NSAIDs generally fall into FDA Category B and most of them can be used safely in the first part of pregnancy, although increased risk [odds ratio (OR) = 1.8, 95% CI 1.0, 3.2] of first trimester miscarriage under NSAID treatment has been reported [⁵⁷]. The risk associated with NSAID use increases with pregnancy duration, since NSAIDs used late in pregnancy can cause premature closure of the ductus arteriosus and increase the risk of neonatal bleeding [⁵⁸]. Moreover, NSAID use in pregnancy has been associated with reversible impairment of fetal renal function and the development of oligohydramnios, in which case immediate discontinuation of NSAID administration is required [⁵⁸]. Most NSAIDs therefore receive an FDA Category C labelling beyond 30 weeks of gestation.

Data on the use of selective cyclo-oxygenase-2 (COX-2) inhibitors during pregnancy are scarce, hence these drugs are not recommended for treating RA inflammatory symptoms during pregnancy [⁵⁸, ⁵⁹]. Moreover, the use of COX-2 inhibitors for prevention of premature labour in at-risk pregnancies was associated with a reversible negative effect on fetal renal function and an increased incidence of delivery before 37 weeks of gestation [⁶⁰].

Most NSAIDs are considered safe during lactation, although they are excreted in milk in low quantities. Aspirin should not be used in doses exceeding 100 mg/day. Feeding immediately before administration of the drug can help to limit drug exposure in infants [⁵⁹, ⁶¹, ⁶²].

CSs at doses up to 15 mg/day (prednisolone equivalent) are believed to be safe throughout pregnancy [⁵⁸]. Higher doses increase the risk of infection and premature delivery [⁵⁹]. Systemic treatment with CSs during the first trimester of pregnancy is believed to slightly augment the risk of oral clefts, from 1 per 1000 live births to 1.3–3.3 per 1000 live births [⁶³]. In the PARA study, the use of CSs during pregnancy in RA patients was associated with lower gestational age at delivery and increased incidence of premature delivery, even at doses <7.5 mg/day. Women taking prednisone delivered on average 1 week earlier [²²].

Pregnant women are preferably treated with prednisone or prednisolone, as these CSs are largely converted to inactive metabolites by placental 11β-hydroxy steroid dehydrogenase, the mechanism in place to protect the fetus from raised levels of maternal cortisol during pregnancy. When the fetus is the target of corticoid therapy, dexamethasone or betamethasone are the compounds of choice, as they cross the placental barrier more efficiently [⁶⁴, ⁶⁵].

Corticotherapy is also considered safe during lactation [⁶²]. Prednisolone is secreted in milk at doses estimated at <0.1% of the maternal dose, which is thought to correspond to <10% of the infants endogenous cortisol level [⁶⁶].

MTX is a known teratogen and as such is contraindicated during pregnancy (FDA Category X; Table 2). MTX causes embryotoxicity and teratogenicity in mice, rats and rabbits. In humans, the aminopterin syndrome, consisting of growth deficiency and major CNS, bone and cardiac abnormalities, has been described in children born after exposure to MTX in the first trimester of pregnancy [⁶⁷]. In view of the long tissue retention time of the active metabolites of MTX, its administration must be stopped 3–6 months before conception [⁶⁸]. MTX is a folic acid antagonist. Therefore, folic acid supplementation after MTX discontinuation is very important. Past exposure to MTX has no negative effects on pregnancy outcomes [⁶⁹]. A recent systematic review found that the risk of minor birth defects after conception while using low-dose MTX (5–15 mg/week) is ∼5%, with no indication of the serious aminopterin syndrome, and questions whether elective abortion is required for pregnancies conceived during low-dose MTX therapy [⁶⁸]. MTX is excreted into breast milk in low concentrations [⁶⁶], but can accumulate in the infant’s tissues, and is therefore contraindicated during lactation.

