Correlation of drug dose estimated from national prescription registers with mean blood level of antiseizure medication in pregnancy

The purpose was to examine the correlation of antiseizure medication drug dose estimated from prescription fill records from prescription registers with blood levels during pregnancy.

• Antiseizure medication dose estimated from prescription fills was a good proxy for blood levels and thus for biological exposure in pregnancy.
• The finding suggests that administrative prescription fill records constitute a valuable resource for estimating drug exposure to antiseizure medication in pregnancy.

Plain Language Summary
Administrative prescription registers provide information on use of drugs and may be used to assess safety of drugs in specific populations, for example, in pregnant women.The information stems from the filling of prescriptions and it is uncertain whether this reflects how much of the drug is ingested.This study assesses whether pregnant women take the medication by studying the relationship between the estimated dose of drugs from prescription register with blood levels.The study found a close relationship between drug dose estimated from the prescription fill and mean blood levels suggesting that information from prescription registers is a good estimate of exposure to drugs in pregnancy.

| INTRODUCTION
Considerable concern surrounds drug treatment during pregnancy 1 ; however, pregnant women are usually excluded from pivotal regulatory trials.Therefore, risk and benefit assessment for maternal and fetal drug treatment stems from post-approval observational studies. 1,24][5] Use of antiseizure medication in pregnancy on the other hand is associated with a number of adverse outcomes in the offspring including reduced fetal growth, 6 congenital malformations, 7,8 and adverse neurodevelopment. 9,10To address the concerns related to treatments with antiseizure medication in pregnancy-especially the risk of congenital malformations-a number of clinical, prospective registers have been established for example, The European and International Registry of Antiepileptic Drugs in Pregnancy, EURAP (1999), 11 The North American Antiepileptic Drug Pregnancy Registry (1997), 12 The UK and Ireland Pregnancy Register, 13 The Australian Pregnancy Register of Anti-epileptic Drugs (1999), 14 and The Kerala Registry of Epilepsy and Pregnancy (1999). 15These registers have provided evidence of increased malformation rates associated with exposure to specific antiseizure medications used in pregnancy. 16,17However, information about risks also stem from national health registers that provide individuallevel register linkage.This allows for assessment of a wide range of risks associated prenatal exposure to antiseizure medication and adjustment for potential confounders, such as socioeconomic indicators, other medical treatments, and underlying co-morbid psychiatric disorders. 2Thus, nation-wide health registers offers an alternative to clinical collection of safety data in prospective registers, 18 and recently the European Medicines Agency issued warnings against specific antiseizures medications 19 based on national register-based cohort studies. 10,20wever, a key challenge in such studies is validating and estimating drug exposure that is usually estimated from prescription fill during pregnancy.A Danish study reported good compliance for antiseizure mediations among pregnant women, 21 but estimates of drug dose in pregnancy based on prescription fill patterns has never been compared with blood level measurements before, during, and after pregnancy. 22Drug dose of antiseizure medication is usually adjusted in pregnancy to account for altered clearance necessitating close monitoring with regular blood level monitoring. 23Using more recently available data on laboratory results linked with other data collected in Danish registers, we have a unique opportunity for assessing the correlation between drug exposure estimated from administrative prescription fill records and actual biological exposure based on blood level measurements in pregnancy, and thus, validate how national health registers may contribute to drug safety surveillance.
Accordingly, we used data from the Danish National Prescription Register 24 linked with data from the Register of Laboratory Results for Research 25 to study the association between estimated drug dose and mean blood levels in pregnancy.

| Study design and population
In a register-based study, we used the Danish Medical Birth Register 26 to identify all pregnancies resulting in still or live birth between 1 January 2014 and 31 December 2018 in Denmark.We excluded pregnancies with missing gestational age data and unrealistic live birth values (≤21 weeks or ≥ 45 weeks).Date of the last menstrual period (LMP) was estimated by subtracting gestational age at birth (in days) from the date of birth recorded in the Medical Birth Register. 26Using the unique Danish personal identification number, 27 we linked data from blood sample analyses in mothers identified in the Register of Laboratory Results for Research 25 to the same mothers who filled prescriptions for antiseizure medication during pregnancy identified in the Danish National Prescription Registry. 24

| Antiseizure medication blood levels
The Register of Laboratory Results for Research 25 is a nation-wide register holding information on biological samples as from 2013.The register lacks complete coverage as individual laboratories started reporting their results at different times.We identified blood level measurements before (LMP À 365 days to LMP), during (LMP to birth), and after (Birth + 365 days) pregnancy for antiseizure medications using the Nomenclature for Property and Unit (NPU) codes 28 NPU03735 (valproate), NPU08732 (lamotrigine), NPU18848 (levetiracetam), NPU01457 (carbamazepine) and NPU09119 (topiramate).For multiple same-day measurements, we used the value from the initial sample. 22Based on antiseizure medication blood levels before, during, and after pregnancy, we estimated the mean blood level during pregnancy as the area under the concentration curve divided by the number of days from LMP to birth (Figure S1).Additionally, we estimated the mean blood levels before pregnancy (LMP À 365 days to LMP), and after pregnancy (date of birth to 365 days after birth) in women with at least one blood level measurement in that time.

