Article, Toxicology

High dose insulin for beta-blocker and calcium channel-blocker poisoning

a b s t r a c t

Background/objectives: High dose insulin (HDI) is a Standard therapy for beta-blocker (BB) and calcium channel- blocker (CCB) poisoning, however human case experience is rare. Our poison center routinely recommends HDI for shock from BBs or CCBs started at 1 U/kg/h and titrated to 10 U/kg/h. The study objective was to describe clin- ical characteristics and adverse events associated with HDI. Methods: This was a structured chart review of patients receiving HDI for BB or CCB poisoning with HDI defined as Insulin infusion of >=0.5 U/kg/h.

Results: In total 199 patients met final inclusion criteria. Median age was 48 years (range 14-89); 50% were male. Eighty-eight patients (44%) were poisoned by BBs, 66 (33%) by CCBs, and 45 (23%) by both. Median nadir pulse was 54 beats/min (range 12-121); median nadir systolic blood pressure was 70 mm Hg (range, 30-167). Forty- one patients (21%) experienced cardiac arrest; 31 (16%) died. Median insulin bolus was 1 U/kg (range, 0.5-10). Median starting insulin infusion was 1 U/kg/h (range 0.22-10); median peak infusion was 8 U/kg/h (range 0.5- 18). Hypokalemia occurred in 29% of patients. Hypoglycemia occurred in 31% of patients; 50% (29/50) experi- enced hypoglycemia when dextrose infusion concentration <=10%, and 30% (31/105) experienced hypoglycemia when dextrose infusion concentration >=20%.

Conclusions: HDI, initiated by emergency physicians in consultation with a poison center, was feasible and safe in this large series. Metabolic abnormalities were common, highlighting the need for Close monitoring. Hypoglyce- mia was more common when less concentrated dextrose maintenance infusions were utilized.

Introduction

Background

Cardiovascular drugs cause a substantial number of poisoning deaths each year, primarily from acute overdose. In 2016, cardiovascular drugs as a class were the second leading cause of poisoning deaths re- ported to the National Poison Data System (NPDS), the vast majority of which were beta-blockers and calcium channel-blockers [1].

High dose insulin (HDI) has become a standard therapy for poison- ing from both beta-blockers and calcium channel-blockers [2]. Despite

? Prior presentations: This work was presented in abstract form as a poster presentation at the 2017 North American Congress of Clinical Toxicology in Vancouver, BC, Canada (abstract #52).

* Corresponding author at: 701 Park Ave, Mail Code: RL.240, Minneapolis, MN 55415, United States.

E-mail addresses: [email protected], @jonbcole2 (J.B. Cole). @jonbcole2 (J.B. Cole).

basic science evidence that HDI is superior to vasopressors for poisoning from both calcium channel-blockers [3,4] and beta-blockers [5,6], con- sensus does not yet exist regarding which therapy to start first after basic supportive care, such as IV fluids and calcium salts, has failed [7,8]. As such, emergency physicians may be unfamiliar with HDI and the implementation of this complex therapy.

In our regional poison center, the clinical guideline for the treatment of beta-blocker and calcium channel-blocker poisoning recommends starting HDI prior to vasopressors after basic supportive care has failed. The guideline (Fig. 1) recommends titrating HDI from 1 to 10 U/kg/h, based on data suggesting cardiac output increases approximately 50% from an HDI dose of 1 U/kg/h to 10 U/kg/h [9].

Importance

Though expert Consensus guidelines recommend HDI if basic sup- portive care has failed [8], there is a relative paucity of human case ex- perience with HDI. In fact, HDI use in humans is poorly described. The

https://doi.org/10.1016/j.ajem.2018.02.004

Check baseline creatinine, glucose, and potassium. Replace potassium if hypokalemia is present.

Administer 25-50 g of dextrose 50% (D50) IV if blood glucose < 200 mg/dL. Administer 1 U/kg of regular insulin IV push.

Begin infusion of 1 U/kg/hr of regular insulin along with an infusion of dextrose; use D20 via a peripheral IV line until central access has been obtained, then start D50 at 25 g/hr. Initial effects of HDI take at least 5-10 minutes to manifest.

Obtain central venous access for safe infusion of concentrated glucose.

Concentrate all fluids to avoid pulmonary edema. Consider using stock D70 as the glucose infusion. The insulin infusion should be concentrated to 10 U/mL.

Increase insulin infusion 1-2 U/kg/hr every 10-15 minutes to a maximum of 10 U/kg/hr until shock improves. Doses up to 22 U/kg/hr have been safely used.

Monitor blood glucose every 10-15 minutes until stable, then every 1-2 hour after steady state is reached. For every hypoglycemic reading, give 25 g of D50 as a bolus and increase the dextrose infusion by 25 g/hr.

Monitor serum potassium every hour during titration, then every 4-6 hours once steady state is reached.

Regarding electrolytes, maintain serum potassium >= 2.8 mEq/L and calcium 1.5-2 times the Upper limit of normal. Monitor magnesium and phosphorus every 4-6 hours and replace as needed. We recommend standard ICU magnesium and phosphorus replacement protocols.

