The tl;dr Pharmacy Journal Club: Once-Weekly Insulin Icodec for Type 2 Diabetes
Steph’s Note: If you don’t know this by now, I LOVE journal club. Which makes it a bit weird that It’s been a little while since we’ve had a good journal club post. Plus, it’s been a little while since we’ve had anything on our seemingly favorite tl;dr topic: diabetes. So we’re here to remedy this, courtesy of a hot-off-the-press article from the New England Journal of Medicine just published last week. See what you think, and if you have additional thoughts or comments, we’d love to hear from you!
When I was trying to figure out what to post about this week, I did a fair amount of internet surfing. Being a bit tired of COVID info everywhere and consuming so much of our inboxes, I was really just craving some good, solid internal medicine. And BAM. There it was, ask and ye shall receive, like a shining beacon:
^^Me when I saw this article (Image)
“Once-Weekly Insulin for Type 2 Diabetes without Previous Insulin Treatment”
Now that ought to get your pharmacist sparks a-flyin!
In case you’re new to this site or want a refresher on insulin products already available in the US, we have a TON of posts on diabetes. Check out this one for everything you need to know about insulin types and this one for learning how to manage insulin pumps. We even have a whole cheat sheet dedicated to diabetes management!
Basically we love discussing diabetes. Which is why this journal club is going to be super fun.
So let’s dive on in.
Long, Longer, and Longest-Acting Insulins (Yet)
If you read the tl;dr everything insulin post, you’ll see there’s a trend towards creation of increasingly longer-acting products. These are developed in hopes of balancing mimicry of natural basal insulin release with improved medication adherence. How few times can a person inject themselves with insulin and still achieve consistent decent basal insulin levels?
Detemir and glargine were the original “long-acting” insulins, although they certainly are not equal in that respect. These are dosed once (or perhaps twice) daily, depending on the patient. Then along came degludec, which is the “ultra long-acting” insulin. It’s still once daily dosing, although its PK allows for more flexibility in dosing times.
But now this article introduces insulin icodec with once WEEKLY dosing. How is this possible?
Icodec is yet another insulin analog. It has 3 substituted amino acids that add stability and protection from enzymatic degradation. It also has a long fatty acid tail that helps with binding to albumin. These structural modifications allow this molecule to have a half life of ~196 hours, which, for y’all too tired at the end of the day to do math, is a little over a week! Much like detemir, icodec creates albumin-bound insulin depots that release fairly steadily over the course of about a 7 day time period.
We know that this reliance on binding to albumin for controlled release of drug can lead to hypoglycemia in patients with hypoalbuminemia in the case of detemir, and there’s some evidence of this with degludec as well. Should we anticipate a similar effect with icodec, and at what albumin levels might this be of concern?
Also, if the half-life is about a week and it generally takes about 5 half-lives to reach steady state, consider the amount of time needed to see the full effect of Icodec…
Sorry, getting ahead of myself. We haven’t even gotten to the journal article yet. But at least we know we want to pay attention to what the patients’ albumin levels might be in the study. And we want to consider when steady state is reached in light of the duration of the study.
Keeping these pieces in mind, let’s move to the article.
Once-Weekly Insulin for Type 2 Diabetes without Previous Insulin Treatment
Purpose: Like any pharmaceutical company with a new product, Novo Nordisk wanted to see if its novel (undoubtably heavily invested in) insulin baby works! That is, in true phase 3 study fashion, Is icodec effective, and is it safe?
(Interestingly though, this is a phase 2 clinical study despite its investigation of both safety and efficacy as compared to a standard treatment. Perhaps the classification as a phase 2 trial is related to other features of the study...we’ll have to wait and see…)
Breaking it down a little more, will once weekly insulin be a useful addition to the insulin arsenal in a population where adherence is key? Will it improve patients’ glycemic control to a degree that’s impactful? How does it compare to the current standard of long-acting insulin therapy - glargine? Even if it’s pharmacokinetically novel, do those properties offer any actual additional benefits for patients compared to what’s already available? If so, which patients may benefit the most?
So many questions…
Funding: Surprise, surprise. Novo Nordisk funded this study. But let’s not put the paper in the trash! (Assuming you printed it. But please don’t. It’s 2020. Save a tree.)