| Prenatal exposure to antiseizure medication from prescription fill records
We obtained information on maternal antiseizure medication use during pregnancy from the Danish National Prescription Registry. 24Information included the date of dispensing and the Anatomical Therapeutic Chemical (ATC) classification code.We identified mothers who had redeemed ≥1 prescription for antiseizure medication with the following ATC codes: N03AG01 (valproate), N03AX09 (lamotrigine), N03AX14 (levetiracetam), N03AF01 (carbamazepine), and N03AX11 (topiramate) from LMP minus 30 days to birth.The mean daily pregnancy dose of antiseizure medication was calculated by dividing the total antiseizure medication amount filled from 30 days before LMP to birth by the corresponding number of days.
The defined daily dose (DDD) is the assumed average maintenance dose per day for a drug used for its main indication in adults. 29Using the DDD, we divided the estimated daily dose of antiseizure medication into low (<50% DDD), medium (50%-100% DDD), and high (>100% DDD).In mothers who had blood level measurements, but no prescription fills in pregnancy, the estimated dose of antiseizure medication in pregnancy was categorized as 0 mg.To further characterize the study population, we assessed the mean drug dose, maternal age, maternal epilepsy, maternal psychiatric co-morbidity, maternal use of psychotropic drugs, maternal firstborn status, year of birth of the child, co-habitation status and maternal education in women who filled prescriptions for antiseizure medication stratified on whether the women had blood level monitoring recorded during pregnancy.

| Statistical analyses
Linear regression was used to estimate the correlation between dose of antiseizure medication estimated from prescription fills and mean blood levels during pregnancy.For the linear correlation between estimated dose and mean blood level in pregnancy, we calculated Pearsons's correlation coefficient. 30In further assessment, we grouped the estimated drug dose into 5 groups for valproate and carbamazepine and 20 groups for lamotrigine and levetiracetam and plotted the estimated drug dose against the mean blood level.Finally, we stratified the estimated dose of antiseizure medication into low, medium, and high dose [31][32][33] and correlated them with mean blood levels.

| Ethics statement
All data were analyzed at Statistics Denmark using encrypted identification numbers with no contact to the individuals.By Danish law, analysis of anonymous data does not require ethical review board approval.

| RESULTS
From the Danish Medical Birth Register, 26 1.The estimated mean dose of antiseizure medication in pregnancy based on prescription fills and the proportion of women who were diagnosed with epilepsy were higher in women who had blood levels measured compared with women who did not have blood levels measured.For valproate and lamotrigine, the proportion of women with psychiatric disorders and proportion of women using psychotropic drugs were higher in women who did not have blood levels measured compared with women who had blood levels measured, whereas there were only minor differences in the proportions with psychiatric disorders for levetiracetam, carbamazepine, and topiramate.For firstborn status, year of birth, married or cohabiting status and maternal education there were only minor differences between women with and without blood level measurements.The distribution of estimated daily antiseizure medication dose is shown in Figure S2.
We identified blood level measurements during pregnancy for 50 pregnancies for valproate, 870 pregnancies for lamotrigine, 330 pregnancies for levetiracetam, and 50 pregnancies for carbamazepine (Table 2).The average number of blood levels analyzed during pregnancy ranged from 3.7 (SD 2.6) to 5.8 (SD 2.7) (Table 2).
Most pregnancies had both blood level measurements and prescription fill in pregnancy (81%-97%) (Table 2).Blood level measurements before or after pregnancy were available for 31%-74% of pregnancies with measurements during pregnancy (Table 2).The distribution of antiseizure medication blood levels in pregnancy is presented in Figure S3.
When estimated daily dose was stratified into groups, the mean blood levels increased with increasing estimated drug dose (Figure 1).
When categorizing estimated drug doses into low, medium, and high based on DDD as in previous studies, 20,31 a clear association was seen between estimated drug dose and mean blood level (Figure 2).
When estimated daily doses before and after pregnancy were stratified into groups, the mean blood levels increased with increasing estimated drug dose (Figure S4A and S5A).

| DISCUSSION
We found a strong correlation between estimated daily dose and mean blood level of antiseizure medication during pregnancy, suggesting that dose estimates from prescription registers mirror biological drug exposure during pregnancy.This finding has implications for the interpretation of studies of prenatal exposure to antiseizure medications relying on administrative prescription fill records 31,32 and possibly for other chronic disorders medications in pregnancy.
Previous studies utilizing prescription registers for drug exposure [31][32][33] had to make assumptions about pregnant women actually ingesting the redeemed antiseizure medication prescriptions.Although a Danish study reported good compliance for antiseizure and other chronic disorder mediations among pregnant women, 21 the present finding supports that prescription fill for antiseizure medication is a reasonable proxy for drug exposure in pregnancy.
For all drugs studied, the regression coefficient for the correlation between drug dose and blood concentration was smaller during pregnancy compared with the regression coefficients before and after T A B L E 2 Measurements of blood levels before, during, and after pregnancy.pregnancy.A lower regression coefficient is an indicator of higher clearance and the difference in the regression coefficient between pregnancy and before/after pregnancy was larger for lamotrigine and levetiracetam compared with valproate, carbamazepine, and topiramate.These findings align with previous studies, 5,23,34 suggesting that lamotrigine and levetiracetam are subject to larger changes in clearance during pregnancy than valproate, carbamazepine, and topiramate.
We used the WHO DDDs to split estimated dose into low, medium, or high dose thus replicating the approach used in previous

