Therapeutic plasma exchange and mortality in critically ill patients with sepsis-associated acute kidney injury: A retrospective cohort study using propensity score matching

Abstract

Objective

To assess the link between therapeutic plasma exchange (TPE) and mortality outcomes in intensive care unite (ICU) patients with sepsis-associated acute kidney injury (AKI).

Methods

This retrospective cohort study was conducted using data from the MIMIC-IV database. Adult ICU patients with sepsis-associated AKI were identified using Sepsis-3 criteria and the Kidney Disease: Improving Global Outcomes (KDIGO) criteria for AKI during ICU admission. Standard TPE was defined as a volume of approximately 1.0–1.5 times the total plasma volume. Kaplan–Meier analysis was used to estimate 30- and 90-day ICU mortality. Cox proportional hazards models were applied to evaluate the associations between TPE and mortality outcomes. Propensity score matching (PSM) was further performed to assess the robustness of the findings.

Results

A total of 17,533 patients with sepsis-associated AKI were included; 204 (1.2%) received adequate TPE. During the 90-day follow-up period, 5,458 patients (31.8%) in the non-TPE group and 119 patients (58.3%) in the TPE group died. In Cox proportional hazards models, TPE was associated with increased 30-day (hazard ratio [HR] = 2.36, 95% confidence interval [CI] 1.94–2.87) and 90-day mortality (HR = 2.30, 95% CI 1.92–2.76). After PSM with good covariate balance (all standardized mean differences <0.1), TPE remained associated with higher 30-day (HR = 1.68, 95% CI 1.22–2.31) and 90-day mortality (HR = 1.59, 95% CI 1.19–2.12).

Conclusions

In ICU patients with sepsis-associated AKI, TPE was associated with higher 30-day and 90-day ICU mortality, underscoring the need for careful patient selection and prospective studies to clarify its clinical role.

Keywords

sepsis acute kidney injury therapeutic plasma exchange propensity score matching intensive care unit mortality

Introduction

Sepsis remains a leading cause of morbidity and mortality in critically ill patients, particularly those admitted to the ICU.¹ The systemic inflammatory response triggered by infection in sepsis patients can lead to multiple organ dysfunction, including AKI.² The incidence of AKI in sepsis patients is approximately 47.1% to 61.7%,³ further complicating patient management. Among these patients, the mortality rate for ICU patients with sepsis complicated by AKI is as high as 41%.⁴ Effective therapeutic strategies are crucial for managing sepsis and its complications, such as AKI.

The pathophysiological mechanisms of sepsis complicated by acute kidney injury are complex, primarily involving the activation of the immune system, leading to the release of large amounts of inflammatory mediators such as cytokines, chemokines, interleukins, and tumor necrosis factors.⁵ Additionally, the systemic inflammatory response induced by sepsis can damage the endothelial cells of the microvasculature, causing endothelial dysfunction⁶ and increased microvascular permeability.⁷ This results in hemodynamic changes in the renal microvasculature, leading to tubular and interstitial edema in the kidneys.⁸ TPE is a blood purification technique that treats diseases by removing pathogenic substances from plasma.⁹ TPE involves drawing the patient’s blood out of the body, using centrifugation or membrane separation techniques to separate plasma from blood cells, discarding plasma containing pathogenic substances (such as autoantibodies, immune complexes, endotoxins, etc.), and replacing it with an equal amount of replacement fluid (such as fresh frozen plasma, albumin, or a mixture of crystalloid fluids).^9–11 Although TPE is believed to reduce systemic inflammation and improve microvascular perfusion, its potential adverse effects, such as hemodynamic risks (e.g., hypotension, arrhythmias),¹² coagulation disorders (risk of bleeding due to loss of coagulation factors),¹³ immune-related risks (increased infection risk due to immunoglobulin loss),¹⁴ and drug clearance issues (removal of antibiotics and other drugs) could exacerbate the condition.

Existing studies have shown that TPE is widely used in various autoimmune diseases, such as systemic lupus erythematosus,¹⁵ rheumatoid arthritis,¹⁶ and Guillain-Barré syndrome.¹⁷ TPE can remove autoantibodies, immune complexes, and inflammatory mediators from the blood, thus alleviating the immune system’s overreaction and reducing clinical symptoms.⁹ TPE also holds significant therapeutic potential for certain toxin poisonings, neuro-immunological diseases, and some blood disorders.^9,18,19 While removing harmful immune factors (such as autoantibodies), TPE may also remove immune factors that help fight infections.²⁰ Additionally, fluctuations in blood volume could exacerbate arrhythmia risk in patients with autonomic dysfunction,²¹ and the removal of coagulation factors (e.g., fibrinogen) may increase the risk of bleeding in patients with thrombocytopenia.²² The use of citrate anticoagulation during treatment may lead to hypocalcemia, which presents a greater risk for patients with renal dysfunction.²³ Currently, the safety and efficacy of TPE in ICU patients with sepsis-associated AKI remain unclear. Therefore, assessing the safety and efficacy of TPE in this patient population is crucial for guiding clinical decision-making and improving ICU treatment outcomes.

Methods

Data source

This retrospective cohort study was conducted using data from the MIMIC-IV database (version 3.0), released on December 19th, 2024 (https://mimic-iv.mit.edu/).²⁴ Developed by MIT’s Computer Science and AI Laboratory in collaboration with Beth Israel Deaconess Medical Center, MIMIC-IV is a publicly accessible database containing detailed ICU patient data from admission to discharge. The study used de-identified data, with approval from the Institutional Review Boards of Beth Israel Deaconess Medical Center (2001-P-001699/14) and MIT (No. 0403000206).^25,26 Local ethical approval and informed consent were not needed due to the de-identified nature of the data, in line with the 1975 Declaration of Helsinki as revised in 2024. Yang Chen, an author, holds a Human Subjects Research Training Certificate (No. 53753450).

