Normative Estimates Of Cross-sectional And Longitudinal Brain Volume Decline In Aging And Ad

Normative estimates of cross-sectional and longitudinal brain volume decline in aging and AD A.F. Fotenos, ScB; A.Z. Snyder, PhD, MD; L.E. Girton, BA; J.C. Morris, MD; and R.L. Buckner, PhD

Abstract—Objective: To test the hypotheses 1) that whole-brain volume decline begins in early adulthood, 2) that cross-sectional and longitudinal atrophy estimates agree in older, nondemented individuals, and 3) that longitudinal atrophy accelerates in the earliest stages of Alzheimer disease (AD). Methods: High-resolution, high-contrast structural MRIs were obtained from 370 adults (age 18 to 97). Participants over 65 (n ⫽ 192) were characterized using the Clinical Dementia Rating (CDR) as either nondemented (CDR 0, n ⫽ 94) or with very mild to mild dementia of the Alzheimer type (DAT, CDR 0.5 and 1, n ⫽ 98). Of these older participants, 79 belonged to a longitudinal cohort and were imaged again a mean 1.8 years after baseline. Estimates of gray matter (nGM), white matter (nWM), and whole-brain volume (nWBV) normalized for head sizes were generated based on atlas registration and image segmentation. Results: Hierarchical regression of nWBV estimates from nondemented individuals across the adult lifespan revealed a strong linear, moderate quadratic pattern of decline beginning in early adulthood, with later onset of nWM than nGM loss. Whole-brain volume differences were detected by age 30. The cross-sectional atrophy model overlapped with the rates measured longitudinally in older, nondemented individuals (mean decline of ⫺0.45% per year). In those individuals with very mild DAT, atrophy rate more than doubled (⫺0.98% per year). Conclusions: Nondemented individuals exhibit a slow rate of whole-brain atrophy from early in adulthood with white-matter loss beginning in middle age; in older adults, the onset of dementia of the Alzheimer type is associated with a markedly accelerated atrophy rate. NEUROLOGY 2005;64:1032–1039

Pathologic brain processes that lead to dementia coexist with normal aging processes that also influence the brain but do not manifest as disease. To better understand the nature of normal brain development in advanced aging and how the earliest stages of dementia of the Alzheimer type (DAT) cause departure from that trajectory, we report here a largesample study of 370 adults age 18 to 97. We sought to characterize the normal development of wholebrain volumes in the absence of dementia and determine, through a combination of cross-sectional and longitudinal estimates, to what degree the presence of early-stage DAT causes departure from normal development. An important feature of the study design is the direct contrast of cross-sectional and longitudinal estimates of brain change. Cross-sectional estimates are efficient in that a single measure can be used as the dependent measure. To the degree that different normal individuals have predictable brain sizes and Additional material related to this article can be found on the Neurology Web site. Go to www.neurology.org and scroll down the Table of Contents for the March 22 issue to find the title link for this article.

changes in brain size, a single point estimate may be informative regarding their likely future course and risk of disease. However, reviews of the structural aging literature highlight the need for longitudinal data because of between-subject variance.1–3 Longitudinal data reduce between-subject variance by using an individual as his or her own baseline and also control for differences that potentially complicate cross-sectional samples. For example, cross-sectional samples may include hidden group heterogeneity (cohort effects), such as environmental differences between when people were born (secular effects). MRI is readily able to obtain longitudinal data through repeated imaging of the same person over time.4 The present design, which combines crosssectional and longitudinal approaches,5 allows three basic questions to be addressed. First, to what extent and at what age does nondemented aging associate with cross-sectional brain-volume reduction? Some volumetric reports suggest that whole-brain volume is stable in nondemented adults under 50,6 –10 whereas others find volume loss in this age range,5,11–15 a difference possibly related to differential contributions of gray- and white-matter loss to

From the Division of Biology and Biomedical Sciences (A.F. Fotenos and Dr. Buckner), Mallinckrodt Institute of Radiology (Drs. Snyder and Buckner), Department of Neurology (Drs. Snyder and Morris), Howard Hughes Medical Institute (L.E. Girton and Dr. Buckner), and Department of Psychology (Dr. Buckner), Washington University, St. Louis, MO. Supported by grants P50 AG 05681 and P01 AG 03991 from the National Institute on Aging, Bethesda, MD; IIRG-00-1944 from the Alzheimer’s Association; the James S. McDonnell Foundation; and the Howard Hughes Medical Institute. Received May 24, 2004. Accepted in final form November 30, 2004. Address correspondence and reprint requests to Dr. Anthony Fotenos, HHMI at Washington University, Psychology Department Campus Box 1125, One Brookings Drive, St. Louis, MO 63108; e-mail: [email protected] 1032

Copyright © 2005 by AAN Enterprises, Inc.

