Conceived and designed the experiments: DB KA NC HC PSS. Analyzed the data: DB. Contributed reagents/materials/analysis tools: DB KA RB WW. Wrote the paper: DB KA NC RB HC WW PSS.
The authors have declared that no competing interests exist.
The association between brain white matter lesions and cognitive impairment in old age is well established. However, little is known about this association in midlife. As this information will inform policy for early preventative healthcare initiatives, we investigated non-periventricular frontal, temporal, parietal and occipital lobe white matter hyperintensities (WMH) in relation to cognitive function in 428 (232 women) community-dwelling adults aged 44 to 48 years.
Frontal white matter lesions were significantly associated with greater intraindividual RT variability in women, while temporal WMH were associated with face recognition deficits in men. Parietal and occipital lobe lesions were unrelated to cognitive performance. These findings did not differ when education and a range of health variables, including vascular risk factors, were taken into account.
Gender differences in WMH-cognition associations are discussed, and we conclude that small vessel disease is present in midlife and has functional consequences which are generally not recognized. Preventative strategies should, therefore, begin early in life.
There is substantial evidence that subcortical white matter lesions are associated with cognitive deficits
Our particular focus was on white matter hyperintensities (WMH). WMH refer to white matter lesions that appear as high signal intensities on T2-weighted MRI. Their neuropathological origins are wide-ranging and include demyelination, gliosis, destruction of axons, and eventual cavitation and infarction. As the myelinated axons within white matter form connective pathways within and between different brain structures, damage to these pathways is likely to have consequences for the efficiency of information transfer within the brain, and therefore, for cognitive function. Indeed, it is likely that white matter alterations differentially affect cognitive function depending on the brain regions involved
Age, together with vascular risk factors, is one of the strongest predictors of WMH burden, and research shows WMH are associated with deficits in a range of cognitive domains including processing speed, executive function and episodic memory
A major objective of the present study, therefore, was to address the paucity of research investigating associations between WMH and cognition in large population-based samples of adults aged below 50 years. Additionally, it was important to establish how far vascular risk factors account for WMH-cognitive associations in middle age. One of the aforementioned studies
All aspects of the study were approved by the Australian National University Human Research Ethics Committee. Written informed consent was obtained from all participants in the study.
This cohort of the
Health histories were obtained through an interview, and included details (prevalence rates in parentheses) of cancer (n = 10; 2.3%), heart disease (n = 13; 3.0%), stroke (n = 4; 0.9%), diabetes (n = 9; 2.1%), thyroid problems (n = 19; 4.4%), and head injury (n = 67; 15.7%). All were coded 1 = Yes, 2 = No. In a minority of cases (<1.7%), missing data were coded ‘2’. This represents a conservative approach to the estimation of disease for missing data. Two readings of resting blood pressure two hours apart with participants sitting were taken by the interviewer using an Omron M4 automatic blood pressure monitor, for which they had received specific training. For present purposes, high blood pressure (BP) was defined as either mean systolic BP>140 mm Hg, or mean diastolic BP>90 mm Hg
Simple and choice reaction tasks were administered, and for both tasks measures of intraindividual mean RT and variability were computed.
These tasks were administered using a small box held with both hands, with left and right buttons at the top to be depressed by the index fingers. The front of the box had three lights: two red stimulus lights under the left and right buttons respectively and a green get-ready light in the middle beneath these. There were four blocks of 20 trials measuring simple reaction time (SRT), followed by two blocks of 20 trials measuring choice reaction time (CRT). For SRT everyone used their right hand regardless of dominance. The interval between the ‘get-ready’ light and the first light of the trial was 2.3 s for both SRT and CRT.
Means were calculated after removing outliers. This was done by firstly eliminating any values over 2000 ms. Next, means and standard deviations were calculated for each individual for each block and values were eliminated which lay outside three standard deviations for each individual. A number of very slow individuals still retained RT scores greater than 1000 ms. In a final step, these values were dropped before the final means per block were calculated for each participant. Here we present the grand mean across blocks for the respective tasks.
Mean absolute residuals (in ms) were calculated for each individual by averaging the deviations from regression models of RT against trial number and block number in each of the simple and choice RT series (Blocks 1–4 inclusive for simple RT, and Blocks 5–6 for choice RT). A quadratic function of trial number was also entered into the model because the decline in RT with practice is not linear. Block number was treated as categorical. These models were designed to remove both intra-block practice effects and the effect of the short rest periods between blocks, leaving residuals that measure only random variation. By contrast, simply using each person's ‘raw’ standard deviation of RT would inflate the apparent variability for participants who showed substantial improvement over the course of their trials.
