Real-World Risk of Drug-Induced Tubulointerstitial Nephritis

Drug-induced drug-induced kidney injury (DIKI) represents a prevalent challenge in clinical practice and novel drug development, with diverse pathological manifestations affecting multiple renal components, including glomeruli, renal vasculature, tubules, and interstitium. Among various forms of DIKI, tubulointerstitial nephritis (TIN) stands as a critical and common pathological subtype. Its pathogenesis is complex, primarily categorized into two types: direct cytotoxic damage to renal tubular epithelial cells by drugs or their metabolites, and immune-mediated hypersensitivity reactions triggered by the host response to drugs. With the expanding variety and clinical application of pharmaceuticals, TIN induced by different drug classes has emerged as a non-negligible public health concern.

Although real-world data contain sporadic reports linking certain drugs to drug-induced TIN (DITIN), and preliminary experiments have confirmed the nephrotoxicity of some agents, existing studies exhibit significant limitations. First, most reports are based on case studies or small-sample retrospective analyses, lacking large-scale population-based systematic evaluations, which hinders accurate quantification of the association strength and actual incidence risk between different drugs and DITIN. Second, research on DITIN risk factors has predominantly focused on single drugs or specific populations, failing to establish a comprehensive risk assessment system covering multiple drug classes and clinical characteristics. This gap leaves clinicians struggling to rapidly identify high-risk patients in complex medication scenarios. Additionally, insufficient exploration of the temporal patterns of DITIN onset exists: while a few studies mention the median time to DITIN for specific drugs, precise descriptions of "drug exposure–injury onset" timing across different drug classes are lacking—information critical for clinical early warning (e.g., duration after drug initiation requiring focused renal function monitoring). These research gaps not only limit the early identification and intervention of DITIN but also impede the development of targeted prevention strategies (e.g., nephroprotective adjuvant therapies, screening protocols for high-risk populations).

Thus, this study integrates large-scale real-world data from the FDA Adverse Event Reporting System (FAERS) spanning Q1 2015 to Q1 2025, combined with multi-dimensional statistical methods including disproportionality analysis and multivariate regression. It systematically screens drug classes and specific agents associated with DITIN, clarifies core risk factors (e.g., sex, age, exposure to specific drugs) and temporal characteristics of onset. This work aims to address the deficiencies in current research regarding "risk quantification", providing more actionable guidance for early DITIN identification, risk stratification, and medication decision-making in clinical practice.

 Data Collection 

The FAERS classifies drug roles into four categories: primary suspect, secondary suspect, concomitant, and interacting. In this study, we specifically examined cases with the adverse event (AE) action code designated as "primary suspect" to minimize confounding variables. AE reports were restricted to the period from the first quarter (Q1) of 2015 to Q1 of 2025. Additionally, to improve analytical accuracy and reduce reporting bias, only cases reported by healthcare practitioners were included, including physicians (MD), health professionals (HP), pharmacists (PH), registered nurses (RN), and other health professionals (OT). Finally, data deduplication was performed following guidelines recommended by the FDA.

Disproportionality Analysis (DPA) 

A two-by-two contingency table was constructed. Four algorithms were used for disproportionality analysis, including reporting odds ratio (ROR), proportional reporting ratio (PRR), Bayesian confidence propagation neural network (BCBPNN), and multi-item gamma Poisson shrinker (MGPS), to evaluate and identify associations between suspected drugs and interstitial nephritis. A drug was considered to have a risk signal if it met the threshold criteria of all algorithms simultaneously. P-adjust refers to the p-value after Fisher’s exact test and Bonferroni correction. A volcano plot was generated with -log(p-adjust) as the x-axis and log(ROR) as the y-axis. 

Regression Analysis 

Complete data on gender and age were extracted for analysis. Age values >120 years were defined as outliers and excluded. Suspected drugs with a >100 and P-adjust <0.01 were included in univariate analysis. Drugs with p<0.01 in the univariate analysis were subjected to least absolute shrinkage and selection operator (LASSO) regression. Multivariate logistic regression was performed using the drugs screened by LASSO combined with basic patient information as independent variables to identify risk factors for drug-induced tubulointerstitial nephritis (DITIN). 

