The Acute STroke Registry and Analysis of Lausanne (ASTRAL) collects all AIS patients admitted to the stroke unit and/or intensive care unit of the Lausanne University Hospital (Centre Hospitalier Universitaire Vaudois [CHUV]) since 2003, as published previously [17]. We conducted the current study on the ASTRAL population, excluding from the analysis all secondary transfers from other hospitals, patients who had undergone brain imaging prior to arrival at our hospital, and in-hospital strokes.

We defined two time-based groups to identify the two paradigms: the MRI-paradigm including all consecutive patients satisfying the inclusion criteria admitted between May 1, 2018 (date when the MRI machine in the ED became operational) and August 31, 2022, and the CT-paradigm comparison group including an equal number of consecutive patients admitted between December 2012 and April 2018. As we aimed to conduct an intention-to-imaging study, the patients who had CT as firstimaging between May 2018 and August 2022 were kept in the MRI-paradigm group for the analysis.

Stroke physicians and study nurses routinely collect data and parameters in a prespecified manner in ASTRAL, both in the acute hospital stay and after discharge. The variables in ASTRAL include demographics; patients’ and stroke characteristics; acute and subacute brain and cervico-cerebral arterial imaging; revascularization treatments, i.e., IVT and EVT; and clinical outcome at 3 months (modified Rankin Scale [mRS]).

In our setting, every patient with suspected acute stroke admitted to the ED is evaluated by an ED physician and the on-call neurologist. The neurological evaluation consists of the National Institutes of Health Stroke Scale (NIHSS) assessment, and neuroimaging is done as soon as possible, usually within 24 hours, regardless of treatment eligibility in all suspected acute strokes with last known well (LKW) less than 24 hours. A stroke code is activated by the ED physician and the on-call neurologist based on a call with the prehospital teams for patients identified as eligible for an IVT or EVT, allowing for an immediate and simultaneous evaluation by a dedicated team on arrival at CHUV.

Both CT and MR scanners are located within the ED, in close proximity to the dedicated room where patients arriving under stroke code are initially received and evaluated. The transport distance from the area of the initial clinical evaluation to either scanner is less than 50 m, and patient transfer typically requires no more than 1 to 2 minutes.

At least one arterial study of extra- and intracranial cerebral arteries is performed on admission, i.e., CT angiography (CTA) and MR angiography (MRA) (time-of-flight [TOF] and contrast MRA).

For multimodal CT-based imaging, we used a 64-multidetector CT scanner (LightSpeed VCT, GE Healthcare, Milwaukee, WI, USA) between December 2012 and November 2015, and a 256-multidetector CT scanner (Revolution CT, GE Healthcare, Milwaukee, WI, USA) thereafter. The institutional standard of care for patients with suspected AIS includes CT imaging protocol including non-contrast CT (NCCT), CT-perfusion (CTP), CTA, and postcontrast series (delayed venous-phase axial CT images acquired following the CTA) [18]. The acquisition of this CT-based protocol including CTP usually takes 8 minutes. For MRI imaging, we use a 3-T scanner (Vida, Siemens, Erlangen, Germany) with the following sequences: sagittal T1 gradient echo, axial diffusion, axial 2D fluid attenuated inversion recovery (FLAIR), axial T2 gradient echo, TOF angiography, gadolinium-enhanced cervical MRA, post-contrast axial T1 gradient echo, and perfusion weighted imaging (PWI), as previously described [19]. Apparent diffusion coefficient (ADC) maps are generated from diffusion-weighted imaging (DWI) sequence. The duration of this MRI protocol is approximately 19 minutes.

