Journal of Cerebral Blood Flow and Metabolism 8:502-512 © 1988 Raven Press, Ltd., New York

Reproducibility of Cerebral Glucose Metabolic Measurements in Resting Human Subjects

Elsa J. Bartlett, Jonathan D. Brodie, * Alfred P. Wolf, *David R. Christman, tEugene Laska, and tMorris Meissner

Department of Psychiatry, NYU Medical Center, New York; *Department tJfChl!inistry, Brookhaven National Laboratory, tStatistical Sciences and Epidemiology Division, Nathan Kline Institute for Psychiatric Research,

Orangeburg, New York, U.S.A.

Summary: Positron emission tomography with IlC-2-de­oxyglucose was used to determine the test-retest vari­ability of regional cerebral glucose metabolism in 22 young normal right-handed men scanned twice in a 24-h period under baseline (resting) conditions. To assess the effects of scan order and time of day on variability, 12 subjects were scanned in the morning and afternoon of the same day (a.m.-p.m.) and 10 in the reverse order (p.m.-a.m.) with a night in between. The effect of anxiety on metabolism was also assessed. Seventy-three percent of the total subject group showed changes in whole brain metabolism from the first to the second measurement of 10% or less, with comparable changes in various cortical and subcortical regions. When a scaling factor was used to equate the whole brain metabolism in the two scans for each individual, the resulting average regional changes

Positron emission tomography (PET) with deoxy­glucose is now a generally accepted method of measuring regional brain glucose metabolism (Rei­vich and Alavi, 1985; Phelps et al. 1986). In many studies, the effect of some stimulation is ascer­tained in comparison with baseline metabolism in a test-retest repeated-measure design. However, to have some assurance that the difference between the two measurements or runs is a valid estimate of the effect of stimulation, one must have some knowledge of the reproducibility of the baseline

Received September IS, 1987; accepted December 24. 1987. Address correspondence and reprint requests to Dr. J. Brodie,

Department of Psychiatry, NY U Medical Center, 550 First Av­enue, New York, NY 10016, U.S.A.

Abbreviations used: PET, positron emission tomography; CDG. Ilc-2-deoxyglucose; ROI, region of interest; CT, com­puted axial tomography; RCMRG, regional cerebral metabolic rate for glucose; CMRG, cerebral metabolic rate for glucose; MANOVA, multivariate analysis of variance.

502

for each group were no mote than 1 %. This suggests that the proportion of the whole brain metabolism utilized re­gionally is stable in a group of subjects over time. Both groups of subjects had lower morning than afternoon me­tabolism, but the differences were slight in the p.m.-a,m. group. One measure of anxiety (pulse at fUll 1) was cor­related with run 1 metabolism and with the percentage of change from run 1 to run 2. No significant run 2 correla­tions were observed. This is the first study to measure test-retest variability in cerebral glucose metabolism in a large sample of young normal subjects. It demonstrates that the deoxyglucose method yields Itlw intrasubject variability and high stability OVer a 24-h period. Key Words: PET - Reprodueibility - Brain, metabolism­Deoxyglucose.

measurements. Ideally, the mean change in a group of individuals whose measurements Were mnde re­peatedly under similar conditiohs wtluld be zero.

The several prior reports of test-retest variability under baseline (resting) conditions have only used small samples (Reivich et aI., 1982; Our et a1., 1987b; Brooks et aI., 1987; Duara et al'! 1987). Ad­ditionally, previous research hfl.!S indicated several potential sources of unwanted bias and/or variance" in such studies. These include variability due to an­atomic repositioning inaccuracies (Chang et al., 1987); partial volume effects (Mazziotta et aI., 1981); differences in subject characteristics such as age (Kuhl et aI., 1982; Duara et aI., 1984), brain size (Hatazawa et aI., 1987), gender, and handedness (Gur et aI., 1982); and differences in the subjects' level of anxiety during scanning (Gur et aI., 1987a). Previous studies in our own laboratory have indi­cated that scan order and time of day may also af­fect test-retest variability (i.e., when subjects were

REPRODUCIBILITY OF CEREBRAL GLUCOSE MEASUREMENTS 503

scanned twice in a single day, the second set of measurements tended to show increased metabolic activity).

The present study is designed to estimate the sta­bility of measurements in a test-retest paradigm conducted during a 24-hour period and thus to ad­dress the suitability of the resting state as a refer­ence state for assessing the effect of stimulation in PET studies.

METHODS

Subjects Twenty-two right-handed male volunteers between the

ages of 18 and 3 7 years (X age = 23.2 ± 3.9) were studied. All were found on medical examination to be clinically healthy and to have no abnormal neurological, metabolic or psychiatric history. Their level of education ranged from 13 to more than 16 years of schooling. Hand­edness was determined by the Edinburgh Handedness Questionaire (Oldfield, 1971).

Twelve subjects (the a.m./p.m. group) were first scanned in the morning and again the same afternoon; ten subjects (the a.m./p.m. group) were scanned in the after­noon and again the following morning.

Pet scan procedures To minimize radiation exposure and still achieve ade­

quate counting statistics, all subjects were scanned in the low resolution mode on the PETT VI. Spatial resolution at full width half maximum under these conditions was 11.8 mm in the plane of section and 14.4 mm in the axial direction.

One to two hours pror to isotope injection, each sub­ject's head was positioned for scanning so that the hori­zontal plane of the image was parallel to the subject's canthomeatal line. A removable plastic headholder was then molded to the contours of the subject's head to in­sure accurate repositioning during subsequent PET scan­ning (Kearfott et aI., 1984). Two seven-slice tomographic transmission scans were performed using a 68Ge/68Ga ring source. These scans were used for attenuation cor­rection and to define the size and center of the brain for each PET image.

Forty-five minutes prior to injection, subjects were seated in an easy chair in a room adjacent to the scanner. Catheters were inserted into a dorsal vein of the left hand for blood sampling and a right antecubital vein for isotope administration. Each subject's left hand was heated to maintain skin temperature at 44°C so as to arterialize venous blood (Phelps et aI., 1979).

