Therapies to control the acute alterations in blood glucose (BG) levels have been the focus of research since Van den Berghe et al. demonstrated in 2001 that intensive insulin therapy (insulin infusion to keep the BG tightly controlled between 80 and 110mg/dL) could significantly lower morbidity and mortality.1 Unfortunately, subsequent studies did not yield similar outcomes.2,3 Glucometrics is defined as the “systematic analysis of inpatient BG data” and various measures have been studied.4

Hyperglycemia, is seen in up to 20–52% of all ICU patients and is associated with increase in morbidity and mortality in the critically ill.5–7 Similarly, hypoglycemia is also harmful by a number of different mechanisms.8,9 Glycemic variability (GV) was recognized later as a dimension of glycemic control of importance among critically ill patients.10–13 Another unifying domain of glycemic control, time in target range (TIR), has emerged and integrates hyperglycemia, hypoglycemia and GV.14,15 It is noteworthy that several studies have shown that while GV is higher in diabetics, this increased variability does not portend a worse outcome, leading some authors to coin the term ‘diabetes paradox’.13,16,17

Strategies like continuous glucose monitoring (CGM) or closed loop glycemic control help to keep glucose levels in target range for a greater proportion of time aided by computer-assisted algorithms and reduce complications, especially hypoglycemia.18–21 Such techniques and individualized approaches were assessed in a few recent trials, none showed proven benefit.19–22

Specifically, no study explores the effects of glucometrics in the acute phase (first week) of critical illness on patient outcomes. We hypothesized that glucometrics is worse during the acute phase week of critical illness and this adversely affects outcome; hence we designed this study to assess the glucometrics in the first week of ICU stay and to correlate it with patient outcomes, including 28 day mortality.

Continuous variables are described as median (25th–75th percentile; i.e., IQR), and categorical variables as number (%). Mann–Whitney U test was used for comparison of continuous variables and the Pearson Chi square test or Fisher's exact test for categorical variables.

During the study period, 286 patients were screened; 100 patients met the inclusion criteria and were included (Figure S1, supplementary material). The median age was 55 years (IQR 41.25–65); 54% were male. The median APACHE II score was 18.5 (16–22) and median SOFA score 8 (6–10.25). Table 1 gives the baseline characteristics, the admitting diagnosis, the various ICU therapies and the ICU outcomes.

Glucometrics in the study cohort

A total of 5762 BG measurements were included with an average of 8.2 readings per patient per day. The median BG at admission was 205.5mg/dL (193.5–225). There were 2253 episodes of hyperglycemia (39%); the maximum number was on day 1 (338/556; 61%) and minimum was on day 7 (217/775; 28%) (Figure S2). These episodes were significantly higher on day 1 as compared to day 2 (338/556, 61% vs. 349/905, 39%; p<0.0001), remaining steady subsequently.

Hypoglycemia was seen in 28 out of 5762 measurements (0.48%); single episode in 23 patients, 2 episodes and 3 episodes in one patient each. None of the 25 patients had severe hypoglycemia (<40mg/dL); the lowest reading was 50mg/dL.

The median BG for the study cohort was 169mg/dL (162–178.75) and the mean 174.30mg/dL. The SD of the BG readings for the entire week was 31mg/dL (26–38.75), the CV was 16.7% (12.6–23.3), the GLI was 718.50 (mg/dL)2/h/week (540.50–1131.50) and the median TIR was 57% (50%–67%).

Similar to the pattern seen in hyperglycemia, median TIR was significantly lower on day 1 (40%) as compared to day 2 (61%) (p=0.003) and remained between 56 and 62% for the rest of the days (Figure S2).

General linear regression model for repeated measures was done for the glucometrics’ values of days 1–7 (Fig. 1). Mean BG, SD, GLI and TIR of each patient for the different days was analyzed. Day 1 values are significantly different from each of the other day's values for mean BG and TIR (repeated measures ANOVA p<0.001 for both).

Figure 1.

Linear regression for repeated measures from day 1 to day 7 for A. Mean blood glucose (mean BG), B. Standard deviation of mean blood glucose (SD), C. Time in blood glucose range (TIR) and D. Glycemic lability index (GLI). BG: blood glucose, SD: standard deviation of mean blood glucose, TIR: time in blood glucose range, GLI: glycemic lability index for each day.

(0.33MB).

In this prospective observational study, we have described glucometrics in the first week of ICU stay in critically ill adults, the factors associated with worse glucometrics and their association with adverse clinical outcomes. Diabetics and patients with higher admission APACHE II had worse glucometrics. Higher SD on day 1, and higher SD, CV and GLI in the first week of ICU admission were associated with mortality. This was similar to that seen in recent retrospective studies assessing GV.10,13,25,26

The prevalence of hyperglycemia in our study was 39%, similar to previous studies ranging from 40 to 54%.27 Analysis of repeated measures revealed a higher mean BG and lower TIR on day 1 as compared to each of the other days. This is probably due to the time taken to stabilize the patient's physiological derangements and titrate the insulin infusion.

