Statins in pre- eclampsia or fetal growth restriction: A systematic review and meta- analysis on maternal blood pressure and fetal growth across species

Background: Several human randomised controlled trials (RCTs) are investigating the effects of statins on pre- eclampsia (PE) and fetal growth restriction (FGR). This cross- species meta- analysis summarises the preclinical evidence of statin use


| I N TRODUC TION
Pre-eclampsia (PE) and fetal growth restriction (FGR) complicate 2%-8% of pregnancies, leading to significant maternal and fetal morbidity and mortality worldwide. 1 PE is a multisystem progressive endothelial disorder defined by the onset of hypertension in combination with either proteinuria, kidney or liver injury, neurological features, haemolysis, thrombocytopenia, or FGR after the 20th week of gestation. 2 The underlying endothelial dysfunction is thought to be the result of anti-angiogenic factors released by the placenta, including soluble Fms-like tyrosine kinase-1 (sFlt-1) and soluble endoglin (sEng) in combination with the loss of vascular protective heme oxygenase-1 (HO-1) enzyme pathway. [3][4][5][6][7] Several animal models with alternations in sFlt-1 levels have shown alternations in HO-1 levels, reduced PIGF and VEGF levels leading to vascular and glomerular damage. [6][7][8][9] Currently, the only definite resolution for PE is the removal of the placenta at birth. 10 Iatrogenic birth remote from term potentially results in adverse neonatal long-term consequences. To prolong pregnancy safely, the maternal condition can be stabilised by antihypertensive medication and magnesium sulphate administration, and the fetal condition can be monitored by ultrasound or cardiotocography. As yet, a treatment for the underlying placental pathology and affected maternal vasculature, which might result in prolonged pregnancy, is lacking. Recently, statins have been proposed to prevent PE or improve disease outcome. 1 Statins increase the activity of HO-1/carbon monoxide (CO) and inhibit the release of sFlt-1 and sEng, thus lowering maternal blood pressure and increasing fetal/birthweight 5,11 (Figure 1).
Clinical trials investigating new drugs or the repurposing of existing drugs in pregnancy are challenging because pregnant women are either excluded from randomised clinical trials (RCT) or reluctant to participate. 12,13 Consequently, RCTs studying pregnant women are often underpowered, preventing firm conclusions. The therapeutic use of statins in PE or FGR has been examined in several animal studies and a few human trials with overall low numbers of participants. The objective of this study was therefore to evaluate the potential effects of statins on maternal blood pressure and birthweight in pregnancies complicated by PE or FGR using all data from human and animal studies in a metaanalysis across species.

| M ET HODS
The review protocol for this combined human and animal meta-analysis was preregistered on PROSPERO on 11 July 2020 (CRD42020194120), and reported according to the PRISMA 2020 guidelines and the MOOSE checklist. 14,15 No amendments to the protocol were made after initial publication.

| Search strategy and inclusion criteria
On 10 February 2021, we originally searched the PubMed and Embase.com databases using the search components 'statins' and 'pregnancy'. The search was updated on 10 May 2022 (search string in Appendix S1). The reference lists of retrieved publications were checked for additional studies meeting the inclusion criteria. Two researchers (MD and LM) independently screened all papers. Discrepancies were resolved by consensus decision.
Inclusion criteria were human RCTs or cohort studies with a control group available and controlled animal studies on statin use during pregnancy. Studies were considered eligible if statins were used in any animal model for PE or FGR, or in women with a previous or current pregnancy complicated by PE or FGR, as defined by study.

| Study selection
Published studies were included if they reported maternal blood pressure and/or birthweight. Studies were excluded in case of (1) combined interventions where it was not possible to deduct the effect of statin therapy alone, (2) use of known teratogenic drugs (i.e. co-medication), (3) intervention of only single dose or single-day treatment or (4) when the full text was in a language other than English or was irretrievable ( Figure S1).

| Data collection and analysis
Data extraction was performed independently by two researchers (MD and LM) and discrepancies were resolved by consensus decision. Data was extracted from included studies on study characteristics, including species, strain, type of animal model (PE/FGR) and maternal weight. For the statin treatment regimen, we extracted data on the route, duration and frequency of administration, type of statin, dose in mg/ kg/day, dose recalculated to atorvastatin in mg/kg/day and gestational days of administration. 16 We also extracted data on bibliographic details, study design and the number of experimental groups per study.
For the outcome maternal blood pressure, we extracted data on method of blood pressure measurement, blood pressure type and gestational day of measurement. Furthermore, we extracted data regarding maternal blood pressure ± standard deviation in mmHg for either mean arterial pressure or systolic blood pressure and data on the number of mothers. For the primary outcome birthweight, we extracted data on birthweight ± standard deviation in grams and the number of mothers.
When data was only presented in graphs, the corresponding authors were contacted via email requesting the raw data. If no response was received after two reminders, data were extracted from graphs using apps.autom eris.io. software.
If data on the pup weight was missing, the litter weight was divided by the litter size to calculate the average pup weight. As the standard deviation of the litter weight cannot be divided by litter size, the highest included standard deviation in this meta-analysis was imputed for the study.

