Subarachnoid hemorrhage is followed by temporomesial volume loss

Study design, patients, and controls.

During February 1, 1995, and December 1999, all consecutive patients with SAH admitted to Kuopio University Hospital were candidates for a prospective, randomized study. According to the study protocol inclusion criteria,^2,3 155 patients with acute aneurysmal SAH were randomly assigned to either endovascular treatment (n = 77) or surgical ligation (n = 78). All randomized aneurysmal SAH patients were scheduled to undergo brain MRI and neuropsychological assessment 1 year after SAH. The current study included 77 aneurysmal SAH patients (41 with originally endovascular, 36 surgical treatment). In addition, 30 age- and sex-matched healthy control subjects were imaged. The baseline characteristics of the study population patients are shown in table 1.

The study design was approved by the ethical committee at our hospital. All participants or their nearest relative gave informed consent for their participation in the study.

MR image acquisition and analysis.

The patients and the control individuals underwent MRI scanning on a Siemens Vision 1.5-T scanner (Erlangen, Germany) with circular polarized head coil. The MRI protocol consisted of T2- and proton density (PD)-weighted images, magnetic resonance angiography, and three-dimensional T1-weighted sequences.³ The areas of high signal on T2- and PD-weighted images were evaluated on consensus reading with an experienced neuroradiologist (R.V.) and an experienced neurosurgeon (T.K.). The experienced neurosurgeon (T.K.) did not perform any of the operations in this study population. Three-dimensional T1-weighted anatomic images were obtained using the magnetization-prepared rapid gradient echo sequence (repetition time = 9.7 milliseconds; echo time = 4 milliseconds; inverse time = 20 milliseconds; flip angle = 12 °; field of view = 250 mm; matrix size = 256 × 256). The oblique coronal slices (thickness 2.0 mm) included the whole cerebrum. The three-dimensional T1 sequence for the volumetry was scheduled at the end of the imaging, if the patient was cooperative and there were no movement artifacts and the patient was willing to stay the extra time in the MR scanner.

Standard neuroanatomic landmarks (such as anterior commissure-posterior commissure line, hemispheres, and orbits) were used to reconstruct slices (thickness 2 mm) perpendicular to the long axis of the left hippocampus. These 16-bit Siemens format images were converted to the ANALYZE format (Mayo Foundation, MN). The volumes of the amygdalae and hippocampi were manually measured by a single observer (A.K.). To maintain blind assessment and confidentiality, data were registered and coded by subject numbers only. The intraclass correlation coefficients for intrarater reliability (bilateral measurements of 10 patients) were 0.94 for the amygdala and 0.88 for the hippocampal volumes. In our study, the point counting method (EasyMeasure, version 1.0, MariArc, Liverpool, Great Britain) was used to calculate the volumes of the hippocampus and the amygdala. The point-counting method consists of overlaying a systematic array of the test points (one point per nine pixels) completely over each slice. Instances in which a point, in fact a cross, lies within the area of interest, are recorded by clicking a computer mouse. The cross was chosen or not chosen according to the following rules: If the upper right quadrant of the cross is inside the area, the cross is recorded; otherwise, the cross is excluded. When the cross is lying in a 45 ° angle to the slice, the cross is accepted, if the right quadrant is within the area. In both cases, the area must extend to the intersection of the cross. Subsequently, the total number of the test points is multiplied by the volume of the test point (17.17 mm³). Boundaries of these nuclei were identified with the use of the neuroanatomy atlas⁴ and the previously published research.⁵ The point counting of the structures proceeded from anterior to posterior. The method of the volumetry is shown in figure 1.

This intracranial area has been noted to give a reliable correlation to the whole brain volume.⁶ To remove the influence of the head size to the volumes of nuclei, the volumes were normalized by dividing the volume of the nucleus with the slice volume of the intracranial area: (volume of the structure/intracranial area in reference slice) × 100.⁷ Normalized volumes were further defined as ipsilateral or contralateral according to the side (right or left) the ruptured aneurysm was filling from. In midline (basilar) aneurysms, ipsilateral side was defined as the side where the surgery or endovascular catheterization was performed.