Study population

The inclusion criteria for the study required patients to be diagnosed with sepsis based on the MIMIC IV version 3.0 ICU admission dataset and to have AKI. Exclusion criteria included patients diagnosed with sepsis more than 24 hours after ICU admission, those with an ICU stay of less than 24 hours, pregnant patients, patients with end-stage heart failure, renal failure, or chronic kidney disease, those with multiple hospital or ICU admissions, patients without AKI, patients undergoing plasma exchange without hematocrit records, and those with a plasma exchange ratio greater than 1.5. Ultimately, 17353 patients met the inclusion criteria for statistical analysis (Figure 1).

Figure 1.

Flowchart of this study. ICU, intensive care unit; MIMIC-IV, Medical Information Mart for Intensive Care IV.

Definition and clinical outcomes

Sepsis was defined as a life-threatening organ dysfunction caused by a dysregulated host response to infection, and the diagnostic criteria included suspected or documented infection and an acute increase in total Sequential Organ Failure Assessment (SOFA) score ≥2 points as a proxy for organ dysfunction.^27,28 The diagnostic criteria for AKI are primarily based on the guidelines published by the Kidney Disease: Improving Global Outcomes (KDIGO) organisation, defined as any of the following: an increase in serum creatinine of ≥26.5 μmol/L (0.3 mg/dL) within 48 hours; an increase in serum creatinine to ≥1.5 times the baseline value; or a urine output <0.5 mL/kg/h for more than 6 hours, baseline serum creatinine was defined as the lowest value recorded within one week prior to ICU admission.²⁹ The exposure variable was defined as TPE during ICU admission. Patients were divided into two groups: those who received TPE during ICU admission and those who did not. Formula for calculating plasma volume was as follows: Estimated plasma volume (in liters) = 0.07 × (1 − hematocrit) × body weight (in kilograms).⁹ The primary outcomes were ICU all-cause mortality rates at 30 days and 90 days after ICU admission.

Data extraction

Data extraction was performed using Structured Query Language. Clinical variables included study participant’s demographic information (age, male, body mass index [BMI]), vital signs at 1st admission (heart rate, systolic blood pressure [SBP], diastolic blood pressure [DBP], respiration rate, oxygen saturation); severity score at 1st admission (SOFA, Simplified Acute Physiology Score II [SAPSII]); comorbidities (myocardial infarct, congestive heart failure, cerebrovascular disease, chronic pulmonary disease, malignant cancer, diabetes mellitus, liver disease); laboratory results at 1st admission (hematocrit, hemoglobin, platelet, red blood cell distribution width [RDW], white blood cell, blood urea nitrogen, calcium, creatinine, sodium, potassium); procedures at 1st admission (renal replacement therapy, vasopressors, mechanical ventilation) and AKI stage.

Propensity score matching

We employed propensity score matching (PSM) to balance baseline characteristics between the plasma exchange and non-plasma exchange groups. To avoid issues of sample imbalance and multicollinearity, and to ensure the stability of the matching and weighting models, we used baseline covariates for propensity score estimation. These variables included age, male, BMI, heart rate, SBP, DBP, respiration rate, oxygen saturation, SOFA, SAPSII, myocardial infarct, congestive heart failure, cerebrovascular disease, chronic pulmonary disease, malignant cancer, diabetes mellitus, liver disease, hematocrit, hemoglobin, platelet, RDW, white blood cell, blood urea nitrogen, calcium, creatinine, sodium, potassium, renal replacement therapy, vasopressors, mechanical ventilation. For PSM, patients who underwent plasma exchange and those who did not were matched using 1:1 nearest neighbour matching. Given the characteristics of the weighted method, the balance of covariates in the weighted sample was also assessed using standardised mean difference (SMD), with SMD <0.1 considered an acceptable balance indicator, indicating good balance between groups.

Statistical analysis

The missing data for each variable are listed in Supplementary Table S1. We used multiple imputation by chained equations (30 imputations) with the ‘mice’ package in R to address missing values. Normality tests indicated that all continuous variables were non-normally distributed, so they are presented as medians with interquartile ranges (IQR). Group differences were assessed using Kruskal-Wallis tests. Categorical variables are shown as counts and percentages, with comparisons made using Fisher’s exact test or Chi-squared test. The Kaplan-Meier survival curves were plotted for the primary outcome in the cohort before and after PSM to visualize the differences in survival probabilities between groups. Subsequently, a Cox proportional hazards model was used to assess the association between plasma exchange and mortality risk in patients who underwent plasma exchange versus those who did not. Hazard ratios (HR) and their 95% confidence intervals (CI) were estimated. For the primary outcome, the analysis was performed in the initial cohort, and three models were developed to explore the association between plasma exchange and mortality risk in patients with sepsis complicated by AKI. Model 1: unadjusted. Model 2: adjusted for age, sex, body mass index, vital signs (heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, oxygen saturation), severity score (sequential organ failure assessment, simplified acute physiology score II). Model 3: Model 2 further adjusted for comorbidities (acute kidney injury stage, myocardial infarction, congestive heart failure, cerebrovascular disease, chronic pulmonary disease, malignant cancer, diabetes mellitus, liver disease), lab tests (hematocrit, hemoglobin, platelet, red cell distribution width, white blood cell, blood urea nitrogen, calcium, creatinine, sodium, potassium), and procedures (dialysis, vasopressors, mechanical ventilation). The analysis was then repeated after PSM. All statistical analyses were performed using SPSS Statistics (version 27) and R software (version 4.4.2). A two-tailed P value < 0.05 was considered statistically significant. This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.³⁰

Results

Baseline characteristics

The baseline characteristics of patients with sepsis-associated AKI who received TPE and those who did not, both before and after PSM the entire cohort, are detailed in Table 1. The study included 17,353 eligible participants, 17,149 of whom did not undergo TPE, and 204 of whom underwent TPE. Before matching in non-TPE group, with a IQR age of 69.02 (58.35, 79.29) years and a male predominance (58.1%), before matching in TPE group, with a IQR age of 64.21 (52.62, 75.33) years and a male predominance (61.3%). Compare to non-TPE group, TPE group had a lower BMI, higher proportion of patients in AKI stage III (69.6%), higher SOFA score (P < 0.001). However, comorbid conditions were more prevalent in the non-TPE group, with significantly higher rates of myocardial infarction, congestive heart failure, cerebrovascular disease, and chronic pulmonary disease.

