Cerebral Blood Flow in Hypertensive Patients
An Initial Report of Reduced and Compensatory Blood Flow Responses During Performance of Two Cognitive Tasks
Abstract—We asked whether the altered cerebral vasculature associated with essential hypertension might dampen or redirect the regional cerebral blood flow (rCBF) response to cognitive work. Relative rCBF was assessed with [15O]water positron emission tomography during a working memory task, a memory span task, and two perceptual control tasks. Unmedicated hypertensive patients and control subjects differed in rCBF response during both memory tasks. Hypertensives showed relatively diminished rCBF responses in right hemisphere areas combined with compensatory activation of homologous areas in the left cerebral cortex. Essential hypertension appears to selectively influence the circulatory reserve of portions of cerebral cortex and secondarily induce recruitment of other cortical areas to process certain tasks.
The cerebral blood vessels of chronically hypertensive patients have an increased vascular resistance that offsets high systemic pressures and maintains global CBF at near normal levels.1 2 Histologically, cerebral parenchymal arterioles show a thickened media and a narrowed lumen. Such changes may reduce the vasodilatory capability of cerebral vessels and thereby limit maximum delivery of oxygen and nutrients to active brain tissue. Cognitive work raises metabolic demand in activated regions of the brain, necessitating concomitant local changes in blood supply. Because of their compromised vasodilatory capacity, we hypothesized that patients with untreated high blood pressure would manifest less rCBF response than control subjects in brain areas activated by cognitive tasks and that these subjects might adopt different, presumably compensatory, topographical patterns of rCBF response.
Nine unmedicated hypertensive individuals were compared with five control subjects who were similar in mean age (hypertensive, 60 to 67 years; control, 59 to 68 years), education (14 versus 16 years), and gender (88% male versus 62% male) but differed in blood pressure (151/91 mm Hg versus 120/76 mm Hg).2 Hypertension was defined as two consecutive clinical screenings with systolic pressures between 140 and 180 mm Hg and diastolic pressures between 90 and 110 mm Hg. Normotensive status was similarly determined as systolic pressures <135 mm Hg and diastolic pressures <85 mm Hg. Hypertensive patients were unmedicated (for a minimum of 8 weeks) and had less than a 2-year lifetime history of taking any hypertensive medication. Exclusion criteria included secondary hypertension; use of any cardiovascular or psychiatric medication; cerebrovascular disease (by report, magnetic resonance imaging, and carotid ultrasound); and history of myocardial infarction, diabetes, cancer, psychiatric disease/alcoholism, or renal and pulmonary disease. One patient reported significant use of his left hand, while other patients reportedly being consistently right-handed. The results reported were essentially unchanged, however, when analyses were repeated without this subject. Informed consent was obtained from all subjects after the nature and possible consequences of participation were explained (following procedures approved by the Institutional Review Board of the University of Pittsburgh).
Two tasks were selected that were known to elicit consistent rCBF responses. Each task was presented at two levels of difficulty. (1) A continuous performance task required subjects to detect the letter “x” in a stream of letters (CPTx); at a higher level of difficulty, the subject responded to any letter that was a repeat of that letter at a position one removed (CPTskip). The CPT was presented on a video screen controlled by a Macintosh IIci computer. Letters were presented for 500 milliseconds each, separated by a 2500-millisecond interval. In the CPTx task, the subjects were instructed to press the button whenever they saw an “x.” In the CPTskip task, the subjects were instructed to press the button whenever a letter was repeated with exactly one intervening nonidentical letter (eg, B Z B, but not BB or B C Z B).3 (2) An auditory free recall task required the subject to remember and repeat back single words; at a higher level of difficulty, 12 words were presented and then repeated back. With use of a tape recorder, the single-word task presented high-frequency nouns at a rate of one per second. Recall (basically word repetition) occurred immediately after each single item. This task was compared with the 12-word task in which 12 words were presented at a rate of one per second followed by a 15-second pause during which the volunteers repeated back verbally in any order as many words as possible.4 Two perceptual tasks were used: visual fixation of a crosshair display and visual observation of a checkerboard pattern alternating colors at a 6-Hz frequency.
All PET scans were acquired in three-dimensional mode (septa retracted) on an ECAT 951R/31 scanner (Siemens/CTI PET Systems). The 951R scanner covers an axial field of 10 cm with a spatial resolution of 6.5 mm, full width at half maximum. A 15-minute transmission scan was collected to provide coefficients for attenuation correction of the emission data. Each activation task was begun 15 seconds before the injection of 7 mCi of [15O]water, and the emission scan was begun 30 seconds after injection, for a scan duration of 60 seconds. No background frame was acquired. All scans were reconstructed in three dimensions without scatter correction and with a pixel size of 1.7 mm, using a Hanning smoothing window and a 0.8 Nyquist cutoff frequency.
