Introduction

Systemic (oral) treatment with the vitamin A derivative isotretinoin (13-cis-retinoic acid) is indicated in moderate to severe acne (e.g., nodular or conglobate acne or acne at risk of permanent scarring) resistant to therapy with systemic antibiotics and topical anti-acne treatment [1]. Systemic isotretinoin is considered to be the clinically most effective anti-acne therapy, achieving long-term remission or significant improvement in many patients [2,3]. At the same time, isotretinoin is one of the strongest human teratogens known today [4]. Given the age distribution of patients with acne, girls and women of childbearing age are among the patients to whom isotretinoin is prescribed.

In 2003, i.e., 20 years after the EU-market authorization for isotretinoin, uniform pregnancy prevention programs were established in order to avoid isotretinoin exposure during pregnancy. In 2018, these measures were evaluated and updated [5]. Girls and women initiating systemic isotretinoin treatment are required to have monthly pregnancy tests and to use 2 complementary contraceptive methods from 1 month before treatment initiation to 1 month after treatment cessation [5]. The continuation of contraception for 1 month after treatment cessation is recommended due to delayed plasma elimination [3]. Because of this delayed elimination, pregnancies beginning shortly after treatment with isotretinoin may also be exposed to potentially harmful plasma concentrations.

Given the potential harm to the unborn child, it is important to monitor the use of isotretinoin in girls and women of childbearing age as well as the frequency of pregnancies exposed to this drug. However, no such data are available from Germany. Also internationally, population-based studies quantifying the frequency of isotretinoin-exposed pregnancies are only available for few countries, namely the United States, Canada, France, and the Netherlands [6–11], and there are no studies on time trends of isotretinoin use among young women. Furthermore, there is a lack of studies systematically exploring the extent to which conceptions occurring in the month after discontinuation of treatment may contribute to the number of exposed pregnancies.

Therefore, the aims of this study were (i) to describe the utilization of isotretinoin in girls and women of childbearing age in Germany including time trends; (ii) to describe the frequency of pregnancies exposed to isotretinoin, considering also potential exposure due to treatment cessation in the month before pregnancy; and (iii) to explore potential malformations among children exposed to isotretinoin in early pregnancy (not to be mistaken with estimating causal effects).

Methods

Results

Prevalence of isotretinoin use among girls and women of childbearing age

Overall, there were 50,936 girls and women with at least 1 prescription of isotretinoin in GePaRD between 2004 and 2019. Across all years, 52% to 67% of users were 16 to 30 years old, and 81% to 90% were ≤40 years old (S1 Table). The overall age-standardized prevalence of isotretinoin use per 1,000 girls and women increased from 1.20 (95% confidence interval [CI]: 1.16, 1.24) in 2004 to 1.96 (95% CI: 1.92, 2.01) in 2019, i.e., by 63% (Fig 1). The prevalences increased particularly among girls and women aged 13 to 30 years. For example, in age group 16 to 20 years they increased by 103% (from 2.07 [95% CI: 1.92, 2.23] to 4.20 [95% CI: 4.00, 4.41] per 1,000) and in age group 21 to 25 years they increased by 96% (from 1.96 [95% CI: 1.82, 2.11] to 3.84 [95% CI: 3.67, 4.02] per 1,000).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Age-specific and age-standardized prevalence with 95% CIs (shaded area) of isotretinoin use per 1,000 girls and women aged 13–49 years between 2004 and 2019 in the GePaRD.

CI, confidence interval; GePaRD, German Pharmacoepidemiological Research Database.

https://doi.org/10.1371/journal.pmed.1004339.g001

In total, 339,408 prescriptions of isotretinoin were dispensed to girls and women aged 13 to 49 years during the study period (S2 Table). The vast majority of those prescriptions (89%) was issued by dermatologists while general practitioners had a share of 6%.

Exposed pregnancies

In the base case analysis, there were 178 pregnancies classified as exposed to isotretinoin during early pregnancy (Table 1). The number of exposed pregnancies per year increased over the course of the study period. On average, there were 7 pregnancies per year from 2004 to 2011 and 15 pregnancies per year from 2012 to 2019. The majority of the 178 pregnancies (62.9%) occurred in the age group 16 to 30 years (S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Number of pregnancies exposed to isotretinoin between 2004 and 2019 in GePaRD, mother’s age at pregnancy beginning and type of pregnancy outcome: base case and sensitivity analyses considering delayed elimination (exposure window assigned to the last dispensation before pregnancy was extended by 1 month), a fixed supply of 30 days, and the possibility that lower or higher doses than the DDD for isotretinoin (30 mg) were used.

https://doi.org/10.1371/journal.pmed.1004339.t001

The mothers’ median age at the beginning of pregnancy was 28 years (IQR 24 to 33 years). About half of these pregnancies were classified as exposed due to a dispensation of isotretinoin within the first 8 weeks of pregnancy, about one quarter because the exposure window assigned to the last dispensation before pregnancy overlapped the beginning of pregnancy and in the remaining, both criteria were fulfilled.

