CT pattern of Infarct location and not infarct volume determines outcome after decompressive hemicraniectomy for Malignant Middle Cerebral Artery Stroke

Malignant middle cerebral artery [MMCA] infarction has a different topographic distribution that might confound the relationship between lesion volume and outcome. Retrospective study to determine the multivariable relationship between computerized tomographic [CT] infarct location, volume and outcomes in decompressive hemicraniectomy [DHC] for MMCA infarction. The MCA infarctions were classified into four subgroups by CT, subtotal, complete MCA [co-MCA], Subtotal MCA with additional infarction [Subtotal MCAAI] and co-MCA with additional infarction [Co-MCAAI]. Maximum infarct volume [MIV] was measured on the pre-operative CT. Functional outcome was measured by the modified Rankin Scale [mRS] dichotomized as favourable 0–3 and unfavourable ≥4, at three months. In 137 patients, from least favourable to favourable outcome were co-MCAAI, subtotal MCAAI, co-MCA and subtotal MCA infarction. Co-MCAAI had the worst outcome, 56/57 patients with additional infarction had mRS ≥ 4. Multiple comparisons Scheffe test showed no significant difference in MIV of subtotal infarction, co-MCA, Subtotal MCAAI but the outcome was significantly different. Multivariate analysis confirmed MCAAI [7.027 (2.56–19.28), p = 0.000] as the most significant predictor of poor outcomes whereas MIV was not significant [OR, 0.99 (0.99–01.00), p = 0.594]. Other significant independent predictors were age ≥ 55 years 12.14 (2.60–56.02), p = 0.001 and uncal herniation 4.98(1.53–16.19), p = 0.007]. Our data shows the contribution of CT infarction location in determining the functional outcome after DHC. Subgroups of patients undergoing DHC had different outcomes despite comparable infarction volumes.


Material and Methods
In a retrospective, multicenter cross-sectional study data for DHC for MMCA infarction, pooled from 3 tertiary referral centres in 3 countries [Qatar, Pakistan and the United Arab Emirates], collected between 2007 and 2014 was analyzed.
Inclusion and exclusion criteria. The study included all patients referred for DHC. The patient inclusion criteria were: National Institutes of Health Stroke Scale score [NIHSS] ≥15 including a score of 1 for item 1a (decreased level of consciousness from the beginning or progressive deterioration]. Cerebral computed tomography [CT] showing infarction involving at least two third of the middle cerebral artery territory[MCA] infarction [>50% but <100%] including basal ganglia with or without internal capsule involvement, signs of oedema with or without additional anterior cerebral artery [ACA] or posterior cerebral artery [PCA] involvement.
Data included demographics, risk factors, imaging studies, hyperosmolar treatment, signs of herniation, time of herniation, and time to surgery. Radiological data. All radiological measurements were made by consensus between 2 Neuroradiologists blinded to clinical data, and a third opinion [stroke Neuroradiologist] was obtained in cases of disagreement. For determining the lesion pattern and maximum pre-operative infarction volume [MIV] the last CT prior to DHC was used. The infarction volume [IV] was measured by open source image analysis software OsiriX version 5.6 9 . Using a closed polygon tool, lesion outline was marked on the first, middle and last CT slice with ischemic stroke. The software then generates the missing polygons that were checked for accuracy and manually corrected if needed. Once corrected, infarction volume was automatically generated by the software. For temporally separate lesions, each lesion was measured separately and added to generate infarction volume.
For CT stroke patterns we used criteria of the randomized control trials, published literature and templates of Tatu et al. as a guide to the topography of stroke by vascular territory [10][11][12][13][14][15][16] . The patients were classified into four subgroups according to CT patterns, subtotal infarction in the territory of the middle cerebral artery [>50% with basal ganglia involvement], complete involvement of the MCA [co-MCA], Subtotal MCA [≥50% MCA] with additional infarction due to ACA and/or PCA involvement [Subtotal MCAAI] and complete MCA with additional infarction due to ACA and/or PCA involvement [Co MCAAI]. The size of the bone flap was measurement included anterior-posterior, vertical measurements from the inner skull table and if middle cranial fossa was opened on postoperative CT scan. For measurement of septum pellucidum shift, a straight line was drawn in the expected location of the septum pellucidum from the posterior-most aspects to the falx on axial images. The shift of the septum pellucidum from this midline was measured on all CT scans and compared to subsequent CT scans to determine any change.
Early signs of uncal herniation on CT scan included medial displacement of the uncus and parahippocampal gyrus of the temporal lobe, encroachment on the suprasellar cistern, narrowing of the contralateral ambient and quadrigeminal plate cisterns with the ipsilateral widening of the ambient and quadrigeminal cisterns 17 .
Patients with only a single imaging study, parenchymal hematoma grade II 18 , haemorrhage with ventricular extension or missing surgical details were excluded from the analysis.
