Diabetes Overtreatment and Hypoglycemia in Older Patients With Type 2 Diabetes on Insulin Therapy: Insights From the HYPOAGE Cohort Study

Recent clinical practice guidelines recommend to avoid “diabetes overtreatment” in older people living with type 2 diabetes (T2D), conceptually corresponding to an inappropriate level of glycemia that places the patient at high risk of unwanted side effects of glucose-lowering treatment (e.g., hypoglycemia), without providing a clinically implementable definition of this concept (1,2). To assess the prevalence and consequences of diabetes overtreatment, several studies used “proxy definitions of diabetes overtreatment,” based on thresholds of HbA_1c (3–6); there is no evidence on the predictive value of these proxy definitions to detect hypoglycemia. This study used data from the HYPOAGE cohort to assess the incidence and predictive value of hypoglycemia associated with proxy definitions of diabetes overtreatment in older insulin-treated patients living with T2D.

This ancillary study of the HYPOAGE prospective cohort (6), conducted from November 2017 to March 2020 in six French diabetes clinics, included outpatients aged ≥75 years, with T2D (duration of ≥1 year), treated by a glucose-lowering therapy including insulin for ≥6 consecutive months, with an HbA_1c measurement at baseline and with ≥2 self-monitoring blood glucose (SMBG) measurements per day (Supplementary Table 1).

All patients agreed to place a continuous glycemic monitoring (CGM) device (FreeStyle Libre Pro [FSLP] Sensor; Abbott Diabetes Care, Wiesbaden, Germany) for 28 consecutive days. CGM data were collected in double-blind conditions.

CGM data are reported as a percentage of the total time spent in different ranges (7): time above range (TAR) for glycemia >180 mg/dL (level 1 TAR [glycemia 181–250 mg/dL] and level 2 TAR [glycemia >250 mg/dL]), time in range for glycemia 70–180 mg/dL, and time below range (TBR) for glycemia <70 mg/dL (level 1 TBR [glycemia 54–69 mg/dL] and level 2 TBR [glycemia <54 mg/dL]). We differentiated diurnal (6 a.m. to midnight) and nocturnal (midnight to 6 a.m.) time.

Sociodemographic (age, sex), clinical, biological, and comprehensive geriatric assessment data were collected at baseline. HbA_1c was expressed as percent (National Glycohemoglobin Standardization Program) and millimoles per mole (International Federation of Clinical Chemistry). Health status was defined according to the three-tiered Blaum classification (8) (Supplementary Fig. 1).

We used two different proxy definitions of diabetes overtreatment from previous studies (1):

• “Fixed proxy definition” (9): HbA_1c <7.0% (<53 mmol/mol)

• “Individualized proxy definition” (3,10): HbA_1c <7.0% (53 mmol/mol), <7.5% (58 mmol/mol), and <8.0% (64 mmol/mol) for patients in good, intermediate, and poor health status, respectively (Supplementary Fig. 1)

The primary outcome was the TBR ≥1% (i.e., ≥15 min/day).

Secondary outcomes were ≥1% of level 1 TBR, ≥1% of level 2 TBR, and the occurrence of ≥1 confirmed hypoglycemic event (<70 mg/dL) measured by SMBG.

Continuous data were expressed as means ± SD or medians [interquartile range], while categorical data were shown as number (percent). Comparisons between overtreated and non-overtreated groups used Student t test or Wilcoxon rank sum test for continuous variables, and Pearson χ² test, or Fisher exact test for categorical variables. The diagnostic performance of overtreatment proxy definitions was assessed using measures of sensitivity, specificity, and accuracy (Supplementary Table 1). Statistical analyses were conducted with SAS version 9.4 (SAS Institute, Cary, NC), with P value <0.05 considered statistically significant.

Of the 141 patients with full follow-up of the HYPOAGE cohort, 134 had HbA_1c measurement at baseline and were included in this ancillary study (Supplementary Fig. 2).

