To evaluate adherence to prescribed antiepileptic drugs (AEDs) in children with epilepsy using a combination of adherence-assessment methods.
To evaluate adherence to prescribed antiepileptic drugs (AEDs) in children with epilepsy using a combination of adherence-assessment methods.
A total of 100 children with epilepsy (≤17 years old) were recruited. Medication adherence was determined via parental and child self-reporting (≥9 years old), medication refill data from general practitioner (GP) prescribing records, and via AED concentrations in dried blood spot (DBS) samples obtained from children at the clinic and via self- or parental-led sampling in children's own homes. The latter were assessed using population pharmacokinetic modeling. Patients were deemed nonadherent if any of these measures were indicative of nonadherence with the prescribed treatment. In addition, beliefs about medicines, parental confidence in seizure management, and the presence of depressed mood in parents were evaluated to examine their association with nonadherence in the participating children.
The overall rate of nonadherence in children with epilepsy was 33%. Logistic regression analysis indicated that children with generalized epilepsy (vs. focal epilepsy) were more likely (odds ratio [OR] 4.7, 95% confidence interval [CI] 1.37–15.81) to be classified as nonadherent as were children whose parents have depressed mood (OR 3.6, 95% CI 1.16–11.41).
This is the first study to apply the novel methodology of determining adherence via AED concentrations in clinic and home DBS samples. The present findings show that the latter, with further development, could be a useful approach to adherence assessment when combined with other measures including parent and child self-reporting. Seizure type and parental depressed mood were strongly predictive of nonadherence.
Adherence is defined as “the extent to which a person's behavior—taking medication, following a diet, and/or executing lifestyle changes—corresponds with agreed recommendations from a healthcare provider” (World Health Organization, 2003). Poor adherence to prescribed medications has been related to poor outcome in adults with epilepsy, for example, reduced seizure control (Jones et al., 2006; Hovinga et al., 2008), increased risk of mortality (Faught et al., 2008), higher incidence of emergency department visits and hospital admissions (Faught et al., 2008; Hovinga et al., 2008), increased health care costs (Davis et al., 2008), and lower quality of life (Hovinga et al., 2008). Despite these major issues, estimated nonadherence in adults with epilepsy ranges from 20–80% (World Health Organization, 2003). This variation in adherence rates is not surprising, considering variations across studies in defining adherence, adherence measurement methodology (Horne, 2005), and approaches used to summarize the data (Dunbar-Jacob & Mortimer-Stephens, 2001).
Little is known about the extent of adherence and factors that affect adherence in children with epilepsy. A study in the United States on adherence of 35 children with new-onset epilepsy, over a 1 month period, indicated an adherence level of 79.4%, as assessed using the Medication Event Monitoring System, TrackCap (Modi et al., 2008). Children of married parents and higher socioeconomic status showed higher 1-month adherence levels. Another study that investigated adherence to antiepileptic drugs (AEDs) in 181 children using self or parental reports (Asadi-Pooya, 2005), showed that factors significantly associated with poor adherence included higher maternal age, positive family history of epilepsy, larger family size, and increased complexity of the AED regimen.
Published research in children to date has used different methods and tested different variables as possible determinants of nonadherent behavior. The aim of the present study was to use a multiple methods approach to better evaluate the extent of adherence to prescribed AEDs in children with epilepsy, including exploration of the use of a novel dried blood spot (DBS) sampling linked with population pharmacokinetics (PopPK) modeling. An additional aim was to identify factors (both child and parental) that influence adherence.
The study was approved by the Office of Research Ethics Committees in Northern Ireland (ORECNI; 09/NIR03/74). A convenience sample of 100 children with epilepsy were recruited into this study from six pediatric epilepsy outpatient clinics in six different centers in Northern Ireland, that is, the Royal Belfast Hospital for Sick Children (RBHSC), Carlisle Health and Wellbeing Centre (CC), Belfast City Hospital (BCH), Craigavon Area Hospital (CAH), South Tyrone Hospital (STH), and Antrim Area Hospital (AAH). The parents/guardians of all children (≤17 years) who attended the participating clinics and who had been prescribed an AED of interest (carbamazepine [CBZ], lamotrigine [LTG], levetiracetam [LVT], sodium valproate [VPA], and topiramate [TPM]) for at least 2 weeks, were invited to take part in the study. Children were included in the study only after their parents or guardians were fully informed and signed a study consent form. Assent was also obtained from children deemed capable of providing this by their doctor.
