Statistics from Altmetric.com
With new effective therapies and less long-term disability most women with inflammatory rheumatic diseases (RDs) can contemplate pregnancy though substantial risk for adverse maternal and fetal outcomes remain particularly in RD with organ involvement. Drug treatment during pregnancy may be required in order to control maternal disease which itself can be a threat for fetal well-being and pregnancy outcome. The risk of leaving active inflammatory RD of the mother untreated for 9 months must be weighed against any potential harm through drug exposure of the fetus.
Adjustment of therapy in a patient planning a pregnancy aims to use medications that support disease control in the mother and are considered safe for the fetus. However, only a limited number of antirheumatic/immunosuppressive drugs fulfil these requirements. With the rapidly increasing number of medications available for the treatment of RD, knowledge on safety in pregnancy lags behind. A consensus paper on use of antirheumatic drugs in pregnancy and lactation was published in 20061 with an update on immunosuppressive drugs in 2008.2 A European League Against Rheumatism (EULAR) task force regarded it timely to collect new available data from the literature and from several databases, and reach expert consensus on their compatibility during pregnancy and lactation, resulting in EULAR points to consider for use of antirheumatic drugs before pregnancy and during pregnancy and lactation.
Participants and methods
The EULAR task force on antirheumatic drugs during pregnancy and lactation is a multidisciplinary committee consisting of 20 members from 10 European countries and the USA (9 rheumatologists, 3 internists, 1 obstetrician, 2 rheumatologist/epidemiologists, 1 specialist in Obstetric Medicine, 1 geneticist, 2 patients with RD as patients’ representatives and 1 research fellow). The objective was to formulate points to consider for the use of antirheumatic drugs during pregnancy and lactation by identifying and critically evaluating recent literature and registry data. The task force followed the procedures outlined in 20043 and updated in 2014.4
Systematic literature review
At the first meeting, the committee decided on the medications to be included in the systematic literature review (SLR): Non-steroidal anti-inflammatory drugs (NSAIDs), glucocorticoids, conventional synthetic DMARDs: methotrexate (MTX), cyclophosphamide, sulfasalazine, leflunomide, antimalarials, azathioprine, colchicine, ciclosporin, tacrolimus, mycophenolate mofetil (MMF), intravenous immunoglobulin (IVIG) and targeted synthetic DMARDs: tofacitinib. Biologic DMARDs included were tumour necrosis factor inhibitors (TNFi) (adalimumab, certolizumab pegol, etanercept, golimumab and infliximab), the T cell costimulation inhibitor abatacept, the anti-B cell agents rituximab and belimumab, the interleukin (IL)-6 receptor-blocking monoclonal antibody tocilizumab, and the IL-1 receptor antagonist anakinra. Biosimilars were not included due to lack of data. Two electronic searches, one for biologic drugs and a separate search for non-biologic drugs were performed in Embase, Medline, PubMed and Cochrane Library from 1 January 2008 to 1 April 2015 by a research librarian at the Norwegian University of Science and Technology University library; Medicine and Health Library, drawing on the Cochrane Musculoskeletal group's strategy for searching for all RDs and adjusting the strategy to make use of the various database search facilities.5 The searches were restricted to effects in pregnancy and/or perinatal effects, and excluded reviews (for details see online supplementary figure S1). References of articles found were screened for additional evidence. The search period of 2008–2015 was chosen because inclusion of publications in the consensus paper of 2006 ended early in 2006, and the update of 2008 ended in 2007. As the update publication2 did not include all non-biologic drugs, an additional search for the period 2006–2008 was performed for 10 drugs; NSAIDs, glucocorticoids, MTX, cyclophosphamide, sulfasalazine, antimalarials, azathioprine, colchicine, ciclosporin and IVIGs. Because of paucity of lactation data, all reports on lactation exposures to antirheumatic drugs published 1970–2015 derived from LactMed, a database in the Toxicology Data Network, were included.
Publications were restricted to the English language and included prospective and retrospective studies, cohort studies, case-control studies, and case reports. In addition abstracts from major international congresses were included. The search was not limited to RD, but all indications for a given drug were included (see online supplementary figure S1). Results of the different databases were combined and duplicates were excluded; issues regarding inclusion or exclusion of articles were resolved by discussion and consensus between the fellow (CGS) and convenor (MØ).
