Single-Photon Emission Computed Tomography Using Hexamethylpropyleneamine Oxime in the Prognosis of Acute Cerebral Infarction
Background and Purpose The role of single-photon emission CT (SPECT) in the prognosis of cerebral infarction is controversial, but most studies report that SPECT using a variety of radiopharmaceutical agents gives useful prognostic information. Only one study has questioned whether acute perfusion deficits independently add to a valid clinical prognostic score. This study was limited to middle cerebral artery territory infarcts and was negative. We present data on the prognostic utility of SPECT using 99mTc–hexamethylpropyleneamine oxime (HMPAO) in cerebral infarction, unselected by site.
Methods Fifty consecutive unselected patients admitted to the hospital with acute cerebral infarction, of whom 10 died and 7 withdrew, had SPECT performed serially at onset and at 1 week and 3 months after stroke onset using 99mTc-HMPAO and the NOVO 810 dedicated high-resolution head tomograph. Clinical severity at presentation and outcome was measured with the Canadian Neurological Scale and the Barthel Index. Infarct volumes were measured from both the SPECT and CT scans. The data for the 43 subjects who completed the study or died were evaluated to determine the most powerful prognostic measures. Predictors were the Canadian Neurological Scale score at onset and 1 week, the Barthel Index at 1 week, the CT infarct volume typically done between 3 and 7 days after stroke onset, and the infarct volumes at the first and second SPECT. Outcome measures were the Canadian Neurological Scale score and Barthel Index score at 3 months, scored as zero for those patients who died.
Results The clinical prognostic indicators correlated with the outcome measures, with coefficients between .617 and .821 (P<.0006 in all cases). The Canadian Neurological Scale score measured at 1 week was the best of these. Infarct volumes measured from SPECT correlated less well (coefficients between −.518 and −.683, P<.0019 in all cases). CT infarct volume was the poorest predictor. Although SPECT infarct volumes predicted outcome, they did so less well than clinical examination. Spontaneous infarct reperfusion did not affect outcome.
Conclusions Although the measurement of infarct volume on SPECT using 99mTc-HMPAO provides a predictor of stroke outcome, it is not a better predictor than the Canadian Neurological Scale score.
When used with agents whose uptake is related to cerebral blood flow, SPECT shows major abnormalities in cerebral infarction. However, it has not achieved an important diagnostic role because both CT and MRI are more sensitive, except perhaps in the hyperacute stage, and give more information regarding differential diagnosis.
Attention has therefore focused on extracting prognostic information from SPECT. That such data should be obtainable is plausible because SPECT gives information about lesion size, reperfusion, remote changes (diaschisis), and the ability of the damaged tissue to retain the radiopharmaceutical agent over short periods, which may be an indicator of tissue viability. Evidence of a role for SPECT as a useful prognostic indicator has been presented by several studies.1 2 3 4 5 6 7 8 9 Many reports now suggest that after cerebral infarction the infarct volume seen on SPECT using a variety of tracers provides useful prognostic information,4 5 7 9 10 11 12 13 14 15 while a few related studies have been negative.1 3 8 16 Much of this work has been criticized.6 The faults include entry of patients too late for the test to be considered prognostic,4 5 10 11 12 13 oversimplified or dichotomized clinical rating scales,3 4 5 7 9 10 11 13 15 retrospective clinical data,4 and dichotomized imaging data.9 Other studies have concluded that SPECT, with the use of a variety of radiopharmaceuticals, is either a poor indicator of outcome1 8 or is not better than other techniques, such as CT1 or simple clinical examination.14
Any new technique should be evaluated critically to see whether it adds useful new information.17 To date, only one report has compared the predictive accuracy of SPECT in stroke with simpler techniques such as clinical examination.14 However, this study was limited to middle cerebral artery infarcts; consequently, its findings cannot be generalized to all cerebral infarction. Furthermore, the infarct volume was measured with the method of Mountz,2 which calculates the zero-perfusion volume equivalent rather than being a simple measurement of the apparent infarct volume. This technique has not been validated, and there is no evidence that it is better than a simple estimate of the volume of the infarct. Not only have other reports not compared the prognostic accuracy of SPECT to other techniques, but they have often themselves been flawed in several other respects.
