We found that postmenopausal status in women with diffuse SSc was associated with a substantially lower mean mRSS (by −2.62 units, 95% CI −4.44, −0.80) compared to premenopausal status. This effect was independent of age, follow-up time, and disease duration. Our findings are supported by previous experimental evidence showing that estrogen increases extracellular matrix protein production in skin fibroblast culture of SSc patients [9]. In addition, observational studies of healthy women showed skin thinning in postmenopausal women compared to women who had not yet reached their menopause [10]. Furthermore, it is well-established that estrogen stimulates normal skin fibroblasts to produce transforming growth factor-beta 1 (TGF-beta 1), as well as monocytes and macrophages to produce platelet-derived growth factor (PDGF). Both TGF-beta 1 and PDGF are key profibrotic cytokines in SSc skin disease [1, 16]. As SSc skin fibroblasts show increased expression of TGF-beta 1 receptor and PDGF receptor, estrogen might play a role in SSc pathogenesis through its stimulatory effect on these two cytokines.
The effect of postmenopausal status on skin thickening was smaller in women with limited SSc compared to women with diffuse SSc, as evidenced by the interaction term between menopausal status and disease subtype. One potential explanation for the differential effect of menopause on skin thickening in women with diffuse and limited SSc might reside in the disease subtypes definition. Indeed, disease subtypes are defined according to the extent and location of skin involvement (for example, in the limited subtype skin is affected in the hands, feet, forearms, and/or face) [17]. Thus, by definition, the mRSS is usually lower in subjects with limited SSc than in those with diffuse SSc, with different maximum scores (that is, respectively 27 and 51 units). This might explain at least in part the smaller (absolute) effect size of menopausal status. Indeed, when contrasted to the baseline mean mRSS, the (relative) effect size of menopausal status was comparable between diffuse and limited subtypes, representing respectively 15% and 10% of the subtype-specific mean baseline mRSS.
As expected and shown in prior studies [2], we observed that disease subtype was the strongest predictor of skin thickening in our cohort, with diffuse SSc being associated with a higher mean mRSS of at least 10 units. We also found more severe skin thickening in women exposed to DMARDs. However, this is likely to have reflected confounding by disease severity (also known as confounding by indication), which occurs when a medication is preferentially prescribed to a group of patients with a worse baseline prognosis [18]. Indeed, SSc subjects with rapidly progressing and/or more extensive skin thickening are more likely to have higher mRSS and be aggressively treated with DMARDs, potentially resulting in a seemingly positive effect of DMARD exposure on the mRSS.
Although not statistically significant, women exposed to OCP and HRT appeared to have less skin thickening than unexposed women. Based on available evidence showing that exogenous estrogen increases skin thickness in normal women, we would have expected the opposite [10, 12]. Still, this finding could also represent confounding by disease severity, in which case, women with less severe skin disease and fewer disease complications (such as pulmonary hypertension or cardiovascular disease) would be more likely to be sexually active (and be on OCP) or have fewer contraindications to OCP and/or HRT.
The magnitude of the effect of postmenopausal status on the mRSS in women with diffuse SSc is very interesting, considering that menopause occurs in all women and SSc predominantly affects women. Very few exposures, including medications, seem to alter the course of skin thickening in SSc. The minimally important difference in the mRSS has been established at 3.2 units, based on a prior randomized controlled trial of d-penicillamine in subjects with diffuse disease [19]. As the CSRG cohort is not an inception cohort, women included in our study had long disease duration. Since it is well-established that skin thickening progresses more rapidly in early than in late disease [1, 2], inclusion of women with longer disease duration might have limited our ability to observe a larger effect of menopausal status on skin thickening. This was evidenced by our subsample analysis of women with disease duration shorter than 5 years, showing a larger effect of postmenopausal status on the mRSS in both disease subtypes. Notably, the magnitude of the postmenopausal effect in diffuse SSc (−3.36 units, 95% CI −5.87, −0.85) exceeded the minimally important difference in the mRSS [19].
Medsger et al. described the natural evolution of skin thickening in SSc subjects [20]. After rapid skin thickening progression in the first 5 years of the disease, in limited SSc, skin thickening slows for a few years to reach a plateau after 10 years of evolution, whereas in diffuse SSc, skin thickening actually regresses, albeit not to the pre-disease state [20]. Interestingly, in our study, the effect estimates for the covariates, time since cohort entry (that is, follow-up time) and disease duration at baseline, reflected this evolution/involution. Indeed, time since cohort entry had an overall positive effect on skin thickening (that is, each additional year of follow up being associated with an average increase of 0.19 unit (95% CI 0.04, 0.34)), because with time skin thickening increases more than it regresses. However, disease duration at baseline had an overall negative effect on the mRSS (that is, each additional year of disease duration at baseline being associated with a lower mean mRSS by −0.05 unit (95% CI −0.07, −0.03)), as skin thickness peaks in women with early disease. Moreover, the negative effect of menopausal status on the mRSS that was objectivated in our study might explain, at least in part, the natural plateau and/or involution of skin thickening reported by Medsger et al., which could have coincided with menopause onset in some subjects [20]. In addition, we did not find an association between age at disease onset and skin thickening after accounting for menopausal status in multivariate analysis, suggesting that menopausal status might be a better predictor of skin thickening compared to age at diagnosis.
Our study has some potential limitations. We used the total mRSS as the outcome measure, which assesses both the extent and degree of skin thickening. Thus, we could not distinguish the association between menopausal status and the extent of skin thickening from the association between menopausal status and the degree of skin thickening. In addition, we were unable to definitively demonstrate an effect of smoking on skin thickening. Previous work from our group has shown the challenges of appropriately modeling smoking exposure in health outcome analyses of smoking effect in SSc patients, and how these analyses can suffer from the healthy smoker effect and/or causality bias, making the smoking effect-estimate more conservative than it should be (that is, biasing the effect estimate towards the null) [21]. Still, this is unlikely to have affected our main effect-estimate as smoking did not appear to be a strong confounder of the effect of menopausal status on skin thickening as shown in the univariate and multivariate analyses. Another potential limitation is that menopausal status was ascertained annually by self-report and not confirmed by any laboratory investigation. However, a previous study has shown high validity of self-reported menopausal status and suggested self-report as a sufficiently accurate measure of menopause in observational studies [22].