OUP user menu

Pregnancy and multiple sclerosis (the PRIMS study): clinical predictors of post‐partum relapse

Sandra Vukusic, Michael Hutchinson, Martine Hours, Thibault Moreau, Patricia Cortinovis‐Tourniaire, Patrice Adeleine, Christian Confavreux
DOI: http://dx.doi.org/10.1093/brain/awh152 1353-1360 First published online: 6 May 2004

Summary

The influence of pregnancy in multiple sclerosis has been a matter of controversy for a long time. The Pregnancy in Multiple Sclerosis (PRIMS) study was the first large prospective study which aimed to assess the possible influence of pregnancy and delivery on the clinical course of multiple sclerosis. We report here the 2‐year post‐partum follow‐up and an analysis of clinical factors which might predict the likelihood of a relapse in the 3 months after delivery. The relapse rate in each trimester up to the end of the second year post‐partum was compared with that in the pre‐pregnancy year. Clinical predictors of the presence or absence of a post‐partum relapse were analysed by logistic regression analysis. Using the best multivariate model, women were classified as having or not having a post‐partum relapse predicted, and this was compared with the observed outcome. The results showed that, compared with the pre‐pregnancy year, there was a reduction in the relapse rate during pregnancy, most marked in the third trimester, and a marked increase in the first 3 months after delivery. Thereafter, from the second trimester onwards and for the following 21 months, the annualized relapse rate fell slightly but did not differ significantly from the relapse rate recorded in the pre‐pregnancy year. Despite the increased risk for the 3 months post‐partum, 72% of the women did not experience any relapse during this period. Confirmed disability continued to progress steadily during the study period. Three indices, an increased relapse rate in the pre‐pregnancy year, an increased relapse rate during pregnancy and a higher DSS (Kurtzke’s Disability Status Scale) score at pregnancy onset, significantly correlated with the occurrence of a post‐partum relapse. Neither epidural analgesia nor breast‐feeding was predictive. When comparing the predicted and observed status, however, only 72% of the women were correctly classified by the multivariate model. In conclusion, the results for the second year post‐partum confirm that the relapse rate remains similar to that of the pre‐pregnancy year, after an increase in the first trimester following delivery. Women with greater disease activity in the year before pregnancy and during pregnancy have a higher risk of relapse in the post‐ partum 3 months. This is, however, not sufficient to identify in advance women with multiple sclerosis who are more likely to relapse, especially for planning therapeutic trials aiming to prevent post‐partum relapses.

  • multiple sclerosis; pregnancy; post‐partum; relapse; prognosis
  • DSS = Kurtzke’s Disability Status Scale; PRIMS = Pregnancy in Multiple Sclerosis Study

Introduction

The influence of pregnancy in multiple sclerosis has been a matter of controversy for a long time. For many years, women with multiple sclerosis were discouraged from contemplating pregnancy (Douglass and Jorgensen, 1948; Tillman, 1950; Sweeney, 1953; Hutchinson, 1993; Sadovnick et al., 1994; Rudick, 1995) due to the possible deleterious effect of pregnancy on multiple sclerosis. Moreover, the pattern of progression of neurological disability over the years can affect family planning (Korn‐Lubetzki et al., 1984). Korn‐Lubetzki and colleagues found, in a large retrospective study, that the frequency of relapses of multiple sclerosis decreased during pregnancy, increased in the immediate post‐partum period and was similar to the frequency observed out of pregnancy when considering the pregnancy year (9 months of pregnancy and the first 3 months post‐partum).

The Pregnancy in Multiple Sclerosis (PRIMS) study was the first large prospective study of the natural history of multiple sclerosis in pregnant women (Confavreux et al., 1998). This observational, European, multicentre study aimed to assess the influence of pregnancy and delivery on the clinical course of multiple sclerosis and showed a significant decline in the relapse rate during pregnancy, most marked in the third trimester, with a rebound increase in the first 3 months post‐partum. There was no apparent acceleration of disability during the puerperium and neither breast‐feeding nor epidural analgesia had any specific adverse effect.

The PRIMS study participants continued to collect data on the cohort of 227 women who had a pregnancy with a live infant, for a total of up to 2 years post‐partum, as initially designed in the protocol. A further analysis was performed of putative clinical predictors which might identify women who were more likely to relapse after delivery and which might be useful to the neurologist advising a pregnant woman with multiple sclerosis regarding prophylactic therapies.