Once stable, wean vasopressors before HDI. Contact the Poison Center regarding weaning of HDI once stability has been reached.

Fig. 1. Implementation of high dose insulin therapy.

largest published series to date is a brief analysis of 46 poison center cases for which HDI was recommended for calcium channel-blocker poisoning [10]. Case reports and smaller series of successful HDI use exist, but frequently are reported by medical toxicologists performing bedside consultation and may suffer from positive reporting bias [11- 16]. As of 2015, published human case experience describing the use of insulin for drug-induced cardiac toxicity was limited to under 100 cases [17]. Emergency physicians may be hesitant to use HDI in lieu of more familiar therapies such as vasopressors given the relative scarcity of published clinical experience. A larger study more clearly describing HDI use, administration, and outcomes may better define the role for HDI in the emergency department (ED).

Goals of this investigation

The primary purpose of this study was to describe the clinical char- acteristics of patients receiving HDI for beta-blocker or calcium channel- blocker poisoning as recommended by our regional poison center in- cluding HDI dosing, concomitant therapies, clinical outcomes, and ad- verse effects associated with HDI.

Methods

Study design and setting

This was a structured [18,19] single poison center chart review of pa- tients treated with HDI for beta-blocker or calcium channel-blocker poi- soning treated from 2000 to 2016. Our human subjects research committee approved this study.

The study setting is an American Association of Poison Control Cen- ters (AAPCC) accredited regional poison center covering 3 U.S. states. In

2016, our poison center handled 64,468 calls, of which 58,838 were ex- posure calls. This poison center is pharmacist-based; 100% of our eligi- ble pharmacists are AAPCC certified specialists in poison information (CSPIs). Board-certified medical toxicologists are available at all times via phone consultation.

Selection of participants

Patients were identified by querying our electronic patient database (Toxicall(R), version 4.7.37, 1999-2013, Computer Automation Systems, Inc., Aurora, CO). Each case in Toxicall(R) consists of categorical data fields (such as gender and exposure route) as well as free text case notes that describe the case in a manner similar to traditional hospital medical records. Patients were identified by searching for patients coded to AAPCC generic substance codes for “Beta blockers” and/or “Cal- cium Antagonists” in the “substance description” data field, “exposure” in the “call type” data field, and the therapy “Insulin” (recommended or performed) coded in the “therapies” data field for the complete years 2000-2016. All cases were handled and documented by trained CSPIs prospectively.

As there is no consensus on minimum dosing for HDI, we defined HDI as a minimum of 0.5 U/kg/h or 25 U/h based upon published re- views of HDI [20,21]. Eligibility was confirmed by review of each patient’s chart; patients were included if (1) the record confirmed in- gestion of a beta-blocker and/or calcium channel-blocker and (2) the patient received HDI (as defined above). Patients were excluded if we confirmed that HDI was recommended but not performed, if a patient received an insulin bolus only with no subsequent infusion, if the pa- tient received an insulin infusion but no infusion dose was recorded, or if it was confirmed the patient received an insulin infusion at a rate below the a priori definition of HDI.

Measurements

Four of the authors, all board-certified medical toxicologists, ab- stracted the data by reviewing the categorical data of each case as well as the poison center case notes. We assessed inter-rater reliability by assessing 20 duplicate charts. We calculated a kappa statistic on the fol- lowing variables: drug class, specific exposure medications, clinical in- terventions, and occurrence of adverse events. All data collection was managed using the electronic data capture software tool, REDCap(R) (Re- search Electronic Data Capture) [22].

Demographic data were recorded but limited to age and gender based on NPDS standard data fields. The specific offending beta-blocker and/or calcium channel-blocker was recorded along with any Clinically meaningful co-ingestions. Vital sign assessment consisted of recording the initial and nadir pulse rate and systolic blood pressure. Insulin dos- ing, specifically the bolus dose, initial infusion rate, and maximum infu- sion rate, were recorded for each patient. Insulin dosing was reported as weight based (U/kg for bolus doses, U/kg/h for infusions), unless only unit-based dosing (U for bolus doses, U/h for infusions) was reported. Patients experiencing cardiac arrest were analyzed separately and not included in the vital sign data, therefore the reported range does not in- clude zero for either parameter. Clinical outcomes (unrelated, minor, moderate, major, death), as defined by NPDS [1,23], were recorded as well as the following clinical effects (also defined by NPDS): “acidosis,” “asystole,” “bradycardia,” “cardiac arrest,” “coma,” “conduction distur- bance,” “creatinine increased,” “drowsy/lethargic,” “dysrhythmia (other),” “dysrhythmia (V-tach/V-fib),” “electrolyte abnormality,” “hy- potension,” “hypoglycemia,” “oliguria,” “renal failure,” and “seizures.” These specific clinical effects were chosen to best match the anticipated effects of beta-blocker and calcium channel-blocker poisoning; defini- tions are noted in Online Supplementary Table 1.

