Predictors of asymptomatic ketonemia in outpatients with diabetic kidney disease receiving SGLT-2 inhibitors: a cross-sectional study Authors: Veysel Baran Tomar, Taha Enes Cetin, Asil Demirezen, Omer Faruk Akcay, Ulver Derici, Galip Guz, Ozant Helvacı Published: 2026 ABSTRACT AbstractBackgroundSodium–glucose cotransporter-2 (SGLT-2) inhibitors are now foundational therapy for diabetic chronic kidney disease (CKD), yet their ketogenic metabolic effects in this population remain incompletely characterised. While euglycaemic diabetic ketoacidosis is a recognised complication, the prevalence and clinical determinants of subclinical asymptomatic ketonemia in patients with diabetic CKD receiving SGLT-2 inhibitors have not been systematically evaluated. We aimed to determine the prevalence of asymptomatic ketonemia, defined by point-of-care capillary β-hydroxybutyrate (β-HB) positivity, and to identify its independent clinical predictors in this population.MethodsThis cross-sectional study enrolled 211 consecutive adult outpatients with diabetic CKD receiving SGLT-2 inhibitor therapy (empagliflozin or dapagliflozin). Capillary β-HB was measured under standardised fasting conditions using a point-of-care ketone meter; positivity was defined as β-HB ≥ 0.6 mmol/L. Univariate comparisons were performed using the Mann–Whitney U test, independent-samples t-test, χ² test, or Fisher’s exact test as appropriate. Given the low number of outcome events (n = 16), Firth’s penalised logistic regression was employed as the primary multivariable analysis to mitigate small-sample bias.ResultsAsymptomatic ketonemia was detected in 16 of 211 patients (7.6%), all of whom were clinically asymptomatic at measurement. In Firth’s penalised logistic regression, heart failure (OR 4.46; 95% CI 1.48–13.51; p = 0.008), absence of statin use (OR 4.94; 95% CI 1.54–15.86; p = 0.007), and higher HbA1c per 1% increment (OR 1.35; 95% CI 1.01–1.81; p = 0.041) were independent predictors of ketone positivity. Systolic blood pressure showed a borderline inverse association (OR 0.97 per mmHg; 95% CI 0.93–1.00; p = 0.063). No ketone-positive patient was receiving thiazide diuretics (0% vs. 30.3%; p = 0.007). SGLT-2 inhibitor type was not independently associated with ketonemia. Findings were robust on sensitivity analysis with standard logistic regression (Hosmer–Lemeshow p = 0.217; Nagelkerke R²=0.234).ConclusionsAsymptomatic ketonemia is present in approximately one in thirteen outpatients with diabetic CKD receiving SGLT-2 inhibitors. Heart failure, absence of statin therapy, and suboptimal glycemic control independently identify a higher-risk metabolic phenotype, while ketonemia appears unrelated to CKD stage. Although these findings are exploratory and require confirmation in larger prospective studies, they suggest that targeted ketone monitoring may be warranted in selected higher-risk patients.Keywords: Asymptomatic ketonaemia, β-Hydroxybutyrate, Chronic kidney disease, Diabetes mellitus, Heart failure, SGLT-2 inhibitors Introduction The widespread adoption of sodium-glucose cotransporter-2 (SGLT-2) inhibitors as foundational therapy for chronic kidney disease (CKD) has substantially altered nephrology practice, with the 2024 KDIGO guidelines recommending these agents for all CKD patients with an estimated glomerular filtration rate (eGFR) ≥ 20 mL/min/1.73 m², regardless of diabetes status [1]. This is supported by accumulating evidence from landmark trials, including EMPA-KIDNEY and DAPA-CKD, which demonstrated significant cardiovascular and renal protective benefits across the spectrum of kidney function [2, 3]. However, SGLT-2 inhibitors induce predictable metabolic changes, most notably mild ketonemia through enhanced hepatic ketogenesis following glucosuria-mediated insulin suppression [4]. While major cardiovascular outcome trials enrolling over 90,000 participants have established the overall safety profile, with diabetic ketoacidosis (DKA) rates of only 0.16–0.76 per 1000 patient-years, concerns persist regarding the potential for ketone accumulation in patients with reduced kidney function [5–7]. The glycosuric