ARTICLE OPEN ACCESS Menarche, pregnancies, and breastfeeding do not modify long-term prognosis in multiple sclerosis Marı́a I. Zuluaga, MD, Susana Otero-Romero, MD, PhD, Alex Rovira, MD, Santiago Perez-Hoyos, BSc, PhD, Correspondence Georgina Arrambide, MD, PhD, Laura Negrotto, MD, Ingrid Galán, MD, Jordi Rı́o, MD, PhD, Dr. Tintoré Manuel Comabella, MD, PhD, Carlos Nos, MD, Marı́a Jesús Arévalo, PhD, Angela Vidal-Jordana, MD, PhD, mtintore@cem-cat.org Joaquin Castilló, MD, PhD, Breogán Rodrı́guez, MD, Luciana Midaglia, MD, Patricia Mulero, MD, Raquel Mitjana, MD, Cristina Auger, MD, Jaume Sastre-Garriga, MD, PhD, Xavier Montalban, MD, PhD,* and Mar Tintoré, MD, PhD* ® Neurology 2019;92:e1507-e1516. doi:10.1212/WNL.0000000000007178 Abstract Objective To investigate the effect of menarche, pregnancies, and breastfeeding on the risk of developing multiple sclerosis (MS) and disability accrual using a multivariate approach based on a large prospective cohort of patients with clinically isolated syndrome (CIS). Methods A cross-sectional survey of the reproductive information of female participants in a CIS cohort was performed. We examined the relationship of age at menarche with the risk of clinically definite MS (CDMS), McDonald 2010 MS, and Expanded Disability Status Scale (EDSS) 3.0 and 6.0. The effect of pregnancy (before and after CIS) and breastfeeding in the risk of CDMS, McDonald 2010 MS, and EDSS 3.0 was also examined. Univariate and multivariate analyses were performed and findings were confirmed using sensitivity analyses and a pro- pensity score model. Results The data of 501 female participants were collected. Age at menarche did not correlate with age at CIS and was not associated with the risk of CDMS or EDSS 3.0 or 6.0. Pregnancy before CIS was protective for CDMS in the univariate analysis, but the effect was lost in the multivariate model and did not modify the risk of EDSS 3.0. Pregnancy after CIS was protective for both outcomes in univariate and multivariate analyses when pregnancy was considered a baseline variable, but the protective effect disappeared when analyzed as a time-dependent event. Breastfeeding did not modify the risk for the 3 outcomes. Conclusions These results demonstrate that menarche, pregnancies, and breastfeeding did not substantially modify the risk of CDMS or disability accrual using a multivariable and time-dependent approach. *These authors contributed equally as last authors. From the Centre d’Esclerosi Múltiple de Catalunya (Cemcat), Department of Neurology/Neuroimmunology (M.I.Z., S.O.-R., G.A., L.N., I.G., J.R., M.C., C.N., M.J.A., A.V.-J., J.C., B.R., L.M., P.M., J.S.-G., X.M., M.T.), and Unitat de RM, Servei de Radiologia (M.I.Z., A.R., L.N., R.M., C.A., X.M.), Hospital Universitari Vall d’Hebron, Universitat Autònoma de Barcelona; Unitat d’Estadı́stica i Bioinformàtica (UEB) (M.I.Z., S.P.-H., L.N., X.M.), Vall d’Hebron Institut de Recerca (VHIR), Barcelona, Spain; Servicio de Neurologı́a (M.I.Z., L.N., X.M.), Instituto Neurológico de Colombia, Medellı́n, Colombia; Raul Carrea Institute for Neurological Research (L.N., X.M.), FLENI, Buenos Aires, Argentina; and Division of Neurology (X.M.), St. Michael’s Hospital, University of Toronto, Canada. Go to Neurology.org/N for full disclosures. Funding information and disclosures deemed relevant by the authors, if any, are provided at the end of the article. The Article Processing Charge was funded by the authors. This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND), which permits downloading and sharing the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal. Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology. e1507 Glossary aHR = adjusted hazard ratio; BMI = body mass index; CDMS = clinically definite multiple sclerosis; CI = confidence interval; CIS = clinically isolated syndrome; DMT = disease-modifying treatment; EDSS = Expanded Disability Status Scale; HR = hazard ratio; MS = multiple sclerosis; OB = oligoclonal bands. Consistent data show that women worldwide are at higher 2007. The patients were seen on a regular basis for clinical risk for multiple sclerosis (MS), with an incidence sex ratio evaluation, which included Expanded Disability Status Scale of 2:1, which increased over the last 2 decades.1 However, (EDSS) evaluation and relapses. The patients who did not there is conflicting evidence on the role of sex in MS prog- attend 2 consecutive follow-up visits were defined as lost nosis. Large natural history studies reported a longer time to follow-up.4 from clinically isolated syndrome (CIS) to second attack and secondary progressive MS in women compared to men.2,3 Female patients from the previously described cohort were However, other studies did not observe a protective role of invited to participate in the MS Gender Project. Their female sex for the risk of second attack and only a mild effect clinical, demographic, CSF, and MRI data were obtained in MS prognosis using a multivariate analysis.4 from the prospectively collected data within the cohort. These patients also completed a self-administered survey The increased MS risk in women suggests differences in the that retrospectively collected reproductive information, ex- immune system or CNS between women and men, which ternal hormonal exposures, and other environmental risk may be caused by genetic dissimilarities, sex hormones, and factors (variables described in table 1). Participants were also environmental exposures. Nonhormonal factors may include asked to report their body size self-perception by selecting smoking, history of Epstein-Barr virus infection, vitamin