OBJECTIVE

The Diabetes Control and Complications Trial (DCCT) identified an inverse relationship between HbA1c and severe hypoglycemia. We investigated the relationship between hypoglycemia and HbA1c in a large type 1 diabetes cohort on multiple daily injection or insulin pump therapy using blinded continuous glucose monitoring (CGM) data. The impact of real-time CGM on these relationships and how these relationships differ with biochemical definitions of hypoglycemia have also been assessed.

RESEARCH DESIGN AND METHODS

CGM data were obtained from the JDRF CGM randomized control trial. Baseline blinded CGM data were used to assess time in hypoglycemia in all individuals. End point data from the CGM intervention group were used to assess the impact of CGM. Percentage of time spent below 3.9, 3.3, 3.0, and 2.8 mmol/L were calculated and quadratic regression plots drawn. Relationships were analyzed visually, and ANOVA was used to assess relationships between glycemia and time below threshold.

RESULTS

J-shaped relationships were observed for all biochemical hypoglycemia thresholds, with the lowest hypoglycemia risk occurring at HbA1c values between 8.1 and 8.6% (65–70 mmol/mol). The use of an average of 5 days/week of CGM flattened the relationships for 3.3, 3.0, and 2.8 mmol/L, and ANOVA confirmed the loss of relationship for the 3.3 mmol/L threshold using CGM.

CONCLUSIONS

The relationship between hypoglycemia and HbA1c in a population with type 1 diabetes is J-shaped. Lower HbA1c values are still associated with increased hypoglycemia risk, although the magnitude of risk depends on biochemical threshold. Real-time CGM may reduce the percentage time spent in hypoglycemia, changing the relationship between HbA1c and hypoglycemia.

The Diabetes Control and Complications Trial (DCCT) study in 1993 (1) demonstrated a substantial reduction in microvascular complication risk in an intensely treated group of adults with type 1 diabetes, and later analysis confirmed a reduction in macrovascular risk (2). However, the intensively treated group had an approximate threefold increase in risk of severe hypoglycemia (SH), of which ∼60% was attributed to HbA1c (3) with a strong inverse relationship between HbA1c and SH.

Hypoglycemia is the main side effect of exogenous insulin, is associated with physical and psychological morbidity, and impacts on treatment outcomes (4,5). It can be clinically categorized into asymptomatic, symptomatic, and severe, with some discussion around which biochemical threshold defines hypoglycemia in type 1 diabetes (5). The International Hypoglycaemia Study Group has recently recommended a glucose threshold <3.0 mmol/L when reporting hypoglycemia in clinical trials. They also propose 3.9 mmol/L as a glucose alert value (6,7). Counterregulatory response can occur <3.3 mmol/L in healthy individuals (8), and cognitive dysfunction occurs at <2.8 mmol/L (9). Continuous glucose monitoring (CGM) has enabled more precise definition of hypoglycemia, requiring a minimum duration of 15 min below threshold with events separated by at least 30 min (10).

Since the DCCT, there have been therapeutic and technological advances in diabetes management that reduce hypoglycemia risk independently of glycemic control. The appearance of insulin analogs (1113), improvements in insulin pump therapy (14), implementation of structured education (15), and multidisciplinary care by specialist teams (16) have all contributed to improvements in type 1 diabetes outcomes, potentially altering the relationships defined in the DCCT. Epidemiological analyses reinvestigating the relationship between SH and HbA1c in pediatric populations have found weak or nonsignificant relationships (1719). The 30-year follow-up data from the DCCT/Epidemiology of Diabetes Interventions and Complications (EDIC) cohort demonstrate a reduction in the frequency of SH over time, with the presence of a previous episode of SH the strongest predictor of future episodes of SH (20). Interestingly, the risk of SH in the intensive group was unchanged between the DCCT and EDIC studies above an HbA1c of ∼8.5% and in the conventional group was higher in the EDIC cohort above a similar HbA1c threshold, suggesting that the impact of advances in therapies on hypoglycemia has differentially addressed risk in those with lower HbA1c values.

