Search tips
Search criteria 


Logo of diaMary Ann Liebert, Inc.Mary Ann Liebert, Inc.JournalsSearchAlerts
Diabetes Technology & Therapeutics
Diabetes Technol Ther. 2009 April; 11(4): 207–210.
PMCID: PMC2842075

Is an Automatic Pump Suspension Feature Safe for Children with Type 1 Diabetes? An Exploratory Analysis with a Closed-Loop System



It has been proposed that the first step towards a closed-loop artificial pancreas might be to use a continuous glucose sensor to automatically suspend the basal insulin delivery based on projected low sensor glucose values.


We reviewed our recent experience with an artificial pancreas system, utilizing a proportional-integrative-derivative (PID) algorithm, in 17 adolescents with type 1 diabetes (T1D) to assess the safety and efficacy of this maneuver.


During 34 h of closed-loop automated insulin delivery, 18 pump suspensions ≥60 min (90 ± 18 min) occurred in eight subjects. Sensor glucose levels fell from 159 ± 42 mg/dL to a nadir of 72 ± 13 mg/dL. Corresponding plasma glucose levels fell from 168 ± 51 to 72 ± 16 mg/dL, with values <60 mg/dL recorded in only four of the 18 events.


These data suggest that automatic pump suspension using the PID algorithm may be an effective means to prevent hypoglycemia in youth with T1D.


Despite the clear imperative to achieve tight glycemic control to prevent complications of type 1 diabetes (T1D), no insulin regimen will be optimal until it is linked to plasma glucose (PG) levels. The availability of continuous glucose monitoring systems has allowed for the development of closed-loop systems, and short-term inpatient studies have demonstrated that feedback-controlled insulin delivery using external sensors and external insulin pumps is feasible.17 Before a fully automated closed-loop (FCL) system becomes available for outpatient management of diabetes, a reasonable incremental step would be to use the glucose sensor to reduce the risk of hypoglycemia by automatically suspending basal insulin delivery for a predetermined period of time. Such programs may be based on reaching threshold glucose levels or on more sophisticated algorithms. The efficacy of such algorithms has just begun to be explored.8 It is unknown whether transient suspension of basal insulin infusion overnight would predispose the wearer to risk of hyperglycemia or ketoacidosis. The data generated during our recent pilot study of the feasibility of an external subcutaneous closed-loop system provided us with the opportunity to examine these questions.

Subjects and Methods

The subjects and methods employed in this study have been reported in detail elsewhere.7


Seventeen subjects with T1D (seven girls, 15.9 ± 1.6 years old, duration of diabetes 6.3 ± 3.7 years, glycosylated hemoglobin 7.1 ± 0.8%) receiving open-loop insulin pump therapy were recruited for the study. Patients with history of chronic medical condition other than hypothyroidism and celiac disease under treatment were excluded. Written informed consent was obtained from subjects ≥18 years old, and for subjects <18 years old, written parental permission and subject assent were obtained. The study was approved by the Yale University Human Investigations Committee, New Haven, CT.

Closed-loop protocol

Subjects were admitted to the Yale Hospital Research Unit on the afternoon prior to the closed-loop study. Two subcutaneous glucose sensors were inserted in the subcutaneous tissue of the anterior abdominal wall, and an intravenous catheter was placed into the antecubital vein for frequent blood sampling. Closed-loop control was started at 7 a.m. the next morning and continued for 34 h. In eight subjects the insulin infusion was under the complete control of the computer (FCL), whereas the other nine subjects received a pre-meal priming bolus (~25–50% of the dose that the subject would have taken for the meal) 15 min prior to the meal, followed by computer-controlled insulin administration (hybrid closed-loop [HCL]).

Venous blood glucose levels measured at the bedside with a YSI 2300 glucose analyzer (YSI Life Sciences, Yellow Springs, OH) were determined every 30 min from 6 a.m. to 10 p.m. and every 60 min from 10 p.m. to 6 a.m. Extra PG measurements were obtained if the plasma or sensor glucose (SG) values fell below 70 mg/dL. Mean ± SD point accuracy of the sensor was 13.2 ± 10.9%.7 Meals were provided at 8 a.m., noon, and 5 p.m. Glucose sensors were calibrated approximately every 12 h.

Closed-loop system

The closed-loop system consisted of three components: a Medtronic Paradigm® 715 insulin pump, a Medtronic continuous glucose sensor, and the Medtronic external Physiological Insulin Delivery (ePID) algorithm (all from Medtronic Diabetes, Northridge, CA). Insulin delivery was based on a model of the beta cell's multiphasic insulin response, consisting of three components: proportional (P), integral (I), and derivative (D). The P component delivers insulin in proportion to the difference between the SG and target glucose, the I component responds to persistent hyper- or hypoglycemia, and the D component responds to acute changes in SG. Expressed more informally, component P increases insulin delivery when glucose is above target and reduces insulin delivery when glucose is below target with no contribution to the delivery when glucose is at the target level. Component I provides insulin when the glucose is at target level, comparable to the basal insulin secretion, and is incremented up or down in parallel to the glucose levels for persistent hyper- or hypoglycemia. Component D increases or decreases insulin delivery in response to the rate-of-change of SG levels. At each time n the sum of these three components, Pn + In + Dn, determines the total insulin delivery and is recalculated every 1 min.46 For this study, the target glucose level was set at 100 mg/dL from 6 a.m. to 10 p.m. and 120 mg/dL from 10 p.m. to 6 a.m. A more detailed explanation of the algorithm and other parameters for this experiment is described elsewhere.7