Table Sample characteristics Old Young No. (cross-sectional) Female/male Age ⫾ SD, y

Middle-aged

CDR 0

DAT (CDR 0.5)

DAT (CDR 1)

127

51

94

69

29

64/63

29/22

67/27

32/37

21/8

23 ⫾ 3

50 ⫾ 8

78 ⫾ 8

78 ⫾ 6

79 ⫾ 6

(18–34)

(35–64)

(65–95)

(65–93)

(69–97)

15 ⫾ 3

14 ⫾ 3

13 ⫾ 3

Education ⫾ SD, y

(8–23) MMSE ⫾ SD

(7–20)

(7–20)

29 ⫾ 1

26 ⫾ 3

22 ⫾ 4

(25–30)

(18–30)

(13–28)

2.9 ⫾ 2.1

2.4 ⫾ 1.9

2.7 ⫾ 2.3

(0–9)

(0–8)

(0–8)

136 ⫾ 18

143 ⫾ 21

146 ⫾ 26

(102–192)

(104–188)

(90–192)

73 ⫾ 10

73 ⫾ 10

77 ⫾ 11

(40–96)

(50–98)

(60–100)

Reported HBP, %

43.0

41.8

48.3

Diabetes, %

10.8

11.8

10.7

Prescriptions, n

Systolic BP, mm Hg

Diastolic BP, mm Hg

No. with follow-up (longitudinal) Female/male Scan interval ⫾ SD, y

38

33

8

30/8

10/23

5/3

1.8 ⫾ 0.5

1.8 ⫾ 0.5

1.8 ⫾ 0.4

(1.1–3.9)

(1.3–3.5)

(1.0–2.4)

The sample consisted of 370 individuals (272 nondemented and 98 with DAT). Mean values given ⫾ SD. Values in parentheses represent the range. Education (n ⫽ 7), MMSE (n ⫽ 11), and clinical data (n ⫽ 16) were not available for some participants. Compared to the older nondemented adults, the older adults with dementia had lower scores on the MMSE (t[179] ⫽ 10.02, p ⬍ 0.001) and slightly fewer years of education (t[183] ⫽ 2.55, p ⬍ 0.05). CDR ⫽ Clinical Dementia Rating, with 0, 0.5, and 1 corresponding to nondemented, very mild, and mild DAT; DAT ⫽ dementia of the Alzheimer type; MMSE ⫽ Mini-Mental State Examination where scores range from 30 (best) to 0 (worst); HBP ⫽ high blood pressure.

brain aging.4 Second, does the cross-sectional rate of atrophy in nondemented older adults match the longitudinal rate? As noted above, a number of potential confounds could lead to a mismatch between crosssectional and longitudinal findings. If the crosssectional observations accurately predict the longitudinal atrophy rate, it is reasonable to assume that cohort and secular effects are minimal and volume loss progresses in a predictable manner in the absence of dementia. Finally, to what extent does the rate of whole-brain atrophy accelerate in early-stage DAT? The available reports addressing this question have found significant acceleration, but differ as to its magnitude.16-18 Methods. Participants. A total of 370 adults (age 18 to 97 at baseline) participated in a structural MR imaging session. Of these individuals, 79 participated on two separate occasions separated by an extended interval to allow for longitudinal data analysis (1.0 to 3.9 year interval; mean ⫽ 1.8 years). Twenty additional individuals were scanned twice at a short interval (mean ⫽ 21 days, range 1 to 64) to allow estimation of measurement reliability. Participants were paid for their participation and gave written informed consent in accordance with guidelines of the Washington University Human Studies Committee. Data from subsets of the participants have been used in previous studies.15,19

Young and middle-aged adults were recruited from the Washington University community. Nondemented and demented older adults were recruited exclusively from the ongoing longitudinal sample of the Washington University AD Research Center (ADRC). ADRC volunteers are more likely than the population of the St. Louis metropolitan area to have a high school education, and volunteers with severe comorbidities such as major depression or disabling stroke are excluded.20 Approximately 40% of ADRC participants who met the study’s clinical criteria (nondemented or dementia restricted to DAT) declined to participate in an MRI; 7% were ineligible based on MRI contraindications. There was no statistically significant difference in age, years of education, or scores on the Mini-Mental State Examination (MMSE21) between ADRC participants who did and did not undergo MRI. Dementia severity was quantified using the Clinical Dementia Rating (CDR22) scale for all ADRC volunteers, and recruitment for MRI was independent of longitudinal clinical progression. The average duration between clinical assessment and participation in the MRI session was 101 days (range ⫽ 3 to 332 days). The 98 participants with DAT exhibited very mild (CDR 0.5, n ⫽ 69) to mild (CDR ⫽ 1, n ⫽ 29) dementia severity. Of the 205 older adults who underwent MRI, 13 (6%; 7 CDR ⫽ 1, 3 CDR ⫽ 0.5, 3 CDR ⫽ 0) did not complete the imaging protocol; one dropped out on repeat imaging. Although several DAT participants had cognitive test scores (e.g., MMSE) that might qualify for classification as mild cognitive impairment, a CDR score of 0.5 or greater in this sample is highly predictive of AD, both in clinical progression and neuropathologic diagnosis at autopsy.23,24 Demographic and clinical data for participants are presented in the table. March (2 of 2) 2005

