Many women with systemic lupus erythematosus (SLE) take hydroxychloroquine (HCQ) to control disease activity. Studies have shown that HCQ can prevent renal and central nervous system lupus and lessen the effects of SLE (1, 2). It has also been demonstrated that cessation of HCQ treatment places a woman at more than twice the risk of a lupus flare in the subsequent 6 months (3, 4). Women with SLE maintain fertility, and most can have a successful pregnancy (5–7). Experts in the care of SLE during pregnancy generally recommend continuing HCQ treatment during pregnancy (8). At the Fourth International Conference on Sex Hormones, Pregnancy, and the Rheumatic Diseases in 2004, the working group on medications during pregnancy recommended continuing HCQ treatment (9). The data to support this recommendation are limited to reports of <300 pregnancies, however. A significant minority of rheumatologists, particularly those who see few lupus pregnancies per year, do not routinely continue HCQ treatment during pregnancy (8).
Early reports of in utero chloroquine toxicity have led to some trepidation about the use of HCQ during pregnancy (10, 11). However, more recent systematic studies of HCQ use during pregnancy have suggested its safety (12–16). There have been no reported cases of fetal malformations or visual or auditory deficits directly related to HCQ exposure in utero.
In this study, we reviewed the impact of HCQ exposure on pregnancy, fetal, and lupus outcomes in a cohort of women with SLE. To our knowledge, this is the first report of the impact of HCQ treatment cessation on SLE activity during pregnancy.
PATIENTS AND METHODS
The Hopkins Lupus Pregnancy Cohort has enrolled consecutive patients seen in the Hopkins Lupus Center since 1987, and this report describes pregnancies seen between 1987 and 2002. Approval for this prospective cohort was obtained from the Johns Hopkins Institutional Review Board, and each woman signed informed consent forms prior to enrollment. At the first visit, the patient's lupus and obstetric history and medications taken prior to and during pregnancy were recorded. At each subsequent visit, generally every 4–6 weeks throughout pregnancy, medications and lupus activity were recorded.
In this study, we used several measures of lupus activity, including the physician's estimate of lupus activity (PEA), also known as the physician's global assessment of disease activity, and the SLE Disease Activity Index (SLEDAI). The PEA is a validated visual analog scale (VAS) for the measurement of current SLE activity based on disease history, physical examination, and prior laboratory findings. The scale ranges from 0 to 3, with a score of 0 indicating no lupus activity, 1 indicating mild activity, 2 indicating moderate activity, and 3 indicating severe lupus activity. Moderate to severe lupus activity was defined as a PEA score ≥2 during pregnancy. A flare was defined as an increase of 1 on this scale within a period of 90 days (17, 18).
The SLEDAI is a validated and widely used measure of lupus activity (17, 19). We defined a SLEDAI score ≥4 as evidence of clinically important lupus activity during pregnancy. We identified pregnancies with high lupus activity through the use of these 2 scales.
The patients in this cohort were divided into 3 groups based on their use of HCQ. The elimination half-life of HCQ is estimated to be 40 days, and the impact of cessation of the drug on SLE activity can take several months (3, 20, 21). Therefore, exposure to and cessation of HCQ treatment in the 3 months prior to pregnancy was taken into consideration when delineating these groups. Group 1 had no exposure to HCQ within the 3 months prior to or during pregnancy. Group 2 took HCQ throughout pregnancy. Group 3 stopped taking HCQ either in the 3 months prior to or during the first trimester of pregnancy. The 6 women who started taking HCQ during pregnancy were not included in this analysis because of the small sample size of this group. Two women first presented to the Lupus Clinic after 35 weeks' gestation. Because they were not at significant risk for preterm delivery or death, they were excluded from this study.
Information about each offspring was obtained from the mother during postpartum followup. The infants did not undergo systematic congenital anomaly examinations or special ophthalmologic or auditory testing beyond what was performed routinely for infant care.