mechanism of SGLT-2 inhibitors induces a state of relative caloric deprivation, promoting a metabolic shift from glucose to fatty acid oxidation and consequently increasing hepatic ketogenesis [4]. This ketogenic propensity has been implicated in the pathophysiology of euglycemic diabetic ketoacidosis (eDKA), an uncommon but serious adverse effect [6]. However, between normal ketone physiology and overt ketoacidosis lies a spectrum of subclinical ketone elevation—asymptomatic ketonemia—whose clinical significance and prevalence remain insufficiently characterized. Theoretical considerations have proposed that impaired renal ketone clearance in advanced CKD could predispose to higher steady-state ketonemia levels, potentially increasing DKA risk [8]. Contrary to these predictions, Ferrannini et al. demonstrated that renal handling of ketones is maintained through the activity of transport proteins such as SMCT1 (SLC5A8) and MCT transporters, supporting the concept that compensatory mechanisms prevent excessive ketone accumulation despite reduced kidney function [9]. Patients with diabetic CKD represent a population of particular interest. Reduced renal ketone clearance, the catabolic milieu of uremia, frequent comorbidities such as heart failure, and polypharmacy may collectively modify ketone metabolism in ways that differ from the general diabetic population [8]. Beyond SGLT-2 inhibitor use, other clinical factors—including the catabolic state of heart failure, insulin deficiency reflected by poor glycemic control, and potentially altered lipid metabolism—may independently modulate ketone production in the diabetic CKD population; yet the relative contribution of these factors has not been systematically evaluated. Most data on SGLT-2 inhibitor-associated ketosis derive from trials that largely excluded patients with advanced CKD, and current consensus documents generally recommend ketone monitoring in the presence of precipitating factors and emphasize “sick day” management rather than CKD stage–specific surveillance [1, 10, 11]. These recommendations are largely based on safety signals from large cardiovascular and renal outcome trials rather than on direct measurements of steady-state ketonemia across CKD stages in routine practice [5–7, 10, 11]. Point-of-care capillary β-hydroxybutyrate (β-HB) measurement provides a practical and reliable method for assessing ketone levels in clinical practice. In the context of SGLT-2 inhibitor therapy, capillary β-HB is the recommended monitoring modality, with established diagnostic thresholds supporting its use for the detection of clinically significant ketonemia in outpatient settings [12–14]. Unlike urine ketone testing, which measures acetoacetate and may underestimate true ketosis in the setting of SGLT-2 inhibitor-altered renal handling, capillary β-HB directly reflects the predominant circulating ketone body. The aim of this study was to determine the prevalence of asymptomatic ketonemia detected by point-of-care capillary β-HB testing in outpatients with diabetic CKD receiving SGLT-2 inhibitors, and to identify independent clinical predictors associated with this condition. Materials and methods Study design and setting This single-center, cross-sectional study was conducted at the Nephrology Department of a tertiary care university hospital between September 2025 and January 2026. The study protocol was approved by the Institutional Clinical Research Ethics Committee on August 1, 2025 (Approval No: 2025 − 139) and conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants prior to enrollment. Participants Consecutive adult patients (≥ 18 years) with a diagnosis of diabetic CKD, defined as a history of diabetes mellitus with persistent albuminuria > 30 mg/g and/or an eGFR < 60 mL/min/1.73 m² (CKD-EPI equation), who were receiving SGLT-2 inhibitor therapy (empagliflozin or dapagliflozin) were enrolled. Patients were required to be clinically stable at the time of assessment, with no acute illness, hospitalization, or medication changes within the preceding 4 weeks. Exclusion criteria included type 1 diabetes mellitus, active