Why Time Below Range Is a Poor Standalone Predictor of Severe Hypoglycemia: Evidence from Six Clinical Trials

Highlights

Baseline CGM metrics for Time Below Range (TBR) levels 1 and 2 demonstrate only marginal predictive accuracy for future severe hypoglycemic (SH) events.

The Area Under the Curve (AUC) for TBR1 and TBR2 was 0.65 and 0.62, respectively, indicating poor discrimination for risk stratification.

Current thresholds for TBR (1-5%) fail to provide a balance of sensitivity and specificity necessary for reliable clinical intervention, underscoring the importance of clinical history and hypoglycemia awareness assessments.

The Hypoglycemia Paradox in Modern Diabetes Management

Severe hypoglycemia (SH) remains the most significant acute complication and the primary limiting factor in the glycemic management of Type 1 Diabetes (T1D). While the advent of continuous glucose monitoring (CGM) has revolutionized our ability to track glucose fluctuations in real-time, the clinical community has long sought a definitive digital biomarker that can forecast the risk of a severe event—defined as an episode requiring third-party assistance. The metrics of Time Below Range (TBR) at level 1 (TBR1: <70 mg/dL) and level 2 (TBR2: <54 mg/dL) have been widely adopted as surrogate markers for hypoglycemic burden. However, their utility as prospective predictors of severe, potentially life-threatening episodes has remained under-evidenced. This study by Montaser et al. provides a critical reality check on the predictive limitations of these metrics.

Study Design and Methodology

To evaluate the predictive capacity of baseline CGM metrics, researchers conducted a secondary analysis of pooled data from six distinct clinical trials. The study population included 1,433 individuals with T1D, representing a diverse age range (median ages across trials from 4 to 43 years). The cohort was 50-62% female and predominantly White (83-96%).

The primary objective was to determine if baseline TBR1 and TBR2 could accurately predict SH adverse events during the follow-up period of the respective trials. Researchers employed the Wilcoxon rank sum test to compare baseline TBR between those who did and did not experience SH, and Spearman correlations to assess the relationship between the number of SH events and TBR levels. Receiver operating characteristic (ROC) analysis was used to calculate the area under the curve (AUC), providing a standardized measure of how well these metrics discriminate between high-risk and low-risk patients.

Key Findings: Statistical Significance vs. Clinical Utility

The results of the analysis reveal a nuanced but ultimately disappointing role for TBR as a standalone predictor. Across the six trials, the baseline median TBR2 ranged from 0.1% to 0.7%, while TBR1 ranged from 1.2% to 4.1%.

1. Comparative Analysis of Baseline Metrics

Participants who went on to develop SH during the follow-up period did indeed exhibit higher baseline TBR levels than those who did not. Specifically, the SH group had a median baseline TBR2 of 0.41% compared to 0.32% in the non-SH group (P = 0.022). Similarly, TBR1 was 2.58% in the SH group versus 2.23% in the non-SH group (P = 0.044). While these differences are statistically significant, the absolute delta is remarkably small, suggesting that at a population level, the overlap between groups is too significant to allow for individual-level prediction.

2. Predictive Accuracy and Discrimination

The ROC analysis further highlighted the limitations. The AUC for TBR2 was 0.62 (95% CI 0.55-0.69), and for TBR1, it was 0.65 (95% CI 0.58-0.71). In clinical diagnostics, an AUC below 0.70 is generally considered to represent poor discriminatory power. This indicates that using TBR alone is only slightly better than a coin flip in identifying who will suffer a severe hypoglycemic event.

3. Sensitivity and Specificity Trade-offs

The study evaluated various TBR thresholds (1% to 5%) to see if a specific cutoff could serve as a clinical “red flag.” For TBR2, a 1% cutoff yielded a sensitivity of 48.9% and a specificity of 75.9%. Increasing the threshold to 5% improved specificity to 95.0% but crashed sensitivity to a mere 18.2%. For TBR1, an 81.8% sensitivity was achieved at the 1% threshold, but at the cost of a very high false-positive rate (specificity of 29.6%). These trade-offs suggest that clinicians cannot rely on a single TBR percentage to accurately triage patients for intensive hypoglycemia prevention interventions.

Mechanistic Insights: Why TBR Falls Short

The lack of strong predictive power for TBR can be attributed to several physiological and behavioral factors. Severe hypoglycemia is not merely a function of “time spent low”; it is often the result of a failure in the body’s counterregulatory response, frequently referred to as Hypoglycemia-Associated Autonomic Failure (HAAF). Patients with long-standing T1D may have high TBR values but remain asymptomatic due to hypoglycemia unawareness, which paradoxically increases their risk for SH while making their CGM data appear chronically “low” without immediate catastrophe.

Furthermore, SH events are often triggered by acute, unpredictable factors such as strenuous exercise, alcohol consumption, insulin dosing errors, or illness. A baseline 14-day CGM snapshot (the typical duration for these trials) may capture chronic glycemic patterns but cannot account for the acute behavioral or environmental volatility that often precipitates a severe event.

Expert Commentary and Clinical Implications

The findings of Montaser et al. emphasize that the “Time in Range” (TIR) framework, while excellent for assessing overall glycemic control and the risk of microvascular complications, is insufficient for assessing the acute risk of SH. Clinicians should be cautioned against a false sense of security when a patient has a low TBR, or conversely, over-reacting to a moderately elevated TBR without context.

Risk stratification for SH must remain multi-factorial. Key indicators that remain more predictive than TBR include:

• A history of SH in the previous 12 months.
• Formal assessment of hypoglycemia unawareness (e.g., Gold or Clarke scores).
• High glycemic variability (Coefficient of Variation > 36%).
• Duration of diabetes and presence of autonomic neuropathy.

In the era of automated insulin delivery (AID) systems, the relationship between TBR and SH may change further, as these systems are specifically designed to minimize TBR. Future research should focus on whether “Time Below Range” remains a relevant metric in the context of advanced hybrid closed-loop algorithms.

Conclusion

The analysis of 1,433 participants across six clinical trials confirms that while baseline TBR1 and TBR2 are associated with SH risk, they do not achieve the level of discrimination required to be used as solo predictive tools. No single threshold of TBR provides a reliable clinical balance of sensitivity and specificity. For clinicians, the takeaway is clear: CGM data is a vital piece of the puzzle, but it must be integrated with clinical history and patient-reported awareness levels to effectively mitigate the risk of severe hypoglycemia.

References

1. Montaser E, Williams C, Shah VN. Assessing Time Below Range as a Predictor of Severe Hypoglycemia: Insights From Six Clinical Trials. Diabetes care. 2026;49(3):483-489. PMID: 41529164.

2. Battelino T, Danne T, Berthold AL, et al. Clinical Targets for Continuous Glucose Monitoring Data Interpretation: Recommendations From the International Consensus on Time in Range. Diabetes Care. 2019;42(8):1593-1603.

3. Cryer PE. Hypoglycemia-associated autonomic failure in diabetes. Diabetes. 2005;54(12):3592-3601.