It is well known that intracerebral hemorrhage (ICH) is the second most common type of stroke, accounting for 10–15% of all stroke events.1 The fatality rate is relatively high, only 12–39% survivors could live independently.2 Early hematoma expansion (HE) occurs in 20–30% of ICH patients,3 accompanied by re-bleeding occasionally. Through the CT scanning and three-dimensional reconstruction analysis of the enlarged hematoma, Liotta4 found that HE was caused by the irregular expansion along the surface of the original hematoma, mainly within 24h after the onset. HE is an independent risk factor for disability and death in patients of ICH, so timely assessment and prevention of HE becomes the focus of current research. But treatments toward HE still exist some disputes and have not been standardized, including intensive blood pressure control, transamin,4 rectify the blood coagulation dysfunction, neurological intensive care unit and minimally invasive surgery. Saving lives, utmost retaining or even restoring neurological function are the fundamental purposes of treatments. Therefore, it is of great value to use simple predictors to screen out the high-risk patients who may have HE and make targeted treatments to curb the early deterioration of ICH. The objective of the study was exploring the risk factors of HE, proposing a prediction score and making a preliminary validation.

We proposed and validated a novel HE score using two large spontaneous ICH cohorts. The score included 4 items: the history of anticoagulants, HG≥2.7ml/h, GCS≤8, and NCCT signs (island sign, black hole sign, blend sign, niveau formation) exist one or more, with a total score ranging from 0 to 11. In previous study,5–11 most of the evaluation methods were comparatively simple. Wielding DCA to assess the clinical utility were rare, which lead to the overlook of harm caused by false negative. In contrast, our study was the first to systematically evaluate the score in multiple dimensions (discrimination, accuracy, clinical utility). The score was divided into the high-risk group and the low-risk group by the cut-off value of 4.5, because the incidence of HE at this point was significantly higher than the average one. It is worth mentioning that no patients with a score of 4 were found, due to the fact that NCCT signs and low GCS scores were more likely to occur when hematomas were large. When the above two coexist, there was a high probability that HG≥2.7ml/h. The specificity of score were 0.85 (development cohort) and 0.83 (validation cohort), which may help the clinical trials focused on hemostatic drugs screen out the groups that benefit most from HE targeted intervention. In comparison with nomogram, the score is simple in form and content, quick in calculation, and combines imaging features with laboratory examination, allowing clinicians to make individualized treatments with limited information.

Veltkamp12 pointed out that the history of antiplatelets was negatively correlated with the prognosis of neurological function, but it's relationship with HE was highly controversial. In our case, the history of anticoagulants was an independent risk factor for HE while the history of antiplatelets was not. The differences may be attributed to the definition of hematoma enlargement, sample size, or demographic factors. Recently, computed tomography angiography (CTA) spot sign has been externally verified and considered as a potential risk factor for HE and poor prognosis, which has been recommended by the relevant American guidelines. However, in the ATACH-II clinical trial,13 >80% of the subjects were not examined by CTA and discrimination of CTA spot sign failed to meet expected theoretical level. In addition, emergency CTA is unavailable in some primary hospitals in Asia. Apart from the problems of allergy and renal insufficiency, spot sign should be distinguished from false positive, such as oligodendroglioma and moyamoya disease. Besides, a randomized clinical trial (SPOTLIGHT)14 using CTA spot sign for hemostasis was prematurely terminated due to low enrollment rates. In contrast, NCCT is the gold standard for the diagnosis of ICH, with strong universality and simple operation. It highlights the value in clinical trials by improving patient compliance and thus increases the enrollment. In our study, NCCT signs (island sign, black hole sign, blend sign, niveau formation) exist one or more was actually a novel quantification of the degree of hematoma heterogeneity, which overcame subjectivity to a certain extent and improved the reliability of prediction. Miyahara5 thought these signs may have undergone similar pathophysiological processes: active bleeding caused by secondary vascular rupture in different phases (avalanche effect). Veltkamp12 reported that edema was an independent risk factor for HE while niveau formation was not, which contradicted our study. The reason may be that it was difficult to identify when different NCCT signs coexist and overlapped. Another possibility was the lack of standardized training for observers. Recently, Li15 found out low GCS score was one of the key predictors of HE (P<0.001). Sakuta9 suggested that compared with GCS, NIHSS may better reflect the degree of mild or moderate neurological impairment. It is indicated that GCS and NIHSS could be complementary according to patients’ conditions, though the deep connection between the above two needs to be further clarified.

