Abstract
Introduction:
Commonly occurring dementias include those of Alzheimer’s, vascular, and mixtures of these and other pathologies. They are believed to evolve over many years, but that time interval has been difficult to establish. Our objective is to determine how many years in advance of a dementia diagnosis cognitive scores begin to change.
Methods:
14,086 dementia-free ARIC participants underwent a cognitive exam at baseline visit 2 (1990–1992, mean age 57±5.72), and in 11,244 at visit 4 (1996–1998), 5,640 at visit 5 (2011–2013), and 3,574 at visit 6 (2016–2017) with surveillance for dementias of all causes combined. Within 5-year intervals after each visit, we compared performance on the Delayed Word Recall Test (DWRT), the Digit Symbol Substitution Test (DSST), the Word Fluency Test (WFT), and the combined mean of three cognitive tests at baseline in participants who were diagnosed with dementia within each interval versus those who survived the interval without a dementia diagnosis. Z-scores were adjusted for demographics and education in separate regression models for each visit. We plotted adjusted z-score means by time interval following each visit.
Results:
During follow-up 3,334, 2,821, 1,218, and 329 dementia cases were ascertained after visits 2, 4, 5, and 6 respectively. Adjusted DWRT z-scores were significantly lower 20–25 years before dementia than those who did not experience dementia within 25 years. DSST z-scores were significantly lower at 25–30 years and 3-test combination z-scores were significantly lower as early as 30–31 years before onset. The difference between dementia and non-dementia group in the visit 2 3-test combination z-score was −0.20 at 30–31 years prior to dementia diagnosis. As expected, differences between the dementia and non-dementia groups increased closer to the time of dementia occurrence, up to their widest point at 0–5 years prior to dementia diagnosis. The difference between dementia and non-dementia groups in the visit 2 3-test combination z-score was −0.90. WFT z-score differences were smaller than for the DSST or DWRT and began later. Patterns were similar in Black and White participants.
Conclusion:
DWRT, DSST and combined 3-test z-scores were significantly lower more than 20 years prior to diagnosis in the dementia group versus the non-dementia group. Findings contribute to our knowledge of the long prodromal period in Blacks and Whites.
Keywords: cognitive decline, dementia, cognitive test score
Introduction:
Previous studies have demonstrated that cognitive test scores start changing years before dementia develops. Most studies tested at only a single time point. [1–6] One publication examined the relationship between trends in cognitive test scores and dementia hospitalization, using data from the Atherosclerosis Risk in Communities Study (ARIC). In this community-based population, dementia incidence was more strongly associated with six-year decline in cognitive test scores than with scores measured only once. [7] That study encompassed a mean of 11 years of follow up, but the ARIC Neurocognitive Study (ARIC NCS) now includes over 30 years of follow up with large enough numbers for race-specific analysis. Our objective is to explore the relationship of cognitive test scores in 5-year intervals prior to dementia diagnosis studied over 30+ years to determine when cognitive changes are first detectable. If the results establish that cognitive changes occur long before dementia is recognized clinically, perhaps earlier than formerly known, this would help to validate associations reported of dementia with risk factor levels measured in middle-age or earlier and to support calls for developing longer-term strategies for dementia prevention.
Methods:
Study population
The Atherosclerosis Risk in Communities (ARIC) Study [8] included 15,792 middle-aged adults from four U.S. communities (Jackson, Mississippi; Washington County, Maryland; selected suburbs of Minneapolis, Minnesota; and Forsyth County, North Carolina). Because of the small numbers of Blacks in the Maryland and Minnesota sites and those who were of neither of White nor Black ethnicity in all centers, members of these groups (n=90) were excluded from our study population. Too few non-White participants were recruited in Washington County or Minneapolis for statistical modeling, so they and other race groups were excluded from the study population (n=90). The first cohort visit occurred in 1987–1989. During this visit, extensive examinations were conducted, a physical examination and collection of information on demographics, medical history, medication use, and health behaviors[9]. Cognitive testing was first done at visit 2 and included 14,348 study population participants between the ages of 46–70 during 1990–1992. Subsequent cognitive testing was done at visit 4 (1996–1998 with 11,656 participants aged 52–75); visit 5 (2011–2013 with 6,538 participants aged 66–90) and visit 6 (2016–2017 with 4,003 participants aged 66–90) (Supplement Table 1). The longer interval between visits 4 and 5 was the result of limited funding for ARIC during that time. Supplemental Table 1 shows how participants were selected for follow up after each visit. It may be worth noting that because persons may not have completed cognitive testing at every visit, some persons eligible at a specific visit may not have been eligible at an earlier visit.