Patients’ medical history, current medications, and biochemical data were recorded from their medical notes. Their seizure severity was assessed by their physician, using the Global Assessment of Severity of Epilepsy (GASE) Scale (Speechley et al., 2008).
DBS sampling has been in use for many years for the diagnosis of inborn errors of metabolism. The methodology involves spotting drops of blood on to absorbent paper. Our group has pioneered the use of DBS sampling in PopPK modeling in neonates and children, for example, in determining PK parameters and dosage recommendations for metronidazole use in neonates (Suyagh et al., 2011). The present study extends the use of DBS sampling to assess adherence. This sampling “matrix” facilitates sample acquisition (e.g., from finger/heel-pricks) and sample storage/transfer (since drugs are generally stable in dried form).
Heel/finger prick blood samples were collected from each child at an outpatient clinic visit, by spotting drops of blood (minimum 1, maximum 3) directly onto a Guthrie card. The date and time of collection were recorded for each sampling time. For children ≥5 years old, parents were asked to take two additional blood samples at home, 4 weeks apart, dry them overnight at room temperature, and mail them to our laboratory in prepaid “mailer-kits.”
Sensitive, selective, microanalytical methods for determination of the AEDs were developed and validated. The method utilized 6-mm disks punched from the DBS samples (equivalent to approximately 12 μl of whole blood). The sample preparation step involved extraction using an aliquot (980 μl) of methanol-to-acetonitrile solvent (3:1, v/v). The sample mixture was then dried under a stream of nitrogen at 40°C for 30 min using a Zymark TurboVap LV Evaporator Workstation (Focus Scientific Solutions Ltd., Stamullen, Ireland) and the residue dissolved in 1 ml water before undergoing solid phase extraction (SPE) using Oasis HLB cartridges (1 ml/30 g; Waters, Dublin, Ireland). For samples analyzed using gas chromatography, the dried residue was derivatized directly using N-methyl-N-trimethylsilyltrifluoroacetamide derivatizing agent without the need for SPE. Samples containing TPM and VPA were analyzed using gas chromatography with mass spectrometry detection (GC-MS), whereas all other AEDs were analyzed using high-performance liquid chromatography (HPLC) with UV detection. The assay limits of quantification for TPM, VPA, LVT, LTG, and CBZ were 0.5, 0.5, 2, 1, 2, and 1 μg/ml, respectively. The intraday and interday accuracy (percentage relative error; %RE) and precision (percentage coefficient of variation; %CV) were within the limits recommended by U.S. Food and Drug Administration (FDA) guidelines (±15%) (FDA, 2001) ranged from −5.54% to 14.33%. During assay development it was shown that variation in volumes of blood used to produce DBS samples did not significantly influence measured AED concentrations. Concentration ranges covered by the assay validation were 0.5–125 μg/ml for VPA and TPM, 1–20 μg/ml for LTG and CBZ, and 2–50 μg/ml for LVT.
The following questionnaires were self-completed during the clinic visit by parents and/or their children (if ≥9 years old, a child's ability to respond to the questionnaires was determined by his/her physician).
Two versions (parent and child) of the questionnaire, to include general (i.e., overview) or specific (i.e., over the last month) adherence, were used (Horne & Hankins, 2008). Parent and child versions have the same items, except for “I don't give it because my child refuses it,” which is only relevant to parents. Scores for each of the six (parent) or five (child) items were summed to give a score ranging from 6 to 30 (parent) or 5 to 25 (child). Higher scores indicate higher levels of self-reported adherence. Children with total scores <23 or whose parents had scores <27 were considered nonadherent according to the MARS (general) or MARS (past 1 month) questionnaires.
The BMQ-specific involves assessment of patients’ beliefs about medication prescribed for a particular illness. Two versions of the questionnaire (parent and child) were used, with similar items and response terms. Higher scores indicate stronger beliefs in the concepts represented by the scale. The questionnaire has two major subscales, that is, “Necessity Beliefs” and “Concerns.” The former assesses the patients’/parents’ views about the necessity of the medication for maintaining or improving health, whereas the Concerns subscale focuses on beliefs about adverse effects of taking medication (Horne et al., 1999).