Five minutes prior to injection, the lights were dimmed and subjects were asked to relax, to refrain from conver­sation, and to remain awake. No other instructions were given. All subjects were studied with their eyes open and their ears unoccluded.

The subjects were injected with a bolus of between 6.3 and 8.6 mCi of IIC-2-deoxyglucose (CDG) prepared as described in MacGregor et al. (1981). Arterialized blood samples were taken and the plasma assayed for radioac­tivity. At 27 min after the injection, the subjects were po­sitioned in the PET scanner. The subjects were scanned at 35 min and again at 45 min postinjection. Each scan resulted in a set of seven images obtained simultaneously

with centers 1.44 cm apart. After the first set of images was obtained, the head was moved 7.2 mm; this resulted in a set of 14 interleaved PET images slightly displaced in time.

Each subject received two isotope injections. Subjects in the a.m.-p.m. group were injected at 10:30 a.m. and again at 1:30 p.m. or at 12: 00 noon and again at 3: 00 p.m. Subjects in the p.m.-a.m. group were first injected at the p.m. times and again at the a.m. times the following morning.

Image analysis Regions of interest for analysis were obtained from the

eight contiguous images that were available for all sub­jects. These were located between approximately 9.6 and 4.7 cm above the canthomeatal line. Regions of interest were outlined by drawing boundaries from the standard neuroanatomical atlas of Matsui and Hirano (1978) onto corresponding metabolic images. Metabolic values were obtained from four large cortical gray matter regions of interest (frontal, parietal, temporal, and occipital lobes); a white matter region of interest (centrum semiovale); and a subcortical region of interest (thalamus). Each re­gion of interest was represented on several PET images. Whole-slice values (which included ventricles and gray and white matter) were also obtained for each of the eight images. Boundaries for whole-slice regions of interest were drawn by computer software at the outermost pixels, representing the 30th percentile of metabolic ac­tivity for a given image. This cut-off best approximates the outermost limit of cortical tissue as determined in our laboratory by transmission scans and computerized to­mography (CT). All pixels within that perimeter were in­cluded in the whole-slice region of interest. A whole brain region of interest, defined as the sum of the eight whole-slice regions of interest, was also obtained.

Regions of interest were chosen a priori to reflect ac­tivity in the major cortical lobes as well as in a subcor­tical and a white matter region large enough to have re­covery coefficients of at least 0.7 (Mazziotta et aI., 1981).

To insure that regions of interest had comparable size and shape across runs, regions of interest drawn on images obtained from one run were copied by computer software onto comparable images from the other run. To control for the effects of drawing versus copying, regions of interest were drawn on images from the first or second run in an alternating order.

Repositioning Accuracy of anatomical repositioning in the axial di­

rection from run I to run 2 was estimated by an axial repositioning ratio. This compared the size (in number of pixels) of the whole-slice region of interest of the top­most image obtained at run 1 and run 2:

(ROI-l- ROI-2)/[(ROI-l + ROI-2)/2]

Accuracy of repositioning in the transaxial direction was estimated by a ratio that compared the sizes (in number of pixels) of the bottom-most temporal pole regions ob­tained at run I and run 2; one pixel equalled 2.8 x 2.8 mm. Data from the left and right temporal pole regions were compared separately;

left-l ieft-2/[(leftl +- left-2)/2]

and

right-l - right-2/[(right-l + right-2)/21

J Cereh Blood FirM Me/ab, Vol. 8, No.4, 1988

504 E. J. BARTLETT ET AL.

The resulting repositioning ratios enabled us to rank­order subjects according to the relative accuracy of their repositioning for each set of anatomically comparable scans.

Data quantitation Regional cerebral metabolic rates for glucose

(RCMROs) are reported as the mean metabolic value cal­culated from the data in the constituent pixels of a given region of interest, using the Sokoloff et al. (1977) model as extended by Huang et al. (1980) and modified for the use of cno by Reivich et al. (1982, 1985). In addition, to study variability in regional metabolism in the presence of change in whole brain metabolism across the two runs, the following adjustment was made. For each subject, we multiplied all run 2 regional values by a global metabolic scaling factor so that the resulting run 2 values would be scaled to a whole brain metabolic rate equal to that of run 1. The factor for each subject was whole-brain- I 1whole­brain-2.

Other measures Other measures of anxiety were also obtained for each

subject at each run. Measures of pulse, blood pressure,

and skin temperature were obtained for each subject at 20 min prior to injection, at injection, and at 20 min after injection for each run. Baseline measures of pulse and blood pressure were also obtained when subjects first ar­rived at Brookhaven National Laboratory. Pulse and blood pressure were measured with a sphygmomanom­eter. Skin temperature was measured at the tip of the middle finger by a Model 403 Temperature Unit (Human Measurement Systems).

The subjective level of stress associated with specific PET procedures was measured by a rating scale adminis­tered to each subject immediately after each run. Sub­jects were asked to rate (on a scale of 0 to 5) the level of stress experienced during each of eight PET procedures: signing the consent form; viewing the scanner for the first time; seeing the headholder; having the head positioned for scanning; insertion of the catheters; having pulse and blood pressure measurements taken; having the isotope injection; having the head inside the scanner. Each sub­ject received a mean stress score for each run, calculated as the average of the ratings assigned to each procedure. The resulting measures enabled us to rank-order subjects according to their subjective stress ratings.

