Epilepsy & Behavior 16 (2009) 634–639
Contents lists available at ScienceDirect
Epilepsy & Behavior journal homepage: www.elsevier.com/locate/yebeh
Enhancing antiepileptic drug adherence: A randomized controlled trial Ian Brown a, Paschal Sheeran a, Markus Reuber b,* a b
Department of Psychology, University of Shefﬁeld, Western Bank, Shefﬁeld, UK Department of Neurology, University of Shefﬁeld, Royal Hallamshire Hospital Shefﬁeld, Glossop Road, Shefﬁeld, UK
a r t i c l e
i n f o
Article history: Received 13 July 2009 Revised 5 September 2009 Accepted 13 September 2009 Available online 27 October 2009 Keywords: Antiepileptic drugs Adherence Randomized controlled trial Compliance Epilepsy Implementation intentions
a b s t r a c t Suboptimal adherence to antiepileptic drug (AED) treatment is commonplace, and increases the risk of status epilepticus and sudden unexplained death in epilepsy. This randomized controlled trial was designed to demonstrate whether an implementation intention intervention involving the completion of a simple self-administered questionnaire linking the intention of taking medication with a particular time, place, and other activity can improve AED treatment schedule adherence. Of the 81 patients with epilepsy who were randomized, 69 completed a 1-month monitoring period with an objective measure of tablet taking (electronic registration of pill bottle openings, Medication Event Monitoring System [MEMS]). Intervention participants showed improved adherence relative to controls on all three outcomes: doses taken in total (93.4% vs. 79.1%), days on which correct dose was taken (88.7% vs. 65.3%), and doses taken on schedule (78.8% vs. 55.3%) (P < 0.01). The implementation intention intervention may be an easy-to-administer and effective means of promoting AED adherence. Ó 2009 Elsevier Inc. All rights reserved.
1. Introduction Approximately 60% of patients with epilepsy achieve full control of their seizures with antiepileptic drugs (AEDs) . Modern medical treatment aims not only to prevent seizures but also to avoid negative effects on cognitive function and emotional, physical, and general well-being. People with epilepsy are most likely to achieve these aims by the regular ingestion of the lowest dose of medication and the smallest number of AEDs necessary . This, in turn, depends on their taking their medication as prescribed. However, research indicates that 30% to 50% of adults with epilepsy adhere poorly to their AED treatment schedules . Adherence problems may be more common in epilepsy than in other medical conditions . Indeed, nonadherence has been identiﬁed as one of the most important causes of treatment failure in patients with epilepsy , It is possible that neurologists underestimate the extent of adherence problems in their own clinical practice because patients do not admit failing to take their medication regularly to their doctor [6–8]. Seventy percent of patients with epilepsy state that they never miss a dose [9,10], and the majority of patients admit to missing only one or two doses per month . However, studies using objective measures have revealed much higher rates of irregular AED use. For instance, a study of 33,658 Medicaid recipients
* Corresponding author. Address: Department of Neurology, University of Shefﬁeld, Royal Hallamshire Hospital Shefﬁeld, Glossop Road, Shefﬁeld S10 2JF, UK. Fax: +44 (0) 114 2713158. E-mail address: [email protected]
(M. Reuber). 1525-5050/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.yebeh.2009.09.014
showed that less than 80% of the AEDs required for full prescription adherence were picked up by participants in 26% of quarters during the observation period . Two studies using the Medical Events Monitoring System (MEMS)—a pill bottle with an electronic cap that registers each occasion the bottle is opened—found that only 76% of doses were taken overall , and that 48% of patients took one-third or fewer of the prescribed AED doses . AED blood levels in the ‘‘therapeutic range” can offer false reassurance: although complete nonadherence can be detected using AED blood levels, there is no reliable correlation between the variability of AED blood levels and irregular medication intake . Poor adherence has been shown to affect important treatment outcomes: in the large Medicaid study mentioned above, the numbers of hospital admissions, inpatient treatment days, and emergency room visits were higher in ‘‘noncompliant” quarters, resulting in increased total health care spending . Another recent study based on more than 10,000 individuals with epilepsy found that 39% picked up less than 80% of the medication required to cover their AED prescription and that hospital admission rates and health care costs were higher in the nonadherent group . Other studies have shown that patients who miss doses may experience additional seizures [7,14], and may be slower to achieve full seizure control . Lack of adherence to AED treatment has been identiﬁed as a potential precipitating cause in 31% of epileptic seizures for which ambulances were called and 13% of seizures requiring emergency hospital admission [16,17]. Patients whose medication intake is irregular are also at increased risk of sudden unexplained death in epilepsy (SUDEP) . After demographic and clinical covariates in the Medicaid study were controlled for,
