Risk Factors for Hypoglycemia During Treatment of Hyperglycemic Crises Free

This was a case-control study of medical patients admitted to the ICU in hyperglycemic crisis at a single Veterans Affairs (VA) health system from 1 January 2013 to 31 March 2020. All patients who were ≥18 years of age and who received insulin therapy as treatment for DKA, HHS, or both were included in the study. Patients were excluded if death occurred within 24 hours of transitioning from IV to SC insulin; no additional exclusion criteria were applied. The case group included patients who developed hypoglycemia at any point during their treatment for hyperglycemic crisis, whereas the control group included patients who did not develop hypoglycemia. Hypoglycemia was defined as any blood glucose reading <70 mg/dL per the American Diabetes Association (ADA) classification of hypoglycemia (1). Clinically important hypoglycemia was defined as any blood glucose reading <54 mg/dL, also in keeping with the ADA classification (7). Per the institution’s protocol, blood glucose was monitored via point-of-care glucose meter every 1–2 hours while patients were on IV insulin infusion and four to six times daily while they were on SC insulin. Additionally, a basic metabolic panel, including blood glucose, was collected every 2–4 hours until resolution of hyperglycemic crises and then at least once daily during the ICU stay. The groups were compared to assess the odds of developing hypoglycemia based on risk factors, including BMI, comorbidities, and type of SC insulin used. Odds ratios (ORs) were calculated to assess the odds associated with developing hypoglycemia while on either IV or SC insulin therapy.

Data were collected retrospectively using patients’ electronic medical records in the VA’s Computerized Patient Records System. As previously described, patients were identified based on the diagnosis of hyperglycemic crisis in the medical record. Additional data collected included demographic characteristics, comorbidities, time to resolution of hyperglycemic crises, laboratory data at admission, and details of IV and SC insulin therapy.

Risk factors assessed for all patients included age >70 years old, presence of chronic kidney disease (CKD), underweight status (BMI <18.5 kg/m²), diagnosis of type 1 diabetes, and presence of concomitant bacteremia (defined as a positive blood culture for a known pathogenic organism). Based on the KDIGO (Kidney Disease: Improving Global Outcomes) definition, CKD was defined as an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² (8). The risk factor assessed for developing hypoglycemia while receiving IV insulin infusion was the presence of an IV insulin bolus dose before the infusion. For patients developing hypoglycemia while on SC insulin, risk factors assessed after resolution of the hyperglycemic crisis included whether the exact chronic basal insulin dose was resumed, whether the exact chronic SC insulin regimen (basal and bolus insulins, if applicable) was resumed, the total daily dose of SC insulin received (>0.5 units/kg/day), the presence of scheduled short-/rapid-acting insulin administration, the type of basal insulin used (NPH vs. glargine insulin), the type of short-/rapid-acting insulin used (regular vs. aspart insulin), and the dose of correctional sliding-scale insulin (SSI) received (high-dose vs. either low-dose or medium-dose SSI). The choice of basal insulin used (NPH vs. glargine) was at the discretion of the treating team; however, factors that may have influenced this choice included the patient’s chronic insulin prescription and the formulary status of each agent at the time. The weight-based dosing value of 0.5 units/kg/day was selected because this is the lower end of the range for weight-based dosing in the ADA consensus statement recommendations (3). These risk factors were chosen because of their association with known or suspected hypoglycemia in previous studies or other situations (3,5,6,9–17).

The OR for each risk factor was calculated using MedCalc (MedCalc Software, Ostend, Belgium) and reported with a 95% CI and P value, with P <0.05 considered statistically significant.

This study received institutional review board approval from the University of Florida and was conducted in accordance with requirements for investigations involving human subjects. Informed consent was waived.

A total of 220 hyperglycemic crises were diagnosed during the study time frame. Four patients were excluded by study criteria, leaving 216 patients for evaluation. The development of hypoglycemia occurred while receiving IV insulin for 11 patients, after transitioning to SC insulin for 44 patients, and both during IV insulin therapy and after the transition to SC insulin for six patients. These 61 cases of hypoglycemia were compared with 155 control subjects. All patients received an insulin infusion until resolution of hyperglycemic crisis, at which point they were transitioned to SC insulin per protocol.

In the control group, the median age was 64 years (interquartile range [IQR] 56.5–70.5 years), 146 patients (94.2%) were male, and the median BMI was 28.0 kg/m² (IQR 24.2–33.3 kg/m²). In the case group, the median age was 63 years (IQR 53–69 years), 56 patients (91.8%) were male, and the median BMI was 23.6 kg/m² (IQR 20.9–27.3 kg/m²). The most identified inciting event for hyperglycemic crises in both groups was insufficient insulin, which included medication or dietary nonadherence, insulin pump malfunction, and a new diagnosis of diabetes. This finding is consistent with what the consensus statement recommendations report as the most common causes of hyperglycemic crises (3).

