Results
Literature Search
Results from our search strategy are summarized in Figure 1. We identified 3,993 publications. Of these, 898 were duplicates between the 2 databases, and an additional 2,823 were excluded on the basis of review of their titles and abstracts. The remaining 141 articles were eligible for abstraction. After reviewing the full texts, we excluded an additional 67 articles. The most common reason for exclusion after full-text review was the inability to obtain or extract the data necessary for analysis. Sixteen studies were excluded on the basis of their low quality scores (<5 points). Six studies using the same patient data were identified. To ensure the data were represented only once, we excluded 5 of the studies from the analysis. In addition, because we intended to stratify the analysis on the basis of trimester of triglyceride measurement, we also excluded 5 studies that did not report the gestational age at the time of blood sampling. After final exclusions, 74 original articles were included in our meta-analyses: 64 for total cholesterol, 60 for HDL-C, 54 for LDL-C, 70 for triglycerides, and 46 for non-HDL-C. Flow diagrams for the studies included in the subgroup analyses of preeclampsia severity are provided in Web Figures 1–4, available at http://aje.oxfordjournals.org/http://aje.oxfordjournals.org/content/180/4/346/suppl/DC1.
(Enlarge Image)
Figure 1.
Flow diagram of study selection for the meta-analysis of the association between lipid measurements during pregnancy and risk of preeclampsia (PE). "Main" refers to the main meta-analysis of lipid levels during pregnancy and risk of any PE. "Severe or mild" refers to the subanalysis of lipid levels during pregnancy and the risk of PE, stratified by PE severity. Overlap between the 2 meta-analyses is possible. Studies were published from the index date of the databases until July 2013. HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol.
Study Characteristics
The studies that met the eligibility criteria for inclusion in the main preeclampsia meta-analysis examined a total of 7,369 participants—1,975 women with preeclampsia and 5,394 healthy pregnant women. Studies that evaluated lipid levels by severity of preeclampsia (for severe preeclampsia, n = 19; for mild preeclampsia, n = 15) included 568 women with severe preeclampsia compared with 1,004 normotensive pregnant controls, as well as 427 women with mild preeclampsia compared with 667 normotensive healthy pregnant women. Characteristics of the studies included in the meta-analyses are shown in Table 1. The 73 included studies were conducted in Asia (52%), Europe (25%), and North America (10%) with sample sizes ranging from 20 to 1,000 pregnant women. The majority of the included studies measured serum lipid levels during the third trimester, and fasting measurements were obtained in 62% of the studies. Within each study, controls were commonly matched to women with preeclampsia on maternal age, gestational age at blood sampling, and/or maternal BMI. In preeclamptic women, mean lipid levels ranged from 162–345 mg/dL, 29–79 mg/dL, 116–300 mg/dL, 96–197 mg/dL, and 100–436 mg/dL for total cholesterol, HDL-C, non-HDL-C, LDL-C, and triglycerides, respectively. In normotensive women, values ranged from 111–317 mg/dL, 33–93 mg/dL, 90–243 mg/dL, 96–197 mg/dL, and 111–269 mg/dL for total cholesterol, HDL-C, non-HDL-C, LDL-C, and triglycerides, respectively. Characteristics of the studies included in the subanalyses of severe and mild preeclampsia can be found in Web Table 1 http://aje.oxfordjournals.org/content/180/4/346/suppl/DC1.
Meta-analyses
Results from the WMD meta-analyses of lipid measurements during pregnancy and preeclampsia are presented in Table 2. Total cholesterol levels measured in the first or second trimester were significantly higher in women who developed preeclampsia than in normotensive pregnant women (WMD = 12.49 mg/dL, 95% CI: 3.44, 21.54). Total cholesterol levels measured in the third trimester were also significantly higher in preeclamptic women compared with normotensive pregnant women (WMD = 20.20 mg/dL, 95% CI: 8.70, 31.70). Although no relationship with preeclampsia was observed for HDL-C levels measured in the first trimester, preeclamptic women had significantly lower HDL-C levels in the third trimester than normotensive women (WMD = −8.86 mg/dL, 95% CI: −11.50, −6.21). This relationship was also observed when stratifying by mild and severe preeclampsia. Non-HDL-C measured in both the first/second trimesters (WMD = 11.57, 95% CI: 3.47, 19.67) and third trimester (WMD = 29.59, 95% CI: 12.13, 47.06) was significantly higher among preeclamptic women than among normotensive pregnant women. LDL-C levels measured in the first/second trimesters (WMD = 3.89 mg/dL, 95% CI: −0.19, 7.97) and third trimester (WMD = 10.92, 95% CI: −0.59, 22.42) were greater among women who developed preeclampsia than among those who remained normotensive throughout pregnancy, though these relationships were only marginally significant (P = 0.06). Triglyceride levels measured in the first/second trimesters were significantly higher in women who developed preeclampsia than in normotensive pregnant women (WMD = 25.08 mg/dL, 95% CI: 14.39, 35.77). Triglyceride levels measured in the third trimester were also significantly higher in preeclamptic women when compared with normotensive pregnant women (WMD = 80.29 mg/dL, 95% CI: 51.45, 109.13). This relationship was also observed when stratifying by mild and severe preeclampsia. Forest plots for all meta-analyses can be found in Web Figures 5–21 http://aje.oxfordjournals.org/content/180/4/346/suppl/DC1.
We detected moderate to significant heterogeneity in nearly all of the meta-analyses we conducted (P ≤ 0.01) with I values ranging from 56% to 99% (Table 2), with the exception of the first/second-trimester total cholesterol and mild preeclampsia analysis (P = 0.45, I = 0%) and the first/second-trimester LDL-C analysis (P = 0.33, I = 13%). Table 3 shows the results from the univariate and multivariate meta-regression analyses designed to identify potential factors that could explain the high heterogeneity. BMI imbalance between the comparison groups is likely an important source of heterogeneity in the LDL-C analysis (P = 0.04, R = 15.2%) and triglyceride analysis (P = 0.08, R = 9.8%) such that, for each 1.0 unit increase in BMI WMD between groups, decreases in the WMD of 5.50 mg/dL and 11.02 mg/dL are expected, respectively. In multivariate meta-regression models with trimester of lipid measurement and fasting status at blood sampling, BMI remained significant as a potential source of heterogeneity for LDL-C but not for triglycerides. Additionally, trimester of lipid measurement was identified as another potential source of heterogeneity in meta-analyses of HDL-C and triglycerides. However, neither BMI imbalance between the comparison groups nor trimester of lipid measurement was detected as a possible source of heterogeneity in the total cholesterol analysis. Fasting status at the time of lipid measurement was not identified as a potential source of heterogeneity for any of the lipid analyses.
On the basis of Egger's test, publication bias was not present in most of the meta-analyses with P values ranging from 0.06 to 0.82 (Table 2), with the exception of the analysis of LDL-C from the third trimester and severe preeclampsia (P = 0.04). However, visual inspection of the funnel plots suggests that publication bias may be present for the total cholesterol, LDL-C, and HDL-C analyses of third-trimester measurements stratified by preeclampsia severity. Funnel plots for all analyses can be found in Web Figures 22–41 http://aje.oxfordjournals.org/content/180/4/346/suppl/DC1.