Immunological disturbances are hypothesised to play a role in recurrent miscarriage (RM) and therefore intravenous immunoglubulins (IVIg) have been tested in RM patients.
The objectives were to investigate the benefits and harms of IVIg versus placebo, no intervention, or treatment as usual in women with RM.
Randomised trials investigating IVIg versus placebo, no intervention, or treatment as usual in women with RM.
Data Collection and Analysis
We undertook meta-analyses of aggregated data and individual patient data using a two-step approach, and we conducted bias domain assessments and trial sequential analyses to assess the risks of systematic and random errors.
We identified 11 randomised clinical trials. No significant difference in the frequency of no live birth was found when IVIg was compared with placebo or treatment as usual (RR 0.92, 95% CI 0.75–1.12, p = 0.42). Trial sequential analysis showed that the required information size of 1,008 participants was not obtained. IVIg compared with placebo seems to increase the risk of adverse events. Subgroup analysis suggests that women with RM after a birth (secondary RM) seemed most likely to obtain a potential beneficial effect of IVIg (RR for no live birth 0.77, 95%CI 0.58–1.02, p = 0.06), however, trial sequential analysis showed that insufficient information is presently accrued.
Citation: Egerup P, Lindschou J, Gluud C, Christiansen OB, ImmuReM IPD Study Group (2015) The Effects of Intravenous Immunoglobulins in Women with Recurrent Miscarriages: A Systematic Review of Randomised Trials with Meta-Analyses and Trial Sequential Analyses Including Individual Patient Data. PLoS ONE 10(10): e0141588. https://doi.org/10.1371/journal.pone.0141588
Editor: D William Cameron, University of Ottawa, CANADA
Received: June 30, 2015; Accepted: October 9, 2015; Published: October 30, 2015
Copyright: © 2015 Egerup et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: The minimal data set can be seen in supporting material, S5 File (Review Manager File).
Funding: The review is funded by The Danish Agency for Science, Technology and Innovation, journal number 0602-00506, and The Fertility Department, Rigshospitalet, Copenhagen University Hospital, Denmark.
Competing interests: One of the review team members, Ole B. Christiansen (OBC), was investigator in three of the randomised clinical trials on the topic and another, Pia Egerup (PE), was investigator in one of the randomised clinical trials on the topic. To avoid potential bias, OBC did not participate in the selecting of relevant trials or in the data analyses and PE did not participate when the data of the trial that she co-authored were analysed. The remaining review team members controlled PE’s work. The remaining review team members (JL and CG) declare that they have no competing interest. This does not alter the authors' adherence to PLOS ONE policies on sharing data and materials.
Recurrent miscarriage (RM) is generally defined as three or more miscarriages before gestational week 24 . Many clinicians define, however, RM as two or more miscarriages . Primary RM refers to a series of miscarriages without a previous live birth whereas secondary RM refers to a series of miscarriages subsequent to a previous birth. Some clinicians also use the term secondary RM if the miscarriages have been preceded by a live birth or stillbirth after gestational week 22 . RM affects 1% to 5% of all women trying to conceive . In only a minority can the condition be explained by parental chromosomal abnormalities, uterine malformations, infective causes, or endocrine or thrombophilic disturbances .
Immunological disturbances are hypothesised to play an important role in RM and therefore various types of immune-based interventions have been tested in RM patients, including intravenous immunoglobulins (IVIg) [3–5].
IVIg have several effects like suppression and neutralisation of autoantibodies, attenuation of natural killer cells, inhibition of complement binding, modification of cytokine production, and expansion of regulatory T lymphocytes [6, 7]. IVIg exhibit a documented effect in many disorders caused by immunological abnormalities . IVIg formulations are made by extracting the IgG fractions from plasma from normal blood donors and therefore there are potential risks of adverse events like allergy and transmission of infections (e.g., HIV, hepatitis). In general, IVIg are well tolerated, and the most frequent adverse reactions, which include headache, fever, and nausea, occur in less than 5% of patients .
The newest systematic review with meta-analysis  was updated in February 2014 and found overall no significant beneficial effect of IVIg versus placebo in improving the live birth proportion. However, this review only assessed one outcome (birth after 20 weeks of gestation) and a single subgroup analysis, it did not include a randomised placebo-controlled trial of IVIg in RM treatment published in 2010 , it did not assess the risk of random errors, and it did not adequately investigate the full effect of benefits or harms of IVIg.
Due to the recent publication of a new randomised trial  and the considerations above, we found it relevant to conduct an up-dated systematic review with meta-analysis and trial sequential analyses (TSA) of randomised trials of IVIg versus placebo, no intervention, or treatment as usual in women with RM. Furthermore, we included individual patient data (IPD) in the meta-analyses whenever possible. IPD allow data checking and re-analysis of data in a consistent way. It also allows categorisation of participants in order to conduct subgroup analyses not feasible using aggregate data. Subgroup analysis is clinically relevant since it allows a more personalised treatment [18–20].
This systematic review was conducted according to our published protocol . The protocol was registered within the International Prospective Register of Systematic Reviews (PROSPERO) as number CRD42014007112.
