Search

We performed a retrospective study to evaluate, under routine circumstances, the tolerance and immunovirological changes associated with antiretroviral regimens that contain nevirapine in 137 patients (88% were antiretroviral experienced). During a mean follow-up of 11 months, 33% of patients reported side effects attributed to nevirapine, and 21% discontinued treatment because of poor tolerance. Administration of antihistamines or corticosteroids at the initiation of treatment was not protective against adverse events (relative risk, 0.82; 95% confidence interval, 0.49-1.38). The proportion of patients with hepatitis C virus (HCV) and/or hepatitis B virus (HBV) coinfection who had alanine aminotransferase levels of >100 IU/L increased from 19.4% at baseline to 42.9% at month 12 of follow-up (P=.02). We noticed a significant increase of the proportion of patients with total cholesterol levels of >5.5 mM (P=.02). We have shown that there is a high level of discontinuation of nevirapine therapy in clinical practice and that side effects were not prevented by administration of antihistamines or corticosteroids. Coinfection with HCV or HBV increased the risk of hepatotoxicity, which lead to the cautious use of nevirapine for such patients.

International audience; OBJECTIVE: To assess the performance of weight related nutritional markers (reported involuntary weight loss greater than 10%, measured weight loss and body mass index-BMI-) in predicting HIV disease progression. DESIGN: Multirisk cohort of HIV-1 infected patients. METHOD: The three nutritional variables were studied in Cox proportional hazard models as time dependant variables. RESULTS: The sample included 2376 subjects (median follow up: 43.1 months), of those 675 experienced an AIDS defining event. After adjustment for well known prognostic factors, the reported weight loss greater than 10% tripled the risk of progression to clinical AIDS (Hazard ratio [HR] 3.0. 95% confidence interval [CI] 2.5-3.7). For measured weight loss under 5%. between 5% and 10% and greater than 10% of baseline weight compared with no weight loss, hazard ratios were respectively 1.8 (CI 1.5-2.2), 2.6 (CI 2.1-3.2) and 5.1 (CI 4.1-6.4). The relative risks of AIDS were 1.7 (CI 1.3-2.2) for BMI between 17 kg/m2 and 18.5 kg/m2, 2.6 (CI 1.7-4.0) for BMI between 16 kg/m2 and 17 kg/m2 and 4.7 (CI 3.0-7.4) for BMI under 16 kg/m2. COMMENTS: Even a limited weight loss measured at a given time during follow up increases the risk of HIV progression; moreover, a simple cross-sectionnal measure of BMI has a good predictive value for subsequent development of clinical disease.

The temporal relationships between plasma human immunodeficiency virus (HIV) RNA levels and evolution of CD4(+) cell counts was studied, using a 2-slope longitudinal mixed model, in 988 patients prospectively enrolled at the initiation of a protease inhibitor--containing regimen of antiretroviral therapy. The short-term slope (baseline through month 4) for mean change in CD4(+) cell count was +21.2 cells/mm(3)/month, and the long-term slope (month 4 through month 24) was +5.5 cells/mm(3)/month. Compared with results from patients without viral response, the long-term slope was 2.5 cells/mm(3)/month higher in patients who had plasma HIV RNA levels of500 copies/mL at month 4 (P.001). It was significantly lower after a rebound in plasma HIV RNA level toor = 500 copies/mL (P.0001), varied according to plasma HIV RNA level at the time of rebound, and was negative only when the plasma HIV RNA level at rebound wasor = 10,000 copies/mL. If CD4(+) cell counts can remain elevated despite virologic treatment failure, such a discrepant response may be transient in patients who have a high plasma HIV RNA level at the time of treatment failure.