D 1 of 9 body silhouettes on the Stunkard figure rating scale13 deficiency, high body mass index (BMI), and high sodium at the age of menarche and CIS, ranging from very thin to intake.5–8 Regarding hormonal factors, previous studies dem- extremely obese, which is a accepted as a proxy of measure onstrated that women with MS exhibited younger age at for BMI. menarche compared to healthy controls,9 and an earlier men- arche was associated with a younger age at CIS.10 The survey was completed from February 2015 until No- vember 2015 during the participants’ regular clinical assess- The short-term effect of pregnancy has been studied pre- ments using a tablet device or a printed version or via email viously; results demonstrate that relapse risk decreases during with the patient’s consent. An invitation to participate with pregnancy, but increases during the postpartum period due the link to access the survey online was sent. A reminder email to hormonal changes occurring during these periods.11,12 was sent up to 3 times if the patient did not fill out the survey However, the role of pregnancy and breastfeeding in the long- after the first email invitation. Any missing or ambiguous data term prognosis of MS remains controversial. were confirmed during clinical visits or telephone. This survey was validated in nonmedical field women and re- We initiated the BARCELONA MS&GENDER project to quired 5–10 minutes to complete. elucidate the role of hormonal factors in the modulation of MS prognosis. The present substudy investigated the role of Descriptive statistics were obtained for the demographic, menarche, pregnancies, and breastfeeding in MS prognosis. clinical, CSF, and MRI variables and specific reproductive information (i.e., age at menarche, pregnancies, and breast- feeding). Spearman correlation examined the relationships Methods of age at menarche with age at CIS and age at menarche with Study design age at MS. Given that a different disease onset (multifocal This study was an observational study based on an ongoing or with encephalopathy) and evolution with high frequency prospective CIS cohort. This cohort began in 1995 and of relapses in the first few years was described for pediatric- includes patients aged <50 years who exhibited a CIS sugges- onset MS,14 a separate analysis of age at menarche was per- tive of CNS demyelination not attributable to other dis- formed in the group of female participants who experienced eases with symptom onset within 3 months of the first clinical their CIS before age 18. evaluation. Clinical, demographic, CSF, and MRI information were prospectively collected using a prespecified protocol. The effect of menarche on the time to CIS for the studied Brain MRIs were performed at baseline and were repeated outcomes CDMS, McDonald 2010, EDSS 3.0 and 6.0) was after 12 months and every 5 years thereafter. A baseline spinal evaluated using a Cox regression analysis adjusted by age at CIS, cord MRI was only performed when patients presented topography, oligoclonal bands (OB), MRI, treatment, and body with spinal cord syndrome from 1995 to 2006, but these size at age at menarche. Confounders were selected based on data were collected for all patients with CIS beginning in the results of the studies previously performed in the same e1508 Neurology | Volume 92, Number 13 | March 26, 2019 Neurology.org/N using inverse probability weighting. These models controlled Table 1 Variables included in the survey for selection bias because participants who became pregnant Reproductive background could exhibit different prognoses. Weights calculated after fitting a logistic model to estimate the probability of becoming Menarche pregnant over follow-up included age, previous pregnancy, No. of pregnancies topography, OB, MRI lesions, DMT, and CDMS before Pregnancy outcomes pregnancy. Breastfeeding The effect of breastfeeding on the time to CDMS, McDonald Menopause 2010 MS, and EDSS 3.0 was evaluated using Cox regression analysis with a 3-categorical variable: nonpregnancy (reference Nonphysiologic hormonal exposure category), pregnancy without breastfeeding, and pregnancy Hormonal contraceptives with breastfeeding. First, we used a variable that classified Fertility treatment participants who were breastfeeding or not at any time during follow-up, and exposure to breastfeeding was considered Hormone replacement therapy present from baseline. Univariate and multivariate Cox models Environmental risk factors adjusted by age at CIS, topography, OB, MRI, and DMT Smoking prior to a second attack (as a time-dependent variable) were obtained. Second, we introduced breastfeeding into the model Body composition at menarche and CIS as a time-dependent covariate. A woman was considered 25OH vitamin D at CIS nonexposed to breastfeeding until the moment of breast- feeding and exposed thereafter (until outcome fulfilment or Cotinine levels at CIS censored). HR was interpreted as the excess or shortage of Abbreviations: CDMS = clinically definite multiple sclerosis; CIS = clinically risk of the studied outcomes for 2 women with the same isolated syndrome. follow-up time from CIS in this second analysis. Sensitivity analyses were performed using the calculated cohort.4 A subgroup analysis compared the age at menarche conception date (delivery date minus 9 months), and preg- between 2 different prognostic profile groups (high-risk and nancy and abortion after CIS were considered. low-risk female participants). The high-risk group (CIS- CDMS) included female participants who had an abnormal Standard protocol approvals, registrations, baseline MRI and