Time Below Range and Its Influence on Hypoglycemia Awareness and Severe Hypoglycemia: Insights From the Association of British Clinical Diabetologists Study

Diabetes, characterized by chronic hyperglycemia resulting from defects in insulin secretion, action, or both, poses significant challenges to health care systems worldwide because of its escalating prevalence and associated morbidity and mortality (1,2). Achieving optimal glycemic control, a cornerstone of diabetes management, is paramount in mitigating the risk of microvascular and macrovascular complications (1). However, the pursuit of tight glycemic targets is often associated with hypoglycemia, an acute complication of abnormally low blood glucose levels (3–5). Hypoglycemia poses a substantial burden on individuals with diabetes, impairing quality of life, increasing health care use, and, in severe cases, culminating in adverse cardiovascular outcomes and mortality (3). Furthermore, impaired awareness of hypoglycemia (IAH), a state wherein individuals often fail to recognize hypoglycemic symptoms, exacerbates the risk of severe hypoglycemia (SH), which can be life threatening and undermines efforts toward achieving glycemic targets (6).

In recent years, the time below range (TBR) percentage has emerged as a valuable metric of glycemic variability, offering insights into periods of hypoglycemia and glycemic fluctuations not captured by traditional glycemic indices (7–10). TBR percentage represents the proportion of time spent below a specified glucose threshold, defined through international consensus as 70 mg/dL or 3.9 mmol/L (10–14). This metric provides an easy-to-understand measure of total exposure to biochemical hypoglycemia over a defined time period, thus complementing traditional measures, such as HbA_1c (10). Despite its growing recognition, the clinical implications of TBR percentage and its relationship with demographic and clinical factors, such as IAH and SH, remain underexplored. Therefore, we sought to explore the predictive utility of TBR percentage in identifying individuals at heightened risk of IAH and SH. By comprehensively examining these associations, we aimed to advance our understanding of glycemic variability and its impact on diabetes outcomes, ultimately informing personalized diabetes management strategies tailored to individual risk profiles.

This study constitutes a post hoc analysis of data collected as part of the ongoing nationwide FreeStyle Libre (FSL) Association of British Clinical Diabetologists (ABCD) audit (15), an initiative to evaluate the real-world clinical utility and outcomes associated with FSL continuous glucose monitoring (CGM) in individuals with diabetes. Data collection commenced in November 2017 through a secure online National Health Service tool, ensuring optimal security and enabling nationwide anonymized data analysis. Data were collected at baseline and follow-up during routine clinical care using secure National Health Service online tools in this nationwide audit. Baseline data included demographics, source of FSL funding, previous structured diabetes education, HbA_1c from the prior 12 months, Gold score for hypoglycemia awareness, history of SH, paramedic callouts, and hospital admissions for hypoglycemia, hyperglycemia, or diabetic ketoacidosis. Data collection relied on both patient recall and medical records where available.

Demographic characteristics, pre- and post-FSL HbA_1c levels, and TBR percentages were among the parameters collected. TBR percentage represents the proportion of time spent below a specified glucose threshold, defined per the Advanced Technologies and Treatments for Diabetes and International Hypoglycaemia Study Group consensus guidelines as 70 mg/dL or 3.9 mmol/L (10,16). Duplicates were meticulously removed, and those with paired baseline and follow-up data were included. Participants with complete loss of hypoglycemia awareness, as indicated by a Gold score of 7, were also identified. Furthermore, SH episodes with decimal point numbers or those exceeding 20 were excluded from the analysis because of clinical nonviability and presumed errors. Hypoglycemia awareness was evaluated using the Gold score, with a score of ≥4 denoting IAH. SH was defined as at least one episode of hypoglycemia necessitating third-party assistance. This was a cross-sectional study conducted at the first follow-up after the initiation of FSL. The study began in 2017, when intermittently scanned CGM in the U.K. did not have alerts for low glucose values. Therefore, the results presented reflect outcomes for intermittently scanned CGM use without the hypoglycemia alert feature.

