Introduction

Clinical trials in drug development generally exclude pregnant women for ethical reasons but leave questions about the safety of new medications on the developing human fetus unanswered upon drug approval and marketing. Despite lingering safety questions, pregnant women may intentionally or inadvertently be exposed to various prescription drugs for pregnancy and nonpregnancy indications. Post-marketing observational studies have revealed associations between many commonly used drugs and various birth defects [1–3]. Current utilization studies that ascertain the most commonly used drugs in pregnancy are important for establishing priorities in birth-defects research with major public health implications [4].

Studies conducted among pregnant women in the U.S. and in some European countries document high rates of exposure to prescription medications, including exposure to medications with known teratogenic potential [5–17]. Engeland and colleagues found that among more than 100,000 pregnant women in Norway in 2004–2006, approximately 57% received a prescription medication [5]. Similarly, Andrade and colleagues found that among more than 150,000 deliveries in the U.S., nearly two-thirds of women received a prescription drug during the pregnancy period, including approximately 5% who received a drug with a U.S Food and Drug Administration (FDA) category X designation, which is reserved for those drugs for which risks of fetal harm outweigh any possible benefit [6].

Country-specific utilization patterns cannot be reliably extrapolated to other countries since utilization patterns differ widely by geography because of differences in drug approval policies, prescribing patterns, and pricing and reimbursement policies. Additionally, utilization patterns differ temporally within countries because drugs are continually added to or removed from markets, and safety and efficacy information and concerns change constantly [4].

Little is know about current drug use during pregnancy in Italy. In 2000, Donati and colleagues published a study based on interviews with women to record data on drugs used during pregnancy [18]. As Olesen and colleagues point out, women may not always disclose use of all drugs during pregnancy [19] so the accuracy of interview-based utilization records is questionable. Despite limitations of their own, automated databases are valuable for research on drug utilization in pregnancy because they can provide detailed prescription information collected prospectively for large numbers of pregnant women [4].

The aim of this study was to provide an updated estimate of the magnitude of prescription drug exposure during pregnancy in Italy, using the Regione Emilia-Romagna (RER) health care database and to describe the extent to which pregnant women are exposed to drugs with potential for fetal harm.

Methods

Data source

The RER database is a population-based longitudinal health care database for the entire region of approximately 4 million RER inhabitants. Since 2000, this comprehensive automated database has prospectively captured information on services rendered in various health care settings, such as hospital discharge abstract data, including diagnosis and procedure codes, admission and discharge dates, and payments based on diagnosis-related groups; and individual prescription-level outpatient pharmacy data. The pharmacy data include records for all drugs reimbursed by the Italian National Health Service (NHS). Data from each file (e.g., hospital discharge abstract data and outpatient pharmacy data) are linkable via anonymized unique patient identifiers.

Study subjects

Only individuals who were RER residents for a full year prior to delivery, and thus had data available in the RER database for this period, were eligible for inclusion in this study. Female residents of RER who delivered a baby in a hospital between January 1, 2004, and December 31, 2004 were identified using inpatient data as those individuals with a recorded International Classification of Diseases-9th revision (ICD-9) code indicating delivery. The first record of any one of these codes was taken as the date of delivery as this was the first available evidence to confirm the commencement of delivery. If a woman delivered more than once during the study period, data for only the first delivery were included in the analysis.

The Chronic Condition Drug Group (CCDG) method was used to identify the most common medical conditions afflicting the cohort of study subjects prior to pregnancy [20]. Adapted from the Chronic Disease Score classification modified by Clark and colleagues [21], CCDGs utilize pharmacy dispensing data to identify up to 31 different common chronic conditions. We tabulated CCDGs in this cohort using pharmacy dispensing data for 1 year prior to the defined date of the beginning of gestation, which is described below. CCDGs were originally developed in the same database as used in this study and have been used to identify chronic disease burden in other published studies [22, 23].

Drug exposure ascertainment

We used the delivery-date algorithm described and used by Andrade and colleagues [6, 7] and validated by Toh and colleagues [24] to define the first day of the exposure period as a proxy for the first day of gestation. In doing so, we linked the cohort of women with deliveries in 2004 from the hospital data file with outpatient pharmacy data and reconciled outpatient prescription drug use in the 360 days prior to the delivery date. We then stratified this gestational period into three 90-day intervals corresponding to trimesters of pregnancy and one 90-day period before pregnancy (i.e., the period between 360 and 271 days before delivery). Pregnancies with abortive outcomes, including both spontaneous and elective abortions, were not considered in this analysis because the pregnancy period would almost certainly be less than 270 days.

