Around the time of launching Fiasp, Novo Nordisk embarked on a wider study of its use in Continuous Subcutaneous Insulin Infusion (CSII), or pumping as we more commonly know it. This study, similar to others in the study of Fiasp, was designated ONSET 5 and launched in July 2016, completing in July 2017.
As many will be aware, the outcomes of the earlier ONSET 1 trial were widely lauded in the medical and diabetes press, with the conclusions drawn that:
Faster aspart effectively improved HbA1c, and noninferiority to IAsp was confirmed, with superior PPG control for mealtime faster aspart versus IAsp. Subjects randomized to postmeal faster aspart for all meals maintained HbA1c noninferior to that obtained with mealtime IAsp.
or in other words, it turned out to be quite the panacea when administered in MDI.
But the EMA approval also had some interesting issues, as I noted in this article:
- The Fiasp group experienced 700% of the premature set changes that Novorapid users saw. There were two premature set changes per 25 users in the Novorapid group versus 14 per 25 in the Fiasp group, none of which could be explained on a technical or occlusion basis
- 40% of Fiasp users suffered unexplained hyperglycaemia as opposed to 25% of the Novorapid users. Of those, six were unexplained, and seven were thought to be occlusion related, which was later disproved.
I also noted that the longer term pump trial would be very interesting in its outcomes given some of these short term issues.
Onset 5 has now completed, and the results of that trial are publicly available on the US National Library of Medicine Clinical Trials website and make for interesting reading. And so far, although the data was published on July 24th 2018, I’ve seen little to no fanfare as to the outcomes. So why is that?
The summary of the data published is that when used with CSII, whilst Fiasp demonstrates a faster action time, the indicators of variability in user glucose levels, different physiological requirements with the insulin and adverse effects on pump sites are still in evidence, and appear to back up the experiences of many people who have reported issues. Read on for the more detailed information.
Firstly, let’s review the positive outcomes from the trial, looking at the secondary outcome of post-prandial glucose measures.
Taking a look at the study, there were initially 472 participants (all of whom had an Hba1C between 7.0% and 9.0%) and the study was double-blind and randomized. 233 completed the Fiasp arm and 230 completed the Insulin Aspart (Novorapid) arm. The study ran for 16 weeks, and the baselines were established using a range of rapid acting insulins.
The Post-Prandial Glucose (PPG) measures, as expected, show an improvement compared to the Novorapid data. The difference, however, might best be described as nominal. Why do I say that? Well the test title is:
Change From Baseline in 30-min, 1-hour, 2-hour, 3-hour and 4-hour PPG (Meal Test)
The description of the test is:
Meal test: The subjects were given a carbohydrate-rich standardised liquid meal immediately after bolus (faster aspart or NovoRapid®) infusion in the morning of the meal test. The subjects were to consume the meal as quickly as possible (within 12 minutes) and blood samples were drawn after 30 minutes, 1, 2, 3 and 4 hours from the start of the meal.
That’s right. We’re bolusing both Novorapid and Fiasp immediately before the meal.
What it shows is that Fiasp reduced post-prandial glucose levels by a maximum of 7.0%, compared to Novorapid having a maximum increase of 4.1%, both versus the baseline. At any point in the data presented, Fiasp’s post-prandial glucose lowering effect was between 8.8% and 10% greater than Novorapid over the first two hours. It also suggests that Novorapid is possibly worse than other rapid-acting insulins that were also in use when the baseline was established.
- Fiasp’s time of action is faster than Novorapid, as demonstrated by the reduction from baseline compared to Novorapid.
- Fiasp’s effect may not be as significant as many first thought, at least when used with CSII.
I’ve just focused on this particular outcome for the speed and area under the curve outcomes, but it’s safe to say that Fiasp performs faster than Novorapid in all of these tests. And that’s what we’re expected to take away.
It is also worth noting that the Fiasp users experienced nearly seven minutes less than their baseline time with glucose levels below 3.9mmol/l compared with the Novorapid cohort that spent just under three minutes more.
It’s when you look at some of the other outcomes that it becomes less clear. So let’s start with the primary outcome.
