This paper describes and tests a largely ignored but important preliminary step for comparative effectiveness research: retrospective evidence syntheses to first establish a knowledge base of condition-based medical conditions. By aggregating and organizing what is already known about a treatment or system, gaps in knowledge can be identified and future research designed to meet those gaps.An information synthesis process may also discover that few knowledge gaps in the knowledge base yet exist, the gaps are negligible, and / or treatment effectiveness and study quality is stable across many years, but is simply not clinically important. A consistent finding of low effectiveness is evidence against more research, including exclusion of a treatment from future comparative effectiveness studies. Though proponents of weak treatments or systems may choose to proceed with further research, use of public funds or resources that eventually increase costs to the public are unwarranted.By first establishing a treatment or system knowledge base, at least three comparative effectiveness research decisions are conceivable: (1) treatment or system should be included in future comparative effectiveness trials to establish relative effectiveness for a given condition, (2) has promise but requires more research in a prospective CER trial, or (3) the treatment is less effective than others for a given condition, making future research unnecessary. Thus, a "retroactive comparative effectiveness research method," rCER, is proposed here to identify which treatments are worth including in future prospective trials and which are known to have small to modest effect sizes and are not worth the time and expense of a closer look.The rCER method herein showed that for non-surgical low back pain any treatments did not improve greatly upon the normal and natural pain trajectory for acute low back pain. Therefore, any advantage in pain reduction by any treatment of acute low back pain over back pain's normal course of resolution without care, is quite small, and as such, the incremental cost for the marginal improvement over no treatment becomes quite large. While the quality of non-surgical low back pain studies over the past 34 years has steadily increased, the effect size has not, leading to the conclusion that future research on non-surgical low back pain treatment is unwarranted.

Background: Cancer and diabetes are the second and twelfth leading global causes of death, respectively. Cancer incidence is increased in diabetics compared to non-diabetics. Common pathobiological pathways are shared by the two diseases: hyperglycaemia, hyperinsulinaemia, chronic inflammation and altered concentrations of endogenous hormones. These pathways can all directly or indirectly be linked to chronic hyperglycaemia. Lifestyle factors also affect cancer, diabetes and hyperglycaemia. Hypothesis: Chronic hyperglycaemia is the common biological pathway linking cancer, diabetes and lifestyle factors. Chronic hyperglycaemia can be assessed by monitoring glycated haemoglobin (HbA1c) levels. Aim: The first aim is to investigate whether the link between diabetes and increased cancer risk can be explained by increasing HbA1c levels. Secondly, glycaemic and overall models of lifestyle factors should be developed and compared to determine the relative influence of lifestyle factors on blood glucose level and, subsequently, cancer risk. This could clarify whether improved glycaemic control via lifestyle factors is sufficient to significantly reduce cancer risk. Method: Dose-response meta-analyses on cancer risk and HbA1c levels were performed and the results communicated via a research article. Statistical glycaemic and overall models were developed from published studies on colorectal cancer (CRC), lifestyle factors and HbA1c, via meta-analysis. Log-linear and restricted cubic spline models were considered for studies relating CRC risk to lifestyle factors or HbA1c. Linear models were considered for studies relating HbA1c to lifestyle factors. Only statistically significant models were compared. Results: Increased cancer risk with increasing HbA1c levels was present for a number of cancers, with some cancer types also showing increased risk in the pre-diabetic and normal HbA1c ranges. Comparison of the glycaemic and overall models revealed that HbA1c significantly affected cancer risk and was significantly affected by lifestyle factors. However, the overall effects of lifestyle factors were much stronger than their glycaemic effects (between 9% and 25% difference in risk between overall effects and glycaemic effects at the exposure levels analysed). Glycaemic and overall models for cigarette smoking and chronic stress revealed increased cancer risk with increasing exposure, but decreased cancer risk for increased dietary fibre intake. The glycaemic model for alcohol consumption displayed decreased cancer risk, while the overall model revealed increased cancer risk, emphasising the strong effect of carcinogenic substances in alcohol. Conclusions: Risk for a number of cancers increased with HbA1c levels in diabetic and non-diabetic persons. Cancer prevention by improved blood glucose control seems plausible. The overall effects of lifestyle factors on cancer risk are much stronger than their glycaemic effects. Lifestyle factors alone do not provide enough reduction in blood glucose levels. Other therapeutic strategies for reducing blood glucose levels, such as pharmacotherapeutics or fasting, should be investigated. The possible harmful effects of reducing blood glucose levels, such as neuroglycopaenia, should be considered before implementation of therapeutic strategies. Although there seems to be a strong association between HbA1c and cancer risk, this does not imply causality. The possibility of residual confounding cannot be ignored, even though the most adjusted estimates were used to develop the models, where possible. / MIng (Electrical and Electronic Engineering), North-West University, Potchefstroom Campus, 2014