A total of 12 emission scans were acquired with an 8-minute interval between injections to allow for the decay of the [15O]water. For each subject, 12 scans were performed with a fixed task order: word repetition, 12-word free recall, checkerboard display, CPTx, CPTskip, visual fixation, CPTskip, CPTx, checkerboard, 12-word free recall, word repetition, and visual fixation. Correction for head movement during the study was made by realigning the images using the Automated Image Registration (AIR) package.5 Images were then converted into the stereotactic coordinate system of a standard brain atlas.6 No measurements of absolute global flow were made because the PET procedure was nonquantitative, and changes in global flow between scans were removed with an ANCOVA procedure. All images were normalized to an average flow of 50 mL · 100 mL−1 · min−1 for each subject. The images were then smoothed with a gaussian filter of 12 mm×12 mm×20 mm.
PET images were analyzed using standard Statistical Parametric Mapping (SPM) software.7 The primary analysis used a contrast to explicitly test three hypotheses. The first and second hypotheses were that rCBF would increase with the increases in the cognitive difficulty of (1) the CPTs and (2) the free recall tasks. Separate analyses were performed for each task and for the hypertensive and control groups. Cognitive difficulty was viewed as minimal for the resting/fixation task, modest for the CPT and one-word recall tasks, and significant for the CPTskip and 12-word recall. Arbitrary weights (chosen a priori) for relative difficulty were assigned to the fixation task (−2), the CPTx and word repetition (−1), and CPTskip and 12-word recall (+3). Use of these weights in an SPM contrast identified voxels that across subjects consistently demonstrated increasing activation in parallel with the assigned weights. We also checked deactivation with increasing cognitive difficulty by reversing the direction of the weights. Contrast analyses were done using ANCOVA in which the covariate was global flow.
The third hypothesis tested was that hypertensive and normotensive groups would differ in the increase of rCBF with increasing cognitive difficulty (ie, an interaction between patient group and cognitive difficulty). The contrast analysis used the same weights for the task conditions, but all subjects were included with contrasting weights for hypertensive and control subjects. In one analysis, voxels were identified in which control subjects showed a greater increase than hypertensives in activation with increase in cognitive difficulty. In a second analysis, weights were reversed so that we identified voxels in which hypertensive subjects showed a greater increase than control subjects in activation with increase in cognitive difficulty. In tests of the third hypothesis, we did not require that voxels show increases across cognitive difficulty in the prior within-group analyses,8 ie, we could identify voxels that changed significantly with increased cognitive difficulty in one group but failed to show any significant change in the other. CBF activation and deactivation in the checkerboard task was analyzed separately within each group, followed by contrasts between hypertensive and control subjects. We interpreted rCBF response comparisons that yielded an SPM Z value associated with an α value of P<0.001, but we emphasize in this report only Z values that remained significant at a value of P<0.05 after correction for multiple comparisons.7
Performance was comparable between groups and generally good: on the CPTs, hypertensive patients averaged 87% correct and control subjects 74%; on the free recall tasks, hypertensives averaged 70% correct, controls 66%. Despite the absence of group differences, the range of individual performance scores required us to ensure that rCBF activation differences between groups were not due solely to differences in performance between individuals. To check this possibility, the percentages of correct scores, as well as reaction time and false-alarm rates from the CPTs, were correlated to rCBF values at areas showing significant differences between groups (see Table 2⇓). No significant correlations were observed. Furthermore, covarying performance did not influence the statistical significance of results in the analyses presented below.
Changes in rCBF during performance were first examined separately for patients and control subjects. Hypertensive patients responded with rCBF changes primarily in left hemispheric areas in the analyses comparing levels of difficulty for both CPT and free recall. In contrast, control subjects showed primarily right hemispheric rCBF responses during performance of both tasks. Table 1⇓ shows that controls activated right frontal, prefrontal, parietal, and temporal areas in both tasks.
Hypertensive patients responded with left parietal and frontal rCBF during the CPT and showed trends toward left frontal, temporal, prefrontal, and parietal rCBF responses in the free recall task.
