Impact of Intraosseous Versus Intravenous Resuscitation During In-Hospital Cardiac Arrest: A Retrospective Study

Abstract

Aim:

To compare outcomes between Intraosseous (IO) and peripheral intravenous (PIV) injection during in-hospital cardiac arrest (IHCA) and examine its utility in individuals with obesity.

Methods:

We performed a retrospective cohort analysis of adult, atraumatic IHCA at a single tertiary care center. Subjects were classified as either IO or PIV resuscitation. The primary outcome of interest was survival to hospital discharge. The secondary outcomes of interest were survival with favourable neurologic status, rates-of-ROSC (ROR) and time-to-ROSC (TTR). Subgroup analysis among patients with BMI ≥ 30kg/m² was performed.

Results:

Complete data were available for 1852 subjects, 1039 of whom met eligibility criteria. A total of 832 were resuscitated via PIV route and 207 via IO route. Use of IO compared to PIV was associated with lower overall survival to hospital discharge (20.8% vs 28.4% p=0.03), lower rates of survival with favourable neurologic status (18.4% vs 25.2% p=0.04), lower ROR (72.2% vs 80.7%) and longer TTR (12:38 min vs 9:01 min). After multivariate adjustment there was no significant differences between IO and PIV in rates of survival to discharge (OR 0.71, 95% CI 0.47 – 1.06, p=0.09) or rates of survival with favourable neurologic status (OR 0.74, 95% CI 0.49 – 1.13, p=0.16). The ROR and TTR remained significantly worse in the IO group. Subgroup analysis of patients with BMI ≥ 30kg/m² identified no benefit or harm with use of IO compared to PIV.

Conclusion:

Intraosseous medication delivery is associated with inferior rates-of-ROSC and longer times-to-ROSC compared to PIV, but no differences in overall survival to hospital discharge or survival with favourable neurologic status during IHCA.

INTRODUCTION

Intraosseous (IO) delivery of emergency medications during cardiac arrest provides safe and rapid access when intravenous (IV) injection is not available^1–4. The American Heart Association and European Resuscitation Council present IO access as an acceptable alternative during resuscitation when peripheral intravenous (PIV) access is difficult or not available^5,6. Despite these recommendations the effectiveness of medication delivery using IO access during states of cardiac arrest is not well characterized. Existing data is mostly limited to pediatric populations, animal models or the pre-hospital setting, and significant knowledge gaps still remain^7,8. Recent trials comparing IO to PIV show mixed results but suggest use of IO delivery during resuscitation for Out-of-Hospital Cardiac Arrest (OHCA) is associated with lower rates of return of spontaneous circulation (ROSC), worse neurologic outcomes and no long-term survival benefit as compared with standard IV resuscitation^7–13. However, outcomes using IO routes of medication delivery during In-Hospital Cardiac Arrest (IHCA) are not known.

One population where IO delivery may have a particular advantage is in individuals with obesity. Prior studies have identified a Body Mass Index (BMI) ≥ 30kg/m² as an independent predictor of difficult PIV access^14–16. The use of IO cannulation in this population has been shown to be both safe and effective as compared to standard PIV methods^14,15. However, no studies have examined the utility of IO resuscitation during IHCA in this unique and growing patient population. We hypothesized that use of IO in individuals with obesity would improve clinical outcomes compared to nonobese individuals.

Our primary study aim was to compare outcomes between IO and PIV injection during IHCA in a general inpatient population with a secondary aim to compare effectiveness in a subgroup of individuals with obesity.

METHODS

Study Design, Population, Setting

We performed a retrospective cohort analysis of IHCA between March 2013 and September 2018 at a single tertiary care center in Birmingham, Alabama. We included adults of age ≥18 years who had atraumatic IHCA. Medical and surgical patients in intensive care, acute care and procedural areas were included.

We excluded individuals who had existing central venous access at the time of IHCA, were on extracorporeal membrane oxygenation (ECMO), or had mechanical circulatory support devices such as a ventricular assist device (VAD) or an intra-aortic balloon pump (IABP). We additionally excluded cases when cardiac arrest occurred in the emergency department (ED), as our data did not allow us to distinguish those patients who were undergoing cardiopulmonary resuscitation upon arrival versus those who had cardiac arrest after arrival to the ED (Figure 1).

Figure 1.

Flowchart of patient inclusion and analysis.

CVL = central venous line; ECMO = extracorporeal membrane oxygenation; ED = emergency department; IABP = intra-aortic balloon pump; IHCA = In-hospital cardiac arrest; IO = intraosseous; PICC = peripherally inserted central catheter; PIV = peripheral intravenous; VAD = ventricular assist device.

