Unintended Insulin Pump Delivery in Hyperbaric Conditions

Introduction

Insulin pumps represent an important tool in type 1 diabetes management. ¹ Pumps allow to replicate pancreas insulin secretion more accurately than the insulin injection by pen. Their use in clinical practice has steeply grown in recent years.

Despite their great utility, pump technology could be affected by technical problems. ² In 2011 it was reported that atmospheric pressure reduction caused predictable, unintended pump insulin delivery. ^3,4 This mechanism is responsible for the creation of air bubbles ⁵ and expansion of existing bubbles. This phenomenon is considered responsible for glycemic perturbation in patients affected by diabetes mellitus under pump insulin treatment, especially in the case of low-basal insulin requirement or high insulin sensitivity. These data were collected in hypobaric conditions miming the changes in the atmospheric pressure during a common flight.

Recently new models of insulin pumps have been introduced in the market: wearable pumps, technologically advanced integrated pumps with continuous glucose monitor, or pumps with preloaded insulin cartridges. Most of them are now waterproof certified for an immersion from 1.3 to 7.6 m, according to the different models. ^{5 –7} It is unclear if the changes in the atmospheric pressure can influence the delivery of insulin also in these new pumps. In particular, it remains to be clarified if unintended insulin pump delivery happens during atmospheric pressure changes in hyperbaric conditions, such as the pressure variations during diving activities.

Our aim is therefore to verify the eventual variation of the insulin delivered by the pumps induced by atmospheric pressure changes in hyperbaric conditions.

Materials and Methods

The study was conducted from January to June 2016. The following pumps were studied: OmniPod (Insulet Corporation, MA; n = 13, of four different Lots: 42552, L42442, L42304, and L41874), Medtronic 640G (Medtronic, MN; n = 4; n NG110588H, NG1070838H, NG1046995H, and NG1070488H), and Roche Insight (Roche, Germany; n = 3; SN48100010536, SN31004489/2014, and SN31003104/2014). All the pumps were loaded by aspart insulin (1.8 mL in 640G, 2 mL in OmniPod). Roche requires 2.0 mL preloaded insulin cartridges. All the pumps were set with 2 U/h insulin flow rate, with the exception of OmniPod at 20 U/h, a flow rate thrice higher than the maximum basal insulin requirement observed in our patients.

The infusion set (or directly the cannula for OmniPod) was connected to microtubules as previously described. ⁴ Microtubules were weighed before and after each test to determine the insulin volume released in each time period. The scale (Sartorius BA110S; Data Weighing Systems, Inc., Elk Grove, IL) has a resolution of 0.1 mg. Four OmniPod pumps were loaded with insulin and with 250 μL of air (12.5% of the cartridge total inner volume), and the insulin volume was reduced accordingly to 1.75 mL.

All the pumps were placed inside a hyperbaric chamber at the Niguarda Hospital in Milan (Italy) at 25°C. Figure 1A describes the pressure changes in our experimental design. After 10 min of basal conditions at a pressure of 1.0 ATA (760 mmHg) inside the hyperbaric chamber, the pressure was increased at 1.3 ATA (988 mmHg) in 10 min. 1.3 ATA corresponds to a depth of about 3.0 m. Thereafter, the pressure was maintained at 1.3 ATA for an additional 10 min and then reduced to 1.0 ATA again in 10 min. The depressurization of 0.3 ATA corresponds to the changes in the atmospheric pressure during a flight takeoff (from 760 mmHg at sea level to 584 mmHg, at about 8000 feet, the usual height of a common flight).

OPEN IN VIEWER

FIG. 1.

Basal rate insulin release of different pump models in different atmospheric pressure conditions. In (A) the pressure variations are showed in the upper part of the figure: the different 10 min phases of our experimental design are indicated with numbers from 1 to 5: 1 indicates basal pressure (1.0 ATA); 2 pressure increase from 1.0 to 1.3 ATA; 3 stable pressure at 1.3 ATA; 4 pressure decrease from 1.3 to 1.0 ATA; and 5 basal pressure again (1.0 ATA). In (B) are showed the 10 min insulin release of OmniPod (in black), Insight (in white), and 640G (in gray) during the different phases of our experimental design. ATA, atmosphere.

To study the exclusive effect of the depressurization on insulin pump delivery, not linked to a previous pressurization, two additional OmniPod pumps were loaded with the insulin directly into the hyperbaric chamber at 1.3 ATA. After the loading the atmospheric pressure was decreased to 1.0 ATA.

Data are expressed as mean ± SD. Generalized hierarchical models for repeated measures were applied to take into consideration the multilevel nature of the data. P = 0.05 was considered statistically significant.

Results

As soon as the atmospheric pressure increases, three out of four Medtronic pumps totally stopped the delivery of insulin.

As a consequence of the increased pressure upon the key device, the pump indicated that “a key is not to be pressed for >3 min.” Data on Medtronic pump are therefore achieved only with one pump (data represent the medium value of three replicated tests on the same device). In hyperbaric conditions the atmospheric pressure variations affected the basal rate of insulin delivery in all the pumps studied (Table 1).

Some additional tests were performed to understand the origin of the unintended insulin delivery.

The role of the air bubbles was explored adding 250 μL of air into four OmniPod pumps at 2.0 U/h. During the compression a significant insulin reduction was observed (0.39 ± 0.12 U during 10 min basal versus 0.02 ± 0.02 U during 10 min compression), and during the decompression a huge insulin increase was verified (0.39 ± 0.12 U versus 6.75 ± 2.05 U, n = 4, P = 0.008).