TABLE 1. Run 1 and scaled and unsealed Run 2 regional glucose metabolic rates for individual subjects in the a.m.-p.m. group

Frontal lobe Parietal lobe Occipital lobe

Subjects 2 3 2 3 2 3

1 43.1 40.4 42.8 44.5 41.8 44.4 49.4 45.3 48.1 2 41.4 44.7 41.3 42.5 45.6 42.1 45.4 48.5 44.8 3 43.0 46.3 41.4 43.9 51.2 45.9 46.6 51.4 46.0 4 45.4 47.7 45.0 47.7 47.3 47.0 51.7 54.2 51.0 5 41.5 38.2 41.4 43.3 38.9 42.1 48.1 43.7 47.3 6 44.7 47.9 44.6 47.3 50.3 46.9 50.5 53.7 50.1 7 46.4 51.1 46.6 50.4 54.8 50.0 52.0 57.7 52.6 8 57.7 66.0 57.5 59.8 67.0 58.4 62.1 69.4 60.5 9 42.1 52.2 42.6 43.8 53.2 43.5 44.0 53.9 44.0

10 38.6 45.2 38.4 42.6 51.6 43.8 43.3 52.8 44.9 11 49.6 48.2 49.2 52.4 50.7 51.7 54.5 53.2 54.3 12 48.9 50.4 49.9 53.1 56.1 55.5 54.3 56.3 55.7

Group Mean 45.2 48.2 45.1 47.6 50.9 47.6 50.2 53.3 49.9 SD 5.0 7.0 5.2 5.4 7.2 5.3 5.3 6.6 5.1

Temporal lobe Centrum T halamus Whole brain

Subjects 2 3 2 3 2 3 2

1 43.7 41.5 44.0 32.0 31.5 33.4 40.1 38.0 40.3 36.4 34.3 2 40.3 42.6 39.3 28.8 30.6 28.3 38.0 42.6 39.3 34.2 37.0 3 41.4 45.6 40.9 32.1 27.9 25.0 39.0 42.3 37.9 35.8 40.0 4 44.6 46.3 43.7 36.4 38.2 36.0 45.1 42.8 40.4 40.3 42.7 5 40.5 37.7 40.6 29.4 28.3 30.7 41.2 42.2 45.7 35.0 32.4 6 44.3 46.2 43.1 34.5 37.8 35.2 42.8 43.2 40.2 38.1 40.9 7 46.4 52.1 47.5 41.1 43.1 39.3 39.9 46.5 42.4 41.2 45.2 8 52.2 59.6 51.9 44.1 53.2 46.4 53.6 60.1 52.4 47.6 54.6 9 39.5 48.8 39.9 33.6 40.0 32.6 35.5 43.6 35.6 34.5 42.3

10 37.9 44.7 38.0 32.3 37.4 31.8 35.4 42.4 36.1 33.3 39.1 11 47.8 46.5 47.5 40.8 39.6 40.4 42.7 43.0 43.8 41.3 40.4 12 47.2 48.7 48.2 42.5 46.1 45.6 45.5 45.4 44.9 42.2 42.6

Group Mean 43.8 46.7 43.7 35.6 37.8 35.4 41.6 44.3 41.6 38.3 41.0 SD 4.1 5.5 4.3 5.2 7.5 6.5 5.0 5.4 4.7 4.3 5.6

Key: 1, run 1 regional metabolic rate; 2, run 2 regional metabolic rate; 3, scaled run 2 regional metabolic rate: ROI-2' = ROI-2 x (whole-brain-llwhole-brain-2). All values are expressed as micro­moles of glucose per 100 g of tissue per minute.

J Cereb Blood Flow Metab, Vol. 8, No.4, 1988

REPRODUCIBILITY OF CEREBRAL GLUCOSE MEASUREMENTS 505

Statistical analysis Analyses were performed on RCMRGs expressed as

micromoles per 100 g of tissue per minute. The effect of scan order and time of day was assessed by a two-group (a.m.-p.m. versus p.m.-a.m.) repeated-measure multivar­iate analysis of variance (MAN OVA) (Davidson and To­porek, 1983) with RCMRGs (frontal, temporal, parietal, and occipital lobes; centrum semiovale; and thalamus) as the multivariate observation vector and runs (l versus 2) as the repeated measure. The model included run effects and group effects, and the null hypothesis (no difference in these effects) was tested. This analysis was performed for both the scaled and the un scaled data. The effect of anxiety and repositioning accuracy on metabolism was assessed by computing Spearman rank-order correlation coefficients (Cohen and Cohen, 1975) for regional meta­bolic values and the various measures of anxiety, accu­racy, and brain size.

RESULTS

The effect of scan order and time of day on metabolic values

The regional metabolic rates for each subject are presented in Tables 1 and 2. The MANOVA as­sessing the effects of scan order and time of day on unsealed metabolic values revealed no significant

main effects for group or run but did show a signifi­cant group x run interaction (F(6,IS) = 2.8S, p =

O.OS). One-way repeated measure MANOVAs per­formed separately on the a.m.-p.m. and p.m.-a.m. groups' regional metabolic rates showed that there was no significant difference between run 1 and run 2 metabolism within either group. When a Bonfer­roni correction (p = 0.OS/6) was applied to the data, one-way MANOVA comparisons across groups showed a trend toward significant differences be­tween morning (F(6,IS) = 3.16, p = 0.03) rates but not between afternoon rates and between run 2 (F(6, IS) = 2.76, p = O.OS) rates but not between run 1 rates. For both groups, average regional whole brain metabolic rates were lower in the morning than in the afternoon (Tables 1 and 2). This was true for 9 of 12 of the a.m.-p.m. and S of 10 of the p.m.-a.m. subjects.

Tables 3 and 4 show the regional percentage of changes from run 1 to run 2 for the individual sub­jects in each group. These were computed as differ­ence ratios:

(ROI-2 - ROI-l)/ROI-l

TABLE 2. Run 1 and scaled and unsealed run 2 regional glucose metabolic rates for individual subjects in the p.m.-a.m. group

Frontal lobe Parietal lobe Occipital lobe

Subjects 2 3 2 3 2 3

1 42.1 45.2 42.3 43.2 48.4 45.2 50.3 54.5 50.9 2 45.0 48.5 45.0 46.6 49.2 45.7 47.5 51.4 47.8 3 47.6 41.0 46.4 47.6 41.9 47.4 54.2 48.3 54.6 4 49.9 48.9 48.7 54.0 52.2 52.0 54.7 54.6 54.3 S 53.4 51.S 53.7 60.2 53.S 55.8 60.5 57.2 59.7 6 58.1 55.3 56.0 55.3 55.1 55.8 60.8 59.6 60.4 7 52.4 45.3 53.4 51.3 44.4 52.4 59.4 51.4 60.6 8 43.3 46.7 43.1 45.0 48.6 44.9 51.6 56.5 52.2 9 41.1 42.7 41.4 40.7 41.5 40.3 46.2 48.1 46.7

