Preresidency research output among US neurological surgery residents

OBJECTIVE

Research productivity is often used to evaluate candidates for neurosurgery residency. Official annual reports describe the mean total number of research products of successful applicants for each match cycle; however, the average number of indexed publications, the highest-valued research product, is not reported separately from other research products. The primary objectives of this study were to describe the distribution of preresidency indexed publication quantity among successful neurosurgery applicants from 2017 to 2021 and determine whether any change in publication quantity across application cycles existed. Secondary objectives included determining the rate at which the average publication quantity is increasing across application cycles, whether this increase is driven by high-output applicants alone, and if a performance ceiling has been reached.

METHODS

US doctor of medicine seniors applying to the senior author's institution between 2017 and 2021 and who successfully matched into any US neurosurgery program were included. Publication quantities were extracted using Scopus. Additional variables were extracted from residency applications. Mean (SD) and median (IQR) publication quantities were used to describe the distribution and compare across years. Applicants were ranked by descending publication count and divided into quartiles. Averages within each quartile were compared with respective quartiles across years. Averages of the top 10% most productive applicants were compared across years to determine if a performance ceiling existed.

RESULTS

Overall, 93.2% of matched applicants were captured. The mean and median total numbers of publications for applicants who matched from 2017 to 2021 were 5.6 ± 8.3 and 3.0 (1.0, 7.0), respectively. The mean and median numbers of publications increased from 3.7 ± 5.3 and 2.0 (0.0, 5.0) in 2016-2017 to 8.1 ± 10.0 and 5.0 (2.0, 11.0) in 2020-2021 (p < 0.001). The distribution of publication quantity was right-skewed. Multivariable analysis determined the application year to be independently and positively correlated with publication quantity (β 1.07 [95% CI 0.71-1.42], p < 0.001). All quartiles observed an increased average number of publications across years (p < 0.001). The mean and median numbers for the top 10% increased from 15.8 ± 8.7 and 13.0 (10.8, 15.5) in 2016-2017, respectively, to 31.3 ± 16.0 and 25.0 (21.0, 35.5) in 2020-2021 (p < 0.001).

CONCLUSIONS

Indexed publications account for a small portion of the total research products that successful neurosurgery candidates list on applications. A high number of publications is not necessary for candidates to match, with approximately 50% of all applicants who successfully matched having ≤ 5 publications and 25% having ≤ 2 publications. The average preresidency publication quantity has been increasing yearly among neurosurgery applicants. This increase was present across the applicant pool. Additionally, no performance ceiling was observed.

Assessing the Accuracy and Reliability of AI-Generated Medical Responses: An Evaluation of the Chat-GPT Model

Background

Natural language processing models such as ChatGPT can generate text-based content and are poised to become a major information source in medicine and beyond. The accuracy and completeness of ChatGPT for medical queries is not known.

Methods

Thirty-three physicians across 17 specialties generated 284 medical questions that they subjectively classified as easy, medium, or hard with either binary (yes/no) or descriptive answers. The physicians then graded ChatGPT-generated answers to these questions for accuracy (6-point Likert scale; range 1 - completely incorrect to 6 - completely correct) and completeness (3-point Likert scale; range 1 - incomplete to 3 - complete plus additional context). Scores were summarized with descriptive statistics and compared using Mann-Whitney U or Kruskal-Wallis testing.

Results

Across all questions (n=284), median accuracy score was 5.5 (between almost completely and completely correct) with mean score of 4.8 (between mostly and almost completely correct). Median completeness score was 3 (complete and comprehensive) with mean score of 2.5. For questions rated easy, medium, and hard, median accuracy scores were 6, 5.5, and 5 (mean 5.0, 4.7, and 4.6; p=0.05). Accuracy scores for binary and descriptive questions were similar (median 6 vs. 5; mean 4.9 vs. 4.7; p=0.07). Of 36 questions with scores of 1-2, 34 were re-queried/re-graded 8-17 days later with substantial improvement (median 2 vs. 4; p<0.01).

Conclusions

ChatGPT generated largely accurate information to diverse medical queries as judged by academic physician specialists although with important limitations. Further research and model development are needed to correct inaccuracies and for validation.

Methods and Impact for Using Federated Learning to Collaborate on Clinical Research

BACKGROUND

The development of accurate machine learning algorithms requires sufficient quantities of diverse data. This poses a challenge in health care because of the sensitive and siloed nature of biomedical information. Decentralized algorithms through federated learning (FL) avoid data aggregation by instead distributing algorithms to the data before centrally updating one global model.

OBJECTIVE

To establish a multicenter collaboration and assess the feasibility of using FL to train machine learning models for intracranial hemorrhage (ICH) detection without sharing data between sites.

METHODS

Five neurosurgery departments across the United States collaborated to establish a federated network and train a convolutional neural network to detect ICH on computed tomography scans. The global FL model was benchmarked against a standard, centrally trained model using a held-out data set and was compared against locally trained models using site data.

RESULTS

A federated network of practicing neurosurgeon scientists was successfully initiated to train a model for predicting ICH. The FL model achieved an area under the ROC curve of 0.9487 (95% CI 0.9471-0.9503) when predicting all subtypes of ICH compared with a benchmark (non-FL) area under the ROC curve of 0.9753 (95% CI 0.9742-0.9764), although performance varied by subtype. The FL model consistently achieved top three performance when validated on any site's data, suggesting improved generalizability. A qualitative survey described the experience of participants in the federated network.

CONCLUSION

This study demonstrates the feasibility of implementing a federated network for multi-institutional collaboration among clinicians and using FL to conduct machine learning research, thereby opening a new paradigm for neurosurgical collaboration.