The JDRF CGM study was a 26-week randomized control trial evaluating the efficacy and safety of CGM in children and adults with type 1 diabetes published in 2008 (2123). It was the first to clearly demonstrate that the continuous use of real-time CGM devices could be associated with improved glycemic control.

We hypothesize that the inverse relationship between HbA1c and SH identified in the DCCT has evolved with advances in treatment and varies with threshold of hypoglycemia. We additionally hypothesize that unblinded CGM has an impact on the interaction between HbA1c and hypoglycemia.

Data Source

Data from the JDRF CGM study (2123) were used in this analysis. The JDRF data set is freely accessible and was obtained from the Jaeb Center for Health Research (http://diabetes.jaeb.org/Dataset.aspx). All data were analyzed using Stata v14.2 (StataCorp, College Station, TX).

The JDRF CGM study was a 26-week randomized control trial evaluating the efficacy and safety of CGM in children and adults with type 1 diabetes. The study protocol and characteristics of the participants enrolled have been detailed elsewhere (2123).

Briefly, a total of 448 people with type 1 diabetes treated by multiple-dose injection (MDI) or continuous subcutaneous insulin infusion (CSII) was randomized to CGM (intervention) or capillary glucose monitoring (control). Prior to randomization, all participants underwent 1 week of blinded CGM recording as a baseline for subsequent analysis. Blinded CGM was repeated in the control group at 13 and 26 weeks. HbA1c was assessed at randomization, 13 weeks, and 26 weeks. HbA1c values are reported as DCCT-aligned percentages.

Analysis Design

Relationships between HbA1c and time in hypoglycemia were assessed using baseline blinded data for all participants. A minimum of 6 days of recording was required to be included in the analysis. To assess the impact of CGM, the 26-week HbA1c value and time below threshold for the last 6 weeks of the intervention period were used. Individuals were excluded if there were no CGM data at the beginning or end of this 6 week-analysis. The 26-week HbA1c value and 26-week blinded CGM from the control group were additionally assessed. Again, individuals were excluded if <6 days of CGM data were available. Relationships were analyzed visually using quadratic regression plots, and a relationship between time spent below threshold (3.9, 3.3, 3.0, and 2.8 mmol/L) was assessed across duodeciles of HbA1c. Glycemic variability (GV) measures were analyzed using MATLAB (MathWorks) to discriminate the subgroups with the highest risk of hypoglycemia (24). Episodes of hypoglycemia for CGM reports defined as time spent <3.0 mmol/L for 15 min separated by a minimum of 30 min were also analyzed.

Statistical Analysis

Data were tested for normality and nonparametric tests were used accordingly. Descriptive data are reported as mean (SD) where normally distributed or as median and interquartile range (IQR) where skewed. Quadratic regression models were used to assess relationships between percentage time below threshold and HbA1c for each threshold in each analysis group, and their curves were plotted. Nonparametric ANOVA was used to assess variance of time below threshold across the range of HbA1c. Statistical tests were two-tailed, and the significance level was set at P < 0.05.

CGM data were available for 451 individuals at baseline, but 3 individuals were excluded due to missing HbA1c values. Of the 448 individuals included in the baseline analysis (54.9% female, age 25.1 ± 15.8 years, and HbA1c 7.44 ± 0.86%), 231 individuals were assigned to the CGM intervention group and 217 to the control group after the randomization. At 26 weeks, 6-week CGM data were available for 155 individuals in the CGM group after exclusion of 76 individuals due to incomplete data, and CGM data with a minimum of 6 days were available for 185 individuals in the control group after exclusion of 32 individuals due to incomplete data. The baseline clinical characteristics of the individuals included in the analyses at baseline and at 26 weeks are shown in Table 1.

