Recent advances in global gene expression measurement and the development of large- scale public repositories for storage of such data have made a wealth of information available to researchers. While one gene expression study may lack sufficient replicates to make statistically significant pronouncements, the combination of studies through meta-analysis can yield results with a much greater likelihood of accuracy. In order to combine multiple sets of data, one must first address the issue of cross-comparison between global gene expression platforms, as well as resolve the issue of repeated measures (multiple probes representing the same gene) within each platform. In this work, I present computational methods for probe reannotation and scoring and for redundant probe consolidation that together allow for greatly improved access to data for meta-analysis. I also present an example of the application of these methods, in the analysis of the gene expression regulated by estrogen across multiple cell types. Estrogen, a steroid hormone, interacts with its receptors to regulate gene transcription in both direct and indirect manners. Estrogen has the effect of increasing proliferation in some tissues, while inhibiting proliferation or increasing apoptosis in others. How estrogen achieves these highly divergent results remains unclear. Through meta-analysis of gene expression experiments across multiple cell types, I show that patterns of estrogen regulation in many tissues involve the same key genes and pathways, including cell cycle, p53 signaling, and TGFβ signaling pathways. However, regulation in different cell types can result from regulation of different genes, or the same genes regulated in different directions. Many patterns of gene regulation support known physiological consequences of estrogen on these tissues. In particular, genes promoting proliferation are upregulated in uterus and certain breast and ovarian cancer cell lines. One gene, thrombospondin-1, is up-regulated in eleven out of nineteen cell types and may be a key player in regulating proliferation in re- sponse to estrogen. Results in other cell types are unexpected. Most notably, neither genes promoting nor inhibiting proliferation are differentially regulated upon estrogen treatment in vascular smooth muscle cells, despite estrogen inhibiting proliferation of these cells.

Le nalméfène représente une option nouvelle dans le traitement de l’alcoolodépendance. Cette molécule a reçu une AMM européenne en 2013 pour réduire la consommation d’alcool des patients adultes ayant une consommation à risque élevé selon la classification de l’OMS (i.e. plus de 60 g d'alcool par jour pour les hommes, ou plus de 40 g par jour pour les femmes). Toutefois, les avis dans la littérature divergent quant à son intérêt. Nous avons donc souhaité réévaluer le rapport bénéfice-risque de cette molécule en appliquant des méthodes appartenant au domaine de la méta-recherche. Une première méta-analyse des essais randomisés contrôlés comparant le nalméfène par voie orale à un placebo nous a permis de mettre en évidence des tailles d’effet faibles sur les critères de consommation, avec un possible biais d’attrition dans les études. Dans une seconde méta-analyse en réseau évaluant l’efficacité, dans l’indication de réduction des consommations, du nalméfène à d’autres traitements également utilisés dans la prise en charge de l’alcoolodépendance, nous avons conclu à l’absence de preuve de haut niveau de l’efficacité des traitements pharmacologiques pour contrôler la consommation d'alcool chez les patients non abstinents souffrant de troubles liés à l’usage. Enfin, nous avons montré, à travers la réalisation d’un grand nombre de méta analyses, que les choix méthodologiques faits dans les essais randomisés contrôlés évaluant le nalméfène et la naltrexone dans la prise en charge de l’addiction à l’alcool entraînaient une variation substantielle de l’estimation de l’efficacité de ces deux traitements, phénomène également appelé « vibration de l’effet » des traitements. / Nalmefene is a new option in the treatment of alcohol dependence. It was approved by the European Medicines Agency to help reduce alcohol consumption in adults who consume more than 60 g of alcohol per day (for men) or more than 40 g per day (for women). However, researchers’ opinions on the interest of the drug are divided. We therefore planned direct and network meta-analyses to enable an objective reappraisal of the efficacy of nalmefene. A first meta-analysis of randomized controlled trials comparing oral nalmefene with placebo showed small effect sizes on consumption criteria with a possible attrition bias in the studies. In a second network meta-analysis evaluating the comparative efficacy of nalmefene to control drinking with other treatments also used in the management of alcohol dependence, we concluded that there was no high-grade evidence for pharmacological treatment to control drinking in non-abstinent patients with alcohol use disorders. Finally, we explored the vibration of effects in a large number of overlapping indirect meta-analyses comparing nalmefene versus naltrexone to reduce alcohol consumption. We demonstrated that the methodological choices made in randomized controlled trials resulted in a substantial variation in the effect sizes of these two treatments.

No description available.

