Automated Insulin Delivery Effects During Driving Among Older Adults with Type 1 Diabetes in a Randomized Trial

Abstract

Dysglycemia among drivers with type 1 diabetes (T1D) is associated with impaired driving performance, and glucose levels “above 5 to drive” are often recommended for insulin-treated drivers. Evidence for diabetes treatments that support euglycemia while driving is minimal, particularly for older drivers. In this randomized, crossover trial involving adults aged ≥60 years with T1D, we used continuous glucose monitoring (CGM) during driving to compare the first-generation closed-loop automated insulin delivery (AID) versus a sensor-augmented pump therapy. There were 1894 trips undertaken by 8 drivers (median age 68 years [IQR: 64–70]). During AID versus sensor-augmented pump, time in range >5.0–10.0 mmol/L was greater (100% [0–100] vs. 81% [0–100]; P = 0.033) and fewer trips had any CGM >16.7 mmol/L (3.5% vs. 6.4%; P = 0.006). Three percent of all trips included CGM <3.9 mmol/L, with no between-stage difference (3.0% vs. 3.5%; P = 0.52). System alerts occurred in 10% of all trips, with no between-stage difference (9% vs. 11%; P = 0.078). First-generation AID reduces hyperglycemic driving among older drivers with T1D, without increasing hypoglycemia. Developing dedicated “driving-mode” settings could prioritize safety while minimizing distraction.

Trial Registration: ACTRN12619000515190.

Introduction

Older drivers have greater vulnerability to injury and death from vehicle crashes than younger drivers.¹ Insulin-treated diabetes is an additional layer of concern for older drivers, with dysglycemia potentially impairing driving performance and increasing crash risk.^2
–4 Closed-loop automated insulin delivery (AID) has been shown to improve glycemia across a range of real-life activities.⁵ To our knowledge, there is no published evidence regarding glycemic effects of AID when driving. However, the first-generation AID has more system alerts than manual insulin dosing with a sensor-augmented pump therapy, potentially a hazardous distraction during driving.⁶

In this randomized, crossover trial involving older adults with type 1 diabetes, we compared glucose levels between the two trial stages: the first-generation closed-loop AID versus a sensor-augmented pump therapy. The present report covers the trial’s secondary outcomes relating to glucose levels while driving. Our hypothesis was that AID would improve glucose levels when driving. We also investigated the frequency of system alerts while driving, considering their impact on driver distraction and performance.⁷

Methods

The OldeR Adult Closed-Loop (ORACL) open-label, randomized, crossover trial compared glucose outcomes overall and during driving among older adults using a closed-loop versus sensor-augmented pump therapy (ACTRN12619000515190). Full ORACL trial eligibility criteria and primary outcome measures are published;⁸ in brief, eligible individuals were aged ≥60 years, with type 1 diabetes ≥10 years, and using an insulin pump. The eligibility for the driving study required individuals to be fully licensed active drivers (driven within previous month) with exclusive access to their vehicle.

Free-living driving was monitored throughout each 4-month trial stage (Supplementary Fig. S1). Data from the final 91 days of each stage were assessed (to avoid treatment initiation and carryover effects). Continuous glucose monitoring (CGM) data (Guardian Sensor3, Medtronic, Northridge, CA) were synchronized with trip information from vehicle logging devices (Drive 51 sat-nav; Garmin, Schaffhausen, Switzerland). MiniMed 670G systems (Medtronic) were used in their automated and manual insulin delivery modes for the closed-loop and sensor-augmented pump trial stages, respectively. System alert settings were clinically individualized. During the sensor-augmented pump stage, low-glucose suspension use was optional (per clinical individualization) and predictive low-glucose suspension was prohibited. Trip duration was calculated from vehicle logging device data, with trips concluding after 10 min of inactivity. Trips shorter than 1 km or without a CGM reading within 5 min before starting were excluded from the analysis.

CGM metrics during driving were calculated for time within, above, and below target glucose ranges. Two time-in-range (TIR) metrics were used: standard (3.9–10.0 mmol/L) and driving specific (>5.0–10.0 mmol/L); the latter was used due to the effectiveness of the “above-5-to-drive” recommendations.^9,10 Time-above-range (TAR) thresholds examined were >10.0 mmol/L, >13.9 mmol/L, and >16.7 mmol/L (hyperglycemic driving). Time-below-range (TBR) thresholds examined were <3.9 mmol/L and <3.0 mmol/L (hypoglycemic driving). Trips were counted in CGM categories if they involved at least one CGM reading beyond the specified threshold. System alerts during driving were assessed by type—related to low or high glucose, or unrelated to low or high glucose.

