Introduction
Alzheimer’s disease (AD) is the most common form of dementia, accounting for approximately 60–70% of the current 55 million dementia cases worldwide (WHO, 2021). Most often, symptoms develop after 65 years of age and known as late-onset AD (LO-AD) (WHO, 1992). However, it can also develop before 65 years of age, called early-onset AD (EO-AD), which accounts for approximately 5.5% of all AD cases (Zhu et al., Reference Zhu2015).
There is evidence that EO-AD and LO-AD differ in clinical presentation. For instance, many researchers have observed that a significantly larger proportion of EO-AD patients exhibit non-amnestic presentations in which their main symptoms involve language deficits, apraxia, and visuospatial deficits, more so than memory complaints, which are typical in LO-AD patients (Gumus et al., Reference Gumus, Multani, Mack and Tartaglia2021). However, non-memory symptoms are not exclusive to EO-AD. Licht et al. (Reference Licht, McMurtray, Saul and Mendez2007) observed poorer verbal fluency and motor-executive scores in LO-AD compared to EO-AD. Differences in the clinical characteristics and disease course have been compared between EO-AD and LO-AD; however, there are no reviews that compared EO-AD and LO-AD. Better understanding of the characteristics of EO-AD relative to LO-AD could facilitate early recognition of EO-AD in clinical settings and, therefore, enable appropriate management. It could also provide patients and their caregivers or families with more information regarding the symptoms and prognostic course for planning care. We undertook a systematic review and meta-analysis to examine if there were any differences in clinical characteristics such as time to diagnosis, cognition, neuropsychiatric symptoms (NPS), activities of daily living (ADLs), quality of life (QoL), and survival time between patients with EO-AD and LO-AD.
Method
Study selection
The review was undertaken according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines (Moher et al., Reference Moher, Liberati, Tetzlaff, Altman and The2009). A systematic search was conducted in the databases Medline, EMBASE, PsycINFO, and CINAHL for papers published until October 26, 2021. More relevant references were identified by the snowballing method, with manual search of reference lists of identified papers. The search terms were “Alzheimer’s disease” (as a sub-heading) AND (“early-onset Alzheimer* disease” OR “young*-onset Alzheimer* disease”) AND “late-onset Alzheimer* disease” AND “age of onset” (sub-heading) OR “Mini-Mental State Examination” OR “cognitive decline” OR “Neuropsychiatric Inventory” OR “Activities of Daily Living” (sub-heading) OR “survival” (sub-heading) OR “patient outcome assessment” (sub-heading).
Studies were included if (1) they included original research comparing diagnosed EO-AD and LO-AD patients in terms of time to diagnosis, cognitive status, cognitive decline, NPS, functional status, quality of time, or survival; (2) they defined EO-AD as any age below 65-years-old at onset and LO-AD as any age from over 65-years-old at onset; and (3) they were published as peer-reviewed journal articles in English language. Studies were excluded if they (1) had EO dementias without sub-typing into EO-AD; (2) were non-English; or (3) published as any publication type other than peer-reviewed journal article such as conference abstracts.
The process of selecting the final studies is demonstrated in the PRISMA chart (Figure 1) (Page et al., Reference Page2021). Studies were independently assessed by two researchers and disagreements resolved through consensus or discussions with a senior researcher.
Figure 1. Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) flow diagram.
Data extraction
A modified version of the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies, Adapted for Prognostic Factor Studies (CHARMS-PF) was used to guide information extraction (Moons et al., Reference Moons2014; Riley et al., Reference Riley2019) (Table 1). Information relating to statistical techniques and model development, study dates, and missing data is addressed in the assessment of risk of bias (Moons et al., Reference Moons2014). Extracted data were then inputted into the software program Review Manager (RevMan) 5.4 (The Cochrane Collaboration, 2020).
Table 1. Table of the included studies’ characteristics
Values are presented as n or mean (standard deviation) unless otherwise specified.
