Propofol-associated QTc prolongation

Abstract

Objectives:

Propofol is a preferred agent for sedation in patients in the intensive care unit (ICU) due, in part, to its established safety profile. Despite this, recent case reports have suggested a potential for prolongation of the corrected QT interval (QTc) in ICU patients receiving propofol, though limited empirical work has been conducted to evaluate this association. As such, the purpose of this study was to assess the relationship between propofol infusion and QTc prolongation in a historical cohort of ICU patients.

Methods:

A single-center, historical, observational, pre-post cohort analysis of medical records from admitted patients ⩾18 years old with cardiovascular disease was conducted, involving cases who received propofol infusion for ⩾3 hours with sequential electrocardiogram monitoring from 2006 to 2012. A multivariable, generalized linear model regression was employed to assess the primary outcome of on-propofol QTc interval (QTc₂), controlling for various demographic and clinical factors.

Results:

A total of 96 patients met inclusion criteria, averaging 56.1 ± 14.1 years of age and 86.1 ± 25.0 kg, with 37.5% being female. A mean prolongation in QTc interval of 30.4 ± 55.5 ms (p < 0.001) was observed during the propofol infusion, with 43.8% of cases exhibiting an on-infusion QTc₂ of ⩾ 500 ms. Regression analyses suggested that prolongation in on-propofol QTc was independently associated with baseline QTc interval and amiodarone use, while weight as inversely associated with QTc₂ (p < 0.05).

Conclusion:

This historical cohort analysis of adult ICU patients receiving propofol suggests that on-infusion QTc prolongation was associated with increasing baseline QTc interval and with amiodarone use. Further research is needed to evaluate the clinical significance and cause-and-effect relationship between potential QTc changes and propofol use in the ICU.

Introduction

Adverse drug events (ADEs) have been shown to occur in as many as 2.3% of all hospital admissions and have a significant impact on hospital costs, length of hospital stay and risk of death [Classen et al. 1997; Sikdar et al. 2012; Nebeker et al. 2005]. The intensive care unit (ICU) setting imparts a particularly high risk of ADEs, as patients are typically severely ill and often require complex medications in the setting of altered pharmacokinetics [Classen et al. 1997; Nebeker et al. 2005; Smith et al. 2012]. Numerous risk factors for ADEs have been identified in critically ill patients such as kidney injury, thrombocytopenia and use of intravenous medications [Kane-Gill et al. 2012]. In addition to these risk factors, the acuity of illness of patients in the ICU may confer a predisposition for ADEs [Classen et al. 1997; Sikdar et al. 2012; Nebeker et al. 2005; Smith et al. 2012; Kane-Gill et al. 2012; Barr et al. 2013]. As such, close monitoring of medications in the ICU is imperative [Kane-Gill et al. 2012; Barr et al. 2013; Drew et al. 2004; Pinsky, 2007].

Electrocardiogram (ECG) monitoring provides critical information about medication safety by providing information regarding cardiac electrophysiology and risk for fatal cardiac arrhythmias, particularly the corrected QT interval (QTc interval) which represents ventricular depolarization and repolarization [Drew et al. 2004; Pinsky, 2007]. Prolongation of the QTc interval is associated with increasing risk for fatal ventricular arrhythmias; therefore, ECG monitoring is recommended if there is a risk of QTc prolongation [Bazett, 1997; Fridericia, 2003; Drew et al. 2010; Yap and Camm, 2003]. The risk of QTc prolongation is greater in the ICU for multiple reasons, including acute or unstable disease, the potential for electrolyte imbalances, and possible exposure to QTc prolonging medications [Drew et al. 2004; Viskin et al. 2003; Beitland et al. 2014]. Medications known to increase the risk of QTc interval prolongation and arrhythmia risk include antiarrhythmic agents, antipsychotics, tricyclic antidepressants, fluoroquinolones, macrolides and azole antifungals [Drew et al. 2010; Yap and Camm, 2003; Viskin et al. 2003; Beitland et al. 2014]. While the potential for QTc prolongation may be known for numerous medications, assessment of QTc interval changes for medications has only been required since 2005 [Darpo, 2010]. Therefore, assessment of medications’ effect on the QTc interval approved prior to this time is important for medication safety.

