A framework for longitudinal latent factor modelling of treatment response in clinical trials with applications to Psoriatic Arthritis and Rheumatoid Arthritis.

OBJECTIVE: Clinical trials involve the collection of a wealth of data, comprising multiple diverse measurements performed at baseline and follow-up visits over the course of a trial. The most common primary analysis is restricted to a single, potentially composite endpoint at one time point. While such an analytical focus promotes simple and replicable conclusions, it does not necessarily fully capture the multi-faceted effects of a drug in a complex disease setting. Therefore, to complement existing approaches, we set out here to design a longitudinal multivariate analytical framework that accepts as input an entire clinical trial database, comprising all measurements, patients, and time points across multiple trials. METHODS: Our framework composes probabilistic principal component analysis with a longitudinal linear mixed effects model, thereby enabling clinical interpretation of multivariate results, while handling data missing at random, and incorporating covariates and covariance structure in a computationally efficient and principled way. RESULTS: We illustrate our approach by applying it to four phase III clinical trials of secukinumab in Psoriatic Arthritis (PsA) and Rheumatoid Arthritis (RA). We identify three clinically plausible latent factors that collectively explain 74.5% of empirical variation in the longitudinal patient database. We estimate longitudinal trajectories of these factors, thereby enabling joint characterisation of disease progression and drug effect. We perform benchmarking experiments demonstrating our method's competitive performance at estimating average treatment effects compared to existing statistical and machine learning methods, and showing that our modular approach leads to relatively computationally efficient model fitting. CONCLUSION: Our multivariate longitudinal framework has the potential to illuminate the properties of existing composite endpoint methods, and to enable the development of novel clinical endpoints that provide enhanced and complementary perspectives on treatment response.

All that Glitters Is not Gold: Type-I Error Controlled Variable Selection from Clinical Trial Data.

Clinical trials are primarily conducted to estimate causal effects, but the data collected can also be invaluable for additional research, such as identifying prognostic measures of disease or biomarkers that predict treatment efficacy. However, these exploratory settings are prone to false discoveries (type-I errors) due to the multiple comparisons they entail. Unfortunately, many methods fail to address this issue, in part because the algorithms used are generally designed to optimize predictions and often only provide the measures used for variable selection, such as machine learning model importance scores, as a byproduct. To address the resulting unclear uncertainty in the selection sets, the knockoff framework offers a model-agnostic, robust approach to variable selection with guaranteed type-I error control. Here, we review the knockoff framework in the setting of clinical data, highlighting main considerations using simulation studies. We also extend the framework by introducing a novel knockoff generation method that addresses two main limitations of previously suggested methods relevant for clinical development settings. With this new method, we empirically obtain tighter bounds on type-I error control and gain an order of magnitude in computational efficiency in mixed data settings. We demonstrate comparable selections to those of the competing method for identifying prognostic biomarkers for C-reactive protein levels in patients with psoriatic arthritis in four clinical trials. Our work increases access to the knockoff framework for variable selection from clinical trial data. Hereby, this paper helps to address the current replicability crisis which can result in unnecessary research efforts, increased patient burden, and avoidable costs.

Lifetime exposure to violence and other life stressors and hair cortisol concentration in women.

Women are exposed to a variety of life stressors, particularly violence, during their lifetime which increases the risk of developing various psychiatric and somatic diseases, with the dysregulated secretion of cortisol as one potential biological mechanism. We examined the association between violence and other life stressors and hair cortisol concentration (HCC) in a population of urban women. We included 470 adult women (age = 21-86 years) attending the Cancer Detection Clinic in Iceland. The Life Stressor Checklist-Revised (LSC-R; 30-items) was used to assess exposure. HCC was measured with liquid chromatography coupled with tandem mass spectrometry. We used linear regression models to assess the association between life stressors and log-transformed HCC. The median HCC (pg/mg) in the study population was 4.9 (range 0.6-616.6). HCC was not associated with background covariates, including age (p = 0.868), education level (p = 0.824), marital status (p = 0.545), income (p = 0.363), occupation (p = 0.192), but associated with current smoking (p = 0.013). We noted a 3.3% (95% CI: 0.17-6.6%) associated increase in HCC per endorsed life stressor after adjusting for age and smoking, while non-violent life stressors were not associated with HCC. Per endorsed violence item, we observed a 10.2% (95% CI: 1.4-19.7%) associated increase in HCC after age and smoking adjustment. Women with lifetime exposure to both physical and sexual violence presented with higher HCC than unexposed women (p = 0.010), after age and smoking adjustment. Lifetime exposure to violence was associated with higher levels of HCC in a community sample of women. These findings need confirmation with prospective studies.

