Abstract
Background:
Using the Faroese Septuagenarian cohort, we aimed to describe the incidence of dementia and assess the validity of neurocognitive tests to predict subsequent dementia diagnosis.
Methods:
In this population-based cohort, 713 Faroese septuagenarians aged 70-74 years without dementia, underwent clinical and neuropsychological examinations. After 10-years of follow-up, information was collected on all participants referred for cognitive evaluations and diagnosed with dementia. Incidence rates were calculated and presented with 95% confidence intervals (CIs), assuming a Poisson distribution. We then performed discriminant analysis to determine the best set of neuropsychological tests to identify those who would develop dementia.
Results:
Over the 10-years, 65 participants (9.1%) were diagnosed with dementia, with a 10-year incidence rate of 1,063 cases per 100,000 person-years (95% CI=825, 1,343). Women had a greater incidence than men (incidence rate ratio (IRR) =1.58; 95% CI=0.93, 2.71). After stepwise selection, gender and six neuropsychological measures were selected to discriminate between those who would and would not develop dementia. Overall, the model was able to correctly identify 82% of those who would not develop dementia (specificity) and 71% of those who would (sensitivity).
Conclusions:
These results indicate that among a greater number of tests covering a broad range of cognitive abilities, tests reflecting verbal and visual learning and recall, visuospatial function, attention, and encoding into and retrieval from long-term memory may be helpful in identifying patients in the pre-symptomatic phase of dementia. Thus, helping care-givers identify patients at a higher risk of developing dementia and adjusting management of care accordingly.
Keywords: Alzheimer’s disease, Dementia, Incidence, Neuropsychological tests, Faroe Islands
1.1. Introduction
Dementia is associated with substantial social and economic burden for both patients and caregivers and is among the leading causes of incapacitation among the elderly. The number of people living with dementia is expected to increase greatly in the coming decades as the population ages and life expectancy increases.[1,2] Currently, an estimated 47 million people live with dementia globally. This is projected to increase to 75 million by 2030, and nearly triple by 2050.[3,4]
Some studies have suggested age-specific incidence rates have declined in Western countries, even as the prevalence increases.[5–7] These declines have been attributed to increased education and improved control of cardiovascular risk factors. However, findings are mixed, and other studies have shown no decline in incidence.[8] And changing prevalence of other upstream dementia related risk factors, such as an increasing type 2 diabetes burden, may influence the incidence of Alzheimer’s and other dementias in the coming decades.[9,10] There is also strong evidence linking cardiovascular health and risk factors in not only vascular dementia, but also Alzheimer’s. Hypertension, hypercholesterolemia, obesity, metabolic syndrome, smoking, and hyperhomocysteinemia, are risk factors linked to dementia[11]. Furthermore, among neurologically healthy individuals, cardiovascular health has been associated with worse performance on neuropsychological measures among[12,13].
Neuropsychological measures can monitor cognitive status and there is great interest in identifying measures sensitive to the cognitive changes caused by AD. Previously, deficits in three main cognitive domains have been shown to be sensitive in predicting conversion to AD: episodic, verbal and visuospatial memory; executive function; and language.[14] While a large number of neuropsychological tests are available, it is not easy to determine which neuropsychological test might be most sensitive in identifying prodromal AD.
In this community-based study, we assessed the incidence of Alzheimer’s disease and other dementias over 10 years of follow-up in the Faroese Septuagenarian cohort and explore gender influences on incidence and age at onset. Furthermore, we assess the performance of the ability of neurocognitive tests performed at baseline to identify those who would develop dementia.
1.2. Methods
1.2.1. Study Population
The Faroese septuagenarian cohort is a study of 713 residents of the Faroes Islands aged 70-74 years, established in 2008. Detail on study procedures has been published previously.[15] Briefly, in this population-based study, all 1,131 Faroese citizens in the Faroese Population Registry born between January 1934 and August 1937 were invited to participate in the study, and 713 people agreed (63%). The Faroese reside on a dozen islands in the Northern Atlantic between Norway and Iceland with the majority (80%) living on four northern islands connected by tunnels or bridges. All subjects underwent a thorough clinical examination with emphasis on neurological and cardiovascular function, using validated and standardized protocols and calibrated instruments[15]. Weight, height, hip and waist circumstance were measured. Systolic and diastolic blood pressure were measured, visual test and audiometry performed. An electrocardiogram (ECG) was obtained using standard procedures and ultrasound measurement of the intima-media thickness and plaque were performed[16,17]. A questionnaire was employed recording current health and past medical history, including medication, risk factors such as smoking and alcohol use, and body weight at age 20 and 30 etc. We also assessed mortality information using vital statistics.
In connection with the 10-year follow-up of the Septuagenarian Cohort, all cohort members referred for cognitive evaluations and diagnosed with dementia over 10-years of follow-up were identified by linking to the Dementia Registry which is registry based on records from the only specialized memory clinic in the Faroe Islands[18]. Thus, analysis is based on the entire Faroese Septuagenarian cohort, without loss to follow-up.
The study protocol was approved by the Faroese ethical committee.
