Article type
Year
Abstract
Background: Publication bias is evident in treatment meta-analyses with the odds for publication of positive compared with negative studies being 2.5 [1]. Diagnostic decision-making has important implications yet there remains the possibility of publication bias [2] or bias due to pooled sensitivity and specificity [3] and it may be more of a concern in diagnostic meta-analyses [3]. Funnel plot methodology, assessing heterogeneity, assumes precision in estimating effect increases with sample size. Small studies scatter widely at the funnel base due to random variation, with larger studies spread narrowly at the top [4] with bias represented by asymmetry and skewness.
Objectives: To assess for the presence of publication bias in a previously published systematic review on echocardiographic LVH diagnosis (cohort studies that recorded subsequent morbidity and mortality) [5].
Methods: We analysed data using a funnel plot graph estimating effect against precision in addition to a line plot/regression of the standard normal deviate (SND), (relative risk/standard error), against precision (1/standard error) [4]. The analysis was also verified using Egger's methodology [2]. The intercept value (a) estimates asymmetry (values >0 indicating trends of higher test accuracy in smaller studies).
Results: The funnel plot was asymmetrical with baseline smaller studies skewed right and two larger studies skewed left at the top. The regression was positive showing asymmetry trends. The line plot of SND against precision had a positive coefficient and p value of 0.03, while Eggers test using diagnostic d was positive and significant. While this asymmetry may represent publication bias, it may be because of poorer quality studies, the non-inclusion of foreign language publications, or due to chance.
Conclusions: There was a trend towards asymmetry of results on the analysis, however the definite cause of this cannot be ascertained.
References:
1. Glasziou P et al. 2001. Systematic reviews in health care: a practical guide.
2. Songa F et al. 2002. Int J Epi 31:88-95.
3. Irwig L et al. 1995. Meta-analytic methods for diagnostic test accuracy. J Clin Epidemiol 48:119-30.
4. Egger M et al 1997. Bias in meta-analysis detected by a simple, graphical test. BMJ 315:629-634.
5. Hilton DJ et al. 2002 Echocardiographic LVH and prognosis. JACC 39:166B
Objectives: To assess for the presence of publication bias in a previously published systematic review on echocardiographic LVH diagnosis (cohort studies that recorded subsequent morbidity and mortality) [5].
Methods: We analysed data using a funnel plot graph estimating effect against precision in addition to a line plot/regression of the standard normal deviate (SND), (relative risk/standard error), against precision (1/standard error) [4]. The analysis was also verified using Egger's methodology [2]. The intercept value (a) estimates asymmetry (values >0 indicating trends of higher test accuracy in smaller studies).
Results: The funnel plot was asymmetrical with baseline smaller studies skewed right and two larger studies skewed left at the top. The regression was positive showing asymmetry trends. The line plot of SND against precision had a positive coefficient and p value of 0.03, while Eggers test using diagnostic d was positive and significant. While this asymmetry may represent publication bias, it may be because of poorer quality studies, the non-inclusion of foreign language publications, or due to chance.
Conclusions: There was a trend towards asymmetry of results on the analysis, however the definite cause of this cannot be ascertained.
References:
1. Glasziou P et al. 2001. Systematic reviews in health care: a practical guide.
2. Songa F et al. 2002. Int J Epi 31:88-95.
3. Irwig L et al. 1995. Meta-analytic methods for diagnostic test accuracy. J Clin Epidemiol 48:119-30.
4. Egger M et al 1997. Bias in meta-analysis detected by a simple, graphical test. BMJ 315:629-634.
5. Hilton DJ et al. 2002 Echocardiographic LVH and prognosis. JACC 39:166B