For exposed pregnancies ending during the observation period (n = 164), the distribution of pregnancy outcomes is summarized in Table 1. Overall, 29.3% (n = 48) of exposed pregnancies ended in live birth, thereof 6.3% (n = 3) were preterm births, 45.1% (n = 74) ended in an induced abortion, and 1.8% (n = 3) in a miscarriage. For 20.7% (n = 34) of pregnancies, no outcome was recorded (i.e., they are assumed to also be abortions).

Table 1 also shows the number of pregnancies and the distribution of pregnancy outcomes observed in the sensitivity analyses. When delayed elimination (washout period of 1 month) was considered, the number of exposed pregnancies increased by 31% (from 178 to 233), while the proportion of live births and induced abortions remained at a level similar to the base case analysis. If a fixed supply of 30 days was assumed, there was an increase of exposed pregnancies by 7% (from 178 to 190). If a daily dose of 40 mg was assumed, the number of exposed pregnancies decreased by 3% (from 178 to 172), if a daily dose of 20 mg was assumed, it increased by 6% (from 178 to 188) and if a daily dose of 10 mg was assumed, it increased by 58% (from 178 to 283) compared to the base case analysis. When the one-month washout period was considered for assumed doses of 10 mg, 20 mg, and 40 mg, the increase in the number of exposed pregnancies was 33% (from 283 to 375), 39% (from 188 to 261), and 30% (from 172 to 224), respectively. In all sensitivity analyses, the proportion of live births was below 40% except for those with an assumed daily dose of 10 mg when delayed elimination was considered (46%).

Characterization of exposed children

In 138 out of 157 (87.9%) pregnancies ending in a live birth and classified as exposed in the base case or the sensitivity analyses, the mother’s and child’s data could be linked. Of these, 6 children had at least 1 major congenital malformation according to EUROCAT definitions (Table 2). Five children had malformations classified as minor according to EUROCAT but are reported here as they required surgical or other intense treatment. Two other children had an atrial septal defect that could not be classified into major or minor according to EUROCAT given that the distinction requires information that is not available in German claims data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Malformations observed in live-born children exposed to isotretinoin during pregnancy in GePaRD between 2004 and 2019.

https://doi.org/10.1371/journal.pmed.1004339.t002

Of the 6 children with major malformations, 3 were classified as exposed in the base case analyses and 3 were classified as exposed in the sensitivity analyses. The major malformations affected the heart, the eye, the skull, and the nose. Furthermore, 1 child had sacral spina bifida with hydrocephalus and 1 child had multiple major malformations including microcephaly, malformation of the ear and heart. Among children with so-called minor malformations, 2 had congenital hypertrophic pyloric stenosis requiring surgical treatment, 2 had undescended testicles requiring surgical treatment and 1 had metatarsus varus with abnormalities of gait and mobility requiring extensive orthopedic treatment. Four of these 5 children were classified as exposed in the sensitivity analyses only.

Table 2 shows—for each of the 13 children with congenital malformations—in which of the analyses they were classified as exposed. In all of them, the exposure window assigned to the last dispensation before pregnancy overlapped the beginning of pregnancy. In the children classified as exposed in the base case analysis, the mother additionally had a dispensation of isotretinoin within the first 8 weeks of pregnancy, while no such dispensations were observed in the mothers of children classified as exposed in the sensitivity analyses only.

Discussion

In this population-based study covering approximately 20% of the German population, we found that the use of isotretinoin among girls and women of childbearing age increased by 63% between 2004 and 2019, from 1.20 per 1,000 in 2004 to 1.96 per 1,000 in 2019. This increase was particularly pronounced in girls and women up to the age of 30 years. Across the whole study period, more than 80% of users were ≤40 years (between 3,629 of 4,456 and 6,723 of 7,504), i.e., in age groups in which pregnancies typically occur. Even though risk minimization measures are in place, we observed—in the base case analysis—178 pregnancies likely exposed to isotretinoin during a time window most critical for fetal development. In sensitivity analyses varying the assumptions on the dose and considering the recommended one-month washout period, this number ranged between 172 and 375 pregnancies. In the base case analysis, at least 74 exposed pregnancies ended in an induced abortion (sensitivity analyses: 74 to 96). Among live births classified as exposed in the base case or the sensitivity analyses, there were 6 children with major malformations.