We searched the relevant published literature using the following databases: the Cochrane Central Register of Controlled Trials (Central December 2014), Medline (1950 to December 2014), Embase (1947 to December 2014), WHO International Clinical Trials Registry Platform (December 2014), and Ovid Medline In-Process and Other Non-Indexed Citations databases (December 2014). The following medical subject headings (MeSH) terms, keywords, and their combinations were used: immunoglobulins; intravenous; immunotherapy; foetal death, abortion; habitual abortion; spontaneous; foetal loss; miscarriage; recurrent abortion; recurrent miscarriage. Appropriate suffixes were used for each database. The Cochrane Collaboration strategy for identifying randomised trials using the relevant MeSH terms and keywords were used. Similar search strategies were used in Central, Embase and Ovid Medline In-Process and other non-indexed citation databases. Relevant abstracts from the annual meetings of American and European Societies of Reproductive Medicine and Human Reproduction were searched. The reference lists of the identified reports were manually searched for other relevant publications. The full search strategy can be seen in S2 and S3 Files.
No language restrictions were applied.
Selection of trials and data extraction
All randomised trials irrespective of publication date, publication type, publication language, and publication status investigating infusions with IVIg in relation to pregnancy in women with RM compared with placebo, no intervention, or treatment as usual were included for assessments of benefits and harms. For assessments of harms, quasi-randomised clinical studies and observational studies that we identified during our search for randomised clinical trials were included.
We used the instructions in The Cochrane Handbook for Systematic Reviews of Interventions  and The Cochrane Hepato-Biliary Group Module  in our evaluation of the methodology and hence bias risk of the included trials [24–29]. A full description on risk of bias assessment can be found in our published protocol .
Two authors (PE and JL) independently identified trials for inclusion, assessed risk of bias, and extracted data from all included trials. The first authors of the original articles were contacted by e-mail with the aim of collecting IPD. If the first e-mail was not answered, a new e-mail was sent out one month later. If the second e-mail was not answered, we tried to contact some of the co-authors of the trial. The original authors were invited to become part of the ImmuReM (Intravenous Immunoglobulins in Recurrent Miscarriage) IPD Study Group and thereby co-authors of the present article, provided they delivered IPD.
- The proportion of women not giving live birth, defined according to the trialists. It is usual practise in Cochrane to choose a negative outcome. This is because RR can be interpreted the same way if all outcomes are negative.
- The proportion of women experiencing a serious adverse event (SAE) defined as any adverse event that results in death, is life-threatening, requires hospitalisation or prolongation of existing hospitalisation, or results in persistent or significant disability or incapacity . SAEs were assessed as a composite of all the above events.
- The proportion of live-born infants experiencing SAEs defined as any adverse event that results in death, is life-threatening, requires hospitalisation in a neonatal care unit or prolongation of existing hospitalisation, results in persistent or significant disability or incapacity or is a congenital anomaly or birth defect . SAEs were assessed as a composite of all the above events.
- The proportion of women experiencing an adverse event (AE) defined as any undesirable medical event occurring to a participant during a clinical trial, which does not necessarily have a causal relationship with the intervention . AEs were assessed as a composite of all the above events.
- The women’s quality of life, as defined by the trialists.
The secondary outcomes for women with a live birth in the trial and their infants were:
- Proportion of women who gives live birth prematurely (<37 weeks).
- Sex of the infant.
- The proportion of infants with low birth weight, i.e., <2500 g.
- The proportion of infants experiencing AEs. AEs were assessed as a composite.
- The infants’ quality of life, as defined by the trialists.
We undertook meta-analyses according to the recommendations of The Cochrane Handbook for Systematic Reviews of Interventions . For binary outcomes we calculated a standard estimation of the risk ratio (RR) and its 95% confidence interval (CI). For all outcomes and according to availability, we combined IPD and published aggregate data into a pooled effect measure using a two stage method . First, IPD for a given trial were analysed to construct summary data, second the summary data were added to the meta-analysis along with published aggregated data from trials without IPD data .
We applied Trial Sequential Analysis (TSA) because cumulative meta-analyses are at risk of producing random errors as the result of sparse data or repetitive testing of accumulating data . TSA can provide a required information size, a threshold for a statistically significant treatment effect (benefit or harm), and a threshold for futility . The required information size indicates the number of patients needed to detect or reject a certain intervention effect. TSA makes it possible to test for statistical significance before the required information size has been reached. The trial sequential monitoring boundaries determine the statistical inference that one may draw regarding a cumulative meta-analysis that have not reached the required information size; if the cumulative Z-cure of the meta-analysis crosses a trial sequential boundary for benefit or harm before the required information size is reached, firm evidence may be established, and further trials may turn out to be superfluous. Futility boundaries are a set of thresholds that reflects the uncertainty of obtaining a negative finding in relation to the strength of the available evidence (e.g., the accrued number of patients). Below the threshold, the test statistic is so low that the likelihood of a significant effect being found if one reaches the required information size becomes negligible. On the other hand, if futility boundaries are not crossed, it is most probably necessary to continue doing trials to detect or reject a plausible intervention effect .
We performed TSA on the primary outcomes, in order to control the risks of type I and type II errors that occur in traditional meta-analyses due to sparse data and repetitive analyses of accumulating data [31–36]. In order to control these risks, we calculated the diversity-adjusted required information size and assessed the eventual breach of the cumulative Z-curve of the relevant trial sequential monitoring boundaries . The calculation was based upon the proportion with the outcome in the control group; a relative risk reduction of 20%; an alpha of 5%; a beta of 20%, and the diversity (D2) of the meta-analysis assessed by the TSA programme .
