CICM Online CCR Journal logo CICM logo

Full Text View

Systematic Review

Accuracy of non-invasive body temperature measurement methods in adult patients admitted to the intensive care unit: a systematic review and meta-analysis

Salvatore L Cutuli, Emily J See, Eduardo A Osawa, Paolo Ancona, David Marshall, Glenn M Eastwood, Neil J Glassford, Rinaldo Bellomo

Crit Care Resusc 2021; 23 (1): 6-13

Correspondence: sl.cutuli@gmail.com

  • Author Details
    • Salvatore L Cutuli 1, 2
    • Emily J See 1, 3
    • Eduardo A Osawa 1
    • Paolo Ancona 1
    • David Marshall 1
    • Glenn M Eastwood 1
    • Neil J Glassford 1
    • Rinaldo Bellomo 1, 4
    1. Department of Intensive Care, Austin Hospital, Melbourne, VIC, Australia.
    2. Dipartimento di Scienze dell’Emergenza, Anestesiologiche e della Rianimazione; UOC di Anestesia, Rianimazione, Terapia Intensiva e Tossicologia Clinica; Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Rome; Istituto di Anestesia e Rianimazione; Università Cattolica del Sacro Cuore, Rome, Italy.
    3. School of Medicine, University of Melbourne, Melbourne, VIC, Australia.
    4. Centre for Integrated Critical Care, School of Medicine, University of Melbourne, Melbourne, VIC, Australia.
  • Competing Interests
    None declared
  • Abstract
    OBJECTIVE: Non-invasive thermometers are widely used in both clinical practice and trials to estimate core temperature. We aimed to investigate their accuracy and precision in patients admitted to the intensive care unit (ICU).
    STUDY DESIGN: Systematic review and meta-analysis.
    DATA SOURCES: We searched MEDLINE, EMBASE and the Cochrane Central Register of Controlled Trials to identify all relevant studies from 1966 to 2017. We selected published trials that reported the accuracy and precision of non-invasive peripheral thermometers (index test) in ICU patients compared with intravascular temperature measurement (reference test). The extracted data included the study design and setting, authors, study population, devices, and body temperature measurements.
    METHODS: Two reviewers performed the initial search, selected studies, and extracted data. Study quality was assessed using the QUADAS-2 tool. Pooled estimates of the mean bias between index and reference tests and the standard deviation of mean bias were synthesised using DerSimonian and Laird random effects meta-analyses.
    RESULTS: We included 13 cohort studies (632 patients, 105 375 measurements). Axillary, tympanic infrared and zero heat flux thermometers all underestimated intravascular temperature. Only oesophageal measurements showed clinically acceptable accuracy. We found an insufficient number of studies to assess precision for any technique. Study heterogeneity was high (99–100%). Risk of bias for the index test was unclear, mostly because of no device calibration or control for confounders.
    CONCLUSIONS: Compared with the gold standard of intravascular temperature measurement, non-invasive peripheral thermometers have low accuracy. This makes their clinical and trial-related use in ICU patients unreliable and potentially misleading.
  • References
    1. Young P, Bellomo R, Bernard G, et al. Fever control in critically ill adults. An individual patient data meta-analysis of randomised controlled trials. Intensive Care Med 2019; 45: 468-76.
    2. Peres Bota D, Lopes Ferreira F, Mélot C, Vincent J. Body temperature alterations in the critically ill. Intensive Care Med 2004; 30: 811-6.
    3. Bhavani S, Carey K, Gilbert E, Afshar M, et al. Identifying novel sepsis subphenotypes using temperature trajectories. Am J Respir Crit Care Med 2019; 200: 327-35.
    4. Laupland K, Zahar J, Adrie C, et al. Determinants of temperature abnormalities and influence on outcome of critical illness. Crit Care Med 2012; 40: 145-51.
    5. Young P, Saxena M, Beasley R, et al. Early peak temperature and mortality in critically ill patients with or without infection. Intensive Care Med 2012; 38: 437-44
    6. Saxena M, Young P, Pilcher D, et al. Early temperature and mortality in critically ill patients with acute neurological diseases: trauma and stroke differ from infection. Intensive Care Med 2015; 41: 823-32.
