Loading summary
Transcript32 lines
- [00:00]
A
This is Endocrine Feedback Loop. I am your host, Chase Hendrickson and welcome you to this Journal Club podcast series brought to you by the Enderkin Society. Thanks for joining us as we explore an important article recently published in one of the Society's clinical journals. Hello and welcome back to the Endocrine Feedback Loop podcast for our 54th episode. For this month's episode, we take a look at a recent JCEM article that provides data in support of metformin being renal protective. Though most attention in type 2 diabetes these days seems focused on the newer therapeutic options, metformin remains a cornerstone of our treatment regimens and such data is obviously highly relevant to many of the patients we see. We look at yet another observational study and so we'll carefully consider the limitations of such a study design and the impact on the conclusions we can draw. Before I introduce our team today, I remind you that I host the Endocrine Feedback Loop and work at the Vanderbilt University Medical center as a general endocrinologist and medical director. Back again to be this month's regular contributor is one of the podcast expert diabetologists, Steve Whitland. He comes to us from the University of Rochester where he works as a general endocrinologist and clinic director. He focuses his clinical work on diabetes in addition to conducting research in diabetes, diabetes technology and innovations in diabetes treatment. With us today is our guest expert is Silvio Inzuki from Yale. You all know him from his numerous publications in the field of diabetes guideline writing and talks. While his busy clinical work centers on diabetes, he too practices general endocrinology and also serves as the Fellowship Program Director. His research in diabetes spans the inpatient and outpatient settings and includes glycemic control and the complications of diabetes. So, as always on the podcast, the perfect pair of endocrinologists joins me today to unpack an article on the treatment of diabetes and the prevention of its complications. As is also always the case, everything we say will be our opinions only and not those of our respective institutions or the endocrine Society. For this edition of the podcast, we review Renal Protective Effect of metformin in type 2 diabetes patients, which is a forthcoming article in the Journal of Clinical Endocrinology and Metabolism. Shi Hao Wang from the Aishu University served as the first author for this paper and was joined by authors from multiple other institutions throughout Taiwan. I will now turn the discussion over to Steve, who will walk us through the main points that the authors make in their introduction, and we'll get Silvio to give us important background information on the treatment of diabetes.
- [02:41]
B
Steve, thank you. So we know that diabetic kidney disease is the leading cause of end stage renal disease and therefore it makes prevention and treatment of the utmost importance. So, before I get into the details of the introduction, I thought Silvio might comment on the impact of diabetes on the risk of developing kidney disease.
- [03:03]
C
Well, overall, obviously, as everyone knows, diabetic kidney disease is one of the most common complications of type diabetes. Both with type 1 as well as type 2 diabetes. The risk is increased in patients who have coexisting hypertension. The risk appears, at least in some studies, to be increased in patients with metabolic syndrome and obesity for a variety of reasons. But it is one of the more feared complications on behalf of our patients because ultimately, when end stage kidney disease is achieved, the only treatments obviously are renal replacement therapy with transplantation or with hemodialysis or peritoneal dialysis. So, as Chase mentioned, there's been a lot of important newer information over the past five to six years, I would say with emerging data indicating very clearly that SGLT2 inhibitors go a long way to preventing the progression of CKD. And just recently, based on the flow trial, at least one GLP1 agonist, the drug known as semaglutide, has also been shown to attenuate the risk of progression. What we know from older studies is that generally speaking, glycemic control matters, such that from the DCCT, although that's clearly a type 1 diabetes study, but also the UK PDS, there appears to be a modest attenuation of the progression of ckd, typically measured in terms of albuminuria, which is an early marker of ckd. When we compare tight versus not so tight glycemic control. So overall the glucose hypothesis appears to be true as related to the progression of ckd. But on top of that, there appears to be a specific advantage from SGLT2 inhibitors, clearly and at least one GLP1 receptor agonist. What we know about metformin and the progression of CKD is actually quite little. In the UK PDS, which is now a study going on some 30 plus years, there appeared to be no specific benefit. In a small cohort of patients that were randomized to metformin within the UKPDs, they did demonstrate a non significant reduction in microvascular outcomes that included retinopathy as well as renal disease, specifically albuminuria, but as mentioned, the results were not statistically significant. A follow up data from other trials such as Adopt, when metformin was compared to rosiglitazone and Glyburide this is a study that looked at the duration of the effect of diabetes therapeutics on hemoglobin A1C durability of glycemic management. They did look at some renal outcomes and there appeared to be no distinctive benefit from metformin. So at least in terms of randomized clinical trials, the data set is not highly robust for any specific benefit from metformin. But perhaps we'll talk about later. There have been some observational studies, mostly interesting, from Korea, suggesting that in the real world, patients that have been taking metformin may have a benefit as regards the progression of ckd. So that's the context that this paper was written in.
