FormalPara Key Points

There have been increasing reports to the World Health Organization linking levetiracetam use to moderate to severe forms of hypokalemia.

We investigated the association between levetiracetam use and hypokalemia by conducting a retrospective cohort study and found that users of levetiracetam had almost twice the risk of developing hypokalemia in comparison to carbamazepine users.

Although this study did not find a statistically significant higher risk of hypokalemia with levetiracetam use in comparison to carbamazepine, this potential risk cannot be ruled out and further research is needed to provide more evidence on this association.

1 Introduction

Hypokalemia may occur following medication use through enhancing renal potassium secretion (e.g., loop diuretics), affecting intracellular potassium shifting (e.g., insulins) or enhancing gastrointestinal loss (e.g., laxatives) [1,2,3,4]. The incidence of drug-induced hypokalemia may vary depending on the medication, dose, and individual patient factors, including age, renal function, hormonal regulation, and underlying medical conditions such as chronic kidney disease or heart failure [1,2,3,4,5]. Hypokalemia is characterized by a blood potassium level below 3.5 mmol/L, which may lead to cardiovascular complications including heart failure or ischemic heart disease, muscle weakness, ileus disease, or renal failure [4,5,6,7,8].

A potential safety signal of hypokalemia following levetiracetam use was published in the World Health Organization (WHO) newsletter based on 57 reported cases, and 21% of these cases were considered moderate to severe hypokalemia [9]. In most cases, hypokalemia developed shortly after patients started use of levetiracetam [9]. However, establishing a causal relationship between levetiracetam use and hypokalemia based on case reports might be challenging owing to the presence of confounders, risk factors, as well as concurrent use of other medications [9,10,11].

Levetiracetam is an anti-seizure medication (ASM) used to treat focal seizures, myoclonic seizures, and primary generalized tonic-clonic seizures in patients with epilepsy, in addition to seizures related to head injuries [12,13,14]. It exerts its effect through binding to the synaptic vesicle protein 2A in the brain and modulating the neurotransmitter release to reduce electrical activity and the likelihood of seizures [12, 13, 15,16,17,18]. The efficacy of levetiracetam in the treatment of the aforementioned conditions has been demonstrated in several clinical trials, with potential benefit that outweighs risks in many patients with seizures.

Although data from clinical trials did not suggest that hypokalemia as a potential risk with levetiracetam use, these trials may not provide a comprehensive safety profile owing to limitation in study population or size [19]. Therefore, the current knowledge on the potential association between levetiracetam use and hypokalemia is limited [9,10,11]. To address this gap, our study aimed to investigate the possible association between the use of levetiracetam and the development of hypokalemia, considering the potential influence of class effects and if patients with other drugs or conditions commonly associated with levetiracetam developed hypokalemia.

2 Methods

2.1 Data Source

This was a retrospective cohort study comprising adult patients (≥ 18 years old) who received either levetiracetam or carbamazepine between 1 January 2016 and 30 April 2022. Study data were retrieved from the Real-world Evidence Research Network (RERN), which serves as an environment for managing, aggregating, analyzing, visualizing, and facilitating the utilization of healthcare-related data from several healthcare institutions in Saudi Arabia [20]. Within this environment, comprehensive de-identified individual-level data encompassing demographics, diagnoses, procedures performed across inpatient or outpatient settings, and prescriptions issued to patients during routine healthcare encounters, are accessible upon granting approval of an institutional review board [20].

2.2 Study Population

Patients without continuous enrollment for at least 6 months prior to the index prescription (i.e. first prescription of levetiracetam or carbamazepine), identified from the active eligibility in their electronic health record (EHR), were excluded from the study [21, 22]. Patients with any evidence of hypokalemia before the index prescription were also excluded. Additionally, we excluded patients initiating both medications at index date or who initiated an intravenous levetiracetam. All patients were followed for up to 6 months from the index date until they were withdrawn for reasons of death, loss to follow-up, receiving levetiracetam intravenously, adding or switching (i.e., modification) to a different studied ASM, or until 31 December 2022.

