A population-based study of maternal use of amoxicillin and pregnancy outcome in Denmark

Abstract

Aims

Amoxicillin is a widely used penicillin, but data on its safety in pregnancy are limited. We examined the association between amoxicillin exposure during pregnancy and birth weight, preterm delivery, congenital malformations, perinatal death, and spontaneous abortion.

Methods

We identified all primiparous women with a live birth, or a stillbirth after the 28th gestational week, from 1 January 1991–31 December 2000 in the County of North Jutland, Denmark. Data on prescriptions for amoxicillin and outcome were obtained from population-based registries. Using a follow-up and a case–control design, we compared pregnancy outcomes between women who had been prescribed amoxicillin during pregnancy and those who had not, adjusting for available potentially confounding factors.

Results

We identified 401 primiparous women who redeemed a prescription for amoxicillin during their pregnancy. The control group consisted of 10 237 primiparous women who did not redeem any prescriptions from 3 months before pregnancy until the end of pregnancy. The adjusted mean birth weight of children born to amoxicillin-exposed mothers was 57 g [95% confidence interval (CI) 9, 105] higher than that of children born to controls. Odds ratios among amoxicillin-exposed relative to controls were: low birth weight 0.63 (95% CI 0.26, 1.53), preterm delivery 0.77 (95% CI 0.49, 1.21), congenital malformation 1.16 (95% CI 0.54, 2.50), and spontaneous abortion 0.89 (95% CI 0.66, 1.18). We did not observe any cases of perinatal death in the amoxicillin-exposed women.

Conclusions

We did not find any increased risk of adverse pregnancy outcome associated with amoxicillin exposure during pregnancy, but additional studies are warranted.

Introduction

In Denmark, 28.7% of all prescriptions redeemed by pregnant women are for antibiotics, of which large proportions are broad-spectrum penicillins [1]. Amoxicillin is the most commonly used drug in this class of antibiotics [2], so any adverse effect of amoxicillin on pregnancy outcome could have considerable public health implications. While several studies have evaluated the effect of other penicillin drugs on pregnancy outcome [3–7], the only available data on the safety of amoxicillin in pregnancy derive from a single randomized trial including 252 pregnant women with probable gonorrhoea, who were treated with either amoxicillin, ceftriaxone, or spectinomycin [8, 9].

Thus, we have conducted a population-based study to examine the association between amoxicillin exposure during pregnancy and birth weight, preterm delivery, congenital malformations, perinatal death, and spontaneous abortion.

Methods

This study was conducted in the County of North Jutland, Denmark, which has approximately 490 000 inhabitants. It was carried out in accordance with the guidelines of the regional science ethics committee for use of clinical and registry data.

Exposure data

The North Jutland Pharmaco-Epidemiological Prescription Database

The National Health Service provides tax-supported health care for all inhabitants of the country. Apart from guaranteeing free access to general practitioners, hospitals, and public clinics, the insurance programme reimburses 50–75% of the costs of selected antibiotics, including amoxicillin. The pharmacies in the County of North Jutland are equipped with a computerized accounting system used primarily to secure reimbursement from the National Health Service. When prescriptions are redeemed, data are also sent from the accounting system to the North Jutland Pharmaco-Epidemiological Prescription Database, which contains data on all redeemed prescriptions since 1 January 1991. These data include the date, the type of drug redeemed (according to the Anatomical Therapeutical Chemical (ATC) classification system), and the customer's personal identification number, which incorporates information on gender and age.

Outcome data

The Danish Medical Birth Registry

This registry contains data on all live births in Denmark since 1 January 1973. The midwives and doctors responsible for the deliveries record data on all births, including birth weight, gestational age, gender, as well as the mother's personal identification number and smoking status.