1.2.2. Neuropsychological Measures
Cognitive function was assessed by an in-person neuropsychological test battery administered by a psychologist assisted by a trained nurse. Overall, 12 tasks were administered, with 25 different variables produced. The test selection was guided by an overall objective to sample broadly from the universe of human mental abilities by specific tests with good psychometric properties in order to represent a number of broad ability domains. Language comprehension was assessed by the Boston Naming Test (BNT), a visual confrontation naming test measuring vocabulary and word retrieval or word finding performance of a subject. Raven’s Colored Progressive Matrices (set A and B, excluding the middle set, Ab) is a non-verbal reasoning test adapted for elderly people. Neurobehavioral Evaluation System2 (NES2), Pattern Recognition assesses visuospatial skills. Warrington’s Recognition Memory for Faces (series B) measures visuo-perceptual and visuospatial memory. Attention was assessed with NES2 Continuous Performance Test (CPT) and NES2 Symbol Digit Modalities Test. Verbal memory was assessed with 10-word list learning with selective reminding. Motor function was measured with NES2 Finger Tapping test and with Purdue Pegboard test. More detail has been published.[15]
1.2.3. Diagnostic procedures
There is one specialized memory clinic on the islands, the Dementia Clinic at the National Hospital in Torshavn. Referrals to the Dementia Clinic for diagnostic evaluation are made through general practitioners who provide the first screening. The dementia diagnoses are established based on ICD-10 criteria as previously described.[15]
1.2.4. Statistical Methods
Differences in the distribution of baseline demographics and health characteristics between those who developed dementia versus those who did not were compared using Student’s t-test (continuous data) and Pearson chi-square test (categorical data).
Incidence rates over the 10-year follow-up period in the Faroese Septuagenarian cohort were calculated and presented with a 95% confidence interval (CI), assuming a Poisson distribution. We calculated person-time at risk for the population, beginning at the date of baseline exam until date of dementia diagnosis or if undiagnosed, death or January 1, 2018. Incidence rates (dementia cases / person-time at risk) are presented per 100,000 person-years. Incidence estimates were also stratified by years of follow-up (incidence rate during 0-5 years and 5-10 years follow-up), age at diagnosis (70-74 years, 75-79 years, and 80+ years), and gender. We also calculated the study period mortality rate in the same manner, with person-time at risk being the date of baseline exam until date of death or January 1, 2018 for those censored.
We performed discriminant analysis to determine the best set of neuropsychological battery predictors (measured at baseline) to distinguish those who would develop dementia versus those who would not.[19] This involved two steps of analysis. In the first analytical step (SAS 9.4, Proc Stepdisc), we determined the best set of neuropsychological measures to explain variance in dementia occurrence, using stepwise regression analysis with stepwise variable selection. Stepwise model selection incorporates both forward and backward selection. Beginning with no variables in the model with dementia as the outcome, variables are sequentially entered. After each addition, the model is examined and the variable may be selected or a variable previously selected may be eliminated because it contributes the least to the discriminatory power of the model. On the other hand, a variable not initially in the model that contributes most to the discriminatory power of the model when entered is retained.[20] Two criteria are used to determine if variables enter or leave the model: 1) the significance level of an F-test from an analysis of covariance, where the variables already chosen act as covariates and the variable under consideration is the dependent variable; 2) the squared partial correlation for predicting the outcome (dementia yes/no), controlling for the effects of the variables already selected for the model.[20] For selection consideration, we included all neuropsychological measures and basic demographics: gender, age at baseline, professional education (none, short, middle, long), current or former smoking indicators (daily or sometimes), and an indicator for brain disease at baseline as potential predictor variables.
In the next analytic step, we then used the selected variables to build a predictive dementia discriminant model based on the observed information, and classified each subject based on this model as predicted of having dementia or not (SAS 9.4, Proc Discrim).[20] The goal being to determine whether certain neuropsychological screening tests can be used to discriminate between those who had an increased risk of developing dementia and those who did not.
We tested the normality of the distributions of each selected classification variable (step 1) within dementia groups to determine whether parametric or non-parametric discriminant methods should be used for analytic step 2. Given the skewness, kurtosis, and Shapiro-Wilk test (supplemental table 1), we could not conclude normality. Therefore, for analytic step 2, we used a k-nearest-neighbor (kNN) method to generate the non-parametric density estimate in each group (each subject is classified into a group based on information about the selected variables from step 1 from the k nearest neighbors based on Mahalanobis distance), with leave-one-out cross-validation, i.e., that each subject is classified as dementia/non-dementia using the k nearest neighbors, excluding the subject under consideration. We present the performance of the discriminant criterion via classification error rates.[19]
We then compared the baseline neuropsychological performance of the subjects based on their classification in discriminant analysis (true positive, true negative, false positive, false negative) using a Wilcoxon test.
1.3. Results
A flowchart for the screening and diagnosis among the Faroese Septuagenarian cohort is shown in Figure 1. A total of 713 septuagenarians, aged 70-74, were enrolled and examined for the study. This represented 63% of the eligible population. Among the participants, slightly over half were male (n=361, 51%), the average age at exam was 72.4 years (SD=1.2), and the majority were not currently smoking (n=123, 17%) though a large proportion did report smoking in the past (n=380, 63%) (table 1).
Figure 1.
Flowchart detailing the characteristics and flow of the Faroese Septuagenarian cohort through dementia screening and evaluation procedures.
Abbreviations: GP=general practitioner; AD=Alzheimer’s disease; DLB=dementia with Lewy bodies; MCI=mild cognitive impairment
Table 1.
Population characteristics of study population by dementia
| Population Characteristics (Mean ± SD / n (%)) |
Full Population (n=713) | Dementia | |
|---|---|---|---|
| Yes (n=65) | No (n=648) | ||
| Male | 361 (51) | 25 (38) | 336 (52)* |
| Age at Exam | 72.4 ± 1.2 | 72.6 ± 1.1 | 72.4 ± 1.2 |
| Current BMI | 28.9 ± 4.7 | 27.9 ± 4.4 | 29.0 ± 4.7* |
| BMI at 20 | 23.4 ± 3.2 | 23.2 ± 3.9 | 23.4 ± 3.1 |
| Current Smoker (daily or sometimes) | 123 (17) | 8 (12) | 115 (18) |
| Former Smoker (daily or sometimes) | 380 (63) | 36 (59) | 344 (63) |
| Professional Education, None | 344 (48) | 37 (57) | 307 (48) |
| Short | 28 (4) | 0 (0) | 28 (4) |
| Middle | 280 (39) | 26 (40) | 254 (39) |
| Long | 60 (8) | 2 (3) | 58 (9) |
| Brain disease | 42 (6) | 5 (8) | 37 (6) |
| Mortality, during study period | 152 (21) | 18 (28) | 134 (21) |
p<0.05; chi-square or t-test
Among the 713 participants interviewed at baseline, 74 were referred to the dementia clinic for cognitive evaluation over the following 10 years. Of these participants, 5 (6.7%) were determined to not have dementia or mild cognitive impairment (MCI), 4 (5.4%) had MCI, 4 (5.4%) had an unspecified or other dementia diagnosis, and 61 (82.4%) had a form of specified dementia. This included Alzheimer’s disease (n=47, 63.5% of those referred), vascular dementia (n=7, 9.5%), mixed (n=4, 5.4%), and dementia with Lewy bodies (n=3, 4%) (supplemental table 2).