As a rough measurement of the ventricular size and possible central type of brain atrophy, the ventricular-to-intracranial width ratio was measured.

Neuropsychological analysis.

Neuropsychological assessments were performed by a single neuropsychologist (H.H.) 12 months after treatment. The comprehensive evaluation included tests of general intelligence, memory and selected language abilities, and assessment of attention and flexibility of mental processing.³

Statistical analysis.

The statistical software SPSS for Windows 11.5 (SPSS Inc., Chicago, IL) was used to analyze the data. In all statistical analyses, the volumes normalized for intracranial area were used. The level of significance was set at p < 0.05. Because each volumetric (right/left and ipsilateral/contralateral) parameter was used twice in subgroup analysis, the Bonferroni correction was used, and the level of significance in these analyses was set at p < 0.025. Ordered categorical data and dichotomous variables were examined by χ² statistics. A t test was used to compare normally distributed data for the volumes for comparisons between the patients and controls and the different treatment groups. The Pearson correlation coefficient was used to analyze the correlation between the volumes of hippocampal and amygdaloid structures, age, ventricular size, and neuropsychological test results.

Descriptive characteristics of the patients.

One year after subarachnoid hemorrhage, 129 of the original 155 randomized SAH patients underwent brain MRI. At that time, 20 patients (12.9%) had died. Claustrophobia prohibited MRI examination in 4 cases (2.6%), and 2 patients (1.3%) declined the examination.

Altogether, 77 patients from the randomized population (33 endovascular, 36 surgical, and 8 combination treatment patients) underwent three-dimensional T1-weighted examinations. Nine randomized patients had additional one or more nonruptured aneurysms, which were treated surgically in 4 patients, endovascularly in 2 patients, and with both treatment modalities in 3 patients before the MRI at the 1-year follow-up. A majority (84.4%) of the patients had good clinical outcome (GOS 5), 11 (14.3%) had moderate disability (GOS 4), and 1 (1.3%) had poor outcome (GOS 3) at 1 year after subarachnoid hemorrhage.

The volumes of the amygdalae of three patients had to be excluded from analysis because of the clip artifact (figure 2) after the surgery. No disturbing artifacts due to clips or coils were detected in the regions of hippocampi. When comparisons between treatment groups (endovascular treatment vs surgery) were performed, crossover patients (n = 8) and those patients whose nonruptured aneurysms had been treated before the 1-year MRI (n = 9) were excluded from the analysis. After exclusion, 30 endovascular and 31 surgical patients were included in the final analysis comparing the actual treatment modalities. The flow diagram of the original cohort of randomized SAH patients is demonstrated in figure 3. The baseline characteristics of these patients (age, sex, education years, preoperative Hunt and Hess grades and Fischer grades, and the location of the ruptured aneurysm) maintained statistically balanced between the different treatment groups. The whole study population was ethnically homogenous.

Temporomesial volumes.

Normalized temporomesial volumes in SAH patients and controls are summarized in table 2. When comparing the normalized volumes of the hippocampi (HC) and amygdalae (AM), differences between SAH patients and controls were seen in the measured volumes of right (23.2 ± 3.9 vs 24.7 ± 3.7; p = 0.072) and left HC (21.3 ± 3.5 vs 23.7 ± 3.6; p = 0.002) and right (18.4 ± 4.9 vs 21.0 ± 3.9; p = 0.012) and left AM (18.8 ± 4.1 vs 20.5 ± 3.7; p = 0.045) after SAH, the last three volumes being significantly smaller in the SAH patients.

