Chantal Guihenneuc-Jouyaux

Combining Expert Opinions in Prior Elicitation

We consider the problem of combining opinions from different experts in an explicitly model-based way to construct a valid subjective prior in a Bayesian statistical approach. We propose a generic approach by considering a hierarchical model accounting for various sources of variation as well as accounting for potential dependence between experts. We apply this approach to two problems. The first problem deals with a food risk assessment problem involving modelling dose-response for Listeria monocytogenes contamination of mice. Two hierarchical levels of variation are considered (between and within experts) with a complex mathematical situation due to the use of an indirect probit regression. The second concerns the time taken by PhD students to submit their thesis in a particular school. It illustrates a complex situation where three hierarchical levels of variation are modelled but with a simpler underlying probability distribution (log-Normal).

Mixed treatment comparison meta-analysis of altered fractionated radiotherapy and chemotherapy in head and neck cancer

OBJECTIVE Different treatments have been investigated in head and neck cancers (HNCs) but not all of them have been appraised using pairwise comparison. This has resulted in failure to directly identify the best treatment with standard methods. Mixed treatment comparison (MTC) meta-analysis allows one to perform simultaneous inference regarding all treatments and select the best among them. STUDY DESIGN AND SETTING We applied MTC models to the Meta-Analyses of Chemotherapy and Radiotherapy in HNC, which pooled individual patient data concerning more than 24,000 patients involved in 102 trials. Fixed- and random-effects models, models with or without consistency factors, possibly adapted to multiarm trials are discussed. RESULTS Altered fractionated concomitant chemoradiotherapy (AF-CRT) leads to the highest probability of survival in nonmetastatic HNC. The probability that AF-CRT is the best treatment is 94% with random-effects models. There was no relevant inconsistency. When only the most recent trials were selected, AF-CRT and concomitant chemoradiotherapy (CRT) were the two best treatments. AF-CRT remains better than CRT but with a lower posterior probability. CONCLUSION MTC is a powerful method for investigating networks of randomized trials. Homogeneity, similarity of trial designs, populations, and the consistency of the network should be thoroughly checked.

Formaldehyde Exposure and Lower Respiratory Infections in Infants: Findings from the PARIS Cohort Study

Background: Certain chemical pollutants can exacerbate lower respiratory tract infections (LRIs), a common childhood ailment. Although formaldehyde (FA) is one of the most common air pollutants found in indoor environments, its impact on infant health is uncertain. Objective: Our aim was to determine the impact of FA exposure on the LRI incidence during the first year of life of infants from the Pollution and Asthma Risk: an Infant Study (PARIS) birth cohort. Methods: FA was measured in a random sample of 196 infants’ dwellings, and exposure to this pollutant was estimated for 2,940 infants using predictive models based on measurements and data about potential determinants of FA levels. Health data were collected from parents by regular self-administered questionnaires. We used multivariate logistic regressions to estimate associations between FA exposure and the occurrence of LRI and wheezy LRI (wLRI), adjusting for potential confounders/risk factors. Results: During the first year of life, 45.8% of infants had at least one LRI, and LRI occurred simultaneously with wheezing in 48.7% of cases. The FA predictive models correctly classified 70% of dwellings as having high or low exposure, and we estimated that 43.3% of infants were exposed throughout the first year to levels of FA > 19.5 µg/m3. FA exposure was significantly associated with LRI and wLRI before and after adjustment for known LRI risk factors/confounders. For an interquartile increase in FA levels (12.4 μg/m3), we estimated a 32% [95% confidence interval (CI): 11, 55] and 41% (95% CI: 14, 74) increase in the incidence of LRI and wLRI, respectively. Conclusion: The findings of this study suggest that infants exposed to FA at an early age have an increased incidence of LRI.

SET translocation is associated with increase in caspase cleaved amyloid precursor protein in CA1 of Alzheimer and Down syndrome patients.

