An evolutionary algorithm for the direct optimization of covariate balance between nonrandomized populations.

Matching reduces confounding bias in comparing the outcomes of nonrandomized patient populations by removing systematic differences between them. Under very basic assumptions, propensity score (PS) matching can be shown to eliminate bias entirely in estimating the average treatment effect on the treated. In practice, misspecification of the PS model leads to deviations from theory and matching quality is ultimately judged by the observed post-matching balance in baseline covariates. Since covariate balance is the ultimate arbiter of successful matching, we argue for an approach to matching in which the success criterion is explicitly specified and describe an evolutionary algorithm to directly optimize an arbitrary metric of covariate balance. We demonstrate the performance of the proposed method using a simulated dataset of 275,000 patients and 10 matching covariates. We further apply the method to match 250 patients from a recently completed clinical trial to a pool of more than 160,000 patients identified from electronic health records on 101 covariates. In all cases, we find that the proposed method outperforms PS matching as measured by the specified balance criterion. We additionally find that the evolutionary approach can perform comparably to another popular direct optimization technique based on linear integer programming, while having the additional advantage of supporting arbitrary balance metrics. We demonstrate how the chosen balance metric impacts the statistical properties of the resulting matched populations, emphasizing the potential impact of using nonlinear balance functions in constructing an external control arm. We release our implementation of the considered algorithms in Python.

Cross design analysis of randomized and observational data - application to continuation rates for a contraceptive intra uterine device containing Levonorgestrel in adolescents and adults.

BACKGROUND: To combine results from a randomized controlled study (RCT) and an observational study (OS) to evaluate discontinuation rate of a levonorgestrel-containing intrauterine contraceptive device (LNG IUD) in a real-life setting. METHODS: We included 253 parous and nulliparous women aged 21-40 years from our own phase II RCT. A total of 1607 women of all ages (including adolescents, < 20 years) were recruited from an OS. We applied the cross design synthesis (CDS) method recommended by the United States General Accounting Office. This method combines the different strengths of RCTs and OSs into one single estimate. RESULTS: Combined continuation rates for parous vs nulliparous women could be estimated more precisely as well as overall continuation rates after one (86.6%) and two years (78.5%), irrespective of age and parity. CONCLUSION: Cross design synthesis allowed more precise estimation of continuation rates of an intrauterine device.

METAINTER: meta-analysis of multiple regression models in genome-wide association studies.

MOTIVATION: Meta-analysis of summary statistics is an essential approach to guarantee the success of genome-wide association studies (GWAS). Application of the fixed or random effects model to single-marker association tests is a standard practice. More complex methods of meta-analysis involving multiple parameters have not been used frequently, a gap that could be explained by the lack of a respective meta-analysis pipeline. Meta-analysis based on combining p-values can be applied to any association test. However, to be powerful, meta-analysis methods for high-dimensional models should incorporate additional information such as study-specific properties of parameter estimates, their effect directions, standard errors and covariance structure. RESULTS: We modified 'method for the synthesis of linear regression slopes' recently proposed in the educational sciences to the case of multiple logistic regression, and implemented it in a meta-analysis tool called METAINTER. The software handles models with an arbitrary number of parameters, and can directly be applied to analyze the results of single-SNP tests, global haplotype tests, tests for and under gene-gene or gene-environment interaction. Via simulations for two-single nucleotide polymorphisms (SNP) models we have shown that the proposed meta-analysis method has correct type I error rate. Moreover, power estimates come close to that of the joint analysis of the entire sample. We conducted a real data analysis of six GWAS of type 2 diabetes, available from dbGaP (http://www.ncbi.nlm.nih.gov/gap). For each study, a genome-wide interaction analysis of all SNP pairs was performed by logistic regression tests. The results were then meta-analyzed with METAINTER. AVAILABILITY: The software is freely available and distributed under the conditions specified on http://metainter.meb.uni-bonn.de. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Novel genetic matching methods for handling population stratification in genome-wide association studies.