Drugs dispensed during each 90-day interval were classified by anatomical group, according to the World Health Organization Collaborating Centre for Drug Statistics Methodology Anatomical Therapeutic Classification/Defined Daily Dose (ACT/DDD) System 2006 [25] and by pregnancy risk category, according to the U.S. FDA pregnancy risk classification system (A, B, C, D, and X; Table 1). FDA risk categories were identified by one investigator (J.G.) by reviewing each drug’s U.S. product label. If the FDA risk classification category was not specified in the product labeling, the category assigned by Briggs and colleagues [26] was used. If a product label with a corresponding FDA risk classification could not be identified and the drug was not listed in the guide by Briggs and colleagues, then the Australian risk classification system [27] was used (Table 1), as suggested by Amann and colleagues [28]. The Briggs classification system and the Australian risk classification system use A, B, C, D, and X categories similar to those of the FDA system. Drugs that could not be identified by any of these resources were considered not classified. If the product labeling indicated different FDA risk classification categories according to trimester of use (e.g., angiotensin-converting enzyme inhibitors), then the drug classification reflected the corresponding trimester of use in the analysis stratified by risk category. Prescription drugs not reimbursed by the Italian NHS (e.g., most anxiolytics), as well as nonprescription drugs (e.g., over-the-counter medications) and complementary medicines (e.g., herbal preparations), are not captured by the RER database and, thus, were not included in this analysis.

Table 1 Descriptions of pregnancy risk classification systems
Full size table

This study was approved by the Institutional Review Board of Thomas Jefferson University, Philadelphia, PA.

Results

We identified 33,343 deliveries during the 1-year study period. The mean age of these women was 32 years and the majority of women (86%) were between ages 25 and 39. The most common chronic conditions afflicting these women prior to pregnancy were thyroid disorders (4.3% of women), cardiovascular diseases (2.6%), and psychiatric conditions (2.4%). For 70% of deliveries (n = 23,480), at least one prescription medication was filled during the pregnancy period (Table 2). In 48% of deliveries (n = 16,007), at least one prescription drug other than a vitamin and mineral product was dispensed during the pregnancy period. Women who received at least one medication received an average of 2.8 drug dispensings and 1.8 compounds with unique ATC codes during pregnancy. A total of 13,577 women (41% of entire cohort) received at least one drug in the first trimester, while 49% (n = 16,364) received at least one drug in the second trimester, and 59% (n = 19,669) received at least one drug in the third trimester.

Table 2 Demographics and characteristics of Regione Emilia-Romagna, Italy, residents who delivered a baby in 2004 (n = 33,343)
Full size table

The most commonly prescribed single chemical entity during pregnancy was iron, with more than one-third of women having been prescribed one of the various formulations of iron supplements (Table 3). Of the 25 most commonly dispensed entities, 9 were oral anti-infectives – amoxicillin, fosfomycin, ampicillin, azithromycin, clarithromycin, erythromycin, cefixime, ciprofloxacin, and spiramycin (Table 3). Female reproductive hormones, such as progesterone (7%, n = 2,222), were also commonly prescribed.

Table 3 Most common prescription drug exposures during pregnancy in Regione Emilia-Romagna, Italy, 2004 (n = 33,343)
Full size table

Based on anatomical groupings, drugs used for conditions of the blood and blood-forming organs were observed in the highest proportion of women (41%). Other groups accounting for large numbers of dispensings included anti-infectives for systemic use (37%), drugs for conditions related to the alimentary track and metabolism (13%), and drugs for the genitourinary system and sex hormones (12%; Table 4).

Table 4 Prescription drug exposures during pregnancy in Regione Emilia-Romagna, Italy, by anatomical group,a 2004 (n = 33,343)
Full size table

Using a combination of the FDA, Briggs, and Australian classification systems, we determined that 49% of women (n = 16,482) received a drug from category A; 48% (n = 15,935) received a drug from category B; 19% (n = 6,476) received a drug from category C; 2% (n = 508) received a drug from category D; 1% (n = 292) received a drug from category X, and 7% (n = 2,207) received a drug that could not be classified during pregnancy (Fig. 1).

