Multimodal CT Imaging in Acute Stroke
To the Editor:
Kilpatrick et al1 retrospectively studied the prospective value of unenhanced CT, CT angiography (CTA), and xenon-enhanced perfusion CT (pCT) in acute stroke on the incidence of infarction on follow-up CT and on discharge disposition. To my knowledge, this is the first observation supporting the view that the assessment of arterial obstruction with CTA could be used as a surrogate for the measurement of cerebral blood flow (CBF). Nine of 10 patients with occlusions of the internal carotid artery or middle cerebral artery (MCA) and no infarctions on baseline CT could not be discharged home (positive predictive value [PPV]=90% [95% confidence interval=60% to 98%]; specificity=92% [65% to 99%]) compared with 9 of 11 patients with compromised CBF <30 mL/100 g per minute (PPV=82% [52% to 95%]; specificity=83% [55% to 95%]). The authors concluded from their observations that the combination of pCT and CTA in conjunction with CT is more capable of predicting new infarction and discharge disposition than is the admission National Institutes of Health Stroke Scale (NIHSS) score plus CT. Because of mistakes in Table 2, the reader is unfortunately not able to duplicate the results and follow this conclusion. The table presents 13 of 13 patients with normal CBF and open CTA discharged to rehabilitation instead of 3 patients. It presents only the results of 12 baseline CT studies within 6 hours of stroke onset of 15 patients with NIHSS score >6. According to the table, 7 new infarcts appeared on follow-up CT and not only 4 new infarcts as reported by the authors. Following the authors’ calculation, a dichotomized NIHSS score has a PPV for new infarcts on follow-up CT of 44% (19% to 73%). In comparison, the PPV of compromised CBF is 55% (28% to 79%) and of arterial obstruction as assessed by CTA is 60% (31% to 83%). The broad overlap of 95% confidence intervals shows that this study does not have the statistical power to justify the conclusion that imaging with CT provides important prognostic information that cannot be derived from the neurological score at admission.
The study bears another problem. The authors defined as normal all CT scans that presented low attenuation in <50% of the MCA territory on 1 CT level. Taking into account that MCA trunk occlusions are often associated with basal ganglia infarctions, this definition means that many of the patients in this study had an infarction on baseline or follow-up CT, but were counted as patients with normal CT. It would be of interest to know whether CTA and pCT really add important prognostic information to the information that can be derived from the unenhanced baseline CT and the neurological score.
- ↵Kilpatrick M, Yonas H, Goldstein S, Kassam AB, Gebel JM, Wechsler LR, Jungreis CA, Fukui MB. CT-based assessment of acute stroke: CT, CT angiography, and xenon-enhanced CT cerebral blood flow. Stroke. 2001; 32: 2543–2549.
We would like to thank Dr von Kummer for his very thoughtful and meticulous review of our article, which examined the potential clinical insights gained by an initial ischemic stroke assessment consisting of a traditional clinical and CT assessment to which was added a CT angiogram and a xenon/CT CBF study. Thirty-one patients were studied between 0 and 6 hours from onset and 19 patients were studied between 6 and 24 hours. The remarkable finding of this relatively small study was that statistically significant improvements in the ability to predict both discharge diagnosis and the occurrence of cortical infarction was provided by the addition of CT angiography or the CBF assessment. Unfortunately, too few patients were available to assess the possible additive insights gained by integrating vascular anatomy and quantitative CBF.
Dr von Kummer was correct in identifying 2 typographical errors in Table 2, but it should be noted that the statistical assessments reported in the article utilized the correct data set and are therefore valid. Dr von Kummer’s decision to perform an assessment of the positive predictive value on this relatively small data set, first with the erroneous tabular data and then with his assessment of the corrected data, is interesting. He rightfully points out that strong statements about the predictive value of the type of data we have assessed will require larger prospective trials.
Dr von Kummer was appropriately concerned because our analysis discussed only cortical infarction without distinguishing the role of infarctions of the basal ganglia. From a reexamination of the films it is apparent that basal ganglion infarction was not a late occurrence but, when present, was there from the time of the initial CT study. Thus, the cortical mantle appears to have been appropriately defined as the territory that may continue to be at increased ischemic risk. Clearly the viability of the cortical MCA territory has an impact on clinical outcome.
The significant occurrence of “no cortical MCA occlusion, no cortical infarction and reversible levels of flow” well beyond 3 hours after occlusion we believe remains the major contribution of this article. Combining three CT-based data sets, not least of which is the absence of CT evidence of cortical infarction, provides a useful basis for focusing efforts to expand the therapeutic window to a far larger group than we are able to treat with current treatment guidelines.