Insulin Pump Use and Diabetic Ketoacidosis Risk in Type 1 Diabetes: Secular Trends over Four Decades

Abstract

Introduction:

Continuous subcutaneous insulin infusion (CSII) in type 1 diabetes has been regarded as a major diabetic ketoacidosis (DKA) risk factor. We aimed to determine secular trends in risk since CSII implementation in the 1980s.

Research Design and Methods:

We assessed the relationship between time-varying CSII use and DKA events from 1983 to 2017 and by each decade in the 1441 Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study participants using crude and adjusted Cox proportional hazards models.

Results:

Time-varying CSII exposure was associated with significantly higher DKA risk in the 1980s (adjusted hazard ratio [HR] 5.81; 95% confidence interval [CI] 3.28–10.29; P < 0.001), but in the 2010s, this risk was not significantly elevated (adjusted HR 1.24; 95% CI 0.73–2.12; P = 0.43).

Conclusions:

DKA risk associated with CSII in type 1 diabetes has declined substantially since the 1980s such that the remaining risk in the past decade appears to be of low magnitude.

Introduction

While continuous subcutaneous insulin infusion (CSII) has been shown to improve glycemic control in patients with type 1 diabetes,¹ it has been established as a major risk factor for diabetic ketoacidosis (DKA).² Common explanations include pump malfunctions or infusion set occlusions,³ which are problematic as pumps administer rapid-acting insulin with a limited duration of action.⁴ Improvements in education and pump technology, such as sensor-augmented pumps and occlusion alarms, may have reduced the DKA risk.^5
–7 Recent short-term observational data suggest a lower risk compared with injections,^8

–11 in contrast to some contemporary trials suggesting higher risk.^12,13 These conflicting findings highlight the need for a longitudinal analysis capable of evaluating secular trends in a consistent type 1 diabetes population.

We aimed to determine trends in DKA risk since CSII implementation in the 1980s over 34 years (1983–2017) in the Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) cohort,¹⁴ which has not yet been examined by the study group. To accomplish this objective, we independently accessed publicly available data from the National Institute of Diabetes and Digestive and Kidney Diseases Central Repository.

Research Design and Methods

We retrospectively analyzed prospectively collected DCCT/EDIC data in an open, interval cohort design. Though previously detailed,¹⁵ the DCCT (1983–1993) enrolled 1441 participants with type 1 diabetes and randomly assigned them to receive either conventional (n = 730) or intensive insulin therapy (n = 711). The EDIC phase enrolled 98% of the surviving cohort, with 94% of the cohort survivors still actively participating. Published DCCT/EDIC data at 27 years of follow-up that approximates the follow-up in our analysis report 125 deaths (8.7%), indicating that 86% of the initial randomized cohort had follow-up for DKA events.¹⁶ We emphasize that none of the participants were on CSII at accrual, and potential trial participants with three or more DKA events during the prior 12 months were excluded from the DCCT.

Exposure definition

CSII status was defined as self-reported use on quarterly questionnaires in the DCCT portion of the trial and yearly questionnaires in the EDIC portion. Throughout the study, many participants used CSII intermittently, and therefore, a single participant could contribute to both the exposed and the unexposed group over follow-up. For instance, 1005 participants had used CSII at some point during the trial, but only 782 of 1304 participants were still using a pump at the most recent follow-up in 2017. Due to the dynamic nature of use, we considered CSII as a time-varying exposure.

Outcome definition

Participants who indicated a DKA event on the annual history and physical examination form were prompted to complete a DKA verification form. Self-reported DKA events triggered a medical records review by study site staff for confirmation. DKA diagnosis was determined by the principal investigator without a central adjudication process, and all events involved emergency or hospital care. All DKA events occurring after trial randomization were classified as outcomes.

Covariates

Our primary analysis adjusted for recognized DKA risk factors⁵: baseline age, sex, and duration of diabetes as well as time-varying glycated hemoglobin (HbA1c), body mass index (BMI), insulin dose, as well as diabetes complications that have been proposed to be associated with DKA risk and CSII use¹⁷: early retinopathy, advanced neuropathy, and major adverse cardiovascular events (MACE). All laboratory measurements were performed in the DCCT/EDIC central biochemistry laboratory.¹⁸ Early retinopathy was defined as Early Treatment Diabetic Retinopathy Study 23-point scale >3, advanced neuropathy as self-reported and physical examination-verified serious diabetic foot ulceration or amputation, and MACE as a composite of nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular disease.