The aims of this paper are to report the 2‐year post‐partum follow‐up and to analyse the factors predictive of relapse in the 3 months after delivery.

Patients and methods

The study population consisted of 227 women who had full‐term delivery of a live infant and who had multiple sclerosis for at least 1 year prior to conception. If a woman had been followed in the PRIMS study for more than one pregnancy, only the first one was considered for analysis. Details of methods can be found in the previous report (Confavreux et al., 1998). The diagnosis of multiple sclerosis was classified according to Poser’s criteria (Poser et al., 1983). A relapse of multiple sclerosis was defined as the appearance, reappearance or worsening of symptoms of neurological dysfunction lasting >24 h. Fatigue alone was not considered as a relapse. Disability was assessed with Kurtzke’s Disability Status Scale (DSS) (Kurtzke, 1961), which is based on data from the neurological examination and the patient’s ability to walk. Scores range from 0 (no abnormality) to 10 (death caused by multiple sclerosis). Confirmed neurological disability was defined as the minimal level of persistent disability recorded on two consecutive examinations at least 3 months apart, excluding any transient worsening of disability related to relapses.

All women had been pregnant for at least 4 weeks but <36 weeks at entry into the study. Using the European Database for Multiple Sclerosis (EDMUS) system (Confavreux et al., 1992), the data were analysed at the EDMUS Coordinating Centre in Lyon, France. The ethics committee of St Vincent’s University Hospital, Dublin, Ireland, approved the study protocol and the women gave informed consent. Demographic data and data on the course of multiple sclerosis prior to entry into the study were recorded and patients were examined at 20, 28 and 36 weeks of gestation and 3, 6 12, 18 and 24 months post‐partum. Telephone assessments were made at 40 weeks of gestation and 1, 9, 15 and 21 months post‐partum. The DSS was assessed retrospectively 1 year prior to pregnancy and at the beginning of the pregnancy, and prospectively at entry to the study, at 36 weeks of gestation, and 3, 6, 12, 18 and 24 months post‐partum. Epidural analgesia and breast‐feeding were recorded.

The relapse rates per woman per year during each 3‐month period during pregnancy and the 2 years post‐partum were compared with the relapse rate during the year before pregnancy by means of paired, two‐sided t‐tests.

The effects of epidural analgesia and breast‐feeding on the course of multiple sclerosis were analysed by univariate logistic regression, as were other clinical variables thought to be related to the presence or absence of a relapse in the first trimester after delivery. These variables were also tested in different multivariate logistic models. Women were then classified according to the results of the best model, with a cut‐off value of 0.5, as having a post‐partum relapse predicted or not, and compared with the observed status. All statistical analyses were performed with SPSS for Windows software version 10.1 (SPSS, 2001).

Results

Data collection

Seventy six per cent of the 227 women were known to their neurologist before entry to the study. Forty two per cent entered the study in the first trimester, 43% in the second trimester and 15% in the third trimester of pregnancy. The follow‐up was at least 6 months post‐partum in 216 women (95.2%), 12 months in 201 (88.5%), 18 months in 182 (80.2%) and 24 months in 163 women (71.8%). Data on epidural analgesia and breast‐feeding were available in 222 and 210 women respectively. Out of these, 42 women (18.9%) chose to have epidural analgesia and 124 (54.6%) to breast‐feed their child.

Only five patients received disease‐modifying drugs in the 6 months after delivery (four azathioprine and one mitoxantrone) and eleven in the second 6‐month period (three azathioprine, six interferon β1b, one mitoxantrone and one cyclophosphamide). Data on immunoactive treatments in the second year post‐partum are not available, as they were not mandatory in the study protocol.

Relapse rate

The annualized relapse rate for each 3‐month period is given in Table 1 and Fig. 1. As previously reported, in comparison with the pre‐pregnancy year, in which the mean ± SD relapse rate was 0.7 ± 0.9, there was a reduction of the relapse rate during pregnancy, most marked in the third trimester (0.2 ± 0.9, P < 0.001), with a marked increase in the first 3 months after delivery (1.2 ± 2.0, P < 0.001). Thereafter, from the second post‐partum trimester onwards and for the following 21 months, the annualized relapse rate fell slightly but did not differ significantly from the annualized relapse rate recorded in the pre‐pregnancy year. Despite the clear increased risk of relapse in the post‐partum 3 months, 160 women (72%) did not experience any relapse during this period.

