Endocrine Dysfunction Criteria in Critically Ill Children: The PODIUM Consensus Conference

The PODIUM collaborative sought to develop evidence-based criteria for organ dysfunction in critically ill children. The present article reports on the systematic review on endocrine-dysfunction scoring tools performed as part of PODIUM, provides a critical evaluation of the available literature, and proposes evidence-based criteria for endocrine dysfunction in critically ill children, as well as recommendations for future research listed in the Supplemental Information. The PODIUM Executive Summary details Population, Interventions, Comparators, and Outcomes questions, search strategies, study inclusion and exclusion criteria, and processes for risk of bias assessment and data abstraction and synthesis and for drafting and developing agreement for criteria indicating endocrine dysfunction.¹¹ 

Of 7027 unique citations published between 1992 and 2020 identified, 212 full texts were assessed for eligibility and 121 met the inclusion and exclusion criteria, as shown in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart (Fig 1), data tables (Supplemental Tables 1 and 2), and risk of bias assessment summary (Supplemental Fig 1). A total 84, 18, and 22 studies pertained to abnormal glucose homeostasis, thyroid dysfunction, and adrenal dysfunction, respectively, including 3 studies representing combinations thereof. Table 1 lists the criteria for diagnosis of abnormal glucose homeostasis, thyroid dysfunction, and adrenal dysfunction in critically ill children.

Criteria for diagnosis of abnormal glucose homeostasis in critically ill children include the following:

1. We recommend that BG concentrations ≥150 mg/dL (≥8.3 mmol/L) be considered as hyperglycemia.

2. We recommend that BG concentrations <50 mg/dL (<2.8 mmol/L) be considered as hypoglycemia.

After excluding preexisting disorders of glucose homeostasis (eg, diabetes mellitus and metabolic disorders affecting glucose metabolism), we reviewed a total of 84 eligible articles (Supplemental Table 1, section A). There were 17 prospective studies, including 2 randomized controlled trials (RCTs), with the remainder being retrospective studies and secondary analyses of previously conducted trials.

There were a total of 84 studies in which researchers addressed the association of hyperglycemia with outcomes (Supplemental Table 2, section A). The definition of hyperglycemia varied across these studies, with majority of researchers defining hyperglycemia as BG ≥150 mg/dL (≥8.3 mmol/L). Of 25 studies in the noncardiac medical and surgical PICU setting, there were 23 studies in which researchers observed an association of hyperglycemia with worse outcomes. In an additional 8 studies in a mixed medical-surgical and cardiac PICU setting, researchers examined prognostic factors associated with hyperglycemia and noted varying, but uniformly positive, associations of hyperglycemia with mortality, organ failure, and hospital-acquired conditions, such as venous thromboembolism and central catheter-associated bloodstream infections.

We reviewed 18 studies of hyperglycemia in children after cardiac surgery. Eight studies revealed positive associations of hyperglycemia with mortality and morbidity. In 6 studies, researchers did not detect any association of hyperglycemia with worse outcomes. In none of 4 additional studies did researchers find an association of hyperglycemia with worse neurodevelopmental outcomes after surgery.

We reviewed 18 retrospective studies on hyperglycemia in children with traumatic brain injury (TBI). In all studies, researchers observed worse morbidity and mortality outcomes in association with elevated BG concentrations ≥150–200 mg/dL (≥8.3–11.1 mmol/L). We reviewed an additional 20 studies with abnormal glucose homeostasis in a variety of disease states (trauma, burns, diarrhea with malnutrition, malaria, submersion injuries, liver transplant, status epilepticus, encephalopathy, and post neurosurgery). Although researchers observed the association of hyperglycemia with worse outcomes in 18 studies in children with severe burn injuries, diarrhea with malnutrition, submersion injuries, and post neurosurgery, in 2 studies (1 in critically ill children presenting to the emergency department and 1 in postoperative liver transplant pediatric patients), researchers did not report any associations.

