Association between intravenous fluid bolus and biomarker trajectory during prehospital care

Abstract

Background:

Patients with acute illness who receive intravenous (IV) fluids prior to hospital arrival may have a lower in-hospital mortality. To better understand whether this is a direct treatment effect or epiphenomenon of downstream care, we tested the association between a prehospital fluid bolus and the change in inflammatory cytokines measured at prehospital and emergency department timepoints in a sample of nontrauma, non-cardiac arrest patients at risk for critical illness.

Methods:

In a prospective cohort study, we screened 4,013 non-trauma, non-cardiac arrest encounters transported by City of Pittsburgh Emergency Medical Services (EMS) to 2 hospitals from August 2013 to February 2014. In 345 patients, we measured prehospital biomarkers (IL-6, IL-10, and TNF) at 2 time points: the time of prehospital IV access placement by EMS and at ED arrival. We determined the relative change for marker X as: ([X_ED – X_EMS] /X_EMS). We determined the risk-adjusted association between prehospital IV fluid bolus and relative change for each marker using multivariable linear regression.

Results:

Among 345 patients, 88 (26%) received a prehospital IV fluid bolus and 257 (74%) did not. Compared to patients who did not receive prehospital fluids, median prehospital IL-6 was greater initially in subjects receiving a prehospital IV fluid bolus (22.3 [IQR 6.4 – 113] vs. 11.5 [IQR 5.5 - 47.6]). Prehospital IL-10 and TNF were similar in both groups (IL-10: 3.5 [IQR 2.2 - 25.6] vs. 3.0 [IQR 1.9 - 9.0]; TNF: 7.5 [IQR 6.4 - 10.4] vs. 6.9 [IQR 6.0 - 8.3]). After adjustment for demographics, illness severity, and prehospital transport time, we observed a relative decrease in IL-6 at hospital arrival in those receiving a prehospital fluid bolus (adjusted β=−10.0, 95%CI: −19.4, −0.6, p= 0.04), but we did not detect a significant change in IL-10 (p=0.34) or TNF (p=0.53).

Conclusions:

Among non-trauma, non-cardiac arrest patients at risk for critical illness, a prehospital IV fluid bolus was associated with a relative decrease in IL-6, but not IL-10 or TNF.

Treatment with intravenous (IV) fluid is first-line therapy for shock. However, the optimal timing, volume and type of fluid remains elusive and may affect outcomes.^1,2,3 For example, early administration of at least 30 mL/kg IV crystalloid for the resuscitation of sepsis patients is recommended,⁴ and several Emergency Department (ED) studies have demonstrated benefit and absence of harm from IV resuscitation bundles incorporating these recommendations.^5–7 Conversely, observational studies have demonstrated harm associated with administration of IV fluids in critically ill patients.^8–10 Nonetheless, recent observational data suggests that fluid administration in the prehospital phase may improve mortality.^11–13 However, the mechanism and true prehospital treatment effects are unknown.

There are many potential mechanisms for the observed treatment effect of prehospital fluid in patients at risk for critical illness. For example, prehospital fluid may directly modify patient inflammatory response and reduce organ injury. On the other hand, the association of prehospital fluid and improved mortality may represent unmeasured factors, such as paramedic skill or judgment. Alternatively, the observed treatment effect may result from more expeditious or complete care in the field and in the ED. Further elucidation of these possibilities may lead to greater focus on direct treatment by prehospital clinicians with IV fluid in randomized trials.

To address this knowledge gap, we tested the association between a prehospital IV fluid bolus and the trajectory of serum levels of TNF, IL-6 and IL-10 in a secondary analysis of a prospective cohort of non-trauma, non-arrest prehospital patients at risk for critical illness.

Methods

The Institutional Review Board of the University of Pittsburgh approved the study (IRB# PR012090482).

Study design

We performed a secondary analysis of a prospective cohort of City of Pittsburgh Bureau of EMS transports from 2013-2014. Pittsburgh has a resident population of 304,000 persons in 2015, with EMS responses dispatched from a central 911 call center. During the study period, EMS responses were single tier, with two paramedics on each ambulance. Medical care follows statewide EMS protocols, supplemented by direct medical oversight from emergency physicians via radio or in-person at the request of EMS providers. During the study period, care for medical hypotension was directed by the Shock/Systemic Inflammatory Response Syndrome (SIRS) Pennsylvania statewide ALS protocol.¹⁴ Per protocol, placement of IV catheter was at the provider’s discretion based on the patient’s clinical condition. For patients in whom systolic blood pressure is less than 100 mmHg, 500 mL boluses of normal saline are to be administered until hypotension has resolved or until 2,000 mL is given. Persistent hypotension despite IV fluid resuscitation requires administration of vasopressors (dopamine or dobutamine) or communication with Medical Command. UPMC is a health system that provides medical oversight for Pittsburgh EMS and operates 5 of the 7 EDs in the city.

