ABSTRACT
When evaluating the effectiveness of a treatment, policy, or intervention, the desired measure of efficacy may be expensive to collect, not routinely available, or may take a long time to occur. In these cases, it is sometimes possible to identify a surrogate outcome that can more easily, quickly, or cheaply capture the effect of interest. Theory and methods for evaluating the strength of surrogate markers have been well studied in the context of a single surrogate marker measured in the course of a randomized clinical study. However, methods are lacking for quantifying the utility of surrogate markers when the dimension of the surrogate grows. We propose a robust and efficient method for evaluating a set of surrogate markers that may be high-dimensional. Our method does not require treatment to be randomized and may be used in observational studies. Our approach draws on a connection between quantifying the utility of a surrogate marker and the most fundamental tools of causal inference-namely, methods for robust estimation of the average treatment effect. This connection facilitates the use of modern methods for estimating treatment effects, using machine learning to estimate nuisance functions and relaxing the dependence on model specification. We demonstrate that our proposed approach performs well, demonstrate connections between our approach and certain mediation effects, and illustrate it by evaluating whether gene expression can be used as a surrogate for immune activation in an Ebola study.
Subject(s)
Models, Statistical , Humans , Biomarkers , Causality , Computer SimulationABSTRACT
PURPOSE: Current prediction models for advanced age-related macular degeneration (AMD) are based on a restrictive set of risk factors. The objective of this study was to develop a comprehensive prediction model applying a machine learning algorithm allowing selection of the most predictive risk factors automatically. DESIGN: Two population-based cohort studies. PARTICIPANTS: The Rotterdam Study I (RS-I; training set) included 3838 participants 55 years of age or older, with a median follow-up period of 10.8 years, and 108 incident cases of advanced AMD. The Antioxydants, Lipids Essentiels, Nutrition et Maladies Oculaires (ALIENOR) study (test set) included 362 participants 73 years of age or older, with a median follow-up period of 6.5 years, and 33 incident cases of advanced AMD. METHODS: The prediction model used the bootstrap least absolute shrinkage and selection operator (LASSO) method for survival analysis to select the best predictors of incident advanced AMD in the training set. Predictive performance of the model was assessed using the area under the receiver operating characteristic curve (AUC). MAIN OUTCOME MEASURES: Incident advanced AMD (atrophic, neovascular, or both), based on standardized interpretation of retinal photographs. RESULTS: The prediction model retained (1) age, (2) a combination of phenotypic predictors (based on the presence of intermediate drusen, hyperpigmentation in one or both eyes, and Age-Related Eye Disease Study simplified score), (3) a summary genetic risk score based on 49 single nucleotide polymorphisms, (4) smoking, (5) diet quality, (6) education, and (7) pulse pressure. The cross-validated AUC estimation in RS-I was 0.92 (95% confidence interval [CI], 0.88-0.97) at 5 years, 0.92 (95% CI, 0.90-0.95) at 10 years, and 0.91 (95% CI, 0.88-0.94) at 15 years. In ALIENOR, the AUC reached 0.92 at 5 years (95% CI, 0.87-0.98). In terms of calibration, the model tended to underestimate the cumulative incidence of advanced AMD for the high-risk groups, especially in ALIENOR. CONCLUSIONS: This prediction model reached high discrimination abilities, paving the way toward making precision medicine for AMD patients a reality in the near future.
