CT angiography predicts use of tertiary interventional services in acute ischemic stroke patients

Regional systems of care have been established in some localities, where acute ischemic stroke patients are preferentially admitted to "stroke centers" [1, 2]. However, no formal guidelines exist for determining which patients should be transferred from a primary stroke center (PSC), capable of administering thrombolysis, to a comprehensive stroke center (CSC), with advanced services including endovascular capabilities. As a result, there can be tremendous heterogeneity in which patients remain at a PSC versus which are transferred to a CSC. Furthermore, many PSCs are likely capable of providing maximal management to stroke patients and may reserve transfer for those who need additional services available only at a CSC [3, 4].

Efficient resource allocation may best be achieved by reserving such transfers for patients who will receive the most benefit. A rapidly available tool that predicts which patients are interventional candidates would help emergency physicians determine who might benefit from transfer to a CSC.

One candidate for such a tool is CT angiography (CTA), which can reliably detect large occlusive thrombi in proximal cerebral arteries [5]. While only 25-35% of patients with acute ischemic stroke have such occlusions, they are disproportionately responsible for high hospital costs, morbidity, and mortality [6, 7]. As intravenous (IV) recombinant tissue plasminogen activator (rtPA) is less effective in recanalizing proximally occluded vessels [8], these individuals may preferentially benefit from advanced therapies at tertiary care centers. In particular, intra-arterial thrombolysis [9, 10], mechanical clot disruption [11, 12], and device-aided thrombus extraction [13‐15] have been shown to recanalize occluded vessels at a rate higher than for IV rtPA, which may lead to better outcome [16]. Since multislice CT scanners are available 24/7 in the majority of US emergency departments [17], it may be that this technology can be harnessed to select patients for transfer.

We hypothesized that the presence of an occlusive thrombus in a proximal cerebral artery on CTA is an accurate predictor of the use of advanced neurointerventional services. We elected to perform an observational study at a center in which virtually all patients undergo emergency CT angiography as a clinical standard of care, in order to minimize selection bias.

During the study period, 290 patients who presented within 24 h of symptom onset were admitted with acute stroke. Of these, seven were excluded for enrollment in the DIAS-2 clinical trial [19] since the intervention was blinded. Another 76 were excluded for not having a CTA performed (61 had MRA for cerebrovascular imaging and 15 had no vessel imaging because of contraindications to both studies), leaving 207 patients for final analysis. The median time to registration in the ED from the time last seen well was 3.9 h (IQR 2-5.8 h). Thirty-three percent of patients presented within 3 h of symptom onset, 75% within 6 h and 90% within 12 h. Of this cohort, 25% of patients received IV rtPA, 2.4% received IA thrombolysis, 6.8% received a mechanical intervention, 3.3% underwent surgery (4 decompressive hemicraniectomies and 3 carotid endarterectomies), and 52% were admitted to the neuroscience ICU.

Table 1 shows patient characteristics among those receiving an advanced neurointervention. Of note, there was no significant difference in rate of IV rtPA use between those who did and did not receive an intervention. Table 2 shows the comparison of patients with and without proximal occlusion. In multivariable logistic regression, proximal occlusion on CTA was an independent predictor of the use of neurointerventional services after controlling for initial NIHSS score (Table 3). Finally, test characteristics for the ability of a proximal cerebral arterial occlusion to predict the need for neurointervention were calculated (Table 4).

We found that proximal cerebral artery occlusion on CTA predicts the use of acute neurointervention. While time to presentation and neurological exam findings are often used in decision-making regarding transfers, this specific radiographic finding appears to add independent value in predicting tertiary care interventions. Use of CTA in selected patients may therefore improve our ability to stratify which patients would benefit from emergent transfer to a CSC.

Although only a quarter of patients with a proximal occlusion actually received a neurointervention, distinguishing those with a large occlusion may be important for two reasons. First, if an occlusion is not seen, it is highly unlikely that a patient will need an intervention. In fact in our study, only 3% received an intervention without a large occlusion on CTA. All of these were patients with critical internal carotid stenosis that received carotid endarterectomies that were not performed on the same day as admission but during that hospital stay. Thus, most of the patients without proximal occlusion could potentially receive appropriate care at PSCs depending upon resources available. On the other hand, if a proximal occlusion is seen on CTA, these patients should be considered for emergent transfer or at least discussed with a CSC via teleradiology or phone consultation to determine whether they are interventional candidates. Even if they are not, they might still benefit from care at a CSC because they will tend to have larger strokes, worse outcomes [6], and may have more complicated care needs.