Propofol is a non-benzodiazepine commonly recommended for sedation in the ICU and was approved for use in the United States in 1989 [Barr et al. 2013]. The Society of Critical Care Medicine Guidelines for Pain, Agitation, and Delirium in the ICU recommend non-benzodiazepines such as propofol, in part, because these agents are associated with reduced time in the ICU, duration of mechanical ventilation, and incidence of delirium compared with benzodiazepines [Barr et al. 2013]. As propofol is historically considered to not prolong the QTc interval, it has been used as an active comparator in studies that evaluate the QTc-prolonging effects of other general anesthetics [Kleinsasser et al. 2000; Michaloudis et al. 1996]. In clinical practice, ECG monitoring is not generally considered routine practice for patients receiving propofol.

Despite historical consensus and an extensive pattern of use, recognition of the potential for propofol to adversely affect the QTc interval has recently emerged. Case reports have suggested propofol induces QTc prolongation in acutely ill patients with risk factors, such as acute cardiovascular disease, hypoalbuminemia or trauma [Sakabe et al. 2002; Irie et al. 2010; Douglas and Cadogan, 2008]. Additionally, propofol has been associated with fatal ventricular dysrhythmias secondary to propofol infusion syndrome [Fong et al. 2008], a rare but potentially fatal adverse reaction to propofol infusions. Several pharmacologic mechanisms have been identified that may support the cogency of propofol’s association with these adverse cardiovascular effects. For example, propofol has been shown to block L-type calcium channels [Fassl et al. 2011], a mechanism for QTc abbreviation unless IKr and L-type calcium channel blockade are present resulting in IKs inhibition and QTc prolongation [Puttick and Terrar, 1992, Heath and Terrar, 1996]. Other mechanisms include inhibition of volume-sensitive chloride channels in coronary artery smooth muscle cells [Masuda et al. 2003], a property which may cause depolarization and block adenosine triphosphate channels in rat cardiomyocytes [Kawano et al. 2004] or polymorphic ventricular tachycardia in rabbits with long QT syndrome type 2 [Odening et al. 2008]. Propofol has also been shown to inhibit sympathetic neuronal activity and decrease baroreflex sensitivity, mechanisms that are believed to predispose other sedative agents to QTc prolongation [Sellgren et al. 1994]. Yet, formal evaluations of propofol on QTc interval changes among ICU patients are lacking.

The impact of propofol on QTc interval changes has been prospectively evaluated among patients undergoing anesthesia induction with conflicting results, though the duration of propofol infusion and patient demographics are considerably different from an ICU population, precluding extrapolation to the ICU setting [Michaloudis et al. 1996; Saarnivaara et al. 1990; Kim et al. 2008; Higashijima et al. 2010; Oji et al. 2013; Hanci et al. 2010]. Further exploration of the relationship between propofol and QTc prolongation in intensive care patients is essential given that propofol continues to be a preferred agent for sedation in the ICU that is not routinely monitored with an ECG [Barr et al. 2013]. Therefore, the purpose of the current study was to assess the relationship between propofol infusion and QTc prolongation in a historical cohort of ICU patients.

Methods

Study design, setting and participants

This single-center, historical, observational, pre-post cohort study was conducted to assess the impact of QTc interval changes in patients receiving propofol in the ICU at the University of Oklahoma Medical Center (OUMC), a 300 bed, tertiary academic medical center and level 1 trauma center serving central Oklahoma. Data were collected using a standardized collection form from the electronic medical record of patients admitted between 1 January 2006 and 31 December 2012. Human subjects’ protection approval was granted from the University of Oklahoma Health Sciences Center Institutional Review Board.

