Association of Alzheimer Disease With Life Expectancy in People With Down Syndrome

This meta-analysis assesses whether the variability in symptom onset of Alzheimer disease in Down syndrome is similar to autosomal dominant Alzheimer disease and its association with mortality.


Introduction
Down syndrome is the most frequent genetic cause of intellectual disability due to the presence of all or part of a third copy of chromosome 21, which occurs approximately in 1:800 births. In Europe and the US, current estimates indicate a population prevalence of 5.6 and 6.7 per 10 000 individuals, respectively, which translates to more than 200 000 people in the US 1 and more than 400 000 people in Europe. 2 Down syndrome is a complex condition with multiple associated comorbidities that vary throughout the lifespan. 3 Improvements in health care, most notably surgery for congenital heart defects, have remarkably increased life expectancy, from approximately 5 years in the 1950s to roughly 60 years in the 2020s. 4,5 However, this still represents a survival difference of about 20 years when compared with the general population.
A consequence of this increased life expectancy has been the emergence of age-related diseases, most importantly Alzheimer disease, which has a risk of more than 90% by the seventh decade for people with Down syndrome. 6 The strong association between Alzheimer disease and Down syndrome has a genetic basis through a gene-dose effect of the amyloid precursor protein (APP) gene located on chromosome 21, which is triplicated in this population. While the triplication of other genes on chromosome 21 can also contribute to amyloid aggregation, 7 APP triplication is necessary and sufficient to cause dementia. 8 Consequently, neuropathologic changes associated with Alzheimer disease, namely amyloid-β (Aß) plaques and neurofibrillary tangles composed of hyperphosphorylated tau, are universally seen in persons with Down syndrome by age 40 years, 9 and biomarker changes follow a predictable course that is strikingly similar to that found in autosomal dominant Alzheimer disease. 10 The latter represents genetic cases, which are caused by fully penetrant mutations in the APP, Presenilin 1 (PSEN1), or Presenilin 2 (PSEN2) genes, leading to earlyonset dementia at a predictable age, typically between age 30 and 60 years, depending on the mutation.
Because of these similarities, Down syndrome has been conceptualized as genetically determined Alzheimer disease, just like the autosomal dominant forms. 11,12 However, the implications of this conceptualization on the predictability of symptom onset in Down syndrome and the limit Alzheimer disease imposes on life expectancy, have not been fully explored. Indeed, many have emphasized the variability in age at onset of dementia in Down syndrome, [13][14][15][16][17][18][19] in sharp contrast with the recognized predictability of symptom onset in autosomal dominant forms. 20 As such, constructs like the estimated years to symptom onset (EYO), which have been developed and applied with success to study the natural history of Alzheimer disease in autosomal dominant forms, 21 have been less used in Down syndrome and remain controversial. This view represents an important setback for the inclusion of adults with Down syndrome in preventive trials, and for the development of appropriate care guidelines for patients and families.
For these purposes, we conducted a meta-analysis to determine the age at onset, age at death, and duration of Alzheimer disease dementia in adults with Down syndrome and compared the variability in disease onset with that reported in autosomal dominant Alzheimer disease. Based on these estimates, we constructed a hypothetical distribution of age at death in people with Down syndrome, under the assumption that Alzheimer disease had full penetrance and was lethal. We compared these results with mortality data from people with Down syndrome obtained from US death certificates in the past 50 years. We also investigated the underreporting of dementia in death certificates by imputing data based on the hypothetical distribution and the US Centers for Disease Control and Prevention (CDC) data and collecting mortality data from the Down Alzheimer Barcelona Neuroimaging Initiative (DABNI) clinical cohort. 10

Systematic Review and Meta-analysis
This meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Metaanalyses (PRISMA) reporting guideline. The review protocol and search strategy were registered in the International Prospective Register of Systematic Reviews (CRD42020203967). We conducted a systematic search to examine the age at onset, age at death, and duration of Alzheimer disease dementia among persons with Down syndrome using the following databases: PubMed/Medline, Embase, Web of Science, and CINAHL (eTable 1 in the Supplement). OpenGray was used for gray literature. Further details can be found in the eMethods in the Supplement.

