| IntroductionWith relatively low incidence accounting for5%to10%of intracerebral hemorrhage (ICH), primary pontine hemorrhage (PPH) is the most devastating hemorrhagic stroke, reported to have an acute mortality ranging from40%to70%. Various factors including coma at admission and hemorrhage volume were found to associate with these diverse outcomes. However, predictors varied when different parameters, population or statistical methods were brought into analysis. As a corollary, it would make cogent sense by combining significant factors to establish a so called outcome predicting system to enhance predictive power and make it quantitative.Generally, an excellent outcome predicting system should have certain calibration, which predicts mortality close to actual mortality. In addition, the system should have high discrimination, which could determine a survivor from non-survivors. Moreover, the system should be simple and easy to use, and should contain concise parameters that could be easily obtained. Also, time required to complete evaluation should be short, and the parameters should be objective and quantitative. Last but not least, the application of the system should be as widely as possible, and the requirements of occation and equipment to carry out evaluation should be as simple as possible. The ICH score, proposed by Hemphill et al in2001is a quantitative outcome predicting system, consisted of age (≥80years=1point,<80years=0point), hemorrhage volume (≥30ml=1point,<30ml=0point), infratentorial origin (yes=1point, no=0point), intraventricular hemorrhage (IVH, present=1point, absent=0point), and Glasgow Coma Scale (GCS,3-4=2points,5-12=1point,13-15=0point). It has been widely validated for its generalization in predicting acute mortality as well as long-term functional outcome in spontaneous ICH. However, the study cohorts used for both model derivation and validation based mainly on supratentorial hemorrhage, and no external validation focused exclusively on PPH has been performed. In addition, the roles of hemorrhage volume and infratentorial origin in the ICH score are meaningless to a patient group of PPH. Moreover, age was reported to be not significant in most studies, while hemorrhage volume was widely regarded as a strong predictor to PPH, with hemorrhage volume grater than5ml reported to associate with poor outcome. Consequently, it is necessary to modify the ICH score to make it more suitable for prediction in PPH patients.In a previous preliminary study, we have compared the Acute Physiology and Chronic Health Evaluation (APACHE) II and the Simplified Acute Physiology Score (SAPS) II with the ICH score in predicting30-day mortality of PPH patients. It was found that the APACHE II had the highest discrimination, whereas the SAPS II had the best collibration. However, both of the APACHE II and the SAPS II contain too many parameters, costing too much time to finish evaluation, and therefore are inconvenient for bedside evaluation. The ICH score, however simple and easy to use, had lower discrimination and collibration than the APCHAE II and the SAPS II. It was implied that the performance of the ICH score would be better if it was modified properly.Objective:To further validate the ICH score in predicting mortality as well as good outcome at30days in an independent patient group of PPH; To determine whether modification could improve the prediction.Materials and methodsA multicenter study was conducted by retrospectively reviewing consecutive patients with first-ever pontine hemorrhage admitted to three teaching hospitals (Nanfang Hospital, Zhujiang Hospital, Southern Medical University; The Second Affiliated Hospital of Guangzhou Medical College) in Guangzhou between January2005and January2012. The inclusion criteria were:(1) a diagnosis of pontine hemorrhage verified by clinical manifestation and CT;(2) aged18to90years;(3) admitted within24hours after symptoms onset;(4) no history of stroke. The exclusion criteria were:(1) missing of clinical date;(2) hemorrhage extended into cerebellum;(3) pontine hemorrhage secondary to head trauma, bleeding diathesis, a cavernous hemangioma, or an arteriovenous malformation;(4) patients that suffered from end-stage malignant diseases before the onset.Medical records were carefully reviewed to obtain the following parameters:age, sex, hypertension, diabetes mellitus, smoking, alcohol abuse, surgery, extraventricular drainage (EVD), and need for mechanical ventilation. Identify with the ICH score, other variables were abstracted from initial ICH evaluation, including temperature, heart rate, respiratory rate, systolic blood pressure, GCS score, and CT features. GCS score was determined at the first time of neurological examination. If GCS score was not specifically recorded, it was calculated from the described neurological examination. Hemorrhage volume was measured with formula ABC/2, where A is the greatest hemorrhage diameter by CT, B is the diameter90degrees to A, and C is the approximate number of CT slices with hemorrhage multiplied by the slice thickness. Patients with IVH and extrapontine extension into midbrain, medulla, or thalamus were also recorded. Hydrocephalus was determined by Diringer’s method.Primary endpoints were mortality and functional outcome at30days after onset. Functional outcome was measured by modified Rankin Scale (mRS), with mRS of0-2regarded as good outcome, mRS of3-5as poor outcome, and mRS of6as death. Information on survival and functional outcome of these who survived when discharged were obtained from family members through telephone interviews by a trained neurologist, who had no access to study data. These of who could not be contacted were carefully judged from medical records by another experienced neurologist who was blinded to the study design.Categorical variables were present by case number (percentage) and compared