Cerebrovascular accidents constitute the second leading cause of death worldwide, with 6.24 million deaths in 2015.1 The European Stroke Association and the American Heart Association/American Stroke Association recommend administering intravenous thrombolysis (IVT) within 3hours of ischaemic stroke onset, with the highest level of evidence,2,3 and within 4.5hours in selected cases.4 Mechanical thrombectomy should be performed within 6hours of symptom onset in patients with anterior circulation stroke secondary to large-vessel occlusion, and can be combined with IVT within 4.5hours of stroke onset in eligible patients.5 Mechanical thrombectomy achieves high rates of recanalisation and functional independence in patients with posterior circulation stroke. Intracranial haemorrhages (ICH) constitute one of the most severe hyperacute complications of revascularisation6; fear of this complication is one of the main reasons that patients refuse this treatment, leading to poorer outcomes.7 The incidence of ICH varies considerably between studies (2%-30%) due to variability in diagnostic criteria.8,9

In view of the great impact of haemorrhagic transformation on prognosis, and the lack of consensus on its definition, this study aimed to analyse ICH incidence and prognosis in patients with acute ischaemic stroke treated with IVT or with mechanical thrombectomy or pharmacological thrombolysis in our centre, and to determine whether the variables available at the time of revascularisation can predict haemorrhagic transformation.

We performed a retrospective observational study with a sample of 235 consecutive patients attended between 1 January 2014 and 30 September 2016.

The patients included were aged 18 years or older and had history of acute ischaemic stroke treated with revascularisation therapy according to the acute stroke management protocol of Hospital Universitario Reina Sofía in Córdoba, Spain.10 We excluded 34 patients with posterior circulation stroke, 17 patients with artery dissections, and one who presented stroke due to septic emboli, in order to obtain a more homogeneous sample comparable to those used in other published series. We opted to exclude patients with posterior circulation stroke due to the differences with carotid artery or middle cerebral artery (MCA) strokes in terms of management. All patients underwent head CT studies within 22-36hours of treatment, or earlier in case of neurological deterioration.4 The final sample included 183 patients.

We gathered data on demographic variables (age and sex) and cerebrovascular risk factors (diabetes mellitus, arterial hypertension, atrial fibrillation, hyperlipidaemia, history of coronary artery disease, history of stroke, peripheral vascular disease, use of anticoagulants or antiplatelets, and alcohol/tobacco use). The National Institutes of Health Stroke Scale (NIHSS) was used to determine each patient's neurological status upon arrival at the emergency department; the scale was administered by a neurologist or an emergency department specialist.4 A radiologist with over 15 years’ experience in neuroradiology determined the Alberta Stroke Program Early CT score (ASPECTS) using the non-contrast head CT images11 obtained at baseline and identified the location of the thrombus using CT angiography of the supra-aortic trunks and intracranial arteries. CT angiography results were classified as follows: no occlusion, thrombus in the extracranial segment of the internal carotid artery (ICA), in the terminal branch of the ICA, in the M1 segment of the MCA, in the M2 segment of the MCA, in the M3 segment of the MCA, or tandem occlusion. We also gathered data on the time elapsed from symptom onset to treatment. The treatment or procedure was classified as follows: IVT (rtPA dosed at 0.9mg/kg, to a maximum of 90mg), mechanical thrombectomy, pharmacological thrombolysis (intra-arterial urokinase, initial bolus of 250000IU in 10minutes, followed by 1000IU/min, to a maximum of 1000000IU), mechanical thrombectomy plus pharmacological thrombolysis, IVT plus mechanical thrombectomy, IVT plus pharmacological thrombolysis, IVT plus mechanical thrombectomy and pharmacological thrombolysis (for the latter 2 treatment combinations, maximum urokinase dose was 500000IU), IVT plus angiography without endovascular treatment, or diagnostic angiography without endovascular treatment. We also evaluated disability at discharge using the modified Rankin Scale (mRS), which was administered by a neurologist; scores ≥ 3 points were considered to indicate dependence.4,12

Haemorrhagic transformation was classified according to the classification of ICH proposed by Fiorelli et al.13 (Table 1, Fig. 1): haemorrhagic infarction type 1 (HI1), haemorrhagic infarction type 2 (HI2), parenchymal haematoma type 1 (PH1), parenchymal haematoma type 2 (PH2), remote parenchymal haematoma type 1 (RPH1), and remote parenchymal haematoma type 2 (RPH2). We also considered patients with subarachnoid haemorrhage (SAH) and those displaying contrast extravasation, defined as a hyperdense area with > 90 Hounsfield units on the follow-up CT scan14; contrast extravasation is the main differential diagnosis for ICH after mechanical thrombectomy.15 Some patients displayed more than one type of haemorrhage; a single patient could therefore be classified under several categories. Follow-up CT images were independently analysed by 2 radiologists with 5 and 3 years’ experience, respectively; disagreements were discussed with a third radiologist with over 15 years’ experience in neuroradiology.

Figure 1.

