Background
There are currently four licensed treatments that provide symptomatic relief for patients with Alzheimer’s disease in England—three acetylcholinesterase (AChE) inhibitors (donepezil, rivastigmine, galantamine) and one
N-methyl-
d-aspartate (NMDA) receptor antagonist (memantine). These drugs are collectively referred to as
drugs for dementia in the British National Formulary, despite their licensing for Alzheimer’s disease only [
1]. Since the first of these drugs became available in 1997, there have been several changes in national guidelines for the treatment of Alzheimer’s disease, as well as several initiatives to encourage better diagnosis and treatment of the disease. Despite this, there has been little research into whether such changes to guidelines and initiatives have directly influenced clinical practice [
2,
3]. We examined how prescription rates in England have changed since the launch of these drugs up to 1st January 2016, using data from the U.K. Clinical Practice Research Datalink (CPRD). We investigated how prescribing was affected by changes in National Institute for Health and Care Excellence (NICE) guidance (including the 2006 guidance that was subject to legal challenges), the addition of dementia to the Quality and Outcomes Framework (QOF), the introduction of ambitious government dementia strategies, and the expiry of drug patents. The timing of each of these changes, which may have influenced aspects of drug prescribing and clinical practice, is discussed further below and summarized in Table
1.
Table 1
Events prior to 1st January 2016 that potentially affected prescription rates
May 1997 | Donepezil first recorded in CPRD |
September 1998 | Rivastigmine first recorded in CPRD |
January 2001 | Galantamine first recorded in CPRD and first NICE guidance released |
December 2002 | Memantine first recorded in CPRD |
November 2006 | NICE recommended restricting drug access |
September 2007 | QOF revised to include dementia |
February 2009 | First National Dementia Strategy launched |
March 2011 | NICE removed recommendation restricting drug access |
January 2012 | Galantamine patent expired |
February 2012 | Donepezil patent expired |
May 2012 | Prime Minister’s Dementia Challenge launched |
July 2012 | Rivastigmine patent expired |
April 2014 | Memantine patent expired |
February 2015 | Prime Minister’s Challenge on Dementia 2020 launched |
NICE guidance on the prescribing of drugs for dementia
In the past NICE guidance has used scores from the Mini Mental State Examination (MMSE), in combination with other measures, to guide whether a patient should be prescribed a drug for dementia. The test, proposed in 1975 by Folstein et al., assesses a patient’s cognition out of a total possible score of 30, where normal cognition is considered as a score of 24 or more [
4]. The original NICE guidance, issued in 2001, on the use of drugs to treat Alzheimer’s disease recommended that the three AChE inhibitors should be used for all patients scoring 12 or above on the MMSE until the drugs were deemed no longer effective [
5,
6]. In November 2006, NICE revised their guidance so that the use of AChE inhibitors was restricted to patients with moderate Alzheimer’s disease; this was defined as patients scoring between 10 and 20 points on the MMSE. The 2006 guidance was also the first to consider the use of the NMDA receptor antagonist memantine, which was recommended for use only in clinical trials for patients with moderate to severe disease [
7]. This revision of the guidance was controversial because of the way in which it assessed cost-effectiveness, which was expected to restrict access to these drugs, and was ultimately the subject of a high court challenge by the Alzheimer’s Society and two drug manufacturers, Eisai and Pfizer [
8‐
10]. This led to a further revision being made to the NICE guidance at the end of March 2011, which recommended AChE inhibitors for patients with mild to moderate Alzheimer’s disease and memantine for patients with moderate to severe Alzheimer’s disease or who could not tolerate AChE inhibitors [
11]. For the duration of our present study, treatment had to be initiated by a specialist and deemed effective as long as there has been ‘an improvement or no deterioration in MMSE score, together with evidence of global improvement on the basis of behavioral and/or functional assessment’ [
6].
Inclusion of dementia on the QOF
QOF is a voluntary incentive program, introduced in 2004, to improve services in primary care [
12]. Dementia first appeared in QOF as an ‘indicator’ in September 2007 [
13]. There are currently three indicators for dementia included in the framework. The first requires that the practice establish and maintain a register of patients diagnosed with dementia, and the other two indicators refer to the ongoing management of the disease [
14]. The inclusion of dementia on the QOF could therefore have encouraged a greater focus on the diagnosis and pharmacological management of the disease in participating practices.
