Portal de Periódicos da CAPES

Haemolysis index: validation for haemolysis detection during extracorporeal membrane oxygenation

2020; Elsevier BV; Volume: 125; Issue: 2 Linguagem: Inglês

10.1016/j.bja.2020.04.080

1471-6771

Elodie Boissier, Karim Lakhal, Thomas Sénage, Philippe Bizouarn, Thierry Lepoivre, Johanna Nicolet, Jean‐Christian Roussel, Bertrand Rozec, Mickaël Vourc’h, Edith Bigot–Corbel,

Heart Failure Treatment and Management

Editor—Extracorporeal membrane oxygenation (ECMO)-induced haemolysis is provoked by blood trauma, release of gaseous microbubbles from degassing, or cavitation of red cells exposed to negative pressure (in the pump head, for instance).1Toomasian J.M. Bartlett R.H. Hemolysis and ECMO pumps in the 21st century.Perfusion. 2011; 26: 5-6Crossref PubMed Scopus (77) Google Scholar,2Omar H.R. Mirsaeidi M. Socias S. et al.Plasma free hemoglobin is an independent predictor of mortality among patients on extracorporeal membrane oxygenation support.PLoS ONE. 2015; 10e0124034Crossref PubMed Scopus (73) Google Scholar When plasma-free haemoglobin (fHb)-scavenging mechanisms are overwhelmed, fHb may cause damage to the kidneys and other organs.2Omar H.R. Mirsaeidi M. Socias S. et al.Plasma free hemoglobin is an independent predictor of mortality among patients on extracorporeal membrane oxygenation support.PLoS ONE. 2015; 10e0124034Crossref PubMed Scopus (73) Google Scholar, 3Lou S. MacLaren G. Best D. Delzoppo C. Butt W. Hemolysis in pediatric patients receiving centrifugal-pump extracorporeal membrane oxygenation: prevalence, risk factors, and outcomes.Crit Care Med. 2014; 42: 1213-1220Crossref PubMed Scopus (92) Google Scholar, 4Lubnow M. Philipp A. Foltan M. et al.Technical complications during veno-venous extracorporeal membrane oxygenation and their relevance predicting a system-exchange—retrospective analysis of 265 cases.PLoS ONE. 2014; 9e112316Crossref PubMed Scopus (83) Google Scholar, 5Lyu L. Long C. Hei F. et al.Plasma free hemoglobin is a predictor of acute renal failure during adult venous-arterial extracorporeal membrane oxygenation support.J Cardiothorac Vasc Anesth. 2016; 30: 891-895Abstract Full Text Full Text PDF PubMed Scopus (25) Google Scholar Therefore, prompt identification of haemolysis is of utmost importance as changes in therapy could be implemented. Measurement of fHb via the conventional spectrophotometry lacks automation, is cumbersome, and is, at best, performed only once a day during working hours. This predisposes to delayed detection of life-threatening haemolysis episodes.6Neal J.R. Quintana E. Pike R.B. Hoyer J.D. Joyce L.D. Schears G. Using daily plasma-free hemoglobin levels for diagnosis of critical pump thrombus in patients undergoing ECMO or VAD support.J Extra Corpor Technol. 2015; 47: 103-108PubMed Google Scholar Newer clinical chemistry analysers display a haemolysis index (HI; a dimensionless value) to assess the reliability of measurements potentially flawed by haemolysis (e.g. K+). The aim of this prospective study was to assess whether the HI, an in vitro tool proposed to detect haemolysis related to sample collection, transport, and processing, could be used to assess in vivo haemolysis in patients on ECMO. We also assessed (1) inter-measurement variability of HI and fHb and (2) impact on HI measurements of icterus and lipaemia (possibly related to propofol infusion) that may interfere with spectrophotometric measurements.7Nikolac Gabaj N. Miler M. Vrtarić A. et al.Precision, accuracy, cross reactivity and comparability of serum indices measurement on Abbott Architect c8000, Beckman Coulter AU5800 and Roche Cobas 6000 c501 clinical chemistry analyzers.Clin Chem Lab Med. 2018; 56: 776-788Crossref PubMed Scopus (20) Google Scholar The ethics committee of the French Society of Anaesthesia and Intensive Care approved the study protocol (00010254-2016-038). The purely observational nature of this study and the lack of clinical data collected enabled patient consent to be waived. In both derivation (March 2015–December 2016) and validation (January 2017–September 2018) cohorts, a daily blood sample from consecutive adult patients on ECMO at our institution was analysed for electrolyte analysis, and lithium heparin tubes were centrifuged at 2200 g for 10 min before analysing the supernatant. HI was measured via a Roche®Cobas 6000 analyser (light absorbance tested at 570 and 600 nm). fHb measurements relied on spectrophotometry (Unicam UV3 UV/Vis Spectrophotometer; PerkinElmer, Hopkinton, Massachusetts, United States); after dilution (1:11) of a plasma sample, the absorption spectrum was measured at different wavelengths before and after reducing oxyhaemoglobin with sodium hydrosulphite.8Shim B.S. Jue D.M. Simple spectrophotometric determination of haptoglobin–haemoglobin complex in haemolysed samples.Scand J Clin Lab Invest. 