Table 1

Baseline characteristics for baseline (n = 448), CGM (n = 155), and control (n = 185) groups

BaselineCGMControl
Females, n (%) 246 (54.9) 87 (56.1) 106 (57.3) 
White race, n (%) 423 (94.4) 149 (96.1) 176 (95.1) 
Age (years), n (%)    
 8–14 142 (31.7) 51 (32.9) 51 (27.6) 
 15–24 141 (31.5) 32 (20.6) 63 (34) 
 >25 165 (36.8) 72 (46.5) 71 (38.4) 
Insulin treatment, n (%)    
 Pump 365 (81.5) 135 (87.1) 146 (78.9) 
 Injections 83 (18.5) 20 (12.9) 40 (21.6) 
HbA1c (%), n (%)    
 <7 128 (28.6) 48 (31) 54 (29.2) 
 7–8.99 301 (67.2) 100 (64.5) 124 (67) 
 ≥9 19 (4.2) 7 (4.5) 8 (4.3) 
Diabetes duration in years, median (IQR) 9.4 (5.5, 17.9) 11 (6.2, 22) 9.7 (5.8, 17.9) 
Individuals with >1 SH in past 6 months, n (%) 38 (8.5) 14 (9) 14 (7.6) 
BaselineCGMControl
Females, n (%) 246 (54.9) 87 (56.1) 106 (57.3) 
White race, n (%) 423 (94.4) 149 (96.1) 176 (95.1) 
Age (years), n (%)    
 8–14 142 (31.7) 51 (32.9) 51 (27.6) 
 15–24 141 (31.5) 32 (20.6) 63 (34) 
 >25 165 (36.8) 72 (46.5) 71 (38.4) 
Insulin treatment, n (%)    
 Pump 365 (81.5) 135 (87.1) 146 (78.9) 
 Injections 83 (18.5) 20 (12.9) 40 (21.6) 
HbA1c (%), n (%)    
 <7 128 (28.6) 48 (31) 54 (29.2) 
 7–8.99 301 (67.2) 100 (64.5) 124 (67) 
 ≥9 19 (4.2) 7 (4.5) 8 (4.3) 
Diabetes duration in years, median (IQR) 9.4 (5.5, 17.9) 11 (6.2, 22) 9.7 (5.8, 17.9) 
Individuals with >1 SH in past 6 months, n (%) 38 (8.5) 14 (9) 14 (7.6) 

Baseline Data

Quadratic regression curves for the percentage time spent in hypoglycemia below four different biochemical thresholds (2.8, 3.0, 3.3, and 3.9 mmol/L) at different HbA1c values are shown in Fig. 1. J-shaped relationships were observed for each threshold. The highest percentage time below threshold occurs at the lowest HbA1c values. Each curve has an HbA1c value that is associated with the lowest time in hypoglycemia for that threshold, indicated by the curve’s nadir point. HbA1c values on either side of this point are associated with progressively increasing hypoglycemia, with the rate of change steeper with decreasing HbA1c values compared with increasing HbA1c values. The regression coefficients, coefficients of determination, and their associated P values were significant for all of the thresholds analyzed (P < 0.05 for all of them).

Figure 1

Quadratic regression plots of percentage time in hypoglycemia for each biochemical threshold. A: Baseline group (n = 448). B: CGM group at 26 weeks (n = 155). C: Control group at 26 weeks (n = 185). D: Overlap control and intervention groups at 26 weeks.

Figure 1

Quadratic regression plots of percentage time in hypoglycemia for each biochemical threshold. A: Baseline group (n = 448). B: CGM group at 26 weeks (n = 155). C: Control group at 26 weeks (n = 185). D: Overlap control and intervention groups at 26 weeks.

Close modal

The 3.9 mmol/L curve has the steepest relationship, showing a greater rate of change for a given change in HbA1c at both sides of the nadir, when compared with the 3.3, 3.0, and 2.8 mmol/L curves. J-shaped relationships were also present in the control group at 26 weeks, shown in Fig. 1C, confirming the relationships seen at baseline.

CGM Group Analysis

CGM changed the nature of the relationships observed in the baseline analysis, as shown in Fig. 1B. The overall relationship between time spent <3.9 mmol/L and HbA1c remained significant; however, the slope was shallower at lower and higher HbA1c values. The shapes of the 3.3, 3.0, and 2.8 mmol/L curves changed substantially with unblinded CGM, becoming flattened U-shaped curves with reduced variation in time spent below hypoglycemia threshold with change in HbA1c. The regression curves for time <3.3, <3.0, and <2.8 mmol/L had low R2 values and were not significant.

