Evidence review

Clinical and technical evidence

Regulatory bodies

A search of the Medicines and Healthcare Products Regulatory Agency website revealed 1 Field Safety Notice (MHRA reference: 2015/009/028/292/033) for the MiniMed 640G insulin pump (MMT‑1711 and MMT‑1712).

This relates to a malfunction of the pump drive motors, which would result in a pump error message alarm notifying the pump user that insulin is no longer being delivered. Medtronic has identified the cause of the issue and it has been corrected in the current manufacturing of the MiniMed 640G. There have been no reported failures from customers as a result of this issue.

No reports of adverse events were identified from a search of the US Food and Drug Administration (FDA) database: Manufacturer and User Device Facility Experience (MAUDE).

Clinical evidence

The search for evidence on MiniMed 640G insulin pumps identified 1 abstract that reports a case series study (Choudhary et al. 2015a). The manufacturer stated that 2 published studies and 1 further abstract reported on the predictive algorithm in the MiniMed 640G insulin pump (Buckingham et al. 2010; Danne et al. 2014, De Bock et al. 2014). The Buckingham et al. (2010) study was excluded because the insulin pump used was from a different manufacturer to that of the MiniMed 640G system. The study by Danne et al. (2014) and abstract by De Bock et al. (2014) were included as proof‑of‑concept studies.

The abstract (Choudhary et al. 2015a) reported findings from a multicentre prospective case series. The information presented was extracted from the full conference poster (Choudhary et al. 2015b; table 1), which has more detail than the published abstract. Forty patients with type 1 diabetes, including 24 adults and 16 children (aged 9–17 years), who had previously used sensor‑augmented insulin pump therapy were included. This study assessed the performance of the MiniMed 640G, its ability to help prevent hypoglycaemia and its acceptability to users. The patients used the MiniMed 640G for 4 weeks.

There were 2402 evaluable pump suspension events, in which the insulin administration was temporarily stopped once glucose levels were predicted to reach a pre‑set low glucose limit within 30 minutes. This was an average of 2.1 events per patient‑day. Of the 2402 suspended events, 2322 (96.7%) were before the pre‑set low was reached. In 83.1% of these, the sensor‑determined glucose value did not subsequently reach the pre‑set low limit. The results from the patient questionnaire suggested that the patients found the system easy to use and it helped them achieve better glucose control. There were 4 mild adverse events reported (2 skin reactions, 1 common cold and 1 urinary tract infection). A full report of this study has been submitted for publication but was not publicly available at the time this briefing was developed.

The study by Danne et al. (2014) reported a single‑centre prospective case series evaluating the potential benefits of a predictive threshold suspend algorithm (presented in table 2). The predictive low glucose management (PLGM) system comprised a Paradigm Veo insulin pump and an Enlite glucose sensor connected to a MiniLink transmitter, with the algorithm installed on a Blackberry phone (Canada). Twenty two patients with type 1 diabetes were given an overnight intravenous infusion of human insulin and glucose. The next morning, they participated in an exercise session while having insulin by continuous subcutaneous infusion. The exercise session was continued for a maximum of 4.5 hours or until reference HemoCue blood glucose reached 80 mg/dl. The exercise session consisted of up to 6 cycles of exercises, lasting between 15 and 30 minutes each, on a stationary bike or treadmill. A 30‑minute predictive horizon with a sensor threshold of 70 or 80 mg/dl was used. The threshold was set at 70 mg/dl for 2 of the experiments (exercise sessions) when hypoglycaemia occurred because of rapidly falling glucose levels during exercise. The remainder had a sensor threshold of 80 mg/dl. Insulin was suspended for up to 120 minutes when sensor glucose was predicted to be at or below the threshold within 30 minutes. Hypoglycaemia was prevented in 12 of the 15 experiments in which the PLGM system was triggered and insulin suspended. Hypoglycaemia (defined as sensor glucose less than 63 mg/dl) was not prevented in 3 of the exercise sessions when the PLGM system was triggered. The system was not triggered in 1 of the experiments for 1 patient, but no information is given about this patient and the reasons for the system failure are not clear.

In another study assessing the SmartGuard algorithm (De Bock et al. 2014; presented as a conference abstract, table 3), 10 patients were given the PLGM system comprising a Veo insulin pump, Enlite glucose sensor, MiniLink transmitter, Bluetooth‑RF translator and a smartphone with the algorithm installed. The study was a randomised, controlled, crossover trial but the information presented in the abstract is limited. Overnight hypoglycaemia was induced by increasing basal insulin rates by 180%. The parameters were set for pump suspension to occur when a sensor glucose level of less than 4.4 mmol/l was predicted to occur in 30 minutes. In the control arm (with SmartGuard off), the glucose level in 9 out of 10 patients fell below the pre‑set 2.8 nmol/l threshold. When SmartGuard was on (the intervention arm), only 2 of the 10 patients' glucose levels reached the 2.8 nmol/l threshold.

Recent and ongoing studies

One ongoing or in‑development trial on the MiniMed 640G system for type 1 diabetes was identified in the preparation of this briefing.

Trial NCT02130284: In‑clinic evaluation of the PLGM system in adult and pediatric insulin requiring patients with diabetes using the Enlite 3 Sensor.

Costs and resource consequences

Type 1 diabetes affects over 370,000 adults and 26,000 children in the UK. Most people with type 1 diabetes manage their blood glucose levels by regular daily self‑monitoring and insulin injections. A small number of people have problems maintaining the correct blood glucose levels by self‑administration of insulin therapy, and so are at risk of the severe complications associated with diabetes. Short‑term complications of type 1 diabetes are hypoglycaemia and diabetic ketoacidosis, which is when the body starts to use fat as an energy source. The longer‑term complications of type 1 diabetes, such as neuropathy, nephropathy, retinopathy and cardiovascular events, are related to hyperglycaemia. If automated continuous glucose monitoring and integrated insulin administration is shown to be a safe and effective option for these people, the incidence of complications associated with hypoglycaemia may be reduced, with associated cost savings for the NHS. There is no evidence that the MiniMed 640G system has any further benefit in reducing the risk of hyperglycaemia compared with any other pump systems.

Strengths and limitations of the evidence

The current evidence for the MiniMed 640G with the SmartGuard algorithm is limited to a single prospective case series presented in a conference poster. The conference abstract and associated poster (Choudhary et al. 2015a and 2015b) provided information on the ability of the MiniMed 640G to prevent hypoglycaemia. The authors reported a mean rate of 2.1 automatic suspend‑before‑low episodes per patient‑day. Most (83%) of these events happened before the sensor glucose levels reached the pre-set low limit suggesting the potential of MiniMed 640G to prevent hypoglycaemia events. This study has limitations inherent to a case series design. For example, it presents only descriptive results and lacks a comparator group. Because the study is presented only in abstract and poster formats, full methodological details are unavailable, and so it is not possible to evaluate the quality of this evidence fully.

The study (Danne et al. 2014) and abstract (De Bock et al. 2014) evaluating the algorithm are informative, although they do not evaluate the full integrated system. These studies were included as proof‑of‑concept for the SmartGuard algorithm. The study by Danne et al. (2014) was a single-centre case series with no comparator group and merely presented descriptive results. It is not possible to evaluate the quality of the randomised, controlled, crossover trial by De Bock et al. (2014) because the study is only presented as an abstract.