Evaluation of Free Valproate Concentration in Critically Ill Patients

OBJECTIVES:

Protein binding of valproate is variable in ICU patients, and the total valproate concentration does not predict the free valproate concentration, even when correcting for albumin. We sought to quantify valproate free concentration among ICU patients, identify risk factors associated with an increasing free valproate concentration, and evaluate the association between free valproate concentration with potential adverse drug effect.

DESIGN:

Retrospective multicenter cohort study.

SETTING:

Two academic medical centers.

PATIENTS:

Patients greater than or equal to 18 years of age with concomitant free and total valproate concentrations collected in the ICU.

INTERVENTIONS:

None.

MEASUREMENTS AND MAIN RESULTS:

Two-hundred fifty-six patients were included in the study, with a median age of 56 years (42–70) and 65% of patients were male. The median total valproate concentration was 53 µg/mL (38–70 µg/mL), the free valproate concentration was 12 µg/mL (7–20 µg/mL), and the free fraction was 23.6% (17.0–33.9%). Therapeutic discordance between the free and total valproate concentration occurred in 70% of patients. On multivariable analysis, increased free valproate concentration was associated with higher total valproate concentration (per 5 µg/mL increase, increase 1.72 µg/mL, 95% CI, 1.48–1.96) and lower serum albumin (per 1 g/dL decrease, increase 4.60 µg/mL, 95% CI, 2.71–6.49). There was no association between free valproate concentration and adverse effects.

CONCLUSIONS:

The valproate total and free concentration was discordant in the majority of patients (70%). Increased valproate free concentration was associated with hypoalbuminemia and total valproate concentration. Clinical decisions based on total valproate concentration may be incorrect for many ICU patients. Prospective, controlled studies are needed to confirm these findings and their clinical relevance.

Valproate is an anti-seizure drug and mood stabilizer with a narrow therapeutic index that often requires total valproate serum concentration monitoring (1, 2). The validity of measuring total valproate concentration in the critically ill has been raised due to the presence of numerous risk factors for altered protein binding, and in turn, this has led to increased interest in free valproate serum concentration monitoring (3).

Valproate is highly protein bound to albumin, and its pharmacologically active free fraction is usually 5–10% (1). Although many factors have been associated with alterations in valproate protein binding including high total valproate concentrations, diurnal variations, hypoalbuminemia, drug-drug interactions (e.g., aspirin or ibuprofen), free fatty acid-containing drugs (e.g., IV lipid emulsion, propofol, and clevidipine), and uremia, the value of monitoring free serum concentrations in acutely ill patients has long been controversial (3–5). Early studies evaluating free valproate concentrations focused on outpatients. This population population few of the many factors that can increase valproate free fractions and the therapeutic discordance between total and free concentrations and which are very often present in critically ill patients (6, 7).

Elevated free valproate concentrations have been associated with hyperammonemia (8), thrombocytopenia (9), and neurotoxicity (10). Increasing indications for valproate administration in critically ill for treatment of seizures and behavioral control highlight the need to clarify if there is an association between valproate's free fraction or concentration and therapeutic effectiveness or adverse drug effects (11, 12).

The objectives of this study were to quantify free valproate concentration in critically ill patients, identify risk factors associated with an increasing free valproate concentration, and to evaluate the association between valproate's free fraction and concentration with potential adverse drug effects.

MATERIALS AND METHODS

Study Design

This retrospective cohort study was conducted at Maine Medical Center (MMC) in Portland, Maine, United States, and Mayo Clinic in Rochester, Minnesota, United States. Both institutions are tertiary, university-affiliated medical centers, with 66 and greater than 200 adult intensive care unit (ICU) beds, respectively. The protocol was approved by the Institutional Review Boards (IRBs) at Mayo Clinic (IRB number 18-011183) on February 14, 2019, and MMC (IRB number 1468271-1) on August 23, 2019. The IRBs provided a waiver of informed consent, and all study procedures were in accordance with the IRBs and the Helsinki Declaration of 1975.

Patient Population

Consecutive patients greater than or equal to 18 years were included if they had simultaneous free and total valproate concentrations collected during their ICU stay. Mayo Clinic patients were included between January 1, 2014, and December 31, 2018. MMC patients were included from September 1, 2015, to December 31, 2018 (no free valproate levels were monitored before 2015 at MMC). Patients were excluded if they had total or free valproate levels less than 3 µg/mL and if they did not give consent for research authorization in the state of Minnesota.

Patient Demographics

Patient demographics included age, gender, race, weight, hospital length of stay, Charlson Comorbidity Index, Acute Physiology and Chronic Health Evaluation (APACHE) scores, discharge disposition, and indication for valproate therapy. Indication for valproate therapy was manually collected at MMC by chart review. International Classification of Diseases, 10th Revision codes for seizures and/or behavioral control were used to classify patients at Mayo Clinic; however, these indications were not manually confirmed for patients’ ICU admission and patients could have been receiving valproate for an additional indication.

