Brain aging research relies mostly on cross-sectional studies, which infer true changes from age differences. We present longitudinal measures of five-year change in the regional brain volumes in healthy adults. Average and individual differences in volume changes and the effects of age, sex and hypertension were assessed with latent difference score modeling. The caudate, the cerebellum, the hippocampus and the association cortices shrunk substantially. There was minimal change in the entorhinal and none in the primary visual cortex. Longitudinal measures of shrinkage exceeded cross-sectional estimates. All regions except the inferior parietal lobule showed individual differences in change. Shrinkage of the cerebellum decreased from young to middle adulthood, and increased from middle adulthood to old age. Shrinkage of the hippocampus, the entorhinal cortices, the inferior temporal cortex and the prefrontal white matter increased with age. Moreover, shrinkage in the hippocampus and the cerebellum accelerated with age. In the hippocampus, both linear and quadratic trends in incremental age-related shrinkage were limited to the hypertensive participants. Individual differences in shrinkage correlated across some regions, suggesting common causes. No sex differences in age trends except for the caudate were observed. We found no evidence of neuroprotective effects of larger brain size or educational attainment.
Introduction
Knowledge about the aging brain is derived mostly from cross-sectional studies [Raz, 2000; Sullivan and Pfefferbaum, 2003; Hedden and Gabrieli, 2004; Raz, 2004]. Such studies estimate the average rate of aging from correlations with age but, unlike longitudinal investigations, are incapable of directly gauging rates of change and individual differences therein. Cross-sectional evidence suggests that in healthy adults, age-related volume reduction is more pronounced in gray [especially prefrontal] matter, and shrinkage of sensory and entorhinal cortices is virtually nil [Raz, 2000]. Thus far, longitudinal studies, with only a few exceptions, have used global indices of brain integrity, and reveal little about regional change [Raz, 2004]. Three exceptions are longitudinal studies revealing significant shrinkage of prefrontal regions with smaller but significant declines of other regions [Pfefferbaum et al., 1998; Resnick et al., 2003; Scahill et al., 2003]. However, in those studies the sample size and/or the number of examined regions were limited. In addition, the extant longitudinal studies, while relying on the samples of generally healthy adults, included some participants with cardiovascular illness, which is common in older persons. The effects of mild vascular conditions, which can exert subtle but detectable negative influence on the brain and cognition [Raz et al., 2003a], have not been examined in the context of longitudinal change.
The studies of brain change have also been limited by reliance on standard linear models that emphasize average trends, are oblivious to measurement issues and disregard individual differences in regional changes, thus obscuring the heterogeneity in brain aging. Those methods assume, without testing, that the same construct is measured over time without separating construct variance from specific variance and measurement error. New longitudinal methods, such as Two-Occasion Latent Difference Modeling [LDM] [McArdle and Nesselroade, 1994], alleviate most of those problems by greatly reducing unreliability of difference scores, examining mean change and individual differences within the same framework, and formally testing measurement equivalence across occasions and groups [cf. Meredith, 1964]. In this study, we used LDM to examine average changes, individual differences in change and covariances of change in multiple brain regions, with attention to departures from linearity. We also assessed the associations of brain volumes at baseline and brain volume changes with age, sex and vascular health [hypertension]. While some average age trends in a portion of this sample have been reported [Raz et al., 2003b,c, 2004a], most cortical regions, variability of change and the influence of health-related factors have not yet been examined.
On the basis of cross-sectional [Raz, 2000; Bartzokis et al., 2001; Jernigan et al., 2001; Raz, 2004] and longitudinal [Pfefferbaum et al., 1998; Resnick et al., 2003] findings, we hypothesized the steepest decline in the lateral prefrontal cortex, with smaller shrinkage of the temporal association cortices and sparing of the primary visual cortex and the inferior parietal lobule. Because of reported nonlinear cross-sectional age trends [Courchesne et al., 2000; Bartzokis et al., 2001, 2004; Jernigan et al., 2001; Raz et al., 2004b], we hypothesized acceleration of white matter shrinkage with age. Connectivity between the prefrontal cortex and the striatum [Alexander et al., 1986], and age-related shrinkage in both [Raz et al., 2003b; Rodrigue and Raz, 2004], suggested significant associations between changes in those regions. In addition, we tested hypotheses that hypertension exacerbates age-related shrinkage in prefrontal regions [Raz et al., 2003a], that women show lesser brain aging than men [Coffey et al., 1998], that larger brain volume is a neuroprotective factor [Satz, 1993], and that higher formal education delays brain aging and ameliorates its course [Stern et al., 1992; Kramer et al., 2004].