The task force obtained access to pregnancy reports from two pharmacovigilance centres and four safety databases from pharmaceutical industries (see online supplementary table S1), and extracted data for all pregnancies with known outcomes.
Data collection sheet
A data collection sheet was constructed to extract relevant data on exposure during pregnancy and lactation. Included were patient characteristics, drug dosing, and exposure time before and during pregnancy/lactation, concomitant medication, and occurrence of pregnancy complications (miscarriages and elective terminations, stillbirth, and preterm delivery) or adverse child outcomes. Congenital malformations, birth weight, neonatal health, infections during the 1st year of life, vaccination responses, and follow-up of childrens’ physical and neurocognitive development were also recorded. Reports that did not disclose the outcome of pregnancy or those for which the temporal association between drug exposure and onset of pregnancy could not be determined were excluded from analysis. Likewise, incomplete reports on breastfeeding exposures were excluded.
We defined as the primary outcome major congenital malformations in live-born children or aborted fetuses. The only secondary outcome included was miscarriages up to 20 weeks gestation. Other outcomes like termination of pregnancy, pre-eclampsia, prematurity, low birth weight, perinatal and postnatal problems were either incompletely documented or imprecise because of confounding factors. For lactation exposure the primary child outcome was defined as any adverse effect (clinical or laboratory).
Experts’ consensus and Delphi rounds
The results of the SLR were presented to the task force members to initiate group discussions and to arrive at statements for the use of antirheumatic drugs in pregnancy and lactation. Statements were based on the consensus papers of 2006/20081 ,2 with added evidence from the new SLR as well as unpublished registry data. In the formulation of statements, emphasis was put on congenital malformations since this was the primary outcome that was consistently reported in all publications included. Given the paucity of high quality data and subjective nature of many decisions, the task force agreed that the practicing clinician would be better served by having each expert stating (dis)agreement with the proposed statement and expressing their practice regarding the use of each medication in daily practice (see online supplementary table S2). The Delphi technique6 was used to reach consensus on the statements and rate of agreement for the propositions for clinical use.
Strength of evidence
The classical ranking of evidence scores defines systematic reviews of randomised controlled trials (RCTs) as providing the highest level of evidence, followed by individual RCT.7 Classical scores of evidence focus on efficacy of an intervention or of drugs, but not on safety. By contrast, evaluation of drugs in pregnancy and lactation has its focus on safety for the embryo/fetus or child, not on efficacy. No adequate ranking system for evaluating strength of evidence (SOE) in regard to safety of drugs in pregnancy and lactation has been developed. After several rounds of discussion, the group decided to use two classical ranking systems in spite of their shortcomings in regard to reproduction issues.
The quality of evidence based on study design was rated according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system and Oxford Centre of Evidence Rating.8 ,9 Data from SLR and data added from registries were subjected to group discussion to establish SOE regarding the statements. For each statement, SOE was graded using a 1–4 ordinal scale for GRADE (see online supplementary table S3) and a 1–5 ordinal scale for Oxford (see online supplementary table S4). The members were then asked to select the proposition that best described their personal current use of each drug during pregnancy and lactation, as described above (see online supplementary table S2). The percentage of consensus for the statements, and agreement for clinical use in pregnancy and lactation were calculated.
In the first meeting the task force defined four overarching principles. In the following two meetings and seven online Delphi rounds (four concerning medication in pregnancy and three concerning medication during lactation) 11 points to consider were developed (table 1).
Result of SLR
The electronic searches identified a total of 5960 references on antirheumatic drugs during pregnancy and lactation. Additional references were added from hand searches. Nine hundred and forty-four duplicates were excluded (see online supplementary figure S1). A total of 319 publications were eligible for analysis: 45 cohort studies (including 7 abstracts), 24 case-control studies (including 1 abstract) and 250 case series/case reports (including 21 abstracts). Unpublished data from six registries were also included (see online supplementary table S1).
Type of studies recorded, references on cohorts and case controls, number of pregnancies included, pregnancy outcomes, and SOE for each drug or group of drugs are presented in table 2. References on case reports and case series are available in online supplementary table S5.