Previous scores5–11 have been published, most of which contained NCCT signs, but all were single signs. Zhang16 concluded in meta-analysis that it was not recommended to use separate NCCT signs to assess the risk of HE in routine clinical practice. Considering the complexity of actual clinical work, our study optimized and improved the single sign to NCCT signs (island sign, black hole sign, blend sign, niveau formation) exist one or more, so as to maximize the clinical information reflected in the limited scoring items. The reasonable infer of the possible mechanism is that, if there are multiple NCCT signs, it implies that there may be multiple bleeding spots around the hematoma, so the risk of HE will also increase. Although some literatures9,12 believed that HE peaked within 6h from the onset, the shorter the time from the onset to the first CT scan, the more likely HE will be detected. Given that patients with mild symptoms tend to delay hospitalization, to increase the enrolled population and make the score more widely applicable, we set the enrollment standard as 24h. The BAT score7 set the time window as 6h from the onset to the first CT scan; the HEP score8 was 12h; the 9 score10 and the NAG score9 were both 24h, consistent with our study. Brouwers10 enrolled more patients with initial hematoma volume >30ml, which was similar to the development cohort of our score from senior medical center. Consistent with the validation cohort of our score, Sakuta9 reported that the majority of the study population had mild symptoms, suggesting that different models may be compatible according to different clinical characteristics. Although the NAG score9 was simpler and more portable, it did not include imaging data. Since CT is the gold standard for the diagnosis of cerebral hemorrhage, its reliability needs to be further verified. Moreover, emergency blood glucose test is not a routine project in several medical institutions, which also limits its clinical application. The HEAVN score5 reckoned as useful in prediction of HE and neural function prognosis. However, the score was slightly complicated and added more subjective items, which required more clinical data to support and brought about deviations to the objective reflection of the disease. In addition, BRAIN score11 derived from the 9-point score also had the same problems, which was not conducive to make decision of clinical strategies rapidly.

Of course, our study had some limitations. 1. The results obtained from the Chinese population were of limited generalizability to other ethnic groups and lack prospective validation. 2. Retrospective study inherently increased the risk of selection bias and the number of patients was smaller in the validation cohort than the development cohort. 3. The time from onset to initial CT scan was sometimes imprecise or even unknown (wake-up strokes). Confounding factors such as scanning machine types and scanning parameters were usually variable. 4. Patients who received initial CT scan later may have developed HE but failed to be detected, which will affect the proportion of HE occurrence. 5. The definition of HE was based on neuroimaging, so the clear relationship between score and functional prognosis was not clarified. 6. A small number of patients were excluded from the study due to early death, abandonment of treatment or emergency surgery. This group was considered to have the highest incidence of HE, excluding them contributed to underestimate its real performance. 7. Given the imparity of academic level of hospitals, patients suffering severe symptoms were more common in development cohort than in the validation cohort, which means the difference of patients’ composition might affect the performance of the score.

Via preliminarily validated externally, our score may provide some references and help for accurately identifying high-risk individuals of HE, swift guiding clinical treatments and also serving clinical trials. But prospective validation is required before score could be applied to routine clinical work.

XYK study concept and design, analysis and interpretation of data, drafting of the manuscript. WQ acquisition of data, analysis and interpretation of data, drafting of the manuscript. JD acquisition of data, analysis and interpretation of data and revision of the drafting of the manuscript. ZYQ obtaining funding, study concept and design, study supervision or coordination, revision of the drafting of the manuscript. All authors read and approved the final manuscript.

This work was supported by the Technology Development Project of Soochow (SYSD2018102) and Postgraduate Practice Innovation Program of Jiangsu Province (SJCX18_0855).

The fund body took no part in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.