To determine the rate of induced abortions in women with rheumatoid arthritis (RA) exposed to methotrexate (MTX) compared with women with RA unexposed to this medication.
To determine the rate of induced abortions in women with rheumatoid arthritis (RA) exposed to methotrexate (MTX) compared with women with RA unexposed to this medication.
We performed a nested case–control study using administrative databases from Quebec. All women with RA ages 15–45 years were identified and cases were defined as women having an induced abortion. Each case was matched to ≥1 controls for age, calendar time, and cohort entry. Exposure was defined as having filled ≥1 prescriptions of MTX ≤16 weeks prior to the index date.
We identified 112 cases of induced abortions in women with RA and 5,855 RA controls. Exposure to MTX occurred in 10.7% of cases and 21.7% of controls. Women exposed to MTX had a lower rate of induced abortions compared with unexposed women (rate ratio [RR] 0.47 [95% confidence interval (95% CI) 0.25–0.89]). In the multivariate analysis, there was a trend toward an increased rate of induced abortions among women exposed to anti–tumor necrosis factor (anti-TNF) agents (RR 2.07 [95% CI 0.81–5.27]).
Women with RA exposed to MTX have a lower rate of induced abortions than unexposed women. Women with RA exposed to anti-TNF agents may have an increased rate of induced abortions compared to unexposed women.
In North America, up to half of pregnancies are unplanned and almost half of unintended pregnancies are terminated (). Women affected by rheumatoid arthritis (RA) during their reproductive period may also be at risk of unplanned pregnancies. In women with RA, specific disease-related factors such as teratogenic drug exposure may influence the decision to end a pregnancy.
Methotrexate (MTX) is a well-recognized teratogen that is associated with fetal aminopterin syndrome, which features skeletal anomalies, microcephaly, and hydrocephalus (). A recent survey of rheumatologists in the US showed that, although the vast majority (96%) recommended an effective method of birth control when initiating MTX in women of childbearing age, only approximately half (55%) reviewed contraceptive use on subsequent visits (). There are currently no data specifically on pregnancy planning in women with RA receiving MTX, but in the general population, half of unintended pregnancies result from contraceptive failure due to inconsistent or incorrect use, while the remainder results from nonuse (). When faced with an unplanned pregnancy, women with RA may be more likely to undergo an induced abortion if they are exposed to MTX (than unexposed women with RA) because of the potential risk of a congenital anomaly.
Although there is a considerable amount of studies in the literature regarding many aspects of pregnancy in women with RA, studies on induced abortion in RA and its potential predictors are scant. Thus, we aimed to determine the rate of induced abortions in women with RA exposed to MTX compared with women with RA unexposed to this medication.
We conducted a nested case–control study using administrative databases from Quebec (the Régie de l'assurance maladie du Québec [RAMQ] physician billing and the Maintenance et exploitation des données pour l'étude de la clientèle hospitalière [MED-ÉCHO] hospitalization databases, from January 1, 1996 to December 31, 2008) that cover all health care beneficiaries. All women with RA ages between 15 and 45 years were identified based on ≥1 hospitalization with either a primary or secondary diagnosis of RA or ≥2 physician claims for RA within any 2-year period (at least 8 weeks apart). Women who had an intrauterine device (IUD) insertion, any procedures leading to sterilization, and/or any conditions causing infertility prior to cohort entry were excluded in the primary analysis (but included in a sensitivity analysis).
Cases were defined as women having an induced abortion based either on 1 procedure code and/or diagnostic code for induced abortion present in the MED-ÉCHO hospitalization and/or RAMQ billing databases. For each case, ≥1 age-matched controls were selected at random from all subjects who entered the cohort on the same month and year as the case and who were born within 12 months of the case birthdate. Controls had to be at risk on the event date of their corresponding case; this date was referred to as the index date.
Exposure was defined as having filled at least 1 prescription for MTX in the 16 weeks prior to the index date, as recorded in the RAMQ prescription database. The RAMQ prescription database provides valid data on the drug exposures for beneficiaries of the public drug insurance plan, including 36% of women between ages 15 and 45 years (). To obtain complete information on medication exposure, we included only cases and controls who had RAMQ prescription plan coverage for the entire 16-week period preceding the index date.
We performed a multivariate conditional logistic regression analysis to assess other potential predictors of induced abortions. Time-dependent predictors assessed at the index date included exposure to anti–tumor necrosis factor (anti-TNF) agents (i.e., any prescription filled in the previous 16 weeks), prednisone use (i.e., any prescription filled in the previous 16 weeks), extraarticular manifestations (i.e., ≥1 diagnostic code prior to the index date and since RA diagnosis for either Felty's syndrome, rheumatoid lung, or RA-related visceral involvement), and number of physician visits and hospitalizations in the preceding year.
