Vårdvetenskapligt Forskningscentrum/Centre for Health Sciences, Örebro University Hospital, County Council of Örebro, Örebro SE-703 62, Sweden

Clinical Chemistry, Department of Clinical and Experimental Medicine, Faculty of Health Sciences, Linköping University, Linköping, Sweden

School of Health and Medical Sciences, Örebro University, Örebro, Sweden

Abstract

Background

Although most animal stroke studies have demonstrated potent neuroprotective effects of estrogens, there are a number of articles reporting the opposite. In 2009, we made the case that this dichotomy was related to administered estrogen dose. Several other suggestions for the discordant results have also been propagated, including the age of the experimental animals and the length of hypoestrogenicity prior to estrogen administration. These two suggestions have gained much popularity, probably because of their kinship with the window of opportunity hypothesis, which is commonly used to explain the analogous dichotomy among human studies. We were therefore encouraged to perform an updated meta-analysis, and to improve it by including all relevant variables in a large multiple regression model, where the impact of confounders could be controlled for.

Results

The multiple regression model revealed an indisputable impact of estrogen administration mode on the effects of estrogens in ischemic stroke. Subcutaneous slow-release pellets differed from the injection and silastic capsule treatments in terms of impact of estrogens on ischemic stroke, showing that the first mentioned were more prone to render estrogens damaging. Neither the use of elderly animals nor the adoption of longer wash-out periods influenced estrogens’ effects on experimental ischemic stroke in rats.

Conclusions

We conclude that the discordant results regarding estrogens’ effects in rat models of ischemic stroke are a consequence of differences in estrogen administration modes. These results are not only of importance for the ongoing debate regarding menopausal hormone therapy, but also have an important bearing on experimental stroke methodology and the apparent translational roadblock for suggested stroke interventions.

Background

Estrogens’ effects in focal cerebral ischemia have been a matter of debate for more than a decade. Among studies performed in humans, the early epidemiological studies, indicating decreased stroke incidence from peri-menopausal hormone therapy

In 2009, a systematic analysis from our lab suggested that the use of different estrogen administration modes may explain the dichotomous results from the animal studies ^{®}), which has been demonstrated to produce extremely high, prolonged serum 17β-estradiol concentrations in rodents

The statistical shortcomings of the previous systematic analysis, the continuing debate on the matter and the publication of several additional original studies since the previous systematic analysis encouraged us to perform an updated and improved meta-analysis, where all methodological differences that reasonably could affect the impact of estrogens on experimental ischemic stroke would be controlled for in a large multiple regression model. The current meta-analysis therefore aimed to address the hypotheses that (A) estrogen administration mode, (B) the age of the experimental animals and (C) the length of hypoestrogenicity affects estrogens’ impact on stroke.

Results

Slow release pellets render estrogens significantly less protective/more damaging

Sixty-one studies, describing 124 pairs of estrogen-treated groups and control groups (subsequently referred to as “group pairs”) of rats in which focal cerebral ischemia was induced, were included (Figure

Three hundred and thirty-three articles were assessed for inclusion according to criteria A to I

**Three hundred and thirty-three articles were assessed for inclusion according to criteria A to I.** Consensus was reached to finally include 61 articles, of which 45 were included in the previous systematic analysis, and 16 were not

**Factor/outcome measure**

**Data type**

**Final categories/unit**

**Reference category for regression analyses**

**Rat properties**

Category

I.

II.

III.

Category

I.

II.

Category

I.

II.

III.

Category

[No]

[No]

[Yes]

Continuous

NA

**Estrogen administration**

Category

I.

II.

III.

Category

I.

II.

III.

IV.

Continuous

μg

NA

Continuous

μg/kg body weight

NA

Continuous

μg

NA

Category

I.

II.

Continuous

Hours

NA

**Focal ischemia procedure**

Category

I.

II.

III.

IV.

V.

Category

I.

II.

III.

Category

[No]

[No]

[Yes]

**Analysis procedure**

Continuous

Hours

NA

Category

I.

II.

III.

**Outcome measures**

Continuous

%

NA

After controlling for confounding factors, it remained clear that slow-release pellets rendered significantly higher EC-ratios than the injection and silastic capsule regimens did (p < 0.001)

**After controlling for confounding factors, it remained clear that slow-release pellets rendered significantly higher EC-ratios than the injection and silastic capsule regimens did (p < 0.001).** The black diamonds mark the mean EC-ratio with 95% confidence interval for each of the three administration modes. Of note, the slow-release pellet diamond reaches over the 100% EC-ratio line, demonstrating the pellets’ potential for harm as well as protection. EC-ratio = infarct size in estrogen treated group divided by infarct size in control group.