Valproate Serum Concentrations

At Mayo Clinic, total and free valproate concentrations were analyzed using ultrafiltration and a homogeneous enzyme immunoassay (Roche valproic reagent; Roche Diagnostic Corp, Indianapolis, IN) by Mayo Clinic Laboratories, Rochester, Minnesota, United States. At MMC, free valproate concentrations were sent to Mayo Clinic in Rochester, Minnesota, United States, from September 2015 to October 2017. After that date, all total and free valproate concentrations were run in-house at MMC using the same process described above. Monitoring of valproate concentrations was not protocolized.

Medications

Concomitant medications interacting with valproate were collected from the electronic medication administration record including aspirin, ibuprofen, ketorolac, propofol, clevidipine, and IV fat emulsion. Medications were considered concomitant if the patient received them during the 24 hours preceding free valproate concentration measurement.

Laboratory Data

Serum albumin, serum creatinine, and blood urea nitrogen (BUN) values were extracted within 72 hours of collecting the free valproate concentration. Total bilirubin, platelets, ammonia, lipase, alanine aminotransferase (ALT), and alkaline phosphate were extracted within 72 hours of collecting the free valproate concentration and for 72 hours after valproate therapy was terminated, until hospital discharge, or death, whichever occurred first, to assess for toxicity. Monitoring of these values was not protocolized.

Adverse Drug Effects

Hepatic dysfunction was defined as a new ALT greater than three times the upper limit of normal (ULN), alkaline phosphate greater than two times the ULN, total bilirubin greater than two times the ULN, or a doubling of the baseline value of these results if they were abnormal prior to valproate initiation (11). Thrombocytopenia was defined as a new decrease in platelet count less than 140,000 cells/mm³ or decrease in platelet count by greater than 50% if the platelet count was lower than 140,000 cells/mm³ before valproate initiation (11). Hyperammonemia was defined as a serum ammonia level of greater than 60 mmol/L not present prior to valproate initiation (11). Monitoring of suspected adverse drug effects was not protocolized.

Statistical Analysis

Continuous data are reported as median (interquartile range [IQR]) and frequencies as n (%). Free fraction of valproate (%) was calculated by dividing the measured free valproate concentration by the simultaneously measured total valproate concentration and then multiplying the quotient by 100. Only the first ICU admission and first free and total valproate concentrations per patient were analyzed.

Total and free valproate concentrations were categorized relative to their published reference ranges of 50–125 µg/mL and 5–15 µg/mL, respectively (1). Concentrations below the lower limit of the reference range (total < 50 µg/mL and free < 5 µg/mL) were categorized as “low” and those above the upper limit (total > 125 µg/mL and free > 15 µg/mL) as “high.” If the total and free valproate concentrations were in the same category (e.g., low total concentration and low free concentration), they were therapeutically concordant, and if they were in different categories (e.g., low total concentration and high free concentration), they were therapeutically discordant (3).

Univariate linear regression analysis evaluated variables associated with an increasing free valproate concentration as a continuous variable. Potentially relevant variables were determined a priori and entered into a multivariable linear regression model: total valproate concentration, serum albumin, BUN, propofol exposure (yes or no), and aspirin exposure (yes or no). Interactions between selected variables were assessed; however, none were found.

Logistic regression was used to determine if valproate free concentration was associated with hepatic dysfunction, thrombocytopenia and hyperammonemia. A free valproate concentration of greater than 15 µg/mL was selected as the upper end of the reference range described in the literature (3). A p value of less than 0.05 was considered statistically significant. All analyses were performed using SAS Version 9.4 (SAS Institute, Cary, NC).

RESULTS

Patient Population and Demographics

A total of 256 patients were included in the study (Supplemental Table 1, http://links.lww.com/CCX/B46). The median age was 56 years (42–70 yr), 65% were males, and the majority were Caucasian (87%). Most patients received valproate for seizures (66%), followed by behavioral control (7%), or a combination of both (19%). The median APACHE III score for patients from Mayo Clinic was 62 (47–79), and the median APACHE IV score for MMC patients was 53.6 (30.8–79.7). Most patients (55.9%) were in the Neurosciences ICU when valproate concentrations were collected. There were 50 patients (19.5%) who received concomitant aspirin and 41 (16%) who received propofol within the 24 hours proceeding valproate levels. Of the 50 who received aspirin, 43 (86%) received 81 mg and 7 (14%) received 325 mg. Median serum albumin was 3.2 g/dL (2.6–3.6 g/dL) and hospital length of stay was 12 days (7–25 d) (Supplemental Table 1, http://links.lww.com/CCX/B46). Only 115 of 256 patients (45%) who had both serum albumin and BUN values at the time of valproate concentration measurement could be included in the multivariable analysis.

