L'émergence du phénomène Big Data a créé un besoin grandissant en analyse de données, mais, bien souvent, cette analyse est conduite par des experts de différents domaines ayant peu d'expérience en science des données. On s'intéresse donc à ce besoin d'assistance à l'analyse de données, qui commence tout juste à recevoir une certaine attention des communautés scientifiques, donnant naissance au domaine de la méta-analyse. Les premières approches du sujet se révélant souvent similaires et peu abouties, on tente en particulier de permettre de nouvelles approches de méta-analyse pour adresser ce problème d'assistance à l'analyse de données. Pour ce faire, une première étape cruciale est de déterminer ce qu'est une méta-analyse performante, aucun standard n'ayant encore été établi dans ce domaine relativement neuf. On propose ainsi un cadre générique d'évaluation de méta-analyse, permettant de comparer et caractériser finement diverses techniques de méta- analyse. Ensuite, afin d'ouvrir de nouvelles voies, on s'intéresse à un verrou majeur de la méta-analyse : la caractérisation de jeu de données. On propose et évalue alors une caractérisation par dissimilarité faisant usage de toute l'information disponible pour autoriser de nouvelles approches de méta-analyse. L'utilisation de cette caractérisation par dissimilarité permettant de recommander facilement des processus d'analyse de données complets, on décrit enfin les nouvelles approches de méta-analyses rendues possibles, ainsi que les processus afférents d'assistance à l'analyse de données. / The emergence of the big data phenomenon has led to increasing demands in data analysis, which most often are conducted by other domains experts with little experience in data science. We then consider this important demand in intelligent assistance to data analysis, which receives an increasing attention from the scientific community. The first takes on the subject often possessing similar shortcomings, we propose to address it through new processes of meta-analysis. No evaluation standard having yet been set in this relatively new domain, we first propose a meta-analysis evaluation framework that will allow us to test and compare the developed methods. In order to open new approaches of meta-analysis, we then consider one of its recurring issue: dataset characterization. We then propose and evaluate such a characterization, consisting in a dissimilarity between datasets making use of a precise topological description to compare them. This dissimilarity allows a new meta-analysis approach producing recommendations of complete data analysis processes, which we then evaluate on a proof of concept. We thus detail the proposed methods of meta-analysis, and the associated process of assistance to data analysis.

This thesis provides a case study on how the wind direction plays an important role in the amount of rainfall, in the village of Somi$acute{o}$. The primary goal is to illustrate how a meta-analysis, together with circular data analytic methods, helps in analyzing certain environmental issues. The existing GLS meta-analysis combines the merits of usual meta-analysis that yields a better precision and also accounts for covariance among coefficients. But, it is quite limited since information about the covariance among coefficients is not utilized. Hence, in my proposed meta-analysis, I take the correlations between adjacent studies into account when employing the GLS meta-analysis. Besides, I also fit a time series linear-circular regression as a comparable model. By comparing the confidence intervals of parameter estimates, covariance matrix, AIC, BIC and p-values, I discuss an improvement on the GLS meta analysis model in its application to forecasting problem in Environmental study.