Patient selection

The original cohort was the Pittsburgh Prehospital LINking Evaluation (PIPeLINE) study (NIH/NIGMS GM104022), which aimed to identify patients at risk for developing in-hospital critical illness during the prehospital interval. PIPeLINE included eligible adults (age ≥ 18 years) transported to 2 tertiary UPMC hospital EDs by Pittsburgh EMS. We excluded transports for cardiac arrest, trauma, burns, ST-elevation myocardial infarction, known pregnancy, or patients who refused participation. For this secondary analysis, we also excluded patients without successful IV access, intra-facility transports, and duplicate encounters.

Clinical data collection

We gathered EMS incident reports for each patient encounter from a computerized database (emsCharts, Inc; Warrendale, PA). Data included dispatch and response times, demographics, first recorded prehospital vital signs, and Glasgow Coma Scale (GCS). Data were used to calculate a validated prehospital risk score, a multivariable score that predicts development of in-hospital critical illness using prehospital clinical variables including age, systolic blood pressure, heart rate, pulse oximetry and GCS.^15,16 A Computer Aided Dispatch program integrated with 9-1-1 dispatch provided date and time data for medical contact and hospital arrival times.

We linked prehospital data with electronic health record data at UPMC hospitals using patient identifiers collected in the ED by research assistants (Cerner Powerchart; Cerner, Kansas City, MO) From the inpatient record, we determined intensive care unit (ICU) admission, hospital and ICU length of stay (LOS), and in-hospital mortality. International Classification of Diseases, 9^th Revision (ICD-9) codes determined the Elixhauser Comorbidity Index.¹⁷

Prehospital and ED biomarker measurement

Prehospital blood samples (20cc) were collected at 2 time points, i.) by prehospital providers while in the field and ii.) by ED staff upon arrival. The first sampling occurred when IV access was established by EMS, and samples were transported on ice at 0°C (Instant Cold Pack, Kimberly-Clark) for processing by the research team on arrival to the ED. Upon arrival in the ED, paramedics placed patients directly in a treatment room, where ED staff completed blood sampling during initial assessment. All blood sampling used pyrogen-free vacuum-evacuated heparin-, citrate-, and EDTA-containing Vacutainer® tubes. Research staff centrifuged samples at 4°C; separated plasma into aliquots; and stored samples frozen at −80°C until assay. We batch analyzed after a single freeze-thaw cycle. IL-6, IL-10 and TNF using the Bio-plex pro human cytokines assay (Bio-Rad Laboratories, Hercules, CA). All assays were measured on the Bio-Rad Bio-Plex® 200 System (Bio-Rad Laboratories, Hercules, CA). Blood processing and analysis were completed by Clinical Research, Investigation, and Systems Modeling of Acute Illness (CRISMA) Clinical Research Biospecimen Core. Research assistants were blinded to fluid group. Accepted upper limits of normal were: IL-6 5.9 pg/mL, IL-10 9.1 pg/mL and TNF 8.1 pg/mL, per the manufacturer’s specifications.¹⁸

Variables

The primary exposure was a prehospital IV fluid bolus (yes/no) of at least 500 mL 0.9% normal saline solution as documented electronically in emsCharts. The primary outcome was the relative change in candidate biomarkers (IL-6, IL-10, TNF) between prehospital and ED measurements defined by ([X_ED – X_EMS] / X_EMS) for each marker.

We measured a priori defined covariates identified as potential confounders that may be associated with a change in biomarkers and IV fluid bolus administration based on prior studies.^18–20 These covariates included demographics such as age, sex, and race (black, white, other); a prehospital sepsis risk score calculated using first prehospital vital signs (range: 0 to 8 points), and time interval from first medical contact (FMC) to ED arrival.

Statistical analysis

We described demographic and clinical characteristics of patients categorized by those who did and did not receive a prehospital IV fluid bolus. Continuous variables are reported as mean (SD) or median (IQR); categorical variables are reported as frequencies or percentages.