Eligibility Criteria and Data Extraction
Studies were considered if they contained numerical results of the following outcomes among persons with Down syndrome: (1) age at onset or age at diagnosis of Alzheimer disease dementia, defined via sufficiently explained methodological criteria and based on expert clinical judgment or dementia criteria; (2) age at death in those with Alzheimer disease; or (3) disease duration (defined as the time from dementia onset/diagnosis to death). We imported records into the Covidence software version 2640 to assist the screening process. Two independent reviewers (MFI and DGC) screened all titles and abstracts to identify studies for full-text review. The risk of bias was assessed using a quality assessment tool (eTable 2 in the Supplement) adapted from the scale of McGrath et al. 22 Further details are included in the eMethods in the Supplement.

Data Synthesis and Analysis
Pooled estimates with corresponding 95% CIs were calculated using random-effects meta-analysis with the DerSimonian-Laird method. 23 Analyses were done in Stata version 15.1 using the metan package version 3.03 (StataCorp). Heterogeneity was assessed with the I 2 index. 24 Further details on the subgroup analyses are in eMethods in the Supplement.
To compare the variability in age at onset of Alzheimer disease in Down syndrome with that of autosomal dominant forms, we used 2 approaches: (1) computing the coefficients of variation of age at onset for each study individually (SD/mean × 100), and (2) calculating the time interval during which 95% of the individuals will develop cognitive symptoms (prediction intervals [95% PI], calculated as mean ± 1.96 × SD). For Down syndrome, these values were calculated from our systematic review of 44 studies on Alzheimer disease dementia. For autosomal dominant Alzheimer disease, these values were calculated from 60 studies (with n Ն 5) with data on age at symptom onset available in a previously published study. 20 Data collection took place from August 2020 to February 2021, and data analysis took place from February 2021 to July 2021.

Mortality Analyses Hypothetical Death Distribution Assuming Full Penetrance of Alzheimer Disease
Data analysis was performed in R version 4.0.4 (R Project for Statistical Computing). We used pooled data from the systematic review calculated with a fixed-effect model 25 to describe a hypothetical normal distribution of mortality in Down syndrome, assuming fully penetrant Alzheimer disease. The hypothetical death distribution was obtained from the sum of 2 independent distributions (age at onset and duration) as follows: mean death = mean onset + mean duration and SD death = sqrt (SD onset 2 + SD duration 2 ). We also examined the distribution using the estimate of age at death in those with Alzheimer disease.

US Mortality Data
We contrasted data obtained from the hypothetical death distribution with data obtained from We next examined the trend in the reporting of congenital heart disease and dementia in people with Down syndrome between 1968 to 2019 (eTable 3 in the Supplement). Given that dementia is typically underreported on death certificates, 26,27 we performed an exploratory analysis to compare the real age at death (CDC data) with the estimated ages at death in the fully penetrant hypothetical distribution by calculating the difference between the 2 for each percentile. Whenever the difference was close to 0, we assumed that Alzheimer disease could be the contributor. When this difference was larger, we attributed it to other causes of death. We plotted this difference for each percentile, and we visually assessed the graph and mathematically tested if there was an inflection point by implementing threshold regressions using the R library "chngpt".

Mortality Data From the DABNI Cohort
We collected mortality data from DABNI, a single-center prospective cohort that is part of a public health plan developed for the screening of Alzheimer disease in adults with Down syndrome in Catalonia, Spain, through clinical and multimodal biomarker assessments. 10 The  in the Supplement) and age at onset did not vary significantly by study quality (eFigure 2 in the Supplement).