We further attempted to identify metabolic adverse effects related to HDI, specifically additional episodes of hypoglycemia and hypokale- mia. In addition to searching for “hypoglycemia” coded by CPSIs,

abstractors manually searched case records for mention of blood glu- cose b70 mg/dL, or case notes describing hypoglycemia that was treated with additional dextrose. Each individual episode of hypoglycemia was recorded and tallied for each patient. Regarding hypokalemia, abstrac- tors recorded the lowest recorded Serum potassium concentration, in- cluding if the case notes merely noted serum potassium was “normal.” Other clinical therapies, also defined by NPDS, were recorded. Spe- cific therapies were searched for based upon the expected toxicity of beta-blockers and calcium channel-blockers. The following therapies were recorded: “antiarrhythmic,” “atropine,” “calcium,” “CPR,” dextrose (“Glucose, N5%”), “ECMO,” “glucagon,” “hemodialysis,” “intubation,” “methylene blue,” “pacemaker,” and “vasopressors.” Regarding vaso- pressors, case notes were analyzed to determine specific vasopressors or inotropes administered via infusion. As it is uncommon for vasopres- sor infusion doses and rates to be documented, we recorded only which vasopressors and inotropes were administered to a given patient and not the infusion dose or rate. In addition, abstractors were instructed to specifically search for and record evidence of intra-aortic balloon pump, intravenous fat emulsion (IFE), and left-ventricular assist de- vices, as none of these three therapies have NPDS-specific codes. All cases treated with IFE were analyzed to determine if a hemodynamic re-

Study Enrollment

506 patients met initial search criteria

307 excluded patients

Coding error (HDI not truly

recommended in case notes)

(n = 103)

HDI recommended, not performed

(n = 76)

Insulin infusion administered, no dose recorded

(n = 48)

Insulin infusion given, confirmed < 0.5 U/kg/hr or 25 U/hr

(n = 45)

Insulin bolus only (n = 35)

199 patients met final inclusion criteria for HDI

(insulin infusion >= 0.5 U/kg/hr or 25 U/hr)

sponse occurred following IFE administration.

Regarding dextrose as a therapy, we specifically assessed for mainte- nance dextrose concentration, as we hypothesized a higher mainte- nance dextrose concentration would be associated with fewer events of hypoglycemia. Abstractors recorded the maximum dextrose infusion concentration for each case (ranging from 5 to 70% dextrose).

Cardiac arrest sub-group analysis

All cases where the clinical effects ‘asystole,’ ‘cardiac arrest,’ or “dys- rhythmia (V-tach/V-fib)” were coded were reviewed by the primary in- vestigator. Once cardiac arrest was confirmed to have occurred, timing

Main results

Insulin dosing

Fig. 2. Study enrollment.

of cardiac arrest (before or after hospital arrival) was recorded in addi- tion to the data above.

Data analysis

Descriptive statistics are reported. Means, medians, inter-quartile ranges, ranges, and confidence intervals were calculated and reported when appropriate. All data were analyzed and graphical outputs were created using STATA(R) (Version 14; StataCorp(R), College Station, TX).

Results

Characteristics of study subjects

A total of 506 patients met initial screening criteria. There were 307 patients excluded leaving 199 patients meeting final inclusion criteria; study enrollment is further delineated in Fig. 2. Kappa for drug class and specific exposure medications was 0.9; kappa for clinical interven- tions and adverse events was 0.93. HDI usage and adherence to poison center recommendations increased steadily over time (Fig. 3).

Median age was 48 years (range 14-89); 50% were male. There were 88 patients (44%) poisoned by beta-blockers, 22 patients (11%) poi- soned by dihydropyridine calcium channel-blockers, 44 patients (22%) poisoned by non-dihydropyridine Calcium-channel blockers (exclu- sively verapamil and diltiazem), and 45 patients (23%) poisoned by both beta-blockers and calcium channel-blockers. Specific drugs and co-ingestions are displayed in Table 1. Median initial pulse rate and sys- tolic blood pressure were 62 beats/min (range, 12-128) and 80 mm Hg (range, 40-223), respectively. Median nadir pulse rate and systolic blood pressure were 54 beats/min (range, 12-121) and 70 mm Hg (range, 30-167).

Median weight-based insulin bolus was 1 U/kg (range 0.5-10, n = 31); median unit-based bolus was 50 U (range 5-400, n = 24). Median starting weight-based insulin infusion was 1 U/kg/h (range 0.22-10, n = 94); median starting unit-based infusion was 53 U/h (range 2.5-700, n = 76). Median peak weight-based insulin infusion was 8 U/kg/h (range 0.5-18, n = 122); median unit-based peak insulin infusion was 80 U/h (range, 25-1200, n = 77). Median number of days on HDI was 2 (range 1-7).

Metabolic events and other clinical data

NPDS outcomes and clinical effects are reported in Table 2. Metabolic abnormalities related to HDI were common. Hypokalemia (b3.5 mEq/L) occurred in 57 cases (29%); in 21 cases (10.5%) potassium nadir was b2.8 mEq/L, the recommended tolerable lower limit of our guideline. Nadir potassium measurement ranges are reported in Table 2.