Background: Cases of lipodystrophy syndrome and metabolic disorders have been described since the onset of highly active antiretroviral therapy in HIV-infected patients. The aim of our study was to estimate the prevalence of lipodystrophy (LD) and to define the associated lipid profile of these patients. Methods: The following were determined for each patient: lipid profile (cholesterol and its subfractions, atherogenicity ratios, and triglycerides), blood glucose, and immunovirological markers (CD4(+) cell count and plasma viral load). Patients were classified into two groups on the basis of whether or not they presented with clinical signs of LD. Results: Among 233 HIV-infected patients included in the study, 61 cases (26.1%) of lipodystrophy (LD) were noted. Compared with non-LD patients (NLD), LD patients were older men (P10(-4)) with a lower CD4(+) lymphocyte cell count (P0.007) and more often at the AIDS stage (P10(-3)) (OR=3.2 (95% CI: 1.47-6.2)). Multivariate analysis showed a correlation between LD cases and age (10 years older) (OR=1.78 (95% CI: 1.23-2.57), P0.002) and the decrease in CD4(+) cell count (100 CD4(+)/mm(3) lower) (OR=1.31 (95% CI: 1.09-1.58), P0.004). An analysis of lipid subfractions and atherogenicity ratios clearly indicated a proatherogenic lipid profile for the LD patients. Conclusions: The underlying physiopathological mechanism of LD is still unknown. However, the lipid profile of HIV-1-infected patients with a LD syndrome appears to place these patients at an increased risk of progression of atherosclerosis.

In most Sub-Saharan African settings the challenge to standardize HIV-decision criteria implies a better understanding of the CD4 count evolution. In their cohort study of HIV-infected untreated adults living in Cape Town South Africa Holmes and colleagues reported a steeper decline of CD4 count with higher baseline CD4 count. The CD4 count annual decline was estimated at 47 31 and 20 cells/mm/3 in patients with > 500 351- 500 and 50-200 CD4/mm/3 at baseline respectively. These estimates could be biased by informative dropout. The authors stated that they did not find evidence of informative dropout although concerns could be raised on the method used to check its existence. (excerpt)

Many previous studies of renal transplant recipients have demonstrated that weight gain post kidney transplantation (KT) is frequent and may predispose to co morbidity. The aim of this prospective study was to evaluate changes in body composition (BC) during the first two years post KT and to determine predictors of these changes, with a special focus on pre KT parameters. When listed for a KT, 41 patients (14 women - 27 men) were included between 2007 and 2008 in a longitudinal study of evaluation of BC. Fat Free Mass (FFM) and Fat Mass (FM) were estimated by Dual-energy X- ray absorptiometry. At the same time, Extra Cellular Water (ECW) was measured by bio impedance spectroscopy. Cellular Active Mass (CAM) was defined as FFM – ECW. Energy and protein intake (EI –PI), physical activity (PA), biochemical and nutritional parameters were also recorded. Patients were evaluated every 6 months before KT, and 15 days, 1, 3, 6, 12 and 24 months after KT. During the first 2 years post KT, FM increase 0.09 kg/month (p=0.007), FFM by 0.06kg/month (p=0.0556) and MCA by 0.04kg/month (p=0.04). Univariate analysis showed that during the first 30 days post KT, FFM is strongly influenced by male gender, higher BMI, higher PIbefore KT, higher PA before KT and lower CRP post KT. During the first 2 years, FFM evolution is associated with male gender, higher EI and PI post KT. Early post KT evolution of FM is related to high BMI and high cumulative dose of corticosteroids. Long term evolution is associated with EI and use of corticosteroids. Pre KT EI and PI, as well as male gender and BMI influenced significantly the early evolution of MCA. In adjusted analyses, BMI and gender remained independently associated with FM, FFM and CAM. Furthermore, higher FFM level was associated with higher EI. We confirm that successful KT is associated with BC modifications; which can be detected very early post KT. These very early changes are strongly associated with energy, protein intake and physical activity level pre KT. Management of post KT weight gain should be anticipated with a special care on nutritional intake and physical activity in patients waiting kidney transplantation.

Yellow fever vaccine (17D, a live attenuated virus vaccine) was effective and safe in 2 human immunodeficiency virus-infected patients without severe immunosuppression, one of whom traveled to Kenya and the other of whom traveled to Senegal.