This procedure is similar to that used in the cognitive aging literature more broadly
In addition to the RT tasks, a battery of cognitive tests was administered to participants. This included a
MRI data were acquired on a 1.5 Tesla Gyroscan scanner (ACS-NT, Philips Medical Systems, Best, The Netherlands). T1-weighted 3-D structural MRI images were acquired in coronal plane using Fast Field Echo (FFE) sequence. About mid-way through this study, for reasons beyond the researchers' control, the original scanner (Scanner A) was replaced with an identical Philips scanner (Scanner B) and single channel RF headcoils. The scanning parameters were kept essentially the same. The first 163 subjects were scanned on Scanner A with TR = 8.84 ms, TE = 3.55 ms, a flip angle of 8°, matrix size = 256×256, slices 160, and field of view (FOV) 256×256 mm. Slices were contiguous with slice thickness of 1.5 mm. For the remaining 268 subjects scanned on Scanner B, the TR = 8.93 ms, TE = 3.57 ms values were slightly different in order to improve image quality, but all other parameters were exactly the same. The fluid-attenuated inversion recovery (FLAIR) sequence was the same for both scanners and acquired with TR = 11,000 ms, TE = 140 ms, TI = 2,600, number of excitations = 2, matrix size = 256×256, and the FOV was 230×230 mm. Slice thickness was 4.0 mm with no gap between slices and in-plane spatial resolution is 0.898×0.898 mm/pixel. To ensure the reliability and compatibility of the data, we compared the subjects scanned on the two scanners on sociodemographic and imaging parameters. There were no differences on age (p = 0.377), or years of education (p = 0.588), but more women were inadvertently scanned on Scanner B than A (p = 0.003). The volumetric measures of total intracranial volume, gray matter volume, white matter volume, or cerebrospinal fluid volume obtained from two scanners did not differ significantly
The image analysis of WMH has been described in detail elsewhere
For missing data on a minority of cognitive variables (mean RT and variability measures for SRT and CRT tasks), values were imputed with the EM algorithm in SPSS
Hierarchical multiple regression was used in the main analyses. WMH variables were regressed onto intracranial volume and total white matter volume at Step 1 in order to take into account individual differences in neuroanatomical structure. At Step 2, the primary effects for the cognitive variables and gender were entered into the equation. As several gender effects involving cognitive variables were evident in the bivariate correlations, at Step 3 the Gender × Cognitive variable interaction (variables were centered prior to this procedure) was entered. Due to the number of regression equations run, alpha was set conservatively at p<.01.
Descriptive statistics for the WMH variables by laterality and gender are presented in
Left Frontal | Right Frontal | Left Temp | Right Temp | Left Parietal | Right Parietal | Left Occ | Right Occ | |||||||||
M | W | M | W | M | W | M | W | M | W | M | W | M | W | M | W | |
% of sample | 7.1 | 6.9 | 8.7 | 11.6 | 1.5 | 0.4 | 2.0 | 0.4 | 15.3 | 15.5 | 16.8 | 20.7 | 1.5 | 0.4 | 1.0 | 0.9 |
Mean vol |
2.23 (12.20) | 2.77 (19.06) | 5.03 (30.98) | 5.46 (24.31) | 0.45 (3.97) | 0.04 (0.59) | 0.37 (2.76) | 0.19 (2.86) | 8.66 (37.21) | 16.28 (57.95) | 7.52 (26.0) | 18.18 (94.59) | 0.39 (3.78) | .06 (.69) | 0.08 (0.91) | .81 (9.67) |
Range |
127.5 | 255 | 346.5 | 244.5 | 45 | 9 | 27 | 43.5 | 337.5 | 415.5 | 190.5 | 1290 | 49.5 | 10.5 | 12 | 139.5 |
Notes.
All values computed within-gender (men = 196; women = 232).
Metric = mm3.
Lowest value = 0.
Temp = Temporal; Occ = Occipital; M = Men; W = Women.
