Limits...
Industry sponsorship and publication bias among animal studies evaluating the effects of statins on atherosclerosis and bone outcomes: a meta-analysis.

Anglemyer AT, Krauth D, Bero L - BMC Med Res Methodol (2015)

Bottom Line: We conducted two independent systematic reviews and meta-analyses identifying animal studies evaluating the effect of statins on reducing the risk of atherosclerosis outcomes (n = 49) and increasing the likelihood of beneficial bone outcomes (n = 45).We also found that inadequate reporting of sponsorship in animal studies is still exceedingly common.In meta-analyses assessing the effects of statins on atherosclerosis and bone outcomes in animal studies, we found evidence of publication bias, though small numbers of industry-sponsored studies limit the interpretation of the trim-and-fill results.

View Article: PubMed Central - PubMed

Affiliation: Department of Clinical Pharmacy, University of California San Francisco, 50 Beale Street, San Francisco, CA, 94143, USA. Andrew.Anglemyer@gmail.com.

ABSTRACT

Background: The effect that sponsorship has on publication rates or overall effect estimates in animal studies is unclear, though methodological biases are prevalent in animal studies of statins and there may be differences in efficacy estimates between industry and non-industry sponsored studies. In the present analysis, we evaluated the impact of funding source on publication bias in animal studies estimating the effect of statins on atherosclerosis and bone outcomes.

Methods: We conducted two independent systematic reviews and meta-analyses identifying animal studies evaluating the effect of statins on reducing the risk of atherosclerosis outcomes (n = 49) and increasing the likelihood of beneficial bone outcomes (n = 45). After stratifying the included studies within each systematic review by funding source, three separate analyses were employed to assess publication bias in these meta-analyses—funnel plots, Egger's Linear Regression, and the Trim and Fill methods.

Results: We found potential evidence of publication bias, primarily in non-industry sponsored studies. In all 3 assessments of publication bias, we found evidence of publication bias in non-industry sponsored studies, while in industry-sponsored studies publication bias was not evident in funnel plots and Egger's regression tests. We also found that inadequate reporting of sponsorship in animal studies is still exceedingly common.

Conclusions: In meta-analyses assessing the effects of statins on atherosclerosis and bone outcomes in animal studies, we found evidence of publication bias, though small numbers of industry-sponsored studies limit the interpretation of the trim-and-fill results. This publication bias is more prominent in non-industry sponsored studies. Industry and non-industry funded researchers may have different incentives for publication. Industry may have a financial interest to publish all preclinical animal studies to maximize the success of subsequent trials in humans, whereas non-industry funded academics may prefer to publish high impact statistically significant results only. Differences in previously published effect estimates between industry- and non-industry sponsored animal studies may be partially explained by publication bias.

No MeSH data available.


Related in: MedlinePlus

Egger’s linear regression method. Data from meta-analyses of atherosclerosis studies (a-d) and bone studies (e-h). Plots show the standardized treatment effect plotted against precision (inverse of standard error). In the absence of funnel plot asymmetry, the slope of the regression line will be zero.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4353470&req=5

Fig2: Egger’s linear regression method. Data from meta-analyses of atherosclerosis studies (a-d) and bone studies (e-h). Plots show the standardized treatment effect plotted against precision (inverse of standard error). In the absence of funnel plot asymmetry, the slope of the regression line will be zero.

Mentions: Funnel plot asymmetry was tested using Egger’s linear regression. Across all studies, (Figure 2; panels a, e) there appears to be bias in both atherosclerosis and bone studies. For atherosclerosis studies, the bias coefficient is −4.17 (95% CI −5.61, −2.73; p value < 0.0001), while for bone studies the bias coefficient is −2.43 (95% CI −4.42, −0.43; p value = 0.021). After stratification by funding source, bias (i.e., funnel plot asymmetry) appears to remain in atherosclerosis and bone studies with no industry sponsorship (Figure 2; panels b, f)—bias coefficients: −4.83 (95% CI −7.25, −2.41; p value < 0.0001), −2.80 (95% CI −5.42, −0.17; p value = 0.043), respectively. However, bias is absent in industry-sponsored studies of atherosclerosis and bone outcomes (bias coefficients: −1.56; 95% CI −4.26, 1.15; p value = 0.270. -2.95; 95% CI −6.60, 0.69; p value = 0.180, respectively) (Figure 2; panels c, g). Similarly, bias is absent in studies with no statement of sponsorship (−2.91; 95% CI −5.96, 0.13; p value = 0.088. 3.30; 95% CI −2.11, 8.71; p value = 0.253, in atherosclerosis and bone studies, respectively) (Figure 2; panels d, h).Figure 2


Industry sponsorship and publication bias among animal studies evaluating the effects of statins on atherosclerosis and bone outcomes: a meta-analysis.

