Evaluation of %CV and average glucose derived from pre‐ and postprandial 6‐point glucose measurements for assessing hypoglycemia risk

Authors: Rika Soga, Keiichi Torimoto, Fumi Uemura, Yosuke Okada, Shingo Nakayamada

Abstract

Aims/Introduction

This study determined the usefulness of coefficient of variation (%CV) and average glucose (AG) from 6‐point pre‐ and postprandial blood glucose measurements in assessing hypoglycemia risk in type 2 diabetes.

Materials and Methods

We retrospectively analyzed 24‐h continuous glucose monitoring data of 233 patients treated with sulfonylureas or insulin. Hypoglycemia was defined as level 1 (<70 mg/dL) or level 2 (<54 mg/dL).

Results

The 6‐point AG and %CV were 182.9 mg/dL and 23.9%, respectively, for all patients, whereas they were 155.8 mg/dL and 29.8% in patients with level 1 hypoglycemia. Receiver operating characteristic analysis determined cutoff values of 167.7 mg/dL for AG and 21.6% for %CV for predicting level 1 hypoglycemia. Using these thresholds as hypoglycemia risk factors, the incidence was evaluated in the three groups according to the number of risk factors. The incidence of both levels 1 and 2 hypoglycemia increased significantly with increase in the number of risk factors. Compared to risk factor‐free patients, the odds ratios for level 1 hypoglycemia were 12.1 and 1.4 in those with two and one risk factor, respectively.

Conclusions

Our findings suggest that the AG and %CV derived from 6‐point glucose measurements can be potentially useful for detecting hypoglycemia risk.

INTRODUCTION

Large‐scale clinical trials have demonstrated that strict glycemic control in patients with diabetes mellitus can effectively prevent the progression of microvascular complications ^1^ , ^2^ , ^3^ . Although early intensive glycemic control is recommended to reduce the risk of macrovascular complications ^4^ , ^5^ , strict glycemic control increases the risk of hypoglycemia ^5^ , ^6^ . Moreover, excessively strict glycemic management, leading to hypoglycemia, has been associated with increased overall mortality ^7^ .

The widespread adoption of continuous glucose monitoring (CGM) has enabled detailed blood glucose tracking, particularly highlighting its usefulness in detecting asymptomatic nocturnal and early morning hypoglycemia ^8^ , ^9^ . Our previous study on hospitalized patients with type 2 diabetes demonstrated that patients with lower average glucose (AG) levels and higher % coefficient of variation (%CV) were more likely to experience hypoglycemia, as identified via CGM ^10^ . In addition, a study involving 999 outpatients with type 2 diabetes demonstrated a higher incidence of hypoglycemia in patients with low AG and high %CV, especially those treated with either sulfonylureas or insulin ^11^ . However, in current clinical practice, it is not feasible to equip all patients with a CGM, and self‐monitoring of blood glucose (SMBG) is widely practiced. Therefore, we explored the potential use of SMBG‐derived AG and %CV as predictors of hypoglycemia, analogous to those obtained using the CGM. Specifically, the aim of this study was to determine the risk of hypoglycemia based on AG and %CV, which were calculated by extracting 6‐point glucose values—pre‐prandial and 2‐hour postprandial, similar to SMBG settings—from the CGM data of patients with type 2 diabetes mellitus (T2DM) treated with either insulin or sulfonylureas, which are considered medications with high hypoglycemic risk.

MATERIALS AND METHODS

Study design and participants

This retrospective study included Japanese patients with T2DM who were treated with either sulfonylureas or insulin. We enrolled T2DM patients who were admitted to our institution or affiliated hospitals for diabetes education and glycemic control between April 2010 and March 2022. All patients underwent CGM within 5 days of hospitalization and did not change the doses or type of medications after wearing the CGM device. Among the 498 patients, we analyzed the data of 137 patients treated with sulfonylureas and 103 patients treated with insulin; whereas seven patients were treated with both sulfonylurea and insulin, with analysis of a total number of 233 patients. No standardized hypoglycemia prevention protocol was implemented in this retrospective study; patients were hospitalized mainly for diabetes education and routine glycemic management.

Anthropometric and biochemical measurements

Several parameters, including age, sex, duration of diabetes, diabetic complications, and medication(s) use, were obtained from the medical records and patient interviews. Clinical examinations, including height and weight measurements, were conducted according to standardized protocols. Blood samples were collected after overnight fasting. Body mass index (BMI) was calculated as weight (kg) divided by height (m)^2^. HbA1c, glucose, and serum creatinine levels were measured using standard methods. HbA1c values were converted to National Glycohemoglobin Standardization Program (NGSP) units using the following HbA1c (NGSP, %) = HbA1c (JDS) × 1.02 + 0.25 (%) ^12^ . Estimated glomerular filtration rate (eGFR, mL/min/1.73 m^2^) was calculated as for men, 194 × (serum creatinine in mg/dL)^(−1.094)^ × (age)^(−0.287)^, and for women, the result was multiplied by 0.739.

