- Top of page
Summary: Purpose: Concerns over teratogenicity of antiepileptic drugs (AEDs) during pregnancy must be balanced against the risks of seizures to the mother and developing fetus. Pharmacokinetic changes and vomiting may alter drug levels, but more important may be the patient's decision to stop medication before or during pregnancy. Compliance assessment traditionally relies either on self-reporting or on AED plasma level monitoring; neither provides reliable information on drug-taking behaviour over an extended interval (e.g., before, during, and after pregnancy).
Methods: We have used hair analysis to assess AED-taking behavior in pregnant women compared with nonpregnant female controls. Twenty-six pregnant women [mean age, 27.5 ± 6.7 (SD) years] and 13 nonpregnant female epilepsy outpatients (mean age, 31.9 ± 8.3 years) were studied. Carbamazepine (CBZ) or lamotrigine (LTG) concentrations were measured in 1-cm hair segments, and the within-subject variance in segmental hair concentrations of these drugs was calculated for each group. The variances of each group were then compared by using a variance ratio test.
Results: The variance of AED concentration in hair differed significantly between the pregnant and nonpregnant groups [variance ratio, 1.59 (p < 0.01)]. Four (15%) of the 26 pregnant patients had little or no AED in their proximal hair segments compared with more distal segments, apparently having discontinued their medication during pregnancy. Only one of these later disclosed having stopped her medication. One pregnant woman whose hair profile was similar to controls died suddenly at 30 weeks of gestation.
Conclusions: This study confirms the perception that pregnant women with epilepsy frequently stop or greatly reduce their prescribed medication, usually without reference or acknowledgement to their clinician.
The teratogenic potential of antiepileptic drugs (AEDs) is a major concern to women planning a pregnancy and to clinicians managing women with epilepsy. Conventional AEDs such as phenytoin (PHT), carbamazepine (CBZ), and sodium valproate (VPA) have been clearly associated with teratogenicity, but as yet there are insufficient human pregnancy data on the newer agents such as lamotrigine (LTG), gabapentin (GBP), tiagabine (TGB), topiramate (TPM), oxcarbazepine (OCBZ), and levetiracetam. Doctors advising young women with epilepsy must consider whether AEDs are necessary and, if indicated, aim for monotherapy at the lowest effective dose (1). The uncertain risk of teratogenicity must be balanced against first, the physical risk of maternal seizures (including sudden unexplained death in epilepsy), and second, the social penalties of having seizures, especially loss of driving licence.
During pregnancy, the plasma levels of AEDs may decline; this is attributed mainly to vomiting, changes in volume of distribution, and changes in protein binding. However, a major factor is likely to be the deliberate reduction or cessation of drug taking by women with epilepsy planning to, or having become, pregnant. Drug-taking behavior is traditionally assessed from patient self-reporting, pill counts, or plasma monitoring. None of these methods is particularly reliable, and they are thought to overestimate the true degree of therapeutic compliance. Blood results during clinic visits reflect the plasma concentration at the time of sampling; they give no insight into the patient's drug-taking behavior since the last clinic visit. Indeed, with most AEDs, a steady-state plasma concentration can be achieved by taking the prescribed dose for only 3 to 4 days before the clinic visit.
Hair analysis offers a different approach, indicating drug-taking behavior retrospectively over the preceding weeks or months. Drugs taken either therapeutically or recreationally become incorporated into hair (2). The most likely mechanism is the passive diffusion of drugs from the systemic circulation supplying the hair follicle (3). Once transferred into the rapidly dividing cells of the dermal papilla, drugs become sequestered and after keratinization become encapsulated into the hair shaft. In time, this shaft emerges above the scalp and becomes available for analysis. Hair collected from the posterior vertex of the scalp grows at an average rate of ∼1 cm/month (4). If the date of sample collection is known, it is therefore possible to relate a particular segment concentration to a corresponding period of growth and thus reflect drug-taking behavior over that period. In a study on a highly supervised inpatient population of patients with epilepsy, prescribed an unchanged dose of CBZ over a 6-month period (5), the mean intrapatient percentage coefficient of variation (CV) was 15.4 ± 5.8, only marginally greater than the analytic CV (6.1 ± 1.2%) over the concentration range measured. The good relation between the prescribed dose and the hair concentration and also the consistency of hair concentration over a period of several months in a compliant population prompted the present study of pregnant patients with epilepsy, among whom there is a perception of erratic compliance.