Somebody has to sponsor new drug studies - they’re certainly not free! And the NIH can’t sponsor all of them.
The manuscript also discloses that the investigators are a mix of academic and industry employees. The authors are quite detailed about who did what with what oversight, as well as who vouches for accuracy of the data. It feels pretty complete, like going around the table and having each person say, “Aye, I pledge my honor!” So, I suppose we should give them credit for this level of detail and disclosure.
This pup knows how to critically assess journal articles #lifegoals (Image)
That being said, Novo Nordisk was pretty involved in the entire process, even down to paying for a medical writing agency to help with the manuscript. So even though it all seems to be above board, we should still assess this critically to ensure no wool’s being pulled over anyone’s eyes.
Design: This was a randomized, double-dummy, double-blind, treat-to-target, active-controlled, parallel-group, multinational phase 2 trial.
Good grief, that’s a mouthful! Let’s break this down.
Essentially, they randomized participants to one of two active, actual treatments and followed both groups over the same time period. No cross over. Also, no placebo group here, which makes it more applicable to real-life where we wouldn’t necessarily want to let our patients’ hyperglycemia go unaddressed for both clinical and ethical reasons. Having an active comparator also makes it harder to demonstrate a significant difference between treatment groups, since of course it would be way easier to show improvements in blood glucose if we compared any insulin product to placebo.
Double-dummy and double-blind means neither group of patients knew which treatment they were receiving, which was possible because they got dummy injections to accompany their active drug injections. Makes it hard to determine which treatment group you’re a part of if you have to do injection regimens representative of both, right? (Although kinda stinks that you know not all the injections you’re doing are actually useful…)
Double-blind also means the investigators following the patients didn’t know which treatment group patients were in. If the investigators knew, that could introduce bias. Imagine this - perhaps an investigator is excited about a new insulin option for his patients, and so consciously or unconsciously, whether for good-hearted or selfish reasons, he could make dosing changes that were impacted by this knowledge. Maybe he gets a smidge more or less aggressive on dosing changes depending on the treatment group… what’s a couple units of insulin here or there…
Just better for everyone not to know. Keeps things more above board and out of an ethical gray area (as well as a biostatistical gray area with bias).
The treat-to-target idea is a relatively newer concept. It describes insulin trials in which regimens are adjusted in order to reach a predefined glycemic goal, usually the HbA1c. This design allows investigators to assess endpoints other than just plain blood glucose values, such as weight changes or hypoglycemia rates. So it’s a little better at addressing the entire “is this insulin worth it” question rather than just addressing “what happens with the blood sugar level” question.
And multinational. Hallelujah, that’s an easy one, no description needed.
Finally, this was a phase 2 study, likely because of the relatively smaller number of patients, the protocolized dosing regimen, and lack of ability to detect significant differences between endpoints. More on this in a few. But just know that even though this study investigated efficacy and safety, it was not the robust assessment required of a phase 3 study.
Inclusion Criteria: Adult patients aged 18-75 years old were included. They had to be insulin-naive (unless previous insulin exposure was for <14 days or for gestational diabetes only). They were diagnosed as diabetic at least 180 days before screening and had been stabilized on metformin with or without a dipeptidyl peptidase 4 inhibitor (DPP4-inhibitor), e.g. sitagliptin. Patients had to have an HbA1c of 7-9.5%.
So let’s take a minute here. This is pretty interesting. Normally when we talk about putting patients on long-acting medications, especially those that are titratable, that usually happens after we’ve stabilized them on the shorter acting version first. The patient with new onset atrial fibrillation generally starts with metoprolol 25mg PO TID and titrates as needed until she is rate controlled and converted to an outpatient-friendly, equivalent ER daily dosing regimen, yay?
Maybe not full on danger just yet… but at least something to watch closely for… (Image)
But here with this study, we’re talking about taking not only new-to-insulin patients, but potentially relatively new-to-diabetes patients, and putting them straight on a once weekly insulin dosing regimen. I mean, even at 180 days of diagnosis, patients may still be learning how to reliably check their blood glucoses using a glucometer and figuring out what foods to eat and not eat. So we’ll need to pay attention to duration of diabetes when we get to baseline characteristics to see just how new to diabetes these patients were.