Hypoglycemia was common, occurring in 63 patients (32%) with 40 patients having a documented blood glucose b70 mg/dL. Another 23 pa- tients had evidence of hypoglycemia either by bolus dextrose doses or the word ‘hypoglycemia’ documented in the case notes. Excluding cases where both beta-blocker and calcium channel-blocker poisoning was reported, hypoglycemia was more common in beta-blocker (36/ 88, 41%) than calcium channel-blocker (17/66, 26%) exposures (differ- ence 15%, 95% confidence interval [CI]: 1-30%). In 60 of 63 cases where hypoglycemia occurred, maximum maintenance dextrose infu- sion also was recorded. The effect of maximum maintenance dextrose infusion on hypoglycemia was analyzed in a trinary fashion. When dex- trose concentration was 10% or less (n = 58), 29 patients (50%, 95% CI: 37-63%) experienced hypoglycemia. When dextrose concentration was 20% – 40% (n = 33), 10 patients (30%, 95% CI: 15-46%) experienced hy- poglycemia. When dextrose concentration was 50% – 70% (n = 75), 21 patients (28%; 95% CI: 18-38%) experienced hypoglycemia. Hypoglyce- mia was more common when 10% maximum maintenance dextrose

Fig. 3. HDI recommendations and usage over time.

Table 1

Demographic data and drug information. Variable

Age, years, median (range) 48 (14-89)

Male gender (%) 100 (50%)

Beta blockers

Metoprolol 40 (20%)

Atenolol 32 (16%)

Propranolol 27 (12%)

Carvedilol 14 (7%)

Metoprolol (XL, SR) 9 (5%)

Labetalol 4 (2%)

Nebivolol 2 (1%)

Propranolol LA 2 (1%)

Sotalol 2 (1%)

Other beta blocker 4 (2%)

calcium channel blockers

Amlodipine 40 (20%)

Diltiazem 22 (11%)

Verapamil 20 (10%)

Diltiazem (ER, XT) 14 (7%)

Verapamil SR 9 (5%)

Nifedipine 5 (3%)

Other calcium channel blocker 4 (2%) Co-ingested drugs

Sedatives/hypnotics	57 (39%)	doses used, our data suggest emergency physicians can feasibly initiate
Antihypertensive (other)	52 (26%)	HDI titrated to a median peak infusion of 8 U/kg/h with the aid of a re-
Ethanol Antidepressants (other) Bupropion/citalopram/venlafaxine Opioids	45 (23%) 39 (20%) 23 (12%) 19 (10%)	gional poison center. Our data also suggest HDI is feasible for physicians providing critical care on Inpatient units as HDI was routinely continued far into the patients’ hospitalizations. Though emergency and critical
Antipsychotics (2nd generation)	17 (9%)	care physicians are generally more familiar with insulin infusions
Antihistamines	13 (7%)	dosed at 0.1 U/kg/h to treat diabetic ketoacidosis [24], in our data phy-
Sympathomimetics Acetaminophen Amiodarone	10 (5%) 8 (4%) 4 (2%)	sicians over time (Fig. 3) were more likely to use insulin at 1 U/kg/h ti- trated up to 10 U/kg/h, or 100 times the typical DKA dose. HDI at this
Alpha-2 agonists	4 (2%)	dose also appears to be safe; though metabolic complications were
Antidepressants (tricyclic)	3 (2%)	common they were easily treatable, and hypoglycemia was less com-
Digoxin Antipsychotics (1st generation)	3 (2%) 2 (1%)	mon when more concentrated dextrose maintenance infusions were used. No deaths occurred that were definitively directly attributable to

infusions were used compared to 20-40% (difference 20%, 95% CI: 0-

40%) or 50-70% dextrose solutions (difference 22%, 95% CI 6-38%). No difference was observed in the proportion of patients experiencing hy- poglycemia when comparing 20-40% solutions to 50-70% solutions (difference 2%, 95% CI: -17-21%).

Selected additional therapies are displayed in Table 3; this table in- cludes therapies received during the patients’ entire hospitalizations and not just the ED. Maintenance dextrose infusion was recorded in 184 (92%) of cases; specific dextrose infusion concentrations are further delineated in Table 3. Median number of hospital days on a dextrose in- fusion was 2 (range 1-10). Median number of dextrose boluses/patient was 1 (range 1-10, n = 48). Concomitant vasopressor use was com- mon, with norepinephrine being the most common vasopressor.

Among cases where cardiac arrest occurred, 33 patients had “cardiac arrest” coded, 16 had “asystole” coded, and 3 had “dysrhythmia (V- tach/V-fib)” coded. After review of all cases to account for duplicate cod- ing, a total of 41 patients experienced a cardiac arrest at some point in their care (see Table 4). A brief account of each case experiencing car- diac arrest is included in Online Supplementary Table 2.

Discussion

Based upon the large number of cases, increased utilization, and high

Clinical effects?