Patients 18 years of age or older who received a propofol infusion lasting at least 3 hours were included in the study. Patients were excluded if they did not have two ECG assessments or if propofol was stopped before the second ECG was measured. The first ECG had to be completed within the 72 hours prior to propofol initiation to establish a baseline; the second ECG had to be completed within the 72 hours after propofol was started but before propofol was discontinued. All ECGs were standard 12-lead recordings with interpretation and manual QT/QTc measurement by OUMC cardiologists. Patients with at least one cardiovascular diagnosis were included to assure the presence of ECG monitoring, since the standard of care for propofol monitoring does not include regular ECG assessment.

The primary outcomes of the current investigation were: (1) the on-infusion QTc interval (QTc₂), measured after at least 3 hours of propofol infusion in patients in the ICU; and (2) the presence of a prolonged on-infusion QTc interval (QTc₂) ⩾500 ms (i.e. yes/no). The independent (i.e. predictor) variables included baseline QTc interval (QTc₁), serum albumin concentration, age, sex, weight, Charlson–Deyo comorbidity index (a validated measure of case-mix severity based on comorbid conditions) and the use of any QTc interval-prolonging medications [Drew et al. 2004, 2010; Yap and Camm, 2003; Viskin et al. 2003; Charlson et al. 1987, 1994]. The QTc-prolonging medications evaluated were those available through the hospital formulary and available for use in the local ICU and are listed in Table 1. Furthermore, based on the observation that a substantially larger quantity of amiodarone use versus other agents had occurred, amiodarone was incorporated in the analysis as a separate variable. Figure 1 presents a general schematic of the study’s methodological framework, including outcomes and independent variables.

Table 1.

Designated QTc-prolonging medications.

Amiodarone	Domperidone	Ibutilide	Quinidine
Chlorpromazine	Dronedarone	Levofloxacin	Ranolazine
Ciprofloxacin	Erythromycin	Moxifloxacin	Risperidone
Citalopram	Escitalopram	Ondansetron	Sotalol
Clarithromycin	Flecainide	Paliperidone	Venlafaxine
Clozapine	Fluconazole	Procainamide	Voriconazole
Disopyramide	Granisetron	Propafenone	Ziprasidone
Dofetilide	Haloperidol	Quetiapine

Figure 1.

Schematic of methodology to assess the on-infusion QTc interval (QTc2), and the presence of a prolonged on-infusion QTc interval (QTc2) > 500 ms.

Statistical analysis

Descriptive statistics were used to analyze baseline patient characteristics, including univariate paired sample t-test for QT interval from baseline to on-propofol infusion (i.e. QTc₁ versus QTc₂) and McNemar’s test for the proportion of cases involving on-infusion QTc₂ ⩾ 500 ms. Measures of central tendency were presented as mean ± standard deviation. A multivariable generalized linear model (GLM) regression analysis was employed to assess the associations between the primary outcome of on-propofol QTc interval time, QTc₂ and the independent variables (i.e. demographic and clinical factors). Notably, the GLM was specified with a normal (Gaussian) distribution and identity link, using Huber–White heteroskedasticity-consistent robust standard error calculations and offsetting for the differential time from initiation of propofol infusion to QTc₂ measurement [Skrepnek, 2005; Skrepnek et al. 2012; Huber, 1967; White, 1980]. Therein, the results of this regression analysis followed a straightforward interpretation: a one-unit increase in a statistically significant independent variable would be associated with a change in QTc₂ in milliseconds designated by the coefficient estimate, $\widehat{\beta }$ [Skrepnek, 2005]. Results of the regression analysis of QTc₂ time followed an interpretation similar to a linear regression: a one-unit increase in a statistically significant independent variable would be associated with a $\widehat{\beta }$ change in QTc₂ in milliseconds (ms) designated by that coefficient estimate,$\widehat{\beta }$ [Skrepnek, 2005]. The regression involving the outcome of QTc₂ time ⩾500 ms would be interpreted as a straightforward logistic regression wherein an odds ratio (OR) = 1.0 would indicate no difference [Skrepnek et al. 2012]. Residual diagnostics and model deviance assessments (e.g. Akaike Information Criteria) were also performed [Skrepnek, 2005; Skrepnek et al. 2012]. Post hoc subgroup analyses were performed for cases that did not involve the use of any QTc-prolonging drugs. An a priori alpha level of 0.05 was used for statistical significance. Finally, post hoc power analyses were also conducted using correlation matrices across the independent variables. All analyses were performed using Stata MP Version 14.0 (College Station, TX).