If it doesn’t, this immediate initiation of once weekly insulin should activate your pharmacist goosebumps. I’m not saying it’s necessarily a bad thing, but your spidey-senses should be on alert for whether this is going to pan out in a safe way.
Also, these patients aren’t the wildly uncontrolled diabetic patients diagnosed with HbA1cs off the EMR chart. Their HbA1cs were between 7-9.5%, corresponding to average blood glucoses of 150-225 mg/dL. Considering our general goal of therapy is to achieve an HbA1c <7%, these patients weren’t starting too far off base. There’s also a big difference between an HbA1c of 7% and one of 9.5%. We’ll want to pay attention to how many patients fell where in this range at the point of enrollment to see how far they had to go to meet goal.
Exclusion Criteria: The list is pretty long. Morbidly obese patients with BMI > 40 kg/m2 were excluded. Also, anyone with current or recent coronary or vascular disease or severe heart failure symptoms was excluded. Significant renal or hepatic dysfunction disqualified patients from participation in the study, as well as most cancers (some skin cancers were allowed). Patients who received any medication for diabetes within the previous 90 days or any medication for >14 days known to affect glucose values or metabolism (thyroid supplementation, steroids, etc) were ineligible.
There was also a pretty ambiguous exclusion criteria of “any disorder…which in the investigator’s opinion might jeopardize the patient’s safety or compliance with the protocol.”
Understandably, it was a phase 2 trial. The investigators didn’t necessarily want the full real-life picture just yet. They wanted to see if their drug worked and was safe in a middle-of-the-bell-curve kind of population. They didn’t really want icodec to be prematurely disqualified from the race because an obese patient with diabetes and a recent history of a heart attack last month had another coronary event while enrolled in the study. They didn’t want to have to try to tease out whether the HbA1c went off track because of ineffective insulin or steroid taper initiation.
Of course, you and I know that those patients may very well need insulin too! But we may have to wait for more studies to figure out whether that choice is icodec.
Methods: Patients were randomized 1:1 to either once-weekly icodec + once-daily placebo (the dummy) injections or once-weekly placebo (the dummy) + once-daily glargine for 26 weeks of treatment. NOTE: Just because there’s a placebo involved does not mean this was a placebo-controlled trial! Both groups received active drug. The placebo injections were just used as dummy injections to maintain blinding.
Stratification was performed according to DPP4-inhibitor use. This helps us to know how much impact the icodec has in the presence versus the absence of this adjunctive medication class.
The scheme for screening, treatment, and follow up is below:
(Image)
Starting doses were fixed: either icodec 70 units once weekly or glargine 10 units once daily. Doses were adjusted weekly for both groups according to a pre-defined algorithm, which was based on the current and previous 2 days’ blood glucose values:
(Image)
Note that this algorithm uses self-monitored blood glucose values. That is, it relied on the patients having sufficient, reliable data in order to make a fixed dose adjustment. If the patients were missing a value, investigators basically just used whatever value(s) they did have. So theoretically, they could have been making adjustments to a weekly insulin regimen based on 1 or 2 blood glucose values.
Perhaps not too far off from real life for some of our patients who just plain don’t check their sugars…but most usually have a little more data than that, especially since glucometers store the data automatically these days. We should pay attention to how many patients had missing glucose data values when we get to the results though!
Doses were adjusted to target fasting blood glucose values of 70-108 mg/dL. Does this seem appropriate to you? Seems pretty reasonable, although I’d prefer my patients’ morning values not to be quite so close to the threshold for hypoglycemia (technically <70 mg/dL).
Interestingly, the once-weekly injections were done with a pen, whereas the once-daily injections were vial and syringe. I’d bet this was done in order to avoid confusion between dosing regimens, but vial and syringe can be more difficult for some patients with dexterity or visual acuity issues. We should consider how adherence to the once daily injections, especially with the glargine group, could have impacted blood glucose outcomes.
Primary Endpoint: The primary endpoint of this study was the change in HbA1c from baseline to week 26. This endpoint was presented in the context of an estimand. (This was a new term for me, and honestly, I had to read the sentence about the estimand about 10 times - and I still gave up and went to another site to investigate what on earth this new term means.)
Apparently, an estimand is a relatively new term for an old concept. That concept is that a clinical study should be designed with a clear objective and defined methods for handling variances from the intended protocol. Easier said than done, of course! This is why we often encounter superimposed analytical frameworks like intention-to-treat and per-protocol analyses - the investigators are trying to deal with deviant or unexpected situations.
