Reporting of hypoglycaemia in clinical trials of basal insulins: A need for consensus

Abstract Hypoglycaemia is a common side‐effect of diabetes therapies, particularly insulin, and imposes a substantial burden on individuals and healthcare systems. Consequently, regulatory approval of newer basal insulin (BI) therapies has relied on demonstration of a balance between achievement of good glycaemic control and less hypoglycaemia. Randomized controlled trials (RCTs) are the gold standard for assessing efficacy and safety, including hypoglycaemia risk, of BIs and are invaluable for obtaining regulatory approval. However, their highly selected patient populations and their conditions lead to results that may not be representative of real‐life situations. Real‐world evidence (RWE) studies are more representative of clinical practice, but they also have limitations. As such, data both from RCTs and RWE studies provide a fuller picture of the hypoglycaemia risk with BI therapies. However, substantial differences exist in the way hypoglycaemia is reported across these studies, which confounds comparisons of hypoglycaemia frequency among different BIs. This problem is ongoing and persists in recent trials of second‐generation BI analogues. Although they provide a lower risk of hypoglycaemia when compared with earlier BIs, they do not eliminate it. This review describes differences in the way hypoglycaemia is reported across RCTs and RWE studies of second‐generation BI analogues and examines potential reasons for these differences. For studies of BIs, there is a need to standardize aspects of design, analysis and methods of reporting to better enable interpretation of the efficacy and safety of such insulins among studies; such aspects include length of follow‐up, glycaemic targets, hypoglycaemia definitions and time intervals for determining nocturnal events.

sleep and cause chronic fatigue, with an overall reduction in healthrelated quality of life. [1][2][3] The resulting fear of hypoglycaemia may cause some individuals to deliberately maintain undesirable hyperglycaemia to minimize the risk and severity of further hypoglycaemia events. 2 Hyperglycaemia is also an important consideration in the management of individuals with diabetes, with poor glycaemic control being associated with increased risk of micro-and macrovascular complications, cardiovascular (CV) risk and all-cause mortality, 4,5 as well as being a burden on healthcare resources. 6 The effective use of insulin requires a sensitive balance between achieving and maintaining glycaemic targets while limiting the risk of hypoglycaemia. 1 Consequently, the assessment and regulatory approval of insulins have depended largely on evidence of glycaemic efficacy combined with incremental reductions in therapy-induced hypoglycaemia, utilizing data commonly derived from randomized controlled trials (RCTs). 7 It is also important to determine whether the hypoglycaemia reductions observed in the tightly regulated conditions observed in RCTs are also observed in real-life clinical practice, in which patient populations are more diverse and clinical monitoring and support is less extensive compared with an RCT. 8 The more diverse patient populations, less rigorous protocols and less intensive patient followup of real-world evidence (RWE) studies 9 may be more representative of clinical practice. 10,11 As such, although the reporting of hypoglycaemia events may be less accurate in RWE studies, observational studies of electronic health records, medical claims and billing data and registries, or prospective RWE studies such as crosssectional surveys are needed to provide a complementary source of information concerning the frequency of hypoglycaemia associated with insulins.
The disparities in definitions, methods of assessment and reporting of hypoglycaemia across RCTs and RWE studies, combined with the differences in trial designs, analyses and populations, present significant challenges when comparing different insulin molecules and formulations. These issues can obfuscate the true differences in the safety of glucose-lowering therapies and may explain the observed inconsistencies across various regulatory and advisory guidelines.
Differences in reporting of BI therapies in the context of hypoglycaemia have been an ongoing challenge, with great variability among early trials of BIs; however, considerable diversity in reporting still exists in the most recent trials of second-generation BI analogues, which precludes true comparisons of their safety across trials.
Although second-generation BI analogues have demonstrated lower rates of hypoglycaemia as compared with first-generation BIs, [12][13][14] the risk of hypoglycaemia has not yet been eradicated and it is therefore important to facilitate interpretation of efficacy and safety among BIs. In addition, as further advances in BI therapies occur, standardization across trials of these newer therapies and technologies would be beneficial, to facilitate interpretation of their hypoglycaemia risk profiles.
Improved understanding of the differences in reporting of hypoglycaemia is required and, ultimately, greater standardization concerning the way hypoglycaemia is defined, measured and analysed would greatly aid the interpretation of the safety of BI therapies across trials. The present review describes the differences in the way hypoglycaemia has been defined, measured and reported in both RCTs and RWE studies, with a focus on the most recent studies of second-generation BIs. Potential explanations for the diversity observed across studies are discussed. Nocturnal and daytime hypoglycaemia, both non-severe and severe, in individuals with type 1 and type 2 diabetes (T1DM and T2DM) are explored.