LEF blocks de novo pyrimidine synthesis by inhibiting dihydroorotate-dehydrogenase, and additionally inhibits protein tyrosine kinase activity. Animal reproduction studies indicate that LEF is both embryotoxic and teratogenic, mainly leading to craniofacial, skeletal and cardiovascular malformations [⁷⁰], which caused the FDA to classify this drug in pregnancy Category X. Due to the long half-life of its metabolites, LEF should be discontinued for 2 years before pregnancy. Alternatively, a washout procedure with cholestyramine should be used until plasma levels are <0.02 µg/ml on two separate measurements at least 2 weeks apart [⁷¹].

A recent prospective study compared pregnancy outcomes in 64 RA patients exposed to LEF during pregnancy (61 of which underwent a cholestyramine washout procedure), 108 RA patients not exposed to LEF and 78 healthy controls. In this study, the overall rate of major structural defects between the studied groups was not significantly different, nor did prenatal LEF exposure give rise to a specific pattern of major or minor anomalies [⁷²]. LEF is secreted into breast milk and therefore continues to be contraindicated during breastfeeding [⁵⁹, ⁷³].

It is generally regarded as safe to keep using SSZ during pregnancy (FDA Category B drug), despite some reports noting a higher incidence of neural tube defects, oral clefts and cardiovascular defects. The outcome of pregnancies exposed to SSZ has mainly been studied in women with IBD. A number of studies have concluded that SSZ use during pregnancy does not give rise to increased rates of birth defects in women with IBD when compared with untreated IBD patients or the general population [⁷⁴, ⁷⁵]. In a recent meta-analysis treatment of IBD patients with 5-ASA, these drugs did not significantly increase the risk of congenital abnormalities (OR = 1.16, 95% CI 0.76, 1.77, P = 0.57), stillbirth (OR = 2.38, 95% CI 0.65, 8.72, P = 0.32), spontaneous abortion (OR = 1.14, 95% CI 0.65, 2.01, P = 0.74), preterm delivery (OR = 1.35, 95% CI 0.85, 2.13, P = 0.26) or low birthweight (OR = 0.93, 95% CI 0.46, 1.85, P = 0.96) [⁷⁶].

SSZ therapy should be accompanied by extra folate supplementation, as this medication halts folate synthesis by inhibiting dihydrofolate reductase. Folic acid supplementation was shown to decrease the augmented risk of oral clefts and cardiovascular anomalies associated with folate antagonist treatment during pregnancy [⁷⁷]. The SSZ metabolite sulphapyridine is secreted into breast milk. SSZ is generally considered safe during lactation, although a case of bloody diarrhoea in an infant has been reported [⁶², ⁷⁸].

Use of HCQ during pregnancy at the low doses used for malaria prophylaxis does not increase the risk of congenital abnormalities or pregnancy loss [⁷⁹]. When used as a DMARD at much higher doses, the continuation of this drug during pregnancy has long been controversial and the FDA classifies HCQ in Category C. This controversy was mainly due to reports on retinal toxicity and ototoxicity observed after treatment with chloroquine [⁸⁰].

HCQ use during pregnancy has been reported in more than 250 pregnancies in SLE patients, with no indications for increased congenital malformations or hearing or vision impairment. A more limited number of children exposed to HCQ have been assessed for retinal toxicity or cardiac conductivity disturbances, again with no indications of adverse effects of the drug. HCQ is secreted into milk at low concentrations, but in view of the transplacental crossing of the drug and the relatively long tissue half-life, it is not deemed necessary to advise against lactation of children exposed to the drug during pregnancy [⁸⁰].

AZA and 6-mercaptopurine (6-MP) are designated as FDA Category D drugs (Table 2), indicating that increased risk for the fetus exists, but the risk must be weighed against the possible benefits of the drug. Although AZA is teratogenic in animals, inducing skeletal and visceral malformations in rabbits and mice at doses equivalent to the human dose, multiple case series and cohort studies in human pregnancy, mostly in transplant recipients and IBD patients, have not revealed increased incidence of congenital anomalies or recurrent patterns of malformations [⁷⁴, ^81–85].

