Therapeutic drug monitoring of antimicrobials in a paediatric setting: A retrospective single‐centre study Authors: Gianluca Gazzaniga, Stefano Agliardi, Riccardo Giossi, Alice Schianchi, Alessandra Romandini, Thomas Virdis, Silvia Nerini Molteni, Maria Luisa De Angelis, Stefania Paccagnini, Cecilia Gechtman, Alice Proto, Marco Fossati, Silvia Oliveri, Giulia Angela Pattarino, Elena Altieri, Agnese Graziosi, Arianna Pani, Romano Danesi, Stefano Martinelli, Costantino De Giacomo, Francesco Scaglione Published: 2026 ABSTRACT AbstractAimsPaediatric pharmacokinetics differ significantly from adults due to age‐related physiological changes, necessitating precise dose adjustments. However, data on therapeutic drug monitoring (TDM) implementation in paediatric setting remain limited. This study assessed TDM use at ASST Grande Ospedale Metropolitano Niguarda, where clinical pharmacologists support therapy optimization, especially in high‐intensity care settings.MethodsThis retrospective study included paediatric patients (<18 years) admitted to our hospital who underwent serum drug concentration (SDC) testing for antimicrobials between 2021 and 2022. The primary objective was to describe TDM utilization, specifically the proportion of patients maintaining therapeutic SDC levels. Secondary objectives included clinical outcomes and factors linked to out‐of‐range SDC levels. Two TDM classifications, binary based on toxicity thresholds and expert‐based, were used.ResultsAmong 427 SDC tests from 90 hospital stays (83 patients), median patient age was 0.2 years, and 41.1% were female. Overall, 22.0% of tests exceeded toxicity thresholds; vancomycin was most frequently monitored. The toxic group, based on binary classification, had a significantly shorter time‐to‐first SDC and higher rates of cardiovascular and renal comorbidities compared with the non‐toxic group. Expert‐based classification showed 50% not‐in‐range, 26.7% in‐range, and 23.3% indeterminate. In‐range patients had higher renal failure, CRRT use, and mortality (20.8% vs. 4.4%, p = .045), though not significant in sensitivity analyses. Kaplan–Meier curves showed no survival difference.ConclusionsTDM frequently identified out‐of‐range SDC, with potentially toxic levels observed in approximately one fifth of hospital stays. In this context, TDM may represent a useful tool for individualized antimicrobial dosing in complex paediatric patients.Keywords: antibiotics, childrenintensive careneonatologytherapeutic drug monitoring What is already known about this subject Therapeutic drug monitoring (TDM) is a useful tool for optimizing antimicrobial therapy. Therapeutic drug monitoring (TDM) is a useful tool for optimizing antimicrobial therapy. Data on antimicrobial TDM in paediatric patients are limited. Data on antimicrobial TDM in paediatric patients are limited. What this study adds A description of the real‐world use of paediatric antimicrobial TDM in a hospital setting with clinical pharmacologist support. A description of the real‐world use of paediatric antimicrobial TDM in a hospital setting with clinical pharmacologist support. Potentially toxic concentrations were observed in approximately one fifth of hospital stays. Potentially toxic concentrations were observed in approximately one fifth of hospital stays. Toxic antimicrobial levels were more frequent in clinically complex patients, particularly those with cardiovascular comorbidities and renal impairment. Toxic antimicrobial levels were more frequent in clinically complex patients, particularly those with cardiovascular comorbidities and renal impairment. INTRODUCTION Therapeutic drug monitoring (TDM) is the measurement of serum drug concentrations (SDC) to optimize therapeutic efficacy and minimize side effects. Pharmacokinetics and pharmacodynamics in paediatric patients can be significantly different compared with adults; indeed, age, weight and organ system development influence the absorption, distribution, metabolism and elimination of drugs, particularly antimicrobials. 1 , 2 Therefore, when administering antimicrobials to paediatric patients, it is necessary to adjust therapeutic doses based on these specific characteristics. 1 , 3 TDM can provide support in this context, allowing for personalized therapy, avoiding overdoses or underdosing and preventing the development of microbial resistance. 1 , 3 This is particularly important for antibiotics with a narrow therapeutic window, such as vancomycin or aminoglycosides, to avoid toxicity and ensure adequate antimicrobial activity. 4 Moreover, TDM can help optimize empirical therapy in severe infections, contributing to reducing recovery times and improving clinical outcomes in children with febrile neutropenia. 5 Furthermore, TDM is particularly important in paediatric patients with renal or hepatic diseases, as they may have altered drug clearance, preventing toxicity and therapeutic failure. 6 TDM in paediatric patients not only ensures effective therapy but can also be a valuable tool for detecting pharmacological interactions. 7 Despite the well‐established rationale behind TDM in paediatric antimicrobial therapy, its clinical implementation remains highly heterogeneous across hospital centres. Although the benefits of TDM are well described, considerable variability exists in monitoring strategies, therapeutic targets, sampling methods and the degree to which TDM is integrated into routine paediatric practice. 8 However, few studies systematically assessed how TDM is used in real‐world paediatric patients, particularly in high‐intensity care settings. 9 Most available literature focuses on pharmacokinetic aspects, whereas data describing actual TDM utilization patterns, frequency and timing of sampling, therapeutic‐range attainment and differences across hospital wards are limited. 10 Moreover, real‐world evidence from hospitals is scarce, and publications rarely describe models that integrate TDM with direct clinical pharmacologist involvement, which would be an element that may significantly influence dosage individualization and clinical decision‐making. 11 Understanding these aspects may help to clarify the real‐world impact of TDM on individualized antimicrobial therapy in paediatric patients and to identify potential areas for improvement. Therefore, given these gaps in the current literature, it may be useful to describe how TDM is implemented among paediatric patients in a large centre with a high intensity of care. This study aims to describe the use of TDM for antimicrobials in paediatric patients followed at the Azienda Socio Sanitaria Territoriale (ASST) Grande Ospedale Metropolitano Niguarda, a hospital centre in the north of Italy that integrates TDM with the advice of a clinical pharmacologist to improve therapy management, especially in wards with high intensity of care. Study objectives The primary objective of this study was to provide a description of the use of TDM for antimicrobials in the study population by means of the percentage of patients who maintained the SDC within the desired range at the peak and trough levels for the drug that required TDM, at each available time point. Secondary objectives included (1) The comparison of patient outcomes (i.e., all‐cause mortality and duration of hospitalization) between those who maintained the therapeutic drug concentration and those who did not; and (2) a comparison of clinical and sociodemographic data in‐between groups to identify variables potentially associated with out‐of‐range TDM values. METHODS Study design and population We performed a single‐centre, observational, retrospective study including all paediatric (i.e., <18 years old) patients admitted to ASST Grande Ospedale Metropolitano Niguarda, Milan (Italy), who underwent at least one SDC for antimicrobials from January 2021 to December 2022. Patients were excluded in case of lack of available clinical data from the medical records (i.e., analyses requested from other hospitals) or invalidated lab test results (e.g., system errors and haemolyzed samples) from laboratory database. Antimicrobials SDC results were automatically retrieved from the laboratory database; age criteria and hospital provenience were initially checked for inclusion. Then, hospital records were searched for final assessment of inclusion criteria and data extraction. Given the retrospective nature of the study and considering the potential use of