I. Brown et al. / Epilepsy & Behavior 16 (2009) 634–639
the mortality risk in nonadherent patients was more than three times higher than that of adherent patients . For these reasons, interventions that can improve AED treatment adherence are of great clinical interest. Previous studies have tested the effectiveness of a range of strategies including the simpliﬁcation of AED regimens to no more than two doses per day , discussion of serum level measurements with patients , and shorter intervals between clinic visits . The most intensive intervention (incorporating counseling, medication containers, self-monitoring, and mailed reminders for prescriptions and appointments) improved adherence in the treatment group and halved seizure frequency . However, this sort of intensive support may not be practical (especially for patients who also miss clinic appointments) or easy to integrate into routine clinical practice. This study tests whether a simple and self-administered worksheet consisting of an implementation intention intervention (III) can increase AED adherence [21,22]. In this III, patients are asked to write down exactly when and where they will take their medication, using the format of an if–then plan (‘‘If it is time X in place Y and I am doing Z, then I will take my pill dose”). IIIs target the problem that holding a strong goal intention (‘‘I intend to take my tablets regularly”) does not guarantee goal achievement, because people may fail to deal effectively with self-regulatory problems during goal striving. Evidence indicates that the act of writing down an if–then plan can help to ‘‘automate” triggering of the intended behavior by sensitizing people to the cues they have written down . This means that they become more likely to complete the intended activity when these cues are encountered . IIIs reduce the burden of having to think about and remember when to act by using environmental cues to trigger the desired behavior. IIIs are not merely a theoretical construct, but have already been proven effective in promoting a range of health behaviors in other areas of medicine including cancer screening, physical activity, and psychotherapy attendance .
randomizer.org). All patients completed a 14-page packet of selfreport measures after they had seen the neurologist (baseline). The intervention consisted of one additional intention implementation worksheet (up to six items on two pages depending on daily dose) which was included in the questionnaire packet that patients randomized to the intervention were asked to complete (i.e., patients in this group completed a total of 15 pages of questionnaires rather than 14). This additional worksheet is shown in Fig. 1. Neither the neurologist nor the clinic or pharmacy staff were aware of the patient’s group allocation. Following the procedure described by Gollwitzer and Sheeran , the III asked participants to specify environmental cues for tablet taking using the format of an ”if–then” plan (i.e., participants wrote down exactly when and where they were intending to take their antiepileptic medication every day, and what they would be doing at the moment they would take their AEDs). All patients picked up a 1-month supply of one of their antiepileptic drugs in an electronic pill-monitoring bottle, which recorded the number and timing of bottle openings (MEMS Aardex Ltd., Switzerland). The electronic monitoring caps can be connected to a personal computer that reads the data from the pill caps’ microprocessor and generates a printout of the patient’s bottle openings. One month after the initial clinic visit (follow-up), we approached patients by letter and asked them to complete an additional set of questionnaires and return the electronic pillmonitoring device. In line with previous studies using this method of monitoring, we measured adherence using three different outcome measures counting each opening as a presumptive dose : percentage of doses taken, percentage of days on which the correct number of doses was taken, and percentage of doses taken on schedule. We designated doses as having been taken on schedule if the MEMS bottle was opened within a ±3-hour target time window for each dose.
2.3. Self-report measures
We administered self-report measures to ensure the equivalence of control and experimental groups and to identify factors that could moderate the impact of the III. At baseline, participants completed the following measures: Theory of Planned Behaviour (TPB, 24 items, ﬁve scales), Brief Illness Perception Questionnaire (BIPQ, 9 items) , Multiple Ability Self-Report Questionnaire (MASQ, 38 items, ﬁve scales) , Hospital Anxiety and Depression Scale (HADS, 14 items, two scales) , Liverpool Seizure Severity Scale (LSSS, 12 items, one scale) [27,28], and a single-item self-estimate of the number of missed doses during the preceding month. At follow-up, participants completed the HADS, LSSS, and Prospective and Retrospective Memory Questionnaire (PRMQ, 7 items, one scale) .