Pregnancy outcomes were obtained through obstetric documentation. A miscarriage was defined as a pregnancy loss prior to 20 weeks' gestation. The miscarriage rate was computed for all pregnant women seen in the Johns Hopkins Lupus Center prior to 20 weeks' gestation. For pregnancies that survived past 20 weeks' gestation, further outcomes were analyzed. A stillbirth was defined as a pregnancy loss after 20 weeks' gestation. Extreme prematurity was a birth between weeks 20 and 27.9. A preterm birth occurred between weeks 28 and 36.9. All deliveries after 37 weeks' gestation were counted as full term. An infant who was small for gestational age weighed less than the 10th percentile for age at birth, based on US national norms (22).
We compared the proportions of pregnancy and disease outcomes of the 3 groups of women. P values to quantify the association between HCQ group and pregnancy outcomes were based on a person-week analysis, such that each week that a woman was observed during pregnancy was used as a data point. For analyses of lupus activity, P values were based on a comparison of the proportions of pregnancies with the outcome. Logistic regression analysis was used to assess the effect of group, controlling for potential confounders. All P values were based on the generalized estimating equations approach to appropriately account for repeated observations of the same woman. All analyses were implemented using SAS software, version 8.02 (SAS Institute, Cary, NC) (23).
HCQ group characteristics.
The Hopkins Lupus Pregnancy Cohort included 257 pregnancies evaluated in 197 women between 1987 and 2002. Of these pregnancies, 56 (22%) had been exposed to HCQ throughout pregnancy, and 163 (63%) had no exposure. In 38 of the pregnancies (15%), the women stopped taking HCQ during the first trimester or within the 3 months preceding pregnancy; 10 women discontinued HCQ prior to pregnancy, and 28 women discontinued it during the first trimester. The medication was usually stopped because of concern over exposure of the fetus to HCQ. One pregnancy in each HCQ group resulted in twins.
The demographic characteristics of each HCQ group are shown in Table 1. The distribution of ethnicity and maternal age was similar among these groups. A similarly high rate of prior lupus nephritis, based on fulfillment of the American College of Rheumatology criteria for lupus nephritis (24), was present in each group. Antiphospholipid syndrome (APS), diagnosed based on the Sapporo criteria (25), was present in 14% of pregnancies without HCQ exposure, 5% with exposure throughout pregnancy, and none of the pregnancies in women who stopped the medication.
Table 1. Demographic characteristics and lupus history*
|Gestational age at first visit|| || || |
| 0–9 weeks||77 (47)||35 (63)||24 (63)|
| 10–19 weeks||62 (38)||17 (30)||13 (34)|
| 20–29 weeks||20 (12)||3 (5)||1 (3)|
| ≥30 weeks||4 (2)||1 (2)||0|
|Ethnicity|| || || |
| Caucasian||98 (60)||35 (63)||23 (61)|
| African American||61 (37)||16 (29)||15 (39)|
| Asian||4 (2)||3 (5)||0|
| Hispanic||0||1 (2)||0|
| Unknown||0||1 (2)||0|
|Age|| || || |
| <25 years||29 (18)||7 (12)||11 (29)|
| ≥25–34 years||109 (67)||38 (68)||22 (58)|
| ≥35 years||25 (15)||11 (20)||5 (13)|
|Duration of SLE|| || || |
| Diagnosed during pregnancy||30 (18)||3 (5)||2 (5)|
| <5 years||58 (36)||18 (32)||18 (48)|
| ≥5 years||75 (46)||35 (63)||18 (47)|
|Antiphospholipid syndrome||23 (14)||3 (5)||0|
|Year of delivery|| || || |
| Before 1995||89 (55)||10 (18)||30 (79)|
| After 1995||74 (45)||46 (82)||8 (21)|
|History of lupus nephritis||66 (40)||20 (36)||19 (50)|
|High-activity lupus in the 6 months prior to conception/no. of women seen in the prior 6 months||8/52 (15)||0/29||4/14 (29)|
The use of HCQ changed over time. The majority of women who conceived while taking the drug prior to 1995 stopped taking it, while the majority of those who conceived while taking HCQ after 1995 continued to take it. A comparison of the pregnancies evaluated prior to and after 1995 revealed several other statistically significant differences. Prior to 1995, more women in the cohort were African American, the average age of the mothers was lower, the rate of increased lupus activity was higher, and more women were taking prednisone at a higher dosage.
