Population pharmacokinetics of intramuscular droperidol in acutely agitated patients

Abstract

Background

Intramuscular droperidol is used increasingly for sedation of aggressive and violent patients. This study aimed to characterise the pharmacokinetics of intramuscular droperidol in these patients to determine how rapidly it is absorbed and the expected duration of measurable drug concentrations.

Methods

We undertook a population pharmacokinetic analysis of a subgroup of patients from a clinical trial comparing droperidol and midazolam: 17 receiving 5 mg and 24 receiving 10 mg droperidol. Droperidol was measured using high‐performance liquid chromatography. Pharmacokinetic modelling was performed under a nonlinear mixed effects modelling framework (NONMEM v7.2). The model was used to simulate concentration time profiles of three typical doses, 5 mg, 10 mg and 10 mg + 10 mg repeated at 15 min.

Results

A two‐compartment first‐order input with first‐order output model fitted the data best. The absorption rate constant was poorly characterised by the data and an estimate of the first order rate constant of absorption when fixed to 10 h^–1 provided a stable model and lowest objective function. This represents extremely rapid absorption with a half‐life of 5 min. The final model had a clearance of 41.9 l h^–1 and volume of distribution of the central compartment of, 73.6 l. Median and interquartile range of initial (alpha) half‐life was 0.32 h (0.26–0.37 h) and second (beta) half‐life was 3.0 h (2.5–3.6 h). Simulations indicate that 10 mg alone provides an 80% probability of being above the lower limit of quantification (5 μg l^–1) for 7 h, 2 h longer than for 5 mg. Giving two 10 mg doses increased this duration to 10 h.

Conclusions

Intramuscular droperidol is rapidly absorbed with high therapeutic concentrations after 5 and 10 mg doses, and supports clinical data in which droperidol sedates rapidly for up to 6 h.

What is Already Known about this Subject

• Intramuscular droperidol is a safe and effective for sedation of aggressive patients

• The pharmacokinetics of intramuscular droperidol is poorly characterised.

What this Study Adds

• Droperidol is rapidly absorbed when given by the intramuscular route.

• Intramuscular droperidol with doses of 5 mg and 10 mg provides good exposure

• Drug concentrations remain above the lower limit of detection for at least 7 h.

Table of Links

LIGANDS
Droperidol

This Table lists key ligands in this article that are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 1.

Introduction

Violence and aggression in hospitalised patients is a growing problem faced by emergency departments and psychiatric units 2, 3, 4, 5, 6, 7. This type of patient behaviour poses significant risks for patients and staff and can cause personal injury or damage to hospital property. Rapid sedation of these agitated patients with suitable medication is required in order to decrease dangerous behaviour and allow necessary diagnosis, investigation and treatment 3. Intravenous (IV) sedation can be difficult in such conditions because it requires sufficient staff to restrain the agitated patient in order to obtain IV access. This may result in needle stick injuries or failed attempts at gaining IV access. Intramuscular (IM) administration of sedative drugs, benzodiazepines or antipsychotics, has become the preferred route of administration for these acutely agitated patients in emergency departments 2. However, there has always been concerns that there will be a delayed onset of effect with IM administration because the medication has to be absorbed first.

Droperidol is a high‐potency butyrophenone that has been used as a sedative agent for decades with great success 8, 9, 10. Despite this, there is limited information on the pharmacokinetics of droperidol via the IM route and in patient sedation. There are a number of studies of IV administration, mainly for anaesthesia 11, 12, 13. There is one small study of male volunteers given 5 mg droperidol by the IV and IM routes, which suggests there is rapid absorption and biexponential decay in drug concentrations after this 14. However, there were a number of limitations of this study including the use of only men, and it not being used in patients. A more recent study of IM versus IV midazolam showed that IM administration is almost as rapid as IV 15.

A black box warning against droperidol in 2001 resulted in a sharp decline in its use despite a long history of safe use 9. In the last 10 years there have been several studies demonstrating the effectiveness of droperidol and its safety in large numbers of patients 4, 5, 6, 7, 16, 17, 18. In Australia, droperidol has remained available and in the last 5 years there has been an increase in its use. However, the higher concentration (5 mg ml^–1) remains a restricted drug because of limited information on efficacy, safety and the pharmacokinetics. A number of randomised controlled trials have demonstrated its effectiveness 4, 5, 6, 7, 18 and a recent large study of over 1000 cases supports its safety and the negligible risk of QT prolongation and torsade des pointes 16.