Results of the searches
Our predefined searches 01-12-2014 identified 856 references, including 256 duplicates. The PRISMA flow diagram illustrating the selection procedure is shown in Fig 1.
Of the 600 remaining individual references, 546 were excluded based on title and abstract. After reviewing full-text articles of the remaining 54 references, 35 were excluded. Reasons for exclusions were that publications were duplicates, clinical trials were not randomised, or the patients or interventions were not correct according to our inclusion criteria.
Table 1 shows the baseline characteristics of the 11 included trials.
Baseline characteristics of the included trials according to IVIg regimen, placebo participants and types of miscarriage.
Two trials enrolled only women with primary RM [11, 12], two trials enrolled only women with secondary RM [15, 17], and six trials enrolled women with primary or secondary RM [5, 9, 10, 13, 14, 39]. One trial did not report whether the women had primary or secondary RM . The experimental intervention was IVIg in all trials. Dosage and number of infusions varied between trials. Eight trials initiated the intervention when the pregnancy was detected by a positive pregnancy test or by ultrasound [5, 10–12, 14, 17, 38, 39] and three trials initiated the intervention before the women became pregnant [9, 13, 15].
Six trials used albumin as placebo [5, 9–12, 17] and three trials used normal saline as placebo [13–15]. Albumin is considered an adequate placebo for IVIg as IVIg and albumin cannot be distinguished if the vials are identical. Saline is not considered an adequate placebo as it can be distinguished visually from the IVIg. Two trials used treatment as usual with multivitamin and subcutaneous heparin plus low-dose aspirin, respectively, as control interventions [38, 39]. These two control interventions could easily be distinguished from IVIg as they are applied differently.
IPD were obtained from four trials, including Jablonowska et al. , Christiansen et al.  Christiansen et al. , and Christiansen et al.. From the trials by Perino et al.  and Stephenson et al.  some of the IPD could be extracted directly from the publications.
Risk of bias
Table 2 illustrates risk of bias in all the included trials.
Risk of bias in the included trials according to a series of domains. In relation to these assessment of these domains the trials were classified as high or low overall risk of bias.
Only one trial  was classified as ‘low risk of bias’ and ten trials were classified as ‘high risk of bias’. Of the trials with ‘high risk of bias’, four trials were classified as ‘low risk of bias’ according to selected domains: allocation sequence, allocation concealment, and blinding of participants and treatment providers [5, 10–12]. Such trials can, however, be biased by other domains.
All meta-analyses were performed both with a fixed-effect and a random-effects model. The following results are based on the random-effects model; however, the fixed-effect model provided similar results.
The proportion of women not giving live birth
Fig 2A shows the meta-analysis for the outcome ‘no live birth’.
A) Forest Plot of the meta-analysis for the outcome no live birth, B) The diversity-adjusted required information size (DARIS) of 1,008 patients was calculated on the basis of type I error of 5%, type II error of 20%, the control group event proportion (PC) of 43%, a relative risk reduction (RRR) of 20%, and the diversity (D2) 0% of the meta-analysis. The cumulative Z-curve does not cross the trial sequential monitoring boundaries for benefits, harms, or futility, and the required information size was not reached.
All 11 trials with a total of 531 women reported adequately on ´no live birth`. Overall, no significant difference in the number of ‘no live birth’ was found when IVIg was compared with placebo or treatment as usual (107/265 (40%) versus 113/266 (42%); RR: 0.92, 95% CI 0.75–1.12, p = 0.42). Funnel plot of the included 11 trials was symmetrical (not included). Fig 2B shows the TSA.
The cumulative Z-curve did not cross any of the trial sequential monitoring boundaries for benefit, harm, or futility and did not reach the required information size of 1,008 participants. The Z-curve was, however, very close to the futility boundary.
The proportion of women experiencing serious adverse events (SAEs)
Eight trials reported on SAEs [5, 10–12, 14, 17, 38, 39]. The meta-analysis of the eight trials with a total of 381 women found overall no significant difference in the number of SAEs when IVIg was compared with placebo or treatment as usual (29/193 (15%) versus 27/188 (14%); RR: 1.06, 95% CI 0.67–1.67, p = 0.81). The description of SAEs included caesarean sections, premature rupture of membranes prior to gestational week 28, and hospitalisation due to infusion-related symptoms. None of the trials reported participants with death, HIV, or hepatitis. TSA showed that the cumulative Z-curve did not cross any of the monitoring boundaries for benefits, harms, or futility and the required information size of 4,410 participants was not accrued (data not shown).
The proportion of live-born infants experiencing SAEs
Nine trials reported on infants experiencing SAEs [5, 9–12, 14, 15, 17, 38]. The meta-analysis of the nine trials with a total of 286 infants found overall no significant difference in the number of SAEs when IVIg was compared with placebo or treatment as usual (20/146 (14%) versus 17/140 (12%); RR: 1.18, 95% CI 0.58–2.42, p = 0.65). The description of SAEs included admission to neonatal care unit, congenital malformations, mental retardation, and death. TSA showed that the cumulative Z-curve did not cross any of the monitoring boundaries for benefits, harms, or futility, and the required information size of 5,251 participants was not obtained (data not shown).
