Severe autoimmune hemolytic anemia in ICU: a place for emergency plasma exchange? A French multicenter retrospective study

Abstract

Background: Autoimmune hemolytic anemia (AIHA) is a rare but potentially life-threatening condition requiring intensive care unit (ICU) admission in severe cases. While corticosteroids and immunosuppressants are standard treatments, their delayed efficacy limits their utility in critical settings requiring rapid hemolysis control. Plasma exchange (PlEx) may offer a rapid intervention, but its effectiveness in severe AIHA remains uncertain. This study aims to assess the clinical characteristics, outcomes, and the potential benefit of PlEx in ICU cases with severe AIHA.

Methods: We conducted a multicenter retrospective cohort study including patients with severe AIHA admitted to 15 ICUs within the Assistance Publique-Hôpitaux de Paris (AP-HP) network between 2017 and 2024. Clinical, biological, and therapeutic data were collected. A multivariate logistic regression model, an analysis adjusted on a propensity score (PS) and on inverse probability of treatment weighting (IPTW) were used to identify predictors of in-ICU mortality and evaluate the association between PlEx and in-ICU mortality.

Results: One hundred forty-eight ICU stays involving severe AIHA hemolytic crises were analyzed. The median age at ICU admission was 61 {48-71} years, with a balanced sex ratio (51% male). Admission median hemoglobin was 5.2 g/dL {4.2-7.2} and in-ICU mortality was 17.6%. Risk factors for in-ICU mortality included age > 60 years, renal replacement therapy, mechanical ventilation, and high-dose intravenous methylprednisolone (HDIM). Therefore, ICU mortality was mainly driven by organ dysfunction rather than anemia severity. Standard corticosteroid therapy (1-2 mg/kg) (OR:0.19 [0.03-0.94], P = 0.05) and PlEx (OR:0.04 [0.004-0.4], P = 0.008) were associated with improved survival. The association between PlEx and outcome persisted after PS (OR: 0.04 [0.001-0.39]; P = 0.012) and IPTW (OR: 0.13 [0.02-0.55]; P = 0.011) adjustment.

Conclusion: In this large ICU cohort, severe AIHA was associated with high mortality, driven by organ failure rather than anemia severity. PlEx are associated with survival, suggesting its potential role as a bridge to immunosuppressive therapy in selected cases. Prospective studies are needed to confirm these findings.

Keywords: Autoimmune hemolytic anemia; Corticosteroids; Intensive care; Mortality; Plasma exchange.

Fig. 1

Symptoms at ICU admission. Reported symptoms at ICU admission, expressed as percentage of patient

Fig. 2

Underlying diseases and/or associated conditions in AIHA. Venn diagram representing the AIHA underlying disease. Overlaps represent comorbidities:*: 2 LGL, 2 B lymphomas, 1 CMML and 1 patient with both LGL and immune deficiency overlap with Evans syndrome; **: 1 CMV infection overlap with Evans syndrome (and LGL and immune deficiency, also represented in £ and *); $: 2 SLE and 1 IBD overlap with Evans syndrome; °: 1 other CTD overlap with tuberculosis; £: 2 T lymphomas overlap with HIV, 3 Castelman’s disease overlap with HIV, 1 LGL overlap with HIV, 1 CMV overlap with immune deficiency (and LGL and CMV infection, also represented in * and **). CMML, chronic myelomonocytic leukemia; CMV, cytomegalovirus; CLL, chronic lymphoid leukemia; CTD, connective tissue disease; EBV, Epstein Barr Virus; HCV, hepatitis C virus; HIV, human immunodeficiency virus; IBD, inflammatory bowel disease; ILD, interstitial lung disease; LGL, Large granular lymphocytic leukemia; MDS, myelodysplastic syndrome; Mycop. Pneum., Mycoplasma pneumoniae; PB19, Parvovirus B19 virus; RA, rheumatoid arthritis; SLE, systemic lupus erythematosus; HSCT, hematopoietic stem cell transplantation; BK, tuberculosis

Fig. 3

In-ICU mortality predictors multivariate analysis. A shows unadjusted multivariate analysis (logistic regression, full model) of in-ICU mortality risk factors. The dots represent the OR; the line through each dot corresponds to the 95% confidence interval. Goodness of fit (Hosmer–Lemeshow statistic): 7.61 (P = 0.47); Turj r2 = 0.67. Calibration: AUC = 0.96 (0.93–0.99); P < 0.0001. B shows OR of in-ICU mortality regarding the realization of PlEx according to the different multivariate logistic regression models. We compared the logistic regression (full model), to logistic regression adjusted on the propensity score, and IPTW. OR, Odd ratio; RRT, renal replacement therapy; HDIM, high-dose intravenous methylprednisolone; PS, propensity score; IPTW, Inverse probability of treatment weighting