A recent cohort study in Denmark, including birth outcomes of 76 exposed pregnancies, showed an increased risk of preterm delivery and low birthweight in women exposed to AZA or 6-MP during pregnancy, but no significant increase in congenital malformations. However, the data suggest that adverse birth outcomes are caused by the underlying disease rather than by use of AZA or MP [⁸⁶].

Thiopurines undergo a complex metabolization process and the placenta forms a partial barrier to their metabolites. Indeed, the active metabolites 6-thioguanine nucleotides (6-TGN) are detectable in fetal red blood cells, whereas 6-methylmercaptopurine (6-MMP) is not [⁸⁷]. Dosing 6-TGN at least once during pregnancy is recommended to avoid excessively high levels, which may cause myelosuppresson in mother and child [⁸⁶]. Traditionally, breastfeeding is believed to be contraindicated under thiopurine treatment, but several recent studies have shown that 6-MP levels in breast milk are low, and conclude that breastfeeding during treatment with AZA seems safe [^87–89].

The currently available TNF inhibitors (etanercept, infliximab, adalimumab, golimumab and certolizumab) are all classified as FDA Category B drugs, indicating that no teratogenic effects of these drugs were observed in animal reproduction studies, but adequate and controlled human safety data are still lacking.

TNF-α is a pleiotropic cytokine that, in addition to being a pivotal cytokine in the pathogenesis of immune-mediated inflammatory disorders, plays a role in various host defence mechanisms and in regulating and maintaining pregnancy. TNF-α and its receptors are expressed in the uterus, placenta and embryo, but surprisingly, pup morphology, litter size and growth of TNF-α knockout mice were not significantly different from wild-type controls [⁹⁰]. When exposed to the teratogen CYC, however, fetuses from TNF-α knockout mice showed significantly higher proportions of malformations, indicating that TNF-α plays a role in protecting the fetus against teratogenic stress [⁹¹]. Administration of the TNF inhibitors during pregnancy and lactation did not influence the number or function of immune-competent cells in mice [⁹²] or macaques [⁹³], although another study reported growth retardation as well as atrophy of the thymus, spleen and lymph nodes in pups born to mice treated with anti-TNF antibodies [⁹⁴].

Despite the roles of TNF in establishing and maintaining pregnancy and protecting the fetus, IgG anti-TNF antibodies can probably be safely administered during the first part of pregnancy, since IgG antibodies do not cross the placental barrier in the first trimester. Indeed, transplacental transport of IgG only starts after the first trimester and mainly increases during the third trimester of pregnancy [⁹⁵]. IgG levels in the maternal circulation decrease during pregnancy, concomitant with a rise of IgG from maternal origin in the fetal circulation starting from Week 13 onwards. At term, IgG levels in the fetus exceed the levels in the maternal circulation, which is indicative of an active transport mechanism (Fig. 2) [⁹⁶]. Transplacental transport was shown to be most efficient for the IgG1 and least efficient for the IgG2 subclass [⁹⁶]. IgG is transported over the placenta via pH-dependent receptor-mediated transcytosis, in which maternal IgG binds to the so-called neonatal immunoglobulin constant fragment (Fc fragment) receptor expressed on the syncytiotrophoblast membrane, followed by transcytosis and release of bound IgG in the fetal circulation (Fig. 3) [^97–99].

Certolizumab is a PEGylated Fab′ fragment of an anti-TNF mAb without Fc fragment. Certolizumab would therefore be expected not to be transferred across the placental barrier according to the mechanism described above. Indeed, animal studies report much lower placental transfer of PEGylated Fab′ antibodies; data in humans are not yet available, but a case of certolizumab administration as rescue therapy in the third pregnancy trimester of an IBD patient had a favourable outcome [¹⁰⁰].