The institutional goals for patients eligible for IVT and/or EVT are as follows: door-to-imaging delay, 15 minutes; DTN delay for IVT, 30 minutes; and DTP delay for EVT, 60 minutes. IVT is recommended for disabling stroke (defined as NIHSS ≥4, or NIHSS <4 but with a deficit deemed disabling by the treating physician) within 4.5 hours without an age limit in patients with a pre-stroke mRS score ≤3 if no clinical, radiological, or laboratory contraindications were identified. To identify later arriving and wake-up patients candidates for IVT, a significant penumbra on perfusion imaging was used since 2008, or a mismatch between FLAIR and DWI on MRI since 2018 [20,21]. EVT was offered in supratentorial AIS to patients up to 6 hours with known onset if NIHSS was ≥6, proximal vessel occlusion and CTP showing >50% of penumbra until 2014. Afterwards, this latter imaging criteria was replaced by Alberta Stroke Program Early CT Score (ASPECTS) ≥5 and the lower NIHSS limit was replaced by the presence of a potentially disabling deficit [22]. Late treatment after 6 hours was offered starting in 2015 to patients with an ASPECTS ≥7 and a CTP mismatch ratio ([penumbra+core]/core) >2.0. From May 2017, patients were treated up to 8 hours without perfusion imaging, and late treatment was offered following DAWN criteria and for an NIHSS of 1-10 and ASPECTS ≥8 [23]. From January 2018, late treatment alternatively followed Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke 3 (DEFUSE-3) criteria [24] and was mainly DWI/PWI based. ASPECTS scores on MRI were assessed as previously described by identifying areas of acute ischemia, defined as hyperintense signal on axial DWI, and applying the ASPECTS methodology using the same topographical approach as with CT: starting from a score of 10, one point was subtracted for each region showing signs of acute ischemia [25]. Basilar artery occlusions were treated with EVT up to 6 hours until April 2017 in the absence of extensive brainstem infarct on radiology. Since May 2017, treatment window was extended to 8 hours if posterior circulation ASPECTS (pc-ASPECTS) was ≥7, and until 24 hours if no transverse irreversible brainstem ischemia was present on MRI, or if pc-ASPECTS was ≥8 [20]. An expert neuroradiologist is available 24/7 for imaging interpretation.

To evaluate the consequences of an MRI-based paradigm on patients’ selection and rates of acute revascularization treatment, we assessed the rates of early and late IVT and EVT treatments, both overall and specifically for posterior circulation strokes, as well as the rate of potentially missed IVT opportunities. The rates of early and late IVT treatment were calculated as the proportion of patients treated with IVT among all potential candidates, defined as patients with baseline NIHSS ≥4 arriving within 4 hours from symptoms onset or LKW for early IVT, and after 4 and 23 hours for late IVT. The 4-hour limit for LKW-to-arrival was chosen based on the conventional treatment window of 4.5 hours from symptom onset to treatment, accounting for the 30-minute in-hospital delay required to initiate treatment, which corresponds to the benchmark DTN time established in performance goals for timely care. The rates of early and late EVT treatment were calculated as the proportion of patients treated with EVT among all potential candidates, defined as patients with baseline NIHSS ≥4 and large vessel occlusion (LVO) arriving within 8 hours from symptoms onset or LKW (early EVT) and between 8 and 23 hours (late EVT). LVO was defined as an occlusion of the first segment of the anterior, middle, or posterior cerebral artery, the M2 segment of the middle cerebral artery, or the basilar artery. The rate of any revascularization treatment for posterior circulation strokes was calculated as the proportion of patients treated with IVT and/or EVT among all those with isolated posterior circulation stroke and an initial NIHSS ≥4, arriving within 23 hours from symptom onset or LKW. Finally, the rate of potentially missed IVT opportunities was defined as the proportion of acute stroke cases that met all institutional eligibility criteria for IVT (as detailed above) but were nonetheless not treated with IVT, over the total number of IVT performed in the same period.

To assess the time metrics and acute stroke workflow in the two paradigms, we evaluated: the median DTN time, defined as the delay in minutes from hospital arrival to start of IVT; the median DTP time, defined as the delay in minutes from hospital arrival to inguinal puncture; and the rate of successful recanalization in EVT, defined as expanded Thrombolysis in Cerebral Infarction (eTICI) score ≥2b67 at the end of the procedure [26].

To assess the consequences of an MRI-paradigm on safety of both IVT and EVT, we measured the rates of symptomatic intracranial hemorrhage (SICH) within 36 hours, according to the European Co-operative Acute Stroke Study (ECASS-2) definition [27], including both parenchymal SICH and subarachnoid hemorrhage (SAH), in patients treated with IVT alone and in those treated with EVT (±IVT). We also assessed the mortality at 7 days in both treatment groups.

As clinical outcomes, we assessed the 3-month disability in patients treated with IVT alone and with EVT (±IVT) and the rate of futile recanalization in patients treated with EVT (±IVT), defined as poor 3-month outcome (mRS 4-6) despite successful intracranial recanalization (eTICI ≥2b67) at the end of EVT. This analysis was performed for the overall EVT cohort, for early EVT (onset-to-inguinal puncture time ≤8 h) and late EVT (>8 h) separately.

Continuous and ordinal variables were expressed as medians (with interquartile range, IQR), and categorical variables as absolute counts (with percentage), unless stated otherwise.

Initially, we compared each outcome between MRI-paradigm and CT-paradigm groups, using unadjusted regression analyses to calculate crude odds ratio (OR) for binary outcomes and beta coefficients (β-coefficient) for continuous variables, along with 95% confidence intervals (CIs).