infection, active malignancy requiring ongoing systemic therapy, pregnancy, advanced liver disease, failure to comply with a ≥ 8-hour fasting period prior to ketone measurement, active diabetic ketoacidosis or screening ketonemia > 1.5 mmol/L, recent acute illness or surgery within 4 weeks, current alcohol abuse or eating disorders, inability to provide informed consent, and adherence to a ketogenic diet or major dietary modifications within the previous 3 months. Data collection and measurements Demographic and clinical data were collected from electronic medical records at the time of enrollment. Recorded variables included age, sex, body mass index (BMI), systolic and diastolic blood pressure, comorbidities (hypertension, coronary artery disease, heart failure) and current medications (renin–angiotensin–aldosterone system [RAAS] blockers, calcium channel blockers, thiazide diuretics, beta-blockers, mineralocorticoid receptor antagonists [MRA], loop diuretics, statins, metformin, insulin). Laboratory parameters obtained from the most recent clinical assessment (within 3 months) included HbA1c, fasting plasma glucose, serum creatinine, eGFR, serum albumin, lipid profile (HDL-cholesterol, LDL-cholesterol, triglycerides), C-reactive protein, complete blood count (hemoglobin, platelets, white blood cell count, lymphocyte count), and liver enzymes (AST, ALT). CKD staging was performed according to the KDIGO 2024 classification based on eGFR categories [1]. Outcome measurement The primary outcome was capillary ketone positivity, defined as a point-of-care capillary β-HB concentration of ≥ 0.6 mmol/L obtained from a fingertip blood sample under standardized fasting conditions (minimum 8-hour overnight fast confirmed prior to measurement). This threshold is consistent with the established categorization of blood β-HB in the context of SGLT-2 inhibitor therapy, in which values below 0.6 mmol/L are considered within the normal physiological range, as defined by the International Consensus on Risk Management of Diabetic Ketoacidosis in Patients Treated with Sodium–Glucose Cotransporter Inhibitors [14]. All measurements were performed using the On Call® GK Dual Blood Glucose & Ketone Monitoring System (ACON Laboratories), a CE-marked electrochemical point-of-care analyzer that measures whole blood β-HB and reports plasma-equivalent concentrations. Measurements were conducted exclusively during morning outpatient visits between 08:00 and 10:00 to minimize diurnal variability. Each measurement was performed in duplicate and the mean value was used for analysis. Analytical quality control included regular device calibration using the manufacturer-supplied control solution prior to each new strip lot, and data integrity was assured through independent auditing of a randomly selected 10% sample of entries by personnel blinded to group assignment. Statistical analysis Continuous variables were assessed for normality using the Shapiro–Wilk test and presented as median (interquartile range [IQR]) or mean ± standard deviation (SD) accordingly. Categorical variables were presented as n (%). Univariate comparisons between ketone-positive and ketone-negative groups were performed using the Mann–Whitney U test or independent-samples t-test for continuous variables, and χ² or Fisher’s exact test for categorical variables. For multivariable analysis, variables with p < 0.10 in univariate analysis were considered as candidates. Given the low number of outcome events (n = 16) relative to candidate predictors, Firth’s penalized logistic regression was employed as the primary analysis to mitigate small-sample bias inherent in standard maximum likelihood estimation [15, 16]. Standard logistic regression was performed in parallel as a sensitivity analysis, with model fit assessed by the Hosmer–Lemeshow test, omnibus χ² test, and Nagelkerke R². Although Firth’s penalized logistic regression can reduce bias related to separation, variables exhibiting complete separation may still yield unstable or poorly interpretable estimates in very small event datasets. Therefore, such variables were excluded from the final multivariable