Procalcitonin levels as a predictor of antibiotic treatment in pediatric diabetic ketoacidosis Authors: Katie D. Dolak, Samuel Davila, Perrin C. White Published: 2026 ABSTRACT BackgroundProcalcitonin is a useful screen for bacterial infection, but its utility in pediatric diabetic ketoacidosis (DKA) is unknown.MethodsWe conducted a retrospective study of DKA patients admitted to two pediatric hospitals in 2019–2023.Results116 admissions had available procalcitonin levels, of which in 32, patients were prescribed a complete antibiotics course and thus presumed to have a bacterial infection. More of the treated group were female (69% vs. 45%, p = 0.036), febrile (47% vs. 22%, p = 0.013), and/or had pneumonia on chest x-ray (30.8% vs. 4.76%, p = 0.023). Area under (AUC) a receiver operating characteristic (ROC) curve for procalcitonin to predict clinically suspected infection was 0.51 (no predictive value), whereas a ROC curve using fever and C-reactive protein (CRP) as predictors had an AUC of 0.79. Including procalcitonin in this model did not improve AUC.ConclusionsCRP and fever are useful predictors for clinically suspected infection in DKA patients, but procalcitonin is not. Background Bacterial infections are a known trigger of DKA [1], accounting for up to 50% of precipitating factors as the triggers of diabetic ketoacidosis (DKA) in adults [2]. Pediatric data suggests a much lower incidence of bacterial infection in DKA, closer to 13% [3]. However, identifying bacterial infection in DKA remains an important clinical question, as individuals with type 1 and 2 diabetes are at higher risk for mortality from infection [4]. Further complicating this question, DKA itself triggers the systemic inflammatory response system even in the absence of a bacterial infection [5]. Thus, a biomarker specific to bacterial infection in the context of DKA could be useful. Procalcitonin is an acute phase reactant released in response to endotoxin and the cytokines interleukin (IL)-1β, tumor necrosis factor-alpha, and IL-6 [6]. It is more specific to bacterial than viral infections since its release is attenuated by interferon [6]. This property of procalcitonin gives it clinical utility in the general pediatric population in detecting bacterial infection as the cause of fever without apparent source [7]. It has similarly proven useful in informing the decision to stop antibiotics in the management of neonatal sepsis [8]. Similar guidance in patients with DKA could reduce unnecessary antibiotic administration and thus antibiotic resistance. It is not known if procalcitonin, as opposed to other inflammatory markers, is also more specific for inflammation due to bacterial infection in the setting of pediatric DKA. Prior studies have evaluated the utility of sepsis markers including procalcitonin in patients with DKA in adults [9–13] with varying conclusions. For example, proposed optimum cut-off values for procalcitonin as a marker of infection ranged from 0.438 to 1.775 ng/mL 9–13. Apart from one study that included patients down to age 13 years [13], no studies have evaluated the utility of procalcitonin as a marker of bacterial infection in the pediatric DKA population. In this study, we aimed to determine if procalcitonin levels in pediatric patients with DKA were correlated with clinically suspected infection as indicated by prescription of a full antibiotics course. Secondarily, we aimed to determine if there were associations between procalcitonin levels and degree of hyperglycemia or acidosis. Methods This study was approved by the Institutional Review Board of UT Southwestern Medical Center, and the present report was designed to adhere to STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines [14]. We conducted a retrospective cohort study of patients admitted to two affiliated pediatric hospitals, Children’s Medical Center Dallas and Children’s Medical Center Plano, from January 1, 2020 through December 31, 2023. These two affiliated centers utilize the same laboratory assays and adhere to the same clinical policies. Patient charts were identified using Epic SlicerDicer (Madison, WI) to interrogate the electronic health record. We included patients aged 18 years or younger admitted with a diabetes diagnosis who had a procalcitonin checked during an inpatient