Accuracy Evaluation of a Novel Reusable Patch Pump Prototype

Abstract

A prototype of an insulin patch pump that operates with a double pump mechanism was tested. Accuracy of bolus delivery of 0.2 U and 1.0 U, and a basal rate of 1.0 U/h were evaluated with a microgravimetric method. In addition, occlusion detection time at basal rates of 0.1 U/h and 1.0 U/h was assessed. Mean deviation from target was lower than 2% for both bolus sizes. Regarding basal rate accuracy, mean deviation over 72 hours was lower than 1%. Occlusion detection occurred in less than 30 minutes with both basal rates. Our study results suggest that the tested pump prototype provides delivery accuracy and occlusion detection that is similar or even better compared to all commercially available pumps tested with a similar experimental approach.

Introduction

Continuous subcutaneous insulin infusion (CSII) is widely recognized as a useful tool for self-management of type 1 diabetes as it is associated with improved glycemic control and enhanced quality of life compared with a multiple daily injection insulin regimen. ¹ CSII requires an insulin pump which is used for basal rate (BR) and bolus delivery. Here, the demand for insulin pumps without visible infusion sets, often referred to as patch pumps, has increased due to their discreetness and ease-of-use. ² Despite this demand, the majority of currently available insulin pumps have visible infusion sets and only few tubeless patch pumps are available.

Sigi^TM (AMF Medical SA, Ecublens, Switzerland) is a novel patch pump currently under development. It features a disposable pad consisting of a pump holder welded on an adhesive and a flexible cannula which is inserted into the skin. The reusable part of the pump houses a standard 1.6 ml prefilled insulin cartridge and is closed with a cap, which is disposable, in order to avoid cross-contamination (Figure 1). Once the main body of the pump is thus assembled, it is clamped to the pump holder on the pad.

Figure 1.

Sigi™ insulin patch pump components (rechargeable pump, insulin cartridge and an easy-lock disposable cap). The pump is assembled and then clipped to a pre-inserted skin adhesive pad (the pad is applied to the patient’s skin through a 1-click pad applicator).

The pump uses an innovative and patented double pump system (Figure 2) composed of a primary classical piston pump and a secondary micro-delivery mechanism. The reason for this combination is that, from experience piston pumps are very accurate on the average delivered volume but may lack precision on each pumping cycle due to the stick-slip of the piston inside the cartridge. On the other hand, micro-pumps are assumed to be very precise on each delivery cycle, but, if not calibrated correctly, will cumulate the error and thus result in less accuracy of the volume injected on the long run. Associating the 2 technologies shall ensure both short- and long-term pumping accuracy and precision. This double pump mechanism is driven by a single motor that, simultaneously, advances the cartridge piston through a curved transmission rod with defined-distance micro-steps while activating the micro-delivery mechanism.

Figure 2.

Double pump mechanism consisting of a plunger with curved transmission rod (inside the reusable part of the pump) and a micro-delivery mechanism (inside the disposable cap) coupled to the occlusion detection system (explained below).

The micro-delivery sequence inside the disposable cap is depicted in Figure 3 (left). The system consists of 3 levers to control the administration of insulin and a Hall sensor to detect occlusions that may occur during dosing. The levers raise and lower, similar to keys on a piano, in a defined sequence to control the flow of insulin from the cartridge to the cannula for delivery into the body. First, lever 1 (input valve) is opened which allows the pressure inside the reservoir to fill the membrane chamber with 0.025 U of insulin, and subsequently closed. After that, lever 3 (output valve) is opened and lever 2 actuates to empty the chamber delivering a single micro-dose of 0.025 U. Lastly, lever 3 is closed and the cycle can repeat. These lever actuations are synchronized with the advancements of the primary mechanism, which together shall ensure precise insulin delivery. Figure 3 provides a schematic of the micro-delivery operation and the defined lever sequence in normal (left) and occluded (right) situations.

Figure 3.

Micro-delivery mechanism components (top), micro-delivery sequence in normal condition (left) and occluded condition (right). A magnet (dark gray rectangle) is attached to lever 2 which applies the necessary emptying pressure to the delivery unit during dose delivery. A Hall sensor detects the movement of lever 2 by sensing the distance between the magnet and the sensor. If there is an occlusion, the delivery unit cannot be emptied (ie, no insulin is delivered into the patient) and the magnet will not move down. Therefore, the distance between the Hall sensor and the magnet will be shorter than expected, indicating an occlusion.

In this investigation, insulin delivery accuracy and occlusion detection time of the described patch pump prototype were assessed.