TMIC-26. USING INTEGRATED MULTIDIMENSIONAL MASS CYTOMETRY AND MULTIPLEX IMMUNOHISTOCHEMISTRY TO INFER SPATIAL RELATIONSHIPS BETWEEN PHENOTYPICALLY DISTINCT GLIOBLASTOMA INFILTRATING IMMUNE CELLS

Glioblastomas (GBM) account for ~60% of adult primary brain tumors. With few advances in therapeutics, median overall survival remains 15-months post-diagnosis. Immunotherapies may provide therapeutic benefit; however, no predictive immune features have informed therapeutic stratification. Radiographic tumor contact with the lateral ventricle (C-GBM) correlates with 7-months worse prognosis compared to patients with ventricle non-contacting GBM (NC-GBM), yet the influence of ventricle contact on anti-tumor immunity is unknown. This study characterized the GBM immune microenvironment and identified targetable mechanisms of immunosuppression correlating with worse outcomes in C-GBM patients.Primary glioblastoma tissue was provided with written informed consent in accordance with the Declaration of Helsinki and with approval of the Vanderbilt Institutional Review Board (IRB #131870). Seventeen patients presented with primary, IDH-wildtype C-GBM and 15 with NC-GBM. Machine learning integrated mass cytometry and matched multiplex immunohistochemistry on FFPE embedded tissue to identify phenotypic, functional, and spatial biomarkers correlating with patient outcome. C-GBM tumors were enriched in STAT3-driven CD32+CD44+HLA-DR+ monocyte-derived macrophages (MDM) compared to NC-GBM (19 ± 8% vs. 6 ± 2%; p&lt; 0.001) and depleted in lymphocytes (2.9 ± 1% vs. 7.6 ± 2%; p&lt; 0.001) and tissue-resident microglia (1.8 ± 0.3% vs. 7 ± 3%; p&lt; 0.001). Exhausted T cells in C-GBM co-expressed checkpoint receptors PD-1 and TIGIT. K-means clustering identified 10 immunological niches in GBM. Macrophage-tumor niches were most common in C-GBM (17.93% of niches), followed by T cell-microglia-tumor niches (17.72%). Within NC-GBM niches, T cell-T cell interactions were more prevalent (log odds ratio = 0.90) and correlated with improved outcome.These findings suggest that factors within the periventricular space negatively influence the immune microenvironment within GBM tumors. Clinically targetable immune biomarkers were identified in C-GBM. Notably, radiologic assessment of lateral ventricle contact may guide clinical trial design for immunotherapies in neuro-oncology based on tumor proximity to the ventricle wall.

IMMU-11. SELECTIVE ENRICHMENT OF SUPPRESSED IMMUNE CELLS IN THE TUMOR MICROENVIRONMENT CORRELATES WITH ESTABLISHMENT OF DISTINCT IMMUNOLOGIC NICHES IN HUMAN GLIOBLASTOMAS CONTACTING THE LATERAL VENTRICLE

Glioblastomas (GBM) account for 60% of adult primary brain tumors. With few advances in therapeutics, median overall survival remains 15-months post-diagnosis. Immunotherapies may provide therapeutic benefit in GBM patients; however, no predictive immune features currently inform therapeutic stratification in GBM. We have shown that, independently of known prognosticators, radiographic tumor contact with the lateral ventricle (C-GBM) correlates with 7-months worse prognosis compared to patients with ventricle non-contacting GBM (NC-GBM). This study sought to characterize the GBM immune microenvironment and identify targetable mechanisms of immunosuppression correlating with worse outcomes in C-GBM. Primary glioblastoma specimens were resected in accordance with the Declaration of Helsinki (IRB #131870). Twelve patients presented with C-GBM and thirteen with NC-GBM. Machine learning tools applied to mass cytometry data characterized tumor-infiltrating immune populations and identified biomarkers correlating with C-GBM and patient survival. C-GBM tumors were enriched in blood-derived macrophages compared to NC-GBM (19 ± 8% vs. 6 ± 2%; p&lt; 0.001) and depleted in lymphocytes (2.9 ± 1% vs. 7.6 ± 2%; p&lt; 0.001) and tissue-resident microglia (1.8 ± 0.3% vs. 7 ± 3%; p&lt; 0.001). Further, T cells in C-GBM co-expressed the checkpoint receptors PD-1 and TIGIT, suggesting acute T cell exhaustion. Multiplex immunohistochemistry (mxIHC) on matched FFPE tissue provided spatial context to risk-stratifying immune populations, and defined structured immunological niches within the TME. Macrophage-tumor niches were most common (36%), followed by T cell-microglia-tumor niches (26%). Within niches, T cell-T cell interactions were more prevalent (log odds ratio = 0.90) whereas T cell-macrophage interactions were less prevalent (log odds ratio = -1.61). These findings suggest that factors within the periventricular space may influence antitumor immunity within tumors, and identify clinically targetable immune biomarkers in glioblastoma. Notably, radiologic assessment of lateral ventricle contact by standard-of-care MRI may guide clinical trial design for immunotherapies in neuro-oncology.

38 Spatial immune profiling of human glioblastoma tissue reveals the presence of aggregated lymphoid niches in the tumor microenvironment

Background

Glioblastomas (GBM) account for 60% of adult primary brain tumors. With few advances in therapeutics, median overall survival remains 15-months post-diagnosis. Immunotherapies may provide therapeutic benefit in GBM patients; however, no predictive immune features currently inform therapeutic stratification in GBM. We have shown that, independently of known prognosticators, radiographic tumor contact with the lateral ventricle (C-GBM) correlates with 7-months worse survival prognosis compared to patients with ventricle non-contacting GBM (NC-GBM). This study sought to characterize the GBM immune microenvironment and identify targetable mechanisms of immunosuppression correlating with worse outcomes in C-GBM.

Methods

Twelve patients presented with pathologically confirmed primary, IDH wildtype C-GBM and thirteen with NC-GBM. Multiplex immunohistochemistry (mxIHC) was performed on formalin-fixed paraffin embedded (FFPE) tissue for each patient interrogating 8 predictive immune markers (CD3, CD4, CD8, FOXP3, CD68, IBA1, PD-1, and PD-L1). Machine learning tools characterized tumor-infiltrating immune populations and identified biomarkers correlating with C-GBM and patient survival. K-means clustering identified immunological neighborhoods within the tissue and a log odds ratio was used to quantify the likelihood of cell-cell interactions in the tissue.

Results

C-GBM tumors were enriched in monocyte-derived macrophages (MDM) compared to NC-GBM (19 ± 8% vs. 6 ± 2%; p<0.001) and depleted in lymphocytes (2.9 ± 1% vs. 7.6 ± 2%; p<0.001) and tissue-resident microglia (1.8 ± 0.3% vs. 7 ± 3%; p<0.001). Further, T cells in C-GBM co-expressed the checkpoint receptors PD-1, suggesting T cell exhaustion in the C-GBM tumor microenvironment. K-means clustering identified 10 immunological niches prevalent in GBM tissue. Macrophage-tumor niches were most common niche in the tissue accounting for 17.93% of all niches, followed by T cell-microglia-tumor niches (17.72%). Conversely, tumor-tumor niches were the least prevalent, accounting for only 2.51% of niches. Within niches, T cell-T cell interactions occurred more frequently than expected by random chance (log odds ratio = 0.90) whereas T cell-macrophage interactions occurred less frequently than expected by random chance (log odds ratio = -1.61). Pathological assessment of the tissue confirmed the presence of lymphoid aggregates in regions of myeloid exclusion in the tissue.

