This study was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [15]. The relevant articles were identified through a comprehensive literature search conducted across multiple databases, including Medline, Embase, Scopus, Web of Science, and the Cochrane Library. The search process included articles published from the inception of the databases through October 4th, 2023. During the search process, specific terms such as “systemic lupus erythematosus”, “pregnancy” and “hydroxychloroquine” were utilized (Supplementary File 1). Two authors (AHL and YWY) worked independently on this step, while another author (XWL) resolved any discrepancies between their results. This systematic review has been registered on PROSPERO (CRD42023485388).

We included studies that met the following criteria: (1) case-control studies, cohort studies, or randomized controlled trials; (2) studies comparing pregnant patients with SLE who received HCQ treatment (HCQ+ groups) versus those without HCQ treatment (HCQ– groups); (3) studies reporting maternal and/or fetal outcomes in both groups; and (4) studies with a minimum sample size of 10 patients in each group.

We excluded studies that met the following criteria: (1) case reports, case series, reviews, or in vivo/vitro studies; (2) studies lacking a control group; (3) studies without available data on maternal and/or fetal outcomes; and (4) duplicate studies.

The following 10 maternal outcomes were measured: preeclampsia, high lupus activity, gestational hypertension, premature rupture of membrane, spontaneous abortion, oligohydramnios, thrombotic disease, gestational diabetes mellitus, hemolysis, elevated liver enzymes, and low platelet count syndrome (HELLP) syndrome, and immune thrombocytopenia. Additionally, we investigated the following 11 fetal outcomes: premature delivery, FGR, birth weight, miscarriage, stillbirth, congenital malformation, intrauterine distress, gestational age at delivery, five-minute Apgar score 7 (APGAR7), neonatal lupus and fetal APOs.

The activity of lupus is commonly assessed using the SLE Disease Activity Index (SLEDAI) [16]. According to the given definition, a pregnant patient with a SLEDAI score of $\ge$4 or experiencing a lupus flare during pregnancy was classified as having high lupus activity [16, 17, 18]. FGR was defined as an estimated fetal weight below or equal to the 10th percentile for gestational age [19]. Preeclampsia was defined as hypertension with proteinuria or other end-organ damage after 20 gestational weeks [20]. Fetal APOs included one or more of the following outcomes: miscarriage, stillbirth, neonatal death, premature delivery, FGR, or APGAR7 [21]. Specific diagnostic criteria, whether referring to international or local guidelines, were deemed acceptable for other APOs [14].

The data from eligible studies were independently extracted by two authors (AHL and YWY) and organized into a two-by-two table. Any discrepancies were resolved through discussions with another author (XWL). The information collected included the author’s name, region, period, design, pregnancy outcomes, sample size of HCQ+ and HCQ– groups, HCQ dose, mean maternal age and body mass index (BMI), proportion of renal involvement history, proportion of patients with antiphospholipid syndrome/antiphospholipid antibody (APS/aPL), and concurrent medications in each group.

The quality of the observational studies was assessed using the Newcastle-Ottawa Scale (NOS), and the quality of the randomized trials was evaluated using the Cochrane’s Quality Assessment Form (CQAF) [22, 23]. A star rating system of items in the NOS and CQAF was utilized to evaluate each study. An observational study was considered to be of high quality when it received a rating of at least 6 stars, and a randomized trial was considered to be of high quality when it received a rating of at least 6 stars [12, 24].

The certainty of evidence for each study outcome was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) handbook [25]. Based on the study design and the factors influencing certainty, the study outcomes were classified as having very low, low, moderate or high certainty.

Pooled odds ratios (ORs) and weighted mean differences (WMDs) with 95% confidence intervals (95% CIs) were calculated for selected maternal and fetal outcomes. Additionally, pooled relative risks (RRs) with 95% CIs were calculated in some of the meta-analyses after excluding the case-control studies. p values $\le$ 0.05 were considered statistically significant.

Heterogeneity was assessed using the Q-test and I² statistic, and p values $\le$ 0.05 in the Q-test were considered indicative of substantial heterogeneity. Furthermore, I² values 50% were defined as indicating low heterogeneity, while I² values $\ge$50% were classified as indicating moderate or high heterogeneity [26]. A fixed-effects model was employed for meta-analysis when I² values were less than 50%, and a random-effects model was utilized under other conditions.

To evaluate the presence of publication bias, this study employed visual funnel plots and Egger’s test [27]. The absence of publication bias was determined when the funnel plot exhibited a symmetrical distribution and/or Egger’s test yielded a p value greater than 0.10.

A sensitivity analysis (leave-one-out analysis) was conducted to assess the robustness of the outputs [28]. One study was excluded at each stage, and the results of all stages were compared to those obtained from the complete analysis. In order to reflect the effect of HCQ on APOs more effectively, subgroup analyses were conducted from three dimensions (different populations, HCQ doses and sample sizes).

All the statistical analyses were conducted using Stata/MP 18.0 software (StataCorp LLC, 4905 Lakeway Drive, College Station, TX, USA). Review Manager 5.4 (Cochrane Collaboration’s Information Management System, London, UK) and GraphPad Prism 9.5 (GraphPad Software LLC, Boston, MA, USA) were utilized for certain aspects of the meta-analysis.

