A closer look at severe acute kidney injury: risk factors and outcomes in PD-1/PD-L1 antibody treatment from a retrospective study

Study design

This retrospective cohort study analyzed patients who received PD-1/PD-L1 antibody therapy at Zhejiang Provincial People’s Hospital between January 2019 and July 2023 using data from the hospital’s electronic medical record system. The patients were categorized into two groups: those who developed severe AKI and those who did not, based on AKI occurrence after treatment with ICIs. Patients with cancer who received PD-1/PD-L1 antibody therapy one or more times. The exclusion criteria were (1) the absence of initial or subsequent measurements of creatinine levels, (2) an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m², (3) age greater than 85 years, and (4) a follow-up period shorter than 7 days. Patients were followed up from the initiation of PD-1/PD-L1 therapy until the occurrence of severe AKI or death.

Since this was a retrospective study, the requirement for informed consent was waived. This study was approved by the Ethics Research Committee of the Zhejiang Provincial People’s Hospital (approval number: QT2024180).

Data collection

Data for each patient were collected using an electronic medical record system, including demographic information (sex, age), medical history (smoking, alcohol consumption), vital signs (blood pressure), comorbidities, extrarenal iRAEs, tumor type, drug combinations, and laboratory data. The recorded comorbidities included diabetes, cardiovascular disease, chronic obstructive pulmonary disease (COPD), and cirrhosis. Extrarenal iRAEs encompassed a range of conditions such as thyroid dysfunction, skin injuries, fever, pneumonia, and drug-induced liver injury. Tumor types were classified, including head and neck tumors, hepatobiliary and pancreatic system tumors, lung cancer, breast cancer, gastrointestinal tract tumors, urinary system tumors, reproductive system tumors, and melanoma. Combination drugs included therapeutic agents such as proton pump inhibitors (PPIs), renin-angiotensin-aldosterone system inhibitors (RAASis), nonsteroidal anti-inflammatory drugs (NSAIDs), nephrotoxic antibiotics, iohexol, H2 receptor antagonists, diuretics (e.g., furosemide, spironolactone), and chemotherapy agents (e.g., platinum-based compounds, gemcitabine, cyclophosphamide, and interleukin therapies). Nephrotoxic drugs include aminoglycosides, sulfonamides, and vancomycin. Laboratory data included markers of anemia, uric acid levels, liver function tests, serum albumin, lipid profiles, serum inorganic phosphate, inflammatory markers, urinary leukocytes, and urinary specific gravity.

AKI was defined within 12 months of the first administration of PD-1/PD-L1 antibodies, based on the internationally accepted Kidney Disease Improving Global Outcomes (KDIGO) standard, and staged according to the magnitude of the increase (Kellum, Norbert & Aspelin, 2012). The stages of AKI were classified as follows: stage 1, a 1.5 to 1.9-fold increase in serum creatinine (SCr) from baseline or an absolute increase of at least 26.5 µmol/L; Stage 2, a 2.0 to 2.9-fold increase in SCr from baseline; and Stage 3, either a 3.0-fold increase in SCr from baseline or an SCr level ≥353.6 µmol/L. Stages 2 and 3 were classified as severe AKI. The baseline creatinine level was the closest creatinine level before starting the first PD-1/PD-L1 antibody treatment.

Statistical analysis

The baseline characteristics of the cohort were summarized using frequency counts (expressed as percentages) for categorical variables and the median (accompanied by the interquartile range [IQR]) for continuous variables. To evaluate differences between groups, categorical variables were analyzed using the chi-squared test, or when appropriate, Fisher’s exact test, whereas continuous variables were compared using the Wilcoxon test. When a variable had less than 20% missing values, multiple imputation by chained equations was performed using the ‘mice’ package in R (version 4.4.0; R Foundation for Statistical Computing, Vienna, Austria). This approach assumes that data were missing completely at random, allowing observed variables to inform the imputation process (Li, Stuart & Allison, 2015; Sterne et al., 2009). The Least Absolute Shrinkage and Selection Operator (LASSO) method was applied to identify significant variables and prevent model overfitting (Li et al., 2022). In this regression model, the coefficients are penalized based on the value of λ, the LASSO regularization parameter. A larger penalty shrinks the estimate of less important factors toward zero, leaving only the strongest predictors. Variables selected by the LASSO regression were then included in a Cox regression analysis to identify risk factors for developing severe AKI in those patients who received PD-1/PD-L1 antibodies. The variable with a p-value of <0.05 in the univariate COX analysis was selected as the candidate risk variable in the multivariate COX analysis. The combination aims to balance model parsimony and statistical power (Tibshirani, 1997). Survival differences of severe AKI according to different risk factors were visualized using Kaplan–Meier survival plots. Statistical analysis was performed using SPSS software (version 26.0; IBM Corp., Armonk, NY, USA) and R software (version 4.4.0; R Foundation for Statistical Computing, Vienna, Austria). R packages used for data analysis and visualization included ‘ggplot2’ (Version 3.5.1) for data visualization, ‘survival’ (Version 3.6-4) for survival analysis, ‘glmnet’ (Version 4.1-8) for LASSO regression, as well as ‘mice’ (Version 3.16.0), ‘missForest’ (Version 1.5), ‘VIM’ (Version 6.2.2), ‘Hmisc’ (Version 5.1-2), ‘rms’ (Version 6.8-0), ‘regplot’ (Version 1.1), ‘tableone’ (Version 0.13.2), ‘survminer’ (Version 0.4.9), and ‘survivalROC’ (Version 1.0.3.1). Statistical significance was at a p-value of <0.05 for all tests.