Figure

Fig. 1 Annualized relapse rate in the year before pregnancy, during pregnancy and in the two years after delivery among 227 women with MS (vertical bars represent means and 95% confidence intervals).

View this table:
Table 1

Relapses in the year before pregnancy, during pregnancy and in the 2 years after delivery

TrimesterNo. of womenNo. of relapsesNo. of women with relapsesAnnualized relapse rateP value*
Mean95% CI
Before pregnancy
 1st trimester22736340.60.5–0.7
 2nd trimester22741380.70.6–0.8
 3rd trimester22742420.70.6–0.9
 4th trimester22745400.80.7–0.9
 Year before pregnancy2271641130.70.6–0.8
Pregnancy
 1st trimester22727270.50.4–0.60.03
 2nd trimester22732290.60.5–0.70.17
 3rd trimester22712110.20.2–0.3<0.001
After delivery
 1st trimester22368631.21.1–1.4<0.001
 2nd trimester21547430.90.8–1.00.17
 3rd trimester20446430.90.8–1.00.15
 4th trimester20131290.60.5–0.70.59
 5th trimester19024220.50.4–0.60.08
 6th trimester18223220.50.4–0.60.11
 7th trimester17022180.50.4–0.60.13
 8th trimester16320190.50.4–0.60.04

CI = confidence interval. *P values for the comparison with the relapse rate in the year before pregnancy.

The annualized relapse rate of the pregnancy year, defined as encompassing the nine months of pregnancy and the first 3‐month period after delivery, was similar (0.6 ± 0.8) to that of the pre‐pregnancy year.

Confirmed disability

Confirmed disability progressed steadily during the whole study period. Pregnancy, delivery and post‐partum did not seem to have any effect on disability progression (Fig. 2).

Figure

Fig. 2 Mean confirmed disability, according to Kurtzke’s Disability Status Scale (DSS), in the year before pregnancy, during pregnancy and in the 2 years after delivery among 227 women with multiple sclerosis (vertical bars represent means and 95% confidence intervals).

Indicators of the occurrence of a relapse in the first 3‐month period after delivery

The distribution of women according to the presence or absence of a relapse in the year before pregnancy, during pregnancy and in the first 3 months after delivery is reported in Fig. 3. The likelihood of having at least one relapse during the first 3 months post‐partum according to demographic and disease‐related factors was estimated by a logistic regression analysis (Table 2). Only three factors significantly correlated with the presence of a post‐partum relapse. There was a 1.7‐fold increased risk with each relapse experienced during the pre‐pregnancy year and 1.8‐fold increased risk for each relapse during pregnancy (P < 0.002 and 0.02 respectively). Patients who had a higher DSS at pregnancy onset were also more likely to have a post‐partum relapse (odds ratio = 1.3, P = 0.04). The occurrence of a post‐partum relapse was not related to breast‐feeding, epidural analgesia, age at multiple sclerosis onset, age at pregnancy onset, disease duration, total number of relapses before pregnancy, number of previous pregnancies or child gender.

Figure

Fig. 3 Flow‐chart of patients regarding the presence or absence of one or more relapses in the pre‐pregnancy year, during pregnancy and in the 3‐month period after delivery among 227 women with multiple sclerosis.

View this table:
Table 2

Indicators of the occurrence of at least one relapse during the 3‐month period after delivery

Explanatory variable n Odds ratio*95% CI P
No. of relapses in pre‐pregnancy year2231.71.2–2.30.002
No. of relapses during pregnancy2231.81.1–3.00.02
DSS score at pregnancy onset2071.31.0–1.50.04
Epidural analgesia
 No177Ref.
 Yes411.40.7–2.90.37
Breast‐feeding208
 No85Ref.
 Yes1230.80.4–1.40.37
No. of relapses before pregnancy2101.11.0–1.20.10
MS duration (years)2231.01.0–1.10.21
Age at MS onset (years)2231.00.9–1.10.46
Age at pregnancy onset (years)2231.00.9–1.10.66
No. of previous pregnancies2161.2–1.60.24
Child sex219
 Male115Ref.
 Female1070.80.4–1.30.29

Univariate logistic regression analysis for 223 women with multiple sclerosis. *Odds ratio per increase of one relapse, 1 point in the DSS or 1 year according to the explanatory variable concerned. CI = confidence interval; DSS = Kurtzke’s Disability Status Scale; MS = multiple sclerosis.