    7. Hassager C, Nagao K, Hildick-Smith D. Out-of-hospital cardiac arrest: in-hospital intervention strategies. Lancet 2018; 391: 989-98.
    8. Bernard S, Gray T, Buist M, et al. Treatment of comatose survivors of out-of-hospital cardiac arrest with induced hypothermia. N Engl J Med 2002; 346: 557-63.
    9. Nielsen N, Wetterslev J, Cronberg T, et al. Targeted temperature management at 33°C versus 36°C after cardiac arrest. N Engl J Med 2013; 369: 2197-206.
    10. Lascarrou J, Merdji H, Gouge AL, et al. Targeted temperature management for cardiac arrest with nonshockable rhythm. N Engl J Med 2019; 381: 2327-37.
    11. Cooper D, Nichol A, Bailey M, et al. effect of early sustained prophylactic hypothermia on neurologic outcomes among patients with severe traumatic brain injury: the POLAR randomized clinical trial. JAMA 2018; 320: 2211-20.
    12. Hypothermia after Cardiac Arrest Study Group. Mild therapeutic hypothermia to improve the neurologic outcome after cardiac arrest. N Engl J Med 2002; 346: 549-56.
    13. Clifton G, Valadka A, Zygun D, et al. Very early hypothermia induction in patients with severe brain injury (the National Acute Brain Injury Study: Hypothermia II): a randomised trial. Lancet Neurol 2011; 10: 131-9.
    14. Andrews P, Sinclair H, Rodriguez A, et al. Hypothermia for intracranial hypertension after traumatic brain injury. N Engl J Med 2015; 373: 2403-12.
    15. Young P, Saxena M, Bellomo R, et al. Acetaminophen for fever in critically ill patients with suspected infection. N Engl J Med 2015; 373: 2215-24.
    16. Mourvillier B, Tubach F, Van de Beek D, et al. Induced hypothermia in severe bacterial meningitis: a randomized clinical trial. JAMA 2013; 310: 2174-83.
    17. Young P, Bailey M, Bass F, et al. Randomised evaluation of active control of temperature versus ordinary temperature management (REACTOR) trial. Intensive Care Med 2019; 45: 1382-91.
    18. Sessler D. Temperature monitoring and perioperative thermoregulation. Anesthesiology 2008; 109: 318‑38.
    19. Eichna L, Berger A, Rader B, Becker W. Comparison of intracardiac and intravascular temperatures with rectal temperatures in man. J Clin Invest 1951; 30: 353-9.
    20. Pandey A, Khera R, Kumar N, et al. Use of pulmonary artery catheterization in US patients with heart failure, 2001–2012. JAMA Intern Med 2016; 176: 129-32.
    21. Wiener R, Welch H. Trends in the use of the pulmonary artery catheter in the United States, 1993–2004. JAMA 2007; 298: 423-9.
    22. Childs C. Body temperature and clinical thermometry. Handb Clin Neurol 2018; 157: 467-82.
    23. Bridges E, Thomas K. Noninvasive measurement of body temperature in critically ill patients. Crit Care Nurse 2009; 29: 94-7.
    24. Carney N, Totten A, O’Reilly C, et al. Guidelines for the management of severe traumatic brain injury, fourth edition. Neurosurgery 2017; 80: 6-15.
    25. Nolan J, Soar J, Cariou A, et al. European Resuscitation Council and European Society of Intensive Care Medicine guidelines for post-resuscitation care 2015: section 5 of the European Resuscitation Council Guidelines for Resuscitation 2015. Resuscitation 2015; 95: 202-22.
    26. O’Grady N, Barie P, Bartlett J, et al. Guidelines for evaluation of new fever in critically ill adult patients: 2008 update from the American College of Critical Care Medicine and the Infectious Diseases Society of America. Crit Care Med 2008; 36: 1330-49.
    27. Cutuli S, Osawa E, Glassford N, et al. Body temperature measurement methods and targets in Australian and New Zealand intensive care units. Crit Care Resusc 2018; 20: 241-4.
    28. Niven D, Gaudet J, Laupland K, et al. Accuracy of peripheral thermometers for estimating temperature. a systematic review and meta-analysis. Ann Intern Med 2015; 163: 768-77.
    29. Jefferies S, Weatherall M, Young P, Beasley R. A systematic review of the accuracy of peripheral thermometry in estimating core temperatures among febrile critically ill patients. Crit Care Resusc 2011; 13: 194-9.
    30. Cutuli S, Ancona P, Osawa E, et al. Accuracy of non-invasive body temperature measurement methods in adult patients admitted to the intensive care unit: a systematic review: PROSPERO; 2017 http://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD42017077838 (viewed Jan 2021).