All cardiac arrest events in our hospital are responded to by our Medical Emergency Team (MET). Per institutional protocol, and in line with ACLS Guidelines, when patients had an in-hospital cardiac arrest event that was responded to by our MET, the default was to use any pre-existing PIV for administration of resuscitation meds.^5–6 If no functioning PIV existed at time of cardiac arrest, providers on-site (e.g., primary nurses or providers) would attempt PIV placement until arrival of MET. If PIV access still was not obtained by the time of arrival of MET, this would be classified as failed attempt at PIV and an IO would be placed as alternative access, without further attempts at PIV. Thus, if available, PIV was the preferred method of medication delivery. IO access was only obtained if no PIV was immediately available upon arrival of the MET nurses, if PIV failed at any point during the code event, or at the discretion of the attending physician. Default location of IO placement was the tibia. Humeral placement was only obtained if there was failure or contraindication at tibial location. Sternal IOs are not placed by the MET at our hospital. When obtained, IO vascular access was performed by trained personnel on the MET using the automated EZ-IO® device. Personnel on this team are trained in both tibial and humeral device placement during onboarding by industry professionals. They must also re-certify yearly using both online modules and simulated code events in the cadaver lab.

Data Collection

Data collected from code events included routes of vascular access, time to first epinephrine administration, total time of cardiac arrest until first ROSC, witnessed status of arrest, and initial cardiac rhythm upon initiation of Advanced Cardiac Life Support (ACLS). Chart review of electronic medical record (EMR) was utilized to obtain demographic information for patients including age, sex and BMI, in addition to specific comorbidities (Table 1). We also utilized EMR to obtain information on survival to hospital discharge and post-arrest neurologic status at discharge based on scoring of the Cerebral Performance Score (CPC). The CPC score was calculated by MET personnel.

Table 1.

Demographic and Clinical Characteristics of Study Population

	All Patients	PIV	IO	PIV vs IO
Characteristic	n = 1039	n = 832	n = 207	p-value
Age, years
All	60.5 ± 8.8	60.4 ± 15.9	61.2 ± 14.6	0.09
Obese		59.1 ± 14.4	60.7 ± 13.7	0.48
Sex, n (%M)
All	638 (61.4)	518 (62.3)	120 (58.0)	0.27
Obese		188 (58.7)	41 (47.1)	0.15
Obese, n (%)	407 (39.1)	320 (38.5)	87 (42.0)	0.35
BMI, kg/m2
All	29.2 ± 15.6	29.2 ± 8.9	29.5 ± 4	0.64
Obese		37.5 ± 8.2	37.3 ± 6.7	0.50
Co-Morbidities
DM, n (%)
All	376 (36.1)	293 (35.2)	82 (39.6)	0.24
Obese		149 (46.5)	45 (51.7)	0.39
CKD/ESRD, n (%) *
All	395 (38.0)	312 (37.4)	84 (40.6)	0.39
Obese		133 (41.5)	43 (49.4)	0.19
CAD, n (%)
All	244 (23.5)	198 (23.7)	47 (22.7)	0.77
Obese		74 (23.1)	19 (21.8)	0.80
CHF, n (%) †
All	301 (28.9)	250 (30.0)	51 (24.5)	0.13
Obese		98 (30.6)	22 (25.3)	0.34
Cirrhosis, n (%)
All	87 (8.4)	75 (9.0)	12 (5.8)	0.14
Obese		27 (8.4)	3 (3.4)	0.11
Cardiac Arrest Data
Non-shockable rhythm, n (%) ‡
All	825 (79.4)	659 (79.2)	167 (80.4)	0.58
Obese		252 (78.7)	68 (78.2)	0.90
Witnessed Arrest, n (%)
All	898 (85.3)	720 (86.5)	166 (80.2)	0.02
Obese		278 (86.9)	65 (74.7)	<0.01
Time to Epinephrine (min)
All	2:38 ± 2:32	2:31 ± 2:29	3:06 ± 3:02	<0.01
Obese		2:21 ± 2:02	3:17 ± 2:51	<0.01

Baseline demographic and clinical characteristics between the two populations. Those resuscitated via IO were more likely to have an unwitnessed event and had statistically significant longer time-to-epinephrine compared to the PIV cohort.

BMI = body mass index; CAD = coronary artery disease; CHF = congestive heart failure; CKD = chronic kidney disease; DM = diabetes mellitus; ESRD = end stage renal disease; IO = intraosseous; PIV = peripheral Intravenous

^*Patients with CKD stage 2–5 and ESRD on dialysis were included in this population.