| DIVERSITY OF HYPOGLYCAEMIA ASSESSMENT AND REPORTING IN RANDOMIZED CONTROLLED TRIALS
Until the advent of treat-to-target trial designs, insulin titration in clinical trials was undertaken largely at the discretion of the investigator. 15 The first treat-to-target trial was conducted in 2003; this trial design used a pre-specified algorithm to titrate either insulin glargine 100 U/mL (Gla-100) or isophane insulin (neutral protamine Hagedorn [NPH] insulin) to achieve and maintain a target fasting plasma glucose of 5.5 mmol/L (100 mg/dL). 16 In this way, treat-to-target trials highlighted the differences in factors such as hypoglycaemia, as blood glucose levels are driven closer to euglycaemia. 17 To demonstrate the diversity in reporting of hypoglycaemia across RCTs, the present review focuses on two treat-to-target trial programmes of second-generation BI analogues, namely, the BEGIN trials, which compared insulin degludec (IDeg) with insulin glargine 100 U/mL (Gla-100), [18][19][20][21][22][23][24][25][26] and the EDITION trials, in which insulin glargine 300 U/mL (Gla-300) and Gla-100 were compared. Some of the older treat-to-target trials of first-generation BI analogues, 12,[27][28][29][30][31] Gla-100 and insulin detemir (IDet) vs NPH insulin have been included for comparison. 16,[32][33][34] The BEGIN and EDITION trials (Table 1 and Table S1) shared certain common design features. For example, both trials were randomized, open-label and treat-to-target trials. However, there were key differences between the trials, such as the starting dose of insulin, titration algorithms, targets for self-monitored blood glucose (SMBG) and hypoglycaemia definitions. These differences are examined in greater detail in the following sections.

| Eligibility criteria
Variability in inclusion and exclusion criteria can influence hypoglycaemia risk results, as various baseline characteristics can be associated with hypoglycaemia. 35  This revealed that definitions of hypoglycaemia were included in only 60% of these trials, 36 and few of these definitions followed American Diabetes Association (ADA) 37 and European Medicines Agency 38 recommendations for definition of hypoglycaemia as blood glucose (BG) of ≤3.9 mmol/L (≤70 mg/dL), or <3.1 mmol/L (<56 mg/dL), which was recommended by the EMA prior to 2012. 36 The differences in BG thresholds for non-severe hypoglycaemia can be seen in the examples of treat-to-target trials shown in

| Duration of follow-up
The duration of patient follow-up in RCTs of BIs has varied widely, ranging from 4 weeks 45,46 to 2 years, 33 and is apparent in the treatto-target trials shown in Table 1 and Table S1. Variation in trial length  Hypoglycaemia threshold set at <3.0 mmol/L (<54 mg/dL) Categories of hypoglycaemia: Major: Requiring external assistance because of severe impairment in consciousness or behaviour with BG <3.0 mmol/L (<54 mg/dL) Minor: Symptomatic episode with BG <3.0 mmol/L (<54 mg/dL) without need for assistance, or asymptomatic episode with BG <3.0 mmol/L (<54 mg/dL) Episodes suggestive of hypoglycaemia: Hypoglycaemia symptoms without corresponding BG measurement.
ADA -Workgroup on Hypoglycaemia 44 2005 Hypoglycaemia threshold set at BG ≤3.9 mmol/L (≤70 mg/dL) Categories of hypoglycaemia: Severe: Any event requiring aid of another person (recorded blood glucose not required) Documented symptomatic: Symptomatic event followed by BG of ≤3.9 mmol/L (≤70 mg/dL) Asymptomatic: BG of ≤3.9 mmol/L (≤70 mg/dL) without symptoms Probable symptomatic: Symptomatic event without BG reading Relative: Hypoglycaemia symptoms but BG of >3.9 mmol/L (>70 mg/dL) At a minimum, incidence and event rates should be reported for first three classifications Reaffirmed guidelines set by ADA in 2005; however, "relative hypoglycaemia" was re-termed "pseudo-hypoglycaemia" The International Hypoglycaemia Study Group 39,40 2017 Consensus of this group stated that a hypoglycaemia threshold indicative of clinically significant hypoglycaemia was required, which needed to be avoided because of its immediate and long-term danger to individuals. This group presented the below threshold recommendations: hypoglycaemia, the period over which hypoglycaemia events are recalled can impact upon the accuracy of hypoglycaemia frequency. The recall of non-severe hypoglycaemic events is inaccurate with a recall period of more than 1 week; however, the accuracy of severe hypoglycaemia recall does not appear to be affected by longer recall periods. 47 Alternatively, event rates of hypoglycaemia could be used; however, in trials with a duration of less than 1 year, annualized event rates are estimated from extrapolation of recorded data. Examples of this include the initial treat-to-target trial, 16  However, participants often maintain BI doses during these extension periods, and, in efforts to avoid hypoglycaemia, they may be less effective in titrating their dose and achieving glycaemic targets.

