Development and application of a simple LC-MS/MS method for therapeutic drug monitoring to guide colistin dosing in critically Ill patients Authors: Yueyuan Wang, Yani Gu, Kai Fan, Chen Feng, XinYun Zhang, Aiming Shi, Yunli Yu Published: 2026 ABSTRACT AbstractObjectivesThis study aims to develop a simple, rapid, and cost-effective liquid chromatography-tandem mass spectrometry (LC-MS/MS) method for determining the concentration of colistin in human serum, including its two active components, colistin A and colistin B, to support routine therapeutic drug monitoring (TDM).MethodsProtein precipitation was performed using pure acetonitrile, with polymyxin B1 as the internal standard (IS). Chromatographic separation was achieved using a Phenomenex KINETEX XB-C18 column (2.6 μm, 3 × 50 mm) with a gradient elution system consisting of 0.1% formic acid (FA) in water and pure acetonitrile as the mobile phase. Detection was carried out using an electrospray ionization (ESI) source in positive ion mode with multiple reaction monitoring (MRM). After applying the validated method to critically ill patients for approximately one year, we used the receiver operating characteristic (ROC) curve to identify the preliminary cut-off value of colistin trough concentration for predicting acute kidney injury (AKI).ResultsThe total runtime for chromatographic separation and mass spectrometric detection was 5 min. Excellent linearity was achieved over the concentration range of 0.27–8.69 µg/mL for colistin A and 0.08–2.56 µg/mL for colistin B, with R2 greater than 0.99 for both analytes. The method demonstrated acceptable matrix effects and recovery rates. In addition‌, the ROC curve analysis of colistin TDM in critically ill patients determined a preliminary cut-off value for AKI at a trough concentration of 2.22 µg/mL. ConclusionThe validated method enables simultaneous and precise quantification of the two active substances, with a simple procedure and minimal analytical cost. The utility of TDM in critically ill patients is paramount for individualizing dosing regimens to optimize outcomes and safety. Specifically, the correlation between trough concentration and AKI risk highlights the importance for clinicians to monitor renal function during colistin treatment to prevent drug-induced AKI.Clinical trial numberNot applicable.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40360-026-01162-8.Keywords: Colistin, Colistin methanesulfonate, Therapeutic drug monitoring, Liquid chromatography-tandem mass spectrometry, Acute kidney injury Introduction Colistin, first isolated in 1950 from a subspecies of Paenibacillus polymyxa, is a cyclic oligopeptide antibiotic. It exerts its antibacterial effect by electrostatically binding to lipopolysaccharides (LPS) on the outer membrane of Gram-negative bacteria, disrupting membrane integrity. This disruption leads to alterations in membrane permeability, the release of intracellular contents, and ultimately bacterial cell death [1]. However, due to its significant adverse effects, colistin was quickly replaced by other, less toxic antimicrobial agents [2]. In recent years, the increasing prevalence of multidrug-resistant (MDR) and extensively drug-resistant (XDR) Gram-negative infections both in China [3] and globally [4, 5] has led to colistin becoming a key treatment option once again. Colistin is a concentration-dependent antibiotic that exhibits a post-antibiotic effect against many bacteria. It consists primarily of two major components, colistin A and colistin B [6]. However, it is associated with considerable nephrotoxicity [7] and neurotoxicity [8]. The incidence of acute kidney injury (AKI) can be as high as 52.6% at conventional doses, representing the most severe and frequent adverse effect limiting its use [9, 10]. The pharmacokinetics of colistin are further complicated by significant inter-individual variability, evidenced by up to 10-fold fluctuations in plasma concentrations [11]. Therefore, accurately predicting the exposure level of active drugs in patients has become inherently challenging [12]. The narrow therapeutic window of colistin, in which effective concentrations partially overlap with those increasing the risk of AKI [9], makes therapeutic drug monitoring (TDM) essential for optimizing treatment outcomes while minimizing toxicity. Given that the trough concentration is a risk factor for colistin-induced AKI [13], TDM is positioned as a vital tool for enabling personalized dosing and mitigating toxicity in clinical practice. A key consensus in the field, articulated in a multinational position paper, suggests a trough concentration threshold of approximately 2.4 µg/mL as a potential risk marker for nephrotoxicity [14]. Translating this pharmacokinetic/pharmacodynamic (PK/PD) target into clinical practice, however, faces a significant operational hurdle: the lack of a simple, rapid, and cost-effective analytical method suitable for