Diabetic Ketoacidosis at Onset of Type 1 Diabetes and Glycemic Outcomes with Closed-Loop Insulin Delivery

Abstract

The presence of diabetic ketoacidosis (DKA) at diagnosis of type 1 diabetes (T1D) is associated with higher glycated hemoglobin levels over time. We evaluated whether hybrid-closed loop (HCL) therapy from onset of T1D could prevent the adverse impact of DKA at diagnosis on long-term glycemic outcomes. This was a posthoc analysis from 51 adolescents using HCL from diagnosis of T1D as part of the CLOuD trial (NCT02871089). We compared glycemic and insulin metrics between adolescents with (n = 17) and without (n = 34) DKA at diagnosis. Participants with and without DKA at diagnosis had similar time in target glucose range 3.9–10.0 mmol/L (70–180 mg/dL), time below range (<3.9 mmol/L, <70 mg/dL) and HbA1c at 6, 12, and 24 months. While insulin requirements at 6 months were higher in those with DKA at diagnosis, this was not statistically significant after adjusting for bodyweight. Residual C-peptide secretion was similar between groups. We conclude that HCL therapy may mitigate against the negative glycemic effects of DKA at T1D diagnosis.

Introduction

Diabetic ketoacidosis (DKA) is a life-threatening and relatively common first presentation of type 1 diabetes (T1D), occurring in ∼25% of children and young people. ^1,2 Large observational studies involving young people with T1D have found that those presenting with DKA at diagnosis have higher glycated hemoglobin (HbA1c) levels over time compared to those who do not present with DKA, independent of demographic and socioeconomic factors. ^{3 –8}

Hybrid-closed loop (HCL) therapy is the most advanced technology currently available to manage T1D, comprising an algorithm which automatically modulates insulin delivery through an insulin pump in response to real-time sensor glucose levels. In young people with new-onset T1D, HCL leads to improved glycemic outcomes over the first 24-months compared to standard insulin therapy. ^9,10 This posthoc analysis assessed whether hybrid closed-loop therapy could additionally mitigate against the adverse impact of DKA at diagnosis on long-term glycemic outcomes.

Materials and Methods

We retrospectively analyzed data from 51 adolescents randomized to use hybrid closed-loop (HCL) for 24 months from onset of T1D as part of the CLOuD multicentre randomized controlled trial. ⁹ The key inclusion criteria for the trial were T1D diagnosis within the previous 21 days and age between 10 and 16.9 years. The CLOuD trial received approval from the Cambridge East Research Ethics Committee (16/EE/0286) and the Medicines and Healthcare Products Regulatory Agency.

The HCL system was the Cambridge model predictive control algorithm (version 0.3.71) run in two hardware configurations: the FlorenceM configuration followed by the CamAPS FX configuration to improve usability and therapy adherence. In both configurations, algorithm-driven insulin delivery was automatically adjusted every 8–12 min in response to real-time sensor glucose.

Glycated hemoglobin (HbA1c) and weight were measured at three monthly intervals from diagnosis. Participants underwent a mixed-meal tolerance test at 6, 12 and 24 months with measurement of plasma C-peptide area under the curve (AUC).

Sensor glucose endpoints (time in range 3.9–10.0 mmol/L [70–180 mg/dL], time below 3.9 mmol/L [70 mg/dL]) and total daily insulin dose were calculated for each participant for each month from diagnosis using GStat software, version 2.3 (University of Cambridge, Cambridge, UK). Days with <60% of time with closed-loop operational were excluded. If a participant had <15 included days in a month, the month was excluded. This threshold was chosen to balance maximizing data availability and ensuring data were reflective of closed-loop use (Supplementary Table S1).

Participants were grouped based on the presence or absence of DKA at diagnosis. Diagnostic criteria for DKA was as per the ISPAD guidelines ¹¹ : (a) hyperglycemia (blood glucose >11 mmo/L), (b) venous pH <7.3 or serum bicarbonate <18 mmol/L, (c) ketonemia (blood β-hydroxybutyrate ≥3 mmol/L) or moderate or large ketonuria.

Unpaired t-tests were used to compare glycemic- and insulin-related endpoints between the groups at 6, 12, and 24 months. Nonnormally distributed data were Winsorized at the 10th and 90th centiles before statistical analysis. P-values were adjusted for multiple comparisons using the Benjamini–Hochberg method. ¹² Statistical analysis was performed using SPSS version 28 (IBM Software, Hursley, UK).

Results

Data from 51 participants in the closed-loop group of the CLOuD study was analyzed. Table 1 shows the demographic characteristics of the study population at baseline: 49% female, mean ± standard deviation (SD) age 12 ± 2 years, mean ± SD HbA1c 94 ± 20 mmol/mol (10.7% ± 1.8%), 17 (33%) with and 34 (67%) without DKA at diagnosis.

Table 1.

Characteristics of the Study Participants at Baseline

	Overall (n = 51)	DKA at diagnosis (n = 17)	No DKA at diagnosis (n = 34)
Age
Mean—years	12.2 ± 1.6	12.5 ± 1.8	12.1 ± 1.5
Distribution—n (%)
10–13 years	41 (80)	14 (82)	27 (79)
14–<17 years	10 (20)	3 (18)	7 (21)
Sex—n (%)
Female	25 (49)	11 (65)	14 (41)
Male	26 (51)	6 (35)	20 (59)
BMI percentile a	53.4 ± 29.3	66.2 ± 24.4	47.1 ± 29.7
Race—n (%) b
White	44 (86)	15 (88)	29 (85)
Black	1 (2)	0 (0)	1 (3)
Asian	2 (4)	1 (6)	1 (3)
More than one race	4 (8)	1 (6)	3 (9)
Glycated hemoglobin level
In millimoles per mole	93.7 ± 19.5	89.5 ± 15.6	95.8 ± 21.2
As a percentage	10.7 ± 1.8	10.3 ± 1.4	10.9 ± 1.9

Plus–minus values are means ± SD. Percentages may not total 100 because of rounding.

a

The BMI is the weight in kilograms divided by the square of the height in meters.

b

Race was reported by the participants or their parents or guardian.