The frequency of missing data was determined for each clinical variable. Missing data was assumed to be missing at random, and we used a flexible multivariable imputation procedure of multiple chained regression equations (multiple imputation by chained equations, i.e. MICE) to generate values for all missing data using the observed data for all patients.²¹ We included all covariates and our primary outcome in the imputation procedure. The imputation procedure generated 10 independent datasets.

We modeled the association between change in candidate biomarkers (IL-6, IL-10, TNF) and administration of a prehospital IV fluid bolus using multivariable linear regression adjusted for covariates specified as potential confounders on imputed datasets, combining estimates using Rubin’s rules.²²

Sensitivity Analysis

We performed a variety of sensitivity analyses to assess the robustness of our findings. First, we hypothesized that changes in cytokine trajectories may not be detectable in short transport times. Thus, we repeated our analyses in a sample limited to those patients with prehospital time intervals greater than 30 minutes. Second, we hypothesized the effect of prehospital fluids may be more pronounced among higher acuity transports. Therefore, we repeated our analyses in a sample limited to patients with a prehospital risk score ≥ 1. Third, we hypothesized that chronic organ dysfunction may be an important confounder, we included the Elixhauser comorbidity index in an additional model.^23–27 Fourth, because changes in cytokine measurements may be modified by temperature, we repeated analyses including prehospital temperature.²⁸ Fifth, in supplemental analyses, we explored the relationship of prehospital time with prehospital fluid volume administered and tested for effect modification of time on volume administered in an additional model, acknowledging that patients with longer transport times may receive more prehospital fluids. We considered a likelihood test of interaction significant if p<0.1. We conducted all analyses with STATA (version 14.2, StataCorp; College Station, TX) with alpha set at 0.05.

Results

Patients

Of 4,013 patients screened, 797 prehospital encounters were included in the PIPeLINE cohort. After exclusion of duplicate encounters, patients without IV catheter and encounters missing biomarker data, 345 unique patients remained for this analysis (Figure 1).

Figure 1.

Patient accrual

A total of 88 patients (26%) received an IV fluid bolus compared to 257 (74%) who did not. These two groups were similar in age (Table 1). Patients receiving an IV fluid bolus had a greater mean heart rate compared to patients who did not. Initial prehospital vital signs, Glasgow coma scale and prehospital risk score were similar in both groups. Clinical outcomes such as hospital length of stay, admission to intensive care and in-hospital mortality were similar in both groups (Table 1).

Table 1.

Patient Characteristics

Variable	No Prehospital Fluid Bolus	Prehospital Fluid Bolus
No. (% of 345)	257 (74%)	88 (26%)
Age [yrs]: mean (SD)	53 (20)	49 (18)
Male sex: no. (%)	129 (50%)	37 (42%)
Elixhauser comorbiditya: mean (SD)	0.25 (0.47)	0.28 (0.48)
Race: no. (%)
White	169 (68%)	53 (63%)
Black	78 (31%)	30 (36%)
Other	3 (1%)	1 (1%)
Prehospital assessmentb
Critical illness risk score: mean (SD)	1.1 (0.9)	1.2 (1.0)
Respiratory rate [breaths per min]: mean (SD)	19 (5)	18 (6)
Systolic blood pressure [mmHg]: mean (SD)	142 (36)	135 (36)
Glasgow coma scale score: mean (SD)	14.6 (1.6)	14.6 (1.1)
Pulse oximetry [%]: mean (SD)	97 (3)	96 (9)
Heart rate [beats per minute]: mean (SD)	91 (20)	99 (28)
Prehospital biomarkers c
IL-6 [pg/mL]: median (IQR)	11.5 (5.5 - 47.6)	22.3 (6.4 – 113)
IL-10 [pg/mL]: median (IQR)	3.0 (19 - 9.0)	3.5 (2.2 - 25.6)
TNF [pg/mL]: median (IQR)	6.9 (6.0 - 8.3)	7.5 (6.4 - 10.4)
Hospital biomarkers d
IL-6 [pg/mL]: median (IQR)	12.5 (6.0 - 41.1)	21.2 (6.8 - 99.4)
IL-10 [pg/mL]: median (IQR)	3.4 (2.0 - 10.4)	4.1 (2.3 - 24.0)
TNF [pg/mL]: median (IQR)	7.4 (6.3 - 8.4)	7.5 (7.1 - 8.9)
Hospital outcomes
Hospital length of stay [days]: median (IQR)	2 (1 – 3)	2 (1 – 4)
ICU admission: no. (%)	14 (5%)	7 (8%)
ICU LOS [days]: median (IQR)	3 (2 - 5)	3 (2 - 5)
In-hospital death: no. (%)	5 (2%)	1 (1%)