routine clinical laboratories. Previously reported methods for quantifying colistin, including immunoassays, microbiological assays, thin-layer chromatography, and capillary electrophoresis, often lack sensitivity and are time-consuming. Although liquid chromatography-tandem mass spectrometry (LC-MS/MS) is the gold standard for colistin quantification due to its specificity and ability to distinguish between colistin A and B, most published protocols are technically complex, time-consuming, or reliant on expensive instrumentation [15], limiting their widespread adoption for routine TDM. Therefore, to bridge the gap between TDM recommendation and routine clinical laboratory implementation, a two-fold approach is necessary: first, to develop a practical analytical solution that lowers the barrier to routine monitoring; and second, to clinically evaluate the proposed safety threshold in a relevant patient population. This study aimed to develop and validate a simplified, rapid, and cost-effective LC-MS/MS method for the simultaneous quantification of colistin A and B in human serum, and apply this method in a cohort of critically ill patients to assess the predictive performance of the consensus-derived preliminary cut-off value of trough concentration for AKI. Materials and methods Instruments and equipment The analytical platform used in this study comprises a liquid chromatography system (Agilent 1260, USA), a triple quadrupole mass spectrometer (Agilent 6465, USA), a low-speed benchtop centrifuge (Xiangyi L530, China), and a high-speed refrigerated benchtop centrifuge (Xiangyi H1750R, China). Reagents The chemical reagents and reference standards used in this study included colistin A (78.4%, CHIA TAI Tianqing, batch no. 20211118-1), colistin B (82.1%, CHIA TAI Tianqing, batch no. 20221203), polymyxin B1 (95%, Aladdin, batch no. F339242), methanol (≥ 99.9%, Merck, product no. 11254407), acetonitrile (≥ 99.9%, Merck, product no. 11250230), FA (≥ 99.0%, ACS, product no. FA-3116-0500), and distilled water (Watsons, batch no. 20240309 C). Preparation of standard solution and internal standard (IS) solution Colistin A and colistin B reference standards were precisely weighed and dissolved in water to prepare stock solutions with concentrations of 8.69 mg/mL (for colistin A) and 2.56 mg/mL (for colistin B), respectively. A volume of 100 µL from each stock solution was combined with 800 µL of methanol to create a mixed stock solution with final concentrations of 869 µg/mL for colistin A and 256 µg/mL for colistin B. This mixed stock solution was then stored at − 80 °C. A series of calibration working solutions were prepared by diluting the mixed stock solution with methanol to achieve the following concentration ranges: 2.7, 5.4, 10.9, 21.7, 43.4, 57.9, and 86.9 µg/mL for colistin A, and 0.8, 1.6, 3.2, 6.4, 12.8, 17.1, and 25.6 µg/mL for colistin B. Similarly, quality control (QC) samples were prepared at final concentrations of 4.1, 16.3, and 65.2 µg/mL for colistin A, and 1.2, 4.8, and 19.2 µg/mL for colistin B. The polymyxin B1 reference standard was accurately weighed and dissolved in purified water to prepare a stock solution with a concentration of 1 mg/mL. An IS working solution with a concentration of 5 µg/mL was then prepared by diluting the stock solution using methanol as the diluent. The resulting IS working solution was stored at − 80 °C for subsequent use. Standard curve and QC samples After adding 10 µL standard solution to 90 µL blank serum and vortex mixing for 1 min, a series of serum calibration samples were prepared with final concentrations of 0.27, 0.54, 1.09, 2.17, 4.34, 5.79, and 8.69 µg/mL for colistin A, and 0.08, 0.16, 0.32, 0.64, 1.28, 1.71, and 2.56 µg/mL for colistin B. Using the same procedure, QC samples were also prepared at concentrations of 0.41, 1.63, and 6.52 µg/mL for colistin A, and 0.12, 0.48, and 1.92 µg/mL for colistin B as low QC (LQC), medium QC (MQC), high QC (HQC) samples, respectively. Sample processing Protein precipitation was utilized for the pretreatment of serum samples. Specifically, 100 µL of serum sample was added into a 1.5 mL centrifuge tube, followed by the addition of 300 µL of pure acetonitrile as the protein precipitant. The mixture was vortexed for 5 min and subsequently centrifuged at 18,525 × g and 4 °C for 5 min using a high-speed refrigerated centrifuge. Finally, 100 µL supernatant was carefully transferred to a 96-well plate for instrumental analysis. Chromatographic and mass spectrometric conditions Chromatographic separation was achieved using a Phenomenex KINETEX XB-C18 column (3.0 × 50 mm, 2.6 μm) equipped with a Phenomenex XB-C18 guard column (3.0 mm). The mobile phase was composed of an organic phase (B: acetonitrile) and an aqueous phase (A: water containing 0.1% FA). The column temperature was maintained at 40 °C, with a flow rate of 0.3 mL/min. The injection volume was 2 µL, and the total analysis time for each