Caspase cleaved amyloid precursor protein (APPcc) and SET are increased and mislocalized in the neuronal cytoplasm in Alzheimer Disease (AD) brains. Translocated SET to the cytoplasm can induce tau hyperphosphorylation. To elucidate the putative relationships between mislocalized APPcc and SET, we studied their level and distribution in the hippocampus of 5 controls, 3 Down syndrome and 10 Alzheimer patients. In Down syndrome and Alzheimer patients, APPcc and SET levels were increased in CA1 and the frequency of both localizations in the neuronal cytoplasm was high in CA1, and low in CA4. As the increase of APPcc is already present at early stages of AD, we overexpressed APPcc in CA1 and the dentate gyrus neurons of adult mice with a lentiviral construct. APPcc overexpression in CA1 and not in the dentate gyrus induced endogenous SET translocation and tau hyperphosphorylation. These data suggest that increase in APPcc in CA1 neurons could be an early event leading to the translocation of SET and the progression of AD through tau hyperphosphorylation.

Laplace Expansions in Markov chain Monte Carlo Algorithms

Complex hierarchical models lead to a complicated likelihood and then, in a Bayesian analysis, to complicated posterior distributions. To obtain Bayes estimates such as the posterior mean or Bayesian confidence regions, it is therefore necessary to simulate the posterior distribution using a method such as an MCMC algorithm. These algorithms often get slower as the number of observations increases, especially when the latent variables are considered. To improve the convergence of the algorithm, we propose to decrease the number of parameters to simulate at each iteration by using a Laplace approximation on the nuisance parameters. We provide a theoretical study of the impact that such an approximation has on the target posterior distribution. We prove that the distance between the true target distribution and the approximation becomes of order O(N−a) with a ∈ (0, 1), a close to 1, as the number of observations N increases. A simulation study illustrates the theoretical results. The approximated MCMC algorithm behaves extremely well on an example which is driven by a study on HIV patients.

Environmental triggers of nocturnal dry cough in infancy: new insights about chronic domestic exposure to formaldehyde in the PARIS birth cohort.

Although formaldehyde is a common indoor pollutant, its impact on respiratory symptoms in childhood remains unclear. The aim of this study was to examine the relation between domestic formaldehyde exposure and occurrence of coughing, one of the most prevalent respiratory symptoms during the first year of life of infants from the PARIS birth cohort involving 3840 healthy full-term babies. The presence of respiratory symptoms, including dry cough at night apart from a cold or chest infection in the past 12 months was reported on a standardized health questionnaire. Formaldehyde exposure was estimated for all infants using a predictive model established from data (both repeated measurements and information about determinants of levels) collected in a random sample of infants from the cohort. An unconditional logistic regression was fitted to study the relation between annual domestic formaldehyde exposure and dry cough at night, adjusting for all potential risk factors/confounders. The prevalence of dry cough at night was 14.9%. Parental history of allergy was found to modify the relation between environmental factors and dry cough. Cockroaches, used mattresses, and family stressor events were associated with dry cough in infants with parental allergy history. Conversely, domestic formaldehyde exposure tended to increase occurrence of dry cough at night only among babies without parental history of allergy (adjusted OR per 10 µg/m(3) increase in levels, single imputation approach: 1.45, 95% CI: 1.08-1.96, and Bayesian approach: 1.12, 0.91-1.36). This study suggests that the impact of indoor environmental exposure on dry cough at night in infancy is different depending on the presence or not of parental history of allergy.

Use in practice of importance sampling for repeated MCMC for Poisson models

Abstract: The Importance Sampling method is used as an alternative approach to MCMC in repeated Bayesian estimations. In the particular context of numerous data sets, MCMC algorithms have to be called on several times which may become computationally expensive. Since Importance Sampling requires a sample from a posteriodistribution, our idea is to use MCMC to generate only a certain number of Markov chains and use them later in the subsequent IS estimations. For each Importance Sampling procedure, the suitable chain is selected by one of three criteria we present here. The first and second criteria are based on the L1 norm of the difference between two posterior distributions and their Kullback-Leibler divergence respectively. The third criterion results from minimizing the variance of IS estimate. A supplementary automatic selection procedure is also proposed to choose those posterior for which Markov chains will be generated and to avoid arbitrary choice of importance functions. The featured methods are illustrated in simulation studies on three types of Poisson model: simple Poisson model, Poisson regression model and Poisson regression model with extra Poisson variability. Different parameter settings are considered.