- [06:27]
B
So the authors then go on to suggest possible effects of the mechanism of action of metformin. And they note decreased intestinal glucose absorption. And more recently they don't note this, but there has been data on concentration of glucose within the intestinal wall. They also note that there's decreased fasting insulin, there's increased insulin sensitivity and the traditional mechanism of action of metformin, which is decreased hepatic glucose production, they suggest via AMP kinase. We also know that metformin has effects on mitochondria and lysosomes. It now has been shown to affect the microbiome. It's also been shown to be anti inflammatory and immunomodulatory. So there are a lot of different mechanisms that could be proposed for how metformin might be renal protective, and that's the object of the study. We know that metformin is inexpensive and it's considered the first line treatment for type 2 diabetes. I wonder if Silvio's thoughts on whether the advent of GLP1 receptor agonists, incretins and SGLT2 inhibitors impact metformin as the first line therapy for type 2 diabetes.
- [07:49]
C
Well, according to the ADA and the EASD, with their physician statements over the past few years, we seem to be getting away from metformin as the standard foundation therapy in all patients with type 2 diabetes. Now, it's a bit more nuanced in that in those patients with compelling indications for a drug specifically targeted at CBD protection or renal protection, kidney protection, that clinicians might favor those medications as the initial therapy. It's an interesting discussion. I think it's a moot point in most circumstances because let's admit it, most of our patients are going to do well on two medications. And the pairing of metformin with an SGLT2 or a GLP1 has become a very standard approach. So whether you start with one drug and add the other, or vice versa. I think at the end of the initial three to six month period, most of our patients are going to be on dual therapy based on the degree of hyperglycemia that they're presenting with.
- [08:51]
B
So the authors go on and note animal data that supports metformin being renal protective. And they note that in a mouse model, there's decreased fibrosis in the kidney and that in a rat model of type 2 diabetes, there's decreased tubular injury in humans. As Silvio noted, there's conflicting data. And Silvio mentioned the retrospective studies that suggested renal protection and the ADOPT trial, which was not designed to look at renal protection, but certainly did not show any renal protection. So this study's purpose was to explore whether metformin has a renal protective effect in patients with type 2 diabetes.
- [09:32]
A
Move on to the methods now and we'll think carefully about the study design. As we mentioned in the introduction, this is an observational study design. All of those come in with baked in limitations. So we want to make sure that we understand what those are so we can then look and see how the authors dealt with them, if indeed they did. So this is a retrospective cohort study. We've looked at a lot of these over the years for this podcast. So as a brief reminder, first of all, how a cohort study works, a cohort study works by splitting subjects into at least two groups. And that's what's done here. And you put people into groups based on an exposure and its simplest form it's do you have an exposure? Do you not have an exposure? And here that exposure will be a medication, metformin, as was mentioned already. You then follow these individuals over time. You have to first establish that they don't have whatever your outcome is of interest. And then you have to establish that you can follow them over time to see if one group versus the other develops the outcome of interest more or less frequently than the other groups. And what makes this retrospective is that all of this data is available by the time you start the study. The prospective study is a bit easier to explain. With a prospective cohort study, you are following these individuals in real time. You start your investigation when only the exposure has occurred and no outcomes have occurred. You follow those people in real time, waiting for the outcomes to develop. Retrospective studies are same in every regard, with the exception of you're starting afterwards. All the data is there. You have to be able to reconstruct that time sequence so that you can prove that the exposure predated the outcome. But all of this is available. You're simply looking back and ended making this determination. So that's what's done here. So the authors, when they start their methods, they immediately move into defining what their exposure and their outcome is. And I really, really like that. A lot of times in observational studies, you have to dig a little bit to really make sure you can understand which is the exposure, what's the outcome, how this is working. The authors label this very clearly. They unpack it later. But initially it's exposure is use or no use of metformin, and then outcome is change in kidney function over time. They then go into detail with all this. So their next step is to define what the patient population is. So their population is all patients with diabetes initially who were treated at the Changung Memorial Hospital. And that's made up of seven institutions throughout Taiwan. And they looked at a long time period starting in January 2006 all the way through December 2016. Now, initially they whittle this down, but initially this is well over 300,000 patients. They got a bunch of exclusion criteria that they apply. So first of all, if there are any individuals who were not treated at all, so those folks get excluded. Turns out to be a fair number of people. If you were treated before this study started, that