The Accuracy of Temperature Measurements Provided by the Edwards Lifesciences Pulmonary Artery Catheter

Abstract

BACKGROUND: 

Pulmonary artery catheters (PACs) are frequently used for monitoring patient temperatures in the intensive care unit. Nevertheless, data regarding the accuracy of these measurements are lacking, and few data testify to the accuracy of temperatures recorded after the PAC has been in place for several days. The absolute values of such measurements are relevant for critical care because patient temperatures are often used as diagnostic criteria for sepsis and antibiotic therapy. We thus hypothesized that the Edwards Lifesciences PAC would accurately measure blood temperature. To test our hypothesis, we compared temperature measurements obtained from PACs inserted in patients for different lengths of time with measurements of a reference platinum resistance thermometer (PRT).

METHODS: 

PACs were removed and analyzed in 39 patients in whom PACs were inserted for 0 to 5 days. The PACs were placed in calibration baths, and 10 consecutive measurements at each of 7 different temperatures were obtained (36°C, 36.5°C, 37°C, 38°C, 38.3°C, 39°C, and 40°C). The temperature measurements obtained using PACs were compared with measurements obtained using a PRT. Bland-Altman statistical analyses were performed. Outliers, defined as PAC temperature measurements that varied more than ±0.3°C from PRT measurements, were identified. We considered a catheter unfit for clinical diagnostic or therapeutic use if ≥15% of data pairs were outliers.

RESULTS: 

A total of 2730 data pairs were analyzed. Overall, the bias was −0.15°C; the precision was +0.13°C; and the limits of agreement were −0.45°C to +0.13°C. The bias and limits of agreement did not differ according to the age of the catheter or the temperature tested. One hundred fourteen data pairs (4.2% [95% confidence interval, 2.0%–6.4%]), involving 13 PACs and mostly from 4 PACs, were outliers.

CONCLUSIONS: 

We conclude that temperature measurements obtained using the Edwards Lifesciences PACs are thus sufficiently accurate to be used for clinical temperature monitoring in critically ill patients.

Fever is frequently observed in critically ill patients, and temperatures above a threshold of 38.3°C can warrant a diagnostic workup to determine the presence of potential infections.¹ In the intensive care unit (ICU), the reference device for patient temperature measurements is often the pulmonary arterial catheter (PAC) because other invasive and noninvasive temperature measurement methods generally produce similar results.^2–6 Nevertheless, data regarding the accuracy of the temperature measurements obtained using PACs are lacking, and the accuracy of a PAC temperature measurement may become inaccurate after several days in situ. Because PACs may be used to assess for the presence of infection, their accuracy should be assessed to avoid inappropriate diagnosis or treatment. Thus, we evaluated the accuracy of temperature measurements provided by PACs collected from ICU patients.

METHODS

In 2014, 39 PACs (Thermodilution Catheter Swan-Ganz™ continuous cardiac output; Edwards Lifesciences LLC, Irvine, CA) used in routine patient care were collected after removal from consecutive patients hospitalized in the ICU at a university hospital. The PACs ranged in usage duration from 0 to 5 days (0 days, n = 1; 1 day, n = 3; 2 days, n = 16; 3 days, n = 7; 4 days, n = 6; 5 days, n = 6). In an accredited laboratory, the PACs were connected to a monitor (HP OmniCare M1165/66A; Hewlett Packard, Andover, MA) and placed in calibration baths of 7 different temperatures (36°C, 36.5°C, 37°C, 38°C, 38.3°C, 39°C, and 40°C). For each temperature, 10 consecutive measurements were taken per PAC 10 minutes after the PAC was immersed in each of the baths. All measurements were performed at a constant room temperature of 23 ± 3°C. This study was approved by the local institutional ethics committee, “Comité d’éthique du CHU de Rennes, France,” which waived the requirement for written informed consent (record number: 15-17).

The bath container included an immersion thermostat (LAUDA Ecoline E100, Koenigshofen, Germany) and contained deionized water that sensed temperatures ranging from +5°C to +90°C. A bath cover was used to minimize evaporation, and an agitator was used for water circulation to maintain a homogeneous temperature.

Reference temperature measurements were determined using a 100-mm PT100 class A (series 3490A) high-temperature platinum resistance thermometer (PRT) that was annually calibrated by an independent company (OCEASOFT) approved by the French committee of accreditation (Comité Français d’Accréditation [cofrac]) and according to ISO/CEI 17025 standards. This temperature probe was connected to a computer and run with Thermo Calibration: V2.1 software (OCEASOFT SA, Montpellier, France). The uncertainty of measurement provided by the PRT was 0.03°C (www.oceasoft.fr).