10 41.7 40.5 40.8 40.5 40.6 40.9 46.0 46.5 46.9 Group

Mean 47.5 46.6 47.1 48.5 47.5 48.0 53.1 52.8 53.4 SD 5.8 4.7 5.6 6.6 5.2 5.7 5.7 4.9 5.5

Temporal lobe Centrum Thalamus Whole brain

Subjects 2 3 2 3 2 3 2

1 43.5 44.9 41.9 26.3 30.5 28.5 40.2 43.2 40.3 36.7 39.3 2 42.7 47.7 44.3 31.3 28.2 26.2 42.1 40.1 37.2 37.0 39.9 3 46.8 41.5 47.0 31.5 32.3 36.6 43.6 40.1 45.7 40.3 35.6 4 48.7 49.4 49.2 42.6 38.2 38.1 46.1 48.7 48.6 42.7 42.9 S SI.9 49.2 51.3 48.4 44.6 46.6 49.5 47.3 49.3 46.4 44.5 6 53.2 51.9 52.6 44.6 39.7 40.2 55.4 54.8 55.5 48.9 48.3 7 52.7 45.8 54.0 41.8 34.8 41.0 51.9 47.1 55.5 46.5 39.4 8 45.1 48.0 44.3 32.7 33.1 30.5 46.9 48.7 45.0 39.3 42.5 9 40.8 40.7 39.5 28.8 31.4 30.4 44.2 45.1 43.7 35.8 36.9

10 41.2 41.5 41.8 33.0 35.7 35.9 40.3 42.1 42.4 35.9 35.7 Group

Mean 46.6 46.1 46.6 36.1 34.9 35.4 46.0 45.8 46.3 41.0 40.5 SD 4.8 3.8 5.0 7.5 4.9 6.4 5.0 4.5 1.9 4.9 4.1

See footnote to Table I.

J Cereb Blood Flow Metab. Vol. 8, No. 4, 1988

506 E. 1. BARTLETT ET AL.

TABLE 3. Percentages of change of unsealed and scaled regional metabolic rates from run I to run 2 for individual subjects in the a.m.-p.m. group

Frontal lobe Parietal lobe Occipital lobe Temporal lobe

Subjects 2 2 2 2

I -6 -1 -6 0 -8 -3 -5 I 2 8 -2 7 -1 7 -1 6 -2 3 8 -4 17 4 10 -I 10 -I 4 5 -2 5 -I 5 -I 4 -2 5 -8 0 -10 -3 -9 -2 -7 I 6 7 0 6 -I 6 -I 4 -3 7 10 0 9 -I II -1 12 2 8 14 0 12 -2 12 -3 14 0 9 24 1 22 -1 22 0 24 1

10 17 0 21 3 22 4 18 0 11 -3 -1 -3 -1 -2 0 -3 -1 12 3 2 6 5 4 2 3 2

Group Mean 7 -0.3 7 0.1 7 0.4 7 -0.3 SD 9 I \0 2 \0 2 9 2

Centrum Thalamus W hole brain

Subjects 2 2

1 0 4 -5 1 -6 2 6 -2 12 3 8 3 -13 -22 8 -3 12 4 5 -I -5 -11 -6 5 -4 4 2 11 -7 6 10 2 I -6 7 7 5 -4 17 6 10 8 21 5 12 -2 15 9 19 -3 23 0 22

10 16 -2 20 2 18 11 -3 -I 1 3 -2 12 9 7 0 -1 1

Group Mean 6 -I 7 0.2 7 SD 10 8 10 6 9

Key: 1, unsealed percentage of change, calculated as (ROI-2 - ROI-I)/(ROI-l x 100); 2, scaled percentage of change, calculated as (ROI-2' - ROI-I)/(ROI-l x 100), where ROI-2' = ROI-2 x

(whole-brain-llwhole-brain-2).

As can be seen, the average percentages of change in whole brain metabolism was 7% in the a.m.-p.m. group and 0.6% in the p.m.-a.m. group. Although changes as high as 22% were observed, it is impor­tant to note that 16 of the 22 subjects in the two groups had changes in whole brain metabolic rates of 10% or less.

Correlations between the regional metabolic rates of runs 1 and 2 for both groups combined are shown in Table 5. Figure lA shows the scatter plot of run 1 and run 2 metabolic values for a typical cortical region. The run I-run 2 correlation coeffi­cients range from 0.61 to 0.86 and are all significant beyond the 0.001 level.

The stability of metabolic relationships To assess the stability of the pattern of regional

metabolism from run 1 to run 2, scaled run 2 re­gional metabolic values were compared with un-

J Cereb Blood Flow Metab, Vol, 8, No.4, 1988

scaled run 1 values. A MANOVA assessing the ef­fect of scan order and time of day revealed no sig­nificant main effects or interactions. The mean regional differences between run 1 and scaled run 2 values in every case amounted to less than 1 J-Lmol glucose/100 g tissue/min (see Tables 6 and 7 and Fig. 1B). As can be seen in Tables 3 and 4, changes in scaled values of cortical regions of interest were uniformly low for all subjects, regardless of the change in their unscaled values; e.g., a.m.-p.m. subjects 9 and 10 show scaled percentage of change values that are no greater than those of subjects 11 and 12. In no case was the average regional change more than 1%. Thus, when run 2 values are scaled by a global metabolic factor that equates the whole brain metabolism of run 1 and run 2, then the re­sulting cortical metabolic values are virtually iden­tical, as can be seen in Fig. lB. That is, the propor­

tion of whole brain metabolic activity observed in

REPRODUCIBILITY OF CEREBRAL GLUCOSE MEASUREMENTS 507

TABLE 4. Percentages of change of unsealed and scaled regional metabolic rates from run 1 to run 2 for individual subjects in the p.m.-a.m. group

Frontal lobe Parietal lobe

Subjects 2

1 7 0 12 2 8 0 5 3 -14 -2 -12 4 -2 -2 -3 5 -3 I -11 6 -5 -4 0 7 -13 2 -13 8 8 0 8 9 4 1 2

10 -3 -2 0 Group

Mean -1 -1 -1 SD 8 2 9

Centrum

Subjects 2

1 16 8 2 -10 -16 3 3 16 4 -10 -11 5 -8 -4 6 -11 -10 7 -17 -2 8 1 -6 9 9 6

10 8 9 Group

Mean -2 -1 SD 11 to

each cortical region under the two scan conditions is virtually identical, since the patterns of regional

activity differ only in amplitude. Changes in scaled values of thalamus and centrum semiovale regions were generally higher than those of cortical re­gions. Inasmuch as thalamus and centrum regions were fairly small, it is likely that these larger changes are due to partial volume effects rather

than metabolic properties of the tissue.