Statistical analysis

Andersen–Gill models were used for recurrent DKA events incorporating time-dependent covariates. Our primary analysis examined the adjusted association between time-varying CSII use as exposure and DKA as the outcome including the aforementioned covariates over 34 years of follow-up using calendar year time scale (randomization dates were used as the left-truncated entry times). We assumed that censoring was noninformative. As the proportional hazards assumption for CSII was not met, we fit models with time-dependent regression coefficients. The first used a natural cubic spline to model dynamic changes in the hazard ratio (HR) in a continuous fashion, and the second used a piecewise constant HR by decade. We report the contemporary remaining risk as the HR from the most recent decade of follow-up. To confirm that the observed trend from our multivariable model is not influenced by changes in variables that may protect from DKA over time,^19,20 in sensitivity analysis we examined crude (unadjusted) association. Where applicable, Poisson regression was used to compare crude rates. An α-level of 0.05 was used for tests of statistical significance. R software (version 4.2.1) and the survival package were used for the analysis.

Results

Baseline and follow-up characteristics of the 1441 participants are described in Supplementary Table S1 according to whether a participant had ever used CSII over 34 years of follow-up. CSII users were more likely to be female (53% vs. 34%; P < 0.001) compared with nonusers, but otherwise their characteristics were similar. As these variables and CSII use itself (over half were using CSII by 2010, e.g., shown in Supplementary Fig. S1) changed over time, we examined DKA risk according to time-varying variables and CSII use. There were 488 DKA events reported throughout follow-up, and the crude DKA rate was not significantly different between CSII users and nonusers (on average 1.3 vs. 1.1 per 100 person-years, P = 0.13).

Temporal trends and residual DKA risk in Cox proportional hazards models

Figure 1 depicts the HR of time-varying CSII exposure and DKA with 95% confidence intervals (CI) from 1983 to 2017 in the multivariable model. For reference, the adjusted single constant HR from this model was 2.43 (95% CI: 1.81–3.28; P < 0.0001), and this figure demonstrates that the HR declined substantially over time, beginning in the 1990s. Most importantly, since 2001, the DKA hazard associated with CSII use was no longer statistically significantly higher than 1.0. Of the covariates, female sex (HR 1.96; 95% CI: 1.49–2.59; P < 0.0001), shorter diabetes duration (HR 1.04; 95% CI: 1.00–1.09 per 1-year lower diabetes duration; P = 0.03), higher HbA1c (HR 1.47; 95% CI: 1.37–1.58; P < 0.0001), higher insulin dose (HR 1.59; 95% CI: 1.25–2.03; P = 0.0002), late neuropathy (HR 1.67; 95% CI: 1.12–2.48; P = 0.01), and MACE (HR 2.26; 95% CI: 1.17–4.37; P = 0.02) were all independently associated with higher DKA risk.

FIG. 1.

Hazard ratio estimates from the regression model using a natural cubic spline with 95% confidence interval of the time-varying exposure of continuous subcutaneous insulin infusion pumps for diabetic ketoacidosis over 34 years (1983–2017). This figure represents the hazard ratio over time holding all covariates (age, sex, duration of diabetes, HbA1c, BMI, insulin dose, early retinopathy [Early Treatment Diabetic Retinopathy Study 23-point scale values >3], amputations or ulcerations, and major adverse cardiovascular events) fixed. The x-axis of this graph begins in August 1983, marking the enrollment of the first participant in the Diabetes Control and Complications Trial, and extends to 2017 (corresponding to ∼39-year average diabetes duration), reflecting the endpoint of the most recent publicly available data. The gray shaded area represents the 95% confidence interval. BMI, body mass index; HbA1c, glycated hemoglobin.

Table 1 demonstrates the absolute DKA rates per 100 person-years by decade along with the crude and adjusted hazard rate estimates from the piecewise constant HR model. CSII-users experienced higher absolute rates per person-year, but by the 2000s the DKA rates fell below 1 per 100 person-years and were not significantly different between groups (HR 1.04; 95% CI: 0.62–1.75; P = 0.87). Notably, the crude model showed lower DKA risk magnitudes, but the trend of declining hazard by decade was consistent across models.

Table 1.