AD = Alzheimer’s disease, EO-AD = Early-onset Alzheimer’s disease, LO-AD = Late-onset Alzheimer’s disease, MMSE = Mini-Mental State Examination, ADL = Activities of Daily Living, NPI = Neuropsychiatric Inventory, I-ADL = Instrumental Activities of Daily Living, QoL-AD = Quality of Life-Alzheimer’s Disease, NINCDS-ARDRA = National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association, IWG = International Working Group, DSM-IV (or DSM-4) = Diagnostic and Statistical Manual of Mental Disorders fourth edition, DSM-IV-TR = Diagnostic and Statistical Manual of Mental Disorders fourth edition text revision, NIA-AA = National Institute of Aging-Alzheimer’s Association, CREDOS = Clinical Research Center for Dementia of South Korea, ADNI GO/2 = Alzheimer’s Disease Neuroimaging Initiative Grand Opportunity/2, CAMD = Coalition Against Major Diseases, MCI = Mild cognitive impairment, PET = Positron emission tomography, UK = United Kingdom, USA = United States of America.
Quality assessment
The risk of bias assessment of individual studies was done according to the PROGRESS framework by the application of the quality in prognostic factor studies (QUIPS) tool (Hayden et al., Reference Hayden, van der Windt, Cartwright, Côté and Bombardier2013; Riley et al., Reference Riley2019). The prognostic factor in this review is considered the age (at onset) variable that is used to separate the groups into EO-AD and LO-AD, where EO-AD is before 65 years and LO-AD is after 65 years. The authors made a priori decisions as to which QUIPS domains are most important to this review (described in the supplementary material).
Instruments
Time to diagnosis was defined as the time from symptom onset to diagnosis (Brück et al., Reference Brück, Wolters, Ikram and de Kok2021). Any studies that reported time to diagnosis in months were converted into years. Cognitive scores as obtained using the Mini-Mental State Examination (MMSE) (Folstein et al., Reference Folstein, Folstein and McHugh1975) assessed baseline cognition and cognitive decline. In AD populations specifically, there is an abundance of evidence suggesting that the MMSE is appropriate (Kørner et al., Reference Kørner, Lauritzen and Bech1996). This, as well as the measure’s wide use in the literature (Stanley and Walker, Reference Stanley and Walker2014; Stanley et al., Reference Stanley, Whitfield, Kuchenbaecker, Sanders, Stevens and Walker2019), is the rationale for selecting this measure for the current review. NPS are psychological and behavioral disturbances that are core features of AD and measured commonly using Neuropsychiatric Inventory (NPI) (Cummings et al., Reference Cummings, Mega, Gray, Rosenberg-Thompson, Carusi and Gornbein1994). Another variable of interest in the current review was functionality, defined as the ability to independently engage in and complete ADLs (Lawton and Brody, Reference Lawton and Brody1969). We also looked at QoL, which, in elderly populations, concerns physical condition, mood, relationships, ability to participate in meaningful activities, financial well-being, and the individual’s perceptions of their QoL, and, in this review, is operationalized as scores on the Quality of Life in Alzheimer’s disease (QoL-AD) scale (Logsdon et al., Reference Logsdon, Gibbons, McCurry and Teri1999). We extracted both self- and informant-reported QoL-AD scores. Finally, the survival time was defined as time from symptom onset or diagnosis to death (Brodaty et al., Reference Brodaty, Seeher and Gibson2012).
Data synthesis and analysis
For continuous variables (time to diagnosis, MMSE scores, mean annual change in MMSE scores, total NPI scores, ADL scores, QoL-AD scores, and survival times), an inverse variance statistical method with a random effects model was used. Random effects models allow for variation in studies’ true effect sizes due to differences in sample characteristics. Standardized mean difference (SMD) was the effect measure to account for differences in scales between studies (Higgins et al., Reference Higgins2022), for example modified or translated versions. RevMan 5.4 automatically calculated the SMD when means, standard deviations (SDs), and sample sizes for each group for each relevant study were inputted. Heterogeneity is assessed in each analysis using the Chi square test and the I 2 statistic. Interpretations of I 2 are based on the cutoff values suggested in the Cochrane Handbook (Higgins et al., Reference Higgins2022). The results of these analyses are illustrated in forest plots (Figures 2 and 3).
