Data collection

Data were collected on the duration of diabetes, history of lifetime episodes of severe hypoglycemia and diabetic ketoacidosis, previous glycated hemoglobin (HbA_1c) values, use of CGM systems, medications, comorbidities, and smoking.

All patients underwent physical examination, which included anthropometric measurement of weight, height, and body mass index (BMI). Blood pressure was measured with a sphygmomanometer. A daily insulin dose was defined as the requirement for insulin per kilogram of body weight per day.

Assessment of hypoglycemia awareness

Awareness of hypoglycemia was evaluated using validated questionnaires (the Clarke scale, the Gold scale, and the HypoA‑Q). The Clarke questionnaire consists of 8 questions assessing prior experience of hypoglycemia, such as a history of severe hypoglycemia and blood glucose levels at which patients begin to perceive the symptoms of hypoglycemia. The questionnaire generates a score (0–7) based on the patient’s responses.²⁰ The Gold questionnaire contains 1 question asking the respondents to report their experience of perceiving hypoglycemic events with a score ranging from 1 (always aware) to 7 (never aware) on the Likert scale.²¹ In the Clarke and Gold questionnaires, the score of 4 or higher indicates IAH, 3 indicates indeterminant awareness, and the score of 2 or lower indicates normal awareness of hypoglycemia.^20,21 HypoA‑Q is a 33‑item complex tool assessing the frequency and severity of hypoglycemic events, health care utilization, blood glucose thresholds at which the symptoms occur, awareness of symptoms, and frequency of checking blood glucose. HypoA‑Q is a profile measure, that is, it yields several item and subscale scores. The subscale of awareness of hypoglycemia of HypoA‑Q consists of 5 questions regarding the symptoms of hypoglycemia experienced in the past, with a score ranging from 0 (never) to 4 (always). The score is the sum of the points (0–20). Higher scores indicate greater impairment of hypoglycemia awareness.²² We received a written permission from the Mapi Research Trust to translate the HypoA‑Q into Polish and conduct our research. We performed the translation and linguistic validation of HypoA‑Q according to the Linguistic Validation Guidance of a Clinical Outcome Assessment. The Mapi Research Trust accepted our translation.

The end point, that is, IAH, was reached if indicated by one of the scales.

Laboratory tests

Eight milliliters of venous blood and a morning urine sample of 100–200 ml were collected from all patients after at least 8‑hour fast. Laboratory analyses were performed on the Cobas 6000 (Roche Diagnostic, Basel, Switzerald), Alinity ci‑series (Abbott Diagnostics, Lake Forest, Illinois, United States), Sysmex UC (Sysmex Corporation, Kobe, Japan), and Sysmex UF (Sysmex Corporation) devices. Whole‑blood HbA_1c level was determined by turbidimetry, and the levels of the other parameters were assessed in the serum. The enzymatic‑colorimetric method was used to determine the concentration of total cholesterol (TC), low‑density lipoprotein cholesterol (LDL‑C), high‑density lipoprotein cholesterol (HDL‑C), and triglycerides. The level of creatinine was determined by the Jaffe method, and estimated glomerular filtration rate was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation. The level of thyroid‑stimulating hormone was measured by electrochemiluminescence. The level of cortisol was measured using a chemiluminescence immunoassay.

Assessment of glycemic control

The glucometers (patients on self‑monitoring of blood glucose [SMBG]) and CGM data were downloaded to evaluate short‑term glycemic control. We analyzed mean glycemia, standard deviation (SD), and coefficient of variation in all patients. Additionally, in individuals utilizing CGM systems, we assessed the percentage of time in range 70–180 mg/dl, time below range (TBR), that is, below 70 mg/dl and 54 mg/dl, and time above range, that is, above 180 mg/dl.²³ We analyzed data from the last 4 weeks.

