Background
Diabetes mellitus is a major cause of morbidity and mortality worldwide, as a result of its impact on cardiovascular system, eyes, kidneys, and nerves [
1]. Optimizing blood glucose control is demanding, since it requires balancing the need for glycemic control with the risk of hypoglycemia [
2‐
4]. In fact, the risk of hypoglycemia is a barrier to optimal treatment of type 1 diabetes (T1DM) and type 2 diabetes (T2DM), especially within the context of insulin therapy [
5].
Hypoglycemic episodes may manifest in different ranges from asymptomatic to severe neurological symptoms, like dizziness, confusion, weakness and loss of consciousness [
6‐
8]. Recurrent asymptomatic episodes characterizes hypoglycemia unawareness, which has been shown to increase the risk of severe hypoglycemic episodes [
6,
9]. Moreover, patient fear of hypoglycemia is frequent and may also affect treatment adherence and, thus, glycemic control [
6,
10].
Some observational studies have reported that individuals with T1DM experience about 42.0–136.8 non-severe hypoglycemic events per patient-year [
11,
12]. Regarding T2DM, non-severe hypoglycemia seems to occur less frequently (0.2 to 48.0 events per patient-year) but, like in T1DM, may increase with longer duration of insulin therapy [
11,
12]. Nonetheless, most studies had a retrospective/cross-sectional design, and are almost limited to North American and European health contexts [
12].
The Hypoglycemia Assessment Tool (HAT) study was a multinational non-interventional study that included 27,585 adult T1DM or T2DM patients, from 24 countries in six world regions. The study results have shown that, during a 4-week prospective period, 83.0% T1DM patients and 46.5% T2DM patients had at least one hypoglycemic episode [
12]. However, the study initially did not include Brazil, one of the top 10 countries in number of people living with diabetes: it was estimated that 11.9 million Brazilian adults were living with diabetes in 2013, which correspond to a prevalence of 11.9% [
13].
Recognizing the scarce data about hypoglycaemia and how this affects diabetes control and management, the HAT study was implemented in Brazil to determine the proportion of patients experiencing hypoglycemic events, to characterize patient awareness, fear and attitudes towards hypoglycemia, and to estimate health resource use and costs of managing hypoglycemic events among patients with T1DM or insulin-treated T2DM.
Methods
Patients and study design
This was a national, non-interventional, multicenter study, with a 6-month retrospective and a 4-week prospective evaluation of hypoglycemic events. The primary endpoint was defined as the percentage of patients experiencing at least one hypoglycemic event during the 4-week prospective period. Included patients were ≥ 18 years-old and had T1DM or T2DM treated with insulin for more than 12 months. Patients were excluded from the study if followed-up at non-ambulatory setting, or not able to answer to the written survey, or if participating in any interventional clinical study. All participants provided their written informed consent prior to any study-related activity. Eligible participants were consecutively enrolled from 21 centers in Brazil, when attending to a routine medical appointment with a diabetes specialist, between October 2014 and February 2015.
Data collection
At baseline, the first questionnaire recorded baseline demographic and treatment information, and evaluate patient knowledge, hypoglycemia unawareness and perceptions of hypoglycemia over the last 6 months. History of severe hypoglycemic events during the same period was also collected, as well as history of symptomatic events over the last 4 weeks. A patient diary was provided to each participant, to register hypoglycemic events during the prospective period. A second questionnaire evaluated the history of both severe and symptomatic hypoglycemia over the 4 weeks after baseline. When a patient recorded more hypoglycemic events in the diary than during the second questionnaire, the patient diary information was used to estimate prevalence and hypoglycemia rates.
Hypoglycemia definitions
Severe hypoglycemia was defined according to the American Diabetes Association definition, as any hypoglycemic event requiring assistance of another person to actively administer carbohydrate, glucagon, or other resuscitative actions [
14]. Non-severe hypoglycemia was defined as events managed by the patient alone, and “any hypoglycemia” resulted from the sum of severe and non-severe hypoglycemia. Other definitions used in this study were nocturnal hypoglycemia (hypoglycemia occurring between midnight and 06:00 h) and hypoglycemia requiring hospital admission.
Hypoglycemia unawareness and fear of hypoglycemia
Hypoglycemia unawareness was evaluated with the question ‘Do you have symptoms when you have a low sugar level?’, where ‘always’ and ‘usually’ denoted normal, ‘occasionally’ denotes impaired, and ‘never’ corresponded to severely impaired awareness. Fear of hypoglycemia was reported on a 10-point scale, from ‘not afraid at all’ to ‘absolutely terrified’.