The proportion of women experiencing adverse events (AEs)
Nine trials with a total of 451 women reported on AEs [5, 10–15, 17, 38]. The meta-analyses indicate a higher frequency of AEs after IVIg treatment compared with placebo or treatment as usual (68/229 (30%) versus 42/222 (19%); random-effects model: RR: 1.55, 95% CI 0.89–2.69, p = 0.12, fixed-effects model: RR: 1.54, 95% CI 1.13–2.11, p = 0.006). The reporting on AEs varied widely between trials. The description of AEs included vaginal bleeding, rash, headache, fever, and itching.
Proportion of women who give live birth prematurely (<37 gestational weeks)
Ten trials with a total of 284 women found no significant difference in the number of premature live births when IVIg was compared with placebo or treatment as usual (12/141 (9%) versus 15/143 (10%); RR: 0.80, 95% CI 0.37–1.71, p = 0.56) [5, 10–15, 17, 38, 39].
Sex of the infant
The proportion of infants with low birth weight
Seven trials reported on low weight infants (<2500 g) [5, 10–14, 17]. The meta-analysis of the seven trials with a total of 238 infants found overall no significant difference in the number of low weight infants when IVIg was compared with placebo (10/120 (8%) versus 10/118 (8%); RR: 1.02, 95% CI 0.43–2.45, p = 0.96).
The proportion of infants experiencing AEs
A total of four participants were lost to follow up. We performed ‘best-worst-case’ scenario and ‘worst-best-case’ scenario analyses , which did not change our results. We also performed a per-protocol analysis, which did not change our results.
As the trials were heterogeneous with regard to control intervention and initiation of the intervention, we further performed post-hoc sensitivity analyses where we excluded the trials that used treatment as usual as control intervention [38, 39], and where we excluded the trials that initiated IVIg intervention before pregnancy was obtained [9, 13, 15]. These sensitivity analyses did not change our results.
An overview of the data used in the subgroup analyses can be found in Table 3.
Data sources for the subgroup analyses in relation to aggregate published data, individual patient data (IPD) from a trial where the randomisation was stratified according to this factor and IPD from a trial where the randomisation was not stratified according to this factor.
Primary compared to secondary RM.
The subgroup analysis for women with primary RM compared to secondary RM is illustrated in Fig 3A.
A) Forest Plot for the outcome no live birth, B) The diversity-adjusted required information size (DARIS) of 698 patients was calculated on the basis of type I error of 5%, type II error of 20%, the control group event proportion (PC) of 53%, a relative risk reduction (RRR) of 20%, and the diversity (D2) 0% of the meta-analysis. The cumulative Z-curve does not cross trial sequential monitoring boundaries for benefits, harms, or futility, and the required information size was not reached.
A statistically significantly test of interaction for outcome ´no live birth`(p = 0.03) was found, indicating that IVIg may have different effects in women with primary RM (35/91 (38%) versus 25/90 (28%); RR: 1.32, 95% CI 0.88–1.98, p = 0.18) compared to women with secondary RM (46/111 (41%) versus 58/110 (53%); RR: 0.77, 95% CI 0.58–1.02, p = 0.06). Fig 3B shows TSA of IVIg compared with placebo or treatment as usual in participants with secondary RM. The cumulative Z-curve did not cross any of the monitoring boundaries for benefits, harms, or futility, and the required information size of 698 was not reached.
Low compared with high IVIg dosage.
The planned total median dosage of IVIg provided from before conception to gestational week 10 in all trials was calculated to be approximately 84 g. Test for subgroup differences for trials providing dosages of IVIg at or over this median dose (RR for ´no live birth`: 0.85, 95% CI 0.64–1.12 p = 0.25) compared to trials providing IVIg doses below the median dose (RR for ´no live birth`: 1.19, 95% CI 0.81–1.75, p = 0.38) was not statistically significantly different (test of interaction, p = 0.17). We post-hoc conducted a meta-analysis in order to evaluate if there was a relation between IVIg dosage and the proportion of women experiencing AEs. The meta-analyses indicate a higher frequency of AEs after IVIg treatment at or over the median dosage (RR 2.33, 95% CI 1.00–5.46, p = 0.05) compared to IVIg treatment below the median dosage (RR 0.85, 95% CI 0.46–1.58, p = 0.61), test for subgroup differences p = 0.06 (random effect model) and p = 0.02 (fixed effect model).
Other subgroup analyses.
Tests for subgroup difference for trials with low risk of bias regarding all domains or regarding the selected domains: allocation sequence, allocation concealment, and blinding of participants and treatment providers compared to trials with ‘high risk of bias’ (p = 0.75); for patients without lupus anticoagulant and/or IgG anticardiolipin compared to patients with lupus anticoagulant and/or IgG anticardiolipin (p = 0.60); for participants with two miscarriages compared to participants with three miscarriages and to participants with four or more miscarriages (p = 0.90) were not statistically significantly different for the outcome ‘no live birth’.