The decision for DHC was made by the treating physicians/neurosurgeons based upon the individual clinical condition and cerebral CT imaging. Patients were generally taken to surgery if their level of consciousness was progressively deteriorating with or without early clinical signs of herniation. All patients underwent hemicraniectomy and duroplasty without removal of the infarcted tissue. The size of the bone flap and opening of the middle cranial fossa was at the discretion of the surgeon. Surgical and postoperative complications were recorded. The bone flap was replaced within 3 months of surgery.
The primary outcome was measured using a modified Rankin Scale (mRS) score at three months by patient examination in the outpatient clinics. Clinical outcome was categorized as favourable (mRS 0-3) and (mRS 0-4) or unfavourable (mRS 4-6) and (mRS 5-6) based on post hoc analysis of the pooled data of the randomized control trials 19 .
All the tertiary referral centres in the study have a well-established comprehensive stroke service including stroke diagnostic, vascular interventional, stroke units, neurosurgical and rehabilitation services. An acute stroke team is available 24 hours a day, seven days a week.
Because of retrospective data evaluation the informed consent was waived by the Institutional Review Board of Hamad Medical Corporation, Qatar, Shifa International Hospital, Pakistan and Rashid Hospital, UAE. Data analysis. All statistical analyses were performed using Statistical Package for Social Sciences Version 22 (SPSS). Descriptive and inferential statistics were used to characterize the study sample and test hypotheses. Descriptive results (including graphical displays) for all quantitative variables (e.g., age) are presented as mean ± standard deviation (SD) (for normally distributed data) or median with interquartile range (for data not normally distributed). Numbers (percentage) were reported for all qualitative variables (e.g., gender). Bivariate analysis was performed using ANOVA, and Kruskal-Wallis test to compare the average for all quantitative variables (e.g., age) among types of infarct location (subtotal MCA infarction, co-MCA, Subtotal MCAAI and Co-MCAAI). Whenever appropriate, while Pearson Chi-Square test or Fisher Exact test as appropriate were used to comparing all the qualitative variables (e.g. gender) among types of the infarct. Multiple comparisons analysis were performed using the Scheffe test if there was an overall statistically significant difference for the average of maximum infarct volume and mRS at three months using One Way ANOVA. Multiple logistic regression models were used to identify significant independent factors associated with functional outcome (mRS 0-4 vs. 5-6) at three months after adjusting for potentially confounding factors. To build the model, a purposeful selection method was used for selecting a subset of covariates considered clinically important, adjusting for confounders and statistical significance. Purposeful selection of covariates begins with a multivariate model that contains all (2019) 9:17090 | https://doi.org/10.1038/s41598-019-53556-w www.nature.com/scientificreports www.nature.com/scientificreports/ variables that are significant in the bivariate analysis at the 20-25 per cent level, as well as any other variables not selected with this criterion but judged to be of clinical importance. We used p-values from the Wald tests of the individual coefficient to identify covariates that might be deleted from the model and p-value of the partial likelihood ratio test confirming that the deleted covariate is not significant. Following the fitting of the reduced model, we assessed whether or not a removal of the covariate will produce an "important" change (about 20%) in the coefficient of the variables remaining in the model. The final model was assessed using the Hosmer-Lemeshow Goodness-of-fit statistics to determine whether the model adequately describes the data. Adjusted Odds ratio and 95% confidence interval for the adjusted odds ratio were reported for each independent predictor. A "P" value < 0.05 (two-tailed) was considered statistically significant.
Compliance with ethical standards. The  Multiple comparisons scheffe test. The relationship between types of the infarct, maximum infarct volume and mRS at three months is shown in Table 2, Fig. 1.

MIV (cm 3 )
. ANOVA Fig. 2]. Most patients in all subgroups had an mRS of 4, irrespective of CT infarction location except co-MCAAI where it was evenly split between mRS 5 or 6. Only one patient with subtotal MCAAI survived with an mRS of 3 out of 57 patients with additional infarction. None of the patients with subtotal MCA infarction had an mRS of 5 or 6. The highest mortality was observed in the subtotal MCAAI and co-MCAAI groups [19/23, 82.60%; p = 0.001]. All deaths happened within two weeks of stroke onset as a result of transtentorial herniation. Table 3] confirmed co MCAAI 7.027 (2.56-19.28), p = 0.000] as independent predictor of functional outcomes at three month after adjusting for confounders [age, gender, MIV, uncal Herniation on CT, SP displacement ≥1 cm, participating sites, hypertension, diabetes, dyslipidemia, CAD, atrial fibrillation, congestive heart failure, NIHSS at admission, intracerebral hemorrhage at the craniotomy site, subgleal and subdural hematoma, infection at the craniotomy site, aspiration pneumonia, and seizures. The MIV did not show any significant association [0.99 (0.99-01.00), p = 0.594] in the presence of MCA with additional infarction after adjusting for confounders.