model and reported descriptively. Thiazide diuretic use was excluded owing to complete separation (none of the 16 ketone-positive patients used thiazides). Insulin use was excluded from the final multivariable model due to collinearity with HbA1c, confirmed by Spearman correlation analysis (rho = 0.233, p = 0.001). HbA1c was retained in preference to insulin use as it is a continuous variable that more directly captures the degree of glycemic dysregulation on a gradient, and has a more immediate mechanistic link to lipolysis and hepatic ketogenesis. Although triglyceride levels met the univariate threshold (p = 0.093), they were excluded from the final multivariable model due to the severely constrained events-per-variable (EPV) ratio (n = 16 events); inclusion of all threshold-meeting variables would have resulted in an EPV well below the recommended minimum of 10, risking overfitting and unreliable effect estimates. A two-sided p < 0.05 was considered statistically significant. Analyses were performed using IBM SPSS Statistics for Windows, version 22.0 (IBM Corp., Armonk, NY, USA) and R (R Foundation for Statistical Computing, Vienna, Austria) with the ‘logistf’ package for penalized regression. Results A total of 211 patients were included in the analysis. The median age was 67 years (IQR 61–74), and 55.9% were female. The median eGFR was 49 mL/min/1.73 m² (IQR 35–69), and 18.0% of patients had CKD stage 4. All patients were receiving SGLT-2 inhibitor therapy at the time of assessment; empagliflozin was used in 50.2% of the cohort and dapagliflozin in the remainder. Median HbA1c was 7.4% (IQR 6.6–8.4). Comorbid hypertension was present in 92.0%, coronary artery disease in 43.1%, and heart failure in 26.1%. Capillary ketone positivity was detected in 16 of 211 patients, yielding a prevalence of 7.6%. All ketone-positive patients were asymptomatic at the time of measurement. Among these 16 patients, the median capillary β-HB concentration was 0.70 mmol/L (IQR 0.60–1.00; range 0.60–1.50 mmol/L); no patient exceeded the 1.5 mmol/L exclusion threshold applied at enrollment, confirming that all detected cases represented subclinical ketonemia and none met criteria for impending or overt ketoacidosis. Baseline characteristics stratified by capillary ketone status are presented in Tables 1 and 2. Ketone-positive patients had significantly lower median systolic blood pressure (122.5 vs. 130.0 mmHg; p = 0.035) and higher median HbA1c (7.9% vs. 7.4%; p = 0.040) compared with ketone-negative patients. Fasting glucose showed a trend toward higher values in the ketone-positive group (157.0 vs. 128.0 mg/dL; p = 0.054). Table 1Demographics and clinical parameters according to capillary ketone statusVariableKetone (−) (n = 195)Ketone (+) (n = 16)p-valueAge, years67.0 (61.0–74.0)65.0 (62.2–70.0)0.700Gender (female), n (%)108 (55.4)10 (62.5)0.772BMI (kg/m²)28.4 (26.5–33.1)28.1 (23.6–32.1)0.575SBP (mmHg)130.0 (120.0–140.0)122.5 (120.0–130.0) 0.035 DBP (mmHg)80.0 (70.0–80.0)77.5 (68.8–80.0)0.338HbA1c (%)7.4 (6.6–8.4)7.9 (7.5–8.8) 0.040 Fasting glucose (mg/dl)128.0 (100.0–163.5)157.0 (124.2–190.2)0.054Creatinine (mg/dl)1.3 (1.0–1.8)1.1 (1.0–1.4)0.306eGFR (mL/min/1.73 m²)48.0 (35.0–69.0)60.0 (45.8–70.0)0.328Albumin (g/dl)4.4 (4.2–4.6)4.5 (4.2–4.6)0.733HDL-cholesterol (mg/dl)45.0 ± 12.242.4 ± 15.70.427LDL-cholesterol (mg/dl)94.0 (73.5–110.5)83.5 (73.0–124.0)0.913Triglycerides (mg/dl)153.0 (109.5–220.0)129.0 (93.2–150.2) 0.093 CRP (mg/l)3.0 (3.0–8.3)3.0 (3.0–10.3)0.678Hemoglobin (g/dl)13.2 (12.3–14.4)12.7 (11.5–14.5)0.605Platelets (×10³/µl)250.0 (216.5–300.0)285.0 (240.8–334.2)0.143WBC (×10³/µl)8.2 (6.7–9.7)7.9 (6.2–8.4)0.288Lymphocytes (×10³/µl)2.0 (1.6–2.6)1.8 (1.4–2.2)0.090AST (U/L)20.0 (16.5–24.0)19.0 (17.2–22.8)0.734ALT (U/L)17.0 (14.0–24.0)17.0 (13.5–20.8)0.747Data are presented as median (interquartile range) or mean ± standard deviation, as appropriate. Mann–Whitney U test or independent-samples t-test was used for continuous variablesALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; CRP, C-reactive protein; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; SBP, systolic blood