Conclusions

These findings suggest that factors within the periventricular space may influence antitumor immunity within GBM, and have identified clinically targetable immune biomarkers in glioblastoma. The prevalence of T cell niches in GBM tumors suggests the establishment tertiary lymphoid aggregates may be targetable to improve patient outcomes. Lastly, radiologic assessment of lateral ventricle contact by standard-of-care MRI may guide clinical trial design for immunotherapies in neuro-oncology.

Acknowledgements

This study was funded by NIH/NCI grant K00 CA212447 and supported by the Translational Pathology Shared Resource at Vanderbilt University (P30 CA068485).

Ethics Approval

Primary glioblastoma tumors obtained in accordance with the Declaration of Helsinki and with institutional IRB approval (#131870) along with patient written informed consent.

IMMU-16. TWO DISTINCT SUBSETS OF NATURAL KILLER CELLS ARE ENRICHED IN THE TUMOR MICROENVIRONMENT AND CORRELATE WITH SURVIVAL OUTCOME IN HUMAN GLIOBLASTOMA

Glioblastomas (GBM) account for up to 60% of all adult primary brain tumors. With few advances in therapeutics, median overall survival (mOS) remains at 15-months post diagnosis. Success of immunotherapy in peripheral solid tumors may offer an alternative therapeutic approach for patients with GBM tumors; however, no predictive immune features currently inform therapeutic stratification for GBM. Recently, we have identified radiographic tumor contact with the lateral ventricle (LV) as a prognostic indicator of OS, as patients with LV+ GBM survive 7 months less than patients with LV- gliomas. This disparity was independent of known prognostic factors (e.g. KPS, extent of resection). Further, we have identified a correlation between greater immune infiltration and the frequency of tumor subtypes with more favorable prognosis. We therefore hypothesized that differences in overall survival between patients with LV+ and LV- are due, in part, to a uniquely immunosuppressive microenvironment within the LV. Using 35-parameter single-cell mass cytometry (CyTOF) we profiled the immune infiltrate of human GBM tissue acquired in accordance with the Declaration of Helsinki and with the approval of the institutional review board (IRB #131870). Computational approaches (tSNE, Citrus, RAPID) correlated natural killer (NK) cell populations correlating with prognosis. NK cells made of 1–14% of the total immune infiltrate in GBM. Ninety percent of NK cells infiltrating LV- tumors were CD16+CD56dim cytotoxic NK cells (cNK). LV+ gliomas, however, were enriched in CD16+CD56bright immunoregulatory NK cells (irNK). The presence of cNK cells correlated with a 2.5-fold improvement of OS, whereas irNK cells correlated with a 2.2-fold reduction in OS. Further, 30–60% of NK cells infiltrating LV+ tumors expressed checkpoint receptors (TIGIT, TIM3, B7-H3) compared to only 10–20% in LV- tumors. These results suggest that NK cells contribute to immunosuppression in the LV and may serve as alternative targets to T cell-based therapies for GBM.

COMP-11. SINGLE CELL MASS CYTOMETRY SIGNALING PROFILES AND A NOVEL COMPUTATIONAL TOOL IDENTIFY HIGH RISK GLIOBLASTOMA CELLS

In glioblastoma, changes in signaling, gene sequence, copy number, or transcript expression can define patient subgroups, but these subgroups are not yet associated with differential outcome for most patients with high-risk, IDH wild-type disease. Single cell interrogation of phospho-protein signaling has successfully revealed novel cell types associated with patient outcomes in blood cancers, suggesting that a comparable approach could be used in brain tumors. The goal of this study was to combine a single cell phospho-protein profiling approach with novel, automated computational analysis to identify abnormal glioblastoma cells that stratify patient clinical risk. Effective tissue dissociation strategies and validated antibody panels were created for mass cytometry analyses of resected glioblastoma tissue. These panels simultaneously measured 45 determinants of neural and glioma cell identity, including transcription factors, phospho-proteins, and surface receptors. 28 glioblastoma tumors were stained and analyzed using traditional gating, existing computational tools, and a new risk assessment population identification algorithm (RAPID, https://www.biorxiv.org/content/10.1101/632208v3). RAPID revealed two malignant cell types closely associated with differential patient outcomes. Glioblastoma negative prognostic (GNP) cells were associated with poor survival and defined by phospho-protein signaling in cells with aberrant neural developmental phenotypes. Glioblastoma positive prognostic (GPP) cells were associated with better progression free survival and defined by increased immunogenic signaling. A Cox proportional-hazards regression model was created to assess the influence of GNP and GPP cells on OS and PFS as continuous variables while accounting for other well-known clinical predictors. Each 1% increase of GNP cells was associated with an 7% increase in annual mortality rate (HR=1.07 [95% CI 1.03–1.12], p=0.001). Tumors containing GNP cells also significantly lacked CD45+ immune cell infiltration (Pearson r=-0.8). The signaling events that define these clinically significant glioblastoma cells represent a useful molecular classification, may indicate responsiveness to immunotherapy, and are themselves important targets of opportunity for new therapeutic approaches.

IMMU-37. SINGLE-CELL SYSTEMS NEUROIMMUNOLOGY REVEALS IMMUNOSUPPRESSIVE CORRELATES WITH VENTRICULAR STEM CELL NICHE CONTACT IN HUMAN GLIOBLASTOMA

