Based on local regulations, neither patient informed consent nor approval by the Institutional Ethics Committee was required given the retrospective design of this study. From December 2012 to December 2016, we collected clinical and imaging data from three different comprehensive stroke centers taking part in the ETIS registry [12]. Prospectively gathered data was then retrospectively analysed.

Patients included in our study presented with an AIS within six hours of stroke onset, and imaging evidence of anterior circulation occlusion. Age, baseline modified Rankin Scale (mRS) score, low or high National Institutes of Health Stroke Scale (NIHSS) and infarct core volume were not considered as exclusion criteria per se. Middle cerebral artery (MCA) occlusions were divided into proximal (M1) and distal (M2). Carotid occlusions were divided into terminal intracranial (carotid T), and tandem (extracranial internal carotid artery occlusion associated with an intracranial artery occlusion). A stroke neurologist decided whether or not to perform IVT, based on conventional inclusion and exclusion criteria for IVT, according to United States and European Union guidelines [4,5].

Contraindications for IVT were: late treatment (between 4.5 and 6 hours after onset) and treatment in patients with a possible high bleeding risk after thrombolytic therapy, including elevated increased international normalized ratio (1.7 to 3.0), thrombocyte count below 90×10⁹/L, history of ICH, severe head injury in the preceding 4 weeks, previous AIS in the preceding 6 weeks and major surgery, gastrointestinal bleeding or urinary tract bleeding within the previous 2 weeks.

All endovascular procedures were performed under a dedicated neuroanesthesic protocol that encompassed conscious sedation or general anesthesia. All procedures were performed using either an aspiration catheter, a stent retriever or both, at the operator’s discretion.

The primary study outcome was the percentage of patients who achieved a favorable 90-day outcome, defined as a mRS score of 0 to 2 or equal to pre-stroke Rankin score. Secondary outcomes included clinical outcomes (excellent 90-day outcome defined as a mRS score of 0 to 1 or equal to pre-stroke Rankin score, 90-day all-cause mortality, any hemorrhagic complications and symptomatic intracerebral hemorrhage [sICH] defined as a hemorrhage on the follow-up computed tomography/magnetic resonance imaging scan associated with an increase of 4 points in NIHSS score) and procedural outcomes (successful reperfusion defined by modified Thrombolysis In Cerebral Infarction [mTICI] 2b/3, excellent reperfusion defined by mTICI 3, rescue therapy, more than two passes, time from groin puncture to successful reperfusion, emboli to uninvolved territory, vessel perforation and vasospasm).

Continuous variables are expressed as mean±standard deviation (SD) in the case of normal distribution or medians (interquartile range) otherwise. Categorical variables are expressed as numbers (percentage). Normality of distributions was assessed using histograms and the Shapiro-Wilk test.

We assessed the effect of the therapeutic approach on clinical outcomes using logistic regression models and calculated the odds ratio (OR) for the MT+IVT relative to the MT group as the treatment effect size. In order to reduce the effects of potential confounding factors in the between-group comparisons, we used propensity-score methods [13-16]. As the main analysis, propensity score was used to assemble well-balanced groups (propensity score-matched cohort) and a generalized linear mixed model was used to take into account the matched design. As secondary analyses, the effect of the therapeutic approach were estimated using propensity score adjustment method (propensity score considered as a covariate in a multivariable logistic regression model) and inverse probability of treatment weighting (IPTW) propensity score method (using stabilized inverse propensity score as weight in simple logistic regression model) were performed. The propensity score was estimated using a non-parsimonious multivariable logistic regression model, with the treatment group as the dependent variable and all of the characteristics listed in Table 1 as covariates. Patients from the MT+IVT group were matched 1:1 to patients in the MT group according to propensity score using the greedy nearest neighbor matching algorithm with a caliper width of 0.2 SD of logit of propensity score [14,16]. To evaluate bias reduction using the propensity score methods, absolute standardized differences (ASDs) were calculated [16]. Because of missing baseline data (Supplementary Table1), we estimated the treatment effect size in propensity score-matched and -adjusted cohorts after handling missing covariate values by multiple imputation using a regression switching approach (chained equations with m=10). Imputation procedure was performed under the missing at random assumption [17] using all variables listed in Table 1 (including treatment group) with a predictive mean matching method for continuous variables and multinomial or binary logistic regression model for categorical variables. In each imputed dataset, we calculated the propensity score and assembled a matched cohort to provide both adjusted and matched ORs. We therefore combined the ORs from each imputed dataset using Rubin’s rules [18].

Finally, we investigated the heterogeneity in treatment effect size for primary outcome across the following subgroups: age (<80 years vs. ≥80 years), admission NIHSS (<10, 10 to 17, ≥18), Alberta Stroke Program Early CT Score (ASPECTS: <5, 5 to 6, ≥7), site of occlusion (carotid terminus vs. tandem vs. M1-MCA vs. M2-MCA), admission mode (mothership vs. drip and ship) and antiplatelet use (yes vs. no). Heterogeneity across subgroups was quantified by introducing a multiplicative term into logistic regression models. Statistical testing was conducted at the two-tailed α-level of 0.05. Data were analyzed using the SAS software version 9.4 (SAS Institute, Cary, NC, USA).