Objective We assessed predictors of changes in systolic (SBP) and diastolic (DBP) blood pressure during follow-up and of the development of hypertension in HIV-infected individuals. Methods International cohort collaborative study (D:A:D) of established prospective cohorts of HIV-1-infected patients. Longitudinal analysis of changes in blood pressure (BP) was performed using mixed effects models in 17170 patients. Predictors of development of hypertension during follow-up (systolic BP ≥140 and/or diastolic BP ≥90 mmHg or initiation of antihypertensive treatment) were assessed using Cox models in 8 984 patients with a normal BP level at baseline. Results 73548 BP measurements with a median of 4 per patient (interquartile range [IQR]: 2–6) were recorded over a median follow-up of 2.3 years (IQR: 1.5–2.6). Risk factors significantly associated with a development of higher systolic BP and diastolic BP (differences ≥5 mmHg and P-values 10 months compared with no exposure; P=0.005). Conclusions Increased blood pressure in HIV-infected individuals is associated with established risk factors for hypertension. There was no evidence for an independent deleterious effect of any class of antiretroviral drugs on BP, although the use of NNRTIs was associated with a lower risk of development of hypertension.

Vaccines have been one of the most successful developments in public health in the last years. However, a major challenge still resides in developing effective vaccines against infectious diseases such as HIV or Ebola virus. This can be attributed to our lack of deep knowledge in immunology and the mode of action of immune memory. Mathematical models can help understanding the mechanisms of the immune response, quantifying the underlying biological processes and eventually developing vaccines based on a solid rationale. First, we present a mechanistic model for the dynamics of the humoral immune response following Ebola vaccine immunizations based on ordinary differential equations. The parameters of the model are estimated by likelihood maximization in a population approach, which allows to quantify the process of the immune response and its factors of variability. In particular, the vaccine regimen is found to impact only the response on a short term, while significant differences between subjects of different geographic locations are found at a longer term. This could have implications in the design of future clinical trials. Then, we develop a numerical tool based on dynamic programming for optimizing schedule of repeated injections. In particular, we focus on HIV-infected patients under treatment but unable to recover their immune system. Repeated injections of an immunotherapeutic product (IL-7) are considered for improving the health of these patients. The process is first by a piecewise deterministic Markov model and recent results of the impulse control theory allow to solve the problem numerically with an iterative sequence. We show in a proof-of-concept that this method can be applied to a number of pseudo-patients. All together, these results are part of an effort to develop sophisticated methods for analyzing data from clinical trials to answer concrete clinical questions.; Les vaccins ont été une grande réussite en matière de santé publique au cours des dernières années. Cependant, le développement de vaccins efficaces contre les maladies infectieuses telles que le VIH ou le virus Ebola reste un défi majeur. Cela peut être attribué à notre manque de connaissances approfondies en immunologie et sur le mode d'action de la mémoire immunitaire. Les modèles mathématiques peuvent aider à comprendre les mécanismes de la réponse immunitaire, à quantifier les processus biologiques sous-jacents et à développer des vaccins fondés sur un rationnel scientifique. Nous présentons un modèle mécaniste de la dynamique de la réponse immunitaire humorale après injection d'un vaccin Ebola basé sur des équations différentielles ordinaires. Les paramètres du modèle sont estimés par maximum de vraisemblance dans une approche populationnelle qui permet de quantifier le processus de la réponse immunitaire et ses facteurs de variabilité. En particulier, le schéma vaccinal n'a d'impact que sur la réponse à court terme, alors que des différences significatives entre des sujets de différentes régions géographiques sont observées à plus long terme. Cela pourrait avoir des implications dans la conception des futurs essais cliniques. Ensuite, nous développons un outil numérique basé sur la programmation dynamique pour optimiser des schémas d'injections répétées. En particulier, nous nous intéressons à des patients infectés par le VIH sous traitement mais incapables de reconstruire leur système immunitaire. Des injections répétées d'un produit immunothérapeutique (IL-7) sont envisagées pour améliorer la santé de ces patients. Le processus est modélisé par un modèle de Markov déterministe par morceaux et des résultats récents de la théorie du contrôle impulsionnel permettent de résoudre le problème numériquement à l'aide d'une suite itérative. Nous montrons dans une preuve de concept que cette méthode peut être appliquée à un certain nombre de pseudo-patients. Dans l'ensemble, ces résultats s'intègrent dans un effort de développer des méthodes sophistiquées pour analyser les données d'essais cliniques afin de répondre à des questions cliniques concrètes.