Patients who did not have documented diabetes at baseline presented with hyperglycemic crisis as a new diabetes diagnosis. The newly diagnosed patients were determined to have type 1 or type 2 diabetes following treatment for hyperglycemic crises and identified as such for the analysis of outcomes. Twenty-one patients had new-onset diabetes; one was determined to have type 1 diabetes, and 20 were determined to have type 2 diabetes. Complete patient characteristics are summarized in Table 1.

Complete results are summarized in Tables 2–4. Odds for developing hypoglycemia were significantly higher for patients diagnosed with CKD or type 1 diabetes and those with a BMI <18.5 kg/m². Likewise, resuming the exact chronic insulin regimen and using NPH versus glargine insulin also increased the risk for hypoglycemia. Seventy-four patients (47.7%) in the control group had CKD compared with 39 patients (63.9%) in the hypoglycemia group (OR 1.94 [95% CI 1.05–3.57], P = 0.03). In the control group, three patients (1.9%) were underweight compared with five patients (8.2%) in the hypoglycemia group (OR 2.82 [95% CI 1.11–7.18], P = 0.03). A total of 27 patients (17.4%) had a diagnosis of type 1 diabetes in the control group compared with 28 patients (45.9%) in the hypoglycemia group (OR 4.02 [95% CI 2.09–7.73], P <0.01). Of the 61 patients who developed hypoglycemia, 24 patients (39.3%) had at least one episode of clinically important hypoglycemia (i.e., blood glucose <54 mg/dL) per the ADA’s hypoglycemia classification (7). The control group included eight patients (8.5%) who resumed the exact same chronic SC insulin regimen after resolution of hyperglycemic crisis compared with 10 patients (21.3%) in the hypoglycemia group (OR 2.91 [95% CI 1.06–7.95], P = 0.04). In the control group, five patients (3.4%) received NPH rather than glargine insulin after resolution of hyperglycemic crisis compared with seven patients (15.2%) in the hypoglycemia group (OR 5.13 [95% CI 1.54–17.06], P <0.01). No significant differences were seen in the other evaluated variables of age, presence of concomitant bacteremia, receipt of an IV insulin bolus before initiation of insulin infusion, resumption of chronic basal insulin dose, presence and type of scheduled short-/rapid-acting insulin, SSI dose, and total daily dose of insulin.

The results of this study show a significantly increased odds of developing hypoglycemia in patients receiving treatment for hyperglycemic crises with several risk factors, including low BMI, presence of CKD or type 1 diabetes, and use of NPH insulin. These findings are consistent with previous findings for development of hypoglycemia in other settings (3,5,6). Based on review of available literature, this is the first study to our knowledge to assess these specific risk factors for developing hypoglycemia during treatment for hyperglycemic crises. Consensus statement recommendations do not specifically address adjusting management based on risk factors for hypoglycemia, but they do note overzealous insulin dosing and use of regular insulin and NPH (compared with an insulin analog) have been shown to increase the risk for hypoglycemia during treatment of hyperglycemic crises.

Although the reported results are significant, this study does have limitations that merit discussion. Because of the retrospective observational study design, some biases are inherent. Selection bias should have been low with our inclusion of all hyperglycemic crisis patients in the defined study period with minimal exclusion criteria; however, our exclusion of patients who died within 24 hours of transitioning to SC insulin may have affected our results. Observer bias was minimized by having all variables be objective and dichotomous, with a clearly defined outcome of hypoglycemia; thus, it would be unlikely that the data were affected by the data collection method. Additionally, some variables such as chronic insulin dosing and the timing of transitioning to SC insulin could have been subject to information bias if not properly documented in the electronic medical record.

Additional confounding variables not accounted for also may have been present. For example, specific dietary and carbohydrate intake, which may have influenced insulin dosing, was not assessed. This study did not evaluate how multiple risk factors may have influenced the development of hypoglycemia. For example, the summative effect of having type 1 diabetes and using NPH insulin was not assessed compared with these factors alone. Development of symptomatic hypoglycemia could not be determined because of the nature of the study design.

The external validity for this study is also limited because of its single-center design and predominantly older adult, male study sample, so the results may not be applicable to other settings or patient populations.

Additionally, this study evaluated hyperglycemic crises as a whole; however, DKA and HHS differ in their etiologies. DKA is typically the result of insufficient insulin, whereas HHS is thought to be more of a result of decreased insulin sensitivity (1,3). Patients with type 1 diabetes are younger and more likely to develop DKA and have lower BMIs. Although we did not specifically evaluate DKA and HHS individually, it would be appropriate to evaluate risk factors for each in further studies.

The use of noninsulin diabetes medications can increase the risk of hypoglycemia. In this study, no such medications were administered to any patient during their ICU stay; therefore, it is unlikely these medications would be responsible for any of the observed hypoglycemia. Likewise, the frequency of blood glucose monitoring was consistent for all patients per the institution’s hyperglycemic crises treatment protocol; thus, it is unlikely the frequency of blood glucose monitoring would be a potential explanation of differences in observed hypoglycemia between groups.