Cognitive Tests
At each of these visits: the Delayed Word Recall Test (DWRT), the Digit Symbol Substitution Test (DSST), and the Word Fluency Test (WFT) were administered [10–12]. The three tests reflect somewhat different cognitive abilities. The Delayed Word Recall Test (DWRT) is a measure of an individual’s episodic memory, and the ability to learn, store, and retrieve information. The Digit Symbol Substitution Test (DSST) gauges processing speed, attention, and working memory. And Word Fluency Test (WFT) tests semantic memory and executive function. We also use the combination scores of these three tests to more closely represent a generalized level of cognitive performance as in previous ARIC reports [7]. We created a z-score for each test, calculated as the difference between the score and its visit 2 population mean and then dividing by its visit 2 population standard deviation. We also created a z-score combination of three tests z-score calculated by summing the mean of the DWRT, the DSST, and the WFT z scores, also standardized to visit 2 by subtracting the mean of the combined z score and then dividing by the standard deviation of the combined score, so that all metrics have a mean of 0 and a standard deviation of 1 at visit 2 [13].
Outcome
The ARIC definition of dementia is concordant with criteria of the National Institute of Aging, Alzheimer Association (NIA-AA) and Diagnostic and Statistical Manual for Mental Disorders, 5th Edition (DSM-5) [13]. For participants diagnosed at an ARIC examination, the initial diagnosis used an algorithm based on either a low MMSE score or failure on tests of more than one cognitive domain plus documented prior cognitive decline. The final diagnosis came from a panel of expert reviewers examining the data. Diagnoses were also made in non-examined participants, and these were based on semi-annual surveillance of the entire cohort using the AD8 (AD8 Dementia Screening interview), the SIS (Six Item Screener) and hospital and death certificate data [13].
Covariates
The comparisons were adjusted for age (years), gender, race-center (Washington-White, Forsyth-White, Forsyth Black, Jackson-Black, and Minneapolis-White), education (< high school; high school graduate or equivalent; college or professional school).
Analysis
Analysis was designed to compare cognitive test scores at fixed intervals prior to dementia in two groups – participants diagnosed with dementia within that interval compared with those who survived that interval without dementia. Five-year intervals were used, so that we could compare scores of participants who did or did not develop dementia within 0–5 years prior to dementia diagnosis, 5–10 years prior, 10–15 years prior, and so on. For visit 2, in the dementia group, we included all dementia-free participants who were diagnosed in the first five years after they were tested and then calculated the mean of their visit 2 test scores and 95% confidence interval. In the next time interval, we included all the 5-year dementia-free survivors who were first diagnosed with dementia 5–10 years after completing visit 2 tests, and so on, up to 25–30 and 30–31 years prior to dementia (the longest follow up was 30.9 years). For non-cases, defined as survivors free of dementia at the end of each time interval, we also calculated for each interval the means of the visit 2 test scores and their 95% confidence intervals. We performed the Welch Two Sample T-test to test the difference between each pair (non-dementia versus dementia groups) within each time interval. Calculations were based on visit 4, 5 and 6 test scores for the data based on these visits, though the final interval was only 20–25 years prior to dementia for the visit 4 score comparisons, and correspondingly shorter for test scores obtained at visits 5 and 6.
We plotted the unadjusted and adjusted z-score means (for DSST, DWRT, WFT, and the 3-test combined z-score) for visits 2, 4, 5 and 6 for both the dementia and non-dementia groups, with time shown on the x-axis, and mean test scores shown on the Y axis within each time interval. We connected the corresponding pair of points with vertical lines from the same visit at each time interval to represent the difference between groups.
The adjustment model included age, gender, race-center and education level. Washington County, Maryland white females with less than high school education level comprised the reference group. We fitted a separate linear regression model for each visit to calculate individual residuals and added the intercept to the average of all individual residuals in each time interval by dementia status to get the adjusted z-scores for each test, interval, and visit. Results of different visits of the same test were shown in one graph. Subgroup analysis for Blacks and Whites used the same methods.