A key element in evidence-based medicine approaches is the ability for clinicians to evaluate the scientific rigor and relevance of research evidence. In the treatment of diabetes, clinicians make increasingly difficult decisions about which drug regimens are best for their patients with limited evidence-based information. While the consensus is that metformin should be the initial drug treatment when diet and exercise are not sufficient, clinicians disagree on whether sulfonylureas should remain a suitable therapy after metformin. While this would be improved with further research investigating the comparative safety of therapeutic options, there is also need for better ways to synthesize available information to guide evidence-based decision-making in health services research. Study 1 summarizes the pre-existing evidence on the long-term safety risks associated with sulfonylurea therapy relative to other drug classes. Results from a series of meta-analyses provide some evidence that sulfonylureas are associated with elevated all-cause mortality and cardiovascular risks relative to several other medications, either as a monotherapy or in combination with metformin. Study 2 analyzes the comparative safety of second-line treatment in diabetic patients in the Veterans Health Administration to address gaps in the literature. Results suggest that second-line use of sulfonylureas is associated with increased risks compared to thiazolidinediones. Results also suggest that changes to existing metformin therapy may lead to differential hazards. Clinicians may disagree about the quality of the evidence as well as the relevancy to their own treatment population. Improvements in methods for evidence-based medicine that take this into account are needed. Study 3 applies an underutilized research method that allows for a more thoughtful synthesis of all available evidence. This framework allows clinicians to incorporate the scientific rigor and relevancy of previous study results when integrating new data into their current knowledge base. Results suggest an elevated risk in all models for sulfonylureas compared to thiazolidinediones and highlight the need to design more focused research to support clinical decision-making around medication safety. This novel application to evidence synthesis shows promise as applied to a health services research problem and has potential as a useful framework in other health services research areas. / 2017-12-09T00:00:00Z

Inúmeros trabalhos avaliando as exigências de lisina podem ser encontrados na literatura, mas seus resultados são heterogêneos. Revisões sistemáticas e a meta-análise permitem integrar diferentes variáveis, gerar novos resultados e obter respostas ajustadas a uma maior diversidade experimental. Objetivou-se explorar as relações da lisina digestível com as variáveis de desempenho, estabelecer a exigência e avaliar a eficiência marginal da lisina segundo a lei de rendimentos decrescentes para suínos em crescimento e terminação. Para os estudos I e II, foram compilados 36 artigos publicados entre 2002 e 2013 avaliando 2.399 suínos pesando entre 15-120 kg. Para o estudo III, foram utilizados 26 artigos que avaliaram 1.820 suínos pesando entre 20-120 kg. No estudo I, o consumo diário de lisina apresentou correlações maiores com a ingestão de energia metabolizável e proteína bruta e com as deposições de proteína e lipídios. O consumo diário de lisina apresentou melhor ajuste nas equações de variância-covariância do que o nível percentual de lisina. No estudo II, o aumento no consumo diário de lisina melhorou o ganho de peso e a deposição de proteína na fase de 70-120 kg. Nas fases de 15-30; 30-70 e 70-120 kg os níveis de lisina digestível determinados para obter as melhores respostas de ganho corresponderam ao consumo diário de 14,8; 20,2 e 18,3 g lisina/dia. Os níveis de lisina digestível determinados para obter as melhores respostas de deposição de proteína corresponderam ao consumo diário de 16,3 e 24,1 g lisina/dia nas fases 15-30 e 30-70 kg. No estudo III, observamos que a medida que o animal se aproxima ou atinge o ponto de máximo desempenho, a eficiência de uso da lisina digestível diminui segundo a lei de rendimentos decrescentes. Dietas formuladas para atender 95% da máxima resposta de ganho de peso permitem uma melhora na eficiência marginal de uso da lisina da ordem de 2,4; 2,5 e 1,5 g de ganho de peso para cada g de lisina ingerida nas faixas de peso vivo 20-50; 50-70 e 70-120 kg. Sugerimos a inclusão de variáveis como a eficiência alimentar e o ganho de peso nas análises de determinação de exigências de lisina, também a escolha de metodologias analíticas adequadas e a inclusão da lei de rendimentos decrescentes da eficiência marginal de uso da lisina nos atuais modelos