Statistical Analysis

Analyses were performed using the Pearson correlation coefficient r and the Bland-Altman plot. The temperature values obtained using the PRT were used as the reference values. Statistical analyses were performed using SAS version 9.3 software (SAS Institute, Cary, NC). Agreement between the 2 measurement methods was assessed using the Bland-Altman analysis.⁷ The Bland-Altman analysis allowed us to determine the bias (defined as the mean difference between the 2 measurement methods), precision (defined as the standard deviation around the bias), and limits of agreement (the limits within which 95% of all the points occurred on either side of the bias; calculated as: ±1.96 × the precision) for the temperature reference value measured using the PRT. Because several measurements were performed for each PAC, a correction was used in the limits of agreement calculation.⁸ Outlier data were defined, locally, as PAC temperatures that were higher than ±0.3°C relative to the corresponding reference measurement using the PRT. This threshold was based on our estimation of clinically relevant thresholds. If >15% of data points were classified as outliers, we judged that catheter unfit for use in clinical practice. Thus, assuming ≥15% outliers and a hypothesis that the number of outliers will be 12.5%, it was necessary (with a β risk of 0.95 and an α risk of 0.05 in a 2-sided test) to have at least 2476 paired data points (NQuery Advisor 6.0; Statistical Solutions Ltd., Cork, Ireland). We determined the bias and limits of agreement for temperature levels <38.3°C and ≥38.3°C. We calculated the sensitivity, specificity, and positive and negative predictive values of the PACs for temperatures ≥38.3°C, which is the threshold for fever diagnosis.¹ The multiple measurements for each PAC were corrected when calculating the confidence intervals (CIs) for proportions. Specifically, we divided the number of subjects (n) by a variance inflation factor, which = 1 + ICC × (m − 1), where ICC is the intraclass correlation coefficient (0.85 in our study, calculated using a proc mixed), and m is the number of repeated measurements by PAC. If we assume that the number of temperatures tested did not influence the ICC, we can choose m = 10 repeated measures. Finally, we calculated the approximate CIs with the widely used asymptotic formula .

RESULTS

A total of 2730 data pairs from 39 PACs were analyzed. A correlation was observed between the temperature measurements obtained using the PRT and those obtained using the PACs (R = 0.99; P < 0.01; Fig. 1).

Figure 1.: Correlation between temperature measurements obtained using the platinum resistance thermometer and measurements obtained using pulmonary artery catheters (PACs).

Overall, the Bland-Altman analysis showed a bias of −0.15°C, a precision of +0.13°C, and limits of agreements ranging from −0.45°C to +0.13°C (Fig. 2). The Bland-Altman analyses conducted based on the length of use of the PACs (ranging from 1 to 5 days) and the temperatures measured are provided in Tables 1 and 2. One hundred fourteen data pairs (4.2% [95% CI, 2.0%–6.4%]), involving 13 PACs, were outliers. The majority (80/114) of these data points were from 4 PACs (Fig. 3).

Table 1.: Bland-Altman Analyses According to the Length of Use (days) of the Pulmonary Artery Catheters (PACs)

Table 2.: Bland-Altman Analyses According to Temperature

Figure 2.: Bland-Altman analyses. PAC = pulmonary artery catheter.

Figure 3.: Number of outliers related to the duration of catheter use. Each bar corresponds to 1 pulmonary artery catheter.

The temperature thresholds of <38.3°C (number of data pairs = 1560) and ≥38.3°C (number of data pairs = 1170) biases (limits of agreement) were −0.13°C (−0.42°C to +0.14°C) and −0.17°C (−0.46°C to +0.12°C), respectively. The sensitivity and specificity of the PACs to register the correct temperature at values ≥38.3°C were 99.9% (95% CI, 99.5%–100%) and 95.7% (95% CI, 92.8%–98.6%), respectively. The positive and negative predictive values for the PACs were 94.0% (95 CI%, 90.0%–98.0%) and 99.9% (95% CI, 99.6%–100%), respectively.