Valproate Serum Concentrations

The median (IQR) total valproate concentration was 53 µg/mL (38–70 µg/mL) and the free valproate concentration was 12 (7–20) (Fig. 1). There were 93 patients (36.3%) with a free valproate concentration greater than 15 µg/mL. Total and free valproate concentrations were therapeutically concordant in 78 patients (30%) and discordant in 178 patients (70%); when discordant, free valproate concentrations were consistently higher than expected based on the total valproate concentration. Of the 70% of patients who were therapeutically discordant, 101 (57%) were due to a “low” total valproate concentration and a “high” or within goal free valproate concentration, while 75 (42%) were due to a total valproate concentration within goal and a “high” free valproate concentration suggesting total valproate concentration underestimated the free concentration among ICU patients.

Figure 1.

Total valproate (VPA) concentration versus free VPA concentration. Total VPA concentration (in µg/mL) is shown along the y-axis with corresponding free VPA concentration (in µg/mL) along the x-axis.

In the univariate analysis, for every 5 mcg/mL increase in total valproate concentration, free concentration increased 2.54 µg/mL (95% CI, 2.41–2.67 µg/mL) (Supplemental Table 2, http://links.lww.com/CCX/B46). The multivariable analysis was conducted on 115 of 256 patients (45%) who had both serum albumin and BUN values at the time of valproate concentration measurement. In the multivariable analysis for every 1 g/dL decrease in serum albumin, the free valproate concentration was estimated to increase 4.60 µg/mL (95% CI, 2.71–6.49 µg/mL) (Table 1). Additionally, for every 5 mcg/mL increase in total valproate concentration, free valproate concentration was estimated to increase 1.72 µg/mL (95% CI, 1.48–1.96 µg/mL) (Table 1).

TABLE 1.

Multivariable Analysis of Variables Associated With an Increasing Valproate Free Concentration

Variable	Estimate (95% CI)	p
Albumin (per 1 g/dL increase)	–4.60 (–6.49 to –2.70)	< 0.001
Total valproate (per 5 mcg/mL increase)	1.72 (1.49 to 1.96)	< 0.001
Blood urea nitrogen (per 1 mg/dL increase)	0.02 (–0.06 to 0.10)	0.59
Propofol (yes)	2.14 (–0.99 to 5.27)	0.18
Aspirin (yes)	1.51 (–1.69 to 4.71)	0.36

Analyzed in 115 of 256 patients (45%) who had both serum albumin and blood urea nitrogen values at the time of valproate concentration measurement.

Adverse Effects

There were 18 of 221 patients (8.1%) who developed hepatic dysfunction, 69 of 255 (27.1%) had thrombocytopenia, and 27 of 128 (21.1%) had hyperammonemia. No association between free valproate concentration and hepatic dysfunction, thrombocytopenia, or hyperammonemia was observed (Table 2).

TABLE 2.

Adverse Drug Effects in Relation to Valproate Free Fraction and Free Serum Concentration

Hepatotoxicity	OR (95% CI)	p
Free concentration (per 1 mg/dL increase)	1.00 (0.99–1.02)	0.94
Free concentration (> 15 vs ≤ 15 mg/dL)	0.53 (0.18–1.55)	0.25
Thrombocytopenia	OR (95% CI)	p
Free concentration (per 1 mg/dL increase)	1.00 (0.98–1.01)	0.58
Free concentration (> 15 vs ≤ 15 mg/dL)	0.92 (0.52–1.62)	0.76
Hyperammonemia	OR (95% CI)	p
Free concentration (per 1 mg/dL increase)	1.00 (0.99–1.01)	0.96
Free concentration (> 15 vs ≤ 15 mg/dL)	1.91 (0.81–4.49)	0.14

OR = odds ratio.

DISCUSSION

This study quantified valproate's free concentration during critical illness, risk factors for increasing valproate free concentration, and the relationship between valproate free fraction and concentration with potential adverse drug effects. The median free valproate concentration was 12 µg/mL (7–20 µg/mL), and 36.3% patients had a free concentration greater than 15 µg/mL. Total valproate concentrations underestimated the free valproate concentrations in 70% of patients indicating that total valproate levels are a poor surrogate for the biologically active free valproate concentration during critical illness. Increasing total valproate concentration was a predictor for elevated free valproate concentration. This is consistent with a previous study in both inpatients and outpatients showing a relationship between increasing total valproate concentrations and increasing free concentrations (13). Hypoalbuminemia was also a predictor of an increasing free valproate concentration, consistent with a prior study of 257 inpatients and outpatients reporting a similar relationship in patients with a serum albumin concentration less than 3.5 g/dL (14).