Subject(s)
Machine Learning , Macular Degeneration/diagnosis , Models, Theoretical , Aged , Area Under Curve , Clinical Decision-Making , Disease Progression , Female , Genetics , Genotype , Humans , Life Style , Male , Middle Aged , Phenotype , Retinal Drusen/diagnosis , Risk FactorsABSTRACT
Electronic Health Records (EHRs) often lack reliable annotation of patient medical conditions. Phenorm, an automated unsupervised algorithm to identify patient medical conditions from EHR data, has been developed. PheVis extends PheNorm at the visit resolution. PheVis combines diagnosis codes together with medical concepts extracted from medical notes, incorporating past history in a machine learning approach to provide an interpretable parametric predictor of the occurrence probability for a given medical condition at each visit. PheVis is applied to two real-world use-cases using the datawarehouse of the University Hospital of Bordeaux: i) rheumatoid arthritis, a chronic condition; ii) tuberculosis, an acute condition. Cross-validated AUROC were respectively 0.943 [0.940; 0.945] and 0.987 [0.983; 0.990]. Cross-validated AUPRC were respectively 0.754 [0.744; 0.763] and 0.299 [0.198; 0.403]. PheVis performs well for chronic conditions, though absence of exclusion of past medical history by natural language processing tools limits its performance in French for acute conditions. It achieves significantly better performance than state-of-the-art unsupervised methods especially for chronic diseases.
Subject(s)
Arthritis, Rheumatoid , Natural Language Processing , Algorithms , Electronic Health Records , Humans , Machine LearningABSTRACT
We report a longitudinal analysis of the immune response associated with a fatal case of COVID-19 in Europe. This patient exhibited a rapid evolution towards multiorgan failure. SARS-CoV-2 was detected in multiple nasopharyngeal, blood, and pleural samples, despite antiviral and immunomodulator treatment. Clinical evolution in the blood was marked by an increase (2-3-fold) in differentiated effector T cells expressing exhaustion (PD-1) and senescence (CD57) markers, an expansion of antibody-secreting cells, a 15-fold increase in γδ T cell and proliferating NK-cell populations, and the total disappearance of monocytes, suggesting lung trafficking. In the serum, waves of a pro-inflammatory cytokine storm, Th1 and Th2 activation, and markers of T cell exhaustion, apoptosis, cell cytotoxicity, and endothelial activation were observed until the fatal outcome. This case underscores the need for well-designed studies to investigate complementary approaches to control viral replication, the source of the hyperinflammatory status, and immunomodulation to target the pathophysiological response. The investigation was conducted as part of an overall French clinical cohort assessing patients with COVID-19 and registered in clinicaltrials.gov under the following number: NCT04262921.
Subject(s)
Betacoronavirus/immunology , Coronavirus Infections/complications , Cytokine Release Syndrome/immunology , Multiple Organ Failure/immunology , Pneumonia, Viral/complications , Respiratory Distress Syndrome/immunology , Aged, 80 and over , Betacoronavirus/pathogenicity , COVID-19 , Coronavirus Infections/blood , Coronavirus Infections/immunology , Coronavirus Infections/therapy , Cytokine Release Syndrome/blood , Cytokine Release Syndrome/therapy , Cytokine Release Syndrome/virology , Fatal Outcome , France , Humans , Longitudinal Studies , Lymphocyte Activation , Male , Multiple Organ Failure/blood , Multiple Organ Failure/therapy , Multiple Organ Failure/virology , Pandemics , Pneumonia, Viral/blood , Pneumonia, Viral/immunology , Pneumonia, Viral/therapy , Prospective Studies , Respiratory Distress Syndrome/blood , Respiratory Distress Syndrome/therapy , Respiratory Distress Syndrome/virology , SARS-CoV-2 , Severity of Illness Index , T-Lymphocytes, Cytotoxic/immunology , Th1 Cells/immunology , Th2 Cells/immunologyABSTRACT
MOTIVATION: In some prediction analyses, predictors have a natural grouping structure and selecting predictors accounting for this additional information could be more effective for predicting the outcome accurately. Moreover, in a high dimension low sample size framework, obtaining a good predictive model becomes very challenging. The objective of this work was to investigate the benefits of dimension reduction in penalized regression methods, in terms of prediction performance and variable selection consistency, in high dimension low sample size data. Using two real datasets, we compared the performances of lasso, elastic net, group lasso, sparse group lasso, sparse partial least squares (PLS), group PLS and sparse group PLS. RESULTS: Considering dimension reduction in penalized regression methods improved the prediction accuracy. The sparse group PLS reached the lowest prediction error while consistently selecting a few predictors from a single group. AVAILABILITY AND IMPLEMENTATION: R codes for the prediction methods are freely available at https://github.com/SoufianeAjana/Blisar. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Subject(s)