Analyses were performed using R Version 4.1.1[13]. A two-sided p-value <0.05 was considered statistically significant.
Results
Out of the 15,305 study participants who were alive at the end of visit 2, 14,348 were seen at examination. After excluding participants who received a dementia diagnosis on or before their visit date, we were left with 14,086 participants, who were dementia-free and had complete cognitive test scores, for further analysis. Corresponding data for visit 4, 5 and 6 are shown in the Supplement table. Baseline characteristics of the study participants at visits 2, 4, 5 and 6 stratified by incident dementia status are shown in Table 1.
Table 1:
Characteristics of ARIC participants free of dementia at visits 2, 4, 5 and 6 stratified by incident dementia occurring over the entire follow up period.
| Visit 2 (1990–1992) |
Visit 4 (1996–1998) |
Visit 5 (2011–2013) |
Visit 6 (2016–2017) |
|||||
|---|---|---|---|---|---|---|---|---|
| Developed dementia after this visit? | No | Yes | No | Yes | No | Yes | No | Yes |
| n | 10752 | 3334 | 8423 | 2821 | 4422 | 1218 | 3245 | 329 |
| Age, years, (mean (SD)) | 56.4 (5.7) | 59.1 (5.4) | 62.0 (5.6) | 65.2 (5.3) | 74.6 (4.7) | 78.4 (5.4) | 79.0 (4.53) | 81.8 (5.3) |
| Male, n (%) | 4933 (45.9) | 1331 (39.9) | 3799 (45.1) | 1147 (40.7) | 1768 (40.0) | 503 (41.3) | 1298 (40.0) | 141 (42.9) |
| Race-center, n (%) | ||||||||
| Washington-White | 2724 (25.3) | 907 (27.2) | 2303 (27.3) | 831 (29.5) | 1174 (26.5) | 346 (28.4) | 812 (25.0) | 92 (28.0) |
| Forsyth-White | 2561 (23.8) | 698 (20.9) | 1961 (23.3) | 584 (20.7) | 958 (21.7) | 205 (16.8) | 747 (23.0) | 56 (17.0) |
| Jackson-Black | 2179 (20.3) | 856 (25.7) | 1471 (17.5) | 627 (22.2) | 835 (18.9) | 332 (27.3) | 653 (20.1) | 96 (29.2) |
| Minneapolis-White | 2998 (27.9) | 787 (23.6) | 2498 (29.7) | 716 (25.4) | 1387 (31.4) | 318 (26.1) | 965 (29.7) | 80 (24.3) |
| Forsyth-Black | 290 (2.7) | 86 (2.6) | 190 (2.3) | 63 (2.2) | 68 (1.5) | 17 (1.4) | 68 (2.1) | 5 (1.5) |
| Current cigarette smoker, n (%) | ||||||||
| 2534 (23.6) | 626 (18.8) | 1286 (15.3) | 361 (12.8) | 240 (5.6) | 72 (6.1) | 214 (7.2) | 30 (10.0) | |
| Current alcohol consumption, n (%) | ||||||||
| 6267 (58.3) | 1684 (50.6) | 4375 (52.0) | 1195 (42.4) | 2255 (53.1) | 425 (38.0) | 1648 (51.8) | 141 (43.8) | |
| Body mass index, Kg/㎡ (mean (SD)) |
27.92 (5.42) | 28.16 (5.39) | 28.80 (5.60) | 28.83 (5.65) | 28.98 (5.64) | 28.51 (6.02) | 28.45 (5.43) | 27.83 (5.59) |
| Total cholesterol, mmol/L (mean (SD)) |
5.40 (1.02) | 5.53 (1.03) | 5.17 (0.96) | 5.25 (0.95) | 4.72 (1.06) | 4.64 (1.10) | 4.52 (1.03) | 4.42 (1.13) |
| Diastolic blood pressure>90) | 3088 (28.8) | 1088 (32.7) | 3104 (37.0) | 1159 (41.3) | 2928 (66.7) | 864 (72.4) | 2228 (70.4) | 233 (74.7) |
| Systolic blood pressure>140) | 3722 (34.7) | 1315 (39.6) | 3857 (45.9) | 1454 (51.8) | 3214 (73.1) | 946 (79.0) | 2526 (79.2) | 270 (84.9) |
| Diabetes, n (%) | 1202 (11.2) | 430 (12.9) | 1328 (15.8) | 516 (18.4) | 1333 (30.8) | 444 (38.3) | 1021 (32.7) | 137 (45.2) |
| Cognitive test z-score (mean (SD)) | ||||||||
| Delayed word recall | 0.04 (0.99) | −0.14 (1.02) | 0.04 (1.01) | −0.27 (1.12) | −0.60 (1.03) | −1.53 (1.22) | −0.74 (1.04) | −1.73 (1.23) |