de determinação de exigência para evitar a superestimação observada nas tabelas de recomendação nutricional. / Several studies evaluating the lysine requirements are avaible in the literature, but its results are heterogeneous. Systematic reviews and meta-analysis allows integrating different variables, generating new results and producing answers adjusted to experimental diversity. This objective of this study was to explore the relationship among dietary lysine and the performance variables, establishing the requirement and assessing the marginal efficiency of dietary lysine according to the law of diminishing returns for growing-finishing pigs. For studies I and II, 36 articles published from 2002 to 2013 evaluating 2399 pigs weighing between 15-120 kg were compiled. And for the study III 26 articles evaluating 1820 pigs weighing between 20-120 kg were used. In the study I, daily intake of lysine showed higher correlations with the metabolizable energy intake and crude protein, with the protein deposition and lipid deposition. The daily intake of lysine showed better adjustment in the variance-covariance equations than the dietary level of lysine. In Study II, increasing the daily intake of lysine improved the weight gain and protein deposition in the 70-120 kg phase. In 15-30; 30-70 and 70-120 kg phase the levels digestible lysine that maximized weight gain reponse corresponded to the daily intake of 14.8; 20.2 and 18.3 g lysine/day. Lysine levels that maximized protein deposition reponse corresponded to daily intake of 16.3 and 24.1 g lysine/day in the phases 15-30 and 30-70 kg. In the study III, it was observed that, as the animal approaches or reaches the maximum performance the efficiency of lysine use decreases according to the law of diminishing returns. Diets formulated to meet 95% of the maximal weight gain response permit a marginal improvement in efficiency of use of the lysine the order of 2.4; 2.5 and 1.5 grams of weight gain for each lysine grams ingested in live weight phases 20-50; 50-70 and 70-120 kg respectively. It suggest the inclusion of variables feed efficiency and weight gain in analysis for the determination of lysine requirements, also the choice of appropriate analytical methods and the inclusion of diminishing returns law of marginal efficiency the dietary lysine in current models for the determination of requirements to avoid the overestimation observed in nutritional recommendation tables.

OBJETIVO: Comparar a eficácia e a tolerabilidade dos fármacos antiepiléticos (FAE) no tratamento em monoterapia de pacientes com epilepsia focal ou generalizada. MÉTODOS: Uma revisão sistemática foi realizada por meio da busca nas bases de dados eletrônicas Pubmed, Scopus, Web of Science e Cochrane Register of Controlled Trials. Foram incluídos os ensaios clínicos controlados com pacientes com epilepsia, em tratamento com FAE, via oral, em monoterapia, e que avaliaram o número de pacientes que atingiram a remissão das crises epilépticas, que interromperam o tratamento devido à ineficácia terapêutica ou à ocorrência de reações adversas (RAM) intoleráveis. Meta-análises de comparação de múltiplos tratamentos foram realizadas por meio do modelo bayesiano de efeitos randômicos que permitiu o cálculo do Odds Ratio meta-analítico para os FAE estudados. Também foi realizado um ranqueamento da probabilidade de cada FAE ser a melhor opção em eficácia e tolerabilidade. RESULTADOS E CONCLUSÕES: A busca identificou 18874 publicações, no entanto apenas 71 estudos foram selecionados, compreendendo 17555 pacientes com epilepsia. Vinte e nove estudos apresentaram os desfechos de eficácia no tratamento de crises focais, 19 em crises generalizadas e 58 apresentaram dados de tolerabilidade. Nesses estudos, 15 FAE foram avaliados. No tratamento das crises focais, os FAE de nova geração levetiracetam (LEV), lamotrigina (LTG), oxcarbazepina, sultiame e topiramato (TPM) demonstraram possuir eficácia equivalente à carbamazepina (CBZ), clobazam e valproato (VPA). No entanto, a CBZ apresentou o pior perfil de tolerabilidade devido à grande probabilidade do paciente abandonar o tratamento devido à RAM intoleráveis. Quanto ao tratamento das crises generalizadas, a LTG, LEV e TPM são tão eficazes quanto o VPA para o tratamento de crises generalizadas tônico-clônicas, tônicas e clônicas. O VPA e a etosuximida constituem as melhores opções para o tratamento de crises de ausências, enquanto que a LTG mostrou-se menos eficaz. Para o tratamento de crises mioclônicas e espasmos infantis mais ensaios clínicos randomizados são necessários para fornecer boas evidências que possam guiar a decisão clínica dos profissionais de saúde. Dentre os FAE com perfil de eficácia adequado, a LTG destacou-se pela menor probabilidade de manifestar