antimicrobial TDM mainly for seriously ill patients in acute‐care setting, direct enrolment and obtaining informed consent were not applicable and the necessity of consent was waived. The study was not sponsored and was conducted in accordance with the ethical principles derived from the Declaration of Helsinki (64th World Medical Association General Assembly, Fortaleza, Brazil, October 2013) and subsequent revisions. This study was approved by the local Ethics Committee (N 332‐29052023) according to regulations governing non‐interventional observational studies. Collected variables For each included patient, data concerning medical outcomes, sociodemographic characteristics and clinical variables were manually collected from medical records. Extracted variables included sex, age, weight at the time of SDC test, time to first SDC request, all SDC results, length of hospital stays (LoS), in‐hospital mortality, comorbidities, invasive treatments such as extra corporeal membrane oxygenation (ECMO), intubation, continuous renal replacement therapy (CRRT), concomitant medications, whether the antimicrobial therapy was empirical or targeted, and, if available, the isolated pathogen. For newborns (i.e., <28 days from birth), variables included weight at birth as well. All data were recorded in an electronic spreadsheet and were anonymized prior to analysis to ensure that the patient's identity could not be traced. All the data were already available in the databases of the computer systems at ASST Grande Ospedale Metropolitano Niguarda (DNLab, Ward Portal) and were generated for clinical purposes. Antimicrobial SDC analyses In this study, we considered all antimicrobials SDC tests performed at our hospital, that is, amikacin, ceftazidime, gentamicin, isavuconazole, linezolid, meropenem, piperacillin/tazobactam, posaconazole, vancomycin and voriconazole. Within the hospital information system used to request chemical analyses, sampling time can be specified as peak or trough, or left unspecified. Trough samples were defined as blood withdrawals performed immediately before the next scheduled dose and no earlier than 60 min prior to dosing. Peak samples were defined as blood withdrawals performed 30–60 min after drug administration. When sampling time was not specified, concentrations were classified as unspecified and considered separately from peak and trough samples. Drug‐ and sampling‐specific threshold values were applied to define therapeutic exposure or potentially toxic concentrations (Table 1). For vancomycin, trough concentrations of 10–20 mg/L were considered therapeutic, while concentrations ≥20 mg/L were considered potentially toxic irrespective of sampling time. For gentamicin, concentrations ≥2 mg/L for troughs and unspecified samples were considered excessive. For amikacin, trough and unspecified samples with concentrations ≥5 mg/L were considered toxic, whereas peak concentrations ≤35 mg/L were considered potentially underdosed. For linezolid, a therapeutic trough range of 2–7 mg/L was adopted. For meropenem and ceftazidime, therapeutic exposure was defined as a trough steady‐state concentration (Css) of 4–8 × MIC, with concentrations ≥44.5 mg/L and ≥20 mg/L, respectively, considered potentially toxic. For voriconazole and isavuconazole, ranges of 1–4.5 mg/L and 1–7 mg/L, respectively, were considered adequate. These threshold values have been routinely applied in clinical practice at our institution for years and were defined through an internal multidisciplinary consensus process involving specialists in laboratory medicine, infectious diseases, clinical pharmacology and other relevant clinical disciplines. TABLE 1Individual SDC results.AntibioticTests, numberSDC, median (IQR)Toxic levels, n (%)Efficacy thresholdToxic thresholdOverallNewbornsNon‐newbornsVancomycinVancomycin, trough1647.4 (4.6–12.2)10 (6.1)5 (6.3)5 (5.9)1020Vancomycin11513.8 (8.0–20.5)30 (26.1)4 (16.7)26 (28.6)20Vancomycin, peak10920.2 (13.9–25.0)56 (51.4)18 (45.0)38 (55.1)20GentamicinGentamicin, trough131.3 (0.9–1.7)3 (23.1)3 (25.0)0 (0.0)2Gentamicin71.1 (0.6–1.9)1 (14.3)1 (14.3)NA2AmikacinAmikacin, trough65.4 (3.2–7.1)3 (50.0)NA3 (50.0)5Amikacin225.4 (3.2–47.6)1 (50.0)NA1 (50.0)5Amikacin, peak16.5 (NA)0 (0.0)NA0 (0.0)35Voriconazole41.3 (1.0–1.6)0 (0.0)NA0 (0.0)14.5Linezolid, trough20.35 (0.0–0.7)0 (0.0)NA0 (0.0)27Meropenem, trough218.1 (16.0–20.1)0 (0.0)NA0 (0.0)Css 4–8 × MIC44.5Ceftazidime10.0 (NA)0 (0.0)NA0 (0.0)Css 4–8 × MIC20Isavuconazole10.7 (NA)0 (0.0)NA0 (0.0)17 Note: Considered efficacy and safety cut‐off levels are reported according for each time of sampling. Efficacy and toxicity thresholds are those routinely used in our hospital and are expressed in mg/l.Abbreviations: Css, steady‐state concentration; IQR, interquartile range; MIC, minimum inhibitory concentration; NA, not applicable; SDC, serum drug concentration. Individual SDC results. Note: Considered efficacy and safety cut‐off levels are reported according for each time of sampling. Efficacy and toxicity thresholds are those routinely used in our hospital and are expressed in mg/l. Abbreviations: Css, steady‐state concentration; IQR, interquartile range; MIC, minimum inhibitory concentration; NA, not applicable; SDC, serum drug concentration. All SDC measurements were performed in the clinical chemistry laboratory of the centre. Vancomycin, gentamicin and amikacin concentrations were measured through a fully automated kinetic interaction of microparticles in a solution (KIMS) method (Cobas 8000, Roche); ceftazidime, meropenem, linezolid and piperacillin SDC were assessed through high‐performance liquid chromatography with UV detector (HPLC‐UV) (Surveyor Plus, Thermo Fisher Scientific), whereas isavuconazole, voriconazole and posaconazole through high‐performance liquid chromatography with mass spectrometry (HPLC‐MS) (TSQ Quantum Access, Thermo Scientific). All analytes were routinely subjected to internal quality controls using commercially available standards (Chromsystems Internal Control ‐ references 0183, 0184, 0253/XT, 0254/XT; Bio‐Rad Liquichek Immunoassay Plus Control). In addition, where required by current regulations, the laboratory participates in an external quality assessment (EQA) programme for pharmacology under the regional health service, ensuring ongoing verification of assay performance and reliability of measured SDC. Antimicrobial TDM classifications Patients were classified using two distinct strategies based on their SDC results. The first classification followed a binary system: If any SDC test for a given patient exceeded the toxicity threshold for the respective antimicrobial, the patient was assigned to the ‘toxic group’. Conversely, patients whose totality of SDC tests remained below the toxicity threshold were categorized as ‘Non‐Toxic’. Additionally, the analysis included the frequency of patients with SDC exceeding the toxicity threshold at the time of the first measurement. As further description, the frequency of patients with antimicrobial levels above toxicity threshold at time of first SDC test was reported. To address the limitations of this initial classification, a second, more clinically relevant classification was developed. Three clinical pharmacology and toxicology specialists (R. G., G. G. and S. A.) independently assigned each patient to one of three categories based on their clinical judgement as ‘in‐range’, ‘not‐in‐range’ or ‘indeterminate’. The assessors were blinded to each other's evaluations and to patient outcomes. In cases of disagreement, consensus was reached through collegial discussion. Although the precise criteria used by each physician may not be strictly documented as they were based on their respective clinical expertise, the following guiding principles were applied: patients were classified as ‘in‐range’ if: (A) the majority of their TDM tests during the hospital stay were below toxicity thresholds and were not underdosed; (B) the number of tests exceeding the toxicity threshold was low, and their levels were only slightly above the limit, carrying minimal or no clinical relevance; or (C) a single SDC test exceeded the threshold early in the hospital stay, but this was promptly corrected and did not recur. Patients were classified as ‘not‐in‐range’ if all or more than one of their SDC tests throughout the hospital stay exceeded the toxicity threshold or were underdosed. All the other cases, when there were alternating