positive OB and presented with a second and patient consents attack during the follow-up. The low-risk group (CIS-CIS) This study was performed with the approval of the Ethics included female participants who had a normal baseline Committee of Hospital Universitari Vall d’ Hebron. All patients MRI, negative OB, and no second attack during the fol- signed a written informed consent form from the inclusion in low-up. the ongoing CIS cohort. The effect of pregnancy (before and after CIS) on the time to Data availability CDMS, McDonald 2010 MS, and EDSS 3.0 and 6.0 was Anonymized data will be shared by request from any qualified evaluated using a Cox regression analysis. Pregnancy after investigator. CIS was analyzed using 2 methods. First, we used a dichoto- mic variable that classified female participants who experi- enced pregnancy or not at any time during follow-up, and exposure to pregnancy was considered present from baseline. Results Univariate and multivariate Cox models adjusted by age at A total of 1,196 patients were enrolled in the prospective CIS, topography, OB, MRI, and disease-modifying treatment CIS cohort study from January 1995 to November 2015. (DMT) prior to a second attack (as a time-dependent vari- Sixty-nine (6.1%) patients did not meet the general inclusion able) were obtained. Second, we introduced pregnancy and criteria of the cohort and were excluded (exceeded entry breastfeeding into the model as time-dependent covariates. window, previous attack, age over 50 years, and alternative A woman was considered nonexposed to pregnancy until diagnosis). the moment of delivery and exposed after pregnancy (until outcome fulfilment or censored). Hazard ratio (HR) was The female/male ratio of the remaining 1,127 patients was interpreted as the excess or shortage of risk of the studied 2:1, of which 771 (68.4%) were female. A total of 501 female outcomes (CDMS, McDonald 2010 MS, EDSS 3.0 and 6.0) participants completed the survey (65% return rate). Patients for 2 female participants with the same follow-up time from who filled out the survey had a similar age at CIS (31.1 vs 31 CIS in this second analysis. Finally, we confirmed our findings years, p = 0.84) but were more prone to exhibit an abnormal by fitting a propensity score model for pregnancy at any time baseline MRI (82% vs 59%, p < 0.005), were more frequently Neurology.org/N Neurology | Volume 92, Number 13 | March 26, 2019 e1509 on DMT (53% vs 16%, p < 0.005), and had a longer follow-up Menarche (9.5 vs 5.3 years, p < 0.005). During follow-up, a total of 53% The mean age at menarche was 12.5 years (SD 1.62), and of patients had a second attack; 61% met the McDonald 2010 the time from menarche to CIS was 18.5 (SD 8.0) years. MRI MS criteria, 13.4% reached confirmed EDSS 3.0, and Comparisons of demographic, clinical, and MRI data between 2.6% EDSS 6.0 (table 2). patients with age at menarche before or after age 12.5 years revealed that the only significant difference was the time from menarche to CIS (18.5 vs 19.7 years, p = 0.001). This difference was explained by the difference in age at menarche Table 2 Patient characteristics but not by the difference in age at CIS (30.9 vs 31.9 years, p = 0.62). We performed correlation tests between the age at Baseline characteristics Values menarche and age at CIS to further confirm this observation Age at CIS, y, mean (SD) (n = 492) 31.1 (7.8) and found no significant correlations (p = 0.94) between age at Age at menarche, y, mean (SD) (n = 501) 12.5 (1.6) menarche and age at CDMS (p = 0.6) or EDSS 3.0 (p = 0.2). Time from menarche to CIS, y, mean (SD) (n = 492) 18.5 (8) Age at menarche was also not related to the risk of second Initial symptom, n (%) attack (HR 1.03, 95% confidence interval [CI] [0.81–1.31], Optic neuritis 160 (32) p = 0.828), McDonald 2010 MS (HR 0.93, 95% CI [0.74–1.16], p = 0.513), or disability accrual (EDSS 3.0 HR Myelitis 147 (30) 1.09, 95% CI [0.67–1.77], p = 0.728; EDSS 6.0 HR 0.60, 95% Brainstem syndrome 135 (26) CI [0.20–1.85], p = 0.377), even after adjusting for age at CIS, topography, OB, MRI, treatment, and body size at menarche Other 59 (12) (table 3). Subgroup analysis comparing the age at menarche Oligoclonal bands, n (%) n = 428 between the high-risk (CIS-CDMS n = 181) and low-risk Positive 273 (64) (CIS-CIS n = 68) group revealed no significant differences (p = 0.281). Negative 155 (36) Basal MRI: T2 lesions, n(%) 481 A separate analysis of female participants with CIS under the age of 18 years (n = 16) demonstrated that the mean age at None 106 (22) CIS was 15.8 (SD 1.3) years. The age at menarche in this 1–3 70 (14.5) subgroup was 12.3 (SD 1.2) years, which was not different 4–9 68 (14) from female participants with CIS after 18 years of age (12.3 vs 12.5, p = 0.55). None of these patients had CIS before ≥10 237 (49) menarche, including the youngest patient at CIS, who was 12 Gadolinium enhancement, median (P25–P75) 0 (0–1) years old. She experienced her menarche at 11 years of age. Barkhof criteria, n (%) n = 489 Eleven (68.8%) patients had a second attack after a median of 12.2 (1.54–122) months. Thirteen (81%) girls met McDo- 0 Barkhof criteria 158 (32) nald 2010 MS after a median of 12.1 (1.54–179) months. One 1–2 Barkhof criteria 108 (22) patient reached EDSS 3.0 36 months after the CIS. 3–4 Barkhof criteria 223 (46) Pregnancy On DMT, n (%) 265 (23) A total of 302 (60%) of the 501 female participants who filled Weight, n (%) n = 501 out the survey had at least one pregnancy prior to a CIS or during follow-up. Comparison of nulliparous female partic- Underweight 280 (56) ipants vs female