Skip to main content

06-11-2019 | Hypoglycemia | ADA 2019 | News

CGM data can predict future severe hypoglycemia

medwireNews: Most hypoglycemia outcomes measured by real-time continuous glucose monitoring (CGM) can be used to predict future severe hypoglycemia in people with type 1 diabetes, with similar levels of accuracy, study findings indicate.

Different thresholds are needed, however, for patients using masked CGM relative to those using self-monitored blood glucose (SMBG) or open CGM, Norbert Hermanns from the Research Institute Diabetes Academy Mergentheim in Germany reported at the 79th ADA Scientific Sessions in San Francisco, California, USA.

The findings were derived from an analysis of baseline data from 127 control group participants of the HypoDE and DIAMOND trials who wore a masked real-time CGM device (Dexcom G4, Dexcom Inc, San Diego, California, USA) for the baseline phase of each study (2–4 weeks) before continuing with SMBG for a further 22–24 weeks.

During that time, there were 44 severe hypoglycemic events, defined as a need for third-party intervention, among 12.6% of participants.

Using the baseline CGM data, Hermanns and team calculated cutoffs that best predicted severe hypoglycemia in the following 6 months according to area under the receiver operating characteristic curve (AUC) analysis for six CGM outcomes:

  • Proportion of glucose values <70 mg/dL
  • Proportion of glucose values <54 mg/dL
  • The number of glucose values <70 mg/dL
  • The number of glucose values <54 mg/dL
  • Low Blood Glucose Index (LBGI)
  • Glucose coefficient of variation (CV)

All but glucose CV significantly predicted future severe hypoglycemia, with the largest AUC (0.75) achieved using the number of events below 54 mg/dL at a threshold of 12.5.

Hermanns pointed out, however, that there was very little difference among the other four measures, with AUCs ranging from 0.71 for LBGI and the proportion of values below 54 mg/dL to 0.74 for proportion of values below 70 mg/dL, when using cutoffs of 1.6%, 2.4%, and 7.0%, respectively.

The team also looked at how well the SMBG and open CGM data predicted severe hypoglycemia and found that the AUCs with SMBG were similar to those with masked CGM, ranging from 0.71 to 0.78, with slightly higher cutoffs.

By contrast, the AUCs and thresholds needed to predict severe hypoglycemia tended to be lower when using data from open CGM. For example, the strongest predictor in this case was LBGI at a cutoff of 0.65 (AUC=0.76), followed by the proportion of values below 70 mg/dL, which had an AUC of 0.72 at a cut off of 3%. The weakest predictor using open CGM data was the proportion of glucose values below 54 mg/dL, which did not reach statistical significance.

Based on these findings Hermanns said that if he had a patient with more than 3% of open CGM values below 70 mg/dL he would consider modifying their treatment because they may have an elevated risk for severe hypoglycemia in the future.

Hermanns concluded that “CGM outcomes are able to predict future severe hypoglycemia” and that the performance of each of these outcomes “is highly comparable, therefore we can question if we need them all or if we can concentrate on one or two.”

He added that the differences observed between thresholds using masked CGM and open CGM were “remarkable.”

By Laura Cowen

medwireNews is an independent medical news service provided by Springer Healthcare. © 2019 Springer Healthcare part of the Springer Nature group

79th ADA Scientific Sessions; San Francisco, California, USA: 7–11 June 2019

More on this topic