NEUROLOGY 64

1033

Figure 1. Whole-brain volume measurement and normalization procedure. (A) Single magnetization-prepared rapid gradient echo (MP-RAGE) image as acquired in native space. (B) Within-participant averaged MP-RAGE image (n ⫽ 4); note the increased contrast-to-noise. (C) Averaged image after registration to a target atlas composed of representative young and old individuals in Talairach and Tournoux29 space. (D) Averaged, atlas-registered image after masking and fieldinhomogeneity correction. (E) Final segmented image; normalized whole-brain volume (nWBV) is defined as the percentage of the brain mask (non-black background) occupied by voxels classified as gray and white matter. Image acquisition. Multiple (three or four) high-resolution structural T1-weighted magnetization-prepared rapid gradient echo (MP-RAGE) images were acquired on a 1.5-T Vision scanner (Siemens, Erlangen, Germany). MP-RAGE parameters were empirically optimized for gray-white contrast (repetition time [TR] ⫽ 9.7 msec, echo time [TE] ⫽ 4 msec, flip angle [FA] ⫽ 10, inversion time [TI] ⫽ 20 msec, delay time [TD] ⫽ 200 msec, 256 ⫻ 256 [1 mm ⫻ 1 mm] in-plane resolution, 128 sagittal 1.25 mm slices without gaps, time per acquisition ⫽ 6.6 minutes). Participants were provided cushioning, headphones, and a thermoplastic face mask for communication and to minimize head movements. Positioning was low in the head coil (toward the feet) to center the field of view on the cerebral hemispheres. The MP-RAGE images were acquired as the second part of a 70-minute protocol that also included fast low angle shot (FLASH) gradient echo, turbo spin echo (TSE), and diffusion tensor imaging (DTI) acquisitions. The DTI data have been reported elsewhere.25 Image analysis. Normalized gray matter (nGM; gray parenchyma within the entire intracranial volume down to approximately the superior arch of C1), white matter (nWM), and whole-brain volume (nWBV; gray plus white parenchyma) were computed for each image session. The procedure was based on a validated, open-source segmentation tool.26,27 Prior to image segmentation, the images were preprocessed to normalize for head size and intensity variation that might affect image segmentation. Preprocessing included multiple steps. Head-size normalization used a validated method based on atlas registration.28 The normalization is proportional to manually measured total intracranial volume (TIV, r ⫽ 0.93) and minimally biased by atrophy. Images were corrected for interscan head movement and spatially warped into the atlas space of Talairach and Tournoux.29 The template atlas consisted of a combined young-and-old target previously generated from a representative sample of young (n ⫽ 12) and nondemented old (n ⫽ 12) adults. The use of a combined template has been shown to minimize the potential bias of an atlas normalization procedure to overexpand atrophied brains.28 For registration, a 12-parameter affine transformation was computed to minimize the variance between the first MP-RAGE image and the atlas target.30 The remaining MP-RAGE images were registered to the first (in-plane stretch allowed) and resampled via transform composition into a 1-mm isotropic image in atlas space. All images were visually inspected to verify appropriate atlas transformation. The result was a single, high-contrast, averaged MP-RAGE image in atlas space (figure 1). Subsequent preprocessing steps included skull removal by application of a loose-fitting atlas mask and correction for intensity inhomogeneity due to nonuniformity in the magnetic field. Intensity variation was corrected across contiguous regions, based on a quadratic inhomogeneity model. Following preprocessing, the segmentation algorithm classified each voxel of the average image as CSF, gray, or white matter.26,27 This segmentation starts with an initial estimation step to obtain and classify tissue parameters. An expectation-maximization algorithm then updates class labels and tissue parameters in order to 1034

NEUROLOGY 64

March (2 of 2) 2005

iterate toward the maximum likelihood estimates of a hidden Markov, random field model. The model uses spatial proximity to constrain the probability with which voxels of a given intensity are estimated to belong to each tissue class. Normalized volumes were computed as the proportion of all voxels within the brain mask occupied by gray (nGM), white (nWM), or gray plus white (nWBV, equivalent to 100 ⫺ %CSF) voxels. The unit of normalized volume is percent, which represents the percentage of estimated TIV. Test-retest measurement reliability. In order to estimate measurement reliability, normalized volumes for the same person were compared over two imaging sessions separated by a brief interval, during which it is reasonable to assume minimal true change. Twenty individuals contributed to the test-retest group (young, n ⫽ 16; middle-aged, n ⫽ 1; older, n ⫽ 3), with a mean delay of 21 days between test and retest (range 1 to 64 days). The mean absolute percentage differences (MAPD, the absolute difference between test and retest volumes divided by the overall mean, expressed in percent) were 0.92% for nGM (CI 0.53 to 1.30), 0.80% for nWM (CI 0.5 to 1.1), and 0.49% for nWBV (CI 0.28 to 0.71). The coefficients of variation (CV, the SD of the difference between test and retest volumes divided by the overall mean, expressed in percent) were 1.24% (nGM), 1.04% (nWM), and 0.68% (nWBV). The interclass correlations between values paired by participant, but randomly assigned to test or retest, were high (r[nGM] ⫽ 0.99, r[nWM] ⫽ 0.98, and r[nWBV] ⫽ 0.99). Cross-sectional analysis. Normalized volumes were plotted against age for the 370 unique participants. Statistical analysis was conducted with the JMP software package (SAS Institute, Cary, NC). Hierarchical polynomial regression was used to test between linear and curvilinear models of cross-sectional volume as a function of age. With the sample restricted to one volume measurement per nondemented individual, higher order terms of the subject’s age at scan were tested until they no longer contributed significantly to the model; the resulting models are referred to as the cross-sectional, nondemented aging curves. Normalized volumes in the older DAT and older nondemented samples were then compared using analysis of covariance with nGM, nWM, or nWBV as the dependent measure and age, sex, and dementia status as cofactors. Longitudinal analysis. The longitudinal analysis was restricted to the most reliable whole-brain data and sought to quantify the whole-brain atrophy rate within older nondemented and DAT individuals. Atrophy rate was computed as the slope of the line connecting nWBV measurements within each individual, divided by baseline nWBV, expressed as percent change per year. For example, in a participant with two scans, atrophy rate was computed as nWBV at scan 2 minus nWBV at scan 1, divided by the interval between measurements, divided by nWBV at scan 1, times 100. Analysis of covariance was again used to test for differences in atrophy rate based on age, sex, and dementia status. Comparison of cross-sectional and longitudinal data. To compare cross-sectional and longitudinal data, atrophy rate was estimated from the cross-sectional, nondemented aging curve and