Perhaps data is missing for some patients. Do we carry the last data point forward or impute some predetermined value to replace lost data? Perhaps some patients only finished part of the follow up period - how are those patients to be compared or incorporated with patients who completed the whole thing? These types of situations introduce the possibility of bias, and the struggle has always been how to minimize bias while still coming up with a reasonable estimate of treatment effect.
Enter the concept of an estimand.
According to this 2020 article, “An estimand is defined in the guidance as ‘..the target of estimation to address the scientific question of interest posed by the trial objective’ and has five attributes – population, treatment, variable (endpoint), intercurrent events (defined as events that can occur post-randomization and preclude or affect the interpretation of the variable) and the summary measure.” So it’s a framework for the study that accounts for all of those 5 pieces. Statistical magic, I suppose.
So in our insulin article, the authors pre-specified that the primary endpoint would be assessed using the trial-product estimand. This particular estimand was defined as a framework to evaluate the treatment effect (between groups difference in HbA1c over 26 weeks) under the following 2 assumptions:
All patients adhered to their assigned, randomized treatment for the entire duration of the trial, and
None of the patients received ancillary treatment.
The purpose of this particular estimand (aka set of conditions for analysis) was to try and isolate the effect of icodec without confounding from either non-adherence or other medications. As we discussed earlier, the investigators weren’t necessarily after real-world results in this study - they purely wanted to know whether icodec itself worked.
Secondary Endpoints: Secondary endpoints included changes in fasting blood glucose values, weight, mean weekly insulin dose in the last 2 weeks of treatment, and time spent with tight glucose control (defined as 70-140mg/dL) as read by a continuous blood glucose monitoring system in the last 2 weeks of treatment.
Safety endpoints included number of adverse events, hypoglycemia (both in general and overnight, specifically), and the number of severe hypoglycemia events (defined as severe cognitive impairment requiring assistance for recovery). Asymptomatic hypoglycemia was included in the adverse event count.
Short of stretching for clinical outcomes like cardiovascular or microvascular impacts, which would be beyond the scope of this phase 2 trial, they’ve pretty much covered the endpoints I would seek in an insulin trial. Sure the estimand makes the external validity a little more difficult, but for assessing icodec, they were fairly comprehensive.
We have efficacy endpoints for change in glucose levels and HbA1c, some exploration with regards to weight changes (since that’s always a question with insulin therapy and diabetes), and some investigation into effective doses between insulin types, which could be helpful in the future when converting between products. There are also safety outcomes with any adverse event and different levels of hypoglycemia.
Stats: This is really where this noticeably becomes a phase 2 trial. Although all analyses were prespecified, which helps to reduce bias - can’t just be like oooo see that result? Let’s take it and run to see what conclusions we can draw from this thread we just incidentally found! - none of the analyses were powered to detect significant differences between any of the endpoints.
So this becomes far more exploratory if we can’t conclude whether observed HbA1c differences between the icodec and glargine groups were due to true treatment effect vs just being chance.
Bummer.
Still useful though for hypothesis generation and future testing.
They also state that they didn’t make any adjustments for multiplicity, meaning they weren’t too worried about the possible bias introduced from analyzing the same data in multiple ways. Usually investigators adjust for making multiple inferences from the same set of data (e.g., Bonferroni or Hochberg adjustments) to avoid incorrectly drawing conclusions about a difference between treatments. So the fact that they didn’t make any of these adjustments should make us wary about the risk of inflating type 1 error rate - or incorrectly determining there’s a difference in treatment effect between icodec and glargine.
Again, it’s a phase 2 trial. We should take the results with a grain of salt, recognizing that it doesn’t have the same power or impact as a large scale phase 3 trial.
The primary endpoint was assessed with a linear mixed model with repeated measures with an unstructured covariance matrix.
Huh?
I basically just take that part of the paragraph to mean they assessed the primary endpoint of change in HbA1c in light of a number of patient-related factors, including use of a DPP4-inhibitor, geographic region, and insulin assignment.