The aim of this study is to characterise the population pharmacokinetics of IM droperidol in patients with severe agitation, to determine how rapidly it is absorbed and the expected duration of measurable drug concentrations to inform future pharmacodynamics studies.

Methods

Study design and setting

This study was a population pharmacokinetic analysis of a subgroup of patients recruited to a blinded randomised controlled trial comparing droperidol vs. midazolam for the sedation of acute behavioural disturbance 7. There were three arms in the trial: 10 mg IM midazolam, 10 mg IM droperidol, and a combination of 5 mg IM midazolam with 5 mg IM droperidol with a 1:1:1 randomisation to each arm. The clinical outcomes for the study have previously been published 7 and here we include 41 patients from the two arms of the trial receiving 5 mg and 10 mg of droperidol.

Ethics approval was obtained from the Hunter and New England Human Research Ethics Committee. Patient consent was waived for both the randomised controlled trial and blood collection because patients required immediate treatment as a duty of care and were unable to consent to a research study in this setting. The clinical trial was registered with the Australian Clinical Trial Registry ‐ ACTRN12607000527460.

Patients

The inclusion criteria for the clinical trial were patients (age 18 years or older) who presented to the emergency department with acute behavioural disturbance and required physical restraint and parenteral sedation. Patients were excluded if they could be successfully verbally de‐escalated or they agreed to oral or IV sedation. For this study, patients were included if it was logistically possible and safe to collect serial blood samples from the patients. However, no patients were completely excluded for safety, but did not have early samples collected. One 16‐year‐old was recruited inadvertently because their age was unknown when they were sedated due to their agitation.

Assay

Serum samples were assayed using high‐performance liquid chromatography (HPLC) with ultraviolet detection (Shimadzu). The system consisted of a LC‐10AT solvent delivery system, SCL‐10AVP system controller and SPD‐M10 A Diode Array Detector. The mobile phase consisted of 220 ml of HPLC grade acetonitrile diluted to 1000 ml with 0.03 mol l^–1 potassium phosphate buffer adjusted to pH 3.5. The mobile phase was pumped through an Agilent SB‐C8 (Zorbax 3.5 μm, 4.6 × 75 mm) column at a flow rate of 0.9 ml min^–1. The detector was set at 205 nm.

To prepare each sample for analysis, 600 μl serum was mixed with 400 μl of 0.03 mol l^–1 potassium phosphate buffer (pH 7.4). To this was added 40 μl of a 10 ng μl^–1 haloperidol solution (internal standard). The sample was mixed and then placed on an Alltech Prevail 1 ml C18 extraction cartridge that had been preconditioned with acetonitrile and water. After the sample was drawn slowly through the cartridge under vacuum, the cartridge was washed with 1 ml of water, 600 μl of water adjusted to pH 3.7, 600 μl of a wash consisting of 25 ml acetonitrile and 10 ml isopropanol in 250 ml potassium phosphate buffer adjusted to pH 3.6. The sample was then eluted using 350 μl of a mix of 40% acetonitrile in 0.03 mol l^–1 phosphate buffer (pH 3.7). The eluted sample was evaporated at room temperature under a stream of air using a Techne DB evaporator. After redissolving the sample in 130 μl of mobile phase, 70 μl of this was injected in to the HPLC system. Quantitation was performed using a standard curve ranging from 5 to 200 μg l^–1. Between and within day coefficients of variation were <10% at 10 μg l^–1 and 75 μg l^–1. The lower limit of quantification was 5 μgl^–1.

Sampling design

At the commencement of data collection for the pharmacokinetics study samples were collected according to an empirical design using a standard geometrically spaced sampling sequence ‐ 5 min, 10 min, 30 min, then 1 h, 2 h, 4 h, and 8 h, where possible. During the initial phase of the clinical trial an optimal design study was undertaken using POPT 19. to determine the optimal sampling windows for sample collection based on a previous pharmacokinetic study of droperidol 14. Sampling windows were defined as a region in which the efficiency of the design was at least 90% of the optimal design. The optimal sampling windows were adapted to meet the practicalities of collection for patients in this setting, including the difficulties in collecting multiple samples in the first 20 min and that many patients were discharged within 6 h. In the second part of the study, these sampling windows were implemented and collection was based on this modified optimal timing at 5 min, 25 min, 40 min, 70 min, 2 h, 4 h and 10 h.