Overall and sex-specific incidence estimates, along with female-to-male ratios for incident dementia are presented in table 2. Sixty-five participants developed dementia over the 10-year follow-up, i.e., 9.1% of the cohort (estimated incidence of 91 per 1,000 people (95% CI=71, 115)). The overall, 10-year incidence rate was 1063 per 100,000 person-years (p-y) (95% CI=825, 1342). The incidence was significantly greater as follow-up time and age increased. For example, according to age at diagnosis, the incidence rate among those aged 70-74 years was 440/100,000 p-y (95% CI=201, 819), 75-79 years was 1140/100,000 p-y (95% CI=807, 1553), and 80+ was 1851/100,000 p-y (95% CI=1168, 2759) (table 2).
Table 2.
Incidence of dementia and mortality among septuagenarians in the Faroe Islands over 10 years of follow-up, by time period and age at diagnosis or death
| Incidence Period | At Risk Population | Events | Total Person-time at risk | Incidence per 1,000 people (95% CI) | Incidence Rate per 100,000 PY (95% CI) | Female Incidence Rate per 100,000 PY (95% CI) | Male Incidence Rate per 100,000 PY (95% CI) | Female to Male IRR (95% CI) |
|---|---|---|---|---|---|---|---|---|
| Dementia | ||||||||
| Cumulative, 10 years of follow-up | 713 | 65 (9.1%) | 6114 | 91.2 (70.8, 115.2) | 1063 (825, 1343) | 1300 (937, 1745) | 823 (541, 1190) | 1.58 (0.93, 2.71) |
| 0-5 years follow-up | 713 | 24 (3.4%) | 3399 | 33.7 (21.9, 49.0) | 706 (460, 1028) | 890 (512, 1442) | 526 (253, 947) | 1.69 (0.69, 4.39) |
| 5-10 years follow-up | 637 | 41 (6.4%) | 2715 | 64.4 (46.6, 86.1) | 1510 (1094, 2020) | 1796(1181, 2595) | 1209 (709, 1902) | 1.49 (0.76, 2.98) |
| 70-74 years | 713 | 8 (1.1%) | 1820 | 11.2 (5.1, 20.9) | 440 (201, 819) | 680 (270, 1377) | 213 (36, 659) | 3.19 (0.57, 32.27) |
| 75-79 years | 683 | 36 (5.3%) | 3159 | 52.7 (37.3, 71.8) | 1140 (807, 1553) | 1325 (836, 1975) | 953 (548, 1520) | 1.39 (0.68, 2.90) |
| 80+ years | 572 | 21 (3.7%) | 1135 | 36.7 (23.2, 54.7) | 1851 (1168, 2759) | 2132 (1173, 3510) | 1524 (697, 2837) | 1.40 (0.54, 3.89) |
| Mortality | ||||||||
| Cumulative, 10 years of follow-up | 713 | 152 (21%) | 6344 | 213.2 (183.7, 245.1) | 2400 (2035, 2797) | 1710 (1297, 2202) | 3102 (2525, 3761) | 0.55 (0.39, 0.78) |
| 0-5 years follow-up | 713 | 52 (7%) | 3442 | 72.9 (54.9, 94.5) | 1511 (1136, 1959) | 1048 (635, 1609) | 1972 (1381, 2711) | 0.53 (0.28, 0.97) |
| 5-10 years follow-up | 661 | 100 (15%) | 2892 | 151.3 (124.8, 180.9) | 3458 (2824, 4180) | 2468 (1756, 3351) | 4522 (3496, 5733) | 0.55 (0.35, 0.83) |
| 70-74 years | 713 | 21 (3%) | 1828 | 29.5 (18.3, 44.7) | 1148 (725, 1712) | 789 (339, 1526) | 1487 (838, 2407) | 0.53 (0.18, 1.40) |
| 75-79 years | 692 | 81 (12%) | 3272 | 117.1 (94.1, 143.4) | 2475 (1975, 3054) | 1687 (1137, 2391) | 3228 (2472, 4240) | 0.51 (0.31, 0.83) |
| 80+ years | 611 | 50 (8%) | 1235 | 81.8 (61.3, 106.5) | 4050 (3028, 5278) | 2985 (1860, 4491) | 5278 (3608, 7398) | 0.57 (0.30, 1.03) |
Person-time: Years from baseline exam to dementia, death prior to 2018 (if participant did not have dementia), or censor (January 1, 2018); 95% CI based on Poisson distribution
Women had an incidence of dementia 1.6 times that of men (95% CI=0.93, 2.71). This was particularly notable among dementia diagnoses prior to age 75 (70-74 years IRR=3.19), though small numbers limit any real conclusions (95% CI=0.57, 32.27). Additionally, overall, women were more likely to be diagnosed with AD (8.5% of women vs 4.7% of men; p=0.04), but not other dementias. Among those who developed dementia, the mean age at diagnosis was 78.5 years (SD=2.6) and there were no gender differences in age at onset (supplemental table 2).
