1-19-2007 forecasting the global burden of alzheimer ' s disease

Background: The goal was to forecast the global burden of Alzheimer’s disease and evaluate the potential impact of interventions that delay disease onset or progression. Methods: A stochastic multi-state model was used in conjunction with U.N. worldwide population forecasts and data from epidemiological studies on risks of Alzheimer’s disease. Findings: In 2006 the worldwide prevalence of Alzheimer’s disease was 26.6 million. By 2050, prevalence will quadruple by which time 1 in 85 persons worldwide will be living with the disease. We estimate about 43% of prevalent cases need a high level of care equivalent to that of a nursing home. If interventions could delay both disease onset and progression by a modest 1 year, there would be nearly 9.2 million fewer cases of disease in 2050 with nearly all the decline attributable to decreases in persons needing high level of care. Interpretation: We face a looming global epidemic of Alzheimer’s disease as the world’s population ages. Modest advances in therapeutic and preventive strategies that lead to even small delays in Alzheimer’s onset and progression can significantly reduce the global burden of the disease. 2 http://biostats.bepress.com/jhubiostat/paper130 INTRODUCTION As the world population ages, enormous resources will be required to adequately care for persons afflicted with Alzheimer’s disease. Research is actively underway to develop interventions to both delay disease onset and slow progression of disease. Effective interventions may significantly reduce the prevalence and incidence of Alzheimer’s disease, improve the quality of life both of the patients and their caregivers, and reduce the resources needed to provide adequate institutional and home health care. Several treatments to help slow disease progression, and prevention strategies including lifestyle changes are being investigated (1). Uncertainty exists in the estimates of the global burden of Alzheimer’s disease and the potential impact of interventions. Recently, Alzheimer’s Disease International, an international consortium of Alzheimer’s associations, produced estimates of the worldwide prevalence of people with dementia (2). These estimates were based on a Delphi consensus study of 12 international experts who systematically reviewed published studies. The consensus method involved a qualitative assessment of evidence by each expert, and then those experts were given an opportunity to revise their estimates of prevalence after reflecting on the input of their colleagues. The resulting Delphi consensus estimates have been considered some of the best currently available estimates of worldwide prevalence. Yet, because the Delphi approach is not based on an underlying quantitative model, the Delphi study cannot be readily used to forecast the potential impact of new interventions on health care needs. Furthermore the study did not take into account the severity of disease. Disease severity is an important consideration for assessing the global burden of Alzheimer’s disease because the resources needed to care for patients with advanced disease are very different than for patients early in the disease process. The objective of this article is to forecast the global burden of Alzheimer’s disease based on a 3 Hosted by The Berkeley Electronic Press mathematical model that incorporates the aging of the world’s population. The model is used to forecast the world-wide prevalence of Alzheimer’s disease, evaluate the impact of interventions, and incorporate disease severity. METHODS The Multi-State Model Our methodology is based on a multi-state probabilistic model for the incidence and progression of Alzheimer’s’ disease. The method extends a single stage disease model used for U.S. projections (3) by including early and late stages of disease. According to the model, healthy persons have an annual probability of onset of Alzheimer’s disease which begins in an early stage and ultimately progresses to late stage disease. Persons with early stage disease have an annual probability of progressing to late stage disease. The definitions of early and late stage disease including the mean durations are discussed below. Persons are at risk of death during each state. The model is illustrated schematically in figure 1. The transition probabilities between states are the probabilities of moving from one state to the next. We allow some of these transition probabilities to depend not only on age but also calendar year to account both for birth cohort effects (e.g. death rates change over time) and the impact of new interventions that could potentially delay disease onset and progression. The model is implemented as a discrete time stochastic model in which transitions occur only at the beginning of a calendar year, and it is possible that persons may have multiple transitions in a year (e.g. disease onset followed by death could occur in the same year). We derived formulas for