DISCUSSION

We found that the Edwards Lifesciences PAC provides generally reliable, sustained, and accurate measurement of patients’ body temperatures. Moreover, our study found that the Edwards PAC was accurate, regardless of the length of time during which the PAC was used and over a wide range of temperatures (36–40°C). Notably, with respect to the ability of the PACs to register temperatures over the fever threshold (≥38.3°C), the positive predictive value was 95.7%, which supports the use of the Edwards PAC for making clinical decisions. Only 4% of the data pairs were outliers, which was below the threshold of 15% defined before the study. These outliers involved 13 PACs, and most (80/114) were from 4 PACs (10.2% of the PACs tested). The latter accounted for 70% of the outliers, equally distributed (24, 19, 24, and 15 outliers, respectively) and were inserted for hemodynamic monitoring during cardiac surgery (2 mitral valve reconstruction, 1 aortic valve replacement, and 1 coronary artery bypass grafting). No problems during catheter insertion and no obvious dysfunction during the course of the catheters’ use (notably, pressure measurements and cardiac output allowed by the computer) were noted. Moreover, the external integrity of all catheters was checked before experimentation, and we observed no plication, stretching, and/or disruption of the catheters. Finally, the dysfunction of these 4 PACs was not associated with the duration of catheter use. These results could suggest that these particular PACs were defective, a possibility that has been previously reported.⁹

Although the PAC has been advocated as the “gold standard” for temperature monitoring in the ICU,^1,10 it has clear drawbacks. The availability of alternative methods of measuring patient temperature is important because when fever (defined as a temperature ≥38.3°C) is observed clinically, general and specific investigations are conducted to determine the cause. Several studies have evaluated alternative methods for temperature measurement in ICU patients and compared the results with measurements obtained using PACs.^2–6 Although temperature measurements obtained using PACs have been described in previous in vitro studies, meticulous laboratory conditions and precise statistical analysis were not used.^11,12 Moreover, few studies have assessed the accuracy of the temperature measurements obtained using PACs with prolonged insertion where rheological strain, infusion of fluids and/or vasopressors, and other factors can distort the measurement. In studies comparing alternative temperature measurements with those obtained using the PAC, the duration of insertion was not specified.^2–6

One strength of our study was the analysis of a large number of data pairs from PACs tested over a wide range of temperatures and inserted for durations ranging from 1 to 6 days. Nevertheless, some limitations were present. First, outliers were defined as values > ±0.3°C in PACs when compared with the PRT, which was based on our estimations that this error threshold was clinically relevant for clinical decision making. However, a meta-analysis of other temperature measurement strategies suggests that our choice was reasonable.¹³ Second, we did not study extreme temperature values (<36°C or >40°C), which may be observed clinically in conditions such as therapeutic hypothermia. Finally, we studied PACs only after removal from the patient and did not study any other temperature measurement strategy.

CONCLUSIONS

We found that temperature measurements provided by the Edwards Lifesciences PACs were clinically accurate across a wide range of temperatures and did not vary with duration of insertion. According to our testing, PACs are a clinically acceptable strategy for body temperature assessments in the ICU and may be used as a reference to establish and evaluate other methods of body temperature measurement.

“TAKE-HOME” MESSAGE

The temperature measurements obtained using PACs were clinically accurate and were not affected by the length of use of the catheter or the range of temperatures assessed.

DISCLOSURES

Name: Yoann Launey, MD.

Contribution: This author helped analyze the data and write the manuscript.

Attestation: Yoann Launey approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.

Name: Raphaëlle Larmet, MD.

Contribution: This author helped conduct the study, analyze the data, and write the manuscript.

Attestation: Raphaëlle Larmet approved the final manuscript.

Name: Nicolas Nesseler, MD.

Contribution: This author helped write and review the manuscript.

Attestation: Nicolas Nesseler approved the final manuscript.

Name: Yannick Malledant, MD.

Contribution: This author helped design the study and write the manuscript.

Attestation: Yannick Malledant approved the final manuscript and attests to the integrity of the original data and the analyses reported in this manuscript.

Name: Clément Palpacuer, PharmD.

Contribution: This author helped analyze the data and write the manuscript.

Attestation: Palpacuer approved the final manuscript.

Name: Philippe Seguin, MD, PhD.

Contribution: This author helped design the study, analyze the data, and write the manuscript.

Attestation: Philippe Seguin approved the final manuscript, attests to the integrity of the original data and the analysis reported in this manuscript, and is also the archival author.

This manuscript was handled by: Avery Tung, MD.

ACKNOWLEDGMENTS

We thank Dr. Frasca Denis for his help with the statistical analyses.