^aMeasured using ICD-9 codes extracted from the electronic health record

^bVital signs are first reported by EMS responders

^cPrehospital biomarkers measured at time of IV catheter placement

Abbreviations: SD, standard deviation; IQR, interquartile range; ICU intensive care unit; LOS, length of stay

Inflammatory cytokines

The median IL-6 at both prehospital and ED measurements was greater in patients who received a prehospital IV fluid bolus compared to patients who did not, (Table 1, eFigure 1). Prehospital IL-10 levels were similar between groups, but ED measurements of IL-10 were greater in patients who received a prehospital IV fluid bolus (Table 1, eFigure 1). Prehospital serum levels of TNF were similar between groups, but ED measurements of TNF were greater in patients who received a prehospital IV fluid bolus (Table 1, eFigure 1).

Association between prehospital fluid and changes in cytokine levels

In unadjusted analyses, the relative change in IL-6, IL-10 and TNF between prehospital and ED measurements was similar between groups (eFigure 2). After adjustment, there was a larger decrease in IL-6 for subjects who received a prehospital fluid bolus than in those who did not (Table 2). We did not observe a significant change in IL-10 or TNF between groups (Table 2).

Table 2.

Adjusted Regression Coefficients for Multivariable Models of Relative Change for Candidate Biomarkers (IL-6, IL-10, TNF)

Model	Predictor	Regression coefficient, β (95% CI)	p-value
Relative change in IL-6	Prehospital fluid bolus	−10.0 (−19.4, −0.6)	0.04
	Male sex	0.5 (−8.5, 9.5)	0.91
	Race
	white	0 (reference)	--
	black	6.5 (−3.7, 16.8)	0.29
	other	−24.5 (−46.2, −2.7)	0.03
	Prehospital risk score	4.4 (−0.8, 9.5)	0.09
	Time from FMC to ED	−15.1 (−45.0, 14.9)	0.31
Relative change in IL-10	Prehospital fluid bolus	4.6 (−5.1, 14.3)	0.34
	Male sex	−0.8 (−9.0, 7.4)	0.99
	Race
	white	0 (reference)	--
	black	1.1 (−7.6, 9.8)	0.9
	other	−34.0 (−66.4, −1.5)	0.03
	Prehospital risk score	3.0 (−1.3, 7.4)	0.12
	Time from FMC to ED	3.0 (−28.9, 27.0)	0.86
Relative change in TNF	Prehospital fluid bolus	−2.1 (−8.3, 4.1)	0.53
	Male sex	−3.8 (−9.1, 1.5)	0.16
	Race
	white	0 (reference)	--
	black	−3.3 (−9.4, 2.7)	0.33
	other	10.0 (−4.3, 24.4)	0.14
	Prehospital risk score	−0.7 (−3.1, 1.8)	0.31
	Time from FMC to ED	−6.7 (−22.5, 9.1)	0.46

Abbreviations: CI, confidence interval; FMC, first medical contact; ED, Emergency Department

Sensitivity Analysis

In sensitivity analyses, we observed similar differences in the change of IL-6 between groups when the sample was restricted to patients with transport times greater than 30 minutes, when the sample was restricted to patients with prehospital risk score ≥1, and when adjusting for Elixhauser Comorbidity Index and prehospital temperature (eTable 1). There was no significant effect modification of prehospital transport time on prehospital IV fluid bolus (p=0.56). We observed no change in results in models for IL-10 and TNF in sensitivity analyses.

Discussion

In this secondary analysis of a prospective sample of prehospital non-trauma, non-arrest patients, we observed a significant decrease in IL-6 from prehospital and ED measurements, but not in IL-10 or TNF, among patients who received a prehospital IV fluid bolus. These results were robust to sensitivity analyses that restricted the sample to patients with longer transport times and greater illness severity.