Vaccines have been one of the most successful developments in public healthin the last years. However, a major challenge still resides in developing effective vaccinesagainst infectious diseases such as HIV or Ebola virus. This can be attributed to our lack ofdeep knowledge in immunology and the mode of action of immune memory. Mathematicalmodels can help understanding the mechanisms of the immune response, quantifyingthe underlying biological processes and eventually developing vaccines based on a solidrationale. First, we present a mechanistic model for the dynamics of the humoral immuneresponse following Ebola vaccine immunizations based on ordinary differential equations.The parameters of the model are estimated by likelihood maximization in a populationapproach, which allows to quantify the process of the immune response and its factors ofvariability. The vaccine regimen is found to impact only the response on a short term,while significant differences between subjects of different geographic regions are found ata longer term. This could have implications in the design of future clinical trials. Then,we develop a numerical tool based on dynamic programming for optimizing schedule ofrepeated injections. In particular, we focus on HIV-infected patients under treatment butunable to recover their immune system. Repeated injections of an immunotherapeuticproduct (IL-7) are considered for improving the health of these patients. The process isfirst modeled by a piecewise deterministic Markov model and recent results of the impulsecontrol theory allow to solve the problem numerically with an iterative sequence. We showin a proof-of-concept that this method can be applied to a number of pseudo-patients. Alltogether, these results are part of an effort to develop sophisticated methods for analyzingdata from clinical trials to answer concrete clinical questions., Les vaccins ont été une grande réussite en matière de santé publique au coursdes dernières années. Cependant, le développement de vaccins efficaces contre les maladiesinfectieuses telles que le VIH ou le virus Ebola reste un défi majeur. Cela peut être attribuéà notre manque de connaissances approfondies en immunologie et sur le mode d’actionde la mémoire immunitaire. Les modèles mathématiques peuvent aider à comprendre lesmécanismes de la réponse immunitaire, à quantifier les processus biologiques sous-jacentset à déveloper des vaccins fondés sur un rationnel scientifique. Nous présentons un modèlemécaniste de la dynamique de la réponse immunitaire humorale après injection d’un vaccinEbola basé sur des équations différentielles ordinaires. Les paramètres du modèle sontestimés par maximum de vraisemblance dans une approche populationnelle qui permet dequantifier le processus de la réponse immunitaire et ses facteurs de variabilité. Le schémavaccinal n’a d’impact que sur la réponse à court terme, alors que des différences significativesentre des sujets de différentes régions géographiques sont observées à plus long terme.Cela pourrait avoir des implications dans la conception des futurs essais cliniques. Ensuite,nous développons un outil numérique basé sur la programmation dynamique pouroptimiser des schémas d’injections répétées. Nous nous intéressons en particulier à despatients infectés par le VIH sous traitement mais incapables de reconstruire leur systèmeimmunitaire. Des injections répétées d’un produit immunothérapeutique (IL-7) sont envisagéespour améliorer la santé de ces patients. Le processus est modélisé par un modèlede Markov déterministe par morceaux et des résultats récents de la théorie du contrôleimpulsionnel permettent de résoudre le problème numériquement à l’aide d’une suite itérative.Nous montrons dans une preuve de concept que cette méthode peut être appliquéeà un certain nombre de pseudo-patients. Dans l’ensemble, ces résultats s’intègrent dansun effort de développer des méthodes sophistiquées pour analyser les données d’essaiscliniques afin de répondre à des questions cliniques concrètes.