Use of health resources and costs
Self-reported use of healthcare resources, related with hypoglycemic events, was collected from the patient diary, with data from the 4-week prospective period. The economic analysis assumed the societal perspective, with the estimates of direct cost based on the healthcare resources required to manage hypoglycemic events (i.e. hospitalization, emergency assistance, extra visits to physician/nurse and outpatient administration of carbohydrates or glucagon), and indirect costs relate with labor absenteeism due to hypoglycemia (i.e., sick leave and delays to work) calculated considering the Capital Human Approach using Brazilian minimum monthly income by gender and age class. Prices for health resources were obtained from official Brazilian sources DATASUS—Base de dados ambulatorial e hospitalar 2014 (hospitalization and emergency visits), SIGTAP—Sistema de Gerenciamento da tabela de Procedimentos, Medicamentos e OPM do SUS (medical visits) and BPS—Banco de Preços em Saúde (use of glucagon). Yearly costs were extrapolated based on the prospective assessment, in Brazilian Reais, for the year 2015.
Statistical analysis
Sample size was calculated based on the primary endpoint of the study. The inclusion of at least 675 patients would enable the estimation of 95% confidence interval (CI) for the percentage of patients experiencing at least one hypoglycemic event during the observation period, with a maximum precision error of 4% and assuming a drop-out rate of 11%.
The incidence of hypoglycemic episodes during prospective period was defined as the percentage of patients experiencing at least one hypoglycemic event during the 4-week observational prospective period. Patients with hypoglycemic events were identified if having reported at least one hypoglycemic event in either the second questionnaire or the patient diary. The incidence of hypoglycemic events, by diabetes type (T1DM or T2DM), was calculated with corresponding 95% CI assuming a binomial distribution. For secondary endpoints, the incidence of various types of hypoglycemia was calculated as the number of events per patient year, as expressed by [total number of events/total follow-up time (patient years)]. The frequency of patients with at least one severe, non-severe, any, nocturnal and hospital-requiring hypoglycemic events, are also presented by diabetes type, and stratified by glycated hemoglobin (HbA1c) at baseline and by age (< 65 years old versus ≥ 65 years old). In addition, the Mann–Whitney test was performed to examine the association between HbA1c and the occurrence of severe hypoglycemia.
To examine the relationship between hypoglycemia and eventual predictors, negative binomial regression models were defined by specifying a log-transformed exposure time offset term and adjusting (stepwise) for all variables in the model, namely: age (years), gender, HbA1c (%), type of diabetes, duration of diabetes (years), duration of insulin therapy (years), type of insulin therapy, frequency of blood glucose testing, hypoglycemia unawareness, knowledge of hypoglycemia before baseline, fear of hypoglycemia, study period (prospective/retrospective) and diabetes type. No imputation of missing data was performed. The negative binomial regression model for the hospital-requiring hypoglycemic events was not presented, as some convergence criteria were not satisfied.
Of those patients who had ever experienced hypoglycemia, the number and proportion with normal, impaired, and severely impaired hypoglycemia awareness, were summarized by diabetes type and by whether or not patients have experienced severe hypoglycemia in the 6 months before baseline. A logistic regression model was performed to examine the relationship between the above mentioned predictor variables and the odds of having an impaired or severely impaired hypoglycemia unawareness.
Patients’ fear of hypoglycemia was also analyzed by diabetes type and whether or not patients have experienced severe hypoglycemia in the 6 months before baseline. The number and proportion of patients who, as a result of fear of hypoglycemia, have and have not consulted their doctor/nurse, increased calorie intake, avoided physical exercise, reduced insulin doses, skipped insulin injections and increased blood glucose monitoring was also determined.
The study was conducted in accordance with the principles of the Declaration of Helsinki and of the Guidelines for Good Pharmacoepidemiology Practices. The study protocol and patient informed consent documents were reviewed and approved by the ethics committees of all study centers.
Discussion
The HAT study in Brazil evaluated the frequency and incidence rate of hypoglycemia in a cohort of patients with insulin-treated diabetes mellitus. It was found that 91.7% T1DM patients and 61.8% T2DM patients had at least one hypoglycemic event during the 4 weeks after baseline, a short period of observation. These frequencies were higher than those observed at the global HAT study, where, overall, 83.0% T1DM and 46.5% T2DM patients had at least one hypoglycemia in the prospective period [
12]. In addition, hypoglycemic rates observed in Brazil were even higher that those reported by the global HAT study for the Latin America region (87.4% T1DM and 43.8% T2DM).