RAM intoleráveis. / OBJECTIVE: To compare the efficacy and tolerability of the antiepileptic drugs (AED) in monotherapy of patients with focal or generalized epilepsy. METHODS: A systematic review was in the Medline/Pubmed, Scopus, Web of Science and Cochrane Register of Controlled Trials databases. We included randomized clinical trials of patients with epilepsy treated with oral monotherapy AED, which evaluated number of patients becoming seizure free at the maintenance treatment period; number of patients which withdrawals from the study because of therapeutic inefficacy and number of patients which withdrawals from the study because of intolerable adverse reaction. Network meta-analyses were performed using Bayesian random effects model. We also carried out a ranking of the probability of each AED be the best option in the efficacy and tolerability outcomes. Sensitivity analyses were conducted in order to check the robustness of the results. RESULTS AND CONCLUSIONS: The research identified 18,874 publications, but only 71 studies were selected, comprising 17555 patients with epilepsy.Twenty-nine trials showed the efficacy outcomes in the treatment of focal seizures, 19 in generalized seizures and 58 showed tolerability data. In the treatment of focal seizures, levetiracetam (LEV), lamotrigine (LTG), oxcarbazepine, sultiame and topiramate (TPM) have demonstrated equivalent efficacy to carbamazepine (CBZ), clobazam and valproate (VPA). LTG, LEV and TPM are as effective as the VPA for the treatment of generalized tonic-clonic, tonic and clonic seizures. VPA and ethosuximide are the best options for the treatment of absence seizures, whereas LTG was less effective. For the treatment of myoclonic seizures and infantile spasms, more randomized clinical trials are needed to provide good evidence to guide the clinical decision of health professionals. Among the AED with adequate efficacy profile, LTG stands out as the AED with the best tolerability profile, suggesting it may be the best option for the treatment of patients with epilepsy.

Orientador: Antonio José Maria Cataneo / Resumo: Introdução: A hiperidrose primária (HP) é um distúrbio caracterizado pela sudorese excessiva, levando a uma piora na qualidade de vida de seus portadores. Objetivo: Determinar a efetividade da radiofrequência (RF) no tratamento da HP. Método: Revisão sistemática de ensaios clínicos randomizados, estudos controlados e estudos observacionais em indivíduos portadores de HP, nos quais foram realizados o tratamento por RF comparado com qualquer outra modalidade de tratamento ou em que foram comparados os resultados antes e após a intervenção no tratamento isolado por RF. Os desfechos avaliados foram: grau de intensidade da hiperidrose, qualidade de vida, sudorese compensatória e recorrência dos sintomas. Resultados: Foram selecionados dez estudos, oito deles (204 participantes) empregando somente a RF (74 microagulhamentos e 130 ablações) e dois (65 participantes) comparando a RF com a simpatectomia videotoracoscópica. Houve redução da intensidade da HP no subgrupo que aplicou microagulhamento (3 estudos, 74 participantes, diferença média -1,24, IC 95% -1,44 a -1,03, I2=50%). Em um estudo em que se realizou ablação simpática por RF (36 participantes) houve maior redução da intensidade da HP do que nos três estudos em que se realizou microagulhamento (-2,56, IC 95% -2,76 a -2,36 versus -1,24, IC 95% -1,44 a -1,03). Houve melhora na qualidade de vida após ablação simpática (2 estudos, diferença média -15,92, IC 95% -17,61 a -14,24, I²= 23%). No microagulhamento também houve melhora ... (Resumo completo, clicar acesso eletrônico abaixo) / Doutor

Les antidépresseurs sont les traitements pharmacologiques de choix de la dépression. Pourtant devant un nombre non négligeable d’études « négatives » et l’absence d’efficacité dans les dépressions peu sévères, la question de leur utilité reste controversée. Cette controverse présuppose que l’efficacité des antidépresseurs se mesure facilement. Pourtant, comme dans toute expérience, le dispositif expérimental peut influencer le résultat mis en évidence. L’objectif de cette thèse est justement d’explorer l’incidence de la procédure expérimentale sur l'effet mesuré des pharmacothérapies antidépressives. Est-ce que leur efficacité mesurée dans les études randomisées reflète leur effectivité (dans la « vraie vie ») ? Pour répondre à cette question, nous avons conduit une méta-analyse d’un large panel d’études de méthodologies très différentes. Une méta-régression a été réalisée pour modéliser la réponse au traitement en fonction des caractéristiques des études. Elle a permis de mettre en évidence 1/ une réponse supérieure dans les essais randomisés par rapport aux études observationnelles (possiblement plus