Automated online deconjugation of antibody-drug conjugate for small molecule drug profiling

Antibody-drug conjugates (ADCs) are designed by chemically linking highly potent cytotoxic small molecule drugs to monoclonal antibodies of unique specificity for targeted destruction of cancer cells. This innovative class of molecules incurs unique developmental challenges due to its structural complexity of having both small molecule and protein components. The stability of the small molecule payload on the ADC is a critical attribute as it directly relates to product efficacy and patient safety. This study describes the use of an end-to-end automated workflow for effective and robust characterization of the small molecule drug while it is conjugated to the antibody. In this approach, online deconjugation was accomplished by an autosampler user defined program and 1D size exclusion chromatography was utilized to provide separation between small molecule and protein species. The small molecule portion was then trapped and sent to the 2D for separation and quantification by reversed-phase liquid chromatography with identification of impurities and degradants by mass spectrometry. The feasibility of this system was demonstrated on an ADC with a disulfide-based linker. This fully automated approach avoids tedious sample preparation that may lead to sample loss and large assay variability. Under optimized conditions, the method was shown to have excellent specificity, sensitivity (LOD of 0.036 µg/mL and LOQ of 0.144 µg/mL), linearity (0.04-72.1 µg/mL), precision (system precision %RSD of 1.7 and method precision %RSD of 3.4), accuracy (97.4 % recovery), stability-indicating nature, and was successfully exploited to analyze the small molecule drug on a panel of stressed ADC samples. Overall, the workflow established here offers a powerful analytical tool for profiling the in-situ properties of small molecule drugs conjugated to antibodies and the obtained information could be of great significance for guiding process/formulation development and understanding pharmacokinetic/pharmacodynamic behavior of ADCs.

Combined PD-L1/TGFβ blockade allows expansion and differentiation of stem cell-like CD8 T cells in immune excluded tumors

TGFβ signaling is associated with non-response to immune checkpoint blockade in patients with advanced cancers, particularly in the immune-excluded phenotype. While previous work demonstrates that converting tumors from excluded to inflamed phenotypes requires attenuation of PD-L1 and TGFβ signaling, the underlying cellular mechanisms remain unclear. Here, we show that TGFβ and PD-L1 restrain intratumoral stem cell-like CD8 T cell (TSCL) expansion and replacement of progenitor-exhausted and dysfunctional CD8 T cells with non-exhausted T effector cells in the EMT6 tumor model in female mice. Upon combined TGFβ/PD-L1 blockade IFNγhi CD8 T effector cells show enhanced motility and accumulate in the tumor. Ensuing IFNγ signaling transforms myeloid, stromal, and tumor niches to yield an immune-supportive ecosystem. Blocking IFNγ abolishes the anti-PD-L1/anti-TGFβ therapy efficacy. Our data suggest that TGFβ works with PD-L1 to prevent TSCL expansion and replacement of exhausted CD8 T cells, thereby maintaining the T cell compartment in a dysfunctional state.

Reduced Fasting Duration in Cynomolgus Monkeys Enhances Animal Welfare During Toxicology Studies

During toxicology studies, fasting animals prior to clinical pathology blood collection is believed to reduce variability in some clinical chemistry analytes. However, fasting adds stress to animals that are already stressed from the administration of potentially toxic doses of the test article. The purpose of this study was to assess the impacts of different fasting durations on cynomolgus monkeys' welfare during toxicology studies. To this end, we assessed the cynomolgus monkeys traditional and ancillary clinical pathology endpoints at different fasting times. We showed that most clinical pathology endpoints were largely comparable between different fasting times suggesting that cynomolgus monkeys could be fasted for as little as 4 hours for toxicology studies, as longer fasting times (up to 20 hours) resulted in stress, dehydration, and significant decreases in blood glucose- changes that impacts animal welfare. Shorter fasting times were associated with higher triglycerides variability among individual animals. Therefore, we propose that shorter fasting time (i.e., 4 hours) should be adequate for most toxicology studies except when: (1) parameters that could be affected by non-fasting conditions are important for safety and pharmacodynamic assessments (i.e., glucose and lipids) and (2) fasting would be needed for the bioavailability of an orally administered test article.