Variability in the Onset of Alzheimer Disease
Despite differences in sample size, cohort characteristics, and methods, the variability in age at symptom onset between Down syndrome and autosomal dominant Alzheimer disease was  Hodgkins et al, 42 1993 Royston et al, 63 1994 Bayen et al, 30    Forest plot of mean age at onset of Alzheimer disease dementia (years) in adults with Down syndrome across 44 eligible studies from the systematic review (N = 2695 individuals). The points represent the mean, the arms indicate the 95% CIs, the light blue diamond indicates the overall pooled estimate (calculated with the DerSimonian-Laird method), and its width indicates the 95% CI. The dotted line indicates the pooled estimate of age at onset (53.8 years). The minimal sample size for a study to be eligible was 5 or more participants. Weights were calculated using random-effects analysis.

Mortality Analyses A Hypothetical Distribution of Age at Death Assuming Fully Penetrant Alzheimer Disease
Using a combined mean (SD) estimate of age at onset of 54.4 (7.1) years and disease duration of 5.1 (3.1) years calculated from the systematic review, we constructed a hypothetical distribution of age at death in Down syndrome assuming Alzheimer disease had 100% penetrance and was lethal. The estimated 10th, 25th, 50th, 75th, and 90th percentiles in this distribution were 49.6 years, 54 Importantly, these percentiles were in close agreement with our fully penetrant model ( Figure 4; eFigure 10 in the Supplement), suggesting that Alzheimer disease is a likely cause for this limit in survival in adults with Down syndrome. We also found racial disparities but no clear sex differences (Figure 4). Age at death was consistently lower for Black individuals, except for the higher percentiles in the 2010s.    The points indicate the mean, the arms indicate the 95% CI, the light blue diamond represents the overall pooled estimate (calculated with the DerSimonian-Laird method), and its width represents the confidence interval. The dotted line indicates the pooled estimates for age at death (58.4 years) and duration (4.6 years). The minimal sample size for a study to be eligible was 5 or more participants. Weights were calculated using random-effects analysis.

Contributing Causes of Death in Down Syndrome
Deaths related to congenital heart disease and dementia were frequent contributing causes of death in Down syndrome in the period 1968 to 2019. Deaths related to congenital heart disease decreased in the last 50 years, while those related to dementia increased up to 30% in 2019 ( Figure 5).
To approximate the percentage of deaths potentially linked to Alzheimer disease, we performed 2 analyses. First, we calculated the difference between the hypothetical mortality distribution and the observed ages at death. Whenever the difference was close to 0, we assumed that Alzheimer disease could be the contributor. A larger difference would likely indicate other causes of death. Figure 5 shows the difference between the hypothetical and CDC data for the year 2019 (and eFigure 12 in the Supplement for 1968 to 2019). We can see that the difference between these estimates is larger until approximately the 20th percentile and minimal after. We found a significant