We reviewed a total of 19 studies in which researchers examined hypoglycemia (Supplemental Table 1, section A) and association with outcomes (Supplemental Table 2, section A). The definition of hypoglycemia varied across studies (from BG <40 mg/dL [<2.2 mmol/L] to BG <60 mg/dL [<3.3 mmol/L]), with researchers in several studies defining hypoglycemia as BG <50 mg/dL (<2.8 mmol/L). In 11 studies, researchers reported the association of hypoglycemia with higher mortality, organ failure, greater length of stay (LOS), and fewer ventilator-free days, whereas in 2 studies, researchers did not report any association. In an additional 3 studies in critically ill children with a variety of disease states (diarrhea with malnutrition, malaria, and emergency department setting), researchers observed the association of hypoglycemia with higher mortality.

In 2 of 3 retrospective studies in the cardiac surgical population, hypoglycemia was associated with EEG seizures and slower EEG recovery and greater mortality. However, in a third study, researchers found no association of hypoglycemia with mortality.

In summary, both hyperglycemia and hypoglycemia are commonly associated with poor outcomes across the breadth of pediatric critical illness. Although hypoglycemia is associated with seizures and EEG changes in survivors, in no study did researchers find an association between critical illness hyperglycemia and poor neurodevelopmental outcomes. It is important to note limitations associated with variable sampling frequency, source, and site in the interpretation of BG concentrations.^12,13 

Criteria for diagnosis of thyroid axis failure in critically ill children^* are as follows:

1. We recommend that serum T4 of <4.2 μg/dL (< 54 nmol/L) be considered thyroid axis failure.

2. No recommendations regarding free thyroxine (FT4), free or total T3, or TSH can be provided on the basis of the existing evidence.

We reviewed 18 eligible articles in critically ill children in which researchers reported on levels of free and total T4, T3, TSH, or rT3 measured on admission, in the perioperative period, and/or at discharge (Supplemental Table 1, section B) and associations with outcomes (Supplemental Table 2, section B). The 2 largest studies included patients across the entire pediatric age range and diagnostic categories and accounted for two-thirds of patients studied. The largest study, in which researchers examined nutrition supplementation in critically ill children, researchers found that lower total T4 on admission was independently associated with a higher risk of death at 90 days (odds ratio [OR] 0.972 [95% confidence interval (CI): 0.953–0.992]; P = .004) and a higher risk of acquiring a new infection (0.987 [0.976–0.998]; P = .02). Total T3, rT3, and the ratio of T3 to rT3 were not statistically significant. Similarly, in a large RCT of glucose control, low T4 on admission was associated with mortality (P = .02). A limitation of both of these studies is that free T4 was not measured and no cutoff for low T4 was analyzed or identified. In the majority of the remainder of the studies, researchers concurred that low T4 levels (and in some instances, low T3 levels) are associated with poor outcomes.

In one study, researchers reported area under the receiver operating curve (AUROC) at 0.81, with sensitivity 75% and specificity 96% of a cutoff value of T4 <4.2 μg/dL at discharge from PICU to predict survival. In a study of patients with sepsis and septic shock, researchers found FT4 and T4 levels substantially lower in nonsurvivors (FT4: 12.77 ± 3.22 vs 20.64 ± 3.48 [P < .001]; T4: 64.5 ± 15.86 nmol/L vs 105.78 ± 19.35 [P < .001]) but did not analyze testing characteristics. A larger study revealed a cumulative increase in mortality if T3, T4, and TSH were each sequentially >2 SD below normal (normal T3 and T4, 3.4% mortality; low T3, 5.4%; low T3 and T4, 10%; and low T3, T4, and TSH, 62.5%). Finally, in a similar-sized cardiac surgical study, researchers found similar associations with postoperative T4 and FT4. In 1 study, researchers did not identify statistically significant associations between T4 levels and survival, and in 1, researchers did not report on mortality as an outcome.

Taken together, correlation of low T4 levels on ICU admission with mortality is a reproducible phenomenon detected in studies of thyroid function in critically ill children across the ranges of age and diagnosis. Although FT4 is more commonly measured in thyroid management, T4 is the analyte reported in all ICU studies. A T4 concentration <4.2 μg/dL (54 nmol/L) is consistent across studies as a sufficiently low value to indicate thyroid failure in the context of nonthyroidal illness syndrome and should not be applied to patients with preexisting primary or central thyroid disease.

Criteria for diagnosis of adrenal axis failure in critically ill children^† include the following:

1. We recommend a peak serum cortisol concentration of <18 μg/dL (500 nmol/L) and/or an increment in serum cortisol concentration of <9 μg/dL (250 nmol/L) post ACTH stimulation be considered a sign of adrenal axis dysfunction in patients with clinical suspicion of primary adrenal insufficiency (eg, unexplained hyponatremia, hyperkalemia, hypoglycemia, and hemodynamic instability).