Statistical considerations for design and analysis of stability, comparability and formulation tests

This article considers designed experiments for stability, comparability, and formulation testing that are analyzed with regression models in which the degradation rate is a fixed effect. In this setting, we investigate how the number of lots, the number of time points and their locations affect the precision of the entities of interest, leverages of the time points, detection of non-linearity and interim analyses. This investigation shows that modifying time point locations suggested by ICH for stability studies can significantly improve these objectives. In addition, we show that estimates of precision can be biased when a regression model that assumes independent measurements is used in the presence of within-assay session correlation. This bias can lead to longer shelf life estimates in stability studies and loss of power in comparability studies. Mixed-effect models that take into account within-assay session correlation are shown to reduce this bias. The findings in this article are obtained from well known statistical theory but provide valuable practical advice to scientists and statisticians designing and interpreting these types of experiments.

Statistical modeling approaches for the comparison of dissolution profiles

Dissolution studies are a fundamental component of pharmaceutical drug development, yet many studies rely upon the f 1 and f 2 model-independent approach that is not capable of accounting for uncertainty in parameter estimation when comparing dissolution profiles. In this paper, we deal with the issue of uncertainty quantification by proposing several model-dependent approaches for assessing the similarity of two dissolution profiles. We take a statistical modeling approach and allow the dissolution data to be modeled using either a Dirichlet distribution, gamma process model, or Wiener process model. These parametric forms are shown to be reasonable assumptions that are capable of modeling dissolution data well. Furthermore, based on a given statistical model, we are able to use the f 1 difference factor and f 2 similarity factor to test the equivalency of two dissolution profiles via bootstrap confidence intervals. Illustrations highlighting the success of our methods are provided for both Monte Carlo simulation studies, and real dissolution data sets.

First-In-Human Phase I Study of the OX40 Agonist MOXR0916 in Patients with Advanced Solid Tumors

PURPOSE

OX40, a receptor transiently expressed by T cells upon antigen recognition, is associated with costimulation of effector T cells and impairment of regulatory T-cell function. This first-in-human study evaluated MOXR0916, a humanized effector-competent agonist IgG1 monoclonal anti-OX40 antibody.

PATIENTS AND METHODS

Eligible patients with locally advanced or metastatic refractory solid tumors were treated with MOXR0916 intravenously once every 3 weeks (Q3W). A 3+3 dose-escalation stage (0.2-1,200 mg; n = 34) was followed by expansion cohorts at 300 mg (n = 138) for patients with melanoma, renal cell carcinoma, non-small cell lung carcinoma, urothelial carcinoma, and triple-negative breast cancer.

RESULTS

MOXR0916 was well tolerated with no dose-limiting toxicities observed. An MTD was not reached. Most patients (95%) experienced at least one adverse event (AE); 56% of AEs, mostly grade 1-2, were related to MOXR0916. Most common treatment-related AEs included fatigue (17%), diarrhea (8%), myalgia (7%), nausea (6%), decreased appetite (6%), and infusion-related reaction (5%). Pharmacokinetic (PK) parameters were dose proportional between 80 and 1,200 mg and supported Q3W administration. The recommended expansion dose based on PK and OX40 receptor saturation was 300 mg Q3W. Immune activation and upregulation of PD-L1 was observed in a subset of paired tumor biopsies. One renal cell carcinoma patient experienced a confirmed partial response. Overall, 33% of patients achieved stable disease.

CONCLUSIONS

Although objective responses were rarely observed with MOXR0916 monotherapy, the favorable safety profile and evidence of tumor immune activation in a subset of patients support further investigation in combination with complementary agents such as PD-1/PD-L1 antagonists.

Bimodal expression of potential drug target CLL-1 (CLEC12A) on CD34+ blasts of AML patients

Objectives

This study aims to retrospectively assess C‐lectin‐like molecule 1 (CLL‐1) bimodal expression on CD34+ blasts in acute myeloid leukemia (AML) patients (total N = 306) and explore potential CLL‐1 bimodal associations with leukemia and patient‐specific characteristics.

Methods

Flow cytometry assays were performed to assess the deeper immunophenotyping of CLL‐1 bimodality. Cytogenetic analysis was performed to characterize the gene mutation on CLL‐1‐negative subpopulation of CLL‐1 bimodal AML samples.

Results

The frequency of a bimodal pattern of CLL‐1 expression of CD34⁺ blasts ranged from 8% to 65% in the different cohorts. Bimodal CLL‐1 expression was most prevalent in patients with MDS‐related AML (P = .011), ELN adverse risk (P = .002), NPM1 wild type (WT, P = .049), FLT3 WT (P = .035), and relatively low percentages of leukemia‐associated immunophenotypes (P = .006). Additional immunophenotyping analysis revealed the CLL‐1⁻ subpopulation may consist of pre‐B cells, immature myeloblasts, and hematopoietic stem cells. Furthermore, (pre)‐leukemic mutations were detected in both CLL‐1⁺ and CLL‐1⁻ subfractions of bimodal samples (N = 3).