The median BG in the cohort was 169mg/dL and the mean was 174mg/dL. Several investigators have combined mean BG with SD, GLI and other similar measures of GV to study the association with adverse outcomes. In earlier studies, it was found that patients with higher GLI, but with lower median BG, had higher mortality.28,29 In our study we found no association between mean or median BG and outcome.

Amongst the factors associated with worse glucometrics at admission, the presence of diabetes mellitus was independently associated with worse SD as well as worse TIR, similar to the studies by Chao et al.,13 and Krinsley et al.16 APACHE II at admission was also associated with a worse SD with an odds ratio of 1.16, suggesting that sicker patients have worse glucometrics.16

Glycemic variability (GV) is considered an important target in the glycemic management of ICU patients. Various measures of variability, i.e., mean BG, SD, CV and GLI have been studied. Egi et al. were the first to study the relationship of GV and outcome in critically ill patients.30 Egi et al. and subsequently Krinsley et al.31 found that the SD of glucose values is strongly associated with ICU and hospital mortality respectively. Similarly, Krinsley et al.31 and Lanspa et al.32 found CV as an important measure of glucose variability. A value above 20mg/dL for SD and 20% for CV have been associated with increased mortality.

In our study, we studied SD, CV, GLI and TIR of a cohort of critically ill patients and found each of these glucometrics associated with different adverse outcomes. Higher SD and CV were associated with higher vasopressor requirement. SD, CV and GLI were associated with 28-day mortality; on multivariate analysis, GLI was associated with almost 3 times higher risk of ICU mortality. The other important finding is the higher risk of BSI in patients whose TIR was below the cohort median.

The median GLI in our study was 718.5[mg/dL]2/h/week (540.50–1131.50), comparable to a recent multicentre study done in Australia and Sweden (GLI was 720 (mg/dL)2/h/week).25 Todi et al. showed that deciles of increasing GLI were associated with increasing ICU mortality.28 In our study, GLI above cohort median was associated with higher mortality (23% in higher GLI vs. 10% in lower GLI); this was significant on univariate and multivariate analysis.

The TIR in our study was 57%, which is similar to previous studies done.14,15 There was no significant difference in requirements of organ supports, length of ICU stay and mortality between patients with higher TIR versus with lower TIR. Diabetics spent less time in target range; however recent research suggests that the target BG levels should be different for diabetics as compared to non-diabetics.13,16,17 A TIR >80% is considered a desirable target in critically ill non-diabetics; however, the TIR in critically ill diabetics needs to be further defined.15

Our study suggests a significant association between low TIR and increased risk of bacteremia. To the best of our knowledge, it is the first to explore this association between TIR and bacteremia. This can be explained by the worsening in immunity and oxidative stress generated by glucose level variation, as shown in previous studies.33–35 Van den Berghe et al. found significant reduction in BSI in surgical ICU patients who received intensive insulin therapy; perhaps because these patients achieved a high TIR.1

Several mechanisms have been postulated in which high GV could be harmful in critically ill patients. Analyses of patients with diabetes mellitus requiring insulin have suggested an association between glycemic variability and end-organ damage through endothelial dysfunction.11 It is likely that poorer outcomes among critically ill patients may be due to increased oxidative stress.11 Such oxidative stress has been found to lead to endothelial dysfunction and vascular injury. Apoptosis of cells has also been observed with rapid correction of hyperglycemia to normoglycemia as against sustained hyperglycemia.11,34

SD, CV and GLI were significantly higher in diabetics, while TIR was lower in diabetics when compared to non-diabetics. This finding was similar to the study done by Lanspa et al. and Krinsley et al. where diabetics had more GV when compared to non-diabetics.15−17,32 This increased GV in diabetic patients does not result in higher mortality, whereas the reverse is true for non-diabetics.16,17 Hence, any glucometric analysis in critically ill patients should be matched for diabetic status and it needs to be personalized in the phase of critical illness.16,25,36 In our study, analysis of adverse outcomes in diabetic patients did not yield results different from outcomes in the entire cohort (results not shown), perhaps due to the small numbers.

Among critically ill patients receiving insulin infusion, glycemic lability index was independently associated with 28-day mortality. Diabetics were found to have worse glucometrics. TIR was significantly associated with the risk of BSI.

VS: Designed the study, collected the data, did the preliminary analysis and wrote the first draft of the manuscript.

BP: Conceived and designed the study, finalized the analysis and the manuscript and is the corresponding author.

SS: Helped in finalizing the analysis and the manuscript.

SKJ: Collected the data and helped VS in analysing the results.

AA: Critiqued the manuscript and suggested modifications.

MG: Helped in designing the study and made major changes to the manuscript.

RKS: Critiqued the manuscript and suggested changes.

AKB: Critiqued the study design, manuscript and suggested changes.

All authors have read the manuscript and approved the manuscript for submission.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

No conflict of interest declared by any of the authors.