This scale was used to measure parents’ perceptions of their ability to manage their child's seizures. Higher scores indicate greater parental confidence (Austin et al., 2004).
The CES-D scale, which enables the assessment of depressed mood in the general population (Radloff, 1977), was administered to parents. Total scores range from 0 to 60 with a score ≥16 representing depressed mood (Austin et al., 2004).
The degree of self-efficacy in the management of seizure disorders in participating children (aged 9–17 years) was assessed using the SSES-C questionnaire. A higher score reflected greater self-efficacy (Austin et al., 2004).
The QOLIE-AD was used to assess health-related quality of life in participating adolescents (11–17 years). It consists of two parts: (i) general health and (ii) effects of epilepsy and AEDs (Cramer et al., 1999). Only the latter part (which addresses epilepsy impact, epilepsy-associated stigma, and attitudes toward epilepsy) was used. Higher scores indicate better functioning and well-being (scores were converted to a scale of 0–100).
To determine whether a child was adherent, a computer simulation method was utilized to estimate the 95% interval of predicted plasma concentrations for each drug at the time of sampling (n = 1,000 sets of simulations using the non-linear mixed effect modeling software package, NONMEM, Icon Development Solutions, Ellicott City, MD, U.S.A.). In this approach, literature values of PopPK parameters for each drug were employed (Hussein & Posner, 1997; Yukawa et al., 1997; Serrano et al., 1999; Reith et al., 2001; Edelbroek, 2002; Shank et al., 2005; Toublanc et al., 2008; Chhun et al., 2009; Milovanovic & Jankovic, 2009; Edelbroek P, personal communication, 2010; Vovk et al., 2010). In addition, significant covariates reported to influence PK parameters for each drug (e.g., age, body weight, dose) were incorporated into the simulation models. For each drug, the simulated plasma concentrations were transformed to corresponding DBS concentrations using published blood-to-plasma correction ratios (Edelbroek, 2002, 2010). For TPM, the correction ratio was calculated using the equation established by Shank et al. (2005) as it is nonlinear across the therapeutic range.
Patients were considered adherent if all of their measured AED concentrations were within the calculated 95% prediction intervals. Patients taking more than one AED of interest were categorized as nonadherent if the measured concentration was below or above the simulated concentration for at least one of the AEDs prescribed.
Adherence was assessed using general practitioner (GP)–held prescribing records for each patient over the last 12 months. MRA was calculated as follows (Hess et al., 2006):
A patient was classified as nonadherent if at least one prescribed AED had an MRA value of <80% (Davis et al., 2008; Zeber et al., 2010). MRA values >120% were not regarded as overadherence but as indicative of oversupply (Andersson et al., 2005).
All analyses were carried out using SPSS version 18 (SPSS Inc., Chicago, IL, U.S.A.). The magnitude of agreement between different adherence assessment methods was determined using the Kappa (κ) coefficient (Landis & Koch, 1977).
Group differences (adherent vs. nonadherent) were explored using Mann-Whitney U analysis for continuous variables. Categorical variables were analyzed using the chi-square or Fisher's exact test, as appropriate. Binary logistic regression was used to evaluate the relative contribution of the potential predictors of adherence categorization. All analyses were two-sided, with the significance level set at 0.05.
From the 173 children who met the inclusion criteria, 102 participated (response rate of 59%). Two patients were excluded as questionnaires were not fully completed. Of the remaining 100 patients, 56 were male and the median age was 7.5 years (range 0.9–16 years). Children who responded to the questionnaires were ≥9 years old. Severity of seizures ranged mainly from mild to moderately severe, Fig. 1. Medical and disease characteristics of participating children are shown in Table 1.