A total of 18,654 women with RA ages between 15 and 45 years were identified during the study period. Of these women, 5,967 had public drug insurance coverage for the 16 weeks preceding the index date. We identified 112 cases of induced abortions and 5,855 corresponding RA controls (Table 1). The overall rate of induced abortions (at the index date) was 6.78 cases per 1,000 person-years (95% confidence interval [95% CI] 5.63–8.16). On the index date, the cases had fewer physician visits and hospitalizations in the preceding year, lower prednisone exposure, and fewer RA extraarticular manifestations than controls (Table 1); all of these findings potentially indicate lower RA activity.
|Characteristics||Cases (n = 112)||Controls (n = 5,855)|
|Age, mean ± SD years||29.4 ± 7.0||33.6 ± 6.8|
|Disease duration, mean ± SD years||4.1 ± 2.6||4.6 ± 2.5|
|Physician visits in the previous year, mean ± SD||10.0 ± 6.7||11.9 ± 14.7|
|Hospitalizations in the previous year, mean ± SD||0.2 ± 0.4||0.3 ± 0.8|
|Exposed to methotrexate, no. (%)|
|Yes||12 (10.7)||1,272 (21.7)|
|No||100 (89.3)||4,583 (78.3)|
|Exposed to anti–tumor necrosis factor agents, no. (%)|
|Yes||7 (6.3)||179 (3.1)|
|No||105 (93.7)||5,676 (96.9)|
|Exposed to prednisone, no. (%)|
|Yes||12 (10.7)||960 (16.4)|
|No||100 (89.3)||4,895 (83.6)|
|Extraarticular manifestations, no. (%)|
|Felty's syndrome||1 (0.9)||10 (0.2)|
|Rheumatoid lung||0 (0.0)||4 (0.1)|
|Rheumatoid arthritis visceral involvement||1 (0.9)||187 (3.2)|
Exposure to MTX occurred in 10.7% of cases and in 21.7% of controls. Women exposed to MTX had a lower rate of induced abortions compared to unexposed women (adjusted rate ratio [aRR] 0.47 [95% CI 0.25–0.89]). The effect estimate was similar in the sensitivity analysis that did not exclude women who had an IUD insertion, procedures leading to sterilization, and/or any conditions causing infertility prior to cohort entry (aRR 0.36 [95% CI 0.1–0.74]).
In the multivariate analysis (Table 2), there was a trend toward an increased rate of induced abortions among women exposed to anti-TNF agents (aRR 2.07 [95% CI 0.81–5.27]). We did not establish an independent effect of prednisone use, extraarticular manifestations, physician visits, or hospitalizations on induced abortions.
|Variables||Rate ratio (95% confidence interval)a|
|Exposure to methotrexate|
|Exposure to anti–tumor necrosis factor agents|
|Exposure to prednisone|
|No. of physician visits in previous yearb||0.99 (0.97–1.01)|
|No. of hospitalizations in previous yearc||0.98 (0.71–1.34)|
We demonstrated that women with RA exposed to MTX have a lower rate of induced abortions compared with unexposed women. There are many potential reasons for this finding; one may be the use of effective contraception. Alternatively, women with RA receiving MTX may have less sexual activity than unexposed women, potentially because of increased disease activity. Furthermore, spontaneous abortions may be increased in women receiving MTX, consequently resulting in a lower induced abortion rate ().
The overall rate of induced abortions observed in women with RA was approximately half the rate of the general population (). Still, induced abortions in women with RA receiving MTX were relatively common (half the rate of unexposed women), suggesting an inappropriately high rate of unplanned pregnancies in women exposed to MTX. This stresses the need to improve contraception counseling in women receiving MTX, not only at the initiation of therapy, but also while therapy is ongoing.
Moreover, we observed that women exposed to anti-TNF agents may be at increased risk of induced abortions, presumably for unplanned pregnancies. Because there is no specific recommendation for contraception in women receiving anti-TNF agents, health care professionals may omit contraception counseling for these women, as suggested in a previous study (). Women receiving anti-TNF agents may be misinformed about the fetal risk following exposure during the conception period or early pregnancy and inadvertently terminate a pregnancy because they perceived the risk to be unduly increased. Furthermore, women with RA receiving anti-TNF agents may have more severe disease, which may affect their childbearing decision when faced with an unplanned pregnancy. Thus, comprehensive contraception counseling should be offered not only to women exposed to teratogenic drugs such as MTX, but also to women with severe disease and/or receiving anti-TNF agents.
The strength of our study was our use of administrative databases from Quebec, which record information on all induced abortions performed in the province. Administrative database research is not affected by the social desirability bias that plagues studies of induced abortions using self-report. Indeed, investigators have shown that <30% of prior induced abortions are voluntarily reported when comparing self-report to administrative data (). Although the validity of the procedures and diagnostic codes used in our study to define cases of induced abortions has not been specifically studied, prior studies have established the validity of other obstetric outcomes such as gestational age, live birth, and congenital anomaly defined within the same administrative databases ([8, 9]).