We further calculated the cumulative incidence rate excluding patients diagnosed during the first year of follow-up to account for potential bias due to late diagnosis. This included one patient who was diagnosed during the first year of follow-up. The cumulative 10-year incidence excluding this participant was 1047 per 100,000 p-y (95% CI=811, 1325) (women: 1267/100,000 p-y (95% CI=910, 1708); men: 823/100,000 p-y (95% CI=541, 1190)).
Over the study period, 152 (21%) participants died (estimated incidence of 213 per 1,000 people (95% CI=184, 245); table 2). A higher proportion of those who developed dementia died than those who did not (28% vs 21%). The cumulative mortality rate was 2400 per 100,000 p-y (95% CI=2035, 2797); the rate strongly increased with age and follow-up time. Furthermore, as expected, women had a significantly lower mortality rate than men (IRR=0.55 (95% CI=0.39, 0.78), which was seen across age ranges and follow-up time.
Individual analysis of each of the single neuropsychological tests related to dementia (logistic and Cox proportional hazards regression) can be found in supplemental table 3. A number of component tests were individually associated with dementia, including immediate and delayed memory tasks (BNT, delayed OR=0.81 (95% CI=0.75, 0.88); Selective Reminding, reproduction (OR=0.71 (95% CI=0.64, 0.80); Face Recognition Memory, immediate (OR=0.94 (95% CI=0.90, 0.98)), and also attention and speed (NES2 Symbol Digit Mean (OR=1.08 (95% CI=1.01, 1.16)), and verbal comprehension (BNT OR=0.95 (95% CI=0.91, 0.99)) (supplemental table 3).
In order to select the best subset of neuropsychological variables to discriminate between those who would develop dementia and those who would not, we performed stepwise discriminant analysis. After stepwise variable selection, gender and 6 of the 25 neuropsychological variables were selected. The six neuropsychological measures were Selective Reminding Recall, Reproduction; Face Reproduction Total, SD for Correct Responses (per 1000); Boston, Delayed Recall;1 Continuous Performance Test (CPT), Mean of RTs for Blocks 1 – 5; Pattern Recognition, number correct; and CPT, Std. Deviation of RTs for Blocks 1 - 5. The association between each of the selected variables and dementia, with each variable included as a covariate in the same logistic model, can be found in table 3. Selective Reminding Recall (Reproduction) score (OR=0.76 (95% CI=0.66, 0.88)) and female gender relative to male (OR=3.28 (95% CI=1.78, 6.03)) were the strongest predictors of dementia among the seven selected variables.
Table 3.
Association between neurocognitive tests selected from stepwise variable selection (shown in order of selection) and dementia. Logistic regression, outcome is dementia, each characteristic included as a covariate in the same model.
| Domain | Characteristic or Neuropsychological test | Description / Reference | Logistic Regression | |
|---|---|---|---|---|
| OR (95% CI) | p-value | |||
| Delayed recall | Selective Reminding, Delayed | SelRem Recall, Reproduction | 0.76 (0.66, 0.88) | 0.0002 |
| Delayed recall | Face Reproduction | Face Reproduction Total, SD for Correct Responses (per 1000) | 1.0002 (1.0001, 1.0004) | 0.01 |
| Female Gender | Male Reference | 3.28 (1.78, 6.03) | 0.0001 | |
| Incidental memory, delayed | BNT, Delayed | Boston, Delayed Recall | 0.86 (0.78, 0.94) | 0.001 |
| Attention, Speed | NES2, Continuous Performance Test with letters | CPT, Mean of RTs for Blocks 1 - 5 | 0.995 (0.989, 1.000) | 0.07 |
| Visuospatial Perception | NES2, Pattern Recognition, Correct | Pattern Recognition - No. Correct | 1.30 (0.98, 1.72) | 0.07 |
| Attention, Speed | NES2, Continuous Performance Test with letters | CPT, Std. Deviation of RTs for Blocks 1 - 5 | 1.01 (0.99, 1.02) | 0.24 |
We used this subset of seven selected variables to build a model to predict dementia status and we show the classification and error rates based on the discriminant analysis (supplemental table 5). Overall, the model, which relied on gender and the six neuropsychological test measures, was able to correctly distinguish 82% of those without dementia (n=512) and 71% of those with dementia (n=42), for an overall accuracy of 77%. Descriptives of all the neuropsychological measures at baseline shown by prediction status are presented in table 4. Those in the false negative, false positive, and true positive groups are each compared to those in the true negative group.
Table 4.
Neuropsychiatric test measures (mean (SD)) by discriminant analysis dementia prediction status. False Negatives (FN), False Positives (FP), and True Positives (TP) compared individually to True Negatives (TN).