Comparisons of the temporomesial volumes between the endovascularly treated patients and controls did not reach significant differences: right AM, 20.7 ± 4.38 vs 20.96 ± 3.91 (p = 0.790); left AM, 19.66 ± 2.90 vs 20.52 ± 3.67 (p = 0.319); right HC, 23.56 ± 4.43 vs 24.71 ± 3.71 (p = 0.281); left HC, 21.79 ± 3.76 vs 23.67 ± 3.64 (p = 0.055). However, the corresponding comparisons between the surgically treated patients and controls revealed the following differences: right AM, 16.31 ± 4.82 vs 20.96 ± 3.91 (p < 0.001); left AM, 18.97 ± 3.69 vs 20.52 ± 3.67 (p = 0.111); right HC, 22.50 ± 3.73 vs 24.71 ± 3.71 (p = 0.024); left HC, 20.81 ± 3.70 vs 23.67 ± 3.64 (p = 0.004). Among all the aneurysmal SAH patients, smaller ipsilateral (17.8 ± 5.0) than contralateral amygdaloid volumes (19.4 ± 3.9) were observed (p = 0.008). In the surgical subgroup (n = 31), the difference between the mean volumes of the ipsilateral and contralateral amygdalae was evident (15.7 ± 5.1 vs 19.4 ± 2.9; p = 0.001), whereas in the endovascular subgroup (n = 30), no such difference in the normalized volumes of the ipsilateral (20.3 ± 4.1) and contralateral (20.0 ± 3.1) amygdalae was found (p = 0.650). There was no difference in the ipsilateral and contralateral volumes of hippocampi. Ipsilateral amygdaloid volumes were (after Bonferroni correction) smaller after surgical (15.7 ± 5.1) than after endovascular (20.3 ± 4.3) treatment of SAH (p < 0.001). There were no differences in volumes of contralateral amygdalae or hippocampi between the different treatment groups. Normalized temporomesial volumes after endovascular and surgical treatment are summarized in table 3 and illustrated in figures 4 and 5.

The possible effects of the other contributing factors to the measured volumes were also analyzed. Patients with lower Hunt and Hess grades (I and II) showed smaller volumes in both right (17.8 ± 4.6) and left (18.2 ± 4.3) amygdalae compared with patients with higher Hunt and Hess grades (III through V) (right, 20.2 ± 5.1; p = 0.04; left, 20.2 ± 2.5; p = 0.04). However, after Bonferroni correction, this difference was not significant. The hippocampal volumes were almost identical in both dichotomized Hunt and Hess grade groups: right, 22.9 ± 4.2 Hunt and Hess grades I and II vs 22.7 ± 4.0 Hunt and Hess grades III through V (p = 0.79); left, 21.1 ± 3.6 Hunt and Hess grades I and II vs 21.1 ± 3.9 Hunt and Hess grades III through V (p = 0.78). In terms of preoperative Fischer grades or hydrocephalus, no differences in the volumes could be found.

Surgical and endovascular patient groups did not differ in regard to complications. The volumetric patient study population consisted of patients with relatively good clinical outcome after SAH, a fact probably explaining the relatively low rate of complications. No postoperative hematomas were seen in this study population. One patient in both treatment groups (2.4% of endovascular and 2.7% of surgical patients; p = 0.92) developed seizures. Twenty-four (39.3%, 11 endovascular, 13 surgical; p = 0.67) of 61 patients had external ventricular drainage (EVD) for treatment of acute hydrocephalus. Two originally endovascular patients (4.9%) and 4 originally surgical patients (11.1%) developed either clinical or subclinical meningitis (positive bacteria in postoperative ventricular drainage or lumbar puncture), indicating no difference (p = 0.31) between the treatment groups. External ventricular drainage placement, a permanent shunt device, ventriculitis, or development of seizures had no effect on measured volumes of amygdalae or hippocampi.

Twenty-six (33.8%, 11 endovascular, 15 surgical; p = 0.17) of 77 patients developed symptoms that were interpreted to be caused by clinical vasospasm. However, the signs of clinical vasospasm were not associated with reduced temporomesial volumes.

In the volumetric study population, parenchymal lesions were seen in 52 (67.5%) of 77 patients (including any high–signal intensity lesions seen in T2- and PD-weighted images, e.g., lesions due to ischemic attacks, hematomas, brain retraction deficits, or lesion associated with ventricular catheterization or permanent shunt placement). The presence of a parenchymal lesion was not significantly associated with any of the measured temporomesial volumes (neither AM nor HC).