Methods

Delivery accuracy and occlusion detection time (ODT) were assessed with 3 individual devices of the newly developed patch pump prototype. All tests were conducted in a laboratory setting at the Institut für Diabetes-Technologie, Forschungs- und Entwicklungsgesellschaft mbH an der Universität Ulm in Ulm, Germany. The pump prototype was not tested in humans; no ethical approval was required. Experimental testing was based on IEC 60601-2-24 and on procedures published previously.^3-5

To assess delivery accuracy, the pump infused commercially procured insulin (NovoRapid^® PumpCart^® Insulin Aspart, Novo Nordisk A/S, Bagsværd, Denmark) from outside the windshield of an analytical balance into a water-filled vessel placed on the balance’s grid weighing pan and weight gain was recorded. The insulin was transferred from the pump into the vessel via a 30 cm long fluoropolymer tube which replaced the pump’s original cannula. For this, the fluoropolymer tube’s end in the vessel was submerged into the aqueous phase which was overcoated with paraffin oil to avoid evaporation. Room temperature and humidity were within a controlled range.

Boluses of 0.2 U and 1.0 U were tested by delivering 25 consecutive boluses in each of 6 runs and recording weight gain 6 minute after a delivery was started. BR delivery of 1.0 U/h was tested over a duration of 72 hour, recording weight gain every 2.5 minute. Experiments were repeated 6 times. Bolus and BR accuracy were determined by comparing the expected weight increase with the recorded weight increase, using a density of 1.005 g/ml for insulin aspart. ⁶

For bolus accuracy evaluation, the relative deviation of each bolus from target was calculated. For BR evaluation, insulin delivery of non-overlapping 1-hour windows was calculated. Additionally, the total delivery of each 72-hour experiment was determined. For both basal and bolus delivery, percentages of boluses and 1-hour windows within ±5% and ±15% from the respective target were calculated.

ODT was determined using the pump’s original cannula. ODT was determined for BR of 0.1 U/h and 1.0 U/h by occluding the cannula with a clamp and measuring the time until an occlusion alarm occurred. The experiment was repeated 6 times per BR and mean, minimum and maximum ODT were calculated.

Results

For each bolus size, a total of 150 delivered boluses were assessed. Mean deviations from target were +1.8% ± 2.5% for the 0.2 U boluses and +0.3% ± 2.4% for the 1.0 U boluses, respectively. The percentage of boluses within ±5% of the target was 89.3% for the 0.2 U bolus and 95.3% for the 1 U bolus (Figure 4). All boluses were within ±15% from target.

Figure 4.

Accuracy of bolus delivery for volumes of 0.2 U and 1.0 U. Boxes represent the distribution of deviations from 150 boluses each.

Figure 5 displays the BR delivery in each 1-hour window during the 72-hour test period. Mean total deviation from target over 72 hours was -0.3% ± 0.3% with 99.8% of 1-hour windows within ±15% from target and 97.7% of 1-hour windows within ±5% from target.

Figure 5.

Basal rate delivery within non-overlapping 1-h windows of 6 experiments (colored dots) over a duration of 72 hour.

Mean ODT of 6 repetitions was 29 minutes at a BR of 0.1 U/h and 10 minutes using a BR of 1.0 U/h (Figure 6). Minimum and maximum ODT were 28 minutes and 30 minutes at a BR of 0.1 U/h and 8 minutes and 18 minutes at a BR of 1.0 U/h, respectively.

Figure 6.

Occlusion detection times for basal rates of 0.1 U/h and 1.0 U/h. Boxes represent the distribution of results for 6 runs per BR. Black boxes represent 50% of values including median. Antennae show the minimum and maximum waiting time until an occlusion alarm occurred.

Discussion

For patients using insulin pumps, reliability of insulin delivery, in particular regarding small doses, especially in the case of neonatal or young pediatric patients, is important to ensure safe and effective therapy measures.

This was the first accuracy evaluation study with this insulin pump which is currently under development. The prototype showed a high level of accuracy regarding the tested BR and both tested boluses sizes.

Comparability to data of previous studies is limited because of differences in the test setup (eg, a smaller number of individual pumps used and fewer repetitions performed). However, the results of this investigation showed that deviations from target were similar or lower compared to other patch pumps currently available on the market and that have been tested in a similar experimental setting.^7,8

In this bench test, occlusion alarms, which are an important safety feature of insulin pumps, were triggered noticeably quicker than for any insulin pump tested so far.^4,7

Conclusions

This first study’s results showed that the tested insulin patch pump prototype provides a delivery accuracy that is similar or even better compared to all commercially available pumps that have been tested in a comparable experimental setting and that ODT was distinctly shorter. Our results suggest that the tested insulin patch pump in its commercially available stage has the potential to represent a promising alternative to the pumps currently available on the market.