Figure 2. (a) and (b) Forest plots comparing symptom onset to diagnosis and cognitive decline of early-onset Alzheimer’s disease (EO-AD) and late-onset Alzheimer’s disease (LO-AD). (c) Forest plot comparing cognitive performance of early-onset Alzheimer’s disease (EO-AD) and late-onset Alzheimer’s disease (LO-AD).
Figure 3. (a–c) Forest plots comparing neuropsychiatric symptoms, functionality, and survival times of early-onset Alzheimer’s disease (EO-AD) and late-onset Alzheimer’s disease (LO-AD).
The Cochrane Group suggests that two studies are adequate to perform a meta-analysis given that those studies are compatible and found similar results (Ryan, Reference Ryan2016). As some of the studies included in the current review were compatible but not with similar results, we performed a meta-analysis where there were three or more studies that reported a mean and SD. When a meta-analysis could not be conducted, the studies were reviewed and reported narratively. When a study had multiple results from the same sample, for example before and after propensity score matching (PSM), analysis was conducted with the original cohort only, but, when a study investigated an outcome after PSM only, the post-PSM result was included in the analysis. Sensitivity analyses were conducted only when the original meta-analysis result is significant (Tawfik et al., Reference Tawfik, Dila, Mohamed, Tam, Ahmed and Huy2019). The factors considered in sensitivity analyses were decided a priori and included study setting, country, and samples’ disease severities when they were reported (Brück et al., Reference Brück, Wolters, Ikram and de Kok2021).
Where variables or measures relevant to the current review were mentioned but no data were reported in a paper, authors were contacted to request for the data, and if we received no response and unable to access the information, the study was not included in the relevant analysis. Where studies used samples stratified by variables in conjunction with age at onset, the data re-stratified by only age at onset with a cutoff value of 65-years-old were requested for from the authors. If provided with raw datasets without summary statistics, we calculated the means and SDs as appropriate. Relevant data from newly stratified versions were only extracted if the equivalent data were published in the original paper. When no inferential test results were provided and the full dataset was available, we used the relevant tests in SPSS version 25 to compare EO- and LO-AD group means. In the case of missing summary statistics data in published papers, they were computed where possible according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., Reference Higgins2022). Mean and SDs were obtained upon request from authors for the survival time variable (Rhodius-Meester et al., Reference Rhodius-Meester2019). If sample size data were missing and unable to be provided, the study was excluded from the analysis (Koedam et al., Reference Koedam, Lauffer, van der Vlies, van der Flier, Scheltens and Pijnenburg2010). All meta-analyzed means and SDs are presented in Supplementary Table 1.
Results
Data selection
Figure 1 (PRISMA flow chart) summarizes the study selection procedure. A total of 42 studies (EO-AD participants, n = 5,544, age = 61.32 ± 2.47; LO-AD participants, n = 16,042, age = 77.45 ± 3.27) were included. Main characteristics and outcome measures of each study are included in Table 1. Overall quality assessment of the 42 studies showed 23 (54.8%) had overall low risk of bias, 16 (38.1%) had overall moderate risk of bias, and 3 (7.1%) had overall high risk of bias (Table 1, Supplementary Figures 1 and 2).
Time to diagnosis
Six studies (EO-AD, n = 1,093; LO-AD, n = 3,000) compared the time to diagnosis of AD from symptom onset to diagnosis in EO and LO participants (Falgàs et al., Reference Falgàs2019; Mendez et al., Reference Mendez, Lee, Joshi and Shapira2012; Park et al., Reference Park2015; Stanley et al., Reference Stanley, Whitfield, Kuchenbaecker, Sanders, Stevens and Walker2019; van Vliet et al., Reference van Vliet2013; van der Vlies et al., Reference van der Vlies, Koedam, Pijnenburg, Twisk, Scheltens and van der Flier2009). The overall effect estimate was not statistically significant (SMD = 0.20, 95% confidence interval (CI) [−0.12, 0.52], Z = 1.21, p = 0.23) (Figure 2a). These data therefore do not provide evidence of a difference in the time period between onset of symptoms and diagnosis for EO-AD and LO-AD patients.
