Stroke Imaging Selection Modality and Endovascular Therapy Outcomes in the Early and Extended Time Windows

The strong benefit of endovascular treatment of large vessel occlusion strokes within the first 6 hours from symptoms onset has now been confirmed in 7 randomized clinical trials.^1–8 In addition, the DAWN (DWI or CTP Assessment With Clinical Mismatch in the Triage of Wake-Up and Late Presenting Strokes Undergoing Neurointervention With Trevo) and DEFUSE 3 (Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke) trials have demonstrated a robust benefit of mechanical thrombectomy (MT) in the 6- to 24-hour window.^9,10 However, the DAWN and DEFUSE 3 trials required the use of computed tomography perfusion (CTP) or brain magnetic resonance imaging (MRI) in all patients. Consequently, the current level-1a evidence for MT in the extended window supports the stringent usage of advanced imaging.¹¹ There is a paucity of prospective data comparing the outcomes of endovascular treatment in patients selected on the basis of noncontrast CT (NCCT)±CT angiography (CTA) alone versus those also undergoing CTP, especially in the extended window. We sought to evaluate the impact of CT imaging selection modality on the clinical outcomes of MT in a large prospective cohort of early and late presenting patients.

Anonymized data from the study are available upon reasonable request to the corresponding author.

A total of 1826 out of the 2008 Trevo Registry patients had intracranial internal carotid artery, M1, or M2 occlusions, with 1673 of them having a premorbid mRS score of 0 to 2. Out of these, 1603 patients underwent treatment within 24 hours from time-last-seen-well (1173 in the 0–6-hour and 430 in the 6–24-hour window) and constituted the primary study cohort (Figure 1). A total of 109 (6.8%) patients underwent imaging selection on the basis of a non-contrast CT alone while 290 (18.1%) got NCCT and CTA and 553 (34.5%) also underwent CTP. Brain MRI was performed in only 13 (0.9%) patients including MR angiography in 9 (0.6%) of them. A total of 244 (15.2%) patients were direct to the angiography suite transfers who had their last image performed beyond 3 hours from arterial puncture. In 394 (24.6%) patients, either core lab adjudication or the imaging acquisition times were missing (eg, undefined imaging modality).

In the early window, a total of 332 patients were selected with NCCT±CTA alone while 373 also underwent CTP. NCCT±CTA alone patients were slightly older (70.3±13.6 versus 68.5±14.5, P=0.093) and had shorter time-last-seen-well to puncture (median, 3.1 [interquartile range, 2.4–4.1] versus 3.4 [2.5–4.5] hours, P=0.044). There were no significant differences in the remaining baseline or procedural characteristics across the 2 groups (baseline NIHSS: 16 [12–20] versus 15 [11–20]; Alberta Stroke Program Early CT Score [ASPECTS]: 8 [7–9] versus 8 [7–9]; IV tPA (tissue-type plasminogen activator) use: 69.9% versus 67.6%). Similarly, no significant differences in the rates of good outcomes (55.9% versus 60.6%, P=0.202), mortality (13.0% versus 10.5%, P=0.302), successful reperfusion (92.8% versus 91.7%, P=0.593), or symptomatic intracranial hemorrhage (1.8% versus 2.4%, P=0.613) were detected across NCCT±CTA alone versus CTP patients (Table 1). After adjusting for identifiable confounders including age, baseline NIHSS, baseline mRS, ASPECTS, diabetes, time to treatment, there were no significant differences in terms of 90-day functional disability (ordinal mRS shift: adjusted odd ratio [aOR], 0.936 [95% CI, 0.709–1.238], P=0.644) or functional independence (mRS score of 0–2: aOR, 1.178 [95% CI, 0.833–1.666], P=0.355) across the CTP and NCCT±CTA groups, respectively (Table 2 and Figure 2).