sample was 5 min. The gradient elution procedure is detailed in Table 1. Table 1Gradient elution programTimeA: 0.1% FA-WaterB: ACN0.09551.070302.55953.55953.519555.0955 Gradient elution program Quantitative analysis was performed by LC-MS/MS using an electrospray ionization (ESI) source in the positive ion mode with MRM. The names of the reference standards, along with their corresponding precursor ions, product ions, declustering potential (DP), and collision energy (CE), are detailed in Table 2. Table 2Mass spectrometry parameters of colistin and Polymyxin B1AnalytesParent ion (m/z)product ion (m/z)DP(V)CE(V)Colistin A391.1385.11005Colistin B386.1380.11055Polymyxin B1402.2101.010018 Mass spectrometry parameters of colistin and Polymyxin B1 Method validation The method was validated in accordance with the ICH M10 guideline [16]. The following parameters were assessed with their respective acceptance criteria: Selectivity Six individual lots of blank human serum were analyzed. No interfering peaks at the retention times of colistin A, colistin B, or the IS were accepted. Linearity and lower limit of quantification (LLOQ) Calibration curves were constructed using eight non-zero concentrations with a 1/X weighting. The LLOQ was defined as the lowest concentration with a signal-to-noise ratio ≥ 10, accuracy within ± 20%, and precision ≤ 20%. The correlation coefficient (R²) was required to be greater than 0.99. Accuracy and precision Four concentration levels (LLOQ, LQC, MQC, HQC) were analyzed in six replicates on a single day (intra-day) and over three separate days (inter-day). Accuracy was expressed as relative error (RE), with acceptance criteria of ± 15% (± 20% at LLOQ). Precision was expressed as relative standard deviation (RSD), with acceptance criteria of ≤ 15% (≤ 20% at LLOQ). Recovery Recovery was calculated by comparing the peak areas of extracted QC samples with those of post-extraction spiked samples at the same concentrations. No specific acceptance criterion is defined by ICH; results were reported as mean ± standard deviation (SD). Matrix effects Matrix effects were evaluated by comparing the peak areas of post-extraction spiked samples with those of neat standard solutions at equivalent concentrations. A matrix factor within ± 15% was considered acceptable. Stability Stability was assessed under various conditions (room temperature for 2 h, 4 °C for 24h, autosampler for 48 h, three freeze-thaw cycles at -80 °C, and long-term storage at -80 °C for 7 months). Stability was considered acceptable if RE was within ± 15%. Carry-over After injecting the highest calibrator, three blank samples were analyzed. Carry-over was considered acceptable if the residual peak area in the blank was < 20% of the LLOQ. Dilution integrity Samples exceeding the upper limit of quantification (ULOQ) were diluted 2-, 5-, and 10-fold with blank serum. Dilution integrity was considered acceptable if the back-calculated concentration (multiplied by the dilution factor) had an RE within ± 15% of the nominal value. Statistical analysis Data were analyzed using Microsoft Excel to calculate the mean, SD, and RSD. Results are expressed as mean ± SD. Intra-day and inter-day precision were evaluated using RSD, while accuracy was assessed by RE. In addition, we conducted Spearman correlation and receiver operating characteristic (ROC) curve analysis using R. Clinical application A total of 54 critically ill adult patients with severe infections were enrolled in the study. Following a loading dose, Colistimethate sodium (CMS, Tianyun) was administered intravenously at 150 mg every 12 h. After at least 48 h of treatment, blood samples were collected on day 3 using inert separation gel tubes at 0.5 h before dosing (Cmin) and immediately after the end (Cₘₐₓ) of intravenous infusion. For this study, the trough concentration is considered equivalent to the minimum concentration and only the first steady-state trough and peak concentrations (measured before any dose adjustment) were included. The samples were centrifuged at 1600 × g for 5 min, and the serum was separated and analyzed on the same day. The area under the concentration-time curve (AUC) was calculated using a formula provided by the Institute of Antibiotics Research, Huashan Hospital Affiliated to Fudan University [17]. The following formulas were applied: (1) (2) (3) Please check and confirm Section headings are correctly identify. Where: Cmax = peak concentration (immediately post-infusion). Ctrough = trough concentration (0.5 h pre-dose). τ = dosing interval. tinfusion = infusion duration. n = number of doses per 24h. Results Development of methods The LC-MS/MS method utilized an electrospray ionization (ESI) source operating in the positive ion mode with multiple reaction monitoring (MRM). Polymyxin B1 was chosen as an ideal IS because of its similar molecular weight, chemical properties, and retention time to colistin. For chromatographic