The primary outcome of this trial is:
Change in Glycosylated Haemoglobin (HbA1c)
Now I’ve railed against this particular choice many times, but leaving that aside for the moment, the, the mean baseline for both Novorapid and Fiasp was 7.49%, with a Standard Deviation (SD) for the Fiasp group of 0.55 and the Novorapid (NR) group of 0.53.
The 16-week results were:
- Fiasp – reduction from baseline of 0.06% SD: 0.50
- Novorapid – reduction from baseline of 0.14% SD: 0.44
In the primary outcome, although the numbers themselves are tiny, Fiasp appears to have performed worse than Novorapid.
In addition, the following are stated in the results:
Statistical Analysis 1 for Change in Glycosylated Haemoglobin (HbA1c)
Groups All groups Statistical Test Type Non-Inferiority Statistical Method ANOVA Treatment difference 0.09 95% Confidence Interval 0.01 to 0.17
In terms of the analysis done on this data, the report states that:
Non-inferiority of faster aspart was considered confirmed if the upper limit of the two-sided 95 % CI for the true treatment-difference D (faster aspart minus NovoRapid®) was below 0.4 %.
So although I don’t have access to this data, based on the above statistical analysis, Fiasp is considered non-inferior to Novorapid, even though the treatment outcome numbers appear to be marginally inferior. More information regarding how non-inferiority tests work and why non-inferiority can be stated can be found in this article.
For the purposes of this article, I’ll group the other outcomes into three groups – variability measures, physiological indicators and adverse indications. There’s quite a lot here, so I’ll endeavour to keep it relatively brief.
In terms of variability measures, I’m referring to indications for Time in Range, time spent in Hyper- or Hypo- glycaemia, and importantly the standard deviations around these. So let’s start with the time in range measure.
For the Fiasp cohort, Time in Range for both 4.0mmol/l-10mmol/l and 4.0mmol/l-7.8mmol/l was not as good as that for the Novorapid cohort, with the added point that the standard deviation was wider on the Fiasp cohort. The difference in time in range works out at around two percentage points in both cases, but the standard deviation shows a roughly 10% wider spread for Fiasp than for Novorapid.
The Fiasp cohort also spent more time in a hyperglycaemic state (described as >10mmol/l and higher) than the Novorapid cohort (2,018, or 6% more incidents >10mmol/l), whilst having a marginally reduced proportion of time in hypoglycaemia.
Summarising these, it appears that the variability measures show that the Fiasp cohort experienced more variability than the Novorapid cohort.
When I refer to physiological indicators, I’m talking about variables such as Total Daily Dose, Carb Ratio, Sensitivity Factor, additional Hba1C outcomes and others on the list of things that are indicators of or requirements for carb counting. These were listed as secondary outcomes in ONSET 5.
Let’s start with the reduced Hba1C measures. Two of the outcomes related to an Hba1C below 7%/53mmol/mol. In both of these, Novorapid performed better than Fiasp. The most notable being where the Hba1C was lowered to <7% without severe hypos, where only 18.6% of participants achieved that on Fiasp while 22.5% achieved it on Novorapid.
There was a lot of data collected looking at insulin units per day for basal, bolus and total. These figures are largely irrelevant, in that absolute unit comparisons fail to take into account the differences between users. We also have mean insulin units per kg of bodyweight per day, which may be a more useful indicator.
|Basal||0.3 (0.12)||0.3 (0.11)|
|Bolus||0.33 (0.17)||0.31 (0.16)|
You can see that there is very little difference between the two insulins in terms of units per kg of bodyweight per day, at either basal or bolus level, although Fiasp does show a wider spread on the variance.
If we then take the mean insulin carb ratio, we see that Fiasp needed more insulin per gram of carbs than Novorapid, however, for once the variability around the mean was greater for Novorapid.
- Fiasp – 9.13g/iu (3.22)
- Novorapid – 9.72g/iu (6.77)
Glucose sensitivity factor is also reviewed, and in that case, Fiasp users showed marginally more sensitivity but with wider variation than the Novorapid users.