Conclusions

C‐lectin‐like molecule 1 bimodality occurs in about 25% of AML patients and the CLL‐1⁻ cell population still contains malignant cells, hence it may potentially limit the effectiveness of CLL‐1‐targeted therapies and warrant further investigation.

Validating Container Closure Integrity with Statistically Based Tests

Two statistically based testing procedures, a zero-failure test and a one-failure test, are proposed for demonstrating that a vial capping process has an acceptably low rate of failed seals, i.e., non-integral container closures, or leaky vials. These tests are developed for use with the standard helium-leak test method that measures the amount of escaped helium from a capped vial. The amount of escaped helium is a continuous measurement and a vial is said to be leaky if the measurement exceeds a threshold and not leaky if it does not. Not leaky measurements are often less than the lower limit of validation, i.e., left-censored. By using the continuous measurements that describe the extent of leaking instead of the binary measurements, leaky or not leaky, the proposed tests are able to reach similar conclusions as tests using binary measurements but with much smaller sample sizes. The proposed tests can handle any number of measurements less than the lower limit of validation.

The Utility of Novel Urinary Biomarkers in Mice for Drug Development Studies

Novel urinary protein biomarkers have recently been identified and qualified in rats for the early detection of renal injury in drug development studies. However, there are few reports on the utility of these renal biomarkers in mice, another important and widely used preclinical animal species for drug development studies. The purpose of this study was to assess the value of these recently qualified biomarkers for the early detection of drug-induced kidney injury (DIKI) in different strains of mice using multiple assay panels. To this end, we evaluated biomarker response to kidney injury induced by several nephrotoxic agents including amphotericin B, compound X, and compound Y. Several of the biomarkers were shown to be sensitive to DIKI in mice. When measured, urinary albumin and neutrophil gelatinase-associated lipocalin were highly sensitive to renal tubular injury, regardless of the assay platforms, mouse strain, and nephrotoxic agents. Depending on the type of renal tubular injury, kidney injury molecule-1 was also highly sensitive, regardless of the assay platforms and mouse strain. Osteopontin and cystatin C were modestly to highly sensitive to renal tubular injury, but the assay type and/or the mouse strain should be considered before using these biomarkers. Calbindin D28 was highly sensitive to injury to the distal nephron in mice. To our knowledge, this is the first report that demonstrates the utility of novel urinary biomarkers evaluated across multiple assay platforms and nephrotoxicants in different mice strains with DIKI. These results will help drug developers make informed decisions when selecting urinary biomarkers for monitoring DIKI in mice for toxicology studies.

Phase I Study of the Indoleamine 2,3-Dioxygenase 1 (IDO1) Inhibitor Navoximod (GDC-0919) Administered with PD-L1 Inhibitor (Atezolizumab) in Advanced Solid Tumors

PURPOSE

IDO1 induces immune suppression in T cells through l-tryptophan (Trp) depletion and kynurenine (Kyn) accumulation in the local tumor microenvironment, suppressing effector T cells and hyperactivating regulatory T cells (Treg). Navoximod is an investigational small-molecule inhibitor of IDO1. This phase I study evaluated safety, tolerability, pharmacokinetics, and pharmacodynamics of navoximod in combination with atezolizumab, a PD-L1 inhibitor, in patients with advanced cancer.

PATIENTS AND METHODS

The study consisted of a 3+3 dose-escalation stage (n = 66) and a tumor-specific expansion stage (n = 92). Navoximod was given orally every 12 hours continuously for 21 consecutive days of each cycle with the exception of cycle 1, where navoximod administration started on day -1 to characterize pharmacokinetics. Atezolizumab was administered by intravenous infusion 1,200 mg every 3 weeks on day 1 of each cycle.