Figure 2. Cross-sectional plots of gray and white matter, normalized for head size. (A) Cross-sectional plot of normalized gray matter (nGM) across the adult lifespan; each data point represents a unique participant from a single scanning session. For participants with follow-up data, the session with nearestin-time clinical data is used (see the table). The best-fit polynomial regression is drawn only for the nondemented individuals (blue) and is represented by the dashed line for women (55.1 ⫺ 0.065[AGE] ⫺ 0.001[AGE2]) and the solid line for men (54.4 ⫺ 0.076[AGE] ⫺ 0.001[AGE2]). (B) Cross-sectional plot of normalized white matter (nWM) across the adult lifespan. The dashed line represents the nondemented, female regression (30.4 ⫹ 0.089[AGE] ⫺ 0.001[AGE2]); the solid line represents the nondemented, male regression (30.3 ⫹ 0.100[AGE] ⫺ 0.001[AGE2]). Note the inflection in the nWM curves (around age 42 for men and 45 for women) and the greater separation between nondemented and DAT individuals (red vs blue) in the nGM plot. DAT ⫽ dementia of the Alzheimer type.

compared with the longitudinal, nondemented atrophy rate. Atrophy rate was estimated from the cross-sectional curve by expressing its slope as a percentage of nWBV at the mean age of the older, nondemented sample (age ⫽ 78, nWBV ⫽ 74.8%). For graphical comparison, trendlines were plotted for nondemented and demented aging. The slope of the nondemented trendline, for example, was determined by the mean atrophy rate of the nondemented, longitudinal sample; the y-intercept was determined by interpolating nWBV from the nondemented, cross-sectional sample (mean age ⫽ 81, nWBV ⫽ 74.1%; DAT trendline drawn equivalently, mean age 78, nWBV ⫽ 72.0%; see figure 3).

Results. Cross-sectional. The cross-sectional dataset is plotted in figures 2 and 3A and summarized in absolute volumes without head-size normalization in table E-1 (on the Neurology Web site at www.neurology.org). Nondemented individuals between 18 and 95 years old exhibited an age-associated decline in normalized gray matter (nGM; r ⫽ ⫺0.91, F[1,270] ⫽ 1311.00, p ⬍ 0.001), white matter (nWM; r ⫽ ⫺0.25, F[1,270] ⫽ 17.71, p ⬍ 0.001), and wholebrain volume (nWBV; r ⫽ ⫺0.88, F[1,270] ⫽ 939.59, p ⬍ 0.001). The age-by-volume correlation remained significant when considering the age range of 18 to 30 for nGM (r ⫽

⫺0.20, F[1,121] ⫽ 4.91, p ⬍ 0.05) and nWBV (r ⫽ ⫺0.19, F[1,121] ⫽ 4.58, p ⬍ 0.05), but not for nWM (F ⬍ 1). Considering the full age range, adding a quadratic term improved the models of nGM (F[2,269] ⫽ 681.24, p ⬍ 0.001, R2 ⫽ 0.84), nWM (F[2,269] ⫽ 18.96, p ⬍ 0.001, ⫽ R2 ⫽ 0.12), and nWBV (F[2,269] ⫽ 533.32, p ⬍ 0.001, R2 ⫽ 0.80). The addition of a cubic term failed to add a significant effect for any model (F ⬍ 1). In addition to the cross-sectional, nondemented aging curves, figures 2 and 3A illustrate that individuals with DAT (CDR ⫽ 0.5 and 1) exhibited volume reduction disproportionate to age. A full-factorial analysis of covariance on the older sample with age, sex, and dementia status as covariates was significant for nGM (F[7,184] ⫽ 20.79, p ⬍ 0.001), nWM (F[7.184] ⫽ 3.14, p ⬍ 0.01), and nWBV (F[7,184] ⫽ 19.80, p ⬍ 0.001), with main effects for all three covariates. Post hoc testing indicated women had more nGM (43.0% vs 42.2%, p ⫽ 0.051), nWM (30.9% vs 30.1%, p ⬍ 0.05), and nWBV (73.9% vs 72.3%, p ⬍ 0.01) than men. The presence of DAT was associated with a decrease in nGM (43.7% vs 41.7%, p ⬍ 0.001), nWM (31.1% March (2 of 2) 2005

NEUROLOGY 64

1035

Figure 3. Cross-sectional and longitudinal plots of whole-brain volume, normalized for head size. (A) Cross-sectional plot of normalized whole-brain volume (nWBV) across the adult lifespan. The line represents the best-fit polynomial regression of all nondemented individuals and is referred to as the cross-sectional, nondemented aging curve (85.3 ⫹ 0.013[AGE] ⫺ 0.002[AGE2]). (B) Longitudinal plot of nWBV in older adults (note scale change); lines connect nWBV at baseline and follow-up scans (or the best fit, for participants with multiple followups), such that the slope of each line as a proportion of baseline nWBV represents an individual’s atrophy rate. The slope of the thick blue line represents the estimated longitudinal rate of change for all of the nondemented individuals and overlaps with the slope of the cross-sectional, nondemented aging curve (shown in black). The slope of the thick red line represents the longitudinal rate of change for all of the individuals with dementia of the Alzheimer type (DAT) and suggests accelerated volume loss in DAT. Lines connected by blue triangles and red squares represent individuals who converted from a Clinical Dementia Rating of 0 to 0.5 during the interscan interval; they are included in the DAT mean.