Clinically relevant patient clusters identified by machine learning from the clinical development programme of secukinumab in psoriatic arthritis.

OBJECTIVES: Identify distinct clusters of psoriatic arthritis (PsA) patients based on their baseline articular, entheseal and cutaneous disease manifestations and explore their clinical and therapeutic value. METHODS: Pooled baseline data in PsA patients (n=1894) treated with secukinumab across four phase 3 studies (FUTURE 2-5) were analysed to determine phenotypes based on clusters of clinical indicators. Finite mixture models methodology was applied to generate clinical clusters and mean longitudinal responses were compared between secukinumab doses (300 vs 150 mg) across identified clusters and clinical indicators through week 52 using machine learning (ML) techniques. RESULTS: Seven distinct patient clusters were identified. Cluster 1 (very-high (VH) - SWO/TEN (swollen/tender); n=187) was characterised by VH polyarticular burden for both tenderness and swelling of joints, while cluster 2 (H (high) - TEN; n=251) was marked by high polyarticular burden in tender joints and cluster 3 (H - Feet - Dactylitis; n=175) by high burden in joints of feet and dactylitis. For cluster 4 (L (Low) - Nails - Skin; n=209), cluster 5 (L - skin; n=283), cluster 6 (L - Nails; n=294) and cluster 7 (L; n=495) articular burden was low but nail and skin involvement was variable, with cluster 7 marked by mild disease activity across all domains. Greater improvements in the longitudinal responses for enthesitis in cluster 2, enthesitis and Psoriasis Area and Severity Index (PASI) in cluster 4 and PASI in cluster 6 were shown for secukinumab 300 mg compared with 150 mg. CONCLUSIONS: PsA clusters identified by ML follow variable response trajectories indicating their potential to predict precise impact on patients' outcomes. TRIAL REGISTRATION NUMBERS: NCT01752634, NCT01989468, NCT02294227, NCT02404350.

Using knockoffs for controlled predictive biomarker identification.

One of the key challenges of personalized medicine is to identify which patients will respond positively to a given treatment. The area of subgroup identification focuses on this challenge, that is, identifying groups of patients that experience desirable characteristics, such as an enhanced treatment effect. A crucial first step towards the subgroup identification is to identify the baseline variables (eg, biomarkers) that influence the treatment effect, which are known as predictive variables. Many subgroup discovery algorithms return importance scores that capture the variables' predictive strength. However, a major limitation of these scores is that they do not answer the core question: "Which variables are actually predictive?" With our work we answer this question by using the knockoff framework, which is a general framework for controlling the false discovery rate when performing prognostic variable selection. In contrast, our work is the first that uses knockoffs for predictive variable selection. We introduce two novel knockoff filters: one parametric, building on variable importance scores derived from a penalized linear regression model, and one non-parametric, building on causal forest variable importance scores. We conduct extensive simulations to validate performance of the proposed methodology and we also apply the proposed methods to data from a randomized clinical trial.

Sequential knockoffs for continuous and categorical predictors: With application to a large psoriatic arthritis clinical trial pool.

Knockoffs provide a general framework for controlling the false discovery rate when performing variable selection. Much of the Knockoffs literature focuses on theoretical challenges and we recognize a need for bringing some of the current ideas into practice. In this paper we propose a sequential algorithm for generating knockoffs when underlying data consists of both continuous and categorical (factor) variables. Further, we present a heuristic multiple knockoffs approach that offers a practical assessment of how robust the knockoff selection process is for a given dataset. We conduct extensive simulations to validate performance of the proposed methodology. Finally, we demonstrate the utility of the methods on a large clinical data pool of more than 2000 patients with psoriatic arthritis evaluated in four clinical trials with an IL-17A inhibitor, secukinumab (Cosentyx), where we determine prognostic factors of a well established clinical outcome. The analyses presented in this paper could provide a wide range of applications to commonly encountered datasets in medical practice and other fields where variable selection is of particular interest.

DNA Methylation Dynamics of Germinal Center B Cells Are Mediated by AID.