Cognition at presentation
Of the 36 studies that compared EO-AD and LO-AD cognitive performance using the MMSE, Koedam et al. (Reference Koedam, Lauffer, van der Vlies, van der Flier, Scheltens and Pijnenburg2010) did not provide exact sample size data for the two groups so was excluded. The remaining 35 studies (EO-AD, n = 3,059; LO-AD, n = 11,524) (Figure 2c) showed a statistically significant overall effect estimate (SMD = −0.22, 95% CI [−0.35, −0.09], Z = 3.37, p < 0.001), implying that EO-AD patients had lower MMSE scores than LO-AD patients at initial presentation. However, there was high heterogeneity (I 2 = 85%, χ 2(34) = 231.32, p < 0.001).
Cognitive decline
Ten of the included studies compared the rate of annual cognitive decline in EO-AD versus LO-AD patients using MMSE. Although Jacobs et al. (Reference Jacobs1994) compared six-monthly rate of change over 2 years, the exact change in MMSE scores for the annual period was not available; hence, we could not include them. However, they did report that EO-AD patients showed a more rapid decline than LO-AD (F = 12.50, p < 0.001). We also could not include Stanley et al. (Reference Stanley, Whitfield, Kuchenbaecker, Sanders, Stevens and Walker2019) as they did not report or provide information on means and SDs. The mean difference of MMSE from the analysis of the remaining seven studies (EO-AD, n = 356; LO-AD, n = 995) was statistically significant (SMD = −0.45, 95% CI [−0.74, −0.17], Z = 3.13, p = 0.002), suggesting that EO-AD patients had a greater decrease in their MMSE score per year than LO-AD patients (Figure 2b). There was however significant heterogeneity (I 2 = 66%, χ 2(7) = 17.40, p = 0.008).
Neuropsychiatric symptoms
Six studies (EO-AD, n = 1,049; LO-AD, n = 2,050) compared mean NPI scores of EO-AD and LO-AD participants. LO-AD participants had a greater total NPI score than the EO-AD group; however, this difference was not statistically significant (SMD = −0.42, 95% CI [−0.91, 0.08], Z = 1.66, p = 0.10) (Figure 3a). There was considerable heterogeneity in the effect estimates (I 2 = 96%; χ 2(5) = 132.74, p < 0.001).
There were three studies that compared EO-AD and LO-AD’s mean scores for the NPI subdomains (Baillon et al., Reference Baillon, Gasper, Wilson-Morkeh, Pritchard, Jesu and Velayudhan2019; Mushtaq et al., Reference Mushtaq2016; Toyota et al., Reference Toyota2007). Toyota et al. (Reference Toyota2007) compared only 10 of the subdomains, so we conducted meta-analysis for the original 10 subdomains (EO-AD, n = 110; LO-AD, n = 357). Overall effect estimates for agitation, disinhibition, and irritability could not be calculated in RevMan because SDs were reported as zero by Mushtaq et al. (Reference Mushtaq2016), leaving too few studies to conduct a meta-analysis for these domains. Of the remaining seven subdomains, there were no significant effect estimates, suggesting that EO-AD and LO-AD do not differ in these subdomains (Supplementary Figure 3) or for the total NPS.
Functional status
A number of studies compared ADL scores in EO-AD and LO-AD patients. However, they used different questionnaires such as Lawton ADL (Lawton and Brody, Reference Lawton and Brody1969), FAQ (Pfeffer et al., Reference Pfeffer, Kurosaki, Harrah, Chance and Filos1982), Bristol Activities of Daily Living (BADL) (Bucks et al., Reference Bucks, Ashworth, Wilcock and Siegfried1996), Alzheimer’s Disease Cooperative Study-Activities of Daily Living (ADCS-ADL) (Galasko et al., Reference Galasko1997), and Barthel Index (Mahoney and Barthel, Reference Mahoney and Barthel1965).