représentatives de la « vraie vie »); 2/ une variation de la réponse en fonction de la durée de l’étude, du nombre de visites de suivi, de la population de patients inclus, du type d’analyse choisie, de la possibilité d’avoir un placebo et du design en aveugle. Une revue de la littérature apporte ensuite une perspective qualitative à la question posée. L’objectif était de décrire les dispositifs expérimentaux actuellement utilisés ou envisageables dans l’évaluation des antidépresseurs. La revue de la littérature 1/ a permis de préciser les limites des études actuelles en termes de validité interne (mesure des critères de jugement, attrition, levée de l’aveugle importance de la réponse placebo) et externe (difficultés majeures de généralisations des études); 2/ conduit à proposer une évaluation multimodale des antidépresseurs dans laquelle les études observationnelles auraient toute leur place. La question de la mesure est ensuite abordée à travers des simulations. Après avoir exploré l’impact du dispositif expérimental au niveau de l’essai en soi, le phénomène est analysé à un niveau plus global. Le choix a été fait d’un raisonnement par l’absurde, sur le terrain de la « comparative effectiveness ». L’objectif était de comparer les placebos d’antidépresseurs grâce à une méta-analyse en réseau. Ces comparaisons indirectes n’ont pas permis de mettre en évidence de différence entre placebos d’antidépresseurs. Un biais de publication a pu être mis en évidence et empêche de conclure formellement sur un résultat pourtant évident. Ce résultat ouvre une discussion sur le degré de confiance que l’on peut apporter à ce que l’on considère généralement comme des preuves scientifiques de l’effectivité de nos pratiques. Nos résultats impliquent un nouvel éclairage à la controverse sur les antidépresseurs en replaçant le débat sur le terrain épistémologique. / Antidepressant drugs have become the most prescribed and used pharmacological treatment for depressive disorders. Yet, their usefulness is still being debated in view of a significant number of “negative” studies and of their lack of efficacy in mild depressive disorders. This controversy presupposes that the efficacy of antidepressants can easily be measured. However, as in any experiment, the experimental method may influence the outcome. The aim of this thesis is precisely to explore the impact of the experimental procedures on the measured effect of antidepressant therapies. Also we ask the question of whether their efficacy measured in randomized trials reflects their “effectiveness” (in a “real life” setting)? To answer these questions, we performed a meta-analysis of a large number of studies with different methodologies. A meta-regression was performed to modelize the response to the treatment according to the studies’ characteristics. It highlighted 1/ a superior response in randomized trials compared to observational studies (possibly more representative of the “real life”); 2/ a variation of the response according to the duration of the study, to the number of follow-up visits, to the type of included patients, to the type of analysis, to the use or not of a placebo and to the use or not of a blind design. A review of the literature provides a qualitative prospect to the question. The objective was to describe the experimental designs used in the assessment of antidepressants and to identify possible innovations. The review of the literature 1/ has permitted to state precisely the limits of the studies in terms of internal validity (measurement of assessment criterions, attrition, unblinding, importance of the placebo response) and external validity (major difficulties with generalization of the studies’ results); 2/ leads to offer a multimodal evaluation of antidepressants in which observational studies would have their own place. Then the issue of measurement is specifically approached, using simulations. After exploring the impact of the experimental procedure at the level of the trial itself, the phenomenon is explored at a more global level. The choice of a reductio ad absurdum on the field of comparative effectiveness was decided. The aim was to compare the placebos of different antidepressants in a network meta-analysis. Indirect comparisons did not show any difference between placebos of antidepressants but a significant publication bias was detected. It does not allow drawing any firm conclusion on a result which is still obvious. This result invites discussion as to whether the evidence that we usually rely on provides us with a reasonable degree of certainty about the nature and effectiveness of our practices. Our results entail a new light on the controversy about the efficacy of antidepressants by replacing the debate on the epistemological ground.