Model building

Pharmacokinetic modelling was performed under a nonlinear mixed effects modelling framework using NONMEM ver 7.2 software 20 with GFortran 2007 compiler. Models were selected based on likelihood ratio tests where the difference in two objective functions of NONMEM (proportional to twice the negative log‐likelihood) was assumed to be approximately and asymptotically Chi‐square distributed. A difference of more than 3.84 was considered significant at α = 0.05. Coingestion of alcohol was considered as a covariate, but weight was not available.

Handling data below the limit of quantitation

In order to accommodate for the existence of the four data points that were below the lower limit of quantitation (LLOQ), the M6 method in NONMEM was implemented for model fitting 21. In this method the first data item (concentration) of a series that was below the limit of quantification was set to half of the LLOQ and the remainder were commented out. We note that four items occurred in four different subjects and hence there were no more than a single value in each setting.

Outliers

Outliers were defined as those observations that were more than 6 standard deviations away from the model prediction (conditional weighted residuals >6). The analysis was repeated in the presence and absence of these observations and if differences in the parameter estimates were >10% between the runs and no explanation could be identified that could account for the error (e.g. incorrect sample time) then the observation was left out of the analysis.

Model evaluation

The model was evaluated using standard goodness of fit plots. In addition, a visual predictive check of the model was performed. In this method the data are simulated 1000 times and the quantiles of the observed data are plotted against the quantiles of the simulated data and both plotted against time. A good agreement is shown when the quantiles of the observed data match the quantiles of the simulated data. In this case the 95% confidence interval of the quantiles of the simulated data are provided to aid comparison.

Simulations

The final model was then used to simulate the concentration time profile of three typical doses used in the emergency department, (1) a single dose of 5 mg, (2) a single dose of 10 mg, and (3) a dose of 10 mg repeated at 15 min. In each case 1000 individuals were simulated and the median and 2.5^th and 97.5^th percentiles reported. In addition, as a guide to likely clinical activity the probability of being above the LLOQ (5 μg l^–1) was plotted versus time. This was computed as the fraction of simulated observations that were above LLOQ at each time point. We are not aware of any data on the minimum concentration of droperidol that has activity (see Schulz and Schmoldt 22 who do not provide a lower value of the typical therapeutic range) or the concentration for which there is 50% effect (EC₅₀). This meant that we had to use the LLOQ which provides an initial basis for comparison of dosing regimens.

Results

Patient data

There were 41 patients recruited to the clinical trial where blood samples were collected, 27 were female and the median age was 33 years (16–62 years). Patient characteristics and admission information are provided in Table 1. Twenty‐four patients received 10 mg droperidol and 17 received 5 mg of droperidol as part of the combination of droperidol and midazolam. Five of the 41 patients were given additional sedation with droperidol and 90% were sedated within 36 min. There were 128 samples available for analysis from the 41 patients with a median of three samples per patient (1–7). Four observations were below the lower limit of quantitation of 5 μg l^–1. The median droperidol concentration was 41 μg l^–1 (Range: below LLOQ to 215 μg l^–1). Only one observation was identified as an outlier and this was found not to be influential.

Table 1.

Demographics and clinical information on 41 patients who were administered droperidol for acute behavioural disturbance

	Median (range)	Number (%)
Age (years)	33 (18 to 62)	41 (100)
Male	–	14 (34%)
Primary reason for presentation:
Threatened or deliberate self‐harm		18 (44%)
Alcohol intoxication		17 (41%)
Drug‐induced delirium		4 (10%)
Psychosis		2 (5%)
Urine drug screen positive		9 (22%)
Breath alcohol levels (mg dl –1 )	0.24 (0.01–0.39)	32
Droperidol dose
5 mg		17 (41%)
10 mg		19 (46%)
>10 mg		5 (12%)
Duration of acute behavioural disturbance (min)	20 (8–90)	41
Time to sedation (min)	15 (5–105)	41
Adverse effects:
Hypotension		1 (2%)
Desaturation (O 2 Sats < 90%)		4 (10%)