The County Hospital Discharge Registry

We identified all congenital malformations, perinatal deaths (stillbirth after the 28th gestational week or death within the first week after birth), and spontaneous abortions through the County Hospital Discharge Registry. This registry was established in 1980 and transfers data to the National Registry of Patients, where 99.4% of all discharges from Danish somatic hospitals are recorded [10]. Recorded information includes personal identification number, dates of admission and discharge, surgical procedures performed, and up to 20 discharge diagnoses. Discharge diagnoses were classified according to the Danish version of the International Classification of Diseases, 8th revision until the end of 1993, and according to the 10th revision thereafter. The children were followed in the County Hospital Discharge Registry until 1 September 2001. The codes for congenital malformations in ICD-8 and ICD-10 were 740.00–759.99 and Q0.00–Q99.9, respectively, but congenital dislocation of the hip and undescended testes were excluded due to poor validity of these diagnoses [11]. The ICD-8 and ICD-10 codes for spontaneous abortion were 634.61, 643.8–9, 645.1, and O02, and O03, respectively.

Record linkage

Records from the North Jutland Pharmaco-Epidemiological Prescription Database were linked with the Danish Medical Birth Registry and the County Hospital Discharge Registry using the mother's personal identification number issued by the Central Office of Civil Registration. However, the child's personal identification number was used to identify congenital malformations in the County Hospital Discharge Registry.

Study design

Cohort analysis

The association between amoxicillin exposure and birth weight, preterm delivery, congenital malformations, and perinatal death was examined in a cohort of primiparous women with either a live birth or a stillbirth after the 28th gestational week between 1 January 1991 and 31 December 2000. We had to restrict the cohort to primiparous women, because prescriptions to children up to 15 years of age were recorded in the mother's name in the Danish National Health Service until 1 April 1996. The exposed group consisted of those who redeemed a prescription for amoxicillin during pregnancy, and these women were further divided by trimester of exposure, based on prescription date recorded in the Pharmaco-Epidemiological Prescription Database and on gestational age as reported to the Danish Medical Birth Registry. The control group consisted of all those who did not redeem any prescriptions in the time period from 3 months before pregnancy to delivery.

Case–control analysis

We used a case–control design to study the association between amoxicillin exposure and spontaneous abortion. Through the County Hospital Discharge Registry, cases were identified as all women whose first pregnancy between 1 January 1991 and 31 December 2000 resulted in hospitalization with spontaneous abortion. None of these women had given birth or had an abortion between 1980 and 1991. Use of amoxicillin in pregnancy was determined as having redeemed a prescription for amoxicillin less than 3 months prior to the date of admission with a diagnosis of spontaneous abortion. The control group consisted of women whose first pregnancy between 1 January 1991 and 31 December 2000 resulted in a live birth or a stillbirth after the 28th gestational week, and who had not given birth or had an abortion between 1980 and 1991. In this group, use of amoxicillin was determined as use of amoxicillin less than 3 months before the 12^th gestational week.

Statistical analysis

Cohort analysis

We used multiple regression analysis to estimate the difference in birth weight between children of women exposed to amoxicillin and controls. The model was adjusted for maternal age (< 25, 25–30, and> 30 years), gestational age (28–33, 34–36, and ≥ 37 weeks), and maternal smoking status (smoker, nonsmoker), using design variables.

We used logistic regression analyses to estimate the association between use of amoxicillin and low birth weight (≤ 2500 g), preterm delivery (< 37 weeks), congenital malformations, and perinatal death. The analyses were adjusted for maternal age and smoking, and these variables were coded as described above. The analysis of low birth weight was restricted to full-term deliveries (≥ 37 weeks), and only prescriptions for amoxicillin redeemed in the first trimester were included in the analysis of congenital malformations.

Case–control analysis

The risk of spontaneous abortion in women exposed to amoxicillin during pregnancy relative to controls was estimated using a logistic regression model adjusted for maternal age divided into three categories (< 25, 25–30, and> 30 years) using design variables.

Results

Cohort analysis

We identified 401 primiparous women who redeemed a prescription for amoxicillin during pregnancy, and 147 of them redeemed their prescription during the first trimester. The control group consisted of 10 237 primiparous women who did not redeem any prescriptions from 3 months before conception until delivery. In the group of women exposed to amoxicillin at any time during pregnancy, average birth weight was 3498 g, 5.0% had a preterm delivery, and 4.0% gave birth to a child with a congenital malformation. The corresponding numbers in the control group were 3429 g, 6.3%, and 4.1% (Table 1). There were no cases of perinatal death in the exposed group, so this outcome was not examined in logistic regression analysis.