Discussion
Our data support the hypothesis that the pattern of CT infarction location and not the MIV is the major determinant of functional outcome after DHC. Subgroups of patients undergoing DHC had different outcomes despite comparable infarction volumes but differed in infarction location. The stroke groups could be arranged in order of the most to least unfavourable functional outcome after DHC were co-MCAAI, Subtotal MCAAI, co-MCA and subtotal MCA infarction. MMCA with additional infarction involving ACA and/or PCA had the worst functional outcome after DHC.
With similar MIV, co-MCA and subtotal MCAAI had significantly different functional outcomes [ Fig. 2]. Interestingly, direct physical factors, such as involvement of ACA and/or PCA arteries apart from MCA, the extent of midline shift and CT signs of uncal herniation bore a significant relationship to the functional outcome. In fact, the outcome was unfavourable if there was any additional stroke [ACA, PCA] irrespective of the size of MCA infarction [Fig. 2]. In HAMLET trial 34% of patients in the control arm had MCAAI compared to 22% who  www.nature.com/scientificreports www.nature.com/scientificreports/ underwent surgery, with increased mortality and unfavourable outcome in the control arm 13 . The ACA and PCA provide important collateral blood flow to the conventionally defined MCA territory and therefore ACA and/or PCA ischemia is likely to be detrimental to the perfusion of the various MCA territories as well. The presence of additional infarcts [ACA, PCA] not only increases the infarction volume and cerebral oedema 20 but add to the cognitive, behavioural impairments and aggravate the functional deficits, thereby, compromise the potential for recovery [21][22][23] .
We accounted for infarction location, infarction volume, as well as clinical characteristics to develop multivariable risk adjustment models, identifying predisposing factors associated with an unfavourable functional outcome. Although it is reasonable to assume that MCA with additional infarction will have lager infarction volumes and therefore, the poor functional outcome as reported recently 24 . In this study, most of the cohort had MCA with additional strokes [49/92; 53.26%] leading to higher infarct volume thus relating poor functional outcome to infarction volume. Amongst the randomized trials of DHC, only DECIMAL used infarction volume in addition to the involvement of MCA >50% for enrolment in the study 10 . Similar to the findings of the DECIMAL trial of a non-significant trend towards a poor outcome with higher infarct volumes at inclusion our outcome regression prediction model did not find any significant relation of MIV to functional outcome. It is also possible that CT infarction location [MCA with additional infarction] covaried with MIV, explaining why MIV was not a significant predictor in the regression analysis.   www.nature.com/scientificreports www.nature.com/scientificreports/ The current study has several limitations including the retrospective nature and the lack of a randomized comparison between groups. We used CT scan lesion patterns to classify patients into groups. The lack of detailed vascular imaging and status of collateral circulation on all patients may have led us to misclassify a complete MCA [due variation in the circulation with ACA and PCA feeding a part of MCA territory] as a subtotal MCA stroke. However, CT lesion patterns [>50% of the MCA territory, DECIMAL and HeADDFIRST, and 2/3 of the territory and including at least part of the basal ganglia, DESTINY and HAMLET] was used in randomized trials of DHC. The data was collected from hospitals in three countries and reflects real life experience rather than clinical trials. There were no local guidelines for DHC patient referral and pre-operative care was not standardized between the participating centres that could have influenced patients' selection and timing of surgery. Moreover, selection bias due to the difference in criteria for surgical selection between treating physicians/surgeons and centres with widely applied standardization protocol for DHC cannot be excluded. Although we have shown that infarction location and not MIV is important but there is a possibility that due to smaller sample size in subgroups there was lack of statistical power to resolve the association of MIV vs. location and outcome. The predominant male population in our study was due to the males expatriate population in the UAE and Qatar. Finally, a major limitation is the short-term (3-month) follow-up and type of rehabilitation received. As most expatriates will leave the country [Qatar and UAE] after treatment, long-term follow-up was not available.
The present study shows the contribution of infarction location to the outcome after DHC. With the same MIV, infarction location showed wide variations in the outcome highlighting the difficulty in outcome prediction. The predictive value of preoperative CT infarction location may help in the decision-making process in the identification of patients at risk of poor functional outcomes, enhance our current strategies of patient selection for DHC and assist in the planning of rehabilitation. Our study is relevant in the context of increasing stroke burden in low and middle-income countries 25 where modern intervention techniques [thrombolysis, thrombectomy] are either too expensive or not available.  Table 3. Multivariate logistic regression model to confirm MCA with additional infarction is a significant independent predictor for functional outcomes at three months. CI: Confidence Interval; P-value has been calculated using binary multiple logistic regression Wald test. ICH intracerebral hemorrhage.