We identified 11 trials accessing the effect of IVIg in women with RM. Only one trial was classified as ‘low risk of bias’ for all domains, and 10 trials as ‘high risk of bias’. We identified four out of these 10 trials that could be considered trials with ‘low risk of bias’ for selected domains according to our protocol . Such trials can of course be biased due to other domains. We were able to obtain IPD from only six trials. Overall, treatment with IVIg did not reduce the risk for the outcome ‘no live birth’ in women with RM when compared with placebo or treatment as usual. TSA on the outcome ‘no live birth’ showed that the cumulative Z-curve did not cross the monitoring boundaries for benefits, harms, or futility and the required information size of 1,008 participants was not reached. Accordingly, we cannot interpret absence of evidence as evidence of absence of intervention effect. The Z-curve was, however, very close to the futility boundary. When the futility boundary is crossed, the test statistic is so low that the likelihood of a statistically significant intervention effect being found if and when the required information size is reached becomes equivalent to the risk of type II error . When the Z-curve is so close to the futility boundary, it is possible that a new trial will cross the futility boundaries. However, we may be quite far from the futility boundaries based on a smaller intervention effect than a relative risk reduction of 20% we used.
The meta-analysis of the proportion of women with AE indicates a greater frequency of AE after IVIg treatment compared with placebo or treatment as usual. The children, however, did not seem to suffer from the intervention, although evidence was sparse.
Test for subgroup differences showed that there was a statistically significant difference in the effect of IVIg in women with primary RM compared to women with secondary RM for the outcome ‘no live birth’. This may indicate that a potential beneficial effect of IVIg may be limited to women with secondary RM. However, TSA of the trials only including women with secondary RM showed that the cumulative Z-curve did not cross any of the trial sequential monitoring boundaries for benefits, harms, or futility, and the required information size of 698 participants was not obtained (so far only 221 women with secondary RM have been randomised), meaning that more trials are needed before any conclusions can be made.
The theoretical explanation for the observed difference in the effect of IVIg in women with primary compared to women with secondary RM may be that immunological disturbances seem to play a larger role in women with secondary RM. It has been reported that there is a higher prevalence of the immunological high responder HLA allele HLA-DRB1*03 in women with secondary RM compared to controls , that maternal carriage of other HLA class II alleles associated with immunity against male-specific minor histocompatibility (HY-) antigens results in a reduced live birth proportion in secondary RM patients with a firstborn boy , and that plasma levels of the proinflammatory and potentially harmful cytokine TNF-α are increased in patients with secondary RM relative to those with primary RM .
Strengths and limitations
Our systematic review has a series of strengths. We conducted the review according to the recommendations stated in The Cochrane Handbook for Systematic Review of Interventions . Furthermore, we followed our published protocol  with predefined participants, interventions, outcomes, and comparisons in order to avoid biases in the review process. We performed an extensive literature search to identify relevant studies based on our predefined inclusion criteria.
We contacted all original authors in order to obtain IPD but were only successful in getting data from four trials and in addition some IPD data could be extracted from the publication of two other trials. To avoid bias, we supplemented the available IPD with aggregate data from those trials, were IPD were not available, which makes the analyses more relevant. We placed participants into subgroups within the trials using IPD. In three instances (women with primary RM compared to women with secondary RM [5, 10]; women without lupus anticoagulant and/or IgG anticardiolipin compared to women with lupus anticoagulant and/or IgG anticardiolipin ; and women with two miscarriages compared to women with three miscarriages and to women with four or more miscarriages [10, 12–14], the original trials had not stratified the randomisation according to the subgroup variable. This means that data will likely resemble more observational data, as the balance obtained by randomisation no longer can be guaranteed. Results from these subgroup analyses must therefore be interpreted with caution.
We conducted meta-analyses based on intention-to-treat principles, whereby all participants randomised in each trial were included in the analyses. It also has to be noted that we due to IPD were able to analyse almost all the predefined outcomes except quality of life.
We did not plan to assess our review with the Grading of Recommendations Assessment, Development and Evaluation system (GRADE) . With GRADE we would have downgraded the evidence to ‘low quality’ or ‘very low quality’ according to the risk of bias, risk of imprecision, and risk of indirectness. We found low heterogeneity and no evidence of publication bias. Our review’s findings and interpretations suffer from limitations ensuring from the low quality of the randomised trials included in our review. After risk of bias assessment, all trials except one were considered with high risk of bias. To try to avoid potential bias from the trials with ‘high risk of bias’, we performed a predefined subgroup analysis where trials with ‘low risk of bias’ (that is low risk of bias in only selected domains) were compared to trials with ‘high risk of bias’. We were unable to find a difference in the effects of IVIg in relation to the trials’ risk of bias. Our analyses for risks of random errors showed lack of sufficient information.
Three prior meta-analyses on the topic were carried out by Wong et al. , Hutton et al. , and Ata et al. . Hutton et al. found a significant treatment effect of IVIg in women with secondary RM whereas Wong et al. and Ata et al. concluded that there was no significant treatment effect of IVIg in women with RM.
In the analysis by Ata et al. two trials by Christiansen et al. (1995, 2002) comprising 92 patients were completely excluded, the main argumentation being that the trials included patients with antiphospholipid antibodies. This argument was rejected as none of the participants in the Christiansen et al. 1995 trial were antiphospholipid positive and in the Christiansen et al. 2002 trial only 5% were positive for these antibodies .
The Cochrane systematic review with meta-analysis by Wong et al. was updated in February 2014 . No new trials of IVIg were included compared with the update in 2006 although a relevant new trial was published already in 2010 , a trial which was, however, included in the Ata et al. meta-analysis. The Cochrane review only includes one outcome (live birth after 20 weeks´ of gestation in all patients) and a single subgroup analysis (trials reporting intention-to—treat analysis).