Glioblastomas make up more than 60% of adult primary brain tumors and carry a median survival of less than 15 months despite aggressive standard therapy. Immunotherapy, which is now standard of care for many solid tumors, offers an appealing therapeutic approach that may improve outcomes for glioblastoma patients. Predictive features in glioblastomas that may inform responsiveness to different immunotherapeutic modalities, however, are still lacking. Recent studies have demonstrated that patients whose tumors show radiographic contact with the lateral ventricles, and thus the stem cell niche of the ventricular-subventricular zone (V-SVZ), have reduced survival outcomes compared to patients whose tumors do not contact the V-SVZ. We therefore hypothesized that tumor contact with the V-SVZ engenders a unique, immunosuppressive microenvironment that promotes tumor growth by suppressing anti-tumor immunity. Glioblastoma tumors, obtained in accordance with the Declaration of Helsinki and with institutional IRB approval (#131870, #030372, #181970), were disaggregated into single-cell suspensions and multi-dimensional single-cell mass cytometry was performed to interrogate >30 immune parameters in thirteen immune populations infiltrating human glioblastomas. Using advanced computational dimensionality-reduction tools (Citrus, t-SNE, FlowSOM, and MEM), we identified distinctions among the abundance and phenotypes of tumor-infiltrating immune cells. Firstly, on the basis of tumor contact with the V-SVZ, Citrus identified differential abundance of five T and myeloid cell subsets among glioblastomas. Secondly, differential expression of five functional immune markers was observed in seven distinct immune cell subsets infiltrating glioblastoma tumors. Further, both immune abundance and marker expression correlated with patient outcome. Manual gating analysis and parallel computational pipelines confirmed that comparable cell subsets could be identified with traditional approaches and unsupervised algorithmic analysis. These results provide key insights into the immune microenvironment of glioblastomas. In addition, several clinically actionable immunotherapeutic targets were uncovered that may be used to optimize treatment strategies for glioblastomas based on V-SVZ contact status.

Sex-specific relevance of diabetes to occlusive vascular and other mortality: a collaborative meta-analysis of individual data from 980 793 adults from 68 prospective studies

BACKGROUND

Several studies have shown that diabetes confers a higher relative risk of vascular mortality among women than among men, but whether this increased relative risk in women exists across age groups and within defined levels of other risk factors is uncertain. We aimed to determine whether differences in established risk factors, such as blood pressure, BMI, smoking, and cholesterol, explain the higher relative risks of vascular mortality among women than among men.

METHODS

In our meta-analysis, we obtained individual participant-level data from studies included in the Prospective Studies Collaboration and the Asia Pacific Cohort Studies Collaboration that had obtained baseline information on age, sex, diabetes, total cholesterol, blood pressure, tobacco use, height, and weight. Data on causes of death were obtained from medical death certificates. We used Cox regression models to assess the relevance of diabetes (any type) to occlusive vascular mortality (ischaemic heart disease, ischaemic stroke, or other atherosclerotic deaths) by age, sex, and other major vascular risk factors, and to assess whether the associations of blood pressure, total cholesterol, and body-mass index (BMI) to occlusive vascular mortality are modified by diabetes.

RESULTS

Individual participant-level data were analysed from 980 793 adults. During 9·8 million person-years of follow-up, among participants aged between 35 and 89 years, 19 686 (25·6%) of 76 965 deaths were attributed to occlusive vascular disease. After controlling for major vascular risk factors, diabetes roughly doubled occlusive vascular mortality risk among men (death rate ratio [RR] 2·10, 95% CI 1·97-2·24) and tripled risk among women (3·00, 2·71-3·33; χ2 test for heterogeneity p<0·0001). For both sexes combined, the occlusive vascular death RRs were higher in younger individuals (aged 35-59 years: 2·60, 2·30-2·94) than in older individuals (aged 70-89 years: 2·01, 1·85-2·19; p=0·0001 for trend across age groups), and, across age groups, the death RRs were higher among women than among men. Therefore, women aged 35-59 years had the highest death RR across all age and sex groups (5·55, 4·15-7·44). However, since underlying confounder-adjusted occlusive vascular mortality rates at any age were higher in men than in women, the adjusted absolute excess occlusive vascular mortality associated with diabetes was similar for men and women. At ages 35-59 years, the excess absolute risk was 0·05% (95% CI 0·03-0·07) per year in women compared with 0·08% (0·05-0·10) per year in men; the corresponding excess at ages 70-89 years was 1·08% (0·84-1·32) per year in women and 0·91% (0·77-1·05) per year in men. Total cholesterol, blood pressure, and BMI each showed continuous log-linear associations with occlusive vascular mortality that were similar among individuals with and without diabetes across both sexes.

INTERPRETATION

Independent of other major vascular risk factors, diabetes substantially increased vascular risk in both men and women. Lifestyle changes to reduce smoking and obesity and use of cost-effective drugs that target major vascular risks (eg, statins and antihypertensive drugs) are important in both men and women with diabetes, but might not reduce the relative excess risk of occlusive vascular disease in women with diabetes, which remains unexplained.

FUNDING

UK Medical Research Council, British Heart Foundation, Cancer Research UK, European Union BIOMED programme, and National Institute on Aging (US National Institutes of Health).

Preclinical TSPO Ligand PET to Visualize Human Glioma Xenotransplants: A Preliminary Study

Current positron emission tomography (PET) imaging biomarkers for detection of infiltrating gliomas are limited. Translocator protein (TSPO) is a novel and promising biomarker for glioma PET imaging. To validate TSPO as a potential target for molecular imaging of glioma, TSPO expression was assayed in a tumor microarray containing 37 high-grade (III, IV) gliomas. TSPO staining was detected in all tumor specimens. Subsequently, PET imaging was performed with an aryloxyanilide-based TSPO ligand, [¹⁸F]PBR06, in primary orthotopic xenograft models of WHO grade III and IV gliomas. Selective uptake of [¹⁸F]PBR06 in engrafted tumor was measured. Furthermore, PET imaging with [¹⁸F]PBR06 demonstrated infiltrative glioma growth that was undetectable by traditional magnetic resonance imaging (MRI). Preliminary PET with [¹⁸F]PBR06 demonstrated a preferential tumor-to-normal background ratio in comparison to 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([¹⁸F]FDG). These results suggest that TSPO PET imaging with such high-affinity radiotracers may represent a novel strategy to characterize distinct molecular features of glioma growth, as well as better define the extent of glioma infiltration for therapeutic purposes.