^†Patients with either reduced or preserved ejection fraction were included in this population.

^‡Non-shockable rhythm was defined as initial rhythm of cardiac arrest being either pulseless electrical activity (PEA), asystole, or bradycardia.

Definitions

Patients were placed in the PIV group if they had pre-existing PIV access at time of cardiac arrest, or if they had PIV access placed after time of cardiac arrest. Patients were placed in the IO group if they had IO access placed at any time during the cardiac arrest event, but not after ROSC. Initial presenting rhythm was classified as either shockable (ventricular fibrillation or ventricular tachycardia) versus non-shockable (asystole, pulseless electrical activity, or bradycardia). Time-to-epinephrine administration and time-to-ROSC were calculated in minutes starting from initiation of ACLS until first epinephrine dose and first ROSC, respectively. A BMI was calculated per the height (in meters) and weight (in kilograms) in the medical record at time of code event. Patients were then placed into the “obese” category if they had a BMI ≥ 30kg/m². Survival with favourable neurologic status was defined as a CPC ≤ 2 at hospital discharge.

Outcomes

The primary outcome of interest was survival to hospital discharge. The secondary outcomes of interest were survival to hospital discharge with favourable neurologic status, rates-of-ROSC (ROR) and time-to-ROSC (TTR). Subgroup analysis among patients with BMI ≥ 30kg/m² was performed.

Statistical Analysis

Descriptive characteristics are reported as means and compared using the Student’s t-test for continuous variables or Chi-Squared for discrete variables as appropriate. Standard Deviation is reported where appropriate. Survival to hospital discharge, survival to discharge with favourable neurologic status, ROR and mean TTR were calculated in PIV and IO groups and compared using multivariable logistic regression analysis in SPSS® software, version 25. Adjustments were made for age, sex, initial rhythm, witnessed status of code event, and time-to-epinephrine. Statistical significance was defined at a p-value of <0.05. Odds Ratios with 95% confidence intervals were calculated.

Subgroup Analysis

Analysis was conducted using the Breslow-Day Test in order to test whether the strength of association between I/O and outcomes of interest varied by obesity status. We then formulated the same test as above in a logistic regression model with an interaction term between I/O and obesity, adding the covariates of age, gender, initial rhythm, witnessed status and time-to-epinephrine.

RESULTS

Complete data were available for 1852 subjects between March 2013 and September 2018. Of these subjects 1039 met eligibility criteria. A total of 832 were resuscitated via PIV route and 207 via IO route. Baseline demographics and comorbidities were well matched across the two groups. However, patients resuscitated via IO were less likely to have a witnessed arrest (80.2% vs 86.5%) and had longer time-to-epinephrine on average (3:06 min vs 2:31 min) compared to those with PIV (Table 1).

Primary Outcome

Use of IO compared to PIV was associated with lower overall survival to hospital discharge (20.8% vs 28.4%; p=0.03) (Table 2). After adjustment, there was no longer a statistically significant difference between the two groups (OR 0.71, [0.47 – 1.06], p=0.09) (Table 2).

Table 2.

Primary and Secondary Outcomes – All Patients

	All (n=1039)	PIV (n = 832)	IO (n = 207)	OR (95% CI)	p-Value
Primary Outcome
Survival to hospital discharge, n (%)	279 (26.9)	236 (28.4)	43 (20.8)	0.71 (0.47 – 1.06)	0.09
Secondary Outcomes
Survival with favourable neurologic status, n (%)	248 (23.9)	210 (25.2)	38 (18.4)	0.74 (0.49 – 1.13)	0.16
ROR, n (%)	821 (79.0)	671 (80.7)	150 (72.2)	0.68 (0.47 – 0.98)	0.04
TTR, min ± SD	9:40 ± 8:56	9:01 ± 8:40	12:38 ± 10:10	--	<0.001

Primary and secondary outcomes for All Patients. Odds ratios and p-values are adjusted for age, gender, initial rhythm, witnessed status and time-to-epinephrine.

IO = intraosseous; PIV = peripheral Intravenous; ROR = Rate-of-ROSC; TTR = Time-to-ROSC.

Secondary Outcomes

Use of IO compared to standard PIV access was associated with lower rates of survival to discharge with favourable neurologic status (18.4% vs. 25.2%; p=0.04), lower ROR (72.2% vs 80.7%, p=0.01) and longer TTR (12:38 min vs 9:01 min, p <0.001). After adjustment, there was no longer a difference in survival with favourable neurologic status between IO and PIV (OR 0.74, 0.49 – 1.13, p=0.16). However, ROR (OR 0.68, 95% CI 0.47 – 0.98), p=0.04) and TTR remained significantly worse in the IO group (Table 2).