Results

A total of 187 adult patients admitted between 2006 and 2012 received a propofol infusion lasting at least 3 hours; 96 met inclusion criteria. The predominant reasons for exclusion were missing ECG data prior to the start of the infusion (n = 20), after the infusion (n = 15), or both (n = 47). On average, patients included in the investigation were 56.1 ± 14.1 years of age, weighed 86.1 ± 25.0 kg, and 37.5% were female. An acute cardiovascular diagnosis was the primary reason for ICU admission in most patients (n = 51 with primary cardiac diagnosis; n = 45 with noncardiac diagnosis as primary diagnosis). Among those admitted for noncardiovascular reasons, the most common cardiovascular comorbidities included coronary artery disease (22.9%, n = 22), hypertension (14.6%, n = 14) and heart failure (9.4%, n = 9). The mean time from propofol initiation to measurement of QTc₂ was 14.3 ± 16.0 hours. The average QTc₁ was 465.6 ± 49.2 ms [95% confidence interval (CI): 455.6–475.5] and the average QTc₂ was 496.0 ± 55.8 ms (95% CI: 484.6–507.3), representing a pre- to on-propofol change of +30.4 ± 55.5 ms (95% CI: 18.7–41.2) (p < 0.001). At baseline, 18 patients (18.8%) had a QTc₁ ⩾500 ms; after propofol initiation, a total of 42 patients (43.8%) had a QTc interval (QTc₂) ⩾500 ms, representing a 25% increase in proportion compared with baseline (p < 0.001). No patients experienced significant ventricular arrhythmias related to QTc prolongation during the study period. Complete descriptive statistics appear in Table 2.

Table 2.

Patient and clinical characteristics (n = 96).

Characteristic	Value
Age, years [mean ± SD (95% CI)]	56.1 ± 14.1 (53.3, 59.0)
Weight, kg [mean ± SD (95% CI) ]	86.1 ± 25.0 (81.1, 91.2)
Serum creatinine, mg/dl [mean ± SD (95% CI)]	1.7 ± 1.7 (1.4, 2.1)
Serum albumin, g/dl [mean ± SD (95% CI) ]	3.1 ± 0.6 (2.9, 3.2)
Pre-infusion QTc1, ms [mean ± SD (95% CI)]	465.6 ± 49.2 (455.6, 475.5)
On-Infusion QTc2, ms [mean ± SD (95% CI)]	496.0 ± 55.8 (484.6, 507.3)
Change from pre- to on-infusion-QTc, ms [mean ± SD (95% CI)]	+30.4 ± 55.5 (18.7, 41.2)
Length of propofol infusion, hours [mean ± SD (95% CI)]	14.3 ± 16.0 (11.1, 17.6)
Charlson comorbidity index [mean ± SD (95% CI)]	1.4 ± 1.3 (1.2, 1.7)
Female (n, %)	36 (37.5%)
Amiodarone use (n, %)	32 (33.3%)
Other QT-prolonging medication use (n, %)	32 (33.3%)
Most common reasons for admission:	51 (53.1%)
Cardiac conditions, n (%)	21 (21.8%)
Acute coronary syndromes	7 (7.3%)
Heart failure exacerbation	4 (4.2%)
Endocarditis	4 (4.2%)
Mitral valve disease	3 (3.1%)
Atrial fibrillation	12 (12.5%)
Other	45 (46.9%)
Non-cardiac conditions, n (%)	8 (8.3%)
Respiratory failure	7 (7.3%)
Pneumonia	4 (4.2%)
Subarachnoid hemorrhage	4 (4.2%)
Intracranial hemorrhage	3 (3.1%)
Sepsis	19 (19.8%)
Other

CI, confidence interval; SD, standard deviation.