optimization, various mobile phase compositions were evaluated, and a gradient program using 0.1% formic acid (FA) in water and pure acetonitrile was ultimately selected. In sample preparation, protein precipitation with acetonitrile was adopted after methanol was found to yield low recovery for colistin. Following thorough optimization of instrumental parameters, a fast and efficient method was established with a total runtime of 5.0 min. Method validation Selectivity and specificity Six individual sources of blank human serum were utilized for the selectivity test. A mixed standard solution is added to each blank serum sample, and the samples are processed following the same procedure as the quality control samples. Figure 1A shows the structures of colistin A, colistin B, and the IS (polymyxin B1). The extracted ion chromatogram of the blank sample demonstrates the absence of interfering peaks at the retention times corresponding to the analytes and the IS (Fig. 1B). Figure 1C displays the extracted ion chromatograms for these three substances at the LLOQ sample. Additionally, to exclude the potential interference of co-eluting analogues near the target peaks, individual stock solutions of colistin A, colistin B, and IS were injected separately, and the resulting total ion current (TIC) chromatograms are shown in Figure S1-S3. Extracted ion chromatograms of colistin and IS. (A) The chemical structure of colistin A, colistin B, and polymyxin B1; (B) Extracted ion chromatograms of colistin A, colistin B, and polymyxin B1 from the blank plasma sample; (C) Extracted ion chromatograms of colistin A, colistin B, and polymyxin B1 from the LLOQ sample Linearity and lower limit of quantification A series of standard serum samples containing the analyte were processed according to the method described in Section "Sample processing " and analyzed under the chromatographic and mass spectrometric conditions specified in Section "Chromatographic and mass spectrometric conditions". The results demonstrated good linearity for both colistin A (0.27–8.69 µg/mL) and colistin B (0.08–2.56 µg/mL) across their respective concentration ranges, with R2 greater than 0.99. The signal-to-noise ratio (S/N) at the LLOQ exceeded 10. Accuracy and precision Six replicates of QC samples at low, medium, and high concentration levels were prepared and processed according to the method described in Section "Sample processing". Analysis was performed under the chromatographic and mass spectrometric conditions outlined in Section "Chromatographic and mass spectrometric conditions". The concentrations of the QC samples were calculated using a freshly prepared calibration curve. Intra-day precision was assessed by analyzing LLOQ and QC samples within a single day, while inter-day precision was evaluated over different days across a two-week period. The results indicated that both intra-day and inter-day RSD for all analytes at LLOQ and each QC level were less than 15%, and the RE were within ± 15%. Detailed results are presented in Table 3. Table 3The intra-day and inter-day accuracy and precision of colistinAnalytesQC levelNominalconcentration(µg/mL)Intra-day (n = 6)Inter-day (n = 24)Precision(RSD, %)Accuracy(RE, %)Precision(RSD, %)Accuracy(RE, %)LLQC0.274.515.962.221.83Colistin ALQC0.4110.883.946.432.72MQC1.635.26-3.254.98-6.28HQC6.527.72-1.806.09-1.13LLQC0.086.92-1.884.576.10Colistin BLQC0.124.079.587.811.01MQC0.481.84-1.374.13-4.14HQC1.926.41-5.183.540.85 The intra-day and inter-day accuracy and precision of colistin Recovery and matrix effects QC samples at low, medium, and high concentrations were prepared using the initial mobile phase (Group A), extracted serum (Group B), and blank serum (Group C), with six replicates for each concentration in each group. Matrix effects were calculated using Group B and Group A samples, and were normalized by the IS. Recovery was determined using Group C and Group B samples. The IS-normalized matrix effects were within ± 15%, and recovery for colistin A and colistin B ranged from 85% to 115%, as shown in Table 4. Table 4The recovery and matrix effects of colistinAnalytesQC levelRecoveryMatrix effectsMean, %RSD, %Mean, %RSD, %LQC94.936.10103.323.90Colistin AMQC102.962.53106.373.26HQC105.402.95113.331.27LQC95.066.82107.443.37Colistin BMQC100.092.08101.540.64HQC99.912.5797.882.17 The recovery and matrix effects of colistin Stability The stability of colistin was evaluated under various conditions: at room temperature for 2 h, at 4 °C for 24h, in the autosampler for 48 h, after three freeze-thaw cycles, and during storage at − 80 °C for 7 months. The stability of colistin was assessed at low, medium, and high concentration levels in serum. The results demonstrated that colistin remained stable under all tested conditions with RE within ± 15%. Detailed stability data are summarized in Table 5. Table 5The stability data of colistin under different storage conditionsStability