Progesterone treatment for experimental stroke: an individual animal meta-analysis

Abstract

Preclinical studies suggest progesterone is neuroprotective after cerebral ischemia. The gold standard for assessing intervention effects across studies within and between subgroups is to use meta-analysis based on individual animal data (IAD). Preclinical studies of progesterone in experimental stroke were identified from searches of electronic databases and reference lists. Corresponding authors of papers of interest were contacted to obtain IAD and, if unavailable, summary data were obtained from the publication. Data are given as standardized mean differences (SMDs, continuous data) or odds ratios (binary data), with 95% confidence intervals (95% CIs). In an unadjusted analysis of IAD and summary data, progesterone reduced standardized lesion volume (SMD -0.766, 95% CI -1.173 to -0.358, P<0.001). Publication bias was apparent on visual inspection of a Begg's funnel plot on lesion volume and statistically using Egger's test (P=0.001). Using individual animal data alone, progesterone was associated with an increase in death in adjusted analysis (odds ratio 2.64, 95% CI 1.17 to 5.97, P=0.020). Although progesterone significantly reduced lesion volume, it also appeared to increase the incidence of death after experimental stroke, particularly in young ovariectomized female animals. Experimental studies must report the effect of interactions on death and on modifiers, such as age and sex.

Figure 1

Flow chart showing study search, and included and excluded studies. Summary data were used for studies where individual animal data were not made available. Studies were excluded if they did not report induction of focal cerebral ischemia, administration of progesterone, or measurement of lesion volume, or did not contain original data; n=number of studies.

Figure 2

Begg's funnel plot for studies of progesterone on lesion volume when combining individual animal and summary data. The funnel plot relates precision (reciprocal of s.e.) to the standardized mean difference (SMD). Asymmetry is present indicating publication bias (Egger's test, P=0.001).

Figure 3

Forrest plot of effect of progesterone on lesion volume using individual animal and summary data. Studies are ordered by animals' sex: male; young ovariectomized female; old female.

Figure 4

Relationship between blood concentration and loading dose of progesterone. Progesterone blood concentration was measured at 1 hour from the start of treatment. Equation of best fit [Progesterone] in blood (ng/mL)=4.156 × progesterone loading dose (mg/kg)+1.304.