The skin content of advanced glycation end products (AGEs) was evaluated to assess long‑term glycemic control. The degree of accumulation of AGEs of proteins in the skin was noninvasively assessed using the fluorescent properties of the tissue. The skin autofluorescence (SAF) ratio was measured using the AGE‑Reader (DiagnOptics, Groningen, the Netherlands). The measurement was performed at the ventral side of the forearm. SAF was measured 3 times in each patient. The result was the arithmetic mean of these 3 measurements.²⁴

Statistical analysis

We compared the 2 groups of patients utilizing different statistical tests based on the scale of measurement for the data. For the interval or ordinal scale data, we employed the Mann–Whitney test, as the data did not follow the normal distribution according to the Shapiro–Wilk test. The results were presented as medians and interquartile range (IQR). When the data were on a nominal scale, we used either the χ² test or the Fisher exact test, depending on the sample size and the expected frequencies. The results were presented as numbers and relative frequencies. Univariable and multivariable logistic regression models were used to assess the influence of significant variables on the occurrence of IAH. For the multivariable analysis, we utilized the least absolute shrinkage and selection operator (LASSO) logistic regression to identify the most significant variables among all those considered. The LASSO regression method was chosen for its ability to perform both variable selection and regularization. This approach prevents overfitting by shrinking the coefficients of less important variables toward 0 and retaining only those variables that contribute meaningfully to the model. Consequently, the LASSO regression is particularly advantageous when dealing with datasets with many potential predictors, as it simplifies the model while maintaining its predictive accuracy.

The Cohen κ coefficient was used to assess the agreement between the studied scales for IAH. To determine the optimal cutoff for the HypoA‑Q scale, we performed receiver operating characteristic (ROC) curve analysis, using the Clarke scale as the gold standard for determining the presence of IAH. We used a threshold of 4 points on the Clarke scale to indicate IAH. The optimal cutoff was identified using the Youden index. The potential prognostic property of the HypoA‑Q scale was assessed by the area under the curve (AUC). Sensitivity and specificity for the cutoff point were also provided.

The analysis was performed using Statistica package version 10 (StatSoft Inc., Tulsa, Oklahoma, United States) and MedCalc Statistical Software version 15.6.1 (MedCalc Software bvba, Ostend, Belgium). All tests were considered significant at a P value below 0.05.

Ethics

All patients signed their written informed consent to participate in the study. The study was approved by the local Bioethics Committee of the Poznan University of Medical Sciences (795/22). The study is registered at the ClinicalTrials (NCT05620927).

DISCUSSION

The primary finding of our study is the confirmation of the usefulness of HypoA‑Q, a new self‑report questionnaire designed to assess awareness of hypoglycemia. The traditionally used scales, such as the Clarke, Gold, and Pedersen‑Bjergaard ones, have limitations.²⁵ There are no clear guidelines on the most accurate noninvasive tool for assessing hypoglycemia awareness status, and none of the methods widely used to date are entirely reliable. Rubin et al²⁶ suggested using 2 questionnaires (the Clarke and the Gold) for better classification accuracy. In their study of 78 adults with T1D, the prevalence of IAH was 33% and 44%, according to the Clarke and Gold scales, respectively. Among the patients who completed both questionnaires, 32% were classified inconsistently, receiving a diagnosis of IAH in only one of them. Additionally, the authors found that IAH diagnosed with the Clarke and the Gold methods poorly predicts the counterregulatory response to hypoglycemia in patients with T1D.²⁶ We observed a comparable occurrence of IAH diagnosed by the Clarke and the Gold scales in the patients with T1D, consistent with findings from other studies.^25,27 The frequency of IAH was 19% and 23%, according to the Clarke and the Gold scales, respectively. Only 15% of the patients were classified inconsistently. Similar to the findings of Rubin et al,²⁶ the Clarke scale detected IAH in fewer people than the Gold scale.

Speight et al²² designed HypoA‑Q as a reliable and accurate tool for the noninvasive assessment of hypoglycemia awareness status. Although they preliminarily validated this questionnaire in patients with T1D, they did not determine the cutoff for indicating IAH. They obtained similar scoring results in HypoA‑Q, that is, the participants with IAH, according to the Gold scale, scored an average of 10.8 (3.42) points, while those without IAH scored 5.09 (3.2) points. In our research, the scores were 10.53 (3.37) points and 3.63 (2.29) points, respectively.

To our knowledge, this study is one of the first to compare the Clarke, the Gold, and the HypoA‑Q questionnaires for assessing hypoglycemia awareness in the patients with T1D, and the first to utilize HypoA‑Q in the Polish population. We found that the scales above detected IAH in different patients, with the Gold scale indicating IAH more frequently than the Clarke scale, which was previously shown.²⁶ Assessing awareness of hypoglycemia based on a single method (Clarke or Gold scale) seems insufficiently reliable.²⁸ In our study, HypoA‑Q with a 9‑point cutoff, showed the highest sensitivity in diagnosing IAH, even identifying patients with undetected IAH according to the Clarke and the Gold scales.