Table 1.

Baseline characteristics of TPE and non-TPE groups before and after PSM.

	Before PSM			After PSM
Variables	Non-TPE	TPE	P	Non-TPE	TPE	P
N	17149	204		195	195
Age, years	69.02 (58.35, 79.29)	64.21 (52.62, 75.33)	< 0.001	62.93 ± 15.99	63.44 ± 15.60	0.751
Male, n (%)	9963 (58.1)	125 (61.3)	0.392	113 (57.9)	120 (61.5)	0.536
Body mass index, kg/m²	28.31 (24.34, 33.46)	26.37 (23.28, 31.94)	0.002	27.35 (23.67, 31.30)	26.40 (23.07, 31.99)	0.783
AKI Stage, n (%)			< 0.001			0.852
AKI stage I	3678 (21.4)	15 (7.4)		13 (6.7)	15 (7.7)
AKI stage II	7496 (43.7)	47 (23.0)		51 (26.2)	47 (24.1)
AKI stage III	5975 (34.8)	142 (69.6)		131 (67.2)	133 (68.2)
Vital signs at 1^st admission
Heart rate, beats/min	89.00 (77.00, 104.00)	91.00 (81.25, 105.00)	0.031	88.00 (78.00, 108.00)	90.00 (81.00, 105.00)	0.421
SBP, beats/min	119.00 (103.00, 137.00)	113.00 (97.00, 130.00)	< 0.001	110.00 (97.00, 125.00)	113.00 (97.00, 131.00)	0.716
DBP, beats/min	65.00 (55.00, 77.00)	60.00 (51.00, 73.00)	0.002	62.00 (51.00, 75.00)	60.00 (51.00, 74.00)	0.935
Respiration rate, beats/min	19.00 (16.00, 24.00)	18.00 (15.00, 23.00)	0.142	18.00 (14.00, 22.00)	19.00 (15.00, 23.00)	0.211
Oxygen Saturation, %	98.00 (95.00, 100.00)	99.00 (95.00, 100.00)	0.138	99.00 (95.00, 100.00)	99.00 (96.00, 100.00)	0.860
Severity score at 1^st admission
SOFA	3.00 (2.00, 5.00)	5.00 (3.00, 8.00)	< 0.001	4.00 (3.00, 6.00)	5.00 (3.00, 7.00)	0.225
SAPSII	42.00 (34.00, 52.00)	50.00 (41.00, 61.00)	< 0.001	48.00 (36.00, 60.00)	49.00 (41.00, 61.00)	0.188
Comorbidities, n (%)
Myocardial infarct	3501 (20.4)	32 (15.7)	0.100	36 (18.5)	32 (16.4)	0.689
Congestive heart failure	6130 (35.7)	55 (27.0)	0.011	52 (26.7)	55 (28.2)	0.821
Cerebrovascular disease	2723 (15.9)	19 (9.3)	0.011	22 (11.3)	18 (9.2)	0.617
Chronic pulmonary disease	4788 (27.9)	42 (20.6)	0.024	46 (23.6)	42 (21.5)	0.716
Malignant cancer	2631 (15.3)	29 (14.2)	0.762	29 (14.9)	27 (13.8)	0.885
Diabetes mellitus	5810 (33.9)	46 (22.5)	< 0.001	37 (19.0)	45 (23.1)	0.384
Liver disease	2867 (16.7)	101 (49.5)	< 0.001	87 (44.6)	92 (47.2)	0.684
Laboratory results at 1^st admission
Hematocrit, %	31.30 (26.80, 36.40)	28.20 (23.93, 33.43)	< 0.001	28.20 (24.30, 33.20)	28.50 (24.10, 33.50)	0.963
Hemoglobin, g/L	10.20 (8.70, 11.90)	9.25 (7.62, 11.00)	< 0.001	9.20 (7.90, 11.00)	9.40 (7.80, 11.10)	0.946
Platelet, ×10^∧9/L	183.00 (126.00, 254.00)	128.00 (73.00, 189.00)	< 0.001	132.00 (88.00, 186.00)	129.00 (78.00, 193.00)	0.368
RDW, %	14.80 (13.70, 16.50)	15.60 (14.20, 17.30)	< 0.001	15.90 (14.20, 18.20)	15.60 (14.20, 17.20)	0.555
White blood cell, ×10^∧9/L	12.10 (8.40, 16.90)	11.45 (7.35, 18.20)	0.405	11.30 (7.40, 17.00)	11.50 (7.60, 18.20)	0.641
Blood urea nitrogen, mg/dL	23.00 (15.00, 38.00)	24.50 (15.00, 39.00)	0.316	23.00 (15.00, 38.00)	24.00 (15.00, 40.00)	0.415
Calcium, mg/dL	8.20 (7.70, 8.70)	8.10 (7.40, 8.80)	0.218	8.30 (7.70, 8.80)	8.10 (7.40, 8.80)	0.431
Creatinine, mg/dL	1.10 (0.80, 1.80)	1.30 (0.90, 2.20)	0.007	1.20 (0.70, 2.20)	1.30 (0.80, 2.20)	0.281
Sodium, mmol/L	139.00 (135.00, 141.00)	139.00 (136.00, 142.00)	0.133	139.00 (136.00, 143.00)	139.00 (136.00, 142.00)	0.259
Potassium, mmol/L	4.20 (3.80, 4.70)	4.20 (3.80, 4.90)	0.184	4.30 (3.80, 4.90)	4.20 (3.80, 4.80)	0.887
Procedures at 1^st admission, n (%)
Renal replacement therapy	1254 (7.3)	48 (23.5)	< 0.001	45 (23.1)	43 (22.1)	0.904
Vasopressors	9624 (56.1)	181 (88.7)	< 0.001	171 (87.7)	172 (88.2)	1.000
Mechanical ventilation	11595 (67.6)	188 (92.2)	< 0.001	177 (90.8)	179 (91.8)	0.858