Sample Size , Least-Squares AnalysisABSTRACT
As gene expression measurement technology is shifting from microarrays to sequencing, the statistical tools available for their analysis must be adapted since RNA-seq data are measured as counts. It has been proposed to model RNA-seq counts as continuous variables using nonparametric regression to account for their inherent heteroscedasticity. In this vein, we propose tcgsaseq, a principled, model-free, and efficient method for detecting longitudinal changes in RNA-seq gene sets defined a priori. The method identifies those gene sets whose expression varies over time, based on an original variance component score test accounting for both covariates and heteroscedasticity without assuming any specific parametric distribution for the (transformed) counts. We demonstrate that despite the presence of a nonparametric component, our test statistic has a simple form and limiting distribution, and both may be computed quickly. A permutation version of the test is additionally proposed for very small sample sizes. Applied to both simulated data and two real datasets, tcgsaseq is shown to exhibit very good statistical properties, with an increase in stability and power when compared to state-of-the-art methods ROAST (rotation gene set testing), edgeR, and DESeq2, which can fail to control the type I error under certain realistic settings. We have made the method available for the community in the R package tcgsaseq.
Subject(s)
Gene Expression , High-Throughput Nucleotide Sequencing/methods , Models, Statistical , Sequence Analysis, RNA/methods , High-Throughput Nucleotide Sequencing/standards , Humans , Longitudinal Studies , Sequence Analysis, RNA/standardsABSTRACT
MOTIVATION: The association between two blocks of 'omics' data brings challenging issues in computational biology due to their size and complexity. Here, we focus on a class of multivariate statistical methods called partial least square (PLS). Sparse version of PLS (sPLS) operates integration of two datasets while simultaneously selecting the contributing variables. However, these methods do not take into account the important structural or group effects due to the relationship between markers among biological pathways. Hence, considering the predefined groups of markers (e.g. genesets), this could improve the relevance and the efficacy of the PLS approach. RESULTS: We propose two PLS extensions called group PLS (gPLS) and sparse gPLS (sgPLS). Our algorithm enables to study the relationship between two different types of omics data (e.g. SNP and gene expression) or between an omics dataset and multivariate phenotypes (e.g. cytokine secretion). We demonstrate the good performance of gPLS and sgPLS compared with the sPLS in the context of grouped data. Then, these methods are compared through an HIV therapeutic vaccine trial. Our approaches provide parsimonious models to reveal the relationship between gene abundance and the immunological response to the vaccine. AVAILABILITY AND IMPLEMENTATION: The approach is implemented in a comprehensive R package called sgPLS available on the CRAN. CONTACT: b.liquet@uq.edu.au SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Subject(s)
Algorithms , Genomics/methods , AIDS Vaccines/immunology , Computer Simulation , Humans , Least-Squares Analysis , Sample SizeABSTRACT
Females have generally more robust immune responses than males for reasons that are not well-understood. Here we used a systems analysis to investigate these differences by analyzing the neutralizing antibody response to a trivalent inactivated seasonal influenza vaccine (TIV) and a large number of immune system components, including serum cytokines and chemokines, blood cell subset frequencies, genome-wide gene expression, and cellular responses to diverse in vitro stimuli, in 53 females and 34 males of different ages. We found elevated antibody responses to TIV and expression of inflammatory cytokines in the serum of females compared with males regardless of age. This inflammatory profile correlated with the levels of phosphorylated STAT3 proteins in monocytes but not with the serological response to the vaccine. In contrast, using a machine learning approach, we identified a cluster of genes involved in lipid biosynthesis and previously shown to be up-regulated by testosterone that correlated with poor virus-neutralizing activity in men. Moreover, men with elevated serum testosterone levels and associated gene signatures exhibited the lowest antibody responses to TIV. These results demonstrate a strong association between androgens and genes involved in lipid metabolism, suggesting that these could be important drivers of the differences in immune responses between males and females.