Our systematic review included more trials than the existing meta-analyses and more patients thus decreasing possible bias in the available evidence. Two of the reviews had according to our knowledge no published protocols prior to the conduct of the reviews [49, 50]. In order to avoid bias in the review process and to ensure transparency we prepared a detailed and comprehensive protocol which we made publically available .
In addition, we conducted TSA, which has not been conducted in any of the previous meta-analyses [16, 49, 50]. TSA deals with the risk of random error and is important to assess how far we are from having reached the required information size in order to conclude the true effects of a specific intervention. Conclusions made using TSA show the potential to be more reliable than those using traditional meta-analysis techniques .
Moreover, we tried to obtain IPD. IPD meta-analyses use a much richer dataset and thus have greater statistical power than conventional meta-analysis. It is also an advantageous methodological approach when subgroup analyses are needed. Subgroup analysis is very clinically relevant since it allows a more personalised treatment.
Based on our results, we have insufficient evidence to recommend or refute IVIg for women with RM. Treatment with IVIg compared with placebo seems to increase the risk of adverse events. Subgroup analysis suggests that women with secondary RM seemed most likely to obtain a potential beneficial effect of IVIg, however, trial sequential analysis showed that insufficient information is presently accrued. IVIg should therefore only be administered in randomised clinical trials, clearly explaining the potential risks to participants. According to our subgroup analyses, women with secondary RM are the group, which most likely should be included in such trials.
S1 File. Characteristics of excluded studies.
We would like to thank Sarah Louise Klingenberg from The Cochrane Hepato-Biliary Group Editorial Team Office for her support in the searches and Xia Yun, Science of Education, Beijing University of Chinese Medicine Subsidiary Dongfang Hospital, Beijing, China, for conducting the Chinese translation into English.
ImmuReM IPD Study Group
ImmuReM IPD Study Group consists of:
Pia Egerup, The Fertility Department, Rigshospitalet, Copenhagen University Hospital, Denmark and Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen University Hospital, Denmark.
Lead author of the ImmuReM IPD Study Group. Contact: firstname.lastname@example.org
Barbara Jablonowska, Department of Obstetrics and Gynaecology, University Hospital Linköping, Sweden.
Jane Lindschou, Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen University Hospital, Denmark.
Christian Gluud, Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen University Hospital, Denmark.
Ole Bjarne Christiansen, The Fertility Department, Rigshospitalet, Copenhagen University Hospital, Denmark.
Conceived and designed the experiments: PE JL OBC CG. Performed the experiments: PE JL. Analyzed the data: PE JL OBC CG. Contributed reagents/materials/analysis tools: PE JL ImmuReM IPD Study Group. Wrote the paper: PE. Revised the manuscript critically: JL OBC CG. Read and approved the final manuscript: PE JL CG OBC and ImmuReM IPD Study Group.
- 1. Rai R, Regan L. Recurrent miscarriage. Lancet. 2006 Aug 12;368(9535):601–11. pmid:16905025
- 2. Christiansen OB. Reproductive immunology. Mol Immunol. 2013;55(1):8–15. pmid:23062611
- 3. Laskin CA, Bombardier C, Hannah ME, Mandel FP, Ritchie JW, Farewell V, et al. Prednisone and aspirin in women with autoantibodies and unexplained recurrent fetal loss. N Engl J Med. 1997;337(3):148–53. pmid:9219700
- 4. Porter TF, LaCoursiere Y, Scott JR. Immunotherapy for recurrent miscarriage. Cochrane Database Syst Rev. 2006(2):CD000112. pmid:16625529
- 5. Christiansen OB, Pedersen B, Rosgaard A, Husth M. A randomized, double-blind, placebo-controlled trial of intravenous immunoglobulin in the prevention of recurrent miscarriage: evidence for a therapeutic effect in women with secondary recurrent miscarriage. Hum Reprod. 2002;17(3):809–16. pmid:11870141
- 6. Schwab I, Nimmerjahn F. Intravenous immunoglobulin therapy: how does IgG modulate the immune system? Nat Rev Immunol. 2013;13(3):176–89. pmid:23411799
- 7. Trinath J, Hegde P, Sharma M, Maddur MS, Rabin M, Vallat JM, et al. Intravenous immunoglobulin expands regulatory T cells via induction of cyclooxygenase-2-dependent prostaglandin E2 in human dendritic cells. Blood. 2013;122(8):1419–27. pmid:23847198
- 8. Negi VS, Elluru S, Siberil S, Graff-Dubois S, Mouthon L, Kazatchkine MD, et al. Intravenous immunoglobulin: an update on the clinical use and mechanisms of action. J Clin Immunol. 2007;27(3):233–45. pmid:17351760