Regarding the increase in the prevalence of isotretinoin use among girls and women of childbearing age, there is hardly any data to which we could compare our findings. There is only a single data source reporting—for each year—the total number of DDDs dispensed to all persons with statutory health insurance in Germany (i.e., no denominator/prevalence estimate, and no age or sex-specific findings). Even though comparability is limited, that report supports our finding regarding an increase of oral use of isotretinoin (ATC code D10BA01) between 2004 and 2019 (2004: 5.2 million dispensed DDDs; 2019: 6.0 million dispensed DDDs) [24,25]. The mainstays of acne treatment have remained largely unchanged over recent years [26], so the increase in the prevalence of isotretinoin use cannot be explained by changes in treatment guidelines. Also, there have been no reports on major changes in the disease prevalence during the past 15 years. We can thus only speculate about reasons for this increase. German guidelines, which expired in 2014 without renewal, recommended systemic isotretinoin as second-line therapy after systemic antibiotics and topical anti-acne therapy, while the European guidelines recommend systemic isotretinoin as first-line therapy in moderate as well as severe acne [1]. In view of the effectiveness of systemic isotretinoin in the treatment of acne and the increasing awareness of the mental health burden associated with acne, German dermatologists might have been inclined to follow the European rather than the German guidelines. Another potential reason could be an increased awareness of antibiotic resistance related to the antibiotic stewardship programs launched during the past decade [27]. Considering recommendations from experts, regulatory authorities, and the WHO to shorten and reduce the number of treatment episodes with antibiotics, the threshold to proceed to second-line treatment with systemic isotretinoin might have become lower when treating acne patients [3,28].

In Germany, there is a long-established pregnancy prevention program for isotretinoin in line with EMA recommendations. In addition, prescriptions of isotretinoin must be filled within 7 days of being issued and the supply per prescription is limited to 30 days of treatment [20]. Despite these measures, we found—in the base case analysis—178 pregnancies likely exposed to isotretinoin in the critical time window. As our database covers approximately one fifth of the German population, it can roughly be estimated that at least 980 such pregnancies occurred across Germany between 2004 and 2019. As a matter of concern, the number of exposed pregnancies per year doubled during the study period, i.e., this problem is continuously gaining relevance. Our sensitivity analyses considering delayed plasma elimination and lower-dose treatment showed that the number of exposed pregnancies may even be considerably higher than estimated in the base case analysis. Of note, irrespective of the daily dose we assumed, the number of pregnancies classified as exposed increased by 30% to 39% if the one-month washout period was considered. Since this suggests that pregnancies often occur too soon after the end of isotretinoin treatment, increasing awareness regarding the component of the pregnancy prevention program that recommends contraception also in the month after treatment cessation seems important.

Our study confirms reports from other countries about isotretinoin-exposed pregnancies occurring despite long-standing pregnancy prevention programs [29]. However, there is a lack of studies quantifying the frequency of these pregnancies. Dividing the number of exposed pregnancies by the total number of users in our study yields (roughly) a ratio between 3.5 (base case analysis) and 6.7 (sensitivity analyses) exposed pregnancies per 1,000 isotretinoin users. This is similar to the results of a study from the US based on pregnancy reports from the FDA Adverse Event Reporting System (FAERS) showing between 3.3 and 6.5 exposed pregnancies per 1,000 users between 1997 and 2017 [10]. It is also similar to a study based on Canadian administrative data (data from the provinces British Columbia, Saskatchewan, Manitoba, and Ontario) from 1996 to 2011, in which this ratio varied between 3.1 and 6.2 pregnancies per 1,000 isotretinoin users depending on the exposure definition (high-specificity versus high-sensitivity) [7]. Regarding Europe, population-based studies estimating the frequency of pregnancies per 1,000 users are only available for France where this ratio ranged between 0.32 and 0.95 when combining data from studies conducted between 1987 and 2011 [9]. A population-based study from the Netherlands determined the proportion of isotretinoin-exposed pregnancies among all pregnancies. It reported that between 1999 and 2007 about 2.5 per 10,000 pregnancies were exposed to isotretinoin in the 30 days before or during pregnancy [11].