Materials and Methods
Participants
The data for this study were collected in a major metropolitan area in the USA. The participants of multiple cross-sectional studies [Raz et al., 1997, 1999, 2000, 2004b] were contacted by mail and telephone, and invited for the longitudinal follow-up. Of 323 eligible persons, 226 [70%] responded to the invitation and 140 [43% of the total eligible pool, or 62% of the responders] agreed to participate in the study. The participants signed a consent form approved by the University Committee for Protection of Human Subjects in Research and by the Hospitals Patients Participation Committee. All participants were screened with a mail-in health questionnaire completed by the participants and augmented by telephone and personal interviews.
Of 140 persons who agreed to participate, 127 [91%; 39% of the initial eligible pool] completed the follow-up study. Screening criteria applied to the follow-up sample were identical to those used to determine eligibility at the first occasion. Persons who reported a history of severe cardiovascular, neurological or psychiatric conditions, head trauma with loss of consciousness for >5 min, thyroid problems, diabetes, treatment for drug and alcohol problems, or a habit of taking three or more alcoholic drinks per day were excluded from the study. Participants with two types of cardiovascular problems were included in the analysis: hypertension controlled by medication [19 cases] and mitral valve prolapse [three cases]. Hypertensive participants took standard hypertension treatment: beta-blockers, calcium channel blockers, ACE-inhibitors and potassium-sparing diuretics. None of the participants used anti-seizure medication, anxiolytics or antidepressants. Persons who suffered or suspected they suffered from claustrophobia were advised not to participate in the study.
Of the 127 subjects who completed the follow-up, 29 [23%] were not included in the data analyses reported here because they no longer met the health screening criteria. The excluded participants ranged in age from 31 to 83 years, and the reasons for exclusion were Parkinson's disease, cerebral hemorrhage, cardiac bypass surgery, angioplasty, hypo- and hyperthyroidism, diabetes mellitus and cancer. The follow-up magnetic resonance imaging [MRI] data on an additional 26 subjects [20%] were either completely lost [four cases] or not suitable for a longitudinal analysis [22 cases that were acquired with a different field of view]. Including those cases would have confounded time of measurement with coarser resolution after re-slicing the acquired images. Thus, the final sample consisted of 72 participants [23% of the eligible cohort, 30 men].
All subjects were screened for dementia and depression using a modified Blessed Information-Memory-Concentration Test [BIMC] [Blessed et al., 1968] with a cut-off of 85% correct, Mini-Mental State Examination [MMSE] [Folstein et al., 1975] with a cut-off of 26 [87% correct] and Geriatric Depression Questionnaire [CES-D] [Radloff, 1977] with a cut-off of 15. The BIMC and CES-D were administered on both testing occasions whereas the MMSE was used only at follow-up. All participants were consistent right-handers, as indicated by a score above 75% on the Edinburgh Handedness Questionnaire [Oldfield, 1971]. An experienced neuroradiologist [J.D.A.] examined the MR scans for space-occupying lesions and signs of significant cerebrovascular disease. Only 10% of the participants smoked tobacco and 63% exercised at least once a week, regardless of their sex or hypertension status [both χ2 < 1]. Some of the participating women [31%] were on hormone replacement therapy [HRT] at the inception of the study.
To estimate the selectivity of the sample, we compared the returnees with the rest of the original sample. The returnees were older than the participants who failed to return for follow-up [52.49 versus 44.82, t = 3.53, P > 0.001] and had somewhat higher vocabulary scores [27.12 versus 24.97, t = 2.15, P < 0.05]. However, the returning sample did not differ from the original one in general cognitive status [BIMC = 90% correct], education [15 years] or sex composition [58% women].
