A high risk of hyperlipidemia in epilepsy patients: a
nationwide
population-based cohort study
Tomor Harnod MD
a,b, Hsuan-Ju Chen MSc
c, Tsai-Chung Li PhD
d,e, Fung-Chang
Sung PhD
c,
Chia-Hung Kao MD
f,g,*Introduction
Epilepsy is a well-recognized brain disease affecting 0.5% to 1.0% of people worldwide [1,2]. The pathogenesis of epilepsy indicates that it is of idiopathic origin or it manifests as symptomatic epilepsy after an insult to the brain. Among the many causes of symptomatic epilepsy in adults, the three major ones are head injury, stroke, and brain tumor [3]. A recent study reported a year-standardized mortality ratio of 2.5 for all-cause mortality among epilepsy patients
relative to that of the general population in Taiwan [4]. Both symptomatic and idiopathic epilepsies have been associated with an increased risk of ischemic heart disease (IHD) and cerebrovascular accident (CVA) in adults [5e8]. IHD and CVA are the major
causes of mortality in epilepsy patients who develop vascular disease before epilepsy [9]. Animal models of epilepsy have provided
strong support for cardiac mechanisms in sudden unexpected death in epilepsy through histologic evidence of myocardial ischemia after an epileptic seizure [10]. However, the involved pathogenesis and mechanism linking epilepsy and vascular disease remains unclear. Therefore, we conducted a nationwide
population-based study by analyzing time series data obtained from the National Health Insurance Research Database (NHIRD) of Taiwan to determine whether the risk of developing hyperlipidemia (HL) is higher in epilepsy patients, and whether epilepsy and elevated risks of IHD and CVA in patients could be linked by means of HL.
Methods Data sources
The NHIRD was obtained from the National Health Research Institutes. In March 1995, the Taiwanese government launched the National Health Insurance (NHI) program. By the end of 2009, the
NHI program provided insurance coverage for approximately 99% of the national population (approximately 23.74 million people). The NHIRD contains the personal data of enrollees, including their sex, date of birth, outpatient and inpatient history, and records of all
diagnosed diseases, which are coded in accordance with the International Classifcation of Diseases, Ninth Revision, Clinical Modifcation
(ICD-9-CM). To protect the privacy of enrollees, all Taiwanese citizens are assigned a unique personal identifcation number that can be used to link all NHI data sets without revealing any personal information. All personal identifcation numbers are cryptographically scrambled to ensure patient anonymity. This study was
approved by the Institutional Review Board of China Medical University, Taiwan (CMU-REC-101-012).
Study participants
We identifed 5012 patients with a history of epilepsy and repetitive seizures (ICD-9-CM Code 345.xx) in the registry of ambulatory and inpatient claims data from 2000 to 2005. Patients were excluded if they were aged younger than 40 years (n ? 2338), had been a previous diagnosis of HL (ICD-9-CM Code 272.xx) (n ? 967) or stroke (ICD-9-CM Codes 430.xxe438.xx) (n ? 576), or were followed up for less than 1 year (n ? 141). For each epilepsy patient in
the epilepsy cohort, we randomly selected four patients from the NHIRD and assigned them to the comparison cohort. Enrollees with no history of epilepsy, HL, or stroke were frequency matched by age (per 5 years), sex, and index-year. Overall, the epilepsy and comparison cohorts comprised 990 and 3960 enrollees, respectively.