participants with at least one pregnancy Normal 192 (38) demonstrated that nulliparous women exhibited CIS at a younger age (28 vs 33 years, p < 0.005), had higher lesion T2 Overweight 29 (6) load in the basal MRI (10 vs 7, p = 0.02), and were more Study outcomes frequently on DMT (62.1% vs 49.3%, p < 0.005). CDMS during follow-up, n (%) 267 (53) Pregnancy prior to CIS McDonald MS during follow-up, n (%) 306 (61) A total of 191 (38%) female participants had at least one EDSS 3.0 during follow-up, n (%) 67 (13) pregnancy prior to their CIS, and the mean age at CIS in this EDSS 6.0 during follow-up, n (%) 13 (2.5) group was 37 years. Univariate analysis revealed that pregnancy prior to CIS was highly protective for CDMS (HR 0.62, 95% Abbreviations: CDMS = clinically definitive multiple sclerosis; CIS = clinically CI 0.47–0.80, p < 0.001) and McDonald 2010 MS (HR 0.68, isolated syndrome; DMT = disease-modifying treatment; EDSS = Expanded Disability Status Scale; MS = multiple sclerosis. 95% CI 0.53–0.86, p = 0.001). This effect was lost in the multivariate analysis when adjusting for age at CIS, topography, e1510 Neurology | Volume 92, Number 13 | March 26, 2019 Neurology.org/N Table 3 Cox regression analysis for age at menarche and clinically definitive multiple sclerosis (CDMS), McDonald 2010 MS, and Expanded Disability Status Scale (EDSS) 3.0 and 6.0 HR 95% CI p Value aHRa 95% CI p Value Time to CDMS Age at menarche, y <12.5 1.0 (Ref) 1.0 (Ref) ≥12.5 1.03 0.81–1.31 0.828 0.94 0.74–1.21 0.651 Time to McDonald 2010 Age at menarche, y <12.5 1.0 (Ref) 1.0 (Ref) ≥12.5 0.93 0.74–1.16 0.513 0.89 0.70–1.12 0.327 Time to EDSS 3.0 Age at menarche, y <12.5 1.0 (Ref) 1.0 (Ref) ≥12.5 1.09 0.67–1.77 0.728 1.15 0.70–1.92 0.579 Time to EDSS 6.0 Age at menarche, y <12.5 1.0 (Ref) 1.0 (Ref) ≥12.5 0.60 0.20–1.85 0.377 0.60 0.19–1.90 0.383 Abbreviations: aHR = adjusted hazard ratio for the multivariate model; CI = confidence interval; HR = hazard ratio; OB = oligoclonal bands. a Adjusted by age at clinically isolated syndrome, topography, OB, MRI, disease-modifying treatment, and body size at menarche. OB, MRI, and DMT for CDMS (adjusted HR [aHR] 1.09, after CIS reached EDSS 6 over follow-up. These results in- 95% CI 0.79–1.51, p = 0.589) and McDonald 2010 MS (aHR dicate that the moment at which a woman becomes pregnant 1.24, 95% CI 0.91–1.69, p = 0.17). Pregnancy prior to CIS did did not affect the development of the 3 outcomes. not modify the risk of attaining confirmed EDSS 3.0 in the univariate (HR 1.05, 95% CI 0.6–1.7, p = 0.837) or multivariate Propensity score analysis confirmed these results for CDMS (aHR 1.3, 95% CI 0.7–2.7, p = 0.392) analyses. (aHR 1.14, 95% CI 0.63–2.06, p = 0.66), McDonald 2010 MS (HR 1.71, 95% CI 0.88–3.33, p = 0.24), and EDSS 3.0 (aHR Pregnancy after CIS 1.03, 95% CI 0.43–2.50, p = 0.94) (figure 1). A total of 142 (28%) women had at least one pregnancy after the CIS, and the mean age at CIS in this group was 27 years. Breastfeeding For the 133 pregnancies that occurred after the CIS but Univariate analysis revealed a protective effect of pregnancy for before EDSS 3.0, 87 female participants breastfed. Differ- CDMS (HR 0.30, 95% CI 0.20–0.46, p < 0.001), McDonald ences in the risk of CDMS, McDonald 2010 MS, and EDSS 2010 MS (HR 0.22, 95% CI 0.14–0.35, p < 0.001), and con- 3.0 were compared to patients who became pregnant and did firmed EDSS 3.0 (HR 0.43, 95% CI 0.24–0.80, p = 0.0073). not breastfeed as a reference. No differences were observed between breastfeeding and nonbreastfeeding groups for all Multivariate analyses using pregnancy as a baseline variable 3 outcomes. The observed HRs were 1.41 (95% CI 0.61–3.23, revealed that the protective effect persisted for CDMS (aHR p = 0.42) for CDMS, 1.42 (95% CI 0.5–3.67, p = 0.46) for 0.25, 95% CI 0.16–0.38, p < 0.001), McDonald 2010 MS McDonald 2010 MS, and 1.21 (95% CI 0.37–3.93, p = 0.75) (aHR 0.23, 95% CI 0.14–0.37, p < 0.001), and disability for EDSS 3.0. accrual (aHR 0.45, 95% CI 0.23–0.89, p = 0.008). However, the protective effect disappeared for CDMS (aHR 1.13, 95% Multivariate analyses of breastfeeding vs nonbreastfeeding CI 0.70–1.82, p = 0.6), McDonald 2010 MS (HR 1.01, 95% groups revealed that no significant differences remained, but CI 0.60–1.73, p = 0.90), and confirmed EDSS 3.0 (HR 0.89, the direction of a marginal effect of breastfeeding appeared 95% CI 0.44–1.80, p = 0.97) when pregnancy was analyzed (CDMS: aHR 0.92, 95% CI 0.39–2.14, p = 0.84; McDonald as a time-dependent event. We could not conduct the survival 2010: aHR 0.73, 95% CI 0.27–1.96, p = 0.53; and EDSS 3.0: analysis for EDSS 6 given that none of the pregnant women aHR 0.85, 95% CI 0.25–2.84, p = 0.79). Neurology.org/N Neurology | Volume 92, Number 13 | March 26, 2019 e1511 Figure 1 Effect of pregnancy after clinically isolated syn- Figure 2 Effect of breastfeeding on the time to clinically drome (CIS) on the time to clinically definite definite multiple sclerosis (CDMS), McDonald multiple sclerosis (CDMS), McDonald 2010 MS, 2010 MS, and Expanded Disability Status Scale and Expanded Disability Status Scale (EDSS) 3.0 (EDSS) 3.0 CI = confidence interval; HR = hazard ratio; OB = oligoclonal bands. *Adjusted by age at clinically isolated syndrome (CIS), topography, OB, MRI, and disease- modifying treatment prior to a second attack (as a time-dependent varia- CI = confidence interval; HR = hazard ratio; OB = oligoclonal bands. *Adjusted by ble).