The mean ± SD follow-up interval was 5.27 ± 0.30 years, with a range from 4.83 to 6.08 years; it did not differ between the sexes [t < 1]. The age of the participants at baseline ranged from 20 to 77 years [mean ± SD = 52.56 ± 14.05 years, 50.33 for men and 54.14 for women, t = 1.07, ns]. Average education was 15.90 ± 2.40 years; there was a trend for men to have more formal education than women: 16.46 versus 15.45 years, t = 1.70, P < 0.10. Despite the MMSE cut-off score of 26 to exclude subjects with dementia, only one person [a 33-year-old normotensive woman] scored below the cut-off at follow-up. Most MMSE scores at the time of second scanning were substantially higher than the cut-off [28.86 ± 1.11], and were unrelated to age [r = −0.06, ns] or sex [t = 0.58, ns]. The participants who were excluded from the analysis because of the wrong field of view parameter did not differ from those included in the sample in age, education, MMSE or BIMC scores [all t < 1.12, ns].
Twelve participants [four men and eight women] with medically controlled hypertension were also included in the sample at baseline. In addition, six women and one man received diagnoses of hypertension during the follow-up period. Thus, a total of 19 treated hypertensive subjects [14 women] were included. The hypertensive subjects were older than the remainder of the sample [mean age at baseline = 63.79 ± 12.99 years, t = 2.29, P < 0.05], but had the same number of years of formal schooling [t = 0.80, ns] and equivalent MMSE scores [t = −1.59, ns].
MRI Protocol
All imaging was performed on 1.5 T Signa scanners [General Electric Co., Milwaukee, WI] installed in the same hospital. However, whereas only one scanner was used for the baseline data collection, two additional scanners were employed for re-testing four of the subjects at follow-up. One subject was re-scanned on an identical GE Signa scanner located in an adjacent room and three subjects were re-scanned on a mobile 1.5 T GE scanner located at the entrance to the imaging center. The scanners were routinely calibrated using the same standard GE phantom.
At baseline and follow-up all subjects were scanned with identical pulse sequences. Sagittal localizer images with repetition time [TR] = 400 ms, echo time [TE] = 16 ms, one excitation and slice thickness = 5 mm were acquired first. Dual-echo fast spin echo [FSE] T2 and proton-density weighted axial images [TR/TR = 3300/90ef or 18ef, slice thickness = 5 mm, and inter-slice gap = 2.5 mm] were acquired to screen for cerebrovascular disease. Volumes were measured on two sets of images acquired using a T1-weighted 3-D spoiled gradient recalled [SPGR] sequence with 124 contiguous axial slices, TE = 5 ms, TR = 24 ms, square field of view = 22 cm, acquisition matrix = 256 × 192, slice thickness = 1.3 mm and flip angle = 30°.
MR Image Processing
Image processing and regional volume measures are described in detail elsewhere [Raz et al., 2003b,c]. The images acquired on both occasions were coded and the order of their tracing was randomized within each subject by a person other than the operators who traced the regions of interest [ROIs]. The operators were blind to the time of acquisition of the specific images and the demographic characteristics of the participants, as well as to the scanner on which the images were acquired. To ensure the blindness of the operators, the baseline measurements previously published in cross-sectional studies [Raz et al., 1997, 1998, 2001, 2004b] were not used and all structures were measured anew. Reliability of ROI measures [intraclass correlation for random raters] [Shrout and Fleiss, 1979] exceeded 0.90, as in Raz et al. [2004b].
The volumes were computed from measured areas of the ROIs on consecutive slices [for details, see Raz et al. 2004b]. The following ROIs were measured: the intracranial vault [ICV], lateral prefrontal cortex [LPFC], orbital frontal cortex [OFC], adjacent prefrontal white matter [PFw], inferior parietal lobule [IPL] and adjacent white matter [IPw], inferior temporal cortex [IT], fusiform cortex [FF], visual [pericalcarine] cortex [VC], hippocampus [HC], entorhinal cortex [EC], striatal nuclei [caudate, Cd] and cerebellar hemispheres [CbH]. The illustrations of the traced ROIs are presented in Figures 1–4. Additional images are available in our previous publications [Raz et al., 2001, 2004b,c].
Examples of ROI demarcation on typical slices of MR images. Top panel: lateral prefrontal cortex, and orbito-frontal cortex, with a line indicating separation between the two. Bottom panel: inferior temporal [IT] and fusiform [FG] cortices.
Examples of ROI demarcation in the posterior cortex. Top panel: a series of slices with traced primary visual cortex [VC]. Bottom panel: a typical slice with inferior parietal lobule [IP] and inferior parietal white matter [IPw] traced.