This diploma thesis presents a meta-analysis of the accumulated empirical evidence on the relationship between financial development and economic growth. So far, hundreds of studies have been written on the role of financial systems in economic growth; however, their results are ambiguous. This is supported both by theory and empirical research. In order to shed some light on the underlying relationship, narrative literature surveys have been conducted. Nevertheless, the authors of these surveys select representative studies for inclusion subjectively and thus build their results on only a limited set of information. Moreover, due to the nature of their analyses, they cannot systematically assess which factors influence the heterogeneity in reported findings or whether the results are driven by the desire to produce only positive and statistically significant results. Thus, the main focus of our work lies in investigating what the role of financial development in economic growth is, adjusted for possible publication selection, and to systematically explain the heterogeneity behind reported results. For this analysis a pool of available studies investigating the underlying relationship was collected. More specifically, our analysis takes into account data from 67 empirical studies with 1334...

Introdução: A constipação funcional corresponde a aproximadamente 90 a 95% dos casos de constipação crônica infantil, sendo uma doença frequente e com grande impacto na qualidade de vida do paciente pediátrico e de sua família. O aumento do consumo de fibras na dieta está associado com estímulo da peristalse gastrointestinal, acelerando o trânsito colônico. Entretanto, ainda não existem evidências claras que corroborem o uso rotineiro da suplementação de fibras na dieta das crianças como parte do tratamento da constipação funcional. Objetivo: Reunir evidências atuais sobre o uso de fibras no tratamento da constipação em pacientes pediátricos. Métodos: Revisão sistemática com metanálise de estudos controlados randomizados identificados por meio de pesquisa nas bases de dados Pubmed, Embase, LILACS e Cochrane. Critérios de Inclusão: Estudos controlados randomizados; pacientes com idade entre 2 a 18 anos e com diagnóstico de constipação não orgânica em uso ou não de tratamento medicamentoso para constipação; artigos publicados em língua portuguesa, inglesa, espanhola, francesa e alemã; artigos documentados e/ou publicados em revistas acessíveis aos pesquisadores. Resultados: Foram encontrados 2.963 artigos na busca e, após avaliação adequada, 9 artigos mostraram-se relevantes frente aos objetivos do estudo. Um total de 680 crianças foram incluídas, sendo 45% meninos. Não foi demonstrado significância estatística da frequência evacuatória, da consistência evacuatória, do sucesso terapêutico, da incontinência fecal e da dor abdominal com o uso de fibras nos pacientes com constipação infantil. Esses resultados devem ser interpretados com atenção devido a alta heterogeneidade clínica entre os estudos e a limitação metodológica dos artigos selecionados para análise. Conclusão: Existe uma grande falta de estudos qualificados para avaliar a suplementação de fibras no tratamento da constipação infantil, gerando um baixo grau de confiança para se estimar o efeito real dessa intervenção na população em questão. Até esse momento, conforme a literatura atual, deve-se apenas recomendar a ingestão adequada de fibras (oriundas de alimento ou suplemento) conforme a idade para as crianças com constipação, não se devendo prescrever a suplementação de fibras como tratamento para crianças e adolescentes constipados. / Introduction: Functional constipation corresponds to approximately 90 to 95% of cases of chronic constipation in children, being a frequent disease with great impact on the quality of life of the pediatric patient and his family. Increase fiber dietary intake is associated with stimulation of gastrointestinal peristalsis, accelerating colonic transit. However, there is still no clear evidence to support the routine use of fiber supplementation in the diet of children as part of the treatment of functional constipation. Objective: To gather current evidence on the use of fiber in the treatment of constipation in pediatric patients. Methods: Systematic review with meta-analysis of randomized controlled trials identified through Pubmed, Embase, LILACS and Cochrane databases. Inclusion Criteria: Randomized controlled trials; Patients aged between 2 and 18 years and diagnosed with non-organic constipation in use or not of drug treatment for constipation; Articles published in Portuguese, English, Spanish, French and German; Articles published in journals accessible to the researchers. Results: A total of 2.963 articles were found in the search and, after adequate evaluation, 9 articles were relevant to the study objective. A total of 680 children were included, of which 45% were boys. No statistical significance was demonstrated for evacuation frequency, bowel consistency, therapeutic success, fecal incontinence, and abdominal pain with fiber in patients with childhood constipation. These results should be interpreted with caution due to the high clinical heterogeneity between the studies and the methodological limitation of the articles selected for analysis. Conclusion: There is a great lack of qualified studies to evaluate fiber supplementation in the treatment of childhood constipation, generating a low degree of confidence in estimating the real effect of this intervention in the population in question. According to the current literature, it is necessary to recommend the adequate intake of fibers (from food or supplement) according to age for children with constipation, and it is not advised to prescribe fiber supplementation as a treatment for constipated children and adolescents.