| Domain | Characteristic or Neuropsychological test | Description / Reference | TN (n=512) | FN (n=18) | FP (n=113) | TP (n=42) |
|---|---|---|---|---|---|---|
| Delayed recall | Selective Reminding, Delayed | SelRem Recall, Reproduction | 8.8 (1.4) | 8.6 (1.9) | 7.9 (2.5)* | 6.6 (3.2)* |
| Incidental memory, delayed | BNT, Delayed | Boston, Delayed Recall | 8.0 (3.3) | 7.6 (4.4) | 7.0 (3.9)* | 5.1 (3.8)* |
| Visuospatial Perception | NES2, Pattern Recognition, Correct | Pattern Recognition - No. Correct | 24.2 (1.3) | 24.4 (1.5) | 24.3 (1.1) | 24.4 (0.9) |
| Delayed recall | Face Reproduction | Face Reproduction Total, RT For Correct Responses | 3687.9 (1396.1) | 3840 (1508.1) | 4823.2 (2268.9)* | 5087.7 (2494.1)* |
| Face Reproduction Total, SD For Correct Responses | 2154.9 (1199.3) | 2460.3 (1400.3) | 3117.4 (1565.5)* | 3271.5 (1694.2)* | ||
| Verbal comprehension | BNT, no Cues | Correct without cue | 17.6 (5.6) | 16.5 (5.8) | 15.8 (6.1)* | 14.6 (5.1)* |
| BNT, with Cues | Correct with stim. & phon. cue | 24.0 (4.4) | 22.9 (4.4) | 22.6 (5.0)* | 22.0 (4.2)* | |
| Non-verbal reasoning | Raven Coloured Progressive Matrices | Raven Coloured Progressive Matrices A+B | 17.3 (2.8) | 17.2 (3.1) | 17.2 (2.9) | 16.4 (2.8)* |
| Perceptual Speed | NES2, Pattern Recognition, time | Pattern Recognition - No. Correct RT | 7.6 (2.8) | 7.5 (1.5) | 8.4 (3.1)* | 9.1 (4.4)* |
| Attention, Speed | NES2, Continuous Performance Test with letters | CPT, Mean of RTs for Blocks 1 - 5 | 457.3 (67) | 442.4 (49.8) | 464.2 (60.4) | 479.6 (65.3)* |
| CPT, Std. Deviation of RTs for Blocks 1 - 5 | 87.1 (25.2) | 81.1 (17.9) | 79.8 (24.4)* | 94.2 (25.9)* | ||
| Attention, Speed | NES2, Symbol Digit Modalities Test | Symbol-Digit- Mean for 40 Responses, Seconds | 5.1 (1.6) | 5.5 (1.3) | 5.9 (2.6)* | 6.5 (3)* |
| Symbol-Digit- SD for 40 Responses | 2.6 (1.3) | 2.9 (1.4) | 3.1 (1.9)* | 3.4 (2.1)* | ||
| Speed of processing | Face Recognition Memory Test, Immediate Time | Face Immediate Total RT For Correct Responses | 4224.2 (1810.1) | 3968.3 (1500.8) | 5520.7 (2682.7)* | 6114.8 (2854.5)* |
| Face Immediate Total SD for Correct Responses | 2444.9 (1472.6) | 2402.2 (1395.2) | 3494.5 (1841.4)* | 3807.7 (1958.1)* | ||
| Immediate memory | Selective Reminding, immediate recall | SelRem Recall, Trial 1 | 4.2 (1.2) | 4.2 (1.1) | 3.9 (1.3) | 3.8 (1.2)** |
| Learning and recall | Selective Reminding, Reminders | SelRem Reminders Total | 19.5 (10.8) | 23.4 (11.6) | 26.2 (15.8)* | 30.4 (16.8)* |
| Immediate memory | Face Recognition Memory Test, Immediate recall | Face Immediate - Total No. Correct | 40 (4.9) | 39.3 (7.5) | 38.1 (6.1)* | 37 (6.7)* |
| Delayed recall | Face Recognition Memory Test, Delayed | Face Reproduction Total - No. Correct | 37.1 (5.1) | 37.9 (5.8) | 35.4 (5.7)* | 34.7 (6.1)* |
| Manual motor speed | NES2, Finger Tapping Test, Dominant | No. Taps with Preferred Hand on Best Trial | 69.3 (13.8) | 71.7 (14.9) | 66.4 (13.1)* | 64 (11)* |
| No. Taps with Non-preferred Hand on Best Trial | 64.3 (12.6) | 65.2 (10.7) | 62.2 (11.7) | 58.5 (11.7)* | ||
| No. Taps with Alternating Hands on Best Trial | 79 (19.9) | 81.3 (15.4) | 80.7 (17.2) | 73.2 (17.6) | ||
| Manual motor speed | Purdue Peg Board | Greatest of 3 Trials w. Right Hand | 13.9 (1.9) | 14.1 (1.8) | 14 (1.9) | 13.5 (2.2) |
| Greatest of 3 Trials w. Left Hand | 12.9 (2) | 13 (2.3) | 13.1 (2) | 12.4 (2.3) | ||
| Greatest of 3 Trials w. Both Hands | 10.2 (1.7) | 10.2 (1.4) | 10.4 (1.7) | 9.9 (2) | ||
Notably, those classified as false negatives (n=18), i.e., those predicted not to develop dementia who did in fact develop dementia, at baseline scored better on multiple neuropsychological measures than those classified as both true positives and false positives, including on the Selective Reminding, delayed; BNT, delayed; Face Reproduction Total, RT For Correct Responses and SD for Correct Responses. This false negative group scored more similarly to the true negative group, i.e. those who were predicted to not develop dementia who in fact did not develop dementia, on these tests. Similarly, those classified as false positives (n=113), i.e. those predicted to develop dementia who had not been diagnosed by 2018, scored worse on these same exams than those classified as true negatives and false negatives.
In order to attempt to elucidate cognitive reserve by our model classification, we compared professional education across classification groups. Over 65% of the false positives had no professional education, versus 62% of the true positive group, 39% of those in the false negative group, and 44% of those in the true negative group. Indicating those in the false positive group had the highest proportion with no professional education and those in the false negative group had the lowest proportion with no professional education. However, forcing professional education into the discriminant model did not improve the accuracy of prediction.
1.4. Discussion
In this study, we estimate the incidence of dementia and AD among the unique septuagenarian cohort in the Faroe Islands, 10 years after baseline examination. The overall incidence of all-cause dementia in this population-based study was 1,063 per 100,000 person-years. As expected, the incidence strongly increased with increasing age and was higher among women than men.