Among the 3 HCQ groups, the pregnancy outcomes, including pregnancy loss rates, the timing of pregnancy loss, gestational age at delivery, and the rate of small for gestational age babies, were statistically similar (Tables 2 and 3). When pregnancies that occurred in women with secondary APS were removed from the analysis, the differences in pregnancy outcome remained statistically similar. Among patients without APS, miscarriage occurred in 5% of pregnancies without HCQ exposure, 10% of those with continued HCQ, and 11% of those in which HCQ was stopped (P = 0.44). Stillbirth occurred in the absence of APS in 6% of women without HCQ exposure, 6% who continued it, and 9% of those who discontinued it (P = 0.88).
Table 2. Miscarriage rate for all pregnancies seen in the Hopkins Lupus Center prior to 20 weeks' gestation*
|Group 1, no HCQ||6/139 (4)|
|Group 2, HCQ continued||7/52 (13)|
|Group 3, HCQ stopped||4/37 (11)|
Table 3. Outcomes for all pregnancies that survived past 20 weeks' gestation*
|Total pregnancies||157||49||34|| |
|Live births||145||47||32|| |
|Stillbirths (pregnancy loss after ≥20 weeks)||13 (8)||3 (6)||3 (9)||0.85|
|Preterm births|| || || || |
| Extreme preterm (20–27.9 weeks)||15 (10)||6 (12)||2 (6)||0.83|
| Preterm (28–36.9 weeks)||49 (31)||13 (27)||16 (47)||0.87|
|Full-term births (≥37 weeks)||93 (59)||30 (61)||16 (47)||0.98|
|Small for gestational age (<10th percentile for age) among live births||29 (20)||11 (24)||7 (23)||0.93|
In this study, there were no fetal abnormalities clearly attributable to HCQ exposure. The congenital anomaly rate for the entire cohort was low and was comparable to that of healthy pregnancies. Among the 79 offspring born to women with HCQ exposure, 1 child had a cleft lip and palate, 1 developed pneumonia and pleural effusions following birth, and another had mild rectal bleeding. All of these problems resolved with medical treatment. Among the 163 pregnancies without HCQ exposure, 3 fetuses had severe, fatal congenital anomalies, 1 child had an abdominal hernia, 3 developed congenital heart block from neonatal lupus, and 4 developed thrombocytopenia following birth. There were no visual or hearing impairments reported in any offspring.
Women who stopped taking HCQ had increased lupus activity and increased lupus flares during pregnancy (Table 4). High-activity lupus, defined as a PEA score ≥2, occurred in twice as many pregnancies in which HCQ treatment was stopped as in those in which HCQ was continued (P = 0.05). The rate of flare was also higher among women who stopped the medication compared with those who either continued taking it or who never took it (P = 0.05). These flares occurred throughout pregnancy, with statistically more in this group than in the others during the first and third trimesters. The risk of lupus flare remained elevated, though not statistically significant, when adjusted for year of delivery, age, ethnicity, the presence of APS, and history of lupus nephritis (P = 0.10). The maximum SLEDAI score was higher among women who discontinued HCQ treatment prior to or during pregnancy (P = 0.06). More women who discontinued HCQ treatment had a SLEDAI score ≥4 during pregnancy (P = 0.007), and this remained elevated when adjusted for year of delivery, APS, age, ethnicity, and prior history of lupus nephritis (P = 0.10).