Various compartmental models were evaluated and a two‐compartment first‐order input with first‐order output model with combined error model was found to provide the best fit to the data. However, it was noted that the estimate of value of the first order rate constant of absorption (ka) varied widely across individual's (1–140 h^–1) without any clear pattern. A number of different absorption processes were considered, including zero‐order, sequential zero‐order/first‐order and a mixture model on first‐order. All were unstable and did not yield better fits to the data. The k _a was therefore fixed and a sensitivity analysis performed to this value. A value of 10 h^–1 and a value of between subject variance of 1 (approximately equivalent to a between subject CV% of 100%) were both stable and provided the lowest value of the objective function. Note that this represents an extremely rapid absorption profile with a half‐life of 5 min. The individual values of k_a with this model varied from 2.27 h^–1 (half‐life of absorption 20 min) to 40 h^–1 (half‐life of 1 min).

The model did not support estimation of random effects for intercompartmental clearance (Q) and V_p. When covariance was estimated between clearance (CL) and volume of distribution of the central compartment (V_c) the correlation was estimated to be 100%. Fixing the variance of V_c provided a worse prediction of the data and allowing V_c to covary with CL provided the best overall fit. Visual inspection of plots of blood alcohol level versus CL and V_c found no association. The final model is given in Table 2. The median and interquartile range of the initial half‐life (alpha phase) was 0.32 (0.26–0.37) h and the median and interquartile range for the second (beta phase) half‐life was 3.0 (2.5–3.6) h.

Table 2.

Final model and parameter estimates

	Parameter estimate (95% CI)	Between subject variability CV% (95% CI)
CL (l h –1 )	41.9 (34.8–49.0)	51% (31.2–64.4%)
Vp (l)	73.6 (51.1–96.1)	51% a (31.2–64.4%)
k a (l –1 )	10 F (−)	100% (−)
Q (l h –1 )	71.5 (42.3–100.7)	–
Vp (l)	79.8 (58.8–100.8)	–
σ (CV%)	22% (8.5–30.3%)
σ add (μg  l –1 )	0.0001 F

CI, confidence interval; CV, coefficient of variance; CL, clearance; V_c, volume of distribution of the central compartment; k_a, the first order rate constant of absorption; Q, intercompartmental clearance; Vp, volume of distribution of the peripheral compartment; F, fixed

^a The same random effect was used for both Vc and CL

Model evaluation

A visual predictive check plot of the final model is presented in Figure 1. The 50^th, 5^th and 95^th percentiles of the model predictions and data are shown. The shaded regions represent the 95% CI of the percentiles of the model predictions which encompass the observed percentiles. The figure indicates that the model provides an adequate description of the data.

Figure 1.

Visual predictive check plot of the median and the 5^th and 95^th percentile of the observed and simulated droperidol concentration‐time curve for the full time course (A) and for the 1^st h (B). The shadowed regions represent the 95% confidence interval of the model predictions

Simulations

Simulations from the model indicate that a dose of 10 mg provides an 80% probability of being over the lower limit of quantification for about 7 h after the dose is administered. This was approximately 2 h longer than for the 5 mg dose (circa the beta‐phase half‐life). Giving two 10 mg doses within 15 min increased this time to 10 h.

Discussion

This study has shown that droperidol is rapidly absorbed when given by the IM route. Using this route of administration it is possible to get high therapeutic droperidol concentrations after both the 5 mg and 10 mg doses used in this study. Importantly, drug concentrations remain above the lower limit of detection for at least 7 h (probability = 0.8) after a 10 mg dose was administered.

The therapeutic range of droperidol has not been clearly defined and it is likely to be different for the various indications of droperidol, including low dose (0.25–1.5 mg) as an anti‐emetic 23, sedation (5–20 mg) 7, 16 and anaesthesia (10–200 mg h^–1) 24. Schulz and Schmoldt report therapeutic concentration up to 50 μg l^–1 22. Sawyer et al. report a mean concentration of 384 μg l^–1 in cardiac bypass surgery, demonstrating that concentrations much higher than 50 μg l^–1 occur in the anaesthetic setting 24. In our study the range of maximum observed concentrations were intermediate to literature findings, with values ranging from 22.4 μg l^–1 to 215 μg l^–1 with the peak concentration occurring within 10 min (Figure 1). The model predicted maximum values that ranged from 20 μg l^–1 to 110 μg l^–1 (95^th percentile for the 5 mg dose), 25 μg l^–1 to 220 μg l^–1 (95^th percentile for the 10 mg dose) and 50 μg l^–1 to 280 μg l^–1 (95^th percentile for the repeated 10 mg dose). Defining a therapeutic range in our patient group (sedating aggressive patients) is not possible without pharmacodynamic outcomes. The concentrations in Figure 2 show that for the currently recommended dose of 10 mg the drug concentrations remain above the LLOQ for over 6 h in almost all patients and is well above the LLOQ for 50% of the patients.