| Digit symbol substitution | 0.06 (0.99) | −0.20 (0.99) | 0.01 (0.93) | −0.32 (0.95) | −0.28 (0.79) | −0.90 (0.75) | −0.44 (0.78) | −1.13 (0.70) |
| Word fluency test | 0.03 (0.99) | −0.10 (1.01) | 0.06 (0.99) | −0.11 (1.02) | 0.12 (0.95) | −0.32 (0.96) | 0.03 (0.95) | −0.47 (0.95) |
| 3-test combination | 0.05 (0.77) | −0.15 (0.78) | 0.04 (0.75) | −0.23 (0.79) | −0.25 (0.69) | −0.89 (0.70) | −0.38 (0.68) | −1.09 (0.64) |
| Education level, n (%) | ||||||||
| <High school | 2132 (19.9) | 905 (27.2) | 1391 (16.5) | 694 (24.6) | 507 (11.5) | 255 (20.9) | 324 (10.0) | 64 (19.5) |
| High school graduate or equivalent | 4491 (41.8) | 1363 (40.9) | 3586 (42.6) | 1180 (41.9) | 1852 (42.0) | 528 (43.3) | 1338 (41.3) | 133 (40.4) |
| College, graduate, or higher | 4113 (38.3) | 1062 (31.9) | 3433 (40.8) | 943 (33.5) | 2055 (46.6) | 435 (35.7) | 1576 (48.7) | 132 (40.1) |
|
| ||||||||
Out of 14,086 participants at visit 2, 3,334 (23.7%) developed dementia by the end of follow up. Following visit 4, 2,821 of the 11,244 participants developed dementia. After visit 5, 1,218 of the 5,640 remaining participants developed dementia. Finally, of the 3,574 participants at visit 6, 329 developed dementia. At each visit, the incident dementia group was older than the non-dementia group. For example, in Visit 2, the average age of the dementia group was 59.1 years, vs. 56.4 years for the non-dementia group. They were also more likely to be Black from Jackson. Likewise, participants who later developed dementia were more likely to have diastolic blood pressure greater than 90mm Hg (at visit 2, 32.7% vs.28.8%; at visit 4, 41.3% vs. 37.0%, etc.) and diabetes (at visit 2, 12.9% vs.11.2%; at visit 4 18.4% vs. 15.8%, etc.) than those without dementia. Participants with less than high school education levels were more likely to develop dementia than those with higher education levels. Cognitive test scores declined over time for both the dementia and the non-dementia group at the later visits. There were significant differences in scores on each test between the dementia and non-dementia groups at each of the baseline visits (p<0.001).
Mean unadjusted DSST scores were higher for the non-dementia than the dementia group at all intervals examined prior to dementia (figure 1A, vertical lines). The differences in scores were larger at times closer to the dementia diagnosis than at earlier dates more remote from the date of diagnosis. The difference between the two group was statistically significant even at the longest interval prior to dementia (30–31 years, p<0.05). Adjustment for age, gender, education level and race-center narrowed the dementia-non-dementia differences (figure 1B), but the differences at 25–30 years, at 20–25 years all intervals closer to dementia remained significant. Differences at intervals of 20–25, 35–30 and 30–31 years were small, but at the 10–15-year interval prior to dementia, the visit 2 scores were −0.21 vs. 0.02, and the difference was larger (−0.23 [−0.35, −0.13]), and visit 4 scores were −0.25 vs. 0.03 (difference: −0.28 [−0.28, −0.17], Table 2). Differences between the dementia and non-dementia groups increased to their widest point at 0–5 years prior to diagnosis.