was also a reason for exclusion. They were looking only at adults, so they excluded anyone under the age of 18. If you were initially not treated with metformin, but then later got started on Metformin, they excluded that. They didn't want anybody crossing over into multiple groups here. If your GFR initially was less than 30, that you were excluded. And then finally, if you were missing data, so some of the particular ones they were looking at, if you didn't have GFR data, urine dipstick data, A1C data, then then you were excluded. So what the authors do then afterwards is matching. So the authors use a couple of different approaches in general to dealing with the issue of confounders. They recognize this is going to be a major issue here. And so they the two big ways to do this are matching and then adjustment. So matching, which is what they did at the first pass is before you do your initial analysis. So what they did is after they excluded all these individuals, they were left with over 56,000 people who were on Metformin and then more than 16,000 people who were not on Metformin. So they then randomly selected patients who were in that metformin group, and then they matched them to people in the non metformin group. And there was several different clinical characteristics that they used to do that matching. Another thing that they used is what's called a propensity score. That's a fairly sophisticated statistical technique where you look at all of the clinical indicators and they list the ones that they use. It's quite a few. And they attempt to use all of those clinical indicators to predict based on those indicators how likely it is that you were going to get started on Metformin. That then generates a score with this technique. And then you can then find people in the other group who have a similar score. So it's an aggregate measure, but it's attempting to find people who are similar. So again, this is before you've done any analysis. You're just trying to make your two groups as similar as possible in the hope that that's going to get rid of a lot of confounders. They use something else. They use adjustment later. So we'll come back to that in a second. But after they've done that matching that one to one matching, they ended up with a little over 13,000 subjects in each group. So one thing that I want us to discuss before we move on further is if to see if anything was left out of that propensity scoring. As you think about confounders and what drives that, it's really just an exercise in creative thinking to try to consider, okay, was there something that should have been included that wasn't? Or maybe there was something that there's not a good way to measure, but it probably actually drove outcomes here. So, Steve, we'll start with you. I know you wanted to make a comment just on something that the author stated regarding not adjusting for GLP1s and SGLT2 inhibitors. But beyond that, any other suggestions that you would make about things that might have been left out of that propensity scoring?
- [14:54]
B
I'll come back to the question you commented on, which is the GLP1 receptors and the SGLT2 inhibitors later when we discuss the results. But there's also the issue of whenever you're doing a study like this, how did the physicians decide when to use Metformin and when not to use metformin? And you never know that, but that's always a question in that you may have some selected patients with the metformin that had another reason for being on metformin or not.
- [15:26]
A
We'll come back to that in just a second. Silvio, anything that you would add as far as things that should have been included in that propensity score?
- [15:32]
C
Obviously in a non randomized study. The prescription of these medications is not given out randomly. So we have to put our heads, or our minds, I should say, into the heads of the clinicians who are prescribing these medications. There's something called a channeling bias in epidemiological studies where patients are channeled, if you will, toward one type of treatment because of another indication or some other comorbidity. So the most obvious one clearly is the renal function at baseline. Right. Clinicians are not going to prescribe Metformin patients who they think have progressive ckd. So by definition, when you look at any non randomized real world data set, GFR in patients who are prescribed Metformin is always going to be higher. And those patients not prescribed Metformin and those other drugs are, at least in this timeframe, are going to be sulfonylureas, insulin and the DPP4 inhibitors in most circumstances. I think there's still a fair amount of use of alpha glucosidase inhibitors in East Asia. So I'm not sure about Taiwan specifically wasn't a lot of use of SGLT2s or GLP1s. So before they did the propensity matching, you can see that the GFR at baseline was much better in patients who were in the metformin group. But obviously if the propensity matching is working out of those 13,000 in both groups, the GFRs are now about the same. I think the median was about 78 or 79. And the percentage of patients in various strata of GFR in terms of the CKD cut points was very well matched. So we can't say that the reason the metformin patients did better is because they have better renal function at baseline because their propensity matching at least tried to address that. Now you never know what's in the head of the clinician. Right? You could have a patient who has a GFR of 30 ish, and many of us may not prescribe metformin in that individual. But if you knew that the person had a recent AKI event and was clearly on the rebound, you might prescribe Metformin. That kind of nuance is lost in these large data sets. But just in terms of plain cutting of the gfr, the groups look pretty well matched.