Subject(s)
Antibodies, Neutralizing/immunology , Gene Expression Regulation/immunology , Influenza Vaccines/immunology , Lipid Metabolism/genetics , Sex Characteristics , Testosterone/immunology , Age Factors , Artificial Intelligence , Cytokines/blood , Enzyme-Linked Immunosorbent Assay , Female , Humans , Leukocytes, Mononuclear , Male , Microarray Analysis , Neutralization Tests , Phosphorylation , STAT3 Transcription Factor/metabolism , Systems Biology , Testosterone/bloodABSTRACT
Gene set analysis methods, which consider predefined groups of genes in the analysis of genomic data, have been successfully applied for analyzing gene expression data in cross-sectional studies. The time-course gene set analysis (TcGSA) introduced here is an extension of gene set analysis to longitudinal data. The proposed method relies on random effects modeling with maximum likelihood estimates. It allows to use all available repeated measurements while dealing with unbalanced data due to missing at random (MAR) measurements. TcGSA is a hypothesis driven method that identifies a priori defined gene sets with significant expression variations over time, taking into account the potential heterogeneity of expression within gene sets. When biological conditions are compared, the method indicates if the time patterns of gene sets significantly differ according to these conditions. The interest of the method is illustrated by its application to two real life datasets: an HIV therapeutic vaccine trial (DALIA-1 trial), and data from a recent study on influenza and pneumococcal vaccines. In the DALIA-1 trial TcGSA revealed a significant change in gene expression over time within 69 gene sets during vaccination, while a standard univariate individual gene analysis corrected for multiple testing as well as a standard a Gene Set Enrichment Analysis (GSEA) for time series both failed to detect any significant pattern change over time. When applied to the second illustrative data set, TcGSA allowed the identification of 4 gene sets finally found to be linked with the influenza vaccine too although they were found to be associated to the pneumococcal vaccine only in previous analyses. In our simulation study TcGSA exhibits good statistical properties, and an increased power compared to other approaches for analyzing time-course expression patterns of gene sets. The method is made available for the community through an R package.
Subject(s)
Computational Biology/methods , Gene Expression Profiling/methods , Models, Biological , Models, Statistical , AIDS Vaccines , Antiretroviral Therapy, Highly Active , Cluster Analysis , Databases, Factual , HIV Infections/drug therapy , HIV Infections/prevention & control , Humans , Influenza Vaccines , Influenza, Human/prevention & controlABSTRACT
We propose an evidence synthesis approach through a degradation model to estimate causal influences of physiological factors on myocardial infarction (MI) and coronary heart disease (CHD). For instance several studies give incidences of MI and CHD for different age strata, other studies give relative or absolute risks for strata of main risk factors of MI or CHD. Evidence synthesis of several studies allows incorporating these disparate pieces of information into a single model. For doing this we need to develop a sufficiently general dynamical model; we also need to estimate the distribution of explanatory factors in the population. We develop a degradation model for both MI and CHD using a Brownian motion with drift, and the drift is modeled as a function of indicators of obesity, lipid profile, inflammation and blood pressure. Conditionally on these factors the times to MI or CHD have inverse Gaussian ([Formula: see text]) distributions. The results we want to fit are generally not conditional on all the factors and thus we need marginal distributions of the time of occurrence of MI and CHD; this leads us to manipulate the inverse Gaussian normal distribution ([Formula: see text]) (an [Formula: see text] whose drift parameter has a normal distribution). Another possible model arises if a factor modifies the threshold. This led us to define an extension of [Formula: see text] obtained when both drift and threshold parameters have normal distributions. We applied the model to results published in five important studies of MI and CHD and their risk factors. The fit of the model using the evidence synthesis approach was satisfactory and the effects of the four risk factors were highly significant.