    31. Liberati A, Altman D, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ 2009; 21: b2700.
    32. Bland J, Altman D. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: 307-10.
    33. Moran J, Peter J, Solomon P, et al. Tympanic temperature measurements: are they reliable in the critically ill? A clinical study of measures of agreement. Crit Care Med 2007; 35: 155-64.
    34. DerSimonian R, Laird N. Meta-analysis in clinical trial. Control Clin Trials 1986; 7: 177-88.
    35. Ahn S, Fessler J. Standard errors of mean, variance, and standard deviation estimators [24 July 2003]. https://web.eecs.umich.edu/~fessler/papers/files/tr/stderr.pdf (viewed Jan 2021).
    36. Niven D, Gaudet J, Laupland K, et al. Accuracy of peripheral thermometers for estimating temperature. Ann Intern Med 2015; 163: 768-77.
    37. Higgins J, Thompson S, Deeks J, Altman D. Measuring inconsistency in meta-analyses. BMJ 2003; 327: 557-60.
    38. Whiting P, Rutjes A, Westwood M, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med 2011; 155: 529-36.
    39. Dahyot-Fizelier C, Lamarche S, Kerforne T, et al. Accuracy of zero-heat-flux cutaneous temperature in intensive care adults. Crit Care Med 2017; 45: e715-7.
    40. Eshraghi Y, Nasr V, Parra-Sanchez I, et al. An evaluation of a zero-heat-flux cutaneous thermometer in cardiac surgical patients. Anesth Analg 2014; 119: 543-9.
    41. Farnell S, Maxwell L, Tan S, et al. Temperature measurement: comparison of non-invasive methods used in adult critical care. J Clin Nurs 2005; 14: 632-9.
    42. Fulbrook P. Core body temperature measurement: a comparison of axilla, tympanic membrane and pulmonary artery blood temperature. Intensive Crit Care Nurs 1997; 13(5): 266-72.
    43. Haugk M, Stratil P, Sterz F, et al. Temperature monitored on the cuff surface of an endotracheal tube reflects body temperature. Crit Care Med 2010; 38: 1569-73.
    44. Lefrant J, Muller L, de la Coussaye J, et al. Temperature measurement in intensive care patients: comparison of urinary bladder, oesophageal, rectal, axillary, and inguinal methods versus pulmonary artery core method. Intensive Care Med 2003; 29: 414-8.
    45. Myny D, de Waele J, Defloor T, et al. Temporal scanner thermometry: a new method of core temperature estimation in ICU patients. Scott Med J 2005; 50: 15-8.
    46. Nierman D. Core temperature measurement in the intensive care unit. Crit Care Med 1991; 19: 818-23.
    47. Nonose Y, Sato Y, Kabayama H, et al. Accuracy of recorded body temperature of critically ill patients related to measurement site: a prospective observational study. Anaesth Intensive Care 2012; 40: 820-4.
    48. Shin J, Kim J, Song K, Kwak Y. Core temperature measurement in therapeutic hypothermia according to different phases: comparison of bladder, rectal, and tympanic versus pulmonary artery methods. Resuscitation 2013; 84: 810-7.
    49. Smith L. Temperature measurement in critical care adults: a comparison of thermometry and measurement routes. Biol Res Nurs 2004; 6: 117-25.
    50. Stavem K, Saxholm H, Smith-Erichsen N. Accuracy of infrared ear thermometry in adult patients. Intensive Care Med 1997; 23: 100-5.
    51. Hooper V, Andrews J. Accuracy of noninvasive core temperature measurement in acutely ill adults: the state of the science. Biol Res Nurs 2006; 8: 24-34.
    52. Chacko B, Peter J. Temperature monitoring in the intensive care unit. Indian J Respir Care 2018; 7: 28-32.
    53. Eyre D, Sheppard A, Madder H, et al. A Candida auris outbreak and its control in an intensive care setting. N Engl J Med 2018; 379: 1322-31.
    54. Legriel S, Lemiale V, Schenck M, et al. Hypothermia for neuroprotection in convulsive status epilepticus. N Engl J Med 2016; 375(25): 2457-67
Body temperature is a key vital sign in intensive care unit (ICU) practice due to its clinical implications. 1 Moreover, body temperature abnormalities are frequent in the ICU, 2 trigger interventions (eg, antibiotics), and allow prognostication. 3, 4, 5, 6, 7 Finally, targeting of specific body temperature levels has been a key feature of multiple studies in patients after cardiac arrest, 8, 9, 10 brain injury, 11, 12, 13, 14 and infection 15, 16, 17 (Online Appendix, 1).
 