vs 30.2%, p ⬍ 0.05), and nWBV (74.7% vs 71.9%, p ⬍ 0.001; figure 4A). Longitudinal. The longitudinal dataset, obtained in older adults, is plotted in figure 3B. The whole-brain atrophy rate in nondemented older adults was ⫺0.45% (SD ⫽ 0.53) per year. The atrophy rate in age-matched individuals with DAT was ⫺0.98% (SD ⫽ 1.0) per year. A fullfactorial analysis of covariance with age, sex, and dementia status at last scan as covariates, and atrophy rate as the dependent measure, was significant (F[7,71] ⫽ 2.16, p ⬍ 0.05), with a main effect for dementia status and a significant interaction between age and dementia status. The longitudinal, nondemented atrophy rate of ⫺0.45% per year showed no correlation with age within the older sample (r ⫽ ⫺0.17, p ⫽ 0.30; see figure 4C) and closely matched the atrophy rate estimated from the crosssectional, nondemented aging curve, which varied from ⫺0.31% to ⫺0.46% per year over the same age range. Of the 43 nondemented (CDR 0) individuals followed longitudinally from their first scan, six declined to a CDR of 0.5 at the time of their last scan. Figure 4B compares 1036

NEUROLOGY 64

March (2 of 2) 2005

the rate of atrophy in individuals based on their CDR scores at first and last scan. The rate of those who started with a CDR of 0 and declined (⫺0.88% per year, SD ⫽ 0.60) matched the rate of those who started with CDR of 0.5 (⫺0.90% per year, SD ⫽ 0.74). Post hoc testing revealed a trend toward a difference (t[41] ⫽ 3.03, p ⫽ 0.09) between the nondemented group (CDR 0 3 0) and the decliner group (CDR 0 3 0.5), although the small sample size limited statistical power.

Discussion. In a large, cross-sectional sample of nondemented adults, significant decline in wholebrain volume was detected in early adulthood and continued into old age, with distinct patterns for gray- and white-matter loss. The cross-sectional rate of decline overlapped the longitudinal rate in the older, nondemented adults. For the longitudinal subset of older adults in the earliest stages of DAT, the rate of whole-brain atrophy (⫺0.98% per year) was more than twice the nondemented rate (⫺0.45% per

Figure 4. Summary data. (A) Mean cross-sectional normalized whole-brain volume (nWBV) for individuals 65 and over separated by Clinical Dementia Rating (CDR) (0, 0.5, and 1). All differences are significant. (B) Longitudinal atrophy rates, expressed in nWBV loss per year relative to baseline, are separated by CDR status at first and last session. Atrophy rate was significantly greater for the group entering the experiment with very mild dementia (CDR 0.5 3 0.5/1) than the group entering without dementia that remained stable (CDR 0 3 0) and resembles the rate for the group that manifested the earliest signs of dementia of the Alzheimer type (DAT) during the experiment (CDR 0 3 0.5), though we await confirmation in a larger sample. (C) Individual atrophy rates are plotted vs age, with the trendline drawn through the CDR 0 3 0.

year), indicating marked acceleration. These main results are elaborated in terms of the three questions posed in the introduction.

The cross-sectional, nondemented aging curve (see quadratic regression, figure 3A) shows that nWBV declined from 85% at age 20 to 74% at age 80, a lifespan atrophy rate of 0.23% per year, in general agreement with prior studies. Table E-2 (on the Neurology Web site at www.neurology.org) summarizes the results from 12 cross-sectional MR reports on healthy aging that cover the adult lifespan and report whole-brain or gray/white-matter estimates as a percentage of head size. A number of other reports are qualitatively similar, but employ quantitatively distinct units that cannot be directly compared to the present results.11,13,31–38 As the median of all estimates shown in table E-2, nWBV declines from 89% at age 20 to 78% at age 80 (median 0.23% per year atrophy). Strong agreement with a recent population-based, volumetric survey of 2,081 individuals, age 34 to 97, argues in favor of the generalizability of this sample and these findings.39 Comparable pathologic estimates fall below the range in table E-2; a quadratic regression of volumetric data from one such study suggests that across the reference age range (20 to 80), brain volume as a percentage of cranial cavity volume decreased from 92% to 85% (0.14% per year atrophy).6 This early study employed a volumetric method involving fluid displacement that did not account for ventricular volume and likely overestimated brain volume and underestimated atrophy. In addition to quantifying the magnitude of volume decline, the present results converge with others on a temporal sequence placing brain volume reduction at or before the start of early adulthood.5,11–15 Whole-brain volume decline was significant within the adult sample when it was restricted to age 18 to 30, although greater volume reductions were noted in the older adults as compared to the young adults. The significant age correlation in this youngest subset argues against a sample contaminated with preclinical AD (i.e., individuals with the pathologic substrate of AD who are not yet sufficiently impaired to be recognized clinically as demented) as the only explanation for atrophy in nondemented older populations. A moderate acceleration of volume loss in nondemented aging occurred in middle age, around the inflection point of the nWM curve at age 44 (see figure 2B). Similar downward inflections7,31 after a period of white volume stability9,11,40,41 or possibly growth10,12,14,42,43 during the third and fourth decades have been attributed to the prolonged and heterochronologic development of brain myelination.44 – 46 This delayed pattern of nWM loss (however, see references 37, 38, and 47) contrasts with the more linear course of nGM decline throughout adulthood, potentially suggesting separate age-related mechanisms for each.4 Sex effects were minimal in our results. For the overall cross-sectional sample, which was not sex balanced across age, men had approximately 12% more brain volume than women prior to head-size correction and 0.3% less after head-size correction. March (2 of 2) 2005