**Variable (reference category)**

**Variable categories**

**Regression coefficient**

**0.95 Confidence interval for regression coefficient**

**p-value**

**Lower bound**

**Upper bound**

Variables excluded in preceding multiple regression analysis (with backward exclusion) due to too low explanatory value:

Constant/baseline – showing the effect of the reference categories

NA

93.4

76.7

110.1

-31.9

-48.5

-15.2

0.000

-38.2

-54.9

-21.6

0.000

-59.3

-129.3

10.7

0.096

33.2

18.8

47.6

0.000

-42.0

-90.8

6.7

0.091

20.6

-53.1

94.2

0.581

2.3

-31.2

35.9

0.891

[Yes]

-15.3

-29.4

-1.2

0.034

-21.8

-54.2

10.6

0.184

-23.1

-39.6

-6.7

0.006

Continuous; hours

0.025

0.011

0.040

0.001

The final multiple regression model included 124 group pairs, and yielded an r^{2}-value of 0.484, hence explaining 48.4% of the variation in EC-ratio.

Higher dose in slow-release pellets increases the risk of estrogen exacerbating ischemic damage

Simple linear regression analyses including only group pairs within one specific estrogen administration mode category were run with dose (μg/pellet, daily injected dosage in μg and μg/silastic capsules, respectively) as the independent factor and EC-ratio as the outcome variable. There was a significant, positive relation between slow-release pellet dose and EC-ratio (y = 72.8 + 0.05x; p = 0.001; N = 39), meaning that increasing the pellet dose increased the likeliness that estrogens would be damaging. This model had an r^{2}-value of 0.26, indicating that it explained 26% of the EC-ratio variation (Figure ^{2}-value (0.26), and there was not even a slight tendency for increased damage in elderly rats (regression coefficient -2.9, with confidence intervals -56.1 to 50.2). A similar analysis to address hypothesis C could not be done since only one single study administered estrogen via pellets more than 14 days after ovariectomy.

By three simple linear regression analyses, it was found that higher pellet doses significantly increased EC-ratio (y = 72.8 + 0.05x; p = 0.001), while there was an opposite trend (not significant) among injection and silastic capsule regimens

**By three simple linear regression analyses, it was found that higher pellet doses significantly increased EC-ratio (y = 72.8 + 0.05x; p = 0.001), while there was an opposite trend (not significant) among injection and silastic capsule regimens.** The numbers of group pairs included in the slow-release pellet, injection and silastic capsule simple regression models were 39, 50 and 31, respectively. Note that the injection graph has been compressed in the upper dose range to accommodate three group pairs being administered very high doses.

For the group pairs administered estrogens via injection or silastic capsules, there were no significant relations between dosage and EC-ratio, however the trend in both models was that increased estrogen dose increased neuroprotection (Injection dose vs. EC-ratio: y = 62.8-0.002x; p = 0.14; N = 50; Silastic capsule dose vs. EC-ratio: 63.2-0.05x; p = 0.17; N = 31).

Descriptive statistics

The frequencies of categories in the 124 included group pairs are depicted in Figure

The frequencies of the different categories in the 124 group pairs are presented as percentages

**The frequencies of the different categories in the 124 group pairs are presented as percentages.** Please note that the Disease and Estrogen type variables were omitted from the statistical analysis because too few studies used diseased animals and other estrogens than 17β-estradiol, respectively.

**Variable**

**Unit**

**Mean**

**Standard deviation**

**Median**

**Inter-quartile range**

16.0

7.3

15

10-20

μg

412.6

53.0

180.0

25.0-1000.0

μg/kg body weight

794.8

3006.2

100.0

11.9-200.0

μg

77.5

124.8

28.0

27.2-28.0

Hours

221.6

600.5

168

0.5-324

Hours

49.6

96.4

24.0

24.0-24.0

%

72.1

42.3

56.2

43.7-93.9

Discussion

The main multiple regression model, controlling for all listed confounders, revealed an indisputable impact of estrogen administration mode on the effects of estrogens in ischemic stroke (p < 0.001). Slow-release pellets significantly differed from the injection and silastic capsule treatments in terms of resulting EC-ratio, showing that slow-release pellets are more prone to render estrogens damaging. Of note, the slow-release pellet confidence interval extended over the 100% EC-ratio line, underscoring the potential for harm as well as benefit (Figure

A plausible explanation for the tendency of estrogens in slow-release pellets to be less protective/more damaging is that this high-dose administration mode causes extremely high, elevated serum 17β-estradiol concentrations. It has repeatedly been shown that the pellets from Innovative Research of America^{®}, albeit with a large portion of unpredictability, render prolonged serum concentrations that are well beyond the physiological spectrum. Silastic capsules produce serum concentrations in a much lower, often physiological, range, while daily injections result in hours-long spikes followed by estrogen-deficient intervals, rendering the 24 h-average serum concentrations low