Among the aneurysmal SAH patients (n = 77), the presence of any high-intensity lesion in either of the temporal lobes seen in T2- and PD-weighted images was associated with smaller amygdalae both on the right (15.9.1 ± 5.2 vs 19.7 ± 4.3; p = 0.01) and on the left (17.6 ± 5.1 vs 19.5 ± 3.4; p = 0.034). However, after Bonferroni correction, this difference on the left side was not significant. The ipsilateral amygdala was smaller (14.7 ± 4.0 vs 19.2 ± 4.0; p < 0.001) if a temporal lobe deficit was present, and on the contralateral side, the difference was not significant (p = 0.07). Furthermore, if a temporal lobe high-intensity lesion was present, the volume of ipsilateral amygdala tended to be smaller in the surgical subgroup (p = 0.04) but not in the endovascular subgroup (p = 0.85). No differences in the hippocampal volumes were noted with respect to temporal lobe high–signal intensity lesions.

Among all aneurysmal SAH patients (including the crossover patients and patients whose unruptured aneurysms were also treated), the presence of a parenchymal retraction injury detected on T2-weighted images was strongly associated with smaller ipsilateral amygdala volumes: 14.7 ± 5.3/19.3 ± 4.0 (p < 0.001). Contralateral amygdala volumes did not correlate with the presence of parenchymal retraction lesions (p = 0.86). When only the patients in the surgical subgroup were included in this analysis (n = 31), the corresponding volumes were 14.3 ± 5.1 and 17.6 ± 4.0 (p = 0.09), not reaching the level of statistical significance. Retraction injury on T2-weighted images had no effect on the hippocampal volumes. Reduced temporomesial volumes are demonstrated in figures 6 and 7.

In the matched control population, the following correlation coefficients between the age of the individual and the measured temporomesial volumes were obtained: right AM, r = −0.41, p = 0.03; left AM, r = −0.52, p = 0.004; right HC, r = −0.60, p = 0.001; left HC, r = −0.60, p = 0.002. Age of the individual also correlated with ventricular/parenchymal ratio in the control group: r = 0.381, p = 0.038.

In the aneurysmal SAH population, correlation between age and both ipsilateral and contralateral hippocampal volumes was somewhat lower: ipsilateral HC, r = −0.31, p = 0.006; contralateral HC, r = −0.33, p = 0.003. The volumes of amygdalae did not correlate with age: ipsilateral AM, r = −0.057, p = 0.61; contralateral AM, r = −0.071, p = 0.538. Age of the SAH patient correlated with the measured ventricular/parenchymal ratio: r = 0.366, p = 0.001. Furthermore, SAH patients had higher mean ventricular/parenchymal ratios (0.23 ± 0.06) than did the control individuals (0.20 ± 0.04; p = 0.021). SAH patients with higher ventricular/parenchymal width ratios showed smaller volumes of contralateral hippocampi (r = −0.278, p = 0.014), but no correlation was demonstrated between the volumes of ipsilateral hippocampus (r = −0.162, p = 0.160) or the volumes of ipsilateral (r = 0.054, p = 0.648) or contralateral amygdalae (r = −0.157, p = 0.172). Furthermore, among SAH patients, education years and hippocampal volumes showed clear positive correlation (right HC, r = 0.408, p < 0.001; left HC, r = 0.367, p = 0.001), but no correlation between the amygdaloid volumes and education years was found (right AM, r = −0.041, p = 0.724; left AM, r = 0.206, p = 0.07).

Neuropsychological results and temporomesial volume correlation.

Among the SAH patients, age of the patient correlated with all neuropsychological parameters. The neuropsychological test results and the correlation to the measured volumes of the hippocampi and amygdalae are seen in table 4. The hippocampal volumes correlated with a visual memory test (Visual Reproduction Subtest¹¹ and several tests of attention, flexibility of mental processing, intellectual ability, and psychomotor speed. A test of verbal episodic memory¹¹ and another visual memory test (Rey)⁹ did not correlate with hippocampal volumes. The volumes of the amygdalae did not significantly correlate with neuropsychological test results.