The estimated incidence is similar to that seen across Europe, with a recent meta-analysis estimating the incidence rate of AD among those aged over 65, at 1,108 cases per 100,000 person-years (95% CI=1,030-1,189).[21] Variation in incidence rates may reflect differences in the modifiable risk factors between the Faroese and other populations. Cognitive reserve (education and occupation), cardiovascular risk factors, lifestyle and psychosocial factors (depression, physical activity and alcohol consumption), and environmental exposure (air pollution) may have contributed to differences in incidence, though methodological and/or diagnostic issues may also have had an impact. Different genetic substructures, even between European populations, also exist and may influence incidence. While the majority of dementia is late-onset and considered to be multifactorial, there is a strong genetic predisposition. Heritability estimates for late-onset AD for example range from 60 to 80%, and to date beyond APOE, an additional 21 genetic variants have been associated with AD.[22]
Primary prevention of dementia focuses on modifiable risk factors. The ability to isolate high-risk, older individuals prior to impairment is critical for both patients and health care workers to identify factors able to successfully aid in prevention or delay of disease onset. Neuropsychological measures can monitor cognitive status. Our findings suggest that, among septuagenarians without dementia, poorer performance on long-term memory (selective reminding, delayed; Face Recognition, delayed; BNT, delayed), attention (the continuous performance test), and visuospatial perception (NES2, pattern recognition) tasks, along with gender, can be used to isolate a large proportion of these high-risk individuals.
Numerous studies have shown that medial temporal lobe-dependent episodic memory tasks are usually affected earliest in disease development and are quite effective in differentiating between very mildly demented patients with AD and normal older adults.[23–26] Beyond the episodic memory domain, poor verbal fluency performance has also been described as a good predictor of future AD in several population-based studies.[25] Still, even though episodic memory and verbal fluency tasks emerge as the most discriminant measures of AD, most studies find that the pre-dementia stage of AD can involve subtle deficits in a broad range of neuropsychological tests. And our results similarly suggest a combination of cognitive measures likely provides the greatest predictive accuracy, rather than a single task.[25]
Among the tests, those found to be most predictive include the Selective Reminding test,[27,28] a test that is capable of evaluating several facets of learning and memory, including acquisition, storage, retention and retrieval,[29] the Boston Naming test which also measures executive and language functions,[30–32] and the Continuous Performance Test, a measure of attention. Both have previously been shown to be sensitive to AD and discriminate well between cognitively normal individuals and those with dementia.[31] AD patients have also been shown to be impaired on tests of executive functions such as the Raven Progressive Matrices.[33,34] This may reflect deficits in executive functioning and visuospatial abilities that have been hypothesized to reflect AD pathology, especially neurofibrillary tangle burden, in the prefrontal cortex.[24]
Our predictive model, based on a subset of neuropsychological tasks, had an accuracy of 77%. This somewhat improves on the accuracy estimates of predictive models from similar populations using the MMSE alone, such as the Monongahela Valley Independent Elders Survey (MoVIES) study, MMSE AUC=0.726.[35] Other studies have reported higher predictive accuracy, though these studies examined patients who were already symptomatic, thus discriminating not who would develop dementia based on scores prior to diagnosis, but those who had dementia. An important strength of our study is the longitudinal nature, allowing us to examine participants with a full neuropsychological battery many years prior to dementia diagnosis. However, while our predictive model does outperform models with the MMSE alone, the usefulness of employing these neuropsychological tests routinely in clinical practice can be discussed, as they require neuropsychological supervision and are more time-consuming than the MMSE.[36]
While the accuracy of our model was similar or better than previous reports,[35] 23% of participants were misclassified and should be discussed. Examining the baseline characteristics and neuropsychological task scores by classification, the false positive group (n=113; 18%) and false negative group (n=18; 29%) are particularly notable. These groups highlight the importance of regular, continued screening by general practitioners and subsequent referral to the memory clinic. Those individuals classified as false negatives, meaning those persons who would develop dementia but were predicted not to, scored similarly to the true negative, no dementia group, and better on many of the neuropsychological tasks than those correctly predicted to develop dementia. This includes on tasks related to immediate and delayed memory. This false negative group showed minimal or even no impairment on many of the neuropsychological tasks at baseline and may represent individuals with more cognitive reserve but more rapid onset of dementia. Conversely, those participants classified as false positives scored more similarly to the true positive, dementia group, and worse on many of the neuropsychological tasks than those classified as true negatives, no dementia. This false positive group may represent participants who started out with lower cognitive reserve but not necessarily vulnerability to dementia, as demonstrated by the lower level of professional education in this group compared with those in the true negative group.
Thus, when considering the discriminative power of cognitive exams in regard to clinical practice and dementia screening, there is a complex history, involving medical factors, cognitive reserve, early and mid-life events or stresses, and family history, that general practitioners can evaluate with patients beyond neurocognitive exams. This taken with cognitive measures will likely continue to perform better than neurocognitive exams alone in dementia screening. Perhaps the most beneficial inclusion to our model would have been information on APOE status or genetic risk, which may have shed light on the group of susceptible participants who scored highly on cognitive exams at the baseline exam prior to their decline, but then declined functionally relatively rapidly. Unfortunately, this information was unavailable. However, the push toward personalized medicine as well as development of circulating biomarkers of biologic aging, may enhance dementia screening with such individual susceptibilities in the future, both in research and clinical practice.
An important strength of our study is the population-based design, including over 60% of all the eligible septuagenarians on the Faroe Islands, although there were only 65 cases of incident dementia. Still, the diagnostic evaluations of dementia are uniform across the Faroe Islands as there is only one specialized dementia clinic and all diagnoses, except those given abroad, have been established by the same clinician.[15] However, referrals to the Dementia Clinic may vary between GPs and localities, influenced, for example, by clinician knowledge and attitudes to the usefulness of a diagnosis, its requirement for treatment initiation, family attitude and physical distance to the Dementia Clinic. The Faroese septuagenarians who did not participate (n=413), may have been more cognitively impaired, leading to underestimated dementia incidence rates. As previously discussed, a unique strength of our study is the longitudinal nature and the administration of a full neuropsychological battery. This allowed us to investigate not which tests discriminate between dementia patients and non-dementia patients as many previous studies have reported on, but among non-demented participants, which tests are able to distinguish who would later develop dementia in a population setting. However, the limited number of dementia patients did not allow us to pursue the discriminative power of these measures against specific dementia subtypes. Given the clinical and behavioral phenotypes associated with different dementias[37], extension of this research to examine specific subtypes would greatly enhance insights.