admission. Patients were considered to have diabetes based on the American Diabetes Association (ADA) criteria for diagnosis of diabetes in non-pregnant individuals; that is, A1c greater than or equal to 6.5%, fasting blood glucose greater than or equal to 126 mg/dL, plasma glucose greater than or equal to 200 mg/dL at 2 h during an oral glucose tolerance test, or a random blood glucose of 200 mg/dL with classic symptoms of hyperglycemia [15]. Patients with diabetes were further categorized as type 1 or type 2 based on type 1 antibody testing and obesity status (with absence of antibodies along with presence of obesity categorized as type 2) [15]. We excluded patients who were not in DKA or who had a type of diabetes other than type 1 or 2. DKA was defined as glucose greater than 200 mg/dL or prior history of diabetes, β-hydroxybutyrate of 3.0 mmol/L, and pH < 7.3 and/or bicarbonate less than 18 mmol/L [16]. Data were collected from the institutional Epic clinical data repository, and included demographic data, vital signs, laboratory data (pH, blood glucose, hemoglobin A1c, bicarbonate, C-reactive protein, procalcitonin, erythrocyte sedimentation rate, alanine aminotransferase (ALT), aspartate aminotransferase (AST), leukocyte counts, blood cultures, urine cultures, chest x-rays, viral polymerase chain reaction tests) as well as clinical information (such as duration and antibiotic course details; see Table 1 for reference ranges of laboratory values where applicable). The investigators were not blinded to procalcitonin level during their review of retrospective data. For patients with multiple admissions for DKA in which a procalcitonin level was checked, we reviewed data from each admission. If multiple inflammatory marker levels were checked during a single encounter, the first level drawn while the patient was still in DKA was reported. Patients were considered febrile if a recorded temperature while in DKA was 38 degrees Celsius or greater. Transaminases (AST and ALT) were collected since procalcitonin is hepatically cleared and thus can be impacted in liver disease [17]. Of note, the vendor for the procalcitonin assay changed from Siemens to Abbot (Atellica IM BRAUMS PCT to Alinity i BRAHMS PCT) in November 2023, The lower limit of detection was < 0.04 ng/mL versus 0.01 ng/mL, and the measuring interval was 0.04-50.00 ng/mL versus 0.02 to 100.00 ng/mL respectively. Table 1Demographic and clinical data comparisonTreatment Antibiotics (N = 32)No Antibiotic Treatment (N = 84)P ValueReference RangeMedian Age (Y)14 (10–16)13 (11–15)0.42-Male Gender (%)10 (31%)46 (55%) 0.036 -Type 1 DM (%)26 (81%)73 (87%)0.56-New Onset? (%)15 (47%)31 (34%)0.40-ALT above reference range? (N = 68) (%)1 (5.9%) (N = 17)9(17.8%) (N = 51)0.4310–50 unit/LAST above reference range? (N = 68) (%)3 (17.6%) (N = 17)11(21.6%) (N = 51)1.010–45 units/LFebrile? (%)15 (47%)19 (22%) 0.013 -Median Initial BG427 (311–602)479 (368.25–637)0.1660–106 mg/dLMedian Initial pH7.07 (6.96–7.20)7.06 (6.97–7.17)0.687.35–7.45Median Initial Bicarb6.5 (4–9)7 (4–10)0.5616–25 mEq/LMedian Hgb A1c (%)12.2 (10.65–13.1)12.2 (10.5–13.5)0.95< 5.7%Median Procalcitonin0.39 (0.18–1.90)0.36 (0.14–2.11)0.88< 0.05 ng/mLMedian CRP(N = 102)5.5 (1.45–9.8) (N = 27)0.8 (0-2.22) (N = 75) < 0.0001 0.1-1.0 mg/dLMedian ESR(N = 27)25 (9.75-38) (N = 6)8 (8–12) (N = 21)0.160–20 mm/hrMedian Initial Leukocyte Count(N = 114)19.8 (11.55–30.45) (N = 31)19.7 (14.05–28.35) (N = 83)0.625.00-14.5 (Thousand/Mm3)Positive Blood Culture (%) (N = 57)2 (10.5%) (N = 19)0 (N = 38)0.1071-Positive Urine Culture (%) (N = 45)4 (28.6%) (N = 14)2 (6.45%) (N = 31)0.065-Pneumonia On Chest X-ray (%) (N = 55)4 (30.8%) (N = 13)2 (4.76%) (N = 42) 0.0232 -Positive Respiratory Viral Panel (%) (N = 90)8 (34.78%) (N = 23)25 (37.3%) (N = 67)1.0000-Antibiotics prior to procalcitonin level (%)8 (25%)15 (18%)0.44-Antibiotics prior to blood culture (%) (N = 57)8 (42%) (N = 19)15 (21%) (N = 38)0.12-Categorical variables compared using Fisher’s exact test, quantitative by Mann-Whitney U. Median values are followed by interquartile ranges. P values < 0.05 are in bold. If fewer than 32 patients in the treatment group or 84 in the non-treatment group had data available, the N for each group is specified Table 1Demographic and clinical data comparisonTreatment Antibiotics (N = 32)No Antibiotic