The incidence of severe hypoglycemia was higher in the present study, with 25.7% T1DM patients and 13.4% reporting at least one event during the prospective period. In the global study, the proportions were 14.4% and 8.9%, respectively [
12]. The same trend was observed for the incidence rates: 4.9 and 2.5 severe events per patient-year in the global study vs. 9.8 and 6.2 severe events per patient-year in this study, for T1DM and T2DM patients, respectively. When considering nocturnal episodes, Brazilian patients also reported higher incidence rates (23.6 and 6.1 events/patient-years among T1DM and T2DM, respectively) compared to the rates observed in the global study (11.3 and 3.7 events/patient-years among T1DM and T2DM, respectively).
The higher hypoglycemia frequencies that were observed, especially among T2DM patients, may reflect differences in diabetes management, even between countries from the same geographic region. A cross-sectional study that included T1DM Brazilian patients reported that 96.8% experienced hypoglycemia at least once in the last 3 months and that 89.6% had at least one hypoglycemic event in the last month [
15].
Predictors of having any hypoglycemia included the higher frequency of blood glucose monitoring, longer duration of insulin therapy, and T1DM. Longer duration of insulin therapy has been described as a predictor of hypoglycemic events, and may reflect the need for treatment adjustment that occurs with the natural course of the disease and counter regulatory hormonal deficiencies related to pancreatic beta-cell failure [
8,
16]. The higher frequency of blood glucose monitoring was also associated with non-severe hypoglycemic events, probably because some of these events were asymptomatic and only identified by blood glucose measurement.
At baseline, about 20% T1DM patients and 35% T2DM patients had impaired or severely impaired hypoglycemia awareness. We hypothesize that T1DM patients could be more knowledgeable about hypoglycemia and monitor glycemia more frequently, even when asymptomatic. Hypoglycemia unawareness was associated with lower rates of reported nocturnal hypoglycemia, which should be further evaluated since prolonged nocturnal events can lead to disturbances of cardiac function [
17]. Unlike other studies that report an association between the hypoglycemia unawareness with older T2DM patients [
9,
18], the results herein obtained suggested that older patients might have a slightly decreased hypoglycemic unawareness than younger patients. Nevertheless, further studies are necessary in order to clarify this association, while adjusting to other confounding variables that may have not been considered in the present study.
Similar to other studies, fear of hypoglycemia was associated with higher risk of severe hypoglycemia, most probably due to reverse causality of previous severe events [
12]. Overall, patients from our study seem to experience more fear than that observed in the global study, with mean scores of 5.9 vs. 4.7 for T1DM patients, and 5.4 vs. 4.4 for T2DM patients [
19]. In the prospective period, the more frequent reaction following hypoglycemic events was to increase blood glucose monitoring (63.8% T1DM patients and 50.9% T2DM patients). In the global study, this reaction was also the most frequent among T1DM patients (69.7%), but requiring any form of medical assistance was the most frequent reaction among T2DM patients (62.0%) [
19]. Furthermore, our study showed that 91.6% T1DM patients with a previous severe hypoglycemia reported the reduction of insulin doses, a reaction that may compromise glycemic control.
Estimated costs for hypoglycaemias in T1DM patients are almost two times the costs in T2DM, most probably due to higher incidence of hypoglycaemias and need of hospitalizations [
20]. Of notice, results of the HAT study for Central and Eastern European countries have reported that T1DM patients had less hospitalizations due to hypoglycaemia compared to T2DM patients (1.2% vs. 2.1%) [
21]. Another study from Canada also confirm hospital admissions as the main cost driver but reported a higher cost of managing hypoglycemias among T2DM patients, as a result of more medical appointments and more lost working days in this group [
22]. These observations, alongside with the high incidence rates of hypoglycemia, reinforce the need of optimizing patient access to healthcare, in order to achieve a better tailored insulin regimen and glycemic control and avoid the need of hospitalizations.