patterns of SDC results, with both toxic and non‐toxic readings, making it unclear whether the antimicrobial safety and efficacy profile was consistently maintained during the hospital stay, were assigned to the ‘indeterminate’ class. Statistical analysis Descriptive statistics were used to summarize the demographic and clinical characteristics of the study population. Continuous variables were presented as medians and interquartile ranges (IQRs). Categorical variables were presented as absolute numbers and percentages. For comparisons between patient groups, Wilcoxon rank‐sum or Kruskal–Wallis tests were used for continuous variables while χ 2 or Fisher's exact tests were used for categorical variables. For the clinical classification comparison, only in‐range and not‐in‐range groups were statistically compared; the indetermined group results were reported only as descriptive statistics. Survival analysis was performed using Kaplan–Meier curves to compare survival rates based on both previously described classifications, with separate analyses considering time from hospital admission and time from antibiotic initiation. Spearman's log‐rank tests were used to assess differences in survival distributions between these groups. Subgroup analyses of newborns and non‐newborns and for patients receiving vancomycin, the most prescribed SDC test in our hospital paediatric wards, were performed. All analyses were performed using STATA 15 software (StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX, USA: StataCorp LLC) and R Statistical Software (version 4.2.3). A two‐tailed significance level of α = .05 was considered for all analyses. Nomenclature of targets and ligands Key protein targets and ligands in this article are hyperlinked to corresponding entries online (http://www.guidetopharmacology.org) and are permanently archived in the Concise Guide to PHARMACOLOGY 2021/22. 12 RESULTS Study population selection and characteristics A total of 5234 SDC tests performed between January 2021 and December 2022 were retrieved from the laboratory database. Of these, 803 tests were excluded because they were performed in patients not admitted to our hospital. The remaining records were filtered to include only paediatric patients; 34 additional tests were excluded due to invalidated laboratory results. The final study population therefore comprised 427 SDC tests performed in paediatric patients across eight antimicrobials: amikacin, ceftazidime, gentamicin, linezolid, meropenem, vancomycin, isavuconazole and voriconazole. No SDC tests were performed in paediatric patients for piperacillin/tazobactam or posaconazole during the study period. Overall, 83 individual patients were included. One patient accounted for four separate hospital admissions, and four patients for two admissions each, resulting in a total of 90 hospital stays included in the comparative analysis (Figure 1). SDC tests selection and flowchart. Among the included hospital stays, 37 (41.1%) were in female patients. The median age was 0.2 years (IQR 0.0–1.6) and the median weight at the time of SDC test was 5.8 kg (IQR 3.1–8.2). Comorbidities were present in 93.3% of patients, with cardiovascular (64.4%) and respiratory (38.9%) conditions being the most common. Critical conditions were common, with invasive mechanical ventilation performed in 64.4% and ECMO in 12.2% of cases. Concomitant medication use was high, with 93.3% of patients receiving additional treatments, including other antimicrobials (76.7%), amines (34.4%), diuretics (32.2%) and sedatives (25.6%). Targeted antimicrobial therapy was administered in 27.8% of cases, and a causative pathogen was isolated in 36.7% of patients, with methicillin‐sensitive Staphylococcus epidermidis (MSSE) being the most frequent (11.1%). The overall median LoS was 34.5 days (IQR 20–66), with a median time to first SDC test of 2 days (IQR 1–3), and the mortality rate was 11.1%. Newborns accounted for 44.4% of the cohort and had a median age of 9.3 days (IQR 5.2–14.0) and a median weight at SDC testing of 2.8 kg (IQR 1.0–3.1). The mortality rate was 10.0%. Non‐newborns patients had a median age of 1.2 years (IQR 0.5–15.8), a median weight of 7.3 kg (IQR 5.8–10.0) and a mortality rate of 12.0%. Individual SDC tests and first TDM classification Among the 427 SDC evaluations performed on paediatric patients, 94 (22.0%) exceeded the respective antimicrobial toxicity threshold. Vancomycin, administered by a 1‐h infusion, was the most frequently monitored antimicrobial, with a total of 164 trough, 115 unspecified and 109 peak measurements available. Median vancomycin concentrations were 7.4 (IQR 4.6–12.2) mg/L for trough levels, 13.8 (IQR 8.0–20.5) mg/L for unspecified assessments and 20.2 (IQR 13.9–25.0) mg/L for peak levels. Among these, 96 tests (24.7%) exceeded the toxicity threshold of 20 mg/L. Gentamicin was the second most frequently monitored antimicrobial, with 13 trough evaluations and 7 unspecified tests. Median concentrations were 1.3 (IQR 0.9–1.7) mg/L for trough evaluations and 1.1 (IQR 0.6–1.9) mg/L for unspecified tests, with 4 (20%) measurements exceeding toxicity thresholds. Boxplots showing the distribution of vancomycin and gentamicin SDCs, stratified by sampling type and age class (newborn vs. older), are available in Figure S1. Amikacin was measured in nine instances, of which four (44.4%) exceeded toxicity thresholds. Other antibiotics (linezolid, meropenem and ceftazidime) and antifungal agents (voriconazole and isavuconazole) were infrequently monitored, with fewer than five assessments each and no recorded instances of toxicity. A detailed description of individual SDC tests is provided in Table 1. Analysis of the first SDC measurement per hospital stay showed potentially toxic antimicrobial concentrations in 19 (21.1%) patients. This included 15 (21.7%) patients treated with vancomycin and 4 (28.6%) treated with gentamicin. Toxic concentrations at first TDM were observed in 6 newborns (15.0%) and 13 non‐newborn patients (26.0%). When restricted to vancomycin‐treated patients, corresponding rates were 8.0% in newborns and 29.6% in non‐newborns. Binary toxic versus non‐toxic classification comparison Comparisons between hospital stays classified as Toxic (n = 36) and Non‐Toxic (n = 54) revealed significant differences (Table 2). The time to first SDC measurement was shorter in the toxic group, with a median of 1 day (IQR 0–2) compared with 2 days (IQR 1–4) in the non‐toxic group (p = .0054). Cardiovascular comorbidities (86.1% vs. 50.0%, respectively; p = .001), renal failure (22.2% vs. 5.6%, respectively; p = .024), ECMO use (22.2% vs. 5.6%; p = .024), and diuretics administration (47.2% vs. 22.2%, respectively; p = .021) were more frequent in the toxic group compared with the non‐toxic group. However, all‐cause mortality did not differ significantly between groups (13.9% vs. 9.3%, respectively; p = .513). TABLE 2Study population and outcome analysis in the overall population.Binary toxic/non‐toxic classificationClinical classificationOverall (n = 90)Non‐toxic (n = 54)Toxic (n = 36) p valueIndeterminate (n = 21)Not‐in‐range (n = 45)In‐range (n = 24) p valueSex, female, n (%)37 (41.1)23 (42.6)14 (38.9).8289 (42.9)20 (44.4)8 (33.3).445Age, years, median (IQR)0.2 (0.0–1.6)0.1 (0.0–12.1)0.5 (0.0–1.1).84980.0 (0.0–1.6)0.3 (0.0–1.0)0.6 (0.0–15.8).2730Weight at TDM, median (IQR)5.8 (3.1–8.2)5.5 (2.7–7.3)7.0 (3.9–8.5).22626.8 (3.3–53.5)5.5 (2.9–7.8)6.4 (4.2–7.9).2672 n = 49 n = 25 n = 24— n = 8 n = 25 n = 16—Clinical classification details, n (%)———NA———NANever in‐rangeNANANA——6 (13.3)——UnderdosedNANANA——36 (80.0)——ToxicNANANA——3 (6.7)——Always in‐rangeNANANA———19 (79.2)—Underdosed than brought in‐rangeNANANA———5 (20.8)—Time to first SDC, days, median (IQR)2 (1–3)2 (1–4)1 (0–2) .0054 1.5 (1–2)2 (1–4)1 (0–4).5038 n = 79 n = 47 n = 32— n = 20 n = 40 n = 19—Comorbidities, n (%)84 (93.3)49 (90.7)35 (97.2).39620 (95.2)44 (97.8)20 (83.3) .046 Cardiovascular58 (64.4)27 (50.0)31 (86.1) .001 14 (66.7)28 (62.2)16 (66.7).796Respiratory35 (38.9)25 (46.3)10 (27.8).08412 (57.1)15 (33.3)8 (33.3)1.000Gastrointestinal14 (15.6)11 (20.4)3 (8.3).1485 (23.8)6 (13.3)3 (12.5)1.000Hepatic failure5 (5.6)3 (5.6)2 (5.6)1.0001 (4.8)1 (2.2)3 (12.5).118Renal failure11 (12.2)3 (5.6)8 (22.2) .024 2 (9.5)2 (4.4)7 (29.2) .007 Neurological26 (28.9)19 (35.2)7 (19.4).1549 (42.9)10 (22.2)7 (29.2).566Onco‐haematological1 (1.1)1 (1.9)0 (0.0)1.0001 (4.8)1 (2.2)0 (0.0)1.000Immunodepression1 (1.1)0 (0.0)1 (2.8).4000 (0.0)0 (0.0)1 (4.17).348Autoimmune0 (0.0)0 (0.0)0 (0.0)NA0 (0.0)0 (0.0)0 (0.0)NAShock14 (15.6)8 (14.8)6 (16.7)1.0002 (9.5)8 (17.8)4 (16.7)1.000ECMO11 (12.2)3 (5.6)8 (22.2) .024 2 (9.5)4 (8.9)5 (20.8).259Intubation58 (64.4)35 (64.8)23 (63.9)1.00015 (71.4)28 (62.2)15 (62.5)1.000CRRT5 (5.6)1 (1.9)4 (11.1).1532 (9.5)0 (0.0)3 (12.5) .039 Concomitant medications, n (%)84 (93.3)49 (90.7)35 (97.2).39621 (100)41 (91.1)22 (91.7)1.000Cardiovascular20 (22.2)9 (16.7)11 (30.6).1314 (19.1)9 (20.0)7 (29.2).550Amines31 (34.4)16 (29.6)15 (41.7).2657 (33.3)14 (31.1)10 (41.7).433Vasodilators23 (25.6)11 (20.4)12 (33.3).2192 (9.5)13 (28.9)8 (33.3).786Diuretics29 (32.2)12 (22.2)17 (47.2) .021 4 (19.1)15 (33.3)10 (41.7).601NSAIDs/paracetamol13 (14.4)11 (20.4)2 (5.6).0672 (9.5)8 (17.8)3 (12.5).736Glucocorticoids9 (10.0)6 (11.1)3 (8.3).7363 (14.3)1 (2.2)5 (20.8) .017 Antivirals2 (2.2)1 (1.9)1 (2.8)1.0000 (0.0)2 (4.44)0 (0.0).540Other antimicrobials69 (76.7)38 (70.4)31 (86.1).12620 (95.2)33 (73.3)16 (66.7).587Opiates/opioids28 (31.1)18 (33.3)10 (27.8).64711 (52.4)11 (24.4)6 (25.0)1.000Sedatives23 (25.6)15 (27.8)8 (22.2).6276 (28.6)9 (20.0)8 (33.3).251Anticonvulsants12 (13.3)10 (18.5)2 (5.6).1144 (19.1)3 (6.7)5 (20.8).116Neuromuscular blockers6 (6.7)4 (7.4)2 (5.6)1.0004 (19.1)1 (2.2)1 (4.17)1.000Immunosuppressants2 (2.2)0 (0.0)2 (5.6).1570 (0.0)1 (2.2)1 (4.17)1.000Anticoagulants/antiplatelets16 (17.8)9 (16.7)7 (19.4).7832 (9.5)10 (22.2)4 (16.7).756Other37 (41.1)28 (51.9)9 (25.0) .016 8 (38.1)22 (48.9)7 (29.2).132Targeted antimicrobial therapy, n (%)25 (27.8)14 (25.9)11 (30.6).6403 (14.3)14 (31.1)8 (33.3)1.000Isolated pathogens, n (%)33 (36.7)21 (38.9)12 (33.3).5926 (28.6)17 (37.8)10 (41.7).753MSSE10 (11.1)5 (9.3)5 (13.9)—1 (4.8)9 (20.0)0 (0.0)— Aspergillus spp.3 (3.3)3 (5.6)0 (0.0)—0 (0.0)0 (0.0)3 (12.5)— Escherichia coli 2 (2.2)1 (1.9)1 (2.8)—1 (4.8)0 (0.0)1 (4.2)— Haemophylus influenzae 2 (2.2)2 (3.7)0 (0.0)—0 (0.0)2 (4.4)0 (0.0)— Klebsiella pneumoniae 2 (2.2)2 (3.7)0 (0.0)—0 (0.0)1 (2.2)1 (4.2)—MRSA2 (2.2)1 (1.9)1 (2.8)—1 (4.8)1 (2.2)0 (0.0)—MRSE1 (1.1)0 (0.0)1 (2.8)—0 (0.0)0 (0.0)1 (4.2)—MSSA2 (2.2)1 (1.9)1 (2.8)—1 (4.8)0 (0.0)1 (4.2)— Staphylococcus hominis 2 (2.2)2 (3.7)0 (0.0)—1 (4.8)1 (2.2)0 (0.0)— Enterococcus faecium 1 (1.1)1 (1.9)0 (0.0)—0 (0.0)0 (0.0)1 (4.2)— Enterococcus fecalis, proteus spp.1 (1.1)0 (0.0)1 (2.8)—0 (0.0)1 (2.2)0 (0.0)— Staphylococcus haemoliticus 1 (1.1)0 (0.0)1 (2.8)—0 (0.0)0 (0.0)1 (4.2)— Staphylococcus warneri 1 (1.1)1 (1.9)0 (0.0)—0 (0.0)1 (2.2)0 (0.0)— Streptococcus lutetiensis 1 (1.1)1 (1.9)0 (0.0)—0 (0.0)1 (2.2)0 (0.0)—Gram‐negative rods1 (1.1)1 (1.9)0 (0.0)—1 (4.8)0 (0.0)0 (0.0)— Clostridium difficile 1 (1.1)0 (0.0)1 (2.8)—0 (0.0)0 (0.0)1 (4.2)—LoS, days, median (IQR)34.5 (20–66)32.5 (22–67)39.5 (17.5–64.5).914720 (16–35)35 (28–64)45 (22–77.5).9047Death, n (%)10 (11.1)5 (9.3)5 (13.9).5133 (14.3)2 (4.4)5 (20.8) .045 Newborns, n (%)40 (44.4)27 (50.0)13 (36.1).27913 (61.9)19 (42.2)8 (33.3).606Sex, newborns, female, n (%)17 (42.5)12 (44.4)5 (38.5)1.0008 (61.5)8 (42.1)1 (12.5).201Age, newborns, days, median (IQR)9.3 (5.2–14.0)10.5 (5.6–14.7)9.0 (3.2–12.5).55383.8 (2.7–5.9)12.5 (8.6–17.7)9.9 (7.9–11.4).2220Weight at birth, kg, median (IQR)2.8 (0.9–3.3)2.6 (0.8–3.3)2.9 (1.7–3.3).58252.3 (1.2–3.5)2.9 (0.8–3.3)2.9 (0.7–3.2).8171 n = 38 n = 26 n = 12— n = 12 n = 19 n = 7—Weight at SDC, newborns, median (IQR)2.8 (1.0–3.1)2.2 (0.8–3.1)3.0 (2.9–3.1).28113.3 (2.6–3.6)2.3 (0.8–3.0)2.1 (1.1–3.1).5906 n = 16 n = 11 n = 5— n = 4 n = 10 n = 2—Time to first SDC, newborns, days, median (IQR)2 (1–3.5)2 (1–3.5)1 (1–3).31762 (1–2.5)2 (1–4)1 (0–2).1823 n = 36 n = 24 n = 12— n = 12 n = 17 n = 7—LoS, newborns, days, median (IQR)36 (21–72)35 (20–79)42 (24–63).942419 (13–27)42 (35–77)83 (42.5–112).2115Death, newborns, n (%)4 (10.0)2 (7.4)2 (15.4).5842 (15.4)1 (5.3)1 (12.5).513Non‐newborns, n (%)50 (55.6)27 (50.0)23 (63.9).2798 (38.1)26 (57.8)16 (66.7).606Sex, non‐newborns, female, n (%)20 (40.0)11 (40.7)9 (39.1)1.0001 (12.5)12 (46.2)7 (43.8)1.000Age, non‐newborns, median (IQR)1.2 (0.5–15.8)12.1 (0.5–16.4)0.8 (0.6–1.8).07499.7 (1.1–16.3)0.9 (0.5–12.1)13.4 (0.6–16.8).1083Weight at SDC, non‐newborns, median (IQR)7.3 (5.8–10.0)7.2 (5.8–60.2)7.4 (5.7–9.1).597253.5 (26.0–70.5)7.3 (5.8–8.7)6.6 (5.7–8.4).4577 n = 33 n = 14 n = 19— n = 4 n = 15 n = 14—Time to first SDC, non‐newborns, days, median (IQR)2 (1–3)2 (1–5)1 (0–2) .0063 1 (0.5–2)2 (1–3)2 (0.5–4.5).8599 n = 43 n = 23 n = 20— n = 8 n = 23 n = 12—LoS, non‐newborns, days, median (IQR)30 (19–51)28 (23.46)34 (16–66).992227 (18–64)30.5 (22–51)31 (20.5–56.5).8358Death, non‐newborns, n (%)6 (12.0)3 (11.1)3 (13.0)1.0001 (12.5)1 (3.9)4 (25.0).061 Note: Concomitant medications are categorized as follows: cardiovascular agents: antihypertensives, antiarrhythmics; amines: adrenaline, dobutamine, dopamine, noradrenaline; vasodilators: fenoldopam, milrinone, sildenafil, prostanoids; diuretics: etacrynic acid, furosemide, spironolactone; sedatives: benzodiazepines, dexmedetomidine, propofol; anticonvulsants: carbamazepine, gabapentin, phenytoin, valproic acid, phenobarbital; neuromuscular blockers: succinylcholine; other agents: levothyroxine, proton pump inhibitors, prokinetics, ursodeoxycholic acid, vitamins. Statistically significant comparisons are emphasized in bold.Abbreviations: CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; IQR, interquartile range; LoS, length of stay; MSSA, methicillin sensitive Staphylococcus aureus; MSSE, methicillin sensitive Staphylococcus epidermidis; MRSA, methicillin resistant S. aureus; MRSE, methicillin resistant S. epidermidis; NSAIDs, non‐steroidal anti‐inflammatory drugs; SDC, serum drug concentration. Study population and outcome analysis in the overall population. Note: Concomitant medications are categorized as follows: cardiovascular agents: antihypertensives, antiarrhythmics; amines: adrenaline, dobutamine, dopamine, noradrenaline; vasodilators: fenoldopam, milrinone, sildenafil, prostanoids; diuretics: etacrynic acid, furosemide, spironolactone; sedatives: benzodiazepines, dexmedetomidine, propofol; anticonvulsants: carbamazepine, gabapentin, phenytoin, valproic acid, phenobarbital; neuromuscular blockers: succinylcholine; other agents: levothyroxine, proton pump inhibitors, prokinetics, ursodeoxycholic acid, vitamins. Statistically significant comparisons are emphasized in bold. Abbreviations: CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; IQR, interquartile range; LoS, length of stay; MSSA, methicillin sensitive Staphylococcus aureus; MSSE, methicillin sensitive Staphylococcus epidermidis; MRSA, methicillin resistant S. aureus; MRSE, methicillin resistant S. epidermidis; NSAIDs, non‐steroidal anti‐inflammatory drugs; SDC, serum drug concentration. Stratified analyses by age group yielded largely comparable results. Among newborns, no significant differences were observed in mortality (15.4% vs. 7.4%; p = .584) or other clinical outcomes. In non‐newborn patients, findings were similarly consistent, with the exception of a significantly longer time to first TDM in the non‐toxic group (p = .006) (Table 2). In the vancomycin‐treated subgroup, overall results were consistent with the main analysis. Differences in ECMO use and renal failure did not reach statistical significance (p = .054 and p = .068, respectively), whereas CRRT was significantly more frequent in the toxic group (p = .036) (Table 3). Survival analyses comparing toxic and non‐toxic groups are reported in Figure 2; no significant differences were observed. TABLE 3Study population and outcome analysis in the vancomycin‐treated population.Binary toxic/non‐toxic classificationClinical classificationOverall (n = 69)Non‐toxic (n = 38)Toxic (n = 31) p valueIndeterminate (n = 9)Not‐in‐range (n = 42)In‐range (n = 18) p valueSex, female, n (%)26 (37.7)15 (39.5)11 (35.5).806—19 (45.2)5 (27.8).258Age, years, median (IQR)0.5 (0.05–1.6)0.2 (0.04–14.7)0.6 (0.07–1.2).55448.4 (0.6–15.5)0.2 (0.0–1.0)0.6 (0.0–15.8).2727Weight at SDC, median (IQR)6.5 (3.1–8.3)5.7 (1.9–7.3)7.0 (4.2–8.7).184453.5 (26.0–70.5)5.5 (2.9–7.8)6.3 (4.2–7.4).3904 n = 40 n = 17 n = 23— n = 4 n = 25 n = 11—Time to first SDC, days, median (IQR)1.5 (1–3)2 (1–4)1 (0–2) .0103 1 (1–2)2 (1–4)1 (0–2).0641 n = 60 n = 33 n = 27— n = 9 n = 38 n = 13—Clinical classification details, n (%)———NA———NANever in‐rangeNANANA——6 (14.3)——UnderdosedNANANA——34 (81.0)——ToxicNANANA——2 (4.8)——Always in‐rangeNANANA———14 (77.8)—Underdosed than brought in‐rangeNANANA———4 (22.2)—Comorbidities, n (%)65 (94.2)35 (92.1)30 (39.8).6228 (88.9)41 (97.6)16 (88.9).212Cardiovascular48 (69.6)20 (52.6)28 (90.3) .001 6 (66.7)27 (64.3)15 (83.3).220Respiratory22 (31.9)16 (42.1)6 (19.4).0682 (22.2)13 (31.0)7 (38.9).564Gastrointestinal10 (14.5)8 (21.1)2 (6.5).1671 (11.1)6 (14.3)3 (16.7)1.000Hepatic failure4 (5.8)2 (5.3)2 (6.5)1.0000 (0.0)1 (2.4)3 (16.7).077Renal failure9 (13.4)2 (5.3)7 (22.6).0681 (11.1)2 (4.8)6 (33.3) .007 Neurological16 (23.2)12 (31.6)4 (12.9).0894 (44.4)8 (19.1)4 (22.2).740Onco‐haematological0 (0.0)0 (0.0)0 (0.0)NA0 (0.0)0 (0.0)0 (0.0)NAImmunodepression1 (1.5)0 (0.0)1 (3.2).4490 (0.0)0 (0.0)1 (5.6).300Autoimmune0 (0.0)0 (0.0)0 (0.0)NA0 (0.0)0 (0.0)0 (0.0)NAShock11 (15.9)6 (15.8)5 (16.1)1.0002 (22.2)7 (16.7)2 (11.1).710ECMO11 (15.9)3 (7.9)8 (25.8).0542 (22.2)4 (9.5)5 (27.8).111Intubation45 (65.2)25 (65.8)20 (64.5)1.0007 (77.8)26 (61.9)12 (66.7).778CRRT4 (5.8)0 (0.0)4 (12.9) .036 2 (22.2)0 (0.0)2 (11.1).086Concomitant medications, n (%)64 (92.8)34 (89.5)30 (96.8).3709 (100)38 (90.5)17 (94.4)1.000Cardiovascular16 (23.2)6 (15.8)10 (32.3).1532 (22.2)9 (21.4)5 (27.8).740Amines23 (33.3)10 (26.3)13 (41.9).2052 (22.2)13 (31.0)8 (44.4).381Vasodilators20 (29.0)9 (23.7)11 (35.5).3011 (11.1)13 (31.0)6 (33.3)1.000Diuretics24 (34.8)9 (23.7)15 (48.4) .043 2 (22.2)14 (33.3)8 (44.4).560NSAIDs/paracetamol11 (15.9)9 (23.7)2 (6.5).0951 (11.1)7 (16.7)3 (16.7)1.000Glucocorticoids5 (7.3)2 (5.3)3 (9.7).6511 (11.1)1 (2.4)3 (16.7).077Antivirals1 (1.5)0 (0.0)1 (3.2).4490 (0.0)1 (2.4)0 (0.0)1.000Other antimicrobials54 (78.3)28 (73.7)26 (83.9).3869 (100)31 (73.8)14 (77.8)1.000Opiates/opioids17 (24.6)10 (26.3)7 (22.6).7843 (33.3)10 (23.8)4 (22.2)1.000Sedatives16 (23.2)10 (26.3)6 (19.4).5742 (22.2)8 (19.1)6 (33.3).319Anticonvulsants5 (7.3)5 (13.2)0 (0.0).0602 (22.2)1 (2.4)2 (1.1).212Neuromuscular blockers2 (2.9)1 (2.6)1 (3.2)1.0000 (0.0)1 (2.4)1 (5.6).514Immunosuppressants2 (2.9)0 (0.0)2 (6.5).1980 (0.0)1 (2.4)1 (5.6).514Anticoagulants/antiplatelets14 (20.3)7 (18.4)7 (22.6).7670 (0.0)10 (23.8)4 (22.2)1.000Other27 (39.1)19 (50.0)8 (25.8).0501 (11.1)21 (50.0)5 (27.8).157Targeted antimicrobial therapy, n (%)20 (29.0)10 (26.3)10 (32.3).6052 (22.2)13 (31.0)5 (27.8)1.000Isolated pathogens, n (%)26 (37.7)15 (39.5)11 (35.5).7345 (55.6)16 (38.1)5 (27.8).443MSSE10 (14.5)5 (13.2)5 (16.1)—1 (11.1)9 (21.4)0 (0.0)— Aspergillus spp.0 (0.0)0 (0.0)0 (0.0)—0 (0.0)0 (0.0)0 (0.0)— Escherichia coli 0 (0.0)0 (0.0)0 (0.0)—0 (0.0)0 (0.0)0 (0.0)— Haemophylus influenzae 1 (1.5)1 (2.6)0 (0.0)—0 (0.0)1 (2.4)0 (0.0)— Klebsiella pneumoniae 1 (1.5)1 (2.6)0 (0.0)—0 (0.0)1 (2.4)0 (0.0)—MRSA2 (2.9)1 (2.6)1 (3.2)—1 (11.1)1 (2.4)0 (0.0)—MRSE1 (1.5)0 (0.0)1 (3.2)—0 (0.0)0 (0.0)1 (5.6)—MSSA2 (2.9)1 (2.6)1 (3.2)—1 (11.1)0 (0.0)1 (5.6)— Staphylococcus hominis 2 (2.9)2 (5.3)0 (0.0)—1 (11.1)1 (2.4)0 (0.0)— Enterococcus faecium 1 (1.5)1 (2.6)0 (0.0)—0 (0.0)0 (0.0)1 (5.6)— Enterococcus fecalis, proteus spp.1 (1.5)0 (0.0)1 (3.2)—0 (0.0)1 (2.4)0 (0.0)— Staphylococcus haemoliticus 1 (1.5)0 (0.0)1 (3.2)—0 (0.0)0 (0.0)1 (5.6)— Staphylococcus warneri 1 (1.5)1 (2.6)0 (0.0)—0 (0.0)1 (2.4)0 (0.0)— Streptococcus lutetiensis 1 (1.5)1 (2.6)0 (0.0)—0 (0.0)1 (2.4)0 (0.0)—Gram negative rods1 (1.5)1 (2.6)0 (0.0)—1 (11.1)0 (0.0)0 (0.0)— Clostridium difficile 1 (1.5)0 (0.0)1 (3.2)—0 (0.0)0 (0.0)1 (5.6)—LoS, days, median (IQR)35 (25–66)35 (27–77)42 (19–66).625027 (19–38)37 (28–67)40.5 (22–66).6630Death, n (%)5 (7.3)2 (5.3)3 (9.7).6511 (11.1)2 (4.8)2 (11.1).576Newborns, n (%)25 (32.2)17 (44.7)8 (25.8).1341 (11.1)18 (42.9)6 (33.3).573Sex, newborns, female, n (%)9 (36.0)7 (41.2)2 (25.0).6611 (11.1)8 (44.4)0 (0.0).066Age, newborns, days, median (IQR)12.4 (9.3–17.7)12.4 (10.8–17.7)10.9 (7.8–19.3).641222.6 (NA)13.0 (9.2–17.7)10.2 (8.9–11.8).3506Weight at birth, kg, median (IQR)2.9 (0.8–3.2)2.4 (7.5–3.2)3.0 (2.9–3.5).08062.4 (NA)2.9 (0.8–3.3)2.9 (2.8–3.1).7941 n = 24 n = 17 n = 7— n = 1 n = 18 n = 5—Weight at SDC, newborns, median (IQR)2.7 (0.8–3.1)0.9 (0.8–2.7)3.1 (2.9–3.3) .0294 NA (NA)2.3 (0.8–3.0)3.1 (NA).2673 n = 11 n = 7 n = 4— n = 0 n = 10 n = 1—Time to first SDC, newborns, days, median (IQR)2 (1–4)2 (1–4)1 (0–4).26564 (NA)2 (1–4)1 (0–1).0582 n = 22 n = 15 n = 7— n = 1 n = 16 n = 5—LoS, newborns, days, median (IQR)63 (35–79)64 (35–81)56 (36–64.5).599427 (NA)45.5 (35–77)87.5 (63–115).1510Death, newborns, n (%)1 (4.0)1 (5.6)0 (0.0)1.0000 (0.0)1 (5.6)0 (0.0)1.000Non‐newborns, n (%)44 (63.8)21 (55.3)23 (74.2).1348 (88.9)24 (57.1)12 (66.7).573Sex, non‐newborns, female, n (%)17 (38.6)13 (61.9)14 (60.9)1.0001 (11.1)11 (45.8)5 (41.7)1.000Age, non‐newborns, median (IQR)1.1 (0.5–15.2)11.0 (0.5–16.4)0.8 (0.6–1.8).20039.7 (1.1–16.3)0.7 (0.5–1.3)8.2 (0.6–16.1).1227Weight at SDC, non‐newborns, median (IQR)7.3 (5.8–9.1)7.2 (5.8–42.0)7.4 (5.7–9.1).696353.5 (26.0–70.5)7.3 (5.8–8.7)6.5 (5.7–7.4).2218 n = 29 n = 10 n = 19— n = 4 n = 15 n = 10—Time to first SDC, non‐newborns, days, median (IQR)1 (0–2)2 (1–3)1 (0–2) .0275 1 (0.5–2)1.5 (1–3)1.5 (0–2).4155 n = 38 n = 18 n = 20— n = 8 n = 22 n = 8—LoS, non‐newborns, days, median (IQR)30 (22–49)28 (25–38)34 (16–66).971927 (18–64)32.5 (25.5–54.5)25 (19–45.5).2902Death, non‐newborns, n (%)4 (9.09)1 (4.8)3 (13.0).6091 (12.5)1 (4.2)2 (16.7).253 Note: Concomitant medications are categorized as follows: cardiovascular agents: antihypertensives, antiarrhythmics; amines: adrenaline, dobutamine, dopamine, noradrenaline; vasodilators: fenoldopam, milrinone, sildenafil, prostanoids; diuretics: etacrynic acid, furosemide, spironolactone; sedatives: benzodiazepines, dexmedetomidine, propofol; anticonvulsants: carbamazepine, gabapentin, phenytoin, valproic acid, phenobarbital; neuromuscular blockers: succinylcholine; other agents: levothyroxine, proton pump inhibitors, prokinetics, ursodeoxycholic acid, vitamins. Statistically significant comparisons are emphasized in bold.Abbreviations: CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; IQR, interquartile range; LoS, length of stay; MSSA, methicillin sensitive Staphylococcus aureus; MSSE, methicillin sensitive Staphylococcus epidermidis; MRSA, methicillin resistant S. aureus; MRSE, methicillin resistant S. epidermidis; NSAIDs, non‐steroidal anti‐inflammatory drugs; SDC, serum drug concentration. Study population and outcome analysis in the vancomycin‐treated population. Note: Concomitant medications are categorized as follows: cardiovascular agents: antihypertensives, antiarrhythmics; amines: adrenaline, dobutamine, dopamine, noradrenaline; vasodilators: fenoldopam, milrinone, sildenafil, prostanoids; diuretics: etacrynic acid, furosemide, spironolactone; sedatives: benzodiazepines, dexmedetomidine, propofol; anticonvulsants: carbamazepine, gabapentin, phenytoin, valproic acid, phenobarbital; neuromuscular blockers: succinylcholine; other agents: levothyroxine, proton pump inhibitors, prokinetics, ursodeoxycholic acid, vitamins. Statistically significant comparisons are emphasized in bold. Abbreviations: CRRT, continuous renal replacement therapy; ECMO, extracorporeal membrane oxygenation; IQR, interquartile range; LoS, length of stay; MSSA, methicillin sensitive Staphylococcus aureus; MSSE, methicillin sensitive Staphylococcus epidermidis; MRSA, methicillin resistant S. aureus; MRSE, methicillin resistant S. epidermidis; NSAIDs, non‐steroidal anti‐inflammatory drugs; SDC, serum drug concentration. Kaplan–Meier survival curves. Kaplan–Meier estimates of time to death from hospital admission (A,B) and from initiation of antimicrobial therapy (C,D). Curves are stratified by toxic vs. non‐toxic drug exposure (A,B) and by clinical classification (C,D). Numbers in parentheses indicate deaths per group. p values were calculated using the log‐rank test. Clinical classification comparison Considering the clinical course, patients were categorized into three groups: indeterminate (n = 21, 23.3%), not‐in‐range (n = 45, 50%), and in‐range (n = 24, 26.7%). Demographic characteristics were comparable across groups. Patients in the not‐in‐range group had a higher prevalence of comorbidities compared with those in the in‐range group (97.8% vs. 83.3%; p = .046). Conversely, renal failure was significantly more common in the in‐range group compared with the not‐in‐range group (29.2% vs. 4.4%; p = .007). Other comorbidities were similarly distributed among the groups. With regard to treatments, glucocorticoid use was significantly higher in