Abbreviations: AKI, acute kidney injury; DBP, diastolic blood pressure; PSM, propensity score matching; RDW, red blood cell distribution width; SAPS II, Simplified Acute Physiology Score II; SBP, systolic blood pressure; SOFA, Sequential Organ Failure Assessment; TPE, therapeutic plasma exchange.

Study outcomes by TPE status

Figure 2 presented mortality outcomes comparing ICU patients who received TPE and those who did not. The 30-day ICU all-cause mortality rate was 25.5%, with the non-TPE group at 25.2% and the TPE group at 50.0% (P < 0.001). For the 90-day ICU mortality, the rate was 32.1% overall, with 31.8% in the non-TPE group and 58.3% in the TPE group (P < 0.001). The length of ICU stay was significantly longer for the TPE group (8.91 days) compared to the non-TPE group (4.31 days) (P < 0.001).

Figure 2.

Distribution of study outcomes by TPE status. ICU, intensive care unit; TPE, therapeutic plasma exchange.

PSM and covariate balance

After performing PSM, a matched cohort of 390 patients was obtained, with 195 in each group, showing a significant reduction in covariate imbalance (Table 1). The quality of the matched samples was confirmed by plotting the probability density distributions of the two groups and evaluating SMDs, as shown in Supplementary Figure 1. In the matched cohort, all covariates exhibited SMD values below and almost 0.1, indicating good balance.

Association between TPE and 30-day ICU all-cause mortality before and after PSM

Before PSM, the 30-day ICU mortality was 25.2% (4319/17149) in the non-TPE group and 50.0% (102/204) in the TPE group. Figure 3 showed the Kaplan-Meier curve for 30-day ICU all-cause mortality according to TPE in the cohort before (P < 0.001) and after (P = 0.0013) PSM. TPE was associated with higher 30-day all-cause mortality (Model 1 HR = 2.36, 95% CI: 1.94-2.87, P < 0.001; Model 2 HR = 1.72, 95% CI: 1.41-2.10, P < 0.001; Model 3 HR = 1.47, 95% CI: 1.20-1.79, P < 0.001) in the initial cohort and the cohort after PSM (HR = 1.68, 95% CI: 1.22-2.31, P = 0.001) (Tables 2 and 3).

Figure 3.

Kaplan–Meier curves for ICU all-cause mortality by TPE status before and after PSM. (A) 30-day mortality before PSM; (B) 90-day mortality before PSM; (C) 30-day mortality after PSM; (D) 90-day mortality after PSM. Differences in survival were assessed using the log-rank test. ICU, intensive care unit; PSM, propensity score matching; TPE, therapeutic plasma exchange.

Table 2.

Associations of TPE with 30-day and 90-day all-cause mortality before PSM.

Outcomes	Group	Events/Samples (%)	Model 1		Model 2		Model 3
Outcomes	Group	Events/Samples (%)	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P
30-day all-cause mortality
Non-TPE		4319/17149 (25.2)	Reference	< 0.001	Reference	< 0.001	Reference	< 0.001
TPE		102/204 (50.0)	2.36 (1.94, 2.87)	< 0.001	1.72 (1.41, 2.10)	< 0.001	1.47 (1.20, 1.79)	< 0.001
90-day all-cause mortality
Non-TPE		5458/17149 (31.8)	Reference	< 0.001	Reference	< 0.001	Reference	< 0.001
TPE		119/204 (58.3)	2.30 (1.92, 2.76)	< 0.001	1.74 (1.45, 2.09)	< 0.001	1.55 (1.29, 1.86)	< 0.001

Model 1: unadjusted.

Model 2: adjusted for age, sex, body mass index, vital signs (heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, oxygen saturation), severity score (sequential organ failure assessment, simplified acute physiology score II).

Model 3: Model 2 further adjusted for commodities (acute kidney injury stage, myocardial infarct, congestive heart failure, cerebrovascular disease, chronic pulmonary disease, malignant cancer, diabetes mellitus, liver disease), lab tests (hematocrit, hemoglobin, platelet, red cell distribution width, white blood cell, blood urea nitrogen, calcium, creatinine, sodium, potassium), procedures (dialysis, vasopressors, mechanical ventilation).

Abbreviations: CI, confidence interval; HR, hazard ratio; ICU, intensive care unit; PSM, propensity score matching; TPE, therapeutic plasma exchange.

Table 3.

Associations of TPE with 30-day and 90-day ICU all-cause mortality after PSM.

Outcomes	Group	Events/Samples (%)	HR (95% CI)	P
30-day ICU all-cause mortality
	Non-TPE	63/195 (32.3)	Reference	0.001
	TPE	94/195 (48.2)	1.68 (1.22, 2.31)	0.001
90-day ICU all-cause mortality
	Non-TPE	81/195 (41.5)	Reference	0.002
	TPE	111/195 (56.9)	1.59 (1.19, 2.12)	0.002

Abbreviations: CI, confidence interval; HR, hazard ratio; ICU, intensive care unit; PSM, propensity score matching; TPE, therapeutic plasma exchange.