Next, we performed multivariable analyses for each above-mentioned outcome. Each time, we first assessed the presence and the shape of temporal trends throughout the study period. To do this we used generalized additive models to fit smooth curves for each outcome with a cubic smoothing spline term for the time variable. We chose generalized additive model given their flexibility in fitting nonlinear patterns in the data, allowing us to identify trends over time that may not have been captured using traditional linear models. Depending on whether the outcome was continuous, binary, or multicategory, we respectively used additive quantile regression model, generalized additive models with binomial distribution, and generalized additive model with ordered categorical outcome. The time variable was obtained by dividing the study period into ten consecutive periods with an equal number of consecutive patients in each period. Results of time trends were expressed as P-values.

Then we added to the models for each outcome the imaging paradigm variable together with potential confounders based on clinical plausibility and the time trend variable (when significant in the previous analysis) to obtain adjusted OR (or adjusted β-coefficients for continuous outcome variables) and 95% CIs. The potential confounders entered as model covariates for each outcome analysis are described as follows. The models for the rates of early and late IVT and EVT were adjusted for age, sex, baseline NIHSS, onset known, onset-to-door delay, and pre-stroke mRS. The same covariates were entered in the models for the rate of IVT and EVT in posterior circulation strokes, the rate of potentially missed IVT, DTN, and DTP. The DTP model also included as covariate IVT prior to EVT. The models for the rates of successful and futile recanalization in EVT were adjusted for the same factors as above, with the addition of stroke mechanism and site of intracranial occlusion. The rate of SICH in the IVT cohort was adjusted for age, baseline NIHSS, sex, baseline blood glucose, onset-to-IVT delay (min), mechanical thrombectomy (yes/no), any antiplatelet, pre-stroke anticoagulants, and baseline ASPECTS. The same factors for adjustment were used for the rate of SICH in the EVT cohort, plus DTP delay (min) and thrombectomy device passes. The disability at 3-month (mRS) and mortality at 7 days in both IVT and EVT cohorts were adjusted for age, baseline NIHSS, sex, pre-stroke mRS, onset-to-IVT delay (min), mechanical thrombectomy, baseline ASPECTS, and type of acute treatment received, plus active cancer for mortality analyses.

The statistical analysis was performed with R version 3.4.2 software or above (R Foundation for Statistical Computing, Vienna, Austria).

ASTRAL follows institutional regulations for clinical and research databases. Data were collected from routine clinical and radiological management and anonymized before analysis. Because only anonymized data were used, neither ethics committee approval nor patient consent was required according to the Swiss Human Research Act (HRA) and applicable data protection legislation. The status of patient consent also did not need to be considered, as this was primarily a quality assurance project of treatment practice in our institutions, which is not covered by the HRA.

Through the study period, of 1,843 potential candidates for IVT (NIHSS ≥4), 52% (n=954) received the treatment, with rates of 75.9% (n=842/1,110) in the early time window and 15.3% (n=112/733) in the late time window (Table 1). Time trend analysis showed a decrease in early IVT rates following the introduction of MRI, while the rate of late IVT progressively increased already during the CT period and reached a plateau during the MRI period (Figure 1A and B). Although this resulted in a reduction of early IVT rate from 78.1% in the CT period to 72.6% in the MRI period, and an increase of late IVT rate from 9.0% to 20.9%, the multivariable regression analysis adjusted for time trend and other potential confounders showed no significant association between the imaging paradigm and early or late IVT rates (Table 1).

Of 829 potential candidates for EVT (NIHSS ≥4, presence of LVO), 64% (n=537) received the treatment. The rates of early and late EVT among potential candidates progressively increased throughout the study period, reaching a plateau in later years (Figure 1C and D). This trend remained unaffected by the imaging paradigm shift, as shown by both visual assessment of the time trend curves and multivariable regression analyses (Table 1). Among patients with isolated posterior circulation strokes, the rate of any IVT and/or EVT within 24 hours remained stable, with no significant difference between the CT and MRI paradigms (52.9% in CT-paradigm vs. 61.7% in MRI-paradigm, OR_adj 1.97, 95% CI 0.66-5.88, P_adj=0.23) (Table 1 and Figure 2A).

The rate of potentially missed IVT opportunities (defined as the proportion of acute strokes not treated by IVT despite having the criteria, over the total number of IVT performed in the same period) was higher during the CT-paradigm (3.1%) compared to the MRI-paradigm (0.8%), a difference that remained significant after multivariable adjustment for confounders (OR_adj 0.22, 95% CI 0.09-0.57, P_adj<0.01) (Table 1 and Figure 2B).