the in‐range group compared with the not‐in‐range group (20.8% vs. 2.2%; p = .017). CRRT use was also significantly more frequent in the in‐range group (12.5% vs. 0.0%; p = .039). A higher all‐cause mortality was observed in the in‐range group, with five deaths (20.8%) compared with two deaths (4.4%) in the not‐in‐range group (p = .045). To explore the robustness of this finding, two sensitivity analyses were conducted. One patient initially classified as in‐range while receiving voriconazole was subsequently switched to isavuconazole (classified as underdosed) and died due to a mycotic aneurysm (Table 4; Patient #1). Excluding this patient or reclassifying them into the not‐in‐range group resulted in non‐significant differences in mortality (17.4% vs. 4.4%, p = .169; and 17.4% vs. 6.5%, p = .211, respectively). Survival analyses for the clinical classification comparison are presented in Figure 2 and did not reveal significant differences. TABLE 4Deceased patients' description.Clinical classificationAge, sexClinical condition/admission reasonsTargeted therapy (isolation matrix)Antimicrobial (classification)Death causeToxic/non‐toxic classificationIn‐range112 years, MBurns in >80% body surface area Apsergillus terreus (skin biopsy)Voriconazole (underdosed than brought in‐range), isavuconazole (underdosed)Mycotic aneurysmNon‐toxic215 years, MCardiac failure, Danon syndromeEmpiricalVancomycin (in‐range), amikacinMOFToxic310 days, FEmergency C‐section due to placental abruption at 25 + 5 weeks of gestational age; intestinal atresia and ascites; partial ileum resection due to intestinal ischaemic necrosisEmpirical, Staphylococcus hominis (blood), Staphylococcus epidermidis (blood)Vancomycin (in‐range), gentamicin (in‐range), meropenemCardiocirculatory failure, septic shock, secondary liver failureToxic41 year, FMitochondrial disease (NDUFS8 mutation), hypertrophic/obstructive cardiomyopathyEmpiricalVancomycin (in‐range)Cardiogenic shockToxic51 month, MCongenital cardiopathy, Danon syndrome, previous stroke Klebsiella pneoumoniae (blood)Vancomycin, amikacin, gentamicin, meropenem (in‐range)Haemorrhagic strokeNon‐toxicNot‐in‐range112 years, FAcute decompensated heart failure, congenital myopathyEmpiricalVancomycin (toxic), amikacin (toxic), piperacillin, tazobactamCardiomyopathyToxic210 days, MPremature 25 weeks + days Klebsiella pneoumoniae (peritoneal fluid)Vancomycin (underdosed), meropenemSmall bowel ischaemic necrosisNon‐toxicIndeterminate12 days, FPremature 22 weeks +6 days, respiratory failure, hypotension, pulmonary hypertension, coagulopathy, necrotizing enterocolitis, brain vessels congestion, and Gram‐ sepsisEmpirical, Staphylococcus haemoliticus (blood)Gentamicin (indeterminate), Vancomycin (underdosed), ampicillinCardio‐respiratory failureToxic22 days, FDelivery at 34 weeks of gestational age, fetal movements absence, pathological cardiotocographic chart, with coagulopathy and hypertensionEmpiricalGentamicin (indeterminate)Anoxic‐ischaemic encephalopathy and cardiocirculatory arrestNon‐toxic32 months, MDelivery at 37 weeks of gestational age, hypotonic and cyanotic at birth, treated with NIV and then mechanical ventilation, right colon resection due to intestinal ischaemic necrosis with ileostomy, aganglionic megacolon, refractory epileptic encephalopathy S. hominis (CVC)Vancomycin (indeterminate), meropenemRespiratory failureNon‐toxicAbbreviations: CVC, central vascular catheter; F, female, M, male, MOF, multi‐organ failure, NIV, non‐invasive ventilation. Deceased patients' description. Abbreviations: CVC, central vascular catheter; F, female, M, male, MOF, multi‐organ failure, NIV, non‐invasive ventilation. Subgroup analysis of newborns and non‐newborns showed no significant differences in most clinical characteristics across groups. Similarly, in the vancomycin‐treated subgroup, results were largely consistent with the main analysis, apart from all‐cause mortality and CRRT use which resulted in non‐significant differences (p = .576 and p = .086, respectively) (Table 3). Death causes A total of 10 deaths were observed among included patients (Table 4). Considering the clinical classification, among the in‐range patients, deaths were registered in the following patients: (1) a 12‐year‐old male admitted in the ICU for burns in more than 80% of body surface area who received voriconazole for cutaneous infection by Aspergillus terreus and died due to mycotic aneurysm; (2) a 15‐year‐old male admitted for cardiac failure associated with congenital cardiomyopathy and Danon syndrome who received empirical vancomycin and amikacin and died due to multi‐organ failure; (3) a 10‐day‐old female admitted after emergency caesarean delivery due to placental abruption and affected by intestinal atresia and ascites who received empirical vancomycin and meropenem and died due to cardiocirculatory failure, septic shock and secondary liver failure; (4) a 1‐year‐old female with mitochondrial disease associated with NDUFS8 mutation and hypertrophic/obstructive cardiomyopathy who received empirical vancomycin and died due to cardiogenic shock; (5) a 1‐month‐old male admitted for congenital cardiopathy and Danon syndrome with previous stroke who received vancomycin, amikacin and gentamicin for Klebsiella pneumoniae and died due to haemorrhagic stroke. Among not‐in‐range patients, deaths were registered in (1) a 12‐year‐old female admitted for acute decompensated heart failure and with congenital myopathy who received empirical vancomycin, piperacillin and tazobactam and died due to cardiomyopathy; (2) a 10‐day‐old male, born premature at 25 weeks + 4 days of gestational age, who received meropenem and vancomycin for K. pneumoniae isolated from peritoneal fluid and died due to small bowel ischaemic necrosis. Among indeterminate patients, deaths were reported in: (1) a 2‐day‐old female, prematurely born at 22 weeks + 6 days of gestational age, admitted with respiratory failure, hypotension, pulmonary hypertension, coagulopathy, necrotizing enterocolitis, brain vessels congestion and Gram‐ sepsis who received empirical gentamicin and ampicillin and subsequently targeted vancomycin (considered underdosed) for Staphylococcus haemoliticus and died due to cardio‐respiratory failure; (2) a 2‐day‐old male admitted after delivery at 34 weeks of gestational age due to fetal movement absence and pathological cardiotocographic chart, with coagulopathy and hypertension, who received empirical gentamicin and died due to anoxic‐ischaemic encephalopathy and cardiocirculatory arrest; (3) a 2 months old male admitted for hypotonia and cyanosis at birth, respiratory distress, right colon resection due to intestinal ischaemic necrosis with ileostomy, aganglionic megacolon, refractory epileptic encephalopathy, who received vancomycin and meropenem for Staphylococcus hominis isolated from central venous catheter, which resolved during hospital stay, and died due to respiratory failure. DISCUSSION In our study, we observed that 22% of paediatric patients undergoing antimicrobial TDM had at least one SDC above toxicity threshold (i.e., toxic group); this frequency was similar when considering only the first SDC determination for each patient, as well as the most represented antimicrobial, which was vancomycin. As for the clinical classification, a large part (45, 50%) of our patients was not‐in‐range, more frequently underdosed, with only 33.3% being in‐range. Similarly, in another study in critically‐ill children and newborns receiving vancomycin, initial trough levels were in‐range in only the 30% of patients, whereas 61% had subtherapeutic levels and 8% had supratherapeutic levels, the latter reaching 15% when considering neonates only. 13 Conversely, we reported more supratherapeutic levels in non‐newborns than in newborns. This difference is difficult to interpret since the two studies were performed with different designs, collected baseline data were diversely reported, and residual bias due to disease severity and comorbidities associated with the limited sample size could not be excluded. However, it is important to note that a significant proportion of SDC measurements fell outside the target range. Even in the retrospective study by Lim et al., 4 which analysed 326 first‐dose TDM samples for gentamicin, amikacin and vancomycin in paediatric patients, 26.4% of measured concentrations fell outside the therapeutic range. Although a higher proportion of treatments were within range after the first dose, these findings are consistent with our observation that a substantial proportion of SDCs fail to meet therapeutic targets, with authors underscoring that achieving appropriate drug exposure remains a widespread challenge in paediatric antimicrobial therapy. Several factors may contribute to this. Among the others, it should be considered that pharmacokinetics varies substantially across different paediatric developmental phases, particularly in newborns, because rapid maturational changes profoundly affect drug absorption, distribution, metabolism and excretion. Neonates have immature hepatic and renal function, higher total body water, lower fat content, and altered plasma protein binding, all of which result in marked inter‐ and intra‐individual variability in drug exposure. 