Table 4. Lupus activity during pregnancy*
|Total pregnancies||163||56||38|| |
|High PEA score||41 (25)||6 (11)||9 (24)||0.051|
|Flare rate||59 (36)||17 (30)||21 (55)||0.053|
|Maximum SLEDAI score, mean ± SD||5.2 ± 3.8||4.2 ± 4.3||6.5 ± 4.0||0.062|
|SLEDAI score ≥4 during pregnancy||101 (62)||29 (52)||32 (84)||0.0075|
|Proteinuria >500 mg/24 hours||48 (29)||10 (18)||10 (26)||0.23|
|Low platelet count (<150,000/μl)||32 (20)||13 (23)||6 (16)||0.64|
|Combination fatigue and arthritis||46 (28)||15 (27)||23 (61)||0.0054|
|Prednisone taken during pregnancy||109 (67)||35 (63)||34 (89)||0.0025|
|Maximum daily dosage of prednisone (excluding pregnancies without prednisone use), mean ± SD mg||23 ± 19||16 ± 12||21 ± 16||0.056|
|High-dose prednisone (≥20 mg/day or pulse therapy)||66 (40)||15 (27)||17 (45)||0.16|
|Azathioprine taken during pregnancy||21 (13)||8 (14)||2 (5)||0.21|
The types of lupus activity that were best controlled by HCQ were arthritis and constitutional symptoms. HCQ did not prevent the more severe complications of proteinuria or thrombocytopenia. These dangerous complications of lupus are associated with pregnancy loss, particularly when they occur early during pregnancy (26). Proteinuria at levels >500 mg/24 hours occurred in 18% of pregnancies in which HCQ treatment was continued, 26% of those in which HCQ was stopped, and 29% of those that were never exposed to it (P = 0.23). The maximum level of proteinuria in the 3 groups of women was also similar (P = 0.65). Thrombocytopenia occurred in 20% of pregnancies that never were exposed to HCQ, 23% of those with continued HCQ exposure, and 16% of those in which HCQ was stopped (P = 0.64). Fatigue not associated with pregnancy and/or arthritis was present in 61% of pregnancies in which HCQ treatment was stopped, but in only 27% of those in which it was continued and 28% of those never exposed (P = 0.005). When adjusted for year of delivery, APS, age, ethnicity, and history of lupus nephritis, the increased rate of fatigue and/or arthritis remained statistically significantly elevated among women who stopped taking HCQ (P < 0.01).
Lupus activity in the months prior to conception had a significant impact on lupus activity during pregnancy and pregnancy survival (27). Ninety-five of the women included in this study were seen at the Hopkins Lupus Center in the 6 months prior to pregnancy. Of the women who discontinued HCQ, 14 were followed up at the Hopkins Lupus Center prior to pregnancy. Four of these women (29%) showed increased lupus activity, based on a PEA score ≥2, in the 6 months prior to conception. Of these 14, 1 woman stopped HCQ treatment prior to this activity and the others stopped HCQ treatment after conception. None of the 29 women who continued HCQ treatment during pregnancy and were seen prior to conception had lupus activity prior to pregnancy. Eight of the 52 women (15%) seen prior to pregnancy who were not taking HCQ had increased lupus activity. Increased lupus activity during pregnancy was more frequent after HCQ cessation, regardless of prior lupus activity. Among women with low-activity lupus prior to conception, 30% (3 of 10) of those who stopped taking HCQ developed high-activity lupus, while only 3% (1 of 29) of those taking HCQ continuously and 7% (3 of 44) of those who never took it developed active lupus. Pregnancy loss was more common in women with high-activity lupus prior to conception. Because of the small number of pregnancies involved in the analysis of lupus activity in the 6 months prior to pregnancy, statistical significance could not be ascertained for this analysis.
Medication during pregnancy.
The majority of patients in the cohort took prednisone during pregnancy. However, more women who discontinued HCQ treatment took prednisone (P = 0.003). Fewer women who continued HCQ required high-dose corticosteroids, defined as either a daily dose of prednisone of at least 20 mg or pulse steroids (P = 0.16). The average maximum daily dose of prednisone was lower among pregnancies in which HCQ treatment was continued than among those that were never exposed to HCQ and those in which HCQ use was stopped (P = 0.06).
A similar rate of azathioprine use during pregnancy was found among women who took HCQ continuously (14%) and those who never took HCQ (13%). Fewer women who discontinued HCQ treatment took azathioprine (5%). This difference was not statistically significant and was likely related to the year in which these pregnancies occurred. Prior to 1995, few pregnancies in the cohort were exposed to azathioprine.
