Gregory J. Matthews

Higher plasma CXCL12 levels predict incident myocardial infarction and death in chronic kidney disease: findings from the Chronic Renal Insufficiency Cohort study

AIMS Genome-wide association studies revealed an association between a locus at 10q11, downstream from CXCL12, and myocardial infarction (MI). However, the relationship among plasma CXCL12, cardiovascular disease (CVD) risk factors, incident MI, and death is unknown. METHODS AND RESULTS We analysed study-entry plasma CXCL12 levels in 3687 participants of the Chronic Renal Insufficiency Cohort (CRIC) Study, a prospective study of cardiovascular and kidney outcomes in chronic kidney disease (CKD) patients. Mean follow-up was 6 years for incident MI or death. Plasma CXCL12 levels were positively associated with several cardiovascular risk factors (age, hypertension, diabetes, hypercholesterolaemia), lower estimated glomerular filtration rate (eGFR), and higher inflammatory cytokine levels (P < 0.05). In fully adjusted models, higher study-entry CXCL12 was associated with increased odds of prevalent CVD (OR 1.23; 95% confidence interval 1.14, 1.33, P < 0.001) for one standard deviation (SD) increase in CXCL12. Similarly, one SD higher CXCL12 increased the hazard of incident MI (1.26; 1.09,1.45, P < 0.001), death (1.20; 1.09,1.33, P < 0.001), and combined MI/death (1.23; 1.13-1.34, P < 0.001) adjusting for demographic factors, known CVD risk factors, and inflammatory markers and remained significant for MI (1.19; 1.03,1.39, P = 0.01) and the combined MI/death (1.13; 1.03,1.24, P = 0.01) after further controlling for eGFR and urinary albumin:creatinine ratio. CONCLUSIONS In CKD, higher plasma CXCL12 was associated with CVD risk factors and prevalent CVD as well as the hazard of incident MI and death. Further studies are required to establish if plasma CXCL12 reflect causal actions at the vessel wall and is a tool for genomic and therapeutic trials.

Candidate Gene Association Study of Coronary Artery Calcification in Chronic Kidney Disease: Findings From the CRIC Study (Chronic Renal Insufficiency Cohort)

OBJECTIVES This study sought to identify loci for coronary artery calcification (CAC) in patients with chronic kidney disease (CKD). BACKGROUND CKD is associated with increased CAC and subsequent coronary heart disease (CHD), but the mechanisms remain poorly defined. Genetic studies of CAC in CKD may provide a useful strategy for identifying novel pathways in CHD. METHODS We performed a candidate gene study (∼2,100 genes; ∼50,000 single nucleotide polymorphisms [SNPs]) of CAC within the CRIC (Chronic Renal Insufficiency Cohort) study (N = 1,509; 57% European, 43% African ancestry). SNPs with preliminary evidence of association with CAC in CRIC were examined for association with CAC in the PennCAC (Penn Coronary Artery Calcification) (N = 2,560) and AFCS (Amish Family Calcification Study) (N = 784) samples. SNPs with suggestive replication were further analyzed for association with myocardial infarction (MI) in the PROMIS (Pakistan Risk of Myocardial Infarction Study) (N = 14,885). RESULTS Of 268 SNPs reaching p < 5 × 10(-4) for CAC in CRIC, 28 SNPs in 23 loci had nominal support (p < 0.05 and in same direction) for CAC in PennCAC or AFCS. Besides chr9p21 and COL4A1, known loci for CHD, these included SNPs having reported genome-wide association study association with hypertension (e.g., ATP2B1). In PROMIS, 4 of the 23 suggestive CAC loci (chr9p21, COL4A1, ATP2B1, and ABCA4) had significant associations with MI, consistent with their direction of effect on CAC. CONCLUSIONS We identified several loci associated with CAC in CKD that also relate to MI in a general population sample. CKD imparts a high risk of CHD and may provide a useful setting for discovery of novel CHD genes and pathways.

Building an NCAA men’s basketball predictive model and quantifying its success

Abstract Computing and machine learning advancements have led to the creation of many cutting-edge predictive algorithms, some of which have been demonstrated to provide more accurate forecasts than traditional statistical tools. In this manuscript, we provide evidence that the combination of modest statistical methods with informative data can meet or exceed the accuracy of more complex models when it comes to predicting the NCAA men’s basketball tournament. First, we describe a prediction model that merges the point spreads set by Las Vegas sportsbooks with possession based team efficiency metrics by using logistic regressions. The set of probabilities generated from this model most accurately predicted the 2014 tournament, relative to approximately 400 competing submissions, as judged by the log loss function. Next, we attempt to quantify the degree to which luck played a role in the success of this model by simulating tournament outcomes under different sets of true underlying game probabilities. We estimate that under the most optimistic of game probability scenarios, our entry had roughly a 12% chance of outscoring all competing submissions and just less than a 50% chance of finishing with one of the ten best scores.