Assessment of the cumulative incidence of SARS-CoV-2 infections is critical for monitoring the course and extent of the COVID-19 epidemic. Here, we report estimated seroprevalence in the French population and the proportion of infected individuals who developed neutralising antibodies at three points throughout the first epidemic wave. Testing 11,000 residual specimens for anti-SARS-CoV-2 IgG and neutralising antibodies, we find nationwide seroprevalence of 0.41% (95% CI: 0.05–0.88) mid-March, 4.14% (95% CI: 3.31–4.99) mid-April and 4.93% (95% CI: 4.02–5.89) mid-May 2020. Approximately 70% of seropositive individuals have detectable neutralising antibodies. Infection fatality rate is 0.84% (95% CI: 0.70–1.03) and increases exponentially with age. These results confirm that the nationwide lockdown substantially curbed transmission and that the vast majority of the French population remained susceptible to SARS-CoV-2 in May 2020. Our study shows the progression of the first epidemic wave and provides a framework to inform the ongoing public health response as viral transmission continues globally., The percentage of national populations infected during the first stages of the COVID-19 pandemic are unclear owing to limited early testing. Here the authors provide a nation-wide prevalence study of SARS-CoV-2 antibodies in France from the first wave of COVID-19 in 2020, including stratification based on age, sex and region.

Background Dolutegravir (DTG)-based antiretroviral therapy (ART) is highly effective and well-tolerated in adults and is rapidly being adopted globally. We describe the design of the ODYSSEY trial which evaluates the efficacy and safety of DTG-based ART compared with standard-of-care in children and adolescents. The ODYSSEY trial includes nested pharmacokinetic (PK) sub-studies which evaluated pragmatic World Health Organization (WHO) weight-band-based DTG dosing and opened recruitment to children Methods ODYSSEY (Once-daily DTG based ART in Young people vS. Standard thErapY) is an open-label, randomised, non-inferiority, basket trial comparing the efficacy and safety of DTG + 2 nucleos(t) ides (NRTIs) versus standard-of-care (SOC) in HIV-infected children Results Between September 2016 and June 2018, 707 children weighing ≥14 kg were enrolled; including 311 ART-naïve children and 396 children starting second-line. 47% of children were enrolled in Uganda, 21% Zimbabwe, 20% South Africa, 9% Thailand, 4% Europe. 362 (51%) participants were male; median age [range] at enrolment was 12.2 years [2.9–18.0]. 82 (12%) children weighed 14 to Conclusions By employing a basket design, to include ART-naïve and -experienced children, and nested PK sub-studies, the ODYSSEY trial efficiently evaluates multiple scientific questions regarding dosing and effectiveness of DTG-based ART in children. Trial registration NCT, NCT02259127, registered 7th October 2014; EUDRACT, 2014–002632-14, registered 18th June 2014 (https://www.clinicaltrialsregister.eu/ctr-search/trial/2014-002632-14/ES); ISRCTN, ISRCTN91737921, registered 4th October 2014.