Treatment (N = 84)P ValueReference RangeMedian Age (Y)14 (10–16)13 (11–15)0.42-Male Gender (%)10 (31%)46 (55%) 0.036 -Type 1 DM (%)26 (81%)73 (87%)0.56-New Onset? (%)15 (47%)31 (34%)0.40-ALT above reference range? (N = 68) (%)1 (5.9%) (N = 17)9(17.8%) (N = 51)0.4310–50 unit/LAST above reference range? (N = 68) (%)3 (17.6%) (N = 17)11(21.6%) (N = 51)1.010–45 units/LFebrile? (%)15 (47%)19 (22%) 0.013 -Median Initial BG427 (311–602)479 (368.25–637)0.1660–106 mg/dLMedian Initial pH7.07 (6.96–7.20)7.06 (6.97–7.17)0.687.35–7.45Median Initial Bicarb6.5 (4–9)7 (4–10)0.5616–25 mEq/LMedian Hgb A1c (%)12.2 (10.65–13.1)12.2 (10.5–13.5)0.95< 5.7%Median Procalcitonin0.39 (0.18–1.90)0.36 (0.14–2.11)0.88< 0.05 ng/mLMedian CRP(N = 102)5.5 (1.45–9.8) (N = 27)0.8 (0-2.22) (N = 75) < 0.0001 0.1-1.0 mg/dLMedian ESR(N = 27)25 (9.75-38) (N = 6)8 (8–12) (N = 21)0.160–20 mm/hrMedian Initial Leukocyte Count(N = 114)19.8 (11.55–30.45) (N = 31)19.7 (14.05–28.35) (N = 83)0.625.00-14.5 (Thousand/Mm3)Positive Blood Culture (%) (N = 57)2 (10.5%) (N = 19)0 (N = 38)0.1071-Positive Urine Culture (%) (N = 45)4 (28.6%) (N = 14)2 (6.45%) (N = 31)0.065-Pneumonia On Chest X-ray (%) (N = 55)4 (30.8%) (N = 13)2 (4.76%) (N = 42) 0.0232 -Positive Respiratory Viral Panel (%) (N = 90)8 (34.78%) (N = 23)25 (37.3%) (N = 67)1.0000-Antibiotics prior to procalcitonin level (%)8 (25%)15 (18%)0.44-Antibiotics prior to blood culture (%) (N = 57)8 (42%) (N = 19)15 (21%) (N = 38)0.12-Categorical variables compared using Fisher’s exact test, quantitative by Mann-Whitney U. Median values are followed by interquartile ranges. P values < 0.05 are in bold. If fewer than 32 patients in the treatment group or 84 in the non-treatment group had data available, the N for each group is specified Demographic and clinical data comparison Median CRP (N = 102) Median ESR (N = 27) Median Initial Leukocyte Count (N = 114) Categorical variables compared using Fisher’s exact test, quantitative by Mann-Whitney U. Median values are followed by interquartile ranges. P values < 0.05 are in bold. If fewer than 32 patients in the treatment group or 84 in the non-treatment group had data available, the N for each group is specified Statistical analysis was conducted using SAS 9.4. (SAS Institute, Cary, NC), except that receiver operator characteristic (ROC) curves were derived in SPSS 26.0 (IBM). Because many quantitative variables were not normally distributed, their distributions of quantitative variables were compared using non-parametric Mann-Whitney U tests. Categorical variables were compared using Fisher’s exact tests rather than chi-square tests because many cell sizes were small. No corrections were made to account for multiple hypothesis testing. Receiver operating characteristic (ROC) curves were generated from generalized linear models with logit link functions to evaluate procalcitonin’s utility in indicating clinically suspected infection. As the primary outcome, we then compared patients who had received a complete treatment course of antibiotics with those who had not; we did not include patients who received antibiotics prophylactically or in whom antibiotics were discontinued after an infection was ruled out. The prescription of a course of antibiotics was used as a proxy for a clinically diagnosed bacterial infection. Treatment courses of antibiotics were started while patients were in the hospital, and if not completed by the time of discharge, continued on an outpatient basis. The decision to continue a full course of antibiotics was invariably made by the primary endocrinology treating team. We also compared patients who did or did not have a documented infectious source (blood culture, urine culture, or pneumonia on chest x-ray), among those patients who had at least one of these three tests performed. Blood and urine cultures were considered positive if any bacteria other than those deemed skin or urogenital flora grew, respectively. Urine culture was considered positive based on growth of bacteria, regardless of clinical symptoms versus asymptomatic bacteriuria. Pneumonia on chest x-ray was based on radiologists’ reads of chest x-rays, not clinical symptoms or physical exam signs. Overlapping infections were not specified. In other words, encounters in the considered to have documented infection could have only 1, any 2, or all 3 of the above. Results We ascertained 348 patients with diabetes mellitus and a procalcitonin level checked during the study period (Fig. 1). 