General aspects of SLR
Adverse pregnancy outcomes
Adverse outcomes other than congenital malformations were not consistently reported; this also applies to miscarriages. Rates of miscarriages may be imprecise since they depend on the time point at which a pregnant patient is included in a study. Only MTX and MMF have been consistently shown to increase the rate of miscarriages. Combination therapies with MTX have sometimes also increased the rate of miscarriages (example table 2, abatacept). The observed correlation between NSAIDs and miscarriage in some studies is controversial because of several confounding factors, including confounding by indication, that often have not been addressed in the studies. The majority of data relate to first trimester exposure. Exposures in the second and third trimesters have been reported for medications either regarded as compatible with pregnancy (examples glucocorticoids, azathioprine, antimalarials) or when serious maternal disease requires therapy in pregnancy (example cyclophosphamide). Drug exposures before conception were included for agents with a long half-life, mainly biologics, with a safety margin five times the half-life.
For cyclophosphamide, ciclosporin, tacrolimus, glucocorticoids and IVIG the number of new publications shown in table 2 is lower than the number of citations retrieved in the literature search. The reason is that these drugs were often administered in combination therapies, and pregnancy outcomes in reports were given for a drug combination, not for each drug separately. The rate of miscarriage and congenital malformations can therefore be given only for studies reporting single drug exposure (table 2).
Several factors limit the completeness and reliability of pregnancy reports from pharmacovigilance centres and from pharmaceutical safety databases: spontaneously reported data often lack preciseness and completeness, and can be biased towards abnormal pregnancy outcomes, particularly in retrospectively collected cases. Information on concomitant medication is frequently absent. A major limitation of global safety databases is the high rate of pregnancies with unknown outcomes and high lost to follow-up rates up to 50%.69
Studies on excretion of drugs into human breast milk are rare and mostly based on single-dose or short-term treatment, therefore grading of evidence for all drugs in table 3 is ‘very low’ (+) according to GRADE (see online supplementary table S3) and score 4–5 according to Oxford evidence rating (see online supplementary table S4). Even when transfer of a drug into milk has been investigated, children were often not breast fed, and the effect of the drug on the nursing infant remains unknown. References concerning lactation are available in online supplementary table S6.
Follow-up of children
Pregnancy exposures in any trimester might have the potential to impair organ function, alter the immune response or influence neurocognitive development in children. Studies published between 2006 and 2015 deal mainly with biologics, have a short follow-up time and show large gaps in reported outcomes (see online supplementary table S7). The available data for several biologics and immunosuppressives show no adverse effects on physical or neurocognitive development nor impaired immunocompetence in children during the 1st year of life.
Biologics are derivatives of IgG, and differ in structure, half-life and placental passage. The half-life ranges from 9 days to 23 days in complete IgG1 monoclonal antibodies and between 4 days and 13 days in Fc-fusion proteins (etanercept, abatacept).72 Active transport of biologics containing the Fc part of IgG1 is mediated by the fetal Fc receptor expressed in the placenta.72 Transfer is thought to be very low during organogenesis, but to increase steadily after week 13 throughout pregnancy. Treatment of the mother with IgG antibodies expressing high affinity to the fetal Fc receptor after gestational week 30 can lead to fetal/cord serum levels equal to or higher than maternal levels.61 IgG has a prolonged half-life, up to 48 days73 in the newborn; they typically disappear from the child’s serum within the first 6 months of life.
The biologics with most pregnancy experience are the TNFi which have been in use for 15 years, including for indications outside rheumatology. For biologics approved <5 years ago, data on pregnancy and lactation are either sparse or completely absent (tables 2 and 3).
Results of Delphi voting
There was 90–100% consensus between experts of the task force on all statements on antirheumatic drugs during pregnancy (tables 4 and 5). Propositions regarding actual use of antirheumatic drugs during pregnancy and lactation in clinical practice received lower levels of agreement (tables 4 and 5). Disagreement between experts on clinical use during lactation was between 10–20% in general, and 25–31% for several biologics without data on transfer into breast milk (abatacept, tocilizumab, belimumab).