Government dementia strategies
The first National Dementia Strategy was launched by the Department of Health in February 2009. The aim of that strategy was ‘to ensure that significant improvements are made to dementia services across three key areas: improved awareness, earlier diagnosis and intervention, and a higher quality of care’ [
15]. This strategy was followed in 2012 by the Prime Minister’s Dementia Challenge, which looked to improve care and research by 2015, and more recently by the Prime Minister’s Challenge on Dementia 2020 [
16,
17]. The most recent strategy aims to build on the work of its predecessors to make England the best place for both dementia care and research. In general such strategies may help to increase the awareness of dementia for both the public and health services [
18,
19].
Drug patents
The King’s Fund charity found that the prescription of generic drugs over their patented alternatives has ‘saved the NHS around £7.1 billion and allowed more than 490 million more items to be prescribed to patients’ between 1976 and 2013 [
20]. AChE inhibitors for the treatment of Alzheimer’s disease became available generically from 2012, whereas NMDA receptor antagonists became available generically from 2014 (Table
2) [
21]. Therefore, in recent years the cost of drugs for dementia has decreased significantly from previous years. This serves as a potential factor in rates of prescribing, particularly in publicly funded health care services such as the NHS in England.
Table 2
Patent information for the drugs used for dementia [
21]
Donepezil | Aricept (Eisai / Pfizer) | AChE inhibitor | January 2012 |
Rivastigmine | Exelon (Novartis) | AChE inhibitor | February 2012 |
Galantamine | Reminyl (Shire) | AChE inhibitor | July 2012 |
Memantine | Ebixa (Lundbeck) | NMDA receptor antagonist | April 2014 |
Discussion
The first trend change for the proportion of patients with probable Alzheimer’s disease receiving their first prescription for an AChE inhibitor occurred in October 2012 (95% CI, September 2011 to April 2013,
p = 0.816). At this time, a long-term steadily increasing trend became a very strong increasing trend. This surge could be related to two factors. Firstly, the patents expired on the three drugs in this class in 2012—galantamine in January 2012, donepezil in February 2012 and rivastigmine in July 2012. Secondly, the Prime Minister’s Dementia Challenge launched in May 2012. It is likely that the reduction in cost of these drugs, which resulted from their patents expiring, in combination with increased awareness of dementia due to the Prime Minister’s Dementia Challenge, led to this substantial change in prescription rates we observed. In addition to these factors, a large amount of literature concerning AChE inhibitors had been published ahead of the revisions to the NICE guidance in 2011. Although this is unlikely to have caused the sharp surge that we observed, it could have contributed to the long-term steadily increasing trend observed prior to this change. A systematic review which covers the literature through November 2014 (i.e., after all join points identified in our analysis but 13 months before the end of our study) summarizes the literature available at that time [
33]. It shows that several studies published between 2003 and 2008 suggested that patients with mild to moderate Alzheimer’s disease could benefit from AChE inhibitors with estimated ‘improvements on the order of 1.5 MMSE (30-point scale)’. We therefore cannot rule out a potential effect of the literature on prescribing, even though the authors of the review questioned whether such an improvement was clinically meaningful when all the evidence was presented together. Further to the support from the literature, the Google Trends data for news searches in England also suggested increased awareness around the time of this trend change. The interest for the search term ‘Alzheimer’s disease’ was at its maximum in September 2012 (based on the data available from January 2008 to December 2015, inclusive), which could indicate interest among the public.