Discussion
This study suggests that the onset of Alzheimer disease in Down syndrome is as variable as in the autosomal dominant forms. Our findings also suggest that most people with Down syndrome now live up to the limit imposed by Alzheimer disease, suggesting near full penetrance of this disease.
Therefore, the lifespan of people with Down syndrome, which is stil approximately 20 years lower than the general population, will not increase further until treatments against Alzheimer disease are available. These findings have considerable implications for public health, policymaking, and counseling to families. Notably, the common assumption has been that the age at onset of dementia in Down syndrome is highly variable, [13][14][15][16][17][18][19] in contrast with the autosomal dominant forms, in which age at onset is considered relatively constant within each pedigree. 20 This view likely stems from the emphasis given to the wide range of disease onset in Down syndrome (ie, 95% PIs) as opposed to the predictability of disease onset ascribed to autosomal dominant forms (ie, 95% CIs). Indeed, given that the diagnosis of symptomatic Alzheimer disease is challenging in Down syndrome, we expected a larger variability in disease onset compared with autosomal dominant Alzheimer disease. However, our analysis revealed that age at onset was consistent across studies, with a similar variability as in the autosomal dominant forms. Importantly, the estimate of age at onset and the variability range was similar to that described in families with APP duplications (52.2; 95% CI, 49.9-54.4). 20 This agrees with the concept that dementia in Down syndrome is primarily driven by the extra copy of the APP gene, which is both sufficient and necessary to cause Alzheimer disease pathology and symptoms. 8,12 While age at onset and its range were consistent across studies, this variability was quite large (25 years The main finding of this study is the observation that the real-world mortality data fitted with a distribution of age at death assuming fully penetrant Alzheimer disease. Treating modifiable conditions has dramatically increased life expectancy in Down syndrome in the past decades; however, Alzheimer disease is lethal and has no cure, and therefore imposes a limit on survival for most adults with Down syndrome. In other words, our study may provide an explanation for the gap in life expectancy between Down syndrome and the general population, and the approximately 15-year stagnation of age at death. Our study further expands these results by suggesting that, although Alzheimer disease diminished the differences in life expectancy between Black individuals and White individuals in the higher percentiles, disparities still persist in lower percentiles. As previously suggested, these differences likely reflect avoidable deaths via better access to specialized health services (eg, pediatric cardiology). 80   The importance of Alzheimer disease as a leading cause of death in Down syndrome is increasingly recognized and is reported in more than 30% of US death certificates in 2019. However, this is likely an underestimate, given that Alzheimer disease is typically underreported in death certificates in the general population. To attempt to quantify this underreporting, we conducted 2 analyses. First, we imputed data between our hypothetical death distribution and the CDC data, which suggested that Alzheimer disease could contribute to death in up to 80% of individuals with Down syndrome. Second, we analyzed mortality data from the DABNI cohort, which revealed that 78.9% of all deaths occurred in those with dementia. Of note, this is consistent with a prior longitudinal cohort study reporting that Alzheimer disease was the proximate cause of death in 70% of adults with Down syndrome. 41 Altogether, these data reinforce the idea that dementia-related deaths are largely underreported in death certificates of people with Down sydrome. 26,27 Notwithstanding the importance of other medical comorbidities, which can greatly affect the quality of life of people with Down syndrome, our findings emphasize a strong association between Alzheimer disease and mortality, which could explain the more than 20-year difference in life expectancy compared with the general population. Furthermore, our study underscores that Alzheimer disease has the same variability in age at onset in Down syndrome and autosomal dominant pedigrees. This demonstrates that Down syndrome is also an optimal population to study the natural history of Alzheimer disease. 10,81 Likewise, the predictability in age at onset (and biomarker changes) 10 seen in Down syndrome, similar to that seen in autosomal dominant forms, support the use of EYO when comparing clinical and biomarker changes in both genetic forms of dementia and can facilitate the enrolment of persons with presymptomatic disease into clinical trials. 12 Many efforts are already in place, including the development of large population cohorts gathering clinical and biomarker data on Alzheimer disease, 10,12,13 and also the launching of trial-ready cohorts across the US and several international sites.

Limitations
This study had limitations. We performed a systematic search and meta-analysis with strict criteria to ensure rigor and excellent methodological quality; however, diagnostic criteria used for dementia diagnosis varied across studies, and samples sizes were often small. Data were predominantly from European and US data sources, and the number of studies that included data on disease duration and age at death was small. Regarding the CDC mortality data, these were extracted from death certificates, which have inherent limitations, such as incomplete reporting, particularly of co-occurring conditions contributing to death. We overcame this limitation by also analyzing mortality data from a prospective clinical cohort (ie, DABNI) designed to study the natural history of Alzheimer disease in Down syndrome. As most participants in the DABNI cohort are of Caucasian ethnicity, we could not perform segregated analyses to match those presented with the CDC data.

Conclusions
These findings suggests that Alzheimer disease should be recognized as a critical medical priority in people with Down syndrome. The mortality data suggest that Down syndrome is compatible with fully penetrant Alzheimer disease. Therefore, increasing life expectancy and closing the gap with the general population will require effective prevention and management for Alzheimer disease.