NEUROLOGY 64

1037

The slightly more downward age-course in men than women did not reach significance for nWBV, nGM, or nWM, but tended in the same direction as reported age-by-sex interactions.8,32,33,37,39 Main sex effects from studies that report no age-by-sex interactions can be compared with the present 0.3% normalized volume difference: four12,40,47,48 report similar sex effects (female ⬎ male), three9,10,42 report no sex effect, and three34,38,41 report sex effects in the opposite direction (male ⬎ female). The small magnitude of any true difference after head-size normalization, the possibility of differential healthfulness between sex cohorts, and methodologic differences likely contribute to inconsistent sex findings. The longitudinal rate of whole-brain atrophy averaged ⫺0.45% per year in the older, nondemented sample. At least seven prior studies have quantified longitudinal, whole-brain volume change in nondemented individuals over a comparable age range. Six18,42,48 –51 have documented annualized rates of about ⫺0.5% (⫺0.37 to ⫺0.88), comparable to the present finding, whereas one52 reported a rate of ⫺2.1%.4 Differences in inclusion/exclusion criteria, scan resolution, and MRI maintenance may explain the divergence of the latter study. Returning to the present findings, the overlap between the longitudinal atrophy rate in nondemented individuals (⫺0.45%) and the cross-sectional estimate (⫺0.31 to ⫺0.46%) for the 65 through 95 age range demonstrates excellent agreement. In addition, longitudinal reports covering the young adult age range5,42 find slower atrophy rates than in this older, longitudinal sample, suggesting that the trend toward accelerated atrophy with age in figure 4C might reach significance with wider longitudinal sampling or increased sample size. Together, our results indicate that secular effects and other confounds minimally influence cross-sectional, whole-brain volume estimates. For instance, if developmental conditions varied among sampled age cohorts, such that people born in more recent years tended to have increased brain volume in proportion to head size than people born in earlier years, we would expect the slope of the cross-sectional aging curve to exceed the longitudinal slope. A difference might similarly result if aging mechanisms were idiosyncratic and either the longitudinal rates formed a multimodal distribution or sampling characteristics differed among the longitudinal and cross-sectional cohorts. Instead, the observed agreement suggests that the brain loses volume with age according to uniform and predictable, though largely unknown, mechanisms.48 It is established that individuals with DAT exhibit decreased brain volume relative to their nondemented peers, with the underlying pathology prominent in regions within the medial temporal lobe.53–55 MRI studies have detected accelerated global and regional volume change in DAT.3,56 In addition, comparing nondemented aging with DAT in figure 2 tentatively suggests that gray matter is more vulnerable than white matter to very mild to mild AD pathology.51 1038

NEUROLOGY 64

March (2 of 2) 2005

Our longitudinal data indicate a ⫺0.98% per year whole-brain atrophy rate in the earliest stages of DAT (CDR 0.5). This rate can be directly compared to recent estimates for nondemented individuals who converted to DAT during follow-up (⫺0.8%) and those with slow-progressing (⫺0.6%) and fastprogressing (⫺1.4%) DAT at baseline.18 Faster whole-brain atrophy rates (⫺5.2%51 and ⫺2.4%57) have been reported in smaller cohorts with more advanced AD (baseline MMSEs ⬍ 20). Other estimates of longitudinal whole-brain change in DAT derive from the brain-boundary shift integral (BBSI), which models boundary changes in serially registered scans.58 With at least one exception,16 atrophy rate estimates based on the BBSI have exceeded ⫺2% per year,17,58 for example, a ⫺2.37% BBSI atrophy rate in 54 DAT patients.49 Divergence in longitudinal atrophy rates may reflect differences in atrophy measurements and DAT cohorts; the DAT sample yielding the ⫺2.37% estimate had a lower age (61 vs 79) and MMSE (20 vs 26) than in the present study and included early onset and familial cases. Providing evidence that the specific DAT sample represents an important factor, in a sample of 5 “preclinical” familial cases followed during their conversion to DAT, the atrophy rate was found to be ⫺1.23% per year using the BBSI and ⫺1.08% per year using TIV correction, very similar to the rates reported here.59 The estimates here and in the literature suggest that nonfamilial, late-onset DAT is characterized by at least a 1% per year volume loss in its earliest clinical presentation with acceleration as the disease progresses.17 This initial rate of brain volume loss represents a doubling from that of nondemented individuals. Our results tentatively suggest that accelerated loss in whole-brain volume may begin in the preclinical phase of AD.60 In particular, the small sample of nondemented individuals who declined over the course of our observation period showed accelerated atrophy rates similar to the individuals with very mild DAT at baseline. The reliability, cross-sectional validity, and sensitivity to clinical progression of automated whole-brain measures such as nWBV, combined with their potential to detect preclinical AD, highlights the promise of global volumetric biomarkers.18 Acknowledgment The authors thank the Washington University ADRC and the Conte Center for clinical assistance and participant recruitment, Elizabeth Grant for database assistance, Chengjie Xiong for discussion of statistical procedures, Daniel Marcus for database development and support, and Susan Larson, Amy Sanders, and Glenn Foster for assistance with MRI data collection.

References 1. Raz N. Aging of the brain and its impact on cognitive performance: integration of structural and functional findings. In: Craick F, Salthouse TA, eds. Handbook of aging and cognition. Mahwah, NJ: Erlbaum, 2000;1–90. 2. Uylings H, de Brabander JM. Neuronal changes in normal human aging and Alzheimer’s disease. Brain Cogn 2002;49:268 –276.