| Clock time definitions
Nocturnal hypoglycaemia has negative effects on quality of life 3 and can incur major economic costs. 53

| Titration protocols
As discussed earlier, the advent of the treat-to-target trial design facilitated comparison of safety profiles, including hypoglycaemia, among BIs. Nevertheless, differences in various clinical trials using the treatto-target design are apparent (Table 1 and Table S1). One such example is a design in which different target BG concentrations have been used. While no correlation between target BG and HbA1c has been shown, 15 adopting a lower target BG may be more likely to increase both the incidence and prevalence of hypoglycaemia events, which may enhance the prospect of observing statistically significant differences among treatments.    (Table 1 and Table S1). However, sometimes only one analysis of between-treatment differences has been reported, often the hypoglycaemia event rate ratios, 54,55,69,70 which hinders assessment of the consistency of results across trials. For instance, the BEGIN trials 18-26 focused on rate ratios between BIs, whereas the EDITION studies 12,27-31 presented both the rate ratio and relative risks.
The methods used to assess these outcomes and adjustments for confounding factors can also vary across trials. For instance, the BEGIN studies employed a negative binomial regression model to assess the rates of hypoglycaemia, with treatment, antihyperglycaemic therapy at screening, gender and geographical region as fixed factors, and with age as a covariate, [18][19][20][21][22][23][24][25][26] whereas the EDITION trials utilized an overdispersed Poisson regression model. 12,14,[27][28][29][30][31] However, the EDITION studies also analysed the number of individuals experiencing more than one hypoglycaemia event, a binary outcome, for which a Cochran-Mantel-Haenszel method, a statistical method that tests the association of a treatment with a specific binary outcome, was used to assess between-treatment differences, stratified by HbA1c at screening and by geographical region. 12,27-31 The variation in analyses described above may influence outcomes and, therefore, standardization in statistical testing of hypoglycaemia endpoints should be applied to trials of BIs.

| RCTs versus real-life clinical practice
RCTs are considered to be the "gold standard" and are essential for demonstrating the efficacy and safety of new therapies. 71 However, inherent limitations in their design may lead to underestimation of rates of hypoglycaemia. 72 For example, RCTs may exclude participants with very high HbA1c levels or those with renal impairment, 8 both of whom are associated with increased risk of hypoglycaemia. 35 Clinical trial participants are often a more engaged and informed subpopulation of individuals with diabetes and, hence, are more likely to adhere to treatment and to accept advice on diabetes self-care. 8 39,40 Additionally, studies increasingly use a standard window of 12:00 to 6:00 AM to define nocturnal hypoglycaemia.
Consensus guidelines concerning the way to measure and report hypoglycaemia events detected by CGM would be beneficial and can be feasibly achieved, but they will require large trials investigating the use of CGM as SMBG to detect hypoglycaemia, which would enable evidence-based opinion concerning the utility of CGM devices and the optimal glycaemic parameters for reporting in CGM-based studies.
However, standardization may not be easily achieved with certain aspects of study design. Inclusion criteria in RCTs may vary, to investigate efficacy and safety in specific populations such as individuals with T2DM undergoing basal-bolus regimens. While this helps to eliminate potential confounders, it also makes interpretation among trials difficult. One approach may be to reduce the number of inclusion criteria and to include more varied populations, similar to those experienced in clinical practice; however, this would require more complex statistical methods to account for potential bias.
While the treat-to-target trial design has facilitated comparisons between BIs for factors other than glycaemic control, including hypoglycaemia, substantial variation in glycaemic targets among trials remains common. Given that the ADA recommends different targets in vulnerable populations, such as children or frail elderly individuals, 37 several standardized glycaemic targets may be required for different age groups or for patients at high risk of hypoglycaemia.
The variation observed in titration algorithms among different treatto-target trials may also be a consequence of the different pharmacokinetics of various BIs.
Study length of RCTs comparing BIs may vary, based on the purpose of the trial, with phase 1 and 2 trials being shorter proof of concept trials, and the length can also be determined by the durability of endpoints of interest. However, it is feasible that study length could be standardized according to the clinical trial phase.
Additionally, standardization of the length of "titration" phases The present review has some limitations. It was not a systematic meta-analysis and, as such, some recent trials of second-generation Bis may have been omitted. In addition, this review focuses on a small subset of diabetes therapies, specifically second-generation BI analogues. However, the noted disparities in hypoglycaemia reporting are also apparent in trials of various antihyperglycaemic drugs.
In conclusion, hypoglycaemia risk profiles of BIs remain important factors in choosing between therapies, but the current diversity in the way hypoglycaemia is reported across RCTs and RWE studies prevents comparisons among studies. The development and application of consensus guidelines denoting the way hypoglycaemia should be defined and reported would contribute to future study design in a way that facilitates interpretation of hypoglycaemia risk profiles among BIs across studies.