Serum Fractalkine (CX3CL1) and Cardiovascular Outcomes and Diabetes: Findings From the Chronic Renal Insufficiency Cohort (CRIC) Study

BACKGROUND Cardiometabolic disease is a major cause of morbidity and mortality in persons with chronic kidney disease (CKD). Fractalkine (CX3CL1) is a potential mediator of both atherosclerosis and metabolic disease. Studies of the relationship of CX3CL1 with risk of cardiovascular disease (CVD) events and metabolic traits are lacking, particularly in the high-risk setting of CKD. STUDY DESIGN Cross-sectional and longitudinal observational analysis. SETTING & PARTICIPANTS Adults with CKD from 7 US sites participating in the Chronic Renal Insufficiency Cohort (CRIC) Study. PREDICTOR Quartiles of plasma CX3CL1 levels at baseline. OUTCOMES Baseline estimated glomerular filtration rate from a creatinine and cystatin C-based equation, prevalent and incident CVD, diabetes, metabolic syndrome and its criteria, homeostatic model assessment of insulin resistance, hemoglobin A1c level, myocardial infarction, all-cause mortality, and the composite outcome of myocardial infarction/all-cause mortality. RESULTS Among 3,687 participants, baseline CX3CL1 levels were associated positively with several CVD risk factors and metabolic traits, lower estimated glomerular filtration rate, and higher levels of inflammatory cytokines, as well as prevalent CVD (OR, 1.09; 95% CI, 1.01-1.19; P=0.03). Higher CX3CL1 level also was associated with prevalent diabetes (OR, 1.26; 95% CI, 1.16-1.38; P<0.001) in adjusted models. During a mean follow-up of 6 years, there were 352 deaths, 176 myocardial infarctions, and 484 composite outcomes. In fully adjusted models, 1-SD higher CX3CL1 level increased the hazard for all-cause mortality (1.11; 95% CI, 1.00-1.22; P=0.02) and the composite outcome (1.09; 95% CI, 1.00-1.19; P=0.04). LIMITATIONS Study design did not allow evaluation of changes over time, correlation with progression of phenotypes, or determination of causality of effect. CONCLUSIONS Circulating CX3CL1 level may contribute to both atherosclerotic CVD and diabetes in a CKD cohort. Further studies are required to establish mechanisms through which CX3CL1 affects the pathogenesis of atherosclerosis and diabetes.

openWAR: An open source system for evaluating overall player performance in major league baseball

Abstract Within sports analytics, there is substantial interest in comprehensive statistics intended to capture overall player performance. In baseball, one such measure is wins above replacement (WAR), which aggregates the contributions of a player in each facet of the game: hitting, pitching, baserunning, and fielding. However, current versions of WAR depend upon proprietary data, ad hoc methodology, and opaque calculations. We propose a competitive aggregate measure, openWAR, that is based on public data, a methodology with greater rigor and transparency, and a principled standard for the nebulous concept of a “replacement” player. Finally, we use simulation-based techniques to provide interval estimates for our openWAR measure that are easily portable to other domains.

Mixed modeling of meta-analysis P-values (MixMAP) suggests multiple novel gene loci for low density lipoprotein cholesterol.

Informing missing heritability for complex disease will likely require leveraging information across multiple SNPs within a gene region simultaneously to characterize gene and locus-level contributions to disease phenotypes. To this aim, we introduce a novel strategy, termed Mixed modeling of Meta-Analysis P-values (MixMAP), that draws on a principled statistical modeling framework and the vast array of summary data now available from genetic association studies, to test formally for locus level association. The primary inputs to this approach are: (a) single SNP level p-values for tests of association; and (b) the mapping of SNPs to genomic regions. The output of MixMAP is comprised of locus level estimates and tests of association. In application of MixMAP to summary data from the Global Lipids Gene Consortium, we suggest twelve new loci (PKN, FN1, UGT1A1, PPARG, DMDGH, PPARD, CDK6, VPS13B, GAD2, GAB2, APOH and NPC1) for low-density lipoprotein cholesterol (LDL-C), a causal risk factor for cardiovascular disease and we also demonstrate the potential utility of MixMAP in small data settings. Overall, MixMAP offers novel and complementary information as compared to SNP-based analysis approaches and is straightforward to implement with existing open-source statistical software tools.

