The present study demonstrated that adding AF to the MSS score increased the predictive value of the MSS score for ICH. We developed and internally validated a nomogram based on AF and the MSS score to predict the probability of ICH in stroke patients treated with alteplase. The new nomogram showed a significantly higher predictive accuracy than the conventional scoring system of the MSS score.
Among the computational models for predicting prognosis, the nomogram is very useful because it is a pictorial representation of a statistical predictive model that generates a numerical probability of a clinical event. It is more accurate than the conventional method using OR . Therefore, we constructed a nomogram that can calculate the probability of ICH for an individual stroke patients undergoing thrombolysis. The parameters constructed in our model are easily available in almost all medical centers and all patients within few minutes of their arrival to the emergency room.
To the best of our knowledge, there have only 2 works carried out on nomograms for individualized prediction of the ICH probability in acute ischemic stroke patients undergoing thrombolysis [11,12]. The STARTING-SICH nomogram including 10 variables was designed to predict sICH in stroke patients treated with intravenous thrombolysis in a large cohort study of Italy , but it has not been external validated in Asian patients and the data on ethnicity are lacking. The other nomogram model including 3 variables (present of AF, NIHSS score and glucose level on admission) was developed in Asian patients . However, the study did not include the information about the total dose of r-PA, and the risk of ICH is reported higher in Asian populations at standard doses .
Our nomogram used only two prognostic factors including AF and an existing scoring system (the MSS score). The MSS score is widely used. History of AF is easily and readily obtainable during the patient’s admission at the hospital. Our nomogram was the first approach to combine AF and the MSS score to predict ICH in stroke patients. This combination approach had more accurate predictive power than the MSS score alone. Researchers recently compared different ICH risk scores and found that ORs based on logistic models and AUC-ROC values for the MSS scores showed improved performance, with values ranging from 0.63 to 0.86 [5,6,8]. The present study demonstrated that the AUC-ROC value for the MSS score was 0.741 for ICH, and the predictive value increased significantly to 0.794 when the MSS score was combined with AF. This easy-to-use simultaneous testing model that adding AF to the MSS score, with noninvasive clinical characteristics, can provide an immediate and reliable estimation of ICH risk in acute ischemic stroke patients who require thrombolysis. This estimate will guide clinicians not only in counseling patients and/or families but also in the early identification of those patients at high risk of ICH as well as support decisions regarding additional treatments or more attention.
Early ischemic signs on CT or even hyperdense cerebral artery signs are difficult to interpret and require experienced personnel or experts to evaluate, as reported by Thanin et al . Therefore, some scoring systems, including the MSS score , Safe Implementation of Thrombolysis in Stroke (SITS)-SICH score , Glucose Race Age Sex Pressure Stroke Severity (GRASPS) score  and Stroke Prognostication using Age and National Institutes of Health (NIH) Stroke Scale (SPAN-100) positive index , have no CT component in their scoring systems. Most scoring systems are derived from Western countries that might have different sets of prognostic parameters. The MSS score was derived from a North American and European study within a 3-hour time window, whereas the present study with a Chinese study employed a 4.5-hour time window. A previous study reported that the MSS score could predict sICH (ECASSII definition) with an AUC of 0.730 in Chinese stroke patients . The AUC for the MSS score alone was 0.741 for ICH in the present study; therefore, we speculated that this result might be due to some potential risk factors not fully captured by the scoring system when external validated in Chinese population.
Studies have reported additional predictive factors for postthrombolysis ICH, including leukoaraiosis, high mean blood pressure, low serum albumin and the neutrophil to lymphocyte ratio [19-22]. However, these identified risk factors accounted for only a proportion of the stroke patients who presented ICH after intravenous thrombolysis. Growing evidence supports that AF is an independent risk factor for ICH events [9,10,23]. However, AF has not been considered in previous risk scoring systems for ICH after thrombolysis [4,7,17,18,24,25]. Yeo et al reported a scoring system using nomogram based on three variables (presence of AF, glucose level and NIHSS score) was a practical tool to predict the risk of ICH after thrombolysis . In our study, the AF prevalence at baseline was higher in patients with ICH and was an independent risk factor for ICH in stroke patients undergoing thrombolysis. Furthermore, we found that AF and the MSS score were correlated, and the patients with AF had higher MSS score. Moreover, adding AF to the MSS score on admission enhanced the predictive value of ICH for stroke patients with thrombolysis.
We included both symptomatic and asymptomatic postthrombolysis ICH as the outcome for the scoring systems, as in previous studies [7,12]. Many reports have demonstrated that both symptomatic and asymptomatic ICH may worsen clinical outcomes [7,26,27], and influence the timing of reintroducing antithrombotic treatment after r-tPA treatment. Furthermore, predicting a higher risk of ICH preceding intravenous thrombolysis may help clinical decision making by slanting treatment toward only mechanical thrombectomy without intravenous r-tPA .
There are some limitations of the study. First, the present study used a retrospective design, so some confounders were not available for inclusion in our multivariate analyses. Second, the study included a single center based sample and a relatively small sample size, which might have limited the statistical power of the results. Finally, although we internally validated our model using bootstrap resampling, our model has not been validated in external cohorts.