Aplicación de una variante del meta-análisis a un estudio de la efectividad de la progesterona vaginal en la reducción del parto pretermino

Abstract

Meta-analysis is the systemic and comprehensive analysis of the information about a group of empirical researches results that analyze the same problem, to reach a synergic result-relevant knowledge- and generalizable conclusions, also to optimize the cognitive, investigative and making-decision processes through quantitative and qualitative methods. Meta-analysis methodology has some limitations causing distorted findings, which depend on the quality of empirical studies, as well as the selection and publication biases. To solve these limitations, a meta-analysis variant is used. It consists on the stratification of the studies from the a posteriori statistic power analysis of the empirical works included in the revision. OBJECTIVE: To increase the consistency and robustness of the meta-analytical study results from the control of the threats of selection and publication biases as well as the heterogeneity regarding the methodological quality of the information. This variant is used in the replication of the work: “Vaginal Progesterone in Women with Asymptomatic Sonographic Short Cervix in the Midtrimester Decreases Preterm Delivery and Neonatal Morbidity: A Systematic Review and Meta-Analysis of Individual Patient Data” due to the risk of preterm delivery is an important health problem at worldwide. METHODS: A VARIANT OF META-ANALYSIS. RESULTS: All studies included in the revision have lower and trivial statistical power. In the “lower stratum” of statistical power (60%-80%), it was obtained a combined relative risk (RR) of 0.57, representing a protector and significant effect of the progesterone; due to the 95% confidence intervals meet not the unit (0.40-0.80). In the case of “less than 60%” of statistic power, it was obtained a RR=0.79, without statistical signification because the 95% confidence intervals contains the unit (0.32-1.91). For this reason, in this status, the vaginal progesterone does not reduce notably the preterm deliveries before 33 weeks. CONCLUSIONS:Our findings demonstrated that the meta-analytical methodology can be feasibly improved, being the a posteriori statistic power analysis, a useful tool to control the heterogeneity of the information quality of the empirical studies included in the revision, allowing to increase the consistency and robustness of the meta-analytical results. Stratification of empirical works, considering the calculated statistic power, allows homogenizing the methodological quality of the intra-stratus information. Also it permits to study and control the threats of various biases types, as well as the comparison of intra and inter-stratums.It is recommended the use of this meta-analytical variant in future revision studies. All studies included in the revision that was replicated showed lower and trivial statistical power.In the “lower stratum” of statistical power (60%-80%), it was obtained a combined relative risk (RR) of 0.57, representing a protector and significant effect of the progesterone; due to the 95% confidence intervals meet not the unit (0.40-0.80). In the trivial stratum (statistic power < 60%), it was obtained a RR=0.79, without statistical signification, because the 95% confidence intervals contains the unit (0.32-1.91).

Date of Award	17 Jul 2012
Original language	Spanish
Supervisor	Lluis Cabero Roura (Director) & Maria Elena Carreras Moratonas (Director)