Association between different acute stroke therapies and development of post stroke seizures

This was a retrospective, multicentre cohort study conducted at the Royal Melbourne Hospital and Jingling Hospital, Nanjing. Subjects at both centres were identified from prospectively maintained clinical databases of patients admitted with an ischemic stroke. The Royal Melbourne Hospital, located in Victoria, Australia, provides IV-tPA therapy to acute ischemic stroke patients who arrive to the hospital within 4.5 h of stroke onset. It also serves as the state-wide referral centre for intra-arterial therapies, including endovascular thrombectomy and intra-arterial urokinase. Jingling Hospital is located in Jiangsu province, China, where most patients received stroke unit care only. Ethical approval of the study was granted by the Melbourne Health Human Research Ethics Committee (project number QA2010089) and patient consent was waived.

We included patients with anterior circulation ischemic stroke admitted to the two hospitals between 2008 and 2015. Patients with a history of epilepsy or seizures prior to their stroke were excluded. Patients included were divided into 4 groups based on the type of acute reperfusion treatment received: 1. IV-tPA only, 2. IAT only, 3. IAT + IV-tPA and 4. stroke unit care only (i.e. no IV-tPA or IAT). For the purpose of analyses, patients receiving ‘stroke unit care only’ were regarded as controls.

Clinical data was collected and entered into the databases at both centres when the patients were admitted to the emergency department, transferred to stroke wards and returned to stroke follow-up clinics. Data included patient demographics, age, sex, pre-morbid modified Rankin score (mRS), admission National Institutes of Health Stroke Scale (NIHSS) score, admission blood pressure and stroke risk factors such as hypertension, atrial fibrillation, diabetes, dyslipidemia, previous stroke or transient ischemic attack (TIA) and smoking. Clinical follow-up information included the modified Rankin Scale (mRS) at 3-months post onset with good outcome defined as mRS 0–2. Hemorrhagic transformation was assessed using the ECASS classification [15] on follow-up 24 h CT brain imaging, blinded to seizure data collection.

Patients with post stroke seizures were identified if they experienced seizures up to 2 years from stroke onset. This cut-off was chosen based on previous findings that suggested that the highest risk of seizure development was within the first year [16]. Occurrence of post stroke seizures was ascertained by reviewing follow-up medical records and via telephone interview using a questionnaire from previous studies [14, 17‐19], modified from a validated screening questionnaire [20]. This questionnaire was translated into Chinese for use at the Jingling Hospital Nanjing. Events were recorded as seizures if the symptoms included motor or autonomic components, with or without impairment of consciousness, as defined by the International League Against Epilepsy [21].

To assess homogeneity between the two sites, the seizure occurrence within 2 years post-stroke between the control groups across the two sites was compared using Fisher’s Exact test with corresponding effect estimated as Odds Ratios (OR) with 95% confidence interval (95%CI). When no significant difference in seizure incidence between the two sites was found, the control groups were combined.

The demographic, clinical and risk factor characteristics for the control group, IAT only, IV-tPA only and IAT + IV-tPA combined groups were summarised as median (IQR) for continuous characteristics and as counts (proportions) for categorical characteristics, and compared using either Kruskall-Wallis test or Fisher’s Exact test depending on the nature of the distribution.

To assess the association between the types of treatment (reperfusion: IAT only, IV-tPA only, IAT + IV-tPA; with controls) and seizure occurrence within 2 years, logistic regression modelling with seizure occurrence as an output and treatment groups as inputs was used. The analysis was adjusted for the following a priori chosen covariates known to be associated with post stroke seizures: age, baseline NIHSS and 3 month mRS [16, 22‐26]. Corresponding effect sizes were summarised as adjusted ORs with 95%CI. For robustness analysis, extra covariates that demonstrated statistically significant association with seizure occurrence on univariate analyses were subsequently included in the model.

To further investigate the robustness of the modelling outcomes, we performed a sensitivity analysis to assess the potential association between thrombectomy and seizures with a control group more closely matching the patients who receive thrombectomy. We used a selection criterion of NIHSS≥6 as per AHA/ASA guidelines for thrombectomy eligibility [27]. We additionally performed a sensitivity analysis including only patients with an NIHSS> 8 to further increase specificity for large vessel occlusion [28].

There was a significant difference in baseline NIHSS across treatment groups median (IQR): IAT + IV-tPA = 17 (13–21), IAT = 18 (13–21), IV-tPA only = 10 (6–17), control = 2 (0–7), p = 0.0001. There was a significant difference in age at stroke across the treatment groups median (IQR): IAT + IV-tPA = 70 (59–78), IAT = 67 (54–77), IV-tPA only = 74 (65–77) control = 62 (52–70), p = 0.0001. There was a significant difference in mRS 0–2 at 3 months across the treatment groups: IAT + IV-tPA = 59/112 (52.7%), IAT = 36/93 (38.7%), IV-tPA only = 191/363 (52.6%) control = 839/1375 (61.0%), p < 0.0001. Baseline characteristics are detailed in Table 1.

The univariate analysis for baseline variables age, NIHSS at baseline and the 3-month mRS of 0-2 and their association with post stroke seizures are provided in Table 2. In multivariable logistic regression adjusted for baseline NIHSS, age, and mRS at 3 months, administration of IV-tPA (without IAT) was significantly associated with increased odds of seizure development: adjusted OR 3.7, 95%CI 1.8–7.4, p < 0.0001. Similarly, there was an independent association between treatment with IAT only and increased odds of seizure development compared to controls: adjusted OR 5.5, 95%CI 2.1–14.3, p < 0.0001. There was also a trend towards increased seizure risk in patients treated with combined IAT + IV-tPA compared to controls: OR 3.4, 95%CI 0.98–11.8, p = 0.05 (Fig. 1). For the purpose of robustness analysis, we included extra baseline adjustment covariates and risk factors that were significantly associated with seizure development in univariate analysis and were recorded for all treatments: sex, hypertension, atrial fibrillation and dyslipidemia (Table 1). Including these variables as extra covariates increased the collinearity in the regression model and rendered some standard error estimates less stable, but the adjusted estimates for the effects for individual treatment groups remain quantitatively similar: IAT only OR 6.6, 95%CI 2.5–17.8, p < 0.001, IV-tPA only OR 2.9, 95%CI 2.3–10.7, p < 0.001, IV-tPA+ IAT OR 4.3, 95%CI 1.2–15.3, p = 0.03. Consistently with largely overlapping confidence intervals, the odds of post stroke seizures in the three treatment groups did not differ (IV-tPA + IAT versus IV-tPA p = 0.89, IV-tPA + IAT versus IAT, p = 0.44, Fig. 1).

For the sensitivity analysis including only patients with an NIHSS ≥6 (to better match the controls with patients eligible for thrombectomy [27]), the results of the logistic regression were similar to the previous results: IV-tPA verses controls OR 3.2, 95%CI 1.4–7.4, p = 0.005, IAT verses controls OR 4.6, 95%CI 1.6–13.0, p = 0.004, IV-tPA + IAT versus controls OR 3.2, 95%CI 0.84–12.1, p = 0.09.

For the sensitivity analysis including only patients with an NIHSS> 8 (to further increase specificity for large vessel occlusion [28]) the results of the logistic regression were similar to the previous results: IV-tPA versus controls OR 3.9, 95%CI 1.5–9.7, p = 0.004, IAT versus controls OR 5.3, 95%CI 1.9–15.3, p = 0.002, IV-tPA + IAT versus controls OR 4.0, 95%CI 1.03–16.2, p = 0.045. Additional file 1 for the univariate and multivariable logistic regression has been presented.