s disease who, if they respond to the treatment, would be most likely to return to near pre-illness levels of ability. Although the magnitude of change may be less, its personal significance will be much greater. It is a perverse interpretation of this finding to delay starting the treatment until the time when the person may have a greater magnitude of response but not be able to return to their preillness levels of ability. Eventually (p 44) the ACD reaches the central point of Patient expert NHS-QIS Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and Prescribing AChE-inhibitors for people with mild Alzheimer s disease does not constitute a cost effective use of NHS resources. See FAD and National Institute for Health and Clinical Excellence Page 26 of 97

27 Alzheimer Scotland s and other patient representative groups earlier submissions to NICE some people with Alzheimer s disease benefit considerably more from AChE inhibitors than others. However, instead of considering the data which can be obtained retrospectively and then analysed, it rejects this, without any substantiating evidence, because retrospective responder analyses could plausibly lead to significant selection bias and related uncertainty in the interpretation (p 43). This takes NICE back to its preoccupation of wanting to define, prospectively, sub-groups of people with Alzheimer s disease who might benefit more than average. Alzheimer Scotland does not believe this to be a reasonable interpretation of the evidence. It is perverse that, having requested a series of retrospective analyses, the Appraisal Committee now deems these same analyses to be insufficiently robust to form the basis of any recommendations, due to their retrospective nature. The basis of this discrimination between mild and moderate AD is the suggestion that data from clinical trials indicates there is a more significant improvement in cognition in moderate dementia. This assessment takes no regard of the Novartis Clinical specialist 3 The Institute has requested further analyses, not the Appraisal Committee (see spx?o= & spx?o=267493). In separate clarification meetings it was made clear to the manufacturers that these were requests by the Institute and that the Appraisal Committee would appraise the evidence independently of the request. See FAD National Institute for Health and Clinical Excellence Page 27 of 97

28 non-linearity of the ADAS-Cog and MMSE scales, which show an S-shaped curve, with the rate of decline in early stages being represented by a less rapidly declining slope. This means that a small numerical improvement may represent considerable benefit in a patient s daily life. The apparently greater response on cognition in moderately impaired patients in comparison with mildly impaired patients can be explained because neither the MMSE nor ADAS-Cog are linear scales. Both scales show a lower rate of deterioration in mildly impaired patients and severely impaired patients in comparison with moderately impaired patients where the rate of change per annum is the greatest. Carer benefits It states on page 40 of the ACD2 that although at any point in time a carer may have a higher utility if the person they cared for was benefiting from drug treatment the effect of the drug would be to delay progression of the condition, in which case the carer would still be faced at some time in the future with the same difficulties caused by disease progression. This statement brings into question the committee s understanding of why carers feel so strongly about the drug treatments. Carers are all too aware that the drugs are not a cure and Alzheimer s is a progressive disease. However, it is the delay in symptom progression that carers value and that allows them to enjoy more time with the person they care for before symptoms are more advanced. The Society acknowledges that data are not available, but as the guide to the methods of technology appraisal points out, patient evidence can identify the limitations in the published research literature. We believe this is the case with carer quality of life and, based on reports RICE See FAD AS See response table to comments on ACD (published January 2006) & FAD , 4.3.9, National Institute for Health and Clinical Excellence Page 28 of 97

29 of carers, greater weighting must be given to improvements in carer quality of life. Section : We acknowledge the lack of quantitative evidence on carers utility gain. However, as we described in the Eisai and Pfizer response to Technical Report no. 2, the utility benefit of 0.01 for carers has not been justified and seems low. No explanation has been provided for the attribution of a utility benefit of 0.01 to carers and it is therefore difficult for consultees to assess whether this figure is appropriate. We ask NICE to provide reasons for its use of this figure. The lack of quantitive evidence about benefits to people with dementia and carers despite there being much qualitative evidence has perhaps weighed against making these drugs available to people earlier in the dementia pathway. The reduction of the start MMSE score to 20, thus eliminating mild AS sufferers is not acceptable. The earlier the start the better the quality-of-life for BOTH sufferer AND carer. The NHS costs in keeping the carer fit have not been adequately considered. Carer costs The guide to the methods of technology appraisal states If the inclusion of a wider set of costs or outcomes is expected to influence the results significantly, such analyses should be presented in addition to the reference case analysis. (para ) The Alzheimer's Society believes reduction in carer time is an important outcome of the drugs that would influence the results of the cost-effectiveness analysis considerably and therefore should be included. Impact for NHS appropriate NO. As above the costs to the Eisai MHF Patient expert AS Patient expert See Technical Report no.1 See response table to comments on ACD (published January 2006) & FAD , 4.3.9, See response table to comments on ACD (published January 2006) & FAD , 4.3.9, See response table to comments on ACD (published January 2006) & FAD See response table to National Institute for Health and Clinical Excellence Page 29 of 97