Available data from the literature and from registries show that a large proportion of medications can be taken by pregnant and lactating women with RD without causing measurable harm to the children. The SLR of the last decade strengthens the evidence for glucocorticoids, sulfasalazine, antimalarials, azathioprine, colchicine, ciclosporin, tacrolimus and IVIGs as being compatible with pregnancy and lactation (table 1). Major changes in regard to the 2006/2008 consensus paper are the following: The SLR and registry data support the use of TNFi in the first half of pregnancy. A study published after the completed Delphi voting showed a slight increase of birth defects at first trimester exposure to TNFi without any pattern of malformations. Given the absence of disease-matched controls the clinical significance of this finding is not yet clear.65
The difference in placental transfer related to molecule structure and half-life needs to be taken into account when selecting a TNFi for women of fertile age (table 5). As a consequence, infliximab and adalimumab may preferentially be stopped at 20 weeks, and etanercept at week 30–32 of pregnancy. The safety of certolizumab in using it throughout pregnancy needs confirmation by extended published reports. Sound evidence for fetal/child safety is still lacking for certolizumab, golimumab, abatacept, tocilizumab, rituximab, belimumab and anakinra, but SLR and registry data do not suggest any evidence of harm from these agents when used before conception or in the first trimester (table 5).
The SLR and registry data showed only cyclophosphamide, MTX and MMF to be teratogenic necessitating their withdrawal before a planned pregnancy. For all other drugs labelled with a statement to discontinue them before or early in pregnancy, the reason is insufficient evidence that they are safe for the fetus, rather than evidence of harm.
Since 30–50% of pregnancies are unplanned, a major question is how to manage pregnancies that occur in women receiving therapy with teratogenic drugs. Some patients opt for immediate termination whereas others contemplate continuation of the pregnancy. Confirmation of pregnancy by a gynaecologist and determination of exact exposure dates for individual risk assessment and counselling are mandatory. A detailed ultrasound examination of the fetus should be offered to all patients who have an unintended pregnancy while taking a teratogenic drug. Macroscopic anomalies can be assessed by experienced fetal medicine specialists at the end of the first trimester and scans should be repeated at later stages of the second trimester. Other prenatal tests like amniocentesis or chorionic villous biopsy are usually not indicated after maternal drug exposure, but might be considered in patients with high risk of chromosomal problems or anomalies at ultrasound examination.
There was 90–100% agreement between experts of the task force with the statements on compatibility of antirheumatic drugs during pregnancy. However, much less agreement was achieved for the use of each drug in clinical practice. In the statements, emphasis was placed on congenital malformations whereas in the propositions for clinical use other considerations come into play including personal experience with a given drug, pharmacological properties of drugs, national preferences, availability of drugs in certain countries and legal issues. Statements on lactation were restricted to compatibility, and included no detailed advice on timing, short-term discontinuation of breast feeding or discarding milk on days of drug administration. As a consequence, great heterogeneity in regard to clinical practice among experts was observed (tables 4 and 5). This reflects the insufficient documentation in the field as well as the propensity to discourage patients in need of therapy from breast feeding although a flexible schedule would allow more women to breastfeed. Lactating mothers may have the opposite view, and would rather breast-feed than receive medications for active disease. Faced with a paucity of studies, pharmacological properties of drugs may act as a guide for decision to allow breast feeding even when there is scarce or no documentation (table 3). Non-ionised and lipophilic agents with a low molecular weight are the most likely to be transferred into breast milk. Highly protein-bound drugs or agents with high molecular weight are unlikely to cross extensively into breast milk.74 Term neonates, older or partially breastfed babies are usually at low risk for side effects of drugs in breast milk. Breast feeding is particularly important for premature and very low birthweight babies, however, no studies on this subgroup and the risks they may encounter by exposure to drugs in breast milk are available.
Studies on the long-term effects of drugs administered during pregnancy and/or breast feeding on child health and development are scarce, and often of low quality (see online supplementary table S7). The data available for azathioprine, ciclosporin and dexamethasone do not indicate immunosuppression in exposed children or raise special concern in regard to physical or neurological development (see online supplementary table S7). By contrast, biologics with extensive placental transfer achieving high serum levels in the child when administered after gestational week 30, might increase the risk of postnatal infection. Children exposed to biologics only before week 22 can receive vaccinations according to standard protocols including live vaccines. Children exposed at the late second and during the third trimester can follow vaccination programmes, but should not receive live vaccines in the first 6 months of life. When available, measurement of child serum levels of the biologic in question could guide the decision for or against a live vaccine.