The multivariable regression analysis of QTc₂ time indicated that baseline QTc₁, weight and amiodarone use were significant predictors of QTc₂ after controlling for other patient demographic and clinical variables (age, sex, serum creatinine, serum albumin, Charlson comorbidity index and QT-prolonging medication use) (p < 0.05) (Table 3, Panel A). The results show that for every ms increase in baseline QTc₁, there was a +0.56 ms increase in QTc₂ ($\widehat{\beta }$ = 0.56, p < 0.001, 95% CI: 0.35–0.78); amiodarone use was associated with a +25.63 ms increase in QTc₂ ($\widehat{\beta }$ = 25.63, p = 0.031 95% CI: 0.63–50.63). However, weight was inversely-related to QTc₂, as every kg increase was associated with a −0.46 ms decrease in QTc₂ ($\widehat{\beta }$ = −0.46, p = 0.037, 95% CI: −0.89 to −0.28). The multivariable analysis assessing the outcome of QTc₂ prolongation ⩾500 ms (Table 3, Panel B) yielded generally consistent results: for every ms increase in baseline QTc₁, there was a 1.03 times higher odds of QTc₂ prolongation ⩾500 ms (OR = 1.03, p < 0.001, 95% CI: 1.02–1.05). Likewise, amiodarone use was associated with a 10.42 times higher odds of QTc₂ prolongation ⩾500 ms (OR = 10.42, p = 0.005, 95% CI: 2.05–52.90). Each kg increase in weight was associated with a 4% reduction in odds of QTc₂ prolongation ⩾500 ms (OR = 0.96, p = 0.019, 95% CI: 0.92–0.99). Overall, the amount of variation explained by both regression models was approximately 21% as measured via the pseudo R²_adjusted. Post hoc power analyses indicated that power exceeded 99% across the main regression analyses.

Table 3.

Multivariable analyses of on-propofol infusion QTc time (in ms) and presence of prolongation ⩾500 ms.

Variable	PANEL A On-infusion QTc2 time$ (overall sample n = 96)		PANEL B On-infusion QTc2 prolongation ⩾500 ms‡ (overall sample, n = 96)
	Coefficient estimate, βˆ$\widehat{\beta }$(95% confidence interval)	p value	Odds ratio (95% confidence interval)	p value
Age	−0.56 (−1.31, 0.19)	0.145	0.97 (0.93, 1.01)	0.124
Female	7.40 (−14.94, 29.74)	0.532	1.97 (0.44, 8.85)	0.378
Weight	−0.46* (−0.89, −0.28)	0.012	0.96* (0.92. 0.99)	0.019
Serum creatinine	3.77 (−3.13, 10.67)	0.130	1.48 (0.93, 2.36)	0.102
Serum albumin	1.03 (−16.08, 18.14)	0.918	1.74 (0.52, 5.89)	0.371
Pre-infusion QTc1	0.56*** (0.35, 0.78)	<0.001	1.03*** (1.02, 1.05)	<0.001
Charlson comorbidity index	1.25 (−7.44, 9.94)	0.809	1.16 (0.57, 2.33)	0.686
Amiodarone use	25.63* (0.63, 50.63)	0.031	10.42** (2.05, 52.90)	0.005
Other QT-prolonging medication use	4.64 (−20.53, 29.80)	0.693	2.18 (0.35, 13.63)	0.404

^*Statistically significant at p < 0.05; ^**Statistically significant at p < 0.01; ^***Statistically significant at p < 0.001.