Table 1.

Characteristics of the study cohort.

	Exposed to amoxicillinin the first trimester	Exposed to amoxicillinduring pregnancy	Controls
Number of women	147	401	10 237
Number of prescriptions	150	445	–
Mean age (range)	27.4 (18–42)	27.0 (16–42)	26.8 (13–45)
Proportion of smokers (%)	35.4	33.4	25.9
Gestational age (weeks)
≥ 37	139	381	9 588
34–36	8	18	450
28–33	0	2	199
Mean birth weight (range)	3518 (2270–4665)	3498 (1815–4675)	3 429 (516–6230)
Number of low birth weight* (%)	2 (1.4)	5 (1.3)	183 (1.9)
Number of preterm deliveries (%)	9 (6.1)	20 (5.0)	649 (6.3)
Number of malformations (%)	7 (4.8)	16 (4.0)	416 (4.1)
Number of perinatal deaths (%)	0 (0)	0 (0)	60 (0.6)

^*Restricted to full term deliveries ≥ 37 weeks).

According to the multiple regression analysis, adjusted mean birth weight of children born to mothers in the exposed group was 57 g [95% confidence interval (CI) 9, 105] higher than that of children born to controls.

In adjusted logistic regression analyses, risk of low birth weight [odds ratio (OR) 0.63, 95% CI 0.26, 1.53] and preterm delivery (OR 0.77; 95% CI 0.49, 1.21) was nonsignificantly reduced, while risk of congenital malformations (OR 1.16; 95% CI 0.54, 2.50) was close to unity among children born to mothers in the exposed group compared with those born to controls (Table 2).

Table 2.

Probability of pregnancy outcomes in the amoxicillin group relative to controls.

		Exposure to amoxicillin
	Crude odds ratio	95% CI	Adjusted odds ratio	95% CI
Cohort analysis
Low birth weight*	0.68	0.28–1.67	0.63†	0.26–1.53
Preterm delivery	0.78	0.49–1.22	0.77†	0.49–1.21
Congenital malformations	1.18	0.55–2.54	1.16†	0.54–2.50
Case–control analysis
Spontaneous abortion	0.92	0.69–1.23	0.89‡	0.66–1.18

^*Restricted to full-term deliveries (≥37 weeks).

^†Adjusted for smoking and maternal age.

^‡Adjusted for maternal age.

Case–control analysis

In the analysis of spontaneous abortion, we found that 52 (1.2%) of the women in the case group redeemed a prescription for amoxicillin compared with 437 (1.3%) of the controls. After adjustment for maternal age, the OR for spontaneous abortion was 0.89 (95% CI 0.66, 1.18) (Table 2).

Discussion

In this study, which we believe is the first follow-up study of the safety of amoxicillin in pregnancy, we found that women who redeemed a prescription for amoxicillin during pregnancy tended to give birth to babies with higher birth weight, but were not at increased risk for adverse pregnancy outcomes such as preterm delivery, congenital malformations, or spontaneous abortion.

Use of the prescription database, which has no more than 0.2% coding errors [12], ensured virtually complete registration of redeemed prescriptions. Furthermore, since patients are required to pay part of the cost of prescribed amoxicillin, the use of prescriptions that were actually redeemed is likely to have reduced misclassification due to noncompliance. However, it is still possible that not all amoxicillin was actually taken [13]. In-hospital use of amoxicillin is an additional source of misclassification as it is not recorded in the prescription database. Any misclassification of amoxicillin use will lead to a conservative bias.

We may have overestimated the effect of amoxicillin on pregnancy outcome, because women in the amoxicillin group could have redeemed prescriptions for other drugs than amoxicillin during pregnancy, whereas the controls did not redeem any prescriptions during pregnancy.

