Rheumatoid arthritis (RA) is a systemic inflammatory disorder characterized by chronic synovitis leading to joint destruction. During pregnancy, spontaneous reduction of disease activity in RA is common, a phenomenon that is also observed in other autoimmune disorders [1–5]. After birth, however, RA deteriorates in the majority of women [3, 4, 6]. Pregnancy is supposed to have immunomodulatory effects, but the exact mechanisms underlying the spontaneous amelioration during pregnancy and the subsequent postpartum flare have still not been elucidated. Several hypotheses have, however, been put forward, including the beneficial effect of maternal-fetal HLA-incompatibility [7, 8] and of increased galactosylation of immunoglobulin G [9–11]. Shifts in T-cell cytokine secretion profiles also have been proposed as a potential mechanism underlying the improvement of RA during pregnancy and the postpartum deterioration [12–15].

In healthy pregnancy, total and free levels of cortisol increase progressively, reaching a peak in the second and third trimesters [16–18]. The improvement in RA starts in the first trimester, and almost half of patients have at least low disease activity (DAS28 <3.2) in the third trimester [4]. Nevertheless, prospectively studied cohorts of pregnant RA patients concurrently evaluating reduction of disease activity with accompanying (free) cortisol levels on an individual basis are lacking. It is known from daily clinical practice, however, that interindividual differences in the degree of pregnancy-induced remission and the postpartum deterioration do exist, with some women reaching complete remission during pregnancy, whereas others have persistent active disease. This discrepancy was already noticed in two early case series in which a cortisol metabolite (that is, 17-hydroxycorticosteroid (17-OHCS)) was measured in pregnant RA women. High levels of 17-OHCS related to improvement of disease activity in only a subset of patients [19, 20]. This variation in clinical responses does not depend solely on the absolute levels of cortisol but might also be explained by differences in individual GC sensitivity.

In the healthy population, a considerable variation in GC sensitivity has been demonstrated by low-dose (0.25 mg) dexamethasone suppression tests and functional in vitro assays [21, 22]. In diseased states, these differences in GC sensitivity are reflected by a wide spectrum of GC therapy efficacy, which may partly be explained by four functional single nucleotide polymorphisms (SNPs) in the glucocorticoid receptor (GR) gene. The minor alleles of the polymorphisms N363S (rs6195) and BclI (rs41423247) are associated with a relative hypersensitivity to GC, whereas the ER22/23EK (rs6189 and rs6190) and 9β (rs6198) SNPs are associated with a relatively decreased GC sensitivity [23]. Previously, we demonstrated that carriers of the ER22/23EK variant more often had erosive disease and more frequently needed tumor necrosis factor-alpha (TNF-α) blocking therapy [24]. Similarly, these GR polymorphisms could explain differences in disease course during pregnancy and postpartum in RA.

Therefore, the aim of our study was to investigate the association between GR gene polymorphisms and level of disease activity and disease course during pregnancy and in the postpartum period in RA patients.

In this nationwide prospective study including 147 pregnant RA patients, we examined for the first time whether GR polymorphisms that modulate GC sensitivity are associated with the level of disease activity and disease course during pregnancy and the postpartum period. We show that GC treated patients in the GC-S group (that is, those with the BclI or N363S or both polymorphisms, associated with relatively increased GC sensitivity) have a significantly lower disease activity in the postpartum period than do patients in the GC-I group (9β or 9β + ER22/23EK, associated with relatively decreased GC sensitivity), as measured by the AUC of the DAS28. In patients not treated with GC, the level of disease activity and disease course during pregnancy or in the postpartum period does not seem to be influenced by differences in GR genotype.

Gestational-induced remission of RA has been recognized for a long time [27] and may in part be attributed to the increase in cortisol production that in turn enhances endogenous immunosuppression. Pregnancy is indeed considered to be a natural variant of hypercortisolism [28, 29] and serum (free) cortisol, urinary free cortisol, salivary cortisol, and cortisol content in hair all have been demonstrated to increase progressively during gestation, followed by a rapid postpartum decrease in cortisol levels [17, 18, 30–37].

Apart from cortisol availability, the ultimate biologic effects of GCs also depend on GC sensitivity, which is modulated by GR polymorphisms [23].

Based on the course of cortisol levels during pregnancy and after delivery, we hypothesized that differences in glucocorticoid sensitivity might in part explain why the beneficial effect of pregnancy on RA disease activity does not occur in all RA patients.

Polymorphisms of the GR gene have been demonstrated to influence disease course in several inflammatory disorders, including Graves ophthalmopathy [38], Crohn disease [39], and multiple sclerosis [40]. We recently demonstrated that the minor alleles of BclI and 9β were associated, respectively, with decreased and increased susceptibility to develop RA. In addition, ER22/23EK carriers had a worse disease phenotype and needed more frequent TNF-α blocking therapy [24]. We extend these data by demonstrating higher levels of disease activity in the postpartum period in GC treated patients in the GC-I group, despite the more frequent use of NSAIDs.