The second trend change in the AChE inhibitor analysis occurred in May 2013 (95% CI, November 2012 to April 2014,
p = 0.789), less than 1 year after the initial change for this drug class and with overlapping 95% CIs. This change signals the end of the surge in prescribing and the start of a decreasing trend in prescriptions. This is not unexpected, because patent expiry may have led to a form of ‘catch-up prescribing’ whereby people who were previously denied access to the drug were granted access at this time owing to its newly reduced cost. This would result in the apparent decreasing trend once ‘catch-up prescribing’ was complete, which is suggested by the trend analysis but is not as clear when considering the raw data points. These results differ from the sensitivity analyses that considered relaxed diagnosis definitions, though the ‘any Alzheimer’s disease’ and ‘any dementia’ analyses were in line with each other. This suggests that prescribing for patients with probable Alzheimer’s disease was more consistent, as one might expect, across the study than for other groups. This could indicate that patients with dementias other than probable Alzheimer’s disease (i.e., with unlicensed indications) were receiving these drugs and that their prescriptions were subject to change over the period studied. Further to this, large increases in prescriptions are observed as the diagnosis definition is relaxed. This could provide further evidence for the possible unlicensed use of this drug class. The literature at that time also reflects ongoing discussion concerning the benefit of these drugs for indications other than Alzheimer’s disease. For example, a 2012 review by Rodda and Carter discusses their use in vascular dementia, dementia with Lewy bodies and Parkinson’s disease dementia [
34]. Alternatively, it could be attributed to the fluctuating course of symptoms that some people with dementia experience or increased recognition of mixed diagnoses where there is evidence of Alzheimer’s disease in addition to other forms of dementia, both of which might lead to treatment changes.
The trend changes in the NMDA receptor antagonist analysis occurred in March 2011 (95% CI, August 2010 to April 2011,
p = 0.891) and June 2011 (95% CI, April 2011 to November 2011,
p = 0.896). Notably, the 95% CI for the first trend change ends in April 2011, which is when the 95% CI for the second trend change begins. This suggests that the trend changes may be related. The first of these trend changes marks the start of a strong increasing trend that changes to a steadily increasing trend following the second trend change. In March 2011 NICE introduced guidelines that recommended the prescription of memantine for patients with moderate to severe Alzheimer’s disease or for those people who could not tolerate AChE inhibitors. This replaced existing guidelines that restricted access to memantine to patients participating in clinical trials. It would therefore seem that these trend changes relate to the transition between the existing guidelines and those introduced in March 2011. In addition, we observed the second highest peak in interest (88% of maximum interest) for the disease term ‘Alzheimer’s disease’ in the Google Trends data for news searches in England in January 2011. In this month, the ‘Final Appraisal Determination on Donepezil, galantamine, rivastigmine and memantine for the treatment of Alzheimer’s disease’ was released. NICE defined this document as ‘the appraisal committee’s final draft guidance about using a treatment or group of treatments in the NHS’, which becomes guidance if not appealed [
35]. The increase in news searches around this time, and its alignment with the release of the final draft guidance, supports the idea of a transition in prescribing practice due to the NICE guidance. Finally, the evidence concerning the use of memantine is summarized in a technology appraisal conducted by NICE in 2011 to support their guidance [
11]. We cannot disentangle the role that this information from several studies published prior to the trend change might have played in changes to prescribing.
Interestingly, neither of the trend changes in the NMDA receptor antagonist analysis aligns with those observed for the AChE inhibitors. This suggests that the NICE guidelines, which were implemented at the same time for both drug classes, may not have been as effective for AChE inhibitors. This is likely due to the fact that these drugs were available outside of clinical trials prior to the restrictive guidelines recommended in 2006. The sensitivity analyses conducted for the NMDA receptor antagonists were consistent with these results, regardless of the diagnosis definition used. The first of the joinpoints for all NMDA receptor antagonist analyses occurred in the 7-month period between August 2010 and March 2011, and the second occurred in the 6-month period between June 2011 and November 2011. This high level of consistency across diagnosis definitions indicates a clear pattern in prescribing, suggestive of a distinct change in practice. This provides additional support for our inferences concerning the impact of the 2011 NICE guidance on the NMDA receptor antagonist drug class.