The effect of anxiety on metabolic activity Prior to examining anxiety in relation to PET

data, we evaluated differences in anxiety measures between the a.m.-p.m. and p.m.-a.m. groups. For each measure, we assessed group differences by a repeated measures MANOVA, with group as the in­dependent variable, run (1 versus 2) as the repeated measure, and time of measurement (20 min prior to injection, injection, 20 min after injection) as the multivariate. These analyses revealed no significant main effects or interactions, which means that the groups could not be differentiated by these anxiety measures.

To assess the effect of anxiety on metabolism, we rank-ordered all 22 subjects according to their un-

2

4 -2

0 �4 -7

1 2

-1 -1

1

-0.6 3

Occipital lobe Temporal lobe

2 2

8 1 3 -4 8 0 12 4

-11 1 -11 0 0 -1 1 1

-5 -1 -5 -1 -2 -1 -2 -1

-13 2 -13 2 4 I 0 -3 4 1 0 -3 1 2 1 1

-0.1 0.6 -0.8 -0.1 8 1 8 2

Thalamus Whole brain

2

7 0 7 -5 -11 8 -7 5 -12

6 5 0 -4 0 -4 -1 0 -1 -9 7 -15

4 -4 8 2 -1 3 4 5 -1

-0.3 0.6 0.6 6 6 8

scaled metabolic rates in each region at each run and according to each stress measure at each run. We then calculated Spearman rank-order correla­tion coefficients. Correlation coefficients were sim­ilarly calculated between measures of anxiety and the percentage of change in regional metabolism. We observed statistically significant negative corre­lations between pulse rate at the time of the run 1 injection and run 1 metabolism in several cortical areas (see Tables 14 and 15). Pulse rates at run 1

TABLE S. Spearman correlation coefficients between unsealed CMRGs of run 1 and 2 for a.m.-p.m. and

p.m.-a.m. groups combined (N = 22)

Region

W hole brain Frontal lobe Parietal lobe Occipital lobe Temporal lobe Centrum semiovale Thalamus

Key: p 0.01 = 0.51; one-tailed.

Correlation coefficients

0.69 0.68 0.67 0.62 0.61 0.86 0.65

J Cereb Blood Flow Metab, Vol. 8, No.4, 1988

508 E. 1. BARTLETT ET AL.

65 1 55 ;:,.

;:,. ;:,.;:,.

;:,. N

C

;:,. d'& �

::J 45 ;:,. "-

;:,.

;:,.;:,. ;:,.

;:,.

35

25������������-r 25 35 45 55 65

run

65

55

c

2 45

35

25������������� 25 35 45 55 65

run

FIG. 1. Scatter plot of run 1 and run 2 scaled and unscaled frontal lobe metabolic rates for all subjects (see Tables 1 and 2). The ordinates are in micromoles of glucose per 100 g of tissue per minute. A: Run 1 rates plotted against unscaled run 2 rates. B: Run 1 rates plotted against scaled run 2 rates. Similar plots were obtained for other cortical areas.

were positively correlated with regional per­centages of change (Tables 8 and 9). No significant correlations between metabolism and the other measures of anxiety were observed.

Effect of anatomic repositioning errors Repositioning accuracy along the axial dimen­

sion, as represented by the average difference be­tween the size (in pixels) of the top-most images obtained at runs 1 and 2, was 6 ± 8% in the a.m.­p.m. group and 6 ± 6% in the p.m.-a.m. group. Transaxial repositioning ratios (which compared differences in the size of the temporal pole regions of runs 1 and 2) had a mean of 12 ± 16% in the a.m.-p.m. group and 16 ± 16% in the p.m.-a.m. group. To assess the effect of repositioning accu­racy on the percentage of change in metabolic values from run 1 to run 2, we rank-ordered the 22 subjects according to their axial and trans axial re-

J Cereb Blood Flow Metab, Vol. 8, No.4, 1988

positioning ratios and their whole brain percentage of change values. We then calculated Spearman rank-order correlation coefficients. None reached significance. Thus, repositioning inaccuracy was not significantly correlated with run 1 to run 2 changes in metabolic rates. The mean repositioning ratios of the a.m.-p.m. and p.m.-a.m. groups were virtually identical, which suggests that reposi­tioning was as accurate after a 22-h interval as after 2 h. This result suggests that with the techniques used, it will be possible to achieve accurate reposi­tioning even after very long intervals.

Effect of the size and laterality of the region

of interest The regions chosen for these analyses are large

and include data from several slices. We considered the possibility that the use of smaller regions might lead to a different pattern of results. Accordingly, we computed a separate MANOVA on unscaled metabolic values for regions obtained from a single slice. We chose for analysis a slice on which data from three lobes (frontal, occipital, and temporal) could be obtained. The slice was located approxi­mately 6.1 cm above the canthomeatal line. This analysis revealed the same pattern of results as was observed with un scaled whole-lobe values. We also considered the possibility that differences between left and right hemisphere regions of interest would affect our results. For each region, we computed a repeated-measures MAN OVA on un scaled values with group as the independent variable, run as the repeated measure, and hemisphere as the multivar­iate. These analyses showed no significant main ef­fects for hemisphere and no significant interactions involving hemisphere, which indicates that metabo­lism in left and right hemisphere regions of interest did not differ significantly.