Analysis conducted on the three studies that used FAQ (EO-AD, n = 588; LO-AD, n = 1,454) showed the effect estimate was not statistically significant (SMD = 0.16, 95% CI [−0.09, 0.40], Z = 1.25, p = 0.21), suggesting that EO-AD and LO-AD patients have similar independence in instrumental ADLs (Figure 3b). The heterogeneity was not significant (I 2 = 63%, χ 2(2) = 5.39, p = 0.07) (Figure 3b). LO-AD patients had greater functional independence using Lawton I-ADL scores (Wattmo and Wallin, Reference Wattmo and Wallin2017), whereas EO-AD were shown to have less severe functional impairment on the ADCS-ADL scores (Grønning et al., Reference Grønning, Rahmani, Gyllenbuorg, Dessau and Høgh2012). On the other hand, some studies showed no difference between the groups using Lawton I-ADL (Carotenuto et al., Reference Carotenuto2012), BADL (Baillon et al., Reference Baillon, Gasper, Wilson-Morkeh, Pritchard, Jesu and Velayudhan2019), and ADCS-ADL (Kaiser et al., Reference Kaiser2012). Two studies compared Barthel Index scores (basic ADLs) between EO-AD and LO-AD patients. Park et al. (Reference Park2015) reported that EO-AD patients (n = 616) scored higher on the Barthel ADL scale than the LO-AD patients (n = 2,351), without any reference to values. Chang, Reference Chang2017) suggested EO-AD had significantly higher ADL scores (n = 331, mean (± SD) = 18.7 (± 3.2)) compared to LO-AD (n = 3,280, mean (± SD) = 17.9 (± 4.1) (p < 0.001)).
Quality of life
Two studies compared EO-AD and LO-AD and reported both the patient-reported and informant-reported QoL-AD scores separately. For the patient-reported data, EO-AD patients’ (n = 52) mean score was 33.2 (± 6.5), similar to LO-AD (n = 155, 34.3 ± 6.2) (Dourado et al., Reference Dourado, Laks and Mograbi2016). For the informant-reported data, EO-AD patients’ (n = 52) mean score of 29.6 (± 6.2) was similar to LO-AD patients (n = 155, 30.4 ± 7.6). A similarly non-significant finding was reported by Kimura et al. (Reference Kimura2018), wherein EO-AD patients had a self-reported QoL-AD score (n = 53, 33.6 ± 6.5) versus LO-AD (n = 57, 32.9 ± 5.8, p = 0.540). The difference between EO-AD and LO-AD’s caregiver-reported QOL-AD scores was also non-significant (p = 0.895).
Survival time
Three studies examined survival in EO-AD and LO-AD patients. Rhodius-Meester et al. (Reference Rhodius-Meester2019) defined survival as time in years from diagnosis to death, whereas Smirnov et al. (Reference Smirnov, Galasko, Hiniker, Edland and Salmon2021) and Spina et al. (Reference Spina2021) defined it as time in years from symptom onset to death. A meta-analysis could be conducted with these studies after Smirnov et al. (Reference Smirnov, Galasko, Hiniker, Edland and Salmon2021) provided summary statistics stratified by the cutoff of 65-years-old upon request (D. Smirnov, personal communication, December 5, 2021). The overall effect estimate was significant (SMD = 0.28, 95% CI [0.14, 0.42], Z = 3.97, p < 0.001) with moderate but non-significant heterogeneity (I 2 = 65%, χ 2(2) = 5.70, p = 0.06), suggesting that EO-AD patients survive for longer than LO-AD patients (Figure 3c).
Discussion
To our knowledge, this is the first systematic review and meta-analysis comparing clinical characteristics of EO-AD to LO-AD. We found people with EO-AD had poorer baseline cognitive scores and faster cognitive decline but longer survival times. They did not differ from people with LO-AD in terms of time to diagnosis, ADLs, and NPS. There were insufficient data to estimate overall effects in QoL.
EO-AD participants showed significantly poorer cognitive performance than LO-AD participants. LO-AD patients typically present mainly with episodic memory complaints, whereas a higher proportion of EO-AD patients exhibit non-amnestic cognitive syndromes that affect domains such as language and visuospatial abilities (Mendez et al., Reference Mendez, Lee, Joshi and Shapira2012), which could explain their lower MMSE scores. EO-AD presents with more non-cognitive symptoms such as depression and anxiety (Gumus et al., Reference Gumus, Multani, Mack and Tartaglia2021), and they may use association areas to compensate for their cognitive difficulties (Solé-Padullés et al., Reference Solé-Padullés2009) and it is not until later when the cognitive symptoms are more obvious and impacting their day-to-day life that they approach for assessment, which may also be a reason for lower scores at presentation. Although the effect size was small, this is of potential theoretical interest and an area to study further.