^$Generalized linear model (GLM): Gaussian distribution with identity link, offset by time from propofol initiation to on-infusion ECG reading, maximum likelihood estimation; log-likelihood = −505.79; model constant = 278.713, p < 0.001, 95% CI (152.7, 404.8); pseudo R²_adjusted = 0.217.

^‡Generalized linear model (GLM): binomial (logistic) distribution with identity link, offset by time from propofol initiation to on-infusion ECG reading, maximum likelihood estimation; log-likelihood = −68.00; model constant = 1.41^E-07, p = 0.001, 95% CI (1.83^E-11,0.01)_); pseudo R²_adjusted = 0.212.

CI, confidence interval; ECG, electrocardiogram.

The post hoc subgroup analysis of patients not receiving amiodarone or any other QTc interval prolonging medications (n = 60) produced similar results as the main analysis, with full subgroup results presented in Table 4. For this subgroup (Table 4, Panel A), each ms increase QTc₁ was associated with a +0.52 ms QTc₂ time ($\widehat{\beta }$ = 0.52,p < 0.001, 95% CI: 0.36–0.68), while each kg increase in weight was associated with a −0.47 ms decrease in QTc₂ ($\widehat{\beta }$ = −0.47, p = 0.016, 95% CI: −0.84 to −0.09). The regression analysis assessing QTc₂ prolongation ⩾500 ms (Table 4, Panel B) also indicated that for every ms increase in baseline QTc₁, there was a 1.03 times higher odds of QTc₂ prolongation ⩾ 500 ms (OR = 1.03, p = 0.002, 95% CI: 1.02–1.05). Each kg increase in weight was also associated with a 4% reduction in odds of QTc₂ prolongation ⩾500 ms (OR = 0.96, p = 0.036, 95% CI: 0.92–0.99).

Table 4.

Post hoc subgroup multivariable analyses of on-propofol infusion QTc time (in ms) and presence of prolongation ⩾ 500 ms among patients not receiving QT-prolonging drugs.

Variable	PANEL A On-infusion QTc2 time$ (No QT prolonging drugs, n = 60)		PANEL B On-infusion QTc2 prolongation ⩾ 500 msec‡ (No QT prolonging drugs, n = 60)
	Coefficient estimate,$\widehat{\beta }$ βˆ(95% confidence interval)	p value	Odds ratio (95% confidence interval)	p value
Age	−0.52 (−1.29, 0.25)	0.182	0.97 (0.93, 1.01)	0.200
Female	3.16 (−19.57, 25.89)	0.785	1.31 (0.30, 5.80)	0.721
Weight	−0.47* (−0.84, −0.09)	0.016	0.96* (0.92. 0.99)	0.036
Serum creatinine	3.77 (−1.55, 9.09)	0.165	1.51 (0.86, 2.65)	0.153
Serum albumin	1.97 (−18.20, 22.15)	0.848	1.77 (0.54, 5.83)	0.346
Pre-infusion QTc1	0.52*** (0.36, 0.68)	<0.001	1.03** (1.01, 1.05)	0.002
Charlson comorbidity index	−1.49 (−8.68, 11.67)	0.773	1.13 (0.59, 2.19)	0.711
Amiodarone use	Excluded		Excluded
Other QT-prolonging medication use	Excluded		Excluded

^*Statistically significant at p < 0.05; ^**Statistically significant at p < 0.01; ^***Statistically significant at p < 0.001.

^$Generalized linear model (GLM): Gaussian distribution with identity link, offset by time from propofol initiation to on-infusion ECG reading, maximum likelihood estimation; log-likelihood = −508.26; model constant = 306.40, p < 0.001, 95% CI (187.69, 425.11); pseudo R²_adjusted = 0.065.

^‡Generalized linear model (GLM): binomial (logistic) distribution with identity link, offset by time from propofol initiation to on-infusion ECG reading, maximum likelihood estimation; log-likelihood = −74.22; model constant = 2.92^E-06, p = 0.003, 95% CI (5.70^E-10,0.01)_); pseudo R²_adjusted = 0.075.