conditionsColistin A (RE, %)Colistin B (RE, %)LQCMQCHQCLQCMQCHQCRoom temperature, 2 h3.69-2.17-0.684.56-2.230.154℃ Refrigerator, 24h2.454.68-1.272.56-1.230.57autosampler 48 h-0.75-7.77-13.023.77-11.67-13.84freeze-and-thaw, -80℃, 3 cycles10.1-11.23-10.336.43-8.64-7.4Freezer (-80 ◦C), 7 months3.568.926.342.455.37-2.89 The stability data of colistin under different storage conditions Carry-over After injecting the highest-concentration standard, three blank samples were analyzed in sequence. The residual peak area ratios for colistin A were 1.83%, 0.66%, and 0.19% of the LLOQ; for colistin B, the ratios were 1.53%, 0.48%, and 0.12%. All values were well below the ICH M10 acceptance criterion of < 20%, indicating negligible carry-over under the rapid 5-minute gradient conditions. Dilution integrity Dilution integrity was assessed by diluting colistin A (26.07 µg/mL) and colistin B (7.86 µg/mL) samples 2-, 5-, and 10-fold with blank serum. The back-calculated concentrations for colistin A were 13.67 µg/mL (2-fold, RE + 4.8%), 5.86 µg/mL (5-fold, RE + 12.5%), and 2.77 µg/mL (10-fold, RE + 6.1%). For colistin B, the back-calculated concentrations were 3.73 µg/mL (2-fold, RE -5.09%), 1.64 µg/mL (5-fold, RE + 4.33%), and 0.81 µg/mL (10-fold, RE + 3.05%). All relative errors were within ± 15%, confirming acceptable dilution integrity. Clinical application The developed LC-MS/MS method was successfully applied to quantify colistin levels in 114 blood samples collected from 54 critically ill patients. For both colistin A and colistin B, when quantified using the standard curve on the day of detection, all QC samples fell within ± 15% of their nominal values. The patients, ranging in age from 19 to 91 years, showed the clinical characteristics detailed in Table 6. The distributions of first steady-state trough and peak concentrations for colistin A and colistin B are illustrated in Fig. 2. Most patients showed trough concentrations of 2.13 ± 1.56 µg/mL for colistin A and 0.46 ± 0.35 µg/mL for colistin B, while peak concentrations were 3.84 ± 1.23 µg/mL for colistin A and 0.75 ± 0.27 µg/mL for colistin B. The linear range of the method proved suitable for monitoring both trough and peak concentrations of colistin in critically ill patients in our hospital. During TDM, clinical pharmacists adjusted colistin dosing regimens based on the drug’s efficacy and the occurrence of adverse effects, thereby supporting personalized and rational use of colistin in critically ill patients. Table 6The general information of critically ill patientsClinical informationStatistical resultsAge (years)65 ± 20.23Man (n)42BMI (kg)23 ± 4.33CICr (mL/min)103 ± 74.43Hypertension (n)15Diabetes (n)10Renal insufficiency (n)10Severe pneumonia (n)23Total trough concentration (µg/mL)2.59 ± 1.91Total peak concentration (µg/mL)4.59 ± 1.48 The general information of critically ill patients The distribution of serum drug concentrations of colistin A and colistin B in critically ill patients. (A) The trough concentration of colistin; (B) The peak concentration of colistin Correlation between trough concentration and acute kidney injury AKI was diagnosed and staged according to the Kidney Disease: Improving Global Outcomes (KDIGO) clinical practice guidelines. Among the 54 patients in our study, 26 (48.15%) were diagnosed with AKI, including 10 with pre-existing chronic renal insufficiency. Most new-onset AKI cases resolved within one to three weeks after discontinuation of colistimethate sodium. To assess the cut-off value of colistin trough concentration for new-onset AKI, the ten patients with pre-existing chronic renal insufficiency were excluded. Spearman correlation analysis was then performed on the remaining 44 patients, using the first steady-state serum trough concentration (Cmin). The analysis revealed a significant positive correlation between Cmin and AKI (P < 0.0001). The ROC curve was generated to evaluate the cut-off value of Cmin for AKI. The area under the curve (AUC) was 0.873 (95% CI: 0.748–0.997). Using the Youden index, the optimal cut-off value was determined as 2.22 µg/mL, yielding a sensitivity of 93.8% and a specificity of 82.1% (Fig. 3). ROC curve of colistin trough concentration for acute kidney injury Discussion Colistin is a complex polypeptide antibiotic that presents significant challenges for the development of reliable analytical methods. Recent advances in mass spectrometry and an improved pharmacological understanding of colistin have facilitated the development of numerous LC-MS/MS methods for quantifying this polypeptide antibiotic [18]. For instance, Zhao et al. described an LC-MS/MS method for detecting colistin using solid-phase extraction (SPE) column pretreatment [15]. However, the sample preparation process using SPE is labor-intensive, time-consuming, and costly due to the expense of SPE columns, making it less suitable for routine TDM. In contrast, protein precipitation offers a simpler