Introduced by Leo Breiman in 2001, random forests are a statistical learning method that is widely used in many fields of scientific research both for its ability to describe complex relationships between explanatory variables and a response variable as well as for its ability to handle high dimensional data. In many health applications, repeated measurements over time are available. These are referred to as longitudinal data. The correlations induced by the measurements of the same individual at different times must be taken into account, which is not the case in the classical random forests method. The aim of this thesis is to adapt this method to the analysis of longitudinal data in a high dimensional context. To do so, two approaches are proposed. The first one is based on a semi-parametric mixed-effects model which allows the intra-individual covariance structure to be taken into account in the construction of the random forest. This method was applied to an HIV vaccine trial and enabled to select 21 gene transcripts for which the association with the HIV viral load was in line with the results observed during the primary infection.The second method takes place in the more general framework of regression on metric spaces. In this context, repeated data are treated as curves. We then introduce the concept of Fréchet random forests, which allows to learn relationships between heterogeneous variables, such as curves, images or shapes, in unordered metric spaces. We describe a new way of splitting the nodes of the trees composing the Fréchet random forest and then we detail the prediction procedure for a non-Euclidean output variable. The classical notions of OOB error as well as the variable importance are adapted to the Fréchet random forest. A consistency theorem for Fréchet regressogram predictor using data-dependent partitions is stated and then applied to Fréchet purely uniformly random trees. A simulation study is then carried out to study the behaviour of this new method within the framework of regression on curves, images and scalars. Finally, Fréchet random forest is applied to the analysis of two high dimensional HIV vaccine trials., Introduites par Leo Breiman en 2001, les forêts aléatoires sont une méthode d’apprentissage statistique largement utilisée dans de nombreux domaines de recherche scientifiques tant pour sa capacité à décrire des relations complexes entre des variables explicatives et une variable réponse que pour sa faculté à traiter des données de grande dimension. Dans de nombreuses applications en santé, on dispose de mesures répétées au cours du temps. On parle alors de données longitudinales. Les corrélations induites entre les mesures d'un même individu à différents temps doivent être prises en compte, ce qui n'est pas le cas dans la méthode classique des forêts aléatoires. L'objectif de cette thèse est d'adapter cette méthode à l'analyse des données longitudinales dans un contexte de grande dimension. Pour ce faire, deux approches sont proposées. La première s'appuie sur l'utilisation d'un modèle semi-paramétrique à effets mixtes qui permet de prendre en compte la structure de covariance intra-individuelle dans la construction de la forêt aléatoire. Cette méthode a été appliquée à un essai vaccinal contre le VIH et a permis de sélectionner 21 transcrits de gènes pour lesquels l'association avec la charge virale du VIH était en adéquation avec les résultats observés lors de l'infection primaire.La seconde se place dans le cadre plus général de la régression sur des espaces métriques. Dans ce contexte, les données répétées sont traitées comme des courbes. Nous introduisons alors le concept de forêts aléatoires de Fréchet qui permet d’apprendre des relations entre des variables de natures diverses, comme des courbes, des images ou des formes, dans des espaces métriques non ordonnés. Nous décrivons une nouvelle manière de découper les noeuds des arbres constituant la forêt de Fréchet puis nous détaillons la procédure de prédiction pour une variable de sortie à valeurs dans un espace non euclidien. Les notions classiques d'erreur OOB ainsi que d'importance des variables sont adaptées aux forêts aléatoires de Fréchet. Un théorème de consistance pour les régressogrammes de Fréchet utilisant des partitions données-dépendantes est énoncé puis appliqué aux arbres de Fréchet purement uniformément aléatoires. Une étude de simulations est ensuite menée pour étudier le comportement de cette nouvelle méthode dans le cadre de la régression sur courbes, images et scalaires. Enfin, la méthode des forêts aléatoires de Fréchet est appliquée à l'analyse de deux essais vaccinaux de grande dimension sur le VIH.

SIVmac251-infected Mauritius cynomolgus macaques presenting a M6 MHC haplotype or challenged with a low inoculum dose by mucosal route are models for natural HIV control. Here we characterized by modeling the dynamics of plasma SIV-RNA and of SIV-DNA in blood cells of 16 macaques of the ANRS SIC study.SIV-RNA kinetics was best fitted using a model where the cytotoxic immune response progressively mounted up and reduced actively infected cells half-life (t1/2) from 5.5 days early on to about 0.3 days. The model predicted that the control was achieved in animals able to mount an effective immune response within three months, and this was corroborated by the longitudinal analysis of the CD8+ T-cell antiviral activity measured ex vivo. The control of SIV-RNA was accompanied in parallel by a slow and biphasic decline of SIV-DNA. This unravels the presence of at least two compartments of non-actively infected cells that are not rapidly eliminated by the immune system, one with a rapid turnover rate (t1/2=5.1 days) and predominant as long as SIV-RNA levels are still large, and one with a slow turnover (t1/2=118 days) consistent with the half-life of memory T-cells, and only visible when control is achieved,.In summary, our analysis suggests that the establishment of an efficient CD8+ T-cell response in the first three months of the infection, and that progressively increases overtime is key to achieve SIV-RNA control in this model. Frequent SIV-DNA quantifications allowed identifying that most cells infected after viral peak have a short t1/2 but do not contribute significantly to viral production.One sentence summaryModeling viral dynamics in SIV natural controller macaques predicts that viral control is primarily driven by the capability to establish an efficient cytotoxic response and the viral decline during control unravels distinct compartments of infected cells.