DISCUSSION

This study measured test-retest variability in young normal male subjects scanned twice in a 24-h period under baseline (resting) conditions. We ob­served average changes in whole brain metabolism of 7% and 0.6% in the a.m.-p.m. and p.m.-a.m. groups, respectively. Seventy-three percent of our total subject group showed average whole brain percentages of change of 10% or less. This is com­parable with change values reported by Reivich et al. (1982) and Brooks et al. (1987) with much smaller samples and is considerably lower than the 24% change reported by Duara et al. (1987) for pairs of resting scans occurring over intervals of about 2 weeks.

When the regional values were rescaled to com-

REPRODUCIBILITY OF CEREBRAL GLUCOSE MEASUREMENTS 509

TABLE 6. Differences between run 1 and run 2 unsealed and scaled regional metabolic rates for individual subjects group

Frontal lobe Parietal lobe Occipital lobe

Subjects 2 2 2

I -2.7 -0.3 -2.6 -0.1 -4.1 -1.3 2 3.3 -0.1 3.1 -0.4 3.1 -0.6 3 3.3 -1.5 7.3 2.0 4.2 -0.6 4 2.4 -0.4 2.5 -0.3 1.8 -0.7 5 -3.3 -0.2 -4.4 -1.2 -4.4 -0.8 6 3.2 -0.1 2.9 -0. 5 3.2 0.4 7 4.7 0.2 4.4 -0.4 5. 6 0.6 8 8.4 -0.1 7.2 -1.4 7.4 -1.6 9 10.0 0.5 9.5 -0.3 9.9 0.0

10 6.6 -0.2 9.0 1.3 9. 5 1.6 11 -1.4 -0.4 -1.8 -0.7 -1. 3 -0.3 12 1.5 1.0 3.0 2.4 1.9 1. 3

Group Mean 3.0 0.14 3.3 0.02 3.2 -0.24 SD 4.1 0.59 4.5 1.22 4.7 0.97

Temporal lobe Centrum T halamus W hole brain

SUbjects 2 2 2

I -2.3 0.2 -0. 5 1.4 -2.0 0.2 -2.1 2 2.2 -1.0 1.8 -0.5 4.5 1.3 2.8 3 4.2 -0.5 -4.2 -7.1 3.3 -1.1 4.2 4 1. 8 -0. 9 1.9 -0.3 -2.3 -4.8 2.5 5 -2.9 0.2 -1.1 1.2 1.0 4.4 -2.6 6 2.0 -1.2 3. 3 0.8 0.4 -2.5 2.8 7 5.7 1.1 1.9 -1.8 6. 6 2.5 4.0 8 7.4 -0.3 9.2 2.3 6.5 -1.2 7.0 9 9.3 0.4 6.4 -0.9 8.0 0.04 7.8

10 6.8 0.1 5. 0 -0.5 7.0 0.7 5.9 11 -1.3 -0.4 -1.2 -0.4 0.3 1.2 -0.8 12 1.5 0.9 3.6 3.1 -0.1 -0.6 0.5

Group Mean 2.9 -0.10 2.2 -0.24 2. 8 0.01 2.6 SD 3.9 0.74 3.7 2.60 3.7 2. 37 3.4

Key: I, ROI-2 - ROI-I; 2, ROI-2' - ROI-I, where ROI-2' = ROI-2 x (whoie-brain-llwhole-brain-2). All values are expressed in micromoles of glucose per 100 g of tissue per minute.

pensate for the effects of changes in whole brain metabolism, the resulting regional mean changes were no more than 1%. Moreover, the scaled re­gional percentages of change in cortical areas were as small for subjects with large global metabolic changes as for those with global changes of less than 5%. These results show that relatively large changes in global metabolism are not necessarily indicative of changes in the proportion of metabolic activity observed in cortical regions; the results also demonstrate that intrasubject changes in global CMRG may occur independently of any change in the cortical RCMRG pattern.

Scaling factors (e.g., regions of interest/whole brain rations, z scores, etc.) are frequently used in PET studies of regional cerebral metabolism to compensate for intersubject variability in whole

brain metabolism and to highlight intersubject com­monalities in relative regional metabolism (Clark et al. , 1985). The present study shows that the test -re­test stability of the relative RCMRG pattern is ex­tremely high.

In order to design an experiment to detect differ­ences after an intervention using the test-retest par­adigm, it is necessary to determine the appropriate sample size. A classical power analysis may be per­formed to compute this sample size using the whole brain or regional variances given in Tables 1 and 2.

In contrast to the low level of intrasubject vari­ability, we found a wide range of regional metabolic rates and change scores across subjects. For ex­ample, the range of whole brain CMRGs for the 22 subjects was 15.7 f-Lmoi/IOO g/min at run 1 and 22.3 j.Lmol/l00 g/min at run 2 (see Tables 1 and 2). This

J Cereb Blood Flow Metab, Vol. 8, No.4, 1988

510 E. 1. BARTLETT ET AL.

TABLE 7. Differences between run 1 and run 2 unsealed and scaled regional metabolic rates for individual subjects in the p.m.-a.m. group

Frontal lobe Parietal lobe Occipital lobe

Subjects 2 2 2

1 3.1 0.2 5.1 1.9 4.2 0.6 2 3.5 0.03 2.6 -0.9 3.9 0.2 3 -6.6 -1.2 -5.7 -0.2 -6.0 0.4 4 -1.0 -1.2 -1.8 -2.1 -0.1 -0.3 5 -1.9 0.3 -6.7 -4.4 -3.3 -0.8 6 -2.8 -2.1 -0.2 0.5 -1.1 -0.3 7 -7.1 1.0 -6.9 1.0 -8.0 1.1 8 3.4 -0.2 3.5 -0.2 4.9 0.5 9 1.6 0.3 0.8 -0.4 1.9 Q.§

10 -1.2 -0.9 0.1 0.4 0.5 0.9 Group

Mean -0.9 -0.37 -0.9 -0.42 -0.3 0.28 SD 3.9 0.94 4.3 1.76 4.4 0.60

Temporal lobe Centrum T halamus W hole brain

Subjects 2

1 1.4 -1.5 4.1 2 5.0 1.6 -3.0 3 -5.3 0.2 0.8 4 0.7 0.5 -4.4 5 -2.7 -0.5 -3.7 6 -1.3 -0.6 -4.9 7 -6.8 1.3 -7.0 8 2.9 -0.8 0.4 9 -0.1 -1.3 2.5

10 0.3 0.6 2.6 Group

Mean -0.6 -0.05 -1.2 SD 3.6 1.07 3.8

See footnote to Table 6.

amounted to 40% and 55% of the combined a.m.­p.m. and p.m.-a.m. group means at run 1 and run 2, respectively.