RESULTS

Patients (n = 157) received navoximod at 6 dose levels (50-1,000 mg) in combination with atezolizumab. The maximum administered dose was 1,000 mg twice daily; the MTD was not reached. Navoximod demonstrated a linear pharmacokinetic profile, and plasma Kyn generally decreased with increasing doses of navoximod. The most common treatment-related AEs were fatigue (22%), rash (22%), and chromaturia (20%). Activity was observed at all dose levels in various tumor types (melanoma, pancreatic, prostate, ovarian, head and neck squamous cell carcinoma, cervical, neural sheath, non-small cell lung cancer, triple-negative breast cancer, renal cell carcinoma, urothelial bladder cancer): 6 (9%) dose-escalation patients achieved partial response, and 10 (11%) expansion patients achieved partial response or complete response.

CONCLUSIONS

The combination of navoximod and atezolizumab demonstrated acceptable safety, tolerability, and pharmacokinetics for patients with advanced cancer. Although activity was observed, there was no clear evidence of benefit from adding navoximod to atezolizumab.

Investigating the Value of Urine Volume, Creatinine, and Cystatin C for Urinary Biomarkers Normalization for Drug Development Studies

Novel urinary protein biomarkers have recently been identified and qualified in rats for the early detection of renal injury in drug development studies. However, there seems to be no standardized normalization method for analyzing these urinary biomarkers, as some users normalize with urinary creatinine (uCr), urine volume (uVol), or leave biomarker un-normalized. More recently, urinary cystatin C is also emerging as a urinary biomarker normalizer, given some of its characteristics as a glomerular filtration marker. The purpose of this study was to identify an optimal drug-induced kidney injury biomarker normalization method that can be adopted more uniformly in the field. To this end, we compared the variability of uVol, urinary cystatin C, and Cr in healthy rats; we evaluated the sensitivity of the renal biomarkers to renal injury after normalization with uVol, uCr, and cystatin C in rats with cisplatin-induced renal injury. We showed that, over time, uCr was less variable than urinary cystatin C and uVol. When the renal biomarkers were normalized with the 3 normalizing end points, the biomarkers showed (1) least variability following normalization with Cr in healthy animals and (2) poor sensitivity when normalized with urinary cystatin C in animals with renal injury. Overall, the results suggested that uCr is better than urinary cystatin C and uVol for normalizing renal biomarkers in rats under controlled preclinical conditions. To our knowledge, this is the first report that compared the variability of uVol, cystatin C, and Cr in the context of renal biomarkers' normalization.

Phase I study of the anti-α5β1 monoclonal antibody MINT1526A with or without bevacizumab in patients with advanced solid tumors

PURPOSE

MINT1526A is a monoclonal antibody that blocks the interaction of integrin alpha 5 beta 1 (α5β1) with its extracellular matrix ligands. This phase I study evaluated the safety and pharmacokinetics of MINT1526A with or without bevacizumab in patients with advanced solid tumors.

METHODS

MINT1526A was administered every 3 weeks (Q3W) as monotherapy (arm 1) or in combination with bevacizumab 15 mg/kg, Q3W (arm 2). Each arm included a 3 + 3 dose-escalation stage and a dose-expansion stage.

RESULTS

Twenty-four patients were enrolled in arm 1 (dose range 2-30 mg/kg) and 30 patients were enrolled in arm 2 (dose range 3-15 mg/kg). Monocyte α5β1 receptor occupancy was saturated at a dose of 15 mg/kg. No dose-limiting toxicities were observed, and the maximum tolerated dose was not reached in either arm. The most common adverse events, regardless of causality, included abdominal pain (25%), diarrhea (25%), nausea (21%), vomiting (21%), and fatigue (21%) in arm 1 and nausea (40%), fatigue (33%), vomiting (30%), dehydration (30%), headache (30%), and hypertension (30%) in arm 2. No grade ≥ 3 bleeding events were observed in either arm. No confirmed partial responses (PR) were observed in arm 1. In arm 2, one patient with thymic carcinoma experienced a confirmed PR and two patients with hepatocellular carcinoma (HCC) experienced durable minor radiographic responses.

CONCLUSIONS

MINT1526A, with or without bevacizumab, was well-tolerated. Preliminary evidence of combination efficacy, including in patients with HCC, was observed, but cannot be distinguished from bevacizumab monotherapy in this phase I study.

Cerebellar arteriovenous malformations: anatomic subtypes, surgical results, and increased predictive accuracy of the supplementary grading system

BACKGROUND

Anatomic diversity among cerebellar arteriovenous malformations (AVMs) calls for a classification that is intuitive and surgically informative. Selection tools like the Spetzler-Martin grading system are designed to work best with cerebral AVMs but have shortcomings with cerebellar AVMs.