Fig. 1
figure 1

Percentage of women with drug dispensings in each 90-day study interval by pregnancy risk category (n = 33,343 deliveries)

Full size image

The most commonly dispensed category D drugs are listed in Table 5 and include atenolol (48 deliveries), carbamazepine (41 deliveries), and phenobarbital (33 deliveries). Three of the top six most commonly dispensed category X drugs were HMG-CoA reductase inhibitors (i.e., statins) – simvastatin (49 deliveries), atorvastatin (28 deliveries), and pravastatin (12 deliveries; Table 5). A total of 95 women (0.3%) were exposed to statins during pregnancy and 262 (0.8%) were exposed to angiotensin-converting-enzyme (ACE) inhibitors.

Table 5 Most common pregnancy risk category D and X prescription drug exposures before and during pregnancy, excluding hormones, in Regione Emilia-Romagna, Italy, 2004 (n = 33,343)
Full size table

Discussion

In this study that included more than 30,000 deliveries in Italy in 2004, we found that a large proportion of women (70 and 48% after excluding vitamin and mineral products) were dispensed at least one drug in the 270 days before delivery, including a considerable number who were exposed to drugs with pregnancy risk designations D and X. This prevalence of exposure is of similar magnitude as those observed in other recent studies in other countries [6, 7, 9, 11]. However, the prevalence of exposure to specific drugs varies among studies. For example, some (e.g., amoxicillin, erythromycin, ampicillin) but not all (e.g., fosfomycin, azithromycin, clarithromycin) of most common exposures to anti-infectives in our study are similar to the most common anti-infectives dispensed during pregnancy in Germany [28].

Experts in the fields of teratology and drug safety agree that research priorities should target the drugs that are most commonly used in pregnancy because they have the potential for the largest public health impact [4, 29]. Although we studied drug use in a specific Italian region, we expect that the results are generalizable to the rest of Italy since the same reimbursement rules apply generally across regions and RER demographics reasonably represent the larger Italian population.

We found that nearly 1 in every 100 women had a record of a category X drug, which is contraindicated in pregnancy. This proportion is slightly smaller than that observed in other studies [6, 10] and provides some reassurance that there is no major concern in current prescribing patterns in RER as compared to other countries in this context. In particular, a small, but measurable, proportion of subjects (0.8%) were exposed to ACE inhibitors, specifically in the first trimester. According to the U.S. FDA pregnancy risk classification system, ACE inhibitors are contraindicated in the second and third trimesters of pregnancy because of an association with an increased risk of fetopathy, but they have pregnancy risk designations of C for the first trimester. A recent study, however, found that infants with only first-trimester exposure to ACE inhibitors had an increased risk of major congenital malformations [risk ratio, 2.71; 95% confidence interval (95% CI), 1.72 to 4.27] as compared to infants who had no exposure to antihypertensive medications [1]. In addition, a meaningful proportion of women (0.3%) were exposed to statins, which are designated pregnancy risk category X because of numerous reports of adverse pregnancy outcomes after exposure to these drugs [30].

Despite evidence of potential fetal harm associated with certain medications, these drugs are still occasionally dispensed to pregnant women as documented in this study and in other studies. From clinical and public health perspectives, understanding how this problem arises is critical to reducing the potential adverse effects of drug use during pregnancy. We offer a possible explanation for why women continue to receive category X drugs during pregnancy and a therapeutic risk management solution. It is unlikely that health professionals prescribe drugs contraindicated in pregnancy to women knowing that they are pregnant, unless no treatment alternatives exist. More likely, women who are not pregnant and who do not intend to become pregnant, but nevertheless become pregnant, consume these medications and continue to do so through at least part of their pregnancies without realizing that they are exposing their unborn babies to these potentially dangerous drugs. This rationale is supported by evidence suggesting that many pregnancies are unplanned [31].