Figure 1:
Cognitive tests at visits 2, 4, 5, and 6 by dementia status and 5-year interval before dementia diagnosis. Mean Z-scores are for Digit Symbol Substitution test (DSST, panel a with adjustment in panel b), Delayed Word Recall Test (DWRT, panel c with adjustment in panel d) and Word Fluency Test (WFT, panel e with adjustment in panel f). Adjustment is to mean age, female sex, Washington County, Maryland white race, and less than high-school education using linear regression within each visit (mean age was 57.0, 62.8, 75.4 and 79.3 years at visits 2,4,5 and 6, respectively). Vertical lines show the difference associated with dementia incidence (all p-values <0.05 except visit 2 DWRT at 30–31 years; WFT at 5–10 and 30–31 years; adjusted WFT at 15–31 years; adjusted DWRT at 25–31 years; adjusted DSST at 30–31 years and adjusted visit 4 WFT at 15–25 years).
Table 2:
Z-score differences for tests for the 10–15 year and 0–5 year follow up intervals
| 10–15 Years Adjusted Cognitive Testing Scores | 0–5 Years Adjusted Cognitive Testing Scores | |||
|---|---|---|---|---|
|
| ||||
| Mean in Dementia vs. Non-Dementia Groups | Mean Difference [95% CI] Dementia minus Non-Dementia Group | Mean in Dementia vs. Non-Dementia Groups | Mean Difference [95% CI] Dementia minus Non-Dementia Group | |
| Visit 2: Both Blacks and Whites DSST |
−0.21 vs. 0.02 | −0.23 [−0.35, −0.13] | −0.71 vs.0.00 | −0.71 [−1.10, −0.32] |
| DWRT | −0.21 vs. 0.03 | −0.24 [−0.38, −0.10] | −1.45 vs. 0.00 | −1.45 [−2.13, −0.77] |
| WFT | −0.16 vs. 0.01 | −0.17 [−0.29, −0.05] | −0.51 vs. 0.00 | −0.51 [−0.87, −0.14] |
| 3−Test Z | −0.19 vs. 0.02 | −0.21 [−0.30, −0.13] | −0.90 vs. 0.00 | −0.90 [−1.31, −0.49] |
| Blacks:3-Test Z | −0.72 vs. −0.52 | −0.20 [−0.32, −0.07] | −1.51 vs. −0.56 | −0.95 [−1.43, −0.47] |
| Whites:3-Test Z | 0.02 vs. 0.20 | −0.18 [−0.28, −0.07] | −0.39 vs. 0.18 | −0.57 [−1.00, −0.13] |
| Visit 4: Both Blacks and Whites DSST |
−0.25 vs. 0.03 | −0.28 [−0.28, −0.17] | −0.83 vs.−0.07 | −0.76 [−0.93, −0.59] |
| DWRT | −0.25 vs. 0.02 | −0.27 [−0.36, −0.17] | −1.11 vs. −0.02 | −1.09 [−1.41, −0.77] |
| WFT | −0.10 vs. 0.05 | −0.15 [−0.22, −0.07] | −0.35 vs. 0.02 | −0.37 [−0.57, −0.17] |
| 3-Test Z | −0.20 vs. 0.01 | −0.21 [−0.27, −0.15] | −0.77 vs. −0.02 | −0.75 [−0.91, −0.58] |
| Blacks:3-Test Z | −0.92 vs. −0.45 | −0.47 [−0.65, −0.28] | −1.18 vs. −0.49 | −0.69 [−1.05, −0.33] |
| Whites:3-Test Z | −0.08 vs. 0.15 | −0.23 [−0.29, −0.17] | −0.57 vs. 0.13 | −0.70 [−0.87, −0.52] |
| Visit 5: Both Blacks and Whites DSST |
/ | / | −0.75 vs.−0.37 | −0.38 [−0.44, −0.33] |
| DWRT | / | / | −1.49 vs. −0.73 | −0.76 [−0.87, −0.64] |
| WFT | / | / | −0.27 vs. 0.06 | −0.33 [−0.41, −0.25] |
| 3-Test Z | / | / | −0.83 vs. −0.34 | −0.49 [−0.56, −0.43] |
| Blacks:3-Test Z | / | / | −1.10 vs. −0.44 | −0.66 [−0.54, −0.77] |
| Whites:3-Test Z | / | / | −0.77 vs. −0.24 | −0.53 [−0.60, −0.46] |
| Visit 6: Both Blacks and Whites DSST |
/ | / | −0.91 vs. −0.46 | −0.45 [−0.52, −0.38] |
| DWRT | / | / | −1.55 vs. −0.76 | −0.79 [−0.93, −0.65] |
| WFT | / | / | −0.31 vs. 0.01 | −0.32 [−0.42, −0.22] |
| 3-Test Z | / | / | −0.92 vs. −0.39 | −0.53 [−0.60, −0.46] |
| Blacks:3-Test Z | / | / | −1.24 vs. −0.72 | −0.52 [−0.64, −0.39] |
| Whites:3-Test Z | / | / | −0.84 vs. −0.36 | −0.48 [−0.56, −0.39] |
Note: Visit 5 and 6 do not have enough long follow-up for the 10–15 year interval.