Table 2 NPDS outcomes and clinical effects.
	Entire cohort	BB	DHP	Non-DHP	BB + CCB
	N = 199	N = 88	N= 22	N= 44	N= 45
Outcomes
Death	31 (16%)	13 (15%)	4 (18%)	7 (16%)	7 (16%)
Major effect	116 (58%)	55 (62%)	13 (59%)	23 (52%)	25 (56%)
Moderate effect	49 (25%)	20 (23%)	3 (14%)	14 (32%)	12 (27%)
Unrelated effect	3 (2%)	0	2 (9%)	0	1 (2%)
Acidosis	75 (38%)	32 (36%)	9 (40%)	16 (36%)	18 (40%)
Bradycardia^a	132 (66%)	57 (65%)	8 (36%)	30 (68%)	37 (84%)
Cardiac arrest	41 (21%)	18 (21%)	4 (18%)	8 (18%)	11 (24%)
Coma	45 (23%)	27 (31%)	4 (18%)	5 (11%)	9 (20%)
Conduction disturbance	59 (30%)	21 (24%)	6 (27%)	19 (43%)	13 (29%)
Creatinine increased	57 (29%)	15 (17%)	10 (45%)	13 (30%)	19 (42%)
Drowsy/lethargic	89 (44%)	45 (51%)	8 (36%)	14 (32%)	22 (49%)
Dysrhythmia (other)	22 (11%)	5 (6%)	5 (23%)	5 (11%)	7 (16%)
Electrolyte abnormality	71 (36%)	28 (32%)	10 (45%)	15 (34%)	18 (40%)
Hypoglycemia^b	63 (32%)	36 (41%)	7 (32%)	10 (23%)	10 (22%)
b 70 mg/dL	40 (20%)	24 (27%)	4 (18%)	7 (16%)	5 (11%)
Hypoglycemia treated	23 (12%)	12 (14%)	3 (14%)	3 (7%)	5 (11%)
Hypotension^c	190 (95%)	83 (94%)	19 (86%)	44 (100%)	44 (98%)
Oliguria	36 (18%)	6 (7%)	6 (27%)	11 (25%)	13 (29%)
Potassium (mEq/L) nadir
K = “normal” or N3.5	17 (9%)	7 (8%)	3 (14%)	4 (9%)	3 (7%)
K = 3.1-3.5	19 (10%)	7 (8%)	3 (14%)	4 (9%)	5 (11%)
K = 2.8-3.0	17 (9%)	9 (10%)	0	6 (14%)	2 (4%)
K = 2.5-2.7	11 (6%)	1 (1%)	2 (9%)	6 (14%)	2 (4%)
K = 2.0-2.4	10 (5%)	4 (5%)	0	3 (7%)	3 (7%)
Renal failure	16 (8%)	4 (5%)	3 (14%)	2 (5%)	7 (16%)
Seizures	15 (8%)	11 (13%)	1 (5%)	1 (2%)	2 (4%)

K = potassium concentration (mEq/L).

^a Bradycardia = pulse b60 beats/min.

^b Hypoglycemia = blood glucose b70 mg/dL.

^c Hypotension = systolic blood pressure (SBP) b90 mm Hg OR N15% decrease from baseline SBP.

* Refer to Online Supplementary Table 1 for definitions.

HDI (see Supplemental Table). HDI appears to be gaining acceptance; in 2016 emergency and critical care physicians administered HDI in 85% of cases where it was recommended by our poison center (Fig. 3), much higher than rates reported previously [25].

HDI improves shock via multiple mechanisms. First, it acts as a po- tent inotrope [9], increasing cardiac output by augmenting calcium pro- cessing and consequently improving myocardial contractility [26]. Second, HDI also improves intracellular energy utilization; stressed

Table 3

Selected other therapies.

Therapy	Entire cohort (n = 199)	BB (n = 88)	DHP (n = 22)	Non-DHP (n = 44)	CCB + BB (n = 45)
Antiarrhythmic	28 (14%)	12 (14%)	6 (27%)	6 (13%)	4 (9%)
Atropine	39 (20%)	20 (23%)	1 (5%)	11 (25%)	7 (16%)
Calcium	141 (71%)	43 (49%)	20 (91%)	40 (91%)	38 (84%)
CPR	18 (9%)	11 (13%)	1 (5%)	2 (5%)	4 (9%)
Dextrose	184 (92%)	83 (94%)	21 (95%)	39 (88%)	41 (91%)
D5	8 (4%)	4 (5%)	2 (9%)	1 (2%)	1 (2%)
D10	50 (25%)	23 (26%)	6 (27%)	9 (20%)	12 (27%)
D20	27 (14%)	11 (13%)	1 (5%)	7 (16%)	8 (18%)
D25/D30/D40	3 (2%)	2 (2%)	1 (5%)	3 (7%)	0
D50	58 (29%)	30 (34%)	4 (18%)	11 (25%)	13 (29%)
D70	17 (9%)	7 (8%)	3 (14%)	3 (7%)	4 (9%)
ECMO	6 (3%)	2 (2%)	1 (5%)	1 (2%)	2 (4%)
Glucagon	84 (42%)	41 (47%)	2 (9%)	18 (41%)	23 (51%)
Hemodialysis	21 (10%)	6 (7%)	3 (14%)	3 (7%)	9 (20%)
Intra-aortic balloon pump	7 (4%)	3 (3%)	2 (9%)	1 (2%)	1 (2%)
Intravenous fat emulsion	14 (7%)	6 (7%)	3 (14%)	3 (7%)	2 (4%)
Intubation	133 (67%)	58 (66%)	13 (59%)	27 (61%)	9 (20%)
Left ventricular assist device	1 (0.5%)	0	1 (5%)	0	0
Methylene blue	12 (6%)	1 (1%)	2 (9%)	4 (9%)	5 (9%)
Pacemaker	13 (6%)	1 (1%)	0	6 (14%)	6 (13%)
Vasopressors/inotropes Norepinephrine	116 (58%)	40 (45%)	17 (77%)	27 (61%)	32 (71%)
Dopamine	86 (43%)	28 (32%)	9 (41%)	22 (50%)	27 (60%)
Epinephrine	71 (36%)	31 (35%)	6 (27%)	15 (34%)	19 (42%)
Vasopressin	46 (23%)	13 (15%)	9 (41%)	10 (23%)	14 (31%)
Phenylephrine	30 (15%)	6 (7%)	8 (36%)	7 (16%)	9 (20%)
Dobutamine	10 (5%)	2 (2%)	0	3 (7%)	5 (11%)
Milrinone	1 (0.5%)	0	0	1 (2%)	0