The mean percentage of sensor use in our intervention group was 72.5% (5.0 days/week). A total of 59.4% of the subjects used the sensor >70% of the total time during the 6-week CGM analyzed at the end of the follow-up. A total of 13.0% of subjects used the CGM <50% (3.5 days/week) of the time.

The percentage reduction in time spent in hypoglycemia in each HbA1c duodecile has been also calculated using the curves from intervention and control groups at 26-week evaluation. Considering the 3.0 mmol/L as the hypoglycemia cutoff point (in line with the International Hypoglycaemia Study Group), the mean reduction in percentage time in hypoglycemia is 57.6%, with a maximum reduction of 75% of time in hypoglycemia between 6.5 and 7.5% of HbA1c. With 3.9 mmol/L as the hypoglycemia cutoff point, the mean reduction in percentage time in hypoglycemia is 46.3% across the HbA1c range of 6.0–9.5% and the maximum reduction of 43% achieved between HbA1c of 6.0 and 7.5%.

Nadir Point

At baseline, the HbA1c values associated with the lowest time spent below hypoglycemia threshold were 8.1% (65 mmol/mol) for 2.8 mmol/L, 8.2% (66 mmol/mol) for 3.0 mmol/L, 8.4% (68 mmol/mol) for 3.3 mmol/L, and 8.6% (70 mmol/mol) for 3.9 mmol/L. The CGM nadir points for the 2.8, 3.0, and 3.3 mmol/L curves were reduced to 7.3% (56 mmol/mol), 7.4% (57 mmol/mol), and 7.7% (61 mmol/mol), respectively, whereas the nadir point for 3.9 mmol/L increased to 9.1% (76 mmol/mol).

At baseline, the time below threshold hypoglycemia experienced at the National Institute for Health and Care Excellence target HbA1c value of 6.5% (48 mmol/mol) was 8.6% for 3.9 mmol/L, 4.5% for 3.3 mmol/L, 3.0% for 3.0 mmol/L, and 2.5% for 2.8 mmol/L. With CGM, the percentage time below hypoglycemia threshold was 4.8% for 3.9 mmol/L, 1.8% for 3.3 mmol/L, 1.1% for 3.0 mmol/L, and 0.8% for 2.8 mmol/L.

ANOVA

The mean time spent below threshold for each HbA1c duodecile is shown in Fig. 2. At baseline, Kruskal-Wallis analysis identified significant variances for time spent below threshold by HbA1c for time <3.9 (P < 0.0001), <3.3 (P < 0.002), <3.0 (P < 0.010), and <2.8 mmol/L (P < 0.05). Significant variance was additionally identified in the control group for 3.9 mmol/L (P < 0.002) and 3.3 mmol/L (P < 0.02), but not for 3.0 mmol/L and 2.8 mmol/L (P = 0.09 and P = 0.06, respectively). However, in the CGM group, Kruskal-Wallis analysis only identified significant difference for the 3.9 mmol/L threshold (P < 0.01), with no variance of hypoglycemia across HbA1c categories identified for the 3.3, 3.0, or 2.8 mmol/L thresholds (Fig. 2).

Figure 2

Bar charts showing mean percentage time below threshold (%) at each HbA1c duodecile (%) for each hypoglycemia threshold by analysis group. A: Baseline group; n = 448. B: CGM group at 26-week end point; n = 155. C: Control group at 26-week end point; n = 185. Kruskal-Wallis analysis for Fig. 2 is also included.

Figure 2

Bar charts showing mean percentage time below threshold (%) at each HbA1c duodecile (%) for each hypoglycemia threshold by analysis group. A: Baseline group; n = 448. B: CGM group at 26-week end point; n = 155. C: Control group at 26-week end point; n = 185. Kruskal-Wallis analysis for Fig. 2 is also included.

Close modal

GV

Measures of GV were analyzed for each HbA1c duodecile. Kruskal-Wallis analysis identified significant variances by HbA1c for SD and mean amplitude of glucose excursions (MAGE; P < 0.0001), with higher values of SD and MAGE occurring with higher duodeciles of HbA1c. Significant variance was also described for low blood glucose index (LBGI), with the lowest duodeciles of HbA1c being associated with a higher LBGI (P < 0.0001). However, the coefficient of variation (CV) showed no variance across HbA1c duodeciles (P = 0.14).

