The next paper will be presented by Dr. Akio Morita, entitled Modified Risk Prediction Model for the Open Treatment of Unruptured Cerebral Aneurysms in Japan, UCAS Treat Result and Model Verification Using UCAS II Outcome. The paper will be discussed by myself and Dr. Seppi Amin-Hanjani. Thank you. Thank you, Chairman, and good morning, ladies and gentlemen. I'm very honored and privileged to be here to present our study. I'm going to talk about Modified Risk Prediction Model for the Open Treatment of Unruptured Cerebral Aneurysms from UCAS Japan Study. I have no COI. Management decision of unruptured cerebral aneurysms should weigh between rupture risk and the management risk and overall patient comorbidity and quality of life. For that purpose, if there is a prediction model, it should support us for the clinical decision-making. According to the tripod statement, there are three types of prediction model. Type one is just internal validation. Type two is make a separate set of data from the same cohort. And type three is external validation. For example, phases score is very important risk prediction model for the rupture, but we did not have external data for this. So this is a type one B prediction model. And on the other hand, we created our own Japanese prediction model from the UCAS Japan Study, then validated it from completely outside data from Japanese cohort. This is type three prediction model. For that treatment-wise, there's no good prediction model for open craniotomy surgery for unruptured aneurysm. So we want to create that from UCAS Japan Study outcome data. Out of 5,720 cases, we had 2,316 cases underwent open craniotomy and clipping. We defined the morbidity as a decline of modified ranking scale two or below. And the factors with p-value less than 0.05 with multivariate model was considered significant. To create prediction model, we incorporated clinically important factors, even p was 0.01. And the prediction model was assessed, validated by internal data and UCAS two data, which is a completely different cohort. Both UCAS Japan and UCAS two is prospective data. And UCAS Japan is a national study incorporating 215 institutions. And the data was entered 2001 to 2004 for the initial data. UCAS two is a volunteer study from 31 institutes which belong to the UCAS Japan Study. And data entry was only for 2006. So this is a completely separate data set. Baseline characteristics of both two data sets are very similar, except for size. Very small aneurysms are less treated in UCAS two. But the medium-sized aneurysms to 5 to 9 millimeter were treated more in the UCAS two group. For the location-wise, MCA and anterior communicating artery aneurysms were treated more in the UCAS two group. But for the vascular and the vertebral aneurysms were less treated, maybe because they are converted to endovascular procedures. As a result, from UCAS Japan Study, that the mobility was related with size more than 7 millimeter and the location of the vascular artery. And if the aneurysms associated with this data set, it favored the outcome. However, the patient factor-wise, age more than 70 years old, and hypertension, diabetes, and the initial modified Rankine scale is poorer, it affects the outcome. And if you do double or three to three aneurysms at the same time, that mobility will be doubled. So for the open procedure size, location, hypertension, diabetes, age, initial modified Rankine scale, and the multiple treatment at one session affected the outcome. But hospital volume, case volume, did not affect for open surgery in Japan. On the other hand, for endovascular procedures, we have very small number of cases at that time. Mobility was similar, 3%. But the factor associated with outcome was only diabetes and initial modified Rankine scale and hospital volume. High volume center had a much better outcome than low volume centers. However, on the other side, size, location, age, which are very important factor for open procedure did not affect endovascular procedures. So if you want to have a prediction model for endovascular, we need to have a separate one. So we can allocate scores for each eight factors. And the factors can be scored or obtained from coefficient from the odds ratio of multivariate model. Then we can sum up the score. If the score was 6 for the patient, it will be open procedure would have 2.8% mobility. We can speculate that. So how about the validation? Internal validation worked pretty well. C-statics was 0.81. However, for the external validation, unfortunately and unexpectedly to us, it did not validate. UCAS 2 cannot be validated that prediction model. C-statics was 0.56. It means 50-50 chance this prediction model was correct. We tried to assess why we failed. First problem would be we had very few number of cases for validation. We should have at least one-third of the case volume for the original UCAS Japan data. Also, another factor would be treatment outcome can be different by the surgical standard. Even though we have the same institution, UCAS Japan and UCAS 2 are five years apart. Surgical standard might have been changed and there might be unknown systematic causes. This is a semantic explanation. If you have a different time frame or different country or different institution, the surgical standard can be different, then outcome can be different. So it will