- [17:56]
B
I think. In summary, what Sylvia and I have been saying is propensity matching assumes you know all the variables that need to be matched and that's not necessarily the case.
- [18:08]
C
Well, you may know that, but you may not have access to in these large data sets, you have Presence of various comorbidities, other meds and lab values, but that's about it. You can't still can't get into the head of the clinician making that prescription to say, why did you prescribe Metformin? Why did you not prescribe Metformin in this other patient?
- [18:27]
A
I will make us 3 for 3 in raising that concern for residual confounding. So the authors, and we'll discuss this at the end, I think the authors do with the information that they have, a very thorough attempt to adjust for confounding. But you always have to worry about things that you just couldn't see, you didn't have access to, was not measured, and that continues to exert an effect. And I think this has been hit on by both Steve and Silvio. But to echo what Silvio just said is the real question is why did you not use Metformin? If you look at all these reasons that clinically you would say, well, this is a reason you don't use Metformin. But if these were all adjusted for, if these were all matched for some method, and yet still the patient didn't end up on Metformin, you have to wonder, what Steve said earlier is why was that person not put on that? What did the clinician know that we don't seem to know, Particularly since from the years of this study, from 06 to 2016, metformin was very clearly the first line treatment. So it would be unusual to not use that in that study. So we'll come back to that. But an important point to think about here. So now onto the outcomes as stated by the author. So they have a few that they're looking at all around those kidney function outcomes. So first of all, it's the doubling of creatinine sustained for at least three months was the outcome definition for a decline in renal function. Also, they were looking at a GFR less than 15, again sustained for at least three months. And then finally, evidence of end stage renal disease. And that was based on an ICD code. So, Silvio, another question for you about this. We're skipping ahead just a little bit to the results, but looking at that, on average, these patients were followed for about three years. It was a study of 11 years, but for individual patients, the average was about three. So a question Steve and I both had for you was, do you think that's long enough to see a meaningful difference in an outcome that we often think of as taking many, many years to develop?
- [20:16]
C
I'd say probably not, you know, two to three years is not a long enough time in most circumstances. So I suspect what's happening here because the, the criteria are pretty strict, right? It's a doubling of creatinine. So those are going to be probably AKI events, I would think, not necessarily progression of CKD over a period of two years. Achieving a GFR under 15, which was the second outcome, is probably going to be achieved in those patients who are well on their way to CKD stage 5. I mean, these are patients probably stage 3B or 4. So you're, you're cherry picking those patients that were already well on their way. And then obviously the ESKD has its own definition and is pretty robust. So the data I think are believable, but I'm not sure we can be generalized to the overall type 2 diabetes population. The patients that were starting metforminon more commonly are healthier younger patients more recently diagnosed who don't have progressive ckd. So it's just, it's, it's interesting that over a period of two, three years, they were able to capture a sufficient number of events to lead to a statistically significant finding in favor of metformin. But I'm not sure those patients who achieve these outcomes are the garden variety type 2 diabetes patients that we see.
- [21:41]
A
We'll wrap up the methods with a few comments on the statistical analysis in particular. So first of all, the authors, after they do this matching and then after they do the initial comparison, then they adjust via a multivariable Cox proportional hazard model. So I mentioned before, there's a couple different ways for accounting for confounders. One is to match, to try to make your groups as similar as possible. So the authors did that, you do that before your initial analysis, then the authors do what can also be done after the initial analysis. And that's to do an adjustment to look at additional clinical factors and to see if there is an imbalance between those potential confounders and the groups and if that is linked to differences in the outcomes and then adjusting for that. So that's what's done here. And another thing that the authors do that I also really, really like is that they show multiple models of how that adjustment was used. A real advantage of that is that gives you as the reader, a sense of how much of an impact the confounders had. So if they present their initial analysis and say there's no adjustment for that and there's a significant difference and it's a clinical meaningful one, but as they serial adjust for more and more and more things if that difference comes closer and closer to one where you'd say, well, there isn't a difference between these two groups, then you would get very concerned that confounders are having a major impact here and that any residual difference may just be from those residual confounders that either the authors didn't think about or weren't able to measure even if they did think about it. So it's really helpful. It's also encouraging then, if with serial adjustments, that number doesn't move much at all. That encourages you that confounders might not be driving it as much. None of these things are absolute, but it gives you a clue as to how much of an effect residual confounding might be having. Finally, the authors do what they describe as a sensitivity analysis. So they continue to be worried, as they should be, that there may be this unmeasured confounders that's having an important impact. And so they say, well, hypothetically, if there was a single such unmeasured confounder and it had the same impact of what we're going to label as the impact of metformin on the results, if we just remove that entirely, would it change our results? So we'll look at that and see what they do. That, of course, assumes that whatever the unmeasured confounder or confounders is or are is only as impactful as metformin and not more than that. But I do think it's a, it's a nice way of looking at that to try to understand what that would look like if there was such an unmeasured confounder. Okay, so a lot of stuff that we still got to think through as we work through the results and the discussion, I'm going to hand it back over to Steve and he's going to walk us through the data and the results as the authors present it.
