Eisai Corporation of North America and Pfizer Inc. Decide not to Appeal NICE Decision and Call for an Expedited Review of Guidance for Alzheimer's Disease
Published: Jul 01, 2009
LONDON, July 1 /PRNewswire/ -- On June 11, the National Institute for Health and Clinical Excellence (NICE) announced that, following consultation with stakeholders on the economic model that underpinned NICE's 2006 guidance on the use of drugs to treat Alzheimer's disease, the guidance remains unchanged despite recognising significant errors within their model.
While the process afforded Consultees the right to appeal against this decision, Eisai Limited, the licence holder of Aricept® (donepezil hydrochloride) and Pfizer Limited, its co-promotion partner, today confirm that they do not intend to pursue an appeal in the interests of patients.
Nick Burgin, Managing Director of Eisai Ltd., commented: "While we shared the disappointment of many on hearing NICE's decision, we have decided today not appeal. The fundamental reason for this is that on June 11th, Andrew Dillon, NICE Chief Executive, committed to commence a review of the existing guidance 'as soon as possible', as long as no appeals were received. It is therefore right that Eisai and Pfizer do all we can to allow NICE to honour its stated commitment and to take into account new data that have become available as well as advances in economic modelling techniques."
Eisai and Pfizer are committed to working with NICE and call upon the Institute to provide a timeline for review as a matter of urgency.
As part of the review process, Eisai commissioned United BioSource Corporation (UBC) to conduct an independent assessment of the NICE model. The UBC team was lead by Professor Jaime Caro, who developed the original 'AHEAD' economic model, which was modified by NICE for use in this appraisal. The UBC team's main findings were:
- The model incorrectly assumes that the likelihood of a patient dying in any year is the same despite evidence that it varies according to factors such as patient's age or the severity of Alzheimer's' disease (Xie 2008). Correcting this error would lead to lower (better) cost- effectiveness estimates than those that were reported by NICE, especially for patients with milder disease.
- The model incorrectly assumes that the annual cost of caring for patients prior to entering full time care is the same for each patient despite evidence that these costs depend on the extent to which the disease has progressed. For example, patients with very mild cognitive deficits are known to be less expensive to care for than those with more severe cognitive deficits (Knapp 2007). Correcting this error would lead to lower (better) cost-effectiveness estimates than those that were reported by NICE, especially for patients with milder disease.
- The model incorrectly assumes that apart from cognitive level, the baseline characteristics are the same for patients with mild disease and with moderate disease: they have the same age, behavioural symptoms and so on, despite evidence that these aspects also relate to disease severity (Knapp 2007, Chatfield 2007, Piccininni 2005, Steinberg 2008). By not taking into account the real difference in baseline characteristics between patients with mild and moderate disease, the model generates results that make treatment of mild patients look less attractive (less cost-effective) than treating patients with moderate disease.
- The model incorrectly assumes that the mean duration of Alzheimer's disease prior to treatment is one year in all analyses, even though studies have shown that from onset of symptoms to diagnosis can approach 3 years in the UK (Bond 2005). The consequence of this assumption is that the model estimates that relatively few patients will show signs of deterioration, including in patients who are not treated. As a result the model makes treatment appear to be less clinically and cost-effective than if it focuses on when treatment is actually started in the UK (between three and seven years from onset of disease) when the rate of deterioration is greater.
- The model incorporates treatment effect estimates derived from six-month clinical trials and ignores information available from randomised, placebo-controlled trials of greater duration. By effectively limiting treatment benefit to a maximum of six months, the model makes treatment appear much less clinically and cost-effective.
- The model assigns a single utility for patients who are not yet in full time care, thus ignoring the slower deterioration with treatment. Rather than addressing this error, the model incorporates a "fix" by adding a quality of life ("augmented") benefit but no justification is made for the significant number of arbitrary assumptions that were made. Despite the significant uncertainty in these assumptions, none of them was varied in the sensitivity analyses.