OBJECTIVE

To define subtypes of cerebellar AVMs that clarify anatomy and surgical management, to determine results according to subtypes, and to compare predictive accuracies of the Spetzler-Martin and supplementary systems.

METHODS

From a consecutive surgical series of 500 patients, 60 had cerebellar AVMs, 39 had brainstem AVMs and were excluded, and 401 had cerebral AVMs.

RESULTS

Cerebellar AVM subtypes were as follows: 18 vermian, 13 suboccipital, 12 tentorial, 12 petrosal, and 5 tonsillar. Patients with tonsillar and tentorial AVMs fared best. Cerebellar AVMs presented with hemorrhage more than cerebral AVMs (P < .001). Cerebellar AVMs were more likely to drain deep (P = .04) and less likely to be eloquent (P < .001). The predictive accuracy of the supplementary grade was better than that of the Spetzler-Martin grade with cerebellar AVMs (areas under the receiver-operating characteristic curve, 0.74 and 0.59, respectively). The predictive accuracy of the supplementary system was consistent for cerebral and cerebellar AVMs, whereas that of the Spetzler-Martin system was greater with cerebral AVMs.

CONCLUSION

Patients with cerebellar AVMs present with hemorrhage more often than patients with cerebral AVMs, justifying an aggressive treatment posture. The supplementary system is better than the Spetzler-Martin system at predicting outcomes after cerebellar AVM resection. Key components of the Spetzler-Martin system such as venous drainage and eloquence are distorted by cerebellar anatomy in ways that components of the supplementary system are not.

Brain arteriovenous malformations associated with hereditary hemorrhagic telangiectasia: gene-phenotype correlations

Hereditary hemorrhagic telangiectasia (HHT) is an autosomal dominant genetic disease with a wide spectrum of vascular malformations (VMs) involving multiple organs. Nine to 16% of patients with HHT harbor brain arteriovenous malformations (AVMs), which can cause intracranial hemorrhage (ICH). Our objective was to study clinical manifestations of brain AVMs in patients with HHT and correlate these with the specific gene mutated. We reviewed records of 171 patients with HHT and brain AVMs. A history of ICH was found in 27% (41/152) patients, with a mean age of 26 ± 18 range, (0-68) years. All of patients with ICH were neurologically asymptomatic prior to ICH. Multiple brain AVMs were found in 23% (170/39) of patients on initial examination. Genetic test results were available in 109 (64%) patients. Mutations in ENG, ACVRL1, and SMAD4 were present in 75 (69%), 18 (17%), and 2 (2%), respectively. A history of ICH was reported in 24% of patients with an ENG mutation and 27% of patients with an ACVRL1 mutation, with a mean age of 26 ± 16 (range, 2-50) and 18 ± 21 (0-48) years, respectively. No statistically significant differences in age at first brain AVM diagnosis, prevalence of ICH history, age at ICH, or other manifestations of brain AVMs were observed among gene groups. In conclusion, no evidence for differences in brain AVM characteristics was observed among HHT gene groups, although we cannot exclude clinically important differences. Larger studies are needed to further guide brain AVM screening decisions in patients with HHT.

Evaluating performance of the spetzler-martin supplemented model in selecting patients with brain arteriovenous malformation for surgery

BACKGROUND AND PURPOSE

Our recently proposed point scoring model includes the widely-used Spetzler-Martin (SM)-5 variables, along with age, unruptured presentation, and diffuse border (SM-Supp). Here we evaluate the SM-Supp model performance compared with SM-5, SM-3, and Toronto prediction models using net reclassification index, which quantifies the correct movement in risk reclassification, and validate the model in an independent data set.

METHODS

Bad outcome was defined as worsening between preoperative and final postoperative modified Rankin Scale score. Point scores for each model were used as predictors in logistic regression and predictions evaluated using net reclassification index at varying thresholds (10%-30%) and any threshold (continuous net reclassification index >0). Performance was validated in an independent data set (n=117).

RESULTS

Net gain in risk reclassification was better using the SM-Supp model over a range of threshold values (net reclassification index=9%-25%) and significantly improved overall predictions for outcomes in the development data set, yielding a continuous net reclassification index of 64% versus SM-5, 67% versus SM-3, and 61% versus Toronto (all P<0.001). In the validation data set, the SM-Supp model again correctly reclassified a greater proportion of patients versus SM-5 (82%), SM-3 (85%), and Toronto models (69%).