|Generalized epilepsy||36 (37.1%)|
|Focal/partial epilepsy||50 (51.5%)|
|Multiple seizure types||11 (11.3%) (3: missing data)|
|Duration of epilepsy|
|Median (range) in years||3 (0.17–12) (1: missing data)|
|Number of AED prescribed|
|Median (range)||1 (1–4)|
|AED of interest prescribed (n [%])||122 (100%)|
|Measured AED concentrations (median [range], μg/ml)|
|Number of other concomitant medicines prescribed (n [%])a|
|Median (range)||0 (0–7)|
|More than three||8 (8.1%) (1: missing data)|
|Seizure frequency (n [%])|
|2–3 Times weekly||16 (16.2%)|
|Less than monthly||46 (46.5%) (1: missing data)|
The majority of parents (92.9%) and children (89.5%) had a strong belief about the necessity of AEDs (BMQ-specific scores above scale midpoint), whereas approximately 40% (parents) and 23% (children) had concerns about AED harmful effects. The median total scores for the necessity and concern subscales were 21 (range 9–25) and 15 (7–24) for parents and 20.5 (12–25) and 13 (5–24) for children, respectively. Reliability coefficients (Chronbach's alpha) for the necessity and concern subscales were 0.84 and 0.73 (parent) and 0.75 and 0.77 (child), respectively, indicating reliability of the methodology.
Although 67.7% parents had “Confidence in Seizure Management” scores ≥4, indicating confidence in managing their child's seizures (reliability coefficient of 0.87), 36.4% had a total CES-D score of ≥16, indicative of depressed mood (Chronbach's alpha 0.93). There was no significant correlation, however, between the level of confidence in seizure management and depressive symptoms in parents in the present study (p > 0.05). In terms of the SSES-C scale, 56.3% of the responding children (n = 18) had scores ≥4, indicating high self-efficacy in relation to seizure management (Chronbach's alpha 0.88). The median total score (range) for each of the QOLIE-AD subscales was 73.1 (26.9–98.1) for epilepsy impact, 72.2 (44.4–100.0) for stigma, and 31.3 (0–26.5) for attitudes toward epilepsy.
Score distribution for the MARS questionnaires is presented in Table 2. Using the parental total MARS questionnaire scores (general and past 1 month), a total of 94 children were deemed adherent. On the other hand, a lower proportion of children were adherent (chi-square analysis; p < 0.01) according to the MARS (child) questionnaire, that is, 32 children were classified as adherent (82.1% from a total of 39 children). The lower number of children who responded to the questionnaire reflected the numbers aged 9 years or above.
|Questionnaire||n||Mean total score (95% CI)||Total score range||Total score indicating non-adherence||No. of patients with total score indicating nonadherence (%)|
|MARS (general; parent)||100||29.23 (29.0–29.5)||23–30||<27||5 (5%)|
|MARS (past 1 month; parent)||99||29.57 (29.4–29.7)||26–30||<27||3 (3%)|
|MARS (general; child)||38||23.55 (23.0–24.1)||20–25||<23||6 (15.8%)|
|MARS (past 1 month; child)||39||23.92 (23.4–24.4)||21–25||<23||6 (15.5%)|
A total of 100 clinic DBS samples were obtained from the recruited children (90 samples being analyzable). For children ≥5 years old (n = 77), a total of 59 (76.6%) and 41 (53.2%) samples were received for the first and second home DBS samples, respectively (89 samples analyzable). Using the combined data sets (clinic, first home, and second home samples), a total of 75 children (80.6% from a total of 93 children with at least one good DBS sample available) were deemed adherent and 18 children nonadherent (either under- [n = 11] or overadherent [n = 7]), that is, an overall adherence rate of 80.6% using this measure. Measured AED concentrations are shown in Table 1.
GP prescribing records were received for 92 (92%) of the participating children. Six records, however, had inadequate information and had to be excluded from the analysis. Of the remaining 86 patients, 17 had an MRA <80% (i.e., classified as nonadherent), that is, an overall adherence rate of 80.2% using this measure.
DBS analysis and calculated MRA captured the highest percentage of nonadherence (~19% for both) followed by the MARS (child) questionnaire (17.9%). The lowest percentage of nonadherence (6%) was observed using the MARS (parent) questionnaire, Fig. 2.