2.3 Exposure Definition

In this study, we adopted an active-control design to minimize confounding by indication bias. We chose carbamazepine as an active comparator given that it has similar clinical use and the absence of documented evidence linking it to hypokalemia [23]. Therefore, we created two mutually exclusive groups: levetiracetam and carbamazepine users. To minimize prevalent-user bias we considered incident users of either drug, which was defined as no recorded dispensing during the 6 months preceding the index date (first dispensing of either study drug).

2.4 Outcome Definition

The study outcome was ascertained through the presence of diagnostic codes or laboratory data, based on clinical guidelines and previous studies [8, 24, 25]. International Classification of Diseases 10th Revision (ICD-10) code E87.6 was used to identify cases of hypokalemia from in- or outpatient data. Serum potassium levels below 3.5 mmol/L were considered as hypokalemia. As a secondary analysis, we specifically focused on moderate to severe hypokalemia (serum potassium levels below 3 mmol/L), as mild hypokalemia typically presents no severe or serious signs and symptoms [24,25,26]. Based on data from case reports and the hypothesized mechanisms, we considered a risk window for developing hypokalemia for up to 6 months after the index date. This exposure window would address the temporal relationship between levetiracetam exposure and occurrence of hypokalemia.

2.5 Covariates and Control for Confounding

We identified an extensive list of covariates to be considered for statistical adjustment during the baseline period (6 months prior to the index date) including age, sex, and history of using drugs that may be associated with hypokalemia, such as loop diuretics, thiazide diuretics, potassium-sparing diuretics, corticosteroids, beta-2 agonists, insulin, antifungal agents (amphotericin B), antibiotics (penicillin, gentamicin), and chemotherapy agents (cisplatin) [8,9,10,11, 27]. Furthermore, the medical conditions that are known to be associated with potassium disturbance, such as diabetes mellitus, chronic kidney disease, cardiovascular disease, gastrointestinal disorders (such as diarrhea or vomiting), alcoholism or drug abuse, magnesium deficiency, and Barter syndrome, were adjusted for [9,10,11]. Additionally, drugs or medical conditions that may interact with levetiracetam or carbamazepine, or impact the patient’s treatment plan, were adjusted for (Table 1) [9,10,11]. These variables were selected based on the available literature and none was assumed to be an instrumental variable. The ICD-10 codes used in identifying these comorbidities are given in Table S1 in the Online Supplementary Material (OSM).

Table 1 Baseline characteristics for levetiracetam and carbamazepine cohorts
Full size table

2.6 Statistical Analysis

To account for confounding factors that may impact the incidence of hypokalemia, we estimated the probability of initiating levetiracetam for each patient regardless of the exposure status based on the aforementioned covariates (propensity score) using a logistic regression model [28, 29]. Following estimation of propensity scores, we adjusted for confounders using inverse probability of treatment weight (IPTW) with stabilization [29, 30]. Performance of propensity score adjustment technique was assessed through examining covariates’ balance after IPTW using the standardized mean difference (SMD), where values more than 0.1 suggest a major difference between the study groups [30].

We used descriptive statistics to summarize the demographic and clinical characteristics of the study population. Hypokalemia unadjusted incidence rate estimation was calculated through dividing the number of identified cases by the cumulative patient-time [28]. To examine the association between levetiracetam and hypokalemia we fitted a weighted Cox regression model to estimate the hazard ratio (HR) of hypokalemia in patients exposed to levetiracetam in comparison to carbamazepine. Proportionality assumption was not violated.

In order to assess the robustness of our findings, we performed several sensitivity analyses. First, we varied the follow-up time to 3 months to address the temporality between initiation of levetiracetam and occurrence of hypokalemia. In the second sensitivity analysis, we focused solely on hypokalemia events as identified through laboratory test, as opposed to the original approach that considered either clinical diagnoses or laboratory criteria. In addition, we accounted for possible cumulative exposure as a sensitivity analysis (see OSM Table S2). All data analyses were conducted using RStudio Version 2023.6.1 Build 524.