Interestingly, the differences in disease activity between carriers of GC-sensitive and GC-resistant polymorphisms were observed only in women treated with GCs. The GC treated patients involve a subgroup of women with high disease activity, as reflected by observed higher DAS28. Our observations may imply that in the postpartum phase, when endogenous cortisol levels decrease, patients with polymorphisms associated with increased GC sensitivity have more benefit from GC therapy. Therefore, in states of relative glucocorticoid deficiency, differences in GC sensitivity due to genetic variability may in part determine variations in disease activity. Conversely, in patients with low disease activity, as characterized by the absence of glucocorticoid therapy in our cohort, endogenous levels of cortisol apparently can prevent uncontrolled inflammatory processes independent of genetic variations of the GR gene, although we did not measure cortisol levels in our patients.

This concept of a "relative glucocorticoid deficiency" might also explain why the observed variation in disease activity seems to be restricted to the postpartum period, because Magiakou and co-workers [41] showed that hypothalamic CRH secretion in healthy pregnant women is transiently suppressed at 3 and 6 weeks, recovering only at 12 weeks postpartum. This suppression of the hypothalamic-pituitary-adrenal (HPA) axis in the postpartum period, which could be even more pronounced in RA in which a preexisting blunted HPA axis is described in nonpregnant states [42], might even further attenuate the ability of the HPA axis to produce sufficient levels of cortisol.

The clinical relevance of this blunted HPA axis in the first 3 months after childbirth is illustrated by a higher incidence or exacerbation of several autoimmune diseases, including postpartum depression, autoimmune thyroid disease, and rheumatoid arthritis itself [41, 43–46]. The lack of differences between GC-I and GC-S patients in disease activity during pregnancy could also be explained by altering levels of glucocorticoid sensitivity, as was suggested by Majzoub and co-workers [47, 48]. Alternatively, patients in the GC-I group tended to need higher daily dosages of GCs during pregnancy, which could have masked a higher level of disease activity in this subgroup of patients. Although we focused on glucocorticoids, absolute levels of estrogens and progesterone also increase progressively during gestation. Both estrogens and progesterone possess antiinflammatory properties and are therefore likely to have substantially influenced the disease course [49]. Similar to differences in GC sensitivity, one could speculate that variation in sensitivity to the immunosuppressing effects of estrogens and progesterone might also contribute to the wide clinical spectrum of changes in disease activity observed in pregnancy and after delivery in RA.

Interestingly, the difference in disease activity between GC-I and GC-S patients persisted during the entire postpartum follow-up period (that is, up to 26 weeks). Future studies should examine at which time points disease activity patterns of both groups converge to prepregnancy levels.

It should be noted that our study also has some limitations. First, genetic-association studies usually require larger numbers of patients. Although this is the largest prospectively studied cohort of pregnant RA patients, additional studies are needed to validate our findings. Second, the presented data are based on Caucasian patients only, who may differ from patients from other geographic areas with different genetic and environmental backgrounds. Third, parameters of HPA axis activity, not measured in this study, could have provided additional information in the non-GC treated patients.

Although the pattern of cortisol levels in pregnancy and after delivery has been extensively documented [17, 18, 30–37], large prospective studies evaluating cortisol levels along with clinical responses during pregnancy and postpartum in RA are currently lacking. Together with new insights in the past two decades supporting a blunted HPA axis in RA, this justifies renewed interest in the precise role of GC in pregnant RA patients and the course of disease [42, 50]. In this context, long-term indices of HPA axis activity, as measured by means of cortisol in hair, together with dynamic functional assays to assess GC sensitivity (that is, GR number, affinity of the GR receptor, and GR-mediated gene transcription) are promising techniques to unravel further the role of GCs and the precise contribution to pregnancy-associated alterations in disease activity in RA.

We demonstrate that differences in GC sensitivity, as determined by GR polymorphisms, might influence the level of disease activity in the postpartum period in GC treated women. The course of the disease itself does not seem to be associated with polymorphisms of the GR. In the light of the relatively small numbers of patients in each genotype group, however, our data should be regarded as an interesting new hypothesis possibly adding to the elucidation of the multifactorial mechanisms underlying pregnancy-induced amelioration and the postpartum flare, but the data do not necessarily prove the genetic association. Therefore, future (larger) studies should validate our hypothesis and examine both parameters of glucocorticoid availability and parameters of glucocorticoid sensitivity in relation to individual disease courses of pregnant RA patients.