Researchers often invoke strong assumptions in empirical analyses to identify significant statistical outcomes. Invoking assumptions that do not sufficiently reflect the occurrence of true phenomenon reduces the credibility of inferences. Literature suggests that the potential effects of assumptions on credibility of inferences can be mitigated by comparing and combining insights from alternative econometric models. I use this recommendation to conduct robustness checks of commonly used methods in environmental and health economics. The first chapter proposes a novel nonparametric regression model to draw credible insights from meta-analyses. Existing literature on benefit-transfer validity is examined as an application. Nonparametric regression is found to be a viable approach for drawing robust policy insights. The second chapter proposes an alternative structural and simulations based framework to understand elicitation effects in survey response data. This analysis explains the structural mechanisms in which response anomalies occur and is important for building credible insights from survey data. The last chapter uses methods in program evaluation to investigate the impacts of institutional child deliveries on long-term maternal health in the context of developing countries. The outcomes of this analysis indicate that institutional deliveries positively affect maternal health in lower socio-economic states. Based on the findings of my three chapters, I recommend that researchers should combine insights from alternative models to mitigate the scope of specification bias in empirical outcomes and inform policy about the potential uncertainty that arises in uncovering the truth using statistical methods. / Ph. D.

faustino.perez@unizar.es / Objective: To assess the effects of vitamin D supplementation during pregnancy on obstetric outcomes and birth variables. Design: Systematic review and meta-analysis of randomized controlled trials (RCTs). Setting: Not applicable. Patient(s): Pregnant women and neonates. Intervention(s): PubMed and 5 other research databases were searched through March 2014 for RCTs evaluating vitamin D supplementation calcium/vitamins/ferrous sulfate vs. a control (placebo or active) during pregnancy. Main Outcome Measure(s): Measures were: circulating 25-hydroxyvitamin D [25(OH)D] levels, preeclampsia, gestational diabetes mellitus (GDM), small for gestational age (SGA), low birth weight, preterm birth, birth weight, birth length, cesarean section. Mantel-Haenszel fixed-effects models were used, owing to expected scarcity of outcomes. Effects were reported as relative risks and their 95% confidence intervals (CIs). Result(s): Thirteen RCTs (n ¼ 2,299) were selected. Circulating 25(OH)D levels were significantly higher at term, compared with the control group (mean difference: 66.5 nmol/L, 95% CI 66.2–66.7). Birth weight and birth length were significantly greater for neonates in the vitamin D group; mean difference: 107.6 g (95% CI 59.9–155.3 g) and 0.3 cm (95% CI 0.10–0.41 cm), respectively. Incidence of preeclampsia, GDM, SGA, low birth weight, preterm birth, and cesarean section were not influenced by vitamin D supplementation. Across RCTs, the doses and types of vitamin D supplements, gestational age at first administration, and outcomes were heterogeneous. Conclusion(s): Vitamin D supplementation during pregnancy was associated with increased circulating 25(OH)D levels, birth weight, and birth length, and was not associated with other maternal and neonatal outcomes. Larger, better-designed RCTs evaluating clinically relevant outcomes are necessary to reach a definitive conclusion. (Fertil Steril 2015;-:-–-. 2015 by American Society for Reproductive Medicine.) / Revisión por pares

Purpose: Adaptive servoventilation (ASV) has been reported to show improvement in patients with sleep-disordered breathing (SDB) and heart failure (HF); however, its role as a second-line or adjunctive treatment is not clear. We conducted a systematic review and meta-analysis of new existing data including cardiac mechanistic factor, geometry, and cardiac biomarkers. Methods: We systematically searched for randomized controlled trials (RCTs) and cohort studies that assessed the efficacy or effectiveness of ASV compared to conventional treatments for SDB and HF in five research databases from their inception to November 2018. Random-effects meta-analyses using the inverse variance method and stratified by study design were performed. Results: We included 15 RCTs (n = 859) and 5 cohorts (n = 162) that met our inclusion criteria. ASV significantly improved left ventricular ejection fraction (LVEF) in cohorts (MD 6.96%, 95% CI 2.58, 11.34, p = 0.002), but not in RCTs. Also, the ASV group had significantly lower apnea-hypopnea index (AHI) in both cohorts (MD − 26.02, 95% CI − 36.94, − 15.10, p < 0.00001) and RCTs (MD − 21.83, 95% CI − 28.17, − 15.49, p < 0.00001). ASV did not significantly decrease the E/e′ ratio in RCTs or in cohorts. Finally, ASV significantly decreased brain natriuretic peptide (BNP) in the cohorts (SMD − 121.99, CI 95% − 186.47, − 57.51, p = 0.0002) but not in RCTs. ASV did not have a significant effect on systolic blood pressure, diastolic blood pressure, and cardiac diameters. Conclusions: ASV therapy is associated with improvements of AHI in comparison to alternative treatments in patients with SDB and HF. ASV did not improve LVEF or E/e′ ratios in randomized trials; other intermediate outcomes did not improve significantly. / Revisión por pares

