Impact of tight glucose control on circulating 3-hydroxybutyrate in critically ill patients

This is a secondary analysis of 3 single-center RCTs performed in the adult surgical, adult medical (NCT00115479) and pediatric ICU (NCT00214916) in Leuven, Belgium, which demonstrated improved clinical outcome by tight glucose control in the ICU [16‐18]. The study was performed in accordance with the 1964 Declaration of Helsinki and later amendments, and written informed consent was obtained from the patient or next of kin, or (for children) from the parents or legal guardian. The detailed study protocols and the effect on the primary outcome have been published [16‐18]. In brief, patients were randomized upon ICU admission to target age-adjusted normal fasting blood glucose concentrations (50–80 mg/dL [2.8–4.4 mmol/L] for neonates and infants, 70–100 mg/dL [3.9–5.6 mmol/L] for children older than 1 year, 80–110 mg/dL [4.4–6.1 mmol/L] for adults) with insulin therapy (tight glucose control), or to liberal blood glucose control, whereby stress hyperglycemia was only treated when it exceeded 215 mg/dL (11.9 mmol/L). All patients received early parenteral nutrition to supplement insufficient or failing enteral nutrition as part of the contemporary standard of care. In all patients, blood samples were taken upon ICU admission, and thereafter daily at 6 ± 2 h am. In patients included in the adult medical ICU (N = 1200) and pediatric ICU study (N = 700), plasma and serum samples were collected. In the adult surgical ICU study (N = 1548), only serum samples were collected. After aliquoting, samples were stored at -80 °C until further analysis.

In this secondary analysis, we aimed to study whether enhanced ketogenesis could have mediated part of the outcome benefit of tight glucose control seen in the parent RCTs [16‐18]. Before studying such potential mediator role through multivariable regression analysis, we performed a time course analysis in 2 matched subsets of critically ill children and adults, to study whether tight glucose control significantly affected ketogenesis in these patients, and to identify a day of maximal effect (if any). To that purpose, we quantified daily circulating 3-hydroxybutyrate (3HB) concentrations from ICU admission until day 3 in the ICU. Patients were eligible if they required intensive care for least 3 days, did not have a history of diabetes mellitus and had available blood samples at each time point (upon admission and the 3 consecutive mornings). To rule out selection bias due to an impact of the intervention on ICU length of stay, and for feasibility reasons, the eligible pediatric (N = 274), adult medical (N = 537) and adult surgical (N = 655) patient cohorts were reduced with propensity score matching to obtain 2 matched cohorts of 100 adult and 100 pediatric patients (50 allocated to tight glucose control and 50 allocated to liberal glucose control, of whom (for the adult cohort) 40 from the medical ICU study, and 60 from the surgical ICU study, to obtain a proportional sample from the respective parent RCTs) (Fig. 1). The groups were matched for demographics (age and sex; weight of children, body mass index of adults), baseline type and severity of illness (Pediatric Logistic Organ Dysfunction (PELOD) score for children, Acute Physiology and Chronic Health Evaluation II (APACHE-II) score for adults), need of extracorporeal membrane oxygenation or mechanical hemodynamic support upon admission (for children) and history of malignancy (for adults).

Blood glucose was measured in whole blood by a blood gas analyzer (ABL700, Radiometer Medical, Copenhagen, Denmark). The concentration of 3HB was quantified with a laboratory assay based upon the oxidation of 3HB to acetoacetate by the enzyme 3HB dehydrogenase and the concomitant reduction of NAD⁺ to NADH, as previously described for plasma samples [9, 19]. Samples with undetectable 3HB concentrations were assigned the detection limit (0.04 mmol/L). In the pediatric cohort, 3HB was determined in plasma samples. As plasma was not available for the adult surgical ICU study, we first evaluated the validity of serum as a replacement for plasma by performing 3HB measurements in paired samples of adult patients included in the medical ICU study (40 patients with 4 paired samples, leading to 160 paired measurements). The mean difference of these paired measurements was 0.015 mmol/L (95% confidence interval ranging from -0.117 to 0.133 mmol/L). Since this confirmed validity of serum as replacement for plasma, 3HB was subsequently quantified in all serum samples of the adult cohort.

The matched cohort of critically ill children consisted of 50 patients randomized to tight glucose control and 50 patients randomized to liberal glucose control (Fig. 1). Baseline characteristics were similar between both groups (Table 1). Upon admission, blood glucose was comparable between both groups (P = 0.08), after which morning blood glucose was significantly lower in tight glucose control patients on all study days (all P < 0.0001) (Fig. 2a), in the presence of higher insulin doses (all P < 0.0001) (Fig. 2b) and similar intake through enteral and parenteral nutrition (Fig. 2c). From admission onward, plasma 3HB concentrations were similarly low in both randomization groups (Fig. 2d), with a significant decrease in concentrations between ICU admission and day 1 (P < 0.0001). Of all 400 3HB determinations, concomitant hypoglycemia (≤ 40 mg/dL [≤ 2.2 mmol/L]) on the corresponding blood glucose measurement was present at 5 occasions (1 blood glucose measurement upon randomization in each randomization group, thereafter 1 measurement at each time point in the tight glucose control group).

The matched cohort of critically ill adults consisted of 100 patients–50 patients per randomization group–of whom 60 were admitted to the surgical ICU and 40 to the medical ICU (Fig. 1). Baseline characteristics were similar (Table 2, Additional file 1: Table S1). As was the case in the total adult study population [20], ICU mortality was significantly lower in tight glucose control patients (P = 0.01). Blood glucose concentrations were not different upon ICU admission (P = 0.6) and were significantly lower in tight glucose control patients throughout the first 3 days in ICU (all P < 0.0001) (Fig. 3a), through infusion of higher insulin doses (all P < 0.0001) (Fig. 3b) and in the presence of similar caloric intake (Fig. 3c). In both study groups, serum 3HB was similarly suppressed from days 1 to 3 (Fig. 3d), with a significant decrease from ICU admission to day 1 (P < 0.0001). Of all 400 3HB determinations, concomitant hypoglycemia was present on one corresponding blood glucose measurement at day 1 in a patient receiving liberal glucose control.