243 patients were excluded: 129 due to lack of type 1 or type 2 diabetes diagnosis (often patients had diabetes secondary to medications or pancreatitis) and 114 who had type 1 or 2 diabetes but were not in diabetic ketoacidosis during the encounter. Review of the remaining 105 patient charts revealed 116 unique encounters in which procalcitonin was checked; in 32 encounters, the patient was prescribed a complete antibiotics course, with indications listed in Fig. 1; Table 2. Six patients had 2 admissions each during the study period; 1 had 3 and 1 had 4 admissions. Only 2 patients with multiple admissions each had 1 completed course of antibiotics; thus, no patient in the study accounted for more than 1 completed course of antibiotics. 107 patient encounters occurred prior to, and 9 encounters during or after November 2023, when our laboratory changed the vendor for the procalcitonin assay (see Methods). Fig. 1Flowchart of recruitment process for cohorts Flowchart of recruitment process for cohorts Table 2Infections treated and corresponding antibioticsInfection (n)Antibiotics given (n)Skin/soft tissue infections (8)Clindamycin (1)Clindamycin, narrowed to amoxicillin (1)Clindamycin, narrowed to cephalexin (1)Clindamycin, vancomycin, and cefepime, narrowed to clindamycin (1)Ampicillin/sulbactam and clindamycin narrowed to amoxicillin/clavulanic acid (1)Ampicillin/sulbactam and cefepime narrowed to cephalexin (1)Ampicillin/sulbactam, narrowed to clindamycin (1)Cefepime, vancomycin, and metronidazole, narrowed to cefuroxime and clindamycin (1)UTI/pyelonephritis (7)Ceftriaxone, narrowed to cephalexin (3)Cephalexin (3)Amoxicillin (1)Streptococcus pharyngitis (3)Amoxicillin (2)Amoxicillin/clavulanic acid (1)Acute otitis media (2)Ceftriaxone (1)Ceftriaxone narrowed to amoxicillin/clavulanic acid (1)Acute otitis externa (2)Cefepime and ciprofloxacin, narrowed to amoxicillin/clavulanic acid (1)Ceftriaxone and ciprofloxacin (1)Necrotizing pancreatitis (2)Piperacillin/tazobactam (1)Ceftriaxone and metronidazole (1)Intra-abdominal infection (2)Ceftriaxone and metronidazole (1)Piperacillin/tazobactam (1)Community acquired pneumonia (2)Amoxicillin/clavulanic acid, narrowed to amoxicillin (1)Ceftriaxone (1)Animal bite (1)Amoxicillin/clavulanic acidBacteremia (1)Ceftriaxone, narrowed to cefuroximeCulture negative sepsis (1)Vancomycin and ceftriaxone (1)Skin/soft tissue infection and acute otitis media (1)Ampicillin/sulbactam, narrowed to amoxicillin/clavulanic acid (1) Table 2Infections treated and corresponding antibioticsInfection (n)Antibiotics given (n)Skin/soft tissue infections (8)Clindamycin (1)Clindamycin, narrowed to amoxicillin (1)Clindamycin, narrowed to cephalexin (1)Clindamycin, vancomycin, and cefepime, narrowed to clindamycin (1)Ampicillin/sulbactam and clindamycin narrowed to amoxicillin/clavulanic acid (1)Ampicillin/sulbactam and cefepime narrowed to cephalexin (1)Ampicillin/sulbactam, narrowed to clindamycin (1)Cefepime, vancomycin, and metronidazole, narrowed to cefuroxime and clindamycin (1)UTI/pyelonephritis (7)Ceftriaxone, narrowed to cephalexin (3)Cephalexin (3)Amoxicillin (1)Streptococcus pharyngitis (3)Amoxicillin (2)Amoxicillin/clavulanic acid (1)Acute otitis media (2)Ceftriaxone (1)Ceftriaxone narrowed to amoxicillin/clavulanic acid (1)Acute otitis externa (2)Cefepime and ciprofloxacin, narrowed to amoxicillin/clavulanic acid (1)Ceftriaxone and ciprofloxacin (1)Necrotizing pancreatitis (2)Piperacillin/tazobactam (1)Ceftriaxone and metronidazole (1)Intra-abdominal infection (2)Ceftriaxone and metronidazole (1)Piperacillin/tazobactam (1)Community acquired pneumonia (2)Amoxicillin/clavulanic acid, narrowed to amoxicillin (1)Ceftriaxone (1)Animal bite (1)Amoxicillin/clavulanic acidBacteremia (1)Ceftriaxone, narrowed to cefuroximeCulture negative sepsis (1)Vancomycin and ceftriaxone (1)Skin/soft tissue infection and acute otitis media (1)Ampicillin/sulbactam, narrowed to amoxicillin/clavulanic acid (1) Infections treated and corresponding antibiotics Clindamycin (1) Clindamycin, narrowed to amoxicillin (1) Clindamycin, narrowed to cephalexin (1) Clindamycin, vancomycin, and cefepime, narrowed to clindamycin (1) Ampicillin/sulbactam and clindamycin narrowed to amoxicillin/clavulanic acid (1) Ampicillin/sulbactam and cefepime narrowed to cephalexin (1) Ampicillin/sulbactam, narrowed to clindamycin (1) Cefepime, vancomycin, and metronidazole, narrowed to cefuroxime and clindamycin (1) Ceftriaxone, narrowed to cephalexin (3) Cephalexin (3) Amoxicillin (1) Amoxicillin (2) Amoxicillin/clavulanic acid (1) Ceftriaxone (1) Ceftriaxone narrowed to amoxicillin/clavulanic acid (1) Cefepime and ciprofloxacin, narrowed to amoxicillin/clavulanic acid (1) Ceftriaxone and ciprofloxacin (1) Piperacillin/tazobactam (1) Ceftriaxone and metronidazole (1) Ceftriaxone and metronidazole (1) Piperacillin/tazobactam (1) Amoxicillin/clavulanic acid, narrowed to amoxicillin (1) Ceftriaxone (1) Our standard protocol is to start a saline fluid bolus as soon as intravenous access is established and an initial blood sample obtained; an intravenous infusion is initiated as soon as the diagnosis of DKA is established, and the initial fluid bolus is completed. A median (interquartile range) of 1 (0–7) hours elapsed between the initial blood sample and the sample analyzed for inflammatory markers; thus, most samples for inflammatory markers were obtained before much treatment of DKA had occurred. Fifty-two samples were obtained in the emergency department, 55 in an intensive care unit, and 9 on a regular hospital floor. There is no institutional policy guiding collection of inflammatory markers; which markers and when to collect are determined based on clinician judgement. Demographic and clinical data were compared for the two groups (those prescribed a full treatment course of antibiotics or not) in Table 1. Of 116 encounters for patients in DKA with a procalcitonin checked, 102 also had CRP and 27 had erythrocyte sedimentation rate (ESR) checked during the admission; 57 encounters had blood culture results available, 45 had urine cultures, and 55 had chest x-rays. Only 17 of the 116 reviewed encounters had data available for all three of: blood culture, urine culture, and chest x-ray. However, 90 of the 116 had at least 1 of these results available. The antibiotics and no-antibiotics groups differed for gender, febrile status, CRP, and pneumonia on chest x-ray (Table 1). There were no other significant differences between groups, including in procalcitonin levels (Fig. 2). In terms of potential confounders, there were no statistically significant differences between the treated and not-treated group in terms of how many patients had received antibiotics prior to arrival (either at a transferring emergency department or in a prior ambulatory course) or had received antibiotics prior to the procalcitonin level or blood cultures being drawn. Table 2 lists the particular antibiotics given, grouped by infection treated. Since most infection types had fewer than 5 cases each, subgroup analysis was not performed. Fig. 2Distribution of procalcitonin levels in patients treated, or not, with a complete course of antibiotics. Boxes represent the interquartile range (IQR) of the data; medians are 0.4 for both groups and the median lines cannot be resolved from the lower margins of the boxes. Whiskers represent the upper bounds of 1.5 times the IQR. Outliers are graphed individually Distribution of procalcitonin levels in patients treated, or not, with a complete course of antibiotics. Boxes represent the interquartile range (IQR) of the data; medians are 0.4 for both groups and the median lines cannot be resolved from the lower margins of the boxes. Whiskers represent the upper bounds of 1.5 times the IQR. Outliers are graphed individually In the group of patients who did not receive a treatment course of antibiotics, 2 had positive urine cultures. Both cases were adolescent-aged females whose urine culture grew Group B streptococcus, were asymptomatic, and thus diagnosed with asymptomatic bacteriuria not requiring treatment. Also, in the group that did not receive a full antibiotics course, 2 patients had chest x-rays with findings consistent with possible developing pneumonia on radiologist overreading. In both cases, the treating team, based on exam and clinical course, did not have concern for bacterial pneumonia. An empirical, distribution-free receiver operating characteristic (ROC) curve for procalcitonin