BACKGROUND: A usually confronted problem in association studies is the occurrence of population stratification. In this work, we propose a novel framework to consider population matchings in the contexts of genome-wide and sequencing association studies. We employ pairwise and groupwise optimal case-control matchings and present an agglomerative hierarchical clustering, both based on a genetic similarity score matrix. In order to ensure that the resulting matches obtained from the matching algorithm capture correctly the population structure, we propose and discuss two stratum validation methods. We also invent a decisive extension to the Cochran-Armitage Trend test to explicitly take into account the particular population structure. RESULTS: We assess our framework by simulations of genotype data under the null hypothesis, to affirm that it correctly controls for the type-1 error rate. By a power study we evaluate that structured association testing using our framework displays reasonable power. We compare our result with those obtained from a logistic regression model with principal component covariates. Using the principal components approaches we also find a possible false-positive association to Alzheimer's disease, which is neither supported by our new methods, nor by the results of a most recent large meta analysis or by a mixed model approach. CONCLUSIONS: Matching methods provide an alternative handling of confounding due to population stratification for statistical tests for which covariates are hard to model. As a benchmark, we show that our matching framework performs equally well to state of the art models on common variants.

Rare variant testing of imputed data: an analysis pipeline typified.

Important methodological advancements in rare variant association testing have been made recently, among them collapsing tests, kernel methods and the variable threshold (VT) technique. Typically, rare variants from a region of interest are tested for association as a group ('bin'). Rare variant studies are already routinely performed as whole-exome sequencing studies. As an alternative approach, we propose a pipeline for rare variant analysis of imputed data and develop respective quality control criteria. We provide suggestions for the choice and construction of analysis bins in whole-genome application and support the analysis with implementations of standard burden tests (COLL, CMAT) in our INTERSNP-RARE software. In addition, three rare variant regression tests (REG, FRACREG and COLLREG) are implemented. All tests are accompanied with the VT approach which optimizes the definition of 'rareness'. We integrate kernel tests as implemented in SKAT/SKAT-O into the suggested strategies. Then, we apply our analysis scheme to a genome-wide association study of Alzheimer's disease. Further, we show that our pipeline leads to valid significance testing procedures with controlled type I error rates. Strong association signals surrounding the known APOE locus demonstrate statistical power. In addition, we highlight several suggestive rare variant association findings for follow-up studies, including genomic regions overlapping MCPH1, MED18 and NOTCH3. In summary, we describe and support a straightforward and cost-efficient rare variant analysis pipeline for imputed data and demonstrate its feasibility and validity. The strategy can complement rare variant studies with next generation sequencing data.

Bleeding Risk Prediction in Patients Treated with Antithrombotic Drugs According to the Anatomic Site of Bleeding, Indication for Treatment, and Time Since Treatment Initiation.

Background Reasons for the relatively poor performance of bleeding prediction models are not well understood but may relate to differences in predictors for various anatomical sites of bleeding. Methods We pooled individual participant data from four randomized controlled trials of antithrombotic therapy in patients with coronary and peripheral artery diseases, embolic stroke of undetermined source (ESUS), or atrial fibrillation. We examined discrimination and calibration of models for any major bleeding, major gastrointestinal (GI) bleeding, and intracranial hemorrhage (ICH), according to the time since initiation of antithrombotic therapy, and indication for antithrombotic therapy. Results Of 57,813 patients included, 1,948 (3.37%) experienced major bleeding, including 717 (1.24%) major GI bleeding and 274 (0.47%) ICH. The model derived to predict major bleeding at 1 year from any site (c-index, 0.69, 95% confidence interval [CI], 0.68-0.71) performed similarly when applied to predict major GI bleeding (0.71, 0.69-0.74), but less well to predict ICH (0.64, 0.61-0.69). Models derived to predict GI bleeding (0.75, 0.74-0.78) and ICH (0.72, 0.70-0.79) performed better than the general major bleeding model. Discrimination declined over time since the initiation of antithrombotic treatment, stabilizing at approximately 2 years for any major bleeding and major GI bleeding and 1 year for ICH. Discrimination was best for the model predicting ICH in the ESUS population (0.82, 0.78-0.92) and worst for the model predicting any major bleeding in the coronary and peripheral artery disease population (0.66, 0.65-0.69). Conclusion Performance of risk prediction models for major bleeding is affected by site of bleeding, time since initiation of antithrombotic therapy, and indication for antithrombotic therapy.

Corrigendum: Bleeding Risk Prediction in Patients Treated with Antithrombotic Drugs According to the Anatomic Site of Bleeding, Indication for Treatment, and Time Since Treatment Initiation.

[This corrects the article DOI: 10.1055/a-2259-1134.].

Integrated genome-wide pathway association analysis with INTERSNP.