CI, confidence interval; ECG, electrocardiogram.

Discussion

This historical cohort analysis of on-infusion QTc intervals among 96 adult ICU patients receiving propofol found a significant prolongation in QTc interval of +30.4 ± 55.5 ms (p < 0.001). After statistically controlling for patient demographics and clinical characteristics, QTc₂ interval prolongation was significantly associated with baseline QTc interval and amiodarone use, while weight was inversely related to QTc₂ (p < 0.05). QTc₂ prolongation ⩾500 ms occurred in 43.8% of the cohort, a 25% increase from baseline (p < 0.001). The likelihood of crossing this clinically important threshold was similarly associated with QTc₁, amiodarone and weight (p < 0.05). A post hoc subgroup analysis among those not receiving any QT-prolonging medications (n = 60) yielded similar results as the study’s main findings. As such, though it was expected that amiodarone would intuitively be associated with QTc prolongation, this subgroup analysis suggests that findings cannot be solely attributed to amiodarone or to other QTc prolonging medications. Though the retrospective design of this study precludes determination of causation, these findings may assist in identifying high risk groups undergoing sedation in the ICU.

The current study’s finding that the baseline QTc₁ interval is associated with QTc₂ prolongation is consistent with a report of propofol-associated cardiac arrhythmia [Rewari and Kaul, 2003]. This case involved a 28-year-old male patient with no underlying cardiovascular diagnosis, no history of syncope and no known family history of long QT syndrome who was undergoing wound debridement surgery. Anesthesia was induced with fentanyl and propofol; 1 hour into the surgery, the patient developed polymorphic ventricular tachycardia alternating with ventricular fibrillation. No causes of QTc prolongation were identified except for the administration of propofol. After cardiopulmonary resuscitation, the patient recovered, was observed in the ICU for 24 hours, and was subsequently referred to cardiology, where his QTc interval was borderline at 460 ms. Further diagnostics revealed a prolonged QTc interval and the patient was found to have long QT syndrome that was undiagnosed before surgery. Our findings corroborate the concerns raised by this case report about propofol’s association with further QTc prolongation among those patients with a borderline QTc interval at baseline.

An additional observation was that QTc₂ and odds of QTc₂ prolongation ⩾500 ms were inversely related with a patient’s body weight, but had no relationship (direct or inverse) with other clinical characteristics such as age, degree of illness or albumin levels. This was unexpected given that the case reports that have associated QTc prolongation with propofol administration included patients with one or more of these characteristics. One case described a 71-year-old woman presenting with an anterior myocardial infarction who was given a bolus and continuous infusion of propofol to allow for mechanical ventilation. Her QTc interval increased from 440 ms to 690 ms 4 hours into the infusion and shortened to 490 ms after withdrawal of propofol; the patient was rechallenged with propofol and QTc was prolonged again (610 ms). No cardiac dysrhythmias developed [Sakabe et al. 2002]. Conversely, a 78-year-old woman developed torsade de pointes after receiving two bolus doses of propofol for sedation for cast placement of a spiral fracture of the distal tibia. An ECG was not obtained during cast placement, though baseline QTc and post event QTc were both within normal range (415 and 405 ms, respectively) [Douglas and Cadogan, 2008]. Finally, a 70-year-old man with multiple acute medical issues and an albumin of 1.4 g/dl was given propofol for mechanical ventilation. His QTc interval increased from 398 ms to 712 ms, with torsade de pointes developing 15 hours after the initiation of propofol [Irie et al. 2010].

Based on these cases, one might have speculated age, hypoalbuminemia or degree of illness to have been significant predictors of QTc₂, though the characteristics of our study’s ICU population did not include a high proportion of elderly patients and may not have captured an adequate number of patients with the characteristics presented in these cases (e.g. profound hypoalbuminemia). Therefore, the current study results cannot rule out an association with the factors presented in these cases. Caution is prudent when using propofol for sedation in intensively ill elderly patients or patients with markedly low albumin levels.