and treatment antibiotics (the primary outcome) was generated (Fig. 3). The area under the curve (AUC) was 0.51 (95% confidence interval, 0.39–0.63), signifying that this test was no better than chance at discriminating clinically suspected infection. Therefore, no optimum cutoff value was calculated for procalcitonin’s sensitivity and specificity in determining if a patient would receive treatment antibiotics. Another ROC curve was generated from a model that included fever and C-reactive protein level (CRP), which resulted in an AUC of 0.79 (0.68–0.90), reflecting moderate discrimination. Age was not a significant contributor to the model (not shown). Of note, CRP and procalcitonin levels were not correlated (rho = 0.093, p = 0.35). To ensure that the 8 patients with multiple admissions did not affect the results, the analysis was repeated including only the first admission for each of these patients, The resulting AUC and confidence intervals for each model were unchanged. Fig. 3A Receiver operating characteristics curve for procalcitonin while in DKA. B Receiver operating characteristics curve for presence of fever (temperature > 38 °C) and C-reactive protein A Receiver operating characteristics curve for procalcitonin while in DKA. B Receiver operating characteristics curve for presence of fever (temperature > 38 °C) and C-reactive protein In terms of comparison of objective tests for infection (blood culture, urine culture, and pneumonia on chest x-ray), 13 of the 116 reviewed encounters revealed positive blood cultures, urine cultures, or chest x-ray with pneumonia. There were an additional 24 encounters in which patients in DKA had inflammatory markers (CRP and/or ESR) checked, but not a procalcitonin level. Eleven of these occurred prior to May 6, 2020, when procalcitonin was not yet performed in on-site laboratories. When the 13 patients who would have met inclusion criteria other than missing a procalcitonin level were included in the logistic regression analysis of fever and CRP in predicting prescription of treatment antibiotics, there was no significant change in the AUC for the ROC curve (not shown). In an exploratory analysis, procalcitonin and initial blood glucose demonstrated a weak positive correlation (Spearman rho = 0.32, uncorrected p = 0.0004) (Fig. 4). There was no correlation between procalcitonin and initial pH or bicarbonate level. Fig. 4Exploratory analysis of correlation between procalcitonin and initial blood glucose (Spearman rho = 0.32, uncorrected p = 0.0004) Exploratory analysis of correlation between procalcitonin and initial blood glucose (Spearman rho = 0.32, uncorrected p = 0.0004) Discussion Based on the ROC curve for procalcitonin versus an antibiotic treatment course (a surrogate marker for bacterial infection), the procalcitonin level had no discrimination for clinically suspected infection in the setting of DKA. Whereas fever and an elevated CRP together were moderately good predictors of clinically suspected bacterial infection, incorporating procalcitonin into the model only minimally improved discrimination (not shown). Thus, the present study failed to identify any diagnostic utility for procalcitonin in determining the likelihood that a child in DKA has a clinically suspected bacterial infection. This is perhaps due to the fact that both DKA and bacterial infection trigger the systemic inflammatory response system and even the same cytokine release [5, 13]. Indeed, recent case reports describe patients with type 1 diabetes in DKA and markedly elevated procalcitonin and no identified infection [18, 19]. In one such case, procalcitonin rose to 15.88 ng/mL [18]. A case series of 5 adult patients also suggests that DKA in itself is associated with elevated procalcitonin levels [19]. Thus, procalcitonin seems to have a limited role in indicating infection in pediatric patients with DKA in the absence of other clinical features to suggest infection. This study included only those encounters in which clinical suspicion for an infection in a patient with DKA was high enough that a clinician ordered a procalcitonin level; there is no formal institutional policy for when (or which) inflammatory markers should be drawn in DKA patients. However, ESR was determined mostly (18/27; 67%) in patients admitted in 2020–2021, whereas