Insulin use was categorized into three basal‐only, bolus (including basal‐bolus therapy), and premixed insulin. Diabetes medications remained unchanged from outpatient treatment during hospitalization and throughout the CGM monitoring.

Continuous glucose monitoring

CGM was performed within 5 days of hospitalization using either the CGMS System Gold (Medtronic, Inc., Fridley, MN) or iPro 2 (Medtronic MiniMed, Inc., Northridge, CA). The CGM data were analyzed over a continuous 24‐h period on either the second or third day of monitoring. The CGM sensor was inserted subcutaneously to measure interstitial glucose concentrations every 5 min, recording up to 288 glucose values per day. These 288 values were referred to as 288‐point glucose levels. From this dataset, pre‐prandial and 2‐hour postprandial glucose values for breakfast, lunch, and dinner were extracted, yielding the 6‐point glucose levels.

The CGM data analysis followed the 2019 international consensus guidelines ^13^ and included assessments of %CV, standard deviation (SD), and AG. The percentage of time during which the glucose levels remained within 70–180 mg/dL (3.9–10.0 mmol/L) per day was defined as time in range (TIR), time below 70 mg/dL (<3.9 mmol/L) as time below range (TBR), and time above 180 mg/dL (>10.0 mmol/L) as time above range (TAR). Hypoglycemia was further categorized as level 1 (54–69 mg/dL, 3.0–3.8 mmol/L) and level 2 (<54 mg/dL, <3.0 mmol/L) in the guidelines. Thus, we defined level 1 hypoglycemia as glucose <70 mg/dL, and level 2 hypoglycemia as glucose <54 mg/dL for both 288‐point and 6‐point glucose levels.

This study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the institutional Ethics Committee (#UOEHCRB21‐105) and is registered at the UMIN Clinical Trials Registry with trial number UMIN000025433. Informed consent was obtained from all participants through an opt‐out process.

Statistical analysis

Continuous variables are presented as mean ± SD (standard deviation), and categorical variables as numbers (%). Cutoff values for level 1 hypoglycemia were determined using receiver operating characteristic (ROC) analysis with AG and %CV as explanatory variables, and the Youden index was used to identify the optimal cutoff points. AG below and %CV above the cutoff were defined as risk factors for hypoglycemia. The patients were then categorized into three groups based on the number of risk factors, and differences among the groups were analyzed using trend tests. The Jonckheere–Terpstra test was used for continuous variables, and the Cochran–Armitage test was used for categorical comparisons. All statistical analyses were performed using IBM SPSS Statistics 28.0 (IBM‐SPSS, Armonk, NY), with P < 0.05 set to imply the presence of statistical significance.

RESULTS

Clinical characteristics of the study participants

Table 1 summarizes the baseline characteristics of the 233 T2DM participants (males: 52.4%). Their mean age was 64.0 ± 12.6 years, and mean HbA1c was 9.0 ± 1.8% (74 ± 4.0 mmol/mol). The mean duration of diabetes was 15.0 ± 10.2 years, and more than half of the participants had diabetic complications. A total of 137 patients (59.0%) were treated with sulfonylurea and 103 (44.0%) with insulin. Furthermore, 27 (11.6%) were treated with basal insulin alone, 48 (20.6%) with bolus insulin (including basal bolus therapy), and 28 (12.0%) with premixed insulin. The most commonly used adjunctive medications were dipeptidyl peptidase‐4 inhibitors (51.0%), followed by biguanides (33.0%) (Table 1).

Comparison of original 288‐point and 6‐point glucose measurements

Figure 1 illustrates the distribution of glucose values recorded with the CGM (288 data points per patient per 24 h) and also the extracted 6‐point glucose values. In the 288‐point glucose measurements, 26 patients (11.2%) experienced level 1 hypoglycemia (<70 mg/dL) and 13 patients (5.6%) had level 2 hypoglycemia (<54 mg/dL). Among the 26 patients with level 1 hypoglycemia, 12 were insulin‐treated, 12 were sulfonylurea‐treated, and the remaining two were treated with both agents (28.6% of dual users experienced hypoglycemia). Hypoglycemia occurred predominantly at night (Figure 1).

In contrast, within the 6‐point measurements, only one individual (0.4%) experienced both level 1 and level 2 hypoglycemia; this individual was treated with premixed insulin without concomitant sulfonylurea therapy.

