Background
Dementia is estimated to affect 36 million people worldwide, costing more than US $600 billion a year [
1]. The ageing demographics of both developed and developing nations mean that its prevalence will double over the next 20 years [
2]. Alzheimer’s disease dementia (AD-dementia) is the most common form, estimated to account for 60 to 80% of cases [
3]. An additional 5-42% of older adults [
4] meet criteria for mild cognitive impairment (MCI), a lesser degree of impairment than dementia, which allows them to continue to function independently [
5].
The current model of Alzheimer’s disease is of a pathological process which starts in mid-life but remains undetected until it causes dementia, at which time a clinical diagnosis of AD-dementia can be made. Recent revisions to diagnostic criteria define three stages of Alzheimer’s disease, namely Alzheimer’s disease dementia [
6], MCI due to AD (MCI-AD) [
7], and preclinical AD [
8]. Broadly speaking, the MCI and dementia stages of AD can be determined by a combination of clinical assessment and biomarkers, while detection of preclinical AD, characterised by the absence of overt symptoms, requires brain-related assays of disease progression [
9‐
11]. Within the MCI population a significant proportion have symptoms due to underlying AD pathology (MCI-AD [
7]). Episodic memory impairment is typically a common cognitive marker in those with MCI who progress to AD-dementia, and each year around 15% of individuals with amnestic (memory-impaired) MCI are diagnosed with AD-dementia [
5,
9].
Technology now exists that can enable the earlier detection of cognitive loss associated with incipient dementia, offering the potential for earlier intervention in the UK health care system ([
12,
13]). Timely detection and intervention in AD-dementia can be cost-effective because even though current treatments have limited efficacy, they nonetheless improve symptoms enough to reduce healthcare costs and keep patients living in the community for longer. Although the debate over cost-effectiveness of current treatment is controversial, some economic studies have estimated that timely treatment is cost-effective [
14]. A UK study based on 2007 costs estimated that over ten years, timely detection and treatment produced savings of £3600 (US $5508) in direct costs and an additional £4150 ($6350) in indirect costs (caregiver time) per patient [
15]. Similar cost-effectiveness has been established for other healthcare systems, including the US [
16]. Despite this, in the UK as in other countries, the majority of people with AD-dementia do not have a diagnosis, with up to three quarters undiagnosed in some regions [
17‐
19].
Improved public awareness of AD-dementia is driving older adults who experience memory problems to seek help at an increasingly early stage. At presentation, many of these individuals may meet criteria for MCI-AD for which, as yet, there are no licensed treatments [
20]. There is considerable debate within the clinical and public health communities as to the stage at which it is most appropriate to assess and treat people who may have, or be worried about, the early signs of AD-dementia. There is generally little economic evidence about pre-dementia stages of AD (MCI-AD) [
21], but it is likely that resource use increases considerably in the years preceding an AD diagnosis [
22].
One key factor determining when detection and diagnosis programmes should be targeted is understanding at what stage in disease the available interventions are most clinically effective. Unfortunately, there is little data on the size of clinical effect at different stages of disease for even the most common treatments such as cholinesterase inhibitors. It is generally thought that these interventions are more effective when used before widespread pathological change has occurred [
23‐
25], but there is little evidence that they are effective in pre-dementia stages such as MCI-AD [
20,
26].
A second major factor is economic: the extent to which healthcare costs and quality of life are changed by intervening at different points in the disease course. In this study we aimed to assess the economic effects of intervening in AD-dementia up to 9 years earlier than that which currently occurs. In the absence of detailed data comparing the efficacy of treatments at different stages, we concentrated only on the economic impact of applying (equally effective) interventions at different points in the disease course. Because few treatments are currently available, we modelled the effects of two hypothetical interventions: one a modestly-effective symptomatic treatment, the other a hypothetical disease-modifying treatment (DMT) that halted cognitive decline for a short period. These two scenarios were chosen as simple and relatively conservative examples of the type of intervention that might plausibly achieve widespread adoption within a healthcare system such as the UK NHS or US Medicare. In this way we aimed to quantify the extent to which the potential costs and savings from an intervention for AD-dementia would be impacted by the time of intervention.
Discussion
In this study we utilised the known natural history of cognitive decline in Alzheimer’s disease to estimate the health economic impact of earlier intervention. We modelled two hypothetical scenarios; one a symptomatic drug with effects similar to those seen for cholinesterase inhibitors, the other a disease-modifying intervention that delayed cognitive deterioration for a period of time. In both cases, results showed that optimally timing an intervention could significantly affect its economic benefits. For the symptomatic treatment, the economic benefits were greater the earlier the intervention was applied; in contrast the disease-modifying intervention was most cost-effective when applied relatively late in the course.
These differences in optimal timing between symptomatic and disease-modifying models can be explained by the effects of mortality on post-diagnostic healthcare costs and by the changing rate of cognitive decline throughout the disease course. For the symptomatic intervention, treatment benefits are immediate and decrease over a fixed time period. Intervention is consequently most effective when applied at the earliest possible timeframe, so that the greatest number of individuals benefit from the intervention before any loss to mortality. However the effect of this intervention is relatively small because it is inherently time-limited. In contrast, in the disease-modifying model, while there is no immediate impact, the effect of intervention is to permanently shift the curve of cognitive decline. As a result, the effect on both healthcare costs and QALYs is considerably larger. Under this scenario, intervention was most economically beneficial when applied around two years prior to the time of standard diagnosis. This is the period during which MMSE decline is most rapid: intervening either earlier or later than this reduces the costs saved by delaying cognitive decline.