To our knowledge, no other reports have described a relationship between propofol administration, weight and QTc prolongation. Propofol dosing is weight-based and would be anticipated to have comparable clinical effects across various weights [van Kralingen et al. 2011]. In fact, it has been reported that obesity (body mass index >30) may be a risk factor for QTc prolongation [Brown et al. 2001]. Propofol is a highly protein-bound drug (97% binding to albumin) and is known to reach higher plasma levels in elderly patients due to age-related decreases in volume of distribution and intercompartmental clearance [Fresenius Kabi, 2014]. Although the association between decreasing body weight and QTc₂ prolongation was also consistently observed in the subgroup analysis, inherent limitations of the sample selection and sample size should also be considered. While further evaluation of this relationship is warranted, it may be prudent to monitor ECGs if a patient’s weight falls to either extreme.

Despite the current study’s analyses and results, there are potential limitations to consider in interpreting or extrapolating findings. The retrospective, nonrandomized nature of this study precludes making statements regarding causation between propofol administration and QTc prolongation given that no control group could be evaluated. The measurement of the QTc interval is potentially unreliable depending on the techniques used to calculate its length or with heart rate extremes during automated ECG assessments [Isbister and Page, 2013; Waring et al. 2010]. The manual interpretation by OUMC cardiologists mitigated variability in QTc measurement and its influence as a confounder in this evaluation. Though the sample size was relatively small, post hoc power analyses indicated sufficient power (i.e. >99%) for the primary endpoints. Additionally, robust statistical approaches (e.g. Huber–White standard errors) were used to mitigate potential violations of any assumptions required in the regression framework. The low patient numbers is also not a reflection of a low ICU census during the study period. Rather, the small number of patients is a result of ECGs not being routinely performed in patients receiving propofol which limited the ability to obtain baseline and on-propofol QTc data.

Prior to this analysis, only case reports have described the potential ADE of QTc₂ prolongation among patients receiving propofol for sedation in the ICU, making our study the largest assessment of propofol’s effect on the QTc interval in the ICU at the time of writing [Sakabe et al. 2002; Irie et al. 2010; Douglas and Cadogan, 2008]. While the use of QTc-prolonging medications was measured during the infusion time period, specific factors including doses, dosing intervals or durations of use may have also been associated with outcomes. Additional patient parameters with known impact on QTc interval changes such as electrolyte imbalances and cardiovascular history may have also influenced the outcomes. Finally, given that this was a single-center study results may not necessarily be generalizable to other ICUs or other patient populations.

By focusing explicitly on patients in an ICU setting, the current research seeks to address an important gap in the literature. Prior to the current analysis, only case reports have described an association between propofol administration and subsequent QTc prolongation in the ICU setting [Sakabe et al. 2002; Irie et al. 2010; Douglas and Cadogan, 2008]. Areas of future research should capture larger samples across multicenter sites and employ additional study methodologies to corroborate the findings of the current study. Continued work should also seek to develop clinical tools by identifying and validating the thresholds of clinical parameters which define increased risk to patients. Recognizing again that the present investigation was limited to patients with at least one cardiovascular diagnosis, additional research is required to establish if propofol is also associated with QTc interval prolongation among those without cardiovascular diagnoses.

Conclusion

This historical cohort analysis of patients receiving propofol for sedation in the ICU found a significant increase in on-infusion QTc prolongation. After controlling for several patient demographics and clinical variables, both longer on-infusion QTc₂ intervals and the presence of prolongation ⩾500 ms were found to be significantly associated with increasing baseline QTc₁ intervals and with amiodarone use, while weight was observed to be inversely related. As potential consequences of QTc prolongation involve fatal ventricular arrhythmias, it may be prudent to monitor ECGs of high-risk patients receiving propofol.