Table 2⇓ shows the statistical results from direct comparison of the two groups on how much their rCBF response increased with increasing task difficulty for the two cognitive tasks. For the CPT, both right frontal and parietal rCBF responses were significantly greater in controls than in hypertensives. Figure 1⇓ shows these results as Z score maps of brain slices and as a graph of mean rCBF responses by group and task level. The brain slices illustrate the extent and statistical strength of the rCBF changes that were greater in controls than in hypertensives. The figures shows horizontal sections at 8-mm intervals beginning at a z value of 12 mm in the Talairach and Tournoux6 nomenclature, 12 mm above the anterior commissure–posterior commissure line. The slices are shown until z=36 mm and illustrate the relatively large extent of the differences in parietal and frontal cortex between groups in the change in rCBF with CPT difficulty. The right panel of Figure 1⇓ shows the mean across subjects of the rCBF counts for the different levels of CPT difficulty. The counts are taken from the centroid of the area of activation as reported by the SPM program and thus correspond directly to the results shown in Table 2⇓. The upper graph corresponds to the parietal activation (the x and y values can be referred to the slices with x=0 defined by the midline and y=0 defined by the vertical traversing the posterior margin of the anterior commissure). The lower graph corresponds to the frontal activation. Both figures show a robust increase in rCBF change for control subjects but little change for hypertensives. In addition, both figures indicate that relative rCBF in the hypertensives is higher than that of the controls at rest but not during task performance. Differences in rCBF for the free recall tasks were not significant after correction for number of comparisons; however, controls showed a similar trend for enhanced right hemispheric prefrontal and temporal rCBF response relative to little change in rCBF response for hypertensives.
Hypertensive patients might be expected to show greater left hemisphere responses than control subjects given the results from the separate analyses within each group. However, this was not seen in the contrast specifically testing for areas in which hypertensives showed greater rCBF change than controls. Table 2⇑ shows that only a left hippocampal (x=−18, y=−32, z=−4) rCBF response showed a statistically significant difference favoring hypertensives relative to controls. Figure 2⇓, constructed similarly to Figure 1⇑, shows slices at z=−8 to z=0 that largely encompass the area yielding the group difference. The accompanying graph for the hippocampal rCBF differences shows that the result is primarily due to a drop in hippocampal rCBF response in controls during the processing of the CPTskip; hypertensive patients maintained a similar rCBF response across CPT difficulty. Table 2⇑ does show a trend for greater left prefrontal rCBF response in hypertensives relative to control subjects; means for this comparison (not shown) demonstrated a greater change in rCBF for hypertensives relative to controls during the CPTskip. Overall, the results show a pattern of memory performance associated with robust right hemispheric prefrontal, parietal, and temporal rCBF response in controls. This response is dampened in hypertensives. Hypertensive patients show little differential response to tasks of differing difficulty but show significant left hemispheric rCBF responses not present in controls. The left hemispheric activations are modest in that we could not show the differences in rCBF responses across task difficulty to be statistically larger in left hemispheric areas of hypertensives relative to controls.
No significant differences in rCBF response or lateralization were observed for a perceptual task (viewing an alternating checkerboard pattern). Both groups bilaterally increased occipital rCBF in response to the checkerboard pattern: controls, 4237 voxels with a centroid at 8, −98, and 4 [Z=6.50, P(Zmax>u)<0.001]; hypertensives, 3631 voxels with a centroid at −16, −100, and 0 [Z=7.21, P(Zmax>u)<0.001]. The absence of differences between hypertensive and control rCBF responses in the visual checkerboard task suggests that lateralization is task- or brain region–specific rather than a general property of hemisphere perfusion in hypertensives.
We have investigated the rCBF response to a verbal episodic memory task and a sustained attention task with an added working memory component. Both tasks were challenging to our participants, and we, as well as earlier investigators, found significant rCBF changes in response to the tasks. The most important difference between hypertensive and control participants was a decreased responsivity to increased task difficulty among hypertensives. An unexpected further difference was in the lateralization to different hemispheres of the primary task-induced changes between hypertensive and control subjects. The unanticipated nature of this finding and the relatively low number of participants in the present study indicate the need for replication of the present results.