Our results demonstrate that among a greater number of tests, covering a broad range of cognitive abilities, low scores on several variables, reflecting verbal and visual learning and recall, vocabulary, visuospatial function, attention and speed of processing as well as manual fine-motor speed, are associated with development of dementia in a subsequent 10-year period from the age of 70 – 74 years. A further discriminant analysis selects a set of six variables in the domains of visuospatial function, attention, and encoding into, and retrieval from, long-term memory with an additional unstable speed of retrieval responses may be helpful in identifying patients in the pre-symptomatic phase of dementia. Thus, helping care-givers identify patients at a higher risk of developing dementia and adjusting management of care accordingly. However, again, the usefulness of employing these rests routinely in clinical practice can be discussed as they require neuropsychological supervision and are more time-consuming than the MMSE, the most commonly used screening instrument for dementia.[36]
Supplementary Material
Acknowledgments
We would like to thank Arni Ludvig psychologist and Hildigunn Steinholm, research coordinator for their contribution at the baseline examinations.
Funding Sources
The study was supported by the Faroese Research Council, National Institute of Health (grant numbers ES013692 & F32ES028087) and the European Commission Sixth Framework Programme for RTD (grant number FOOD-CT-2006-016253, PHIME).
Footnotes
Publisher's Disclaimer: This Author Accepted Manuscript is a PDF file of an unedited peer-reviewed manuscript that has been accepted for publication but has not been copyedited or corrected. The official version of record that is published in the journal is kept up to date and so may therefore differ from this version.
Disclosure of Potential Conflict of Interest: None
Compliance with Ethical Standards
Research involving Human Participants and/or Animals: Yes, human subjects. All procedures have been approved by the Faroese Ethical Committee.
Informed Consent: Informed consent was obtained from all study participants
References
- 1.Wu Y-T, Fratiglioni L, Matthews FE, Lobo A, Breteler MMB, Skoog I, et al. Dementia in western Europe: epidemiological evidence and implications for policy making. Lancet Neurol. 2016;15(1):116–24. [DOI] [PubMed] [Google Scholar]
- 2.Shah H, Albanese E, Duggan C, Rudan I, Langa KM, Carrillo MC, et al. Research priorities to reduce the global burden of dementia by 2025. Lancet Neurol. 2016;15(12):1285–94. [DOI] [PubMed] [Google Scholar]
- 3.Prince M, Bryce R, Albanese E, Wimo A, Ribeiro W, Ferri CP. The global prevalence of dementia: A systematic review and metaanalysis. Alzheimer’s Dement. 2013;9(1):63–75. [DOI] [PubMed] [Google Scholar]
- 4.Guerchet M, Prina M, Prince M. Policy Brief for Heads of Government: The Global Impact of Dementia 2013–2050. Policy Br Heads Gov Glob Impact Dement 2013–2050 Publ by Alzheimer’s Dis Int (ADI), London December 2013. 2013;1–8. [Google Scholar]
- 5.Sposato LA, Kapral MK, Fang J, Gill SS, Hackam DG, Cipriano LE, et al. Declining Incidence of Stroke and Dementia: Coincidence or Prevention Opportunity? JAMA Neurol. 2015;72(12):1529. [DOI] [PubMed] [Google Scholar]
- 6.Ahmadi-Abhari S, Guzman-Castillo M, Bandosz P, Shipley MJ, Muniz-Terrera G, Singh-Manoux A, et al. Temporal trend in dementia incidence since 2002 and projections for prevalence in England and Wales to 2040: Modelling study. BMJ. 2017;358 DOI: 10.1136/bmj.j2856 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Grasset L, Brayne C, Joly P, Jacqmin-Gadda H, Peres K, Foubert-Samier A, et al. Trends in dementia incidence: Evolution over a 10-year period in France. Alzheimer’s Dement. 2016;12(3):272–80. [DOI] [PubMed] [Google Scholar]
- 8.Prince M, Ali GC, Guerchet M, Prina AM, Albanese E, Wu YT. Recent global trends in the prevalence and incidence of dementia, and survival with dementia. Alzheimer’s Res Ther. 2016;8(1). DOI: 10.1186/s13195-016-0188-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kivipelto M, Mangialasche F, Ngandu T. Lifestyle interventions to prevent cognitive impairment, dementia and Alzheimer disease. Nat Rev Neurol. 2018;14(11):653–66. [DOI] [PubMed] [Google Scholar]
- 10.Norton S, Matthews FE, Barnes DE, Yaffe K, Brayne C. Potential for primary prevention of Alzheimer’s disease: An analysis of population-based data. Lancet Neurol. 2014;13(8):788–94. [DOI] [PubMed] [Google Scholar]
- 11.Santos CY, Snyder PJ, Wu WC, Zhang M, Echeverria A, Alber J. Pathophysiologic relationship between Alzheimer’s disease, cerebrovascular disease, and cardiovascular risk: A review and synthesis. Alzheimer’s Dement Diagnosis, Assess Dis Monit. 2017. DOI: 10.1016/j.dadm.2017.01.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Bratzke-Bauer LC, Pozehl BJ, Paul SM, Johnson JK. Neuropsychological patterns differ by type of left ventricle dysfunction in heart failure. Arch Clin Neuropsychol. 2013. DOI: 10.1093/arclin/acs101 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ylikoski R, Ylikoski A, Raininko R, Keskivaara P, Sulkava R, Tilvis R, et al. Cardiovascular diseases, health status, brain imaging findings and neuropsychological functioning in neurologically healthy elderly individuals. Arch Gerontol Geriatr. 2000. DOI: 10.1016/S0167-4943(00)00045-5 [DOI] [PubMed] [Google Scholar]