The unit of analysis was each individual trip. Trips with complete CGM data were eligible. Drivers with at least one eligible trip in both trial stages were included. No data were imputed. To accurately align trip times with 5-min CGM intervals, glucose data were linear interpolated to 1-min intervals. Group comparisons were evaluated using rank-sum tests (continuous variables) or Fisher’s Exact tests (categorical variables). Due to skewness and small overall sample size, it was not possible to adjust for within-participant correlation. No adjustments were made for multiple comparisons; two-sided P < 0.05 value was considered statistically significant. Analyses were performed using Stata 17 (StataCorp LLC).

Results

There were 1894 eligible trips (trial stages: closed-loop n = 904, and sensor-augmented pump n = 990; Supplementary Fig. S2) driven by eight drivers who were randomized evenly to the crossover order of trial stages. These drivers were six women and two men, with median age 68 years (IQR 64–70), type 1 diabetes duration 34 years (21–44), and driving experience 48 years (45–50; Supplementary Table S1). All eight drivers had low-glucose suspension activated during the sensor-augmented pump stage. A total of 46,928 driving minutes were eligible for the analysis (trial stages: closed-loop n = 21,126, sensor-augmented pump n = 25,802). No adverse events occurred during driving in either stage.

Driving time with CGM below 3.9 mmol/L was low overall (both stages had median TBR 0% [IQR 0–0] and fewer than 4% of trips with any CGM reading below 3.9 mmol/L), with no difference between stages (Table 1). Time spent in the driving-specific range was increased with closed-loop intervention (median 100% [IQR 0–100] vs. 81% [0–100]; P = 0.033). The proportion of time spent in the standard glucose range was median 100% (0–100) for both stages, with no between-stage differences. During the closed-loop versus sensor-augmented pump stage, there was less TAR (above 13.9 mmol/L and 16.7 mmol/L) and a lower percentage of trips with any CGM readings exceeding these thresholds (all P < 0.05); the percentage of trips with any CGM above 16.7 mmol/L was nearly halved (3.5% vs. 6.4%; P = 0.006; Table 1; Supplementary Fig. S3). Most trips had minimal variation in glucose levels (CV < 5%).

Table 1.

Continuous Glucose Monitoring Metrics and System Alerts during Driving by Trial Stage

	Closed-loop stage (n = 904)	Sensor-augmented pump stage (n = 990)	P value
CGM metrics during driving^a
Percentage of time at glucose concentration
3.9–10.0 mmol/L, %	100 (0.0, 100)	100 (0.0, 100)	0.11
>5.0–10.0 mmol/L, %	100 (0.0, 100)	80.8 (0.0, 100)	0.033
>10.0 mmol/L, %	0.0 (0.0, 100)	0.0 (0.0, 100)	0.19
>13.9 mmol/L, %	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.032
>16.7 mmol/L, %	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.005
≤5.0 mmol/L, %	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.31
<3.9 mmol/L, %	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.49
<3.0 mmol/L, %	0.0 (0.0, 0.0)	0.0 (0.0, 0.0)	0.88
Mean glucose concentration, mmol/L	8.8 (7.3, 11.0)	8.9 (7.0, 11.6)	0.45
SD of glucose concentration, mmol/L	0.16 (0.07, 0.35)	0.16 (0.06, 0.39)	0.62
CV of glucose concentration, %	1.8 (0.7, 4.4)	1.7 (0.6, 4.0)	0.41
Trip count (n) with glucose concentration:
All CGM readings 3.9–10.0 mmol/L	510 (56%)	530 (54%)	0.21
All CGM readings >5.0–10.0 mmol/L	469 (52%)	479 (48%)	0.13
Any CGM reading >10.0 mmol/L	367 (41%)	425 (43%)	0.31
Any CGM reading >13.9 mmol/L	106 (12%)	148 (15%)	0.043
Any CGM reading >16.7 mmol/L	32 (3.5%)	63 (6.4%)	0.006
Any CGM reading ≤5.0 mmol/L	68 (7.5%)	86 (8.7%)	0.40
Any CGM reading <3.9 mmol/L	27 (3.0%)	35 (3.5%)	0.52
Any CGM reading <3.0 mmol/L	6 (0.7%)	6 (0.6%)	>0.99
System (pump and CGM) alerts during driving^b
Total drive time (min)	21,126	25,802
Number of alerts (n)
All alerts	128	181
Low glucose–related alerts	31/128 (24%)	45/181 (25%)	>0.99
High glucose-related alerts	60/128 (47%)	103/181 (57%)	0.084
Other alerts (not related to either low or high glucose)	37/128 (29%)	33/181 (18%)	0.038
Trip count (n) with alerts:
Trips with any alerts	79 (9%)	111 (11%)	0.078
Trips with any alert related to low glucose	27 (3%)	29 (3%)	>0.99
Trips with any alert related to high glucose	37 (4%)	66 (7%)	0.015
Trips with only alerts unrelated to low or high glucose	28 (3%)	30 (3%)	>0.99
Rate of alerts (number of alerts per person per 30 min)^c
All alerts	0.13 (0.08, 0.23)	0.13 (0.03, 0.30)	0.89
Alerts related to low glucose	0.06 (0.0, 0.09)	0.04 (0, 0.12)	0.56
Alerts related to high glucose	0.03 (0.004, 0.09)	0.04 (0, 0.08)	0.93
Alerts unrelated to low or high glucose	0.04 (0.01, 0.07)	0.04 (0.02, 0.05)	0.49