Subject(s)
Myocardial Infarction/epidemiology , Coronary Disease/epidemiology , Coronary Disease/etiology , Humans , Incidence , Life Tables , Male , Models, Cardiovascular , Models, Statistical , Myocardial Infarction/etiology , Risk FactorsABSTRACT
In response to the COVID-19 pandemic caused by SARS-CoV-2, governments have adopted a wide range of non-pharmaceutical interventions (NPI). These include stringent measures such as strict lockdowns, closing schools, bars and restaurants, curfews, and barrier gestures such as mask-wearing and social distancing. Deciphering the effectiveness of each NPI is critical to responding to future waves and outbreaks. To this end, we first develop a dynamic model of the French COVID-19 epidemics over a one-year period. We rely on a global extended Susceptible-Infectious-Recovered (SIR) mechanistic model of infection that includes a dynamic transmission rate over time. Multilevel data across French regions are integrated using random effects on the parameters of the mechanistic model, boosting statistical power by multiplying integrated observation series. We estimate the parameters using a new population-based statistical approach based on a Kalman filter, used for the first time in analysing real-world data. We then fit the estimated time-varying transmission rate using a regression model that depends on the NPIs while accounting for vaccination coverage, the occurrence of variants of concern (VoC), and seasonal weather conditions. We show that all NPIs considered have an independent significant association with transmission rates. In addition, we show a strong association between weather conditions that reduces transmission in summer, and we also estimate increased transmissibility of VoC.
ABSTRACT
[This corrects the article DOI: 10.1016/j.isci.2021.102711.].
ABSTRACT
Objective: The aim of this study was to develop an accurate regional forecast algorithm to predict the number of hospitalized patients and to assess the benefit of the Electronic Health Records (EHR) information to perform those predictions. Materials and Methods: Aggregated data from SARS-CoV-2 and weather public database and data warehouse of the Bordeaux hospital were extracted from May 16, 2020 to January 17, 2022. The outcomes were the number of hospitalized patients in the Bordeaux Hospital at 7 and 14 days. We compared the performance of different data sources, feature engineering, and machine learning models. Results: During the period of 88 weeks, 2561 hospitalizations due to COVID-19 were recorded at the Bordeaux Hospital. The model achieving the best performance was an elastic-net penalized linear regression using all available data with a median relative error at 7 and 14 days of 0.136 [0.063; 0.223] and 0.198 [0.105; 0.302] hospitalizations, respectively. Electronic health records (EHRs) from the hospital data warehouse improved median relative error at 7 and 14 days by 10.9% and 19.8%, respectively. Graphical evaluation showed remaining forecast error was mainly due to delay in slope shift detection. Discussion: Forecast model showed overall good performance both at 7 and 14 days which were improved by the addition of the data from Bordeaux Hospital data warehouse. Conclusions: The development of hospital data warehouse might help to get more specific and faster information than traditional surveillance system, which in turn will help to improve epidemic forecasting at a larger and finer scale.
ABSTRACT
Knowledge of the mechanisms underpinning the development of protective immunity conferred by mRNA vaccines is fragmentary. Here, we investigated responses to coronavirus disease 2019 (COVID-19) mRNA vaccination via high-temporal resolution blood transcriptome profiling. The first vaccine dose elicited modest interferon and adaptive immune responses, which peaked on days 2 and 5, respectively. The second vaccine dose, in contrast, elicited sharp day 1 interferon, inflammation, and erythroid cell responses, followed by a day 5 plasmablast response. Both post-first and post-second dose interferon signatures were associated with the subsequent development of antibody responses. Yet, we observed distinct interferon response patterns after each of the doses that may reflect quantitative or qualitative differences in interferon induction. Distinct interferon response phenotypes were also observed in patients with COVID-19 and were associated with severity and differences in duration of intensive care. Together, this study also highlights the benefits of adopting high-frequency sampling protocols in profiling vaccine-elicited immune responses.