Five consultant neurologists recruited patients consecutively from their outpatient clinic at the Royal Hallamshire Hospital in Shefﬁeld, United Kingdom, between January and June 2007. All patients had a clinical diagnosis of epilepsy, were taking antiepileptic drugs once or twice daily, and were attending the neurology clinic for a follow-up visit. The diagnoses had been made by a consultant neurologist on the basis of clinical history and neurological examination. Neuroimaging, EEG, or video/EEG telemetry had been carried out if clinically indicated. Patients were included only if they were: taking at least one of the AEDs that could be dispensed in the monitoring bottle once or twice daily (carbamazapine, clonazepam, gabapentin, lamotrigine, levetiracetam, oxcarbazepine, phenytoin, topiramate, or zonisamide); at least 16 years of age; able to read and write English; and responsible for taking their own medication. Patients were excluded if they indicated that they were already using a method of ensuring adherence that could be compromised if they took part in the study (e.g., weekly tablet dispensers), if they were receiving a diagnosis of epilepsy for the ﬁrst time, or if they had learning difﬁculties. The study was approved by the North Shefﬁeld Ethics Review Committee, and all patients gave written informed consent. 2.2. Study design We randomized patients to the intervention or control group using a computerized random number generator (http://www.
2.4. Statistical analysis We conducted an analysis of variance (ANOVA) on the continuous, cognitive, clinical, and demographic variables measured at baseline to ensure the equivalence of (1) participants who completed both baseline and follow-up measures and participants who completed the baseline measures only (representativeness check), and (2) participants in the intervention and control groups (randomization check). We used v2 tests to compare the respective groups on categorical variables. We used an ANOVA to determine the effect of intervention on the three measures of AED adherence. Because adherence is measured using continuous (0–100%) scales, moderated regression analysis is the appropriate test for identifying possible interactions between condition (intervention
I. Brown et al. / Epilepsy & Behavior 16 (2009) 634–639
Fig. 1. Implementation intention worksheet.
vs. control) and patient characteristics in predicting adherence , that is, for discovering what factors moderate the effects of the intervention on adherence outcomes. The three outcome measures (doses taken, days on which the correct number of doses was taken, and doses taken on schedule) were consolidated into a reliable overall index of adherence (Cronbach’s a = 0.95) to reduce the number of statistical comparisons and the risk of making type 2 errors. We accomplished this consolidation by standardizing and then averaging the three measures. For each potential moderator variable (i.e., all demographic, clinical, and cognitive variables), adherence was regressed on condition (coded 0 = control, 1 = intervention) and the focal moderator variable (standardized) at step 1, and the interaction term (condition multiplied by the standardized moderator variable) was added to the model at step 2. If the interaction term improved the ﬁt of the model according to the DR2 and DF statistics and the b coefﬁcient for the interaction term was signiﬁcant, moderation was demonstrated (i.e., the impact of the intervention on adherence varies according to scores on the moderator variable). The nature of the interaction was probed using simple slopes analysis. In particular, we computed the b coefﬁcients describing the effect of the intervention on adherence at high levels of the moderator (conventionally deﬁned as the mean + 1SD) and at low levels of the moderator (mean 1SD). Simple slopes analyses indicate precisely what types of patients obtain greater (vs. less) beneﬁt from the intervention.
3. Results 3.1. Participants We recruited 81 patients with epilepsy from the outpatient clinic at baseline. Of these, 79 (98%) collected their medication in MEMS pill monitors from the hospital pharmacy; 2 patients in the control group left the hospital without collecting their medication. At follow-up, 11 individuals (14%) did not return their bottles (3 (4%) could not be contacted, and 8 (10%) said they had returned their bottle but the bottles were not received at the pharmacy). Comparisons between participants who completed both the baseline and follow-up measures (n = 69) and participants who completed the baseline questionnaire only (n = 12) did not reveal any relevant clinical or demographic differences. This indicates that the ﬁnal sample (n = 69) satisfactorily represents the population from which it was drawn. Table 1 lists the demographic and clinical characteristics of the ﬁnal sample. Comparisons between participants in the intervention and control conditions on demographic, clinical, and cognitive variables revealed signiﬁcant differences on only two variables: intervention participants reported more frequent symptoms and expressed greater concern about epilepsy on the BIPQ compared with control participants (P < 0.02 for both variables). Type 2 error cannot plausibly account for these ﬁndings (d = 0.03).