In the initial search across five databases, a total of 1332 studies were identified based on the search strategies determined in the protocol. During the second phase, a total of 456 duplicate articles were subsequently excluded. After conducting a thorough screening of titles, abstracts, and full texts, 17 studies were ultimately selected for meta-analysis [17, 21, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43]. The flow chart is presented in Fig. 1.

Fig. 1.

The preferred reporting items for systematic reviews and meta-analyses (PRISMA) flow chart. PRISMA, preferred reporting items for systematic reviews and meta-analyses.

The design types and the selected maternal and fetal outcomes in each study are summarized in Table 1 (Ref. [17, 21, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43]). This meta-analysis included 17 studies, which consisted of 2 randomized trials, 11 cohort studies, and 4 case-control studies. These studies involved a total of 2238 pregnancies with SLE. Among these pregnancies, 1165 patients were in the HCQ+ groups and 1073 patients were in the HCQ– groups. The patients were enrolled from 12 regions worldwide, including the United States of America, the United Kingdom, Italy, Germany, China, and others during the period spanning from 1987 to 2020. The HCQ dose ranged from 100 to 400 mg/d, while some studies did not provide precise dose information. Based on the available data, there were no significant differences observed in most studies between the HCQ+ and HCQ– groups regarding mean maternal age (11/12) or BMI (4/4), proportion of renal involvement history (8/9), APS/aPL+ (6/7), or concurrent medications (8/13) (Table 2, Ref. [17, 21, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43]).

The NOS scores for cohort and case-control studies ranged from 7 to 8, while the CQAF scores for randomized trials were between 6 and 7 (Table 3, Ref. [17, 21, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43]). Consequently, all the studies included in the quality assessment were deemed to possess a high level of quality and thus were incorporated into the meta-analysis.

The meta-analysis revealed statistically significant differences in the incidence of preeclampsia (OR = 0.45, 95% CI: 0.32–0.65, p 0.0001), high lupus activity (OR = 0.62, 95% CI: 0.46–0.83, p = 0.001), and gestational hypertension (OR = 0.41, 95% CI: 0.18–0.91, p = 0.03) between the HCQ+ and HCQ– groups (Fig. 2). Additionally, no significant differences were observed in the risk of premature rupture of membrane (OR = 0.58, 95% CI: 0.30–1.12, p = 0.10), spontaneous abortion (OR = 1.60, 95% CI: 0.76–3.35, p = 0.21), oligohydramnios (OR = 0.46, 95% CI: 0.17–2.25, p = 0.13), thrombotic disease (OR = 1.19, 95% CI: 0.27–5.33, p = 0.82), gestational diabetes mellitus (OR = 0.52, 95% CI: 0.01–32.10, p = 0.76), HELLP syndrome (OR = 0.88, 95% CI: 0.19–3.96, p = 0.86) or immune thrombocytopenia (OR = 0.36, 95% CI: 0.08–1.67, p = 0.19) between the two groups (Fig. 2). The I² values exceeded 50% for 2 maternal outcomes (gestational hypertension and gestational diabetes mellitus). For the other maternal outcomes (8/10), I² values were less than 50% with p values $>$ 0.05 in the Q-test (Supplementary File 2).

Fig. 2.

Meta-analyses of adverse pregnancy outcomes in pregnant patients with SLE receiving HCQ vs. controls. *p 0.05. ^†Random-effects model. OR, odds ratio; WMD, weight mean difference; 95% CI, 95% confidence interval.

For further analysis, the RRs were calculated for adverse maternal outcomes with positive results, after excluding the case-control studies. Significant differences were observed for preeclampsia (RR = 0.51, 95% CI: 0.37–0.71, p 0.0001) and high lupus activity (RR = 0.77, 95% CI: 0.64–0.92, p = 0.005) between the HCQ+ and HCQ– groups (Supplementary Fig. 1A). Although the RRs for gestational hypertension (RR = 0.49, 95% CI: 0.23–1.07, p = 0.07) were not significantly different, the primary result (OR = 0.41, 95% CI: 0.18–0.91, p = 0.03) was considered to be closer to the true value (Supplementary Fig. 1B).

The meta-analysis revealed significant differences in the incidence of premature delivery (OR = 0.59, 95% CI: 0.46–0.75, p 0.0001) and birth weight (WMD = 0.21 kg, 95% CI: 0.09–0.33, p = 0.0008) between the HCQ+ and HCQ– groups (Supplementary Fig. 1C). Furthermore, there were no significant differences detected in the risk of FGR (OR = 0.68, 95% CI: 0.41–1.14, p = 0.14), miscarriage (OR = 1.45, 95% CI: 0.93–2.28, p = 0.10), stillbirth (OR = 1.05, 95% CI: 0.51–2.18, p = 0.89), congenital malformation (OR = 0.51, 95% CI: 0.18–1.50, p = 0.22), gestational age at delivery (WMD = 0.59 week, 95% CI: –0.81–1.99, p = 0.41), intrauterine distress (OR = 0.60, 95% CI: 0.33–1.09, p = 0.09), APGAR7 (OR = 0.46, 95% CI: 0.17–1.24, p = 0.13), neonatal lupus (OR = 1.22, 95% CI: 0.35–4.27, p = 0.75), or fetal APOs (OR = 0.58, 95% CI: 0.30–1.15, p = 0.12) between the two groups (Fig. 2). The I² values exceeded 50% for 3 fetal outcomes (FGR, gestational age at delivery and fetal APOs). For the other fetal outcomes (8/11), I² values were less than 50% with p values $>$ 0.05 in the Q-test (Supplementary File 2).