**Adjusted by age at CIS, topography, OB, MRI, and disease-modifying age at CIS, topography, OB, MRI, and disease-modifying treatment prior to a sec- treatment prior to a second attack (as a time-dependent variable) and ond attack (as a time-dependent variable). **Adjusted by age at CIS, topography, breastfeeding as a time-dependent covariate. OB, MRI, and disease-modifying treatment prior to a second attack (as a time- dependent variable) and pregnancy as a time-dependent covariate. ***Propensity score model for pregnancy at any time using inverse probability weighting. Analysis of breastfeeding as a time-dependent event revealed Discussion that no significant differences remained for CDMS (aHR The effect of hormonal changes during the female re- 1.09, 95% CI 0.47–2.54, p = 0.84) or McDonald 2010 MS productive life in the short and long term prognosis of MS (HR 0.86, 95% CI 0.32–2.29, p = 0.77) and confirmed the have been previously addressed. EDSS 3.0 result (HR 0.97, 95% CI 0.29–3.26, p = 0.97) (figure 2). These results were confirmed in the sensitivity Regarding menarche, the evidence of the effect of age at analyses, which considered new T2 lesions during follow-up menarche on age at MS onset and the risk and prognosis of MS and time to CDMS (CDMS: HR 1.10, 95% CI 0.47–2.59, p = is inconclusive. It is known that puberty encompasses several 0.83; McDonald 2010: aHR 0.54, 95% CI 0.20–1.45, p = 0.22; years and begins in girls between 8 and 12 years. Menarche EDSS: HR 0.74, 95% CI 0.21–2.59, p = 0.64). represents the end of puberty and is regulated by genetic and e1512 Neurology | Volume 92, Number 13 | March 26, 2019 Neurology.org/N several environmental factors (BMI, diet, and exercise).15,16 It parity and the risk of disability accrual or secondary pro- has been hypothesized that early menarche or earlier sexual gressive MS.26,27 However, none of these studies considered maturation will involve more prolonged exposure to sexual the interaction of other known confounding factors, such as hormones, which might be related to an increased risk of MS. A age at CIS, CIS topography, OB, burden of T2 lesion, and the Canadian study reported an increased risk of MS in women use of DMT. with a younger age at menarche—menarche in controls oc- curred at 12.6 years compared to 12.4 years in patients with MS Postpartum is a critical time for women with MS. The pro- (p = 0.00017)—but did not find any association between age at tective effect of pregnancy is lost, and mothers must decide menarche and age at MS onset.9 Our study did not include whether to restart early treatment or breastfeed.28 Some controls and cannot address this specific question; however, we studies suggested that breastfeeding modestly reduces the risk did not find a relationship between age at puberty and age at of postpartum relapses29–32; others have reported that MS onset or long-term clinical course of the disease (risk of breastfeeding was not related to postpartum relapses, and second attack, McDonald MS, risk of disability) even after relapse rates before or during pregnancy were the main factors controlling for age at CIS, topography, OB, MRI, treatment, that determined the risk of relapse during the postpartum and body size self-perception at menarche. A complementary period.33 The potential protective effect of breastfeeding in subgroup analysis comparing the high-risk group (CIS-CDMS) MS prognosis may be at least partially explained by selection and the low-risk group (CIS-CIS) did not observe differences bias because women with milder disease activity and less in the age at menarche and age at CIS.10 disability would be more prone to breastfeed their children compared to patients with more active disease.28 Again, A separate analysis of the female participants in our cohort controlling for previously known prognostic factors and per- with a CIS before the age of 18 years was performed. None of forming time-dependent analyses may help to avoid selection these patients with pediatric onset had their CIS before bias and weight accurately the real effect of breastfeeding. menarche, and the age at menarche was not different from the rest of the group. However, a selection bias cannot be ex- Our study has 2 important strengths. The first strength is the cluded because patients under the age of 12 years were likely characteristics and number of patients included in this report. underrepresented in our cohort. Another potential limitation These female participants are part of a prospective, open CIS of our study is recall bias because the age of menarche was cohort with a long follow-up, and they are evaluated on collected retrospectively. Women may remember the year but a regular basis. Demographic, clinical, MRI, and treatment not the exact date of the menarche; however, previous studies information is recorded prospectively under a prespecified demonstrated good reliability of the approximated age at protocol. These high-quality data allow us to perform adjusted menarche.17 This should be taken into account since the analyses for possible confounding factors. previous studies reported significant differences based on a very small difference (months) of onset of menarche be- The second strength lies in the statistical analyses used; with tween the groups.9 It is therefore difficult to establish a causal our time-dependent approach, the HR reflects the effect of relationship between that small difference in age at menarche pregnancy from the moment in which the patient becomes reported in previous studies and the risk of such a complex pregnant instead of treating it like a baseline covariate. In disease as MS. To overcome this limitation in the future, we addition, breastfeeding was modeled analogously. In time- have started to collect this information