Juan Li and Yang Si contributed equally to this article.
This randomized intervention trial was to determine whether the implementation of a practical intervention was effective in enhancing medical compliance and improving seizure control among patients with convulsive epilepsy in rural communities in western China.
Two of four areas were randomly selected for this study and assigned to be the intervention group (IG) and the control group (CG), respectively. An intervention package with four components (intensive education, consultation services, maintenance of an epilepsy tracking card, and repeated reminders) was formulated. Medical compliance included antiepileptic drug (AED) adherence and lifestyle; each was graded on a 6-point scale with possible scores. Medical compliance and seizure control were measured and compared between the groups before and after the intervention. In addition, correlation of both changes in medical compliance and seizure frequency were investigated.
After 1-year follow-up, 183 patients in the IG (105 male) and 177 in the CG (99 male) remained for the analysis. At the end of the study, the average number of seizures in the IG declined 18.3% compared to that prior to the intervention (after 6-month phenobarbital monotherapy), nearly twice as much as in CG (9.1%) with statistical difference (p = 0.023). The proportion of patients with a reduction in seizures >50% (including those who were seizure-free) rose to 79.8% in the IG compared to 61.0% in the CG (p < 0.05). With regard to medical compliance, the majority of the IG members were rated as excellent or very good, but medical compliance remained nearly unchanged for the CG. A moderate correlation was found between the changes in AED adherence and seizure control (r = 0.4, p < 0.05), and a weaker correlation was found between lifestyle and seizure control (r = 0.328, p < 0.05).
This intervention package proved to be efficient in enhancing medical compliance and improving seizure control in rural communities of resource-poor areas.
Epilepsy is one of the most common chronic neurologic disorders, and epilepsy patients carry an increased risk of morbidity and mortality compared to the general population (Neligan et al., 2011). Around the world, nearly 50 million people are affected by this disease, and 80% of them live in low-income countries (WHO, 2004). In China, approximately 9 million people are estimated to have epilepsy; the lifetime prevalence is 7.0/1,000 (Wang et al., 2003). Uncontrolled seizures, which may lead to status epilepticus and sudden unexplained death, are believed to be the primary driver of the heightened morbidity and mortality in this population (Walczak et al., 2001; Sperling, 2004). Although antiepileptic drugs (AEDs) offer effective seizure prevention in approximately 70% of patients with epilepsy when the most effective regimen is followed (Kwan & Brodie, 2000), unfortunately, early reports suggest that AED nonadherence is highly prevalent, with estimates ranging from 20% to 80% (Burkhart & Sabaté, 2003). Current evidence also indicates that adherence to medications among patients with epilepsy remains suboptimal (Briesacher et al., 2008; Davis et al., 2008; Ettinger & Baker, 2009). Poor medical compliance indicates that more patients will have an enduring predisposition to chronic seizures, exposing themselves to a relatively high risk of morbidity and mortality. AED adherence was intensively studied in developed countries for decades, but in developing countries, reports have been relatively limited. Because there were few randomized or quasi-randomized controlled trials of adherence-enhancing interventions, the findings were varied, and the results were inconsistent from patient to patient (Al-Aqeel & Al-Sabhan, 2011). A recent review suggests that education and counseling for patients with epilepsy have shown mixed success, and behavioral interventions had better effects on adherence (Al-Aqeel & Al-Sabhan, 2011).
In 1997, under the Global Campaign Against Epilepsy launched by the World Health Organization (WHO) in cooperation with the International League Against Epilepsy (ILAE) and the International Bureau for Epilepsy, a project that included an epidemiologic survey, an intervention trial, and an educational program was initiated to study epilepsy in China (Sander, 2002). In 2000, another large-scale project supported by the government was launched in the northern and eastern rural areas of China to identify and manage active convulsive epilepsy (ACE) in these communities. In 2005, a project with the same protocol and design as the project launched in 2000 was extended to the rural areas of western China. After 2 years of preparation, an operational network for managing patients with ACE was established to guarantee successful management of patients with convulsive epilepsy in rural areas (Liu et al., 2010). Every 3–6 years, new targeted areas in western China were incorporated into our on-going epilepsy project. Routine procedures for screening and identifying convulsive epilepsy in those areas were initiated. Thereafter, AEDs were provided, and follow-up was conducted.