A total number of 1,507 patients with an arterial occlusion in the anterior circulation were treated by MT at three comprehensive stroke centers. Of these, 975 (64.7%) who received prior IVT and 407 matched pairs (mean value of the 10 imputed datasets) could be included (Supplementary Figure 1). Table 1 shows the baseline characteristics after handling missing values by multiple imputation according to the two study groups before and after propensity score-matching (Supplementary Table 1 for baseline characteristics before matching and handling missing values). Before matching, several meaningful differences (ASD >10%) were found; as expected, the stronger difference was observed for anticoagulation therapy at stroke onset (34.1% in MT vs. 7.5% MT+IVT) which was a contraindication to use IVT. These differences were reduced after propensity score-methods with a maximum ASD of 10% for first-line thrombectomy strategy in matched-propensity score (Table 1, Supplementary Table 2, and Supplementary Figure 2).

Favorable outcome at 90 days (i.e., primary outcome) was achieved in 49.4% (n=745) of patients overall, with a significant difference between the two groups in favor of the IVT+MT approach, with an unadjusted OR of 1.61 (95% confidence interval [CI], 1.29 to 2.01). However, this difference remained not significant in matched-propensity score cohort (OR, 1.21; 95% CI, 0.90 to 1.63) although it remained when propensity score adjustment (OR,1.31; 95% CI, 1.02 to 1.68) (Figure 1) or IPTW (OR, 1.37; 95% CI, 1.09 to 1.73) (Supplementary Table 3) method were used. Similar to favorable outcome, excellent outcome was achieved more frequently in MT+IVT than in MT group (Figure 1). This difference remained significant after propensity-score matching (OR, 1.45; 95% CI, 1.05 to 2.01), propensity score adjustement (OR, 1.63; 95% CI, 1.25 to 2.11), and IPTW (OR, 1.71; 95% CI, 1.33 to 2.21) analyses.

Regarding safety outcomes, all-cause mortality within 90 days after symptom onset occurred less often in IVT+MT patients (17.6%) than in MT patients (24.9%; unadjusted OR, 0.64; 95% CI, 0.49 to 0.84). However, this difference did not remain significant regardless of the propensity score method used (Figure 1 and Supplementary Table 3). There was no difference between the two study groups regarding any ICH or sICH before and after using propensity score methods.

Successful reperfusion was achieved more frequently in IVT+MT (79.1%) than in MT patients (69.1%; unadjusted OR, 1.69; 95% CI, 1.33 to 2.15). This difference in favor of the IVT+MT approach remained significant in propensity-score matched, adjusted and IPTW analyses (Figure 1 and Supplementary Table 3). However no such difference was observed with excellent reperfusion. Regarding other efficacy procedural outcomes, cases of emboli to uninvolved territory and of more than two thrombectomy device passes were less frequently observed in IVT+MT than in MT patients with an unadjusted OR of 0.66 (95% CI, 0.44 to 0.99) and 0.74 (95% CI 0.60 to 0.93), respectively. Conversely, successful reperfusion was more often achieved within 45 minutes after groin puncture in the IVT+MT group than after MT alone, with an unadjusted OR of 1.32 (95% CI, 1.06 to 1.65). These differences did not remain significant in the propensity-score matched-cohort but they are close to significance for the number of passes more than 2 (adjusted OR, 0.80; 95% CI, 0.63 to 1.01; P=0.059) and for time from groin puncture to successful reperfusion ≤45 minutes (adjusted OR, 1.27; 95% CI, 1.00 to 1.62; P=0.053) in the propensity score-adjusted cohort. In IPTW analyses, the association of IVT+MT with number of passes more than two and time from groin puncture to successful reperfusion ≤45 minutes remained significant (Supplementary Table 3).

When the analysis in primary outcome (favorable) was stratified according to key subgroups, we found no significant heterogeneity in the treatment effect size (Figure 2). In the propensity score-adjusted cohort, we found only a non-significantly better effect of IVT+MT treatment in patients on antiplatelet drugs compared to patients not on antiplatelet drugs at stroke onset (P for heterogeneity=0.068); the adjusted OR for IVT+MT was 1.68 (95% CI, 1.06 to 2.68; P=0.029) in patients on antiplatelet drugs and 1.20 (95% CI, 0.90 to 1.61; P=0.21) in patients not on antiplatelet drugs. In IPTW analyses, we found only a significantly better effect of IVT+MT treatment in patients with ASPECTS ≥7 (OR, 1.67; 95% CI, 1.26 to 2.22) compared to patients with ASPECTS <5 (OR, 1.02; 95% CI, 0.47 to 2.24) (Supplementary Table 4).