- [24:17]
B
So, just general comment. The groups were well matched generally, and as you discussed in the methods, that's an important point. However, the metformin group had more patients with coronary artery disease, more patients on statins, more patients that had ACEs or ARBs, and more patients with calcium channel blockers and beta blockers. The non metformin group, on the other hand, had more stroke, gout autoimmune diseases, sulfonylureas, and importantly NSAID use outcomes. So in the doubling of creatinine, there were 1,038 versus 1,186 patients favoring the metformin with a hazard ratio of 0.7:1. The confidence interval was 0.65 to 0.77 and they did a Kaplan Meier plot where the p was less than 0.0001 which was highly significant. As far as GFR less than 15mls per minute. There were 376 patients in the metformin group versus 439 in the non metformin group. Favoring the metformin group. Again, the hazard ratio was 0.61 with a confidence interval of 0.53 to 0.71 and a highly significant P in the Kaplan Meier analysis. I wanted to point out that this is a small difference. These are 63 patients difference between the two groups, although you would expect in a very very large study like this or small differences in medications between the groups, et cetera, not to matter. There were 71 patients more on ACEs and ARBs in the metformin group and there were 18 fewer NSAID patients. And actually there were more than 200 patients more on calcium channel blockers in the metformin group versus the non metformin group. So the fact that there was only a 73 patient absolute difference between the two groups raises the question that although these groups were very well matched, might this not in part account for the difference that was seen between the two groups? Finally, in terms of the end stage renal disease, there were 248 patients in the metformin group and 303 in the non metformin group. The hazard ratio was 0.55 with a confidence interval of 0.07 0.47 to 0.66. And again, the Kaplan Meier plot was highly statistically different. However, if one actually looks at the Kaplan Meier data, by year two to three you've lost about 50% of the patients, which puts the data somewhat into question. And by year five nearly 75% of the patients are not included. That is the follow up is short. Also, by time the end stage kidney disease curves separate. There are more than 50% of the patients no longer in the study. So again that's a problem with a retrospective study, also with a high dropout rate, which is not their fault. It's a retrospective study. In the subgroup analysis the metformin was uniformly better except that in the end stage kidney disease group for patients with malignancy liver disease in H less than 45. In the EGFR group the malignancy non DPP4 use and liver disease were different. And in the doubling of creatinine the stroke group was different. They did a sensitivity analysis looking at hemoglobin A1C and it did not significantly affect the outcome. And we know that glycemic control is important in preventing diabetic nephropathy. This apparently is not a factor and their analysis for potential confounders did not significantly change the outcome. I found interesting. The number needed to treat analysis the number needed to treat to prevent doubling of creatinine was 89, to prevent an EGFR less than 15 was 208 and to prevent end stage kidney disease was 239. When I saw this, knowing how inexpensive metformin can be, I did a quick Google search and the reported annual cost of metformin is $240 per annum. So assuming an average follow up of four years, that's $229,440 per case of end stage kidney disease prevented or $55,000 per annum to prevent a case of end stage kidney disease, which I believe is cost effective.