Some success was achieved in controlling seizures by administering free phenobarbital to treat epilepsy (Wang et al., 2006; Liu et al., 2010). However, during those years of epilepsy management, we observed that the medical compliance of the patients in rural areas was not optimal. We wondered whether medical nonadherence could be reduced, consequently reducing seizures as much as possible, by conducting an intervention to enhance the adherence of patients with epilepsy in rural areas. To answer this question, we designed a study to assess the medical adherence of patients with convulsive epilepsy in rural communities of western China. We used an adherence-enhancing intervention to determine some basic characteristics of medical compliance in this cohort. In addition, the present study fills a research gap in the literature and provides guidance for epilepsy management in low-income countries.
Design and areas
This randomized intervention trial was conducted in rural communities of western China between September 2009 and December 2012. Under the rural epilepsy management project supported by the government, four new target areas were recruited into the epilepsy management project in 2009. Baseline characteristics (e.g., socioeconomic level, education, and primary health care facilities) of those areas were compared previously, and no obvious differences were found. We randomly chose two of four newly selected target areas to complete this prospective trial (Fig. 1). In addition, the study designers, local physicians, patients, and data analyst were blinded to the intervention.
Procedures and participants
The Institutional Ethics Committee of West China Hospital gave ethical approval for the epilepsy project and our study. The entire study process consisted of three phases. The first phase included identifying and managing the patients with convulsive epilepsy following the standard rules of the epilepsy project, a process that has been described in previous publications (Wang et al., 2006; Liu et al., 2010). In brief, patients with convulsive epilepsy were identified initially in the primary health care unit using a clue survey method and confirmed by the senior neurologic physicians. To be included in the study, the patients had to be older than 2 years, with at least two convulsive seizures in the previous 12 months (active epilepsy), and who were not on standard AED treatment. Next, all of the patients entering the study were treated with phenobarbital monotherapy for free and underwent continuous surveillance during the follow-up period (including dosage adjustment, side effects monitoring, and withdrawal from the study). In the first stage, the patients were required to undergo 6 months of phenobarbital monotherapy, and then we continuously enrolled 200 participants from each target area into the intervention group (IG) and the control group (CG) to study medical compliance (phase two). Before conducting the intervention, the factors that could possibly affect medical compliance were studied, and a practical intervention program was formulated. In the second phase, the inclusion criteria for the participants were as follows: (1) age ≥13 and <65 years; (2) retention in the project and constant receipt of phenobarbital monotherapy (note that in this period, the patients had reached an appropriate dosage after the previous 6 months of drug adjustment); (3) agreeing to participate and providing written consent to participate in the project (for children under the age of 18, consent was given by their guardians). The exclusion criteria included the following: (1) patients with severe mental retardation or neurologic diseases or psychosis; and (2) patients receiving another one or two AEDs in addition to phenobarbital as additional therapy. In this phase, interventions were implemented for the IG. The CG continued to receive the standard management previously described; therefore, some routine measures such as phenobarbital and routine public education were available (Liu et al., 2010). The third phase was the 1-year follow-up period after the intervention was implemented. This period was divided into two parts (the first 6 months and the next 6 months) for observation. Figure 2 shows the general research process.
Randomization and concealment
Initially four new areas with similar features (such as demographic characteristics, economic status, and so on) and so on. in our epilepsy project were chosen as candidates for this study. In phase 2, we selected two areas out of four by a simple randomization (by random selection software). Then we determined the intervention known as IG, also by a simple randomization. For each group, different management protocols were formulated (intervention for IG and routine measures for CG). Local physicians were responsible for the implementation of the protocols. The whole study schedule was not revealed to local physicians and patients during the entire process. We only required those physicians to be involved into an epilepsy project named as “management of active convulsive epilepsy in rural area.” Furthermore we took measures to avoid any possible direct communication between the primary physicians from each area in order to reduce some possible exposure of intervention for the CG. This project ensured that those running the study, including outcome assessors and those administering the intervention, were blinded to allocation.
The intervention components were partially derived from the evaluation of the factors influencing medical compliance during phase 2 of the study. Aiming for efficiency and practicality, adjustments were discussed and implemented by experts in our epilepsy center and epidemiologists in the field. Feedback from primary care physicians also contributed considerably to the decision-making process. Ultimately, the intervention program was established with four components. First, the intervention provided intensive education that included an elaborate explanation of this type of disease and the potential precipitating factors, emphasizing the importance of receiving appropriate AED treatment, taking medications regularly, and reporting side effects, as well as the hazards of superstition or distorted beliefs. Second, consultation services ensured that clinical providers and telephone support were available for the patients at any time. Well-trained rural community health workers or primary care physicians were hired to fill these roles. Third, the intervention used reminders by keeping a simple record with our designed card (Fig. 3). Fourth, patients received repeated (>3 times at each attending clinic) reminders about medical adherence every month when they picked up the 1-month supply of their AED at the primary care rural community clinic. The patients in the CG received routine management (Liu et al., 2010), including a phenobarbital supply, public education (such as distributed pamphlets and media messages), and free community medical consultations. (These services were not like the intensive individual education and medical consultation). The patients in the IG could obtain the same public education and consultation services as the control patients.