only 41/116 (35%) of procalcitonin levels and 36/102 (also 35%) of CRP levels were from these years. This reflects the increasing recognition that ESR is less useful than CRP in diagnosing acute infections [20]. It is noteworthy that most (88/116, 76%) patients with procalcitonin levels had an ICU stay; in contrast, only 26% of DKA patients treated in our main Dallas campus require ICU care [21], suggesting that most of the patients who had procalcitonin levels determined appeared unusually ill. There is unlikely to be any bias in which of these ill-appearing patients had a procalcitonin levels determined, given that there were only 13 similar encounters who would have met the inclusion criteria if they had otherwise had a procalcitonin level checked (that is, clinical suspicion was high enough to check other inflammatory markers but procalcitonin was not included). As noted, including these 13 patients did not change the results of the ROC analysis . Limitations of this study include its retrospective nature. Also, prescription of a full course of antibiotics was used as a proxy for the presence of bacterial infection, since clinical rather than objective criteria were used to diagnose bacterial infections in most of the included pediatric cases. Thus, to allow for more robust statistical analysis, clinical diagnosis as ascertained by prescription of a full treatment course of antibiotics as a proxy outcome for bacterial infection was used as the primary outcome. However, a secondary analysis of the 90 encounters that included at least one of blood culture, urine culture, and chest x-ray, using documented infection as an outcome, yielded very similar results: procalcitonin was useless as a predictor (AUC 0.42 [95% CI 0.26–0.58]) whereas CRP plus a fever was a moderately good predictor (AUC 0.82 [0.64–0.99]). Additionally, patients with either type 1 or type 2 diabetes were included in the study, despite these two diseases processes having different pathophysiology. However, only a minority of patients had type 2 diabetes, and there was no significant difference in the percentage of type 1 versus type 2 diabetes between the treated and non-treated group (Table 1). Finally, 8 (25%) of procalcitonin levels in the treated group and 15 (18%) in the non-treated group were drawn after the patient had already received antibiotics, either as empiric/sepsis rule-out or as part of a preceding ambulatory treatment regimen (Table 1). While ideally the studied procalcitonin values would have all been drawn prior to any recent antibiotic exposures, prior data is mixed as to whether pre-treatment with antibiotics significantly impacts procalcitonin level during acute infection [22–25]. For example, whereas in one study of inflammatory markers in adults with community acquired pneumonia (CAP) reported lower procalcitonin values in those patients pre-treated with antibiotics than those not [22], a similar study of usefulness of procalcitonin level in CAP in adults showed no statistically significant difference in previous antibiotic therapy between patients with and without elevated procalcitonin levels [23]. A prospective study of adults with suspected infection or inflammatory disease of any etiology showed no difference between those pre-treated with antibiotics with regard to false negative and true positive procalcitonin results, though sensitivity in those pre-treated was lower24. The present study also found no significant difference in the proportions of procalcitonin samples collected before or after administration of antibiotics between the group of patients who received a treatment course of antibiotics and those who did not (Table 1). As previously noted, in an exploratory analysis outside our primary aim, a ROC curve was generated that included fever and C-reactive protein level (CRP), which resulted in an AUC of 0.79 (0.68–0.90), reflecting moderate discrimination. A retrospective study of 180 children similarly found CRP to be significantly more elevated in children in DKA with infection than those in DKA but without infection [26]. Future study of inflammatory markers in pediatric DKA might further explore the clinical utility of CRP and fever as predictors of bacterial infection in children with DKA, especially in the development of automated decision support tools.