We previously reported a prehospital IV fluid bolus was associated with improved in-hospital mortality among patients with sepsis on ED arrival. ^11,12 There are a variety of explanations for this observation, including i.) a direct treatment effect from the prehospital fluid, ii.) an epiphenomenon, where “better” prehospital care occurred among patients receiving IV fluid, iii.) unmeasured confounding in regression models or iv.) more efficient care after ED arrival resulting from prehospital IV access placement. Though septic patients were not specifically studied here, our finding of an association between a prehospital IV fluid bolus and relative decrease in IL-6 levels support the hypothesis that prehospital fluids provide a direct treatment effect through modification of the host inflammatory response. These data are consistent with reports that hospital-based fluid resuscitation modifies cytokine trajectory perhaps through improved organ perfusion and mean circulating volume.¹⁹

Yet, there are other potential explanations for a relative decrease in IL-6 among patients receiving a prehospital IV fluid bolus. First, median values of IL-6 were greater at the EMS time point among subjects who received a prehospital IV fluid bolus. Though we accounted for baseline cytokine measurements by assessing relative change in biomarkers instead of absolute difference, the observed relative decrease associated with a prehospital IV fluid bolus could be due to unmeasured greater illness severity among the fluid group. We note in sensitivity analyses, however, our findings were consistent to restricting cohort to highest risk patients (prehospital critical risk score ≥1). Second, previous work has illustrated a temporal variability in certain biomarkers as a function of illness stage and severity.^18,29 In some patients, the prehospital care interval could have overlapped natural biomarker trajectories, including those experiencing a relative decrease in inflammation, irrespective of fluid administration. In others, the prehospital interval may correspond with early stages of illness, and the changes in IL-6 in response to a prehospital IV fluid bolus may represent a treatment effect. Thus, the observed difference in treatment effect may plausibly represent variability in illness as opposed to variability in response to intervention. However, previous work has demonstrated that in the setting of illness, IL-6 tends to naturally increase and remain persistently elevated over days to weeks, far beyond the treatment interval of prehospital care.^18,29 Our observed decrease in response to a prehospital fluid bolus, therefore, more likely represents a treatment effect as opposed to a natural denouement of inflammation.

We did not observe an association between a prehospital IV fluid bolus and change in IL-10 or TNF. Hotchkiss and colleagues describe the inflammatory response to infection as an orchestrated activation of signaling pathways with variable expression of common gene classes including cytokines like IL-6, IL-10 and TNF, where the expression of cytokines is tuned in response to perceived infectious threat.^30,31Though elevated serum levels of TNF and IL-10 have been observed in inflammatory conditions such as sepsis, association between trajectory, clinical severity and prognosis is uncertain.^32,33 The observed variability in the trajectory of IL-6 compared to IL-10 and TNF in our study is similar to observations previously reported.

Our data have important implications. First, blood sampling by EMS and subsequent measurement of biomarkers is feasible, even in urban settings. Further investigation of panels of biomarkers that capture different underlying mechanisms in acute illness, including abnormal coagulation, endothelial dysfunction, or tissue perfusion may be warranted. The translation of panels to point-of-care devices could allow for recognition and initiation of treatment in the prehospital interval. Second, these data suggest that prehospital fluid may have a direct treatment effect in medical patients, as recently reported in a Canadian cohort study.¹³ Further investigation into the treatment mechanism as well as downstream implications on clinical illness severity, ED care, and clinical outcomes is warranted, perhaps using randomized trials.

Limitations

These findings have several limitations. First, we were limited in our ability to adjust for other potentially important confounders due to sample size. Next, because biomarkers are not measured in routine emergency care, our blood sample assays were not performed upon acquisition but after a single freeze-thaw cycle. Other investigators, however, have described the stability of cytokine sampling after a single freeze-thaw cycle and all samples were subject to similar conditions.^{34, 35} Third, due to the observational nature of our study, patients could have been misclassified in an incorrect fluid group due to documentation error. Our prior work on a similar prehospital cohort, however, has shown adequate documentation of administered fluid volumes by comparing recorded volumes with mass-derived measurements, therefore, this limitation is unlikely to significantly affect our results.³⁶ Finally, the non-experimental design and covariate imbalances at baseline limits direct causal inference.

Conclusion

Among non-trauma, non-cardiac arrest patients, a prehospital fluid bolus is associated with a relative decrease in IL-6, but not IL-10 or TNF, from prehospital to ED measurement.

Figure 2. Heatmap of Expected Relative Change in IL-6 from Multivariable Linear Regression Model.

Heat map depicts expected relative change in IL-6, defined as $\frac{(\text{biomarkerED}-\text{biomarkerEMS})}{\text{biomarkerEMS}}$ from multivariable linear regression model. Green shading represents expected decrease in IL-6 in response to prehospital IV fluid bolus. Red shading represents expected relative increase in IL-6 in response to prehospital IV fluid bolus. Each cell represents an individual patient.