Subject(s)
COVID-19 Vaccines , COVID-19 , Humans , COVID-19/prevention & control , RNA, Messenger/genetics , Vaccines, Synthetic , Interferons , mRNA VaccinesABSTRACT
OBJECTIVE: Large amounts of health data are becoming available for biomedical research. Synthesizing information across databases may capture more comprehensive pictures of patient health and enable novel research studies. When no gold standard mappings between patient records are available, researchers may probabilistically link records from separate databases and analyze the linked data. However, previous linked data inference methods are constrained to certain linkage settings and exhibit low power. Here, we present ATLAS, an automated, flexible, and robust association testing algorithm for probabilistically linked data. MATERIALS AND METHODS: Missing variables are imputed at various thresholds using a weighted average method that propagates uncertainty from probabilistic linkage. Next, estimated effect sizes are obtained using a generalized linear model. ATLAS then conducts the threshold combination test by optimally combining P values obtained from data imputed at varying thresholds using Fisher's method and perturbation resampling. RESULTS: In simulations, ATLAS controls for type I error and exhibits high power compared to previous methods. In a real-world genetic association study, meta-analysis of ATLAS-enabled analyses on a linked cohort with analyses using an existing cohort yielded additional significant associations between rheumatoid arthritis genetic risk score and laboratory biomarkers. DISCUSSION: Weighted average imputation weathers false matches and increases contribution of true matches to mitigate linkage error-induced bias. The threshold combination test avoids arbitrarily choosing a threshold to rule a match, thus automating linked data-enabled analyses and preserving power. CONCLUSION: ATLAS promises to enable novel and powerful research studies using linked data to capitalize on all available data sources.
Subject(s)
Algorithms , Medical Record Linkage , Bias , Databases, Factual , Diagnostic Tests, Routine , HumansABSTRACT
The identification of patients with coronavirus disease 2019 and high risk of severe disease is a challenge in routine care. We performed cell phenotypic, serum, and RNA sequencing gene expression analyses in severe hospitalized patients (n = 61). Relative to healthy donors, results showed abnormalities of 27 cell populations and an elevation of 42 cytokines, neutrophil chemo-attractants, and inflammatory components in patients. Supervised and unsupervised analyses revealed a high abundance of CD177, a specific neutrophil activation marker, contributing to the clustering of severe patients. Gene abundance correlated with high serum levels of CD177 in severe patients. Higher levels were confirmed in a second cohort and in intensive care unit (ICU) than non-ICU patients (P < 0.001). Longitudinal measurements discriminated between patients with the worst prognosis, leading to death, and those who recovered (P = 0.01). These results highlight neutrophil activation as a hallmark of severe disease and CD177 assessment as a reliable prognostic marker for routine care.
ABSTRACT
BACKGROUND: Brain lipid metabolism appears critical for cognitive aging, but whether alterations in the lipidome relate to cognitive decline remains unclear at the system level. METHODS: We studied participants from the Three-City study, a multicentric cohort of older persons, free of dementia at time of blood sampling, and who provided repeated measures of cognition over 12 subsequent years. We measured 189 serum lipids from 13 lipid classes using shotgun lipidomics in a case-control sample on cognitive decline (matched on age, sex and level of education) nested within the Bordeaux study center (discovery, n = 418). Associations with cognitive decline were investigated using bootstrapped penalized regression, and tested for validation in the Dijon study center (validation, n = 314). FINDINGS: Among 17 lipids identified in the discovery stage, lower levels of the triglyceride TAG50:5, and of four membrane lipids (sphingomyelin SM40:2,2, phosphatidylethanolamine PE38:5(18:1/20:4), ether-phosphatidylethanolamine PEO34:3(16:1/18:2), and ether-phosphatidylcholine PCO34:1(16:1/18:0)), and higher levels of PCO32:0(16:0/16:0), were associated with greater odds of cognitive decline, and replicated in our validation sample. INTERPRETATION: These findings indicate that in the blood lipidome of non-demented older persons, a specific profile of lipids involved in membrane fluidity, myelination, and lipid rafts, is associated with subsequent cognitive decline. FUNDING: The complete list of funders is available at the end of the manuscript, in the Acknowledgement section.