Statistical learning objective is to learn from observed data in order to predict the response for a new sample. In the context of vaccination, the number of features is higher than the number of individuals. This is a degenerate case of statistical analysis which needs specific tools. The regularization algorithms can deal with those drawbacks. Different types of regularization methods can be used which depends on the data set structure but also upon the question. In this work, the main objective was to use the available information with soft-thresholded empirical covariancematrix estimations through SVD decompositions. This solution is particularly efficient in terms of variable selection and computation time. Heterogeneous typed data sets (coming from different sources and also called multiblock data) were at the core of our methodology. Since some data set generations are expensive, it is common to down sample the population acquiring some types of data. This leads to multi-block missing data patterns. The second objective of our methodology is to deal with those missing values using the response values. But the response values are not present in the test data sets and so we have designed a methodology which permits to consider both the cases of missing values in the train or in the test data sets. Thanks to soft-thresholding, our methodology can regularize and select features. This estimator needs only two parameters to be fixed which are the number of components and the maximum number of features to be selected. The corresponding tuning is performed bycross-validation. According to simulations, the proposed method shows very good results comparing to benchmark methods, especially in terms of prediction and computation time. This method has also been applied to several real data sets associated with vaccine, thomboembolic andfood researches.; L’apprentissage statistique consiste à apprendre à partir de données mesurées dans un échantillon d’individus et cherche à prédire la grandeur d’intérêt chez un nouvel individu. Dans le cas de la vaccination, ou dans d’autres cas dont certains présentés dans ce manuscrit, le nombre de variables mesurées dépasse le nombre d’individus observés, c’est un cas dégénéré d’analyse statistique qui nécessite l’utilisation de méthodes spécifiques. Les propriétés des algorithmes de régularisation permettent de gérer ces cas. Il en existe plusieurs types en fonction de la structure des données considérées et du problème qui sont étudiés. Dans le cas de ce travail, l’objectif principal a été d’utiliser l’information disponible à l’issue de décompositions en éléments propres des matrices de covariances transformées via un opérateur de seuillage doux. Cette solution est particulièrement peu coûteuse en termes de temps de calcul et permet la sélection des variables d’intérêt. Nous nous sommes centrés sur les données qualifiées d’hétérogènes, c’est à dire issues de jeux de données qui sont provenant de sources ou de technologies distinctes. On parle aussi de données multiblocs. Les coûts d’utilisation de certaines technologies pouvant être prohibitifs, il est souvent choisi de ne pas acquérircertaines données sur l’ensemble d’un échantillon, mais seulement sur un sous-échantillon d’étude. Dans ce cas, le jeu de données se retrouve amputé d’une partie non négligeable de l’information. La structure des données associée à ces défauts d’acquisition induit une répartition elle-même multibloc de ces données manquantes, on parle alors de données manquantes par blocs. Le second objectif de notre méthode est de gérer ces données manquantes par blocs en s’appuyant sur l’information à prédire, ceci dans le but de créer un modèle prédictif qui puisse gérer les données manquantes aussi bien pour les données d’entraînement que pour celles de test. Cette méthode emprunte au seuillage doux afin de sélectionnerles variables d’intérêt et ne nécessite que deux paramètres à régler qui sont le nombre de composantes et le nombre de variables à sélectionner parmi les covariables. Ce paramétrage est classiquement réalisé par validation croisée. La méthode développée a fait l’objet de simulationsla comparant aux principales méthodes existantes. Elle montre d’excellents résultats en prédiction et en termes de temps de calcul. Elle a aussi été appliquée à plusieurs jeux de données.