Among the 134 patients aged 81.6 ± 5.4 years, with 41% female, 67 (54.5%) met the Fried criteria for frailty, 57 (42.5%) had malnutrition, and 29 (21.6%) were nursing home residents (Table 1). According to Blaum classification, global health was good in 16 (12%), intermediate in 67 (50%), and poor in 51 (38%) patients (criteria provided in Supplementary Fig. 1).

At baseline, the mean HbA_1c was 7.9 ± 1% (63 ± 11 mmol/mol). All participants were on insulin therapy, and the glucose-lowering regimen is detailed, with 22 (16%) on sulfonylurea or repaglinide (Supplementary Table 2).

Overall, 25 patients (18.7%) were classified as overtreated based on the fixed proxy definition and 53 patients (39.6%) based on the individualized proxy definition (Table 1).

The overtreated patients (for both definitions), as compared with the other patients, had lower HbA_1c, lower eGFR, and lower use of metformin. Overtreated patients according to the individualized proxy definition also showed lower functional independence (by definition). No other significant difference was observed between overtreated and non-overtreated patients (Table 1 and Supplementary Table 2).

Figure 1 (and Supplementary Table 2) shows the CGM metrics according to the presence or the absence of diabetes overtreatment. Trends for higher median TBR were observed in overtreated patients compared with non-overtreated patients using the fixed proxy definition (6.3% [1.1; 12.1] vs. 2.6% [0.9; 6.4]; P = 0.059) and the individualized proxy definition (4.1% [1.1; 12.1] vs. 2.6% [0.9; 5.1]; P = 0.070). Statistically significant differences were observed between overtreated participants and non-overtreated ones in the median time spent in level 1 TBR using the fixed definition (4.3% [1.1; 6.1] vs. 2.0% [0.6; 4.1]; P = 0.034) and the individualized proxy definition (3.6% [0.9; 6.1] vs. 2.0% [0.6; 3.6]; P = 0.038). The median time in range was higher in overtreated patients than in non-overtreated patients using the fixed proxy definition (82.4% [73.1; 90.0] vs. 65.7% [50.0; 78.3]; P < 0.001), and the individualized proxy definition (79.6% [64.2; 86.1] vs. 62.6% [46.7; 75.6]; P < 0.001). The median TAR was lower in overtreated patients than in non-overtreated patients using the fixed definition (5.9% [2.7; 16.6] vs. 28.0% [15.4; 44.7]; P < 0.001) and the individualized definition (15.4% [4.0; 26.7] vs. 29.9% [18.3; 51.2]; P < 0.001). The same pattern of CGM metrics distribution was found for level 1 TAR and level 2 TAR.

In overtreated participants, the median coefficient of variation indicating glycemic variability was 31.7% [29.0; 35.7] and 32.6% [29.2; 37.7] using the fixed and individualized proxy definitions, respectively, not significantly different from non-overtreated participants (Supplementary Table 2).

In participants with diabetes overtreatment, the incidence of hypoglycemia, defined as TBR ≥1%, was 100% using the fixed proxy definition and 94.3% using the individualized proxy definition. A significant difference in TBR ≥1% during nocturnal time between overtreated and non-overtreated patients was observed using the fixed proxy definition.

The incidence of level 1 TBR ≥1% was 100% and 92.5%, and of level 2 TBR ≥1% was 92.0% and 81.1%, respectively, according to the fixed and individualized proxy definitions (Table 2). Level 2 TBR ≥1% was more frequent in overtreated patients than in non-overtreated patients with the fixed proxy definition (92.0% vs. 72.5%, P = 0.039) but not with the individualized proxy definition (Table 2).

Overall, diabetes overtreatment based on an HbA_1c threshold had a low sensitivity, specificity, and accuracy for the prediction of hypoglycemia (e.g., TBR ≥1%). The individualized proxy definition of diabetes overtreatment, as compared with the fixed proxy definition, showed higher sensitivity and higher accuracy for the prediction of hypoglycemia (Table 3).