Bivariate correlations for all the main variables in the study are presented in
M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | |
1.Gender | − | − | − | |||||||||||||||||||
2.Years Educ | 14.56 | 2.40 | −.08 | − | ||||||||||||||||||
3.Immed Rec | 8.18 | 2.27 | .24 |
.10 |
− | |||||||||||||||||
4.Del Rec | 7.54 | 2.48 | .22 |
.09 | .84 |
− | ||||||||||||||||
5.Digit Back | 5.80 | 2.22 | −.10 |
.12 |
.21 |
.16 |
− | |||||||||||||||
6.Face Recog | 9.44 | 1.44 | .09 | .11 |
.14 |
.13 |
.10 |
− | ||||||||||||||
7.Lex Dec Making | 51.65 | 4.85 | −.12 |
.41 |
.20 |
.19 |
.28 |
.09 | − | |||||||||||||
8.ISD SRT | 0.044 | .019 | .06 | −.07 | −.06 | −.05 | −.14 |
−05 | −.09 | − | ||||||||||||
9.Mn SRT (ms) | 240 | 42.5 | .17 |
−.08 | −.04 | −.01 | −.18 |
−.07 | −.09 | .59 |
− | |||||||||||
10.ISD CRT | 0.046 | .015 | .07 | −.02 | −.02 | −.04 | −.03 | −.01 | −.03 | .37 |
.29 |
− | ||||||||||
11.Mn CRT (ms) | 292 | 41.4 | .15 |
.01 | .00 | .02 | −.08 | −.02 | −.06 | .38 |
.67 |
.61 |
− | |||||||||
12.ICV | 1449 | 136 | −.66 |
.14 |
−.16 |
−.13 |
.12 |
−.13 |
.15 |
.00 | −.14 |
−.08 | −.12 |
− | ||||||||
13.Tot WM vol | 463 | 55.1 | −.65 |
.13 |
−.16 |
−.12 |
.11 |
−.09 | .11 |
−.02 | −.17 |
−.14 |
−21 |
.86 |
− | |||||||
14.Front WMH L | 2.52 | 16.26 | .02 | .01 | .01 | .01 | −.02 | −.03 | .03 | .11 |
.01 | .12 |
.06 | .09 | .08 | − | ||||||
15.Front WMH R | 5.26 | 27.53 | .01 | −.01 | −.03 | −.04 | −.01 | .09 | −.02 | −.01 | .01 | −.04 | −.01 | −.04 | −.03 | .12 |
− | |||||
16.Temp WMH L | 0.23 | 2.72 | −.08 | −.05 | −.06 | −.09 | .00 | −.12 |
−.11 |
−.02 | −.05 | −.01 | −.01 | .11 |
.09 | −.01 | .15 |
− | ||||
17.Temp WMH R | 0.27 | 2.81 | −.03 | .05 | −.10 |
−.11 |
.06 | .01 | .12 |
.03 | .08 | .00 | .04 | .12 |
.07 | −.02 | .07 | .02 | − | |||
18.Par WMH L | 12.79 | 49.63 | .08 | −.07 | .06 | .06 | .05 | −.01 | .02 | −.06 | .00 | −.02 | .03 | −.06 | −.06 | .19 |
.19 |
.00 | .05 | − | ||
19.Par WMH R | 13.30 | 71.95 | .07 | .04 | −.01 | .03 | .00 | .03 | −.01 | −.06 | −.02 | −.05 | .00 | −.04 | −.04 | .01 | .10 |
−.02 | .01 | .26 |
− | |
20.Occ WMH L | 0.20 | 2.61 | −.07 | .01 | −.03 | −.06 | −.04 | −.01 | .06 | −.05 | −.03 | −.04 | −.03 | .02 | .02 | −.01 | .12 |
.13 |
.35 |
−.01 | −.01 | − |
21.Occ WMH R | 0.48 | 7.15 | .05 | .02 | .07 | .04 | .06 | .05 | .02 | −.04 | −.06 | −.04 | −.08 | −.05 | −.04 | −.01 | −.01 | −.01 | .00 | −.02 | .00 | −.01 |
*P<.05,
**p<.01.
ISD = Intraindividual variability; SRT = Simple RT; CRT = Choice RT; ICV = Intracranial volume; WMH = white matter hyperintensities (ICV and WMH = mm3); Gender, 1 = male, 2 = female.