Characteristics of patients with severe AKI

This study included 1,840 patients treated with PD-1/PD-L1 antibodies between January 2019 and July 2023. Patients were excluded for the following reasons: 82 patients lacked baseline creatinine data, 431 had no follow-up creatinine measurements, 48 had a baseline eGFR <60 mL/min/1.73 m², 17 were over 85 years old, and 263 had follow-up periods <7 days (Fig. 1). After these exclusions, the final cohort comprised 907 patients, including 660 men (72.8%) and 247 women (27.2%) with a median age of 64 years (IQR: 58–72) and a median baseline eGFR of 103.14 mL/min/1.73 m² (IQR: 87.85–121.76). In this cohort, 29 patients (3.2%) developed severe AKI, with 23 cases (79.3%) classified as stage 2 and six cases (20.7%) as stage 3. Mortality was significantly higher among patients with severe AKI, with a death rate of 20.7%, compared to 4.1% in those without AKI (p < 0.001). The Kaplan–Meier survival analysis showed significantly lower survival in patients with severe AKI (Fig. 2).

Lung cancer was the most prevalent tumor type, accounting for 304 cases (33.5%), followed by gastrointestinal tumors in 249 cases (27.5%) and hepatobiliary and pancreatic tumors in 205 cases (22.6%). Chemotherapy was administered to 673 patients (74.2%), with platinum-based agents being the most commonly used (316 patients, 34.8%), followed by gemcitabine (79 patients, 8.7%) and pemetrexed (56 patients, 6.2%). Over the follow-up period, 42 patients (4.6%) died, including 20.7% (six patients) in the severe AKI group (Table 1).

No statistically significant differences were observed between the non-AKI and severe AKI groups regarding sex, age, smoking status and alcohol consumption, or comorbidities, including diabetes mellitus, hypertension, cirrhosis, chronic obstructive pulmonary disease (COPD), and cardiovascular disease (p > 0.05). Additionally, there were no significant differences between the groups in the occurrence of extrarenal iRAEs, with the exception of dysfunction. In the severe AKI group, thyroid dysfunction was present in six patients (20.7%). Among tumor types, only gastrointestinal tumors showed a significant difference between groups, with 6.9% in the severe AKI group compared to 28.1% in the non-AKI group (p = 0.021). Differences were also noted in the use of certain combined medications: ACEI/ARB use was more frequent in the severe AKI group (24.1% vs. 9.7%, p = 0.026), as were NSAIDs (89.7% vs. 53.2%, p < 0.001) and diuretics (62.1% vs. 25.3%, p < 0.001). Regarding antitumor drug use, ICI usage did not differ significantly between the groups (p > 0.05). However, cytotoxic drug use was more frequent in the severe AKI group (6.9% vs. 0.9%, p = 0.033). Baseline parameters, including eGFR, alanine aminotransferase, aspartate transaminase, total cholesterol, low-density lipoprotein cholesterol (LDL-C), serum inorganic phosphate, glucose, monocytes, eosinophils, and urinary specific gravity, showed no statistically significant differences between the two groups. In the severe AKI group, elevated C-reactive protein levels (21.2 vs. 8.0, p = 0.015) were observed, alongside decreased levels of hemoglobin (11.1 vs. 12.1, p = 0.014), lymphocytes (0.92 vs. 1.20, p = 0.012), uric acid (234 vs. 298, p = 0.01), and albumin (33.1 vs. 36.2, p = 0.025).

Risk factors for severe AKI

The risk factors associated with severe AKI were further examined in these patients. Using LASSO regression analysis, 56 characteristic variables were included (Fig. S1A). This analysis identified 18 factors that were significantly associated with the occurrence of severe AKI (Fig. S1B & Table 2). These 18 factors included age, high blood pressure, diabetes mellitus, gastrointestinal tumor, head neck tumor, other tumors, gamma-glutamyl transferase (GGT), albumin, hemoglobin, white blood cell in the urine, ACEI/ARB use, NSAID use, iohexol use, diuretics, platinum drugs, pemetrexed, cytotoxic drugs, and proton pump inhibitor. These factors were then analyzed using univariate and multivariate Cox regression, which finally identified four independent risk factors for severe AKI: GGT (hazard ratio (HR): 1.17; 95% confidence interval CI [1.01–1.03], p = 0.008), NSAIDs (HR: 4.58; 95% CI [1.36–15.46], p = 0.014), diuretics (HR: 3.61; 95% CI [1.62–8.07], p = 0.002), and cytotoxic drugs (HR: 5.04; 95% CI [1.12–22.81], p = 0.036) (Fig. 3). The dynamic alterations of the concordance index of severe AKI are shown (Fig. 4).

When GGT > 250 U/L was used as the threshold, the Kaplan−Meier survival curve analysis showed a significantly higher incidence of severe AKI in patients exceeding this threshold (p < 0.0001) (Fig. 5A). Similarly, the incidence of severe AKI was notably elevated among patients receiving diuretics compared to those who did not (p < 0.0001) (Fig. 5B). Additionally, NSAID use was associated with higher severe AKI incidence than non-use (p = 0.00023) (Fig. 5C). Lastly, severe AKI incidence was significantly higher in patients treated with cytotoxic drugs than in those not receiving these medications (p < 0.0001) (Fig. 5D).

Prognosis of patients

The 29 patients diagnosed with severe AKI were monitored for an average duration of 92 days. Among these patients, two achieved full recovery, nine experienced partial recovery, 13 did not recover, and five were lost to follow-up after being diagnosed with severe AKI. In the subgroup of 23 patients with stage 2 AKI, two recovered fully, seven had partial recovery, 10 did not recover, and the remaining four were lost to follow-up. Of the six patients with stage 3 AKI, two recovered partially, three did not recover, and one was lost to follow-up (Fig. 6).