The use of HCQ during pregnancy in women with SLE appears to be safe and decreases lupus activity during pregnancy. Based on the findings of this study, we encourage the continuation of HCQ treatment in women who are taking the drug at the time of conception.
The safety of HCQ during pregnancy has been confirmed in several other studies. The largest study compared 133 pregnancies in women taking HCQ for rheumatic disease with 70 pregnancies in women with similar illness but not taking HCQ (11). There was no difference in the live birth rate (88% versus 84%) or the rate of preterm birth (28% versus 33%). The rate of congenital abnormalities was no higher than that found in the general population. The PR interval on electrocardiograms performed soon after birth was the same in the 2 groups of infants. In followup to a mean age of 26 months, there was no evidence of visual, hearing, growth, or developmental problems in any of the children. Further studies, which combine to provide a total of >250 infants exposed to HCQ during pregnancy, have found similar results, with no evidence of consistent health problems or developmental delay in offspring exposed to HCQ in utero (13, 14, 28–30). There are isolated reports of patent ductus arteriosus, atrial septal defect, Down syndrome, and a deletion of the short arm of chromosome 6 in offspring exposed in utero to HCQ (31). However, the rate of congenital abnormalities in all studies has been comparable to that seen in the general population.
In the 1960s there were several reports of congenital ocular toxicity in offspring exposed in utero to high doses of chloroquine (28). There is also evidence that links HCQ to ocular toxicity in adults taking the medicine long term (32). Therefore, several studies have included ophthalmologic examinations in the offspring exposed in utero to HCQ. The only ocular abnormality reported was that of 2 infants with retinal hemorrhage at birth that resolved within a month, both thought to be secondary to birth trauma (16). No ocular disease has been reported in almost 200 children, of whom 58 have had systematic ophthalmologic examinations (12, 15, 28).
In this cohort of 79 live-born infants exposed to HCQ during pregnancy, there was no increase in congenital abnormalities compared with pregnancies without HCQ exposure. One case of cleft lip and palate was identified. However, the fetus was also exposed to prednisone, which is known to increase the risk of this abnormality (33). There are no other reported cases in the literature of cleft lip and palate in an infant exposed to HCQ.
Though we did not collect detailed information on the frequency of breastfeeding in this cohort, others have suggested the safety of HCQ during lactation. An estimated 0.06–0.2 mg/kg would be administered to an infant through breast milk, which is much less than the suggested dose of 6.5 mg/kg administered to patients (31).
This is the first study to look at cessation of HCQ treatment in a large cohort of pregnant patients. Our findings demonstrate that stopping this drug during or just prior to pregnancy leads to increased lupus activity, just as it does in nonpregnant patients. The Canadian Hydroxychloroquine Study Group randomized 47 nonpregnant lupus patients to continue or stop HCQ treatment. The study revealed that discontinuation of HCQ treatment doubles the risk for a lupus flare over the next 3 years (4). After 6 months without HCQ, 73% of patients had had a lupus flare versus 36% of those who continued HCQ (3). We demonstrate a similar increase in lupus activity in pregnant lupus patients who stopped HCQ treatment, and this risk for lupus activity continues for the duration of pregnancy.
In this study, we have presented several measures of increased lupus activity. The PEA is a validated VAS that estimates lupus activity, with scores that range from 0 to 3. Once pregnancy was recognized, symptoms that could be attributed to pregnancy were not identified as lupus-related unless the physician believed they were secondary to a flare. The flare rate presented in this study was based on the PEA score at each visit. Any increase in PEA score of ≥1 point over a 3-month period was counted as a flare. Because a flare is defined as an increase in lupus activity, women cannot be classified as having a flare at their first office visit. Therefore, this measure missed 26 women who had a PEA score ≥2 at their first visit. The SLEDAI score was also assessed at each visit and is reported in this study. Although modifications related to pregnancy have been suggested for the SLEDAI, these were not used in this study, since this would have been done retrospectively (34). We used a SLEDAI cutoff score of 4 to identify women with clinically relevant lupus activity. Each of these measures of activity, the PEA, flare rate, and SLEDAI, were higher in women who stopped taking HCQ during pregnancy.