- Top of page
Ethical approval for the study was granted by respective local research ethics committees. All pregnant women currently prescribed CBZ or LTG for epilepsy, either as monotherapy or in combination with other AEDs, consecutively presenting to local medical antenatal clinics between 1998 and 2000, were invited to participate. Patients receiving VPA monotherapy were excluded, as hair analysis for VPA was, at the time, unavailable. The need for continued AEDs in pregnancy was evaluated and discussed with the patients, and after delivery, patients were asked to complete a written questionnaire indicating personal perceptions of their drug compliance and seizure control during pregnancy.
Twenty-eight pregnant women with epilepsy were invited to participate; two declined to provide a hair sample. Thus 26 pregnant women (mean age, 27.5 ± 6.7 years) taking CBZ or LTG either alone or in combination with other AEDs were recruited. Their clinical details and prescribed medication are given in Table 1. The controls comprised 13 female epilepsy outpatients of childbearing potential (mean age, 31.9 ± 8.3 years).
Table 1. Clinical data and the prescribed antiepileptic dose for the 26 pregnant patients
|Patient||Age (y)||Gravida||Gestation (wk) at sampling||Epilepsy type||Seizure frequency||Drug 1 (mg/day)||Drug 2|
|1||28||4||22||Juvenile myoclonic||C||Lamotrigine 400|| |
|2||16||1||19||Idiopathic generalised||A||Lamotrigine 50|| |
|3||34||1||38||Idiopathic generalised||B||Lamotrigine 100|| |
|4||34||1||22||Localisation related||B||Carbamazepine 1,000|| |
|5||24||1||14||Idiopathic generalised||A||Lamotrigine 150|| |
|6||25||1||20||Idiopathic generalised||B||Lamotrigine 150|| |
|7||27||2||14||Idiopathic generalised||A||Lamotrigine 300||Valproate|
|8||36||3||31||Localisation related||A||Lamotrigine 200||Vigabatrin|
|9||25||2||38||Juvenile myoclonic||A||Lamotrigine 250||Valproate|
|10||40||1||18||Localisation related||A||Carbamazepine 1,200|| |
|11||22||2||5||Idiopathic generalised + NEAD||B||Lamotrigine 300|| |
|12||36||2||17||Localisation related||A||Lamotrigine 350|| |
|13||19||2||18||Localisation related||B||Lamotrigine 350|| |
|14||37||3||20||Localisation related||A||Lamotrigine 200|| |
|15||37||1||20||Localisation related||B||Lamotrigine 175|| |
|16||17||1||14||Idiopathic generalised||A||Lamotrigine 200|| |
|17||21||1||36||Idiopathic generalised||A||Carbamazepine 400|| |
|18||26||2||19||Localisation related||A||Carbamazepine 400|| |
|19||25||1||9||Idiopathic generalised||A||Carbamazepine 400|| |
|20||22||2||18||Localisation related||B||Carbamazepine 600|| |
|21||29||3||22||Unclassified||B||Carbamazepine 200|| |
|22||25||2||30||Juvenile myoclonic||B||Lamotrigine 250|| |
|23||24||2||38||Localisation related||B||Carbamazepine 600||Valproate|
|24||27||3||31||Localisation related||C||Carbamazepine 800|| |
|25||35||1||40||Localisation related||A||Carbamazepine 800|| |
|25||22||1||12||Localisation related||B||Carbamazepine 900|| |
About 15–20 hairs (∼100 mg) were cut as closely as possible to the scalp at the posterior vertex and placed in aluminium foil marked to enable the identification of the root end. For analysis, the sample was removed from the foil and cut into 1-cm segments by using a specially designed template. Each segment was weighed and then digested in 1 ml of 1.5 M sodium hydroxide for 2 h at 40°C. The AED concentration in each segment was measured by using high-performance liquid chromatography (6).
Correction of individual segment concentration due to “washout effect”
It is well established that a decline in drug concentration occurs along the proximal–distal axis of the hair sample (i.e., the concentration declines the farther away from the scalp that the segment is taken) (7). Although washout is a function of individual personal hygiene behavior, the decline is remarkably linear (Fig. 1a). This decline can therefore be successfully corrected by using linear regression analysis (Fig. 1b). Corrections in Fig. 1b are based on a calculated linear regression slope of 2.593 (i.e., the mean segment concentration declines at a rate of 2.6 ng/mg per segment). Linearization is achieved by adding the multiple (slope × segment number) to the uncorrected segment concentration. For example, uncorrected concentration at segment 4 is 21.7 ng/mg; corrected concentration is 21.7 + (4 × slope) = 32.1.
Figure 1. The decline in hair-segment drug concentrations (mean + SD) of nonpregnant female controls (n = 13). B: Data from (A) corrected for “washout” effect (see text).