Insulin suspension event

Insulin delivery data during the 34 h of closed-loop control were recorded and stored on the laptop computer that ran the insulin delivery algorithm. We manually reviewed the datasets of all 17 subjects for time periods of at least 60 min for which insulin delivery was suspended; these were defined for the purposes of this analysis as “suspension events.” It should be noted that the suspension of insulin delivery by the PID controller is not determined by a “prediction” of hypoglycemia. Rather, insulin delivery is suspended on a minute-to-minute basis if the sum of the three components (Pn, In, and Dn) at any time (n) is ≤0, and this sum Pn + In + Dn is recalculated every minute; once the sum is >0, insulin delivery resumes for that minute. The SG level at the start of each suspension and the nadir SG value during the suspension event were compared with corresponding PG values. Glucose rate-of-change was calculated at the time of interest over the previous 15 min. A representative suspension event is shown in Figure 1.

FIG. 1.
Representative glucose and insulin tracing for one subject. (Top panel) Glucose readings with the blue line as the SG values and red dots that represent corresponding PG measurements. Green triangles represent meals. (Bottom panel) Demonstration of three ...

Statistical analysis

Data are expressed as mean ± SD values. Pearson correlations were used for comparisons of glucose rate-of-change with nadir and ΔSG levels. All analyses were performed using GraphPad (San Diego, CA) Prism® version 5.0.


During 34 h of closed-loop automated insulin delivery in 17 subjects, there were 18 insulin suspension events ≥60 min in eight subjects (mean duration 90 ± 18 min, range 60–132 min). As shown in Figure 2, mean SG fell from 159 ± 42 mg/dL to a nadir of 72 ± 13 mg/dL (range 41–92 mg/dL). Corresponding PG levels fell from 168 ± 51 mg/dL to a nadir of 72 ±16 mg/dL (range 49–105 mg/dl). Nadir SG fell below 70 mg/dL in eight of 18 events and below 60 mg/dL in two of 18; however, during these latter two episodes, corresponding PG values were both normal at the time of low SG values (80 vs. 41 and 96 vs. 51 mg/dL). True nadir PG values <60 mg/dL occurred in four of 18 suspension events. In three of these events low PG values would have been detected by the sensor if the low glucose alarm was set at 70 mg/dL (SG vs. PG: 65 vs. 49, 63 vs. 54, and 69 vs. 57 mg/dL). None of the automatic suspensions led to hyperglycemia (mean PG at the end of suspension 85 ± 61 mg/dL, range 59–170 mg/dL) or ketosis. Results were similar for FCL and HCL patients. Rate of fall in SG during the 15 min prior to suspension did not correlate with the SG level at the time of the suspension event or with the nadir SG (data not shown).

FIG. 2.
SG levels at the time of automatic pump suspension (triangles) and subsequent nadir SG levels (squares). Mean values are indicated by the horizontal bars, and corresponding suspension–nadir pairs are connected by straight lines. The dotted line ...


This study demonstrates the feasibility of incorporating an automatic pump suspension function to reduce the risk of hypoglycemia in pump-treated patients with T1D. Using the PID algorithm to automatically suspend the insulin infusion during periods of rapidly falling SG levels (in which the predominance of a large negative D component causes the sum P + I + D to be <0) or when glucose levels are below target (predominance of negative P component) was sufficient to avoid reductions in PG below 60 mg/dL in 78% of the suspensions. Since the automatic suspension is intended to supplement rather than the replace hypoglycemia alarm functions, it is noteworthy that all but one of the episodes in which PG levels fell below 60 mg/dL would have been detected if the hypoglycemia alarm was set at a SG value of 70 mg/dL and a confirmatory blood glucose measurement had been obtained. Because 10 of the 18 suspensions would have prevented hypoglycemia even before triggering a sensor alarm at 70 mg/dL, it is reasonable to speculate that use of the automatic suspension would reduce the frequency of false alarms, one of the main reasons why patients stop using these devices.9

It appears that the PID algorithm identifies two distinct conditions in which to suspend insulin delivery. The first typically occurred in the late postprandial period, as glucose levels fell rapidly after the prandial peak. During this condition, the large negative D component makes the sum P + I + D  0, thereby suspending insulin infusion. Resumption of insulin delivery would not occur until the rapid rate of change resolved. The other condition occurred when glucose levels were below target, even if not changing rapidly. Under this circumstance, when glucose levels were relatively flat and the D component is close to 0, the negative P component predominates.