Measurements
The demographic factors examined in this study were age and sex. Patients were categorized by age into three groups: 40 to 49, 50 to 59, and _60 years. We considered IHD (ICD-9-CM Codes
410.xxe414.xx), hypertension (ICD-9-CM Codes 401.xxe405.xx), and diabetes (ICD-9-CM Code 250.xx) as comorbidities that were potential confounders in the association between epilepsy and HL. The main outcome (occurrence of HL) was determined by linking ambulatory and inpatient care data. All patients were observed from the index date until the date of HL diagnosis, withdrawal from the NHI program, or December 31, 2011 (whichever occurred frst). Statistical analysis
The distribution of sociodemographic status and comorbidities between patients with and without epilepsy were analyzed using a c2 test for categorical variables and t test for continuous variables. The sex-, age-, and comorbidity-specifc incidence density rates of HL were calculated for the follow-up period until December 31, 2011, the date of HL diagnosis, death, or loss to follow-up. Patients who had less than 1 year of follow-up were excluded from the analysis. We used the KaplaneMeier (KeM) estimation method to
plot the cumulative incidence curves of HL for both cohorts. Subsequently, we performed a log-rank test to check for statistically
signifcant differences between the KeM curves. Cox proportional hazards models were used to assess the independent effects of epilepsy by adjusting the model to account for the effects of other variables in the models. We calculated the hazard ratios (HRs) and 95% confdence intervals (CIs) after adjusting the models to account for the effects of age, sex, and comorbidities. All statistical analyses were performed using SAS Version 9.3 (SAS Institute Inc., Cary, NC), with the signifcance level set to P < .05 for a two-tailed test. Results
The mean age of the 990 patients in the epilepsy cohort was
57.25 years (standard deviation ? 13.32) and that of the 3960 patients in the comparison cohort was 56.91 years (standard
deviation ? 13.53). The mean follow-up period was 6.63 years for the epilepsy cohort and 7.49 years for the comparison cohorts.
Table 1 shows the baseline demographic factors and comorbidities
of the study participants according to epilepsy status. The distribution of age and sex at entry were the same in both cohorts. The
epilepsy cohort exhibited a higher prevalence of IHD (19.39% vs. 12.85%), hypertension (41.62% vs. 25.91%), and diabetes (5.29% vs. 5.98%) compared with the patients without epilepsy.
Figure 1 presents the cumulative incidence curves of HL according
to epilepsy status.We conducted a log-rank test to compare the cumulative incidence of HL between the two cohorts and observed a signifcantly higher HL incidence in the epilepsy cohort (P < .001).
The incidence rate and HR for HL were calculated according to epilepsy status and stratifed based on demographic factors and comorbidities (Table 2). The incidence rate of HL was higher in the
epilepsy cohort than it was in the comparison cohort (34.15 vs. 26.96 per 1000 person-years), and the HR was 1.17 (95% CI, 1.01e1.36) after adjusting the model to account for the effects of age, sex, and comorbidity. The adjusted HR was considerably higher in the epilepsy cohort than that of the comparison cohort when the patients aged 50e59 years were considered (adjusted HR, 1.35; 95% CI, 1.01e1.79). For the patients with no comorbidity (i.e., without IHD, hypertension, and diabetes), the risk of developing HL was higher in the epilepsy cohort than in the comparison cohort after adjusting the multivariate model (HR, 1.27; 95% CI, 1.02e1.58). Among the patients without hypertension, those with epilepsy were more likely to develop HL compared with those without epilepsy (adjusted HR, 1.32; 95% CI, 1.08e1.61). Among the patients
with diabetes, those with epilepsy were also more likely to have HL compared with those without epilepsy (adjusted HR, 1.65; 95% CI,
1.07e2.54).
Figure 2 shows the combined effect of epilepsy and the three
comorbidities (IHD, hypertension, and diabetes) on the risk of HL relative to the patients with neither epilepsy nor comorbidity. The calculated HRs for developing HL in patients with epilepsy and IHD (HR, 1.56; 95% CI, 1.16e2.10), epilepsy and hypertension (HR, 1.82; 95% CI, 1.46e2.27), or epilepsy and diabetes (HR, 2.75; 95% CI, 1.96e3.87) were higher than those of the patients with neither epilepsy nor any comorbidity. The primary effects of epilepsy were statistically signifcant and had narrow 95% CIs, and were similar across risk stratifcation according to the three comorbidities. IHD, hypertension, and diabetes were signifcant risk factors for developing HL.