pressure; WBC, white blood cell count Demographics and clinical parameters according to capillary ketone status Data are presented as median (interquartile range) or mean ± standard deviation, as appropriate. Mann–Whitney U test or independent-samples t-test was used for continuous variables ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; CRP, C-reactive protein; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; SBP, systolic blood pressure; WBC, white blood cell count Table 2Comorbidities and medications according to capillary ketone statusVariableKetone (−) (n = 195)Ketone (+) (n = 16)p-valueHypertension180 (92.3)14 (87.5)0.624Coronary artery disease83 (42.6)8 (50.0)0.753Heart failure47 (24.1)8 (50.0) 0.035 eGFR > 45 mL/min/1.73 m²109 (55.9)12 (75.0)0.222CKD stage 436 (18.5)2 (12.5)0.742Hypervolemia48 (24.6)6 (37.5)0.248RAAS blocker154 (79.0)11 (68.8)0.350Calcium channel blocker106 (54.4)9 (56.2)1.000Thiazide diuretic59 (30.3)0 (0.0) 0.007 Beta-blocker107 (54.9)8 (50.0)0.908MRA23 (11.8)2 (12.5)1.000Loop diuretic56 (28.7)4 (25.0)1.000Statin115 (59.0)4 (25.0) 0.018 Metformin78 (40.0)4 (25.0)0.359Insulin82 (42.1)12 (75.0) 0.022 Data are presented as n (%). Chi-square test or Fisher’s exact test was used as appropriateCKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; MRA, mineralocorticoid receptor antagonist; RAAS, renin–angiotensin–aldosterone system Comorbidities and medications according to capillary ketone status Data are presented as n (%). Chi-square test or Fisher’s exact test was used as appropriate CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; MRA, mineralocorticoid receptor antagonist; RAAS, renin–angiotensin–aldosterone system Regarding comorbidities and medications, ketone-positive patients had significantly higher prevalence of heart failure (50.0% vs. 24.1%; p = 0.035), higher insulin use (75.0% vs. 42.1%; p = 0.022), lower statin use (25.0% vs. 59.0%; p = 0.018), and complete absence of thiazide diuretic use (0% vs. 30.3%; p = 0.007). No significant differences were observed in age, BMI, eGFR, serum albumin, lipid parameters, inflammatory markers, or use of RAAS blockers, calcium channel blockers, beta-blockers, loop diuretics, or metformin. Seven variables met the univariate threshold (p < 0.10) for inclusion in multivariable modeling: systolic blood pressure, HbA1c, heart failure, thiazide diuretic use, statin use, insulin use and triglycerides. Thiazide diuretic use was excluded owing to complete separation, and insulin use was excluded to avoid collinearity with HbA1c, as detailed in the statistical analysis section. Triglycerides met the univariate threshold but were excluded from the final model owing to EPV constraints; retaining all threshold-meeting variables would have risked overfitting given the limited number of outcome events (n = 16). The final model included four variables: heart failure, statin use, HbA1c, and systolic blood pressure. Firth’s penalized logistic regression identified three independent predictors of capillary ketone positivity (Table 3). Heart failure was a strong predictor (OR 4.46; 95% CI 1.48–13.51; p = 0.008), followed by absence of statin use (OR 4.94; 95% CI 1.54–15.86; p = 0.007) and higher HbA1c (OR 1.35 per 1% increase; 95% CI 1.01–1.81; p = 0.041). Systolic blood pressure demonstrated a borderline inverse association (OR 0.97 per mmHg; 95% CI 0.93–1.00; p = 0.063). Standard maximum likelihood logistic regression produced consistent results. Model fit was adequate: omnibus χ² = 21.58 (p < 0.001), Hosmer–Lemeshow p = 0.217, and Nagelkerke R² = 0.234. Table 3Firth’s penalized logistic regression for predictors of capillary ketone positivityVariableBSEp-valueOR (95% CI)Heart failure1.4960.562 0.008 4.46 (1.48–13.51)No statin use1.5980.595 0.007 4.94 (1.54–15.86)Systolic blood pressure (per mmHg)−0.0350.0190.0630.97 (0.93–1.00)HbA1c (per %)0.3030.148 0.041 1.35 (1.01–1.81)Firth’s penalized likelihood method was applied as the primary multivariable analysis owing to the low number of outcome events (n = 16). Standard maximum likelihood logistic regression was performed as a sensitivity analysis and produced