Our observations contrast with a study by Matus et al,²⁹ who were the first to propose a diagnostic threshold of 12 points to recognize IAH using HypoA‑Q. In their study, a group of 21 adults with T1D (men, 48%; median age, 36 years; and median T1D duration, 21 years) completed the HypoA‑Q, the Clarke, and hypoglycemia severity questionnaires, and underwent continuous glucose monitoring and hyperinsulinemic‑hypoglycemic clamp testing. Their findings also suggest the Clarke scoring should be modified to 5 or more points to identify IAH. We used a commonly applied and validated threshold of 4 points, resulting in lower thresholds for both the Clarke scale and HypoA‑Q in our analysis. Additionally, the population characteristics were different in our study and that of Matus et al.²⁹ Their IAH group was older (51 years), had a longer duration of diabetes (38 years), and exhibited a lower HbA_1c value (6.6%) than our study group. According to our data and other research, these factors might be associated with greater frequency of IAH.⁶ The limitation of the cited study was a small number of patients, but its strength was using hyperinsulinemic‑hypoglycemic clamp testing.²⁹

Our results are also consistent with recent data presented in adults with type 2 diabetes (T2D). Henao‑Carrillo et al³⁰ evaluated the validity of HypoA‑Q for assessing awareness of hypoglycemia in 406 insulin‑treated patients with T2D. They obtained a similar median (IQR) HypoA‑Q score of 6 (3–7) points as in our study (6 [3–9] points). They proposed a cutoff of 10 points in HypoA‑Q to diagnose IAH, whereas our data suggest a diagnostic threshold of 9 points. However, we examined individuals with T1D, not T2D. The limitation of their research was the exclusive use of the Clarke scale to assess IAH.³⁰

Our work included a large, homogenous group of patients with T1D from Poland, suggesting that the findings may be generalizable to a broader population with T1D. We demonstrated in the multiple regression analysis that the presence of IAH is related to longer diabetes duration. This is consistent with the findings of Olsen et al,³¹ who reported that the intensity of autonomic symptoms and awareness of hypoglycemia decline with increasing duration of T1D. They suggested that subjective symptoms of hypoglycemia change over time.³¹ Likewise, Stefanon et al³² reported that the occurrence of IAH increases with a longer duration of diabetes. This is a significant clinical finding, confirming the need for repeated education of patients with long‑term T1D as well as increased alertness to the risk of hypoglycemia.

Our study also showed that the patients with IAH were more likely to have hypertension. Previous research demonstrated an association between hypoglycemia and elevated blood pressure. Severe hypoglycemia triggers counterregulatory mechanisms, such as the activation of the sympathoadrenal system, which leads to systolic blood pressure elevation.³³ Feldman‑Billard et al³⁴ supposed that hypoglycemia‑induced hypertension is more likely to occur in individuals with recurrent and severe hypoglycemia. Additionally, hypertension is a risk factor for both microvascular and macrovascular complications of diabetes.³⁵ Therefore, more health abnormalities can be expected in the group of patients with hypertension.

Other findings indicated that the patients with IAH had lower LDL‑C and non–HDL‑C levels. Comparable results were presented in the DCCT (Diabetes Control and Complications Trial) and confirmed in later studies, showing a positive correlation between poor glycemic control and dyslipidemia.^36-40 Poor glycemic control leads to quantitative and qualitative changes in lipid profiles, resulting in atherogenic lipid patterns. Higher HbA_1c values are associated with increased TC, LDL‑C, and triglyceride levels.^36-40 In our study, the individuals with IAH had lower HbA_1c levels, corresponding to lower LDC and non–HDL‑C levels.

Patients with T1D have a higher risk of developing other autoimmune diseases, the most common being hypothyroidism.^41,42 This endocrinopathy disrupts the counterregulatory response to hypoglycemia. Mohn et al⁴³ reported that even subclinical hypothyroidism is linked to an increased risk of hypoglycemia. Moreover, hypothyroidism can cause cognitive impairment, potentially affecting recognition of hypoglycemic symptoms. In line with these findings, we observed that IAH was related to hypothyroidism, highlighting the importance of screening for thyroid dysfunction in patients with T1D.