Uremia may also alter valproate's protein binding through alterations in the physiochemical characteristics of albumin or competitive inhibition due to uremic toxins (15). In a study of 24 patients with nondialysis dependent chronic kidney disease (mean BUN of 40 mg/dL and serum creatinine 2.6 mg/dL), valproate's free fraction correlated with creatinine (r = 0.65) and BUN (r = 0.64), although multivariate regression analysis was not performed (16). Similar to the report by Gibbs et al (14), our study (with a median BUN of 17 mg/dL and creatinine of 0.9 mg/dL) did not demonstrate an association between BUN or creatinine and free valproate concentration. The relationship between increasing free valproate concentration and the degree of renal impairment remains uncertain.

Drug-drug interactions can also alter valproate protein binding. Antipyretic doses of aspirin are known to result in a four-fold increase in valproate's free fraction, and similar effects have been seen with lower cardioprotective doses (17, 18). Ibuprofen has been shown to produce a similar effect (18). In our study, no relationship between aspirin, ibuprofen or ketorolac was observed, but the number of patients receiving nonaspirin NSAIDs was low.

Valproate is an 8-carbon branched-chain carboxylic acid that is structurally similar to free fatty acids that have high affinity for albumin and may displace highly protein bound medications. Stearic, palmitic, oleic, and linoleic acid, which represent 80% of free fatty acids in humans, can increase valproate's free fraction in a concentration-dependent manner by 19–118% (16). Although our study suggests that propofol administration is not associated with an elevated valproate free, only 16% of patients received propofol, and therefore the analysis may have been underpowered.

Several adverse drug effects have been reported during valproate therapy, including hepatotoxicity, which may occur in 5–10% of valproate-treated patients. We found no relationship between an increasing free valproate concentration and hepatic dysfunction. Thrombocytopenia occurs in 5–18% of valproate-treated patients and is potentially due to bone marrow suppression or autoantibody production (11). In a study of 98 patients by Tseng et al (9), including 47 (48%) in an ICU, free valproate concentration was associated with thrombocytopenia (1.10 [1.01–1.19]) and a concentration greater than 14.7 µg/mL had strong discriminating power for this association (area under the receiver operating characteristic curve, 0.77; 95% CI, 0.66–0.88). In our study, free valproate concentration greater than 15 µg/mL was not associated with thrombocytopenia.

Hyperammonemia during valproate therapy may result from direct inhibition of carbamoyl phosphate synthetase I by valproate's metabolites, including valproyl-CoA, but this mechanism requires confirmation (1). An early study of 19 patients by Itoh et al (8) found a moderate association between free valproate concentration and ammonia concentration. These results were corroborated by Tseng et al (9) who found that while hyperammonemia occurred in 32% of their patients, it was not associated with free valproate concentration, consistent with our results. Our data are consistent with this finding (8); however, it is important to highlight that since we evaluated the first valproate concentration measured during a patients’ ICU stay hyperammonemia that occurred later during admission may have been missed.

There are several limitations to this study. Since we evaluated the initial concentration drawn during patients’ ICU stay, we were unable to characterize valproate steady state conditions in our population. Steady state levels could impact protein binding and clearance of valproate, so our results should be interpreted with caution (19). We did not evaluate valproate effectiveness for the clinical indication and cannot comment about free valproate concentration and efficacy. The multivariable model for an association between clinical variables and an increasing valproate free concentration included only 45% of patients because the majority of patients had missing albumin values. The very low number of patients receiving IV fat emulsion (n = 1) hampered our ability to assess any impact on valproate free fraction. No protocol for monitoring free valproate concentration or toxicity existed at the time of the study, which may have introduced selection bias. We examined a limited number of previously described adverse drug effects and other toxicities may have occurred. We did not control for other variables in critically ill patients (e.g. sepsis, heparin use, dialysis) that may have led to toxicities. We also did not collect potential drug interactions (most notably phenytoin), which should be further explored. Lastly, we did not characterize the timing of concentrations (trough or peak). Despite these limitations, our study represents the largest cohort of critically ill patients with reported valproate free fractions. Future studies should address the important limitations from our study to better inform the influence of free valproate concentrations on valproate efficacy and toxicity. The discordance between total and free valproate concentrations highlights the need to monitor free valproate concentrations in critically ill patients and expand the availability of this technology in more healthcare settings.

CONCLUSIONS

This study emphasizes the potential importance of being able to routinely measure and monitor free valproate concentrations in critically ill patients since free concentrations are often discordant with total concentrations, especially when albumin is low. For these patients, total valproate concentrations are poor surrogates for biologically active free valproate and may lead to inappropriate therapeutic decisions. Prospective, controlled studies are needed to confirm these findings.