Research investigating the role of two evolutionary constructs – perceptions of defeat and entrapment – in various psychological problems and processes has burgeoned over recent years. This meta-analysis quantitatively summarised the findings from 38 studies (11,343 participants) which examined relationships between perceptions of defeat and entrapment and four psychological problems commonly encountered in NHS clinical services: depression, suicidality, anxiety problems and Post-Traumatic Stress Disorder (PTSD). All correlations between defeat and entrapment and the four psychological problems were large by Cohen’s (1988) criterion. Correlations between defeat and entrapment and depression were larger than those for the other psychological problem groups, and significantly larger than those for anxiety problems and PTSD. The magnitude of the observed correlations introduces the possibility that defeat and entrapment, and perhaps other evolutionary constructs, may be integral components or driving forces behind all psychological problems. A robust approach to sensitivity analysis provided confidence that the population effect size estimates are robust and were not severely inflated by unpublished studies not included in the meta-analysis. As there was no significant between-study heterogeneity, moderator analyses were undertaken on an exploratory basis. Findings are generally consistent with theoretical predictions from the Involuntary Defeat Strategy, the theoretical model underpinning the literature. Overall, perceptions of defeat and entrapment appear to be strong risk factors for the four psychological problems examined, perhaps representing transdiagnostic processes that are common across various psychological problems. The potential role of defeat and entrapment in mental health assessment, formulation, intervention and evaluation, is considered in detail and limitations of this meta-analysis and of the literature on which it is based are discussed, highlighting areas of research where future work is needed.

Systematic reviews and meta-analyses have been a central pillar of evidence-based practice and policy-making in healthcare over recent years. Traditionally, this has focused on effectiveness evidence from trials.. There is increasing understanding, however, that other study designs also provide essential information and this has led to interest in developing ways to review and synthesis such evidence. Qualitative research has unique potential to illuminate the patient experience. This research has three aims: 1) To review and compare the proposed methods of systematic review and synthesis of qualitative research. 2) To develop and assess two methods of systematic review and synthesis of qualitative research. 3) To compare these two methods and suggest how they might be used in a policy-making context. In addressing these aims, this thesis substantially contributes to debates about the purpose and practice of systematic review and synthesis of qualitative research, particularly in the context of health technology assessment and related pOlicy-making. I undertake a unique critical comparison of the methods suggested for reviewing and synthesising qualitative research, based on their approach to key stages of systematic review. This is used to produce a comprehensive framework for good practice~ I use the framework in two systematic reviews, one about heavy menstrual bleeding using meta-ethnography, and one about hysterectomy using meta-study. These two reviews allow a comparison of the two methods, and in particular explore the impact of expanding the meta-ethnography approach through meta-study, which adds explicit steps to assess the impact of study methodology and theory on findings. The ability of meta-study to unpack the procedures and theories that produce particular findings is key and illuminates the importance of theory in systematic reviews of qualitative research. Through the two systematic reviews, my thesis also contributes to understanding of these reproductive health topic areas through the creation of new insights and concepts from the synthesis. The synthesis of heavy menstrual bleeding studies produced a detailed patient illness model based on women's experiences. In addition, it allowed an understanding of elements that contribute to women's certainty or uncertainty about whether or not their periods could be seen as problematic and requiring medical help. This helps to establish the limitations of the medical model for doctors, as well as women who suffer from heavy menstrual bleeding. The synthesis of hysterectomy studies produced a detailed description of the journey that women make to, and through, hysterectomy, based on their experiences. I also created a theoretical framework, which shows that hysterectomy needs to be understood in the context of personal, physical experiences, together with sociocultural forces that affect the way in which hysterectomy is experienced, and that the interaction of these micro- and macro-concerns mediate through, and affect, relationships with key other people. Methodology affected the research questions posed and the conclusions of research. Comparing the two methods of review and synthesis showed the importance of taking account of the methods and theories that produce research findings. However this additional detail may be at the expense of certainty and requires additional resources.