Changes in DNA methylation are required for the formation of germinal centers (GCs), but the mechanisms of such changes are poorly understood. Activation-induced cytidine deaminase (AID) has been recently implicated in DNA demethylation through its deaminase activity coupled with DNA repair. We investigated the epigenetic function of AID in vivo in germinal center B cells (GCBs) isolated from wild-type (WT) and AID-deficient (Aicda(-/-)) mice. We determined that the transit of B cells through the GC is associated with marked locus-specific loss of methylation and increased methylation diversity, both of which are lost in Aicda(-/-) animals. Differentially methylated cytosines (DMCs) between GCBs and naive B cells (NBs) are enriched in genes that are targeted for somatic hypermutation (SHM) by AID, and these genes form networks required for B cell development and proliferation. Finally, we observed significant conservation of AID-dependent epigenetic reprogramming between mouse and human B cells.

Variability in DNA methylation defines novel epigenetic subgroups of DLBCL associated with different clinical outcomes.

Diffuse large B-cell lymphoma (DLBCL) is the most common aggressive form of non-Hodgkin lymphoma with variable biology and clinical behavior. The current classification does not fully explain the biological and clinical heterogeneity of DLBCLs. In this study, we carried out genomewide DNA methylation profiling of 140 DLBCL samples and 10 normal germinal center B cells using the HpaII tiny fragment enrichment by ligation-mediated polymerase chain reaction assay and hybridization to a custom Roche NimbleGen promoter array. We defined methylation disruption as a main epigenetic event in DLBCLs and designed a method for measuring the methylation variability of individual cases. We then used a novel approach for unsupervised hierarchical clustering based on the extent of DNA methylation variability. This approach identified 6 clusters (A-F). The extent of methylation variability was associated with survival outcomes, with significant differences in overall and progression-free survival. The novel clusters are characterized by disruption of specific biological pathways such as cytokine-mediated signaling, ephrin signaling, and pathways associated with apoptosis and cell-cycle regulation. In a subset of patients, we profiled gene expression and genomic variation to investigate their interplay with methylation changes. This study is the first to identify novel epigenetic clusters of DLBCLs and their aberrantly methylated genes, molecular associations, and survival.

A hybrid mechanism of action for BCL6 in B cells defined by formation of functionally distinct complexes at enhancers and promoters.

The BCL6 transcriptional repressor is required for the development of germinal center (GC) B cells and diffuse large B cell lymphomas (DLBCLs). Although BCL6 can recruit multiple corepressors, its transcriptional repression mechanism of action in normal and malignant B cells is unknown. We find that in B cells, BCL6 mostly functions through two independent mechanisms that are collectively essential to GC formation and DLBCL, both mediated through its N-terminal BTB domain. These are (1) the formation of a unique ternary BCOR-SMRT complex at promoters, with each corepressor binding to symmetrical sites on BCL6 homodimers linked to specific epigenetic chromatin features, and (2) the "toggling" of active enhancers to a poised but not erased conformation through SMRT-dependent H3K27 deacetylation, which is mediated by HDAC3 and opposed by p300 histone acetyltransferase. Dynamic toggling of enhancers provides a basis for B cells to undergo rapid transcriptional and phenotypic changes in response to signaling or environmental cues.

Mechanism-based epigenetic chemosensitization therapy of diffuse large B-cell lymphoma.

UNLABELLED: Although aberrant DNA methylation patterning is a hallmark of cancer, the relevance of targeting DNA methyltransferases (DNMT) remains unclear for most tumors. In diffuse large B-cell lymphoma (DLBCL) we observed that chemoresistance is associated with aberrant DNA methylation programming. Prolonged exposure to low-dose DNMT inhibitors (DNMTI) reprogrammed chemoresistant cells to become doxorubicin sensitive without major toxicity in vivo. Nine genes were recurrently hypermethylated in chemoresistant DLBCL. Of these, SMAD1 was a critical contributor, and reactivation was required for chemosensitization. A phase I clinical study was conducted evaluating azacitidine priming followed by standard chemoimmunotherapy in high-risk patients newly diagnosed with DLBCL. The combination was well tolerated and yielded a high rate of complete remission. Pre- and post-azacitidine treatment biopsies confirmed SMAD1 demethylation and chemosensitization, delineating a personalized strategy for the clinical use of DNMTIs. SIGNIFICANCE: The problem of chemoresistant DLBCL remains the most urgent challenge in the clinical management of patients with this disease. We describe a mechanism-based approach toward the rational translation of DNMTIs for the treatment of high-risk DLBCL.