be very difficult to create universally applicable simple prediction model. To overcome this, we need to create combined cohort, international cohort with standard data set. Standardized data should be very important. Then we can extract that, then we can create at least a type 2 prediction model. Even type 1 prediction model supports our clinical decision making. For example, this 65-year-old male came with 9-millimeter ECOM annulation with DOTASAC. He has a history of hypertension. From our study, that prediction showed the three-year rupture risk will be 5.7%. On the other hand, phases score showed the five-year rupture risk is 7.2%. So here we have 1.5 to 2% annual rupture risk. On the other hand, what's the outcome for the treatment? There's no prediction model so far. But our data can show to the patient it will be scored 6, so mobility will be close to 2.8%. Then we can show to the patient, discuss, we can just say this is indicative data. Then we can decide. Unfortunately, this patient chose not to be treated, but he got a rupture two years later. The limitation and conclusion, tripod type 3 prediction model could not be created from our data. Instead, tripod type 1 prediction model was supplied, and such prediction model should support communication with patient and the shared decision making. For the endovascular procedures, we need a separate prediction model because the factors for outcome are different. For the future, we need systematically collected large-scale data with standardized data set in the various occasions in various countries to make better and reliable prediction model. Thank you very much for your attention. Thank you, Dr. Morita. The UCAS investigators built a risk prediction model for the morbidity of open repair of unruptured aneurysms, and then tested their model in a separate, similarly treated patient group enrolled two to six years later. On internal validation, shown in the left panel, using 65 events in 2,300 treated patients, the model performed well. Predicted and observed risk were in good agreement, with data points tightly clustered along the diagonal. In the second cohort, the validation cohort, shown on the right, using 17 events in 452 patients, predicted and observed risk were almost unrelated. The authors should be commended for testing their model in this most rigorous fashion. This failure of validation could reflect three different problems. The validation set could be too small to be a fair test of the model. The true risk of surgery could have changed as more patients had endovascular treatment. Or the model itself could be unstable. With 65 events and 12 variables tested, the number of events per predictor in model building here was about five. About 10 to 25 are usually recommended to build a stable model. Further research will be needed to explain these findings. So with this paper, Marita and colleagues have reported data that's really relevant to us for the most challenging aspect of decision making and management of patients with unruptured intracranial aneurysms, namely the balance of the risk of future rupture versus the risk of treatment. Although we have prediction scoring systems for rupture risk in recent years derived from large prospective cohorts, scoring systems for treatment risks have primarily been based on single institution small experiences. So with this study, the UCAS team has turned their attention to predicting the treatment risk side of the equation in a large multicenter prospective cohort of over 2,000 patients in Japan. The take home points from this, if we can advance the slides, is that the overall morbidity of the procedure in this series was 3 to 4%, which is an excellent outcome relative to prior reports and a significant improvement relative to the international study of unruptured intracranial aneurysms, the issue of study which was published some 15 years ago and demonstrated a surgical morbidity closer to 8%. Additionally, the most predictive factors seen here in this data were aneurysm size, location, initial Rankin score, hypertension, diabetes, and multiple treatments as predictive of morbidity. Although they did not find aneurysm case volume to be predictive of outcome, a finding which is not parallel to that seen in other prior studies, although this could be due to the different cutoff threshold of greater than 60 aneurysm cases used here as compared to a threshold of about 20 to 30 annual case volumes seen in prior volume outcome data. The final model failed external validation and thus still needs additional data, including from large additional populations. And as they noted, a new model would be required to predict the morbidity of endovascular management. The remaining and unfortunate challenge for us remains that the same risk factors which appear to lead to increased risk of morbidity are unfortunately the same ones that can be associated with an increased risk of rupture, such that patients most in need of treatment are those that are prone to higher treatment risk also. I congratulate the authors on their ongoing efforts to better define these risk prediction tools for unruptured aneurysms, a task that we should also aspire to undertake with similar vigor here in the U.S. Thank you.

Dr. Akio Morita

Modified Risk Prediction Model

Open Treatment

Unruptured Cerebral Aneurysms

UCAS Japan Study