Predicting Aneurysm Rupture

Transcription

1 Predicting Aneurysm Rupture Robert E. Harbaugh, MD, FAANS, FACS, FAHA Director, Penn State Institute of the Neurosciences University Distinguished Professor & Chair, Department of Neurosurgery Professor, Department of Engineering Science & Mechanics Penn State University - M.S. Hershey Medical Center

2 Predicting Aneurysm Rupture Many intracranial aneurysms are found incidentally Imaging studies % of population Autopsy studies - 1-9% of population Aneurysms may remain asymptomatic, but if they rupture there is a 30-50% one month mortality All aneurysm treatments carry a risk of neurological injury and death How can we choose only those aneurysms with elevated risk of rupture for treatment?

3 ISUIA Aneurysm Size and Annual Rupture Rates - Prospective Data Aneurysm Size Group 1 (No SAH) Group 2 (Previous SAH) 2-6 mm ~0.1% 1/1000/year ~0.5% 1/200/year 7-9 mm ~0.5% 1/200/year ~0.5% 1/200/year mm 7% in year #1 then 1%/year 1.0% 1/100/year 25+ mm 17% in year #1 then 2%/year No data

4 ISUIA Aneurysm Size and Risk of Rupture: Multivariate Analysis of the Combined Cohort Size vs 2-6 mm RR P Value 7-9 mm mm 5.9 < mm 9.7 < mm 27.6 <0.0001

5 ISUIA Aneurysm Location and Risk of Rupture Multivariate Analysis of the Combined Cohort Site vs ICA RR P Value Cavernous MCA PCoA Vertebrobasilar Basilar Apex

6 Predicting Rupture Risk factor Key findings Investigators Size Larger UIAs have greater RR Aneurysm size is an independent predictor of RR Defining a critical size threshold remains difficult ISUIA investigators Wiebers et al Ishibashi et al Enlargement IAs are often larger at time of rupture than at diagnosis Larger UIAs are more likely to grow Larger UIAs greater growth risk increased RR Yasui et al Burns et al Matsubara et al Previous SAH Prior history of aneurysmal SAH increases RR Aneurysms <7mm have greater RR with prior history of SAH ISUIA investigators Wiebers et al Location Posterior circulation aneurysms are more hazardous Intracavernous IAs are more benign Weir et al Wermer et al Kupersmith et al Patient Age, sex and co-morbidities influence aneurysmal RR Female sex and cigarette smoking are independent predictors of aneurysm formation, growth and rupture Nahed et al Juvela et al Morphology Multiple lobulations or loculations increase RR High dome:neck ratio increases RR Aneurysm angle from parent vessel is a predictor of rupture Quantified irregular aneurysm shape is a predictor of rupture Aneurysm shape determines hemodynamic stress and is associated with biological behavior of aneurysm wall Hademenos et al Beck et al Dhar et al Harbaugh, Raghavan et al. 2004, 2005 Raghavan, Harbaugh, Laaksamo et al. 2007, 2010, 2012

7 Predicting Rupture: Aneurysm Geometry Size (maximum dimension) is most commonly used Easy to quantify and objective Is shape also an important factor? Difficult to quantify and subjective Single-lobe vs. multi-lobular Neck-to-height ratio Ratio of neck to maximum diameter Regular vs. irregular

8 Quantifying Geometry: Overview Acquire 3-D digital data from CTA/MRA/DSA Develop algorithms for surface mesh refinement Quantify aneurysm size, including aneurysm volume and surface area Quantify aneurysm shape using various indices

9 Order of Predictive Capabilities Order ROC deviation from null Index Type p value p < Isoperimetric Ratio 3D, shape TRUE Convexity Ratio 3D, shape TRUE Mean Curvature 3D, shape TRUE Aspect Ratio 2D, shape TRUE Neck Diameter 1D, size FALSE Bottleneck Factor 2D, shape FALSE Bulge Location 2D, shape FALSE Height 1D, size FALSE Volume 3D, size FALSE Maximum Diameter 1D, size 0.91 FALSE Surface Area 2D, size FALSE

