Are point-of-care measurements of glycated haemoglobin accurate in the critically ill?

Is it time to personalise glucose targets during critical illness?

PURPOSE OF REVIEW

Dysglycaemia complicates most critical care admissions and is associated with harm, yet glucose targets, particularly in those with preexisting diabetes, remain controversial. This review will summarise advances in the literature regarding personalised glucose targets in the critically ill.

RECENT FINDINGS

Observational data suggest that the degree of chronic hyperglycaemia in critically ill patients with diabetes attenuates the relationship between mortality and several metrics of dysglycaemia, including blood glucose on admission, and mean blood glucose, glycaemic variability and hypoglycaemia in the intensive care unit. The interaction between acute and chronic hyperglycaemia has recently been quantified with novel metrics of relative glycaemia including the glycaemic gap and stress hyperglycaemia ratio. Small pilot studies provided preliminary data that higher blood glucose thresholds in critically ill patients with chronic hyperglycaemia may reduce complications of intravenous insulin therapy as assessed with biomakers. Although personalising glycaemic targets based on preexisting metabolic state is an appealing concept, the recently published CONTROLLING trial did not identify a mortality benefit with individualised glucose targets, and the effect of personalised glucose targets on patient-centred outcomes remains unknown.

SUMMARY

There is inadequate data to support adoption of personalised glucose targets into care of critically ill patients. However, there is a strong rationale empowering future trials utilising such an approach for patients with chronic hyperglycaemia.

The Effect of a Liberal Approach to Glucose Control in Critically Ill Patients with Type 2 Diabetes: A multicenter, parallel-group, open-label, randomized clinical trial

Rationale Blood glucose concentrations affect outcomes in critically ill patients but the optimal target blood glucose range in those with type 2 diabetes is unknown. Objective To evaluate the effects of a 'liberal' approach to targeted blood glucose range during intensive care unit (ICU) admission. Methods This mutlicenter, parallel-group, open-label, randomized clinical trial included 419 adult patients with type 2 diabetes expected to be in the ICU on at least three consecutive days. In the intervention group intravenous insulin was commenced at a blood glucose >252 mg/dL and titrated to a target range of 180 to 252 mg/dL. In the comparator group insulin was commenced at a blood glucose >180 mg/dL and titrated to a target range of 108 to 180 mg/dL. The primary outcome was incident hypoglycemia (<72 mg/dL). Secondary outcomes included glucose metrics and clinical outcomes. Main Results At least one episode of hypoglycemia occurred in 10 of 210 (5%) patients assigned the intervention and 38 of 209 (18%) patients assigned the comparator (incident rate ratio: 0.21 (95% CI, 0.09 to 0.49); P<0.001). Those assigned the intervention had greater blood glucose concentrations (daily mean, minimum, maximum), less glucose variability and less relative hypoglycaemia (P<0.001 for all comparisons). By day 90, 62 of 210 (29.5%) in the intervention and 52 of 209 (24.9%) in the comparator group had died (absolute difference 4.6 percentage points (95%CI, -3.9 to 13.2%); P=0.29). Conclusions A liberal approach to blood glucose targets reduced incident hypoglycemia but did not improve patient-centered outcomes. Clinical trial registration available at www.anzctr.org.au, ID: ACTRN12616001135404.

Using the AUSDRISK score to screen for pre-diabetes and diabetes in GP practices: a case-finding approach

OBJECTIVE

To identify the optimal AUSDRISK threshold score to screen for pre-diabetes and diabetes.

METHODS

A total of 406 adult patients not diagnosed with diabetes were screened in General Practices (GP) between May and October 2019. All patients received a point of care (POC) HbA1c test. HbA1c test results were categorised into diabetes (≥6.5% or ≥48 mmol/mol), pre-diabetes (5.7-6.4% or 39-47 mmol/mol), or normal (<5.7% or 39 mmol/mol).

RESULTS

Of these patients, 9 (2%) had undiagnosed diabetes and 60 (15%) had pre-diabetes. A Receiver Operator Characteristic (ROC) curve was constructed to predict the presence of pre-diabetes and diabetes; the area under the ROC curve was 0.72 (95%CI 0.65-0.78) indicating modest predictive ability. The optimal threshold cut point for AUSDRISK score was 17 (sensitivity 76%, specificity 61%, + likelihood ratio (LR) 1.96, - likelihood ratio of 0.39) while the accepted cut point of 12 performed less well (sensitivity 94%, specificity 23%, +LR=1.22 -LR+0.26).

CONCLUSIONS

The AUSDRISK tool has the potential to be used as a screening tool for pre-diabetes/diabetes in GP practices. A cut point of ≥17 would potentially identify 75% of all people at risk and three in 10 sent for further testing would be positive for prediabetes or diabetes. Implications for public health: Routine case-finding in high-risk patients will enable GPs to intervene early and prevent further public health burden from the sequelae of diabetes.

Are point-of-care urine albumin-creatinine ratio measurements accurate in the critically ill?

BACKGROUND

In the critical care environment, elevated albuminuria values show capacity to reflect illness severity and predict mortality and hence assessing albumin/creatinine ratio (ACR) at the bedside has potential clinical benefit Point-of-care (POC) analysers offer rapid results but may be less accurate then laboratory analysis.

METHODS

Critically ill adult patients with a urinary catheter in situ had albumin, creatinine, and ACR measurements performed via laboratory and POC analysis. Data are presented as mean (standard deviation) or median [interquartile range]. Measurement agreement was assessed by Lin's concordance correlation coefficient, Bland Altman 95% limits of agreement, and classification by Cohen's kappa statistic.

RESULTS/FINDINGS

Albumin, creatinine, and ACR analysis was performed for 30 patients. Lin's correlation coefficient showed 'substantial' agreement for albumin and ACR and 'almost perfect' agreement for creatinine for POC vs laboratory analysis. POC vs laboratory analysis also showed poor agreement for identification of normal ACR (>1 mg/mmol) and mild urine ACR (1-3 mg/mmol) and 'substantial' agreement for moderately increased urine ACR (3-30 mg/mmol).

CONCLUSIONS

ACR POC values appear to provide an accurate and rapid method that has potential to provide an early indication of injury severity and mortality risk in the critically ill.