All the variables were examined in the multivariate analysis, whatever the result of the univariate analysis. In the best multivariate model, only the number of relapses in the pre‐pregnancy year, the number of relapses during pregnancy and disease duration at pregnancy onset correlated independently with the presence of a post‐partum relapse (Table 3). The area under the ROC curve corresponding to this multivariate analysis was 0.72, which is considered to be acceptable discrimination. This mathematical model makes it possible to classify women according to their baseline characteristics, in relation to the probability of experiencing a relapse in the first trimester after delivery. The model predicted a relapse for only 16 (7%) of the 223 women, but eight of these women did not relapse in the first 3 months post‐partum. By contrast, out of 207 women for which it predicted no relapse, 55 (27%) actually relapsed in the post‐partum period. When comparing the predicted and observed status, only 160 out of 223 (72%) of the women were correctly classified by this model and 63 out of 223 (28%) were misclassified (Fig. 4).

Figure

Fig. 4 Individual prediction of the presence or absence of a relapse in the 3‐month period after delivery among 223 women with multiple sclerosis, according to the best multivariate model.

View this table:
Table 3

Indicators of the occurrence of at least one relapse during the 3‐month period after delivery

Explanatory variableOdds ratio95% CI P
No. of relapses in pre‐pregnancy year1.941.35–2.80<0.0001
No. of relapses during pregnancy1.871.12–3.130.02
MS duration (years)1.111.03–1.200.01

Multivariate logistic regression analysis among 223 women with multiple sclerosis. CI = confidence interval; MS = multiple sclerosis.

The following formula can be used to calculate the individual probability (Pr) of having a relapse during the three months post‐partum =

In Figure 4, with a cut‐off limit of 0.50, a women is supposed to have a relapse if her probability is higher than 0.50.

Discussion

The PRIMS study affords a unique opportunity to improve our knowledge on interactions between pregnancy and multiple sclerosis. With 227 pregnant women with multiple sclerosis and a full‐term delivery of a life infant, from 12 European countries, followed prospectively during pregnancy and up to the end of the second year post‐partum, this is the largest natural history study ever conducted in this setting. The study population can be considered as representative of women with multiple sclerosis contemplating childbearing: they were young, with a mean age of 30 years at pregnancy onset, and had a relapsing–remitting disease with mild disability (mean DSS 1.3). The mean relapse rate in the year before pregnancy was 0.7, very similar to the relapse rate observed in natural history studies for this early stage of the disease (Confavreux et al., 1998). Therefore, a selection bias is unlikely. However, the validity of pre‐pregnancy data can be questioned, as the study formally started when the women were pregnant. In fact, 76% of the women were known to their neurologist before the study and 52% of the women had been followed prospectively by the participating neurologists before the study. This reduces the probability of recall bias. Sixty four (28%) women were lost to follow‐up before the end of the second year post‐partum, as they did not return to their neurologist at the scheduled visits. There were no differences between the women who were followed and those lost to follow‐up with respect to demographic and clinical data (DSS scores and relapses before pregnancy, during pregnancy and in the first trimester post‐partum), and no differences in the frequency of breast‐feeding and epidural analgesia. Excluding women lost to follow‐up from the analyses should therefore not have introduced any systematic error.

