Patient-Specific Modeling with Personalized Decision Paths.

Predictive models can be useful in predicting patient outcomes under uncertainty. Many algorithms employ "population" methods, which optimize a single model to perform well on average over an entire population, but the model may perform poorly on some patients. Personalized methods optimize predictive performance for each patient by tailoring the model to the individual. We present a new personalized method based on decision trees: the Personalized Decision Path using a Bayesian score (PDP-Bay). Performance on eight synthetic, genomic, and clinical datasets was compared to that of decision trees and a previously described personalized decision path method in terms of area under the ROC curve (AUC) and expected calibration error (ECE). Model complexity was measured by average path length. The PDP-Bay model outperformed the decision tree in terms of both AUC and ECE. The results support the conclusion that personalization may achieve better predictive performance and produce simpler models than population approaches.

Self-Reported Sleep Apnea and Dementia Risk: Findings from the Prevention of Alzheimer's Disease with Vitamin E and Selenium Trial.

OBJECTIVES: To investigate the association between baseline sleep apnea and risk of incident dementia in the Prevention of Alzheimer's Disease with Vitamin E and Selenium (PREADViSE) study and to explore whether the association depends on apolipoprotein E (APOE) ɛ4 allele status. DESIGN: Secondary analysis based on data collected during PREADViSE. SETTING: Participants were assessed at 128 local clinical study sites during the clinical trial phase and later were followed by telephone from a centralized location. PARTICIPANTS: Men enrolled in PREADViSE (without dementia or other active neurological conditions that affect cognition such as major psychiatric disorders, including depression; N = 7,547). MEASUREMENTS: Participants were interviewed at baseline for sleep apnea. The Memory Impairment Screen (MIS) was administered to each participant annually. Subjects who failed this initial screen were tested with secondary screening tests. Medical history and medication use were determined, and the AD8 dementia screening instrument was used. RESULTS: The effect of self-reported sleep apnea on dementia risk depended on APOE ɛ4 status. When the allele was absent, baseline self-reported sleep apnea was associated with a 66% higher risk of developing dementia (95% confidence interval = 2-170%), whereas self-reported sleep apnea conferred no additional risk for participants with an ɛ4 allele. CONCLUSION: Sleep apnea may increase risk of dementia in the absence of APOE ɛ4. This may help inform prevention strategies for dementia or AD in older men with sleep apnea. Registration: PREADViSE is registered at ClinicalTrials.gov: NCT00040378.

Repair of a Complicated Calvarial Defect: Reconstruction of an Infected Wound With rhBMP-2.

BACKGROUND: Management of the previously infected craniofacial defect remains a significant clinical challenge, posing obstacles such as wound healing complications, lack of donor site availability, and predisposition to failure of the repair. Optimal therapy would reconstruct like with like, without donor site morbidity. The purpose of this study was to compare the efficacy of recombinant human bone morphogenetic protein-2 (rhBMP-2)-mediated bone regeneration with the current standard of autologous bone graft for repair of previously infected calvarial defects. METHODS: Nineteen adult New Zealand white rabbits underwent subtotal calvariectomy. Bone flaps were inoculated with Staphylococcus aureus and replanted. After 1 week of infection, bone flaps were removed, and wounds were debrided, followed by 10 days of antibiotic treatment. After 6 weeks, animals underwent scar debridement followed by definitive reconstruction in 1 of 4 groups: empty control (n = 3), vehicle control (buffer solution on absorbable collagen sponge [ACS], n = 3), autologous bone graft (n = 3), or rhBMP-2 repair (rhBMP-2/ACS, n = 10). Animals underwent computed tomography imaging at 0, 2, 4, and 6 weeks postoperatively, followed by euthanization and histological analysis. Percent healing was determined by 3-dimensional analysis. A (time × group) 2-way analysis of variance was performed on healing versus treatment group and postoperative time. RESULTS: At 6 weeks postoperatively, rhBMP-2/ACS and autologous bone graft resulted in 93% and 68% healing, respectively, whereas the empty and vehicle control treatment resulted in 27% and 26% healing (P < 0.001). Histologically, compared to autologous bone graft, bone in the rhBMP-2/ACS group was more cellular and more consistently continuous with wound margins. CONCLUSIONS: The rhBMP-2 therapy is effective in achieving radiographic coverage of previously infected calvarial defects.

Lasting Effects of the Breast and Cervical Cancer Early Detection Program on Breast Cancer Detection and Outcomes, Ohio, 2000-2009.