The authors also define that they imputed missing patient data, which was in alignment with their trial-product primary estimand, in which all patients completed all follow up. So they chose to essentially ensure all patients had complete data, even if they had to impute data on patients’ behalves, for the purpose of maintaining their pre-defined estimand. In keeping with our theme so far, not necessarily real-world data, but perhaps useful for exploratory analyses.
Baseline Characteristics: There were 247 patients included in the full analysis data set. So who were these patients? We reviewed the inclusion criteria to see who was eligible, but who did they actually enroll? And were the icodec and glargine groups well-balanced?
The majority of patients were their late 50s, decently split male and female. Although the standard deviation was quite large in both icodec and glargine groups, the majority of patients had been diagnosed with diabetes for almost a decade. The icodec group had a slightly longer duration of disease than the glargine group, but it was still 8 vs 10 years. So our earlier concern about these being relatively newly diagnosed diabetics still trying to figure out how to check their blood glucoses probably wasn’t a huge issue.
Most were overweight with a mean BMI of ~30 kg/m2. The mean baseline HbA1c was about 8%, so decently well-controlled. So even though the inclusion criteria allowed for HbA1cs up to 9.5%, most patients were better controlled than that. About half of the patients were taking a DPP4-inhibitor. The rates of complications from diabetes were relatively low with only about 20% in each group having any sort of noted complication.
So let’s take a step back and picture this patient.
He or she is a middle-aged adult, dealing with diabetes for a significant portion of their adulthood. And honestly, doing a pretty decent job of it. Perhaps their HbA1c wasn’t quite at the goal of <7%, but 8% isn’t too shabby, especially considering only about half were even using a DPP4-inhibitor on top of metformin! (I wish more of my patients would achieve that kind of HbA1c control with metformin alone - or even with metformin + a DPP4-inhibitor - but HbA1cs of 8% don’t just come around every day. Clearly this crew was doing something right other than just taking their minimal medication regimen.)
Remember how we discussed the possible effects of hypoalbuminemia on icodec depots?
Unfortunately, we don’t have a lot of information on which to base any assumptions of hypoalbuminemia. Albumin levels were not assessed. When considering conditions that may accompany or predispose to hypoalbuminemia, such as malnourishment or undernourishment or chronic kidney disease, we don’t have much info on these possibilities either! Even though the majority of patients were overweight, it is possible for someone to still be malnourished even if they’re not undernourished. With regards to kidney disease, they only tell us that ~2-3% of patients in each group had diabetic nephropathy. But it would be a LARGE leap to assume hypoalbuminemia from that alone. So we likely won’t be able to make any connections between observed hypoglycemia and low albumin levels from this study.
Efficacy Outcomes: With regards to the primary endpoint, the icodec group’s mean HbA1c decreased 1.33 percentage points by week 26, whereas the glargine group decreased 1.15 percentage points (between groups difference 0.18 percentage points, 95% CI -0.38 to 0.02, p=0.08). Not statistically significant, but remember, this study wasn’t powered to detect significant differences between groups.
So maybe there’s a true difference there, and there just aren’t enough data points to see it.
Or maybe there’s just no difference!
Or maybe it’s still only a difference of 0.18 percentage points…so even if it was statistically significant, would it be clinically significant?
Looking at the HbA1c curves over the 26 weeks for both insulin types, the decrease in HbA1c for the icodec patients seems to begin diverging from glargine at ~6-8 weeks, which thinking back to its half life and kinetics, seems to make sense with reaching steady state. That’s probably when the icodec was reaching full effect, which is further illustrated by the stabilization of the weekly icodec dose around that same time. The glargine dose continued to require further adjustment to higher doses. So the initial idea that the icodec dose is equivalent to the daily glargine dose x 7, as visualized in the dosing adjustment protocol, may not be quite accurate.
About 72% of icodec patients met the HbA1c goal of <7% by week 26 compared to 68% of glargine patients. Again, pretty similar.
With regards to some of the secondary endpoints, the authors note increased percentage of time spent with tight glycemic control with icodec, although the difference between groups was only 5% extra time in the tight range. I think it’s more encouraging to note that this endpoint was comparable to glargine, as in people didn’t lose control at the end of the icodec dosing interval, etc.
Changes in fasting plasma glucose levels and body weight were similar between treatment groups. This is encouraging as weight gain is often of concern for patients starting insulin. Again, at least icodec seems to be comparable to glargine without any obvious adverse weight change signals.