ECMO = extracorporeal membrane oxygenation, DHP = dihydropyridine.

Table 4

Patients experiencing cardiac arrest (n = 41).

Died (n = 31) Survived (n = 10)

1999 [12]. HDI was successfully used as a Rescue therapy in 5 patients; dosing ranged from 0.1 to 1 U/kg/h. Case reports and small case series documenting the success of HDI soon followed, typically dosed at 0.5-

Arrived to hospital in arrest 7 (22%) 6 (60%)

2 U/kg/h [13,14,31]. As animal evidence emerged that larger doses of

Initial vitals (if present) Pulse rate (beats/min)

Systolic blood pressure (mm Hg) Max HDI dose (U/kg/h)

Median (range) 55 (30-96)

80 (40-139)

Median (range) 46 (30-50)

60 (60-95)

HDI may be beneficial [9,32], use of larger doses occurred in human poi- sonings. Holger et al. reported a consecutive series of 12 patients treated with HDI administered as a 1 U/kg IV bolus followed by an infusion at

1 U/kg/h, rapidly titrated up to 10 U/kg/h [11]. Mean HDI dosing in

(median, range) 10 (1-15) 10 (1-18) Overdose type

Beta blocker	12 (39%)	6 (60%)	Page et al. published a multi-center series of 22 patients receiving HDI
Dihydropyridine	4 (13%)	0	for beta-blocker and calcium channel-blocker poisoning; median dose

this series was 8.4 U/kg/h, with 11 of 12 patients surviving. Recently

Intravenous fat emulsion

Non-dihydropyridine	8 (26%)	0 in this series was 150 U/h, however doses ranged up to 1500 U/h
Beta & calcium channel blocker	7 (22%)	⁴ ^(40%) (15 U/kg/h) [16]. Three patients (14%) died in this series; metabolic ad-
Specific drugs
Atenolol	4 (13%)	₀ verse events were common but easily treated.
Amlodipine	6 (19%)	1 (10%) Though animal data show insulin to be a superior therapy in isolated
Diltiazem	7 (22%)	1 (10%) overdose models, HDI is not a panacea. Multiple published case reports
Metoprolol	7 (22%)	4 (40%) describe HDI as insufficient to resuscitate a patient and the patient ei-
Propranolol	7 (22%)	² ^(20%) ther died [33] or was rescued with therapies such as vasopressors, intra-
Verapamil	3 (10%)	2 (20%)
Other beta or calcium channel blocker	3 (10%)	₄ _(40%) venous fat emulsion [34], methylene blue [35], or mechanical devices
Max # vasopressors & inotropes		such as LVADs [36] or ECMO [37]. We observed several such cases in
(median, range)	3 (1-5)	1 (1-4) our study. The inotropic and vasodilatory effects of HDI may also be det-
Other notable therapies a	13 (42%)	1 (10%) rimental in some patients, particularly those with underlying heart con-
Glucagon infusion	8 (26%)	1 (10%)	ditions. For instance, HDI produced worsening hemodynamic effects in
Pacemaker^b	5 (16%)	2 (20%)	a patient with Hypertrophic cardiomyopathy who had overdosed on
Methylene blue	4 (13%)	0	metoprolol, diltiazem, and amiodarone [38]. Inotropes may create out-

Intra-aortic balloon pump	5 (16%)	0
ECMO	2 (6%)	0

^a In all but one case, IFE was specifically noted in the case notes to have no positive effect.

^b All patients with a pacemaker were documented to have obtained capture.

myocardial cells use glucose for energy rather than free fatty acids, their usual preferred energy source [27]. High doses of insulin provide an abundant fuel source for stressed cardiac myocytes. Third, HDI acts as a vasodilator (not a vasopressor) via enhancement of endothelial nitric oxide synthase; this vasodilation improves the microvascular dysfunc- tion associated with cardiogenic shock [28] and results in increased car- diac output [9]. Last, HDI also improves the metabolic dysfunction and subsequent hyperglycemia seen with calcium channel-blocker poison- ing [29]. HDI also appears to exhibit a dose-response relationship with cardiac output; data from swine poisoned with propranolol suggest car- diac output increases 50% over a 6-h resuscitation between 1 and 10 U/kg/h [9].