10 Summary: Geometry and Hemodynamics Shape indices are better correlated with the ruptured state than size indices All 3D shape indices and one 2D shape index showed statistically significant differences between the ruptured and unruptured groups No size indices show statistically significant differences Aneurysm geometry determines aneurysm hemodynamics Pressure is the dominant hemodynamic load on the aneurysm - shear stress is no more than 1% of pressure load. The maximum shear stress value can be larger than that regarded to cause endothelial damage. Do areas of low shear stress result in wall thinning and areas of high shear stress result in atherosclerotic change?

11 Next Steps to Determine Best Treatment What is the best option for this patient with this aneurysm in my hands? Randomized Controlled Trial? Decision Analysis? Registry with Propensity Analysis?

12 Methodological Considerations - RCT Patients are evaluated by cerebrovascular specialists In some, treatment is recommended. Why? In others, surveillance is recommended. Why? Do these two groups have the same natural history? One explanation of the ISUIA results is that evaluators select low rupture risk aneurysms for observation. Would an RCT of observation versus clipping or coiling yield reliable results?

13 Problems with an RCT for UIAs Lack of equipoise 40 year old woman, positive FH of aneurysm rupture, cigarette smoker, 11 mm, irregular, basilar apex aneurysm - would you randomize? 65 year old woman, no FH of aneurysm rupture, non-smoker, 7 mm, regular, ophthalmic artery aneurysm - would you randomize? Duration of study 2 years? 5 years? 15 years?

14 Decision Analysis Modeling Patients start out neurologically well at age 40 Natural history annual rupture rate is 1.46% Clipping has 11.2% morbidity/mortality, decreases risk of hemorrhage by 95% Coiling has 5.6% morbidity/mortality, decreases risk of hemorrhage by 75% Actuarial risks from U.S. Health Statistics Standard discount rate for later years of life QALYs assigned via Monte Carlo method

15 Decision Analysis Modeling One year from entry Observe QALY Coil QALY Clip QALY Five years from entry Coil QALY Clip QALY Observe QALY Lifetime Clip QALY Coil QALY Observe QALY Crossover point for clipping vs. coiling is 10.5 years How reliable are the data on which the model is based?

16 The Power of the Registries Registries are cost-effective, easily scaled and document care in real world environments involving a variety of practices. Propensity analysis can be used to infer causal relationships and perform comparative effectiveness research if the registries are properly designed

17 Problems with Registries Registries must be properly designed, audited and maintained Infrastructure is needed for data collection and management Interdisciplinary registries may be needed for CER Long term follow-up across practice sites may be needed

18 Designing Registries for Propensity Analysis What randomized experiment do we want to model? Who are the decision makers for treatment assignment? What are the key covariates used to assign treatment? Can we measure the key covariates well? What clinically meaningful outcomes will we measure? What sample sizes will be needed? If we can address the issues listed above we will be able to draw a reliable causal inference from registry data. This may correct some of the inadequacies of the present EBM algorithm.

19 Designing a Registry for UIAs What randomized experiment do we want to model? Observation vs invasive treatment for patients with UIAs Who are the decision makers for treatment assignment? Physicians, patients and family members What key covariates do they use to decide? Patient-specific factors (patient age, prior aneurysm rupture, co-morbidities, social history, family history, patient preferences) aneurysm-specific factors (aneurysm size, shape and location) and physician-specific factors (endovascular specialist, open surgical specialist, both, neither, years of experience, practice setting) Are the key covariates well measured? Define and quantify key covariates What are the clinically meaningful outcomes we want to measure Mortality, aneurysm rupture, functional health status, QOL What sample sizes will be needed? Traditional power calculations

20 Thank You for Your Attention Questions?