and more cost-effective pretreatment procedure, better aligned with the needs of high-volume clinical colistin testing. In our study, method validation confirmed that there was no significant difference in the quantitative results of colistin between protein precipitation and SPE-based pretreatment. It is worth noting that CMS, the prodrug, was not quantified in this study, as it has no direct antibacterial activity and its measurement offers limited clinical value for TDM. The active colistin forms (colistin A and B) are the primary drivers of efficacy and toxicity, and accurate quantification of these components is sufficient for clinical decision-making. In the initial stage of method development, we took many factors into consideration. Polymyxin B1 was selected as the IS in this method for two main reasons. Firstly, its physicochemical properties and retention behavior are similar to those of colistin, making it well-suited for accurate quantification. Secondly, isotopically labeled colistin is not commercially available. The use of a structurally analogous IS effectively compensates for matrix effects and procedural losses, which is particularly critical for ensuring accuracy in the absence of a stable isotope-labeled analog. Furthermore, comparisons among different blood collection tubes revealed that serum obtained using inert separation gel promoting tube yielded higher analyte responses during mass spectrometry detection. As a result, these tubes were adopted for TDM sampling. CMS is an inactive prodrug that has been reported to be gradually converted to colistin in vivo [19]. Additionally, it is susceptible to temperature-dependent degradation in vitro. To assess the stability of CMS in serum, we tested various sample handling and storage conditions. Our results indicated that there were no significant differences between samples that were transported and processed immediately on ice and those kept at room temperature for up to two hours before processing. In the selection of blood collection tubes, we also evaluated the stability of colistin in whole blood, and the results showed that colistin in whole blood is stable for at least 2 h at room temperature. The stability of CMS demonstrated in our study is sufficient to meet the routine requirements for TDM. The patients included in our study are critically ill patients with complex conditions. Critically ill patients frequently experience dynamic physiological alterations, primarily due to the presence of comorbidities and polypharmacy. These changes result in highly variable pharmacokinetics, especially among the elderly [20]. During colistin therapy, neurotoxicity and nephrotoxicity are of significant concern. Given the difficulty in diagnosing neurotoxicity in critically ill patients, nephrotoxicity has become the main indicator to judge whether colistin-related adverse drug reactions occur. Although most cases of colistin-associated AKI are reversible [10], and the creatinine levels of the patients we enrolled also recovered within 1–3 weeks, AKI can exacerbate the condition of patients by causing fluid overload, cardiac dysfunction, and electrolyte imbalances, which may adversely affect both short- and long-term outcomes [9]. Even transient AKI is associated with increased mortality, longer hospital stays, and higher costs, reinforcing that prevention is far superior to management. Therefore, standardizing colistin quantification and implementing routine TDM are essential for improving clinical outcomes and preventing the emergence of colistin-resistant strains. Furthermore, the prevention of toxicity through TDM may indirectly combat resistance by avoiding subtherapeutic dosing due to fear of side effects, which is a common clinical dilemma. Colistin is a concentration-dependent antibiotic, and its PK/PD index is the area under the concentration-time curve to minimum inhibitory concentration (AUC/MIC). According to the 2019 International Consensus Guidelines for the Optimal Use of Colistin, the recommended target is an AUCss, 24h of 50 mg·h/L and a Css, avg of 2 µg/mL [21]. In our clinical practice, some patients undergo both trough and peak concentration monitoring for AUC/MIC calculation. However, the majority of patients receive trough concentration monitoring only, as it requires a single blood sample and is more feasible for routine TDM. Additionally, in China, LC-MS/MS testing is fully out-of-pocket; monitoring a single time point reduces patient costs by approximately 50% compared to multiple sampling. Therefore, we do not mandate multiple sampling. Previous studies have demonstrated that adding nebulized colistin to intravenous therapy offers no additional benefit [22], whereas a loading dose combined with a long course of intravenous therapy improves clinical outcomes in critically ill patients with carbapenem-resistant Acinetobacter baumannii (CRAB) infections [23]. However, loading dose therapy also