Regarding hypothesis B, no impact of the use of elderly animals on estrogens’ effects in stroke was seen. In the preceding Backward multiple regression analysis, the variable was the first one to be excluded, without even a trace of higher EC-ratio in elderly rats (regression coefficient before exclusion: -2.6, 95% confidence intervals -29.2 to 24.0; p = 0.85). If anything, the regression coefficient suggested

Similarly for hypothesis C, the second (dummy) variable to be excluded in the Backward analysis was the wash-out category

A few other variables were found to significantly affect the EC-ratio, however since these did not address the main hypotheses, we refrain from drawing conclusions about them, and refer to Table

Strengths and weaknesses

An inherent draw-back with linear regression analysis is that linear relations are assumed, which evidently is not always true. This imperfection must be kept in mind when assessing the results. Another weakness of the current study is that the 124 group pairs were described in only 61 articles, which in turn were published by even fewer research groups. To be perfectly stringent, group pairs from the same article or the same research group should not be regarded as independent. However, creating dummy variables for each article or research group would have made the analysis totally devoid of statistical power, and thus impossible to perform.

The analysis presents a composite result of data gathered from 124 group pairs and more than 1900 rats, handled in a rich variety of experimental conditions. The main strength of the current meta-analysis is that all factors that have been suggested to be responsible for the discrepant results were tested in parallel, thus potently correcting for confounders.

Conclusions

We conclude that the discordant results regarding estrogens’ effects in rat models of ischemic stroke are a consequence of differences in estrogen administration modes, corroborating the earlier systematic analysis from our laboratory

Methods

Article search and inclusion

To define articles to include in the meta-analysis, Medline was the 5^{th} of March 2013 searched with the search line

A. Article written in English

B. Original research article

C. The experiment was performed in adolescent, adult or elderly rats

D. The rats were males or ovariectomized and/or reproductively senescent females

E. Each estrogen treatment group had a corresponding control group, and no factor except the estrogen administration per se differed between the groups

F. No other treatments were administered in parallel to the estrogen regime

G. One single focal cerebral ischemic lesion was induced in the animals

H. The estrogens were administered prior to or at the time of MCAo

I. Infarct sizes were assessed and results presented

Consensus was reached to include 61 studies, of which 45 were included in the previous systematic analysis

It was initially the intention to expand the meta-analysis by setting up parallel models for studies in mice. However, the search line

Data extraction

Eighteen pre-defined method and result variables were extracted from the 61 articles describing the 124 group pairs. Method variables were chosen with the aim to encompass all methodological aspects that theoretically could influence estrogens’ effects on ischemia. When extracting the method data, the principle “If it was not described, it was not performed” was strictly adhered to. All extracted variables are presented in Table

Processing of data

Variable definitions and categorizations

Sprague Dawley and Wistar were the only strains that were sufficiently well-used (at least 5 group pairs) to deserve separate categories. All other registered strains (Lister hooded rats, Diabetes type 1 rats, Spontaneously hypertensive rats – stroke prone, Spontaneously hypertensive rats and rats of unknown strain) were put in an

The category

Too few studies used diseased animals for the

The continuous variable

Slow-release pellets from the company Innovative Research of America^{®}, various injections and subcutaneous silastic capsules defined the three

The variable

The three variables

To deal with the fact that male rats are not ovariectomized, and that no fields in the multiple regression analysis are allowed to be empty, the variable

In the variable

Regarding the

Regarding the category

Edema correction when calculating infarcts sizes can be performed according to at least two principles. Swanson et al.

This method assumes that all edema is in the infarct, and not outside it, and defined the category

The assumption here is instead that the edema is equally distributed in the entire infarcted hemisphere, and group pairs in which this procedure had been adopted were registered in the category

The outcome variable

Statistical analyses

To identify which methodological factors significantly affected the EC-ratio, multiple linear regression analysis was used. As abovementioned, this analysis was the most important improvement from our previous systematic analysis

All statistical calculations were performed in SPSS (Version 20, IBM Corporation, Armonk, NY, USA). P-values <0.05 were considered statistically significant.

Protocol violations

As mentioned earlier, the variables

One study

Competing interests

The authors state that they have no competing interests to report.

Authors’ contributions

JOS came up with the idea, contributed to the design, performed part of the article inclusion process, performed all analyses and drafted the manuscript. EI contributed to the design, performed part of the article inclusion process and revised the manuscript. Both authors read and approved the final manuscript.

Acknowledgements

We gratefully acknowledge the expert advice of statistician Karl Wahlin (PhD). This study was supported by the County Council of Örebro and Linköping University, Sweden.