Vancomycin remains a cornerstone of modern antibiotic therapy, serving as the first-line treatment for methicillin-resistant Staphylococcus aureus infections. Yet the drug carries a well-known liability: a narrow therapeutic range in which the concentration that suppresses bacteria can edge dangerously close to concentrations that damage the kidneys. To manage this tension, clinicians rely on therapeutic drug monitoring, or TDM, in which blood samples are drawn and vancomycin concentrations measured so that doses can be adjusted. A new retrospective study from Japan, published in the journal Drugs – Real World Outcomes, has examined a specific and often overlooked piece of this monitoring puzzle: whether performing a follow-up TDM measurement within one week of the initial assessment is associated with a lower occurrence of vancomycin-induced nephrotoxicity, abbreviated VIN.
The research team, led by Tomoyuki Yamada of Osaka Medical and Pharmaceutical University, drew on the Japan Medical Data Center claims database, an anonymized health insurance dataset that aggregates records from more than 200 hospitals across Japan. The study received ethics approval from the Osaka Medical and Pharmaceutical University Ethics Committee, and the requirement for informed consent was waived because the analysis used routinely collected retrospective data. The investigators identified adults who received intravenous vancomycin between January 2020 and December 2021 and who remained on the drug for at least one week after their initial TDM measurement. Patients were excluded if they were younger than 18, were receiving renal replacement therapy, lacked baseline or on-treatment serum creatinine measurements, had their initial TDM performed after nephrotoxicity had already been diagnosed, or had none of the risk factors that guidelines flag for earlier follow-up monitoring: a vancomycin dose change after the initial TDM, use of vasopressors, use of loop diuretics, or unstable renal function or high risk of kidney injury.
After these exclusions, 492 patients who underwent an initial TDM formed the study cohort. Of these, 403 patients, or 81.9 percent, had a follow-up TDM performed within one week of the initial assessment. The remaining patients either had no further TDM at all, 45 individuals, or had their next measurement one week or later after the initial assessment, 44 individuals. The definition of follow-up TDM was deliberately strict: a measurement conducted within one week after the initial TDM. Nephrotoxicity was diagnosed according to the 2009 Vancomycin Consensus Statement of the Infectious Diseases Society of America, which defines VIN as either an increase in serum creatinine of at least 0.5 milligrams per deciliter or a rise of at least 1.5-fold from the baseline concentration. Serum creatinine was monitored until the second day after vancomycin was discontinued.
Comparing the two groups revealed meaningful baseline differences. Follow-up TDM was more frequently performed in patients who were older, had higher white blood cell counts, had pneumonia as the infection type, used vasopressors, had cancer as a comorbidity, were admitted to the intensive care unit, experienced a vancomycin dose change after the initial TDM, or had different average vancomycin trough concentrations. Implementation also varied with hospital size, with lower rates observed in facilities with fewer beds. These patterns suggest that follow-up monitoring was concentrated in sicker patients treated in larger, better-resourced institutions, a reality that the investigators had to confront when interpreting their results.
The primary analysis used a Cox proportional hazards model, adjusted for estimated glomerular filtration rate below 30 milliliters per minute per 1.73 square meters, concomitant use of nephrotoxic drugs, vasopressors, loop diuretics, and an initial vancomycin trough concentration exceeding 15 milligrams per liter. In this model, follow-up TDM was associated with a lower occurrence of VIN, with a hazard ratio of 0.47 and a 95 percent confidence interval of 0.25 to 0.86, and a p value of 0.015. A multivariable logistic regression performed under the same conditions pointed in the same direction, yielding an odds ratio of 0.49 with a 95 percent confidence interval of 0.25 to 0.97 and a p value of 0.035. In other words, after adjusting for the major recognized pharmacological and clinical risk factors for kidney injury, patients whose monitoring included a follow-up measurement within a week experienced roughly half the rate of vancomycin-associated nephrotoxicity in this primary model.
Recognizing that observational studies of this kind are vulnerable to bias, the researchers ran two sensitivity analyses. The first addressed immortal time bias, the artifact that arises when exposure classification ignores the time a patient must survive before becoming exposed. Using a time-varying covariate model in which patients were treated as unexposed from the start of therapy until the day of their follow-up TDM and exposed thereafter, the association held: the hazard ratio was 0.46 with a 95 percent confidence interval of 0.25 to 0.84 and a p value of 0.012. The second sensitivity analysis used inverse probability of treatment weighting, or IPTW, a propensity score method that additionally adjusted for institutional characteristics, including hospital bed count of 300 or more, the presence of an infection control team or antimicrobial stewardship team, and ward pharmacy services, alongside age-adjusted Charlson Comorbidity Index and the baseline clinical risk factors from the primary model. Here the picture changed. The association was attenuated and did not reach statistical significance, with a hazard ratio of 0.62, a 95 percent confidence interval of 0.31 to 1.23, and a p value of 0.167.