The number of exposed pregnancies ending in a live birth in our study was 48 of 164 (29%) in the base case analysis and increased to 157 of 340 (46%) in a sensitivity analysis where a dose of only 10 mg and delayed elimination were considered. This illustrates that the proportion of live births is sensitive to the exposure definition, which may hamper comparability between existing studies. Furthermore, it is influenced by the extent to which incomplete pregnancies are captured in the respective database, which may also vary between studies. This may explain variation in the proportion of live births reported in different studies. For example, a German study using data collected in the context of counseling pregnant women and their health care providers reported a proportion of 18 live births of 91 pregnancies (20%) [30], a study from the United States reported proportions of 68 of 138 (49%) [8], while much lower proportions of 118 of 1,473 (8%) were reported by a Canadian study [7] and 85 of 553 (15%) in French studies [9].

While our study was not designed to quantify risks, it was still striking that 13 live-born children classified as exposed to isotretinoin in early pregnancy in the base case or the sensitivity analyses had congenital malformations, 6 of them with major malformations. Many of these malformations involved organ systems known to be affected by the so-called retinoic acid embryopathy. This finding and the fact that at least 45% of exposed pregnancies (74 of 164) in our study ended in an induced abortion—as compared to 4% when all pregnancies in GePaRD are considered [17]—underline the importance of improving adherence to pregnancy prevention programs.

Some limitations should be considered in the interpretation of our results. First, German claims data do not include the dose prescribed by the physician. Consequently, treatment duration has to be estimated based on the DDD. The DDD represents the dose for adults, but lower doses may be used, particularly in girls and adolescents, or sometimes a higher dose might be used. To address this issue, we performed comprehensive sensitivity analyses varying the exposure windows assigned to each dispensation. Second, as in all pharmacoepidemiological studies, there is uncertainty whether patients filling a prescription are actually taking the drug. It is also uncertain whether they always start taking the drug after filling the prescription or whether they may partly start later. Third, while our study was designed to describe prevalence of isotretinoin use and pregnancies occurring under treatment with isotretinoin, our database would not have been suited to assess whether risk minimization measures were followed on an individual level. This would have required comprehensive information on contraceptive measures, which is limited in GePaRD as in most other claims databases [31]. Fourth, with regard to pregnancy outcomes and malformations, our study was merely descriptive, so causal conclusions cannot be drawn. Estimating causal effects would have required a different design including the consideration of relevant confounders, as well as a larger sample of exposed children. The latter might be achieved by a consortium of large databases. We strongly advise against simplified calculations in which the proportion of live births with malformations in our study is compared to the corresponding proportion reported for all live births in Germany in order to estimate risks. Such a comparison could be very misleading for various reasons (e.g., due to differences in age and thus presumably also in the prevalence of comorbidities, different proportions with induced abortions and thus missing malformation status). Fifth, as in most databases used to investigate drug utilization and safety during pregnancy, pregnancy outcomes clearly classified as “miscarriages” are underrepresented in our database, i.e., the frequency of this outcome was underestimated in our study. To address this limitation, we also searched for incomplete pregnancies with no outcome recorded. It seems plausible that these were miscarriages not requiring medical treatment or induced abortions without medical indication, i.e., not reimbursed by health insurances. Although not perfect as it still remained unclear whether it was a miscarriage or induced abortion, we think this approach was valuable to capture the number of exposed pregnancies more completely. Sixth, to assess the presence of malformations in children exposed during pregnancy, we conducted an in-depth patient profile review based on all diagnoses and procedure codes available in GePaRD but did not have additional clinical data.

A main strength of our study is the large claims database that has been shown to be representative of persons with statutory health insurance in Germany in terms of drug dispensations [32]. The available data allowed us to assess trends in isotretinoin dispensations over a 15-year period. Due to the use of claims data, our analyses were not affected by recall or non-responder bias. Furthermore, the sophisticated methods developed for GePaRD (i) to identify pregnancy outcomes [17], which were further optimized to capture incomplete pregnancies; (ii) to link mothers’ and babies’ data [19]; and (iii) to estimate the beginning of pregnancy—predominantly based on the estimated date of delivery—which is expected to minimize misclassification of gestational age [18], are strengths of our study. We also consider it a strength of our study that we conducted sensitivity analyses to systematically assess whether pregnancies may have started during the recommended one-month washout period after treatment cessation.

The prevalence of isotretinoin use among girls and women of childbearing age increased in Germany between 2004 and 2019, and there were a considerable number of pregnancies likely exposed to isotretinoin in a critical period. This highlights the importance of monitoring compliance with the existing risk minimization measures for isotretinoin in Germany.

Supporting information

Acknowledgments

The authors would like to thank Marieke Niemeyer and Philipp Alexander Volkmar for statistical programming of analysis datasets and double-independent programming of results as well as all statutory health insurance providers which provided data for this study, namely AOK Bremen/Bremerhaven, DAK-Gesundheit, Die Techniker Krankenkasse (TK), and hkk Krankenkasse.