It has also been suggested that measurements of infarct volume made with different techniques may be measuring different physiological processes, including reperfusion and the ischemic penumbra, so that ratios of these measurements might provide additional prognostic information.1 4 12
To clarify the role of HMPAO SPECT infarct volume measurements in determining the prognosis after cerebral infarction, we analyzed the data from 43 consecutive unselected patients with acute cerebral infarction who were examined as part of a prospective study of the role of HMPAO SPECT in stroke.
Subjects and Methods
All patients over the age of 18 years admitted to our district general hospital with first stroke were eligible for the study. Reasons for exclusion from entry or continuation in the study were pregnancy, previous stroke or cranial neurosurgery, primary intracerebral hemorrhage, non-English speaker, current severe psychiatric illness, epilepsy, dementia, alcohol abuse, drug abuse, severe intercurrent illness, and any other factor that might interfere with the appearances of the 99mTc-HMPAO SPECT scans or result in a high risk of the patient defaulting from follow-up.
The first clinical examination and SPECT were carried out as soon as possible after the infarct, and subsequent studies, including SPECT and clinical examination, were done at about 1 week and 3 months after stroke onset. On each occasion, an assessment for the CNS score,18 a functional evaluation (the BI19 ), and SPECT scanning were carried out within 2 hours of each other. The CNS is a simple assessment of level of consciousness, impairment of language, and motor deficits that can be carried out in a few minutes by suitably trained nurses. Similarly, the BI is a brief assessment of the patient’s actual functional achievements, which is carried out by taking a functional history and can also be done in a few minutes by suitably trained paramedical personnel. The modified scoring scheme for the BI was used, with scoring up to 20. CT scanning was usually performed between 3 and 7 days after the infarct. MRI was performed later if the CT scan did not reveal the lesion. The study was approved by the hospital ethics committee.
Fifty consecutive patients with cerebral infarction were recruited to the study. Of these, 33 completed the protocol, 10 died during the study, and 7 withdrew, usually because they did not tolerate lying still for the duration of the scan. This report is based on the 43 patients who completed the study or died; data from those withdrawing have not been included in the analyses presented here.
SPECT scanning began 15 to 30 minutes after the injection of 750 MBq of 99mTc-HMPAO (Ceretec, Amersham International). Scanning was started 30 mm below the infraorbitomeatal line; slices were 10 mm apart, and each of 8 to 12 slices covering the whole brain was collected for 5 minutes with the NOVO 810 scanner. This is a dedicated multidetector tomographic head scanner with a resolution of 9 mm, full width at half maximum, in the plane of the scan and 14 mm axially, full width at half maximum. A representative scan of a normal subject is shown in the Figure⇓. The attenuation correction was a linear correction operating with circular symmetry from the center of the field.20 Because this can cause substantial artifactual right/left differences if the patient’s head is not exactly centered, all analyses were done on data reconstructed with no attenuation correction.
SPECT infarct volumes were measured from the SPECT scans with the volume estimation facility provided in the graphic analysis package Analyze (Biodynamics Research Unit, Mayo Clinic). This technique works on a proportionate basis using randomly selected voxels (termed “seeds”) scattered throughout the volume of the scan. If, for example, 10% of the seeds are found within the volume of interest, in this case the infarct, the infarct occupies approximately 10% of the total scan volume. To increase the accuracy of this technique with small infarcts, we used 10 000 seeds in the 160 000 voxels of the total scan volume. Seeds located within the infarct were interactively selected on a two-dimensional slice-by-slice basis. Because the borders of the lesions were generally sharp, there was rarely any difficulty in determining whether a seed was in the lesion or not. When there was doubt, the location of the seed was examined in three dimensions; if doubt persisted, the seed was included within the lesion. The infarct volume was calculated from the number of seeds selected. Each scan was analyzed without reference to other scans of the same patient. Areas of hyperemic reperfusion were not classified as infarcts on the scan in which the hyperemia was seen, since such regions do not necessarily proceed to complete infarction. This method for measuring SPECT infarct volumes was validated by comparison, using Spearman’s rank correlation, of the volumes measured at different times for the same patient and also comparison of the SPECT volumes with the CT volumes, which provides an external frame of reference. CT infarct volumes were measured by planimetry with an Apple Graphics Tablet.