The results of the hierarchical regressions are presented in
Left Frontal | Right Frontal | Left Temp | Right Temp | Left Parietal | Right Parietal | Left Occ | Right Occ | |
Beta | Beta | Beta | Beta | Beta | Beta | Beta | Beta | |
1a.WM vol | .03 | .02 | −.02 | −.13 | −.04 | −.02 | .03 | .04 |
IC vol | .07 | −.06 | .13 | .23 | −.02 | −.02 | −.01 | −.09 |
2b. Immed Rec (IR) | .01 | −.03 | −.04 | −.10 | .04 | −.03 | −.01 | .06 |
Gender | .14 | −.03 | .00 | .09 | .06 | .10 | −.09 | .02 |
3c. Gender × IR | −.12 | .05 | .03 | .06 | −.03 | −.03 | .03 | .06 |
2b. Del Rec (DR) | .00 | −.04 | −.07 | −.11 | .04 | .01 | −.05 | .03 |
Gender | .14 | −.02 | .01 | .09 | .06 | .09 | −.09 | .03 |
3c. Gender × DR | −.11 | .09 | .06 | .03 | −.02 | −.01 | .07 | .02 |
2b. Digit Back (DB) | −.03 | .00 | −.01 | .05 | .06 | .01 | −.04 | .07 |
Gender | .14 | −.03 | −.01 | .06 | .07 | .09 | −.10 | .04 |
3c. Gender × DB | −.05 | .02 | .00 | .05 | .02 | .00 | .04 | .06 |
2b. Face Rec (FR) | −.03 | .08 | −.10 | .03 | −.01 | .02 | −.01 | .04 |
Gender | .14 | −.03 | −.01 | .06 | .07 | .09 | −.10 | .03 |
3c. Gender × FR | −.04 | −.03 | .14 |
−.04 | −.01 | .02 | .01 | .03 |
2b. Lexical DM (LDM) | .02 | −.01 | −.13 |
.10 | .03 | .00 | .05 | .03 |
Gender | .14 | −.03 | −.02 | .07 | .07 | .09 | −.09 | .04 |
3c. Gender × LDM | −.02 | .05 | .13 |
−.02 | −.01 | .00 | −.07 | .04 |
2b. ISD SRT (ISRT) | .11 | −.01 | −.02 | .02 | −.06 | −.06 | −.04 | −.04 |
Gender | .14 | −.03 | −.01 | .06 | .08 | .10 | −.09 | .04 |
3c. Gender × ISRT | .09 | .01 | .02 | −.06 | −.04 | −.03 | .05 | −.03 |
2b. MSRT | .02 | .01 | −.03 | .08 | −.02 | −.03 | −.02 | −.07 |
Gender | .14 | −.03 | −.01 | .05 | .07 | .09 | −.09 | .04 |
3c. Gender × MSRT | .08 | .06 | .03 | −.14 |
−.05 | −.02 | .05 | −.05 |
2b. ISD CRT (ICRT) | .13 |
−.04 | .00 | .01 | −.02 | −.05 | −.04 | −.05 |
Gender | .14 | −.03 | −.01 | .06 | .07 | .10 | −.10 | .03 |
3c. Gender × ICRT | .16 |
.05 | −.02 | −.02 | −.05 | −.03 | .05 | −.03 |
2b. MCRT | .07 | −.01 | .00 | .04 | .02 | −.01 | −.02 | −.08 |
Gender | .14 | −.03 | −.01 | .06 | .07 | .09 | −.09 | .04 |
3c. Gender × MCRT | .09 | .02 | .00 | −.08 | .02 | −.01 | .03 | −.07 |
Notes:
*p<.01; a = df 2, 425; b = df 2, 423; c = df 1,422.
WM vol = White matter volume; IC vol = Intracranial volume; ISD = Intraindividual standard deviation; MSRT = Mean Simple RT; MCRT = Mean Choice RT; Temp = Temporal; Occ = Occipital.
Therefore, where significant Gender × Cognition interactions were found, regressions were rerun for men and women separately. With one exception, all interactions stemmed from stronger effects in men. The exception was the association between left frontal WMH and variability in the choice RT task. Although this was the only primary effect to attain significance across the whole sample (14 men and 16 women exhibited left frontal WMH), when the regression was rerun for men, it was statistically unreliable. However, for women, that regression was significant, df = 1,228, beta = .23, p<.001, indicating the association between WMH and variability to be stronger in this group (see
When the remaining regressions producing significant Gender × Cognitive variable interactions were rerun within gender, for women, all associations were nonsignificant. By contrast, however, for men, significant associations indicated that greater WMH burden was associated with poorer cognitive performance: Left temporal lobe and face recognition, df = 1,192, beta = −.17, p = .014; left temporal lobe and Spot-the-Word scores, df = 1,192, beta = −.19, p = .008; right temporal lobe and simple mean RT, df = 1,192, beta = .24, p<.001.
We then repeated the analyses taking years of education into account. This had little bearing on the initial regression findings. Importantly, as health status, and in particular vascular risk factors, may influence white matter-cognition relations, we then statistically controlled (by entering health factors individually at Step 1 of the hierarchical multiple regression) for histories of cancer, thyroid problems, head injury, diabetes, stroke, heart disease, and high blood pressure (blood pressure variables were entered as both dichotomous and continuous variables). Notably, none of these variables altered our original findings.