Discussion
Epilepsy patients are at higher risk of premature death
compared with people in the general population. A previous study showed that epilepsy patients were at higher risk of acute
myocardial infarction with a specifc odds ratio of 4.92, and an even greater risk of death after a myocardial infarction [11]. To eliminate the major confounding factors of preexisting diseases, we excluded patients with a history stroke or HL on entry, and we considered IHD, hypertension, diabetes as potential confounders in the association between epilepsy and HL. The results indicate that a statistically
signifcant association exists between epilepsy and
increased risk of HL. This fnding indicates the existence of a potential relationship between epilepsy and myocardial or cerebral
vascular diseases, particular in patients aged 50e59 years. Table 2
and Figure 2 categorize the participants in the two cohorts according
to the three comorbidities (IHD, hypertension, or diabetes). The epilepsy patients with no comorbidity exhibited a signifcantly higher risk of developing HL, indicating that epilepsy could be an independent risk factor of subsequent HL development, which could explain the increased risk of IHD and CVA in epilepsy patients.
Alternatively, because HL is typically accompanied by subtle cerebrovascular problems, such as lacunar infarction, the risk of epileptic seizure might be higher during the period between the cerebrovascular incident and CVA diagnosis. To eliminate this reverse causality, we used the KeM estimation method to plot cumulative incidence curves of HL and subsequently performed a log-rank test for statistically signifcant differences between the KeM curves of the epilepsy and comparison cohorts. The results in
Figure 1 indicate an increased risk of developing HL in the epilepsy
cohort over time, thus reducing the likelihood of reverse causality. The etiology or mechanism of HL may vary among epilepsy patients. Few genetic syndromes of epilepsy are related to weight gain, and genetic factors are not considered to play a role in the lipid
metabolism of patients with epilepsy [12]. Currently, three possible mechanisms of HL in epilepsy patients have been proposed. First,
compared with healthy patients, those with epilepsy engage in fewer physical activities of daily living are heavier and have a higher body mass index, even when their seizures are under control
[13,14]. Thus, the lipid profles of epilepsy patients might be higher
than those of healthy people. Second, a previous study reported that altered lipid metabolism might be associated with epilepsy although the alteration might not be infuenced by thyroid hormones alonedit might result from hypothalamusepituitary dysregulation caused by the insult from repetitive seizures [15].
Third, either monotherapy or polytherapy with antiepileptic drugs (AEDs), including carbamazepine, phenytoin, gabapentin, pregabalin, and vigabatrin, could be the primary mechanism
affecting the lipid profles of patients [12,16e18]. In particular, previous research reported signifcantly lower serum triglyceride and cholesterol levels in patients receiving valproate exhibit [16]. Moreover, several studies have shown that changes in the serum lipid profles that did not correlate with AED plasma levels although they did correlate with the induction of the cytochrome P450 system, particularly when older AEDs were used [16,18]. These reports indicate that further categorization according to the AED mechanism would be necessary to determine the overall effect of AEDs on HL. However, it would have been challenging to directly compare the AEDs to determine which drug was superior because our study design depended on the diagnoses listed in the NHI claims data. Moreover, directly contacting the patients to obtain their prescription details would have been impossible because the data are
depersonalized. This issue was the major limitation of this study. Although the fndings of this study are strengthened by the nationwide population-based design and representativeness, several other limitations were encountered while conducting this study. First, the NHIRD does not contain crucial information, such as the patients’ smoking habits, alcohol consumption, body mass index, socioeconomic status, and family history, all of which are confounding factors that may have affected HL development in the two cohorts. Second, compared with randomized and controlled trials, the evidence derived from cohort studies is typically weaker, specifcally because a cohort study design is subject to numerous biases related to the model adjustment when accounting for confounding factors. Despite our meticulous study design and adequate control over confounding factors, a key limitation of this study is that bias may exist if unmeasured or unknown confounders are present. However, several studies have demonstrated the high accuracy and validity of diagnoses based on the ICD-9-CM codes, as well as those obtained by studies with a similar design to this one, in which the NHIRD was used as the data source [19,20]. Thus, the results of this study provide a valuable insight into the relationship between epilepsy and HL.
Conclusions
The results of this population-based cohort study indicate a statistically signifcant association between epilepsy and increased
risk of HL, particularly in middle-aged patients. Based on these results, we conclude that epilepsy is an independent risk factor of HL development, which could explain the high risk of cerebral and cardiovascular diseases in epilepsy patients.