Multi-institutional validation of a preoperative scoring system which predicts survival for patients with glioblastoma

Glioblastoma is the most common and aggressive type of primary brain tumor in adults. Average survival is approximately 1 year, but individual survival is heterogeneous. Using a single institutional experience, we have previously identified preoperative factors associated with survival and devised a prognostic scoring system based on these factors. The aims of the present study are to validate these preoperative factors and verify the efficacy of this scoring system using a multi-institutional cohort. Of the 334 patients in this study from three different institutions, the preoperative factors found to be negatively associated with survival in a Cox analysis were age >60 years (p<0.0001), Karnofsky Performance Scale score ≤80 (p=0.03), motor deficit (p=0.02), language deficit (p=0.04), and periventricular tumor location (p=0.04). Patients possessing 0-1, 2, 3, and 4-5 of these variables were assigned a preoperative grade of 1, 2, 3, and 4, respectively. Patients with a preoperative grade of 1, 2, 3, and 4 had a median survival of 17.9, 12.3, 10, and 7.5 months, respectively. Survival of each of these grades was statistically significant (p<0.05) in log-rank analysis. This grading system, based only on preoperative variables, may provide patients and physicians with prognostic information that may guide medical and surgical therapy before any intervention is pursued.

Segmentation of malignant gliomas through remote collaboration and statistical fusion

PURPOSE

Malignant gliomas represent an aggressive class of central nervous system neoplasms. Correlation of interventional outcomes with tumor morphometry data necessitates 3D segmentation of tumors (typically based on magnetic resonance imaging). Expert delineation is the long-held gold standard for tumor segmentation, but is exceptionally resource intensive and subject to intrarater and inter-rater variability. Automated tumor segmentation algorithms have been demonstrated for a variety of imaging modalities and tumor phenotypes, but translation of these methods across clinical study designs is problematic given variation in image acquisition, tumor characteristics, segmentation objectives, and validation criteria. Herein, the authors demonstrate an alternative approach for high-throughput tumor segmentation using Internet-based, collaborative labeling.

METHODS

In a study of 85 human raters and 98 tumor patients, raters were recruited from a general university campus population (i.e., no specific medical knowledge), given minimal training, and provided web-based tools to label MRI images based on 2D cross sections. The labeling goal was characterized as to extract the enhanced tumor cores on T1-weighted MRI and the bright abnormality on T2-weighted MRI. An experienced rater manually constructed the ground truth volumes of a randomly sampled subcohort of 48 tumor subjects (for both T1w and T2w). Raters' taskwise individual observations, as well as the volume wise truth estimates via statistical fusion method, were evaluated over the subjects having the ground truth.

RESULTS

Individual raters were able to reliably characterize (with >0.8 dice similarity coefficient, DSC) the gadolinium-enhancing cores and extent of the edematous areas only slightly more than half of the time. Yet, human raters were efficient in terms of providing these highly variable segmentations (less than 20 s per slice). When statistical fusion was used to combine the results of seven raters per slice for all slices in the datasets, the 3D agreement of the fused results with expertly delineated segmentations was on par with the inter-rater reliability observed between experienced raters using traditional 3D tools (approximately 0.85 DSC). The cumulative time spent per tumor patient with the collaborative approach was equivalent to that with an experienced rater, but the collaborative approach could be achieved with less training time, fewer resources, and efficient parallelization.

CONCLUSIONS

Hence, collaborative labeling is a promising technique with potentially wide applicability to cost-effective manual labeling of medical images.

Temporal trends in ischemic and hemorrhagic strokes in Northern Italy: results from the cardiovascular monitoring unit in Northern Italy population-based register, 1998-2004

BACKGROUND

We compared rates and case fatality from a population-based stroke register in Northern Italy between 1998 and 2004 to assess changes over time and to evaluate changes in case diagnosis and management.

METHODS

The WHO Multinational Monitoring of Trends and Determinants in Cardiovascular Disease criteria were used to identify suspected fatal or nonfatal events occurring among residents 35-74 years of age. Data on in-hospital treatments, symptoms and diagnostic tools were extracted. Out-of-hospital deaths were also investigated. The annual average relative change (ARC) in death rate, attack rate and case fatality were derived from Poisson models.

RESULTS

Death rates due to ischemic stroke (IS) decreased [men: ARC -12.7, 95% confidence interval (CI) -21.3 to -3.2; women: ARC -14.0, 95% CI -23.3 to -3.5]. These reductions are attributable to decreases in case fatality; attack rates of nonfatal IS increased (men: ARC 3.6, 95% CI 0.5-6.7; women: ARC 4.1, 95% CI 0.0-8.2). IS patients showed a higher prevalence of dyslipidemia and hypertension and underwent MRI more frequently in 2004. Both findings may explain the increased proportions of less severe cases. Case fatality and attack rates for hemorrhagic strokes (HS) were stable, with an observed increased prevalence of patients under anticoagulant/antiplatelet treatments.

CONCLUSIONS

In this low-IS-incidence population, death rates decreased substantially during the investigated period. More accurate diagnostic tools increase the probability of detecting less severe cases. HS remains a frequently fatal disease with a stable incidence.

COLLABORATIVE LABELING OF MALIGNANT GLIOMA

Malignant gliomas represent an aggressive class of central nervous system neoplasms which are often treated by maximal surgical resection. Herein, we seek to improve the methods available to quantify the extent of tumors as seen on magnetic resonance imaging using Internet-based, collaborative labeling. In a study of clinically acquired images, we demonstrate that teams of minimally trained human raters are able to reliably characterize the gadolinium-enhancing core and edema tumor regions (Dice ≈ 0.9). The collaborative approach is highly parallel and efficient in terms of time (the total time spent by the collective is equivalent to that of a single expert) and resources (only minimal training and no hardware is provided to the participants). Hence, collaborative labeling is a very promising new technique with potentially wide applicability to facilitate cost-effective manual labeling of medical imaging data.

Collaborative Labeling of Malignant Glioma with WebMILL: A First Look

Malignant gliomas are the most common form of primary neoplasm in the central nervous system, and one of the most rapidly fatal of all human malignancies. They are treated by maximal surgical resection followed by radiation and chemotherapy. Herein, we seek to improve the methods available to quantify the extent of tumors using newly presented, collaborative labeling techniques on magnetic resonance imaging. Traditionally, labeling medical images has entailed that expert raters operate on one image at a time, which is resource intensive and not practical for very large datasets. Using many, minimally trained raters to label images has the possibility of minimizing laboratory requirements and allowing high degrees of parallelism. A successful effort also has the possibility of reducing overall cost. This potentially transformative technology presents a new set of problems, because one must pose the labeling challenge in a manner accessible to people with little or no background in labeling medical images and raters cannot be expected to read detailed instructions. Hence, a different training method has to be employed. The training must appeal to all types of learners and have the same concepts presented in multiple ways to ensure that all the subjects understand the basics of labeling. Our overall objective is to demonstrate the feasibility of studying malignant glioma morphometry through statistical analysis of the collaborative efforts of many, minimally-trained raters. This study presents preliminary results on optimization of the WebMILL framework for neoplasm labeling and investigates the initial contributions of 78 raters labeling 98 whole-brain datasets.