I. Brown et al. / Epilepsy & Behavior 16 (2009) 634–639
Table 1 Demographic and clinical characteristics of the intervention and control groups (N = 69). n (%)
Demographic data Married Employed Male
17 (46) 10 (31) 15 (42)
15 (47) 10 (35) 12 (38)
v2 = 0.01 v2 = 0.15, ns v2 = 0.07, ns
Education No secondary school Secondary school to 16 years Secondary school to 18 years Degree Mean age (SD)
12 (40) 9 (30) 3 (10) 6 (20) 41.9 (15.4)
5 (23) 9 (41) 4 (18) 4 (18) 44.10 (16.4)
v2 = 0.52, ns F = 0.33, ns
Percentage of Participants 5
F = 0.19, ns F = 1.98, ns F = 1.23, ns
Note. There were missing data for 8, 17, and 3 cases for employment, education, and gender, respectively, because of incomplete records.
3.2. Intervention effects on AED adherence Table 2 indicates that the intervention had a signiﬁcant positive impact on all three measures of AED adherence. Relative to controls, participants in the intervention group took 18.1% more prescribed doses, took the correct number of doses on 35.9% more days, and took 42.5% more of their doses in the correct 6-hour time window. Fig. 2 illustrates the impact of the intervention on taking AED medication on schedule. Whereas adherence exhibited a Vshaped curve among control participants (18.8% took <30% of doses on schedule, and 24.9% took >90% of doses on schedule), the distribution of responses for intervention participants is highly skewed in the direction of greater adherence (2.7% took <30% of doses on schedule, and 48.6% took >90% of doses on schedule) (Fig. 2).
0 -7 61
0 -6 51
v2 = 0.09, ns v2 = 1.70, ns v2 = 3.46, ns
v = 1.67, ns
26 (81) 14 (44) 11 (34) 18.6 (15.4) 12.6 (3.3) 16.8 (3.0)
29 (78) 22 (60) 21 (57) 20.3 (18.1) 11.2 (4.6) 15.8 (3.9)
Last seizure <1 year Polytreatment Comedication Duration of seizures, years HADS depression score HADS anxiety score
25 (78) 2 (6) 5 (16)
26 (70) 6 (16) 5 (14)
Clinical data Epilepsy diagnosis Focal Idiopathic generalized Unclassiﬁed
Percentage of Doses Taken on Schedule Fig. 2. Distribution of doses taken on schedule in the control and implementation intention intervention conditions.
adherence. Simple slopes analyses showed that the implementation intention intervention did not beneﬁt adherence when participants had high scores on prospective memory (B = 0.31, SE = 0.27, P = 0.26), estimated having missed few doses at baseline (B = 0.27, SE = 0.36, P = 0.46), or were concerned about their epilepsy (B = 0.37, SE = 0.30, P = 0.22. Conversely, the III had a signiﬁcant positive impact on adherence when participants had poor memory for intended actions (B = 1.17, SE = 0.27, P < 0.001), when participants reported having missed many doses at baseline (B = 1.49, SE = 0.43, P < 0.001), and among participants who were less concerned about their epilepsy (B = 1.14, SE = 0.30, P < 0.001). Thus, the III had particular beneﬁt for AED adherence among participants who were at greatest risk of poor adherence (see Table 3).
4. Discussion 4.1. Main ﬁndings Although 85% of patients with epilepsy in our study stated that they considered it ‘‘very important” to take their AEDs regularly
3.3. Moderators of the intervention effects on AED adherence Moderated regression analyses that tested interactions between condition (intervention vs. control) and demographic, clinical, and cognitive variables revealed three signiﬁcant effects: prospective memory scores, participant-estimated number of missed doses at baseline, and one item from the BIPQ (illness concern) each moderated the impact of the intervention on the overall index of AED
Table 3 Moderators of the effect of the implementation intention intervention on AED adherence.