Unadjusted DWRT scores were also consistently higher in the non-dementia than in the dementia group (Figure 1). After adjustment (Figure 1D), the differences between the two groups were small at intervals of more than 20 years prior to dementia. Differences were not significant at the 31–30 or 25–30 year intervals but became significant at the 20–25-year interval. Differences were larger at intervals more proximal to dementia diagnosis.
The mean WFT scores of the non-dementia group were consistently higher than for the dementia group, but differences were smaller than for DWRT or DSST (Figure 1). Adjustment narrowed the differences (figure 1F). Although WFT test results first showed statistically significant differences at the 10–15-year interval, the differences fluctuated over the intervals studied. The differences between the scores for the dementia and non-dementia groups were statistically significant at all baseline visits only at the 0–5-year interval.
Supplement figures 2A and 2B show mean z-scores for the 3-test combination. Even after adjustment, scores were consistently higher for the non-dementia versus the dementia group, and they were statistically significant even at the 30–31-year interval prior to dementia, e.g., for visit 2 the difference between was −0.20 [−0.39, −0.02]). Over the 20 to 31-year intervals, differences were small, but they became substantial at the 15–20-year interval prior to dementia onset and much larger at the 0–5-year interval (Table 2). For visit 2 the z-score for the difference between dementia and non-dementia group was −0.90 [−1.31, −0.49].
Patterns in the difference between those with and without incident dementia were similar in Blacks and Whites. While Black participants had consistently lower adjusted 3-test Z scores than Whites, the scores were lower in the incident dementia group than in the non-dementia group for all time intervals in both Blacks and Whites, with very similar time-to-dementia patterns (Supplement figure 1). The magnitude of dementia vs non-dementia differences in scores were similar in Blacks and Whites and confidence intervals overlapped widely (Table 2).
Discussion
Although adjusted differences between persons with and without dementia were statistically significant early on, at 25–30 years or more prior to dementia, the magnitudes of differences at intervals remote from dementia dates were small. Cognitive test scores would hardly succeed as markers of dementia risk 20 or more years later. However, we believe the study provides clear evidence of the very long prodromal period with respect to cognitive performance for dementia, all-causes combined, in a general population. We found that differences in adjusted DWRT and DSST scores were associated with dementia incidence more than 20 years later. Differences in WFT z-scores were smaller. When the three tests were combined into a single adjusted score, differences also first appeared more than 20 years prior to diagnosis. Patterns were similar in Blacks and Whites.
Understanding the long premonitory phase of dementia is important in supporting claims about risk factors for dementia that are identified in mid-life. Our data support ongoing population efforts to change lifestyles, e.g., reducing sedentary behavior and smoking and making dietary changes [15]. Past research has demonstrated that physical activity can help improve cognitive function and reduce and delay progression to dementia [16]. In theory, such changes might lessen cognitive impairments due to cerebrovascular changes. There is less evidence regarding prevention of neurodegenerative diseases (e.g., Alzheimer’s disease, characterized by the accumulation of amyloid beta in the form of extracellular plaques and intracellular neurofibrillary tangles) [17].
Our results extend a previous study using ARIC data which looked at associations between cognitive tests and dementia over a mean of 11 years. With fewer dementia cases, that report could not detect significant unadjusted differences in scores between participants with or without dementia more than 7 years prior to dementia diagnosis, and it could not examine race group separately [7]. In the present study scores were adjusted for education and demographic factors, reducing confounding, but resulting in an attenuation of observed dementia-non-dementia differences. But despite the attenuation, our longer follow-up period with many more dementia cases allowed us to show significant differences at most intervals studied.
