Supplement to Research Letter Intensive Care Medicine

Supplement to Research Letter Intensive Care Medicine

Associations Between Dynamics of the Blood Glucose Level after Hypoglycemia and Intensive Care Unit Mortality: A Retrospective Multicenter Study

Roosmarijn T.M. van Hooijdonk MD1, Jan M. Binnekade PhD1, Ameen Abu–Hanna PhD2, Floris van Braam Houckgeest MD3, Lieuwe S. Hofstra MD, PhD4, Janneke Horn MD PhD1, Michael A. Kuiper MD, PhD5, Nicole P. Juffermans MD, PhD1,6, Huub L.A. van den Oever MD7,Johannes P. van der Sluijs MD, PhD8, Peter E. Spronk MD, PhD1,9, Marcus J. Schultz MD, PhD1,6 for the Tight Glucose Control Group

Academic Medical Center, Amsterdam, The Netherlands:

1Department of Intensive Care Medicine,

2Department of Medical Informatics,

6Laboratory of Experimental Intensive Care and Anesthesiology (L·E·I·C·A)

Tergooi Hospitals, location Hilversum, Hilversum, The Netherlands

3Department of Intensive Care Medicine

Scheper Hospital, Emmen, The Netherlands

4Department of Intensive Care Medicine

Medical Centre Leeuwarden, Leeuwarden, The Netherlands

5Department of Intensive Care Medicine

Deventer Hospital, Deventer, The Netherlands

7Department of Intensive Care Medicine

Medical Center Haaglanden, The Hague, The Netherlands

8Department of Intensive Care Medicine

Gelre Hospitals, location Lukas, The Netherlands

9Department of Intensive Care Medicine

Methods

Study design and ethical approval

This is a retrospective analysis of pooled cohorts of consecutive patients admitted to the ICU from seven hospitals in the Netherlands from January 2007 till December 2012. The Institutional Review Board of the Academic Medical Center, Amsterdam, The Netherlands, approved the study protocol and waived the need for ethical approval and individual patient consent to collect and analyze data from registries that exclude patient–identifying information.

Study settings and local guidelines for blood glucose control

All ICUs were mixed medical–surgical ICUs with their patients under the direct care of a team consisting of board–certified ICU nurses and ICU physicians. Recommendations in the local guidelines are presented in Table 1.

Table 1. Characteristics of the guidelines for blood glucose control in use
Guideline / 1 / 2 / 3
Blood glucose targets / 8.3 mmol/L / 5–8 mmol/L / 4.4–6.1 mmol/L
Insulin administration / Continuous intravenous infusion or subcutaneous administration / Continuous intravenous infusion / Continuous intravenous infusion of insulin
Correction of hypoglycemia / No recommendation, correction was often established with abrupt stops of insulin infusions and the administration of intravenous boluses of dextrose / Insulin infusion was stopped, and a bolus of glucose was usually given when the blood glucose level dropped < 3.4 mmol/L / Gradual recovery from hypoglycemia, explicitly avoiding abrupt stops of intravenous insulin infusions and avoiding of administration of intravenous boluses of dextrose
Number of ICUs and years in use / 6 ICUs – 3 years
3 ICUs – 3 years / 1 ICU – 6 years / 3 ICUs – 2 years

Inclusion and exclusion criteria

Readmitted patients and patients < 18 years old were excluded from the present analysis. Patients were included when developing hypoglycemia, defined as a blood glucose level below 3.9 mmol/L [1–3]. Patients in whom hypoglycemia occurred but in whom no follow–up blood glucose measurement within eight hours were performed were also excluded from the analysis.

We used a cut–off of 3.9 mmol/Lfor hypoglycemia, because two previous studies showed that a blood glucose level below 3.9 mmol/Lwas associated with increased mortality, an association that was independent of local guidelines for blood glucose control[2, 3]. A period of eight hours after hypoglycemia was chosen, for two reasons: first, we expected all corrections, including possible transitions from hypoglycemia to hyperglycemia, to be captured within this timeframe; second, sufficient numbers of blood glucose levels are needed to calculate the metrics of blood glucose variability used in the present analysis; the number of blood glucose measurements, including the index measurement was expected to be three or more within eight hours.

Data collection

Blood glucose measurement data were retrieved from the hospitals’ clinical information systems and linked to patient data. Prospectively collected admission diagnosis, demographic data, Acute Physiology and Chronic Health Evaluation (APACHE) II scores and Simplified Acute Physiology Scores (SAPS) II, and outcome data were extracted from the Dutch National Intensive Care Evaluation (NICE) registry [4].

Endpoint and definitions

The endpoint was ICU mortality, defined as death in the ICU.

Categorization of blood glucose levels following hypoglycemia

We categorized blood glucose levels following hypoglycemia using quartiles of the peak blood glucose level (the peak glucose), the delta of blood glucose levels (the delta glucose) and the standard deviation of blood glucose levels (the SD glucose), all calculated from blood glucose levels obtained within eight hours after the first blood glucose measurements revealing hypoglycemia. Peak glucose was defined as the highest blood glucose level within that time frame; delta glucose was defined as the difference between the lowest and highest blood glucose level within that time frame; and SD glucose was defined as the standard deviation of all blood glucose levels within that time frame [1, 5].