3 Results

3.1 Cohort Characteristics

A total of 8,982 patients met the inclusion and exclusion criteria (Fig. 1). Patients exposed to levetiracetam contributed 2273 patient-years to the study, while those receiving carbamazepine had a total exposure time of 955 patient-years. Moreover, 1,404 (15.6%) patients were censored on switching from one medication to another. The baseline characteristics of the two groups were generally comparable, with a few exceptions (Table 1).

Fig. 1
figure 1

levetiracetam vs carbamazepine cohort entry details

Full size image

In comparison to carbamazepine users, levetiracetam-exposed patients had a higher frequency of cerbral ischemia or stroke diagnoses (9.0% vs. 4.3%), and dementia (2.3% vs. 0.6%). Conversely, the carbamazepine-exposed group had a greater frequency of diagnoses related to hyperlipidemia (15.6% vs. 5.8%) and hypertension (12.4% vs. 7.7%). The frequency of loop diuretics (14.8% vs. 8.4%) and insulin (9.6% vs. 6.4%) use was higher in the levetiracetam cohort. On the other hand, the carbamazepine-exposed group had a greater frequency of use of anticoagulants (20.1% vs. 1.0%), statins (21.7% vs. 15.0%), and antiarrythmic (11.8% vs. 6.6%). However, after applying IPTW, the baseline comorbidities and medications were balanced between the exposure groups, and all standardized difference values were below 0.1 (Table 1).

3.2 Hypokalemia Events

During the 6-month follow-up period, 104 cases of hypokalemia occurred, and the majority of cases (80.8%) were ascertained based on serum potassium level, while one patient had both a serum potassium level and an ICD-10 diagnosis of hypokalemia. Among these cases, 56 were mild, 28 were moderate to severe forms of hypokalemia. Moreover, the median time to event for levetiracetam was shorter (25 days) compared to carbamazepine (71 days). The mean age of the cases was 59 years (standard deviation (SD): ± 20.64), and males outnumbered females (52.9% vs. 47.1%). Glucocorticoids (N = 41, 39.4%) and loop diuretics (N = 40, 38.5%) were the most often used concomitant medications at the baseline. The most common comorbidities were cerebral ischemia or chronic stroke (N = 11, 10.6%) and cancer (N = 10, 9.6%).

The incidence rate of hypokalemia was 303 cases per 10,000 patient-years in the levetiracetam-exposed cohort, while it was 57 cases per 10,000 patient-years among carbamazepine users (Table 2). The crude HR of hypokalemia with levetiracetam in comparison to carbamazepine was 1.89 (95% confidence interval (95% CI), 1.05–3.40), while the adjusted HR was 1.99 (95% CI 0.88–4.49). By restricting the definition of hypokalemia cases to moderate or severe hypokalemia we observed a crude HR of 2.03 (95% CI 0.82–5.04), and an adjusted HR of 2.17 (95% CI 0.93–5.03).

Table 2 Association between levetiracetam use and hypokalemia
Full size table

When we shortened the follow-up to 90 days, as a sensitivity analysis, the adjusted HR was 1.99 (95% CI 0.78–5.04), regardless of the hypokalemia severity. Moreover, the adjusted HR for moderate or severe hypokalemia was 2.14 (95% CI 0.62–7.46). The adjusted HR for the sample in which hypokalemia was identified solely by lab results regardless of its severity was 2.41 (95% CI 1.04–5.60). Additional information is available in Table 2 and Fig. 2. After accounting for cumulative exposure, the HRs were consistent with the main analysis (OSM Table S2).

