Correlation between LDH and plasma-free haemoglobin in microaxial flow pump support: rethinking haemolysis criteria

Temporary mechanical circulatory support (tMCS) is increasingly used in patients with cardiogenic shock (CS). Haemolysis remains a frequent and clinically significant complication associated with adverse outcomes (e.g. acute kidney injury), making early identification essential. The haemolysis INTERMACS definition— (LDH) ≥2.5×ULN and plasma free haemoglobin (PFHb) ≥20mg/dL—is widely used, yet biochemical markers and thresholds vary across institutions. Although LDH is the most commonly used haemolysis marker, its elevation can reflect non-haemolytic cell injury (e.g. myocardial necrosis), limiting its specificity. Whether LDH accurately identifies haemolysis in microaxial flow pump (mAFP) patients compared with PFHb, the current gold standard, remains uncertain.

We retrospectively analysed all laboratory data from CS patients treated with mAFPs for ≥12 hours at our quaternary cardiac centre between May 2017 and January 2025. We defined haemolysis as PFHb ≥0.30mg/dL. Pearson’s correlation coefficient was used to assess relationships between PFHb and LDH.

A total of 193 patients were included (80% male, mean age 51.7±14.3 years). Most presented with SCAI-D CS. 62% (n=120) were supported with mAFP alone and 38% (n=73) with combined V-A ECMO and mAFP ("ECMELLA"). The most commonly used device was the Impella CP (81% of cases), followed by the Impella 5.0 (15%) and Impella 5.5 (4%). The median (IQR) duration of support was 6.0 (3.0–11.5) days with a median (IQR) flow of 3.0 (2.5–3.4) L/min. The predominant aetiologies of shock were acute myocardial infarction (45%) and dilated cardiomyopathy (40%)

A total of 884 paired PFHb–LDH measurements (53% of all available time points) were analysed. Based on our haemolysis definition, haemolysis was found in 40% of cases (355/884). PFHb and LDH showed weak correlation (r=0.398, p<0.001; Figure 1a). Receiver operator curve analysis (Figure 1b) showed an area under curve of 0.77 for LDH in predicting haemolysis, with sensitivity 73.5%, specificity 68.1%, positive predictive value 60%, and negative predictive value 79.4%. The optimal LDH cut-off was 1439 U/L. Using the INTERMACS haemolysis definition (LDH ≥2.5xULN, i.e., 1250U/L), sensitivity increased to 79.2% while specificity decreased to 59.7% (PPV=56.7%, NPV=81.1%), indicating a higher false-positive rate at this lower threshold.

In patients supported with mAFPs, LDH shows only a weak correlation with PFHb, likely reflecting underlying illness severity in relation to multiple organ failure/cell necrosis rather than haemolysis itself. Reliance on LDH alone may detect only around ≈60% of haemolysis cases and lead to a high rate of false positives. The optimal LDH cut-off in our cohort was higher than ≥2.5×ULN (1439 vs 1250U/L). These findings reinforce PFHb as the reference biochemical marker and highlight the need for caution when relying solely on LDH for haemolysis monitoring in mAFP patients.