Persistence with Basal Insulin and Frequency of Hypoglycemia Requiring Hospitalization in Patients with Type 2 Diabetes

Due to the progressive nature of type 2 diabetes (T2DM), initiation of treatment with insulin is a necessary step towards achieving optimal glycemic targets. It is well established that an adequate glycemic control reduces diabetes-related microvascular and macrovascular complications [1, 2]. The American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) mandate an early initiation of basal insulin therapy in patients with T2DM with glycated hemoglobin (HbA1c) > 11% associated with hyperglycemia symptoms and/or evidence of ongoing catabolism, such as weight loss, polyuria and polydipsia [3]. The most recent French guidelines recommend insulin therapy when treatment with oral antidiabetic drugs (OADs) and non-insulin agents have failed to achieve glycemic targets [4]. However, only 25% of all patients treated for diabetes in France are receiving insulin [5]. Therapeutic inertia is observed at every step of the management of T2DM, but this delay in intensification is even worse when insulin therapy is being considered for insulin-naïve patients [6].

An additional difficulty is that patients with T2DM are often challenged when attempting to manage their antidiabetic medications, including insulin [7]. Such difficulties may result in a low level of persistence to the insulin therapy, with this persistence defined as the proportion of patients who remain on treatment for a specific time or the duration of time from initiation to discontinuation of therapy [8]. A previous French study using the same methodology as the above-mentioned study showed that among survivors, only 75% were still treated with insulin after 12 months (81.6% for patients treated with a basal scheme) [9]. In Germany, a study conducted in the primary care setting among patients with T2DM who were initiating basal insulin supported oral therapy [10] found that 2-year persistence was 65, 53 and 59% in users of glargine 100 U/mL (Gla-100), detemir (IDet) and neutral protamine hagedorn (NPH), respectively (p < 0.001). In the setting of intensified conventional therapy, persistence was higher without differences between groups: 84, 85 and 86% in those on Gla-100, IDet and NPH, respectively (p = 0.536) [10]. Several other studies have been conducted on insulin persistence, mainly in USA. Wei et al. reported a lower rate (65%) of persistence with Gla-100 or IDet within 1 year after initiation [11]. Baser et al. conducted a retrospective study using a large claims database in the USA and found that 55.9% of patients remained on Gla-100 within 1 year after treatment initiation [12].

To date, few real-world studies have been conducted on the recently marketed second-generation, long-acting insulins, such as glargine 300 U/mL (Gla-300). Gla-300 was developed to extend the duration of action achieved with first-generation basal insulin analogues [13] and reduce the risk of hypoglycemia in patients treated with insulin. Gla-300 has a flatter and more extended time-action profile compared to Gla-100, leading to a more stable and sustained glycemic control over a 24-h period [13]. Phase III clinical trials have demonstrated a lower incidence of hypoglycemia in patients with T2DM treated with Gla-300 [14] compared to those treated with Gla-100 with a basal supported oral therapy [15] or those using basal and mealtime insulin [16]. It is well known that hypoglycemia or the fear of hypoglycemia are major barriers to the initiation and optimal use of insulin [17].

The study reported here was conducted roughly 2 years after the launch of Gla-300 in France. It is one of the first studies using the large French comprehensive national healthcare system claims databases to assess persistence to this new-generation basal insulin and the frequency of hospitalizations for severe hypoglycemia among patients with T2DM in a real-life setting versus first-generation basal insulin analogues (branded or biosimilar Gla-100 combined and Idet). Of note, at the time of the study, degludec was not marketed in France.

A retrospective analysis was conducted using prescription claims data from the SNDS (Système National de Données de Santé), a French administrative healthcare database covering 99% of the French population (66 million inhabitants, 3.5 million with diabetes). Patients who had initiated basal insulin therapy ± OAD or GLP-1 RA from January 2016 to December 2017 were included in order to document treatment persistence without discontinuation.

The SNDS is the main French prescription claims database and includes information from other systems, such as the national hospital discharge database (PMSI). It is one of the largest health databases in the world, with data on more than 1.2 billion reimbursed care entries, 500 million medical procedures and 11 million hospital stays per year. The SNDS database contains data on all reimbursed healthcare expenditure (inpatient, outpatient and cash payments) for the entire population living in France. It also includes sociodemographic, medical and administrative data on these beneficiaries (age, gender, diagnoses of long-term diseases eligible for 100% reimbursement, diagnoses reported during hospitalizations, town of residence, date of death) [18]. This study was authorized through two French legal Ethic Committees: Expertise Committee for Research, Studies and Evaluations in the field of Health (CEREES Dossier TPS 37634bis, approval on 12 April 2018) and the National Commission for Data Protection and Liberties (CNIL-France No. 918140 approval on 31 July 2020). Informed consent was not required due to the retrospective anonymized nature of this study.