A French study of 442 participants with Alzheimer’s disease and matched nested controls with 20-year follow up demonstrated that in a higher-educated group, cognitive decline occurred 15–16 years prior to dementia with accelerated decline 7 years before dementia [18], similar to the pattern we report here. In a smaller less-educated group, the dementia cases showed cognitive scores significantly lower than those of controls only 7 years prior to dementia. Another study with cases limited to dementia of presumed Alzheimer’s disease completed a mean of 7-years of annual evaluations with a few participants followed up to 16 years found that cognitive decline accelerates during the 5 years prior to dementia due to Alzheimer’s disease [18]. Authors did not compare this trajectory against non-dementia controls.
An important limitation of our study is the fact we did not have data with which to separate specific causes of dementia, such as vascular or Alzheimer’s disease; we use the well-defined outcome of all-cause dementia. It is difficult to know whether test differences would show similar time associations for separate dementia types. Also, use of dementia diagnosis as the outcome date does not accurately reflect the date when dementia might be first ascertained.
Study attrition occurred over time, particularly in the 13-year interval between visits 4 and 5, and we have not adjusted for potential informative censoring. We believe there would be a higher dementia incidence among individuals who may have escaped our surveillance, which included hospital and death certificate codes even among non-examined persons. These methods maximized our capture of cases compared with ascertainment limited to examined cases alone. However, we cannot identify those who may have escaped the surveillance and our ascertainment is not likely to be 100%.
In summary, with over 30 years of follow-up of large study population first tested at age of 48–67 years, and use of comparable individuals who did not go on to develop dementia, we were able to demonstrate rapid decline in population-based dementia cases beginning at least 10 years prior to dementia ascertainment and statistically significant differences in test scores from controls even 25–30 years prior to the dementia, with similar finding in Blacks and Whites from different communities. These results quantify the long prodromal period of cognitive decline prior to dementia and suggest the need to understand this period better.
Supplementary Material
Figure 2:
Cognitive tests at visits 2, 4, 5, and 6 by dementia status and 5-year interval before dementia diagnosis. Mean Z-scores are for 3-test combination (panel a with adjustment in panel b). Adjustment is to age, female gender, Washington County white race-center and less than high-school education using linear regression within each visit (mean age was 57.0, 62.8, 75.4 and 79.3 years at visits 2,4,5 and 6). Vertical lines show the difference associated with dementia incidence (all p-values <0.05).
Funding Sources
The Atherosclerosis Risk in Communities Study is conducted as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700005I, and HHSN268201700004I). Neurocognitive data are collected by U01 2U01HL096812, 2U01HL096814, 2U01HL096899, 2U01HL096902, and 2U01HL096917 from the NIH (NHLBI, NINDS, NIA, and NIDCD). And with previous brain MRI examinations funded by R01-HL70825 and biomarkers by R01-HL134320 from the NHLBI.
Statements
Statement of Ethics
The subjects have given their written informed consent, and the study protocol was approved by the Institute’s Committee on Human Research (IRB00012998).
Footnotes
Conflict of Interest Statement
No conflicts of interest.
Data Availability Statement
The ARIC data are not publicly available due to confidentiality issues. Investigators can access data from the ARIC study by submitting a manuscript proposal to aricpub@unc.edu.