Our study has some limitations. Although the cases of induced abortion seemed to have lower disease activity, as shown by lower prednisone exposure, fewer extraarticular manifestations, and fewer physician visits or hospitalizations than controls, we were not able to demonstrate an independent effect of these potential predictors on induced abortions. Prior studies aiming to identify a valid marker or index of RA disease severity for use in administrative database research did not confirm the usefulness of the above potential predictors as such (). Currently, there is no validated measure of RA disease severity in administrative database research.
In a nested case–control study, subjects must be at risk of the outcome for the RR to be accurately estimated. Thus, in the present study, we excluded women who had an IUD from the analysis because the failure rate with this contraceptive method is extremely low (<0.14% at 1 year) and comparable to sterilization (). However, we did not exclude women receiving reversible forms of contraception such as oral contraceptives and barrier methods (which have a 1-year failure rate superior to 9%) because half of unplanned pregnancies occur in women using these forms of contraception (). Because women receiving MTX may be more likely to use an IUD for contraception than unexposed women, we performed a sensitivity analysis including women who had an IUD, which showed results similar to the primary analysis. Because the Quebec administrative databases do not provide complete information on reversible forms of contraception, we could not investigate the independent effect of these forms of contraception on the risk of induced abortions in women with RA.
We used a 16-week exposure window before the index date to capture drug exposure occurring during the conception period and/or early pregnancy. Notably, while it is advised that MTX be discontinued at least 12 weeks before attempting to conceive because of its long half-life, there is no consensus on discontinuation (or continuation) of anti-TNF agents, even though these agents are classified as Food and Drug Administration category B in pregnancy (). The British Society for Rheumatology recommends that anti-TNF agents should be stopped prior to conception, but does not specify the duration (). Because induced abortions are usually performed at a gestational age >6 weeks (i.e., approximately 4 weeks postconception) (), the 16-week period allowed us to define a subject as exposed if the exposure occurred ≤12 weeks before conception. However, this cutoff may have led us to classify cases of induced abortions occurring at a late gestational age as unexposed, but with an exposure early in pregnancy. Still, this exposure misclassification, which was likely nondifferential, would have produced more conservative estimates than the ones we observed.
The 16-week medication exposure window did not allow us to assess exposure to other teratogenic drugs that have a longer half-life, such as leflunomide. However, few patients are exposed concomitantly to leflunomide and MTX, or leflunomide and an anti-TNF agent (). Furthermore, as reported in a previous study using the same source population, <1% of patients with RA who were on the RAMQ prescription plan were taking leflunomide (). Therefore, our lack of evaluating leflunomide exposure is unlikely to have affected our results.
We did not include exposure to nonsteroidal antiinflammatory drugs (NSAIDs) as a predictor of induced abortions in our multivariate model. Although the Quebec administrative databases contain valid data on prescription medication exposures, NSAID exposure is particularly difficult to ascertain accurately because these drugs can be obtained over the counter (in which case exposure is not captured in the database) and are frequently prescribed as needed, which increases the difficulty in assessing the timing of exposure relative to a specific event. However, NSAIDs can cause luteinized unruptured follicle syndrome, which results in secondary infertility (). Thus, women with RA exposed to NSAIDs could potentially have decreased fertility and be less likely to have an unplanned pregnancy, and this might consequently reduce their risk of induced abortions compared to unexposed women. Furthermore, this could have introduced bias in our effect estimate if the women receiving MTX had differential exposure to NSAIDs than women unexposed to MTX. However, we previously used Quebec administrative databases to examine the frequency of MTX use and did not find any difference in concomitant NSAID use (traditional and selective) between MTX users and nonusers (). Therefore, we believe that failure to adjust for NSAIDs is unlikely to have introduced substantial bias in our current results.
Exposure to MTX and anti-TNF agents was defined based on filled prescriptions, which might not have reflected actual intake. However, it is likely that most women who filled a prescription for MTX and/or an anti-TNF agent took at least 1 dose because, within the RAMQ prescription plan, beneficiaries need to cover part of their medication cost. Thus, we believe this would not have invalidated our findings.
In summary, women with RA exposed to MTX appeared to have a lower (not higher) rate of induced abortions than unexposed women. Women with RA exposed to anti-TNF agents may have an increased rate of induced abortions compared to unexposed women. These results raise concerns about the rate of unplanned pregnancies in women with RA, particularly in those receiving anti-TNF agents, and should prompt future research on counseling, contraception use, and unplanned pregnancies in these women.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Vinet had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Vinet, Kuriya, Pineau, Clarke, Bernatsky.
Acquisition of data. Vinet, Pineau, Clarke, Bernatsky.
Analysis and interpretation of data. Vinet, Kuriya, Pineau, Clarke, Bernatsky.