Examples of ROI demarcation in the medial temporal cortex. Top panel: a series of slices with the hippocampus [HC] outlined. Bottom panel: a series of slices with entorhinal cortex outlined.
Examples of non-cortical ROIs demarcation. Top panel: a series of slices used in measuring cerebellar volume. Bottom panel: a typical slice with the caudate nucleus outlined.
Data Conditioning
Before conducting the LDM analyses, we examined the data for possible sources of systematic error. The effects of the scanner [scanner 1, scanner 2 and the mobile scanner] on the intracranial volume measured at time 2 were examined using two separate linear models. In these models ICV was the dependent variable, scanner and participant's sex were grouping factors, and age was a continuous independent variable. The results of these analyses demonstrated that changes in scanner occurring between baseline and follow-up did not affect measured intracranial volume, i.e. produced no systematic bias. The mean intracranial volume remained stable across the five-year delay, showing a total mean change of only −0.3% [t < 1, ns].
Intracranial volume differed between the sexes, with men having larger crania [t[70] = 6.82 and 6.85 for baseline and follow-up, both P < 0.001]. It was not, however, correlated with age: r = −0.04 and −0.05 for baseline and follow-up, respectively, both ns. Therefore, ICV was used to adjust the regional volumes for sex differences in body size. The adjustment was performed on each ROI volume in each hemisphere via a formula based on the analysis of covariance approach: adjusted volume = raw volume − b × [ICV − mean ICV], where b is the slope of regression of an ROI volume on ICV. The adjusted volumes were used as dependent variables in the analyses presented below.
Latent Difference Modeling
We applied LDM [McArdle and Nesselroade, 1994] to assess mean changes as well as variances and covariances of change in regional volumes. Difference scores based on observed variables are especially vulnerable to the consequences of less than perfect reliability of measurement [Cronbach and Furby, 1970; cf. Rogosa and Willett, 1985; Baltes et al., 1988]. LDM circumvents those problems by handling mean change and individual differences [i.e. variance] in change at the level of latent factors [see Fig. 5; cf. McArdle and Nesselroade, 1994]. In the present set of LDM analyses, ROIs were defined as latent factors representing the variance common to the two hemisphere of a given ROI and measurement occasion [see Fig. 5]. For example, a latent factor of hippocampal volume at baseline [time T1] was estimated by the volume of the left and right hippocampi measured at that occasion. The same was done for follow-up [time T2], and the difference between hippocampal volumes at T1 and T2 was expressed as the difference between T1 and T2 latent factors of the hippocampus. LDM were established for each ROI to compute average change and individual differences in change [i.e. the variance of change]. Specifically, at the latent level, five parameters were estimated: [i] mean volume at T1; [ii] mean volume change from T1 to T2; [iii] variance in volume at T1; [iv] variance in volume change from T1 to T2; and [v] the covariance between volume at T1 and change in volume between T1 and T2. In Figure 5, these five latent parameters are designated as α, β, γ, δ and ε, respectively.
Measurement model for the assessment of two-occasion changes in regional brain volume. Squares represent observed variables, circles represent latent variables and the triangle serves to represent information regarding means and intercepts. Free parameters are indicated by an asterisk. Parameters with equal sign and the same subscript are constrained to be equal to each other. T1 = baseline; T2 = follow-up; VL = regional volume of left hemisphere; VR = regional volume of right hemisphere; V = regional latent volume; ΔV = difference in regional volume between first and second occasions; α = latent mean of regional brain volume at first occasion [baseline]; β = mean difference between latent regional brain volumes at first and second occasions; γ = variance [individual differences] in latent regional brain volume at first occasion; δ = variance [individual differences] in latent regional brain volume changes between first and second occasions; ε = covariance between individual differences in regional brain volume at first occasion and individual differences in regional brain volume changes. The model has four observed variables, 10 free parameters and four degrees of freedom. For further information on two-occasion latent difference modelling in general, see McArdle and Nesselroade [1994].