Severe poisoning from beta-blockers and calcium channel-blockers typically results in hypotension, bradycardia, and cardiogenic shock. In beta-blocker poisoning, these effects occur from antagonism of beta-ad- renergic receptors resulting in decreased activity of phosphokinase A and subsequently less calcium-induced calcium release. Calcium chan- nel-blocker poisoning causes similar effects, though from downstream blockade of L-type Calcium channels. Though all calcium channel- blockers bind L-type calcium channels, dihydropyridine poisoning occa- sionally results in reflex tachycardia as subtle binding differences by dihydropyridines of the ?1c subunit of L-type calcium channels results primarily in vasodilation and subsequent reflex tachycardia. These sub- tle binding difference are frequently lost in large overdoses, however, resulting ultimately in bradycardia [30]. Regardless, in both beta- blocker and calcium channel-blocker poisoning, shock and ultimately death result in large part from decreased inotropy.

Animal models (primarily swine and canines) have clearly demon-

strated HDI to be superior to more tradition therapies such as vasopres- sors or glucagon both in terms of hemodynamic parameters and mortality; [3] this holds true for both calcium channel-blockers [3,4] and beta-blockers [5,6]. Human case experience with HDI, however is more limited. HDI use in humans was first reported by Yuan, et al., in

flow obstruction in hypertrophic cardiomyopathy, or worsen obstruc- tion in existing obstructive cardiomyopathy. As HDI’s primary benefit is as an inotrope, patients with preexisting cardiomyopathy (both ische- mic or obstructive) may receive little benefit from HDI and may even worsen due to HDI’s vasodilatory properties. The metabolic derange- ments associated with HDI should also be taken into account when choosing HDI. For instance, a patient who overdoses on sotalol may present with cardiogenic shock, however sotalol overdoses frequently results in a prolonged QT interval and ventricular dysrhythmias [39]. Hypokalemia from HDI may increase the risk of life-threatening ventric- ular dysrhythmias in sotalol poisoning. In our study only 2 patients overdosed on sotalol, however one developed ventricular tachycardia and cardiac arrest that may have been exacerbated by hypokalemia from HDI. In general, the patient most likely to benefit from HDI is one with a heart that has a baseline normal ejection fraction that now has drug-induced cardiogenic shock.

Metabolic abnormalities were extremely common in our study. In the largest series on HDI to date, Espinoza, et al., reported on 46 patients receiving HDI dosed at 0.5-1 U/kg/h for calcium channel blocker poi- soning without a single episode of hypoglycemia. This study examined only poison center data, however, and likely grossly underestimates metabolic events. Examination of series using hospital data reveals met- abolic events are common, similar to our data. Holger, et al. found 6/12 patients (50%) experienced hypoglycemia and 8/12 experienced hypo- kalemia (b3.0 mEq/L) (67%) [11]. Page, et al. found hypoglycemia (16/ 22, 73%) and hypokalemia (18/22, 82%) to be common as well, though similar to our data they observed hypokalemia to be mostly mild (nadir K 2.5-3.4 mEq/L in 16/22 patients) requiring minimal interven- tion [16]. Our observed rates of hypoglycemia (32%) and hypokalemia (29%) highlight the need for close monitoring of glucose and electro- lytes while utilizing HDI, preferably in a critical care environment. Our data suggest use of concentrated maintenance dextrose solutions, 20% or higher, were associated with fewer events of hypoglycemia. Use of concentrated dextrose solutions also may help minimize the risk of vol- ume overload, a serious complication previously reported in cases where HDI was used [33].

We found hypoglycemia was more common in beta-blocker than

calcium channel-blocker poisoning, consistent with the known endo- crine effects of both drugs. In overdose calcium channel-blockers non- specifically block calcium channels in beta-islet cells, impairing insulin secretion causing hyperglycemia; a trend toward hyperglycemia is

even associated with adverse hemodynamic effects [40]. Beta-blockers, conversely, are reported to cause hypoglycemia via inhibition of glyco- genolysis and gluconeogenesis, though hypoglycemia is quite rare in overdose [41,42]. A theoretical concern exists that refractory hypoglyce- mia could occur in beta-blocker poisoning treated with HDI if the offending beta-blocker and high insulin doses were synergistic. While we found beta-blockers poisonings received more dextrose, hypoglyce- mia was still easily treated suggesting this theoretical synergistic risk is minimal to non-existent.

Limitations

This study has several limitations. First, it has the typical limitations of a retrospective chart review, including inability to properly assess the frequency of complications because of lack of a comparative group, con- venience sampling inherent to poison center data, and the possibility of unmeasured bias. For example, we were unable to account for the ef- fects of vasopressors or other therapies. Frequently our poison center was contacted regarding HDI dosing after vasopressors had already been initiated, as such we were unable to assess if starting HDI or vaso- pressors first had any effect on clinical outcomes.