Privacy protection and aggregate health data: a review of tabular cell suppression methods (not) employed in public health data systems

Public health research often relies on individuals’ confidential medical data. Therefore, data collecting entities, such as states, seek to disseminate this medical data as widely as possible while still maintaining the privacy of the individual for legal and ethical reasons. One common way in which this medical data is released is through the use of Web-based Data Query Systems (WDQS). In this article, we examined WDQS listed in the National Association for Public Health Statistics and Information Systems (NAPHSIS) specifically reviewing them for how they prevent statistical disclosure in queries that produce a tabular response. One of the most common methods to combat this type of disclosure is through the use of suppression, that is, if a cell count in a table is below a certain threshhold, the true value is suppressed. This technique does work to prevent the direct disclosure of small cell counts, however, primary suppression by itself is not always enough to preserve privacy in tabular data. Here, we present several real examples of tabular response queries that employ suppression, but we are able to infer the values of the suppressed cells, including cells with 1 counts, which could be linked to auxiliary data sources and thus has the possibility to create an identity disclosure. We seek to stimulate awareness of the potential for disclosure of information that individuals may wish to keep private through an online query system. This research is undertaken in the hope that privacy concerns can be dealt with preemptively rather than only after a major disclosure has taken place. In the wake of a such an event, a major concern is that state and local officials would react to this by permanently shutting down these sites and cutting off a valuable source of research data.

Assessing the privacy of randomized vector-valued queries to a database using the area under the receiver operating characteristic curve

As the amount of data generated continues to increase, consideration of individuals’ privacy is a growing concern. As a result, there has been a vast quantity of research done on methods of statistical disclosure control. Some of these methods propose to release a randomized version of the data rather than the actual data. While methods of this type certainly offer some layer of protection, there is still the potential for private information to be disclosed. Quantifying the level of privacy provided by these methods is often difficult. In the past, a method for assessing privacy using the receiver operating characteristic (ROC) curve based on ideas related to differential privacy was proposed. However, the method was only demonstrated for univariate randomized releases. Here, the ROC-based privacy measure is extended to the release of randomized vectors.

A comparison of machine learning techniques for taxonomic classification of teeth from the Family Bovidae

Abstract This study explores the performance of machine learning algorithms on the classification of fossil teeth in the Family Bovidae. Isolated bovid teeth are typically the most common fossils found in southern Africa and they often constitute the basis for paleoenvironmental reconstructions. Taxonomic identification of fossil bovid teeth, however, is often imprecise and subjective. Using modern teeth with known taxons, machine learning algorithms can be trained to classify fossils. Previous work by Brophy et al. [Quantitative morphological analysis of bovid teeth and implications for paleoenvironmental reconstruction of plovers lake, Gauteng Province, South Africa, J. Archaeol. Sci. 41 (2014), pp. 376–388] uses elliptical Fourier analysis of the form (size and shape) of the outline of the occlusal surface of each tooth as features in a linear discriminant analysis (LDA) framework. This manuscript expands on that previous work by exploring how different machine learning approaches classify the teeth and testing which technique is best for classification. In addition to LDA, four other machine learning techniques were considered (neural networks, nuclear penalized multinomial regression,random forests, and support vector machines) with support vector machines and random forests performing the best in terms of log loss and classification rate.

Confidence Intervals for the Area Under the Receiver Operating Characteristic Curve in the Presence of Ignorable Missing Data: Confidence Intervals for AUC with Missing Data

Receiver operating characteristic curves are widely used as a measure of accuracy of diagnostic tests and can be summarised using the area under the receiver operating characteristic curve (AUC). Often, it is useful to construct a confidence interval for the AUC; however, because there are a number of different proposed methods to measure variance of the AUC, there are thus many different resulting methods for constructing these intervals. In this article, we compare different methods of constructing Wald-type confidence interval in the presence of missing data where the missingness mechanism is ignorable. We find that constructing confidence intervals using multiple imputation based on logistic regression gives the most robust coverage probability and the choice of confidence interval method is less important. However, when missingness rate is less severe (e.g. less than 70%), we recommend using Newcombes Wald method for constructing confidence intervals along with multiple imputation using predictive mean matching.