Number of DKA Events, DKA Rates per 100 Person-Years, and HRs for Time-Varying Continuous Subcutaneous Insulin Infusion Pump Exposure for Each Decade of Follow-up in the Crude and Adjusted Piecewise Constant HR Models with 95% Confidence Intervals and Associated P-Values

Decades^a	Insulin pump use^b (n)	Number of DKA events	DKA rate per 100 person-years	Crude hazard ratio (95% CI)	P-value	Adjusted hazard ratio (95% CI)^c	P-value
1980s	No (1150)	78	1.94	2.40 (1.32–4.35)	0.004	5.81 (3.28–10.29)	<0.001
1980s	Yes (291)	30	4.69	2.40 (1.32–4.35)	0.004	5.81 (3.28–10.29)	<0.001
1990s	No (914)	127	1.17	1.71 (1.21–2.42)	0.002	3.11 (2.22–4.37)	<0.001
1990s	Yes (521)	62	1.92	1.71 (1.21–2.42)	0.002	3.11 (2.22–4.37)	<0.001
2000s	No (638)	59	0.77	1.20 (0.74–1.94)	0.47	1.60 (0.99–2.61)	0.06
2000s	Yes (751)	55	0.95	1.20 (0.74–1.94)	0.47	1.60 (0.99–2.61)	0.06
2010s	No (522)	33	0.84	1.04 (0.62–1.75)	0.87	1.23 (0.73–2.11)	0.43
2010s	Yes (782)	44	0.87	1.04 (0.62–1.75)	0.87	1.23 (0.73–2.11)	0.43

The 1980s decade includes data from 1983 to 1989, and the 2010s decade includes data from 2010 to 2017.

Participants were classified based on their most recent reported continuous subcutaneous insulin infusion pump use status at the time of the DKA event, while n refers to the number of ever or never CSII users by that time.

The adjusted model includes the following covariates: age, sex, and duration of diabetes as well as time-varying HbA1c, BMI, insulin dose, early retinopathy (Early Treatment Diabetic Retinopathy Study 23-point scale values >3), amputations or ulcers, and major adverse cardiovascular events.

BMI, body mass index; CI, confidence interval; CSII, continuous subcutaneous insulin infusion; DKA, diabetic ketoacidosis; HbA1c, glycated hemoglobin; HR, hazard ratio.Bold values indicate p-values <0.05, which met our threshold for statistical significance.

Discussion

DCCT/EDIC data from 1983 to 2017 demonstrate a major decline in DKA risk associated with CSII use and a low magnitude of residual risk in the latest decade (HR 1.23; 95% CI: 0.73–2.11; P = 0.43 in the 2010s). While they confirm the historic concerns about DKA (adjusted risk >5-fold in the 1980s), these findings should reassure clinicians about modern-day implementation of CSII therapy in the type 1 diabetes population.

Our analysis is the first to clearly define the temporal decline in DKA risk associated with CSII within a single longitudinal cohort. Regardless of CSII use, the rate of DKA declined in the whole cohort over time (as demonstrated in Table 1), which could be attributed to increasing age and advancements in basal insulin formulations.⁵ Among CSII users, the additional risk reduction is likely explained by the observations that overall risk decreases with age, with advancements in insulin formulations, as well as the evolving pump technology (i.e., tubing and infusion set occlusion alarms in the 1990s and sensor-augmented CSII in the mid-2000s), greater understanding of infusion set problems (i.e., cannula dislodgement), and patient education.^5
–7 Prior studies have not established secular trends. Rather, they report on the contemporary magnitude of risk in the 2010s.^8

–12 However, the estimates have been variable. Specifically, recent observational studies^8

–11 have suggested a lack of DKA risk in CSII users. On the contrary, a cluster-randomization trial from the 2010s¹² demonstrated substantially greater risk for CSII-users (17/156 in 24-month follow-up, corresponding to ∼5.4 events/100 person-years) compared with multiple daily injections (5/161 events, ∼1.6 events/100 person-years). Though we can conclude that the magnitude of DKA risk has declined, we cannot completely dismiss remaining risk associated with CSII use as negligible since the DCCT/EDIC population may not be fully representative of the type 1 diabetes population.