This analysis of the HYPOAGE cohort found no association between proxy definitions of diabetes overtreatment and occurrence of hypoglycemia in insulin-treated older adults living with T2D. Both fixed and individualized proxy definitions had low predictive value for hypoglycemia in this population. Our findings align with previous studies on diabetes overtreatment prevalence (3,9,11) but contradict those associating overtreatments with hypoglycemia (12). The high prevalence of hypoglycemia we found may be attributed to our use of CGM, differences in study population, settings, or outcomes definitions.

Our results therefore challenge the use of these proxy definitions of diabetes overtreatment for the detection of hypoglycemia, which is nevertheless assumed by various studies (1,3,9,11). One of the main issues of these proxy definitions is their reliance solely on (sometimes physiological) HbA_1c levels. While HbA_1c is effective for monitoring glycemic control over several weeks, previous research has demonstrated the weak link between low HbA_1c and hypoglycemia (13,14). Furthermore, recent evidence and progress in glucose-lowering treatment, including insulin therapies, allow for improved glycemic control (with lower HbA_1c) without significantly increasing hypoglycemia risk (15–17).

This finding underscores the necessity for a revised proxy definition of diabetes overtreatment. The prospect of a new CGM-based proxy definition is promising, yet it is premature to assert its validity. Our results require further validation, and the association between CGM metrics and adverse clinical outcomes must be thoroughly explored in this older population.

This study is the first to use CGM data to explore diabetes overtreatment and hypoglycemia in older adults with T2D, benefiting from prospective high-quality multicenter cohort data. However, all patients were treated with insulin, possibly underestimating the predictive values of overtreatment definitions. Moreover, the used CGM device might have introduced a measurement bias, overestimating hypoglycemia prevalence (18). Nevertheless, this bias is consistent across all patients, likely having minimal impact on outcomes. In addition, similar conclusions were found when using SMBG-confirmed hypoglycemia.

In conclusion, this study underscores the urgent need to explore alternative approaches for identifying older patients with T2D at risk for hypoglycemia from intensive glucose-lowering therapy. Addressing these issues is pivotal in optimizing the management of diabetes in older patients and mitigating the adverse consequences of glucose-lowering treatment.

Acknowledgments. We thank the investigators of HYPOAGE: Guillaume Chapelet and Matthieu Wargny (Centre Hospitalier Universitaire de [CHU] Nantes, France); Ingrid Allix, Claire Briet, and Cédric Annweiller (CHU Angers, France); Nathalie Cervantes and Alfred Penfornis (Centre Hospitalier Sud Francilien, Corbeil-Essonnes, France); Isabelle Delabrière and Rachel Litke (Centre Hospitalier Régional Universitaire de Lille, France); Pierre Gourdy and Sophie Guyonnet (CHU Toulouse, France); and Mac Paccalin and Pierre-Jean Saulnier (CHU Poitiers, France). We are indebted to Marielle Joliveau, Pamela Hublain, and Carole Agasse (CHU Nantes, France) for their assistance in patient screening and enrollment.

Funding. This Investigator Sponsored Study was funded by Sanofi.

Sanofi had no role in the design, methods, recruitment, data collections, or the analysis and preparation of the paper.

Duality of Interest. B.C. reports grants from Amgen, LVL-Air Liquide, Sanofi, and Ypsomed, and personal fees from Abbott, Amgen, AstraZeneca, Bristol-Myers Squibb, Eli Lilly and Company, LVL-Air Liquide Novartis, Novo Nordisk, Sanofi, and Ultragenyx. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.C., A.-S.B., S.H., and B.C. designed the ancillary study. A.C. and B.G. performed the statistical analyses. A.-S.B., L.d.D., S.H., and B.C. recruited and followed study patients. A.C. drafted the first version of the manuscript. All coauthors critically reviewed and edited the manuscript. B.C. 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.

Prior Presentation. The results of this study were presented during a commented poster session at the Annual Meeting of the Société Francophone du Diabète, Toulouse, France, 19–22 March 2024.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Vanita R. Aroda.