Aberration in DNA methylation in B-cell lymphomas has a complex origin and increases with disease severity.

Despite mounting evidence that epigenetic abnormalities play a key role in cancer biology, their contributions to the malignant phenotype remain poorly understood. Here we studied genome-wide DNA methylation in normal B-cell populations and subtypes of B-cell non-Hodgkin lymphoma: follicular lymphoma and diffuse large B-cell lymphomas. These lymphomas display striking and progressive intra-tumor heterogeneity and also inter-patient heterogeneity in their cytosine methylation patterns. Epigenetic heterogeneity is initiated in normal germinal center B-cells, increases markedly with disease aggressiveness, and is associated with unfavorable clinical outcome. Moreover, patterns of abnormal methylation vary depending upon chromosomal regions, gene density and the status of neighboring genes. DNA methylation abnormalities arise via two distinct processes: i) lymphomagenic transcriptional regulators perturb promoter DNA methylation in a target gene-specific manner, and ii) aberrant epigenetic states tend to spread to neighboring promoters in the absence of CTCF insulator binding sites.

methylKit: a comprehensive R package for the analysis of genome-wide DNA methylation profiles.

DNA methylation is a chemical modification of cytosine bases that is pivotal for gene regulation, cellular specification and cancer development. Here, we describe an R package, methylKit, that rapidly analyzes genome-wide cytosine epigenetic profiles from high-throughput methylation and hydroxymethylation sequencing experiments. methylKit includes functions for clustering, sample quality visualization, differential methylation analysis and annotation features, thus automating and simplifying many of the steps for discerning statistically significant bases or regions of DNA methylation. Finally, we demonstrate methylKit on breast cancer data, in which we find statistically significant regions of differential methylation and stratify tumor subtypes. methylKit is available at http://code.google.com/p/methylkit.

Base-pair resolution DNA methylation sequencing reveals profoundly divergent epigenetic landscapes in acute myeloid leukemia.

We have developed an enhanced form of reduced representation bisulfite sequencing with extended genomic coverage, which resulted in greater capture of DNA methylation information of regions lying outside of traditional CpG islands. Applying this method to primary human bone marrow specimens from patients with Acute Myelogeneous Leukemia (AML), we demonstrated that genetically distinct AML subtypes display diametrically opposed DNA methylation patterns. As compared to normal controls, we observed widespread hypermethylation in IDH mutant AMLs, preferentially targeting promoter regions and CpG islands neighboring the transcription start sites of genes. In contrast, AMLs harboring translocations affecting the MLL gene displayed extensive loss of methylation of an almost mutually exclusive set of CpGs, which instead affected introns and distal intergenic CpG islands and shores. When analyzed in conjunction with gene expression profiles, it became apparent that these specific patterns of DNA methylation result in differing roles in gene expression regulation. However, despite this subtype-specific DNA methylation patterning, a much smaller set of CpG sites are consistently affected in both AML subtypes. Most CpG sites in this common core of aberrantly methylated CpGs were hypermethylated in both AML subtypes. Therefore, aberrant DNA methylation patterns in AML do not occur in a stereotypical manner but rather are highly specific and associated with specific driving genetic lesions.

Predicting clinical outcomes after radical nephroureterectomy for upper tract urothelial carcinoma.