Descriptive statistics, including means, SDs, and percentages, were used to characterize baseline demographics between individuals. Logistic regression analysis was used to identify the independent effect of TBR on hypoglycemia unawareness and SH. The Youden J index (17) was used to identify the optimal TBR percentage cutoffs for detecting IAH and SH. Through its calculation, which considers both sensitivity and specificity, the Youden J index provides a single metric to assess the overall performance of a diagnostic test across various threshold values. This optimal threshold maximizes the discriminative ability of the test, ensuring both high sensitivity and specificity while minimizing misclassification errors (17).

Caldicott Guardian approval has been obtained for the nationwide ABCD audit program, designating it as audit work rather than research. Guided by established guidelines, all data collection occurred during routine clinical visits, with only anonymized data submitted to the central database to ensure patient confidentiality and privacy.

Table 1 presents a comparative analysis of the demographics and clinical characteristics of individuals with diabetes stratified by those achieving TBR <4% (n = 2,132) and those with TBR ≥4% (n = 2,897). Forty-two percent of the study participants achieved TBR <4%. Patients with TBR ≥4% had longer diabetes duration, with mean durations of 25.7 years for TBR ≥4% and 22.1 years for TBR <4% (P = 0.04). Furthermore, individuals with TBR ≥4% were more likely to be using continuous subcutaneous insulin infusion therapy (21% vs. 16%; P < 0.0001). Completion rates for structured education programs like Dose Adjustment for Normal Eating (DAFNE) did not significantly differ between the groups. Notably, those with TBR ≥4% had lower baseline HbA_1c levels (65.9 vs. 75.3 mmol/mol; P < 0.0001). TBR ≥4% also correlated with higher mean Gold score (2.34 vs. 2.11; P < 0.0001). A higher incidence of IAH was observed in the TBR ≥4% group (17% vs. 12%; P < 0.0001).

The prevalence of IAH and SH in this population was 17.2% and 4%, respectively. We analyzed factors associated with IAH and SH using logistic regression analysis, as shown in Table 2. For IAH, age was a weak but significant predictor, with an odds ratio (OR) of 1.020 (95% CI 1.014–1.026; P < 0.001), indicating that each additional year of age increased the likelihood of hypoglycemia unawareness by 2%. Sex also showed a significant effect, with female patients being more likely to experience IAH compared with male patients (OR 1.257; 95% CI 1.064–1.485; P = 0.007). TBR percentage was another significant factor (OR 1.014; 95% CI 1.004–1.024; P = 0.005) associated with IAH, with a 1% increase in TBR associated with a 1% increase in IAH. Baseline BMI and baseline HbA_1c were not significant predictors of IAH. TBR was significantly associated with SH (OR 1.013; 95% CI 1.016–1.052; P < 0.001); age and sex were not.

The optimal TBR cutoffs for identifying follow-up IAH and SH, as determined by the Youden J index, were 3.35% and 3.95%, respectively. The receiver operating characteristic (ROC) curves for detecting IAH and SH are shown in Fig. 1A. The area under the curve (AUC) was 0.597 for IAH and 0.598 for SH, indicating that the model’s poor discriminative ability was marginally better than random guessing (AUC 0.5). Both ROC curves are close to the diagonal line, suggesting limited effectiveness in accurately predicting hypoglycemic events. Figure 1B illustrates the impact of incorporating the Gold score into the model predicting SH, resulting in an improved AUC of 0.74. This improvement suggests that combining TBR with the Gold score could offer a synergistic benefit in predicting SH. In a model using only the Gold score as a predictor of SH, the AUC of the ROC analysis was 0.73 (Fig. 1C). This demonstrates that the Gold score alone has good predictive ability, and the increase in AUC compared with models incorporating additional factors, such as TBR, is modest.