Analysis plan

First, we determined ICU mortality in the groups of quartiles of the peak glucose, the delta glucose and the SD glucose, and in electively admitted patients versus non–elective patients.

Then, we performed three separate multivariate logistic regression models examining the independent associations between the three metrics for blood glucose variability and ICU mortality. The quartiles with the lowest ICU mortality were used as reference categories. Continuous covariates were fitted using restricted cubic splines. We defined a covariate as a confounder when it changed the odds ratio for one of the quartiles by at least 10% when entered into each model [6]. As such, the APACHE II score, elective versus non–elective admission, the median blood glucose level during stay in ICU, the lowest blood glucose level in the hypoglycemic period, duration of hypoglycemia in minutes and one versus more than one hypoglycemic episode were considered significant confounders. As all these covariates showed no collinearity (the Pearson correlation coefficients were all under 0.5), they were all included in multivariate analysis. Other covariates tested for confounding but not changing the Odds ratio by at least 10%, and thus not included in multivariate analysis, included, the local guideline for blood glucose control (see eTable 1), and the standard deviation of the mean blood glucose level calculated from blood glucose levels obtained during the entire stay in ICU.

Statistical analysis

Data were reported as means with their standard deviations, medians with their interquartile ranges, or percentages, where appropriate. Data were compared using the Student’s t–test, the Mann–Whitney U test or the Chi–Square test. Blood glucose metrics calculated from blood glucose levels obtained within eight hours after hypoglycemia and blood glucose metrics calculated from blood glucose levels obtained during the entire stay in ICU were summarized for the total cohort and compared for patients who died or survived the ICU. Analyses were performed using R (version: 3.1.1; R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was considered to be at a P–value < 0.05. When appropriate, statistical uncertainty was expressed by the 95% confidence levels.

Results

Patients

The seven cohorts consisted of 19,505 patients, of which 4,512 patients (23%) were included in the present analysis. Demographic data and commonly used metrics of blood glucose control are presented in Table 2. Metrics of blood glucose variability within the eight hours following hypoglycemia are presented in Table 3.

Table 2. Patient characteristics and blood glucose metrics calculated from blood glucose levels obtained during over the entire stay in ICU
Died in ICU
N = 966 / Survived ICU
N = 3,546 / P–value / All patients
N = 4,512
Age – years, median [IQR] / 70 [59–78] / 67 [56–75] / <0.001 / 67 [56–76]
Male gender, n (%) / 563 (58) / 2063 (58) / 0.983 / 626 (58)
Non–elective admission, no (%) / 874 (91) / 2629 (75) / <0.001 / 3503 (78)
APACHE II scores, median [IQR], 12 missing / 28 [22–34] / 20 [15–26] / <0.001 / 22 [16–28]
SAPS II median [IQR] / 62 [50–75] / 44 [35–55] / <0.001 / 48 [37–60]
ICU LOS – days, median [IQR] / 4.8 [2.1–11.8] / 5.6 [2.7–10.9] / N.A. / 5.4 [2.5-11.1]
Number of blood glucose measurements, median [IQR] / 48 [25–103] / 48 [24–90] / 0.356 / 48 [24–92]
Mean blood glucose level – mmol/L, median [IQR] / 6.8 [6–7.8] / 6.8 [6.2–7.8] / 0.153 / 6.8 [6.1–7.8]
Median blood glucose level – mmol/L, median [IQR]7.4 / 6.4 [5.6–7.4] / 6.4 [5.8–7,4] / 0.014 / 6.4 [5.8–7.4]
Delta glucose – mmol/L, median [IQR] / 10.5 [7.9–14.2] / 10 [7.7–13] / 0.002 / 10.1[7.8–13.3]
Standard deviation of blood glucose level – mmol/L, median [IQR] / 2.3 [1.7–3.2] / 2.2 [1.7–2.9] / 0.001 / 2.2 [1.7–2.9]
Maximum blood glucose level – mmol/L, median [IQR] / 13.2 [10.7–16.9] / 12.9 [10.7–15.9] / 0.043 / 13 [10.7–16]
Minimum blood glucose level – mmol/L, median [IQR] / 2.8 [2.2–3.3] / 3.1 [2.5–3.5] / <0.001 / 3.0 [2.4–3.4]
Abbreviations: IQR, interquartile range; BMI, body–mass index; APACHE, Acute Physiology and chronic Health Evaluation; SAPS, Simplified Acute Physiology Score LOS, length of stay; ICU, intensive care unit; NA, not applicable
Table 3. Blood glucose metrics
Time between ICU admission and first hypoglycemic event – hours, median [IQR] / 26 [11–60]
Time between last hypoglycemic event and death – hours, median [IQR] / 24 [9-66]
Time between the blood glucose measurement showing hypoglycemia and first follow-up measurements – minutes, median [IQR] / 79 [52-128]
Patients with more than one hypoglycemic event, no (%) / 2,280 (51)
ICU mortality for patients with one hypoglycemic event, no (%) / 455 (20)
ICU mortality for patients with more than one hypoglycemic event, no (%) / 511 (22)
Number of blood glucose measurements within 8 hour time frame, median [IQR] / 4 [3–6]
Mean number of blood glucose measurements per 8 hours over entire stay in ICU, median [IQR] / 3 [2–4]
Time interval between measurements within 8 hour time frame – minutes, median [IQR] / 114 [83–158]
Time interval between measurements over entire stay in ICU – minutes, median [IQR] / 136 [111–179]
Abbreviations: IQR, interquartile range; ICU, intensive care unit; NA, not applicable