2. No recommendations for random cortisol concentrations as a measure for adrenal axis function in critically ill children can be provided on the basis the existing evidence.

We reviewed 22 eligible articles on adrenal axis function in critically ill children (Supplemental Table 1, section C). In 2 studies, researchers assessed receptor-level function, and in 3 studies, researchers assessed cortisol levels in postoperative cardiac surgery patients after high-dose corticosteroid administration. In 15 remaining studies, researchers assessed cortisol levels and included children from 1 month of age to adolescence. In no study did researchers evaluate cortisol levels in the context of an RCT, and only 4 studies were conducted prospectively. Six studies occurred in the general PICU population, whereas the others were performed in children with septic shock, acute respiratory distress syndrome, and meningococcemia.

Two studies suggested worse clinical outcomes in patients with low admission cortisol levels (Supplemental Table 2, section C). Menon et al found that a baseline cortisol level <5 µg/dL (138 nmol/L) was associated with increased number of catecholamine infusions (P = .001) and increased duration of infusion (P < .001), whereas Bone et al found that 4 of 5 children who died had baseline cortisol levels <7 µg/dL (200 nmol/L). In other studies, rated as having medium to high risk of bias, however, researchers did not find an association of lower cortisol levels with worse outcomes or examine higher cortisol cutoffs, rendering it difficult to make definitive recommendations regarding the clinically significant lower cutoff for cortisol. No study was adequately powered to determine the association of low cortisol levels with mortality.

In several studies, researchers assessed the association of high baseline cortisol levels on admission with hemodynamic outcomes, but in only 4 studies did researchers assess their association with mortality. In 2 studies, researchers found an increased mortality rate with cortisol levels >21.7 µg/dL (600 nmol/L) and 30 µg/dL (828 nmol/L), respectively, whereas in the other 2, researchers found no association of cortisol levels with mortality. All studies were limited by small sample sizes and were rated as moderate or high risk of bias, rendering definitive recommendations regarding upper cortisol levels difficult.

In several studies, researchers reported the association of the results of ACTH stimulation testing with a need for vasoactive-inotropic support. In the largest study of 389 patients, researchers found that a peak serum cortisol level <18 µmol/dL (500 nmol/L) and/or increment in serum cortisol level of <9 µmol/dL (250 nmol/L) post ACTH stimulation was associated with an increased need for catecholamines and more fluid boluses. In 2 studies, researchers found an increment in serum cortisol level of <9 µmol/dL (250 nmol/l) post ACTH stimulation to be associated with an increased risk of catecholamine-resistant shock, whereas in 1 study, researchers found an increased need for catecholamines to be associated with an increment in cortisol level of <7 µmol/dL (200 nmol/L). None of the studies revealed an association of adrenal insufficiency with mortality but may have been limited by their small sample sizes or low mortality rate.

In making recommendations regarding criteria for adrenal dysfunction in critically ill children, it is important to note several practical limitations. These include the limited availability of real-time cortisol testing in some centers,¹³  the high cost of administered ACTH,¹⁴  and the confusion surrounding interpretation of adrenal testing.¹⁵ 

Abnormalities of glucose homeostasis, thyroid function, and adrenal function are common in critically ill children and variably associated with worse clinical and functional outcomes. After a systematic review of 108 published articles via a modified Delphi process, we present criteria for multiple organ dysfunction syndrome (MODS)-associated endocrine dysfunction in critically ill children that include measures of BG, T4, and peak cortisol and/or increment in cortisol post ACTH stimulation testing. Future research priorities should include identification of additional upstream and downstream biomarkers of endocrine dysfunction in critically ill children.

ACTH

adrenocorticotropic hormone

AUROC

area under the receiver operating curve

BG

blood glucose

CI

confidence interval

FT4

free thyroxine

LOS

length of stay

MODS

multiple organ dysfunction syndrome

OR

odds ratio

RCT

randomized controlled trial

rT3

reverse triiodothyronine

T3

triiodothyronine

T4

thyroxine

TBI

traumatic brain injury

TSH

thyroid-stimulating hormone