To ensure interpretability and identification of the models, the following equality constraints were imposed on the latent difference models: [i] residual means of the left hemisphere across time to accommodate possible time-invariant hemisphere differences in volume; [ii] unique variances of the left hemisphere across time; [iii] unique variances of the right hemisphere across time; [iv] factor loadings of the right hemisphere across time; and [v] autocorrelated residuals of the two hemispheres. To define the metric of latent factors, the factor loading of the left hemispheres were set to unity at both occasions. With four observed variables and 10 free parameters, this measurement model is overidentified with four degrees of freedom, which means that it can be estimated empirically [e.g. Kline, 1998]. Tenability of this model is consistent with the assumption of strict metric invariance [Meredith, 1964], which implies that the relation between observed and latent variables does not change over time so that differences at the latent level can be interpreted with confidence. Note also that a mathematically equivalent autoregressive counterpart exists for each two-occasion LDM [McArdle and Nesselroade, 1994].
We first computed univariate LDMs separately for each of the 12 ROIs to examine baseline volume, mean change, variance in change, and associations of initial level and change with linear age, quadratic age, sex and diagnosis of hypertension by the follow-up time. The latter variables were specified as time-invariant covariates measured without error. Then, for the 11 ROIs showing reliable inter-individual differences in change, we specified a single overall multivariate LDM to explore covariance relations across ROIs. To obtain numerically balanced matrices, chronological age and cerebellar volumes were divided by 10.
Results
Descriptive Statistics
Descriptive statistics for ICV-adjusted regional volumes [mean, SD and coefficient of variation[CV]], as well as their correlations with age at baseline and at the five-year follow-up, are presented in Table 1. Values refer to observed scores averaged over hemispheres. For each ROI and for each scanning occasion, we gauged the strength of association between the adjusted volume and age, as well as the change in volume over time. The coefficients of stability [Pearson correlations between T1 and T2 measures of all ROI volumes] were high, ranging between r = 0.89 for the fusiform cortex to r = 0.97 for the inferior parietal lobule, median r = 0.91. The plots of changes in cortical and subcortical ROI volumes as a function of age are presented in Figures 6–8. The figures show age-related differences in the magnitude and shape of age–volume associations at both measurement occasions, individual differences in change between the occasions and the differences in magnitude and rate of change of the cortical ROIs.
Longitudinal changes in adjusted volumes of the lateral prefrontal, orbito-frontal, inferior temporal and fusiform cortices as a function of baseline age.
Longitudinal changes in adjusted volumes of the inferior parietal and primary visual [pericalcarine] cortices as well as prefrontal and inferior parietal white matter as a function of baseline age.
Longitudinal changes in the adjusted volumes of the hippocampus, the entorhinal cortex, the caudate nucleus and the cerebellar hemispheres as a function of baseline age.
Table 1
Descriptive statistics and age-related differences in regional brain volumes at two measurement occasions