This study examines the relationship between service learning innovations and improved academics, self-concept, and social or personal growth in middle and high school students. Meta-Analysis is employed to arrive at effect-size estimates for each construct. A historical overview of service learning is presented and a detailed description of the study selection process is provided. The data revealed a moderate relationship between service learning participation and academics, self-concept and social or personal growth in middle and high school students. The findings are presented, and some appropriate conclusions are drawn. A discussion of the implications of these findings and recommendations for future research are also provided.

Introduction et objectifs: Les préparations coliques sont utilisées pour nettoyer le côlon avant de subir une coloscopie. Il existe plusieurs types de préparations coliques; celles-ci contiennent du polyéthylène glycol (PEG), du phosphate de sodium (NaP), du picosulfate (PICOS) ou une solution orale contenant du sulfate (OSS). L'administration en doses fractionnées est recommandées par plusieurs études randomisées mais une étude approfondie de la littérature n'a pas encore été effectuée. Notre objectifs est de déterminer l’efficacité des doses fractionnées comparativement à d’autres régimes posologiques, ainsi que le produit optimal et le volume le plus efficace. Méthode: Nous avons procédé à une recension systématique de la littérature publiée entre janvier 1980 et mars 2014, à partir des bases de données MEDLINE, Embase, Scopus, CENTRAL et ISI Web of Knowledge. Une méta-analyse a été effectuée en incluant les études randomisées qui comparent des doses fractionnées à des doses administrées uniquement la veille de l’examen (non fractionnées). Ont été exclues les études portant sur une population pédiatrique ainsi que celles portant sur des patients hospitalisés ou atteints d’une affection intestinale inflammatoire. L’issue primaire consistait à déterminer la propreté du côlon. Les issues secondaires sont les effets indésirables,les issues procédurales, la volonté des patients à répéter la préparation colique et le temps requis pour reprendre leurs activités normales. Résultats: Quarante-sept études randomisées répondant aux critères d’inclusion ont été répertoriées (n = 13 487 patients). La propreté du côlon était signficativement supérieure avec les doses fractionnées [rapport de cotes (RC) = 2,51; intervalle de confiance à 95 % (IC) = (1,86-3,39)]. Le côlon était significativement plus propre avec les doses fractionnées de PEG [RC=2,60 (1,46-4,63)] de même qu’avec le NaP [RC=9,34 (2,12-41,11)] et le PICOS [RC=3,54 (1,95-6,45)] comparativement aux doses non fractionnées du même produit. La solution de 3 L ou plus de PEG en doses fractionnées a démontré une propreté supérieure à celle du PEG en doses fractionnées de faible volume [RC=1,89 (1,01-3,46)], mais seulement dans les analyses par intention de traitement. La volonté de répéter la préparation colique était supérieur chez les patients ayant reçu des doses fractionnées comparé aux doses non fractionnées [RC=1,90 (1,05-3,46)]; de même, que ceux ayant reçu <3 L de PEG en doses fractionnées comparativement à ceux qui ont PEG ≥ 3L en doses fractionnées [RC=4,95 (2,21-11,10)]. Les différences n’étaient pas significatives pour les autres issues. Conclusion: Les doses fractionnées augmentent le degré de propreté du côlon, en plus d’être le mode d’administration que les patients préfèrent, comparativement aux doses non fractionnées. D’autres études sont nécessaires afin d’évaluer L'OSS et les volumes réduits de PEG. / Abstract : Background & Aims Bowel preparations are used before a colonoscopy to clean the bowel. There are different regimens of preparing the colon for colonoscopy, including