Reperfusion was defined here as either (1) an absolute increase in HMPAO uptake in the infarct (as demonstrated by CT scan or MRI) compared with the homologous contralateral region more than 2 SDs above the normal side-to-side variability found in control subjects21 or (2) a relative increase in HMPAO uptake in the infarct as shown in one or more HMPAO SPECT scans in a given case compared with the same volume in an earlier SPECT scan, the increase exceeding 2 SDs above the scan-to-scan variability in these subjects.
The CNS scores at the first and second examinations and the BI at the second examination were tested as predictors of the CNS score and BI at the third examination (outcome measures) by Spearman’s rank correlation. Infarct volumes were also assessed as predictors of outcome. Because certain mathematical derivations of the CT and SPECT volumes have been reported to be good prognostic indicators,12 these derivations were also tested. Reperfusion as a predictor of outcome was assessed by comparing the final CNS and BI in those cases with and without reperfusion. Each of these analyses was carried out for the patients who completed the study alone and then with the addition of those who died, using zero as their score on the final CNS score and BI. The importance of the prognostic variables in determining the outcome variables was assessed using forward stepwise multiple logistic regression. The final CNS score and BI were dependent variables, and the early CNS scores, BI, and CT and SPECT volumes were independent variables. Reperfusion was not used in this part of the analysis because relative reperfusion, which accounted for most of the observations of reperfusion, requires two SPECT scans for its detection.
The first SPECT scan and clinical examination were performed at a median of 1.1 days from stroke onset, the second at 6.8 days, and the third at 95 days. CT scanning was performed in the 50 patients at a median of 4.6 days (range, 0.2 to 9.8 days) from stroke onset. The infarct volumes were not normally distributed. For the CT scans, the median was 12 cm3 (range, 0 to 273 cm3); for the SPECT scans, the median values were for the first scan, 18 cm3 (range, 0 to 221 cm3); for the second, 6 cm3 (range, 0 to 211 cm3); and for the third, 19 cm3 (range, 0 to 250 cm3).
Data comparing the infarct volumes measured by SPECT at different times and by CT are given in Table 1⇓. These coefficients were between .57 and .79, in each case with values of P<.0001.
Neither the CNS score nor the BI were normally distributed. Furthermore, the stroke scales cannot be assumed to be interval data sets; they must be assumed to be ordinal. Consequently, in assessing the relationships between the stroke scores, Spearman’s rank correlation was used.
Initially, with the analysis limited to the 33 patients who completed the study, the CT and SPECT scan infarct volumes and the early scores on the CNS and the BI all correlated significantly with clinical outcome (see Table 2⇓). The correlation coefficients were, however, not high (between .57 and .74). The ratio of the CT and SPECT infarct volumes correlated very poorly with clinical outcome. Further derivations such as the log of the SPECT volume12 were therefore not tested. The best prognostic indicator was the CNS score for the examination carried out at 1 week. When data were included from the 10 patients who died (with those clinical examinations that were due after death scored as zero), clinical examination with the CNS correlated more closely with outcome, but prediction of outcome from infarct volumes did not improve.
Spontaneous reperfusion was seen in 14 cases; this was absolute (above normal) in 8 cases and was seen on the first scan in 4 cases and on the second scan in the remainder. In 6 patients, reperfusion was seen on the second SPECT scan relative to the deficit seen on the first scan. By definition, relative reperfusion cannot be seen on the first scan. Reperfusion did not affect outcome as measured by the final CNS score (reperfused median, 6.0; not reperfused, 8.5; Mann-Whitney U test, P=.501) or BI (reperfused median, 20.0; not reperfused median, 20.0; P=.464) in those patients who survived to the final SPECT. Three of the 14 reperfused patients died. Reanalysis of the data, scoring the outcome variables as zero in those who died, did not change the findings (final CNS score: reperfused median, 5.0; not reperfused, 6.0; Mann-Whitney U test, P=.684; final BI: reperfused median, 9.8, not reperfused median, 11.9; P=.745). These relationships were not changed by adjusting for infarct volume with an ANCOVA.