Analyses performed on a per-trip basis. Results presented as median (IQR) or n (%). Comparison using rank sum test (continuous variables) or Fisher’s Exact test (categorical variables).

Analyses performed on a per-alert and per-trip basis for number of alerts and number of trips, respectively; analyses performed per driving time for alert rate. Results presented as median (IQR) or n (%). Comparison of number of alerts and trips using Fisher’s Exact test.

Analysis performed per total driving time; negative binomial regression was used for comparison of alert rate (accounting for the total drive time).

System alerts occurred in approximately 10% of all trips (190/1894), with three-quarters unrelated to low-glucose levels (Table 1). Overall alert rates during driving were comparable between the trial stages. However, consistent with the reduction in hyperglycemic driving during the closed-loop stage, there were fewer trips with alerts related to high glucose in the closed-loop versus sensor-augmented pump stage (n = 37 [4%] vs. n = 66 [7%]; P = 0.015). Conversely, there were proportionally more system alerts unrelated to either low or high glucose in the closed-loop versus sensor-augmented stage (n = 37 [29%] vs. n = 33 [18%]; P = 0.038). In both trial stages, low-glucose alerts predominantly occurred within the first 30 min of trips (Fig. 1A and 1B). As trip time progressed, there were increasing numbers of system alerts unrelated to either low or high glucose during the closed-loop stage (Fig. 1C); whereas alerts related to low glucose and to high glucose became more prevalent during the sensor-augmented pump stage (Fig. 1D).

FIG. 1.

Sensor glucose levels and system alerts during driving by trial stage. Panels A and B: gray lines represent sensor glucose trace during individual trips for the closed-loop stage (A) and sensor-augmented pump stage (B), respectively; dots denote individual alerts (blue dot = alert related to low glucose; red dot = alert related to high glucose; green dot = alert unrelated to either low or high glucose). Panels C and D: summary distribution of alert rate (per 100 trip minutes) since trip start for the closed-loop stage (C) and sensor-augmented pump stage (D), respectively; colored bars denote alert by type (blue = alert related to low glucose; red = alert related to high glucose; green = alert unrelated to either low or high glucose).

Discussion

In this randomized trial involving older adults with long-duration type 1 diabetes, a closed-loop AID system reduced the proportion of time spent in, and number of trips with, a hyperglycemic state compared with a sensor-augmented pump therapy. Reassuringly, there was no increase in hypoglycemic driving with AID, nor any issues related to driving safety, albeit among a small sample size. To our knowledge, this is the first trial to evaluate the impact of diabetes technology on glycemia when driving. Our data build upon previous studies showing hyperglycemic driving is far more common than hypoglycemic among drivers with type 1 diabetes.³ With minimal hypoglycemic driving in the present study overall, and active low-glucose alerts throughout both trial stages, it is unsurprising that the impact of AID intervention was only observed within elevated glucose levels. Although the connection between chronic hyperglycemia and diabetes-related complications is well-established,¹¹ studies have also highlighted impact of acute hyperglycemia on cognitive functioning and driving ability. Severe hyperglycemia has been linked to both impaired driving performance³ and unsafe stopping.¹² Therefore, evidence showing AID’s effectiveness in reducing hyperglycemic driving holds crucial importance for drivers with type 1 diabetes, their healthcare teams, and other road users.