As previously reported (Confavreux et al., 1998), the relapse rate decreased significantly during pregnancy, especially in the third trimester, and increased in the first trimester after delivery. From the second trimester onwards, in the 2 years post‐partum, the relapse rate did not differ from that in the pre‐pregnancy year. We cannot exclude the possibility that minor relapses could have been not recorded during the pre‐study period, but the bias introduced in this way would have underestimated the true reduction in the relapse rate during pregnancy. Confirmed disability seemed not to be influenced by pregnancy, delivery and the post‐partum period. With the addition of data for the second year post‐partum, the follow‐up is now long enough to exclude a pregnancy‐related increase in confirmed disability that could not have been detected at 1 year. Again, the validity of the pre‐pregnancy data can be questioned. As stated above, a majority of the women were already known by their neurologist before their enrolment in the study. It could also be questioned whether treatments that were potentially introduced after delivery could have had an impact on these findings. Only five patients received a disease‐modifying drug in the first 6 months post‐partum and 11 in the second 6‐month period. In the first year, when assessment regarding treatments was reliable and complete, the number of treated patients was small and should not have changed the overall results for this period. Regarding the second year post‐partum, we have no data on immunoactive treatments as the study had not been designed to collect them. In this second year, immunoactive treatment may have reduced the relapse rate, but the length of the treatment period was too short for us to expect a significant change in the measure of confirmed disability.

Neither breast‐feeding nor epidural analgesia seem to have any effect on relapses and disability, with the proviso that the study was not designed to assess this risk. In this respect, it is important to note a possible misunderstanding. Women who chose to breast‐feed experienced fewer relapses and had milder disability scores in the year before and during pregnancy, in comparison with women who chose not to breast‐feed (Confavreux et al., 1998). In an editorial accompanying the original report, Whitaker (Whitaker, 1998) suggested that ‘the finding… that breast‐feeding had a beneficial effect by reducing relapses during the course of the study is unexpected’. The interpretation of the data on breast‐feeding by the PRIMS study group is that multiple sclerosis patients who chose to breast‐feed had a milder form of multiple sclerosis from the beginning of the study period, 12 months before conception, and that patients with more active disease decided not to breast‐feed. It cannot be concluded from our data that breast‐feeding has a protective effect per se on disease activity in the post‐partum period.

The relapse rate over the period from month 4 to month 24 post‐partum did not differ significantly from the relapse rate in the pre‐pregnancy year. Thus, the period of increased risk of relapse relates only to the first 3 months after delivery. While clearly this is a period of high risk, only 28% of the cohort had a relapse during this time. Two factors appeared predictive of an increased likelihood of a relapse in the 3 months post‐partum: the number of relapses in the year before pregnancy and the number of relapses during pregnancy. The DSS score at pregnancy onset just reached the significance level in the univariate analysis, but was not independently correlated to the presence of a post‐partum relapse as it did not remain significant in the multivariate analysis. Surprisingly, the higher the DSS score, the higher the risk of relapse. One should, however, consider that this analysis was done in patients who had low disability scores, with a mean of 1.3, and that results should not be extended to patients with higher DSS scores. By contrast, multiple sclerosis duration added some information in the multivariate model even if it was not significant alone: the longer the multiple sclerosis duration, the higher the likelihood of a relapse. Other disease or pregnancy indices were not related to the presence or absence of the post‐partum relapse.

The more active the disease in the pre‐pregnancy year and during pregnancy, the higher the risk of relapse in the post‐partum period. However, of 33 women who had had at least one relapse in each period (pre‐pregnancy year and pregnancy) and who were followed at least until the end of the first trimester after delivery, 20 (61%) did not relapse in the first trimester after delivery. By contrast, 14 (16%) of the 87 women without any relapse in these periods experienced at least one relapse after delivery. Moreover, the risk of having a relapse in the first trimester after delivery could not be reliably predicted even by the mathematical models for a given individual. In the best proposed multivariate model, 28% of the women were misclassified. In other words, if, by using the multivariate model, patients were chosen for a putative post‐partum therapy to prevent relapses, then eight of 16 patients in whom a relapse was predicted would have been treated needlessly. Similarly, of the 207 women predicted by this model not to relapse, 55 who actually experienced a relapse would not have been treated. However, depending on the choice of the treating physician and the patient to over‐ or undertreat, it is possible to modify this figure by changing the cut‐off limit in the same model. These results have clear consequences for planning future therapeutic trials aiming to prevent post‐partum relapses. There is no clinical tool for identifying with certainty in advance the women with multiple sclerosis who will eventually experience a relapse in the first 3 months post‐partum.

Acknowledgements

We are indebted to the patients for their participation in the PRIMS study and to all the investigators for their active contribution to the study.

Appendix

Study design and steering committee

A. Alpérovitch, Villejuif, France; H. Carton, Leuven, Belgium; M. B. d’Hooghe, Melsbroek, Belgium; O. Hommes, Nijmegen, the Netherlands; M. Hutchinson, Dublin, Ireland.