Time Interval

Time Intervalwas defined as the time interval between the initiation of the suspected drug and the onset of DITIN, with the median and interquartile range recorded. Reports with incorrect date entries, inconsistencies, or missing data were excluded.

Statistical Analysis  

Descriptive analysis was performed to summarize and present the clinical characteristics of patients in the DITIN report. Multivariate regression analysis with Bonferroni correction was applied to identify independent risk factors for DITIN. All data analyses were conducted using R software (4.5.0).

Baseline characteristics of DITIN  

A total of 6784 DITIN-related reports were included in the analysis. Of these, 3122 (46.0%) were from female patients and 3582 (52.8%) from male patients. By age, 3718 reports (54.8%) were from young and middle-aged individuals, and 3066 (45.2%) from elderly patients aged 65 years or older. The most common reported outcome was hospitalization, accounting for 3949 cases (58.2%), with 199 reports (2.9%) documenting death. The overall reporting trend showed a consistent increase over time (data not shown).  

Drugs associated with TIN  

Seventy-nine TIN-related drugs were identified (data not shown). The top 10 categories were as follows: antibacterial drugs (26/79), antineoplastic drugs (9/79), gastric acid secretion inhibitors (8/79), analgesic drugs (7/79), antihypertensive drugs (4/79), antiatherosclerotic drugs (4/79), antiepileptic drugs (3/79), immunomodulatory drugs (3/79), diuretic drugs (3/79), and antituberculosis drugs (2/79).  

Suspected drugs for TIN  

Drugs meeting the criteria of >100 case reports, a lower 95% confidence interval (CI) of the reporting odds ratio (ROR) >1, and adjusted p<0.01 were extracted for univariate analysis. As shown in Figure 1a, 11 such drugs were identified: omeprazole, ibuprofen, pantoprazole, ciprofloxacin, pembrolizumab, nivolumab, vancomycin, carboplatin, pemetrexed, tacrolimus, and esomeprazole.  

The relationship between TIN and these suspected drugs is illustrated in the volcano plot (Figure 1a). In this plot, the x-axis represents the logarithm of ROR; a positive x-value indicates more frequent reporting of drug-related TIN compared with other adverse reactions. The y-axis represents the negative logarithm of the adjusted p-value (after Fisher’s exact test and Bonferroni correction), with a positive y-value denoting a highly significant difference. The color of each dot corresponds to the logarithm of the number of case reports, with redder shades indicating a higher number of reports. Thus, drugs located in the upper right quadrant of the plot exhibit both strong signal strength and significant differences. The thresholds of four data mining algorithms for these 11 drugs are shown in Figure 1b.  

Risk factors for DITIN  

LASSO regression analysis was performed on drugs with p<0.01 in the univariate analysis, yielding 9 candidate drugs (Figures 1c, 1d). Multivariate logistic regression analysis incorporating patient gender and age was then conducted (Figure 1e). The results indicated that male sex, age ≥60 years, and 9 drugs (omeprazole, ibuprofen, pantoprazole, ciprofloxacin, pembrolizumab, vancomycin, carboplatin, pemetrexed, and esomeprazole) were independent risk factors for DITIN. The model’s predictive accuracy, as measured by ROC-AUC, was 0.71 (Figure 1f).  

Time interval from drug administration to TIN onset  

The time interval from drug administration to TIN onset was evaluated (Figure 1g). The median onset time for TIN associated with the 9 drugs ranged from 4 days (fastest) to 129 days (slowest), with statistical differences detailed in Figure 1h. Further subgroup analyses of drugs within the same category (Figure 1i-k) showed that approximately 50% of reported cases among the 3 gastric acid secretion inhibitors occurred within 50 days of administration; ~90% of cases among the 2 antibacterial drugs occurred within 20 days; and ~75% of cases among the 3 antineoplastic drugs occurred within 200 days.

In this study, a comprehensive list of drugs associated with TIN was established using FAERS data. Individuals aged over 60 years, male, or those who have taken 9 specific drugs including omeprazole are at a higher risk of developing DITIN. The findings of this study can provide important information for the early identification of drug-related acute interstitial nephritis and lay a foundation for future research on the pathogenesis of DITIN.