This retrospective study was designed to assess persistence with basal insulin therapy by patients with T2DM following treatment initiation using data from the French national insurance claims database, SNDS, that covers the entire French population. Our analysis revealed that when all insulin regimens were considered, 72% of patients continued to use basal insulin therapy after 12 months; this percentage increased to 75% of patients when only those on the basal insulin-only scheme were considered. Insulin persistence in the patients in our study is very similar to that observed in the same database in France in 2012/13 [9], but is generally better than results reported from other countries. Wei et al. reported 65% persistence with insulin glargine or IDet within the first year after initiation [11]. Baser et al. conducted a retrospective study using a large claims database in the USA and found that 55.9% of patients remained on insulin glargine within 1 year after treatment initiation [12]. Ascher-Svanum et al. conducted a study using a US Commercial Claims and Encounters database and reported that only 18% of patients continued their insulin therapy after 1 year [20]. These differences across studies could be explained by the lack of homogeneity in defining and measuring persistence with insulin therapy as well as by differences in healthcare systems.

Our results from the adjusted analyses show that persistence with Gla-300 was significantly better than that IDet and Gla-100. This result may be partly explained by a lower rate of hypoglycemic events in patients treated with Gla-300, a possibility supported by Bolli et al. who reported a lower incidence of nocturnal hypoglycemia with Gla-300 compared to Gla-100 in the EDITION 3 study [21]. A similar finding was also reported by Yki-Järvinen et al. in the EDITION 2 trial comparing the same treatments [15]. Interestingly, lower nocturnal hypoglycemia with Gla-300 was associated with a significant improvement in HbA1c [14], which can help more patients achieve an optimal glycemic control. It is well known that hypoglycemia is a major limiting factor in glycemic management in patient with diabetes. Thus, the fear of hypoglycemia should be considered when initiating a patient on insulin therapy, and treatments with a lower risk of hypoglycemia should be prescribed. Interestingly, our study did not show any differences across basal insulin treatment regimens until treatment intensification, but this approach was limited to assessing the addition of another non-basal insulin or GLP-1 RA (if the GLP-1 RA was not present at the time or within 3 months of insulin initiation) and therefore does not consider the potential for better titration of insulin.

However, several other factors may have contributed to the better persistence observed with Gla-300, including the novelty of this insulin, more frequent prescriptions of a new insulin by the specialist or some baseline characteristics not considered in the statistical adjustment conducted in the maintenance analysis.

Surprisingly, Gla-300 was less frequently initiated in patients aged ≥ 74 years, a patient population that is more susceptible to hypoglycemia. During the period of our observations, physicians had only a limited experience with Gla-300 in France in elderly patients, and the lack of evidence from clinical trials at that time may have influenced the physicians’ decision. Subsequently, however, new data from randomized controlled trials have demonstrated the safety of Gla-300 in elderly patients (≥ 75 years) [22].

Our study was not designed to determine the reasons for non-persistence with insulin therapy. However, a knowledge and understanding of these underlying reasons can prevent early discontinuation and achieve better outcomes. Factors associated with an early discontinuation of insulin therapy in a US study were younger age, female sex, comorbid conditions, visits to the ER and higher medical costs [20]. In a German study, predictors of discontinuation of basal insulin therapy were ≥ 1 documented hypoglycemia before discontinuing and diagnosed depression [23].

The crude percentage of patients hospitalized for hypoglycemia was significantly lower for those treated with Gla-300 (0.5%) than for patients treated with IDet (0.7%) or Gla-100 (0.8%). The adjusted analysis shows, however, that the trend towards a lower frequency of hypoglycemia requiring hospitalization with Gla-300 is not statistically significant. Finally, the risk of having to go to the ER (whatever the reason) during the treatment period was lower for patients treated with Gla-300 than for those treated with Gla-100 (OR 0.547, 95% CI 0.523–0.571), even after adjustment for baseline characteristics.

Our study using data from a claims database present some limitations inherent to this type of data collection. Isolated diagnosis codes reported within the database can be unreliable since diagnosis is only documented in case of hospitalizations and long-standing diseases. The database does not contain any information on clinical status, nor are the results of paraclinical examinations recorded. In addition, reasons and indications for which a specific medication was given are not documented. Nevertheless, the SNDS database is reliable and representative of all insured patients in France. Missing data are limited in the database since a prescription is required in order to initiate insulin therapy. Thus, all of the patients treated with insulin during the study period were included in the database.

Among patients with T2DM who initiate basal insulin therapy, early discontinuation of insulin may signal that the patients are experiencing difficulties with this therapy, possibly related to injections, side-effects or a lack of efficacy. This real-world study conducted in a very large population in France found that persistence by patients with T2DM was significantly better for Gla-300 than for Gla-100 and IDet. This difference could be partly related to a lower frequency of hospitalization for hypoglycemia and fewer ER visits by patients on Gla-300. Further work is needed to better understand the underlying reasons for such differences in persistence over time.