Fig. 2
figure 2

Combined plot of survival probability hypokalemia for levetiracetam and carbamazepine cohorts. Panel A shows the Kaplan Meier (KM) curve for the main analysis (up to 6 months of follow-up). Panel B shows the sensitivity analysis (3 months of follow-up). Panel C shows the secondary outcome (moderate to severe hypokalemia) within 6 months of follow-up. Panel D shows for secondary outcome within 3 months

Full size image

4 Discussion

In this study, we show that the incidence of hypokalemia with levetiracetam use in comparison to carbamazepine was statistically comparable. Although the study yielded insignificant results, the finding of elevated hazard ratios (1.99; 95% CI 0.88–4.49) from the main analysis and from the sensitivity analyses (2.03; 95% CI 1.05–3.40) may suggest that hypokalemia following levetiracetam initiation cannot be ruled out and the insignificant findings might be due to the limited sample size.

This is the first study on levetiracetam-related hypokalemia, therefore, we were unable to compare our findings to other studies. However, our findings align with increasing reports of hypokalemia occurrence following levetiracetam use [9]. Our study population had a comparable sex and age distribution to the cases reported to the WHO [9]. Additionally, we found glucocorticoids to be a common concomitant medication, and was identified as a co-suspected medication in two reported cases [9]. However, lacosamide and proton pump inhibitors appeared as co-suspected medications in several case reports, while in the current study none of the patients had a prescription for these medications [9].

The plausible mechanism underlying this potential association is unknown, and due to the observational nature of our study, the exploration of the mechanism of this association was unattainable. A possible mechanism would be the transcellular potassium shift, which is the transfer of potassium from the extracellular to the intracellular fluid [9, 11, 31]. Metabolic alkalosis might be another mechanism, whereby the intravascular volume contraction inhibits the excretion of the excess bicarbonate in the urine, leading to several symptoms including hypokalemia [9, 11, 32]. Gastrointestinal losses or renal tubular disorder might an additional possible mechanism of levetiracetam-induced hypokalemia [9,10,11, 33].

It is noteworthy that when we restricted the risk (follow-up) period to 3 months, as sensitivity analysis, the incidence rate was higher than what we found in the main analysis, which may suggest that hypokalemia is likely to occur shortly after initiation of levetiracetam. Our findings indicate an excess of 246 cases of hypokalemia per 10,000 patients-years when treating adult patients with levetiracetam in comparison to carbamazepine. This excess risk may warrant consideration for counseling patients about symptoms of hypokalemia and the need to consult with their physicians when these symptoms appear.

A notable strength of this study is the adaptation of an active-comparator incident-user approach to minimize confounding bias. This approach also reduces the potential of time-related biases by restricting the study to treatment initiators. The utilization of a real-world setting provides the study with the capacity to incorporate a representative population, accompanied by a reasonable follow-up period. RERN extended the advantage of incorporating structured EHR data. This structured format empowered the identification of diagnoses through the ICD-10 codes, and the retrieval of serum potassium levels, as an essential parameter in case ascertainment.

However, the study is subject to some limitations. We were not able to capture patient encounters in a facility not integrated with RERN, which may lead to missing documentation of hypokalemia events; therefore, the study might underestimate the true incidence of these events equally in both exposure groups, which may result in non-differential outcome misclassification. An underestimation of hypokalemia risk associated with levetiracetam use could be attributed to our inability to ensure patients’ adherence to their medication. During conceptualization of the study design, we considered that diagnostic codes (ICD-10) may have lower specificity compared to potassium levels available from laboratory data. Therefore, we defined hypokalemia a priori to performing the analysis as either having diagnostic code (ICD-10) or potassium level data to avoid the possibility of underestimating the incidence rate of hypokalemia. Therefore, we considered a sensitivity analysis by restricting the outcome definition based on the laboratory value, which resulted in a statistically significant higher risk of hypokalemia with levetiracetam use in comparison to carbamazepine. Further studies using different a data source (i.e., study population) may provide more evidence on the association between levetiracetam and hypokalemia by addressing the sample size and generalizability limitations of this study.

5 Conclusion

In this study, the hazard ratio of hypokalemia between levetiracetam and carbamazepine was statistically comparable. We cannot rule out the possible association between levetiracetam use and hypokalemia given that the main and several sensitivity analyses were suggestive of elevated hazard ratios. Thus, further studies may provide a more precise assessment of this association.