polyethylene glycol (PEG), sodium phosphate (NaP), picosulfate (PICO), or oral sulfate solutions (OSS). Split-dose are recommended in several randomized trials but a thorough literature review has not yet been performed. Our aim was to determine the efficacy of split-dose vs other colon preparation regimens, the optimal products for use, and the most effective preparation volumes. Methods We performed systematic searches of MEDLINE, EMBASE, Scopus, CENTRAL, and ISI Web of Knowledge databases, from January 1980 to March 2014. A meta-analysis was done by including all randomized trials that assessed split-dose regimens vs day before colonoscopy preparation. We excluded studies that included pediatric or hospitalized patients, or patients with inflammatory bowel disease. The primary outcome was efficacy of bowel cleansing. Secondary outcomes included side effects, outcomes of procedures, patients’ willingness to repeat the procedure, and the amount of time required for patients to resume daily activities. Results We identified 47 trials that fulfilled our inclusion criteria (n=13,487 patients). Split-dose preparations provided significantly better colon cleansing than all day-before preparations (odds ratio [OR], 2.51; 95% confidence interval [CI], 1.86–3.39), as well as day-before preparations with PEG (OR, 2.60; 95% CI, 1.46–4.63), sodium phosphate (OR, 9.34; 95% CI, 2.12–41.11), or picosulfate (OR, 3.54; 95% CI, 1.95–6.45). PEG split-dose preparations of 3L or more yielded greater bowel cleanliness than lower-volume split-dose regimens (OR, 1.89; 95% CI, 1.01–3.46), but only in intention-to-treat analysis. A higher proportion of patients were willing to repeat split-dose vs day-prior cleansing (OR, 1.90; 95% CI, 1.05–3.46) and low-volume split-dose preparations vs high-volume split-dose preparation (OR, 4.95; 95% CI, 2.21–11.10). There were no differences between preparations in other secondary outcome measures. Conclusion Based on meta-analysis, split-dose regimens increase quality of colon cleansing and are preferred by patients, compared with day-before preparations. Additional research is required to further evaluate oral sulfate solution-based and PEG low-volume regimens.

Recent advances in genotyping technology coupled with an improved understanding of the architecture of linkage disequilibrium across the human genome have resulted in genome-wide association studies (GWAS) becoming a useful and widely applied tool for discovering common genetic variants associated with both quantitative traits and disease risk. After each GWAS was completed, it left behind a set of genotypes and phenotypes, often including anthropometric measures used as covariates. Genetic associations with anthropometric measures are not well characterized, perhaps due to lack of power to detect them in the sample sizes of individual studies. To improve power to detect variants associated with complex phenotypes such as anthropometric traits, data from multiple GWAS can be combined. This thesis describes the methods and results of several such analyses performed as part of the Genome-wide Investigation of ANThropemtric measures (GIANT) consortium, and compares various different methods that can be used to perform combined analyses of GWAS. In particular, the comparisons focus on comparing differences between meta-analysis methods, in which only summary statistics that result from within-study association testing are shared between studies, and mega-analysis methods in which individual-level genotype and phenotype data is analysed together. Finally, a brief discussion of technological means that have the potential to help overcome some of the challenges associated with performing mega-analyses is offered in order to suggest future work that could be undertaken in this area.