Forward stepwise multiple logistic regression using either the final CNS score or BI as the dependent variable and the early CNS scores, BI, and CT and SPECT volumes as independent variables showed that only the CNS score carried out at 1 week entered the model, using F-to-enter and F-to-remove. ANOVA for the regression model showed that it was statistically highly significant but that the model accounted for relatively little of the variance (Table 3⇓).
This prospective study of serial unselected patients with first stroke has confirmed that infarct volume measured in the first week with CT or SPECT correlates with the simple standard measures of neurological and functional outcome used here (ie, the CNS and the BI), with statistically highly significant coefficients of between .48 and .68.
However, most studies considering the prognostic role of SPECT have, with one exception,14 failed to consider the possibility that other simpler methods of prognostication may be better than SPECT infarct volumes and related measurements. This is crucial because new technologies should not be promoted or put into routine use unless they add useful new data or have other advantages over older techniques.17
Clinical examination is a well-established means of predicting the outcome of cerebral infarction.22 23 24 25 In this study, we have shown that the CNS score measured at 1 week has correlation coefficients of .76 with the outcome CNS score and .82 with the outcome BI when fatalities are included and scored as zero. These figures are better than those obtained using SPECT or CT infarct-volume data in the same patients. Furthermore, multiple regression showed that of all the prognostic factors considered here, only the CNS score at 1 week was needed in the regression model.
Davis et al14 reached similar conclusions in finding that clinical examination was a better predictor of outcome than the calculated hypoperfusion volume on SPECT, which is an unvalidated method of assessing infarct volume described by Mountz.2 Validation of the measurement of SPECT infarct volume is difficult, since there is no gold standard and the volume will fluctuate with time. Our technique for measuring SPECT volumes is in essence a sampling technique. Because the SPECT infarct volumes measured at different times in the same patient correlated closely with each other, as they did with the CT infarct volume, it seems likely that the technique is valid. Furthermore, the SPECT infarct volumes correlated with outcome better than the CT infarct volumes, suggesting that our technique is at face value a measure of stroke severity superior to CT. The work of Davis et al14 was confined to infarcts in the territory of the middle cerebral artery, which may limit its general applicability. Our work included infarcts in all sites and so confirms and generalizes the observations of Davis et al. We found the early CNS score to be a better predictor of final outcome than did Davis et al, who found the Allen scale22 to be a better predictor. They found that the hypoperfusion volume was a useful addition to the CNS score as a predictor of outcome, but it did not add to the predictive power of the Allen scale. Apart from the difference in the distribution of lesions that were studied, the timing of the clinical evaluations may be a very important factor in explaining the higher degree of correlation that we saw with the CNS; we found it to be most accurate when administered at 1 week, whereas it was administered within 72 hours by Davis et al.14
Not only could we identify no special role for SPECT infarct volumes, but we found that CT infarct volume alone was only marginally inferior to SPECT as a predictor of outcome. CT volume measurements have rarely been compared with SPECT in this way before, and in some work15 CT infarct volume subclassification has produced misleading conclusions. Rango et al15 classified their CT lesions as none, small, and large, where large was defined as over 2 cm3. They found no correlation between CT deficit and clinical status but did between SPECT deficit and clinical status, using this to argue for some special role for SPECT scanning. It seems more likely that the absent correlation of clinical deficit with CT was an artifact of inappropriate classification of CT lesion size.