In two OmniPod pumps where the basal infusion rate was increased up to 20 U/h unexpectedly, no insulin variation was observed during the compression (data not shown), while the unintended insulin delivery during the depressurization still occurred (3.03 ± 0.12 U during 10 min basal versus 3.62 ± 0.12 U during 10 min decompression, n = 2).

In the two OmniPod pumps loaded directly inside the hyperbaric chamber at 1.3 ATA, the unintended insulin delivery was again confirmed during the depressurization to 1.0 ATA (0.30 ± 0.04 U during 10 min basal versus 0.57 ± 0.06 U during 10 min decompression, n = 2).

Discussion

We showed for the first time that atmospheric pressure variations affected insulin pump delivery in hyperbaric conditions. In particular, an unintended insulin delivery occurred when pressure decreased from 1.3 to 1.0 ATA and a decreased insulin release happened when pressure increased from 1.0 to 1.3 ATA. In our model of hyperbaric conditions, we therefore confirmed what was observed in response to similar pressure variations in hypobaric conditions. ⁴ This phenomenon was observed at a different extent in all the new three different models of insulin pumps tested.

The unintended insulin delivery could depend on a mechanic pump dysfunction consequent to the atmospheric pressure changes or on the volume changes of air bubbles inside the pump tubing and cartridge induced by pressure modifications, according to the Boyle's law. ³ The presence of air bubbles was previously suggested to be responsible for the unintended insulin release in hypobaric chamber. ⁴ The presence of air bubble within the insulin cartridge could be the consequence of an incorrect insulin loading. Alternatively, in a recent article, de novo bubble formation was described simply as the consequence of an atmospheric pressure change. ⁸ In our study two data seemed to support the role of air bubbles as responsible for the unintended insulin delivery. First, the modifications of the insulin delivery induced by pressure changes appeared limited in the devices with preloaded insulin cartridge (Insight), where the presence of preexisting air bubbles is virtually absent. Second the addition of a large amount of air (250 μL) into the reservoir (total final volume 2.0 mL) greatly amplified the phenomenon inducing an unintended insulin release of >1000%. According to the Boyle law, P1 × V1 = P2 × V2, where V1 and P1 are the volume and the pressure at 1.0 ATA basal condition, and P2 and V2 are corresponding values after the depressurization from 1.3 to 1.0 ATA. When 250 μL was introduced in the cartridge (V1 = 250 μL), the above equation becomes 250 μL × 1 ATA = Y × 1.3 ATA, where Y is the volume of air bubbles at 1.3 ATA that results to be equal to 190 μL. The final volume variation is therefore 60 μL (250–190 μL), which was the volume we expected to vary, when atmospheric pressure changes to 0.3 ATA (−60 μL during compression and +60 μL during decompression). This volume corresponded to the 6 U insulin released, as observed during the decompression. During the compression the decrease of insulin release sometimes did not follow the Boyle's law: when the OmniPod pumps were loaded with air bubbles the insulin decrease was lower than expected according to Boyle's law. The reason could be the limited accuracy of our method for volume below 1 μL, having our balance at readability of 0.1 mg.

In the case of the OmniPod pumps, the system is designed to minimize the possibility of air bubbles with the use of proprietary technology that reduces their occurrence during the filling of the reservoir. However, this technology is only effective if the OmniPod is fully filled as per the instructions for use. If the OmniPod is partially filled, then air bubbles may persist in the reservoir with the unfortunate effect described by the authors. Therefore, the users do need to follow the instructions for use to reduce the risk of these effects. In our experimental design the unintended insulin release during depressurization appeared to be independent of the previous decreased insulin release during the pressurization. In fact, the unintended insulin release appeared even when pumps were loaded directly at 1.3 ATA.

Three out of four Medtronic 640G pumps stopped to release insulin when exposed to an increase of the atmospheric pressure from 1 to 1.3 ATA (corresponding to a depth less than 2 m). This evidence appeared to be due to the effects of the atmospheric pressure on the pump keyboard. This might be taken into account by the patients in the case of diving activities. In the case of water activities, it is important to note that insulin pumps are designed to comply with IPX8 standard for waterproofing. This is a waterproof rating that is typically used for routine bathing and recreational swimming and not diving. In the case of routine bathing or recreational swimming, no significant changes in insulin delivery are expected. Sustained immersion at a depth of 25 feet (7.6 m) could occur mainly with scuba diving, which is certainly a special case.

Our observations have some important clinical implications. Insulin pump release can be modified during a flight or in the case of a dive at a depth <3 m and the patients should be advised to take into account the risks of unexpected variation in basal pump insulin rate. It is worthwhile to note that for most patients the magnitude of the changes is likely to be insignificant compared with the effect of increases in insulin resistance associated with the lack of physical activity during air travel. Some patients may experience even greater insulin resistance due to counter-regulatory hormones associated with the perceived stress of travel. However, in patients with low doses of insulin needs and high insulin sensitivity, such as children, small women, or in general patients with type 1 diabetes with an insulin sensitivity factor of 1:100 or superior, these findings are important.

In addition, patients should pay particular attention to avoid air bubble formation during the insulin loading procedures. The use of the preloaded insulin cartridge seems to be useful in reducing this phenomenon. Also healthcare professionals should advice their patients of the possible interference of atmospheric pressure changes and pump insulin release, particularly in children and patients with low insulin requirement. Patients should be acknowledged about the potential increased risk of hypoglycemia in between 1 and 2 h after the takeoff of a plane (when the depressurization takes place) and of hyperglycemia in between 1 and 2 h after the landing of a flight (when there's a pressurization) during pressurization.