A review of more than 120 000 adverse events recorded in the FDA drug surveillance database revealed 61 congenital anomalies in 41 children born to women treated with etanercept or infliximab during pregnancy. These anomalies could be classified into 34 different types, 19 of which were associated with the so-called VACTERL syndrome [¹⁰¹]. VACTERL is an acronym for congenital syndromes characterized by the following birth defects: vertebral, anal atresia, cardiac abnormalities, tracheoesophageal fistula, oesophageal atresia, renal abnormalities and limb abnormalities. This study was criticized for the inherent selection bias in the reporting of anomalies, the lack of sensitivity analysis taking into account co-existing risk factors, the fact that the congenital anomalies reported were among the most common in the population and the fact that none of the reported cases showed three anomalies, necessary for formal VACTERL diagnosis. These criticisms cast doubts on the validity of the assumption that TNF inhibitors cause congenital defects of the VACTERL type [¹⁰², ¹⁰³].

Apart from the VACTERL controversy, accumulating data on anti-TNF treatment in RA during pregnancy has not yielded any indications of teratogenicity or adverse outcomes [^104–111]. Although the indications that anti-TNF therapy is not teratogenic are growing as the number of pregnancies exposed to anti-TNF during preconception and early pregnancy increases, experience with anti-TNF therapy in pregnant arthritis patients is still quite limited [⁵⁸, ¹¹²].

Probably rheumatologists can learn from the more extensive experience with anti-TNF treatment during pregnancy that exists already for IBD. In the latter disease context, the need for effective treatment during pregnancy is higher than in RA, because IBD disease activity is less suppressed by pregnancy [¹¹³, ¹¹⁴]. Additionally, the abdominal location of the disease raises the stakes of maintaining disease remission during pregnancy in IBD, as IBD disease flares during pregnancy carry a high risk of adverse birth outcomes, including prematurity, low birthweight and congenital abnormalities [¹¹⁵]. These considerations have prompted gastroenterologists to build more extensive experience with anti-TNF treatment during pregnancy. A number of studies report intentional use of anti-TNF treatment during pregnancy [¹⁰⁸, ¹¹⁶, ¹¹⁷], and the 2010 European Crohn's and Colitis Organisation (ECCO) guidelines state that ‘medical treatment for Crohn’s disease (except methotrexate) should generally continue during pregnancy, because the benefits outweigh the risk of medication’ [¹¹⁸].

Although experience is still limited, the fact that anti-TNF antibodies do not cross the placenta in the first trimester of pregnancy together with the overall unremarkable outcome of pregnancies in women treated with TNF inhibitors lead us to the opinion that TNF blockers can be useful in RA patients to keep disease activity under control in the preconception period after withdrawal of teratogenic DMARDs and in the first trimester of pregnancy.

Abatacept is a cytotoxic T-lymphocyte antigen 4 (CTLA4) human immunoglobulin fusion protein that selectively blocks T-cell co-stimulation through the CD80/CD86 pathway. Abatacept crosses the placenta, but animal studies with abatacept have revealed no congenital abnormalities. Human pregnancy data are still lacking. Hence this drug received an FDA Category C classification and it is advised to discontinue therapy for 10 weeks before conception [⁵⁸]. Since no data are available currently on the secretion of abatacept into milk, breastfeeding is contraindicated during abatacept treatment [¹¹⁹].

Animal reproduction studies did not observe fetotoxicity caused by the anti-CD20 mAb rituximab, which depletes B cells. Scant data are available on the outcome of human pregnancies exposed to rituximab. Consequently the FDA classifies this drug in Category C.

The eight case reports of rituximab treatment during pregnancy have reported reversible B-cell depletion and lymphopenia in the neonate, but no congenital malformations. Due to its long-lasting effects, it is recommended that rituximab treatment be stopped 1 year before conception [⁷⁹]. Currently data on secretion of rituximab in breast milk are still lacking, hence the safety of this biological during lactation cannot be guaranteed [¹¹⁹].

Preclinical studies have not observed adverse effects of tocilizumab on fertility or fetal development. In cynomolgus monkeys, small increases in abortion rates and fetal mortality rates were observed, but these effects only occurred at tocilizumab doses >100-fold the dose used in humans. No pregnancies under tocilizumab therapy have been reported up to now [¹¹⁷]. Currently it is advised to discontinue tocilizumab therapy 3 months before planned pregnancy [⁵⁸]. Data on excretion of tocilizumab in breast milk are not available at present [¹²⁰].