Body temperature is classified as “core”, which refers to the temperature of organs, measured by invasive methods (eg, oesophageal probes), and “peripheral”, which refers to the temperature of external body surfaces, measurable by non-invasive tools (eg, cutaneous) 18 (Online Appendix, 2). Intravascular thermometers are considered to be the gold standard for core body temperature measurement, although pulmonary artery catheter (PAC) 19 use is decreasing over time. 20, 21 Other tools have been proposed for clinical use or trials. 22, 23 Although invasive tools have been advised in severely ill patients, 24, 25, 26 peripheral thermometers are widely used in the ICU. 27 However, concerns about their low accuracy 28, 29 and the impact of such possible inaccuracy on clinical management remain. Despite such concerns and the importance of accurate temperature measurement in ICU patients, the available evidence for these patients has never been systematically reviewed and analysed.
 
Accordingly, we aimed to survey the literature with the intention to systematically assess the accuracy and precision of both peripheral and invasive thermometers in the ICU.
 

Methods

Study protocol

The protocol for this systematic review and meta-analysis was published 30 before commencement and was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 31
 

Search strategy

MEDLINE, EMBASE and the Cochrane Central Register of Controlled Trials were searched to identify relevant studies published from 1966 to 2017 (Online Appendix, 3). Electronic database searches were supplemented by hand-searching of reference lists of retrieved articles, previous reviews and international guidelines.
 

Study inclusion and exclusion criteria

We included observational studies and randomised controlled trials published in English in peer-reviewed journals. These studies were performed in adult (aged ≥ 18 years) ICU patients and investigated accuracy and precision of non-invasive peripheral and invasive body temperature methods, using intravascular measurements as the comparator. We excluded unpublished articles and commentaries. Furthermore, we excluded animal studies and articles that did not use the Bland–Altman approach to investigate accuracy of different body temperature measurement methods.
 

Outcome measures

The primary outcome was the assessment of accuracy and precision of non-invasive peripheral thermometers compared with intravascular methods.
 

Assessment of accuracy and precision

Accuracy was assessed by the Bland–Altman approach. 32 The mean difference (reference test minus index test) between measures is the bias. The 95% confidence interval (CI) of the differences between measures is the limits of agreement (LoA), a measure of the variance between body temperature measurements. Precision was assessed by the Lin concordance correlation coefficient. 33
 

Study selection and data extraction

Two reviewers (SLC and DM) independently performed the initial search and the study selection by title, abstract and full text. All disagreements were managed through discussion. In the case of consensus not being reached, a third reviewer (NJG) was consulted.
 
Study design and setting, authors, study population, device, and body temperature measurements were independently extracted by four reviewers (SLC, EAO, DM and PA) using a standardised online spreadsheet (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia). One additional reviewer (EJS) checked all data abstractions.
 