Subgroup Analysis

Subgroup analysis of patients with BMI ≥ 30kg/m² identified no statistically significant benefit or harm with use of IO compared to PIV in terms of overall survival to hospital discharge (18.4% vs. 23.4%; OR 0.91 [0.48 – 1.73] p= 0.77), survival with favourable neurologic status (15.9% vs 21.3%; OR 0.94 [ 0.48 – 1.86], p=0.86), or ROR (73.6% vs 81.6%, OR 0.73 [0.41 – 1.31], p=0.30). We did identify an association between IO and longer TTR as compared with PIV (9:56 min vs 13:58 min, p<0.01) (Table 2).

DISCUSSION

The aim of our study was to compare outcomes between IO and PIV administration of medications during IHCA. We present evidence suggesting an association between IO medication delivery and inferior rates-of-ROSC and times-to-ROSC among subjects with IHCA compared to standard PIV methods. After adjusting for potential confounders, IO use was not independently associated with differences in overall survival or survival with favourable neurologic status as compared with PIV. Additionally, the hypothesis that IO delivery would assist resuscitation efforts in patients who had BMI ≥ 30kg/m² was not supported. IO resuscitation was associated with longer TTR compared to PIV but otherwise provided no identifiable benefit or harm among a sub population of individuals with obesity.

There are a number of possible explanations to consider for our observed outcomes. These findings could suggest a true physiologic difference in drug administration when using IO versus PIV. Various models suggest bioequivalence in terms of maximum serum concentration and time to peak effect when drugs are injected through IO and PIV routes^17,18. These studies were mostly done in animal models and/or in controlled settings. It is possible such results do not apply to vascular beds in humans during states of cardiac arrest. Drug administration via IO routes may result in inferior pharmacokinetics due to poorly perfused sinusoids within bone or absorption of drug itself into bone. Additionally, data suggest that anatomic site of IO access matters, both due to the type of bone (e.g. those with red vs yellow marrow) and proximity to central venous system^18,19. For example, humeral and sternal routes seem to have superior pharmacokinetic performance and greater flow rates compared to tibial route, despite tibial placement being one of the most common^19,20. In contrast, sternal and humeral IO placement seem to have lower initial success rates of placement^7,21. What could be gained with superior pharmacokinetics may simply be counterbalanced by more difficult placement. If true, providers may need to be more discerning in regard to IO site placement or perhaps alter drug dosage or flush volume administered when using different IO routes. Finally, it may be that a focus on obtaining access via IO route hinders quality of, or time spent on, those aspects we know help in achieving ROSC during cardiac arrest – namely, good CPR with limited interruptions, early defibrillation for shockable rhythms and early identification of reversible causes of cardiac arrest. Taken together our data seems to suggest that institutions should prioritize acquiring and maintaining venous access during hospitalization prior to unforeseen IHCA events. Whether our results represent an actual biologic difference in medication delivery or patient level confounders cannot be determined in this retrospective study and further work is needed.

Lower rates-of-ROSC and longer times-to-ROSC in the IO group did not correspond to differences in survival or neurologic outcomes as compared to PIV. As mentioned, recent studies investigating these outcomes during OHCA have been mixed^7–13. Additionally, use of epinephrine itself during cardiac arrest has been somewhat controversial, especially with respect to long term survival and neurologic outcomes^22–27. Thus, the optimal route, dosing and timing of epinephrine delivery is not yet determined. One study by Donnino et. al. suggests earlier administration of epinephrine for non-shockable rhythms during IHCA results in higher probability of ROSC, survival, and neurologically intact survival²⁸. However, the effect could not be narrowed down to a precise time, meaning “early” is still a relative term in the ACLS literature. In our study the difference in time-to-epinephrine between IO and PIV was statistically significant (about 30 seconds longer in the IO group). However, the absolute difference was relatively small, and its clinical significance is unclear. One explanation for why survival and neurologic outcomes did not differ could be that although epinephrine administration was delayed, it was still within a critical “early” time period to allow for the beneficial effects. Alternatively, given the overall trend towards decreased survival and decreased survival with favourable neurologic status in the IO group, this study may simply have been underpowered to detect these subtle differences.