Association between TPE and 90-day ICU all-cause mortality before and after PSM

Before PSM, the 90-day ICU mortality was 31.8% (5458/17149) in the non-TPE group and 58.3% (119/204) in the TPE group. Figure 3 showed the Kaplan-Meier curve for 90-day ICU all-cause mortality according to TPE in the cohort before (P < 0.001) and after (P = 0.0014) PSM. TPE was also associated with higher 90-day all-cause mortality (Model 1 HR = 2.30, 95% CI: 1.92-2.76, P < 0.001; Model 2 HR = 1.74, 95% CI: 1.45-2.09, P < 0.001; Model 3 HR = 1.55, 95% CI: 1.29-1.86, P < 0.001) in the initial cohort and the cohort after PSM (HR = 1.59, 95% CI: 1.19-2.12, P = 0.002) (Tables 2 and 3).

Discussions

In this study, we examined the outcomes of ICU patients with sepsis-associated AKI who received TPE and compared them with those who did not undergo this treatment. Our analysis revealed a significant association between TPE and higher ICU mortality rates at both 30 and 90 days, even after performing PSM. Specifically, the 30-day ICU all-cause mortality rate was notably higher in the TPE group (50.0%) compared to the non-TPE group (25.2%), with similar trends observed at the 90-day mark (58.3% in the TPE group vs. 31.8% in the non-TPE group). These differences persisted across both unadjusted and adjusted models, supporting the consistency of this association across analytical approaches.

Previous literature has suggested that TPE may offer clinical benefits in septic patients through the removal of pro-inflammatory mediators, immune complexes, and other circulating pathogenic substances, thereby potentially mitigating systemic inflammation and improving microvascular perfusion.⁹ However, our findings indicate that in real-world ICU settings, the application of TPE in sepsis patients with AKI is associated with increased short- and long-term mortality. This could be partially explained by several pathophysiological and clinical fact. Firstly, immune dysfunction is commonly observed in sepsis patients.³¹ TPE can provide short-term relief from inflammation by clearing inflammatory mediators and immune complexes; however, it may excessively suppress the immune system.³² Long-term or frequent TPE may increase the risk of secondary infections, thereby worsening the patient’s condition.³³ Secondly, sepsis patients with AKI may already have electrolyte imbalances, and during TPE, important ions such as sodium, potassium, calcium, and magnesium are removed from the plasma.³⁴ If not properly supplemented, this could exacerbate hyponatremia, hypokalemia, and other conditions, potentially leading to arrhythmias or muscle weakness.^35,36 Moreover, sepsis patients often suffer from hypotension and inadequate perfusion, and TPE may further exacerbate renal ischemia, leading to the deterioration of AKI.³⁷ TPE requires catheterization or intubation procedures, which increase the risks of vascular injury, infection, and phlebitis.⁹ Additionally, thrombus formation or bleeding, indicative of coagulation disorders, may also occur.³⁸ TPE is a technique that relies heavily on specialized procedures, and failures in treatment can occur, leading to prolonged hospital stays and an increased healthcare burden.³⁸ Finally, the cohort of patients receiving TPE presented with more severe illness at baseline, as evidenced by higher SOFA scores and a higher proportion of AKI stage III. While PSM minimized baseline imbalances, residual confounding due to unmeasured factors such as clinician judgment, timing of TPE initiation, and underlying immunological status may have contributed to worse outcomes in the TPE group.

In addition, the relatively small proportion of patients receiving TPE in our cohort suggests that this therapy may have been selectively applied to the most critically ill patients. In real-world ICU practice, extracorporeal therapies such as TPE are often used as rescue or last-line interventions in patients with refractory organ failure. Consequently, the observed association between TPE and increased mortality may partly reflect confounding by indication rather than a direct harmful effect of the intervention itself. Although propensity score matching and multivariable adjustment were performed to balance baseline characteristics, residual confounding related to clinician decision-making, treatment timing, and underlying immunological status cannot be fully excluded. Furthermore, detailed information regarding the timing of TPE initiation, session number, and replacement fluids was not available in the MIMIC IV database, which further limits causal interpretation of the observed associations. In particular, the MIMIC IV database does not allow reliable identification of early initiation of TPE after ICU admission, which has been emphasized in several ongoing prospective trials and may represent an important factor influencing treatment efficacy.

Our findings differ from some earlier studies, which suggested the potential benefits of TPE in sepsis.³⁹ Several small clinical trials have reported that TPE can improve short-term outcomes in patients with thrombocytopenic multiple organ failure (TAMOF),⁴⁰ disseminated intravascular coagulation (DIC),¹³ or high inflammatory phenotypes in SARS patients.⁴¹ For children with critical heart disease clinically diagnosed with TAMOF, TPE may be associated with improvements in organ failure and platelet count.⁴² However, a study including 2,772 critically ill patients found that TPE was not associated with improvements in organ failure and mortality in patients with severe sepsis, and it may be related to prolonged ICU length of stay.⁴³ In contrast, this study, based on a large real-world ICU cohort, found that TPE was associated with increased ICU mortality at 30 and 90 days in patients with sepsis complicated by acute kidney injury, even after propensity score matching and adjustment for confounding factors. In particular, differences compared with studies using the same database, such as Yan et al.’s study, may reflect variation in patient selection and data granularity.⁴⁴ Their study focused on a more homogeneous subgroup, whereas our cohort captures broader real-world heterogeneity. These differences may be further explained by several factors. First, previous studies often included highly selective patients with specific indications, such as TAMOF or severe DIC, whereas our cohort reflects a broader ICU population with sepsis-associated AKI. The lack of biomarker-guided stratification in real-world practice may have diluted the potential benefits of TPE observed in more targeted populations. Second, patients who received TPE in this study had significantly higher disease severity, as indicated by higher SOFA scores and a greater proportion of stage III AKI, suggesting that TPE may have been used as a rescue therapy in more critically ill patients. This selection bias may partially explain the observed higher mortality. Third, prior studies often emphasized early initiation and standardized protocols, whereas our retrospective design could not account for key treatment factors such as timing, frequency, or replacement fluid composition, all of which are important determinants of treatment outcomes.