Invasive extravascular measurement methods included urinary bladder, oesophageal, rectal, nasopharyngeal and tracheal body temperature-sensing probes. Non-invasive peripheral measurement methods included axillary, tympanic infrared, temporal scanner, oral, inguinal and zero heat flux (ZHF) thermometers. ZHF is a dot-shaped electronic thermometer that measures body temperature about 1–2 cm below the skin surface.

Statistical analysis

Continuous variables were pooled and reported as median (interquartile range [IQR]) or mean (standard deviation [SD]), as appropriate. Categorical variables were reported as frequencies and percentages. Pooled estimates of the mean bias between index and reference tests and the standard deviation of mean bias were synthesised using DerSimonian and Laird random effects meta-analyses. 34 Temperature-sensing intravascular catheters were considered the reference test for all pooled analyses, with data for each invasive or peripheral index test summarised separately. In this review, we will refer to temperature-sensing intravascular catheter as “PAC-derived”, because of the paucity of data on other intravascular temperature-sensing probes. Pooled analyses were performed when at least three comparisons were available. The 95% CI for the mean bias and the SD were calculated using the appropriate equations. 35 The LoA between index and reference tests were calculated as the pooled mean bias ± 1.96 times the pooled SD. 32 It was decided a priori that the clinically acceptable mean bias was within ± 0.2°C and LoA within ± 0.5°C. 29, 36

Between-study heterogeneity was evaluated by the Cochran Q test and quantified by the I2 statistic. 37 Heterogeneity was considered to be low, moderate or high at I2 values of 25%, 50% and 75% respectively. High heterogeneity was investigated by sensitivity analyses and meta-regression. We performed a sensitivity analysis excluding studies that did not fully account for repeated measures within individual subjects. Moreover, we performed a further sensitivity analysis to explore the accuracy of electronic thermometers set in “core mode”, which means adjusting temperatures measured at one peripheral site of the body (eg, ear canal) to estimate core temperature by manufacturer-made conversion algorithms that add automatically a fixed number to the temperature taken. 29 Meta-regression analyses were performed to investigate the impact of study-level characteristics on the mean bias, including year of publication, number of patients and measurements, patient age, proportion of male patients, illness severity score (Acute Physiological and Chronic Health Evaluation [APACHE] II or Simplified Acute Physiology Score [SAPS] 2), or the requirement for neuromuscular blockade, vasopressor, or mechanical ventilation.
 
Study quality was independently assessed by two reviewers (SLC and DM) using the QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies 2) tool 38 (Online Appendix, 4).
 
Publication bias was assessed using funnel plots and by Egger test. Data were analysed using Stata/SE14.0 (College Station, TX). A two-sided P < 0.05 was considered significant.
 

Results

Characteristics of included studies

We identified 601 unique citations (Figure 1), among which, 13 studies from 1991 to 2017 33, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50  (632 patients; 105 375 measurements) were eligible for inclusion (Online Appendix, table S1).

The median number of patients was 42 (IQR, 21–60) per study and the median number of index-reference comparisons per study was 529 (IQR, 66–2000). The mean patient age was 64 years (SD, 7) and the percentage of male patients ranged from 47% to 86%. The percentage of mechanically ventilated patients ranged from 64% to 100% (n = 4), while the percentage of patients requiring vasopressor support ranged from 35% to 71% (= 5). Three studies reported mean APACHE II scores that ranged from 15 to 25, while one study reported a mean SAPS 2 score of 43.4.

Intravascular measurements were performed by PAC (n = 12) or temperature-sensing catheters placed in the iliac artery (n = 1). The mean PAC-derived temperature ranged from 34.7°C to 37.7°C.
 

Accuracy and precision of non-invasive peripheral thermometers compared with PAC

Thirteen studies (632 patients) compared PAC with non-invasive peripheral thermometers (Figure 2). The results of the random effect meta-analysis are shown in Table1 and in the Online Appendix, figures S1 and S2. We did not find enough data to conduct a meta-analysis on the accuracy of cutaneous, oral thermometers and temporal scanner. None of the peripheral thermometers showed clinically acceptable mean bias and LoA.