After excluding the case-control studies, significant differences in the incidence of premature delivery (RR = 0.71, 95% CI: 0.60–0.84, p 0.0001) were still observed between the HCQ+ and HCQ– groups (Supplementary Fig. 1A).

Visual inspection of the funnel plots did not reveal any significant asymmetry among all maternal or fetal outcomes (Supplementary File 3). Egger’s test was used for meta-analyses involving more than 10 studies and indicated no significant publication bias in the analyses of preeclampsia, premature delivery, or FGR (Fig. 3A,C,E) [44].

Fig. 3.

Funnel plots and sensitivity analyses of three adverse pregnancy outcomes. (A) Funnel plots of preeclampsia with Egger’s test (p = 0.396). (B) Sensitivity analysis (leave-one-out analysis) of preeclampsia. (C) Funnel plots of premature delivery with Egger’s test (p = 0.485). (D) Sensitivity analysis (leave-one-out analysis) of premature delivery. (E) Funnel plots of fetal growth restriction with Egger’s test (p = 0.605). (F) Sensitivity analysis (leave-one-out analysis) of fetal growth restriction. Red arrow: the study reported by Do et al. in 2019 [32].

For meta-analyses that included more than 10 studies, robustness was assessed through leave-one-out analysis. The analysis demonstrated the robustness of the findings regarding preeclampsia and premature delivery (Fig. 3B,D). However, the analysis of FGR revealed a significant instability (Fig. 3F).

Through the examination of funnel plots and full-texts, a potential problem was identified in the results of FGR in the study reported by Do et al. (Fig. 4A). [32]. The baseline characteristics revealed that the occurrence rates of most categories of previous SLE manifestations in the HCQ+ group were higher than those in the HCQ– group (12/15 categories). Therefore, the SLE diseases were considered to be more severe in the HCQ+ group to a certain extent. This probably concealed the true impacts of HCQ on FGR in pregnant patients with SLE.

Fig. 4.

Meta-analysis of fetal growth restriction after excluding one study. (A) Forest plot meta-analysis of fetal growth restriction in pregnant patients with SLE receiving HCQ vs. controls, using the random-effects model. (B) Forest plot meta-analysis of fetal growth restriction in pregnant patients with SLE receiving HCQ vs. controls using the fixed-effects model, after excluding the study reported by Do et al. in 2019 [32]. (C) Funnel plots of fetal growth restriction with Egger’s test (p = 0.962), after excluding Do et al’s study. (D) Sensitivity analysis (leave-one-out analysis) of fetal growth restriction, after excluding Do et al’s study. Red arrow: the study reported by Do et al. in 2019 [32].

After excluding the study by Do et al. [32], the meta-analysis revealed significant differences in the incidence of FGR (OR = 0.61, 95% CI: 0.43–0.85, p = 0.003) between the HCQ+ and HCQ– groups (Fig. 4B). The funnel plot and Egger’s test showed no apparent evidence of publication bias (p = 0.962, Fig. 4C). Additionally, the sensitivity analysis demonstrated significant stability in the analysis of FGR (Fig. 4D).

Due to the limited number of included studies, subgroup analyses were conducted for maternal and fetal outcomes that included more than 10 studies (Supplementary Figs. 45–62). The effects of HCQ on three APOs (preeclampsia, premature delivery, and FGR) were similar in Asian populations as well as in European and North American populations. For preeclampsia and premature delivery, there were no apparently different effects of HCQ shown between two dose groups (200–400 mg/d group, and group with no precise dose provided). Similarly, no significant differences in the effects of HCQ were observed between two sample-size groups (20–100 group and 100–200 group).

However, the subgroup meta-analysis of FGR revealed no significant effects of HCQ in the group with no precise HCQ dose provided (Supplementary Fig. 59), as well as in the sample-size group of 100–200 or $>$200 (Supplementary Fig. 61). Due to the limited number of studies included in this subgroup analysis, there was no sufficient evidence to analyze these results further more.

The final certainty analysis revealed that the results of preeclampsia, high lupus activity, premature delivery and birth weight yielded high-quality evidence. Moderate quality evidence was found for gestational hypertension, premature rupture of membranes, spontaneous abortion, miscarriage, stillbirth and intrauterine distress. After excluding the study by Do et al. [32], high quality evidence was also obtained for FGR. Other APOs received low or very low scores for certainty (Table 4).