- [29:18]
A
Now going to move on to the discussion and wrestle with those numbers and then try to figure out what to do with this. I'm going to start with a couple of quotes from the authors where they begin their discussion. So first of all, as the authors summarize their conclusions, they say patients with diabetes who took metformin were revealed to have better renal outcomes including lower incidence of doubling of serum creatinine, EGFR 15 or less and end stage kidney disease. And I'll comment on that. I thought that was very nuanced and I really like that because the authors aren't claiming here that it's necessarily the metformin that's doing it, it's just that those two things are associated. However, the authors then go on to say that the subgroup analyses showed a consistent reno protection effect in almost all groups. That starts nudging a bit more towards the sides of implying that the metformin is actually causing that, that that reno protection is suggestive of that. And, and I think maybe loses a little bit of the nuance that the authors appropriately had earlier on, but we'll think about that a little bit more. The authors do try to build their case that it makes sense that metformin could be exerting that effect for those patients and they do a couple of different things and we'll take them one at a time at a relatively high level. So first of all, the authors review those conclusions. We mentioned them early on. Both Silvio and Steve talked about the previous studies regarding that possible renal protective impact of metformin. Silvio, any final comments? Things that we need to keep in our minds about those previous studies and how that could potentially help us understand the results of this investigation?
- [30:49]
C
Well, whenever you have conflict between observational studies and randomized clinical trials, you have to ask yourself, why is that? And often, as in the history of pharmaceutical therapy and in medicine, I think we tend to believe the RCT is more than the observational trials because of the inherent biases and, you know, other confounders that might influence the results of the latter. On the other hand, you know, those that are supporting real world data as a complement to RCT data would point to the fact that rwd, or real world data is comprised of real world patients. And those that get into clinical trials may not reflect the complexity of the patients that we see in everyday clinical practice. So it's easy to point to a clear advantage from a pharmaceutical product when there's concordance between the RCT and the observational data. When they're in conflict, I think we don't know what to do. Although obviously as clinicians and scientists, we tend to believe the RCT data. The problem with the randomized clinical trials to date, as we've already alluded to, is it been quite small. The UK PDS was less than 4,000 patients, but importantly, in that metformin substudy of the UK PDS, there was a ridiculously small proportion of patients on metformin. I think it was under 400. So you really can't say much. Even though the UK PDS was conducted over a longer period of time than this specific observational study comprised, the numbers were just too small. The Adopt trial, I actually forget, Steve, the number of patients, but I think it was some like 500 or 600 in each arm. We don't have a lot of large RCT data with metformin, surprisingly. So we really can't say for sure in the UK pds. I was looking the paper up, this is published almost 30 years ago now. There was this trend toward improvement in microvascular outcomes. It just did not achieved statistical significance and was involving retinopathy data. So it was really hard to tease out, at least in the primary publication what was renal and what was ophthalmological. So there may be a trend there. But, you know, I think that the study is striking in terms of the hazard ratios are quite impressive. I think they've done a pretty good job looking at the propensity. Matching, for instance, was very well done. I think that the models that they used to actually adjust for any residual differences even after propensity matching, seem to show very, very consistent data. So I think that the study does suggest that there probably is a benefit from metformin. But whether it's simply an association that healthier patients are given metformin or there's a specific effect of metformin on the kidney, I think is still unknown. And unfortunately, I don't think there'll ever be a large randomized trial using or not using metformin to look at renal outcomes. These are very, very long and expensive trials to conduct.
- [33:56]
A
And Sylvia, the authors, I think, anticipated a question like you asked being asked. And so they wanted to propose some potential mechanisms that metformin could actually be directly exerting this effect. Steve mentioned several of them in the introduction, the ones that the authors mentioned, and Steve also added several as well. On top of that, what are your thoughts? Or the authors review these in much more detail in their discussion. Does it seem plausible that these effects could be real, that it really could be causing these clinical outcomes that we're seeing or maybe doesn't fully explain that?
- [34:28]
C
I think whenever we don't know about mechanisms, we throw around words like anti inflammatory and reduction in oxidative stress. I mean, how many times have we seen that in the preclinical data? I don't think we really know. We don't even know what the mechanism of actual metformin is in terms of its glucose lowering effect. Whether it's gut or the liver. I think most data would point to reduction in hepatic glucose production. There's some effect on GLP1 synthesis from the GI tract. There's also evidence of improvements in peripheral insulin sensitivity. It's clearly a mitochondrial drug, but there are probably multiple mechanisms of action of glucose lowering. So who knows if there's a purported beneficial kidney effect. Who knows what the mechanism is there?
- [35:13]
A
I think those comments highlight an interesting comment that the authors make where they say that a comprehensive understanding of these processes is essential for optimizing metformin therapy. I would at least agree with a statement that it's ideal to know, but I think, Silvio, your point's a really good one, is we've used metformin for decades and actually have not really fully understood what that mechanism is. And so maybe it's not essential for us to use metformin, but certainly ideal if we're looking at other benefits that we may not have fully recognized in the past.