The correlations between the prognostic indicators and outcome measures were affected by the inclusion of outcome data from those patients who died. Inclusion of these patients improved the correlation between the CNS score and the BI done at 1 week compared with that achieved by the CNS score when carried out on the first or second day. This probably simply reflects the consequences of increasing cerebral edema, pneumonia, hypoxia, etc, on neurological function and consciousness, given the close association of these adverse developments with mortality. Early SPECT was a better predictor of outcome than SPECT scans done at 1 week, probably because these later SPECT scans more often showed reperfusion that causes an artifactual lowering of the apparent infarct volume.21 The inclusion of fatal cases caused the correlation between the second SPECT and outcome to decrease, while the correlation improved slightly between the first SPECT and outcome measures. This deterioration in the correlation between the second SPECT volume and outcome is probably an exaggeration of the effect of reperfusion to decrease infarct volume, since mortality was closely related to infarct volume measured at the first SPECT (nonfatal cases, infarct volume of 42 cm3; fatal cases, 120 cm3; P=.0051, Mann-Whitney U test) but not when measured at the second SPECT (volumes of 33 and 30 cm3, respectively; NS).
The ratio of the CT and first SPECT infarct volumes correlated less well with outcome measures than either of these volumes alone. These observations contradict the findings of Launes et al4 and do not support the suggestion that the difference between the SPECT and CT scans may represent some evolving process of prognostic importance. Methodological reasons may account for the differences: Launes et al carried out their prognostic scans up to 120 days after stroke onset, obtained clinical information by retrospective chart review, and used very simple measures of outcome. Theoretically, the appearance of a penumbra might affect the measured SPECT infarct volume and could produce a positive result. However, we did not observe SPECT changes that were consistent with a penumbra.21 26 This may have been because our scans were too late to detect the penumbra (in which case the penumbra is a very transient phenomenon), because the penumbra is sufficiently narrow that it falls below the resolution of the scanner, because the penumbra occurs only rarely, or finally, because HMPAO SPECT is a poor technique for detecting the penumbra. Only this last explanation negates our findings, and since HMPAO SPECT is a well-validated proportional indicator of cerebral perfusion at the levels of blood flow that would be expected in a penumbra, this seems unlikely. In the cases of all the other possible explanations, our negative findings argue against HMPAO SPECT detecting an evolving process of prognostic importance.
Various aspects of HMPAO, particularly its nonlinear relationship to cerebral blood flow and its redistribution,27 28 29 30 31 may affect its accuracy in demarcating infarcts; newer radiopharmaceuticals may overcome these limitations. Promising agents will need to be assessed in a similar way.
We conclude that infarct volumes measured from HMPAO SPECT do provide prognostic information but that this information is only marginally better than that obtainable from CT scanning and is less accurate than a simple clinical rating scale. The use of infarct volume measurements from HMPAO SPECT cannot be recommended as a routine prognostic tool in cerebral infarction.
Selected Abbreviations and Acronyms
|CNS||=||Canadian Neurological Scale|
|SPECT||=||single-photon emission computed tomography|
This study was supported by the Stroke Association (Dr Bowler).
Reprint requests to Dr J.V. Bowler, Academic Unit of Neuroscience, Charing Cross and Westminster Medical School, St Dunstan’s Rd, London W6 8RP, UK. E-mail email@example.com.
- Received May 15, 1995.
- Revision received September 26, 1995.
- Accepted September 26, 1995.
- Copyright © 1996 by American Heart Association
Hayman LA, Taber KH, Jhingran SG, Killian JM, Carroll RG. Cerebral infarction: diagnosis and assessment of prognosis by using 123IMP-SPECT and CT. Am J Neuroradiol. 1989;10:557-562.
Limburg M, van Royen EA, Hijdra A, Verbeeten B. rCBF-SPECT in brain infarction: when does it predict outcome? J Nucl Med. 1991;32:382-387.
Limburg M, van Royen EA, Hijdra A, de Bruine JF, Verbeeten BWJ. Single-photon emission computed tomography and early death in acute ischemic stroke. Stroke. 1990;21:1150-1155.
Giubilei F, Lenzi GL, Di Piero V, Pozzilli C, Pantano P, Bastianello S, Argentino C, Fieschi C. Predictive value of brain perfusion single-photon emission computed tomography in acute ischemic stroke. Stroke. 1990;21:895-900.