Safety and cost effectiveness of early discharge following microscopic trans-sphenoidal resection of pituitary lesions

Background:

Inpatient hospitalization following trans-sphenoidal resection of a pituitary neoplasm has traditionally involved a hospital stay of 2 days or more. It has been the policy of the senior pituitary neurosurgeon (GSA) since February 2008 to allow discharge home on postoperative day (POD) 1 if thirst mechanism is intact and the patient is tolerating oral hydration. The goal of this study was to evaluate the safety and cost-effectiveness of this practice.

Methods:

We reviewed the charts of 30 patients, designated the early discharge group, who consecutively underwent microscopic trans-sphenoidal resection from February 2008 to December 2009. We then reviewed the charts of 30 patients, designated the standard discharge group, who consecutively underwent trans-sphenoidal resection from May 2007 to February 2008 before discharge home on POD1 was considered an appropriate option. Safety and cost-effectiveness of the two patient groups were retrospectively evaluated.

Results

Patients in the early discharge group went home, on average, on POD 1.3. Following exclusion of two outliers, the average date of discharge of patients in the standard discharge group was POD 2.2. The policy of early discharge saved an average of $1,949 per patient-approximately 4% the total cost of the procedure. Trends toward decreased costs did not reach statistical significance. While no patient suffered any measurable morbidity as a result of early discharge home, 1 in 3 patients in the early discharge group required unscheduled postoperative re-evaluation-a figure significantly higher than the standard discharge group.

Conclusions:

At a dedicated pituitary center with the resources to closely monitor outpatient endocrinological and postsurgical issues, early discharge home following trans-sphenoidal surgery is a safe option that is associated with an increase in the number of unscheduled postoperative visits and a trend toward lower costs.

Hemostatic factors and subclinical brain infarction in a community-based sample: the ARIC study

BACKGROUND

Previous data are conflicting as to whether imbalance between hemostatic factors is associated with clinical strokes. We evaluated the association between hemostatic factor levels and subclinical lacunar infarcts in a nested sample from a subset of the Atherosclerosis Risk in Communities (ARIC) cohort.

METHODS

196 cases without clinical strokes had lacunar infarcts by MRI, and 214 controls without radiographic infarcts were frequency-matched by age group and sex. Logistic regression models were fitted to assess the association between levels of hemostatic markers and case status.

RESULTS

In age-, race- and sex-adjusted models, von Willebrand factor (vWF) and D-dimer were positively associated with case status, with odds ratios for the highest vs. lowest tertile of 2.0 (95% CI 1.2-3.6) for vWF and 1.76 (95% CI 1.02-3.0) for D-dimer. Plasminogen had nonsignificant inverse associations with presence of silent lacunar infarcts.

CONCLUSIONS

vWF and D-dimer were positively associated, and plasminogen was nonsignificantly inversely associated with subclinical radiographic infarct. Further studies on the role of these hemostatic factors in the development of silent lacunar infarcts may help elucidate the mechanisms behind this injury and may even point to potential targets for future intervention.

Systematically missing confounders in individual participant data meta-analysis of observational cohort studies

One difficulty in performing meta-analyses of observational cohort studies is that the availability of confounders may vary between cohorts, so that some cohorts provide fully adjusted analyses while others only provide partially adjusted analyses. Commonly, analyses of the association between an exposure and disease either are restricted to cohorts with full confounder information, or use all cohorts but do not fully adjust for confounding. We propose using a bivariate random-effects meta-analysis model to use information from all available cohorts while still adjusting for all the potential confounders. Our method uses both the fully adjusted and the partially adjusted estimated effects in the cohorts with full confounder information, together with an estimate of their within-cohort correlation. The method is applied to estimate the association between fibrinogen level and coronary heart disease incidence using data from 154 012 participants in 31 cohorts.† Copyright © 2009 John Wiley & Sons, Ltd.

The cardiovascular risk factor profile in the study area Augsburg. Results from the first MONICA survey 1984/85

The main objective of the first survey was the investigation of the prevalence and distribution of hypertension, hypercholesterolemia and cigarette smoking in the study area. The data-collection phase lasted from October 1984 to May 1985. A two-stage cluster sample of 5312 persons of German nationality was drawn from a population of 282,279 inhabitants, aged 25-64. The data were gathered through interview, physical examination and self-administered questionnaire. A response of 79% was achieved. Sixteen percent of men and 10% of women had high blood pressure (BP) values (greater than or equal to 160/95 mmHg). Only 16% of male and 34% of female hypertensives had controlled BP values. The prevalence of hypercholesterolemia (greater than or equal to 6.72 mmol/L = greater than or equal to 260 mg/dl) was 26% in men and 22% in women. Forty percent of men and 22% of women reported they were current cigarette smokers. Among participants, aged 25-44, cigarette smoking was the most prevalent risk factor in men and women. Thirty-seven percent of men and 52% of women, aged 25-64, had none of the three major risk factors.

Neighborhood of residence and incidence of coronary heart disease

BACKGROUND

Where a person lives is not usually thought of as an independent predictor of his or her health, although physical and social features of places of residence may affect health and health-related behavior.

METHODS

Using data from the Atherosclerosis Risk in Communities Study, we examined the relation between characteristics of neighborhoods and the incidence of coronary heart disease. Participants were 45 to 64 years of age at base line and were sampled from four study sites in the United States: Forsyth County, North Carolina; Jackson, Mississippi; the northwestern suburbs of Minneapolis; and Washington County, Maryland. As proxies for neighborhoods, we used block groups containing an average of 1000 people, as defined by the U.S. Census. We constructed a summary score for the socioeconomic environment of each neighborhood that included information about wealth and income, education, and occupation.

RESULTS

During a median of 9.1 years of follow-up, 615 coronary events occurred in 13,009 participants. Residents of disadvantaged neighborhoods (those with lower summary scores) had a higher risk of disease than residents of advantaged neighborhoods, even after we controlled for personal income, education, and occupation. Hazard ratios for coronary events in the most disadvantaged group of neighborhoods as compared with the most advantaged group--adjusted for age, study site, and personal socioeconomic indicators--were 1.7 among whites (95 percent confidence interval, 1.3 to 2.3) and 1.4 among blacks (95 percent confidence interval, 0.9 to 2.0). Neighborhood and personal socioeconomic indicators contributed independently to the risk of disease. Hazard ratios for coronary heart disease among low-income persons living in the most disadvantaged neighborhoods, as compared with high-income persons in the most advantaged neighborhoods were 3.1 among whites (95 percent confidence interval, 2.1 to 4.8) and 2.5 among blacks (95 percent confidence interval, 1.4 to 4.5). These associations remained unchanged after adjustment for established risk factors for coronary heart disease.