Figure 2 shows diagnostic performance measures with TBR cutoff for the optimal TBR cutoffs for SH and IAH. We observed that although the sensitivity for SH and IAH was moderate (65% and 73%, respectively), indicating the test’s ability to correctly identify true positive cases, the specificity was relatively low (43% and 43%, respectively), reflecting the test’s tendency to misclassify true negative cases. Interestingly, the positive predictive value (PPV) for IAH and SH was notably low (9% and 15%, respectively), indicating a high rate of false positives among individuals identified as positive for these conditions. Conversely, the negative predictive value (NPV) was high for both SH and IAH (97% and 85%, respectively), suggesting a low rate of false negatives among individuals identified as negative for these conditions.

In this nationwide study, we evaluated the link between TBR, IAH, and SH, finding that although TBR had a weak correlation with these outcomes, it had limited use in predicting them. However, its high NPV provides reassurance in ruling out IAH and SH. The term TBR gained prominence with the widespread adoption of CGM technology in the late 2000s and early 2010s (18). The most significant formal recognition occurred with the international consensus on time in range (TIR) publication in 2019 (10). This consensus report standardized TBR and other CGM measures, such as TIR and time above range (TAR). Subsequently, TBR was recognized as an important metric in managing diabetes (8–10,19).

The relationship between TBR and hypoglycemic risk has been investigated previously. A study by Thomas et al. (20) showed that TBR with CGM was significantly associated with a reduced epinephrine response during hypoglycemic clamps, indicating impaired counterregulation. This impaired epinephrine response may weaken the body’s ability to defend against future hypoglycemic episodes. The authors then concluded that CGM metrics, particularly TBR, could be valuable in identifying patients at heightened risk of SH and guiding preventive clinical interventions.

In our study, we showed that 42% of individuals with diabetes managed to meet the TBR of <4%, indicating the prevalent issue of hypoglycemia in this population. These findings agree with those of another study conducted using deidentified user accounts from LibreView, where only 47% to 55% of participants achieved the recommended TBR of <4% (21).

Our analysis revealed a positive correlation between increased TBR and IAH and SH. This correlation persisted even after adjusting for confounding variables, such as age, sex, and BMI. The results of our study are consistent with previous research, which showed that IAH is associated with higher percentages of values <3.9 and <3.0 mmol/L compared with those in patients with normal hypoglycemia awareness (11). Nevertheless, this study’s findings contradict those of a previous study (22), which indicated that CGM did not differentiate between individuals with impaired and normal awareness. The lack of correlation in this study might be attributed to the small sample size (22).

The Youden J index identified 3.35% and 3.9% cutoff values for IAH and SH, respectively. However, the specificity and predictive value of these cutoffs were low, at ∼50%, suggesting that TBR alone is insufficient to reliably predict individuals with IAH or those at high risk of SH. Despite this, the NPV was high, indicating that a TBR of <4% correlates with a very low risk of SH. This discrepancy may be explained by the ability of CGM technology to enhance a user’s awareness of glucose levels, enabling corrective actions to be taken before hypoglycemia becomes severe, even when there is significant exposure to low glucose levels. Therefore, although TBR may not predict SH directly, it remains clinically valuable for identifying low-risk patients and guiding management strategies.

One reason for the weak association between TBR and SH or IAH might be that it could take longer for reductions in hypoglycemia exposure from CGM interventions to affect IAH in individuals with longer diabetes duration. A longitudinal follow-up study by Rickels et al. (23) demonstrated that although endogenous glucose production in response to insulin-induced hypoglycemia in those using CGM did not change from baseline to 6 months, significant improvements were observed after 18 months. Furthermore, another study (24) showed that intervention with automated insulin delivery improved the epinephrine response during hypoglycemic clamps in individuals with IAH, with reductions in TBR contributing to this improvement. This highlights the potential of TBR not only as a risk marker but also as a modifiable target in clinical interventions to restore hypoglycemia counterregulation in individuals with IAH.