There was a significant association between all adherence assessment methods and overall adherence (p < 0.05 for each comparison; chi-square or Fisher's exact tests). Adherence assessment using DBS analysis and the MRA showed the strongest agreement (κ = 0.306, p = 0.006), with 79.8% of the patients being classified the same using both methods Table 3. Upon combining all methods of adherence assessment, 33% of all patients recruited were classified as nonadherent.
|Pairs of adherence measure||κ coefficient|
|DBS vs. prescribing records (GPs)||0.306*|
|DBS vs. MARS (overall; parent)||0.228*|
|DBS vs. MARS (overall; child)||0.253|
|DBS vs. PMRs (community pharmacies)||0.011|
|MARS (overall; parent) vs. prescribing records (GPs)||0.401*|
|MARS (overall; parent) vs. PMRs (community pharmacies)||0.117|
|MARS (overall; child) vs. prescribing records (GPs)||0.242|
|MARS (overall; child) vs. PMRs (community pharmacies)||−0.013|
|MARS (overall; parent) vs. MARS (overall; child)||0.552*|
|Prescribing records (GPs) vs. PMRs (community pharmacies)||0.318*|
Univariate analysis indicated that overall adherence was associated with the “necessity’ subscale of BMQ-specific (parent version) and the “epilepsy impact” subscale of the QOLIE-AD questionnaire. Two separate logistic regressions (backward LR) were performed. The first model included patient demographics, disease characteristics, and parental variables, whereas the second model included child-completed questionnaire data.
These analyses indicated that four variables were significantly and independently associated with nonadherence, that is, (i) increased age, (ii) diagnosis of generalized epilepsy, (iii) higher scores for the BMQ-specific (parent) necessity subscale, and (iv) parental depressed mood (CES-D score ≥16), Table 4. None of the child-reported parameters (i.e., the BMQ-specific, the SSES-C and the QOLIE-AD questionnaires) approached statistical significance.
|Parameter||B||SE||Odds ratio (OR)||95% CI for OR|
|Increasing age (in years)||0.18||0.08||1.20*||1.03–1.40|
|Generalized epilepsy (reference: focal epilepsy)||1.54||0.62||4.66*||1.37–15.81|
|Increasing BMQ-specific (necessity subscale; parent) score||0.26||0.10||1.30*||1.08–1.57|
|Total score ≥16 (presence of depressed mood) (Reference: Total score <16; i.e., absence of depressed mood)||1.29||0.58||3.60*||1.16–11.41|
The highest percentage of adherence was observed with parental reporting (94% adherence) reflecting the known overestimation of adherence using this method (Daniels et al., 2011; Modi et al., 2011a). Modi et al. suggested a mean correction factor of 0.83 to adjust parent-reported adherence to their so-called “gold standard” of electronically monitored adherence. Application of this correction yielded a value of 78.0%, that is, close to the 80.2% and 80.6% adherence levels documented using the GP prescribing and DBS methodology, respectively.
Assessment of adherence to AEDs in children with epilepsy using plasma/serum AED concentrations has been reported (Lisk & Greene, 1985; Takaki et al., 1985; Shope, 1988; Hazzard et al., 1990; Gomes & Maia Filho, 1998; Mitchell et al., 2000). However, a drawback with such assessment is the intraindividual and interindividual variability associated with measured concentrations (Paschal et al., 2008). These uncertainties were minimized in the present study through the application of PopPK models, to predict AED concentrations according to the relevant covariates present at individual child level (i.e., age, body weight, and dose and drug interactions due to concomitant medicines). This in silico approach has not been applied to adherence studies to date, but has been utilized for the prediction of plasma concentrations (Maitre et al., 1988) and to recommend a dose to bring concentrations to within the therapeutic range after missed dose(s) (Reed & Dutta, 2004; Ahmad et al., 2005). Published PopPK models for AEDs in children are limited and have been based mainly on trough levels, which give good estimations of clearance (Cl) but poor estimations of volume of distribution (Vd) and absorption rate constant (Ka). Fixed values of Vd and Ka from the literature were used in the present study whenever there was a large uncertainty in the population estimates of these parameters.