OBJECTIVES: Pathway association analysis (PAA) tests for an excess of moderately significant SNPs in genes from a common pathway. METHODS: We present a Monte-Carlo simulation framework that allows to formulate the main ideas of existing PAA approaches using a self-contained rather than a competitive null hypothesis. A stand-alone implementation in INTERSNP makes time-consuming communication with standard GWAS software redundant. By additional parallelization with the OpenMP API, we achieve a reduction in running time for PAA by orders of magnitude, making a power simulation study for PAA feasible. Our approach properly accounts for linkage disequilibrium and is robust with respect to residual λ inflation. RESULTS: We demonstrate that under simple, realistic disease models, PAA can actually strongly outperform the GWAS single-marker approach. PAA methods that make use of the strength of the SNP association (GenGen, Fisher's combination test), in general, perform better than ratio-based methods (ALIGATOR, SNP ratio), whereas the relative performance of gene-based scoring (ALIGATOR, GenGen) and pathway-based scoring (SNP ratio, Fisher's combination test) depends on the architecture of the assumed disease model. Finally, we present a new PAA score that models independent signals from the same gene in a regression framework and show that it is a reasonable compromise that combines the advantages of existing ideas.

Identification of International Society on Thrombosis and Haemostasis major and clinically relevant non-major bleed events from electronic health records: a novel algorithm to enhance data utilisation from real-world sources.

Introduction: In randomised controlled trials (RCTs), bleeding outcomes are often assessed using definitions provided by the International Society on Thrombosis and Haemostasis (ISTH). Information relating to bleeding events in real-world evidence (RWE) sources are not identified using these definitions. To assist with accurate comparisons between clinical trials and real-world studies, algorithms are required for the identification of ISTH-defined bleeding events in RWE sources. Objectives: To present a novel algorithm to identify ISTH-defined major and clinically-relevant non-major (CRNM) bleeding events in a US Electronic Health Record (EHR) database. Methods: The ISTH definition for major bleeding was divided into three subclauses: fatal bleeds, critical organ bleeds and symptomatic bleeds associated with haemoglobin reductions. Data elements from EHRs required to identify patients fulfilling these subclauses (algorithm components) were defined according to International Classification of Diseases, 9th and 10th Revisions, Clinical Modification disease codes that describe key bleeding events. Other data providing context to bleeding severity included in the algorithm were: 'interaction type' (diagnosis in the inpatient or outpatient setting), 'position' (primary/discharge or secondary diagnosis), haemoglobin values from laboratory tests, blood transfusion codes and mortality data. Results: In the final algorithm, the components were combined to align with the subclauses of ISTH definitions for major and CRNM bleeds. A matrix was proposed to guide identification of ISTH bleeding events in the EHR database. The matrix categorises bleeding events by combining data from algorithm components, including: diagnosis codes, 'interaction type', 'position', decreases in haemoglobin concentrations (≥ 2 g/dL over 48 hours) and mortality. Conclusions: The novel algorithm proposed here identifies ISTH major and CRNM bleeding events that are commonly investigated in RCTs in a real-world EHR data source. This algorithm could facilitate comparison between the frequency of bleeding outcomes recorded in clinical trials and RWE. Validation of algorithm performance is in progress.

Risk of chronic kidney disease in patients with acute kidney injury following a major surgery: a US claims database analysis.

Background: Acute kidney injury (AKI) is a common complication after major surgery. This study assessed the risk of developing or worsening of chronic kidney disease (CKD) and other clinical outcomes in patients experiencing AKI after major surgery. Methods: This retrospective observational study used Optum's de-identified Clinformatics Data Mart Database to investigate cardiorenal outcomes in adult patients at the first AKI event following major surgery. The primary outcome was CKD stage ≥3; secondary outcomes included myocardial infarction (MI), stroke, heart failure, all-cause hospitalization, end-stage kidney disease, need for dialysis or kidney transplant and composite measures. Follow-up was up to 3 years. Additionally, the effect of intercurrent events on the risk of clinical outcomes was assessed. Results: Of the included patients (N = 31 252), most were male (61.9%) and White (68.9%), with a median age of 72 years (interquartile range 64-79). The event rates were 25.5 events/100 patient-years (PY) for CKD stage ≥3, 3.1 events/100 PY for end-stage kidney disease, 3.0 events/100 PY for dialysis and 0.1 events/100 PY for kidney transplants. Additionally, there were 6.9 events/100 PY for MI, 8.7 events/100 PY for stroke and 49.8 events/100 PY for all-cause hospitalization during follow-up. Patients with AKI relapses as intercurrent events were more likely to develop CKD stage ≥3 than those with just one AKI event after major surgery. Conclusion: This analysis demonstrated that patients experiencing AKI following major surgery are at high risk of developing severe CKD or worsening of pre-existing CKD and other cardiorenal clinical outcomes such as MI and stroke.