A post hoc analysis identified three main groups defined by cut points of HbA1c of <6.5% (48 mmol/mol), between 6.5% (48 mmol/mol) and 8.5% (69 mmol/mol), and >8.5% (69 mmol/mol) with significant differences in GV. The analysis was then repeated through the three main groups, and Kruskal-Wallis showed significant variances for SD (P < 0.0001), MAGE (P < 0.0001), LBGI (P < 0.0001), and high blood glucose index (HBGI; P < 0.0001). The percentage of CV showed no variance across the groups (P = 0.06). In addition, the number of episodes of hypoglycemia was calculated for the three HbA1c groups. Significant variance was identified with an HbA1c of <6.5% (48 mmol/mol) associated with a higher number of hypoglycemia episodes. The results are shown in Table 2.

Table 2

Measures of GV through the three main groups of HbA1c

HbA1c <6.5%HbA1c 6.5–8.5%HbA1c >8.5%P value
Mean glucose, mmol/L (mg/dL) 7.3 (132) 9.2 (166) 11.5 (207)  
SD 51.89 (43.60, 63.83) 66.17 (55.92, 76.28) 82.73 (69.81, 96.26) <0.0001 
%CV 39.44 (33.93, 49.30) 39.81 (34.83, 45.37) 39.66 (34.62, 47.19) 0.058 
MAGE 105.72 (89.11, 124.60) 130.19 (110.41, 150.88) 160.68 (137.53, 182.74) <0.0001 
LBGI 1.87 (1.13, 4.10) 1.07 (0.43, 2.25) 0.53 (0.18, 1.41) <0.0001 
HBGI 3.87 (2.16, 5.43) 8.29 (5.86, 11.64) 16.30 (12.42, 20.04) <0.0001 
Episodes of hypoglycemia 2 (0, 7.5) 1 (0, 4) 0 (0, 3.5) <0.05 
HbA1c <6.5%HbA1c 6.5–8.5%HbA1c >8.5%P value
Mean glucose, mmol/L (mg/dL) 7.3 (132) 9.2 (166) 11.5 (207)  
SD 51.89 (43.60, 63.83) 66.17 (55.92, 76.28) 82.73 (69.81, 96.26) <0.0001 
%CV 39.44 (33.93, 49.30) 39.81 (34.83, 45.37) 39.66 (34.62, 47.19) 0.058 
MAGE 105.72 (89.11, 124.60) 130.19 (110.41, 150.88) 160.68 (137.53, 182.74) <0.0001 
LBGI 1.87 (1.13, 4.10) 1.07 (0.43, 2.25) 0.53 (0.18, 1.41) <0.0001 
HBGI 3.87 (2.16, 5.43) 8.29 (5.86, 11.64) 16.30 (12.42, 20.04) <0.0001 
Episodes of hypoglycemia 2 (0, 7.5) 1 (0, 4) 0 (0, 3.5) <0.05 

Data are median (IQR) unless otherwise indicated. Episodes of hypoglycemia are episodes measured by CGM <3.0 mmol/L for at least 15 min separated by at least 30 min.

A multiple linear regression analysis was performed to identify baseline independent factors associated with time <3.0 mmol/L or LBGI >2. The analysis was conducted through the whole cohort and in the group with an HbA1c >8.5% (69 mmol/mol). Age, type 1 diabetes duration, HbA1c, sex, BMI, and type of treatment at baseline (MDI or CSII) were included in the analysis. For time <3.0 mmol/L, the regression showed an R2 of 0.114 with HbA1c, and treatment modality significantly associated with the time under the threshold. In the analysis of the HbA1c >8.5% group (69 mmol/mol), R2 was 0.145. For this group, CSII treatment was associated with a reduction in time <3.0 mmol/L by 10% (P = 0.05). The analysis showed similar results when LBGI >2 was selected as the dependent variable. Again, higher HbA1c and CSII treatment were significantly associated with lower LBGI across the cohort. For the group with an HbA1c >8.5% (69 mmol/mol), the model showed an R2 of 0.212, and treatment with CSII was independently associated with a reduction in LBGI of 0.2 (P < 0.05).