Outcome measurements and definitions
Compliance is generally defined as the extent to which the patients’ treatment-related behaviors match the provider-recommended treatment plan (Paschal et al., 2008). In the present study, medical compliance included two primary components: AED (phenobarbital) adherence and avoiding lifestyle-precipitated seizures. Because the patients were enrolled in the epilepsy project, all of the patients were required to attend the designated clinics in the rural area monthly to pick up a 1-month supply of their AED; therefore, clinic attendance was not taken into account in the assessment of compliance. Medical compliance was evaluated at the beginning of phase 2 after all enrolled participants had completed 6 months of AED treatment. These data provided a baseline to compare to the reassessment data gathered 6 months after intervention and at the end point of the study. With regard to AED adherence, there is no consensus about a gold standard measurement method (De Geest & Sabaté, 2003). Following the suggestions of a previous study (Preston & Colman, 2000), rating scales were used to assess drug adherence. Six response options were available (Lu et al., 2008). The categories were very poor, poor, fair, good, very good, and excellent, which were, respectively, assigned scores of 0, 20, 40, 60, 80, and 100 (with 100 reflecting the best adherence). We defined the following categories: excellent, no missing doses in the 3-month period prior to the interview; very good, 1 or 2 missed doses (an average of less than once per month); good, 3–6 missed doses (an average of one to two times per month); fair, 7–12 missed doses (an average of more than once per 2 weeks to once per week); poor, 13–24 missed doses (an average of more than once per week to less than twice per week); very poor, >24 missed doses (an average range of more than twice per week). With regard to lifestyle or habits, six similar ratings were applied to measure the frequency of seizure-provoking events. The events that reflected nonadherence were drinking tea, coffee, and alcohol, staying up late, becoming overly fatigued (like watching TV and using a computer for a long time). Slightly differently from the self-reported estimated categories, we quantified the categories by calculating the medical compliance events (frequency of missed doses, and so on,) recorded on the epilepsy tracking card for IG. However, in the CG, medical compliance ratings (excluding AED adherence) were derived from self-reported data and we calculated AED adherence by counting the remaining pills to count the number of missed doses.
The seizure assessment was conducted at the beginning of phase 1, the beginning of phase 2, and the end of phase 3. The number of seizures that occurred in the 3-month period before the assessment was counted as the seizure frequency. Seizure frequency in the 3 months before the beginning of phase 1 served as the baseline, which was used to compare the seizure frequency in phase 2 and at the end point of phase 3. We calculated the seizure frequency in the last 3 months prior to phase 2 or 3, and we divided the frequency by the baseline seizure frequency. From this calculation, we determined whether there had been a >50% reduction in seizure frequency. We defined a >50% reduction as effective seizure control. The patients were considered seizure-free if they had no seizures for the 12 months of follow-up (as measured in phase 3). The seizure assessment at the beginning of phase 1 was based on self-report. The subsequent seizure assessment for the IG was obtained from the epilepsy tracking card. In CG the seizure assessment was based on self-report monthly.
The analyses were conducted with SPSS 17.0 (SPSS, Chicago, IL, U.S.A.). The chi-square test or corrected chi-square test or Fisher's exact test and a one-way analysis of variance (ANOVA) were used to conduct statistical comparisons of categorical and continuous variables, respectively. Before conducting the ANOVA, the distribution of the data was tested by assessing skewness and kurtosis. The data were transformed to produce a more normal distribution when necessary and possible. Second, a test of homogeneity of variance was performed to guarantee the validity of the ANOVA. When the ANOVA was not appropriate, the Mann-Whitney test was considered for continuous variables. For bivariate analyses, Spearman's rank correlation was used when at least one of the variables was an ordered variable. In the present study, we used that method to analyze the correlation between the changes in seizure frequency and the changes in the medical compliance score. More specifically, seizure control at the end point of the study was calculated as previously described. However, the number of seizures in the 3 months prior to the end point was compared to the number of seizures in the 3 months prior to phase 2 instead of phase 1. Six categories were created to rate seizure reduction: grade 1, no seizure reduction or aggravation; grade 2, 0 < seizure reduction < 25%; grade 3, 25% < seizure reduction < 50%; grade 4, 50% ≤ seizure reduction < 75%; grade 5, 75% ≤ seizure reduction < 100%; grade 6, seizure reduction = 100%. Changes in the medical compliance score were calculated before and after the intervention by subtracting the score at the end point from the score in phase 2.