O presente estudo teve como objetivo avaliar a performance dos métodos visual, radiográfico e baseados em fluorescência quanto à detecção de lesões de cárie em dentes permanentes e decíduos, realizar meta-análise com a performance global dos métodos e investigar as possíveis fontes de heterogeneidades. Foram realizadas três revisões distintas relacionadas a cada método diagnóstico. Quatro examinadores avaliaram os trabalhos, sendo pelo menos dois por método. Estes efetuaram a busca e seleção de artigos de modo independente; um pesquisador com experiência em trabalhos de diagnóstico de cárie resolveu os casos de conflito ou dúvida. As bases de dados eletrônicas utilizadas foram MEDLINE, EMBASE e Scopus, visando reduzir o viés de publicação, também foram buscados trabalhos acadêmicos e documentos não publicados, através do OpenSigle e Anais do Congresso ORCA. Após a seleção dos estudos por critérios de inclusão e exclusão bem definidos, os dados foram coletados e anotados, de maneira independente por cada revisor. Foi realizada a avaliação da qualidade metodológica através da ferramenta Quality Assessment of Diagnostic Accuracy Studies (QUADAS) e a análise estatística através do Paired Forest Plot, Summary Receiver Operating Characteristics analysis (SROC), relative diagnostic odds ratio (RDOR), análise de sensibilidade e gráfico de funil, além da investigação e exploração das possíveis fontes de heterogeneidade. Para o método visual, 102 artigos preencheram os critérios de inclusão. Em geral, a análise demonstrou que o método visual teve uma boa acurácia para a detecção de lesões de cárie. Embora estudos laboratoriais e clínicos tenham apresentado acurácias semelhantes, os valores de especificidade obtidos clinicamente foram maiores. Observamos, também, moderada a elevada heterogeneidade e evidência de viés de publicação na maioria dos estudos. Além disso, o uso de sistemas de escores visual resultaram em valores significativamente melhores de acurácia em comparação com estudos que utilizaram os seus próprios critérios. Quanto à relevância clínica dos estudos de acurácia do método visual, 96 estudos foram incluídos. Em geral, a maioria dos estudos falha em considerar alguns aspectos clínicos relevantes: somente 1 estudo incluiu avaliação da atividade das lesões; nenhum estudo considerou o seu prognóstico; 79 estudos não consideraram um desfecho clínico relevante e apenas 1 avaliou um desfecho centrado no paciente. Quanto à qualidade metodológica, a maioria dos estudos presentes apresentou alto risco de viés na seleção da amostra. Quanto ao método radiográfico, 116 artigos foram selecionados. Os métodos radiográficos convencional e digital apresentaram desempenho semelhante, mostrando valores maiores de especificidade comparados aos de sensibilidade, além de maiores valores de AUC no limiar de lesões mais avançadas. Quanto aos métodos baseados em fluorescência, 75 estudos preencheram os critérios de inclusão. Em geral, a análise demonstrou uma tendência a valores de acurácia semelhantes para todos os tipos de dentes, superfícies ou tipos de estudo. Houve uma tendência de melhor desempenho na detecção de lesões de cárie mais avançadas. Observamos, também, moderada a alta heterogeneidade e evidência de viés de publicação. Em conclusão, o método visual de detecção de lesões de cárie tem um bom desempenho geral, mas os estudos deveriam considerar desfechos clínicos relevantes, além da acurácia. Além disso, eles deveriam ser conduzidos com maior qualidade metodológica, principalmente em relação à seleção da amostra. Ainda, a utilização de índices detalhados e validados parece melhorar a acurácia do método. O método radiográfico não apresentou diferença entre os modos de aquisição de imagem (digital ou convencional), demonstrando uma tendência a valores superiores de especificidade e melhor desempenho em lesões avançadas. Quanto aos dispositivos baseados em