Raynaud C, Rancurel G, Samson Y, Baron JC, Soucy JP, Kieffer E, Cabanis E, Majdalani A, Ricard S, Bardy A, Bourgignon M, Syrota A, Lassen A. Pathophysiologic study of chronic infarcts with I-123 isopropyl iodo-amphetamine (IMP): the importance of periinfarct area. Stroke. 1987;18:21-29.
Mountz JM, Modell JG, Foster NL, DuPree ES, Ackermann RJ, Petry NA, Bluemlein LE, Kuhl DE. Prognostication of recovery following stroke using the comparison of CT and technetium 99m HM-PAO SPECT. J Nucl Med. 1990;31:61-66.
Gupta S, Bushnell DL, Mlcoch A, Eastman G, Barnes WE, Fisher SG. Utility of late N-isopropyl-p-(iodine-123)-iodoamphetamine brain distribution in predicting outcome following cerebral infarction. Stroke. 1991;22:1512-1518.
Davis SM, Chua MG, Lichtenstein M, Rossiter SC, Binns D, Hopper JL. Cerebral hypoperfusion in stroke prognosis and brain recovery. Stroke. 1993;24:1691-1696.
Smith FW, Donald RT, Morris AJ, Sharp PF, Gemmell HG. The study of regional cerebral blood flow in stroke patients using technetium 99m HMPAO. Br J Radiol. 1988;61:358-361.
Caplan LR. Question-driven technology assessment: SPECT as an example. Neurology. 1991;41:187-191.
Coté R, Hachinski VC, Shurvell BL, Norris JW, Wolfson C. The Canadian Neurological Scale: a preliminary study in acute stroke. Stroke. 1986;17:731-737.
Mahoney FI, Barthel DW. Functional evaluation: the Barthel Index. Maryland State Med J. 1965;14:61-65.
Holm S. Performance analysis of a new multidetector brain SPECT scanner. In: Schmidt HAE, Emrich D, eds. Nuklearmedizin: Clinical Demands on Nuclear Medicine. Stuttgart, Germany: FK Schatlauer Verlag; 1986:158-160.
Bowler JV. Cerebral Infarction and99Tcm HMPAO SPECT. London, UK: University of London; 1993. Thesis.
Allen CMC. Predicting the outcome of acute stroke: a prognostic score. J Neurol Neurosurg Psychiatry. 1984;47:475-480.
Hantson L, De Weerdt W, De Keyser J, Diener HC, Franke C, Palm R, Van Orshoven M, Schoonderwalt H, De Klippel N, Herroelen L. The European stroke scale. Stroke. 1994;25:2215-2219.
Bowler JV, Wade JPH, Jones BE. Stroke, SPECT and the ischaemic penumbra. J Neurol Neurosurg Psychiatry. 1992;55:420. Abstract.
Yonekura Y, Nishizawa S, Mukai T, Fujita T, Fukuyama H, Ishikawa M, Kikuchi H, Konishi J, Andersen AR, Lassen NA. SPECT with 99mTc-d,l-hexamethyl-propylene amine oxime (HM-PAO) compared with regional cerebral blood flow measured by PET: effects of linearization. J Cereb Blood Flow Metab. 1988;8(suppl):S82-S89.
Andersen AR, Friberg HH, Schmidt JF, Hasselbalch SG. Quantitative measurements of cerebral blood flow using SPECT and 99mTc-d,l-HM-PAO compared to xenon-133. J Cereb Blood Flow Metab. 1988;8(suppl):S69-S81.
Nakano S, Kinoshita K, Jinnouchi S, Hoshi H, Watanabe K. Comparative study of regional cerebral blood flow images by SPECT using xenon-133, iodine-123 IMP, and technetium-99m HM-PAO. J Nucl Med. 1989;30:157-164.
Bullock R, Patterson J, Park C. Evaluation of 99mTc-hexamethylpropyleneamine oxime cerebral blood flow mapping after acute focal ischemia in rats. Stroke. 1991;22:1284-1290.