Other health economic models of disease modifying treatments have also estimated higher costs with DMT over time, due to additional costs of DMT and the increased survival with DMT [
38].
Of the two interventions, the results from the symptomatic model are more immediately applicable because moderately effective symptomatic treatments are already available in the UK for patients with mild to severe AD -dementia (MMSE of 26 or below). In this cohort, individuals already had a mean MMSE score indicating AD-dementia treatment up-to nine years before diagnosis, so these results are in one sense supporting not ‘early’, but rather ‘on time’ intervention. While we have not included treatment costs in the model, generic AD-dementia drugs now cost just pennies per day. These results therefore support previous analyses demonstrating the cost-effectiveness of intervention and treatment using symptomatic treatments in the NHS, and extend this by arguing that intervention will be more cost-effective when applied earlier than currently occurs for the majority of patients.
The disease-modifying results should not be interpreted to mean that we can wait until frank dementia before intervening. On the contrary, to maximise the cost-effectiveness of a disease-modifying strategy, the benefits of such a treatment must be being felt before patients hit the period of rapid decline that precedes a diagnosis in this cohort. In order to prevent this precipitous decline it will, at a minimum, be necessary to detect and closely monitor individuals with milder cognitive impairment so that intervention can be certain to occur prior to that point of inflection. In all likelihood, the mode of action of such an intervention, be it a lifestyle prevention strategy or anti-amyloid immunisation, would probably need intervention to start many years or decades earlier in order to have an impact on cognitive course.
Our model included full direct health and social care costs and the cost of the diagnostic procedure but excluded any costs associated with informal caregiver time [
39,
40], and, importantly, the cost of the intervention itself. This might range from a few pounds per month for a generic drug such as a cholinesterase inhibitor to many thousands for a future biological therapy. Alternatively, modification of disease course and its impact, could potentially be obtained from non-pharmacological strategies such as lifestyle intervention which might be required decades earlier (i.e. as a primary prevention strategy) in order to prevent the disease process from occurring [
41]. Although excluding the cost of informal care narrows the viewpoint somewhat, excluding these costs takes a conservative approach. Including caregiver costs would cause the results to be more favourable towards early detection. A sensitivity analysis demonstrated that the inclusion of a cost associated with the intervention has a linear impact on the model’s findings. That is, if the cost of the intervention is £50, then the total cost savings associated with early detection will be reduced by £50. If the cost of the intervention is £200, the cost savings are reduced by £200.
We did not include in our models a specific cost of treatment, so these results provide a framework for estimating the likely bearable costs to the UK health system for a sufficiently effective anti-dementia intervention. The results show that cost-effectiveness might be achieved with either a cheap symptomatic treatment or a more expensive disease-modifying one. The treatment effects modelled here were purposefully modest and reflect the relatively small effects achieved thus far in both approved and developing compounds. While both the healthcare costs and QALY values used here are taken from the UK system, the principles are likely to apply widely within European and North American countries, where under-diagnosis is common [
17‐
19] and where any intervention which improved the function of AD-dementia patients would be expected to have a proportionally similar effect on healthcare costs [
42].
An important limitation of the model is the assumption that the efficacy of the intervention does not depend on when it is applied, either with respect to illness severity, or to cohort age. At present little data exists addressing this for existing treatments, although for future anti-amyloid therapies intervention is likely to be more effective when applied earlier in disease course. Age at treatment start is particularly important because at the older ages, mortality plays an increasing role in exaggerating the economic differences between early and standard intervention. In the absence of published data, we ignored the effect of treatment on survival, whilst in reality disease-modifying interventions, in particular, would be expected to play a role in reducing mortality. All of these limitations are conservative in nature and would lead, if anything, to an underestimation by the present model of the benefits of earlier intervention.
The model also does not take into account the possible effect of false positives. As diagnoses of AD-dementia shifts to earlier disease states the difference between pathological symptoms and normal age-related cognitive decline is likely to be subtle, with the potential for greater diagnostic uncertainty, leading to an increased risk of a false positive diagnosis (Albert, 2011). Due to a lack of data this was not included in the model; however the likely impact of including this would be that in earlier diagnosis there are more false positives, leading to treatment costs increasing, while no extra benefit is incurred. This should be considered in future studies and when translating our analyses.
Conclusions
In summary, we demonstrate that for both symptomatic and disease-modifying interventions, timing is crucial in determining the economic benefits of a treatment, even leaving aside the issue of treatment efficacy. Early intervention is clearly indicated for current symptomatic treatments, which are likely to be most cost-effective when applied as early as it is possible to diagnose AD-dementia. For a disease-modifying intervention, maximal cost-effectiveness would be achieved by intervening early enough to anticipate the point at which cognition begins a rapid decline. Taken together, these results suggest that cost-effective early detection and intervention should be an achievable goal in earlier stages of AD (MCI-AD), and that a range of different intervention effects and costs might be manageable or even beneficial in terms of overall healthcare costs. Public policy including screening, diagnostic and prescribing guidelines should aim to reflect the optimal cognitive stage for cost-effective intervention, which will be several years prior to current standard practice.
Competing interests
JB & AB are employees of, and shareholders in, Cambridge Cognition, a company which produces cognitive tests for use in research and healthcare. LL and MT report no potential conflicts.
Authors’ contributions
JB and AB conceived the idea for the study. LL and MT conceived and built the economic model. JB led, and all authors contributed to, the drafting of the manuscript. All authors read and approved the final manuscript.