No literature data were available to support sample size estimation in the present study directly due to different population recruited and methods used. We expected the percentage of subjects with a medical compliance score ≥60 after 12-month intervention in IG were at least 20% more than that in CG. Therefore, to achieve a statistical power of 90% and a two-tailed 95% significance level, a sample of 200 subjects for each group was adequate.
This prospective intervention study included a sample of 200 patients with ACE for each group (IG and CG). After a 12-month follow-up, 183 cases (91.5%) were retained in IG and 177 (88.5%) in CG. A total of 40 cases withdrew from the project including 21 dropouts migrated from the study area, 5 cases due to severe adverse effect, 8 cases because of their perception of “cure” and refusing to receive continuous medication, 3 deaths, and 3 lost to follow-up without any specific reasons.
Table 1 presents the general information about the cohort. At baseline, the two study groups were similar to each other, with no obvious significant differences before the intervention.
Table 1. General characteristics of the cohort
The initial evaluation of monetary data was performed using Chinese currency, renminbi (RMB). For reference, the exchange rate was USD 1 = RMB 6.821 in 2008.
Between-group differences calculated using chi-square test.
Between-group differences calculated using Mann-Whitney U-test.
Figure 4 shows the changes in the average number of seizures in both study groups. After 6 months of monotherapy with phenobarbital, there was a 41.2% reduction in seizures in the IG and a 38.3% reduction in the CG. Over the next year of follow-up after the intervention, the number of seizures in the IG further declined 18.3%, nearly twice as much as in CG (9.1%).
Table 2 presents some detailed information about seizure control and medical compliance in both groups before and after the intervention. Before the intervention, the baseline numbers of patients with >50% seizure reduction after 6-month phenobarbital monotherapy and AED adherence were similar in both groups. However, there was a difference in the number of patients with lifestyle ratings of fair and very poor. After the intervention, the proportion of patients with >50% seizure reduction (including those who were seizure-free) rose to 79.8% in the IG compared to 61.0% in the CG (p < 0.05). Nevertheless, both groups had a >50% seizure reduction, and there was a statistically significant difference in self-control method (all p < 0.05). With regard to medical compliance, there was a remarkable change for the IG; the majority of the group was rated excellent or very good. However, medical compliance remained nearly unchanged for the CG.
Table 2. Assessment of seizure control and medical compliance during the follow-up period
IG (n = 183)
CG (n = 177)
Cases with >50% reduction at end point in seizure frequency include those who are seizure-free.
Between-group differences calculated using chi-square test.
Between-group differences calculated using Mann-Whitney U-test.
Figures 5 and 6 show that the medical compliance scores increased more rapidly in the initial 6 months after the intervention.
Table 3 shows the correlation between the changes in seizure control at the end point of the study compared to phase 2 and the changes in medical compliance before and after intervention. The table shows a moderate correlation between AED adherence and seizure control (r = 0.4, p < 0.05), but a weaker correlation between lifestyle and seizure control (r = 0.328, p < 0.05).
Table 3. Correlation between seizure control ratings and changes in medical compliance
Spearman's rank correlation was used.
AEDs are predominantly used to control seizures, and the successful control of seizures partially depends on medical compliance. The reported rates of noncompliance with AEDs remained high (30–60%; Green & Simons-Morton, 1988; Leppik, 1990; Jones et al., 2006). To date, medical compliance has been intensively investigated in high-income countries. Previous research has identified a number of factors affecting noncompliance, including side effects, interference with daily life, poor memory, duration of the disease, cost, medical barriers, and the patients’ understanding of the disease. Based on these factors, various strategies have been developed in high-income countries. However, there is little knowledge on the topic in developing countries. The present study did not concentrate on the factors contributing to medical compliance; we focused on whether our practical intervention strategy could enhance compliance and whether this intervention improved seizure control to an extent that justified the development of the intervention.