1986; 46: 45-51Crossref PubMed Scopus (10) Google Scholar The relationship between fHb and HI was assessed (Lin's concordance correlation coefficient [CCC]). In the derivation cohort, using receiver operating characteristics (ROC) curve analysis, we evaluated the ability of HI to detect fHb exceeding a critical threshold of 100 mg dl−1.3Lou S. MacLaren G. Best D. Delzoppo C. Butt W. Hemolysis in pediatric patients receiving centrifugal-pump extracorporeal membrane oxygenation: prevalence, risk factors, and outcomes.Crit Care Med. 2014; 42: 1213-1220Crossref PubMed Scopus (92) Google Scholar,6Neal J.R. Quintana E. Pike R.B. Hoyer J.D. Joyce L.D. Schears G. Using daily plasma-free hemoglobin levels for diagnosis of critical pump thrombus in patients undergoing ECMO or VAD support.J Extra Corpor Technol. 2015; 47: 103-108PubMed Google Scholar In the validation cohort, we tested the cut-off of HI we determined in the derivation cohort. Beforehand, the bootstrap technique created a large set of 1000 samples in each cohort. To assess the inter-measurement variability, the coefficient of variation of HI and fHb was determined in 30 repeated measurements on four home-made fHb solutions (25, 50, 100, and 200 mg dl−1). Those solutions were prepared and frozen for a few weeks at –20°C after we ensured they were stable for more than 12 months. They were then thawed in a random order, and measurement of fHb and HI started simultaneously. The Roche®Cobas 6000 analyser provides spectrophotometric measurements of icterus and lipaemic indices, surrogates for plasma total bilirubin and triglycerides, respectively.7Nikolac Gabaj N. Miler M. Vrtarić A. et al.Precision, accuracy, cross reactivity and comparability of serum indices measurement on Abbott Architect c8000, Beckman Coulter AU5800 and Roche Cobas 6000 c501 clinical chemistry analyzers.Clin Chem Lab Med. 2018; 56: 776-788Crossref PubMed Scopus (20) Google Scholar An in vitro experimental exploration was performed in five patients. First, we evaluated the impact on HI of a gradual increase of icterus index via manual addition of bilirubin. Second, we tested the impact of an increase in triglyceride concentration. Last, we tested the effect of a gradual increase in propofol concentration on the HI. Owing to the limited size of these preliminary experiments, no statistical comparisons were made. In our clinical cohort, a logistic regression assessed the performance of a model including HI, icterus, and lipaemic indices. In the derivation cohort, 560 samples from 67 patients were analysed: median fHb was 3 mg dl−1 [interquartile range (IQR)=0; 20], ranging from 0 to 1992 mg dl−1 with fHb ≥100 mg dl−1 in 58 (10%) samples. Median HI was 5 [IQR=2; 12], ranging from 0 to 1655. There was a strong relationship between HI and fHb: CCC=0.93 [95% confidence interval (95% CI), 0.92–0.94]. For the detection of fHb >100 mg dl−1 (n=58 [10%]), the performance of HI was excellent: area under the ROC curve (AUCROC) of 0.98 (95% CI, 0.85–1.0), sensitivity=86% (95% CI, 57–100) and specificity=97% (95% CI, 91–100) for HI >20. In the validation cohort, 700 samples from 95 patients were analysed. Median fHb was 4 mg dl−1 [IQR=0; 14], fHb ranged from 0 to 1526 mg dl−1, with fHb ≥100 mg dl−1 in 50 (7.1 %) samples. Median HI was 4 [IQR=2; 8], ranging from 0 to 1011. There was also a strong relationship between HI and fHb (CCC=0.95; 95% CI, 0.95–0.96). For the detection of fHb ≥100 mg dl−1 (n=50 [7%]), HI ≥20 had good diagnostic performance: AUCROC=0.99 (95% CI, 0.95–1), negative predictive value 0.97 (95% CI, 0.93–1), positive predictive value 0.86 (95% CI, 0.50–1), rate of correct classification of 96% (95% CI, 92–99%). The HI tended to have lower inter-measurement variability than fHb: for 25, 50, 100, and 200 mg dl−1 fHb solutions, the coefficient of variation was 9.4%, 9.0%, 6.6%, and 3.9% for fHb and 5.1%, 4.8%, 3.1%, and 0.4% for HI, respectively. In vitro, for a given degree of haemolysis (HI of 30), the gradual increase in icterus