Safety Outcomes: Now that we have some idea about efficacy of icodec, it’s time to check out the other side of the coin - safety. We had all these concerns when we were first going through the methodology about giving new-to-insulin patients a long-acting, more slowly titratable product. Were any of these concerns validated?
About half of the patients in each group experienced an adverse event of some sort. One patient in each insulin group accounted for significant reports of serious events and injection reactions. Must have been very thorough reporters - hard to take much away from the overall numbers when one patient reported 20 events of mild erythema and there were only 28 injection reaction events total in the entire icodec group.
Clinical Pearl: The Rule of 15 for treatment of hypoglycemia (Image)
So then as far as hypoglycemia (FINALLY)…
Hypoglycemia of any severity was numerically more common in the icodec group compared with glargine. Mild hypoglycemia (54-70 mg/dL) was observed in 53.6% of icodec patients versus 37.7% of glargine patients over the course of the study and follow up period. Clinically significant or severe hypoglycemia was observed in 16% of icodec patients versus 9.8% of glargine patients. Not exactly a good sign, although the portion of those icodec patients who had severe hypoglycemia was small compared to glargine (0.8% vs none) and that one severe case was noted to have been a blood glucose of 58 mg/dL that recovered with oral administration of carbohydrates only. (So was it reeeeeally severe?)
I guess the takeaway is that the majority of hypoglycemia with icodec was either mild or clinically significant rather than meeting criteria for full-on severe.
Still, when comparing to glargine… even without statistical significance, there definitely seemed to be more hypoglycemia overall with icodec compared with glargine.
One might think that this increased hypoglycemia could have been due to increased icodec adherence with the pen formulation compared with glargine’s vial and syringe set up. But the glargine group not only had a similar change in HbA1c but also had similar numerical HbA1c values as the icodec group, indicating that they were indeed taking their doses too even if the daily vial and syringe was cumbersome.
It’s also possible that the fixed dosing adjustment protocol used for icodec was too aggressive. Perhaps the number of units per adjustment needs revision. Also, further individualization based on an adequate number of blood glucose values could impact the success of this dosing protocol. Remember what glucose values were used to make these adjustments and the fact that it was possible for a patient to have an insulin dose adjustment made from 2 or even 1 blood glucose value per week.
On that note, we never circled back around to how many patients had sufficient data points vs how many had to have imputed data points! It’s not in the published manuscript, but it is in the supplemental appendix. Only 3 patients in the icodec group and 9 patients in the glargine group had imputed data, so these patients were actually pretty darn good at having sufficient self-monitored blood glucose values. Guess we can’t blame imputed values or lack of data points for shaky insulin dose adjustments, can we… So perhaps the protocol was just too aggressive for this new product.
Authors’ Conclusions: According to the authors, once-weekly icodec provided similar efficacy and safety to glargine; however, future, larger studies are needed to fully evaluate for significant differences. Also, further analysis of Icodec’s hypoglycemia, especially if used with certain anti-diabetic agents like sulfonylureas, is needed. Finally, additional exploration of the comparative insulin doses required to achieve glycemic control is needed. The weekly dosing curves diverged between icodec and glargine despite comparable HbA1c lowering, but icodec patients also experienced more hypoglycemia. So further determinations of how to initiate and adjust this medication are needed.
tl;dr Conclusions: Basically, I agree with all of the above. This is definitely an intriguing once-weekly product worth exploring further, but the optimal initiation and adjustment strategy needs further elucidation in order to get those hypoglycemia rates to be more comparable to our current glargine standard of care. I’d like to see additional details about patients’ baseline characteristics, including albumin levels and comorbidities, to evaluate whether these play any role in hypoglycemia and could help us determine who is or isn’t an ideal candidate for this insulin type. I’d like to see a more practical, real-world study with patients who miss doses, don’t have all their blood glucose values, are on multiple anti-diabetic agents, and take weekly doses of medications on variable days. I also don’t know that I necessarily see the big difference that the authors note with regards to time in a tight glycemic control range with icodec, especially if it’s at the price of increased hypoglycemia.
Keeping all of this in mind, once-weekly icodec is certainly an exciting product in the pipeline to monitor!
Hope you all enjoyed this journal club and return to diabetes! I look forward to our next article evaluation together!