A safety concern is that suspension of insulin for 1–2 h could potentially result in hyperglycemia, especially if starting PG values exceed target glucose levels. In our study, there were six instances in which the pump was suspended when SG levels were >180 mg/dL (Fig. 2). In all of these suspensions, PG and SG levels fell, and in no case did PG levels increase to the hyperglycemic range after the insulin infusion was suspended. These observations are consistent with previous studies that examined changes in PG during the first few hours after intentional suspension of basal insulin infusion in pump-treated patients.10 This phenomenon is undoubtedly related to the delays in absorption of subcutaneous insulin delivery. We previously reported that plasma insulin concentrations after a subcutaneous insulin bolus do not peak until almost 60 min, and peak insulin action does not occur until almost 100 min, after the bolus.11 Even with suspension of insulin delivery, therefore, there exists a pool of “residual insulin” infused during the previous hour that is pharmacologically active maintaining or lowering glucose levels during the time of the suspension.

It should be noted that use of the PID algorithm to temporally suspend the insulin infusion during closed-loop control is not strictly comparable to automatically suspending an insulin pump for a fixed interval of time based on an absolute threshold glucose level or the current rate of change of SG levels projected forward for a predetermined time horizon. Additional studies are needed to more clearly define the control parameters to activate such a function during open-loop control. Nevertheless, our findings support the potential utility of adding such a glucose-responsive program to an open-loop platform, which could serve as an important first step towards the goal of a fully closed-loop system for treatment of T1D.


This study is supported by grants from NIH-CTSA (RR 023423), the Juvenile Diabetes Research Foundation, and the Stephen I. Morse Pediatric Diabetes Research Fund.

Author Disclosure Statement

W.V.T. is a member of the speakers bureau for Eli Lilly and Co. and Novo Nordisk and is a consultant for Novo Nordisk. G.M.S. was previously employed by Medtronic. S.A.W. is a member of the speakers bureau for Eli Lilly and Co. E.C., K.L.S., and A.T.S. declare no competing financial interests exist.


1. Bequette BW. A critical assessment of algorithms and challenges in the development of a closed-loop artificial pancreas. Diabetes Technol Ther. 2005;7:28–47. [PubMed]
2. Hovorka R. Continuous glucose monitoring and closed-loop systems. Diabet Med. 2006;23:1–12. [PubMed]
3. Hovorka R. Chassin LJ. Wilinska ME. Canonico V. Akwi JA. Federici MO. Massi-Benedetti M. Hutzli I. Zaugg C. Kaufmann H. Both M. Vering T. Schaller HC. Schaupp L. Bodenlenz M. Pieber TR. Closing the loop: the ADICOL experience. Diabetes Technol Ther. 2004;6:307–318. [PubMed]
4. Steil GM. Panteleon AE. Rebrin K. Closed-loop insulin delivery—the path to physiological glucose control. Adv Drug Deliv Rev. 2004;56:125–144. [PubMed]
5. Steil GM. Rebrin K. Darwin C. Hariri F. Saad MF. Feasibility of automating insulin delivery for the treatment of type 1 diabetes. Diabetes. 2006;55:3344–3350. [PubMed]
6. Steil GM. Rebrin K. Janowski R. Darwin C. Saad MF. Modeling beta-cell insulin secretion—implications for closed-loop glucose homeostasis. Diabetes Technol Ther. 2003;5:953–964. [PubMed]
7. Weinzimer SA. Steil GM. Swan KL. Dziura J. Kurtz N. Tamborlane WV. Fully automated closed-loop insulin delivery versus semiautomated hybrid control in pediatric patients with type 1 diabetes using an artificial pancreas. Diabetes Care. 2008;31:934–939. [PubMed]
8. Buckingham BA. Cobry E. Clinton P. Gage V. Luckey K. Kunselman E. Wilkinson J. Chase PH. Preventing hypoglycemia using predictive alarm algorithms and insulin pump suspension [abstract] Diabetes. 2008;57(Suppl 1):A66–A67.
9. Chase HP. Beck R. Tamborlane W. Buckingham B. Mauras N. Tsalikian E. Wysocki T. Weinzimer S. Kollman C. Ruedy K. Xing D. A randomized multicenter trial comparing the GlucoWatch Biographer with standard glucose monitoring in children with type 1 diabetes. Diabetes Care. 2005;28:1101–1106. [PubMed]
10. Attia N. Jones TW. Holcombe J. Tamborlane WV. Comparison of human regular and lispro insulins after interruption of continuous subcutaneous insulin infusion and in the treatment of acutely decompensated IDDM. Diabetes Care. 1998;21:817–821. [PubMed]
11. Swan KL. Weinzimer SA. Dziura JD. Steil GM. Voskanyan GR. Steffen AT. Martin ML. Tamborlane WV. Effect of puberty on the pharmacodynamic and pharmacokinetic properties of insulin pump therapy in youth with type 1 diabetes. Diabetes Care. 2008;31:44–46. [PubMed]

Articles from Diabetes Technology & Therapeutics are provided here courtesy of Mary Ann Liebert, Inc.