Table 2 Effects of the implementation intention intervention on three measures of AED adherence. Intervention
Percentage of doses taken Percentage of days correct doses taken Percentage of doses taken on schedule Overall adherence
93.4 (12.3) 88.7 (15.1)
79.1 (28.1) 65.3 (35.6)
78.8 (23.5) 0.35 (0.55)
55.3 (34.8) 0.40 (1.15)
Note. Values are means (SD). Overall adherence scores were generated by standardizing and then averaging the three percentage measures. a P < 0.01. b P < 0.001.
0.36b 0.40a 0.39b
0.61a 0.39a 0.31b
Missed doses at baseline Condition Missed doses condition
Illness concern Condition Illness concern condition
Prospective memory Condition Prospective memory condition
Step 1 1 2
P < 0.001. P < 0.05. One-tailed test.
I. Brown et al. / Epilepsy & Behavior 16 (2009) 634–639
and 90% indicated that they ‘‘deﬁnitely” intended to take their medication, our study conﬁrmed that many patients do not adhere to their AED treatment schedules. In our control group, only 55.3% of all doses were taken on schedule, the correct number of doses was taken on 65.3% of days, and only 79.1% of doses were taken overall. These high rates of nonadherence are not unusual. In fact, they fall in the middle of the range suggested by studies in epilepsy using other assessment methods , and are virtually identical to rates obtained in a previous study using the same methodology . The ﬁndings from the intervention arm of our study suggest that implementation intentions can produce substantial improvements in AED adherence. In fact, the effect size for our intervention (d = 0.78) compares very favorably with that of adherence improvement initiatives for other diseases . Importantly, the III beneﬁted mainly those participants who were less concerned about their illness, who had missed taking their medication in the past, and who had poor prospective memory—in other words, the patient groups at greatest risk of missing doses . These ﬁndings are line with previous studies demonstrating the positive impact of IIIs on action initiation and goal attainment , and with evidence that this type of planning is especially advantageous for people with cognitive and emotional difﬁculties [33,34]. The levels of the depression, anxiety, and memory functioning observed in this study document how serious the cognitive and emotional problems are in this particular patient group. 4.2. Limitations This study has a number of limitations. Most importantly, this ﬁrst study of an III in patients with epilepsy did not include a period of baseline monitoring prior to the intervention. This means that we cannot rule out with certainty that the different adherence levels we found in the intervention and control groups are not related to the documented differences in illness perceptions. In view of our positive ﬁndings, future studies of IIIs involving a more complex design (including a baseline monitoring period) are now justiﬁed. Next, tablet taking was monitored for only 1 month after the III. In view of the fact that the intervention aimed to make tablet taking more ‘‘automatic” and that each successful implementation of the formed intention would strengthen the link between the environmental clues and the act of tablet taking, it would be reasonable to hypothesize that the intervention effect should be maintained. Indeed, IIIs in other areas have shown sustained improvements in performance over periods of 1 year . Nevertheless, the long-term effectiveness of our intervention remains unproven. A further limitation of our ﬁndings to date is that our study was not designed or powered to detect differences in important outcomes such as the frequency of seizures, the percentage of patients achieving full seizure control, and the frequency or severity of side effects. However, a link between increased adherence and reduction in symptomatology has been suggested by other studies [7,14,36,37], and the present study also obtained a signiﬁcant correlation between the extent of adherence and scores on the Liverpool Seizure Severity Scale at follow-up (r = 0.23, P < 0.03, onetailed). Although the MEMS system is considered the best method for adherence research , and has been shown to be superior to pill counting, the coefﬁcient of variability of AED blood levels, and selfreport measures in patients with epilepsy [12,39], it is associated with some important drawbacks. For instance, choosing the MEMS system for this study meant that patients taking large hydrophilic tablets (such as the most important brands containing sodium valproate) had to be excluded. However, there is every reason to suppose that the III used here would be effective in promoting tablet
