Correlation between early methotrexate TDM and AKI

High-dose methotrexate (HDMTX) therapy has been proven to be effective in treating acute lymphoblastic leukaemia (ALL), lymphoma and osteosarcoma. However, its application may be limited due to the potential for acute kidney injury (AKI), which can lead to delayed methotrexate elimination (DME) and various methotrexate-related toxicities. Despite supportive measures such as urine alkalisation, 2–33% of patients receiving HDMTX therapy develop AKI. Therefore, early detection of AKI is crucial for preventing irreversible MTX-related toxicity.

Serum creatinine (Scr) is a common clinical indicator of acute kidney injury (AKI), but this method requires serial monitoring and significant elevations may not appear until 48–72 hours after AKI onset. To overcome this limitation, researchers have developed novel biomarkers, such as neutrophil gelatinase-associated lipocalin (NGAL), which shows promise in patients treated with HDMTX. However, as a new indicator, additional testing costs are incurred, and clinical validation is required across different age groups, diagnostic categories, and other variables.

In fact, monitoring methotrexate concentration (MTXC) is standard practice during HDMTX therapy, and is also used to evaluate the renal clearance of methotrexate. Current guidelines recommend adjusting glucuronidase therapy based on MTXC levels at 36, 42 and 48 hours after the HDMTX infusion. However, if AKI occurs early in the infusion period, these time points may be too late to administer agents that could prevent permanent toxicity. The aim of this study is to determine whether early MTX clearance following HDMTX therapy can serve as an early biomarker for AKI, which will be achieved by evaluating a large dataset of HDMTX treatments.

The study sample comprised 169 patients from two independent Spanish institutions. MTX concentrations were measured in all patients at least twice within 16 hours after completion of the MTX infusion. Acute kidney injury (AKI) was assessed using the AKIN criteria, with severe AKI defined as meeting stage 2–3 criteria. DME was evaluated using two criteria. One criterion employed the micromolar threshold: DME was defined as MTXc >1μM at any time point 42 hours or later after HDMTX infusion initiation. The second criterion used a 20-standard-deviation (SD) threshold. DME was diagnosed if the MTXc value measured by the MTXPKorg model deviated by more than two SDs from the population mean at 42 or 48 hours post-infusion. All MTXc values measured 16 hours after infusion completion were input into the MTXPKorg model to predict the MTXc value at 42 or 48 hours, which served as the standard for assessing DME. Changes in MTX concentration over time are divided into three independent periods: The 'first' period (within 2.5 hours), the 'second' period (2.5 to 8.5 hours) and the 'third' period (8.5 to 16 hours). If two or more measurements occurred within the same time period, the value closest to the end of the MTX infusion was used. For 24-hour infusions, concentrations collected within the last five hours before infusion completion were assigned to the 0 hour period. Secondly, two independent early MTX clearance biomarkers are calculated by combining MTX concentration data from any two of the three time periods (i.e. 'first' and 'second', 'first' and 'third', 'second' and 'third'): elimination half-life and slope (rate of MTXe concentration decline).

The study population comprised 169 patients who collectively received 556 courses of HDMTX therapy. The median age was 10.3 years (interquartile range 4.2–37.3), and 114 patients (67.5%) were diagnosed before the age of 18. Two-thirds (66.9%) of patients were diagnosed with acute lymphoblastic leukaemia of the B-cell lineage. Of the 556 treatment cycles for which data was available, 329 (59.2%) underwent MTX concentration testing at all three designated time points, and 83% of cycles utilised extended-duration infusion (24 hours). The median sampling times for the first, second and third MTX tests were 0.17 hours (interquartile range [IQR] 0.00–0.68), 6.28 hours (IQR 4.00–6.62) and 12 hours (IQR 11.8–12.8), respectively, post-infusion (see Figure 1).

Figure 1: Classification of Metronidazole Measurement Time Intervals Following Completion of Infusion

ROC analysis revealed that the methotrexate (MTX) elimination half-life measured between the second and third time periods demonstrated the highest area under the curve (AUC) value for acute kidney injury (AKI) across different time interval combinations. As an early clearance biomarker, MTX demonstrated optimal performance in terms of the area under the curve (AUC) for all AKI grades: 0.62 (interquartile range 0.56–0.69) for AKI Grade II and 0.65 (interquartile range 0.54–0.77) for AKI Grade I, with particularly high discriminatory power for severe AKI. The MTX elimination half-life also demonstrated high discriminatory ability for diabetic macular oedema (DME) at the micromolar concentration standard (AUC 0.79, interquartile range 0.54–0.77). It effectively distinguishes DME with an AUC of 0.79 (interquartile range 0.73–0.84) under the micromolar standard and an AUC of 0.86 (interquartile range 0.73–1.00) under the standard deviation standard (Table 1, Figure 2). The median elimination half-lives obtained from 24-hour and 4-hour infusions in the second and third time periods were 2.95 hours (interquartile range 2.51–3.47) and 2.51 hours (interquartile range 2.34–2.80), respectively.

Table 1: Area under the curve (AUC) for early clearance biomarkers (slope and elimination half-life)

Figure 2 shows the receiver operating characteristic (ROC) curves for methotrexate elimination half-life (DME), calculated from measurements taken during the second and third time periods. (A) DME at micromolar standards; (B) DME at standard deviation standards; (C) acute kidney injury (AKI) at any grade; (D) grade 2–3 AKI.

After adjusting for age, sex, dose, infusion time, HDMTX regimen and baseline eGFR, the elimination half-life between the second and third periods remained statistically significant for acute kidney injury (AKI), with a ratio of 1.29 (95% confidence interval [1.03, 1.65], p = 0.031). A likelihood ratio test comparing models with and without the early clearance biomarker showed that the biomarker significantly improved overall performance (p=0.006).

In sensitivity analyses predicting AKI occurrence, the half-life measured during the second to third phase showed consistent performance as an early clearance biomarker. The area under the curve (AUC) for this biomarker across all AKI severity levels was 0.61 (interquartile range (IQR) 0.55–0.68), and 0.64 (IQR 0.52–0.76) for severe AKI. For predicting DME, the AUC was 0.78 (IQR 0.72–0.83) using the micromolar standard and 0.84 (IQR 0.67–1.00) using the SD standard.

The aim of this study was to determine whether early elimination of methotrexate (MTX) after infusion completion during HDMTX therapy could serve as an early clearance biomarker for acute kidney injury (AKI). The results showed that changes in MTX concentrations during the second and third phases were associated with the occurrence of AKI. As this detection occurs well before serum creatinine elevation, it provides a more effective means of identifying AKI risk at an early stage than serum creatinine does. This temporal advantage is crucial given that guidelines indicate that the optimal window for glucuronidase-based AKI prevention is 48–60 hours after the initiation of HDMTX infusion; otherwise, drug toxicity may become irreversible.

The key strength of this study is its comprehensive approach to screening and evaluating biomarkers for early methotrexate clearance. First, a receiver operating characteristic (ROC) analysis was performed to evaluate and select the most promising candidate biomarkers. This analysis then established the discriminatory capacity of methotrexate elimination half-life as an early clearance biomarker, revealing the trade-off between sensitivity and specificity. Recognising that this analysis considers only one variable, the study also used a multivariate logistic regression model. This approach enhanced the reliability of the conclusions by accounting for various confounding factors.