A post hoc sensitivity analysis for studies comparing PAC with tympanic infrared set in core mode found a clinically acceptable mean bias (-0.02°C), but the LoA was still wide (-0.88 to 0.84°C).

We found an insufficient number of studies on precision of peripheral body temperature measurement methods to perform a meta-analysis.

Study heterogeneity was very high (99–100%). Study variability was partly explained by year of publication (coefficient 0.05; 95% CI, 0.01–0.09; P = 0.03) and mean variation in number of patients (coefficient 0.01; 95% CI, 0.00–0.02; P = 0.04).

Accuracy and precision of invasive extravascular thermometers compared with PAC

Seven studies (331 patients) compared PAC with invasive extravascular thermometers (Figure 3). The results of the random effect meta-analysis are shown in Table1 and in the Online Appendix, figures S3 and S4. All invasive extravascular thermometers showed clinically acceptable mean bias, although only oesophageal probes had LoA within the a priori set range. We did not find enough data to conduct a meta-analysis on endotracheal tube and nasopharyngeal body temperature sensing-probes.

Study heterogeneity was very high (99–100%). Study variability was not explained by any year of publication or number of patients.

We found an insufficient number of studies on precision of invasive body temperature measurement methods to perform a meta-analysis.

Sensitivity analyses

When excluding studies that did not fully account for repeated measures within individual subjects, 33, 41, 44, 48, 50 there were no qualitative differences in findings or heterogeneity (Online Appendix, table S2).

Quality assessment

The results of the QUADAS-2 evaluations are provided in the Online Appendix, table S3 and figure S5. The risk of bias for the index test was unclear, mostly because device calibration was not undertaken, 33, 43, 44, 45, 46, 47, 48, 50 investigators skills in using index tests were not verified, 39, 40, 43, 44, 46, 48, 50 and there was no control for possible confounders 33, 39, 40, 41, 43, 44, 45, 46, 49, 50 (Online Appendix, table S4).
 

Publication bias

There was no funnel plot asymmetry in any of the analyses to support the presence of small study effects (Online Appendix, figures S6a and S6b).
 

Discussion

Main findings

This systematic review and meta-analysis investigated the accuracy of non-invasive peripheral body temperature measurement methods in ICU patients, most of whom were mechanically ventilated and/or haemodynamically unstable. Core temperature ranged from moderate hypothermia to just above normothermia. We found that none of the non-invasive peripheral thermometers had clinically acceptable mean bias and LoA. Moreover, all provided a variable underestimation of core temperature. When tympanic infrared thermometers were set in “core mode”, the mean bias decreased ten-fold and became clinically acceptable, but the LoA remained wide. We found an insufficient number of studies on oral, skin surface, and temporal scanner to perform a meta-analysis on the accuracy of such methods.

Among invasive extravascular methods, only temperature-sensing oesophageal probes were clinically acceptable. Study heterogeneity was high and the risk of bias was unclear. We did not find enough studies to assess accuracy of endotracheal tube and nasopharyngeal body temperature-sensing probes. Finally, we found an insufficient number of studies on precision to perform a meta-analysis.
 

Relationship with previous studies

To date, no study has conducted a systematic review and meta-analysis of the accuracy of body temperature measurement methods in ICU patients. In 2006, Hooper and Andrews 51 conducted a systematic review of 23 articles on the accuracy of peripheral thermometers in a mixed population of acutely ill hospitalised adults. The authors concluded that only oral measurements were accurate in estimating core body temperature, although high study heterogeneity warranted further investigation. Because of differences in inclusion criteria (not conducted in an ICU setting, not only involving adult patients, or not assessing accuracy with the Bland–Altman approach), most of these studies were not included in our review.