| Quality evaluation
The risk of bias in each included study was assessed by two independent investigators (MD and LM) and discrepancies were resolved by consensus decision. For animal studies, the risk of bias was assessed using the SYRCLE Risk of Bias Tool. 17 The risk of bias in human studies was assessed using the Cochrane Risk of Bias Tool. 18

| Statistical methods
Meta-analysis was performed using R (version 4.0.3). When studies reported multiple comparisons between control and intervention groups, these were treated as independent comparisons in the meta-analysis. If a study compared multiple intervention groups with the same control group, the number of subjects in the control group was corrected for multiple comparisons by dividing by the number of comparisons made.
For both maternal blood pressure in mmHg and birthweight in grams, the standardised mean difference (SMD) with a 95% confidence interval was calculated to compare blood pressure presented as either mean arterial pressure or systolic blood pressure and birthweight across species. As heterogeneity was expected, a random effects model was used.
To investigate potential sources of heterogeneity, subgroup analyses on species, statin type, dose, timing and route of administration were performed if each subgroup included at least three independent studies. For the subgroup analysis on dose, studies were categorised based on doses either higher or lower than 1 mg/kg/day. Timing was translated to human trimesters and split into first, second or third trimester. Route of administration was divided into oral and intraperitoneal.
Administered statin dosage was recalculated to atorvastatin in mg/kg/day. 16 This meant that maternal body mass index (BMI) was recalculated to kilograms. When animal studies did not report maternal weight, average weight for the specific animal model of fertile age was extracted from the Jax databank. 19

| Study selection
The flow chart of the study selection process is depicted in Figure S1. The systematic literature search resulted in 3339 hits. After removing duplicates, 2531 references underwent three phases of screening; after application of the inclusion and exclusion criteria, 13 studies were included in this meta-analysis.

| Study characteristics
Study characteristics were extracted for the 13 included studies ( Table 1). Twelve of the included studies were animal studies and one was a human RCT. Four of the animal studies were rat studies, the other studies used mice. Animal models to mimic PE and/or FGR varied from DOCA salt to sFlt-1 adenoviruses to transduced placentas. All animal studies, apart from Bauer (2013) and Chimini (2018), started statin administration before what translates to the human second trimester. 1,[20][21][22] In the human trial, either statins or a placebo were prescribed to women diagnosed with PE between 24 +0 and 31 +6 weeks of gestation. 23

| Quality evaluation
A number of key indicators of study quality were not reported in the majority of the animal studies ( Figure S2A). Although all studies described randomisation at some experimental phase (typically at the allocation of animals to the experimental groups), none of the studies reported the method of randomisation used. Blinding at any experimental phase was reported in three (25%) studies. None of the studies accurately reported a sample size calculation.
The low reporting quality, especially for randomisation and blinding, led to the vast majority of risks of bias being assessed as unclear ( Figure S2B). Only four studies had similar baselines at time of selection, only two reported adequately on dropouts and only four were at low risk of selective outcome reporting.
The risk of bias for the single human study is displayed in Figure S3. There were some concerns about the risk of bias considering the selection of reported results. The study was powered on the primary outcome mean maternal serum sFlt-1, rather than clinical parameters.