taking no matter whether tablets are dispensed via a bottle, a blister pack, or another type of dispenser. Another potential drawback is the lack of additional biological measures of adherence. The use of MEMS pill bottles does not prove that tablets were actually taken, only that the medication bottle was opened at a particular time. It would make sense in future studies to combine the MEMS system with a measure such as hair analysis that can provide biochemical evidence of medication ingestion over time . 5. Conclusion Despite these limitations the present study conﬁrms that nonadherence to AED treatment is a common problem. Our data suggest that cognitive or memory problems are a more signiﬁcant reason for nonadherence than are unwillingness to take medication and carelessness. The overwhelming majority of patients were keen to take their medication regularly. In this context, our simple intention implementation tool is a promising intervention and was associated with a large treatment effect in this randomized trial. Further research should look at treatment schedules with more frequent doses where adherence has been shown to be a much greater problem . Sufﬁciently powered future studies should include a period of baseline monitoring, be designed to provide further conﬁrmation that the intervention really is effective, and use a longer period of postintervention monitoring to demonstrate that the beneﬁts are sustained. Future studies also need to prove that improved adherence really produces better seizure control, fewer side effects, and, ultimately, reduced mortality. It also remains to be established how best to deliver the intervention: in the physician’s ofﬁce, at the pharmacy, or incorporated into medication package design. Acknowledgments The authors thank the following organizations that provided ﬁnancial support to purchase equipment and consumables for this study: Janssen Cilag provided $6582, Epilepsy Action provided $2950, and the University of Shefﬁeld provided $985. The funders had no role in study design; the collection, analysis, and interpretation of data; the writing of the report; and the decision to submit the article for publication. The authors also thank the staff and patients at the Royal Hallamshire Hospital outpatient neurology clinic, especially Dr. Grünewald, Dr. Howell, Dr. Hills, and Dr. Sarasama; the inpatient pharmacy staff, Helen Bowler and Edna Webster; and Brian Parkinson for his design advice on the graphic for the implementation intention. References  Kwan P, Brodie MJ. Early identiﬁcation of refractory epilepsy. N Engl J Med 2000;342:314–9.  Schuele SU, Luders HO. Intractable epilepsy: management and therapeutic alternatives. Lancet Neurol 2008;7:514–24.  Leppik IE. How to get patients with epilepsy to take their medication: the problem of noncompliance. Postgrad Med 1990;88:253–6.  DiMatteo MR. Variations in patients’ adherence to medical recommendations: a quantitative review of 50 years of research. Med Care 2004;42:200–9.  Chen JWY, Wasterlain CG. Status epilepticus: pathophysiology and management in adults. Lancet Neurol 2006;5:246–56.  Buelow JM, Smith MC. Medication management by the person with epilepsy: perception versus reality. Epilepsy Behav 2004;5:401–6.  Cramer JA, Glassman M, Rienzi V. The relationship between poor medication compliance and seizures. Epilepsy Behav 2002;3:338–42.  Williams J, Myson V, Steward S, et al. Self-discontinuation of antiepileptic medication in pregnancy: detection by hair analysis. Epilepsia 2002;43:824–31.  Buck D, Jacoby A, Baker GA, Chadwick DW. Factors inﬂuencing compliance with antiepileptic drug regimes. Seizure 1997;6:87–93.