3. Kantarci K, Jack CR. Neuroimaging in Alzheimer disease: an evidencebased review. Neuroimaging Clin N Am 2003;13:197–209. 4. Raz N. The aging brain observed in vivo: differential changes and their modifiers. In: Cabeza R, Nyberg L, Park DC, eds. Cognitive neuroscience of aging: linking cognitive and cerebral aging. London, UK: Oxford University Press, 2004;1–39. 5. Giedd JN, Blumenthal J, Jeffries NO, et al. Brain development during childhood and adolescence: a longitudinal MRI study. Nat Neurosci 1999;2:861– 863. 6. Davis P, Wright EA. A new method for measuring cranial cavity volume and it application to the assessment of cerebral atrophy at autopsy. Neuropathol Appl Neurobiol 1977;3:341–358. 7. Miller AK, Alston RL, Corsellis JA. Variation with age in the volumes of grey and white matter in the cerebral hemispheres of man: measurements with an image analyser. Neuropathol Appl Neurobiol 1980;6: 119 –132. 8. Matsumae M, Kikinis R, Morocz IA, et al. Age-related changes in intracranial compartment volumes in normal adults assessed by magnetic resonance imaging. J Neurosurg 1996;84:982–991. 9. Guttmann C, Jolesz FA, Kikinis R, et al. White matter changes with normal aging. Neurology 1998;50:972–978. 10. Ge YL, Grossman RI, Babb JS, Rabin ML, Mannon LJ, Kolson DL. Age-related total gray matter and white matter changes in normal adult brain. Part I: volumetric MR imaging analysis. Am J Neuroradiol 2002;23:1327–1333. 11. Pfefferbaum A, Mathalon DH, Sullivan EV, Rawles JM, Zipursky RB, Lim KO. A quantitative magnetic resonance imaging study of changes in brain morphology from infancy to late adulthood. Arch Neurol 1994; 51:874 – 887. 12. Courchesne E, Chisum HJ, Townsend J, et al. Normal brain development and aging: quantitative analysis at in vivo MR imaging in healthy volunteers. Radiology 2000;216:672– 682. 13. Rovaris M, Iannucci G, Cercignani M, et al. Age-related changes in conventional, magnetization transfer, and diffusion-tensor MR imaging findings: study with whole-brain tissue histogram analysis. Radiology 2003;227:731–738. 14. Sowell ER, Peterson BS, Thompson PM, Welcome SE, Henkenius AL, Toga AW. Mapping cortical change across the human life span. Nat Neurosci 2003;6:309 –315. 15. Salat DH, Buckner RL, Snyder AZ, et al. Thinning of the cerebral cortex in aging. Cereb Cortex 2004;14:721–730. 16. Cardenas VA, Du AT, Hardin D, et al. Comparison of methods for measuring longitudinal brain change in cognitive impairment and dementia. Neurobiol Aging 2003;24:537–544. 17. Chan D, Janssen JC, Whitwell JL, et al. Change in rates of cerebral atrophy over time in early-onset Alzheimer’s disease: longitudinal MRI study. Lancet 2003;362:1121–1122. 18. Jack CR, Shiung MM, Gunter JL, et al. Comparison of different MRI brain atrophy rate measures with clinical disease progression in AD. Neurology 2004;62:591– 600. 19. Head D, Snyder AZ, Girton L, Morris, Buckner RL. Frontalhippocampal double dissociation between normal aging and Alzheimer’s disease. Cereb Cortex 2005 (in press). 20. Villareal DT, Grant E, Miller JP, Storandt M, McKeel DW, Morris JC. Clinical outcomes of possible versus probable Alzheimer’s disease. Neurology 2003;61:661– 667. 21. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state.” A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12:189 –198. 22. Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology 1993;43:2412–2414. 23. Berg L, McKeel DW, Miller JP, et al. Clinicopathologic studies in cognitively healthy aging and Alzheimer disease—relation of histologic markers to dementia severity, age, sex, and apolipoprotein E genotype. Arch Neurol 1998;55:326 –335. 24. Morris JC, Storandt M, Miller JP, et al. Mild cognitive impairment represents early-stage Alzheimer disease. Arch Neurol 2001;58: 397– 405. 25. Head D, Buckner RL, Shimony JS, et al. Differential vulnerability of anterior white matter in nondemented aging with minimal acceleration in dementia of the Alzheimer type: evidence from diffusion tensor imaging. Cereb Cortex 2004;14:410 – 423. 26. Zhang YY, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectationmaximization algorithm. IEEE Trans Med Imaging 2001;20:45–57. 27. Smith SM. Fast robust automated brain extraction. Hum Brain Mapp 2002;17:143–155. 28. Buckner RL, Head D, Parker J, et al. A unified approach for morphometric and functional data analysis in young, old, and demented adults using automated atlas-based head size normalization: reliability and validation against manual measurement of total intracranial volume. Neuroimage 2004;23:724 –738. 29. Talairach J, Tournoux P. Co-planar stereotaxic atlas of the human brain: an approach to medical cerebral imaging. New York: Thieme Medical Publishers, 1988. 30. Snyder AZ. Difference image vs ratio image error function forms in PET-PET realignment. In: Bailey D, Jones T, eds. Quantification

31.

32. 33.

34. 35. 36. 37. 38.

39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57.