ASSESSMENT OF EXPOSURE TO PERSISTENT ORGANOCHLORINE COMPOUNDS IN EPIDEMIOLOGICAL STUDIES ON BREAST CANCER: A LITERATURE REVIEW AND PERSPECTIVES FOR THE CECILE STUDY

Abstract Breast cancer is the most frequent neoplastic disease in women representing 50,000 new cases each year in France. The well-established risk factors, as those related to the reproductive history, cannot account for all cases of breast cancer. Other environmental or lifestyle factors need to be explored in depth. Persistent organochlorine compounds (OCs) have attracted attention because of their endocrine disrupting properties that make them possible risk factors for breast cancer, but most epidemiological studies did not report an association between OC concentrations in blood or adipose tissue and breast cancer risk. In these studies, OC levels were measured in biological samples obtained at the time of cancer diagnosis or only a few years before. In this paper, we review the studies on dichlorodiphenyltrichloroethane (DDT) and polychlorobiphenyl (PCB) exposures in relation to breast cancer. We discuss the relevance of OC biological measurements as lifelong exposure indicators, and we describe a new method for assessing exposure to OCs in epidemiological studies. Most studies were carried out recently and reported OC concentrations that were substantially lower than those reported during the 1960s and 1970s. We make the assumption that these OC levels were not reliable indicators, as they were not measured during etiologically relevant periods in a woman’s lifetime, i.e. during the prenatal period, the puberty or the period before a first full-term pregnancy, which are regarded as key periods of vulnerability of mammary gland cells to carcinogens.This may have resulted in non differential exposure misclassification and hence in the absence of an observed association between OC levels and breast cancer in most epidemiological studies. Physiologically-based pharmacokinetic (PBPK) models allow estimating persistent organic pollutant lifetime toxicokinetics profiles retrospectively in women, by taking into account individual differences in metabolism and key events that affect OC kinetics such as lactation and weight variations. PBPK models will be applied to the participants of a large French population-based case-control study including 1080 cases and 1055 controls. Exposure misclassification could have prevented from observing an association between exposure to OCs and breast cancer risk. PBPK models could be used as a novel way of assessing exposure to OCs and to investigate the impact of internal exposure at different time windows on breast cancer incidence.

Convergence Assessment in Latent Variable Models: Application to the Longitudinal Modelling of a Marker of HIV Progression

Infection’ with Human Immunodeficiency Virus type-1 (HIV-1), the virus that leads to AIDS, is associated with a decline in CD4 cell count, a type of white blood cell involved in the immune system. In order to monitor the health status and disease progression of HIV infected patients, CD4 counts have thus been frequently used as a marker. In particular, Markov process models of the natural history of HIV play an important part in AIDS modelling (Longini et al., 1991, Freydman, 1992, Longini, Clark and Karon, 1993, Gentleman et al., 1994, Satten and Longini, 1996).

Dynamics of the risk of smoking-induced lung cancer : A compartmental hidden markov model for longitudinal analysis

Background: To account for the dynamic aspects of carcinogenesis, we propose a compartmental hidden Markov model in which each person is healthy, asymptomatically affected, diagnosed, or deceased. Our model is illustrated using the example of smoking-induced lung cancer. Methods: The model was fitted on a case-control study nested in the European Prospective Investigation into Cancer and Nutrition study, including 757 incident cases and 1524 matched controls. Estimation was done through a Markov Chain Monte Carlo algorithm, and simulations based on the posterior estimates of the parameters were used to provide measures of model fit. We performed sensitivity analyses to assess robustness of our findings. Results: After adjusting for its impact on exposure duration, age was not found to independently drive the risk of lung carcinogenesis, whereas age at starting smoking in ever-smokers and time since cessation in former smokers were found to be influential. Our data did not support an age-dependent time to diagnosis. The estimated time between onset of malignancy and clinical diagnosis ranged from 2 to 4 years. Our approach yielded good performance in reconstructing individual trajectories in both cases (sensitivity >90%) and controls (sensitivity >80%). Conclusion: Our compartmental model enabled us to identify time-varying predictors of risk and provided us with insights into the dynamics of smoking-induced lung carcinogenesis. Its flexible and general formulation enables the future incorporation of disease states, as measured by intermediate markers, into the modeling of the natural history of cancer, suggesting a large range of applications in chronic disease epidemiology.