Reports from other laboratories had led us to ex­pect that some of the between-subject variability could be accounted for by individual differences in

anxiety. For example, Gur et al. (1987a) reported a significant negative correlation between a self-re­

port measure of anxiety and cortical metabolism.

Such a pattern was also evident in our data. Evi­

dence that moderate anxiety was associated with a

decrease in metabolism during the first but not the

second scanning experience was provided by the negative correlation between pulse rate at the run 1

injection and run 1 cortical metabolism. Relatively higher anxiety, as evidenced by a higher pulse rate, was associated with relatively lower metabolism at run 1. The lack of correlation between physiolog­

ical measures of anxiety and metabolism at run 2 suggests that the more anxious subjects may have

habituated to the stress of the scanning experience

by the time of the second run. This conjecture is supported by the fact that pulse at run 1 was posi-

J Cereb Blood Flow Metab. Vol. 8. No.4. /988

2 2

2.1 3.0 0.2 2.6 -5.0 -2.0 -4.8 2.8

5.1 -3.2 2.1 -4.7 -4.5 2.6 2.5 0.2 -1.8 -2.2 -0.2 -1.9 -4.4 -0.6 0.1 -0.6 -0.8 -4.8 3.6 -7.0 -2.1 1.8 -1.9 3.3

1.6 0.9 -0.4 J.l 2.9 1.8 2.1 -0.2

-0.69 -0.3 .31 -0.5 3.50 2.7 2.45 1.0

tively correlated with the percentage of change in

metabolic activity from run 1 to run 2. Thus, rela­

tively higher anxiety at run 1 was also associated

with relatively larger increases in metabolism from run 1 to run 2. To our knowledge, this is the first demonstration of a significant correlation between an autonomic measure of arousal and regional brain metabolism.

We also investigated the effect of scan order and

time of day on intersubject variability. A MAN OVA revealed no main effects for scan order or time of

day but did show a significant interaction. While

both groups showed a lower mean metabolism in

the morning than in the afternoon, the difference was very small in the p.m.-a.m. group. This pattern of results may reflect the effects of both anxiety and habituation as well as such systemic biological influences such as diurnal variation. For example,

if there is a tendency for cerebral metabolic activity

to be somewhat lower in the morning than in the

afternoon, then the observed pattern of increased

metabolic rate in the afternoon scans of a.m.-p.m.

subjects may reflect the influences of both dimin-

REPRODUCIBILITY OF CEREBRAL GLUCOSE MEASUREMENTS 511

TABLE 8. Spearman rank-order correlation coefficients calculated between pulse (as measured at the time of isotope injection) and measures of unsealed regional

glucose metabolism in 22 subjects

Region

Run 1 W hole brain Frontal lobe Parietal lobe Occipital lobe Temporal lobe Centrum semiovale Thalamus

Run 2 W hole brain Frontal lobe Parietal lobe Occipital lobe Temporal lobe Centrum semiovale T halamus

a p < 0.05. b P < 0.01.

Pulse at Pulse at run 1 injection run 2 injection

-0.47a -0.29 -0.43a -0.32 -0.46a -0.30 -0.49a -0.33 -0.53b -0.34 -0.25 -0.11 -0.35 -0.10

-0.Q7 0.05 0.01 0.14 0.03 0.11

-0.14 0.08 -0.08 0.10 -0.17 0.01 -0.Q7 0.05

ished anxiety and diurnal increases in metabolism. On the other hand, the effects of anxiety and habit­uation may have been attenuated by the effect of

diurnal increases in metabolism in p.m.-a.m. sub­jects to produce the higher metabolism observed at

their first (or afternoon) scan. The possible effect of

anxiety and diurnal variation on intersubject vari­

ability merits further examination.

In conclusion, this is the first study to measure cerebral glucose metabolism in a sample large

enough to provide reliable estimates of test-retest

variability. The study was designed to minimize several potential sources of measurement error, in­cluding anatomic repositioning inaccuracies and

differences in various subject characteristics. The study also estimated the effect on metabolism of

TABLE 9. Spearman rank-order correlation coefficients calculated between pulse (as measured at the time of

isotope injection) and percentages of change from run 1 to run 2 of unsealed metabolic rates in 22 subjects

Region

Whole brain Frontal lobe Parietal lobe Occipital lobe Temporal lobe Centrum semiovale Thalamus

a p < 0.05. b P < 0.01.

Pulse at run 1 injection

0.54b 0.45a 0.54b 0.54b 0.50a 0.34 0.57b

Pulse at run 2 injection

0.28 0.29 0.21 0.30 0.28 0.125 0.24

subjects' level of anxiety as well as the effects of scan order and time of day. A trend toward signifi­

cant between group differences indicates that some

of the observed variability in global CMRGs may

be due to circadian metabolic changes or other sys­temic biological influences. When a scaling factor is used to compensate for the effect of changes in whole brain glucose metabolism, the resulting mean RCMRG changes were no more than 1 %. This dem­onstrates that the relationship among RCMRGs at rest is reliable and stable within a group of indi­

viduals over time and that intrasubject changes in

global CMRGs may occur independently of

changes in the RCMRG pattern. This suggests that

pattern analyses of scaled metabolic data may be

particularly appropriate for perturbation (activa­tion) experiments that use a repeated-measures de­

sign.