I. Brown et al. / Epilepsy & Behavior 16 (2009) 634–639  Thatcher Shope J. Compliance in children and adults: review of studies. Epilepsy Res Suppl 1988;1:23–47.  Faught E, Duh MS, Weiner JR, Guerin A, Cunnington MC. Nonadherence to antiepileptic drugs and increased mortality: ﬁndings from the RANSOM Study. Neurology 2008;71:1572–8.  Cramer JA, Mattson RH, Prevey ML, Scheyer RD, Ouellette VL. How often is medication taken as prescribed? A novel assessment technique. JAMA 1989;261:3273–7.  Davis KL, Candrilli SD, Edin HM. Prevalence and cost of nonadherence with antiepileptic drugs in an adult managed care population. Epilepsia 2008;49: 446–54.  Peterson GM, McLean S, Millingen KS. A randomised trial of strategies to improve patient compliance with anticonvulsant therapy. Epilepsia 1984;25:412–7.  Takaki S, Kurokawa T, Aoyama T. Monitoring drug noncompliance in epileptic patients: assessing phenobarbital plasma levels. Ther Drug Monit 1985;7: 87–91.  Stanaway L, Lambie DG, Johnson RH. Non-compliance with anticonvulsant therapy as a cause of seizures. N Z Med J 1985;98:150–2.  Irving P, Al DA, Srinivasan A, Greenwood R. An audit of admissions of patients with epilepsy to a district general hospital. Seizure 1999;8:166–9.  Williams J, Lawthom C, Dunstan FD, et al. Variability of antiepileptic medication taking behaviour in sudden unexplained death in epilepsy: hair analysis at autopsy. J Neurol Neurosurg Psychiatry 2006;77:481–4.  Sherwin AL, Robb JP, Lechter M. Improved control of epilepsy by monitoring plasma ethosuximide. Arch Neurol 1973;28:178–81.  Wannamaker BB, Morton Jr WA, Gross AJ, Saunders S. Improvement in antiepileptic drug levels following reduction of intervals between clinic visits. Epilepsia 1980;21:155–62.  Gollwitzer PM, Sheeran P. Implementation intentions and goal achievement: a meta-analysis of effects and processes. Adv Exp Soc Psychol 2006;38:119.  Gollwitzer PM. Implementation intentions: strong effects of simple plans. Am Psychol 1999;54:493–503.  Webb TL, Sheeran P. How do implementation intentions promote goal attainment? A test of component processes. J Exp Soc Psychol 2007;43: 295–302.  Broadbent E, Petrie KJ, Main J, Weinman J. The brief illness perception questionnaire. J Psychosom Res 2006;60:631–7.  Seidenberg M, Haltiner A, Taylor MA, Hermann BB, Wyler A. Development and validation of a Multiple Ability Self-report Questionnaire. J Clin Exp Neuropsychol 1994;16:93–104.
 Zigmont AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr Scand 1983;67:361–70.  Scott-Lennox J, Bryant-Comstock L, Lennox R, Baker GA. Reliability, validity and responsiveness of a revised scoring system for the Liverpool Seizure Severity Scale. Epilepsy Res 2001;44:53–63.  Baker GA, Smith DF, Jacoby A, Hayes JA, Chadwick DW. Liverpool Seizure Severity Scale revisited. Seizure 1998;7:201–5.  Crawford JR, Smith G, Maylor EA, Della SS, Logie RH. The Prospective and Retrospective Memory Questionnaire (PRMQ): normative data and latent structure in a large non-clinical sample. Memory 2003;11:261–75.  Aiken LS, West SG. Multiple regression: testing and interpreting interactions. Newbury Park, CA: Sage; 1991.  McDonald HP, Garg AX, Haynes RB. Interventions to enhance patient adherence to medication prescriptions: scientiﬁc review. JAMA 2002;288: 2868–79.  Jones RM, Butler JA, Thomas VA, Peveler RC, Prevett M. Adherence to treatment in patients with epilepsy: associations with seizure control and illness beliefs. Seizure 2006;15:504–8.  Lengfelder A, Gollwitzer PM. Reﬂective and reﬂexive action control in patients with frontal brain lesions. Neuropsychology 2001;15:80–100.  Sheeran P, Aubrey R, Kellet S. Increasing attendance for psychotherapy: implementation intentions and the self-regulation of attendance-related negative affect. J Consult Clin Psychol 2007;75:853–63.  Sheeran P, Milne S, Webb TL, Gollwitzer PM. Implementation intentions and health behaviours. In: Conner M, Norman P, editors. Predicting health behaviour: research and practice with social cognition models. Buckingham: Open Univ. Press; 2005. p. 276–323.  Mattson RH, Cramer JA, Collins JF. Aspects of compliance: taking drugs and keeping clinic appointments. Epilepsy Res Suppl 1988;1:111–7.  Leppik IE. Variability of phenytoin, carbamazepine and valproate concentrations in a clinic population. Epilepsy Res Suppl 1988;1:85–90.  Dunbar-Jacob J, Mortimer-Stephens MK. Treatment adherence in chronic disease. J Clin Epidemiol 2001;54(Suppl. 1):S57–60.  Cramer J, Vachon L, Desforges C, Sussman NM. Dose frequency and dose interval compliance with multiple antiepileptic medications during a controlled clinical trial. Epilepsia 1995;36:1111–7.  Williams J, Patsalos PN, Mei Z, Schapel G, Wilson JF, Richens A. Relation between dosage of carbamazepine and concentration in hair and plasma samples from a compliant inpatient epileptic population. Ther Drug Monit 2001;23:15–20.