- [35:44]
C
You know, one point I wanted to make is that I would have loved to have seen in the subgroup Figure I would have loved to have seen subgroups by baseline gfr. And for some reason it's such an obvious thing that readers would have wanted. And I did not see that in that table or figure, I should say. And the reason that's important is because what I said before about these pretty hardcore renal outcomes, doubling of creatinine, EGFR under 15, the development of ESKD, and I think it was probably those patients who had the progressive CKD to the point of stage 3B or even 4. And you wonder why those patients were prescribed Metformin to begin with. So there was something about those patients with very advanced CKD to the point of needing to stop the metformin that had their metformin continued. And if we saw that most of these events were in the patients with GFRs under 30, but not much of an effect in patients with GFRs of 90, for instance, I think that would have been very telling. And I'm just curious as to why the authors did not include that in that figure. Such an obvious one that you see in all these renal studies is the effect of the intervention based on GFR or uacr, which is another important variable as far as the kidneys are concerned.
- [37:03]
A
We'll get to Steve's comment here in just a minute, but we'll list the author's own limitations that they report and it's several of the ones that we've been talking about throughout this discussion. They are very clear this is an observational study and carries with it all the limitations that are intrinsic to those types of investigations. They are very upfront about the unmeasured confounding. They talked about all the things that they did to try to account for that, but recognize that there is residual confounding in all of these studies. They also point out their reliance on ICD codes in this investigation and then finally that they couldn't really determine which patients were most likely to benefit from metformin. Where the authors end things is with the statement that in this extensive multi center retrospective cohort study, metformin was identified to exhibit renal protective effects. And then so we'll think about that a little bit more. Silvio's given us his thoughts on potential good things about this study, the impact that it may have, but also some of these residual concerns that we have. Steve, other comments that you would make just about the quality of this report overall.
- [38:07]
B
I think for what it is, it was done very well. Obviously, as you mentioned, it's a retrospective study and yes, it's based on ICD9 claims, but the endpoint is also based on ICD9 claims, which can get tricky. But also, as I commented in the results section, the end stage kidney disease doesn't separate out until you've lost half the patients and it's within three years. So I don't know what that means. Mechanistically, Silvio's outlined all the pitfalls that can occur with an analysis like this, and I support that. We'd love to see EGFR listed by subgroups, but there's no further subgroup analysis. And in this age of personalized precision medicine, it'd be nice to know if you're going to use metformin for kidney disease. Might there be a subgroup that benefits versus a group where you oughtn't be using it? Certainly. I would guess that if your EGFR is greater than 90mls per minute, you're going to need a long study to show differences in outcome. For example, the design doesn't allow for mechanistic speculation, even though they do it. Finally, the question that always comes up in single center or single system studies is this generalizable to all nationalities and all ethnicities? But we really don't know that from this study. Overall, for what they set out to do, I think they did it well.
- [39:31]
A
So, Steve, let's stay with you then. So, as you think about this, I think we're all on the same page here. I think it's some very interesting, compelling data. Do all have big concerns about the impact of that residual confounding, recognizing that that probably is affecting that result. But. But what do you think? So we're listening to this, we're reading this, we're trying to decide, okay, I see a lot of patients and I'm trying to decide which medication to start. Should metformin continue to be one of those first ones? Because I now think that it may provide some renal protection. So. So what are your thoughts, Steve? Should this change our clinical practice now or not?
- [40:03]
B
Yet I find it interesting to look at this historically because until 10 or 15 years ago, people were frightened to death to give metformin to anyone with a mildly reduced egfr. Then we found out that you can adjust dosing based on egfr. And now we've. The pendulum's come all the way across to where we're talking about. Is it renal protective or not? And it's possible both groups are right, that if you reach a certain egfr, maybe it's not a good idea. If you don't, maybe it is A good idea. From this we can't conclude anything. I certainly think that there's nothing in this study that would further disqualify Metformin from its lofty cost effective benefit as a first line drug for type 2 diabetes. I don't know that there's data here that makes it look like an SGLT2 inhibitor, which almost by accident was found to be a super kidney drug.