increases the risk of nephrotoxicity [24], highlighting the need for TDM and the identification of a reliable AKI prediction threshold. Therefore, finding a trough concentration threshold will be of greater help for the clinical application of colistin. Our study also established an AKI prediction model utilizing the colistin trough concentration in critically ill patients at our hospital. A preliminary cut-off value of 2.22 µg/mL was identified, which is almost consistent with findings reported in the literature [14]. Notably, earlier monitoring of colistin concentrations after steady-state achievement enables earlier warning of AKI risk, facilitating timely clinical interventions. Future studies could explore whether this threshold is equally predictive across specific subpopulations, such as those on renal replacement therapy or with severe burns. While a rapid and convenient analytical method provides the technical foundation for TDM, other factors are equally critical for successful clinical implementation. These factors include the correct application of TDM by clinicians, the time required for nurses to collect blood samples, sample storage conditions, accurate calculation of AUCss, 24h and Css, avg, as well as the correct interpretation of TDM reports. We have therefore established an integrated workflow covering the entire process from TDM application to result reporting. The process begins with training clinicians and nurses. After blood collection at specified time points, samples are stored at 4 °C before transport. Upon arrival at the laboratory, TDM personnel log and centrifuge the samples, followed by sample preparation and LC–MS analysis according to a standardized operating procedure (SOP). Trough and peak concentrations, AUC values, patient demographics (e.g., age, gender, weight), serum creatinine levels, and antimicrobial susceptibility results are reviewed to generate a final TDM report. Clinical pharmacists then interpret the report and provide dosing recommendations based on patient’s condition. This comprehensive system has enhanced the rational use of colistin and has been well received in clinical practice. The success of this multidisciplinary model highlights that TDM is not merely a laboratory test but a clinical service. Its value is fully realized only when embedded within a collaborative care framework involving physicians, pharmacists, nurses, and laboratory scientists. Sustainability of such a program requires institutional support and ongoing education. Real-world data from our TDM program have unveiled substantial interindividual variability in colistin concentrations. Notably, 48.15% of patients exhibited increased serum creatinine levels within approximately one week of initiating colistin therapy, with older adults being particularly affected. Given these observations, initial dosing based on renal function, followed by timely TDM once steady-state concentrations are achieved, and subsequent dose adjustments guided by TDM results, renal function, and clinical symptoms are essential for balancing efficacy and safety. This “initiate-monitor-adjust” paradigm represents a pragmatic and proactive approach to personalized antimicrobial therapy in critical care. It is important to acknowledge that, despite the established TDM workflow at our center, a considerable number of patients exhibited first steady-state trough concentrations exceeding the identified preliminary cut-off value for toxicity. This observation does not indicate that TDM is ineffective; rather, it reflects several clinical realities. First, TDM is primarily a monitoring tool rather than a prospective preventive intervention. Blood samples for trough concentration measurement were typically collected after several days of therapy, at which point AKI may have already developed or was already in progress. Second, clinicians managing critically ill patients with multidrug-resistant infections face a difficult trade-off between efficacy and toxicity. In some cases, accepting a certain risk of nephrotoxicity may be considered necessary to achieve infection control. Third, there is an inherent time lag in the TDM process—from sample collection and transport to laboratory analysis, result reporting, and dose adjustment. Even when workflows have been optimized, delays of 24 to 48 h are often unavoidable. During this time, additional doses may be administered based on the previous regimen. Fourth, the preliminary cut-off value of 2.22 µg/mL is not an absolute dividing line. Some patients can tolerate higher trough concentrations without developing AKI, while others experience renal injury at lower concentrations due to individual risk factors such as baseline renal insufficiency, concomitant nephrotoxic agents, or sepsis-induced hemodynamic instability. Therefore, we view TDM as a continuous quality improvement tool. Its value lies not in the immediate elimination of all preventable AKI