Anglemyer AT, Krauth D, Bero L - BMC Med Res Methodol (2015)

Egger’s linear regression method. Data from meta-analyses of atherosclerosis studies (a-d) and bone studies (e-h). Plots show the standardized treatment effect plotted against precision (inverse of standard error). In the absence of funnel plot asymmetry, the slope of the regression line will be zero.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4353470&req=5

Fig2: Egger’s linear regression method. Data from meta-analyses of atherosclerosis studies (a-d) and bone studies (e-h). Plots show the standardized treatment effect plotted against precision (inverse of standard error). In the absence of funnel plot asymmetry, the slope of the regression line will be zero.
Mentions: Funnel plot asymmetry was tested using Egger’s linear regression. Across all studies, (Figure 2; panels a, e) there appears to be bias in both atherosclerosis and bone studies. For atherosclerosis studies, the bias coefficient is −4.17 (95% CI −5.61, −2.73; p value < 0.0001), while for bone studies the bias coefficient is −2.43 (95% CI −4.42, −0.43; p value = 0.021). After stratification by funding source, bias (i.e., funnel plot asymmetry) appears to remain in atherosclerosis and bone studies with no industry sponsorship (Figure 2; panels b, f)—bias coefficients: −4.83 (95% CI −7.25, −2.41; p value < 0.0001), −2.80 (95% CI −5.42, −0.17; p value = 0.043), respectively. However, bias is absent in industry-sponsored studies of atherosclerosis and bone outcomes (bias coefficients: −1.56; 95% CI −4.26, 1.15; p value = 0.270. -2.95; 95% CI −6.60, 0.69; p value = 0.180, respectively) (Figure 2; panels c, g). Similarly, bias is absent in studies with no statement of sponsorship (−2.91; 95% CI −5.96, 0.13; p value = 0.088. 3.30; 95% CI −2.11, 8.71; p value = 0.253, in atherosclerosis and bone studies, respectively) (Figure 2; panels d, h).Figure 2

Bottom Line: We conducted two independent systematic reviews and meta-analyses identifying animal studies evaluating the effect of statins on reducing the risk of atherosclerosis outcomes (n = 49) and increasing the likelihood of beneficial bone outcomes (n = 45).We also found that inadequate reporting of sponsorship in animal studies is still exceedingly common.In meta-analyses assessing the effects of statins on atherosclerosis and bone outcomes in animal studies, we found evidence of publication bias, though small numbers of industry-sponsored studies limit the interpretation of the trim-and-fill results.

View Article: PubMed Central - PubMed

Affiliation: Department of Clinical Pharmacy, University of California San Francisco, 50 Beale Street, San Francisco, CA, 94143, USA. Andrew.Anglemyer@gmail.com.

ABSTRACT

Background: The effect that sponsorship has on publication rates or overall effect estimates in animal studies is unclear, though methodological biases are prevalent in animal studies of statins and there may be differences in efficacy estimates between industry and non-industry sponsored studies. In the present analysis, we evaluated the impact of funding source on publication bias in animal studies estimating the effect of statins on atherosclerosis and bone outcomes.

Methods: We conducted two independent systematic reviews and meta-analyses identifying animal studies evaluating the effect of statins on reducing the risk of atherosclerosis outcomes (n = 49) and increasing the likelihood of beneficial bone outcomes (n = 45). After stratifying the included studies within each systematic review by funding source, three separate analyses were employed to assess publication bias in these meta-analyses—funnel plots, Egger's Linear Regression, and the Trim and Fill methods.

Results: We found potential evidence of publication bias, primarily in non-industry sponsored studies. In all 3 assessments of publication bias, we found evidence of publication bias in non-industry sponsored studies, while in industry-sponsored studies publication bias was not evident in funnel plots and Egger's regression tests. We also found that inadequate reporting of sponsorship in animal studies is still exceedingly common.

Conclusions: In meta-analyses assessing the effects of statins on atherosclerosis and bone outcomes in animal studies, we found evidence of publication bias, though small numbers of industry-sponsored studies limit the interpretation of the trim-and-fill results. This publication bias is more prominent in non-industry sponsored studies. Industry and non-industry funded researchers may have different incentives for publication. Industry may have a financial interest to publish all preclinical animal studies to maximize the success of subsequent trials in humans, whereas non-industry funded academics may prefer to publish high impact statistically significant results only. Differences in previously published effect estimates between industry- and non-industry sponsored animal studies may be partially explained by publication bias.

No MeSH data available.


Related in: MedlinePlus