ROI . Regional volumes. . . . . . . . . T1. . . . T2. . . . . Mean. SD. CV. rage. Mean. SD. CV. rage. LPFC 9.12 1.04 0.11 −0.55*** 8.69 1.12 0.13 −0.59*** OFC 5.03 0.61 0.12 −0.41** 4.80 0.60 0.13 −0.37* IT 6.07 0.69 0.11 −0.20 5.85 0.74 0.13 −0.26 FF 9.87 0.86 0.09 −0.31 9.63 0.93 0.10 −0.29 IPL 6.31 1.23 0.20 −0.08 6.03 1.20 0.20 −0.13 VC 2.56 0.31 0.12 −0.11 2.54 0.29 0.12 −0.10 PFw 20.37 2.40 0.12 −0.27 19.83 2.74 0.14 −0.41** IPw 5.47 1.06 0.19 −0.11 5.33 1.06 0.20 −0.01 HC 3.43 0.39 0.11 −0.41** 3.29 0.39 0.11 −0.50*** EC 1.38 0.18 0.12 −0.01 1.35 0.18 0.13 −0.12 Cd 4.61 0.61 0.12 −0.38* 4.43 0.55 0.12 −0.42** CbH 67.55 6.28 0.09 −0.36* 65.36 5.95 0.09 −0.40**
ROI . Regional volumes. . . . . . . . . T1. . . . T2. . . . . Mean. SD. CV. rage. Mean. SD. CV. rage. LPFC 9.12 1.04 0.11 −0.55*** 8.69 1.12 0.13 −0.59*** OFC 5.03 0.61 0.12 −0.41** 4.80 0.60 0.13 −0.37* IT 6.07 0.69 0.11 −0.20 5.85 0.74 0.13 −0.26 FF 9.87 0.86 0.09 −0.31 9.63 0.93 0.10 −0.29 IPL 6.31 1.23 0.20 −0.08 6.03 1.20 0.20 −0.13 VC 2.56 0.31 0.12 −0.11 2.54 0.29 0.12 −0.10 PFw 20.37 2.40 0.12 −0.27 19.83 2.74 0.14 −0.41** IPw 5.47 1.06 0.19 −0.11 5.33 1.06 0.20 −0.01 HC 3.43 0.39 0.11 −0.41** 3.29 0.39 0.11 −0.50*** EC 1.38 0.18 0.12 −0.01 1.35 0.18 0.13 −0.12 Cd 4.61 0.61 0.12 −0.38* 4.43 0.55 0.12 −0.42** CbH 67.55 6.28 0.09 −0.36* 65.36 5.95 0.09 −0.40**
Table 1
Descriptive statistics and age-related differences in regional brain volumes at two measurement occasions
ROI . Regional volumes. . . . . . . . . T1. . . . T2. . . . . Mean. SD. CV. rage. Mean. SD. CV. rage. LPFC 9.12 1.04 0.11 −0.55*** 8.69 1.12 0.13 −0.59*** OFC 5.03 0.61 0.12 −0.41** 4.80 0.60 0.13 −0.37* IT 6.07 0.69 0.11 −0.20 5.85 0.74 0.13 −0.26 FF 9.87 0.86 0.09 −0.31 9.63 0.93 0.10 −0.29 IPL 6.31 1.23 0.20 −0.08 6.03 1.20 0.20 −0.13 VC 2.56 0.31 0.12 −0.11 2.54 0.29 0.12 −0.10 PFw 20.37 2.40 0.12 −0.27 19.83 2.74 0.14 −0.41** IPw 5.47 1.06 0.19 −0.11 5.33 1.06 0.20 −0.01 HC 3.43 0.39 0.11 −0.41** 3.29 0.39 0.11 −0.50*** EC 1.38 0.18 0.12 −0.01 1.35 0.18 0.13 −0.12 Cd 4.61 0.61 0.12 −0.38* 4.43 0.55 0.12 −0.42** CbH 67.55 6.28 0.09 −0.36* 65.36 5.95 0.09 −0.40**
ROI . Regional volumes. . . . . . . . . T1. . . . T2. . . . . Mean. SD. CV. rage. Mean. SD. CV. rage. LPFC 9.12 1.04 0.11 −0.55*** 8.69 1.12 0.13 −0.59*** OFC 5.03 0.61 0.12 −0.41** 4.80 0.60 0.13 −0.37* IT 6.07 0.69 0.11 −0.20 5.85 0.74 0.13 −0.26 FF 9.87 0.86 0.09 −0.31 9.63 0.93 0.10 −0.29 IPL 6.31 1.23 0.20 −0.08 6.03 1.20 0.20 −0.13 VC 2.56 0.31 0.12 −0.11 2.54 0.29 0.12 −0.10 PFw 20.37 2.40 0.12 −0.27 19.83 2.74 0.14 −0.41** IPw 5.47 1.06 0.19 −0.11 5.33 1.06 0.20 −0.01 HC 3.43 0.39 0.11 −0.41** 3.29 0.39 0.11 −0.50*** EC 1.38 0.18 0.12 −0.01 1.35 0.18 0.13 −0.12 Cd 4.61 0.61 0.12 −0.38* 4.43 0.55 0.12 −0.42** CbH 67.55 6.28 0.09 −0.36* 65.36 5.95 0.09 −0.40**
To check further the degree of selectivity in the longitudinal sample we compared the magnitude of association with age exhibited by the regions studied in the longitudinal sample among the returnees and participants who did not return. In 65% of the examined ROIs, the association between volume and age was numerically, though not statistically, stronger in the remainder of the sample compared to the returnees, and in none was the reverse observed.
Univariate LDM
Results are summarized in Table 2. Standardized factor loadings of hemispheric indicators on the latent volume factors ranged from 0.88 [fusiform gyrus] to 0.99 [cerebellum], with a mean of 0.92. Also, inspection of the path testing for time-invariant differences in volume between right and left hemispheres indicated that the assumption of equally large hemispheres could not be rejected for any of the regions. By conventional criteria [Kline, 1998], all models, with the exception of those for the IPL and VC, fit the data well. Specifically, for the former 10 models, comparative fit indexes [CFI] were 0.98 or higher, standardized root-mean-square residuals [SRMRs] were