TPE should also be interpreted within the broader context of extracorporeal blood purification therapies (EBPT) in sepsis. In addition to TPE, other EBPT approaches such as hemoperfusion, adsorption-based devices, and renal replacement therapy–based blood purification techniques have been explored for the removal of circulating inflammatory mediators in septic patients.⁴⁵ However, these modalities differ substantially in their mechanisms and therapeutic targets. TPE is a non-selective technique that removes and replaces plasma components, whereas adsorption-based therapies aim to eliminate circulating mediators without plasma replacement. Current evidence suggests that the clinical efficacy of these approaches remains uncertain and may depend strongly on patient selection, timing of initiation, and treatment protocols.⁴⁶ Therefore, our findings should be interpreted specifically in the context of TPE rather than extrapolated to EBPT as a whole.

In addition, future studies may also evaluate short-term physiological responses to TPE, such as changes in vasopressor requirements, hemodynamic parameters, or lactate clearance, which may more directly reflect the potential immunomodulatory and circulatory effects of TPE. Such dynamic indicators could provide complementary mechanistic insights beyond mortality outcomes and help clarify which patient subgroups might derive the greatest benefit from this therapy. Renal outcomes, including recovery of kidney function after TPE, may also represent important endpoints for future investigation.

However, this study has several limitations. First, as a retrospective observational study, despite the use of propensity score matching and multivariable analysis, there may still be unmeasured confounders. Second, the relative relationship between the timing of TPE treatment and the onset of disease cannot be determined, and therefore, causal relationships cannot be established. Third, the dynamic changes in the clinical condition of sepsis patients with acute kidney injury over time may introduce variability into our prognostic analysis. In addition, some important clinical variables such as vasopressor exposure were captured as binary variables rather than quantitative measures, which may not fully reflect treatment intensity. Fourth, detailed information regarding TPE treatment characteristics, including indications, timing of initiation, number of sessions, treatment duration, frequency, substitution fluids, and treatment modality, could not be reliably obtained from the MIMIC IV database, which may influence patient outcomes and limit the interpretation of treatment effects. Fifth, information regarding the specific technical details of TPE, such as the materials or devices used (e.g., haemoperfusion columns or filters), was not available in the MIMIC IV database. Sixth, although missing data were handled using multiple imputation (a method that could theoretically introduce bias), the proportion of missing values was relatively low. Finally, the relative relationship between the detailed information regarding TPE’s indications, duration, frequency, and replacement fluids cannot be precisely determined, which may be associated with patient mortality and affect the interpretation of prognosis. Therefore, future multicenter prospective clinical trials are needed to clarify the relationship between TPE and prognosis in patients with sepsis complicated by AKI.

Conclusions

In this large cohort of ICU patients with sepsis-associated acute kidney injury, TPE use was consistently associated with higher ICU all-cause mortality at both 30 and 90 days, as well as prolonged ICU length of stay. Although these associations persisted after extensive multivariable adjustment and propensity score matching, the findings likely reflect a combination of residual confounding and treatment selection toward the sickest patients. Clinically, our results suggest that TPE identifies a very high-risk subgroup among sepsis-associated AKI patients, and underscore the need for careful patient selection and clearer indications. Prospective studies are required to determine whether specific phenotypes of sepsis-associated AKI may derive benefit from TPE, and to clarify its optimal timing and therapeutic role.

Supplemental material

Supplemental material - Therapeutic Plasma Exchange and Mortality in Critically Ill Patients With Sepsis-Associated Acute Kidney Injury: A Retrospective Cohort Study Using Propensity Score Matching

Supplemental material for Therapeutic Plasma Exchange and Mortality in Critically Ill Patients With Sepsis-Associated Acute Kidney Injury: A Retrospective Cohort Study Using Propensity Score Matching by Shu Zhang, Yang Chen, Kaihe Xie, Na Xu, Ting Liu, Jipeng Liu, Lin Wang and Qingzhen Han in Science Progress.

Footnotes

Acknowledgements

We gratefully acknowledge the MIMIC-IV research team and sincerely thank all participants who contributed to this study.

ORCID iDs

Shu Zhang

Yang Chen

Ethical considerations

Ethical approval for data collection was approved by the institutional review boards of the BIDMC (2001-P-001699/14) and Massachusetts Institute of Technology (No.0403000206). In this work, all data were anonymised and could therefore be exempt from local ethics review committee.

Authors contributions

Shu Zhang: Data curation, Formal analysis, Conceptualization, Methodology, Writing-original draft. Yang Chen: Data curation, Formal analysis, Conceptualization, Methodology, Writing-review & editing. Kaihe Xie: Writing-review & editing. Na Xu: Writing-review & editing. Ting Liu: Writing-review & editing. Jipeng Liu: Writing-review & editing. Lin Wang: Conceptualization, Methodology, Formal analysis, Writing – review & editing, Supervision. QingZhen Ha: Conceptualization, Methodology, Formal analysis, Writing – review & editing, Supervision. All named authors meet the criteria for authorship as outlined by the International Committee of Medical Journal Editors, take collective responsibility for the integrity of the work, and have provided their approval for its publication.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Suzhou Industrial Park Medical Innovation Research Project (Grant No.CXYJ2024A08)

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Our data was obtained from MIMIC-IV3.0. This data can be found here: MIMIC-IV v3.0 (physionet.org). Data used in this study will be made available on reasonable request to corresponding author.*

Supplemental material

Supplemental material for this article is available online.