Subject(s)
Aging/blood , Aging/psychology , Cognitive Dysfunction/blood , Cognitive Dysfunction/epidemiology , Lipidomics , Lipids/blood , Aged , Aged, 80 and over , Biomarkers , Case-Control Studies , Cognitive Dysfunction/diagnosis , Cohort Studies , Comorbidity , Female , Geriatric Assessment , Humans , Lipidomics/methods , Male , Public Health Surveillance , Reproducibility of ResultsABSTRACT
RNA-seq studies are growing in size and popularity. We provide evidence that the most commonly used methods for differential expression analysis (DEA) may yield too many false positive results in some situations. We present dearseq, a new method for DEA that controls the false discovery rate (FDR) without making any assumption about the true distribution of RNA-seq data. We show that dearseq controls the FDR while maintaining strong statistical power compared to the most popular methods. We demonstrate this behavior with mathematical proofs, simulations and a real data set from a study of tuberculosis, where our method produces fewer apparent false positives.
ABSTRACT
Electronic medical records data are valuable resources for discovery research. They contain detailed phenotypic information on individual patients, opening opportunities for simultaneously studying multiple phenotypes. A useful tool for such simultaneous assessment is the phenome-wide association study, which relates a genomic or biological marker of interest to a wide spectrum of disease phenotypes, typically defined by the diagnostic billing codes. One challenge arises when the biomarker of interest is expensive to measure on the entire electronic medical record cohort. Performing phenome-wide association study based on supervised estimation using only subjects who have marker measurements may yield limited power. In this paper, we focus on the setting where the marker is measured on a small fraction of the patients while a few surrogate markers such as historical measurements of the biomarker are available on a large number of patients. We propose an efficient semi-supervised estimation procedure to estimate the covariance between the biomarker and the billing code, leveraging the surrogate marker information. We employ surrogate marker values to impute the missing outcome via a two-step semi-non-parametric approach and demonstrate that our proposed estimator is always more efficient than the supervised counterpart without requiring the imputation model to be correct. We illustrate the proposed procedure by assessing the association between the C-reactive protein and some inflammatory diseases with an electronic medical record study of inflammatory bowel disease performed with the Partners HealthCare electronic medical record database where C-reactive protein was only measured for a small fraction of the patients due to budget constraints.
Subject(s)
Data Interpretation, Statistical , Electronic Health Records , Genome-Wide Association Study , Algorithms , Bias , Biomarkers , Inflammatory Bowel DiseasesABSTRACT
Evaluation of immunogenicity is a key step in the clinical development of novel vaccines. T-cell responses to vaccine candidates are typically assessed by intracellular cytokine staining (ICS) using multiparametric flow cytometry. A conventional statistical approach to analyze ICS data is to compare, between vaccine regimens or between baseline and post-vaccination of the same regimen depending on the trial design, the percentages of cells producing a cytokine of interest after ex vivo stimulation of peripheral blood mononuclear cells (PBMC) with vaccine antigens, after subtracting the non-specific response (of unstimulated cells) of each sample. Subtraction of the non-specific response is aimed at capturing the specific response to the antigen, but raises methodological issues related to measurement error and statistical power. We describe here a new statistical approach to analyze ICS data from vaccine trials. We propose a bivariate linear regression model for estimating the non-specific and antigen-specific ICS responses. We benchmarked the performance of the model in terms of both bias and control of type-I and -II errors in comparison with conventional approaches, and applied it to simulated data as well as real pre- and post-vaccination data from two recent HIV vaccine trials (ANRS VRI01 in healthy volunteers and therapeutic VRI02 ANRS 149 LIGHT in HIV-infected participants). The model was as good as the conventional approaches (with or without subtraction of the non-specific response) in all simulation scenarios in terms of statistical performance, whereas the conventional approaches did not provide robust results across all scenarios. The proposed model estimated the T-cell responses to the antigens without any effect of the non-specific response on the specific response, irrespective of the correlation between the non-specific and specific responses. This novel method of analyzing T-cell immunogenicity data based on bivariate modeling is more flexible than conventional methods, and so yields more detailed results and enables accurate interpretation of vaccine-induced response.