Topiramate
F I G U R E 1 Mean blood level as a function of estimated daily drug dose for valproate, lamotrigine, levetiracetam, carbamazepine, and topiramate.Individuals were divided into 20 equal-sized groups for lamotrigine and levetiracetam and 5 groups for valproate and carbamazepine and 4 groups for topiramate based on the estimated daily dose.The plots show the mean blood level within each of these groups with 95% confidence intervals.The dose estimates are centered for each group.
studies. 20,31There is no universal agreed way of defining drug exposure levels (i.e., low, medium, and high) in pregnancy, but our dose categories align with those used in previous studies of dosedependent risk of malformations with antiseizure medication. 35The aim of our study was not to define cutoff values for drug exposure in pregnancy over which a certain effect could be observed, or exposure be considered safe-but merely to show that the estimated drug dose correlated with both blood levels when analyzed as continued measure (the regression analyses) and when categorized into high, medium, and low levels.For valproate and lamotrigine, the proportion of women who were diagnosed with epilepsy was higher in women who had blood levels measured compared with women who did not have blood levels measured, which contrasted with women with psychiatric disorders where the opposite was found.This suggests that most women who have their blood levels for valproate and lamotrigine measured in pregnancy drugs received treatment primarily for epilepsy.For the other antiseizure medications studied, the picture was less clear, but for levetiracetam the only indication for use is epilepsy, which was also reflected by the high proportion with blood level measurements in pregnancy (93%).

Topiramate
F I G U R E 3 Correlation of the estimated daily dose of antiseizure medication with mean blood levels of valproate, lamotrigine, levetiracetam, carbamazepine, and topiramate in pregnancy.For privacy reasons, it is not possible to show scatterplots indicating individual people.The plots show the linear regression with 95% confidence intervals for the correlation between the estimated daily dose (mg) and the mean blood level (μmol/L) of antiseizure medication in pregnancy.
This study carries several limitations.First, the women included with measurements of blood levels may exhibit better treatment adherence than those without.Second, therapeutic drug monitoring 22 may be less used for other indications than epilepsy, that is, the proportion of women who had measurements of blood levels was higher for levetiracetam, where epilepsy is the only indication, compared with lamotrigine that has several indications in addition to epilepsy.
Third, we had limited information on women who discontinued antiseizure medication for example, valproate, after learning that they were pregnant.Fourth, some antiseizure medications, especially lamotrigine, are subject to significant increase in clearance in pregnancy, which is clinically compensated for by increasing the drug dose thereby maintaining stable blood levels.This was also reflected in our data, with smaller regression coefficients (i.e., smaller changes in the blood level by unit increase in the estimated dose from prescription fill), but the correlation between the two measures remained relatively stable (Pearson's correlation for lamotrigine before: 0.57; during: 0.63; and after: 0.59 pregnancy).Fifth, the alteration of the clearance of antiseizure medication in pregnancy is highly drug specific 36 and varies between individuals. 37We did not account for the timing of blood levels in pregnancy for example, by trimester, which therefore may be source of uncertainty of the mean blood level estimate.However, even with this uncertainty, there was a good correlation between estimated drug dose and blood concentration.Sixth, for topiramate there was a close correlation between estimated drug dose and mean blood levels.However, most of the women who filled prescriptions for topiramate did not have measurements of blood levels in pregnancy and most of these women did not have a diagnosis of epilepsy suggesting that these women were using topiramate for other indication than epilepsy, for example, migraine.

| CONCLUSION
Post-approval observational studies are inherently subject to various biases, 1,38 however, we find that estimated drug dose based on prescription fill data from prescription registers correlated highly with blood levels, suggesting that the dose estimates may reasonably approximate actual drug exposure during pregnancy.
we identified 298 560 births including 1040 stillbirths, from 1 January 2014 to 31 December 2018.Among these pregnancies, we identified pregnancies with recorded prescription fill from the prescription register for valproate (N = 90), lamotrigine (N = 1360), levetiracetam (N = 340), and carbamazepine (N = 60), and topiramate (N = 100).In these pregnancies, T A B L E 1 Characteristics of women who filled prescriptions for antiseizure medication in pregnancy.

2
The estimated dose of antiseizure medication stratified into low, medium, and high dose with mean blood level for valproate, lamotrigine, levetiracetam, carbamazepine, and topiramate.The boxplot indicates the interquartile range, the line in the box indicates the median, the dot in the box indicates the mean, and the whiskers indicate the mean of the top/bottom 5 observations in each dose group."Low" indicates an average estimated daily dose <50% DDD, "Medium" indicates 50%-100% DDD, and "High" indicates >100% DDD.DDD, defined daily dose.