Our findings have limitations. While there was potential for selection and volunteer bias, we focused on secular trends rather than absolute magnitude of risk. As changing protective factors or risk factors over time could impact observed DKA risk, we highlight that crude risk demonstrates the same secular trend as adjusted risk. Our study population also comes from an open cohort recruited during the 1980s; it was not a dynamic cohort and therefore we were unable to include people who were diagnosed with type 1 diabetes after this time. Newer diabetes technologies, such as continuous glucose monitoring and automated insulin delivery systems, were not assessed as the follow-up time preceded their availability in the United States and Canada where the study was conducted.

Although there was over a 5-fold greater DKA risk associated with CSII use in the 1980s, this risk has substantially declined and is of low magnitude since the early 2000s. Accordingly, clinicians should feel confident that DKA risk should not deter initiation of CSII in suitable patients with type 1 diabetes.

Footnotes

Acknowledgments

The authors are grateful to the study participants whose time and effort are critical to the success of the DCCT/EDIC research program that we accessed through the National Institute of Diabetes and Digestive and Kidney Diseases Central Repository Central Repository. D.R.B. was supported with protected research time by the University of Toronto Department of Medicine Scholarly Activity Subcommittee. The authors are grateful for unrestricted support from the Savlov Family Endowed Fund, BMO Bank of Montreal, and the Barwise Family Fund.

Authors’ Contributions

D.R.B., B.A.P., and L.E.L. researched data in this analysis and wrote the article. B.A.P. is the guarantor of this work. All authors have reviewed the article and have approved the final version of this article. D.R.B., B.A.P., and L.E.L. had full access to the data and take responsibility for the integrity of the data analysis.

Author Disclosure Statement

B.A.P. has received honoraria for educational events from Medtronic, Novo Nordisk, Sanofi, Insulet, and Abbott. His research institute has received funding from BMO Bank of Montreal and Novo Nordisk for research Support. He has served as an advisor to Boehringer Ingelheim, Sanofi, Insulet, Abbott, and Vertex. D.Z.I.C. has received honoraria from Boehringer Ingelheim-Lilly, Merck, AstraZeneca, Sanofi, Mitsubishi-Tanabe, Abbvie, Janssen, Bayer, Prometic, BMS, Maze, Gilead, CSL-Behring, Otsuka, Novartis, Youngene, Lexicon, Inversago, GSK, and Novo-Nordisk and has received operational funding for clinical trials from Boehringer Ingelheim-Lilly, Merck, Janssen, Sanofi, AstraZeneca, CSL-Behring, and Novo-Nordisk. D.Z.I.C. is supported by a Department of Medicine, the University of Toronto Merit Award, and receives support from the Canadian Institutes of Health Research, Diabetes Canada, and the Heart & Stroke/Richard Lewar Centre of Excellence in Cardiovascular Research. D.Z.I.C. is also the recipient of a 5-year Canadian Institutes of Health Research–Kidney Foundation of Canada Team Grant award. A.J. has received honoraria for educational events from Abbott, Acerus, AstraZeneca, Amgen, Bausch Healthcare, Bayer, Boehringer Ingelheim, Care to Know, Connected in Motion, Dexcom, Diabetes Canada, Eli Lilly, GSK, HLS Therapeutics, Janssen, MDBriefcase, Merck, Medtronic, Moderna, Novartis, Novo Nordisk, Pfizer, Sanofi Aventis, Timed Right, and WebMD. He has also served as an advisor to Abbott, Acerus, AstraZeneca, Amgen, Bausch Healthcare, Bayer, Boehringer Ingelheim, Dexcom, Eli Lilly, Gilead Sciences, GSK, HLS Therapeutics, Insulet, Janssen, Medtronic, Novo Nordisk, Partners in Progressive Medical Education, Pfizer, PocketPills, Roche, Sanofi Aventis, Takeda, and Ypsomed and has received research funding from Abbott, Amgen, and Novo Nordisk. M.F. was a consultant for ProofDx, a start-up company creating a point-of-care diagnostic test for COVID-19, and is an advisor for SIGNAL1, a start-up company deploying machine-learned models to improve inpatient care. The remaining authors have no conflicts of interest to declare.

Funding Information

This study was supported by Diabetes Canada (Operating Grant OG-3–21-5572-BP).

Supplementary Material

Supplementary Figure S1

Supplementary Table S1

References

Pickup

, Mattock

, Kerry

. 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(7339):705.