| Synthesis of results
For the outcome birthweight, the highest standard deviation was imputed for three studies, because average pup weight was calculated by dividing litter weight by litter size. [24][25][26] Seven animal studies (n = 97 statin and n = 75 controls) reported on maternal blood pressure. Prenatal administration of statins significantly reduced maternal blood pressure during pregnancy with an overall SMD of −2.49 [95% CI −4.26 to −0.71], p = 0.01 ( Figure 2). Heterogeneity (I 2 80%; p < 0.01) between studies was significant.
We performed subgroup analyses for dose, species and route of administration for the outcome maternal blood pressure by meta-regression. The residual heterogeneity remained high: I 2 was 90% for dose, 89% for species and 89% for route. The R 2 for dose was 0%, for species 1.7% and for route 0%, meaning that the heterogeneity in our metaanalysis cannot be attributed to either dose, species or route. Lastly, the test of moderators' coefficients was 2.17 (p < 0.2), 1.2 (p < 0.3) and 2.1 (p < 0.2) for dose, species and route of administration, respectively, indicating that none of the subgroup variables significantly influenced the primary outcome maternal blood pressure.
Ten studies reported on birthweight: nine animal studies (n = 347, 52 statin and n = 355, one control) and the single human study (n = 30 pravastatin and n = 31 controls). As mean litter weight was divided by mean pup count in several animal studies, the total number of animals contains decimals. There was no significant overall effect of statin treatment on birthweight (SMD 0.69 [95% CI −0.65 to 2.03], p = 0.28; Figure 3). Significant heterogeneity was present between studies (I 2 91%, p < 0.01).

F I G U R E 2
Effect of statins on maternal blood pressure.

F I G U R E 3 Effect of statins on birthweight.
A subgroup analysis on statin dose demonstrated no significant difference in effect between a high and a low dose regimen. In the regression analysis on dose for the outcome birthweight, heterogeneity remained high (I 2 90%). R 2 was 4.26%, meaning that a negligible 4.26% of the heterogeneity in our meta-analysis could be attributed to dose. The test of moderators showed a coefficient of 0.02 (p < 0.9), from which we conclude that dose does not influence the SMD significantly.
Subgroup analyses on species, statin type, timing and route of administration could not be performed since the respective subgroups did not include more than three studies.

| Main findings
This meta-analysis across species demonstrated that statins reduce maternal blood pressure in animal models for PE but do not affect birthweight in a combined human and animal meta-analysis.

| Strengths and limitations
A cross-species meta-analysis provided insight into knowledge gaps that need further investigation and may prevent unnecessary (animal) research. The relevance of this methodology was demonstrated in prior studies investigating the use of Sildenafil and amino acid supplementation in pregnancy and the importance of this research method was discussed in a recent paper. [27][28][29] To our knowledge, we are the first to pool data of human as well as animal studies on statin use during pregnancy. This unique method of analysis allows the review of all existing data, which facilitated inclusion of a variety of studies using different statins, animal models, doses, timing and route of administration. The results of our meta-analysis across species regarding the outcome birthweight are in line with the results of the single human trial included in this meta-analysis in terms of the direction, magnitude and precision of the effects measured. 23 This increases our confidence that pooling data across species, including humans, is acceptable and informative, provided that expected heterogeneity is accounted for by using a random effects model and standardised mean difference.
The potential therapeutic effect of statins on maternal blood pressure is hypothesised to arise from the statin-induced inhibition of the sFlt-1 release. 5 SFlt-1 is thought to contribute causally to blood pressure elevation in pregnancy. 30 Maternal blood pressure was significantly lower in the statin intervention groups (SMD −2.49 [95% CI −4.26 to −0.71]). Although the SMD suggests a strong effect, 31 the clinical relevance of this finding is questionable at best. First, the confidence interval is broad and nearly includes zero, suggesting a marginally significant effect. Secondly, the hypothesised direction of the statin effect is not consistent throughout all studies. Lastly, several studies contributing to our pooled analysis stem from a single research group. None of the subgroup analyses (statin dose, species and route of administration) revealed a significant difference in maternal blood pressure.
As statins inhibit the pathogenic pathways contributing to placental insufficiency, the hypothesised effect of improved placental function would lead to improved fetal growth and therefore higher birthweight. 32 However, maternal statin administration during pregnancies complicated by PE and FGR left birthweight unchanged across species. The subgroup analysis that could be performed for birthweight was on dose. The regression analysis on dose showed that dose does not influence the effect of statins on fetal growth.
A limitation of our meta-analysis is the low number of studies, limiting the possibility for subgroup analyses to explore causes of heterogeneity. A potential cause of this heterogeneity is the different animal models used. Although these animal models are frequently used PE or FGR models, different animal models tie into different factors in the multifactorial pathogenesis of PE, making pooling of data difficult. However, spontaneously developing PE animal models do not exist. To translate rodent models of placental insufficiency to the human placental originated multi-factorial disease, integration of multiple (spontaneous) models can be of benefit for the field. 33 In pooling the data from the different animal models, we provide an overview of all existing data displaying the trends observed when treating PE and FGR with statins.
Besides the different type of animal models, the models also mimic different timing in human pregnancy, resulting in potential heterogeneity but also blurring timing of statins as either prophylaxis or treatment of PE and FGR. Rat gestation takes approximately 23 days, whereas mice are pregnant for approximately 20 days. 20,21 The human first trimester translates to gestational day 1-8 in rats and day 1-16 days and 15 hours in mice. 20,21,34 This means that apart from Bauer and Chimini, all rat and mice studies begin statin intervention before the second trimester. 1,21,22 In human studies, statins are also both investigated as preventive agents as well as for treatment purposes. [35][36][37][38][39][40][41] A third factor possibly contributing to the heterogeneity is the different statins used. All studies used pravastatin, apart from Dong, 42 who used simvastatin. Therefore, we expect the contribution of different statins to the observed heterogeneity to be small. Pravastatin has the most potential in forming a treatment for PE and FGR because of its favourable safety profile and no reports of congenital anomalies. 43,44 Recalculation to atorvastatin was done purely for analysis purposes.
The last factor potentially contributing to the observed heterogeneity is the large risk of bias present in the available studies ( Figures S2 and S3).