Background: Associations between dietary fats and mortality are unclear. Methods: We evaluated the relationship between quartiles of total fat, mono-unsaturated (MUFA), polyunsaturated (PUFA) and saturated fatty acid (SFA) consumption, and all-cause, coronary heart disease (CHD), stroke, and type 2 diabetes (T2D)-associated mortality in 24,144 participants from the National Health and Nutrition Examination Surveys (NHANES) 1999–2010. We added our results to a meta-analysis based on searches until November 2018. Results: In fully adjusted Cox-proportional hazard models in our prospective study, there was an inverse association between total fat (HR: 0.90, 95% confidence interval 0.82, 0.99, Q4 vs Q1) and PUFA (0.81, 0.78–0.84) consumption and all-cause mortality, whereas SFA were associated with the increased mortality (1.08, 1.04–1.11). In the meta-analysis of 29 prospective cohorts (n = 1,148,117) we found a significant inverse association between total fat (0.89, 0.82–0.97), MUFA (0.93, 0.87–0.99) and PUFA (0.86, 0.80–0.93) consumption and all-cause mortality. No association was observed between total fat and CVD (0.92, 0.79–1.08) or CHD mortality (1.03 0.99–1.09). A significant association between SFA intake and CHD mortality (1.10, 1.01–1.20) was observed. Neither MUFA nor PUFA were associated with CVD or CHD mortality. Inverse associations were observed between MUFA (0.80, 0.67–0.96) and PUFA (0.84, 0.80–0.90) intakes and stroke mortality. Conclusions: We showed differential associations of total fat, MUFA and PUFA with all-cause mortality, but not CVD or CHD mortalities. SFA was associated with higher all-cause mortality in NHANES and with CHD mortality in our meta-analysis. The type of fat intake appears to be associated with important health outcomes. / European Society of Cardiology / Revisión por pares

Chronic pain is common and frequently interferes with people’s regular functioning and reduces quality of life. Though pharmacological approaches are used most frequently to treat pain-related issues, the side effects of these medications often lead to other problems. Group therapy has been used and studied for decades in treating pain though it’s general efficacy for addressing pain is not clear. Objectives: to determine group therapy’s efficacy for patients with pain-related issues and whether the effects are moderated by study, patient, leader, or group characteristics. Method: potential articles were selected from searches completed in major databases based on a set of inclusion criteria. A random effects meta-analysis was conducted, and potential moderators were analyzed. Results: we analyzed 57 studies representing 8,933 patients receiving group therapy for pain which produced a significant, small effect (g = 0.28) for reducing pain intensity. Various secondary outcomes such as pain frequency, interference with activities of daily living, physical functioning, catastrophizing, self-efficacy, anxiety, depression, and quality of life were also found to improve significantly. Four significant moderating variables were found to include pain measure used, gender composition, number of sessions, and presence of pain diagnosis. Discussion: results are discussed and compared to those of past meta-analyses regarding both chronic pain and group therapy. Implications for practice and research are provided.

In this article, a meta-analysis was conducted using correspondence tests from 2004 and later, researching gender discrimination in the labour markets hiring process. A total of 19 correspondence tests were found from 12 different countries across the world. Comparisons of call-back rates for job applications between male and female applicants were made. I created a call-back ratio, dividing the female applicants’ call-back rate by the male applicants’ call-back rate for each study and regressed on two different variables, “equality score” and “work incentives rating”, meant to measure gender equality of countries. My hypothesis is that females are still discriminated against in the world’s labour markets. Statistical discrimination is my main theory of why discrimination of female applicants would occur in the labour market. The method of meta-analysis is useful for observing overall trends and drawing broad conclusions regarding a subject. The equality score is significantly correlated with higher call-back rates for both genders, bot it is not correlated with the call-back ratio, and therefore, does not seem to be correlated with gender discrimination. Providing more work incentives are correlated with lower call-back rates for both genders, but these findings are not significant at any level. The call-back ratio is unaffected by the work incentives rating, showing no correlation with discrimination. The main findings show no discrimination of female applicants, but rather points to discrimination of male applicants. This could be to a perceived image of men being more threatening.

The association between economic freedom and economic growth has been largely explored by researchers and the overall ndings indicate a signi cant and positive relationship. The empirical literature, however, is subject to suer from bias. In this paper we collect 16,070 estimates from 69 studies and using recently developed meta-analytic techniques investigate the eect of publication and speci cation biases on the reported results. While our baseline analysis re- ports some evidence for publication bias, but not very strong and robust, and con rms the speci cation bias reported by previous reviews, we also nd that these results are aected by the inclusion of three in uential outliers in the data set. Once we trim these studies, there is no evidence of speci cation bias anymore and we nd evidence of a robust and strong publication bias. Further, after controlling for the bias, we nd that the true eect of economic freedom on growth is substantially smaller than the eect reported by the empirical literature. JEL Classi cation O43; P10; P12; C52 Keywords 'economic freedom', 'economic growth', 'publi- cation bias', 'speci cation bias', 'meta-analysis' Title Economic freedom and economic growth: A Meta-Analysis