The attenuation in the weighted analysis carries an important interpretive message, and the authors were candid about it. Although propensity score weighting reduced the major initial imbalance in hospital bed size and substantially reduced the imbalance in vasopressor use, residual imbalance remained for some covariates: the weighted standardized mean differences for loop diuretics, ward pharmacy services, and vasopressor drugs all exceeded the conventional 0.1 threshold. The investigators suggest that the clinical benefit observed in the primary analysis may not rest solely on performing follow-up TDM itself, but may be closely linked to abundant institutional resources and the overall quality of clinical support systems, such as specialized antimicrobial stewardship and infection control activities and ward pharmacy services, which are highly prevalent in facilities that actively implement follow-up monitoring. Because high-risk patients are more frequently managed in well-resourced facilities, fully isolating these interconnected factors remains a challenge.
Beyond the primary comparison, the team conducted an exploratory receiver operating characteristic analysis to ask how frequently TDM measurements per week after the initial assessment might discriminate between patients who developed VIN and those who did not. The cut-off frequency that best discriminated was 1.009 measurements per week, with an area under the ROC curve of 0.673, a 95 percent confidence interval of 0.597 to 0.749, sensitivity of 0.618, and specificity of 0.661. The authors emphasize that this finding should be interpreted with caution: TDM frequency is strongly influenced by treatment duration and the patient’s clinical course, and the area under the curve was modest, so the analysis remains exploratory and the optimal monitoring frequency warrants prospective evaluation.
The study’s limitations deserve equal attention. As a retrospective observational analysis of claims data, it cannot establish causation. The site of infection was not recorded in more than 30 percent of patients, and data on pathogens and infection severity were unavailable, although analyses of ICU admission, vasopressor use, and Charlson Comorbidity Index scores may have partially addressed severity. Drug administration was inferred from prescription records, so actual usage could not be verified, and the database did not capture exact administration or blood sampling times. Notably, the dataset contained no information on whether area under the concentration–time curve, or AUC-guided, monitoring was performed, even though vancomycin practice has shifted from trough-based monitoring toward AUC-guided approaches because AUC relates more closely to both efficacy and nephrotoxicity. The findings therefore apply within the framework of concentration-based monitoring. Because the cohort excluded patients who developed nephrotoxicity before the initial TDM or stopped vancomycin within a week, the treated population had relatively long therapy durations, with a median of 14 days, and a correspondingly late median onset of VIN at 12 days, raising the possibility that some creatinine elevations stemmed from factors other than vancomycin. Unmeasured clinical variables such as daily urine output, transient hypotension, and fluid balance could not be controlled.
Within these constraints, the study offers a real-world, multi-facility perspective on follow-up TDM practice and supports current guidelines recommending follow-up monitoring within a week of the initial assessment, particularly for patients with vancomycin dose changes, vasopressor or diuretic use, unstable renal function, or high kidney injury risk. The authors conclude that healthcare providers should consider conducting follow-up TDM to potentially lower the occurrence of vancomycin-induced nephrotoxicity, while stressing that further prospective studies using standardized, AUC-guided monitoring protocols are needed to clarify the independent effect of follow-up monitoring and to account for the institutional confounding that this claims-based analysis could not fully remove.
Subject of Research: Association between follow-up therapeutic drug monitoring and vancomycin-induced nephrotoxicity in Japanese adults
Article Title: Association Between Follow-Up Therapeutic Drug Monitoring and Vancomycin-Induced Nephrotoxicity: A Claims Database Study
Article References: Yamada, T., Imai, S., Minami, K., Ogawa, T., Ukimura, A., Nishihara, M., & Neo, M. (2026). Association Between Follow-Up Therapeutic Drug Monitoring and Vancomycin-Induced Nephrotoxicity: A Claims Database Study. Drugs – Real World Outcomes, 13(3), 383-394. https://doi.org/10.1007/s40801-026-00575-5
Image Credits: AI Generated
DOI: 10.1007/s40801-026-00575-5
Keywords: vancomycin, therapeutic drug monitoring, nephrotoxicity, claims database, retrospective study, Japan, kidney injury, antimicrobial stewardship, pharmacokinetics, hospital care, serum creatinine, antibiotics
News Source: Ophelia Keating. (October 29, 2026). Follow-Up Vancomycin Drug Monitoring Linked to Lower Kidney Injury Risk in Japanese Claims Study. Scienmag.
