Whilst watching EASD2020 this week, during the session on developing better insulins, a question was asked by multiple participants in relation to faster insulins. This question was “What are the clinical benefits of faster insulins?”. None of the presenters was easily able to answer that question.
Additionally, there have been posters comparing the use of Lispro and URLi (Ultra Rapid Lispro (URLi) demonstrates similar time in target range to Lispro with the Medtronic MiniMed 670G hybrid closed-loop system; Bruce Bode et al, source) and Aspart and Fiasp (Postprandial glucose control using the Medtronic Advanced Hybrid Closed Loop System: faster-acting insulin aspart vs insulin aspart; Melissa H Lee et al, source) in commercial closed loop systems on display. These have arrived at various conclusions, which we’ll go into shortly. In the meantime, I have also been using purely ultra-rapid insulins in a DIYAPS system and therefore have another set of data that can be added to this discussion and may help to start answering this question. We’ll start by looking at the research studies.
Firstly, the comparison of Lispro and URLi that was undertaken by Bruce Bode (2 sets of four weeks with crossover, n=42) showed no difference between URLi and Lispro in the 670G, as shown in figure 1, however, in the conclusions, the team stated that they thought that this was due to use of a basal only hybrid closed loop and that in a more advanced system, it would provide greater benefits.
When Aspart and Fiasp were compared in the Melissa Lee’s Medtronic Advanced Hybrid Closed Loop (AHCL) (each user had 6 weeks of each insulin in closed loop with a prior 2 week open loop period, n=12), the study showed an improvement over the breakfast postprandial period and generally a slightly lower postprandial glucose level. Full CGM data is shown in figure 2.
In another session, during the Q&A process, one of the presenters mentioned that an earlier version of the AHCL had been able to adjust to the variations in time to peak (TTP) and insulin action time (DIA) seen with the faster insulins, however neither of these studies suggest that there current commercial systems see that much difference with faster insulins.
As an OpenAPS user who had previously used Fiasp mixed with Humalog (at a 1:1 ratio) to reduce the effects of site issues caused by Fiasp, and as recently documented, I had decided to try mixing the two faster insulins (Fiasp and Lyumjev (URLi)) to see if I could gain the benefits of purely faster acting insulin. By using OpenAPS, I also have the tools to adapt the insulin model in the system to take into account the differences between traditional rapid analogue insulins and the newer, faster insulins.
This provides a retrospective review of the data for one month prior to using Lyumjev on its own and one month after switching to a mix of Fiasp and Lyumjev, with a two day tuning period to allow Autotune to adjust basal rates, sensitivity factor and carb ratios after the issues seen with Lyumjev.
The settings for the algorithm are listed in the following table.
|Insulins used||Fiasp and Lyumjev|
|Time to Peak||45 minutes|
|Duration of Insulin Action||6 hours|
|Algorithm||oref1 with SMB, UAM and Autotune|
|Starting sensitivity factor||2.1 mmol/l/u|
|Starting carb ratio||9.8 g/u|
Table 1: oref1 configuration details
In the month prior to using the mix of the two ultra-rapid insulins, the ambulatory glucose profile (AGP) shown in figure 3 was observed.
In the month of use of the ultra-rapid mix, the AGP shown in figure 4 was observed.
Comparing the two, there isn’t a great deal of difference. The 4am-8am slot appears to have a slightly tighter distribution while the post lunch 25%-75% band appears to be sightly tighter around lunch time and have a lower variance throughout the day.
|Time <3.9 mmol/l (%)||6.6||3.9|
|Time 3.9-8 mmol/l (%)||74.8||80.8|
|Time 3.9-9 mmol/l (%)||82.7||88.6|
|Time 3.9 – 10 mmol/l (%)||87.7||91.9|
|Mean glucose (mmol/l)||6.4||6.4|
|Median glucose (mmol/l)||6.1||6.2|
|Estimated HbA1C (mmol/mol)||38||38|
|Glycemic Variability Index – 3.9-10mmol/l (%)||1.37||1.29|
|Patient Glycemic Status – 3.9-10mmol/l (%)||19.21||12.02|
Table 2: Outcome data comparison
Table 2 shows the numerical details. In this context, the differences between the two are more noticeable. The Fisap/Lyumjev mix shows reduced time spent below 3.9 mmol/l and also greater time in range at all levels measured. TIR 3.9-8mmol/l increased under the mix by 6 percentage points, which is an overall 8% improvement.
Whilst the mean (and HbA1C estimate) remained the same across both, the reduced time both hypoglycaemic and hyperglycaemic accounts for this. The measures of variance included show a reduced standard deviation (-10%), reduced patient glycaemic status (-37%) and reduced glycaemic variability index (-6%), all indicating that use of the ultra-rapid mix resulted in reduced glycaemic variation.
Whilst the ability to react to high levels more quickly would be expected to bring hyperglycaemia down, it isn’t obvious why hypoglycaemia would be reduced. The hypothesis is that with the faster clearance provided by Lyumjev, as we move down the tail and into insulin stacking range, the amount of insulin on board would be lower and would therefore result in fewer hypoglycaemic events.
Whilst the research studies aren’t generally able to show significant differences, the observational experiment, whilst being only n=1, does appear to highlight that the major benefit of using faster insulins is the reduction in glycaemic variation that takes place. Some might argue that this variation is marginal, however studies have shown that increased glycaemic variation contributes to vascular damage, so a reduction even of 5% would appear to be beneficial. The RESCUE study, also presented at EASD2020, suggests that a lower time in range increases the risk of microvascular complications, so the increased TIR that was demonstrated in the n=1 study is potentially beneficial in that respect.
Additionally, whilst there are no changes to estimated Hba1C across the two insulin configurations, the increased time in range and reduced variation appear to have been possible as a result of the flexibility built into DIYAPS systems that allow adjustment of peak time and DIA within the algorithm. Whilst some of the commercial systems are able to observe this and adjust accordingly, results from the two commercial system studies suggest that this is something that is perhaps not fully effective in what is currently available, or at least, with which these insulins have been tested.
Overall then, what’s the clinical benefit of using an ultra-rapid insulin? The breakfast period of the study with the AHCL and my n=1 observations appear to answer this question:
Faster insulins used in a suitable automated insulin delivery system that is able to account for the changes in in the insulin characteristics would allow a reduction in glycaemic variability and an increase in time in range, with all the benefits that these changes bring, .