The strengths of this study include the extensive SLR, inclusion of until now unpublished pharmacovigilance and registry data, and evaluation of data by experts from different specialties. Limitations of the study are the great variability in quality of reports in the literature and in registries. There is variety in disease indications and drug dosage. Assignment of an adverse pregnancy outcome to a particular drug can be influenced by confounders. Disease type, disease activity during pregnancy, extent of systemic inflammation and organ involvement, comorbidities, and concomitant drug therapy may all contribute to negative outcomes. When combinations of immunosuppressive and cytotoxic drugs are used defined pregnancy outcomes cannot be assigned to each of these classes of drugs separately. For recently approved biologics the adverse effect of concomitant use of MTX confounds the rate of miscarriages and of congenital malformations occurring after first trimester exposure in unintended pregnancies (table 5). In studies without carefully matched non-exposed control groups it is difficult to separate adverse drug effects from the above-mentioned confounders. Control groups are lacking in a majority of reports. The malformation rate is nearly always reported for live birth but does not include information on miscarriages or terminations. Therefore malformation rates are best derived from studies that include comparator groups of women with the same disease unexposed to the drug under consideration as well as non-exposed healthy pregnant women.
Treating a pregnant woman with RDs during pregnancy and lactation is a challenge since the well-being of two individuals, the mother and her child, has to be considered. Decisions on therapy during pregnancy and lactation have often been confounded by medical and legal concerns.75 The general cautious attitude to drug treatment during pregnancy and lactation has resulted in the withholding of necessary therapy, often at considerable risk for mother and child.75 Updating of knowledge in the field and dissemination of new insights is therefore of great importance in order to ensure implementation into daily practice and counselling. A publication based on SLR and available registry data is a first step that must be followed by dissemination of the new data through congresses, conferences, workshops and educational courses that include all types of healthcare providers (HCPs). Dissemination should target national societies of specialists in rheumatology, internal medicine, gynaecology and obstetrics, family medicine, paediatrics and pharmacology as well as national teratology information services. Disseminating the data through internet accessible websites would reach a large audience of the different HCPs who care for patients with RDs. Ideally the new insights should also be communicated to the patients at congresses, conferences and via national patient associations. Information needs on childbearing issues are great in women with RDs.76 There is a considerable gap in written material and educational resources that could meet this need. Development of evidence-based information on drugs in pregnancy/lactation, tailored for the lay public and accessible on the internet, would help patients make informed decisions.
Recommendations for future research
Despite various international efforts, there is still limited evidence on the safety of a substantial number of drugs in pregnancy and lactation. The following are points for a research agenda:
All pharmaceutical companies should give academic institutions access to data on drug exposure during pregnancy and lactation from long-term extension studies of randomised trials and from registries. Independent assessment of the available data would be crucial.
Current initiatives for establishing pregnancy registers should be continued on a long-term and international basis. Specifically for the more recently licensed drugs, data collection should be intensified. Individual pregnancy registers are not likely to yield enough exposure and observation time to draw valid conclusions. Therefore, joint approaches among several countries which enable collaborative data analyses are recommended. EULAR could be an umbrella organisation for the harmonisation of approaches in establishing pregnancy registries.
Data collection should follow a protocol and be prospective, starting in early pregnancy or preferably when a pregnancy is planned and with high follow-up rates throughout pregnancy, lactation and during at least the 1st year of life of the child. Studies should include comparator groups of disease-matched women and their children unexposed to the drug under consideration as well as non-exposed healthy pregnant women.
The major gap in the documentation of transfer of drugs into human breast milk and the effect of drugs in breastfed children, including risk groups of premature and very low birthweight children, requires new and detailed studies.
Management of female patients with RDs during pregnancy and lactation requires weighing risks of withholding treatment from the mother against any risk to the fetus/child via exposure to drugs during pregnancy or breast feeding. Restrictions in use apply for the few proven teratogenic drugs and the large proportion of medications for which insufficient safety data for the fetus/child are available. The points to consider presented in this review show that, in spite of limitations, effective drug treatment of active RD is possible with reasonable safety for the fetus/child during pregnancy and lactation.
The authors thank Solveig Isabel Taylor; Medicine and Health Library, NTNU University Library Knowledge Center, St Olavs Hospital, Trondheim, Norway for help with the literature search; Ane S Simensen; The Norwegian Medicines Agency, Pharmacovigilance, Oslo, Norway for access to pregnancy reports; May Ching Soh; Division of Women’s Health, Women’s Health Academic Centre, King’s College London, King’s Health Partners, London, UK, for collecting data on lactation; Cecilia Nalli, Department of Clinical and Experimental Science Rheumatology and Clinical Immunology Unit, Spedali Civili and University of Brescia, Italy for collecting data on child outcome.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.