14 , 15 Additionally, the age‐dependent variability we observed is also consistent with prior pharmacokinetic literature. For example, Kearns et al. 2 reported substantial age‐related differences in crucial pharmacokinetic parameters, such as volume of distribution and clearance, particularly pronounced in neonates. These findings support the interpretation that the discrepancies we detected between newborns and non‐newborns are biologically plausible and reflect known maturational processes affecting drug disposition. Indeed, according to Guido et al., 16 in the neonatal intensive care unit, the proportion of patients achieving target concentrations at the first measurement is significantly lower (15%) compared with other non‐neonatal wards, where target attainment, although still low, ranges from 30% to 33%. Regardless of the underlying cause, the key message for clinicians is to carefully consider this variability in their clinical decision‐making. In our study we observed an association with cardiovascular comorbidities, renal failure (and CRRT in vancomycin‐treated patients), ECMO use and diuretic use in patients included in the toxic group. These findings suggest that, in more severely ill patients, particularly those with renal involvement, altered pharmacokinetics may increase the likelihood of antimicrobial accumulation (in our case, especially vancomycin). Importantly, these associations should be interpreted as reflecting underlying illness severity and organ dysfunction, rather than as a direct effect of drug exposure per se. In a study by Cheng et al. 17 involving ECMO patients with or without concomitant CRRT receiving vancomycin, dialyzed patients were exposed to a markedly increased risk of overdosing, with probabilities of toxicity exceeding 92%, even when lower‐dose regimens were used. Furthermore, Bobrowski et al. 18 reported that lower vancomycin doses are required in ECMO patients to achieve serum concentrations comparable to those observed in non‐ECMO subjects, and that reduced dosing regimens should be preferred in ECMO patients with impaired renal function at treatment initiation. When considering the clinical classification, renal failure, glucocorticoid use and CRRT were more common in the in‐range group, probably driven by patient severity and intensity of clinical management, which also influenced the frequency and timeliness of TDM interventions. In fact, this initially seems counterintuitive, as these variables would typically be expected to correlate with the not‐in‐range group, but a plausible explanation may be that patients with renal failure and those requiring CRRT tend to be more critically ill and, in our centre, are more rigorously monitored through TDM. The combination of closer monitoring and proactive dose adjustments likely facilitated target attainment in these patients. As further confirmation, patients in the not‐in‐range group were more commonly underdosed. Therefore, outcome differences, including mortality, should not be interpreted exclusively as exposure‐driven effects, but rather as reflecting residual confounding related to disease severity and indication for TDM. All‐cause mortality was significantly higher in the in‐range group in the overall population; however, this finding might be due to chance and the small sample size as shown by the Kaplan–Meier curves, the loss of significance in the subgroup analysis for vancomycin, and after the two sensitivity analyses in the overall population. Additionally, patients exhibiting toxic concentrations likely represent a subgroup with greater baseline severity, in whom drug accumulation and worse outcomes coexist rather than being causally linked. Although antimicrobial dosing was corrected following TDM, mortality in these patients may have been driven by unrelated factors such as underlying comorbidities rather than antimicrobial toxicity or efficacy. Notably, a reverse‐causality paradox has been described in the literature, whereby more critically ill patients experience worse outcomes and higher rates of clinical failure despite achieving higher serum antibiotic concentrations and exhibiting lower rates of underdosing. 19 Furthermore, the challenges of demonstrating the clinical impact of TDM on mortality in dedicated studies have been well recognized, particularly due to difficulties in patient enrolment and sample size limitations. 20 Antifungal TDM in paediatrics was rarely performed in our study population. Although no occurrences of toxicity were recorded, this does not justify the absence of TDM for this drug class, given the small sample size and the well‐known toxicity profile of antifungal agents. On the other hand, a retrospective study on voriconazole by Brüggemann et al. 21 reported that the first measured voriconazole concentration was below the therapeutic target in 44% of patients. Thus, the implementation of antifungal TDM may enable rapid dose optimization to achieve clinical efficacy, further supporting the need for routine antifungal TDM in complex paediatric cases, while the low number of antifungal measurements in our cohort likely reflects underutilization rather than a low clinical demand. A solid correlation between TDM implementation and improved clinical outcomes in paediatric settings has not yet been conclusively demonstrated for antimicrobial therapy. 8 However, such a correlation has been well established in other clinical fields, 22 , 23 is supported by a strong pharmacological rationale 24 and is further suggested by some publications reporting promising results. 25 A systematic review and meta‐analysis evaluating 1288 adult patients from 10 included randomized controlled trials reported improvements in patients who received personalized antimicrobial dosing, with significant enhancements observed in target attainment, treatment failure prevention, and nephrotoxicity reduction. Although non‐statistically significant, a tendency towards reduction in all‐cause mortality, clinical cure and microbiological outcomes was observed. 3 These findings suggest that microbiological cure may not be the main determinant to reduce mortality, especially in critical patients, where concomitant conditions severely affect the risk of mortality. Nevertheless, in this scenario, not achieving infection resolution would be further detrimental for the clinical outcome and TDM may be a useful tool to improve antimicrobial therapy. Target therapeutic concentrations may be more rapidly attained in patients undergoing TDM after the first dose of amikacin and gentamicin, although this was not observed in patients receiving vancomycin. 4 Interestingly, the aforementioned meta‐analysis showed significantly increased clinical cure in patients who received personalized antimicrobial dosing 3 and a cost–benefit analysis showed that the combination of pharmacy TDM with healthcare practitioners education resulted in 60% decrease in annualized hospital costs. 26 Nevertheless, antimicrobial TDM in paediatrics is sparsely implemented and the introduction of more structured programmes, also including software‐based tools embedded in electronic clinical records, could result in improved outcomes for children in critical conditions. 