Acknowledgment: T his research was supported by Grant NS-J5638 from the NINCDS and by the U.S. Department of Energy, Office of Health and Environ­mental Research. In addition, we deeply appreciate the contributions of F. Barouche, C. Barrett, R. Carciello, C. Codd, J. S. Fowler, R. Gard, E. Jellett, T. Johnson, K. Karlstrom, P. King, N. Netusil, C. Redvanly, 1. A. G. Russell, H. Sachs, D. J. Schlyer, M. Smith, P. Van Gelder, and D. Warner.

REFERENCES

Brooks, RA, Di Chiro G, Zukerberg BW, Bairamian D, Larson SM (1987) Test-retest studies of cerebral glucose metabolism using fluorine-18 deoxyaglucose: validation of method. J Nucl Med 28:53-59

Chang JY, Barker W W, Apicella A, Emran A, Duara R (1987) Test-retest paradigms in PET : errors arising from subject re­positioning. J Nuc Med 28:681-682

Clark C, Carson R, Kessler R, Margolin R, Buchsbaum M, De­Lisi L, King C, Cohen R (1985) Alternative statistical models for the examination of clinical positron emission to­mographylfluorodeoxyglucose data. J Cereb Blood Flow Metab 5:142-150

Cohen J, Cohen P (1975) Applied Multiple Regression/Correla­tion Analysis for the Behavioral Sciences. New York: L. E. Erlbaum

Davidson M, Toporek J (1983) P4V: general univariate and mul­tivariate analysis of variance and covariance, including re­peated measures. In BMDP Statistical Software (WJ Dixon et aI., eds), Berkeley, University of California Press, pp 93-122

Duara R, Grady C, Haxby J, Ingvar D, Sokoloff L, Margolin RA, Manning RG, Cutler NR, Rapoport SI (1984) Human brain glucose utilization and cognitive function in relation to age. Ann Neural 16:702-713

Duara R, Gross-Glenn K, Barker W W, Chang JY, Apicella A, Loewenstein D, Boothe T (1987) Behavioral activation and the variability of cerebral glucose metabolic measurements. J Cereb Blood Flow Metab 7:266-271

Gur RC, Gur RE, Obrist W D, Hungerbuhler JP, Younkin D, Rosen AD, Skolnick BE, Reivich M (1982) Sex and handed­ness differences in cerebral blood flow during rest and cog­nitive activity. Science 217:659-660

Gur RC, Gur RE, Resnick SM, Skolnick BE, Alavi A, Reivich M (1987a) T he effect of anxiety on cortical cerebral blood flow and metabolism. J Cereb Blood Flow Metab 7:173-177

J Cereb Blood Flow Metab, Vol. 8, No.4, 1988

512 E. J. BARTLETT ET AL.

Gur RE, Resnick SM, Gur RC, Alavi A, Caroff S, Kushner M, Reivich M (1987b) Regional brain function in schizophrenia: II. Repeated evaluation with positron emission tomography. Arch Gen Psychiatr 44: 126-129

Hatazawa J, Brooks RA, Di Chiro G, Bacharach SL (1987) Glu­cose metabolic rate versus brain size in humans. J Cereb Blood Flow Metab 7:S301

Huang S-C, Phelps ME, Hoffman EJ, Sideris K, Selin CJ, Kuhl DE (1980) Noninvasive determination of local cerebral met­abolic rate of glucose in man. Am J Physiol 238:E69-E82

Kearfott KJ, Rottenberg DA, Knowles RJR (1984) A new head­holder for PET, CT, and NMR imaging. J Comp Assist To­mogr 8:1217-1220

Kuhl DE, Metter EJ, Riege WH, Phelps ME (1982) Effects of human aging on patterns of local cerebral glucose utilization determined by the (18F) fluorodeoxyglucose method. J Cereb Blood Flow Metab 2:163-171

MacGregor RR, Fowler JS, Wolf A, Shive CY, Lade RE, Wan CN (1981) A synthesis of 2-deoxy-o-(J-1IC)glucose for re­gional metabolism studies. J Nucl Med 22:800-803

Matsui T, Hirano H (1978) An Atlas of the Human Brain for Computerized Tomography. Tokyo: Igaku-Shoin

Mazziotta JC, Phelps ME, Plummer D, Kuhl DE (1981) Quanti­tation in positron emission computer tomography: 5. Phys­ical-anatomical effects. J Comput Assist Tomogr 5:734-743

Oldfield RD (1971) The assessment and analysis of handedness: T he Edinburgh Inventory. Neuropsychiatry 9:97 -113

Phelps ME, Huang S-C, Hoffman EJ, Selin C, Sokoloff L, Kuhl

J Cereb Blood Flow Metab. Vol. 8. No.4, 1988

DE (1979) Tomographic measuremnt of local cerebral glu­cose metabolic rate in humans with (F-18)2-fluoro-2-o-de­oxyglucose: validation of method. Ann NeuroI6:371-388

Phelps ME, Mazziotta JC, Schelbert HR, eds. (1986) Positron Emission Tomography and Autoradiography: Principles and Applications for the Brain and Heart. New York, Raven Press

Reivich M, Alavi A, eds. (1985) Positron Emission Tomography. New York, Raven Press

Reivich M, Alavi A, Wolf A, Greenberg JH, Fowler J, Christman D, MacGregor R, Jones SC, London J, Shiue C, Yonekura Y (1982) Use of 2-deoxy-o(J-1IC)glucose for the determination of local cerebral glucose metabolism in humans: variation within and between subjects. J Cereb Blood Flow Metab 2:307-319

Reivich M, Alavi A, Wolf A, Fowler J, Russell J, Arnett C, MacGregor RR, Shiue CY, Atkins H, Anand A, Dann R, Greenberg JH (1985) Glucose metabolic rate kinetic model parameter determination in humans: the lumped constants and rate constants for (l8-F)fluorodeoxyglucose and (IIC)deoxyglucose. J Cereb Blood Flow Metab 5:179-192

Sokoloff L, Reivich M, Kennedy C, Des Rosiers CS, Patlak CS, Pettigrew KD, Sakurada 0, Shinohara M (1977) T he (I4C)de­oxyglucose method for the measurement of local cerebral glucose utilization: theory, procedure and normal values in the conscious and anesthetized albino rat. J Neurochem 28:897-916