In HIV infected patients, antiretroviral therapy suppresses the viral replication which is followed in most patients by a restoration of the CD4+ T cells (CD4) pool. However, it is not the case for some patients called low immunological responders. Injections of interleukin-7 (IL7), a cytokine involved in the CD4 homeostasis, are considered in order to maintain the CD4 levels above 500 cells per microL, the level at which life expectancy is similar to that of the non-infected. INSPIRE trials evaluated the effect of repeated injections of IL7 on low immunological responders.We present a few mechanistic models of the effect of IL7 injections on CD4, which include random effects to account for inter-individual variability. Using these models with a Bayesian approach, the individual parameters of a new patient are sampled, which allows us to make predictions about its CD4 dynamics and thus to personalize the treatment. We propose four adaptive protocols that limit the time spent under 500 CD4 per microL, without increasing the number of injections. Those protocols are implemented into a Shiny app with an easy to use interface, and they could be tested during clinical trials.The viral reservoir, mainly made up of quiescent infected CD4, is the main obstacle to HIV eradication. IL7 injections induce an increase of the level of CD4, hence of the viral reservoir ; the question is then to determine if the injections induce new cell infections or if the reservoir increases in the same way as CD4. We conclude that while some patients presented marker dynamics consistent with the occurrence of new cell infections, this is not the case for the majority of patients. Confirmation of these events and characterization of potential at-risk patients requires additional measurable data in a clinical trial., Chez les patients infectés par le VIH, les traitements antirétroviraux empêchent la réplication virale, ce qui est suivi, dans la plupart des cas, par une restauration de la population des lymphocytes T CD4+ (CD4). Néanmoins ce n’est pas le cas pour certains patients appelés patients à faible réponse immunitaire. Des injections d’interleukine-7 (IL7) exogène, une cytokine impliquée dans l’homéostasie des CD4, sont considérées afin de maintenir les taux de CD4 au-dessus de 500 cellules par microL, taux au-dessus duquel les patients ont une espérance de vie comparable aux personnes non infectées par le VIH. Les essais INSPIRE ont évalué l’effet d’injections répétées d’IL7 chez les patients à faible réponse immunologique.Nous présentons plusieurs modèles mécanistes de l’effet des injections d’IL7 sur les CD4, qui incluent des effets aléatoires afin de tenir compte de la variabilité inter-individuelle. En utilisant ces modèles avec une approche Bayésienne, les paramètres individuels d’un nouveau patient sont échantillonnés, ce qui nous permet de faire des prédictions sur sa dynamique de CD4 et donc de personnaliser le traitement. Nous proposons quatre protocoles adaptatifs permettant de limiter le temps passé sous 500 CD4 par microL, sans pour autant augmenter le nombre d’injections. Ces protocoles ont été implémentés dans une application Shiny présentant une interface facile d’utilisation,et pourront être testés lors d’essais cliniques.Le réservoir viral, principalement constitué de CD4 quiescentes infectées, est la première barrière à l’éradication du VIH. Les injections d’IL7 entraînent une augmentation du nombre de CD4 et donc du réservoir viral ; la question est alors de savoir si les injections provoquent de nouvelles infections cellulaires ou si le réservoir augmente de la même façon que les CD4. Nous concluons que si quelques patients ont présenté des dynamiques de marqueurs compatibles avec la survenue de nouvelles infections de cellules, ce n’est pas le cas de la majorité des patients. La confirmation de ces phénomènes et la caractérisation de potentiels patients à risque nécessite des données supplémentaires mesurables dans un essai clinique.