As much as possible, the administration of high risk drugs should be avoided in all women of childbearing potential regardless of pregnancy status and intention to become pregnant. However, we acknowledge that even when evidence of fetal risk does exist for certain drugs, in some circumstances clinicians must prescribe these drugs during pregnancy if, for example, alternative treatments are not available, if other drugs have been tested and failed, or if patients cannot stop taking the drugs because withdrawing from them would have serious adverse effects. In these scenarios, a woman and her partner should be counseled about the potential risk of fetal harm associated with the medication and a thorough risk-benefit discussion should ensue with documentation to follow. Typically, the drugs involved in these scenarios are those with category D designations. For example, several anticonvulsant drugs were among the most commonly dispensed category D drugs. These drugs are associated with an increased risk of major congenital abnormalities [32], but are critical in epilepsy management. The major limitation of the current risk classifications systems is that they are “based on the degree to which available information has ruled out risk to the fetus, balanced against the drug’s potential benefits to the patient” [33, 34], and for most medications adequate safety research has not yet been conducted making accurate risk classification difficult, if not impossible, for many drugs. Furthermore, pregnancy risk classification systems may have limited utility in clinical settings in which individual risk-benefit scenarios vary. Nevertheless, such systems can serve as a beacon to guide general treatment decisions and have been used in similar studies of drug use in pregnancy, thereby enhancing the comparability of this study to others from other countries [7, 10, 11, 28].

Category X drugs, however, carry a definite risk of fetal harm which outweighs any possible benefit. Generally, these drugs should be avoided in all women of childbearing potential. In the rare circumstance that a drug that is contraindicated during pregnancy is required for a woman of childbearing potential, health care professionals should either perform a pregnancy test to confirm that the woman is not pregnant with follow-up documentation or should document that appropriate contraception is being used.

A limitation to this study is that, although using the RER database avoided the potential for maternal underreporting of drug exposures, we were not able to ascertain whether drugs that were dispensed before delivery were taken during pregnancy. One of the major limitations of administrative pharmacy data is that they serve only as a proxy for drug exposure as they cannot ensure that medications are actually consumed. It is possible that a dispensing during pregnancy might be meant for maternal use after delivery (e.g., contraceptive hormones dispensed during the third trimester of pregnancy), leading to a possible overestimation of the burden of drug use during pregnancy. On the other hand, it is possible that some of the medications not classified in our study could be potential teratogens and thus could contribute to our prevalence estimates of category D and X drugs as underestimates. Also, nonprescription drugs such as over-the-counter analgesics and herbal products were not captured in the database, further contributing to a possible underestimate of risk exposure. Furthermore, the proportion of women reportedly exposed to folic acid in this study is likely a significant underestimate since, prior to 2005, folic acid was reimbursed in Italy only for the treatment of megaloblastic anemia. Thus, women who used folic acid before or during pregnancy for prevention of congenital malformations in 2004 were not captured in the RER database.

We applied a delivery-date algorithm to ascertain drug exposures during the gestational period that has been used by other researchers [6, 7] and recently validated [24]. Toh and colleagues found that the sensitivity of this approach in identifying any drug use in pregnancy is approximately 90.0% (95% CI, 86.6–92.7) with a specificity of 99.3% (95% CI, 99.0–99.6). The sensitivity and specificity of our approach could have been slightly enhanced by only including deliveries that were not associated with ICD-9 codes indicative of conditions commonly related to preterm births [24].

A limitation of applying this algorithm to examine drug utilization during pregnancy is that it begins by identifying only those women who delivered. Pregnancies with abortive outcomes, therefore, were not included. Despite being ignored, pregnancies with abortive outcomes may be important in the larger picture of assessing drug utilization during pregnancy because it is possible that medications could have been complicit in the abortive outcome. However, in a descriptive study such as this one, it would be impossible to ascertain whether the outcome was due to the drug use or to the condition for which the drug was used, or due to some other explanation all together.

Conclusions

More than two-thirds of women who delivered a baby in RER, Italy, in 2004, were exposed to one or more prescription medications during pregnancy. Nearly 1 in every 100 pregnant women was prescribed a contraindicated medication with a pregnancy risk designation of X. As much as possible, administration of high risk medications should be avoided in women of childbearing potential to obviate subsequent use during pregnancy. In the rare occasion when this is not possible, other appropriate therapeutic risk management strategies may be warranted. Additional research is required to elucidate associations between medications and birth defects and should focus on the drugs most commonly prescribed during pregnancy.