BMI, body-mass index; SD, standard deviation.

Glycemic outcomes over time are presented in Figure 1 and Supplementary Table S2. Mean time in range 3.9–10.0 mmol/L (70–180 mg/dL) gradually decreased over the 24 months (Fig. 1A) but remained above 70% in both groups at all time-points. There were no significant between-group differences throughout the study period (71.9% ± 9.0% in those with DKA and 73.0% ± 9.4% in those without DKA at 24 months). Median time below range (<3.9 mmol/L, <70 mg/dL) was below 4% in both groups over the 24 months (Fig. 1B) and was similar between groups at each time point compared. Mean glycated hemoglobin was similar in both groups over time (Fig. 1C) (52.5 ± 11.5 mmol/mol [7.0% ± 1.1%] in those with DKA at diagnosis and 51.3 ± 11.8 mmol/mol [6.8% ± 1.1%] in those without at 24 months). Comparing those with and without DKA at diagnosis, there was no significant difference in glycemic endpoints at 6, 12, or 24 months.

FIG. 1.

Glycemic and insulin data for first 24 months from diagnosis of T1D. (A) The percentage time in range 3.9–10.0 mmol/L (70–180 mg/dL) (mean ± SD) of sensor glucose levels from 2 to 24 months after diagnosis of T1D. (B) The percentage time in hypoglycemia <3.9 mmol/L (<70 mg/dL) (median, IQR) from 2 to 24 months after diagnosis. (C) The HbA1c level (mean ± SD) at baseline and at 6, 12, and 24 months after diagnosis. (D) The total daily insulin dose (median, IQR) from 2 to 24 months after diagnosis of T1D. (E) The AUC for the plasma C-peptide level in response to a mixed-meal tolerance test (median, IQR) at baseline and at 6, 12, and 24 months after diagnosis. AUC, area under the curve; HbA1c, glycated hemoglobin; IQR, interquartile range; SD, standard deviation; T1D, type 1 diabetes.

Insulin-related endpoints are presented in Figure 1 and Supplementary Table S2. Total daily insulin dose increased over time after diagnosis (Fig. 1D). At 6 months insulin requirements were significantly higher in the group with DKA at diagnosis: median (IQR) 42.8 (24.1, 57.6) units/day compared to 22.9 (19.0, 37.4) units/day in those without DKA (P = 0.05). When adjusted for body weight, insulin doses were still numerically greater in those with DKA at diagnosis (0.74 ± 0.26 units/kg/day vs. 0.58 ± 0.19 units/kg/day), but this was not statistically significant (P = 0.19). The between-group difference in insulin requirements narrowed with time and was no longer significant at 12 or 24 months. C-peptide AUC following mixed-meal tolerance test decreased with time in both groups (Fig. 1E), with no significant difference between groups at 6, 12, or 24 months.

Discussion

In this study of children and young people using hybrid-closed loop from the onset of T1D, 33% first presented with DKA, similar to the incidence reported in other cohorts. ^1,2 Those who presented with DKA at diagnosis had similar glycemic outcomes to those who did not present in DKA over the first 2 years postdiagnosis, with both groups sustaining a mean time in range greater than the 70% recommended by international guidelines. ¹³ This contrasts with previous large studies which found that DKA at diagnosis was associated with worsening long-term glycemic outcomes observed as early as 1–2 years postdiagnosis, ^{3 –5} and suggests that automated insulin delivery technology may be able to mitigate against the negative effects of DKA.

In our cohort, there was no difference in residual endogenous insulin secretion between those with and without DKA at diagnosis. AUC for stimulated C-peptide level was similar between groups. While total insulin needs were greater in the group presenting with DKA at 6 months, potentially suggesting a reduced honeymoon phase, this effect was not statistically significant after adjusting for body weight. Some previous studies have found lower C-peptide levels and increased insulin needs in those presenting with DKA. ^3,14,15 However, others, like us, found no difference in C-peptide levels based on presence or absence of DKA. ¹⁶ It is important to note that a range of other factors including ethnicity, body mass index, and preceding infection also affect the likelihood of someone presenting with DKA. ¹⁷

Notably, in the SEARCH for diabetes in youth study which prospectively followed up 1396 young people with newly diagnosed T1D, DKA at diagnosis was associated with increasingly higher HbA1c levels over time, independent of baseline fasting C-peptide. ³ The authors propose that the effects of DKA at onset on long-term glycemic outcomes operate through mechanisms in addition to those related to residual insulin secretion. This is in keeping with our findings, where closed-loop therapy was able to mitigate against the negative effects of DKA at diagnosis, despite not having an effect on residual C-peptide secretion. ⁹

A limitation of our analysis is that it is not possible to assess the additional impact of closed-loop therapy on top of its component technology. Some studies have found that insulin pump therapy in itself mitigates against adverse glycemic effects of DKA at diagnosis, ^5,7,18 although results are inconsistent. ³ Other limitations include the modest sample size (n = 51), lack of ethnic diversity in the study population (86% white), and that not all participants had sufficient closed-loop use to be used in the analysis every month. Longer follow-up of this cohort would be of interest to see if this effect is sustained.

Conclusions

In children and young people using hybrid closed-loop therapy from the onset of T1D, glycemic outcomes over the first 24-months did not differ based on the presence or absence of DKA at diagnosis. Closed-loop therapy may be a useful tool to mitigate against the negative glycemic effects of DKA at first presentation.