consistent results. Model fit (standard logistic regression): omnibus χ² = 21.579 (p < 0.001); Hosmer–Lemeshow p = 0.217; Nagelkerke R² = 0.234; classification accuracy = 92.4%. Variables were coded as binary (1 = presence/absence as indicated)B, regression coefficient; CI, confidence interval; OR, odds ratio; SE, standard error Firth’s penalized logistic regression for predictors of capillary ketone positivity Firth’s penalized likelihood method was applied as the primary multivariable analysis owing to the low number of outcome events (n = 16). Standard maximum likelihood logistic regression was performed as a sensitivity analysis and produced consistent results. Model fit (standard logistic regression): omnibus χ² = 21.579 (p < 0.001); Hosmer–Lemeshow p = 0.217; Nagelkerke R² = 0.234; classification accuracy = 92.4%. Variables were coded as binary (1 = presence/absence as indicated) B, regression coefficient; CI, confidence interval; OR, odds ratio; SE, standard error Discussion In this cross-sectional study of 211 outpatients with diabetic CKD receiving SGLT-2 inhibitors, asymptomatic ketonemia detected by point-of-care capillary β-HB measurement was observed in 7.6% of patients. Heart failure, absence of statin therapy, and higher HbA1c were independently associated with ketone positivity. These findings provide a clinically relevant framework for identifying higher-risk phenotypes among SGLT-2 inhibitor–treated patients with diabetic CKD. Although ketonemia was clinically silent at the time of assessment, it may reflect an underlying state of metabolic susceptibility rather than a benign biochemical fluctuation, particularly in individuals with concurrent heart failure or suboptimal glycemic control. Whether this phenotype represents a precursor to euglycemic DKA or an adaptive metabolic shift remains uncertain and warrants prospective evaluation. Notably, ketonemia was not associated with eGFR or CKD stage in our stable outpatient cohort, suggesting that reduced renal clearance alone may not be the primary determinant of steady-state β-HB levels. Although total renal ketone clearance declines proportionally with GFR, adaptive mechanisms appear to preserve ketone homeostasis [9]. Experimental studies indicate that SGLT2 inhibitor–related elevations in β-HB are largely driven by increased hepatic ketone production rather than impaired renal elimination [4]. In addition, extra-renal metabolism, particularly peripheral tissue utilization, constitutes the major pathway of ketone disposal, thereby limiting excessive accumulation even in advanced CKD [8]. Collectively, these mechanisms provide a physiological explanation for the stable ketonemia observed across CKD stages in our cohort. The 7.6% prevalence of asymptomatic ketonemia in our diabetic CKD cohort lies at the higher end of estimates reported in broader diabetic populations receiving SGLT-2 inhibitors. Importantly, all cases were clinically silent, underscoring that routine clinical assessment alone may fail to detect subclinical biochemical ketosis. Large cardiovascular and renal outcome trials, including CREDENCE, DAPA-CKD, and EMPA-KIDNEY, have consistently reported very low absolute rates of diabetic ketoacidosis across a wide spectrum of kidney function [2, 3, 17]. In EMPA-KIDNEY, which included patients with eGFR down to 20 mL/min/1.73 m², only one case of ketoacidosis was observed among 3304 empagliflozin-treated participants [3]. Importantly, these trials did not perform systematic capillary β-HB measurements; DKA events were captured only upon symptomatic presentation, and subclinical ketonemia would therefore have remained undetected regardless of its true prevalence. Taken together, these data suggest that while mild ketonemia may be relatively frequent at a biochemical level, progression to overt ketoacidosis remains rare in stable CKD populations. Heart failure emerged as the strongest independent predictor, conferring an approximately 4.5-fold increase in the odds of ketonemia. This association is biologically coherent and mechanistically multifactorial. First, heart failure is itself a ketosis-prone