Second, as a poison center study, the usual limitations of a retrospec-

tive study are compounded. Multiple studies have shown poison center data have limited clinical utility [43,44]. Poison center data may be in- complete, or even inaccurate [45]. The coding of our cardiac arrest cases, for example, demonstrates this limitation: some cases had asystole coded but not cardiac arrest. Poison center case notes are not complete medical records; vital signs and laboratory values are fre- quently not recorded and are subject to the reporting bias and mistakes of caregivers at the bedside reporting information via phone. This makes assessing for complications related to HDI, such as hypoglycemia and hypokalemia, problematic, as these events are likely under-reported. These limitations notwithstanding, poison center data also have strengths. Though HDI has become standard of care in many guidelines, there are still a limited number of cases reported in the literature. A poi- son center-based study allowed us to gather a large sample size on a rel- atively infrequently described therapy; our study more than doubles the number of reported cases of HDI in the literature. It is also possible our poison center recorded adverse events more commonly than other centers; for example we detected a substantial number of adverse events in contrast to Espinoza, et al., another study on HDI that utilized poison center data.

Third, the insensitive nature of poison center data makes assessing for more rare complications difficult or even impossible. While hypokalemia is well-described with HDI, hypomagnesemia and hypophosphatemia may also occur [13]. Magnesium and phosphorus concentrations are rarely recorded in poison center case notes; thus we were unable to assess for derangements of these electrolytes. Poison center data also do not rou- tinely assess for certain cardiovascular and pulmonary complications such as volume overload or Acute respiratory distress syndrome . Pa- tients on HDI typically receive multiple infusions, putting them at risk for volume overload, pulmonary edema, and ARDS [33]. Though the use of concentrated maintenance dextrose infusions such as D50 and concen- trating insulin to 10 or even 16 U/mL [46] (rather than the typical 1 U/mL used in diabetic ketoacidosis [24]) may reduce the risk of volume over- load, the limitations of poison center data made assessing the true fre- quency of these complications infeasible. However, even series utilizing exclusively hospital data [11,16], one of which was prospective [13], failed to address these particular complications of HDI. These gaps remain in the literature.

Fourth, the intermittent nature of poison center data collection also makes assessing for titration-dependent adverse effects difficult. For in- stance, because HDI is titrated rapidly, we were unable to assess if higher doses of HDI were associated with more episodes of hypoglyce- mia. Laboratory data on pigs poisoned with propranolol and treated with HDI, however suggest glucose utilization does not increase

substantially from 1 to 10 U/kg/h of HDI despite an increase in cardiac output [9]. Page, et al., also found no relationship between blood glucose and maximum HDI infusion rate in human cases [16].

Fifth, responsible poisons in all cases were confirmed only by his- tory; in no cases were urine drug screen results available to confirm the exposure. Therefore our patients’ exposures (Table 1) are more properly thought of as purported exposures. While drug screening was not available in our patients, we believe this limitation was minimized by the clinical nature of beta-blocker and calcium channel-blocker poi- soning. Based on the vital signs exhibited by our patients on initial pre- sentation, coupled with the numerous critical care modalities required, we believe these exposures were very likely true poisonings.

Lastly, selection bias may have played a role in the perceived effec- tiveness of HDI. As a regional poison center covering 3 states, clinicians may have been more likely to consult the poison center for cases refrac- tory to standard therapies, or for patients who were more ill. This selec- tion bias may have made HDI appear less effective. Examination of previous literature suggests this may be the case. Though beta-blockers are the most common cause of cardiovascular poisoning [1], calcium channel-blockers [21] – particularly the non-dihydropyridines diltia- zem and verapamil [47] – carry the highest case mortality rates among cardiovascular drugs in poisoning. As an example, in one of the largest series on verapamil poisoning, of 65 patients admitted to an ICU, 11% experienced cardiac arrest with 8% ultimately dying [48]. In our study, 21% of patients experienced cardiac arrest, with 6.5% of pa- tients arriving to the hospital already in arrest. Though our overall mor- tality rate was 16%, 24% of patients experiencing cardiac arrest ultimately survived, including 6 of 13 patients who arrived to hospital already in arrest. It is possible our higher doses of insulin may have mit- igated the effects of a reporting bias toward sicker patients; randomized trials are needed in the future to address these confounders. It is note- worthy our observed mortality rate of 16% was remarkably similar to the recently published data by Page, et al. (14%).

Conclusions

In summary, initiation by emergency physicians of HDI started at 1 U/kg/h titrated up to 10 U/kg/h to treat shock from beta-blocker and calcium channel-blocker poisoning, in consultation with a regional poi- son center, was feasible and safe. No deaths were clearly attributable to HDI or its metabolic complications. Hypoglycemia and hypokalemia were common, however hypoglycemia was easily treatable and oc- curred less frequently when more concentrated maintenance dextrose infusions were used. As iatrogenic metabolic abnormalities were ex- pectedly common, HDI should be utilized when critical care resources are readily available. A randomized trial in humans is needed to clarify if HDI is superior to other therapies.

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2018.02.004.

Conflicts of interest/declarations

No authors report any relevant conflicts of interest.

References

AJEM