In 2011, Jefferies and colleagues 29 completed a review of three articles on the accuracy of peripheral thermometry in estimating core temperature among febrile ICU patients. The authors concluded that both tympanic and oral thermometry provide an accurate measure of core temperatures within the febrile range. However, this study was underpowered to provide a pooled estimation of accuracy and did not evaluate the degree of heterogeneity among such articles. These factors significantly limit the applicability of the above findings. In a 2015 study, Niven and colleagues 28 reviewed the accuracy of peripheral thermometers compared with invasive thermometers and pooled the results of 75 studies on children and adults. All invasive devices carried clinically acceptable mean bias and LoA compared with the PAC. However, the authors found that peripheral thermometers did not have clinically acceptable accuracy and discouraged their clinical use. Unfortunately, this systematic review and meta-analysis may be biased by studies from different clinical settings (45% ICU, 27% general ward/mixed, 16% emergency department, 7% operating theatre, 4% outpatient clinic). Specifically, physiological and iatrogenic conditions of critical illness (eg, fluid shifts, haemodynamic instability, nasogastric feeding, and parenteral infusions) may create local and rapid systemic temperature alterations, which make body temperature in each anatomical district different from those seen in non-ICU patients. 23, 52 Moreover, recent evidence identified skin-surface axillary temperature probes as main factors involved in the transmission of Candida auris infection, 53 raising additional concerns on their use in the ICU.

Implications

The lack of accuracy of peripheral thermometers in estimating core temperature implies the need to limit the use of such methods in severe ICU patients due to the risk of obtaining misleading clinical information. Moreover, our study implies that, among invasive body temperature measurement methods, only oesophageal probes have acceptable accuracy and should, therefore, be preferred in clinical practice. Finally, the lack of evidence on precision among all measurement techniques as well as the low accuracy of several peripheral and even invasive extravascular body temperature methods are of great concern. This is particularly relevant to trials focusing on temperature control in patients after cardiac arrest 8, 9, 10, 12 or after traumatic brain injury 11, 13, 14 or convulsive status epilepticus 54 or patients with infection, 15, 16, 17 where core temperature was estimated by many different extravascular invasive methods (Online Appendix, 4). Furthermore, the axillary route was considered the standard method for measuring body temperature in the REACTOR trial, 17 which evaluated acetaminophen safety for fever control in ICU patients. Our study highlights that, in most of these trials, the technology of body temperature measurement was likely flawed.

Strengths and limitations

To our knowledge, we are the first to perform a comprehensive systematic review and meta-analysis of daily used body temperature monitoring tools in ICU patients. In addition, we shed light on the major shortcomings of peripheral thermometers in clinical practice. Finally, we provide strong evidence to inform the research agenda in this field, specifically when temperature management is intended to be an intervention that modifies patient-centred outcomes in specific clinical conditions, such as cardiac arrest, traumatic brain injury, and sepsis.

However, we acknowledge some limitations. We restricted our research to articles that evaluated the accuracy of peripheral thermometers using the Bland–Altman approach. Although many statistical methods have been used in this kind of research 33 and their inclusion might add information on other body temperature measurement tools (eg, oral or temporal scanner), we selected such a statistical approach because of its wide acceptance and to provide homogeneity to the data interpretation. Furthermore, we did not contact the authors of each study to obtain original data and perform a patient-level meta-analysis. Nevertheless, the observational nature of the articles included and the unclear control for possible confounders would have limited the additional value of such analysis. Moreover, we acknowledge the validity that LoA calculated as the pooled mean bias  1.96 times the pooled SD increases with larger sample sizes. However, this is the standard method to calculate LoA and there is no specific threshold of sample size below which it is invalid. Finally, we did not investigate the accuracy of body temperature measurement methods in the setting of marked hypothermia or fever, but this assessment had been previously attempted and found to be difficult due to the small number of studies. 29

Conclusion

In our systematic review and meta-analysis, peripheral thermometers showed a clinically unacceptably low degree of accuracy compared with PAC. In contrast, oesophageal measurements showed clinically acceptable accuracy in ICU patients, suggesting that their use may be justified. However, in most of the trials of temperature management in clinical conditions in which its accurate measurement matters, the technology of body temperature measurement was flawed. Finally, the lack of data for most thermometer types and the low quality of the evidence available remain a major problem.
Acknowledgements: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
 

TOP