EDMUS Coordinating Center, Lyon, France

P. Adeleine, A. Biron, P. Cortinovis‐Tourniaire, J. Grimaud, M. Hours, T. Moreau, S. Vukusic, C. Confavreux.

Participants

Numbers of pregnancies studied are shown in parentheses. France (71): Lyon, G. Chauplannaz, C. Confavreux, P. Cortinovis‐Tourniaire, J. Grimaud, D. Latombe, T. Moreau; Toulouse, M. Clanet, G. Lau; Besançon, L. Rumbach; Brest, J. Y. Goas, F. Rouhart; Lille, A. Mazingue; Paris, E. Roullet; Saint Brieuc, M. Madigand; Lomme, P. Hautecoeur; Rennes, P. Brunet, G. Edan, C. Allaire; Tours, G. Riffault; Vendome, J. Leche; Villepinte, T. Benoit; Arles, C. Simonin; Belfort, F. Ziegler; Caen, J. C. Baron, Y. Rivrain; Dijon, R. Dumas, D. Loche; Draguignan, J. C. Bourrin; Epinal, B. Huttin; Lens, B. Delisse; Le Puy en Velay, I. Gibert; Mulhouse, C. Boulay; Nantes, M. Verceletto; Nice, G. Durand; Orléans, G. Bonneviot; Poitiers, R. Gil; Pointe‐à‐Pitre, M. A. Hedreville; Reims, C. Belair; Restigné, R. J. Poitevin; Angoulême, J. L. Devoize; Saint Quentin, P. Wyremblewski; Vannes, F. Delestre; Villefranche‐sur‐Saône, A. Setiey. Italy (42): Milan, G. Comi, M. Filippi, A. Ghezzi, V. Martinelli, P. Rossi, M. Zaffaroni; Ferrara, M. R. Tola; Florence, M. P. Amato; Pisa, C. Fioretti, G. Meucci; Genoa, M. Inglese, G. L. Mancardi; Chietti, D. Gambi, A. Thomas; Modena, M. Cavazzuti; Pavia, A. Citterio. Denmark (25): Roskilde, A. Heltberg; Aarhus, H. J. Hansen. Spain (25): Malaga, O. Fernandez, F. Romero; Barcelona, T. Arbizu, J. J. Hernandez; Madrid, C. De Andres de Frutos; Castellon, D. Geffner Sclarky; Las Palmas, Y. Aladro Benito, P. Reyes Yanes; Terrassa, M. Aguilar; Valencia, J. A. Burguera, R. Yaya; Valladolid, W. Bonakim Dib, D. Arzua‐Mouronte. Belgium (23): Melsbroek, M. B. d’Hooghe; Brussels, C. J. M. Sindic; Leuven, H. Carton; Diepenbeek, R. Medaer; Bonheiden, H. Roose; Ekeren, K. M. J. Geens; Fraiture, D. Guillaume; Brugge, M. Van Zandycke; Lewen, J. Janssens; Seraing, M. Cornette; Turnhout, L. Mol. Germany (21): Würzburg, F. Weilbach, P. Flachenecker, H. P. Hartung; Magdeburg, J. Haas, I. Tendolkar; Bochum, E. Sindrn; Erfurt, H. W. Kölmel, D. Reichel; Bielefeld, M. Rauch; Emden, S. Preuss; Göttingen, S. Poser; Schwendi, E. Mauch. Austria (18): Graz, S. Strausser‐Fuchs; Vienna, H. Kolleger. UK (18): Belfast, S. Hawkins; Sheffield, S. J. L. Howell; Haywards Heath, J. E. Rees; London, A. Thompson; Leeds, M. Johnson; Stoke‐on‐Trent, M. Boggild; Swindon, R. P. Gregory; Newcastle upon Tyne, D. Bates; Glasgow, I. Bone. Ireland (14): Dublin, M. Hutchinson. The Netherlands (7): Amsterdam, C. Polman; Nijmegen, S. Frequin, P. Jongen, O. Hommes. Portugal (4): Lisbon, J. Correia de Sa; Porto, M. E. Rio. Switzerland (1): Basel, S. Huber, J. Lechner‐Scott, L. Kappos.

References

View Abstract