events, but in the systematic identification of risk, analysis of deviations, and iterative optimization of dosing strategies to progressively reduce the incidence of supratherapeutic exposure. This perspective, we believe, offers a realistic and constructive framework for interpreting TDM outcomes in routine clinical practice. The ROC model established in this study identified a trough concentration preliminary cut-off value with high sensitivity and specificity for predicting AKI, confirming the clinical value of monitoring colistin levels to guide therapy. Looking forward, several research directions are worth pursuing. Although concomitant medications were collected, most patients received colistin with β-lactams or tigecycline (low nephrotoxic), while diuretics (e.g., spironolactone) were commonly used but not adjusted for due to limited sample size; future larger-cohort studies should account for these confounders. Additionally, the integration of such PK/PD targets into clinical decision support systems within electronic health records could automate alerts and recommendations, further standardizing and improving care. Finally, pharmacoeconomic studies are warranted to formally evaluate the cost-effectiveness of implementing this comprehensive TDM strategy versus empirical dosing, considering the costs of AKI management and prolonged hospitalization. Several limitations of this study should be acknowledged. First, the sample size was relatively small, and multivariate regression analysis was not performed to adjust for potential confounders. Second, the preliminary cut-off value identified in this study requires external validation in an independent cohort. Additionally, potential factors during sample collection and transportation, such as prolonged handling time or temperature fluctuations, may affect quantification accuracy [25], although we followed standardized procedures to minimize these risks. Future studies should address these limitations with larger, multicenter cohorts and validated sampling protocols. Conclusion To meet clinical demands for colistin detection, we have developed a rapid, robust, and reliable LC–MS method that is capable of simultaneously quantifying colistin A and colistin B within just 5 min. The method utilizes acetonitrile-mediated protein precipitation, which is characterized by its simplicity, convenience, and cost-effectiveness. This approach not only supports high-throughput clinical monitoring but also provides a robust technical foundation for the rational dosing of colistin. Furthermore, the implementation of TDM for colistin not only highlights the crucial role of TDM in guiding personalized drug delivery to optimize treatment efficacy and improve patient safety, but also confirms the application value of colistin trough concentration detection in clinical treatment. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We would like to express our gratitude to all the staff in the therapeutic drug monitoring room of the Second Affiliated Hospital of Soochow University. We also thank all the participants in this study. Author contributions Yunli Yu designed the study and conducted experimental operations, Yueyuan Wang carried out the experiments and performed the statistical analysis and wrote the manuscript. Chen Feng and Kai Fan collected clinical samples, Aiming Shi and XinYun Zhang conducted clinical sample processing, and Yani Gu conducted clinical sample testing. All of the authors discussed the results, reviewed and are responsible for the manuscript. Funding This work was supported by the Jiangsu Pharmaceutical Association (A202428) and Medical Scientific Research Foundation (Grant No. YWJKJJHKYJJ-TYU066D). Suzhou Health Workforce Program (GSWS2022040), Discipline Promotion Program of the Second Affiliated Hospital of Soochow University (XKTJ-XK202406), Hunan Provincial Key Laboratory of Anti-Resistance Microbial Drugs (No:2023TP1013). Data availability The data presented in this study are available within the article and its supplementary materials. Additional data can be made available upon request from the corresponding author. Declarations Human ethics and consent to participate This study was approved by the Ethics Committee of the Second Affiliated Hospital of Soochow University (Approval number: JD-LK2024171-IR01). The study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants or their legal guardians, as required by the approved protocol (Informed consent form version 3.0, dated 10 April 2025). Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Aiming Shi, Email: sam740411@163.com. Yunli Yu, Email: haoyyl0902@163.com. References Associated Data Data Citations Supplementary Materials Supplementary Material 1 Data Availability Statement The data presented in this study are available within the article and its supplementary materials. Additional data can be made available upon request from the corresponding author.