As a non-interventional self-reported study, there were some limitations such as potential misunderstanding of questions, and erroneous self-perception of hypoglycemic events by the patient, besides the possible recall bias, especially for baseline questions related to the previous 6 months. Missing data was frequent in some questions, namely when patients who reported having hypoglycemic events in the patient diary did not report the number of events that occurred in second questionnaire. When discrepancies were found between patient diary and second questionnaire regarding the registries of hypoglycemia, it was assumed the highest frequency stated on either of these forms to estimate frequency and incidence rate but the higher reporting in the patient diary may be partially explained by an increased awareness about hypoglycemia and the reminder effect of the patient diary.
The HAT study methodology enables the comparison of data across several regions and countries, thus providing real-world information about hypoglycemia rate and risk factors. The self-reporting of hypoglycemia, although prone to misunderstanding and overestimation of hypoglycemia, provides the collection of events that patients could, otherwise, forgot to report or neglect due to the absence of symptoms. The evaluation of several patient-centered dimensions, such as hypoglycemia awareness and fear, is expected to support clinicians on tailoring insulin treatment and improving patient education and diabetes management.
Authors’ contributions
RNL was responsible for data collection and study supervision and made critical revisions of the manuscript. BG, SOL, RM, ARC and LZ were responsible for data collection. MF drafted the manuscript and reviewed statistical analysis. FOG and MNE were responsible for study implementation and supervision. All authors read and approved the final manuscript.
Acknowledgements
The authors acknowledge the study participants and investigators, and Eurotrials Scientific Consultants for data management, statistical analysis (Helga Garcia), economic analysis (Catarina Silva) and medical writing assistance.
The HAT Brazil study group
Rodrigo Nunes Lamounier (Centro de Diabetes de Belo Horizonte, Belo Horizonte, Brazil), Bruno Geloneze (LIMED – UNICAMP, Campinas, Brazil), Silmara Oliveira Leite (Hospital Cruz Vermelha - Curitiba. Centro de Diabetes de Curitiba, Curitiba, Brazil), Renan Montenegro Jr. (Faculdade de Medicina da Universidade Federal do Ceará, Fortaleza, Brazil), António Roberto Chacra (Departamento de Medicina – UNIFESP, São Paulo, Brazil), Lenita Zajdenverg (Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil), Milene Fernandes (Eurotrials, Scientific Consultants SA, Lisboa, Portugal), Fabiano de Oliveira Griciunas, Mariana Narbot Ermetice (Novo Nordisk A/S, Brazil), Adriana Costa e Forti (Centro de Estudos em Diabetes e Hipertensão - Secretaria de Saúde do Estado do Ceará, Ceará, Brazil), Ana Priscila Soggia (Instituto de Ensino e Pesquisa - Hospital Sírio-Libanês, São Paulo, Brazil), Daniela Espíndola Antunes (Hospital das Clínicas da Universidade Federal de Goiás, Goiás, Brazil), Flavia Coimbra Pontes Maia (Santa Casa de Belo Horizonte; NUPEM - Núcleo de Pesquisa em Endocrinologia e Metabologia, Minas Gerais, Brazil), Giselle Fernandes Taboada (Unidade de Pesquisa Clínica - Hospital Universitário Antônio Pedro - UFF, Rio de Janeiro, Brazil), Hugo Roberto Kurtz Lisboa (Hospital São Vicente de Paulo, Rio Grande do Sul, Brazil), Jorge de Faria Maraschin (Hospital Nossa Senhora da Conceição, Santa Catarina, Brazil), Jorge Luiz Gross† (Centro de Pesquisas em Diabetes, Rio Grande do Sul, Brazil), Manuel dos Santos Faria (Centro de Pesquisas Clínicas - Hospital Universitário da UFMA, Maranhão, Brazil), Márcio Antonio Pereira (Instituto São José dos Campos em Pesquisas Médicas – ISPEM, São Paulo, Brazil), Maria Elizabeth Rossi da Silva (Centro de Pesquisas Clinicas - ICHC – FMUSP, São Paulo, Brazil), Miguel Nasser Hissa (Centro de Pesquisa em Diabetes e Doenças Endocrino-Metabólicas, Ceará, Brazil), Mirian Takahashi (CLIPEM - Clinica e Pesquisa em Endocrinolgia de Maringá, Paraná, Brazil), Rosane Kupfer (Instituto Estadual de Diabetes e Endocrinologia Luiz Capriglione, Rio de Janeiro, Brazil), Tania Maria Bulcão Lousada Ferraz (Instituto de Estudo e Pesquisa – HGF, Ceará, Brazil).