In our study we document the feasibility of conducting CTP as part of routine clinical practice. All patients with suspicion of acute stroke underwent CTP. Twenty-five percent of these patients received intravenous thrombolysis. We demonstrated that our median door-to-needle time for intravenous thrombolysis was only 20 minutes. Such finding has important clinical implications because some previous studies documented that CTP is prolonging initiation of IVT and thus should not be used before initiation of IVT (9). Our study documents, that despite adding few minutes for CTP imaging (usually 4–5 minutes), which is typically done in our center together with unenhanced CT and CTA, CTP in principle did not pose a barrier for achieving very short DNT. Main reason is that patients are directly transported to CT scan from ambulance and IVT is initiated in CT scanner as we publish before [7, 12]. Moreover, our study documented that CTP had important added value in decision making as demonstrated in the Results and discussed below.
Our first major finding is that once CTP is positive, regardless of time window, all patients had a stroke and none had stroke mimics. Such information is very useful, especially because it eliminates any diagnostic doubts that physicians might have in general and especially in the short amount of time needed to decide to initiate intravenous thrombolysis. Most likely, the value of stroke confirmation will be increased even more in those hospitals that lack the expertise in stroke. We could not confirm the results of some previous studies documenting that hypoperfusion (and even hyperperfusion) on CTP imaging can also be found in patients with other diagnosis then stroke such as status epilepticus or post-paroxysmal deficit (10, 11, 12, 13). Twenty-three patients in our cohort had epileptic seizures and all were correctly diagnosed by RAPID as stroke mimic. Different results can be explained by the fact that in previous studies no automatic software was used, no thresholds were defined, and numbers of patients were low. Our results are therefore applicable for clinical practice as long as certified postprocessing tools such as RAPID are used for evaluation of CTP.
CTP is used to define penumbra and accordingly our results documented that core or penumbra presence at baseline predicted infarction on follow up imaging. False negative results were obtained mainly in patients with lacunar supratentorial infarction (24 patients) or vertebrobasilar stroke (66 patients). Also, 10 infarcts were in the area of the brain which was outside of coverage by baseline CTP. These are known limitations of brain perfusion imaging (14, 15). Calculated specificity against radiological diagnosis of stroke is comparable with largest published meta-analysis (16). We found numerically just a bit lower sensitivity (74%) against radiological diagnosis as reported in previous largest and the most recent analyses (80%) (16). If CTP covers whole brain instead of 8cm width, as in our study, accuracy of CTP will slightly improve due to capture of infarctions in any part of the brain.
In patients with negative CTP, many patients (57%) suffered a stroke, which is not surprising because lacunar and vertebrobasilar strokes cannot be diagnosed by CTP as discussed above. However, 43% of those with negative CTP, which is quite a substation number, had stroke mimics and were discharged with another diagnosis such as vertigo of non-vascular cause (23%), epilepsy (18%), orthostatic collapse or others as described in Table 2. Patients with negative CTP with stroke mimics were younger (68 vs. 73, p < 0.001) and with lower NIHSS (0 vs. 4, p < 0.001) compared to patients with negative CTP and a stroke. Stroke mimic patients also had fewer cardiovascular risk factors, most likely due to younger age. The overall prevalence of stroke mimics in our study was 27% which is nearly the same as in metanalysis of 29 studies documenting 26% of stroke mimics (9, 17). Also, stroke mimics diagnoses were very similar to published data as shown in Table 2. Therefore, our study documented similar number and causes of stroke mimics and found that if CTP is negative, it should raise a red flag in younger patients with minor neurological deficit and fewer cardiovascular risk factors because they might have stroke mimics. In cases presenting with lacunar syndrome or suspicion for stroke in the vertebrobasilar territory, negative CTP should not preclude treatment with intravenous thrombolysis.
Limitations of our study are retrospective design and missing follow-up imaging in 37% of cases. Such missing cases could limit conclusion about CTP accuracy against radiological but not clinical diagnoses of stroke. In our clinical practice, follow up CT is omitted only when diagnoses of stroke or stroke mimic is established. In clinically uncertain cases, we always perform MRI. Therefore, missing follow up imaging should not limit validity of our study although confirmation of our results by future studies will be re-assuring.
For generalizability of our results, especially in terms of high proportion of stroke mimics, it might be important to consider referral pattern that is applied in our hospital as described in the Methods. Hospitals with different referral pattern might have less stroke mimics, especially if some triage criteria are applied before imaging and will refer to CT only the most obvious stroke candidates. Our results related to stroke mimics (negative predictive value) are thus generalizable to unselected patients’ population with acute neurological deficit. Our results related to 100% positive predictive value of stroke should be, however, applicable to any hospital regardless of referral pattern.
The major strengths of our study are the consistency of the use of multimodal imaging not limited to any patients' subgroups. This is the advantage over previous studies that included only patients with stroke in anterior circulation (18), patients in a short time window (19), or patients with MRI follow-up imaging only (20). Other advantage of our study is a relatively large number of patients, which was much less (usually up to 120 cases) in previous studies (18, 21, 22, 23).