fluorescência, estes mostraram desempenho global semelhante; no entanto, uma melhor acurácia na detecção de lesões de cárie mais avançadas tem sido observada. / This study aimed to evaluate the performance of visual, radiographic and fluorescence-based methods in the detection of caries lesions in permanent and primary teeth, to perform meta-analysis of the overall performance of the methods and to investigate possible sources of heterogeneity. Three distinct systematic reviews related to each diagnostic method were performed. Four observers evaluated the articles, at least two per method. They performed the search and study selection independently; a researcher with experience in caries diagnostic studies solved the cases of conflict or doubt. Electronic databases used were MEDLINE, EMBASE and Scopus, to reduce publication bias, were also pursued academic papers and unpublished documents through OpenSigle and the Annals of ORCA Congress. After the selection of studies by well-defined inclusion and exclusion criteria, data were collected and recorded, independently by each reviewer. Methodological quality assessment was performed through Quality Assessment of Diagnostic Accuracy Studies tool (QUADAS) and statistical analysis using Paired Forest Plot, Summary Receiver Operating Characteristics analysis (SROC), relative diagnostic odds ratio (RDOR), sensitivity analysis and funnel plot, in addition we explored possible sources of heterogeneity. Regarding the accuracy of visual method, 102 articles met the inclusion criteria. In general, visual analysis showed that the method had a good accuracy in the detection of caries lesions. Although laboratory and clinical studies have shown similar accuracies, clinical studies have obtained higher specificities. We also observed moderate to high heterogeneity and evidence of publication bias in most studies. In addition, the use of visual scoring systems resulted in significantly better accuracy compared to studies using their own criteria. Concerning the clinical relevance of accuracy studies of visual method, 96 studies were included. In general, most studies failed to consider some relevant clinical aspects: only 1 study included assessment of activity of the lesions; no study considered their prognosis; 79 studies did not consider a relevant clinical outcome and only 1 evaluated a patient-centered outcome. With regard to the methodological quality, most of the present studies showed a high risk of bias in the selection of the sample. With concern to the radiographic method, 116 articles were selected. Conventional and digital radiographic methods showed similar performance, showing greater specificity values compared to sensitivity, and higher AUC values at the threshold of more advanced caries lesions. Regarding fluorescence-based methods, 75 studies met the inclusion criteria. In general, the analysis demonstrated that the fluorescence-based method tend to have similar accuracy for all types of teeth, dental surfaces or settings. There was a trend of better performance of fluorescence methods in detecting more advanced caries lesions. We also observed moderate to high heterogeneity and evidenced publication bias. In conclusion, the visual method of detecting caries lesions has a good overall performance, but studies should consider relevant clinical outcomes, as well as accuracy. In addition, they should be conducted with greater methodological quality, especially in relation to sample selection. Still, the use of detailed and validated indices seems to improve the accuracy of the method. Radiographic method showed no difference between the image acquisition modes (digital or conventional), showing a tendency to higher values of specificity and performance in advanced caries lesions. Fluorescence-based devices showed similar overall performance; however, a better accuracy in detecting more advanced caries has been observed.