Using the largest available CGM data set, we have shown that the relationship between HbA1c and hypoglycemia, defined at any threshold, is J-shaped, with excess time spent below threshold at lower HbA1c values. Furthermore, we have demonstrated that the relationship between HbA1c and hypoglycemia is weakened, and may even be abolished, by CGM and that CGM enables achievement of a lower HbA1c and reduced hypoglycemia risk simultaneously. The increased hypoglycemia risk at higher HbA1c values has been identified in other data sets (19) and is consistent with the persistent risk of SH above an HbA1c of 8.5% (69 mmol/mol) in the DCCT/EDIC follow-up data (20).

From our data, two groups of people are identified at increased risk of hypoglycemia: those with tight glycemic control manifested by a lower HbA1c and reduced glucose variability assessed by SD and MAGE and people with higher HbA1c values and increased glucose variability with higher SD, MAGE, and HBGI values. We hypothesize that this second group is more labile and an important group to recognize, as it is not consistently recognized that people with higher HbA1c values may be at risk for hypoglycemia. This may inhibit access to treatments like CGM that we have shown can address hypoglycemia risk.

Additionally, we have shown that achieving an HbA1c value between 8.1 and 8.6% may be associated with the lowest hypoglycemia risk, and therefore, this may be a safe initial objective for those individuals with a higher HbA1c and a history of severe or challenging hypoglycemia.

Associations between HbA1c and some measures of GV have previously been described in the JDRF data set (25). Furthermore, as demonstrated in the DCCT/EDIC cohort (20), insulin pump use in the JDRF data set was associated with a lower risk of hypoglycemia across the whole range of HbA1c. As 87% of all of the patients included in the JDRF analysis were on a pump, our results cannot be extrapolated to a population solely treated with MDI. Similar analysis using MDI populations would be of interest.

Rates of SH in the JDRF study were low and cannot be assessed by HbA1c, as overall rates were low, and direct comparison with SH in the DCCT and EDIC cohorts is not possible. However, CGM has allowed more detailed description of hypoglycemia at differing thresholds and with intervention. Percentage time below hypoglycemic thresholds is a valid measure of hypoglycemia exposure that reduces counterregulatory responses in a dose-response relationship (26), increasing the risk of SH.

CGM Impact on Relationships

CGM abolished the relationship between HbA1c and hypoglycemia <3.3, <3.0, and <2.8 mmol/L, both in the quadratic regression curves and in ANOVA. The relationship between HbA1c and time spent <3.9 mmol/L remained after 26 weeks of CGM, suggesting that mild biochemical hypoglycemia is likely to remain more frequent at the extremes of HbA1c but that it is not more likely to progress to more significant hypoglycemia and may therefore be managed appropriately. This finding has implications for the selection of hypoglycemia thresholds in clinical practice in which, with the support of CGM, 3.9 mmol/L may be considered a safe threshold for action.

The decoupling of the relationship between HbA1c and hypoglycemia that is seen with CGM between the thresholds of 3.9 and 3.3 mmol/L may be explained by the alerts and alarms for hypoglycemia and glucose rate of change. These may be personalized and commonly set between 3.5 and 4 mmol/L, a level that may not prevent glucose values <3.9 mmol/L but may reduce the risk of glucose concentrations <3.3 mmol/L.

Our results further support the role of CGM in enabling and empowering patients to intensify glycemic control without increasing the risk of hypoglycemia. For people achieving the National Institute for Health and Care Excellence target HbA1c of 6.5% (48 mmol/mol) in the JDRF study, at baseline, 64 min/day was spent with glucose <3.3 mmol/L and 36 min/day <2.8 mmol/L. With CGM, these times fell to 26 and 12 min, respectively.

A potential limitation to our study is that the data were derived from early generation CGM devices that are less accurate than contemporary devices, especially in the hypoglycemia range.