state; neurohormonal activation characterized by elevated catecholamines and glucagon suppresses insulin signaling, activates hormone-sensitive lipase, and provides augmented free fatty acid supply for hepatic ketogenesis independent of SGLT-2 inhibitor use [4, 18]. Second, experimental and human studies have demonstrated markedly increased myocardial ketone utilization in advanced heart failure, reflecting a metabolic substrate switch toward ketone bodies as an alternative energy source [19, 20]. In this context, SGLT-2 inhibitor–induced ketonemia in heart failure patients likely amplifies a pre-existing ketogenic drive rather than initiating a de novo metabolic shift. The “thrifty substrate hypothesis” proposes that this convergence may enhance cardiac energetic efficiency [4], though whether it represents adaptive physiology or incremental metabolic risk warrants prospective clarification. The nearly fivefold increased odds of ketonemia among patients not receiving statins represents a novel observation. Inhibition of HMG-CoA reductase may theoretically influence hepatic ketone body synthesis pathways, given shared upstream lipid metabolic intermediates [21]. In addition, statins exert anti-inflammatory and insulin-sensitizing effects that could indirectly attenuate the ketogenic drive. However, clinical evidence supporting a direct anti-ketogenic effect of statins is limited. Therefore, this association should be interpreted cautiously and considered hypothesis-generating. Furthermore, the inverse association between statin use and ketonemia observed in this study is susceptible to confounding by indication; patients not receiving statins may represent a distinct metabolic and cardiovascular risk phenotype, and the observed association may therefore reflect differences in underlying comorbidity burden rather than a direct anti-ketogenic pharmacological effect of statins. Higher HbA1c independently predicted ketonemia, with each 1% increase conferring 35% higher odds. This finding is consistent with established pathophysiology, whereby poor glycemic control reflects relative insulin deficiency, leading to disinhibition of hormone-sensitive lipase, enhanced lipolysis, and subsequent ketogenesis [8]. In addition, greater glycosuria-induced caloric loss under SGLT-2 inhibition may further accentuate the metabolic shift toward fatty acid oxidation in individuals with suboptimal glycemic control. Notably, 75% of ketone-positive patients in our cohort were receiving insulin therapy, suggesting a more advanced metabolic phenotype in this subgroup. An additional observation was the complete absence of thiazide diuretic use among ketone-positive patients (0% vs. 30.3%; p = 0.007). Although this variable could not be included in multivariable modeling due to complete separation, the strength of the univariate association merits consideration. One possible explanation is confounding by indication: thiazide use in this population may reflect a less advanced cardiorenal phenotype, whereas patients with more severe heart failure or volume overload are more likely to receive loop diuretics. Given the limited number of outcome events and the complete separation precluding multivariable adjustment, this finding warrants cautious interpretation. Current consensus statements emphasize patient education and recognition of precipitating factors rather than routine universal ketone monitoring in individuals treated with SGLT-2 inhibitors [11]. The ADA/EASD 2022 Consensus Report underscores the importance of structured sick-day management to mitigate the risk of ketoacidosis in this population [11]. Real-world data further indicate that euglycemic DKA events are uncommon and are typically triggered by transient stressors such as infection, surgery, insulin omission, or dehydration [22, 23]. In this context, our findings suggest that a targeted approach to capillary ketone testing may be reasonable in higher-risk phenotypes—such as patients with heart failure or suboptimal glycemic control—rather than systematic screening of all stable outpatients. Several limitations should be acknowledged. First, the