CONCLUSIONS

Even after controlling for personal income, education, and occupation, we found that living in a disadvantaged neighborhood is associated with an increased incidence of coronary heart disease.

Occupational risk factors for mortality from stomach and lung cancer among rubber workers: an analysis using internal controls and refined exposure assessment

OBJECTIVES

To determine occupational risk factors for stomach and lung cancer among workers in the German rubber industry.

METHODS

A cohort of all male German rubber workers (n = 11,633) who had been employed for at least one year in one of five study plants and who were alive and actively employed or retired on 1 January 1981 was followed for mortality from 1 January 1981 through December 1991. A subcohort of n = 8,933 workers who were hired after 1 January 1950 was defined to focus on working conditions in the more recent rubber industry. Work histories were reconstructed using routinely documented 'cost centre codes' and classified into six work areas as well as subgroups of some work areas. The work areas are: 1. preparation of materials, 2. production of technical rubber goods, 3. production of tyres, 4. storage and dispatch, 5. maintenance, 6. others. Standardized mortality ratios (SMR) and Cox proportional hazards models were calculated for each of the work areas (>1 year of employment in the respective work area). Hazard rate ratios were adjusted for age (time marker) and stratified for year of hire (1950-1959, > or =1960) and years of employment in the respective work area (1-9 years, > or =10 years); years of employment were lagged 10 years to account for latency.

RESULTS

Compared to the national reference population mortality from cancer of stomach (observed 44, SMR = 117; 95% CI: 85-157) and lung (observed 154, SMR = 123; 95% CI: 104-144) was slightly increased. Using internal controls we observed excess deaths from stomach cancer in work area 1 (relative risk [RR] = 2.3; 95% CI: 1.2-4.2) and from lung cancer in work areas 1 (RR = 1.7; 95% CI: 1.2-2.3), 2 (RR = 1.5; 95% CI: 1.1-2.1), and 3 (RR = 1.3; 95% CI: 0.9-1.8). On the basis of cumulative years of employment an exposure response relationship was observed for mortality from both cancer sites among a subcategory of work area 1: weighing and mixing. Increased risks were also seen for lung cancer among workers employed in production of technical rubber goods.

CONCLUSION

Our results support an association between an excess mortality from stomach and lung cancer and employment in early production stages of rubber manufacturing, especially weighing and mixing. This may point to an aetiologic role of asbestos or carbon black. For stomach cancer additional risk factors, e.g. exposure to dust and talc, deserve further investigation. The results of the present study do not support a causal role of nitrosamines for stomach or lung cancer.

An occupational cohort mortality study of women in the German rubber industry: 1976 to 1991

Few previous studies of workers in the rubber industry have focused on women. We examined patterns of mortality among 2871 women employed in one of five German rubber plants for at least 1 year on or after January 1, 1976, and observed through December 31, 1991. All-causes mortality was near that expected (standardized mortality ratio [lsqbSMR], 101; 95% confidence interval [CI], 87 to 118), but cancer mortality was decreased (SMR, 90; 95% CI, 70 to 115). Nevertheless, excesses were observed for mortality from stomach cancer (SMR, 156; 95% CI, 63 to 322), lung cancer (SMR, 140; 95% CI, 56 to 289), and lymphatic system cancers (SMR, 175; 95% CI, 48 to 448). Stronger associations were observed among sub-cohorts defined by time period hired. Despite limited numbers of deaths, modest excesses of mortality due to specific cancers were observed and are consistent with previous studies.

Comparison of medical care and survival of hospitalized patients with acute myocardial infarction in Poland and the United States

Few studies have evaluated between-country differences in medical care and survival after acute myocardial infarction, and none have compared the US with countries from Eastern Europe. Comparable data from the US (Atherosclerosis Risk in Communities Study [US-ARIC]) and Poland (Multinational Monitoring of Trends and Determinants in Cardiovascular Disease project [Pol-MONICA]) were developed. From 1987 through 1993, a total of 3,694 patients were hospitalized with acute myocardial infarction events in the 2 Pol-MONICA communities and 4,801 in the 4 US-ARIC communities. Patients in the US-ARIC were 1.7 times more likely to be treated in a coronary care unit and received cardiac procedures, calcium channel blockers, and thrombolytic agents significantly more often than patients in the Pol-MONICA. The use of antiplatelet agents, nitrates, angiotensin-converting enzyme inhibitors, and beta blockade agents was similar in both countries. Case fatality (28-day) rates after hospitalized acute myocardial infarction were nearly identical (men, 7% in Pol-MONICA vs 6% in US-ARIC; women, 9% in Pol-MONICA vs 8% in US-ARIC). However, when fatal coronary heart disease events not associated with a hospitalized myocardial infarction were included, the US-ARIC rates were less than half than those seen in Pol-MONICA. Substantial differences in treatment of hospitalized acute myocardial infarction between countries did not translate into a survival advantage for patients reaching clinical attention. Differences in case severity, arising from the high out-of-hospital coronary death rate in Poland may play an important role in this finding.

Workplace risk factors for cancer in the German rubber industry: Part 2. Mortality from non-respiratory cancers

OBJECTIVES

To determine the mortality from non-respiratory cancers by work area among active and retired male workers of the German rubber industry.

METHODS

A cohort of 11,633 male German workers was followed up for mortality from 1 January 1981 to 31 December 1991. Cohort members were active (n = 7536) or retired (n = 4127) on 1 January 1981 and had been employed for at least one year in one of five study plants producing tyres or technical rubber goods. Work histories were reconstructed from routinely documented "cost centre codes" and classified into six categories: I preparation of materials; II production of technical rubber goods; III production of tyres; IV storage and dispatch; V general service; VI others. Standardised mortality ratios (SMRs) and 95% confidence intervals (95% CIs), controlling for age and calendar year and stratified by work area (employment in respective work area for at least one year) and time related variables (year of hire, lagged years of employment in work area) were calculated from national mortality rates as the reference.

