Systematic review of animal studiesThe case of statins

Pecoraro V, Moja L, Dall'Olmo L, Cappellini G, Garattini S. Most appropriateanimal models to study the efficacy of statins: a systematic review. Eur J Clin Invest. 2014;44(9):848-71. doi: 10.1111/eci.12304.

Moja L, Pecoraro V, Ciccolallo L, Dall'Olmo L, Virgili G, Garattini S. Flawsin animal studies exploring statins and impact on meta-analysis. Eur J ClinInvest. 2014;44(6):597-612. doi: 10.1111/eci.12264.

Our work evaluates the efficacy of statins in animal studies to provide an overall synthesis that facilitate the interpretation of aggregated data from the basic research conducted so far.

Methods: We have aggregated the findings in animal models - mice, rats and rabbits - using the technique of systematic review and meta-analysis.

We included 161 animal studies and we analysed 120 studies, accounting for 2432 animals.

Total Cholesterol (TC): reduction in all animal models

Rats-10%

Mice-20%

Rabbits-30%

Pecoraro et al. Eur J Clin Invest. 2014;44(9):848-71

Pecoraro et al. Eur J Clin Invest. 2014;44(9):848-71

Pecoraro et al. Eur J Clin Invest. 2014;44(9):848-71

Pecoraro et al. Eur J Clin Invest. 2014;44(9):848-71

Pecoraro et al. Eur J Clin Invest. 2014;44(9):848-71

Overall estimate for blood pressure reduction and infarct size in each animal model and strain

Blood pressure reduction

Animal model

Strain Studies included Studies Total sample

Overall estimate (mmHg) Heterogeneity†

Test for subgroup differences*

N N MD 95% CI P I² P

MICE Apo E-/- All ‡ 3 84 -4 [-8 ; -1] 0,01 0%

RATS Spontaneously Hypertensive

All 14 239 -17 [-24 ; -10] <0,001 95%

St diet 11 201 -20 [-32 ; -9] <0,001 94% <0,001

HC diet 3 38 -6 [-15 ; 2] 0,2 91%

Sprague Dawley

All ‡ 10 201 -5 [-12 ; 1] 0,1 49%

Wistar All ‡ 5 82 -10 [-26 ; 7] 0,3 86%

Infarct size

Animal model

Strain Included studies Studies Total sample

Overall estimate (%) Heterogeneity†

N N SMD 95% CI P I²

RATS Sprague Dawley

All 6 101 -1 [-2 ; -0,3] 0,009 75%

Wistar All 3 65 -0,5 [-1,6 ; 0,6] 0,4 77%

Statins lowered the total cholesterol across all species, although with large differences in the effect size: -30% in rabbits, -20% in mice and -10% in rats.

The reduction was larger in animals fed on a high-cholesterol diet.

Statins reduced infarct volume but did not consistently reduce the blood pressure or effect the overall survival. Few studies considered strains at high risk of cardiovascular diseases or hard outcomes.

Main results

Review authors’ judgments about each risk of bias item presented as percentages across all included studies

Moja et al. Eur J Clin Invest. 2014;44(6):597-612

Meta-analysis graph for Rats using Ratio Of Means

•Point estimate•Confidence interval•Prediction interval•Fixed vs random effects model

Moja et al. Eur J Clin Invest. 2014;44(6):597-612

Meta-analysis graph for Mice using Ratio Of Means

Moja et al. Eur J Clin Invest. 2014;44(6):597-612

Meta-analysis graph for Rabbits using Ratio Of Means

Moja et al. Eur J Clin Invest. 2014;44(6):597-612

Counter-enhanced funnel plots to explore small studies effects in Mice, Rabbits and Rats

Moja et al. Eur J Clin Invest. 2014;44(6):597-612

The funnel plots of the three animal models suggested that the effect size of studies were highly variable, overall, in both direction and magnitude. Most of the studies had high precision; this was clearly displayed in the general funnel plot in which most of them gathered at the top

Main results• Our systematic review highlights the inadequate experimental

reporting: over half did not report design quality components (randomisation and blinding), many omitted information about animal gender, age or weight, all did not report the sample size.

• The description of characteristics was consistently unsatisfactory and it can result in such studies being un-interpretable and difficult to reproduce.

• Fixed- and random-effects models gave different results: ratio of effect size increased by !! five folds !!

• Ratio Of Means model occasionally preferable in animal studies.• Heterogeneity was consistently substantial. Accounting for

covariates had minimal impact. • Publication bias is highly suspected across studies.

Taylor 2012, The Cochnale Library Issue 5

Humans

Pecoraro et al. Eur J Clin Invest. 2014;44(9):848-71

Mor

talit

yRats

Mice

Conclusions

Heterogeneity was consistently substantial, raising questions about the appropriateness of a cumulative approach. Results of meta-analyses in animal studies – statins - were difficult to interpret and reproduce. Different meta-analytic approaches were proved to be largely inconsistent, suggesting tha reliable meta-analystic approaches cannot be not widely used in animal studies. Policies that address these issues are required from investigators, editors and institutions that care about the quality standards and ethics of animal research.