cross-sectional design precludes causal inference and does not allow assessment of temporal relationships between ketonemia and clinical outcomes. Second, the relatively low number of events (n = 16) limits multivariable modeling and results in wide confidence intervals around effect estimates. Although Firth’s penalized logistic regression reduces small-sample bias in low event-per-variable settings, the findings should be interpreted as exploratory. Capillary β-HB was measured at a single time point, precluding assessment of intra-individual variability or persistence of ketonemia; future longitudinal studies with repeated measurements are needed to determine whether the subclinical ketonemic phenotype identified here is stable or transient. We did not measure serum insulin or C-peptide levels; endogenous insulin reserve is a key determinant of ketogenic risk in the context of SGLT-2 inhibitor therapy, and its absence precludes stratification by beta-cell insufficiency. Future prospective studies should incorporate C-peptide measurement to refine risk stratification beyond the clinical predictors identified here. Finally, as a single-center study, generalizability to broader and more diverse CKD populations may be limited. Future studies should prospectively clarify the clinical trajectory of asymptomatic ketonemia in patients with diabetic CKD receiving SGLT-2 inhibitors, particularly during periods of metabolic stress. Larger multicenter prospective studies with repeated β-HB measurements are needed to validate the predictors identified in this exploratory analysis and to determine whether asymptomatic ketonemia translates into clinically meaningful outcomes. Continuous ketone monitoring and data-driven risk stratification approaches may further refine individualized assessment in selected higher-risk populations [24, 25]. Importantly, the consistency of our findings with major outcome trials demonstrating a low absolute risk of ketoacidosis provides a reassuring safety context for SGLT-2 inhibitor use in stable CKD populations [26]. Conclusion Asymptomatic ketonemia, detected by point-of-care capillary β-HB measurement, is present in approximately one in thirteen outpatients with diabetic CKD receiving SGLT-2 inhibitors. Heart failure, absence of statin therapy, and suboptimal glycemic control independently identify a higher-risk metabolic phenotype, whereas ketonemia was not associated with CKD stage in stable outpatients. The absence of thiazide use among ketone-positive patients represents a hypothesis-generating observation. These findings support a risk-adapted approach to ketone monitoring during routine nephrology care. Prospective studies are warranted to clarify the long-term clinical implications of asymptomatic ketonemia in diabetic CKD. Acknowledgements Not applicable. Abbreviations Alanine aminotransferase Aspartate aminotransferase β-Hydroxybutyrate Body mass index Chronic kidney disease C-reactive protein Diabetic ketoacidosis Euglycemic diabetic ketoacidosis Estimated glomerular filtration rate Events per variable Interquartile range Mineralocorticoid receptor antagonist Odds ratio Renin–angiotensin–aldosterone system Standard deviation Sodium–glucose cotransporter-2 White blood cell count Author contributions VBT conceived and designed the study, collected data, performed the statistical analysis, and drafted the manuscript. TEC and AD contributed to data collection. OFA contributed to data interpretation. UD and GG supervised the study and critically revised the manuscript. OH supervised the study and gave final approval. All authors read and approved the final manuscript. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Data availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate All the authors contributed significantly to the research and the manuscript. The manuscript has been read and approved by all authors, and we confirm that each author meets the authorship criteria outlined in the journal’s guidelines. Furthermore, all the authors believe that the manuscript represents honest and original work. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.