The commonly used practice of relying on clinical findings and noncontrast head CT for management decisions may provide inadequate information for triaging stroke patients for advanced therapies. For example, large artery occlusive strokes may not respond well to IV rtPA, but show better response to IA therapies [20, 21]. In addition, vascular imaging provides independent information regarding the patient's prognosis [18]. As a result, current American Heart Association (AHA) guidelines endorse vascular imaging in the initial evaluation of the patient with acute ischemic stroke symptoms [22].

Our data confirm findings from others that patients with proximal occlusions tend to have higher NIHSS scores [23‐26]. This raises the question of whether the NIHSS score alone can select those patients requiring advanced intervention. We conclude, however, that CTA does add independent value. First, one recent prospective study found that NIHSS alone has a poor negative predictive value for proximal occlusion amenable to intervention [27]. Second, we found that CTA provides independent information even when controlling for NIHSS. In particular, NIHSS is known to be influenced by location because it is so heavily weighted toward language function, with posterior circulation occlusions leading to a lower initial NIHSS but a worse clinical outcome [28, 29].

The major limitation of our study design is that it is a single center retrospective cohort. We chose this design for our initial analysis because our center routinely performs CTA on almost all stroke patients, minimizing selection bias. However, patients presenting to an academic center with available tertiary care services may not reflect the full range of ischemic stroke patients that present to community hospitals. More than half of the patients that had proximal occlusions on CTA or received neurointervention were transferred from outside hospitals; this likely reflects a concentrating effect providing a population of more severe strokes than that which might present to any single community hospital. While this enriched our cohort for those that achieved the primary outcome, improving our statistical power, a multicenter study in a larger cohort will be necessary to validate these findings in a more representative population. There may be logistical, financial, and ethical considerations in consenting stroke patients for CTA in other practice settings where it is not routine, but our results appear to provide justification for such a larger, prospective study of the use of CTA to guide transfer decisions.

Another limitation was the exclusion of those who were unable to undergo CTA, most often due to IV contrast allergy and renal insufficiency. While many such patients would also be excluded from interventional neuroradiological procedures, it is possible that some would still have been candidates. Also, there is the possibility that CTA, performed at centers unaccustomed to acquiring it during acute stroke or at off hours, might perform an inadequate study that could delay treatment or transfer decisions, or preclude repetition of the study at the receiving facility.

Finally, the CTA findings were used in clinical decision making, potentially confounding our analysis. This likely overestimates the association of CTA proximal occlusion and neurointervention. Unfortunately, it would likely be unethical to "blind" clinical decision makers to CTA findings. In addition, our primary goal was to aid emergency physicians in predicting clinical options that would ultimately be offered to patients, and in a real world setting such decisions are expected to incorporate all available clinical and radiographic data.

Several factors must be considered prior to incorporating the use of emergency CTA in transfer decisions. AHA guidelines highlight that decision-making regarding IV thrombolytics should not be delayed for vascular imaging such as CTA, and protocols would need to be in place to ensure that treatment decisions for IV rtPA are made prior to initiation of further imaging [1, 22]. Options can include only performing this test after IV thrombolysis in eligible patients, or only for those in whom decision-making would be changed based on the results. A rapid CTA can take less than 10 min to acquire, and the source images are available immediately on the CT scanner workstation. These images can then be rapidly processed and examined to detect proximal artery occlusion, and further studies should validate the ability of plain radiography technicians to generate the images and general radiologists or emergency physicians to reliably diagnose these occlusions. Another concern is the use of IV contrast, which can carry the risk of allergic reaction or contrast-induced nephropathy (CIN). Although traditionally thought to occur in 2-3% of cases, the risk of nephropathy after stroke or hospitalization is similar even without contrast, and many cases of CIN may simply be due to the nephropathy associated with hospitalization [30‐36]. Finally, protocols should be in place to ensure that the study would not need to be repeated upon arrival to a tertiary care center, either due to an inadequate initial study or problems with image transfer between facilities. Prearranged transfer agreements, or even remote consultation via telephone or telemedicine [37], can ensure appropriate usage and communication.

In summary, the finding of a large proximal cerebral arterial occlusion on CTA predicts the use of neurointerventional services in patients with acute ischemic stroke. Thus, our results provide justification for conducting future prospective studies on using CTA as a rapid decision-making tool to select patients who may be candidates for endovascular therapies at CSCs.