the age-specific prevalence rates of early stage and late stage disease in terms of the model in figure 1. The transition probabilities are inputs into these 4 http://biostats.bepress.com/jhubiostat/paper130 formulas. We performed a number of analyses and systematic reviews of published literature, to estimate the transition probabilities (described below.) Then, we forecast disease prevalence by multiplying the formulas for age–specific prevalence rates by demographic population projections. We used the United Nations worldwide population projections (4). Those projections are in terms of 5 year age groups which we interpolated to obtain projections by single year of age. We performed analyses separately by gender, and for each of six regions of the world. Then, we evaluated the potential effects of interventions that delay disease onset, delay disease progression or both by modifying the transition probabilities under different scenarios. We multiplied the transition probabilities by various factors (relative risks) to model the potential effects of the interventions. We translated these relative risks into average delays in disease onset and progression (in the absence of competing causes of death) as an alternative way to express the efficacy of intervention programs. We considered the impact of interventions that begin in the year 2010. The technical details including the formulas for the age specific prevalence rates and computing software are available from the authors at www.biostat.jhsph.edu/project/globalAD/index.htm. Transition Probabilities In this section, we discuss inputs for each of the transition probabilities of figure 1. Incidence rates We estimated age-specific probabilities of disease onset by performing a systematic review of published Alzheimer’s disease incidence rates. Jorm and colleagues (5) reviewed the worldwide literature on Alzheimer’s disease incidence rates. We updated the Jorm review to include additional recent studies reporting age-specific incidence rates of Alzheimer’s disease. 5 Hosted by The Berkeley Electronic Press We fit a linear regression equation to the log of the age-specific incidence rate for each of 27 studies in our review because incidence rates appeared to grow exponentially with age. We then averaged the rates from the fitted regression lines to obtain an equation for the age-specific incidence rate. We found that the annual age-specific incidence of Alzheimer’s disease at age t expressed in per cent per year (for t greater than 60) is given by: Incidence rate (% per year) = ( ) .121 60 .132e t− . (1) Equation 1 implies that incidence grows exponentially with a doubling time of about 5.7 years. We found no significant geographic differences in the doubling times of Alzheimer’s incidence (p=.3), suggesting that any geographic variation may be due to different criteria and thresholds for diagnosis. We used equation 1 for the incidence rates (rt,y in figure 1) in our analyses. We accounted for uncertainty in equation 1 by performing a sensitivity analysis that used a range based on the upper and lower 10 percentiles of the distribution of fitted incidence rates from all the studies. This range spanned from about half to double the incidence estimates from equation 1. For example, the predicted annual incidence at age 80 is 1.48 % per year with range of 0.67% to 3.41%. The ranges we cite in the results section account for this uncertainty in incidence rates. We also performed sensitivity analyses to the assumption that incidence continues to grow exponentially at the oldest ages by holding incidence rates constant after age 90. Disease Progression Alzheimer’s disease is a progressive disease and persons who have the disease longer often require a higher level of care. Considerable variability exists in the world’s literature on the rate of Alzheimer’s disease progression which results from differences in definitions of severe disease among studies, and heterogeneity in the disease course among patients. The Consortium to Establish a Registry for Alzheimer’s disease suggested that 6 years is the mean 6 http://biostats.bepress.com/jhubiostat/paper130 time from mild to severe disease using the Clinical Dementia Rating scale (5). Similarly, a study examining the time for patients needing care equivalent to placement in a health related facility, such as a nursing home, also obtained an estimate of about 6 years (7). We defined late stage disease to refer to the period when patients need such a high level of care. We used an annual transition probability from early to late stage disease of .167 in our model which corresponds to a mean duration of early stage disease of approximately 6 years. The model accounts for variability in the duration of early disease course (the 25, 50 and 75 percentiles of the distribution of durations of early stage disease are approximately 1.7, 4.2 and 8.3 years respectively). We performed sensitivity analyses to the underlying disease progression rate (γ) We recogni