Conclusion

We present data that suggest that the relationship between HbA1c and time spent in hypoglycemia in adults and children with type 1 diabetes is J-shaped. Lower HbA1c values remain associated with increased hypoglycemia risk, although the magnitude of risk depends on the biochemical hypoglycemia threshold. Higher HbA1c values are additionally associated with increased hypoglycemia, which may be attributable to greater glucose variability. This has the potential to change the paradigm when deciding suitability of candidates for CGM in terms of clinical benefit. Hypoglycemia risk is minimized at an HbA1c value of ∼8.1–8.6% (65–70 mmol/mol) and continuous use of CGM (5 days/week) may reduce the percentage time spent in hypoglycemia, changing the relationship between HbA1c and hypoglycemia for time <3.3 mmol/L while enabling optimization of HbA1c.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.G. and N.O. directed and conducted the statistical analysis, interpreted the data, and wrote the manuscript. A.J.T. conducted statistical analysis, interpreted the results, and wrote the manuscript. M.R. and I.C. contributed to interpreting the results and reviewing the manuscript. V.M. conducted the statistical analysis and contributed to interpreting the results. M.G. and N.O. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Nathan
DM
,
Genuth
S
,
Lachin
J
, et al.;
Diabetes Control and Complications Trial Research Group
.
The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus
.
N Engl J Med
1993
;
329
:
977
986
[PubMed]
2.
Nathan
DM
,
Cleary
PA
,
Backlund
JY
, et al.;
Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) Study Research Group
.
Intensive diabetes treatment and cardiovascular disease in patients with type 1 diabetes
.
N Engl J Med
2005
;
353
:
2643
2653
[PubMed]
3.
The Diabetes Control and Complications Trial Research Group
.
Hypoglycemia in the Diabetes Control and Complications Trial
.
Diabetes
1997
;
46
:
271
286
[PubMed]
4.
Cryer
PE
.
Mechanisms of hypoglycemia-associated autonomic failure in diabetes
.
N Engl J Med
2013
;
369
:
362
372
[PubMed]
5.
Boyle
PJ
,
Schwartz
NS
,
Shah
SD
,
Clutter
WE
,
Cryer
PE
.
Plasma glucose concentrations at the onset of hypoglycemic symptoms in patients with poorly controlled diabetes and in nondiabetics
.
N Engl J Med
1988
;
318
:
1487
1492
[PubMed]
6.
International Hypoglycaemia Study Group
.
Glucose concentrations of less than 3.0 mmol/L (54 mg/dL) should be reported in clinical trials: a joint position statement of the American Diabetes Association and the European Association for the Study of Diabetes
.
Diabetes Care
2017
;
40
:
155
157
[PubMed]
7.
Seaquist
ER
,
Anderson
J
,
Childs
B
, et al
.
Hypoglycemia and diabetes: a report of a workgroup of the American Diabetes Association and the Endocrine Society
.
Diabetes Care
2013
;
36
:
1384
1395
[PubMed]
8.
Amiel
SA
,
Sherwin
RS
,
Simonson
DC
,
Tamborlane
WV
.
Effect of intensive insulin therapy on glycemic thresholds for counterregulatory hormone release
.
Diabetes
1988
;
37
:
901
907
[PubMed]
9.
Cryer
PE
.
The barrier of hypoglycemia in diabetes
.
Diabetes
2008
;
57
:
3169
3176
[PubMed]
10.
Schnell
O
,
Barnard
K
,
Bergenstal
R
, et al
.
Role of continuous glucose monitoring in clinical trials: recommendations on reporting
.
Diabetes Technol Ther
2017
;
19
:
391
399
[PubMed]
11.
Brunelle
BL
,
Llewelyn
J
,
Anderson
JH
 Jr
,
Gale
EA
,
Koivisto
VA
.
Meta-analysis of the effect of insulin lispro on severe hypoglycemia in patients with type 1 diabetes