13 , 27 , 28 , 29 In this context, our study contributes valuable real‐world data in a group of underrepresented patients in existing literature. Limitations Our study has limitations. The retrospective design and limited data availability may introduce bias and limit the possibility to control for confounding variables. However, the scarcity of data highlights the limited adoption of TDM in paediatric settings, even in large centres like ours. Despite the small dataset, our approach was systematic, including all available SDC tests performed at our hospital during the study period. A significant limitation of our study was the inability to access certain crucial data elements from the medical records. Specifically, information on the exact time of drug administrations and blood withdrawals, drug dosages, dosing schedules and any subsequent changes in these parameters were not consistently documented in the hospital medical records. Instead, these data were automatically recorded and stored in dedicated software systems to which we did not have access, resulting in the unavailability of essential information for our analyses. In some wards, only brief narrative descriptions of dose adjustments were available; however, these lacked sufficient detail to allow extraction of usable data for a retrospective analysis. In other cases, medication records were maintained in separate paper‐based systems that were not retrievable at the time of the study. These limitations highlight the need for more integrated and user‐friendly hospital information systems that consolidate clinical, laboratory and medication data into a single accessible source. Such systems would facilitate higher‐quality research, simplify clinical workflows for physicians, and most importantly, improve patient care. A notable challenge in our study was the varying attributes of the exams requested by clinicians (i.e., peak, trough and ‘no attribution’). This heterogeneity in data presentation made it difficult to interpret the results accurately. Particularly, in the case of drug tests without specific attributions, it was not possible to discern the time elapsed since the last drug dosage administration. This inherent limitation was due to the extraction of data from the database and the way in which SDC tests were requested by clinicians. Consequently, the correct attribution (peak/trough) for ‘unspecified’ tests could not be determined, and these SDCs were analysed separately. However, it is crucial to emphasize that this limitation does not significantly impact the determination of toxicity levels, as the threshold values remain and are not dependent on the time since administration. Trying to classify the clinical courses of admitted patients posed several challenges. In fact, reconciling the need for a rigorous and objective classification with the need for clinical utility is difficult, as reality is far more complex than a binary classification. This is the reason why two different classifications were proposed, and why a considerable number of cases were classified as ‘indeterminate’ in the clinical classification. The first approach, based on a binary distinction between toxic and non‐toxic courses, may be considered overly simplistic, as even a single SDC outside the reference range is sufficient to classify a patient as toxic. Nevertheless, this approach has the advantage of being fully objective, as it does not rely on clinical judgement or interpretation. In contrast, the clinical classification was designed to address these limitations by incorporating clinical reasoning and contextual interpretation of laboratory results—an approach we believe is essential in clinical pharmacology consultations for TDM. However, some bias cannot be excluded, such as potential confounding by disease severity and a certain level of subjectivity. To mitigate these limitations, we implemented a blinded evaluation process involving multiple clinical pharmacologists, who applied general rules agreed upon in advance and assessed each patient's overall clinical course. Additionally, it should be considered that the two classifications may be not concordant because one patient could be classified as having toxic levels whether he or she demonstrated any SDC results above toxicity threshold, even if they were minimal. On the contrary, when applying the clinical classification, the assessor may judge as in‐range a patient who demonstrated sporadic results minimally out of therapeutic or toxic levels if their SDC results were overall adequate. We believe that this aspect may help depict the complexity of hospital stay of each patient. Therefore, while categorizing patients' clinical courses is by definition challenging, we believe that the combination of these two classification approaches represents the best option to depict real‐world situations, as each compensates for the limitations of the other. CONCLUSION In this retrospective paediatric cohort, antimicrobial TDM revealed a high frequency of out‐of‐range SDC, particularly for vancomycin and aminoglycosides. Potentially toxic concentrations were observed in approximately one fifth of hospital stays, often already at the first TDM assessment, and were more common in clinically complex patients with cardiovascular comorbidities, renal failure and greater illness severity. Although no significant differences in short‐term all‐cause mortality were observed between patients with in‐range and out‐of‐range drug concentrations across different classification, TDM may still provide clinically relevant information by identifying early and frequent alterations in antimicrobial exposure in vulnerable paediatric populations. In this context, TDM can support individualized dosing decisions in paediatric care and the recognition of potentially harmful drug concentrations, particularly in patients with high clinical complexity. Further studies are required to confirm these findings and to better elucidate the associations between antimicrobial exposure, adverse drug reactions and additional clinically meaningful outcomes beyond mortality. AUTHOR CONTRIBUTIONS Gianluca Gazzaniga: Conceptualization; investigation; methodology; data curation; writing the original draft. Stefano Agliardi: Conceptualization; investigation; data curation; writing the original draft. Riccardo Giossi: Conceptualization; investigation; data curation; methodology; formal analysis; writing the original draft; visualization. Alice Schianchi: Investigation; data curation; writing review and editing. Alessandra Romandini: Investigation; writing review and editing. Thomas Virdis: Investigation; data curation; writing review and editing. Silvia Nerini Molteni: Investigation; data curation; writing review and editing. Maria Luisa De Angelis: Investigation; writing review and editing. Stefania Paccagnini: Investigation; writing review and editing. Cecilia Gechtman: Investigation; writing review and editing. Alice Proto: Investigation; writing review and editing. Marco Fossati: Investigation; writing review and editing. Silvia Oliveri: Investigation; writing review and editing. Giulia Angela Pattarino: Investigation; writing review and editing. Elena Altieri: Investigation; writing review and editing. Agnese Graziosi: Investigation; writing review and editing. Arianna Pani: Investigation; writing review and editing. Romano Danesi: Investigation; writing review and editing. Stefano Martinelli: Investigation; writing review and editing. Costantino De Giacomo: Investigation; writing review and editing. Francesco Scaglione: Conceptualization; methodology; supervision; writing review and editing. CONFLICT OF INTEREST STATEMENT R. G. received support for congress participation from Mylan, acted as a consultant for C.I.R.M.—Consorzio Italiano per la Ricerca in Medicina and Daiichi‐Sankyo—and received fees for speaking from Alexion–AstraZeneca; he was sub‐investigator in clinical trials sponsored by Sanofi, argenx, Alexion–AstraZeneca and Momenta–Johnson&Johnson, outside this work. All other authors declare no conflicts of interest related to this work. Supporting information Figure S1: Serum drug concentrations for vancomycin (Panel A) and Gentamicin (Panel B) stratified by sampling type and age class. ACKNOWLEDGEMENTS Open access publishing facilitated by Azienda Socio Sanitaria Territoriale Grande Ospedale Metropolitano Niguarda, as part of the Wiley – SBBL agreement. DATA AVAILABILITY STATEMENT The data that support the findings of this study are available from the corresponding author upon reasonable request. REFERENCES Associated Data Supplementary Materials Figure S1: Serum drug concentrations for vancomycin (Panel A) and Gentamicin (Panel B) stratified by sampling type and age class. Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.