Calculation of individual patient variation in corrected segmental drug concentration
After correcting the individual segment concentrations for the washout effect, we calculated the mean of all segments analyzed for an individual, and each corrected segment concentration was expressed as a percentage of this mean value. This allowed comparison of data from patients with differing concentrations of AEDs. It also avoided the difficulties that would have arisen from percentage transform relative to root segments, some of which were drug free.
Within-subject variance was calculated for each group and compared by using a variance ratio test.
The time–concentration profile, based upon an assumed growth rate of 1 cm per month, was plotted for each patient. The duration of the pregnancy, extrapolated from the expected date of delivery, was superimposed above the respective time–concentration profile.
- Top of page
The individually corrected segment concentrations (ng/mg) for each pregnant patient are illustrated in Table 2. Fifteen per cent of patients (four of 26; i.e., patients 1, 7, 10, and 21), show marked declines in drug concentration in the more proximal segments, suggesting a change in drug-taking behavior over this interval. No such obvious changes are seen among the nonpregnant group except in patient 10, where there is evidence of erratic drug-taking behavior, rather than self-discontinuation. Figure 1a illustrates the mean (±SD) decline in drug concentration along the proximal–distal axis of hair segments 1–10 of the nonpregnant group (n = 13). Figure 1b shows these data corrected for the washout effect by linear regression.
Table 2. Individually corrected segment drug concentrations (ng/mg) of patients (1–26) and controls (C1–13)
|3||143.9||162.8||109.7||118.6||162.5||142.4|| || || || ||140.0||22.0||15.7|
|4||75.7||62.5||60.1||71.5||68.3||70.0|| || || || ||68.0||5.8||8.5|
|7||24.8||40.1||80.7||108.7||122.7||133.6|| || || || ||85.1||44.7||52.6|
|8||32.0||36.0||32.3||32.9||33.6|| || || || || ||33.4||1.6||4.8|
|9||45.8||30.5||45.7||33.4||31.6||37.6||42.0||41.2|| || ||38.5||6.1||15.9|
|10||0||0||26.2||139.6||189.4||216.8|| || || || ||55.3||58.5||103.8|
|14||20.9||25.8||27.7||24.6||20.7||26.7||22.6|| || || ||24.1||2.8||11.6|
|15||125.5||113.0||126.5||125.0||112.6||126.1|| || || || ||121.5||6.7||5.5|
|16||18.8||19.7||19.5||18.3||17.0||16.9||21.1|| || || ||18.7||1.5||8.1|
|19||12.7||11.5||12.8||12.7||11.5||12.7|| || || || ||12.3||0.6||5.2|
|20||18.6||19.0||19.5||18.3||17.0||17.0||20.8|| || || ||18.6||1.4||7.3|
|22||33.2||40.4||45.0||39.10||33.0||42.0||35.9|| || || ||38.4||4.5||11.8|
|23||59.3||41.2||60.5||63.9||67.4||52.3||55.7||52.0|| || ||56.5||8.2||14.5|
|24||12.1||14.5||13.5||12.4|| || || || || || ||13.1||1.1||8.4|
|26||29.0||33.7||33.3||30.1||34.2||33.1||29.1|| || || ||31.8||2.3||7.2|
|C3||48.4||54.4|| ||57.3||55.4||53.0||48.6|| || || ||52.9||3.7||6.9|
|C5||35.0||32.2||33.8||35.9||33.2|| || || || || ||34.0||1.5||4.3|
|C9||54.6||40.3||36.3||40.3||46.2||50.5|| || || || ||44.7||7.0||15.6|
|C12||27.8||25.3||23.8||24.0||25.0||26.1||26.9|| || || ||25.6||1.5||5.8|
|C13||16.0||18.7||21.2||19.6||18.3|| ||16.6||16.4||20.4|| ||18.4||1.9||10.6|
The within-subject variance of the corrected segmental hair concentrations for the pregnant group (n = 26) was 550.8 [181 degrees of freedom (DF)], with a coefficient of variation (CV) of 23.47. For the nonpregnant group (n = 13), the variance was 346.2 (96 DF) with a CV of 18.61. Thus the variance ratio (pregnant to nonpregnant group) was 1.59, significant at the p < 0.01 level. If the four patients suspected of discontinuing their medication were excluded from the pregnant group, the variance among the remaining pregnant patients was 146.9 (153 DF), with a CV of 12.12. The variance ratio here is 2.35 (nonpregnant to the remaining pregnant patients), significant at the p < 0.001 level. Thus the pregnant group overall demonstrates significantly more variability in hair concentrations than the nonpregnant group, indicating significantly less consistent AED-taking behaviour among the pregnant women. With the four pregnant patients suspected of self-discontinuation removed from the analysis, however, the variability among the pregnant group is significantly less than that in the nonpregnant group.