Contexte : Les répercussions de l’alcool au niveau des fonctions cognitives sont bien documentées. Certaines hypothèses suggèrent que l’alcool affecte des fonctions cognitives spécifiques alors que d’autres avancent l’hypothèse de déficits diffus. Cependant, une ambigüité persiste concernant quelles fonctions cognitives sont le plus touchées par l’alcool, et à quel point la durée d’abstinence affecte la récupération cognitive. Nous avons procédé à une des premières méta-analyses qui s’intéresse aux différentes fonctions cognitives touchées par la consommation problématique d’alcool et à la durée d’abstinence requise pour une récupération au niveau des cognitions. Méthodes : Une recherche de la littérature a permis d’identifier 62 études évaluant les cognitions chez les personnes présentant des troubles liés à l’utilisation d’alcool. Les estimations de la taille d’effet ont été calculées avec la Comprehensive Meta Analysis –V2 pour les 12 domaines cognitifs suivants : quotient intellectuel, fluidité verbale/langage, vitesse de traitement de l’information, mémoire de travail, attention, résolution de problème/fonctions exécutives, inhibition/impulsivité, apprentissage verbal, mémoire verbale, apprentissage visuel, mémoire visuelle, habiletés visuo-spatiales. Parmi ces 12 domaines cognitifs, 3 estimations de la taille d’effet ont été calculées selon les durées d’abstinences suivantes : court- (<1 mois), moyen- (2 à 12 mois) et long- (>1 an) termes. Résultats : Les résultats ont révélé la présence de dysfonctions modérées dans 11 domaines cognitifs durant l’abstinence à court terme, et dans 10 domaines cognitifs pour le moyen-terme. Des dysfonctions cognitives minimales furent notées durant l’abstinence à long-terme. Conclusions : Ces résultats révèlent des déficits cognitifs significatifs et diffus durant la première année d’abstinence. Déficits qui se normalisent après un an. Ces résultats soutiennent l’hypothèse de déficits cognitifs diffus reliés à l’alcoolisme et suggèrent que la cognition devrait faire partie intégrante du traitement d’alcoolisme. / Background: The cognitive repercussions of alcoholism are well documented. However, the literature remains somewhat ambiguous with which distinct cognitive functions are more susceptible to impairment in alcoholism and to how duration of abstinence affects cognitive recovery. Some theories claim alcohol negatively affects specific cognitive functions while others assert that deficits are more diffuse in nature. We performed the first meta-analysis to examine cognition in alcoholism and how duration of abstinence affects cognitive recovery. Methods: A literature search yielded 62 studies assessing cognitive dysfunction among alcoholics. Effect size estimates were calculated using the Comprehensive Meta-Analysis V2, for the following 12 cognitive domains: intelligence quotient, verbal fluency/language, speed of processing, working memory, attention, problem solving/executive functions, inhibition/impulsivity, verbal learning, verbal memory, visual learning, visual memory, and visuo-spatial abilities. Within these 12 domains, 3 effect size estimates were calculated based on abstinence duration and partitioned into short- (<1 month), intermediate- (2 to 12 months) and long- (>1 year) term abstinence. Results: Findings revealed moderate impairment across 11 cognitive domains during short term abstinence with moderate impairment across 10 domains during intermediate term abstinence, and overall small effect size estimates during long term abstinence. Conclusions: Results suggest significant cognitive dysfunction during the first year following abstinence from alcohol and that long term abstinence yields near normalisation of cognitive function. These findings support the diffuse brain deficits hypothesis. Clinical implications suggest that cognition may need to be considered an integral part of the treatment of alcoholism.