AG and glycemic variability metrics were similar between the AG was 173.5 ± 45.5 for the 288‐point data and 182.9 ± 46.1 mg/dL for the 6‐point data; SD was 40.0 ± 14.5 and 42.9 ± 16.7 mg/dL, respectively; and %CV was 23.6 ± 8.2% and 23.7 ± 8.4%, respectively. In patients who experienced level 1 hypoglycemia in the 288‐point measurements (n = 26), the corresponding 6‐point AG, SD, and %CV were 155.8 ± 31.1, 44.9 ± 11.2 mg/dL, and 29.8 ± 10.8%, respectively. For those with level 2 hypoglycemia (n = 13), the 6‐point AG was 148.7 ± 29.6 mg/dL, SD was 44.0 ± 12.9 mg/dL, and %CV was 30.7 ± 13.6%. To further evaluate the relationship between the 6‐point glucose metrics and CGM‐derived metrics, we performed Spearman's rank correlation analysis. The 6‐point AG correlated significantly with CGM‐derived AG (ρ = 0.883, P < 0.001), and the 6‐point %CV also correlated significantly with the CGM‐derived %CV (ρ = 0.634, P < 0.001) (Table S1).

Determination of cutoff values for 6‐point AG and %CV in predicting level 1 hypoglycemia (<70 mg/dL)

ROC analysis determined the optimal cutoff values of 167.7 mg/dL for AG (AUC = 0.70, 95% CI: 0.60–0.80) and 21.6% for %CV (AUC = 0.72, 95% CI: 0.64–0.80) for the prediction of level 1 hypoglycemia (<70 mg/dL) (Figure 2). In addition, fasting plasma glucose (FPG) was significantly lower in patients who experienced level 1 hypoglycemia (P < 0.001), and ROC analysis identified 144 mg/dL value as the optimal FPG cutoff value for detecting level 1 hypoglycemia (AUC = 0.72, 95% CI: 0.61–0.81). Furthermore, ROC analysis incorporating AG plus %CV showed improved discrimination, with an AUC of 0.77 (95% CI: 0.66–0.86) (Figure 2). Using these threshold values, AG <167.7 mg/dL and %CV >21.6% were identified as independent risk factors for hypoglycemia.

Subsequently, the patients were stratified into three groups based on the number of risk factors for hypoglycemia (Table 2). Among patients with zero risk factors (n = 57), level 1 hypoglycemia was noted in two cases (3.5%); among those with one risk factor (n = 117), six (5.1%) experienced hypoglycemia; and among patients with both risk factors (n = 59), 18 (30.5%) experienced hypoglycemia, with the incidence significantly increasing with increases in the number of risk factors (P < 0.001). Similarly, for level 2 hypoglycemia, the prevalence was 0% in patients with zero risk factors, 2.6% in those with one risk factor and 16.9% in those with both, also significantly increasing with increases in the number of risk factors (P < 0.001). The odds ratio for level 1 hypoglycemia was 1.4 (95% CI: 0.3–7.6) for one risk factor and 12.1 (95% CI: 2.7–55.0) for two risk factors, compared with the reference group (zero risk factors).

We also examined the relationship between hypoglycemic risk factors and medication use (Table 3). Patients treated with both insulin and sulfonylurea (n = 7) were excluded from this analysis due to a small sample size. Among sulfonylurea‐treated patients (n = 130), level 1 hypoglycemia occurred in 0% of patients with zero risk factors, 3.0% with one risk factor, and 29.4% with two risk factors (P < 0.001). Similarly, among insulin‐treated patients (n = 96), the incidence of hypoglycemia was 4.0% in those with zero risk factor, 8.5% in those with one risk factor, and 29.2% in those with two risk factors (P = 0.008). Similar patterns were observed for level 2 hypoglycemia, with significant increases in prevalence with increased number of risk factors (sulfonylurea, P = 0.012; insulin, P = 0.005).

DISCUSSION

This study demonstrated that AG and %CV values derived from 6‐point glucose measurements obtained with the CGM correctly assessed the risk of hypoglycemia in patients with T2DM treated with sulfonylureas or insulin. Furthermore, the concurrent presence of AG <167.7 mg/dL and %CV >21.6% was associated with a markedly increased risk of hypoglycemia.

Our findings indicate that the AG and %CV values calculated from the 6‐point glucose measurements obtained using the CGM serve as reliable indicators of hypoglycemia risk. In our previous study, we demonstrated that the AG and %CV derived from 96 to 288 glucose data points collected through the CGM correctly predicted the risk of hypoglycemia. In another study involving 294 hospitalized patients with T2DM, AG <152 mg/dL and %CV >22% from 24‐h CGM monitoring predicted hypoglycemia ^10^ . Similarly, another investigation involving 62 insulin‐treated T2DM patients reported that AG <150.9 mg/dL and SD >41.1 mg/dL were effective predictors of hypoglycemia (glucose <70 mg/dL) based on 24‐h CGM data ^14^ . Additionally, a study of 999 outpatients with T2DM identified AG <129.42 mg/dL and %CV ≥30.30% as risk factors for hypoglycemia (glucose <54 mg/dL), exceeding 1% of CGM monitoring time over a 14‐day period ^11^ . The present study evaluated the AG and %CV using 6‐point glucose data, which are commonly measured in clinical practice with SMBG, and demonstrated that lower AG and higher %CV values are risk factors for hypoglycemia.