The median DTN time for IVT-treated patients increased from 31 minutes (IQR=24-48) during the CT-paradigm to 43 minutes (IQR=33-58) during the MRI-paradigm, a difference that remained significant after multivariable adjustment (β coefficient 15.63 (95% CI 11.27-20.0, P_adj<0.01) (Table 1 and Figure 2C). The median DTP time for EVT-treated patients decreased in the initial study years and then remained stable (Figure 2D), with no significant difference between the CT (101 min, IQR= 90- 121) and MRI paradigms (104 min, IQR=85-129) after multivariable adjustment (β coefficient 9.95, 95% CI -2.24-22.14, P_adj=0.11) (Table 1).

The rate of successful recanalization (eTICI ≥2b67) in EVT (±IVT) patients was similar within the two imaging paradigms (81.4% in CT vs. 88.3% in MRI, OR_adj 0.74, 95% CI 0.28-1.91, P_adj=0.53) (Table 1 and Supplementary Figure 2A).

The rate of SICH after IVT was characterized by slight linear reduction over time (Figure 3A), resulting in a lower rate in the MRI-paradigm (5.6% vs. 3.2%); however, this difference was not significant in the multivariable analysis (OR_adj 1.00, 95% CI 0.29-3.40, P_adj=0.99). The rate of SICH after EVT (±IVT) was similar across both paradigms (6.5% vs. 4.2%, OR_adj 0.62, 95% CI 0.14-2.68, P_adj=0.52) (Table 1 and Figure 3B).

Mortality rate at 7 days among IVT-only patients was comparable between paradigms (5.6% vs. 6.1%, OR_adj 0.51, 95% CI 0.16-1.56, P_adj=0.23) (Table 1 and Supplementary Figure 2B). Mortality at 7 days after EVT showed a significant constant linear increase over the course of the study (Supplementary Figure 2C), resulting in a higher rate in the MRI paradigm (5.0% vs. 8.3%). However, this difference appeared to reflect the underlying temporal trend rather than an effect of the imaging paradigm change, which was not significant in the multivariable analysis (OR_adj 0.96, 95% CI 0.24-3.78, P_adj=0.95) (Table 1).

Disability in IVT patients decreased over time (Figure 3C), though no clear association with imaging paradigm was found in multivariable analysis (OR_adj for favourable mRS shift, MRI vs. CT=1.26, 95% CI=0.77-2.06, P_adj=0.36) (Table 1). Among EVT-treated patients, disability remained stable over time and across imaging paradigm (OR_adj=1.22, 95% CI=0.66-2.26, P_adj=0.52) (Table 1 and Figure 3D).

The rate of futile recanalization (defined as 3-month mRS=4-6 despite eTICI ≥2b67) in EVT patients was similar between the two paradigms (22.8% vs. 29.2%, OR_adj 1.03; 95% CI 0.41-2.56, P=0.95), with similar findings when analyzing early and late time windows separately (Table 1 and Supplementary Figure 2D).

Among patients in the MRI-paradigm, those who underwent CT due to MRI non-feasibility were older, had more vascular risk factors, more severe strokes (higher NIHSS), more frequent vigilance alterations at baseline, and arrived earlier at the hospital (onset-to-door time: median 173 min, IQR 73-605 vs. 219 min, IQR 83-640; P<0.001) (Supplementary Table 2). The proportion of IVT administration was similar between groups, while EVT was more frequently performed in patients who did not undergo MRI—likely reflecting their more severe clinical profile. Despite similar median values, these patients exhibited a distribution shifted toward shorter treatment delays. Specifically, DTN for IVT was 43 minutes (IQR 29.7-60.7) vs. 43 minutes (IQR 34.0-57.3) in patients who underwent MRI (P<0.001), and DTP for EVT was 96 minutes (IQR 80.0-116.2) vs. 102 minutes (IQR 91.0-122.1), respectively (P<0.001).

The study period included two COVID-19 waves, both occurring during the MRI period and significantly impacting hospital care and workflow. The first wave lasted from February 25 to April 20, 2020, and the second from October 7 to November 30, 2020. During these periods, we observed a significant decline in weekly admissions for AIS, from 6.5 to 5.2 patients per week (t=2.295, P=0.03). The 94 patients admitted during the two COVID-19 waves, beside younger age distribution, had baseline characteristics comparable to those admitted outside these periods, and rates of IVT and EVT administration remained unchanged (Supplementary Table 3). Patients admitted during the COVID-19 waves presented later to the hospital (median onset-to-door time: 338 [118.8-793.1] min vs. 210 [82-642.5] min), but in-hospital workflow metrics were unaffected, except for a slightly shorter DTN time for IVT during the COVID period (median 41 [31.7-51.7] min vs. 43 [33-59.3] min, P<0.001).