La qualité de la viande englobe des critères d'importance différente suivant l'espèce animale considérée. Chez le porc, la qualité de la viande a un impact économique important et de nombreux travaux ont pu montrer que les interactions entre le type génétique (particulièrement les gènes halothane et Rendement Napole), les conditions d'élevage et les conditions d'abattage des animaux déterminent la qualité de la viande de porc. L'Europe, longtemps leader du marché mondial des produits de la viande de porc, se trouve aujourd'hui confronté à la concurrence d'autres pays, notamment le Brésil, les Etats-unis et la Chine. Ce projet de thèse s'inscrit dans le cadre d'un projet européen multidisciplinaire portant sur l'amélioration de la qualité de la viande de porc et des produits dérivés pour le consommateur. Un des modules est dédié à l'intégration des connaissances passées et celles à venir afin de construire des modèles de prédiction des qualités technologiques et sensorielles de la viande de porc à partir des données se rapportant à l'animal, ainsi qu'aux conditions d'élevage et de pré-abattage. L'objectif de la thèse est d'intégrer l'ensemble des données bibliographiques et expérimentales relatives aux principaux facteurs influençant la qualité de viande dans le cadre d'une méta-analyse, qui est une démarche statistique combinant les résultats d'une série d'études indépendantes sur un problème donné. Les résultats obtenus ont permis de préciser l'effet du gène halothane, des temps de jeûne, de transport et de repos à l'abattoir sur les principaux paramètres d'évaluation de la qualité technologique de la viande de porc. En effet, nos résultats ont pu confirmer l'effet significatif du gène halothane sur l'ensemble des paramètres de qualité étudiés. Cependant, l'effet significatif du gène halothane a été associé à celui de la race uniquement pour expliquer les variations des paramètres suivants : L* (la réflectance), b* (indice de jaune) et les pertes d'exsudat (DL). La méta-régression a permis de mettre en évidence l'effet du poids à l'abattage sur les différences de pH ultime (pHu) entre les génotypes homozygotes NN et nn. Parmi les facteurs de pré-abattage analysés, nos résultats indiquent que le temps de jeûne affecte le pHu et le DL mesurés au niveau du muscle longissimus, alors que le temps de repos à l'abattoir affecte le pHu du muscle semimembranosus. Le transport et l'interaction transport x temps de jeûne n'avaient d'effet significatif que sur les variations de DL. L'effet significatif de l'apport en vitamine E sur le processus d'oxydation des lipides pendant la conservation de la viande, et par conséquent, sur la qualité technologique de la viande a pu être démontré en utilisant un modèle de régressionméta-analytique non linéaire. D'un point de vue statistique, la comparaison des méthodes statistiques de méta-analyse par approche bayésienne et fréquentiste a permis de mettre en évidence l'avantage de la première quant à l'obtention de résultats plus probants lorsqu'il s'agit de méta-analyse sur un nombre de données relativement limité. Parallèlement à l'approche méta-analytique, une étude exploratoire par analyse multivariée a été réalisée afin d'identifier les corrélations qui existent entre un très grand nombre de variables structurées en groupes distincts. Nos résultats ont montré que les variables discriminant les cinq classes de race (Large White et Basque) x système d'élevage (conventionnel, alternatif et extensif) étaient les mêmes que celles qui discriminaient les individus entre eux, en l'occurrence les variables transcriptomiques, chimiques et technologiques. Le groupe composé des variables relatives à la réactivité à l'abattage discriminaient principalement les systèmes d'élevage, en particulier la classe des porcs basques élevés dans le système extensif. / Meat quality includes criteria of different importance depending on the animal species. Meat quality is a significant economic factor in pork production, and numerous publications have shown that it is strongly influenced by the interaction between genetic type (especially the halothane gene (Hal) and Napole yield gene (RN)), breeding conditions, and pre-slaughter and slaughter conditions of animals. The European Union, longtime leader of the globalmarket of pork and pork products, now faces increasing competition from other countries,including Brazil, the United States and China. This thesis project is part of a multidisciplinary European project on improving the quality of pork and pork products for the consumer. One of the modules of this project is dedicated to the integration of existing and upcoming data to produce a model for predicting technological and sensory pork quality, based on animal,production, and slaughter information. The objective of this thesis is to synthesize existing knowledge derived from bibliographic and experimental data relating to the main factors influencing meat quality within a meta-analysis, which is a statistical approach combining the results of a series of independent studies on a given problem. The results of this work have confirmed the effect of the halothane gene, as well as fasting, transport and lairage times on most