A notable finding in this study was the increased incidence of hypoglycemia after transitioning from IV insulin to a patient’s exact chronic insulin regimen. Per consensus statement recommendations, patients with known diabetes may be given insulin at the dosage they were receiving before the onset of DKA or HHS as long as it was controlling glucose properly before the event (3). This is a consistent recommendation across multiple updates to the consensus guidelines, with few data to the contrary (3,10). Because insufficient insulin was the most reported inciting event for hyperglycemic crisis in this study, it may be a reflection of poor glucose control before the onset of hyperglycemic crisis for those patients. With insulin nonadherence being a major concern, further investigation may be warranted to determine whether an adjustment to these recommendations would be considered best practice to reduce the incidence of hypoglycemia. Additionally, decreased dietary intake during hospitalization is common and could explain the finding of exact chronic insulin regimen resumption being associated with hypoglycemia. This scenario may highlight the need for sound clinical judgment when resuming chronic insulin dosing in the inpatient setting.

Another notable risk factor reported in our study was the increased incidence of hypoglycemia with the use of NPH versus glargine insulin. The higher rate of hypoglycemia with NPH may be explained by the inadequate duration of therapy requiring twice-daily dosing, combined with an undesirable peak effect seen 4–6 hours after administering the dose (11). Additionally, there is variability in the rate of absorption, onset, and duration of activity for NPH based on lipodystrophy, local blood supply, and body temperature (12). Prior studies of patients with type 1 diabetes have reported patients treated with NPH and regular insulin after resolution of DKA have about a 25% increased incidence of hypoglycemia compared with patients being treated with glargine and glulisine (3,11). A similar study evaluating hazard ratios (HRs) for the use of NPH versus both detemir and glargine in older patients with type 2 diabetes not in hyperglycemic crisis found a significantly decreased risk of hypoglycemia when using glargine (HR 0.71 [95% CI 0.63–0.80]) or detemir (HR 0.72 [95% CI 0.63–0.82]) (13). Our results, combined with similar reported results, warrants consideration of whether it should be advised to avoid using NPH insulin after resolution of hyperglycemic crises to reduce the risk of hypoglycemia.

Patients who have type 1 diabetes or CKD or were underweight have an increased risk of hypoglycemia, as seen in this study and others (3,5,6). CKD patients are at an increased risk of hypoglycemia because there is a reduction of renal gluconeogenesis in addition to a decrease in the degradation and elimination of insulin (14). The increased risk of hypoglycemia in underweight patients could be partially explained by these patients’ increased risk for hyperglycemic crises.

Previous studies assessing duration of insulin use in patients with type 1 or type 2 diabetes have shown a significant increase in hypoglycemia rates for individuals with type 1 diabetes, which is thought to be the result of variations in the sensitivity of insulin receptors from prolonged insulin use (15).

Although infection is one of the most common precipitating factors for hyperglycemic crises, this study’s finding of a lower percentage of patients with concomitant bacteremia in the hypoglycemic group may have been explained by the frequent correlation between infections and hyperglycemia (3,16). These results also showed a higher proportion of younger patients developed hypoglycemia, contrary to the commonly reported increased hypoglycemia rates in elderly patients. Although other studies have not assessed age as a risk factor among a similar population or among individuals receiving treatment for hyperglycemic crises, one explanation for this finding could be the increased insulin sensitivity and reduced decline in β-cell function seen in younger patients with a relatively newer diabetes diagnosis (17). Additionally, the percentages of patients receiving a total daily dose of SC insulin >0.5 units/kg/day, those placed on high-dose SSI, and those who received an insulin bolus before the insulin infusion were all greater in the hypoglycemic group, as would have been expected given that all of these circumstances would lead to an increase in the amount of insulin received; however, none of these were statistically significant risk factors. A larger sample size may have been beneficial to more definitively determine whether these are significant risk factors for developing hypoglycemia and should be considered for future studies.

The odds of developing hypoglycemia during treatment for hyperglycemic crises may be associated with low BMI, CKD, type 1 diabetes, resumption of the exact same chronic insulin regimen, and use of NPH insulin. Further investigation is warranted because some of these findings oppose or have not yet been reported in consensus statement recommendations. These findings may help ICU clinicians prevent this complication of hyperglycemic crisis management or generate hypotheses for further studies.

All work associated with this article was performed at the North Florida/South Georgia Veterans Health System. This material is the result of work supported with resources and facilities of the U.S. Department of Veterans Affairs. The contents do not represent the views of the U.S. Department of Veterans Affairs or the U.S. government.

No potential conflicts of interest relevant to this article were reported.

R.C.R. contributed to the acquisition/analysis/interpretation of data and drafted the manuscript. N.S.V. contributed to the acquisition/analysis/interpretation of data. K.W., A.L.H., J.B., and E.D.T. contributed to the conception/design and reviewed/edited the manuscript. A.J.F. contributed to the conception/design, reviewed/edited the manuscript, and oversaw the completion of the study. A.J.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Portions of this project were reported virtually in abstract form at the Society of Critical Care Medicine 2022 Annual Critical Care Congress, 18–21 April 2022.