index (to 5, 10, 20, and 40) lowered the HI (–17%, –37%, –74%, and –100%, respectively). In contrast, in plasma with a baseline HI of 30, the gradual increase in lipaemic index (to 70, 125, 225, and 500) increased the HI (+11%, +14%, +25%, +56%, and +100%, respectively). The under-/overestimation of haemolysis by HI during severe icterus/hypertriglyceridaemia, respectively, was found with various degrees of haemolysis (baseline HI of 30, 60, 120, 250, or 500), but its magnitude tended to fall as haemolysis increased. Increasing propofol concentration (3, 6, and 10 μg ml−1) had an irrelevant impact on the HI. In our derivation and validation cohorts, median icterus index and lipaemic indices were 2 [IQR=1–5] and 24 [IQR=15–42], respectively. A model including HI, icterus, and lipaemic indices did not outperform HI alone (AUC of 0.99 vs 0.99). In our study, HI was a reliable marker of haemolysis during ECMO with low inter-measurement variability. A preliminary study (100 samples from in-house patients, not containing high degrees of haemolysis) found an excellent correlation (r=0.98) between HI and fHb.9Petrova D.T. Cocisiu G.A. Eberle C. et al.Can the Roche hemolysis index be used for automated determination of cell-free hemoglobin? A comparison to photometric assays.Clin Biochem. 2013; 46: 1298-1301Crossref PubMed Scopus (20) Google Scholar This was confirmed by a recent retrospective study in patients on ECMO.10Bosma M. Waanders F. Van Schaik H.P. et al.Automated and cost-efficient early detection of hemolysis in patients with extracorporeal life support: use of the hemolysis-index of routine clinical chemistry platforms.J Crit Care. 2019; 51: 29-33Crossref PubMed Scopus (2) Google Scholar Our prospective study goes further than these studies because we included more samples from more patients (1260 samples from 162 patients). We also assessed the performance of HI for the detection of fHb below or above a critical threshold of 100 mg dl−1.3Lou S. MacLaren G. Best D. Delzoppo C. Butt W. Hemolysis in pediatric patients receiving centrifugal-pump extracorporeal membrane oxygenation: prevalence, risk factors, and outcomes.Crit Care Med. 2014; 42: 1213-1220Crossref PubMed Scopus (92) Google Scholar,6Neal J.R. Quintana E. Pike R.B. Hoyer J.D. Joyce L.D. Schears G. Using daily plasma-free hemoglobin levels for diagnosis of critical pump thrombus in patients undergoing ECMO or VAD support.J Extra Corpor Technol. 2015; 47: 103-108PubMed Google Scholar We used a rigorous design with derivation and validation cohorts, assessed the inter-measurement variability of HI, and assessed the impact of icterus, hypertriglyceridemia, and propofol on the diagnostic performance of HI. Limitations of the study include: (1) the study was monocentric; (2) caution should be exercised before extrapolating the good performance we report for HI to other chemistry analysers or to haemolysis not related to ECMO. However, the assessment of the degree of haemolysis seems to be similar from one analyser to another and observing a lower performance of the HI with other haemolysis triggers seems unlikely7Nikolac Gabaj N. Miler M. Vrtarić A. et al.Precision, accuracy, cross reactivity and comparability of serum indices measurement on Abbott Architect c8000, Beckman Coulter AU5800 and Roche Cobas 6000 c501 clinical chemistry analyzers.Clin Chem Lab Med. 2018; 56: 776-788Crossref PubMed Scopus (20) Google Scholar,11Fernandez P. Llopis M.A. Perich C. et al.Harmonization in hemolysis detection and prevention. A working group of the Catalonian Health Institute (ICS) experience.Clin Chem Lab Med. 2014; 52: 1557-1568PubMed Google Scholar,12Lippi G. Favaloro E.J. Franchini M. Haemolysis index for the screening of intravascular haemolysis: a novel diagnostic opportunity?.Blood Transfus. 2018; 16: 433-437PubMed Google Scholar; and (3) we did not collect clinical data because this preliminary study was not designed to assess the relationship between high HI and patient outcomes. Our encouraging results pave the way for such future studies. In summary, as a fully automated, repeatable, and affordable gauge for ECMO-induced haemolysis, HI showed good performance and may replace non-routine measurements of fHb. Caution should be exercised in case of severe icterus and hypertriglyceridemia, but these conditions are also automatically quantified and displayed. The authors declare that they have no conflicts of interest.