Benjamin A. Goldstein

Clinical Interpretation and Implications of Whole-Genome Sequencing

IMPORTANCE Whole-genome sequencing (WGS) is increasingly applied in clinical medicine and is expected to uncover clinically significant findings regardless of sequencing indication. OBJECTIVES To examine coverage and concordance of clinically relevant genetic variation provided by WGS technologies; to quantitate inherited disease risk and pharmacogenomic findings in WGS data and resources required for their discovery and interpretation; and to evaluate clinical action prompted by WGS findings. DESIGN, SETTING, AND PARTICIPANTS An exploratory study of 12 adult participants recruited at Stanford University Medical Center who underwent WGS between November 2011 and March 2012. A multidisciplinary team reviewed all potentially reportable genetic findings. Five physicians proposed initial clinical follow-up based on the genetic findings. MAIN OUTCOMES AND MEASURES Genome coverage and sequencing platform concordance in different categories of genetic disease risk, person-hours spent curating candidate disease-risk variants, interpretation agreement between trained curators and disease genetics databases, burden of inherited disease risk and pharmacogenomic findings, and burden and interrater agreement of proposed clinical follow-up. RESULTS Depending on sequencing platform, 10% to 19% of inherited disease genes were not covered to accepted standards for single nucleotide variant discovery. Genotype concordance was high for previously described single nucleotide genetic variants (99%-100%) but low for small insertion/deletion variants (53%-59%). Curation of 90 to 127 genetic variants in each participant required a median of 54 minutes (range, 5-223 minutes) per genetic variant, resulted in moderate classification agreement between professionals (Gross κ, 0.52; 95% CI, 0.40-0.64), and reclassified 69% of genetic variants cataloged as disease causing in mutation databases to variants of uncertain or lesser significance. Two to 6 personal disease-risk findings were discovered in each participant, including 1 frameshift deletion in the BRCA1 gene implicated in hereditary breast and ovarian cancer. Physician review of sequencing findings prompted consideration of a median of 1 to 3 initial diagnostic tests and referrals per participant, with fair interrater agreement about the suitability of WGS findings for clinical follow-up (Fleiss κ, 0.24; P < 001). CONCLUSIONS AND RELEVANCE In this exploratory study of 12 volunteer adults, the use of WGS was associated with incomplete coverage of inherited disease genes, low reproducibility of detection of genetic variation with the highest potential clinical effects, and uncertainty about clinically reportable findings. In certain cases, WGS will identify clinically actionable genetic variants warranting early medical intervention. These issues should be considered when determining the role of WGS in clinical medicine.

An application of Random Forests to a genome-wide association dataset: Methodological considerations & new findings

BackgroundAs computational power improves, the application of more advanced machine learning techniques to the analysis of large genome-wide association (GWA) datasets becomes possible. While most traditional statistical methods can only elucidate main effects of genetic variants on risk for disease, certain machine learning approaches are particularly suited to discover higher order and non-linear effects. One such approach is the Random Forests (RF) algorithm. The use of RF for SNP discovery related to human disease has grown in recent years; however, most work has focused on small datasets or simulation studies which are limited.ResultsUsing a multiple sclerosis (MS) case-control dataset comprised of 300 K SNP genotypes across the genome, we outline an approach and some considerations for optimally tuning the RF algorithm based on the empirical dataset. Importantly, results show that typical default parameter values are not appropriate for large GWA datasets. Furthermore, gains can be made by sub-sampling the data, pruning based on linkage disequilibrium (LD), and removing strong effects from RF analyses. The new RF results are compared to findings from the original MS GWA study and demonstrate overlap. In addition, four new interesting candidate MS genes are identified, MPHOSPH9, CTNNA3, PHACTR2 and IL7, by RF analysis and warrant further follow-up in independent studies.ConclusionsThis study presents one of the first illustrations of successfully analyzing GWA data with a machine learning algorithm. It is shown that RF is computationally feasible for GWA data and the results obtained make biologic sense based on previous studies. More importantly, new genes were identified as potentially being associated with MS, suggesting new avenues of investigation for this complex disease.

Self-help on-line: an outcome evaluation of breast cancer bulletin boards.

Many breast cancer patients find help from on-line self-help groups, consisting of self-directed, asynchronous, bulletin boards. These have yet to be empirically evaluated. Upon joining a group and 6 months later, new members (N= 114) to breast cancer bulletin boards completed measures of depression (CES-D), growth (PTGI) and psychosocial wellbeing (FACT-B). Improvement was statistically significant on all three measures. This serves as a first validation of Internet bulletin boards as a source of support and help for breast cancer patients. These boards are of particular interest because they are free, accessible and support comes from peers and not from professional facilitators.