prospectively. fixed models one assumes that the exposure (in this case, pregnancy or breastfeeding) is present since the beginning of The effect of pregnancy on the risk and evolution of MS has follow-up until the end of the course of the patient. Thus, been widely studied. Previous evidence supports a protective ignoring the time-dependent nature of an exposure could lead effect of pregnancies on risk and short-term prognosis, de- to time-dependent bias.34 A propensity score over the preg- creasing clinical inflammatory activity, especially during the nancy status was performed to avoid a probable selection bias. last trimester.20,21 For instance, a woman with a mild progression of the disease could be more likely to be pregnant than a woman with Higher parity has been associated with a decreased risk of a more aggressive course. This approach has been performed having a first clinical demyelinating event18,19 and pregnancy in other previous studies.35 However, in our study we have was associated with a decreased risk of disability progression considered not only clinical variables for the multivariate (EDSS 6.0) compared to a nulliparous status.20–24 The pro- analysis but biological (oligoclonal bands) and MRI data, and tective effect of pregnancy after CIS on long-term outcomes both factors are known to be important at the time to predict was recently addressed in a very large real-life cohort the course of MS.4 (MSBase). This study considered pregnancy as a baseline and not a time-dependent variable and found that female partic- The results of this study provide us with useful information for ipants with at least one pregnancy exhibited a lower EDSS women with MS regarding reproductive counseling. The score over 10 years after a propensity score adjustment.25 In main message is that the prognosis of MS will not be signif- contrast, this protective long-term effect was not found in icantly affected by pregnancy once all other variables are other studies that examined the possible association between considered. In this regard, the clinical activity, MRI findings, Neurology.org/N Neurology | Volume 92, Number 13 | March 26, 2019 e1513 and early initiation of treatment are the most important member of clinical trials or participated in advisory boards of prognostic factors that should be considered when addressing clinical trials in the past with Actelion, Amirall, Bayer, Biogen, reproductive decisions. Celgene, Genzyme, Hoffmann-La Roche, Novartis, Oryzon Genomics, Sanofi-Genzyme, and Teva Pharmaceutical. M. It is known that the risk of MS is multifactorial, and hormones Tintoré has received compensation for consulting services may play an important role in the risk of the disease, but this and speaking honoraria from Bayer Schering Pharma, Merck- effect may be diminished once the disease is established. Serono, Biogen-Idec, Teva Pharmaceuticals, Sanofi-Aventis, Future studies considering the exact time of menarche and the Novartis, Almirall, Genzyme, and Roche. Go to Neurology. characteristics of breastfeeding in the overall disease course org/N for full disclosures. are required to further establish the association between re- productive background and MS prognosis. Publication history Received by Neurology June 3, 2018. Accepted in final form November Study funding 21, 2018. This project was supported by FIS PI15/0070 from Ministry of Economy and Competitiveness of Spain, Fundación Gen- Appendix Authors zyme research grant, and ECTRIMS clinical fellowship awarded to Marı́a I. Zuluaga. Name Location Role Contribution Marı́a I. Centre d’Esclerosi Author Designed and Disclosure Zuluaga, Múltiple de Catalunya conceptualized MD (Cemcat); Department the study, analyzed M. Zuluaga has received compensation for consulting services of Neurology/ the data, drafted the from Novartis, Stendhal, Genzyme, Merck-Serono, and Bayer. Neuroimmunology, manuscript for Hospital Universitari intellectual content S. Otero-Romero reports no disclosures relevant to the man- Vall d’Hebron, uscript. A. Rovira serves on scientific advisory boards for Bio- Barcelona, Spain gen Idec, Novartis, Genzyme, and OLEA Medical, and on Susana Centre d’Esclerosi Author Designed and the editorial boards of the American Journal of Neuroradiology, Otero- Múltiple de Catalunya conceptualized Romero, (Cemcat); Department the study, analyzed Neuroradiology, and European Radiology; has received speaker PhD of Neurology/ the data, drafted the honoraria from Bayer, Genzyme, Sanofi-Aventis, Bracco, Neuroimmunology, manuscript for Merck-Serono, Teva Pharmaceutical Industries Ltd., OLEA Hospital Universitari intellectual content Vall d’Hebron, Medical, Stendhal, Novartis, and Biogen Idec; and has re- Barcelona, Spain search agreements with Siemens AG. S. Perez-Hoyos reports Alex Rovira, Unitat de RM, Servei Author Major role in the no disclosures relevant to the manuscript. G. Arrambide has MD de Radiologia, acquisition of data received compensation for consulting services from Biogen- Hospital Universitari Vall d’Hebron; Idec, research support from Novartis, and speaking hono- Universitat Autònoma raria from Sanofi-Aventis. L. Negrotto and I. Galán report no de Barcelona, Spain disclosures relevant to the manuscript. J. Rı́o has received S. Perez- Unitat d’Estadı́stica i Author Interpreted the data, speaking honoraria and personal compensation for partici- Hoyos Bioinformàtica (UEB); revised the Vall d’Hebron Institut manuscript for pating on advisory boards from Almirall, Bayer-Schering de Recerca (VHIR), intellectual content Healthcare, Biogen-Idec, Genzyme, Merck-Serono, Novartis, Barcelona, Spain