In our study group, most patients were treated with intensive insulin therapy as per current guidelines.⁴⁴ The benefits of this model of insulin therapy were demonstrated by the DCCT and the EDIC (Epidemiology of Diabetes Interventions and Complications) follow‑up study, showing better glycemic control and a lower risk of microvascular complications of diabetes in comparison with the patients receiving a conventional insulin treatment.^45,46 Despite the proven advantages, not every patient is able or willing to be treated with intensive insulin therapy. In the current analysis, only 7% of the participants used conventional insulin therapy. However, fewer patients with than without IAH were treated with intensive insulin therapy, reinforcing the idea that this method leads to better treatment outcomes.

Similarly to the findings of Matus et al,²⁹ our study showed that the individuals with IAH required less insulin per day, suggesting that factors beyond insulin dosage alone play a significant role in the awareness of hypoglycemia.

The patients with IAH reported more lifetime episodes of severe hypoglycemia, which is understandable given that recurrent exposure to hypoglycemia can lead to the development of IAH. Our observations align with those published previously.^25,31 Additionally, we found that the episodes of severe hypoglycemia were associated with a 31% higher risk of diagnosing IAH. Recurrent hypoglycemia may generate a vicious cycle, where the recent episode weakens the body’s counterregulatory mechanism. This phenomenon, known as hypoglycemia‑associated autonomic failure, reduces the sympathoadrenal response to subsequent hypoglycemia due to decreased epinephrine response and reduced awareness of hypoglycemia through impaired sympathetic neural response.⁴⁷ In our study, the presence of IAH was related to lower glycemia, SD, and HbA_1c values. This highlights the need for regular patient education, even for those with strict glycemic control, to prevent hypoglycemia and ensure safe insulin therapy practices. A previous study by Lin et al⁴⁸ showed that IAH is also related to TBR below 54 mg/dl.⁴⁸ Our data suggest that low blood glucose may play a significant role in the development of IAH. Ali et al⁴⁹ noticed that widespread use of CGM might have a positive effect on reducing the occurrence of IAH in T1D patients. However, in our research, only 37% of the participants utilized CGM, which may have affected our results. Our study’s timing coincided with the introduction of CGM reimbursement in the Polish health care system, which was a confounding factor. Prior to the reimbursement, the cost was a limiting factor for many patients. According to the guidelines, CGM should be used for every patient with T1D, especially those with IAH.⁴⁴ It was previously reported that hypoglycemia awareness can be restored by avoidance of hypoglycemia. Sepulveda et al⁵⁰ found that while CGM reduces severe hypoglycemia, it does not improve endogenous hypoglycemia awareness.⁴ Similarly, Lin et al⁴⁸ showed that despite using CGM, IAH remains an important clinical concern in T1D patients.

Interestingly, despite better glycemic control in the individuals with IAH, they had a significantly higher accumulation of AGEs in the skin; however, only on the Clarke scale. These data suggest that the participants with IAH had poorer diabetes control in the past, as SAF reveals long‑standing glycemic control over a much longer period than the HbA_1c value. These findings align with prior research linking AGEs to the occurrence of both micro- and macroangiopathic complications of diabetes.⁵¹ This may be related to metabolic memory and the negative long‑term impact of poor glycemic control, even after improvements have been made.⁵² Thus, good glycemic control from the onset of diabetes may be essential in preventing IAH.

Our study has some limitations, as it did not verify the assessment of awareness of hypoglycemia by hyperinsulinemic‑hypoglycemic clamp that might have increased the sensitivity of diagnosing IAH.

To conclude, IAH presents a significant challenge for individuals with T1D. Our study highlighted the variability in IAH detection across different self‑report questionnaires. To our knowledge, this is one of the few studies proposing a threshold for diagnosing IAH with HypoA‑Q. The HypoA‑Q 9‑point cutoff demonstrated the highest sensitivity in diagnosing IAH. Thus, HypoA‑Q can be considered the most valuable screening test for IAH. Our study confirmed that the presence of IAH is related to multiple factors. Further research is needed to determine the optimal HypoA‑Q cutoff and identify additional risk factors for IAH.