Etiological paradigms of depression: The relationship between perceived causes, empowerment, treatment preferences, and stigma

Background: There is a growing trend to view depression as a biological illness rather than a psychosocial condition, even though there is no consensus as to what causes depression. Furthermore, there are mixed data on the impact of advocating the biological model. Aims: This study examined public perceptions concerning the etiology of depression as well as the relationship between such perceptions and treatment preferences, empowerment, and stigma. Method: Survey techniques were used to assess how 66 college students view the etiology of depression. Etiology beliefs, as well as demographic data, were regressed upon measures of treatment preference, empowerment, and stigma. Results: Factor analysis produced three distinct models of etiology: biological, psychological, and environmental. Regression analyses showed that endorsement of the biological model was associated with increased empowerment, preference for psychotherapy, and decreased stigma. Endorsing the psychological model was associated with an increased belief that people can help themselves and increased stigma. Endorsing the environmental model was associated with a mixture of positive and negative beliefs concerning depression. Conclusions: Endorsement of each etiological model is associated with both positive and negative consequences. The current public emphasis on viewing depression as biologically based should thus be viewed with some caution. Declaration of interest: None.

Cardiovascular disease mortality in Asian Americans.

BACKGROUND Asian Americans are a rapidly growing racial/ethnic group in the United States. Our current understanding of Asian-American cardiovascular disease mortality patterns is distorted by the aggregation of distinct subgroups. OBJECTIVES The purpose of the study was to examine heart disease and stroke mortality rates in Asian-American subgroups to determine racial/ethnic differences in cardiovascular disease mortality within the United States. METHODS We examined heart disease and stroke mortality rates for the 6 largest Asian-American subgroups (Asian Indian, Chinese, Filipino, Japanese, Korean, and Vietnamese) from 2003 to 2010. U.S. death records were used to identify race/ethnicity and cause of death by International Classification of Diseases-10th revision coding. Using both U.S. Census data and death record data, standardized mortality ratios (SMRs), relative SMRs (rSMRs), and proportional mortality ratios were calculated for each sex and ethnic group relative to non-Hispanic whites (NHWs). RESULTS In this study, 10,442,034 death records were examined. Whereas NHW men and women had the highest overall mortality rates, Asian Indian men and women and Filipino men had greater proportionate mortality burden from ischemic heart disease. The proportionate mortality burden of hypertensive heart disease and cerebrovascular disease, especially hemorrhagic stroke, was higher in every Asian-American subgroup compared with NHWs. CONCLUSIONS The heterogeneity in cardiovascular disease mortality patterns among diverse Asian-American subgroups calls attention to the need for more research to help direct more specific treatment and prevention efforts, in particular with hypertension and stroke, to reduce health disparities for this growing population.

Random Forests for Genetic Association Studies

The Random Forests (RF) algorithm has become a commonly used machine learning algorithm for genetic association studies. It is well suited for genetic applications since it is both computationally efficient and models genetic causal mechanisms well. With its growing ubiquity, there has been inconsistent and less than optimal use of RF in the literature. The purpose of this review is to breakdown the theoretical and statistical basis of RF so that practitioners are able to apply it in their work. An emphasis is placed on showing how the various components contribute to bias and variance, as well as discussing variable importance measures. Applications specific to genetic studies are highlighted. To provide context, RF is compared to other commonly used machine learning algorithms.

National decline in donor heart utilization with regional variability: 1995-2010.

The severe shortage of donor hearts limits the availability of transplantation for the growing population of patients with end‐stage heart disease. We examined national trends in donor heart acceptance for transplant. OPTN data were analyzed for all potential adult cardiac organ donors between 1995 and 2010. Donor heart disposition was categorized as transplanted, declined for transplant or other. We studied changes in the probability of donor heart acceptance according to demographic and clinical characteristics, nationwide and by UNOS region. Of 82 053 potential donor hearts, 34% were accepted and 48% were declined (18% used for other purposes). There was a significant decrease in donor heart acceptance from 44% in 1995 to 29% in 2006, and subsequent increase to 32% in 2010. Older donor age, female sex and medical co‐morbidities predicted non‐acceptance. Donor age and co‐morbidities increased during the study period, with a concomitant decrease in acceptance of hearts from donors with undesirable characteristics. Overall, predictors of heart non‐use were similar across UNOS regions, although utilization varied between regions. Regional variation suggests a potential to improve heart acceptance rates in under‐performing regions, and supports research and policy efforts aimed at establishing evidence‐based criteria for donor heart evaluation and acceptance for transplantation.