Teva, and Sanofi-Aventis. M. Comabella has received com- Georgina Centre d’Esclerosi Author Major role in the pensation for consulting services and speaking honoraria from Arrambide, Múltiple de Catalunya acquisition of data, Bayer Schering Pharma, Merck Serono, Biogen-Idec, Teva MD, PhD (Cemcat); Department interpreted the data, of Neurology/ revised the Pharmaceuticals, Sanofi-Aventis, and Novartis. C. Nos has re- Neuroimmunology, manuscript for ceived funding for travel from Biogen Idec and F. Hoffmann-La Hospital Universitari intellectual content Vall d’Hebron, Roche, Ltd., and speaker honoraria from Novartis. M. Arevalo Barcelona, Spain reports no disclosures relevant to the manuscript. A. Vidal- Laura Centre d’Esclerosi Author Major role in the Jordana has received speaking honoraria and consulting Negrotto, Múltiple de Catalunya acquisition of data fees from Novartis, Roche, and Sanofi-Aventis. J. Castilló, MD (Cemcat); Department of Neurology/ B. Rodrı́guez-Acevedo, L. Midaglia, P. Mulero, and R. Mitjana Neuroimmunology, report no disclosures relevant to the manuscript. C. Auger Hospital Universitari Vall d’Hebron, has received speaking honoraria from Novartis, Biogen, and Barcelona, Spain Stendhal. J. Sastre-Garriga has received compensation for Ingrid Centre d’Esclerosi Author Major role in the participating on advisory boards, speaking honoraria, and Galán, MD Múltiple de Catalunya acquisition of data travel expenses for scientific meetings, consulting services, (Cemcat); Department of Neurology/ or research support from Celgene, Novartis, Biogen, Teva, Neuroimmunology, Merck, Almirall, and Genzyme. X. Montalban has received Hospital Universitari speaking honoraria and travel expenses for participation Vall d’Hebron, Barcelona, Spain in scientific meetings and has been a steering committee e1514 Neurology | Volume 92, Number 13 | March 26, 2019 Neurology.org/N Appendix (continued) Appendix (continued) Name Location Role Contribution Name Location Role Contribution Jordi Rı́o, Centre d’Esclerosi Author Major role in the Patricia Centre d’Esclerosi Author Major role in the MD, PhD Múltiple de Catalunya acquisition of data Mulero, MD Múltiple de Catalunya acquisition of data (Cemcat); Department (Cemcat); Department of Neurology/ of Neurology/ Neuroimmunology, Neuroimmunology, Hospital Universitari Hospital Universitari Vall d’Hebron, Vall d’Hebron, Barcelona, Spain Barcelona, Spain Manuel Centre d’Esclerosi Author Interpreted the data, Raquel Unitat de RM, Servei Author Major role in the Comabella, Múltiple de Catalunya revised the Mitjana, MD de Radiologia, acquisition of data MD, PhD (Cemcat); Department manuscript for Hospital Universitari of Neurology/ intellectual content Vall d’Hebron; Neuroimmunology, Universitat Autònoma Hospital Universitari de Barcelona, Spain Vall d’Hebron, Barcelona, Spain Cristina Unitat de RM, Servei Author Major role in the Auger, MD de Radiologia, acquisition of data Carlos Nos, Centre d’Esclerosi Author Interpreted the data, Hospital Universitari MD Múltiple de Catalunya revised the Vall d’Hebron; (Cemcat); Department manuscript for Universitat Autònoma of Neurology/ intellectual content de Barcelona, Spain Neuroimmunology, Hospital Universitari Jaume Centre d’Esclerosi Author Interpreted the data, Vall d’Hebron, Sastre- Múltiple de Catalunya revised the Barcelona, Spain Garriga, (Cemcat); Department manuscript for MD, PhD of Neurology/ intellectual content Marı́a Jesús Centre d’Esclerosi Author Interpreted the data, Neuroimmunology, Arévalo, Múltiple de Catalunya revised the Hospital Universitari PhD (Cemcat); Department manuscript for Vall d’Hebron, of Neurology/ intellectual content Barcelona, Spain Neuroimmunology, Hospital Universitari Xavier Centre d’Esclerosi Author Interpreted the data, Vall d’Hebron, Montalban, Múltiple de Catalunya revised the Barcelona, Spain MD, PhD (Cemcat); Department manuscript for of Neurology/ intellectual content Angela Centre d’Esclerosi Author Major role in the Neuroimmunology, Vidal- Múltiple de Catalunya acquisition of data Hospital Universitari Jordana, (Cemcat); Department Vall d’Hebron, MD, PhD of Neurology/ Barcelona, Spain Neuroimmunology, Hospital Universitari Mar Centre d’Esclerosi Author Designed and Vall d’Hebron, Tintoré, Múltiple de Catalunya conceptualized Barcelona, Spain MD, PhD (Cemcat); Department study, analyzed the of Neurology/ data, drafted the Joaquin Centre d’Esclerosi Author Major role in the Neuroimmunology, manuscript for Castilló, Múltiple de Catalunya acquisition of data Hospital Universitari intellectual content MD, PhD (Cemcat); Department Vall d’Hebron, of Neurology/ Barcelona, Spain Neuroimmunology, Hospital Universitari Vall d’Hebron, Barcelona, Spain References Breogán Centre d’Esclerosi Author Major role in the 1. Trojano M, Lucchese G, Graziano G, et al. Geographical variations in sex ratio trends Rodrı́guez, Múltiple de Catalunya acquisition of data over time in multiple sclerosis. PLoS One 2012;7:e48078. MD (Cemcat); Department 2. Koch M, Kingwell E, Rieckmann P, Tremlett H; UBC MS Clinic Neurologists. The of Neurology/ natural history of secondary progressive multiple sclerosis. J Neurol Neurosurg Neuroimmunology, Psychiatry 2010;81:1039–1043. Hospital Universitari 3. Debouverie M. Gender as a prognostic factor and its impact on the incidence of Vall d’Hebron, multiple sclerosis in Lorraine, France. J Neurol Sci 2009;286:14–17. Barcelona, Spain 4. Tintore M, Rovira À, Rı́o J, et al. Defining high, medium and low impact prognostic factors for developing multiple sclerosis. Brain J Neurol 2015;138:1863–1874. Luciana Centre