Because we sought to show that hypertensive volunteers differed from control subjects in rCBF, it is important to demonstrate that our controls showed typical rCBF changes. This is particularly important because we studied a relatively small number of healthy normotensive controls. Although few have studied middle-aged individuals screened for hypertension, we can compare the present results to findings for two age-appropriate groups examined with the same tasks in our PET facility, as well as to a number of other studies using comparable tasks but in younger college-age volunteers. Carter et al (unpublished observations, 1997) used the CPT and CPTskip to study schizophrenics with the same PET techniques that we used. The control group for this study had a mean age of 39 years. CPT performance in the control subjects was associated with prefrontal and frontal activation (Brodmann’s areas 44 and 46) that was predominantly lateralized to the right hemisphere, as well as a bilateral parietal (Brodmann’s area 40) activation. Becker et al,4 with a similar PET technique, used the verbal episodic memory task to study patients with Alzheimer’s disease and had a normal control group with a mean age of 66 years. They compared eight-word and three-word free recall tasks and observed bilateral frontal activation differences in their control group, but the area showing differences for the eight- and three-word recall was larger in the right hemisphere. Parietal activation was not observed in their study. Thus, the two comparable studies from the Pittsburgh laboratory generally support the right prefrontal/frontal activation observed in the present study, but they did not find the parietal involvement that we observed.
Cabeza and Nyberg10 recently reviewed all PET studies of cognition with normal predominantly young adult subjects. They reported six comparisons in the literature between a control and a sustained attention condition, comparable to our CPTs. In agreement with our findings, four of these comparisons showed right frontal activation and three of them right parietal activation. There were no reports of left frontal activation and only one of left parietal activation. Cabeza and Nyberg also reported on two comparisons of verbal episodic memory with a control task. Of these 12 studies, eight reported right frontal activation, one reported left frontal activation, and the remainder of the activations were bilateral. Our finding of predominantly right parietal change in control subjects is not, however, consistent with the literature reviewed; there was only one report of right parietal activation and two reports of left parietal activation, with the clear majority of comparisons showing bilateral activation. Cabeza and Nyberg’s review10 also suggested that working memory (involved in our CPTskip) evoked primarily bilateral frontal and parietal activations. Other reviews11 12 draw similar conclusions to those of Cabeza and Nyberg.10 Overall, the evidence suggests that our control subjects showed rCBF activations that were consistent with the majority of prior investigations of sustained attention and verbal episodic memory; our evidence for right frontal/prefrontal activation is very consistent with previous reports, as is parietal activation, but right lateralized parietal activation has not been widely observed.3 4 12 13 14 15
The failure of hypertensive subjects to demonstrate increased rCBF response with increasing task difficulty is significant given the representativeness of the rCBF task responses in our control subjects. We cannot claim that we have demonstrated a failure of neural activation or even metabolic activation in response to these tasks. Fox et al16 suggested that brain activation differentially alters brain glucose uptake, oxygen metabolism, and blood flow. This suggestion remains controversial,17 18 but it prevents definitive interpretation of blood flow changes in terms of metabolic or neural activation. Furthermore, hypertension may alter vascular anatomy or metabolism such that the relationship is altered between blood flow and neural activation. For example, hypertension may change the permeability of the vascular wall to vasoregulatory compounds and thus alter the responsivity of the vasculature to metabolic needs of the neural tissue. Further work will be needed to determine whether rCBF changes in hypertension represent a change in neural activation of the areas examined.19 Our working hypothesis is, however, that hypertension alters the responsiveness of the cerebrovasculature to neural activation. Responsiveness may decline because of chronic arteriolar vasoconstriction and reduced distensibility of nutritive vessels resulting from the hemodynamic adjustment of the brain to systemic hypertension.1 2 20 21 22 23 24 In our results, control subjects showed greater relative rCBF than hypertensive patients only during task processing but not during a resting visual-fixation condition.
Hypertensive patients showed rCBF responses to the CPT and verbal recall tasks that were in many instances lateralized to left hemisphere areas. The predominantly left frontal and prefrontal changes in rCBF during the tasks differed from the right lateralized changes in both our controls and in the literature just reviewed. Activation of functionally homologous areas in the left hemisphere may compensate for the inability to further activate right hemisphere areas among hypertensives. Quantitative rCBF measurements will be required, however, to test our working hypothesis that hypertensives show relatively enhanced nutritive CBF during rest and to assess the reasonableness of our compensation argument. Others have suggested that compensatory activation of regions occurs with added task difficulty because of either task requirement or physiological impairment. Smith et al13 reported the recruitment of lateralized homologous structures with increasing difficulty of short-term memory tasks. Becker et al4 have made similar suggestions for Alzheimer’s disease patients. With the present data, we can suggest such compensation but not prove that the enhanced left hemisphere rCBF response of hypertensives is in fact compensatory. The demonstration of compensation would be difficult; ideally, measurements should be available for the same individuals before the establishment of hypertension, during a hypertensive phase, and after reversal of the hypertension. Such data would permit a separation of strictly hemodynamic effects on rCBF responses from effects of other consequences of essential hypertension.