Opportunities and challenges in developing risk prediction models with electronic health records data: a systematic review

Objective: Electronic health records (EHRs) are an increasingly common data source for clinical risk prediction, presenting both unique analytic opportunities and challenges. We sought to evaluate the current state of EHR based risk prediction modeling through a systematic review of clinical prediction studies using EHR data. Methods: We searched PubMed for articles that reported on the use of an EHR to develop a risk prediction model from 2009 to 2014. Articles were extracted by two reviewers, and we abstracted information on study design, use of EHR data, model building, and performance from each publication and supplementary documentation. Results: We identified 107 articles from 15 different countries. Studies were generally very large (median sample size = 26 100) and utilized a diverse array of predictors. Most used validation techniques (n = 94 of 107) and reported model coefficients for reproducibility (n = 83). However, studies did not fully leverage the breadth of EHR data, as they uncommonly used longitudinal information (n = 37) and employed relatively few predictor variables (median = 27 variables). Less than half of the studies were multicenter (n = 50) and only 26 performed validation across sites. Many studies did not fully address biases of EHR data such as missing data or loss to follow-up. Average c-statistics for different outcomes were: mortality (0.84), clinical prediction (0.83), hospitalization (0.71), and service utilization (0.71). Conclusions: EHR data present both opportunities and challenges for clinical risk prediction. There is room for improvement in designing such studies.

Trends in the Incidence of Atrial Fibrillation in Older Patients Initiating Dialysis in the United States

Background— One sixth of US dialysis patients 65 years of age have been diagnosed with atrial fibrillation/flutter (AF). Little is known, however, about the incidence of AF in this population. Methods and Results— We identified 258 605 older patients (≥67 years of age) with fee-for-service Medicare initiating dialysis in 1995 to 2007, who had not been diagnosed with AF within the previous 2 years. Patients were followed for newly diagnosed AF. Multivariable proportional hazard regression was used to examine temporal trends and associations of race and ethnicity with incident AF. We also studied temporal trends in the mortality and risk of ischemic stroke after new AF. Over 514 395 person-years of follow-up, 76 252 patients experienced incident AF for a crude AF incidence rate of 148/1000 person-years. Incidence of AF increased by 11% (95% confidence interval, 5–16) from 1995 to 2007. Compared with non-Hispanic whites, blacks (−30%), Asians (−19%), Native Americans (−42%), and Hispanics (−29%) all had lower rates of incident AF. Mortality after incident AF decreased by 22% from 1995 to 2008. Even more pronounced reductions were seen for incident ischemic stroke during these years. Conclusions— The incidence of AF is high in older patients initiating dialysis in the United States and has been increasing over the 13 years of study. Mortality declined during that time but remained >50% during the first year after newly diagnosed AF. Because data on warfarin use were not available, we were unable to study whether trends toward better outcomes could be explained by higher rates of oral anticoagulation.

Donor Predictors of Allograft Use and Recipient Outcomes After Heart Transplantation

Background—Despite a national organ-donor shortage and a growing population of patients with end-stage heart disease, the acceptance rate of donor hearts for transplantation is low. We sought to identify donor predictors of allograft nonuse, and to determine whether these predictors are in fact associated with adverse recipient post-transplant outcomes. Methods and Results—We studied a cohort of 1872 potential organ donors managed by the California Transplant Donor Network from 2001 to 2008. Forty-five percent of available allografts were accepted for heart transplantation. Donor predictors of allograft nonuse included age>50 years, female sex, death attributable to cerebrovascular accident, hypertension, diabetes mellitus, a positive troponin assay, left-ventricular dysfunction and regional wall motion abnormalities, and left-ventricular hypertrophy. For hearts that were transplanted, only donor cause of death was associated with prolonged recipient hospitalization post-transplant, and only donor diabetes mellitus was predictive of increased recipient mortality. Conclusions—Whereas there are many donor predictors of allograft discard in the current era, these characteristics seem to have little effect on recipient outcomes when the hearts are transplanted. Our results suggest that more liberal use of cardiac allografts with relative contraindications may be warranted.