Figure 2 illustrates the hair concentration/time profile of patient 4 prescribed CBZ, 1,000 mg daily, for the 6 months before hair sampling. The profile clearly illustrates that at the time of sample collection, the patient was 5 months pregnant. The profile is indicative of consistent drug-taking behavior, with a mean segment concentration of 68.0 ± 5.8 ng/mg and a %CV of 8.5 over a 6-month period. This CV% is well within the range experienced from studying compliant inpatients (5). The patient also provided a sample of baby hair at 6 weeks after delivery. Although the sample was insufficient for segmental analysis, the CBZ concentration in the baby hair was 65.6 ng/mg, remarkably similar to that of the mother. In prospective studies now under way, the collection of baby hair at 6 weeks after delivery has been drafted into the protocol.
Figure 3 shows the hair concentration–time profile of patient 10, prescribed CBZ, 1,200 mg daily. The sample collected was sufficient for the analysis of six segments. The profile clearly illustrates a decline in segment concentration with no detectable levels of CBZ in the proximal two segments, suggesting discontinuation of therapy. The gradual decline in hair AED concentration to zero over three to four segments suggests the medication was discontinued slowly rather than abruptly. The patient later acknowledged having discontinued her medication at 6 weeks' gestation (without seizure recurrence), restarting medication after pregnancy. A plasma sample collected during her pregnancy at a clinic visit confirmed this, showing no trace of CBZ.
Figure 4 illustrates the profile of patient 1, prescribed LTG, 400 mg daily. At the time of sample collection, she was 22 weeks pregnant. This profile suggests that her drug-taking behavior had been consistent until the first trimester, with a mean segment concentration of 26.3 ± 3.3 ng/mg and a %CV of 12.6, well within the range experienced with compliant inpatients. In months 4 and 5 of pregnancy, however, segment concentrations decreased significantly to 12.0 ± 9.0 ng/mg, CV% 74.8. She did not acknowledge any changes in her medication dose or seizure frequency during her pregnancy.
Patient 7, (Fig. 5), prescribed LTG, 300 mg daily, with VPA, showed a drug concentration–time profile very similar to that of the previous patient, demonstrating good compliance until the onset of pregnancy, but with clear evidence of drug discontinuation on becoming pregnant. The original sample was taken early in pregnancy; she provided an additional sample well into her pregnancy in which LTG was undetectable in all segments. After pregnancy, she stated that she had discontinued VPA while continuing LTG (although LTG was undetectable in her proximal hair segments) and had remained seizure free.
Patient 21 (Fig. 6) showed a profile indicating erratic drug-taking behavior, with a mean concentration 22.4 ± 4.2 ng/mg, and %CV, 18.9, during the 7 months before becoming pregnant. On becoming pregnant, however, there is clear evidence of self-discontinuation with no CBZ detected in the two most proximal segments. She reported increased seizure frequency during pregnancy but did not acknowledge any change in her CBZ dosage.
Patient 22, (Fig. 7), with juvenile myoclonic epilepsy; was found dead at 30 weeks of gestation and diagnosed as sudden unexplained death in epilepsy (SUDEP) after a coroner's postmortem examination. She had been followed up regularly in the epilepsy clinic, experiencing approximately one generalized tonic–clonic seizure per month without significant changes in seizure control during pregnancy. She had been prescribed LTG, 250 mg daily, and her hair concentration–time profile derived from hair taken both at antenatal booking and after death was consistent with that of regular drug-taking behavior with a mean concentration of 38.4 ± 4.5 ng/mg, %CV 11.8, over the 7-month period analysed. The data presented were derived from the postmortem sample, and suggest that her SUDEP was unlikely to have resulted from altered drug-taking behavior.
- Top of page
We report the first application of hair analysis to assess AED-taking behavior in pregnancy. Our data suggest that a significant minority of patients, four (15%) of 26, had stopped their medication during pregnancy, yet only one of these subsequently acknowledged having done so.
Hair analysis offers a novel and noninvasive method of assessing drug-taking behavior, and can be applied in principle to any area of therapeutic medicine. It has the advantage over blood testing of providing a perspective of drug-taking behaviour retrospectively over a period of several months, depending on hair length. The technique readily identifies patients who have previously stopped their medication but, like other techniques used for the assessment of compliance, is not sufficiently sensitive to detect minor variations in drug-taking behavior. The technique would probably fail to detect minor changes attributable to altered volume of distribution, vomiting of the occasional missed tablet, etc.