- 9. Coulam CB, Krysa L, Stern JJ, Bustillo M. Intravenous immunoglobulin for treatment of recurrent pregnancy loss. Am J Reprod Immunol. 1995;34(6):333–7. pmid:8607936
- 10. Christiansen OB, Mathiesen O, Husth M, Rasmussen KL, Ingerslev HJ, Lauritsen JG, et al. Placebo-controlled trial of treatment of unexplained secondary recurrent spontaneous abortions and recurrent late spontaneous abortions with i.v. immunoglobulin. Hum Reprod. 1995;10(10):2690–5. pmid:8567794
- 11. The German RSA/IVIG Group. Intravenous immunoglobulin in the prevention of recurrent miscarriage. Br J Obstet Gynaecol. 1994;101(12):1072–7. pmid:7826961
- 12. Perino A, Vassiliadis A, Vucetich A, Colacurci N, Menato G, Cignitti M, et al. Short-term therapy for recurrent abortion using intravenous immunoglobulins: results of a double-blind placebo-controlled Italian study. Hum Reprod. 1997;12(11):2388–92. pmid:9436670
- 13. Stephenson MD, Dreher K, Houlihan E, Wu V. Prevention of unexplained recurrent spontaneous abortion using intravenous immunoglobulin: a prospective, randomized, double-blinded, placebo-controlled trial. Am J Reprod Immunol. 1998;39(2):82–8. pmid:9506206
- 14. Jablonowska B, Selbing A, Palfi M, Ernerudh J, Kjellberg S, Lindton B. Prevention of recurrent spontaneous abortion by intravenous immunoglobulin: a double-blind placebo-controlled study. Hum Reprod. 1999;14(3):838–41. pmid:10221723
- 15. Stephenson MD, Kutteh WH, Purkiss S, Librach C, Schultz P, Houlihan E, et al. Intravenous immunoglobulin and idiopathic secondary recurrent miscarriage: a multicentered randomized placebo-controlled trial. Hum Reprod. 2010;25(9):2203–9. pmid:20634190
- 16. Wong LF, Porter TF, Scott JR. Immunotherapy for recurrent miscarriage. Cochrane Database Syst Rev. 2014;10:CD000112. pmid:25331518
- 17. Christiansen OB, Larsen EC, Egerup P, Lunoee L, Egestad L, Nielsen HS. Intravenous immunoglobulin treatment for secondary recurrent miscarriage: a randomised, double-blind, placebo-controlled trial. BJOG 2014;
- 18. Sud S, Douketis J. ACP Journal Club. The devil is in the details…or not? A primer on individual patient data meta-analysis. Ann Intern Med 2009; 151(2):JC1–JC3.
- 19. Clarke MJ. Individual patient data meta-analyses. Best Pract Res Clin Obstet Gynaecol 2005; 19(1):47–55. pmid:15749065
- 20. Higgins JPT, Green S (eds). Cochrane Handbook for Sytematic Reviews of Interventions. Version 5.1.0. [updated March 2011], chapter 18. The Cochrane Collaboration. 2011. Available from www.cochrane-handbook.org.
- 21. Egerup P, Lindschou J, Gluud C, Christiansen OB. The effects of immunotherapy with intravenous immunoglobulins versus no intervention, placebo, or usual care in patients with recurrent miscarriages: a protocol for a systematic review with meta-analyses, trial sequential analyses, and individual patient data meta-analyses of randomised clinical trials. Systematic reviews. 2014;3(1):89.
- 22. Higgins JPT, Green S (eds). Cochrane Handbook for Sytematic Reviews of Interventions. Version 5.1.0. [updated March 2011], The Cochrane Collaboration. 2011. Available from www.cochrane-handbook.org.
- 23. Gluud C, Nikolova D, Klingenberg S, Alexakis N, Als-Nielsen B, Colli A, et al. About The Cochrane Collaboration (Cochrane Review Groups (CRGs)). 2011(10).
- 24. Gluud C, Nikolova D, Klingenberg S, Alexakis N, Als-Nielsen B, Colli A, et al. About The Cochrane Collaboration (Cochrane Review Groups (CRGs)). 2014. Issue 7. Art. No.: LIVER.
- 25. Kjaergard LL, Villumsen J, Gluud C. Reported methodologic quality and discrepancies between large and small randomized trials in meta-analyses. Ann Intern Med. 2001;135(11):982–9. pmid:11730399
- 26. Lundh A, Sismondo S, Lexchin J, Busuioc OA, Bero L. Industry sponsorship and research outcome. Cochrane Database Syst Rev. 2012;12:MR000033. pmid:23235689
- 27. Savovic J, Jones H, Altman D, Harris R, Juni P, Pildal J, et al. Influence of reported study design characteristics on intervention effect estimates from randomised controlled trials: combined analysis of meta-epidemiological studies. Health Technol Assess. 2012;16(35):1–82. pmid:22989478
- 28. Savovic J, Jones HE, Altman DG, Harris RJ, Juni P, Pildal J, et al. Influence of reported study design characteristics on intervention effect estimates from randomized, controlled trials. Ann Intern Med. 2012;157(6):429–38. pmid:22945832
- 29. Wood L, Egger M, Gluud LL, Schulz KF, Juni P, Altman DG, et al. Empirical evidence of bias in treatment effect estimates in controlled trials with different interventions and outcomes: meta-epidemiological study. BMJ. 2008 Mar 15;336(7644):601–5. pmid:18316340
- 30. ICH-Guideline for Good Clinical Practice. 2002. [http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2009/09/WC500002874.pdf]