Registration of births and stillbirths was of high quality. The predictive value and completeness of diagnoses of congenital malformations in the County Hospital Discharge Registry have been estimated at 90% (unpublished data), but we do not believe that either predictive value or completeness could be associated with use of amoxicillin. However, we were not able to identify malformed fetuses detected at prenatal examinations and consequently aborted, and we do not know whether these malformations were associated with exposure to amoxicillin. Lack of completeness and of data from prenatal examinations may explain the slightly lower occurrence of congenital malformations in this study (4.1% among controls) compared with the 4.8% (after subtraction of congenital dislocation of the hip and undescended testes) reported in a Hungarian study [14].

If amoxicillin does cause congenital malformations, it would probably cause only a few specific malformations, and it would take a large number of these specific malformations to produce a statistically significant increase in congenital malformations overall [15]. We had the power to detect only a 2.5-fold increased risk of congenital malformations overall.

With respect to spontaneous abortion, we were able to identify only spontaneous abortions which resulted in hospitalization. Thus, we cannot report on the association between amoxicillin exposure and early spontaneous abortion that did not require hospitalization.

We assumed that all spontaneous abortions occurred at 12 weeks of gestation, because we did not know the exact gestational age at the time of the abortion. If it occurred before week 12 of gestation, we could have included prescriptions redeemed before conception. Conversely, if the spontaneous abortion occurred after week 12 of gestation, we may have missed prescriptions redeemed early in pregnancy. Thus, we may have misclassified amoxicillin exposure, resulting in a conservative bias.

Smoking is a risk factor of spontaneous abortion [16], but we were unable to adjust for smoking in our case–control analysis, because we did not know the smoking status of women with a spontaneous abortion. However, we did find a higher proportion of smokers among women exposed to amoxicillin in our cohort analysis (33.4%vs 25.9% among controls), suggesting that smoking is unlikely to explain the slightly protective effect of amoxicillin on spontaneous abortion.

Our finding of a higher birth weight among children born to mothers exposed to amoxicillin is consistent with findings from other studies of antibiotics in pregnancy [3, 4], but the opposite has also been reported [5]. However, we believe that the magnitude of the weight difference found in this study (57 g) was too small to have any prognostic impact. One explanation for our finding could be that antibiotic treatment during pregnancy prevents infections that would otherwise cause lower birth weight. However, confounding by socio-economic status, whereby women of low socio-economic status, which may be associated with low birth weight [17], did not receive appropriate antibiotic treatment, cannot be ruled out.

A previous randomized trial, including 252 pregnant women with probable gonorrhoea, found similar risk of congenital malformations among those who were treated with amoxicillin and in those treated with ceftriaxone or spectinomycin [8]. To our knowledge, there are no studies to date of the association between amoxicillin and either low birth weight, perinatal death, or spontaneous abortion. Augmentin is a combination of amoxicillin and clavulanic acid, and a Hungarian case–control study including 52 (0.75%) augmentin-exposed women among 6935 who had offspring with congenital malformations reported a nonsignificantly increased odds ratio for congenital malformations overall of 1.4 compared with 10 238 controls [7]. The risks of cleft lip or palate (OR 2.1; 95% CI 0.6, 7.1), poly/syndactyly (OR 2.5; 95% CI 0.4, 16.3), cardiovascular malformations (OR 2.6; 95% CI 1.1, 6.0), malformations of diaphragm (OR 7.0; 95% CI 0.4, 135.5), and hypospadias (OR 4.3; 95% CI 1.2, 15.4) were increased after exposure to augmentin at any time during pregnancy, and not only during the critical period for congenital malformations. As the data on augmentin exposure were based on questionnaires, recall bias might have contributed to the findings [18]. Pivampicillin and ampicillin are closely related to amoxicillin, and another Hungarian study of ampicillin use in 61 016 pregnant women, based on the same case–control dataset, did not report increased risk of serious congenital malformations [6]. Similarly, a Danish study of 199 women who received pivampicillin in the first trimester did not find any increased risk of malformations, preterm delivery, or low birth weight [4].

In conclusion, amoxicillin use during pregnancy does not appear to increase the risk of adverse pregnancy outcome, but even this large study cannot rule out an effect.