People with EO-AD showed a more rapid rate of annual decline in global cognitive function compared to LO-AD, with a medium effect size, which is potentially clinically meaningful and needs to be established in future studies. This may be due to the faster, more severe neuropathological changes that have been observed in EO-AD patients (Sakai et al., Reference Sakai2013). The faster EO-AD decline has been observed with multiple measures of cognition (Schneider et al., Reference Schneider, Kennedy, Wang and Cutter2015), so it would not have been affected by our choice of limiting the cognitive measure to the MMSE. The faster cognitive decline has also been noted beyond the period of 24 months from baseline. Sakai et al. (Reference Sakai2013) found that EO-AD patients’ MMSE decreased at an average rate of 1.9 (SD = 1.0) points per year, up to 90 months from baseline, which was significantly greater than LO-AD patients whose MMSE decreased by 1.1 (SD = 0.8) points on average per year (p < 0.001).
We did not see a difference in time to diagnosis from symptom onset to getting diagnosed between the two groups. Previous evidence that EO dementia has longer time to diagnosis may be influenced by the other subtypes of EO-dementias such as frontotemporal dementia which are more prevalent in younger age, more non-amnestic presentations, total number of specialist services consulted which increased the time to diagnosis, and maybe also lack of competence even in specialist services (Kvello-Alme et al., Reference Kvello-Alme, Bråthen, White and Sando2021; Loi et al., Reference Loi2022). However, when we focussed on EO-AD versus LO-AD there seems to be no difference, as also seen when time to diagnosis was compared for different types of young onset dementia, with shorter time to diagnosis in people with AD compared to the “other” dementia group (Loi et al., Reference Loi2022). This needs to be further investigated and compared between different dementia subtypes for the age groups. A recent study found delay in time to diagnosis of people with EO-AD who were mostly diagnosed using biomarkers (Kvello-Alme et al., Reference Kvello-Alme, Bråthen, White and Sando2021). This study however did not compare with LO-AD group. It would be useful and important for future studies to have accurate diagnosis using biomarkers which are now more available to compare EO-AD versus LO-AD.
Our meta-analysis of studies investigating EO-AD and LO-AD’s ADL using FAQ suggested that there was no difference in functional dependence, indicating the illness affects both younger and older adults similarly. In the literature, there is high variability in use of measures to assess functional status in AD. Future studies should aim to use consistent measures.
There was no significant difference in NPS between EO-AD and LO-AD. This is consistent with recent studies that also found no significant difference between LO-AD and EO-AD for overall NPI scores (Altomari et al., Reference Altomari2022; Falgàs et al., Reference Falgàs2022). We also found no significant group differences in any of the NPI sub-domain scores. Most NPI sub-domain difference effect sizes were negligible to small; however, the effect size for the group difference in aberrant motor behavior was large, but because of the small sample of studies limited the analysis’ power to detect statistical significance.
Two identified studies suggested that QoL for EO-AD and LO-AD patients was not significantly different regardless of who was reporting it, patient or the caregiver. More research in QoL comparing EO-AD versus LO-AD is needed (Ducharme et al., Reference Ducharme, Lachance, Kergoat, Coulombe, Antoine and Pasquier2016).
The lack of difference in NPS, functional dependence, and QoL possibly demonstrates that AD impacts people in similar ways irrespective of age at onset.
There were three studies comparing survival times (Rhodius-Meester et al., Reference Rhodius-Meester2019; Smirnov et al., Reference Smirnov, Galasko, Hiniker, Edland and Salmon2021; Spina et al., Reference Spina2021). Contrary to the assumption that EO-AD progresses quickly with a short survival period, the effect estimate suggested that EO-AD patients survive longer than LO-AD patients. The explanation could be that younger people are in better physical health and medical conditions compared to elderly with LO-AD (Spina et al., Reference Spina2021). EO-AD patients may also have protective factors, such as more commonly experiencing atypical presentations such as language symptoms or executive dysfunction, which are associated with longer survival (Pavisic et al., Reference Pavisic2020). It could also be because the studies included in the review were sporadic EO-AD cases which are unlike familial EO-AD which is known to have rapid progression (Loeffler, Reference Loeffler2021). Future research could investigate survival in different types of EO-AD cases compared to LO-AD.