Our randomized trial is the first to consider insulin pump and CGM alerts from a road safety perspective, with approximately 10% of trips involving at least one system alert. Low-glucose alerts during driving are clearly beneficial for road safety. However, the auditory similarity of critical low-glucose alerts to other noncritical alerts may lead drivers to underappreciate importance of these alerts. Auditory distractions for drivers is an ongoing area of road safety research.¹³ We propose prioritizing a dedicated “driving mode” for diabetes technology including insulin pumps and CGM systems with alerts. While the issue of disrupting alerts and alarms from diabetes technology has been reported,⁶ the way it impacts driver performance has not been examined and warrants further investigation. To our knowledge, only two papers (both in 2024 by the same group) have explored the effectiveness of hypoglycemic in-vehicle alerts (multimodal alerts appear to be most effective).^14,15 This is also an important area of future research that should ideally influence design considerations for next-generation CGM devices.

This study’s major strength is its long-term monitoring of older drivers with type 1 diabetes, providing real-world trips and glycemic states within a randomized trial. None of these drivers were frail, and all were pre-existing pump users with overall glycemia within consensus recommendations for older adults with type 1 diabetes. Whether AID could benefit more vulnerable drivers (e.g., drivers with frailty) or with less well-managed glycemia (e.g., TIR below 50%) are important areas for future research. Younger drivers, with potentially increased crash risk than older drivers,¹⁶ who also have type 1 diabetes would be an important group to include in future research. The study’s small sample size and largely short drives during COVID-19 lockdowns call for caution when interpreting these findings. Larger, more diverse studies, and use of instrumented vehicle methodologies to monitor actual driving performance, will help fully elucidate the impact of AID during driving.

In summary, closed-loop AID improves glycemia among older drivers with type 1 diabetes compared with a sensor-augmented pump therapy. Future technology development should prioritize driving-specific modes to minimize nonglucose alerts, and research should involve other potentially vulnerable groups including youth and frail older drivers.

Footnotes

Acknowledgments

The authors would like to thank the study volunteers for their participation. The authors also acknowledge contributions of the clinical trial staff at St Vincent’s Hospital Melbourne and The Royal Melbourne Hospital, Melbourne, Australia.

Authors’ Contributions

S.A.M. led the study. S.A.M., S.T., P.C.G., M.H.L., R.J.M., D.N.O., N.A.O., V.S., and G.M.W. co-designed the randomized trial. S.T., H.J.K, S.V., and S.A.M. analyzed and interpreted the data. S.T. wrote the first draft of the article with input from S.A.M. and S.V. All authors critically reviewed the report and approved the final version.

Author Disclosure Statement

S.F. reports receiving honoraria for lectures from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Novo Nordisk, and Sanofi; and serving on advisory boards for Medtronic, Mylan, Pfizer, Sanofi, and Viatris. M.H.L. reports speaker honoraria from Medtronic. R.J.M. reports research grants from Novo Nordisk, Servier, Medtronic, and Gray Innovation; honoraria for lectures from Eli Lilly, Novo Nordisk, Sanofi, Astra-Zeneca, Merck Sharp & Dohme, Novartis, and Boehringer Ingelheim; travel support from Novo Nordisk, Sanofi, and Boehringer Ingelheim; serving on advisory boards for Novo Nordisk, Sanofi, Boehringer Ingelheim and Eli Lilly Diabetes Alliance, and AstraZeneca; and being principal investigator for industry-sponsored clinical trials run by Novo Nordisk, Bayer, Johnson-Cilag, and Abbive. D.N.O. reports serving on advisory boards for Abbott, Medtronic, MSD, Novo Nordisk, Roche, and Sanofi; receiving research support from Medtronic, Novo Nordisk, Roche, Lilly, and Sanofi, and travel support from Novo Nordisk and MSD. S.A.M. reports receiving speaker honoraria from Eli Lilly Australia and Sanofi-Aventis Australia; serving on advisory boards for Medtronic and Ypsomed; and that her institution has received support for research from Medtronic. All other authors declare no competing interests.