Appendix

References

Rhee

Dantes

Epstein

, et al. Incidence and Trends of Sepsis in US Hospitals Using Clinical vs Claims Data, 2009-2014. Jama 2017; 318: 1241–1249. https://doi.org/10.1001/jama.2017.13836

Srzić

Nesek Adam

Tunjić Pejak

. SEPSIS DEFINITION: WHAT'S NEW IN THE TREATMENT GUIDELINES. Acta clinica Croatica 2022; 61: 67–72. https://doi.org/10.20471/acc.2022.61.s1.11

Ling

Zhang

Chang

, et al. Population Sepsis Incidence, Mortality, and Trends in Hong Kong Between 2009 and 2018 Using Clinical and Administrative Data. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America 2025; 80: 91–100. https://doi.org/10.1093/cid/ciad491

Aguilar

AlHussen

Gandhi

, et al. Sepsis-Associated Acute Kidney Injury: Pathophysiology and Treatment Modalities. Cureus 2024; 16: e75992. https://doi.org/10.7759/cureus.75992

Cao

Chai

. Pathological alteration and therapeutic implications of sepsis-induced immune cell apoptosis. Cell death & disease 2019; 10: 782. https://doi.org/10.1038/s41419-019-2015-1

Molema

Zijlstra

van Meurs

, et al. Renal microvascular endothelial cell responses in sepsis-induced acute kidney injury. Nature reviews Nephrology 2022; 18: 95–112. https://doi.org/10.1038/s41581-021-00489-1

Souza

Yuen

Star

. Microparticles: markers and mediators of sepsis-induced microvascular dysfunction, immunosuppression, and AKI. Kidney international 2015; 87: 1100–1108. https://doi.org/10.1038/ki.2015.26

Peerapornratana

Manrique-Caballero

Gómez

, et al. Acute kidney injury from sepsis: current concepts, epidemiology, pathophysiology, prevention and treatment. Kidney international 2019; 96: 1083–1099. https://doi.org/10.1016/j.kint.2019.05.026

Cervantes

Bloch

Sperati

. Therapeutic Plasma Exchange: Core Curriculum 2023. American journal of kidney diseases : the official journal of the National Kidney Foundation 2023; 81: 475–492. https://doi.org/10.1053/j.ajkd.2022.10.017

10.

Ahmed

Kaplan

. Therapeutic Plasma Exchange Using Membrane Plasma Separation. Clinical journal of the American Society of Nephrology : CJASN 2020; 15: 1364–1370. https://doi.org/10.2215/cjn.12501019

11.

Sanchez

Balogun

. Therapeutic Plasma Exchange in the Critically Ill Patient: Technology and Indications. Advances in chronic kidney disease 2021; 28: 59–73. https://doi.org/10.1053/j.ackd.2021.03.005

12.

Lahmer

Messer

Schnappauf

, et al. Impact of Therapeutic Plasma Exchange on Hemodynamic Parameters in Medical Intensive Care Unit Patients: An Observational Study. Artificial organs 2017; 41: 204–209. https://doi.org/10.1111/aor.12734

13.

Weng

Chen

Fang

, et al. Therapeutic Plasma Exchange Protects Patients with Sepsis-Associated Disseminated Intravascular Coagulation by Improving Endothelial Function. Clinical and applied thrombosis/hemostasis : official journal of the International Academy of Clinical and Applied Thrombosis/Hemostasis 2021; 27: 10760296211053313. https://doi.org/10.1177/10760296211053313

14.

Stahl

Bikker

Seeliger

, et al. Effect of Therapeutic Plasma Exchange on Immunoglobulin Deficiency in Early and Severe Septic Shock. Journal of intensive care medicine 2021; 36: 1491–1497. https://doi.org/10.1177/0885066620965169

15.

Soyuöz

Karadağ

Karaağaç

, et al. Therapeutic plasma exchange for refractory SLE: A comparison of outcomes between different sub-phenotypes. European journal of rheumatology 2018; 5: 32–36. https://doi.org/10.5152/eurjrheum.2017.17088

16.

Green

Gonzalez

Alperin

, et al. The utility of therapeutic plasma exchange in Hyperviscosity syndrome associated with juvenile rheumatoid arthritis: A case report. Journal of clinical apheresis 2021; 36: 658–663. https://doi.org/10.1002/jca.21903

17.

Haridy

Shehab

Khedr

. Long-term outcomes of plasma exchange versus intravenous immunoglobulin for the treatment of Guillain-Barré Syndrome: A double-blind, randomized clinical trial. Restorative neurology and neuroscience 2023; 41: 203–217. https://doi.org/10.3233/rnn-231369

18.

Schutt

Ronco

Rosner

. The role of therapeutic plasma exchange in poisonings and intoxications. Seminars in dialysis 2012; 25: 201–206. https://doi.org/10.1111/j.1525-139X.2011.01033.x

19.

Rattanathamsakul

Siritho

Viswanathan

, et al. Facilities, selection, outcome measurement, and limitations of therapeutic plasma exchange for neuroimmunological disorders: The South East Asian survey study. Journal of clinical apheresis 2023; 38: 437–446. https://doi.org/10.1002/jca.22047

20.

Fodil

Urbina

Bredin

, et al. Bloodstream infections among patients receiving therapeutic plasma exchanges in the intensive care unit: a 10 year multicentric study. Annals of intensive care 2024; 14: 117. https://doi.org/10.1186/s13613-024-01346-7

21.

Berkelmans

GFN

Kuipers

Westerhof

, et al. Comparing volume-clamp method and intra-arterial blood pressure measurements in patients with atrial fibrillation admitted to the intensive or medium care unit. Journal of clinical monitoring and computing 2018; 32: 439–446. https://doi.org/10.1007/s10877-017-0044-9

22.

Lim

Leung

Manitta

, et al. A comparison of global coagulation assays between normal controls and patients with thrombocytopenia. International journal of laboratory hematology 2019; 41: 184–191. https://doi.org/10.1111/ijlh.12941

23.

Hsu

Bai

, et al. Regional citrate versus heparin anticoagulation for continuous renal replacement therapy: a meta-analysis of randomized controlled trials. American journal of kidney diseases : the official journal of the National Kidney Foundation 2012; 59: 810–818. https://doi.org/10.1053/j.ajkd.2011.11.030

24.