Clinical characteristics and disease outcomes in non-diabetic chronic kidney disease: retrospective analysis of a US healthcare claims database.

BACKGROUND: The observational, real-world evidence FLIEDER study aimed to describe patient clinical characteristics and investigate clinical outcomes in non-diabetic patients with chronic kidney disease (CKD) using data collected from routine clinical practice in the United States. METHODS: Between 1 January, 2008-31 December, 2018, individuals aged ≥ 18 years, with non-diabetic, stage 3-4 CKD were indexed in the Optum® Clinformatics® Data Mart US healthcare claims database using International Classification of Diseases-9/10 codes for CKD or by laboratory values (estimated glomerular filtration rate [eGFR] 15-59 mL/min/1.73 m2). The primary outcomes were hospitalization for heart failure, a composite kidney outcome of end-stage kidney disease/kidney failure/need for dialysis and worsening of CKD stage from baseline. The effects of the intercurrent events of a sustained post-baseline decline in eGFR ≥ 30%, ≥ 40%, and ≥ 57% on the subsequent risk of the primary outcomes were also assessed. RESULTS: In the main study cohort (N = 504,924), median age was 75.0 years, and 60.5% were female. Most patients (94.7%) had stage 3 CKD at index. Incidence rates for hospitalization for heart failure, the composite kidney outcome, and worsening of CKD stage from baseline were 4.0, 10.3, and 4.4 events/100 patient-years, respectively. The intercurrent event analysis demonstrated that a relative decline in kidney function from baseline significantly increased the risk of cardiorenal events. CONCLUSIONS: This real-world study highlights that patients with non-diabetic CKD are at high risk of serious adverse clinical outcomes, and that this risk is amplified in patients who experienced greater post-baseline eGFR decline.

Worsening of kidney function in patients with atrial fibrillation and chronic kidney disease: evidence from the real-world CALLIPER study.

OBJECTIVE: Evidence is needed on the impact of anticoagulation therapy on kidney function in patients with atrial fibrillation (AF). The objective of this analysis, which is part of the CALLIPER study, was to investigate the risk of worsening kidney function with rivaroxaban 15 mg once daily compared with warfarin in patients with AF and moderate-to-severe chronic kidney disease (CKD) in routine clinical practice in the United States. METHODS: CALLIPER was an observational, retrospective, new-user cohort study. Adult patients with AF in the US IBM Watson MarketScan databases who newly initiated anticoagulation with rivaroxaban 15 mg once daily or warfarin between January 2013 and December 2017 were included. Comparative analysis was performed using Cox proportional hazards regression after adjustment for potential confounding by the stabilized inverse probability of treatment weighting approach and propensity score matching. One of the main study outcomes was worsening kidney function (composite of progression to CKD stage 5, kidney failure, or need for dialysis), besides traditional AF-related outcomes. RESULTS: The cohort included 7368 patients: 5903 (80.1%) initiating warfarin and 1465 (19.9%) initiating rivaroxaban 15 mg once daily. Rivaroxaban 15 mg was associated with a significant 47% reduction in the risk of worsening kidney function versus warfarin (hazard ratio 0.53; 95% confidence interval 0.35-0.78). Similar results were observed in the subgroup of patients with type 2 diabetes. CONCLUSIONS: Rivaroxaban 15 mg may be associated with a lower risk of worsening kidney function as compared with warfarin in the atrial fibrillation population with moderate-to-severe CKD. TRIAL REGISTRATION NUMBER: NCT03359876.

Agreement between coding schemas used to identify bleeding-related hospitalizations in claims analyses of nonvalvular atrial fibrillation patients.