Measurement of multiple AEDs concurrently from a single DBS sample was achieved in the present study. When coupled with the use of published PopPK models for predicting AED blood concentrations, this approach was shown to be useful for estimating levels of adherence in children participating in the study. The emerging, minimally invasive technology of DBS sampling, which can be performed by nonexperts, therefore, offers opportunities for improved patient care. The feasibility of home DBS sampling by parents was explored in the present study (two samples requested, 4 weeks apart). On both occasions more than half of the requested home DBS samples were received. However, a proportion of the samples was not analyzable due to poor quality of returned DBS samples. Overall, 57 parents of a total 77 parents who had children ≥5 years old were able to provide at least one analyzable home DBS sample. This exploratory work indicates that, with further training (e.g., utilizing video/DVD training sessions), home sampling effectiveness could be further enhanced.
The overall nonadherence rate to the AEDs, derived from the different measures, was 33%, which is within the reported range in international literature on the nonadherence to treatment in children with epilepsy of 3.5–58.0% (Lisk & Greene, 1985; Takaki et al., 1985; Shope, 1988; Hazzard et al., 1990; Gomes & Maia Filho, 1998; Kyngäs, 2000; Mitchell et al., 2000; Asadi-Pooya, 2005; Conn et al., 2005; Modi et al., 2008, 2011a,b).
Both increasing age and diagnosis of generalized epilepsy were associated with increased risk of nonadherence. Although increasing age was statistically significant, the OR value was close to 1, implying that the influence was marginal. Children with generalized epilepsy were, however, approximately five times more likely to be classified as nonadherent compared to children with focal epilepsy. This supports previous findings that young adults with generalized tonic–clonic seizures are less likely to be adherent with AEDs compared to other seizure types (Specht et al., 2003).
Unexpectedly, nonadherence was associated with the parents’ beliefs toward a greater need for AEDs in managing their child's epilepsy. However, again the OR was close to 1, indicating that the overall influence on adherence was marginal. Data linking parental beliefs about necessity of medicines in their children and adherence is limited. One such study by Conn et al. (2005) reported that parental beliefs about the need for controller asthma medication were not related to adherence. Furthermore, studies in adults with epilepsy in the UK found that beliefs about AEDs were not related to treatment adherence (Jones et al., 2006; Kemp et al., 2007).
In the present study, children of parents with depressed mood were approximately four times more likely to be classified as nonadherent to AEDs compared to children of parents without depressed mood. Maternal depression was also shown to correlate with poor medication adherence in children with epilepsy in a small prospective UK cohort study of 21 children with epilepsy (Otero & Hodes, 2000). A study by Mitchell et al. (Hazzard et al., 1990) in the United States, however, reported greater adherence to treatment in children (4–13 years old) with epilepsy in families experiencing stressful life events. A meta-analysis (DiMatteo et al., 2000) identified depressed adult patients with epilepsy as three times more likely to be nonadherent, compared to patients who were not depressed.
The sample size in the present study was relatively small, although larger than many cohorts of children published to date. This may have diminished the inability to identify some predictors initially hypothesized to potentially influence adherence in this population. A further potential limitation of the study is that electronic medication bottles were not used. Electronic medication monitors are capable of providing detailed insights into patients’ adherence behavior as they document the pattern of taking medication over multiple points in time, hence detecting possible occurrence of white-coat adherence prior to clinic visits (Modi et al., 2008). Although these devices are precise and their results are easily quantified, they are expensive and still considered an indirect method of measuring adherence because they do not document whether the patient had actually ingested the correct dose of prescribed medication (Osterberg & Blaschke, 2005). The novel DBS sampling, on the other hand, has the advantage of being both a direct and an objective method for measuring adherence. However, home DBS sampling was incomplete, indicating the need for a more robust training of parents on this technique.
The present study utilized a multiple-methods approach to better evaluate the adherence to AED therapy in children. For the first time, DBS sampling, coupled with PopPK modeling, was shown to be useful in estimating adherence. This novel approach, including home sampling, has the potential to significantly facilitate further research on adherence in children across a range of disease states (although the sampling approach needs further refinement). The present study also highlights the difficulties associated with adherence classification in children. Logistic regression analysis, however, showed that seizure type and parental depressed mood were strongly predictive of nonadherence in the present population.
The authors would like to acknowledge the funding received from the British Council under the Prime Minister's II Initiative award and from Atlantic Philanthropies. The authors wish also to thank Prof Rob Horne for his permission to use the MARS and BMQ questionnaires; and the following research nurses who participated in the research: Ms. Patricia McCreesh and Mrs. Sara Gilpin.
None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.