The type of lupus activity that increased upon cessation of HCQ tended not to be life and pregnancy threatening. The rate of lupus nephritis and thrombocytopenia was similar in all 3 groups in this study. We have already demonstrated that having proteinuria and/or thrombocytopenia, particularly early during pregnancy, puts a pregnancy in jeopardy (26). We suggest that the similar rates of these complications may explain why HCQ does not provide a pregnancy survival advantage in this study.
The rate of lupus-related fatigue and arthritis were twice as high among women who stopped taking HCQ. Though these symptoms are not devastating to a fetus, they cause discomfort to the mother and may demand further therapy during pregnancy. The absence of these symptoms may have allowed the women who continued HCQ treatment during pregnancy to have a lower average prednisone dose.
This is a prospective database in which the use of HCQ was dependent on the discretion of the treating physicians and patient. Therefore, confounding by indication cannot be excluded. The decision to stop taking HCQ was based on recommendations of the obstetrician and the preference of the patient, not the Hopkins Lupus Center. In the mid-1990s, evidence of the benefits of HCQ in lupus was published. For this reason, the majority of women who conceived while taking HCQ prior to 1995 stopped taking the drug, while those who conceived while taking HCQ after 1995 continued it. This introduces a secular bias into the study. It is possible that other factors that are determined by the era of the delivery, such as changes in obstetric and neonatal care, may confound our findings. There are also significant changes in the ethnicity, maternal age, degree of lupus activity, and use of prednisone within this cohort over time. Pregnancy outcomes are similar for these 2 time periods, with the exception that more pregnancies were delivered at full term after 1995 (51% prior to 1995 and 63% after 1995). In order to control for secular changes, the year of delivery was included in the statistical modeling used to determine adjusted risks.
Another possible confounder might be the increased rate of lupus activity in the 6 months preconception in patients who ultimately discontinued HCQ during pregnancy. Several studies have demonstrated the great importance of lupus activity prior to pregnancy on lupus activity during pregnancy and on pregnancy outcomes (27, 35). Due to the fact that not all patients were followed up prior to pregnancy in the Hopkins Lupus Center, we are unable to make a statistically reliable determination as to whether the cessation of HCQ or the lupus activity prior to pregnancy was responsible for increased lupus activity. However, our analysis of the patients with inactive lupus seen prior to pregnancy suggests that when these patients stop taking HCQ, they increase their risk of experiencing increased lupus activity during pregnancy.
A combination of several factors places women who continued taking HCQ at lower risk of poor pregnancy outcomes in this cohort. No women who continued taking HCQ had increased lupus activity recorded prior to pregnancy. Few of these women had APS. Women who continued taking HCQ had lower rates of lupus activity, as measured by the PEA, flare rate, and SLEDAI. Fewer of these women took prednisone during pregnancy, and when they did, the dose was lower than in the other groups. Despite these potential benefits, the pregnancy outcomes were not significantly better among women who continued taking HCQ. This suggests that the benefit of HCQ during pregnancy is more symptom relief than an alteration in the underlying pregnancy pathology associated with SLE.
This study is not powered to discern small differences between the groups. This makes it impossible to conclude the absolute safety of HCQ during pregnancy. A larger study is necessary to confirm these findings. In order to discern a statistical difference in pregnancy loss similar to that found in our study (5%), based on a power calculation with an alpha error of 0.05 and a beta error of 0.8, a randomized controlled trial would require >400 pregnancies per study arm.
Based on this study, HCQ does not appear to cause toxicity to the fetus or harm pregnancy outcomes. Its continued use promotes the maintenance of quiet lupus activity during pregnancy. Larger studies, preferably randomized and blinded, will be required to confirm these findings. In the meantime, we advocate the continuation of HCQ treatment during pregnancy for women with lupus.