We hypothesized that IO placement would result in improved resuscitation outcomes in individuals with obesity, but this was not observed. As with our main cohort, use of IO compared to PIV in those with obesity was associated with lower rates of survival to hospital discharge, less favourable neurologic outcomes, and lower rates-of-ROSC, although not meeting statistical significance. Use of IO in this cohort did result in significantly longer times-to-ROSC even after adjustment. One explanation for this observation could be that obesity is associated with more difficult IO catheter placement, as evidenced by longer time-to epinephrine in this cohort. Although first-attempt success rates of IV and tibial IO placement seem to be relatively comparable in recent studies, no differentiation based on BMI was available^7,21. Other data show that anatomic barriers, such as poorly defined landmarks and increased subcutaneous tissue may lead to improper placement in those with obesity^15,29. Some of these barriers may be overcome through use of ultrasound or longer IO needle length, though study populations did not include cardiac arrest patients, and such strategies may not be feasible in those with IHCA. With lack of randomization, we cannot rule out the possibility that obesity may simply be an independent risk factor for worse outcomes during cardiac arrest regardless of route of medication delivery²⁹.

LIMITATIONS

Our study had several limitations. The primary concerns are related to the observational nature and include selection bias and confounding. First, we were not able to completely characterize or adjust for all potential confounders between groups. Specifically, we had no data to compare severity of illness between groups nor did we adjust for all comorbidities. Additionally, we had incomplete data related to features of the cardiac arrest itself, such as which drugs were administered, management of airway, time-to-defibrillation for shockable rhythms and nature of post arrest care (e.g., rates of targeted temperature management in each group). Similarly, we could not define characteristics related to placement of catheters, such as reason for use of IO, success rates of initial attempts, nor location of IO catheter (e.g., tibial vs. humeral). Each of these factors could lead to confounding and may independently correlate with differing outcomes during cardiac arrest^7,19–21,30. Therefore, we can infer no direct causality, and worse outcomes related to use of IO catheter may be related to undefined patient-level characteristics that are independent risk factors for poor outcomes rather than use of IO cannulation itself. Second, our data set was derived from a single, urban, tertiary care center in the United States which limits the generalizability of our results. Third, grouping of patients into PIV or IO was based on whether an IO was placed at any time during cardiac arrest, but we could not differentiate between those with pre-existing PIV access at time of IHCA and those who had PIV placed after IHCA. As a result, our study is not able to directly compare the entirety of the resuscitation process, but rather primarily reflects the pharmacokinetics of IO vs PIV medication delivery. Similarly, when IO access was placed, we had no subsequent information as to how this access was used during ACLS efforts, however, our institutional protocol is to use IO only if PIV is unavailable which suggests that IO was used as the primary means of medication delivery when present. Inevitably this means some patients may have been misclassified within groups but given the relatively large size of our study the effect of such misclassification on outcomes is likely to be small. Fourth, although total sample size was relatively large, the number of patients in the IO group and the subgroup of individuals with obesity was relatively small, thus we may not have had the power to capture true inter-group differences in outcomes.

Given these limitations we do not suggest IO access should be abandoned altogether. Rather that its current utilization as a back-up to PIV in emergency scenarios is appropriate, regardless of BMI, until further studies to directly compare the two modalities can be performed.

CONCLUSION

Use of IO compared to PIV during IHCA was associated with worse rates-of-ROSC and times-to-ROSC but no difference in overall survival or survival with favourable neurologic status. Our data suggests IO placement should remain an alternative during resuscitation efforts after failure of PIV, regardless of BMI, and considered a bridge to more definitive access routes until further randomized studies to investigate the effective clinical use of IO access can be performed.

Table 3.

Primary and Secondary Outcomes – Obese Cohort

	All (n=407)	PIV (n = 320)	IO (n = 87)	OR (95% CI)	p-Value
Primary Outcome
Survival to hospital discharge, n (%)	91 (22.3)	75 (23.4)	16 (18.4)	0.91 (0.48 – 1.73)	0.77
Secondary Outcomes
Survival with favourable neurologic status, n (%)	81 (19.9)	68 (21.3)	13 (15.9)	0.94 (0.48 – 1.86)	0.86
ROR, n (%)	325 (79.8)	261 (81.6)	64 (73.6)	0.73 (0.41 – 1.31)	0.30
TTR, min ± SD	10:44 ± 9:58	9:56 ± 9:22	13:58 ± 11:38	--	<0.01

Primary and secondary outcomes for a subgroup of individuals with obesity, defined as BMI ≥ 30kg/m² Odds ratios and p-values are adjusted for age, gender, initial rhythm, witnessed status and time-to-epinephrine.

IO = intraosseous; PIV = peripheral Intravenous; ROR = Rate-of-ROSC; TTR = Time-to-ROSC.