d’Esclerosi Author Major role in the 5. Belbasis L, Bellou V, Evangelou E, Ioannidis JPA, Tzoulaki I. Environmental risk Midaglia, Múltiple de Catalunya acquisition of data factors and multiple sclerosis: an umbrella review of systematic reviews and meta- MD (Cemcat); Department analyses. Lancet Neurol 2015;14:263–273. of Neurology/ 6. Kavak KS, Teter BE, Hagemeier J, Zakalik K, Weinstock-Guttman B; New York State Neuroimmunology, Multiple Sclerosis Consortium. Higher weight in adolescence and young adulthood is Hospital Universitari associated with an earlier age at multiple sclerosis onset. Mult Scler 2015;21:858–865. Vall d’Hebron, 7. Farez MF, Fiol MP, Gaitán MI, Quintana FJ, Correale J. Sodium intake is associated Barcelona, Spain with increased disease activity in multiple sclerosis. J Neurol Neurosurg Psychiatry 2015;86:26–31. Neurology.org/N Neurology | Volume 92, Number 13 | March 26, 2019 e1515 8. Fitzgerald KC, Munger KL, Hartung HP, et al. Sodium intake and multiple sclerosis 22. D’hooghe MB, Nagels G, Uitdehaag BMJ. Long-term effects of childbirth in MS. activity and progression in BENEFIT. Ann Neurol 2017;82:20–29. J Neurol Neurosurg Psychiatry 2010;81:38–41. 9. Ramagopalan SV, Valdar W, Criscuoli M, et al. Age of puberty and the risk of multiple 23. Ramagopalan S, Yee I, Byrnes J, Guimond C, Ebers G, Sadovnick D. Term preg- sclerosis: a population based study. Eur J Neurol 2009;16:342–347. nancies and the clinical characteristics of multiple sclerosis: a population based study. 10. Sloka JS, Pryse-Phillips WEM, Stefanelli M. The relation between menarche and the J Neurol Neurosurg Psychiatry 2012;83:793–795. age of first symptoms in a multiple sclerosis cohort. Mult Scler 2006;12:333–339. 24. Masera S, Cavalla P, Prosperini L, et al. Parity is associated with a longer time to reach 11. Confavreux C, Hutchinson M, Hours MM, Cortinovis-Tourniaire P, Moreau T. Rate irreversible disability milestones in women with multiple sclerosis. Mult Scler 2015; of pregnancy-related relapse in multiple sclerosis: Pregnancy in Multiple Sclerosis 21:1291–1297. Group. N Engl J Med 1998;339:285–291. 25. Jokubaitis VG, Spelman T, Kalincik T, et al. Predictors of long-term disability accrual 12. Vukusic S, Hutchinson M, Hours M, et al. Pregnancy and multiple sclerosis (the in relapse-onset multiple sclerosis. Ann Neurol 2016;80:89–100. PRIMS study): clinical predictors of post-partum relapse. Brain J Neurol 2004;127: 26. Koch M, Uyttenboogaart M, Heersema D, Steen C, De Keyser J. Parity and secondary 1353–1360. progression in multiple sclerosis. J Neurol Neurosurg Psychiatry 2009;80:676–678. 13. Stunkard AJ, Sørensen T, Schulsinger F. Use of the Danish Adoption Register for the 27. D’Amico E, Leone C, Patti F. Offspring number does not influence reaching the study of obesity and thinness. Res Publ Assoc Res Nerv Ment Dis 1983;60:115–120. disability’s milestones in multiple sclerosis: a seven-year follow-up study. Int J Mol Sci 14. Huppke B, Ellenberger D, Rosewich H, Friede T, Gärtner J, Huppke P. Clinical 2016;17:234. presentation of pediatric multiple sclerosis before puberty. Eur J Neurol 2014;21: 28. Airas L, Jalkanen A, Alanen A, Pirttilä T, Marttila RJ. Breast-feeding, postpartum and 441–446. prepregnancy disease activity in multiple sclerosis. Neurology 2010;75:474–476. 15. Waubant E. Effect of puberty on multiple sclerosis risk and course. Mult Scler 2018; 29. Hellwig K, Rockhoff M, Herbstritt S, et al. Exclusive breastfeeding and the effect on 24:32–35. postpartum multiple sclerosis relapses. JAMA Neurol 2015;72:1132–1138. 16. Chitnis T, Graves J, Weinstock-Guttman B, et al. Distinct effects of obesity and 30. Pakpoor J, Disanto G, Lacey MV, Hellwig K, Giovannoni G, Ramagopalan SV. puberty on risk and age at onset of pediatric MS. Ann Clin Transl Neurol 2016;3: Breastfeeding and multiple sclerosis relapses: a meta-analysis. J Neurol 2012;259: 897–907. 2246–2248. 17. Castilho SD, Nucci LB, Assuino SR, Hansen LO. The importance of memory bias in 31. Hellwig K, Haghikia A, Rockhoff M, Gold R. Multiple sclerosis and pregnancy: obtaining age of menarche by recall method in Brazilian adolescents [in Portuguese]. experience from a nationwide database in Germany. Ther Adv Neurol Disord 2012;5: Arq Bras Endocrinol Metabol 2014;58:394–397. 247–253. 18. Ponsonby AL, Lucas RM, van der Mei IA, et al. Offspring number, pregnancy, and risk 32. Langer-Gould A, Huang SM, Gupta R, et al. Exclusive breastfeeding and the risk of a first clinical demyelinating event: the AusImmune Study. Neurology 2012;78: of postpartum relapses in women with multiple sclerosis. Arch Neurol 2009;66: 867–874. 958–963. 19. Nielsen NM, Jørgensen KT, Stenager E, et al. Reproductive history and risk of 33. Portaccio E, Ghezzi A, Hakiki B, et al. Breastfeeding is not related to postpartum multiple sclerosis. Epidemiology 2011;22:546–552. relapses in multiple sclerosis. Neurology 2011;77:145–150. 20. Runmarker B, Andersen O. Pregnancy is associated with a lower risk of onset and 34. Munoz-Price LS, Frencken JF, Tarima S, Bonten M. Handling time-dependent var- a better prognosis in multiple sclerosis. Brain 1995;118:253–261. iables: antibiotics and antibiotic resistance. Clin Infect Dis 2016;62:1558–1563. 21. Verdru P, Theys P, D’Hooghe MB, Carton H. Pregnancy and multiple sclerosis: the 35. Karp I, Manganas A, Sylvestre M-P, Ho A, Roger E, Duquette P. Does pregnancy alter influence on long term disability. Clin Neurol Neurosurg 1994;96:38–41. the long-term course of multiple sclerosis? Ann Epidemiol 2014;24:504–508. e1516 Neurology | Volume 92, Number 13 | March 26, 2019 Neurology.org/N