This study demonstrates that combining the Gold score with TBR resulted in the highest AUC for predicting SH. Prior research has shown that combining IAH and CGM metrics significantly improves the prediction of individuals with absent autonomic symptom recognition during insulin-induced hypoglycemia. Flatt et al. (25) demonstrated that the Clarke score, combined with CGM measures of hypoglycemia exposure, had a strong predictive ability (AUC ≥0.80) for identifying absent autonomic symptom recognition during hypoglycemic clamp experiments. Importantly, a composite threshold of IAH (Clarke score ≥4) alongside CGM measures of hypoglycemia exposure increased the specificity and predictive value for identifying individuals at risk of SH. This aligns with our findings, which highlight that the combination of TBR measurement and IAH assessment provides enhanced prediction of SH. However, our analysis also showed that the Gold score alone could predict SH, and there was modest improvement in the AUC with the addition of TBR.

It is clear that the Gold score and TBR do not capture the full complexity of hypoglycemia risk. Future research should assess how these measures can work synergistically to improve the accuracy of hypoglycemia risk prediction and guide better clinical decision-making. Recent data from the HypoMETRICS study (26) found similar rates of TBR between individuals with and without IAH. Notably, even among those with good hypoglycemia awareness, as indicated by low Gold scores (1–2), up to half of CGM-detected low glucose events were asymptomatic. This high rate of asymptomatic hypoglycemia, even in individuals classified as being at low risk of SH based on the Gold score, may explain the low predictive value of TBR alone in identifying those with IAH or SH.

Exploring additional metrics and approaches may provide greater predictive accuracy to improve the PPV of TBR. These could include incorporating a more comprehensive analysis of glycemic variability using TIR and TAR, alongside TBR, to enhance the prediction of hypoglycemia and overall glycemic control. Advances in machine learning and predictive algorithms that factor in glucose variability, insulin dosing patterns, and meal timing may also offer more precise predictive metrics than TBR alone. Additionally, combining CGM data with physiologic parameters, such as heart rate variability, skin temperature, and electrodermal activity, available from wearable devices, could further improve real-time hypoglycemia predictions.

We also explored the relationship between age, duration of diabetes, and risk of SH. The relationship between TBR percentage, age, and duration of type 1 diabetes suggests a nuanced interaction between glycemic targets and hypoglycemia risk. Older patients may have higher A_1c goals to reduce the risk of hypoglycemia, which could explain the association with lower TBR percentage. In contrast, patients with longer disease duration, who may be more adept at managing tight glucose control, tend to have higher TBR percentages. This could also be influenced by IAH, which becomes more prevalent with longer diabetes duration. The competition between these two effects, higher A_1c targets for older patients and tighter glucose management for patients with long-term diabetes, may play a key role in hypoglycemia risk as patients age. Future research should further explore these dynamics to inform better personalized glycemic targets that account for both age and disease duration while minimizing hypoglycemia risk.

Two important limitations of the FSL system must be acknowledged. First, as an intermittently scanned CGM system, FSL requires the user to scan the sensor to capture glucose data manually. This reliance on user interaction may lead to significant data gaps when scans are infrequent, potentially introducing errors in the calculation of TBR and leading to an inaccurate assessment of hypoglycemic exposure. Second, evidence suggests that the FSL system is less accurate in detecting glucose levels in the hypoglycemic range. Galindo et al. (27) showed that that the FSL system tended to overestimate hypoglycemia compared with capillary glucose testing in hospitalized patients with type 2 diabetes. Similarly, Alitta et al. (28) demonstrated an overestimation of hypoglycemia in long-term care home residents using the system. These inaccuracies may result in inflated TBR values and limit the precision of TBR as a standalone metric for assessing hypoglycemic risk. When interpreting our findings, these limitations should be considered and underscore the importance of combining CGM data with clinical judgment in hypoglycemia management.