Reference
- 1.Chen P, Ratcliff G, Belle SH, Cauley JA, DeKosky ST, Ganguli M. Patterns of cognitive decline in presymptomatic Alzheimer disease: a prospective community study. Archives of general psychiatry. 2001. Sep 1;58(9):853–8. [DOI] [PubMed] [Google Scholar]
- 2.Alonso A, Mosley TH, Gottesman RF, Catellier D, Sharrett AR, Coresh J. Risk of dementia hospitalisation associated with cardiovascular risk factors in midlife and older age: the Atherosclerosis Risk in Communities (ARIC) study. Journal of Neurology, Neurosurgery & Psychiatry. 2009. Nov 1;80(11):1194–201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bäckman L, Small BJ, Fratiglioni L. Stability of the preclinical episodic memory deficit in Alzheimer’s disease. Brain. 2001. Jan 1;124(1):96–102. [DOI] [PubMed] [Google Scholar]
- 4.Amieva H, Jacqmin-Gadda H, Orgogozo JM, Le Carret N, Helmer C, Letenneur L, Barberger-Gateau P, Fabrigoule C, Dartigues JF. The 9 year cognitive decline before dementia of the Alzheimer type: a prospective population-based study. Brain. 2005. May 1;128(5):1093–101. [DOI] [PubMed] [Google Scholar]
- 5.Elias MF, Beiser A, Wolf PA, Au R, White RF, D’Agostino RB. The preclinical phase of Alzheimer disease: a 22-year prospective study of the Framingham Cohort. Archives of neurology. 2000. Jun 1;57(6):808–13. [DOI] [PubMed] [Google Scholar]
- 6.Riley KP, Snowdon DA, Desrosiers MF, Markesbery WR. Early life linguistic ability, late life cognitive function, and neuropathology: findings from the Nun Study. Neurobiology of aging. 2005. Mar 1;26(3):341–7. [DOI] [PubMed] [Google Scholar]
- 7.Schneider AL, Gottesman RF, Mosley T, Alonso A, Knopman DS, Coresh J, Sharrett AR, Selvin E. Cognition and incident dementia hospitalization: results from the atherosclerosis risk in communities study. Neuroepidemiology. 2013;40(2):117–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Wright JD, Folsom AR, Coresh J, Sharrett AR, Couper D, Wagenknecht LE, Mosley TH Jr, Ballantyne CM, Boerwinkle EA, Rosamond WD, Heiss G. The ARIC (Atherosclerosis Risk in Communities) study: JACC focus seminar 3/8. Journal of the American College of Cardiology. 2021. Jun 15;77(23):2939–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Forms and Form Codebooks [Internet]. [cited 2023 Jun 23]. Available from: https://sites.cscc.unc.edu/aric/cohort-forms-forms?order=field_view_pub1&sort=desc
- 10.Knopman DS, Ryberg S. A verbal memory test with high predictive accuracy for dementia of the Alzheimer type. Archives of neurology. 1989. Feb 1;46(2):141–5. [DOI] [PubMed] [Google Scholar]
- 11.Wechsler D. WAIS-R manual: Wechsler adult intelligence scale-revised. Psychological Corporation; 1981. [Google Scholar]
- 12.Benton AL. Development of a multilingual aphasia battery: Progress and problems. Journal of the neurological sciences. 1969. Jul 1;9(1):39–48. [DOI] [PubMed] [Google Scholar]
- 13.Knopman DS, Gottesman RF, Sharrett AR, Wruck LM, Windham BG, Coker L, et al. Mild Cognitive Impairment and Dementia Prevalence: The Atherosclerosis Risk in Communities Neurocognitive Study (ARIC-NCS). Alzheimers Dement (Amst). 2016;2:1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.R version 4.1.1 (2021–08-10) Copyright(C) 2021 The R Foundation for Statistical Computing
- 15.Kivipelto M, Mangialasche F, Snyder HM, Allegri R, Andrieu S, Arai H, Baker L, Belleville S, Brodaty H, Brucki SM, Calandri I. World-Wide FINGERS Network: a global approach to risk reduction and prevention of dementia. Alzheimer’s & Dementia. 2020. Jul;16(7):1078–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Nimmo MA, Leggate M, Viana JL, King JA. The effect of physical activity on mediators of inflammation. Diabetes, Obesity and Metabolism. 2013. Sep;15(s3):51–60. [DOI] [PubMed] [Google Scholar]
- 17.Joe E, Ringman JM. Cognitive symptoms of Alzheimer’s disease: clinical management and prevention. bmj. 2019. Dec 6;367. [DOI] [PubMed] [Google Scholar]
- 18.Amieva H, Mokri H, Le Goff M, Meillon C, Jacqmin-Gadda H, Foubert-Samier A, Orgogozo JM, Stern Y, Dartigues JF. Compensatory mechanisms in higher-educated subjects with Alzheimer’s disease: a study of 20 years of cognitive decline. Brain. 2014. Apr 1;137(4):1167–75. [DOI] [PubMed] [Google Scholar]
- 19.Wilson RS, Leurgans SE, Boyle PA, Bennett DA. Cognitive decline in prodromal Alzheimer disease and mild cognitive impairment. Archives of neurology. 2011. Mar 14;68(3):351–6. [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.
Supplementary Materials
Data Availability Statement
The ARIC data are not publicly available due to confidentiality issues. Investigators can access data from the ARIC study by submitting a manuscript proposal to aricpub@unc.edu.