RESULTS

Significant increases in mortality were found for pharyngeal cancer in work area IV (three deaths, SMR 486, 95% CI 101 to 1419), oesophageal cancer in work area III (11 deaths, SMR 227, 95% CI 114 to 407), and leukaemia in work areas I (11 deaths, SMR 216; 95% CI 108 to 387) and II (14 deaths, SMR 187; 95% CI 102 to 213). Furthermore, increased SMRs were found for stomach cancer in work area I (22 deaths, SMR 134; 95% CI 84 to 203), colon cancer in work area II (27 deaths, SMR 131, 95% CI 86 to 191), prostatic cancer in work area V (27 deaths, SMR 152, 95% CI 99 to 221), and bladder cancer in work areas IV (six deaths, SMR 253; 95% CI 93 to 551) and V (12 deaths, SMR 159, 95% CI 82 to 279). Mortality from cancer of the liver or gall bladder, pancreas and kidney, and from lymphomas was not substantially increased in any of the work areas.

CONCLUSIONS

Mortality from cancer of several sites was associated with specific work areas. Some of these associations have been reported previously. Future analyses of our study will have to determine the role of specific exposures in the aetiology of these cancers.

Workplace risk factors for cancer in the German rubber industry: Part 1. Mortality from respiratory cancers

OBJECTIVES

To determine the cancer specific mortality by work area among active and retired male workers in the German rubber industry.

METHODS

A cohort of 11,663 male German workers was followed up for mortality from 1 January 1981 to 31 December 1991. Cohort members were classified as active (n = 7536) or retired (n = 4127) as of 1 January 1981 and had been employed for at least one year in one of five study plants producing tyres or technical rubber goods. Work histories were reconstructed with routinely documented "cost centre codes" which were classified into six categories: I preparation of materials; II production of technical rubber goods; III production of tyres; IV storage and dispatch; V maintenance; and VI others. Standardised mortality ratios (SMRs) adjusted for age and calendar year and 95% confidence intervals (95% CIs), stratified by work area (employment in respective work area for at least one year) and time related variables (year of hire, lagged years of employment in work area), were calculated from national reference rates.

RESULTS

SMRs for laryngeal cancer were highest in work area I (SMR 253; 95% CI 93 to 551) and were significant among workers who were employed for > 10 years in this work area (SMR 330; 95% CI 107 to 779). Increased mortality rates from lung cancer were identified in work areas I (SMR 162; 95% CI 129 to 202), II (SMR 134; 95% CI 109 to 163), and V (SMR 131; 95% CI 102 to 167). Mortality from pleural cancer was increased in all six work areas, and significant excesses were found in work areas I (SMR 448; 95% CI 122 to 1146), II (SMR 505; 95% CI 202 to 1040), and V (SMR 554; 95% CI 179 to 1290).

CONCLUSION

A causal relation between the excess of pleural cancer and exposure to asbestos among rubber workers is plausible and likely. In this study, the pattern of excess of lung cancer parallels the pattern of excess of pleural cancer. This points to asbestos as one risk factor for the excess deaths from lung cancer among rubber workers. The study provides further evidence for an increased mortality from laryngeal cancer among workers in the rubber industry, particularly in work area I.

Population versus clinical view of case fatality from acute coronary heart disease: results from the WHO MONICA Project 1985-1990. Multinational MONItoring of Trends and Determinants in CArdiovascular Disease

BACKGROUND

The clinical view of case fatality (CF) from acute myocardial infarction (AMI) in those reaching the hospital alive is different from the population view. Registration of both hospitalized AMI cases and out-of-hospital coronary heart disease (CHD) deaths in the WHO MONICA Project allows both views to be reconciled. The WHO MONICA Project provides the largest data set worldwide to explore the relationship between CHD CF and age, sex, coronary event rate, and first versus recurrent event.

METHODS AND RESULTS

All 79,669 events of definite AMI or possible coronary death, occurring from 1985 to 90 among 5,725,762 people, 35 to 64 years of age, in 29 MONICA populations are the basis for CF calculations. Age-adjusted CF (percentage of CHD events that were fatal) was calculated across populations, stratified for different time periods, and related to age, sex, and CHD event rate. Median 28-day population CF was 49% (range, 35% to 60%) in men and 51% (range, 34% to 70%) in women and was particularly higher in women than men in populations in which CHD event rates were low. Median 28-day CF for hospitalized events was much lower: in men 22% (range, 15% to 36%) and in women 27% (range, 19% to 46%). Among hospitalized events CF was twice as high for recurrent as for first events.

CONCLUSIONS

Overall 28-day CF is halved for hospitalized events compared with all events and again nearly halved for hospitalized 24-hour survivors. Because approximately two thirds of 28-day CHD deaths in men and women occurred before reaching the hospital, opportunities for reducing CF through improved care in the acute event are limited. Major emphasis should be on primary and secondary prevention.

Gender, racial, and geographic differences in the performance of cardiac diagnostic and therapeutic procedures for hospitalized acute myocardial infarction in four states

This study compared rates of performance of cardiac procedures in relation to gender, race, and geographic location in patients hospitalized for myocardial infarction. The Atherosclerosis Risk in Communities (ARIC) study provides population data and standardized data collection methods. Hospital records of eligible people aged 35 to 74 years were abstracted in communities of 4 states in the United States: North Carolina, Mississippi, Maryland, and Minnesota. Between January 1987 and December 1991, 5,462 "definite" hospitalized patients with myocardial infarctions were identified. Women treated in nonteaching hospitals were less likely than men to have coronary angiography (odds ratio [OR] 0.7, 95% confidence interval [CI] 0.5 to 1.0), coronary artery bypass graft surgery (CABG) (OR 0.6, 95% CI 0.4 to 0.8), and thrombolytic therapy (OR 0.8, 95% CI 0.6 to 1.0), after controlling for age, race, severity of myocardial infarction, co-morbidity, and geographic area. Findings were similar in teaching hospitals. Blacks in the biracial communities were significantly less likely than whites to have coronary angiography, percutaneous transluminal coronary angioplasty, coronary artery bypass graft surgery, and thrombolytic therapy. After controlling for age, race, severity of myocardial infarction and co-morbidity, no consistent geographic differences were observed, except for Forsyth whites having the highest and Washington County the lowest odds for coronary angiography. Appropriate outcome measures would serve to evaluate the effect, if any, of the differences described on the ARIC population.