Figure 2.

The upper panels represent the dosing scenarios: (1) a single 5 mg dose, (2) a single 10 mg dose, (3) a 10 mg dose repeated at 15 min. The lower panel represents the probability of the concentration being greater than the lower limit of quantitation for each dosing regimen over 12 h

A major concern with droperidol has been whether its association with QT prolongation, particularly at the doses used in our study. A more recent study of over 1000 patients found that QT prolongation was rare 16 and appeared to be no more common than in a population of patients not taking QT prolonging drugs 25. Electrocardiogram data from the same patient group used for this pharmacokinetic study, found that droperidol was not associated with QT prolongation and patients had QT values that were the same as those in the group not administered droperidol (i.e. patients given only midazolam) 7. This suggests that the droperidol concentrations measured in these patients are unlikely to be associated with QT prolongation, and it would not be possible to undertake a pharmacokinetic–pharmacodynamic analysis of the QT data because of the lack of abnormal QT intervals.

There were a number of limitations to the study, including patient weight not being available and that few samples were available in the absorption and early distributional phase. Weight is usually an important covariate in pharmacokinetic studies and inclusion would allow a dose adjusted for weight to be used. However, just as it was very difficult to obtain the weight of the patients in this study due to their agitated and aggressive behaviour, it would be similarly difficult to obtain the weight to determine a dose in a future clinical setting. Therefore, from a practical point of view using a fixed dosing regimen was needed in this therapeutic setting. The simulation demonstrates that the concentrations are within a reasonable range (peak: 25–225 μg l^–1; 6 h: >5–50 μg l^–1) for a single dose of 10 mg and that these concentrations are likely to be sufficiently high for 5–7 h postdose. Doubling the dose from 5 mg to 10 mg increased the duration of high concentrations by approximately 2 h. Finally, none of the patients experienced significant side‐effects 7.

The lack of samples collected in the first 10 min postadministration of droperidol explains the difficulty in estimating k_a in this study and may also potentially affect estimation of the intercompartmental clearance (or re‐parameterised as the distributional half‐life). Despite this, the data demonstrate that absorption was very rapid and peak concentrations occurred within minutes. This means that IM administration results in almost as rapid peak drug concentrations as IV administration, without the need for IV access. A previous study has demonstrated that on average there is a delay of up to 9 min in sedating patients for IV administration due to difficulties in gaining IV access, compared to immediate IM administration of medication 2. The figures show that peak levels occur within 9 min so that any benefit of IV administration in time to peak is lost in the time to obtain IV access. In a much larger clinical study the median time to sedation was 20 min in 1403 patients administered droperidol 16.

Another limitation of the study was the diverse range of drugs ingested by some of the patients who presented with deliberate self‐poisoning. These included numerous prescription drugs where for many cases only one patient ingested a particular drug, making any sort of covariate analysis impossible. Twenty‐two percent of patients had a urine drug screen (Table 1), which only detected drugs of abuse, and again because there was such a varied combination of drugs ingested, it was not possible to undertake a covariate analysis. Alcohol intoxication was common but did not appear to affect the pharmacokinetics of droperidol.

Further study is required to link drug concentrations to level of sedation to determine the appropriate concentrations for sedation and confirm that 10 mg droperidol is appropriate. This study provides further support to clinical data on droperidol showing that it is safe and effective for sedation of aggressive patients for short periods of time (up to 4–6 h).

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organisation for the submitted work.

G.K.I. is supported by an NHMRC Senior Research Fellowship 1 061 041.

Foo, L. ‐K. , Duffull, S. B. , Calver, L. , Schneider, J. , and Isbister, G. K. (2016) Population pharmacokinetics of intramuscular droperidol in acutely agitated patients. Br J Clin Pharmacol, 82: 1550–1556. doi: 10.1111/bcp.13093.