Finally, WMH data are highly skewed and this, together with outliers, may have influenced the findings. In order to reduce the influence of these sources of variance, we reran the main analyses using sequential logistic regression having recoded the WMH variables (0 = no WMH, and 1 = >0 WMH). For left frontal WMH and variability in the CRT task, entry of variability significantly raised the probability of the presence of WMH, B = 0.38, OR = 1.46, CI = 1.09–1.96, p = .011. As subsequent entry of the Gender × Variability interaction term approached significance at conventional levels (p = .068), we ran logistic regression within men and women. For men, the regression was nonsignificant. For women however, greater variability was associated with the presence of left frontal WMH, B = 0.61, OR = 1.83, CI = 1.19–2.82, p = .006. These findings are consistent with the earlier linear regressions.
For left temporal WMH and face recognition, the findings were similar to the earlier analyses. When face recognition was entered into the equation, statistics indicated the presence of left temporal WMH were associated with poorer face recognition B = −0.69, OR = 0.50, CI = 0.26–0.97, p = .039. Although entry of the Gender × Face recognition interaction was nonsignificant, further analysis revealed the trend was stronger in men. By contrast, logistic regressions examining left temporal WMH and spot-the-word, and right temporal WMH and simple mean RT, were inconsistent with the earlier analyses.
This is one of the first investigations to focus on WMH and cognitive function in a large population-based sample of middle-aged adults. Several important findings suggested the possible presence of neuropathology in this relatively young and independently functioning group of 44 to 48 year olds living in the community. First, frontal lobe white matter lesions were associated with increased intraindividual variability, and temporal lobe WMH with deficits in face recognition. Second, these findings were left-lateralized, and the frontal lobe associations stronger in women, while the temporal lobe associations were stronger in men. Finally, statistically controlling for a range of health variables, including vascular risk factors, made no difference to those findings.
That WMH were associated with cognitive deficits was not in itself unusual, and is consistent with findings elsewhere
The results were selective in that left frontal lobe lesions were associated with within-person variability, while left temporal WMH were associated with spot-the-word and face recognition performance. The former finding is in line with the proposition that intraindividual variability indexes neurobiological disturbance
It is of note that, although the association between left frontal WMH and intraindividual variability was stronger in women, the primary effect was also significant for this regression (see
It is important to emphasize that while the measure of within-person variability was sensitive to WMH presence, the alternative measure of central tendency (mean RT) for the same choice RT task, was not. This finding adds to work showing a dissociation between measures of mean RT and within-person variability from the same task, with the latter variable being sensitive to possible neuropathology
The association between left temporal WMH and spot-the-word and recognition performance was in line with, respectively, work implicating that lobe in the processing of nouns
It is important to note that statistically controlling for a range of health variables, including histories of cancer, heart disease, thyroid problems, diabetes, stroke, head injury, and high blood pressure, had no bearing on the findings. The analyses controlling for vascular risk factors were of particular note as evidence suggests that non-periventricular WMH are associated with ischaemia [e.g.,
There are a number of limitations to the present research that should be acknowledged. First, the study was cross-sectional, and we are therefore unable to give any indication of causality. Moreover, the use of a narrow cohort design allows individual differences in characteristics such as WMH to be investigated without the confounding of age differences. However, this means that we cannot generalise our results beyond the ages of 44 to 48 years. Second, at present we do not have any information concerning the future neurological status of participants. Planned long-term follow-ups in this group will provide valuable information on how far the present findings represent the early manifestation of eventual age-related neurological conditions. Finally, due to the young age of the sample, there were relatively few participants with significant white matter lesion load. Therefore, despite the sample being arguably the largest to investigate WMH and cognition in persons in their mid−40 s, even larger samples are required to enable detailed evaluation of the small group of individuals who demonstrate significant pathology in this age group.
To conclude, the finding that cognitive deficits were associated with non-periventricular WMH in a community sample aged 44 to 48 years having taken into account a range of health variables has important implications. From a lifespan perspective, the findings suggest that cognitive deficits may have a neuropathological basis that manifests in some individuals during middle age. From a healthcare perspective this underlines the view that population-based preventative strategies should start in early adulthood and not wait until mid or later life. Not only are the costs of such initiatives likely to be offset by long-term healthcare savings, but also by associated benefits to the quality of life and extended independence of vulnerable persons living in the community.
The authors are grateful to Anthony Jorm, Bryan Rodgers, Chantal Reglade-Meslin, Patricia Jacomb, Karen Maxwell, and the PATH interviewers.