It is interesting that our meta-analysis showed that EO-AD had lower cognitive performance at presentation and greater cognitive decline but longer survival rate. Higher age at diagnosis, higher number of medical comorbidities, and greater disability have been shown to predict shortened survival better than cognitive impairment (Lichtenstein et al., Reference Lichtenstein2018). Later age at diagnosis and greater disease severity at presentation have been associated with shorter survival time (Brodaty et al., Reference Brodaty, Seeher and Gibson2012; Schaffert et al., Reference Schaffert2022). Future studies should examine the role of cognitive impairment in predicting life expectancy in those with milder dementia using more sensitive neuropsychological measures (Schaffert et al., Reference Schaffert2022).
Our review and meta-analysis are limited by methodological weaknesses in the included studies as identified during quality assessment, specifically pertaining to sample size, selective outcome reporting, and variability of questionnaires used in the studies. Of the included studies, 18 were investigating neuroimaging and/or biomarkers and, although we excluded studies that matched groups by MMSE scores, there may have been potential confounders in some studies that may have clinically homogenous groups.
Our review was limited to studies that directly compared EO-AD and LO-AD. Some studies identified in the literature search divided their samples into EO-AD, middle-onset AD (MO-AD), and LO-AD (Stanley et al., Reference Stanley, Whitfield, Kuchenbaecker, Sanders, Stevens and Walker2019). Their findings suggest that MO-AD may be another separate sub-type of AD, as found that MO-AD patients have a significantly different monthly rate of MMSE decline compared to EO-AD and LO-AD patients (Stanley et al., Reference Stanley, Whitfield, Kuchenbaecker, Sanders, Stevens and Walker2019). Future meta-analyses could also compile research comparing EO-AD, MO-AD, and LO-AD in order to provide a further elucidated and more specific understanding of the effects of age of onset on AD characteristics if there are enough studies. We conducted a meta-analysis for the survival time including studies which defined them either as diagnosis to death or symptom onset to death similar to the criteria previously used (Brodaty et al., Reference Brodaty, Seeher and Gibson2012). However, future studies should have uniform and similar definitions for better comparisons. More research comparing EO-AD and LO-AD patients, particularly in the domains of basic ADLs and QoL, is needed to improve clinical knowledge of how these conditions differ based on age of onset. We did not test the influences of prevalence of APOE ϵ4 allele, presence of co-pathologies, or compare biomarkers such as blood and CSF markers, or neuroimaging (brain volumes). Future reviews should examine influences of these too.
Conclusion
Our meta-analysis suggests that people with EO-AD have poorer cognitive performance at presentation, faster cognitive decline, and longer survival times compared to people with LO-AD but did not differ in time to diagnosis, functional dependence, and total NPS. This implies that the AD condition is similar and affects people in similar way irrespective of age of onset. However, further research is warranted for clarification about differences in EO-AD and LO-AD’s QoL. It is important to understand EO-AD’s characteristics to facilitate early identification, diagnosis, and management and to alleviate the burdensome social, emotional, and financial, as well as medical consequences. A greater understanding of EO-AD symptoms, course, and prognosis will also aid patients and their caregivers and families for future care needs and planning. Better understanding of clinical features along with their underlying neuropathologies would also support precision medicine with appropriate pharmacological and non-pharmacological interventions.
Conflict of interest
None.
Description of author(s)’ roles
LM and PS did the literature review. PS did data extraction, assessed risk of bias, carried out the analysis, and drafted the manuscript. LV conceived the project, was involved in study selection, assessed risk of bias, supervised data extraction and analysis, interpretation of the results, and finalized the manuscript.
Acknowledgements
The authors would like to thank Sara Pisani for help with R codes.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1041610223000509.
Open access