Funding Information

The ORACL trial was funded by Breakthrough T1D (3-SRA-2018–667-M-R), the Diabetes Australia Research Program and St Vincent’s Hospital (Melbourne) Research Endowment Fund. Medtronic supplied discounted insulin pumps and glucose monitoring devices for the study. S.A.M. has been supported by a Breakthrough T1D Research Award and University of Melbourne Paul Desmond Clinical Research Fellowship. S.V. was supported by the St Vincent’s Hospital (Melbourne) Research Endowment Fund. M.H.L. has been supported by a National Health and Medical Research Council (NHMRC) postgraduate scholarship, co-funded by Diabetes Australia. The funders of this study had no role in trial design, data collection, data analysis, data interpretation, or writing of the report. Breakthrough T1D; Diabetes Australia; St Vincent’s Hospital Melbourne

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Table S1

References

Rakotonirainy

, Steinhardt

, Delhomme

, et al.: Older drivers’ crashes in Queensland, Australia. Accid Anal Prev, 2012; 48:423–429.

Cox

, Ford

, Gonder-Frederick

, et al.: Driving mishaps among individuals with type 1 diabetes: a prospective study. Diabetes Care, 2009; 32(12):2177–2180.

Merickel

, High

, Smith

, et al.: Driving safety and real-time glucose monitoring in insulin-dependent diabetes. Int J Automot Eng, 2019; 10(1):34–40.

Liu

, Chihuri

, Andrews

, et al.: Diabetes mellitus and hard braking events in older adult drivers. Inj Epidemiol, 2024; 11(1):22.

Sherr

, Heinemann

, Fleming

, et al.: Automated insulin delivery: benefits, challenges, and recommendations. A consensus report of the joint diabetes technology working group of the European association for the study of diabetes and the American diabetes association. Diabetes Care, 2022; 45(12):3058–3074.

Chakrabarti

, Trawley

, Kubilay

, et al.: Closed-loop insulin delivery effects on glycemia during sleep and sleep quality in older adults with type 1 diabetes: results from the ORACL Trial. Diabetes Technol Ther, 2022; 24(9):666–671.

Karthaus

, Wascher

, Falkenstein

, Getzmann

. The ability of young, middle-aged and older drivers to inhibit visual and auditory distraction in a driving simulator task. Transportation Res Part F: Traffic Psychol Behav. 2020; 68:272–284.

McAuley

, Trawley

, Vogrin

, et al.: Closed-loop insulin delivery versus sensor-augmented pump therapy in older adults with type 1 diabetes (ORACL): a randomized, crossover trial. Diabetes Care, 2022; 45(2):381–390.

Trawley

, Stephens

, McAuley

, et al.: Driving with type 1 diabetes: real-world evidence to support starting glucose level and frequency of monitoring during journeys. Diabetes Technol Ther, 2022; 24(5):350–356.

10.

Kwon

, Trawley

, Vogrin

, et al.: Driving-related glucose patterns among older adults with type 1 diabetes. Diabetes Technol Ther, 2024; 26(5):335–340.

11.

Ceriello

. Postprandial hyperglycemia and diabetes complications: is it time to treat? Diabetes, 2005; 54(1):1–7.

12.

Barnwal

, Chakraborty

, Sharma

, et al.: Sugar and stops in drivers with insulin-dependent type 1 diabetes. Accid Anal Prev, 2022; 173:106692.

13.

Regan

, Oviedo-Trespalacios

. Driver distraction: mechanisms, evidence, prevention, and mitigation. In: Björnberg KE, Hansson SO, Belin M-Å, Tingvall C., eds. The Vision Zero Handbook. Springer: Cham, 2022, pp. 1–62.

14.

Bérubé

, Lehmann

, Maritsch

, et al.: Effectiveness and user perception of an in-vehicle voice warning for hypoglycemia: development and feasibility trial. JMIR Hum Factors, 2024; 11:e42823.

15.

Bérubé

, Maritsch

, Lehmann

, et al.: Multimodal in-vehicle hypoglycemia warning for drivers with type 1 diabetes: design and evaluation in simulated and real-world driving. JMIR Hum Factors, 2024; 11:e46967.

16.

Arnett

. Developmental sources of crash risk in young drivers. Inj Prev, 2002; 8(Suppl 2):ii17–ii23; discussion ii21–ii13.

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

0.18 MB

0.17 MB

0.36 MB

0.54 MB