BACKGROUND: Schemas to identify bleeding-related hospitalizations in claims data differ in billing codes used and coding positions allowed. We assessed agreement across bleeding-related hospitalization coding schemas for claims analyses of nonvalvular atrial fibrillation (NVAF) patients on oral anticoagulation (OAC). HYPOTHESIS: We hypothesized that prior coding schemas used to identify bleeding-related hospitalizations in claim database studies would provide varying levels of agreement in incidence rates. METHODS: Within MarketScan data, we identified adults, newly started on OAC for NVAF from January 2012 to June 2015. Billing code schemas developed by Cunningham et al., the US Food and Drug Administration (FDA) Mini-Sentinel program, and Yao et al. were used to identify bleeding-related hospitalizations as a surrogate for major bleeding. Bleeds were subcategorized as intracranial hemorrhage (ICH), gastrointestinal (GI), or other. Schema agreement was assessed by comparing incidence, rates of events/100 person-years (PYs), and Cohen's kappa statistic. RESULTS: We identified 151 738 new-users of OAC with NVAF (CHA2DS2-VASc score = 3, [interquartile range = 2-4] and median HAS-BLED score = 3 [interquartile range = 2-3]). The Cunningham, FDA Mini-Sentinel, and Yao schemas identified any bleeding-related hospitalizations in 1.87% (95% confidence interval [CI]: 1.81-1.94), 2.65% (95% CI: 2.57-2.74), and 4.66% (95% CI: 4.55-4.76) of patients (corresponding rates = 3.45, 4.90, and 8.65 events/100 PYs). Kappa agreement across schemas was weak-to-moderate (κ = 0.47-0.66) for any bleeding hospitalization. Near-perfect agreement (κ = 0.99) was observed with the FDA Mini-Sentinel and Yao schemas for ICH-related hospitalizations, but agreement was weak when comparing Cunningham to FDA Mini-Sentinel or Yao (κ = 0.52-0.53). FDA Mini-Sentinel and Yao agreement was moderate (κ = 0.62) for GI bleeding, but agreement was weak when comparing Cunningham to FDA Mini-Sentinel or Yao (κ = 0.44-0.56). For other bleeds, agreement across schemas was minimal (κ = 0.14-0.38). CONCLUSIONS: We observed varying levels of agreement among 3 bleeding-related hospitalizations schemas in NVAF patients.

Analysis of the progression of systolic blood pressure using imputation of missing phenotype values.

We present a genome-wide association study of a quantitative trait, "progression of systolic blood pressure in time," in which 142 unrelated individuals of the Genetic Analysis Workshop 18 real genotype data were analyzed. Information on systolic blood pressure and other phenotypic covariates was missing at certain time points for a considerable part of the sample. We observed that the dropout process causing missingness is not independent of the initial systolic blood pressure; that is, the data is not missing completely at random. However, after the adjustment for age, the impact of systolic blood pressure on dropouts was no longer significant. Therefore, we decided to impute missing phenotype values by using information from individuals with complete phenotypic data. Progression of systolic blood pressure (∆SBP/∆t) was defined based on the imputed phenotypes and analyzed in a genome-wide fashion. We also conducted an exhaustive genome-wide search for interaction between single-nucleotide polymorphisms (7.14 × 10(10) tests) under an allelic model. The suggested data imputation and the association analysis strategy proved to be valid in the sense that there was no evidence of genome-wide inflation or increased type I error in general. Furthermore, we detected 2 single-nucleotide polymorphisms (SNPs) that met the criterion for genome-wide significance (p≤5 × 10(-8)), which was also confirmed via Monte-Carlo simulation. In view of the rather small sample size, however, the results have to be followed-up in larger studies.

A one-degree-of-freedom test for supra-multiplicativity of SNP effects.

Deviation from multiplicativity of genetic risk factors is biologically plausible and might explain why Genome-wide association studies (GWAS) so far could unravel only a portion of disease heritability. Still, evidence for SNP-SNP epistasis has rarely been reported, suggesting that 2-SNP models are overly simplistic. In this context, it was recently proposed that the genetic architecture of complex diseases could follow limiting pathway models. These models are defined by a critical risk allele load and imply multiple high-dimensional interactions. Here, we present a computationally efficient one-degree-of-freedom "supra-multiplicativity-test" (SMT) for SNP sets of size 2 to 500 that is designed to detect risk alleles whose joint effect is fortified when they occur together in the same individual. Via a simulation study we show that the SMT is powerful in the presence of threshold models, even when only about 30-45% of the model SNPs are available. In addition, we demonstrate that the SMT outperforms standard interaction analysis under recessive models involving just a few SNPs. We apply our test to 10 consensus Alzheimer's disease (AD) susceptibility SNPs that were previously identified by GWAS and obtain evidence for supra-multiplicativity ([Formula: see text]) that is not attributable to either two-way or three-way interaction.