using the two-sided chi-square analysis or Fisher’s exact test. Partition of chi-squared was performed when comparing multiple groups (hematoma locations, hemorrhage volume, GCS). Continuous variables with normal distribution were presented with mean±SD and compared by using the two-sided Student’s t test, whereas those without normal distribution were presented with median (quartile) and compared using the Mann-Whitney U test. Similar to the development of the ICH score, variables that showed significance in univariate analysis were drawn into multivariate logistic analysis. Points were assigned by weighting their influence (Odds Ratio, OR) and a model was developed. Cuzick’s trend test was used to assess the correlation between models with30-day mortality and30-day good outcome of PPH. Receiver operating characteristic (ROC) curve analysis was performed to measure the discrimination of the models. Area under the curve of "0.7-0.8" was regarded as "acceptable","0.8-0.9" was regarded as "excellent", and "0.9-1.0" was regarded as "outstanding" discrimination. Areas under the curves were compared non-parametrically using the Hanley and McNeil method. Maximum Youden Index (sensitivity+specificity-1) was identified to denote each score’s optimal predictive cutoff value and sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated at each score’s maximum Youden Index. All the analyses were performed using the SPSS Ver.13.0. A significant level was established at P<0.05.ResultsA total of171patients (126male,45female) met the criteria were finally included, with52(30.4%) of them from Nanfang Hospital,72(42.1%) from Zhujiang Hospital, and47(27.5%) from the Second Affiliated Hospital of Guangzhou Medical College. The median age was53(45-60) years, and median duration of hospital stay was16(4-35) days. Sixty-eight (39.8%) patients died and52(30.4%) had good outcome within30days. Of the non-survivors,37(54.4%) of them were taken off life support based on perceived poor prognosis. Of all the patients,23(13.5%) could not be contacted and3(1.8%) were regarded as death at30days according to the medical records. The median hemorrhage volume was5.0(2.7-9.6) ml, with57(33.3%) patients had CT evidences of hemorrhage extension into the midbrain,15(8.8%) into the medulla, and17(9.9%) into the thalamus.Mulitivariate logistic analysis was performed on variables that were significant on univariate analysis. It was found that only GCS and hemorrhage volume were independently related to30-day mortality as well as30-day good outcome. As GCS and hemorrhage volume showed similar weight on the model, same points were assigned and a model named PPH score was established. Although GCS was generally categorized at8, we agreed with Hemphill et al that classifying GCS into3-4,5-12, and13-15was more clinically relevant. In addition, the classification of hemorrhage volume of the original ICH score (≥30ml or<30ml) is inappropriate to a patient group of PPH. According to literature review and clinical findings, hemorrhage volume was classified into three group, including<5ml,5-10ml, and>10ml. Univariate analysis demonstrated that patients with hemorrhage volume bigger than10ml showed significantly higher30-day mortality than the other two groups (P<0.001), whereas those who smaller than5ml showed lowest30-day mortality but highest ratio of30-day good outcome (both P<0.001). In brief, the PPH score was consisted with only2parameters, namely GCS (3-4=2points,5-12=1point,13-15=0point) and hemorrhage volume (<5ml=0point,5-10ml=1point,>10ml=2points).No patient had a score of0or6on the ICH score, and all patients with ICH scores of5died within30days. Thirty-day mortality for patients with ICH scores of1,2,3, and4were1.8%,34.1%,61.4%, and92.3%, whereas ratio of30-day good outcome (mRS≤2) were71.4%,25.0%,2.3%, and0%, respectively. All patients with PPH scores of4died within30days. Thirty-day mortality for patients with PPH scores of0,1,2,3were1.7%,17.2%,40.0%, and81.1%, whereas ratio of30-day good outcome were74.1%,24.1%,8.0%, and0%, respectively. It is shown in the Cuzick’s trend test that30-day mortality increased as ICH scores and PPH scores increased, while30-day good outcome decreased dramatically (both P<0.001). In terms of areas under the curve, the ICH score was excellent in predicting both30-day mortality (0.874,95%confidence interval [95%CI]0.822-0.925) and30-day good outcome (0.879,95%CI0.828-0.930). However, the PPH score was more discriminative than the ICH score in predicting both30-day mortality (u=4.20, P<0.001) and30-day good outcome (u=1.94, P=0.026) through non-parametric comparison. For30-day mortality, the cutoff values were set at1.5for the PPH score and2.5for the ICH score, respectively. The PPH score had higher Youden Index, sensitivity, PPV, and NPV than the ICH score, whereas the ICH score had higher specificity. For30-day good outcome, the cutoff values were set at0.5for the PPH score and1.5for the ICH score, respectively. The PPH score had higher Youden Index, sensitivity, specificity, PPV, and NPV than the ICH score.ConclusionsAccurate outcome prediction in critically ill neurologic patients is necessary to distinguish who would benefit from special care or who should limit the usage of medical resources. Undoubtedly, prognostic estimates are still only estimates, and no single clinical outcome measure can predict all dimensions of mortality or recovery. Nevertheless, the ICH score and the PPH score were both found to be suitable for outcome prediction in PPH patients. Once again, the ICH score was validated to be effective in an independent patient group of PPH, demonstrating its potential of improving consistency in clinical care and clinical research. Worthy of note, the PPH score, modified from the original ICH score, was more accurate and may be preferred when dealing with a specific patient of PPH. |
PDF Full Text Request