Baseline head CT scan performed within 36hours of revascularisation. (a) Haemorrhagic infarction type 1: small petechiae around the infarcted area. (b) Haemorrhagic infarction type 2: confluent petechiae in the infarcted area with no mass effect. (c) Parenchymal haematoma type 1: haematoma in ≤ 30% of the infarcted area, with no mass effect. (d) Parenchymal haematoma type 2: parenchymal haematoma open to the ventricular system, observed in > 30% of the infarcted area, with a significant mass effect. (e) Remote parenchymal haematoma type 1: small haemorrhage with no association with the infarcted area (left middle cerebral artery territory). (f) Haemorrhagic infarction type 1+subarachnoid haemorrhage+contrast uptake: small petechiae around the infarcted area, haematoma in the subarachnoid space, and contrast uptake.

(0.41MB).

To identify the predictors of ICH and PH, we performed a univariate analysis, using the chi-square test and Fisher exact test for qualitative variables and the t test for quantitative variables. The 2 dependent variables (ICH and PH) were modelled using binary logistic regression. To create the regression line, we analysed the independent variables that displayed an association with the dependent variables; these were identified using contingency tables and the chi-square test, with the critical value established at 0.1. Having obtained these variables, we conducted a binary logistic regression analysis, following the enter method, to obtain the regression line. To validate the regression line we calculated the Cox and Snell R2 coefficient, the Nagelkerke R2 coefficient, and the percentage of correct predictions of the model. The latter is the most important value for validation; regression lines with a percentage below 50% should not be used.

We also designed other binary logistic regression models to determine the prognosis and mortality rates associated with ICH and PH.

Statistical analysis was performed using SPSS version 23 for Windows.

A total of 183 patients met our inclusion criteria. Mean age (standard deviation) was 67 (12) years; 63% of patients were men. Mean NIHSS score upon arrival at the emergency department was 14 (5) points; mean ASPECTS score was 8.4 (1.4) points.

CT angiography revealed no alterations in 22% of patients; thrombi were located in the extracranial segment of the ICA in 12%, in the terminal branch of the ICA in 12%, in MCA segment M1 in 31%, in MCA segment M2 in 8%, and in MCA segment M3 in 2%; and tandem occlusions were found in 14% of patients.

Regarding treatment, 35% of patients received IVT, 23% underwent mechanical thrombectomy, 1% received pharmacological thrombolysis, 3% mechanical thrombectomy plus pharmacological thrombolysis, 27% IVT plus mechanical thrombectomy, 3% IVT plus pharmacological thrombolysis, 4% IVT plus mechanical thrombectomy and pharmacological thrombolysis, 4% received IVT and underwent angiography but no endovascular procedure, and 1% underwent angiography and no endovascular procedure. Mean time from symptom onset to treatment onset was 174.2 (81.8) minutes.

Haemorrhagic transformation was observed in 31.1% of patients; 16.4% were PH (PH2: 14.2%). HI2 was the most common type of haemorrhagic infarction, accounting for 8.7% of cases. Symptomatic ICH, (i.e., ICH accompanied by neurological deterioration [NIHSS≥ 4 points]) was observed in 10.9% of patients, 80% of whom had PH2. Fourteen percent of the patients who received IVT developed ICH (6% with symptomatic ICH), as did 41% of those who underwent endovascular treatment with or without IVT (13% with symptomatic ICH). SAH and contrast extravasation were observed in 3.2% and 1.6% of patients, respectively; both complications were associated with haemorrhagic infarction in all cases.

At discharge, 45.9% of patients were dependent according to the mRS. Sixteen (25%) of the patients who received IVT were discharged with mRS scores indicating dependence, 4 of whom (6%) presented ICH. Sixty-eight (58%) of the patients who underwent endovascular treatment with/without IVT were discharged with mRS scores ≥ 3 points, of whom 33 (28%) presented ICH. The overall mortality rate was 16.9%. Four (6%) of the patients who received IVT died, of whom 2 (3%) presented ICH. Twenty-seven (23%) of the patients who received endovascular treatment with/without IVT died, 16 of whom (14%) presented ICH.

Table 2 shows the univariate analysis of the predictors of any type of ICH. Mechanical thrombectomy (OR=3.3; 95% CI, 1.42-7.63; P=.005) and IVT plus mechanical thrombectomy (OR=3.39; 95% CI, 1.52-7.56; P=.003) were found to be independent predictors of ICH in the binary logistic regression model (Table 3), whereas high ASPECTS scores (OR=0.71; 95% CI, 0.55-0.91; P=.007) were independently associated with lower risk of ICH. The model created to predict ICH has a percentage of correct predictions of 72.4%.

Table 4 shows the results of the univariate analysis of predictors of PH. Older age (OR=1.07; 95% CI, 1.02-1.13; P=.006) and thrombus location in the terminal branch of the ICA (OR=4.03; 95% CI, 1.35-11.99; P=.012) were found to be independent predictors of PH in the binary logistic regression model, whereas treatment with IVT (OR=0.24; 95% CI, 0.08-0.68; P=.008) was independently associated with lower risk of PH (Table 5). The model created to predict PH showed a percentage of correct predictions of 80.1%.

After controlling for all the variables measured, ICH was not associated with mRS scores ≥ 3 at discharge or with higher mortality rates. After adjusting for all the variables measured, PH was found to be associated with mRS scores ≥ 3 (OR=3.2; 95% CI, 1.17-8.76; P=.02) and higher mortality rates (OR=5.06; 95% CI, 1.65-15.5; P=.005).

The studies published in the literature propose different definitions of haemorrhagic transformation and different methodologies for patient selection, which makes it difficult to analyse and extrapolate findings.

The authors have no conflicts of interest to declare.