In the extended window, a total of 67 patients were selected with NCCT±CTA alone while 180 also underwent CTP. NCC±CTA-selected patients were less often females (41.8% versus 58.9%, P=0.016) and had a trend towards lower rates of conscious sedation (38.8% versus 52.8%, P=0.084). There were no significant differences in the remaining baseline or procedural characteristics across the 2 groups (age, 65.8±15.1 versus 66.7±14.5; baseline NIHSS, 16 [9–20] versus 15 [10–19]; ASPECTS, 8 [7–9] versus 8 [7–9]; and time to treatment, 9.3 [6.8–14.5] versus 10.2 [7.5–14]). Similar to the patients in the early window, there were no significant differences in the rates of 90-day functional independence (60.6% versus 54.7%, P=0.412), 90-day mortality (9.0% versus 11.1%, P=0.624), successful reperfusion (95.5% versus 93.9%, P=0.622), or symptomatic intracranial hemorrhage (1.5% versus 0.6%, P=0.470; Table 1). After adjusting for the aforementioned co-variates, there were no significant differences in terms of 90-day functional disability (ordinal mRS shift: aOR, 0.983 [95% CI, 0.581–1.662], P=0.949) or functional independence (mRS score of 0–2: aOR, 0.640 [95% CI, 0.318–1.289], P=0.212) across the CTP and NCCT±CTA groups, respectively (Table 2 and Figure 2).

Notably, there was no interaction between the treatment time window (0–6 versus 6–24 hours) and CT selection modality (CTP versus NCCT±CTA) in terms of the degree of functional disability at 90 days (P=0.45). Subgroup analysis of the DAWN-like patients yielded similar results with no relationship between imaging selection modality and outcomes in either the early or extended time windows (Table I in the Data Supplement).

To our knowledge, no other study based on a large prospective cohort of MT patients has compared the effects of imaging selection modality within and across the early and extended time windows. In this analysis of patients with large vessel occlusion stroke with predominantly favorable imaging profile on presentation, we did not find a relationship between diagnostic CT imaging modality and outcome in either time epoch. While these represent preliminary and exploratory results, they put in question the current consensus guideline recommendations mandating the usage of advanced imaging for extended window thrombectomy patients and suggest that the currently adopted extended window selection criteria may be over stringent. As such, our findings have potentially broad implications to clinical practice as the absence of outcome differences across patients selected with CTP versus NCCT±CTA alone needs to be considered in the background of the additional contrast and radiation exposure, extra costs and potential treatment time delays that are typically associated with use of CTP. More importantly, as advanced imaging is not universally available, the use of non-contrast CT±CTA alone may represent a reasonable option for extended window selection, especially in environments that lack CTP and acute phase MRI capabilities.

DAWN and DEFUSE 3 completely relied on the use of advanced imaging (eg, CT/MR perfusion or diffusion-weighted imaging) for patient selection. Unfortunately, due to technological, logistical, and financial constraints, this has created a significant challenge to the implementation of extended window protocols in many centers across the globe. In addition, as the inclusion and exclusion criteria used in DAWN and DEFUSE-3 were, at least on retrospect, too strict and given their extremely high treatment effect sizes, more inclusive selection paradigms could potentially result in the treatment of a larger proportion of the late presenting patients while still maintaining a significant benefit. As such, the validation of simpler and more inclusive imaging selection, ideally relying solely on NCCT and CTA, has become critical to a more widespread adoption of MT in the extended time window. Concurrently, the utility of perfusion imaging for MT has become increasingly controversial. In the SWIFT PRIME trial (Solitaire With the Intention for Thrombectomy as Primary Endovascular Treatment), the use of perfusion imaging did not improve the efficacy of MT but was associated with potential time delays.¹⁵ Likewise, there was no significant interaction between CTP-derived parameters or CTP mismatch and treatment effect in the MR CLEAN trial (Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands).¹⁶ A recent analysis of the HERMES cohort (Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke Trials) demonstrated that (1) CTP mismatch volume was not independently associated with functional outcomes and (2) as compared with diffusion-weighted imaging, CTP was independently associated with lower functional independence and less functional improvement.¹⁷ As early window outcomes are highly time-dependent and ≈85% of the patients presenting within 0 to 6 hours have favorable imaging on NCCT (ASPECTS ≥6),^18,19 the trade-offs between treatment speed versus the additional data provided by CTP must be considered. Furthermore, patients may be mistakenly excluded from treatment due to CTP infarct volume over-estimation.²⁰ Our study showed similar functional outcomes with the use of NCCT±CTA versus CTP in both the early and extended windows. This remained the case after the adjustment for identifiable confounders. Although counter-intuitive at first glance, the comparable performance of NCCT±CTA versus CTP in the extended window is somewhat expected since the performance of NCCT for the detection of ischemia improves over time.²¹ In this setting, imaging selection paradigms based on NCCT Clinical-ASPECTS Mismatch may represent a good option to centers that lack advanced imaging capabilities.²²

