Prognostic value of different lactate metrics in patients treated with peripheral veno-arterial extracorporeal membrane oxygenation (VA-ECMO)

Abstract Background Lactate has been associated with mortality in VA-ECMO, but performance varies by timing and metric (absolute value, early change or percentage clearance). Which measure offers the most clinically meaningful discrimination remains uncertain. Purpose To compare lactate-derived metrics for predicting in-hospital mortality in peripheral VA-ECMO. Methods Retrospective single-centre study of 72 consecutive adults treated with femoral VA-ECMO (2022–2025). Baseline characteristics, aetiology, management and laboratory data were recorded at initiation. Lactate was assessed at 0 h, 24 h and as peak within the first 24 h. We evaluated absolute 0→24 h change, 24-h percentage clearance, lactate at initiation, and 24-h peak lactate as predictors of in-hospital mortality. Receiver-operating characteristic (ROC) curves were constructed and area under the curve (AUC, 95% CI) estimated for each metric; complete 24-h data were available in 57 patients. We also built a basic logistic regression model (age, sex, diabetes mellitus, extracorporeal cardiopulmonary resuscitation ECPR, renal replacement therapy RRT) and assessed the incremental effect of adding each lactate: metric, comparing AUCs. Results 72 patients were included (median age 57 years; 82% men); ECPR occurred in 45/72 (62.5%). In-hospital mortality was 38/72 (52.8%). Patient characteristics are shown in Table 1. Survivors had longer duration of ECMO support (6 vs 2 days; p=0.0008), were younger (49.1 vs 56.1 years; p=0.049) and had lower baseline prothrombin time ratio (1.3 vs 1.7; p=0.049). Peak lactate within 24 h was higher in non-survivors (12.6 vs 6.15 mmol/l, p=0.0008) and showed the best standalone discrimination for in-hospital mortality (AUC 0.73, 95% CI 0.61–0.85; Figure 1A). In the complete-case subset (n=57), peak lactate tended to outperform 24-h percentage clearance (AUC 0.69 vs 0.52; p=0.069; Figure 1B). Absolute 0→24 h change did not improve upon clearance. In the multivariable model (age, sex, diabetes, ECPR, RRT), adding 24-h peak lactate increased discrimination from AUC 0.747 (95% CI 0.63–0.86) to 0.801 (95% CI 0.70–0.91; p=0.19) and remained independently associated with mortality (adjusted OR 1.21; 95% CI 1.05–1.40; p=0.009; Figure 1C). Incorporating absolute or relative 24-h lactate reduction produced minimal AUC changes (e.g., 0.721→0.741; p=0.44) and neither was independently associated (Figure 1D). Conclusions In this single-centre VA-ECMO cohort, lactate metrics differed in prognostic value. Peak lactate within the first 24 h was the only metric independently associated with in-hospital mortality and yielded the largest (though statistically non-significant) improvement in model discrimination, outperforming 24-h percentage clearance and absolute 0→24 h change, which showed little discriminative capacity either alone or when added to a clinical model.Table 1.Baseline characteristics Figure 1.ROC curves