Article type
Year
Abstract
Background: These years there were a number of developments in the methodology used in systematic reviews or meta-analysis of screening tests. But the epidemic characteristics and quality of meta-analysis or systematic reviews in screening tests has not been evaluated.
Objectives: To explore the epidemic characteristics and quality of meta-analysis and systematic reviews in screening tests.
Methods: Eight databases (including MEDLINE and EMBASE) were searched. Outcome analysis methods, methodological quality, statistical analysis software and conflicts of interest reported were evaluated.
Results: The searches yielded 4,726 relevant articles. After application of inclusion criteria, 98 studies were included. Cancer was the most common category of target disorder (33% or studies), followed by depression (6%) and alcohol abuse disorders (6%). The main methods used for outcome analysis were sensitivity (96% 94/98 of included studies), specificity (96% 94/98), positive predictive value (57.1% 56/98), negative predictive value (52% 51/98), positive likelihood ratio (30.6% 30/98), negative likelihood ratio (29.6% 29/98), diagnostic odds ratio (32.7% 32/98), pre-test odds (8.2% 8/98), post-test odds (6.1% 6/98), receiver operating characteristic curves (SROC: 8.2% 8/98). Statistical analysis reported were: fixed-effects model (8.2% 9/98 of included studies), random-effects model (15.7% 19/98), publication bias (5.6% 9/98), bayesian bivariate hierarchical mode (6.1% 6/98), sensitivity analysis (22.4% 22/98), heterogeneity analysis (35.7% 35/98). The included reviews assessed methodological quality through QUADAS (30.6% 30/98 of included studies), not reported (56.1% 55/98), and others (13.3% 13/98) (Figure1), and used the following statistical softwares: Meta-DiSc (19.4% 19/98 of included studies), STATA (14.3% 14/98), SAS (4.1% 4/98), RevMan (6.1%, 6/98), others (13.3% 13/98).
Conclusions: The number of meta-analyses and systematic reviews on screening tests has increased over time (Figure2). Sensitivity, specificity, positive predictive value and negative predictive value remain the preferred methods for analysis. Although quality assessment in meta-analyses has improved with the introduction of QUADAS in 2004, most screening test reviews still do not useit.
Objectives: To explore the epidemic characteristics and quality of meta-analysis and systematic reviews in screening tests.
Methods: Eight databases (including MEDLINE and EMBASE) were searched. Outcome analysis methods, methodological quality, statistical analysis software and conflicts of interest reported were evaluated.
Results: The searches yielded 4,726 relevant articles. After application of inclusion criteria, 98 studies were included. Cancer was the most common category of target disorder (33% or studies), followed by depression (6%) and alcohol abuse disorders (6%). The main methods used for outcome analysis were sensitivity (96% 94/98 of included studies), specificity (96% 94/98), positive predictive value (57.1% 56/98), negative predictive value (52% 51/98), positive likelihood ratio (30.6% 30/98), negative likelihood ratio (29.6% 29/98), diagnostic odds ratio (32.7% 32/98), pre-test odds (8.2% 8/98), post-test odds (6.1% 6/98), receiver operating characteristic curves (SROC: 8.2% 8/98). Statistical analysis reported were: fixed-effects model (8.2% 9/98 of included studies), random-effects model (15.7% 19/98), publication bias (5.6% 9/98), bayesian bivariate hierarchical mode (6.1% 6/98), sensitivity analysis (22.4% 22/98), heterogeneity analysis (35.7% 35/98). The included reviews assessed methodological quality through QUADAS (30.6% 30/98 of included studies), not reported (56.1% 55/98), and others (13.3% 13/98) (Figure1), and used the following statistical softwares: Meta-DiSc (19.4% 19/98 of included studies), STATA (14.3% 14/98), SAS (4.1% 4/98), RevMan (6.1%, 6/98), others (13.3% 13/98).
Conclusions: The number of meta-analyses and systematic reviews on screening tests has increased over time (Figure2). Sensitivity, specificity, positive predictive value and negative predictive value remain the preferred methods for analysis. Although quality assessment in meta-analyses has improved with the introduction of QUADAS in 2004, most screening test reviews still do not useit.
Images