Our results raise a number of general issues about the interpretation of PET results in patients with vascular disease but also suggest that PET results may be useful in showing cerebrovascular and cognitive sequelae of hypertension. More work will be required to verify and extend the present results. Such work should define the relative role of neural/vascular changes correlated with hypertension, further exploring such factors as the degree of extracranial and intracranial atherosclerosis and the role of small white matter lesions.22 23 Our findings provide an impetus for the further examination of these issues and may elucidate earlier studies that demonstrated neuropsychological deficits in hypertensive compared with normotensive subjects.25 26 27
Selected Abbreviations and Acronyms
|CPT||=||continuous performance task|
|PET||=||positron emission tomography|
|rCBF||=||regional cerebral blood flow|
|SPM||=||Statistical Parametric Mapping software|
The support of National Institutes of Health grants HL57529 and HL40962 is gratefully acknowledged.
- Received December 15, 1997.
- Revision received December 31, 1997.
- Accepted January 9, 1998.
Edvinsson L, MacKenzie ET, McCulloch J. Cerebral Blood Flow and Metabolism. New York, NY: Raven Press; 1993.
Becker JT, Mintun MA, Aleva K, Wiseman MB, Nichols T, DeKosky ST. Compensatory reallocation of brain resources supporting verbal episodic memory in Alzheimer’s disease. Neurology. 1996;46:692–700.
Talairach J, Tournoux P. A Stereotactic Co-Planar Atlas of the Human Brain. Stuttgart, Germany: Thieme; 1988.
Grady CL, Miasog JM, Horwitz B, Ungerleider LG, Mentis MJ, Salerno JA, Pietrini P, Wagner E, Haxby JV. Age-related changes in cortical blood flow activation during visual processing of faces and location. J Neurosci. 1994;14:1450–1462.
Deleted in proof.
Cabeza R, Nyberg L. Imaging cognition: an empirical review of PET studies with normal subjects. J Cog Neurosci. 1997;9:1–26.
Smith EE, Jonides J. Working memory in humans: Neuropsychological evidence. In: Gazzianga MS, ed. Handbook of Cognitive Neurosciences. Cambridge, Mass: MIT Press; 1995:1009–1020.
Fiez JA, Raife EA, Balota DA, Schwarz JP, Raichle ME, Petersen SE. A positron emission tomography study of the short-term maintenance of verbal information. J Neurosci. 1996;16:808–822.
Smith EE, Jonides J, Koeppe RA. Dissociating verbal and spatial working memory using PET. Cereb Cortex. 1996;6:11–20.
Fox PT, Raichle ME, Mintun M, Dence C. Nonoxidative glucose consumption during focal physiologic neural activity. Science. 1988;241:462–464.
Roland PE, Eriksson L, Stone-Elander S, Widen L. Does mental activity change the oxidative metabolism of the brain? J Neurosci. 1987;7:2373–2389.
Baringa M. What makes brain neurons run? Science. 1997;276:196–198.
Mentis MJ, Salerno J, Horwitz B, Grady C, Schapiro MB, Murphy DGM, Rapoport SI. Reduction of functional neuronal connectivity in long-term treated hypertension. Stroke. 1994;25:601–607.
Sugimori H, Ibayashi S, Irie K, Ooboshi H, Nagao T, Fugii K, Sadoshima S, Fujishima M. Cerebral hemodynamics in hypertensive patients compared with normotensive volunteers: a transcranial Doppler study. Stroke. 1994;24:1384–1389.
Strassburger TL, Lee HC, Daly EM, Szczepanik J, Krasuski JS, Mentis MJ, Salerno JA, DeCarli C, Schapiro MB, Alexander GE. Interactive effects of age and hypertension on volumes of brain structures. Stroke. 1997;28:1410–1417.
Salerno JA, Murphy DGM, Horwitz B, DeCarli C, Haxby JV, Rapoport SI, Schapiro MB. Brain atrophy in hypertension. Hypertension. 1992;20:340–348.
Salerno JA, Grady C, Mentis M, Gonzalez-Aviles A, Wagner E, Schapiro MB, Rapoport SI. Brain metabolic function in older men with chronic essential hypertension. J Gerontol Med Sci. 1995;50A:M147–M154.
Shapiro AP, Miller RE, King HE, Ginchereau CH, Fitzgibbon K. Behavioral consequences of mild hypertension. Hypertension. 1982;4:355–360.