- 31. Thorlund K, Engstrøm J, Wetterslev J, Brok J, Imberger G, Gluud C. User manual for trial sequential analysis (TSA). Copenhagen Trial Unit, Centre for Clinical Intervention Research, Copenhagen, Denmark. 2011. p. 1–115. Available from www.ctu.dk/tsa
- 32. Wetterslev J, Thorlund K, Brok J, Gluud C. Trial sequential analysis may establish when firm evidence is reached in cumulative meta-analysis. J Clin Epidemiol. 2008;61(1):64–75. pmid:18083463
- 33. Brok J, Thorlund K, Gluud C, Wetterslev J. Trial sequential analysis reveals insufficient information size and potentially false positive results in many meta-analyses. J Clin Epidemiol. 2008;61(8):763–9. pmid:18411040
- 34. Thorlund K, Devereaux PJ, Wetterslev J, Guyatt G, Ioannidis JP, Thabane L, et al. Can trial sequential monitoring boundaries reduce spurious inferences from meta-analyses? Int J Epidemiol. 2009;38(1):276–86. pmid:18824467
- 35. Brok J, Thorlund K, Wetterslev J, Gluud C. Apparently conclusive meta-analyses may be inconclusive—Trial sequential analysis adjustment of random error risk due to repetitive testing of accumulating data in apparently conclusive neonatal meta-analyses. Int J Epidemiol.2009;38(1):287–98. pmid:18824466
- 36. Wetterslev J, Thorlund K, Brok J, Gluud C. Estimating required information size by quantifying diversity in random-effects model meta-analyses. BMC Med Res Methodol. 2009;9:86. pmid:20042080
- 37. Thorlund K, Engstrøm J, Wetterslev J, Brok J, Imberger G, Gluud C. User manual for trial sequential analysis (TSA), pp. 58–104. Copenhagen Trial Unit, Centre for Clinical Intervention Research, Copenhagen, Denmark. 2011. p. 1–115. Available from www.ctu.dk/tsa
- 38. Triolo G, Ferrante A, Ciccia F, Accardo-Palumbo A, Perino A, Castelli A, et al. Randomized study of subcutaneous low molecular weight heparin plus aspirin versus intravenous immunoglobulin in the treatment of recurrent fetal loss associated with antiphospholipid antibodies. Arthritis Rheum. 2003;48(3):728–31. pmid:12632426
- 39. Mahmoud F, Diejomaoh M, Omu A, Abul H, Haines D. Effect of IgG therapy on lymphocyte subpopulations in the peripheral blood of Kuwaiti women experiencing recurrent pregnancy loss. Gynecol Obstet Invest. 2004;58(2):77–83. pmid:15114031
- 40. Stricker RB, Steinleitner A, Bookoff CN, Weckstein LN, Winger EE. Successful treatment of immunologic abortion with low-dose intravenous immunoglobulin. Fertil Steril. 2000;73(3):536–40. pmid:10689009
- 41. Christiansen OB, Mathiesen O, Lauritsen JG, Grunnet N. Intravenous immunoglobulin treatment of women with multiple miscarriages. Hum Reprod. 1992 May;7(5):718–22. pmid:1639992
- 42. Stricker RB, Winger EE. Update on treatment of immunologic abortion with low-dose intravenous immunoglobulin. Am J Reprod Immunol. 2005 Dec;54(6):390–6. pmid:16305665
- 43. Carp HJ, Toder V, Gazit E, Ahiron R, Torchinski A, Mashiach S, et al. Further experience with intravenous immunoglobulin in women with recurrent miscarriage and a poor prognosis. Am J Reprod Immunol. 2001 Oct;46(4):268–73. pmid:11642675
- 44. Riley RD, Simmonds MC, Look MP. Evidence synthesis combining individual patient data and aggregate data: a systematic review identified current practice and possible methods. J Clin Epidemiol. 2007 May;60(5):431–9. pmid:17419953
- 45. Kruse C, Steffensen R, Varming K, Christiansen OB. A study of HLA-DR and -DQ alleles in 588 patients and 562 controls confirms that HLA-DRB1*03 is associated with recurrent miscarriage. Hum Reprod. 2004 May;19(5):1215–21. pmid:15070884
- 46. Nielsen HS, Steffensen R, Varming K, Van Halteren AG, Spierings E, Ryder LP, et al. Association of HY-restricting HLA class II alleles with pregnancy outcome in patients with recurrent miscarriage subsequent to a firstborn boy. Hum Mol Genet. 2009 May 1;18(9):1684–91. pmid:19223392
- 47. Piosik ZM, Goegebeur Y, Klitkou L, Steffensen R, Christiansen OB. Plasma TNF-alpha levels are higher in early pregnancy in patients with secondary compared with primary recurrent miscarriage. Am J Reprod Immunol. 2013 Nov;70(5):347–58. pmid:23656493
- 48. Guyatt GH, Oxman AD, Vist GE, Kunz R, Falck-Ytter Y, Alonso-Coello P, et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008 Apr 26;336(7650):924–6. pmid:18436948
- 49. Hutton B, Sharma R, Fergusson D, Tinmouth A, Hebert P, Jamieson J, et al. Use of intravenous immunoglobulin for treatment of recurrent miscarriage: a systematic review. BJOG 2007;114(2):134–42. pmid:17166218
- 50. Ata B, Tan SL, Shehata F, Holzer H, Buckett W. A systematic review of intravenous immunoglobulin for treatment of unexplained recurrent miscarriage. Fertil Steril. 2011;95(3):1080–5 e1–2. pmid:21232738
- 51. Christiansen OB, Larsen EC, Husth M, Pedersen B. Intravenous immunoglobulin and recurrent miscarriage. Fertil Steril. 2011 Jun;95(7):e35; author reply e6. pmid:21457961