Supervised inpatients taking an unchanged dose of medication show remarkably consistent drug concentrations between hair segments collected over a 6-month period (5). Furthermore, the correlation of these concentrations with their respective trough plasma concentrations is excellent. Similar studies (8) demonstrated significantly greater variability in segment concentrations among outpatients when compared with the highly supervised inpatient group. For meaningful comparison with our pregnant group, therefore, we used outpatient controls rather than supervised inpatients.
The extent of drug incorporation into hair varies between individuals, and there is a substantial variation among patients prescribed the same dose. This does not discredit the technique because each patient acts as her own control, allowing intrapatient changes to be evaluated. As a result of this interpatient variation, hair analysis (like plasma monitoring) cannot be used to predict the concentration given the prescribed dose.
A particular area of concern is the assumption in this study that hair grows at a rate of 1 cm per month, and all segment concentrations were related to this growth rate. Hair growth from the posterior vertex has been recorded at 0.8 to 1.5 cm/month (4), although growth rates are relatively constant within individuals. Despite these limitations, our data clearly show that hair analysis can provide a useful retrospective overview of drug-taking behavior.
Our results emphasise that pregnancy is a time when women are particularly likely to discontinue their AEDs. Furthermore, most of those who stopped their medication did not acknowledge that they had done so, even when asked directly on a postpregnancy questionnaire. The self-discontinuation rate in this study (15%), however, almost certainly underestimates the extent of noncompliance among pregnant women with epilepsy, for several reasons. First, our results apply only to women taking CBZ or LTG, because VPA hair levels were not measured. The teratogenic potential of VPA is well known, and the risks, especially of neural tube defects (9), are routinely discussed with women in our clinic. Patients planning to become or becoming pregnant might therefore be more inclined to discontinue VPA than to discontinue medications with less perceived teratogenic potential. Second, other patients may have attempted to reduce or stop their treatment, only to restart after deterioration in seizure control. Brief cessation of medication (e.g., a few days) would not necessarily show on hair analysis. The patients showing evidence of significantly impaired AED compliance reported no change in seizure control during pregnancy. This suggests either that continued good control provided the incentive to continue their medication discontinuation or that they concealed seizures to avoid being advised to take more medication. Third, the concept of satisfactory drug-taking behavior is based on the variability in hair concentration rather than on absolute values, and thus patients reducing their medication to a constant level, rather than stopping it, would appear as compliant on our analysis. Fourth, case selection from hospital clinics may overestimate compliance, because those most at risk of discontinuing medication might also be most likely to default from hospital appointments. Finally, it is quite possible that compliance was poorer in the two patients who declined participation in the study.
The sudden unexplained death of one patient would seem to illustrate in the boldest terms the importance of maintaining seizure control in pregnancy. Epilepsy remains the second commonest cause of maternal death in the U.K., with 19 of 134 maternal deaths documented during 1997 through 1999 being attributed to epilepsy (10). There is concern that covert discontinuation of medication by women in pregnancy might increase their risk of maternal death. In the case described here, however, retrospective analysis of the hair sample showed no evidence of altered drug-taking behavior over the months preceding her death. Similarly, postmortem hair from several other nonpregnant SUDEP cases has so far failed to identify significant evidence of AED noncompliance as a contributory factor (11). This evidence, therefore, fails to provide reassurance that SUDEP is a preventable cause of maternal death.
Our results suggest that most patients who stop their medication do so after rather than before conception. In practice, this is too late to prevent the major known teratogenic effects of AEDs (e.g., neural tube defects). Furthermore, <50% of pregnant women with epilepsy plan their pregnancy (12), one reason being oral contraceptive failure from enzyme induction by certain AEDs. This serves to emphasise the importance of an open discussion and checking of patient knowledge of medication-related teratogenic risks before the decision to become pregnant.
We have shown that not only is compliance in pregnancy worse than that in age-matched controls but also that patients' reporting of their medication taking in pregnancy is unreliable. Previous studies of drug-taking behavior in pregnancy have depended on patient reporting and occasionally on blood monitoring. On direct questioning, ∼60% of pregnant women with epilepsy admit lapses of compliance (12). However, it is not known how many of these simply reduce rather than completely discontinue their medication, or how this figure compares with that in nonpregnant patients. AED plasma levels give an idea of compliance at a single point in time and not the overall view offered by hair analysis. A patient who stops and restarts her medication, or takes it more regularly in the days before a hospital clinic appointment, might appear to have satisfactory drug-taking behavior based on blood testing, but would be identified as “noncompliant” on hair testing.
In conclusion, pregnancy remains a pivotal point in the lives of women with epilepsy. Further work is needed to understand and allay the fears of these women, and we hope to avoid potentially dangerous patient-initiated medication changes.