of the analyzed parameters of technological pork meat quality. Indeed, our meta-analysis confirmed the significant effect of the halothane gene on all the analyzed meat quality attributes. However, the halothane gene effect was associated with the breed effect only for the following variables: L*-value (reflectance), b*-value (yellowness) and DL (drip loss). Slaughter weight contributed significantly only to the explanation of differences in ultimate pH between homozygous genotypes: NN and nn. Regarding the preslaughter conditions, the results of our meta-analysis showed that fasting time had a significant effect on pHu and DL measured in longissimus muscle. While, lairage affected only the pHu in semimembranosusmuscle. Interestingly, we found that DL was the lonely attribute significantly affected by transport time and its interaction with fasting time. The significant effect of the vitamin E supplementation on lipid oxidation process in pork and, therefore, affect the technological meat quality, was demonstrated by using a non linear meta-analytic regression model. Statistically, the comparison between the Bayesian and frequentist meta-analytic approaches has highlighted the ability of the hierarchical Bayesian model to reduce uncertainty in parameter estimates, especially when the number of studies included in the meta-analysis is relatively limited. In addition to the meta-analytic approach, an exploratory study by multivariate analyses was performed to identify the correlations among a large number of variables structured into separate groups. Our results showed that the variables which discriminated the five classes of breed (Large White and Basque pigs) x rearing system (conventional, alternative and extensive) were the same as those discriminating individuals. Indeed, a first group composed of transcriptomic, chemical and technological variables discriminated breeds. Slaughter reactivity variables discriminated rearing systems, especially the Basque pigs reared in the extensive system.

Un des enjeux de la filière bovine est la maîtrise et la prédiction de la tendreté de la viande, critère important pour les consommateurs. Inscrite dans le programme européen ProSafeBeef, ma thèse avait pour objectif de mieux prédire la tendreté par méta-analyse à partir des caractéristiques biochimiques du muscle. Pour cela, mon travail de thèse s’est appuyé sur la base de données BIF-Beef regroupant des données individuelles issues de plusieurs programmes de recherche, allant de l’animal à la viande en passant par la carcasse et le muscle. Une première méta-analyse a montré que le muscle Semitendinosus (ST) est de type plus rapide glycolytique que le Longissimus thoracis (LT) chez les mâles entiers et les femelles mais pas chez les mâles castrés. Après avoir identifié par une approche par classe de tendreté les caractéristiques musculaires associées à la tendreté, nous avons montré que ces caractéristiques sont différentes entre muscles et types d’animaux. Dans le muscle LT des taurillons, la surface moyenne des fibres musculaires est la variable qui joue le principal rôle sur la tendreté sensorielle où elle explique 2% de la variabilité des notes de tendreté. Principalement dans le muscle ST, les teneurs en collagène total et insoluble et l’activité enzymatique du métabolisme glycolytique expliquent au plus 6% chacun de la variabilité de la force de cisaillement. Malgré ces faibles parts de variabilité expliquée de la tendreté, ces conclusions validées sur un grand volume de données sont importantes pour préparer de nouveaux projets visant à compléter cette démarche en prenant en compte d’autres caractéristiques telles que des biomarqueurs génomiques. / The control of beef tenderness is essential for beef producers and retailers to deliver a consistently high quality product to consumers. Being part of the European program ProSafeBeef, my thesis aimed to predict beef tenderness by meta-analysis approaches using biochemical characteristics of muscles. To achieve this goal, we used data available in the BIF-Beef data warehouse which contains animal, carcass, muscle and meat measurements from different research programs. From available data on Longissimus thoracis (LT) and Semitendinosus (ST) muscles, we demonstrated that ST was faster and more glycolytic than LT in both entire males and females but not in steers. With a cluster analysis, we identified muscle biochemical traits associated with meat tenderness. Then, we demonstrated that no specific muscle biochemical characteristic can be a predictor of tenderness for all muscles and animal types. In LT muscle of young bulls, mean muscle fibre area explained 2% of the variability in sensory tenderness score. Mainly in ST muscle, total and insoluble collagen content and enzymatic indicators of glycolytic metabolism each explained about 6% of the variation in shear force. Although we were only able to explain a relatively small proportion of the total variance in tenderness, these results will form an important basis for the design of future experiments and the identification of new genomic markers of tenderness to be combined with muscle biochemistry in order to better predict beef quality.