Pala

, Dicembrini

, Mannucci

. Continuous subcutaneous insulin infusion vs modern multiple injection regimens in type 1 diabetes: An updated meta-analysis of randomized clinical trials. Acta Diabetol, 2019; 56(9):973–980.

Ross

, Milburn

, Reith

, et al. Clinical review: Insulin pump-associated adverse events in adults and children. Acta Diabetol, 2015; 52(6):1017–1024.

Scheen

, Castillo

, Jandrain

, et al. Metabolic alterations after a two-hour nocturnal interruption of a continuous subcutaneous insulin infusion. Diabetes Care, 1984; 7(4):338–342.

Ehrmann

, Kulzer

, Roos

, et al. Risk factors and prevention strategies for diabetic ketoacidosis in people with established type 1 diabetes. Lancet Diabetes Endocrinol, 2020; 8(5):436–446.

Kesavadev

, Saboo

, Krishna

, Krishnan

. Evolution of insulin delivery devices: From syringes, pens, and pumps to DIY artificial pancreas. Diabetes Ther, 2020; 11(6):1251–1269.

Alsaleh

, Smith

, Keady

, Taylor

KMG

. Insulin pumps: From inception to the present and toward the future. J Clin Pharm Ther, 2010; 35(2):127–138.

Foster

, Beck

, Miller

, et al. State of type 1 diabetes management and outcomes from the T1D exchange in 2016-2018. Diabetes Technol Ther, 2019; 21(2):66–72.

Gandhi

, Ebekozien

, Noor

, et al. T1D Exchange Quality Improvement Collaborative. Insulin pump utilization in 2017–2021 for more than 22,000 children and adults with type 1 diabetes: A multicenter observational study. Clin Diabetes, 2024; 42(1):56–64.

10.

Auzanneau

, Karges

, Neu

, et al. Use of insulin pump therapy is associated with reduced hospital-days in the long-term: A real-world study of 48,756 pediatric patients with type 1 diabetes. Eur J Pediatr, 2021; 180(2):597–606.

11.

Karges

, Schwandt

, Heidtmann

, et al. Association of insulin pump therapy vs insulin injection therapy with severe hypoglycemia, ketoacidosis, and glycemic control among children, adolescents, and young adults with type 1 diabetes. JAMA, 2017; 318(14):1358–1366.

12.

Castro

, McCoy

, Cagan

, et al. Relative effectiveness of insulin pump treatment over multiple daily injections and structured education during flexible intensive insulin treatment for type 1 diabetes: Cluster randomised trial (REPOSE). BMJ, 2017; 356:j1285.

13.

Beck

, Riddlesworth

, Ruedy

, et al. Effect of initiating use of an insulin pump in adults with type 1 diabetes using multiple daily insulin injections and continuous glucose monitoring (DIAMOND): A multicentre, randomised controlled trial. Lancet Diabetes Endocrinol, 2017; 5(9):700–708.

14.

Nathan

, Group

, DCCT/EDIC Research Group . The diabetes control and complications trial/epidemiology of diabetes interventions and complications study at 30 years: Overview. Diabetes Care, 2014; 37(1):9–16.

15.

The 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(14):977–986.

16.

Lachin

, Bebu

, Nathan

, et al. Mortality in type 1 diabetes in the DCCT/EDIC versus the general population. Diabetes Care, 2016; 39(8):1378.

17.

Meunier

, Aguadé

, Videau

, et al. Age, male gender, and social deprivation are associated with a lower rate of insulin pump therapy initiation in adults with type 1 diabetes: A population-based study. Diabetes Technol Ther, 2021; 23(1):8–19.

18.

Karger

, Eckfeldt

, Rynders

, et al. Long-Term Longitudinal Stability of Kidney Filtration Marker Measurements: Implications for Epidemiological Studies and Clinical Care. Clin Chem, 2021; 67(2):425–433.

19.

Ladd

, Sharma

, Rahme

, et al. Comparison of socioeconomic disparities in pump uptake among children with type 1 diabetes in 2 Canadian provinces with different payment models. JAMA Netw Open, 2022; 5(5):e2210464–e2210464.

20.

Agarwal

, Schechter

, Gonzalez

, Long

. Racial–ethnic disparities in diabetes technology use among young adults with type 1 diabetes. Diabetes Technol Ther, 2021; 23(4):306–313.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.11 MB

0.13 MB