CONCLUSIONS

The SM-Supp model demonstrated better discrimination and risk reclassification than several existing models and should be considered for clinical practice to estimate surgical risk in patients with brain arteriovenous malformation.

An Admission Bioclinical Score to Predict 1-Year Outcomes in Patients Undergoing Aneurysm Coiling

Background and Purpose—

A number of scores were developed to predict outcomes after clipping for subarachnoid hemorrhages, yet there is no score for patients undergoing endovascular treatment. Our goal was to develop, compare, and validate a predictive score for 1-year outcomes in patients with coiled subarachnoid hemorrhage.

Methods—

We studied 526 patients for 1 year after intensive care unit discharge. We developed an admission bioclinical score (ABC score), which integrated biomarkers such as troponin I and S100β, with the Glasgow Coma Scale. Using the receiver operating characteristic curve (95% CI), the ABC score was compared with the Glasgow Coma Scale, World Federation of Neurosurgical Societies score, and Fisher score in the derivation cohort and further validated in an independent cohort.

Results—

In the derivation cohort (from 2003–2007, n=368), multivariate logistic regression analysis showed that only Glasgow Coma Scale ( P <0.001), high S100β ( P <0.001), and high troponin ( P <0.02) were independently associated with 1-year mortality. Troponin, S100β, and Glasgow Coma Scale were thus integrated to derive the ABC score. In the derivation cohort, the ABC score reached an receiver operating characteristic curve of 0.82 (0.77–0.88, P <0.001) and was significantly greater than the receiver operating characteristic curves of the Glasgow Coma Scale, World Federation of Neurosurgical Societies, and Fisher scores for predicting 1-year mortality. In the validation cohort (from 2008–2009, n=158), the ABC score's receiver operating characteristic curve of 0.76 (0.67–0.86, P <0.001) remained superior to the 3 other scores for predicting 1-year mortality.

Conclusions—

The ABC score improves 1-year outcome prediction at admission for patients with coiled subarachnoid hemorrhage. Our study provides large cohort-based evidence supporting integration of individual biomarkers and clinical characteristics to predict outcomes.

Clinical Trial Registration—

URL: www.clinicaltrials.gov . Unique identifier: NCT01357057.

Silent intralesional microhemorrhage as a risk factor for brain arteriovenous malformation rupture

BACKGROUND AND PURPOSE

We investigated whether brain arteriovenous malformation silent intralesional microhemorrhage, that is, asymptomatic bleeding in the nidal compartment, might serve as a marker for increased risk of symptomatic intracranial hemorrhage (ICH). We evaluated 2 markers to assess the occurrence of silent intralesional microhemorrhage: neuroradiological assessment of evidence of old hemorrhage-imaging evidence of bleeding before the outcome events-and hemosiderin positivity in hematoxylin and eosin-stained paraffin block sections.

METHODS

We identified cases from our brain arteriovenous malformation database with recorded neuroradiological data or available surgical paraffin blocks. Using 2 end points, index ICH or new ICH after diagnosis (censored at treatment, loss to follow-up, or death), we performed logistic or Cox regression to assess evidence of old hemorrhage and hemosiderin positivity adjusting for age, sex, deep-only venous drainage, maximal brain arteriovenous malformation size, deep location, and associated arterial aneurysms.

RESULTS

Evidence of old hemorrhage was present in 6.5% (n=975) of patients and highly predictive of index ICH (P<0.001; OR, 3.97; 95% CI, 2.1-7.5) adjusting for other risk factors. In a multivariable model (n=643), evidence of old hemorrhage was an independent predictor of new ICH (hazard ratio, 3.53; 95% CI, 1.35-9.23; P=0.010). Hemosiderin positivity was found in 36.2% (29.6% in unruptured; 47.8% in ruptured; P=0.04) and associated with index ICH in univariate (OR, 2.18; 95% CI, 1.03-4.61; P=0.042; n=127) and multivariable models (OR, 3.64; 95% CI, 1.11-12.00; P=0.034; n=79).

CONCLUSIONS

The prevalence of silent intralesional microhemorrhage is high and there is evidence for an association with both index and subsequent ICH. Further development of means to detect silent intralesional microhemorrhage during brain arteriovenous malformation evaluation may present an opportunity to improve risk stratification, especially for unruptured brain arteriovenous malformations.