- [41:02]
C
Although the point estimates are not dissimilar. Right. We're talking about if there is a cause and effect scenario here, it looks like it's about at least a 35, if not a 45% risk reduction. And for ESKD it was a 50%. Those are in the same general scale that we see with the SGLT2 inhibitors. Maybe not the GLP1 agonist, but obviously we're talking about randomized clinical trials versus observational data sets. But I'm just making a point that the point estimate is not, are not dissimilar. So if the effect is real, it could be very important in the grand scheme of things. I think that the guideline writers recently have gotten away from metformin as foundation therapy, as we used that term before. I've always been a little uncomfortable with that. Most of the patients in the cardiovascular outcome trials that led to the changes in SGLT2 inhibitors and GLP1 agonists as primary therapy in those patients who are at very high risk for cardiovascular complications, 70 to 75% of those patients were on Metformin. Now, there was no heterogeneity in terms of the treatment effect in those patients who were taking, were not taking metformin at baseline. But as mentioned before, you often will need two drugs. And I think metformin is a great pair with any of these agents. I think what happens in clinical practice, at least in endocrinology, is a scenario where a patient is on both an SGLT2 inhibitor and a GLP1 agonist at very high cost, I might add. You know, with approaching $2,000 a month, if you're paying out of pocket and is achieving lost a lot of weight, is achieving a 1Cs that we've never seen before, sometimes under 6% and they're still taking their Metformin, They've been on metformin for a long time. And then the question is, hey doc, do I still need my metformin? These other drugs are working so well and, you know, I'm not sure what to do in that circumstance. I kind of leave it up to the patient to some degree as to Whether they're tolerating and would mind staying on medications. Clearly patients like to minimize the number of drugs they're taking, so I empathize with that as well. But if they're doing well on a medication and renal function is normal. You know, I personally believe that metformin has a cardiovascular benefit. It was the data from that small study in the uk PDS showing less myocardial infarction release versus conventional diet therapy. So I think there's something there. Unfortunately, just like the kidney outcomes, we'll just never have a large robust study using metformin for cardiovascular complications. So we'll never know. But my heart of hearts, I do think that there is probably an additional cardiovascular benefit from metformin and this study would suggest that maybe a renal benefit from metformin. So my tendency is to keep at least a little metformin on board, maybe just to make me feel better. Even in those patients that are achieving terrific A1Cs. But they're more potent and more modern therapies with SGLT2s and GLP1.
- [43:56]
B
Silvio, from a public health perspective, a year of metformin is $240. GLP1s and SGLT2s are more than 10 times that. Yeah, if, and it's a big if just based on this paper. But if metformin in fact is reno protected, isn't that a strong argument from a cost benefit point of view to using metformin upfront still?
- [44:22]
C
Yeah, along with a number of other. I mean it's hard to make recommendations based on non randomized clinical trained observational studies. But yeah, in fact it's probably cheaper. In some of the large commercial pharmacies you can get metformin for $4 a month. So that's, I mean you can do the math. It's, that's under $50 a year for sure. They're, they're the cost. And, and no one's ever shown that an SGLT2 inhibitor or GLP1 agonist is better than metformin. I'm talking about a monotherapy study. We don't know how those drugs would, would look. Remember the mace effect with both the GLP1s and the SGLT2s are not home runs. Right. These are 13 to 14% relative risk reductions. They're impressive because historically we've never been able to show any benefit from any glucose lowering drugs. So they take 13 and 14%. And you know, I'm obviously a big proponent of SGLT2 inhibitors as well as GLP1 agonists. But when you actually look at the overall effect of these drugs on Mace major adverse cardiovascular events. If they were compared head to head with metformin in a monotherapy trial, I'm not certain that they would win. The studies that were done were based on adding these drugs to background therapy, much of which included metformin in both groups, the placebo group as well as the active therapy group. But we don't have the data to know whether these drugs are actually better than metformin as monotherapy. So more reason to leave the metformin alone and leave it on board.
- [45:53]
A
And with that, I would like to thank Steve Whitland and Silvio Inzuki for joining me for this month's edition of Endocrine Feedback Loop. I hope that you all learned as much as I did and that you will join us again next month. And now you're in the loop. This has been Endocrine Feedback Loop. Endocrine Feedback Loop is brought to you by the Endocrine Society with Production Oversight by Brandy Brown and Andrew Harmon. If you want to like and subscribe, you can find us on Apple, Spotify, or wherever you get your podcasts. We'd love to hear your feedback on this episode or the podcast itself. Please email us@podcastren.org Endocrine Feedback Loop is a free service of the Endocrine Society.
- [46:42]
B
To learn more or to become a.
- [46:43]
A
Member, visit the society's website at www.endocrine.org.