- 14.Verma M, Howard RJ. Semantic memory and language dysfunction in early Alzheimer’s disease: A review. Int J Geriatr Psychiatry. 2012;27(12):1209–17. [DOI] [PubMed] [Google Scholar]
- 15.Petersen MS, Debes F, Grandjean P, Weihe P. Gender differences in cognitive performance and health status in the Faroese Septuagenarians cohort. Eur J Public Health. 2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Choi AL, Weihe P, Budtz-Jorgensen E, Jorgensen PJ, Salonen JT, Tuomainen TP, et al. Methylmercury exposure and adverse cardiovascular effects in Faroese Whaling men. Environ Health Perspect. 2009. DOI: 10.1289/ehp.11608 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Salonen JT, Seppänen K, Lakka TA, Salonen R, Kaplan GA. Mercury accumulation and accelerated progression of carotid atherosclerosis: A population-based prospective 4-year follow-up study in men in eastern Finland. Atherosclerosis. 2000. DOI: 10.1016/S0021-9150(99)00272-5 [DOI] [PubMed] [Google Scholar]
- 18.Petersen MS, Restorff M, Stora T, Waldemar G, Joensen S. Trend in the Incidence and Prevalence of Dementia in the Faroe Islands. J Alzheimer’s Dis. 2019. DOI: 10.3233/JAD-190341 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Fernandez G Discriminant analysis, a powerful classification technique in data mining. Proceedings of the SAS Users International Conference 2002; pp 247–56. [Google Scholar]
- 20.SAS. SAS/STAT 9.3. SAS Inst Inc, Cary, NC: 2011;8640. [Google Scholar]
- 21.Niu H, Alvarez-Alvarez I, Guillen-Grima F, Aguinaga-Ontoso I. Prevalence and incidence of Alzheimer’s disease in Europe: A meta-analysis. Neurol (English Ed. 2017. DOI: 10.1016/j.nrleng.2016.02.009 [DOI] [PubMed] [Google Scholar]
- 22.Van Cauwenberghe C, Van Broeckhoven C, Sleegers K. The genetic landscape of Alzheimer disease: Clinical implications and perspectives. Genet Med. 2016;18(5):421–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Bondi MW, Jak AJ, Delano-Wood L, Jacobson MW, Delis DC, Salmon DP. Neuropsychological contributions to the early identification of Alzheimer’s disease. Neuropsychol Rev. 2008;18(1):73–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Weintraub S, Wicklund AH, Salmon DP. The neuropsychological profile of Alzheimer disease. Cold Spring Harb Perspect Med. 2012;2(4). DOI: 10.1101/cshperspect.a006171 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Bastin C, Salmon E. Early neuropsychological detection of Alzheimer’s disease. Eur J Clin Nutr. 2014;68(11):1192–9. [DOI] [PubMed] [Google Scholar]
- 26.Belleville S, Fouquet C, Hudon C, Zomahoun HTV, Croteau J. Neuropsychological Measures that Predict Progression from Mild Cognitive Impairment to Alzheimer’s type dementia in Older Adults: a Systematic Review and Meta-Analysis. Neuropsychol Rev. 2017;27(4):328–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Buschke H Selective reminding for analysis of memory and learning. J Verbal Learning Verbal Behav. 1973;12(5):543–50. [Google Scholar]
- 28.BUSCHKE H, FULD PA. Evaluating storage, retention, and retrieval in disordered memory and learning. Neurology. 1974;24(11):1019–1019. [DOI] [PubMed] [Google Scholar]
- 29.Spreen O, Strauss E. A compendium of neuropsychological tests: Administration, norms, and commentary, 2nd ed 1998. [Google Scholar]
- 30.Goodglass H WS E. The Boston Naming Test. Philadelphia Lea Febiger. 1983 [Google Scholar]
- 31.Williams BW, Mack W, Henderson VW. Boston naming test in Alzheimer’s disease. Neuropsycho logia. 1989;27(8):1073–9. [DOI] [PubMed] [Google Scholar]
- 32.Katsumata Y, Mathews M, Abner EL, Jicha GA, Caban-Holt A, Smith CD, et al. Assessing the discriminant ability, reliability, and comparability of multiple short forms of the boston naming test in an Alzheimer’s disease center cohort. Dement Geriatr Cogn Disord. 2015;39(3–4):215–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Raven J, Raven JC, Court J. Manual for Raven’s progressive matrices and vocabulary scales. 1998. DOI: 10.1006/cogp.1999.0735 [DOI] [Google Scholar]
- 34.A FI, I A, R B, C S, C G, I L, et al. Qualitative patterns at Raven’s colored progressive matrices in mild cognitive impairment and Alzheimer’s disease. Aging Clin Exp Res. 2016;28(3):561–5. [DOI] [PubMed] [Google Scholar]
- 35.Chen P, Ratcliff G, Belle SH, Cauley JA, DeKosky ST, Ganguli M. Cognitive tests that best discriminate between presymptomatic AD and those who remain nondemented. Neurology. 2000. DOI: 10.1212/WNL.55.12.1847 [DOI] [PubMed] [Google Scholar]
- 36.Roebuck-Spencer TM, Glen T, Puente AE, Denney RL, Ruff RM, Hostetter G, et al. Cognitive Screening Tests Versus Comprehensive Neuropsychological Test Batteries: A National Academy of Neuropsychology Education Paper. Arch Clin Neuropsychol. 2017. DOI: 10.1093/arclin/acx021 [DOI] [PubMed] [Google Scholar]
- 37.Reul S, Lohmann H, Wiendl H, Duning T, Johnen A. Can cognitive assessment really discriminate early stages of Alzheimer’s and behavioural variant frontotemporal dementia at initial clinical presentation? Alzheimer’s Res Ther. 2017. DOI: 10.1186/s13195-017-0287-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.