Univariate and multivariate analysis

In the univariate analysis, only the lowest quartile of the peak glucose, the delta glucose and the SD glucose were associated with increased ICU mortality compared to the reference categories (Table 4). In the multivariate analysis of the peak glucose, when including the APACHE II score and elective versus non–elective admission as confounder, the lowest quartile remained to be associated with ICU mortality. However, when including the APACHE II score, elective versus non–elective admission, the median blood glucose level calculated from blood glucose levels obtained during the entire stay in ICU, duration of hypoglycemia and one versus more than one hypoglycemic episode, this association disappeared. Using a similar approach in multivariate analyses of the delta glucose and the SD glucose, the lowest quartiles of the delta glucose and the SD glucose remained to show a significant association with ICU mortality.

Table 4. Results for univariate and multivariate analysis
Univariate analysis / Multivariate analysisa / Likelihood ratio test
Odds ratio [95% CI] / Odds ratio [95% CI]
Peak glucose
< 5.6 mmol/L / 1.62 [1.33-1.98] / 1.21 [0.95-1.53] / 0.02195b
5.6–7mmol/L / 1 (reference) / 1 (reference)
7–8.9mmol/L / 1.06 [0.85-1.31] / 1.05 [0.83-1.33]
> 8.9mmol/L / 1.22 [0.99-1.50] / 1.07 [0.84-1.36]
Delta glucose
< 2.4mmol/L / 1.41 [1.16-1.73] / 1.28 [1.01-1.62] / <0.001c
2.4–3.8mmol/L / 1 (reference) / 1 (reference)
3.8–5.9mmol/L / 1.00 [0.81-1.24] / 1.00 [0.80-1.27]
> 5.9mmol/L / 1.20 [0.98-1.48] / 0.99[0.78-1.27]
SD glucose
< 1.1mmol/L / 1.63 [1.33-1.99] / 1.54 [1.22-1.94] / <0.001d
1.1–1.7mmol/L / 1.14 [0.93-1.40] / 1.20 [0.97-1.49]
1.7–2.6mmol/L / 1 (reference) / 1 (reference)
> 2.6mmol/L / 1.10 [0.89-1.35] / 0.97 [0.78-1.20]
Abbreviations: APACHE II score, Acute Physiology and Chronic Health Evaluation II score; CI, confidence interval;
aThe multivariate model for peak glucose included the following confounders: the APACHE II score,elective versus non–elective admission, the median blood glucose level during stay in ICU, duration of hypoglycemia in minutes and one versus more than one hypoglycemic episode;
the multivariate model for delta glucose included the APACHE II score, elective versus non–elective admission, the median blood glucose level during ICU, the lowest blood glucose level during the hypoglycemic episode, the duration of hypoglycemia and one versus more than one hypoglycemic episode;
the multivariate model for SD glucose included the median blood glucose level during ICU, the lowest blood glucose level during the hypoglycemic period and the duration of hypoglycemia.
bP–value for likehoodratio test multivariate model peak glucose versus multivariate model delta glucose
cP–value for likehoodratio test multivariate model delta glucose versus multivariate model SD glucose
dP–value for likehoodratio test multivariate model SD glucose versus multivariate model peak glucose

References

1. Finfer S, Wernerman J, Preiser J-C, et al. (2013) Clinical review: Consensus recommendations on measurement of blood glucose and reporting glycemic control in critically ill adults. Crit Care 17:229.

2. Finfer S, Liu B, Chittock DR, et al. (2012) Hypoglycemia and risk of death in critically ill patients. N Engl J Med 367:1108–18.

3. Krinsley JS, Schultz MJ, Spronk PE, et al. (2011) Mild hypoglycemia is independently associated with increased mortality in the critically ill. Crit Care 15:R173.

4. Arts D, de Keizer N, Scheffer G-J, de Jonge E (2002) Quality of data collected for severity of illness scores in the Dutch National Intensive Care Evaluation (NICE) registry. Intensive Care Med 28:656–9.

5. Eslami S, Taherzadeh Z, Schultz MJ, Abu-Hanna A (2011) Glucose variability measures and their effect on mortality: a systematic review. Intensive Care Med 37:583–93.

6. McNamee R (2003) Confounding and confounders. Occup Environ Med 60:227–34; quiz 164, 234.

1