INTRODUCTION: The National Breast and Cervical Cancer Early Detection Program (BCCP) in Ohio provides screening and treatment services for uninsured low-income women aged 40 to 64. Because participation in the BCCP might engender greater self-efficacy for cancer screening, we hypothesized that breast cancer and survival outcomes would be better in BCCP participants who become age-eligible to transition to Medicare than in their low-income non-BCCP counterparts. METHODS: Linking data from the 2000 through 2009 Ohio Cancer Incidence Surveillance System with the BCCP database, Medicare files, Ohio death certificates (through 2010), and the US Census, we identified Medicare beneficiaries who were aged 66 to 74 and diagnosed with incident invasive breast cancer. We compared the following outcomes between BCCP women (n = 93) and low-income non-BCCP women (n = 420): receipt of screening mammography in previous year, advanced-stage disease at diagnosis, timely and standard care, all-cause survival, and cancer survival. We conducted multivariable logistic regression and survival analysis to examine the association between BCCP status and each of the outcomes, adjusting for patient covariates. RESULTS: Women who participated in the BCCP were nearly twice as likely as low-income non-BCCP women to have undergone screening mammography in the previous year (adjusted odds ratio, 1.77; 95% confidence interval, 1.01-3.09). No significant differences were detected in any other outcomes. CONCLUSION: With the exception of screening mammography, the differences in outcomes were not significant, possibly because of the small size of the study population. Future analysis should be directed toward identifying the factors that explain these findings.

Application of Bayesian logistic regression to mining biomedical data.

Mining high dimensional biomedical data with existing classifiers is challenging and the predictions are often inaccurate. We investigated the use of Bayesian Logistic Regression (B-LR) for mining such data to predict and classify various disease conditions. The analysis was done on twelve biomedical datasets with binary class variables and the performance of B-LR was compared to those from other popular classifiers on these datasets with 10-fold cross validation using the WEKA data mining toolkit. The statistical significance of the results was analyzed by paired two tailed t-tests and non-parametric Wilcoxon signed-rank tests. We observed overall that B-LR with non-informative Gaussian priors performed on par with other classifiers in terms of accuracy, balanced accuracy and AUC. These results suggest that it is worthwhile to explore the application of B-LR to predictive modeling tasks in bioinformatics using informative biological prior probabilities. With informative prior probabilities, we conjecture that the performance of B-LR will improve.

Selective model averaging with bayesian rule learning for predictive biomedicine.

Accurate disease classification and biomarker discovery remain challenging tasks in biomedicine. In this paper, we develop and test a practical approach to combining evidence from multiple models when making predictions using selective Bayesian model averaging of probabilistic rules. This method is implemented within a Bayesian Rule Learning system and compared to model selection when applied to twelve biomedical datasets using the area under the ROC curve measure of performance. Cross-validation results indicate that selective Bayesian model averaging statistically significantly outperforms model selection on average in these experiments, suggesting that combining predictions from multiple models may lead to more accurate quantification of classifier uncertainty. This approach would directly impact the generation of robust predictions on unseen test data, while also increasing knowledge for biomarker discovery and mechanisms that underlie disease.

Decision path models for patient-specific modeling of patient outcomes.

Patient-specific models are constructed to take advantage of the particular features of the patient case of interest compared to commonly used population-wide models that are constructed to perform well on average on all cases. We introduce two patient-specific algorithms that are based on the decision tree paradigm. These algorithms construct a decision path specific for each patient of interest compared to a single population-wide decision tree with many paths that is applicable to all patients of interest that are constructed by standard algorithms. We applied the patient-specific algorithms to predict five different outcomes in clinical datasets. Compared to the population-wide CART decision tree the patient-specific decision path models had superior performance on area under the ROC curve (AUC) and had comparable performance on balanced accuracy. Our results provide support for patient-specific algorithms being a promising approach for predicting clinical outcomes.

A fast algorithm for learning epistatic genomic relationships.

Genetic epidemiologists strive to determine the genetic profile of diseases. Epistasis is the interaction between two or more genes to affect phenotype. Due to the often non-linearity of the interaction, it is difficult to detect statistical patterns of epistasis. Combinatorial methods for detecting epistasis investigate a subset of combinations of genes without employing a search strategy. Therefore, they do not scale to handling the high-dimensional data found in genome-wide association studies (GWAS). We represent genome-phenome interactions using a Bayesian network rule, which is a specialized Bayesian network. We develop an efficient search algorithm to learn from data a high scoring rule that may contain two or more interacting genes. Our experimental results using synthetic data indicate that this algorithm detects interacting genes as well as a Bayesian network combinatorial method, and it is much faster. Our results also indicate that the algorithm can successfully learn genome-phenome relationships using a real GWAS dataset.