Abstract 2701: Multicenter Study For Prediction And Validation Of Intracranial Hemorrhage Risk In The Natural History Of Brain Arteriovenous Malformation Patients

Background: BAVM is an important cause of intracranial hemorrhage (ICH) in younger persons. Accurate and reliable prediction models for determining ICH risk in the natural history course of BAVM patients are needed to help guide management. The purpose of this study was to develop a prediction model of ICH risk, and validate the performance independently using the Multicenter AVM Research Study (MARS).

Methods: We used 3 BAVM cohorts from MARS: the UCSF Brain AVM Study Project (n=726), Columbia AVM Study (COL, n=640), and Scottish Intracranial Vascular Malformation Study (SIVMS, n=218). Cox proportional hazards analysis of time-to-ICH in the natural course after diagnosis was performed, censoring patients at first treatment, death, or last visit, up to 10 years. UCSF served as the model development cohort. We chose a simple model, including known risk factors that are reliably measured across cohorts (age at diagnosis, gender, initial hemorrhagic presentation, and deep venous drainage); variables were included without regard to statistical significance. Tertiles of predicted probabilities corresponding to low, medium, and high risk were obtained from UCSF and risk thresholds were validated in COL and SIVMS using Kaplan-Meier survival curves and log-rank tests (to assess whether the model discriminated between risk categories).

Results: Overall, 82 ICH events occurred during the natural course: 28 in UCSF, 41 in COL, and 13 in SIVMS. Effects in the prediction model (estimated from UCSF data) were: age in decades (HR=1.1, 95% CI=0.9-1.4, P=0.41), initial hemorrhagic presentation (HR=3.6, 95% CI=1.5-8.6, P=0.01), male gender (HR=1.1, 95% CI=0.48-2.6; P=0.81), and deep venous drainage (HR=0.8, 95% CI=0.2-2.8 P=0.72). Tertiles of ICH risk are shown in the Figure , demonstrating good separation of curves into low, medium and high risk after 3 years in UCSF (left, log-rank P=0.05). The model validated well in the COL referral cohort with better discrimination of curves (middle, P<0.001). In SIMVS, a population-based study, the model separated curves in the earlier years but a consistent pattern was not observed (right, P=0.51), possibly due to the small number of ICH events.

Conclusion: Our current prediction model for predicting ICH risk in the natural history course validates well in another referral population, but not as well in a population cohort. Inclusion of additional cohorts and risk factors after data harmonization may improve overall prediction and discrimination of ICH risk, and provide a generalizable model for clinical application.

Brain arteriovenous malformation multiplicity predicts the diagnosis of hereditary hemorrhagic telangiectasia: quantitative assessment

BACKGROUND AND PURPOSE

The purpose of this study was to quantitatively estimate the relationship between multiplicity of brain arteriovenous malformations (bAVMs) and the diagnosis of hereditary hemorrhagic telangiectasia (HHT).

METHODS

We combined databases from 2 large North American bAVM referral centers, including demographics, clinical presentation, and angiographic characteristics, and compared patients with HHT with non-HHT patients. Logistic regression analysis was performed to quantify the association between bAVM multiplicity and odds of HHT diagnosis. Sensitivity, specificity, positive and negative predictive value, and positive and negative likelihood ratios were calculated to determine accuracy of bAVM multiplicity for screening HHT.

RESULTS

Prevalence of HHT was 2.8% in the combined group. bAVM multiplicity was present in 39% of patients with HHT and was highly associated with diagnosis of HHT in univariate (OR, 83; 95% CI, 40-173; P<0.0001) and multivariable (OR, 86; 95% CI, 38-195; P<0.001) models adjusting for age at presentation (P=0.013), symptomatic presentation (P=0.029), and cohort site (P=0.021). bAVM multiplicity alone was associated with high specificity (99.2%; 95% CI, 98.7%-99.6%) and negative predictive value (98.3%; 95% CI, 97.6%-98.8%) and low sensitivity (39.3%; 95% CI, 26.5%-53.2%) and positive predictive value (59.5%; 95% CI, 42.1%-75.2%). Positive and negative likelihood ratio was 51 and 0.61, respectively, for diagnosis of HHT. HHT bAVMs were also more often smaller in size (<3 cm), noneloquent in location, and associated with superficial venous drainage compared with non-HHT bAVMs.

CONCLUSIONS

Multiplicity of bAVMs is highly predictive of the diagnosis of HHT. The presence of multiple bAVMs should alert the clinician to the high probability of HHT and lead to comprehensive investigation for this diagnosis.