| Interpretation in the light of other evidence
Several studies regarding statins in human pregnancy did not meet our inclusion criteria because statin treatment was combined with another intervention. These studies illustrated how pravastatin prolonged pregnancy, reduced PE symptoms and led to higher birthweight, but left maternal sFlt-1and PIGF levels unchanged. 23,[35][36][37]45,46 Yet, women in these studies were also treated with either low-dose aspirin (LDA), low molecular weight heparin (LMWH), antihypertensive drugs or l-arginine. 23,[35][36][37]45,46 This made it impossible to isolate the effect of pravastatin, making these studies unsuited for our analysis. Two other recent studies illustrated no beneficial effects of statins. Döbert et al. demonstrated that the occurrence of PE was not reduced by pravastatin. 39 Another large retrospective cohort including 469 women exposed to statins showed that birthweight was lower for these women than for non-exposed age-matched controls. 47 As of November 2021, eight human trials regarding statins in pregnancy have been registered online. 38, 40 Six of these trials aim to investigate a similar research question as the current paper. The first trial registration dates from 2011. At present, 10 years after the first trial registration, only two research groups have published data, with the results of one trial expected in 2022. 38, 41 The combined prospected number of pregnancies included in these trials is 1025. In all, 775 of these pregnancies will be included in a single trial with an estimated study completion date set at June 2031. The expected low total number of pregnancies to be included in these trials, the slow rate of inclusion and the fact that the vast amount of data (i.e. 75%) is expected 10 years from today, emphasises the need for research methods other than traditional RCTs when investigating new drugs or a new indication of existing drugs in pregnancy. This underlines the importance and the added value of performing a crossspecies meta-analysis such as ours before the conduct of human trials. This proposed research method could also be performed to evaluate the therapeutic effects of any drug currently investigated as treatment for PE, e.g. MZe786, metformin, hydroxychloroquine, l-citrulline and antithrombin gamma. 6,48-51

| CONCLUSIONS
Our systematic review and meta-analysis combining human and animal data demonstrates that statins only induce lower maternal blood pressure in rodent pregnancies complicated by PE or FGR and have no effect on birthweight in either animals or humans. The marginally significant effect of statins on maternal blood pressure was accompanied by a broad confidence interval, an inconsistent direction of the observed effect, large risk of bias and high heterogeneity present in the few available studies.
Ideally, high quality studies using different types of animal models, and thus mimicking multiple aspects of PE and FGR, provide robust evidence that inform the appropriate design of human RCTs. 33 Currently, a series of human trials on statins is being conducted without such a base. 38 Although there is little evidence from our cross-species meta-analysis to suggest a positive effect of statins in hypertensive pregnancy, we await the completion of the human trials with interest.

AU T HOR C ON T R I BU T ION S
LM and MD collected all data and performed the metaanalysis. KEW checked the methodology and statistical analysis. LM and MD drafted the first paper. All authors contributed equally to the final article.

AC K NOW L E D GE M E N T S
None.

F U N DI NG I N FOR M AT ION
This research was funded in part by a grant from ZonMw (114024144).

C ON F L IC T OF I N T E R E S T S TAT E M E N T
None declared. Completed disclosure of interest forms are available to view online as supporting information.

DATA AVA I L A BI L I T Y S TAT E M E N T
Data will be shared upon request by contacting our corresponding author, M. Depmann (m.depmann-3@umcutrecht.nl).

E T H IC S A PPROVA L
Ethical approval was not required for this research.