BACKGROUND: Novel prognostic factors for patients after radical nephroureterectomy (RNU) for upper tract urothelial carcinoma (UTUC) have recently been described. OBJECTIVE: We tested the prognostic value of pathologic characteristics and developed models to predict the individual probabilities of recurrence-free survival (RFS) and cancer-specific survival (CSS) after RNU. DESIGN, SETTING, AND PARTICIPANTS: Our study included 2244 patients treated with RNU without neoadjuvant or adjuvant therapy at 23 international institutions. Tumor characteristics included T classification, grade, lymph node status, lymphovascular invasion, tumor architecture, location, and concomitant carcinoma in situ (CIS). The cohort was randomly split for development (12 centers, n=1273) and external validation (11 centers, n=971). INTERVENTIONS: All patients underwent RNU. MEASUREMENTS: Univariable and multivariable models addressed RFS, CSS, and comparison of discrimination and calibration with American Joint Committee on Cancer (AJCC) stage grouping. RESULTS AND LIMITATIONS: At a median follow-up of 45 mo, 501 patients (22.3%) experienced disease recurrence and 418 patients (18.6%) died of UTUC. On multivariable analysis, T classification (p for trend <0.001), lymph node metastasis (hazard ratio [HR]: 1.98; p=0.002), lymphovascular invasion (HR: 1.66; p<0.001), sessile tumor architecture (HR: 1.76; p<0.001), and concomitant CIS (HR: 1.33; p=0.035) were associated with disease recurrence. Similarly, T classification (p for trend<0.001), lymph node metastasis (HR: 2.23; p=0.001), lymphovascular invasion (HR: 1.81; p<0.001), and sessile tumor architecture (HR: 1.72; p=0.001) were independently associated with cancer-specific mortality. Our models achieved 76.8% and 81.5% accuracy for predicting RFS and CSS, respectively. In contrast to these well-calibrated models, stratification based upon AJCC stage grouping resulted in a large degree of heterogeneity and did not improve discrimination. CONCLUSIONS: Using standard pathologic features, we developed highly accurate prognostic models for the prediction of RFS and CSS after RNU for UTUC. These models offer improvements in calibration over AJCC stage grouping and can be used for individualized patient counseling, follow-up scheduling, risk stratification for adjuvant therapies, and inclusion criteria for clinical trials.

DNA methyltransferase 1 and DNA methylation patterning contribute to germinal center B-cell differentiation.

The phenotype of germinal center (GC) B cells includes the unique ability to tolerate rapid proliferation and the mutagenic actions of activation induced cytosine deaminase (AICDA). Given the importance of epigenetic patterning in determining cellular phenotypes, we examined DNA methylation and the role of DNA methyltransferases in the formation of GCs. DNA methylation profiling revealed a marked shift in DNA methylation patterning in GC B cells versus resting/naive B cells. This shift included significant differential methylation of 235 genes, with concordant inverse changes in gene expression affecting most notably genes of the NFkB and MAP kinase signaling pathways. GC B cells were predominantly hypomethylated compared with naive B cells and AICDA binding sites were highly overrepresented among hypomethylated loci. GC B cells also exhibited greater DNA methylation heterogeneity than naive B cells. Among DNA methyltransferases (DNMTs), only DNMT1 was significantly up-regulated in GC B cells. Dnmt1 hypomorphic mice displayed deficient GC formation and treatment of mice with the DNA methyltransferase inhibitor decitabine resulted in failure to form GCs after immune stimulation. Notably, the GC B cells of Dnmt1 hypomorphic animals showed evidence of increased DNA damage, suggesting dual roles for DNMT1 in DNA methylation and double strand DNA break repair.

Effects of inbreeding and genetic modification on Aedes aegypti larval competition and adult energy reserves.

BACKGROUND: Genetic modification of mosquitoes offers a promising strategy for the prevention and control of mosquito-borne diseases. For such a strategy to be effective, it is critically important that engineered strains are competitive enough to serve their intended function in population replacement or reduction of wild mosquitoes in nature. Thus far, fitness evaluations of genetically modified strains have not addressed the effects of competition among the aquatic stages and its consequences for adult fitness. We therefore tested the competitive success of combinations of wild, inbred and transgenic (created in the inbred background) immature stages of the dengue vector Aedes aegypti in the presence of optimal and sub-optimal larval diets. RESULTS: The wild strain of Ae. aegypti demonstrated greater performance (based on a composite index of survival, development rate and size) than the inbred strain, which in turn demonstrated greater performance than the genetically modified strain. Moreover, increasing competition through lowering the amount of diet available per larva affected fitness disproportionately: transgenic larvae had a reduced index of performance (95-119%) compared to inbred (50-88%) and wild type larvae (38-54%). In terms of teneral energy reserves (glycogen, lipid and sugar), adult wild type mosquitoes had more reserves directly available for flight, dispersal and basic metabolic functions than transgenic and inbred mosquitoes. CONCLUSIONS: Our study provides a detailed assessment of inter- and intra-strain competition across aquatic stages of wild type, inbred, and transgenic mosquitoes and the impact of these conditions on adult energy reserves. Although it is not clear what competitive level is adequate for success of transgenic strains in nature, strong gene drive mechanisms are likely to be necessary in order to overcome competitive disadvantages in the larval stage that carryover to affect adult fitness.