58. 59. 60.

of brain function using PET. San Diego: Academic Press, 1996; 131–137. Harris GJ, Schlaepfer TE, Peng LW, Lee S, Federman EB, Pearlson GD. Magnetic resonance imaging evaluation of the effects of ageing on grey-white ratio in the human brain. Neuropathol Appl Neurobiol 1994; 20:290 –293. Christiansen P, Larsson HB, Thomsen C, Wieslander SB, Henriksen O. Age dependent white matter lesions and brain volume changes in healthy volunteers. Acta Radiol 1994;35:117–122. Murphy DG, DeCarli C, McIntosh AR, et al. Sex differences in human brain morphometry and metabolism: an in vivo quantitative magnetic resonance imaging and positron emission tomography study on the effect of aging. Arch Gen Psychiatry 1996;53:585–594. Raz N, Gunning FM, Head D, et al. Selective aging of the human cerebral cortex observed in vivo: differential vulnerability of the prefrontal gray matter. Cereb Cortex 1997;7:268 –282. Passe TJ, Rajagopalan P, Tupler LA, Byrum CE, MacFall JR, Krishnan K. Age and sex effects on brain morphology. Prog NeuroPsychopharmacol Biol Psychiatry 1997;21:1231–1237. Jernigan TL, Archibald SL, Fennema-Notestine C, et al. Effects of age on tissues and regions of the cerebrum and cerebellum. Neurobiol Aging 2001;22:581–594. Sullivan EV, Rosenbloom M, Serventi KL, Pfefferbaum A. Effects of age and sex on volumes of the thalamus, pons, and cortex. Neurobiol Aging 2004;25:185–192. Raz N, Gunning-Dixon F, Head D, Rodrigue KM, Williamson A, Acker JD. Aging, sexual dimorphism, and hemispheric asymmetry of the cerebral cortex: replicability of regional differences in volume. Neurobiol Aging 2004;25:377–396. DeCarli C, Massaro J, Danielle H, et al. Measures of brain morphology and infarction in the Framingham heart study: establishing what is normal. Neurobiol Aging 2005 (in press). Blatter DD, Bigler ED, Gale SD, et al. Quantitative volumetric analysis of brain MR: normative database spanning 5 decades of life. Am J Neuroradiol 1995;16:241–251. Good CD, Johnsrude IS, Ashburner J, Henson R, Friston KJ, Frackowiak R. A voxel-based morphometric study of ageing in 465 normal adult human brains. Neuroimage 2001;14:21–36. Liu R, Lemieux L, Bell GS, et al. A longitudinal study of brain morphometrics using quantitative magnetic resonance imaging and difference image analysis. Neuroimage 2003;20:22–33. Jernigan TL, Fennema-Notestine C. White matter mapping is needed. Neurobiol Aging 2004;25:37–39. Bartzokis G. Age-related myelin breakdown: a developmental model of cognitive decline and Alzheimer’s disease. Neurobiol Aging 2004;25:5–18. Bartzokis G. Quadratic trajectories of brain myelin content: unifying construct for neuropsychiatric disorders. Neurobiol Aging 2004;25:49 – 62. Sowell ER, Thompson PM, Toga AW. Mapping changes in the human cortex throughout the span of life. Neuroscientist 2004;10:372–392. Van Laere KJ, Dierckx RA. Brain perfusion SPECT: age-and sexrelated effects correlated with voxel-based morphometric findings in healthy adults. Radiology 2001;221:810 – 817. Resnick SM, Pham DL, Kraut MA, Zonderman AB, Davatzikos C. Longitudinal magnetic resonance imaging studies of older adults: a shrinking brain. J Neurosci 2003;23:3295–3301. Chan D, Fox NC, Jenkins R, Scahill RI, Crum WR, Rossor MN. Rates of global and regional cerebral atrophy in AD and frontotemporal dementia. Neurology 2001;57:1756 –1763. Wang DM, Doddrell DM. MR image-based measurement of rates of change in volumes of brain structures. Part 1: method and validation. Magn Reson Imaging 2002;20:27– 40. Thompson PM, Hayashi KM, de Zubicaray G, et al. Dynamics of gray matter loss in Alzheimer’s disease. J Neurosci 2003;23:994 –1005. Tang Y, Whitman GT, Lopez I, Baloh RW. Brain volume changes on longitudinal magnetic resonance imaging in normal older people. J Neuroimaging 2001;11:393– 400. Hubbard BM, Anderson JM. A quantitative study of cerebral atrophy in old age and senile dementia. J Neurol Sci 1981;50:135–145. Braak H, Braak E. Frequency of stages of Alzheimer-related lesions in different age categories. Neurobiol Aging 1997;18:351–357. Price JL, Ko AI, Wade MJ, Tsou SK, McKeel DW, Morris JC. Neuron number in the entorhinal cortex and CA1 in preclinical Alzheimer disease. Arch Neurol 2001;58:1395–1402. Jernigan TL, Salmon DP, Butters N, Hesselink JR. Cerebral structure on MRI, part II: specific changes in Alzheimer’s and Huntington’s diseases. Biol Psychiatry 1991;29:68 – 81. Wang DM, Chalk JB, Rose SE, et al. MR image-based measurement of rates of change in volumes of brain structures. Part II: application to a study of Alzheimer’s disease and normal aging. Magn Reson Imaging 2002;20:41– 48. Fox NC, Freeborough PA. Brain atrophy progression measured from registered serial MRI: validation and application to Alzheimer’s disease. J Magn Reson Imaging 1997;7:1069 –1075. Schott JM, Fox NC, Frost C, et al. Assessing the onset of structural change in familial Alzheimer’s disease. Ann Neurol 2003;53:181–188. DeKosky ST, Marek K. Looking backward to move forward: early detection of neurodegenerative disorders. Science 2003;302:830 – 834. March (2 of 2) 2005

NEUROLOGY 64

1039