Because the reasons for nonadherence are complex and multilayered (and might have a direct or indirect impact on the effects of our practical intervention), we had several inclusion criteria to ensure that the two target groups were identical at baseline. Another feature of our design was the beginning of the intervention after 6 months of monotherapy without the intervention. This 6-month period might help us to observe the effect of the intervention by excluding the direct effect of phenobarbital on reducing seizures, which means further seizure control might be more closely related to the level of medical compliance. However, in this study, we continuously enrolled participants instead of selecting those with suboptimal seizure control who might be more likely to have refractory epilepsy or medical noncompliance. This type of cohort reflects the general population with convulsive epilepsy in those areas.
In our study, 1 year after the intervention, the IG had better seizure control; the total number of seizures decreased approximately twice as much as the decline among the control subjects (17.8% vs. 9.2%). Although this decrease was not immense, we still revealed an obvious difference by comparing the number of patients with a reduction >50% between the groups.
Previous randomized intervention trials demonstrated that there was no consensus about enhancing compliance through various measures. DiIorio et al. (2009) reported that motivational interviewing aimed at enhancing self-management practices failed to show an effect on adherence. Peterson et al. (1984) found that patient reminders, such as reminders mailed together with a counselling leaflet, produce promising effects on adherence. Verbal and written educational materials improved knowledge scores but produced no significant increase or decrease in the mean serum levels of a prescribed antiepileptic (Pryse-Phillips et al., 1982). In comparison, our simple package of intervention techniques, including education and reminders, proved to be effective in enhancing AED adherence and preventing seizure-precipitating lifestyle habits. Patients may have overestimated their adherence in their self-reported data (Stephenson et al., 1993).
By calculating the percentage of patients who were adherent over a cumulative follow-up interval, the study of Manjunath et al. shows a precipitous decline in the first 4 months after AED initiation (Manjunath et al., 2009). However, by using the scoring method, there were no obvious changes in the total AED adherence score in the CG. The heterogeneity of the enrolled population might partially explain this phenomenon because our cohort did not include those who withdrew from the project because of the adverse effects of AEDs within the first 6 months of exposure to the AED. We note that the regular supply of free phenobarbital in our project might be the strongest motivating factor for the patients to be AED adherent. In our study, we confirmed that medical adherence was greatly improved during the first 6 months following the intervention. This result might prompt the question of whether the subsequent 6-month intervention following the first 6-month intervention is necessary to retain a high compliance level.
The results about the relationship between medical adherence and epilepsy control are not conclusive, owing to the nature of the illness and the efficacy of AEDs (Sweileh et al., 2011). In our study, we did not predict a relationship between seizure frequency and medical adherence. Instead, we focused on the relationship between changing medical compliance and seizure frequency after the implementation of an intervention. In other words, we investigated whether the enhancement in medical compliance was associated with the reduction in seizures. We avoided certain aspects of bias and confounding factors by using monotherapy of phenobarbital and formulating inclusion criteria to minimize the internal complexity of influencing factors. We observed that a relationship existed, although it failed to reach the level of statistical significance (AED adherence r = 0.4, lifestyle r = 0.328).
There are limitations inherent in collecting self-reported data at the beginning of phase 1 and using self-reported data for CG members in the subsequent follow-up period, although we collected objective data from the IG through the epilepsy tracking card during the follow-up period. Other methods, such as the evaluation of serum medication levels or the use of electronic monitoring caps, might be more objective (Hess et al., 2006; Paschal et al., 2008). The nonadherence scoring system used in our survey varied from the measures used in other studies because there is no gold standard for measuring nonadherence. In the present study, the intervention was developed for local practicality based on the experience of experts, but it lacks adequate evidence to prove its effectiveness. Although the local physicians who were responsible for the preliminary evaluation of patients were trained before the project started, there may still have been differences in individual knowledge about epilepsy and the adherence to the intervention procedures and follow-up protocol, even though all local physicians were required to qualify to participate in the intervention to minimize these potential biases.
Considering that the purpose of this trial was to test the effectiveness of a practical intervention package, we could not distinguish whether some of the four intervention items were ineffective or determine the relationship between the items. Nevertheless, the intervention package increased patients’ medical compliance and enhanced seizure control, demonstrating its practicality and effectiveness in rural communities. Whether it should be popularized in less-developed areas merits further discussion.
We sincerely thank all of the participants in this project and all of our colleagues in the local primary heath care centers for their cooperation. We also thank the experts at the CDC of Sichuan province for their assistance.
The authors declare no conflict of interest. The authors 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.