SMBG is the cornerstone of diabetes management, particularly for preventing hypoglycemia in insulin‐treated patients ^15^ . However, SMBG has a lower sensitivity for detecting hypoglycemia than CGM ^16^ , and severe hypoglycemic episodes occur more frequently in users of SMBG than in those using CGM ^17^ . Consequently, proactive adoption of CGM is advised for high‐risk patients. Nevertheless, the specific cutoff values and optimal SMBG measurement frequencies remain inadequately defined. Our study is the first to demonstrate that AG and %CV values derived from 6‐point glucose measurements obtained with the CGM can be useful for evaluating the risk of hypoglycemia. Thus, incorporating CGM in patients meeting these criteria via SMBG are potentially useful for preventing unrecognized and severe hypoglycemic events by facilitating appropriate therapeutic interventions.

In this study, although approximately one in three patients with AG <167.7 mg/dL and %CV >21.6% experienced hypoglycemia, only one case (0.4%) was directly detected using the 6‐point glucose data. This discrepancy may be attributed to the predominance of nocturnal hypoglycemia, which is not adequately captured by daytime 6‐point glucose measurements. Consistent with previous reports on insulin‐treated T2DM ^14^ , hypoglycemic episodes frequently occurred between midnight and 7 a.m. without noticeable symptoms. Nocturnal hypoglycemia is influenced by several factors such as daytime physical activity, alcohol intake, medications, and diminished counter‐regulatory hormone responses in the early morning. Furthermore, autonomic responses to hypoglycemia are diminished during sleep, making the symptoms less noticeable ^18^ . Given the association of hypoglycemia with increased risks of cardiovascular events, mortality ^19^ , ^20^ , and arrhythmias ^21^ , preventing nocturnal hypoglycemia is imperative and clinically important. Because nocturnal hypoglycemia is often asymptomatic and difficult to detect using SMBG alone, patients treated with insulin or sulfonylurea with a high risk of hypoglycemia should be monitored actively for pre‐ and postprandial glucose levels using SMBG. If AG <167.7 mg/dL and %CV >21.6% are noted, CGM should be proactively considered to confirm and manage nocturnal hypoglycemia, regardless of the presence of symptoms or hypoglycemia detection with SMBG.

This study has several limitations. First, the inclusion of cases dating back over a decade resulted in a higher proportion of patients who had used mixed insulin formulations, which may not reflect current clinical practice. Second, as the study population consisted of hospitalized patients, factors such as early evening meals (at 6 p.m.) and prolonged overnight fasting might have contributed to the higher incidence of nocturnal hypoglycemia compared with outpatient settings. Third, the evaluation relied on glucose data extracted from the CGM, which may differ from the actual SMBG measurements. CGM readings correlate well with capillary blood glucose values in T2DM patients hospitalized in the general ward ^22^ . However, they can exhibit delays during rapid glycemic changes ^15^ and may have reduced accuracy in the hypoglycemic range ^8^ . Future prospective studies using SMBG are warranted to validate these findings in routine clinical practice.

In conclusion, our study demonstrated 30% prevalence of hypoglycemia among insulin‐ or sulfonylurea‐treated patients with T2DM, who presented with AG <167.7 mg/dL and %CV >21.6%, calculated from 6‐point glucose data recorded by the CGM. These findings suggest that the AG and %CV values, which can be easily obtained from pre‐ and postprandial glucose measurements commonly used in clinical practice, may be potentially useful for the detection of hypoglycemia risk.

AUTHOR CONTRIBUTIONS

All authors contributed to the study design and collection of clinical data. RS drafted the manuscript. YO, KT, FU, and SN collected, analyzed, and interpreted the data; reviewed and edited the manuscript; and approved the final version of the manuscript. YO and KT are the principal guarantors of this work; they have full access to all study data and take responsibility for the integrity of the data and the accuracy of the data analysis. All the authors have read and agreed to the publication of this manuscript.

DISCLOSURE

The authors declare no competing interests.

Approval of the research The study was approved by the Ethics Committee of the University of Occupational and Environmental Health, Japan (Approval No. UOEHCRB21‐105).

Informed Informed consent was obtained from all participants through an opt‐out process.

Registry and the registration no. of the study/ This study was registered with the UMIN Clinical Trials Registry (UMIN000025433).

Animal N/A.

Supporting information