Johnson

AEW

Bulgarelli

Shen

, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data 2023; 10: 1. https://doi.org/10.1038/s41597-022-01899-x

25.

Chen

Yang

Liu

, et al. Prognostic value of glycaemic variability for mortality in critically ill atrial fibrillation patients and mortality prediction model using machine learning. Cardiovascular Diabetology 2024; 23: 426. https://doi.org/10.1186/s12933-024-02521-7

26.

Chen

Ming

, et al. Red Blood Cell Distribution Width as a Risk Factor for 30/90-Day Mortality in Patients with Gastrointestinal Bleeding: Analysis of the MIMIC-IV Database. Dig Dis Sci 2024; 69: 1740–1754. https://doi.org/10.1007/s10620-024-08295-y

27.

Singer

Deutschman

Seymour

, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). Jama 2016; 315: 801–810. https://doi.org/10.1001/jama.2016.0287

28.

Liu

Chen

Yang

, et al. Association between alactic base excess on mortality in sepsis patients: a retrospective observational study. J Intensive Care 2025; 13: 20. https://doi.org/10.1186/s40560-025-00789-9

29.

KDIGO 2024 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease. Kidney international 2024; 105: S117-s314. https://doi.org/10.1016/j.kint.2023.10.018

30.

von

Altman

Egger

, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med 2007; 147: 573–577. https://doi.org/10.7326/0003-4819-147-8-200710160-00010

31.

Prescott

Angus

. Enhancing Recovery From Sepsis: A Review. Jama 2018; 319: 62–75. https://doi.org/10.1001/jama.2017.17687

32.

Liu

Wei

, et al. Plasmapheresis, immunosuppressive therapy and anti-GBM disease prognosis: a cohort study of 107 patients. Renal failure 2024; 46: 2400539. https://doi.org/10.1080/0886022x.2024.2400539

33.

François

Daubin

Menouche

, et al. Adverse Events and Infectious Complications in the Critically Ill Treated by Plasma Exchange: A Five-Year Multicenter Cohort Study. Critical care explorations 2023; 5: e0988. https://doi.org/10.1097/cce.0000000000000988

34.

Guo

Yang

, et al. Effect of propofol versus midazolam on short-term outcomes in patients with sepsis-associated acute kidney injury. Frontiers in medicine 2024; 11: 1415425. https://doi.org/10.3389/fmed.2024.1415425

35.

Baeg

Lee

Jeon

, et al. Management for Electrolytes Disturbances during Continuous Renal Replacement Therapy. Electrolyte & blood pressure : E & BP 2022; 20: 64–75. https://doi.org/10.5049/ebp.2022.20.2.64

36.

Cabral

BMI

Edding

Portocarrero

, et al. Rhabdomyolysis. Disease-a-month. DM 2020; 66: 101015. https://doi.org/10.1016/j.disamonth.2020.101015

37.

Sun

Zhang

Tian

, et al. Mitochondria in Sepsis-Induced AKI. Journal of the American Society of Nephrology : JASN 2019; 30: 1151–1161. https://doi.org/10.1681/asn.2018111126

38.

Mahmoud

Buhler

Chu

, et al. Drug Dosing in Patients Undergoing Therapeutic Plasma Exchange. Neurocritical care 2021; 34: 301–311. https://doi.org/10.1007/s12028-020-00989-1

39.

Lee

OPE

Kanesan

Leow

, et al. Survival Benefits of Therapeutic Plasma Exchange in Severe Sepsis and Septic Shock: A Systematic Review and Meta-analysis. Journal of intensive care medicine 2023; 38: 598–611. https://doi.org/10.1177/08850666231170775

40.

Çeleğen

. Management of thrombocytopenia-associated multiple organ failure: plasma infusion vs plasma exchange. The Turkish journal of pediatrics 2022; 64: 1077–1085. https://doi.org/10.24953/turkjped.2022.494

41.

Porosnicu

Sandesc

Jipa

, et al. Assessing the Outcomes of Patients with Severe SARS-CoV-2 Infection after Therapeutic Plasma Exchange by Number of TPE Sessions. Journal of clinical medicine 2023; 12: 1743. https://doi.org/10.3390/jcm12051743

42.

Chong

Lopez-Magallon

Saenz

, et al. Use of Therapeutic Plasma Exchange during Extracorporeal Life Support in Critically Ill Cardiac Children with Thrombocytopenia-Associated Multi-Organ Failure. Frontiers in pediatrics 2017; 5: 254. https://doi.org/10.3389/fped.2017.00254

43.

Luo

Lai

, et al. Therapeutic plasma exchange in patients with sepsis: Secondary analysis of a cluster-randomized controlled trial. Journal of clinical apheresis 2023; 38: 55–62. https://doi.org/10.1002/jca.22027

44.

Yan

Yao

Xie

, et al. THE THERAPEUTIC EFFICACY OF PLASMAPHERESIS FOR SEPSIS WITH MULTIPLE ORGAN FAILURE: A PROPENSITY SCORE-MATCHED ANALYSIS BASED ON THE MIMIC-IV DATABASE. Shock 2024; 61: 685–694. https://doi.org/10.1097/shk.0000000000002254

45.

De Rosa

Marengo

Fiorentino

, et al. Extracorporeal blood purification therapies for sepsis-associated acute kidney injury in critically ill patients: expert opinion from the SIAARTI-SIN joint commission. J Nephrol 2023; 36: 1731–1742. https://doi.org/10.1007/s40620-023-01637-5

46.

Kuklin

Sovershaev

Bjerner

, et al. Influence of therapeutic plasma exchange treatment on short-term mortality of critically ill adult patients with sepsis-induced organ dysfunction: a systematic review and meta-analysis. Crit Care 2024; 28: 12. https://doi.org/10.1186/s13054-023-04795-x

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.31 MB