This study provides valuable insights into the relationship between TBR, IAH, and SH; however, several limitations warrant consideration. First, our study relies on retrospective analysis of data collected from the nationwide FSL ABCD audit, introducing the possibility of selection bias and incomplete data capture. Furthermore, the determination of optimal TBR cutoffs for identifying IAH and SH may be influenced by study population characteristics and may not be generalized to other populations. Finally, the observational nature of our study precludes causal inference, and additional prospective studies are warranted to validate our findings and elucidate the mechanistic underpinnings of the observed associations. Despite these limitations, our study offers several strengths that enhance its clinical relevance and utility. Leveraging real-world data from a large cohort of individuals with diabetes using CGM provides valuable insights into glycemic control patterns and hypoglycemia risk in routine clinical practice. By identifying optimal TBR cutoffs for detecting IAH and SH and evaluating diagnostic performance measures, our study contributes to the refinement of hypoglycemia risk assessment tools and informs clinical decision-making in diabetes management. Moreover, our findings underscore the importance of minimizing TBR to reduce the incidence of IAH and SH, highlighting the clinical significance of glycemic variability metrics in optimizing diabetes care and improving outcomes.

In conclusion, our study highlights the critical role of TBR in managing diabetes and preventing hypoglycemic events. By understanding the factors contributing to increased TBR and implementing targeted interventions, health care providers can better care for those at risk of IAH and SH, ultimately improving the outcomes of individuals living with diabetes.

Acknowledgments. The authors thank all the clinicians and support staff who participated in the nationwide study, listed at https://abcd.care/Resource/ABCD-Freestyle-Libre-Audit-Contributors.

Funding. T.S. serves on the speakers bureau for Novo Nordisk Foundation.

Duality of Interest. The nationwide FSL ABCD audit is supported by a grant from Abbott Laboratories. E.G.W. has received personal fees from Abbott, AstraZeneca, Dexcom, Eli Lilly, Embecta, Insulet, Medtronic, Novo Nordisk, Roche, Sanofi, Sinocare, and Ypsomed and research support from Abbott, Embecta, Insulet, Novo Nordisk, and Sanofi. C.W. has a spouse/partner serving on the advisory panel for Celgene and on the speakers bureaus for LEO Pharma and Novartis. R.E.J.R. serves on the advisory panel for Novo Nordisk and on the speakers bureau for BioQuest. T.S. reports a relationship with Bristol-Myers Squibb, Eli Lilly, and Sanofi. P.C. has received consulting fees from Medtronic, Dexcom, Insulet Corporation, Abbott Diabetes, Lilly Diabetes, and Sanofi; honoraria or payment for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Novo Nordisk, Medtronic, Insulet Corporation, Lilly Diabetes, Sanofi Diabetes, and Glooko; payment for expert testimony and support for travel and attending meetings from Abbott Diabetes; participation in data safety monitoring boards or advisory boards for Medtronic; is the chair of the Diabetes Technology Network–UK and the lead for Type 1 Diabetes Midlands UK; and was supported by National Institute for Health and Care Research (NIHR) Wellcome Trust clinical research facility at King’s College Hospital and the NIHR patient recruitment center at University Hospitals Leicester. No other potential conflicts of interest relevant to this article were reported.

The FSL audit was independently initiated and performed by ABCD, and the authors remained independent in the analysis and preparation of this report.

Author Contributions. H.D. wrote the first draft of the manuscript and performed the data analysis. H.D., E.G.W., P.C., C.W., R.E.J.R., and T.S. conceived the presented idea. H.D., E.G.W., C.W., R.E.J.R., and T.S. contributed to the data analysis. All authors contributed to the writing of the manuscript; made extensive comments, criticisms, and changes to the final draft of the paper; and reviewed the final version of the manuscript. H.D. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Portions of this article were presented in abstract form at the 84th Scientific Sessions of the American Diabetes Association, 21–24 June 2024.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were John B. Buse and Jeremy Pettus.