.
Diabetes Care
1998
;
21
:
1726
1731
[PubMed]
12.
Pedersen-Bjergaard
U
,
Kristensen
PL
,
Beck-Nielsen
H
, et al
.
Effect of insulin analogues on risk of severe hypoglycaemia in patients with type 1 diabetes prone to recurrent severe hypoglycaemia (HypoAna trial): a prospective, randomised, open-label, blinded-endpoint crossover trial
.
Lancet Diabetes Endocrinol
2014
;
2
:
553
561
[PubMed]
13.
DeWitt
DE
,
Hirsch
IB
.
Outpatient insulin therapy in type 1 and type 2 diabetes mellitus: scientific review
.
JAMA
2003
;
289
:
2254
2264
[PubMed]
14.
Pickup
J
,
Mattock
M
,
Kerry
S
.
Glycaemic control with continuous subcutaneous insulin infusion compared with intensive insulin injections in patients with type 1 diabetes: meta-analysis of randomised controlled trials
.
BMJ
2002
;
324
:
705
[PubMed]
15.
DAFNE Study Group
.
Training in flexible, intensive insulin management to enable dietary freedom in people with type 1 diabetes: Dose Adjustment for Normal Eating (DAFNE) randomised controlled trial
.
BMJ
2002
;
325
:
746
[PubMed]
16.
Nathan
DM
.
Diabetes: Advances in diagnosis and treatment
.
JAMA
2015
;
314
:
1052
1062
[PubMed]
17.
Karges
B
,
Rosenbauer
J
,
Kapellen
T
, et al
.
Hemoglobin A1c Levels and risk of severe hypoglycemia in children and young adults with type 1 diabetes from Germany and Austria: a trend analysis in a cohort of 37,539 patients between 1995 and 2012
.
PLoS Med
2014
;
11
:
e1001742
[PubMed]
18.
Pedersen-Bjergaard
U
,
Pramming
S
,
Heller
SR
, et al
.
Severe hypoglycaemia in 1076 adult patients with type 1 diabetes: influence of risk markers and selection
.
Diabetes Metab Res Rev
2004
;
20
:
479
486
[PubMed]
19.
Haynes
A
,
Hermann
JM
,
Miller
KM
, et al.;
T1D Exchange, WACDD and DPV registries
.
Severe hypoglycemia rates are not associated with HbA1c: a cross-sectional analysis of 3 contemporary pediatric diabetes registry databases
.
Pediatr Diabetes
2017
;
18
:
643
650
[PubMed]
20.
Gubitosi-Klug
RA
,
Braffett
BH
,
White
NH
, et al.;
Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) Research Group
.
Risk of severe hypoglycemia in type 1 diabetes over 30 years of follow-up in the DCCT/EDIC Study
.
Diabetes Care
2017
;
40
:
1010
1016
[PubMed]
21.
JDRF CGM Study Group
.
JDRF randomized clinical trial to assess the efficacy of real-time continuous glucose monitoring in the management of type 1 diabetes: research design and methods
.
Diabetes Technol Ther
2008
;
10
:
310
321
[PubMed]
22.
Tamborlane
WV
,
Beck
RW
,
Bode
BW
, et al.;
Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group
.
Continuous glucose monitoring and intensive treatment of type 1 diabetes
.
N Engl J Med
2008
;
359
:
1464
1476
[PubMed]
23.
Beck
RW
,
Hirsch
IB
,
Laffel
L
, et al.;
Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group
.
The effect of continuous glucose monitoring in well-controlled type 1 diabetes
.
Diabetes Care
2009
;
32
:
1378
1383
[PubMed]
24.
Service
FJ
.
Glucose variability
.
Diabetes
2013
;
62
:
1398
1404
[PubMed]
25.
El-Laboudi
AH
,
Godsland
IF
,
Johnston
DG
,
Oliver
NS
.
Measures of glycemic variability in type 1 diabetes and the effect of real-time continuous glucose monitoring
.
Diabetes Technol Ther
2016
;
18
:
806
812
[PubMed]
26.
Galassetti
P
,
Tate
D
,
Neill
RA
,
Richardson
A
,
Leu
SY
,
Davis
SN
.
Effect of differing antecedent hypoglycemia on counterregulatory responses to exercise in type 1 diabetes
.
Am J Physiol Endocrinol Metab
2006
;
290
:
E1109
E1117
[PubMed]
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.

Supplementary data