Our study has all the limitations associated with a post hoc retrospective registry analysis including a greater susceptibility to selection bias. This is exacerbated by the fact that, despite its encouragement, consecutive enrollment was not mandated or monitored for. Moreover, given the small sample size of some of the study subgroups, our analysis lacked enough statistical power to properly exclude the possibility of a benefit from CTP and this may have led to a type II error. As our results relate specifically to the defined subpopulation of patients with intracranial internal carotid artery or middle cerebral artery M1- or M2-segment occlusions they cannot be generalized to other populations. Finally, there was no standardization for the imaging protocols across the different sites and many centers did not employ the more sophisticated automated CTP software, which are known to reduce the times for imaging processing/interpretation and to eliminate inter-operator variability. As such, our results should be interpreted with caution and as hypothesis-generating only. Indeed, only well-conducted randomized controlled trials will be able to accurately evaluate the need of CTP for patient selection in endovascular therapy. Fortunately, the MR CLEAN LATE (Endovascular treatment of acute ischemic stroke in the Netherlands for late arrivals; ISRCTN19922220) and the RESILIENT-Extended (Randomization of Endovascular Treatment in Acute Ischemic Stroke in the Extended Time Window; URL: https://www.clinicaltrials.gov; Unique identifier: NCT04256096) trials are currently investigating simpler paradigms for imaging selection in the 6- to 24-hour time window.

In conclusion, our data add to the increasing body of evidence challenging the utility of CTP in the early window and suggest that extended window patients may be safely treated in the absence of CTP or MRI data. However, as our study was based on a post hoc analysis of a registry, these findings must be properly confirmed in prospective trials.

We would like to acknowledge all the Trevo Registry and DAWN Trial investigators. Dr Nogueira was responsible for the study design, data interpretation, and writing of the article. Drs Nogueira, Gupta, Liebeskind, Budzik, B. Baxter, Krajina, B. Bartolini, English, and Veznedaroglu contributed to the Trevo Registry oversight. Drs Nogueira, Jovin, Liebeskind, BB, Gupta, and Bonafe contributed to the DAWN trial oversight. Drs Nogueira, Haussen, Jovin, Gupta, Jadhav, Budzik, Baxter, Krajina, Malek, Narata, Bartolini, English, and Veznedaroglu contributed to the Trevo Registry participation and enrollment of patients. Drs Nogueira, Haussen, Jovin, Jadhav, Budzik, BB, Bonafe, English, Frankel, and Veznedaroglu contributed to the DAWN trial participation and enrollment of patients. Dr Nogueira, Dr Zhang, and R. Shields contributed to the article data analysis, figures, and tables. P. Morgan, R. Shields, and Dr Zhang contributed to the Trevo Registry and DAWN trial data collection and study management. Dr Liebeskind contributed to the Trevo Registry and DAWN trial imaging interpretation. All authors contributed to critical appraisal of important intellectual content and approved the final version of the article. The first drafts of the article were written by the first author, with input from all the authors and with no external writing assistance.