In addition, the “Active Insulin Time” for both was given as 3.6 hours. Given the half life (according to Novo Nordisk) for both insulins is 57 minutes, this would be the point at which 10% of the insulin remained in the average person’s body. I guess what we can take from this is that the researchers felt that once you get below 10% you will not see any insulin activity (even though, as we’ve found from looping experience and larger boluses, this isn’t necessarily the case).
All in all, the physiological indicators appear to show little difference between the two, and if anything, it’s the wider variation in the Fiasp data that stands out.
The EMA report highlighted a significant increase in pump site issues compared to Novorapid, so in this section we look at the adverse outcomes section of the clinical trial and see if these were perpetuated.
Onset 5 defines treatment emergent adverse events as:
an event that has an onset date on or after the first day of exposure to randomised treatment (in week 0), and no later than seven days after the last day of randomised treatment (i.e., maximum week 16 + 7 days). The results are based on the on-treatment period.
Fiasp had 440 compared to 412 for Novorapid, or 6.8% more.
If we dig further into these, we find that there were 44 infusion site reactions with Fiasp compared to 32 with Novorapid. That’s 37.5% more, and in the Other Adverse Events section of the study results, we find the following data:
|Infusion site reaction|
|# participants affected / at risk||16/236 (6.78%)||7/236 (2.97%)|
We see from this that more than twice as many Fiasp cohort participants suffered infusion site reactions than Novorapid cohort participants and that on a per person basis, Fiasp users had 13.7% more incidents per affected person.
There were 1,185 unexplained cases of hyperglycaemia for Fiasp versus 1,058 for Novorapid, which is a 12% increase. Compared to what was seen in the EMA 6-week trial, however, this is substantially lower.
The other indicator of interest is the number of participants with at least one non-regular change of infusion sets. As per the table below, we can see there were differences:
|[Units: Number of subjects]|
|1. Perceived Occlusion||50||50|
|2. Any problem related to infusion set||108||75|
|3. Any technical issue with pump||23||17|
|4. Changes in insulin solution in set or reservoir||3||8|
|5. High BG with no other explanations||66||60|
|6. Infusion Site Reaction||16||8|
What stands out here is that 44% more Fiasp users had at least one issue with infusion sets than Novorapid users for reason 2, and that in general, 23% more Fiasp users made at least one irregular change to infusion sets than Novorapid users for any reason.
Notably, in this trial, there were similar numbers of perceived occlusions between the two cohorts, unlike the EMA results, where the Fiasp cohort perceived significantly more.
There’s a great deal of data in the study results, and I’ve not delved into all of it here. The three things that I take away from reading them are:
- Even using CSII, in general, Fiasp works more quickly than Novorapid, in line with the MDI results.
- Variability of results across the Fiasp cohort on CSII is noticeably higher than that for the Novorapid cohort.
- The number of adverse events, especially relating to site issues, is higher in the Fiasp cohort than the Novorapid cohort.
The greater variability in many of the results, displayed by a larger standard deviation, and the increase in adverse effects suggests that Fiasp isn’t a one-for-one swap with Novorapid, and similarly is unlikely to be for Humalog or Apidra. There is plenty of “in the wild” user experience to back this up.
There’s also an aside in the Novorapid vs Baseline measures that suggests that Novorapid is probably slower than other rapid-acting insulins in its action, as the Novorapid vs Baseline post-prandial readings resulted in worse outcomes for Novorapid than baseline.
Overall, I’m not sure the outcomes of this trial present an overwhelming argument to move the majority of users onto Fiasp. Whilst for a significant proportion of users, it clearly works without problems, even this controlled trial shows that there are differences between it and the older rapid-acting insulins that are likely to make it more difficult for some to use.
The significant increase in unexpected infusion set changes with Fiasp use is something worth noting, and I think demonstrates some commonality with real world experience of Fiasp. It would have been good to see not only how many participants had unexpected changes, but how many there were per participant.
I’d suggest that education about some of the issues that were presented in the trial should be given to users before they switch to Fiasp in order to fore-arm them and give them some guidance as to what might work best.
But overall, Onset 5 is not a compelling case to move pump users to Fiasp in its own right, and given the faster onset that should, in theory provide better performance, that’s mighty disappointing. And that maybe why the results have taken far longer than I was expecting to see the light of day.