Strengths and limitations
The key strength of this study is the large sample of primary care data with prescribing information, provided by the CPRD. The CPRD is ‘broadly representative of the UK general population’ and was generally comparable to the last census in 2011 for age, sex and ethnicity despite young people and smaller practices tending to be slightly underrepresented [
36,
37]. Our data extract contains 40,202 patients diagnosed with dementia in England up to 1st January 2016 (note that data are restricted to practices with a last data collection date in 2016), including 10,651 with probable Alzheimer’s disease and a further 12,167 with possible Alzheimer’s disease. A further strength of our study is the long follow-up of patients that allowed us to consider patients who did not receive immediate treatment. This is important because pharmacological interventions for Alzheimer’s disease have historically considered severity as part of the prescribing decision, so there is likely to be a treatment delay after initial diagnosis for those presenting with mild disease.
The main limitation of our study is the likelihood of missed diagnoses. This is demonstrated within our dataset, because there were 1231 patients receiving one of the treatments of interest who did not have any form of recorded dementia diagnosis. Missed diagnoses are likely to be due to (1) outdated or non-specific diagnoses (i.e., type of dementia is not updated once established), (2) diagnoses received outside of primary care (i.e., from a specialist service) and (3) unrecorded diagnoses in primary care (i.e., a diagnosis is given but not added to a record). Missed diagnoses have been explored in sensitivity analyses by testing the sensitivity and specificity of our diagnosis definitions (Additional file
2) and by relaxing the diagnosis definition from ‘probable Alzheimer’s disease’ to include other less certain codes for the disease and other types of dementia (Additional file
3). Neither of these sensitivity analyses provided any cause for concern. A final limitation of this study is the difficulty in determining the lag time between an event and a trend change to assess the impact of the event. To allow for this, we have focused on events that are considered to be of greatest impact—for example, changes at a national level—and so we expect any effect associated with them to be evident if present. However, this prevents us from covering all the factors that may influence prescriptions for drugs for dementia during the study period; for example, we cannot comment on all papers concerning these drugs published during this time.
Conclusions
Analysis of both drug classes indicates that inclusion of dementia in QOF had no effect on prescribing trends and the other factors had mixed effects. NICE guidance on the prescribing of drugs for dementia aligned with trend changes for NMDA receptor antagonists but not AChE inhibitors. The guidance that had the noticeable effect was released in March 2011 and allowed the NMDA receptor antagonist memantine to be used outside of clinical trials. All other guidance for both this drug and AChE inhibitors, including that which recommended restricting access, did not align with trend changes. Government dementia strategies also appear to have had mixed results, with the Prime Minister’s Dementia Challenge (launched May 2012) being the only strategy to align with a trend change. Although this strategy is likely to have increased awareness of dementia around the time of the October 2012 trend change for AChE inhibitors, we believe that the more likely cause of this change is the patent expiry of the drugs in this class. This will have reduced the cost of these drugs and potentially led to a surge in prescribing, such as that observed in our trend analysis. The events considered here highlight the many factors that may have influenced prescribing rates and the challenges in assessing the impact of a given event. Overall it would seem that the proportion of patients receiving prescriptions increased over the period studied, regardless of changing guidelines and other initiatives. Furthermore, given the increase in diagnoses of dementia and, more specifically, Alzheimer’s disease reported in the CPRD (Fig.
1), the absolute number of prescriptions has increased considerably over the period studied.
To our knowledge, there are two other studies that have considered prescribing trends, and these were focused mainly on the impact of the National Dementia Strategy [
2,
3]. Our study extends the findings of these previous studies because it considers trends since the launch of these drugs and implements a joinpoint model as a hypothesis-free approach for the factors affecting prescribing. We have observed that prescription rates in England do not always respond to factors such as regulatory guidance, recommendations or patent expiry, and when they do, not necessarily in a predictable way. This suggests that communication with clinicians may need to be improved to use drugs for dementia more cost-effectively. In addition to this, the present study provides insight into the factors that may have influenced prescription rates of drugs for dementia in England since their launch in 1997. This is essential for accurate assessment of the effectiveness of these treatments and to adjust for them in other forms of analyses, particularly as factors that may modify the rates of disease progression. This study may also help to inform the handling of regulatory guidance and recommendations concerning drugs for dementia in the future.
Acknowledgements
We acknowledge Dr. Elizabeth Coulthard (University of Bristol) for input during the initial stages of this study.