Are new diabetes drug making a difference at the population level? Study reports rise in GLP-1 agonist, DPP-4 inhibitor use from 2006-2013 alongside increasing A1c and flat hypoglycemia rates – October 19, 2016

Executive Highlights

  • Between 2006-2013, DPP-4 inhibitor use increased from 0.5% to 15% of patients with type 2 diabetes, GLP-1 agonist use increased from 3% to 5% (granted a big percentage increase but a very low base), TZD use declined by 29% to 6%, sulfonylurea use declined from 39% to 31% (hurrah!), and severe hypoglycemia remained flat at 1.3 events per 100 person-years and the proportion of patients reaching target A1c <7% fell from 56% to 54%, all according to a recent paper by Dr. Kasia Lipska and colleagues. We’re not that surprised that A1cs didn’t decline but there are many pieces here to unpack – the fact that this small dataset would be viewed as definitive in any way, shape, or form, and is prompting headlines “about new drug classes” is worrisome.
  • In a controversial conclusion, study authors speculate that “newer” diabetes drugs such as DPP-4 inhibitors and GLP-1 agonists may not be cost-effective. It’s an incredibly unwarranted conclusion in our view, particularly given A1c as the “end all be all” and particularly given the “new drugs” were just 20% of the total and they have nothing on SGLT-2s.
  • We see room for further sub-analyses of this data, and we note that the diabetes drug landscape has already changed quite a bit since 2013, which may limit the applicability and validity of these findings.

In a recent Diabetes Care publication, Yale University’s Dr. Kasia Lipska and colleagues report on trends in diabetes drug use, A1c, and severe hypoglycemia between 2006 and 2013. The analysis drew from claims data from privately insured and Medicare Advantage patients with type 2 diabetes (n=1,657,610). Within this population, use of incretin-based agents increased substantially during the study span, with DPP-4 inhibitors exploding from use in 0.5% to 15% of adult type 2 diabetes patients (p<0.001) and GLP-1 agonists rising from use in 3% to 5% of patients (p<0.001). As noted, the growth in GLP-1 agonist use was from a low base. GLP-1 agonists have substantial room for growth and expect their use has increased further just in the years since 2013. We also don’t want to be discouraging, but we don’t think most doctors have time in the current system to help patients do well on insulin and we think that also may have contributed, ironically, to the absence of progress. Looking to older medications, metformin use rose from 48% to 54% (p<0.001). There was a sharp decline in use of TZDs, falling from 29% to 6% (p<0.001) – in hindsight, this was Avandia “drama and trauma” that wound up being completely unfounded. As well, there was as a decline in use of sulfonylureas from 39% to 31% (p<0.001). On the insulin side, overall use increased from 17% in 2006 to 23% in 2013 (p<0.001), driven by greater adoption of basal and rapid-acting insulin analogs, up from 11% to 19% and from 7% to 12%, respectively (p<0.001 for both). Use of human insulin dropped from 12% to 6% (p<0.001).

Despite more adults with type 2 diabetes taking glycemic-dependent drugs and fewer taking sulfonylureas, severe hypoglycemia rate was flat from 2006 to 2013 at 1.3 events per 100 person-years (p=0.72 for trend over time). We’d have to know more overall about how this rate compares to other institutions, and how much there was for this to fall. The proportion of severe hypoglycemia episodes leading to hospitalization declined marginally from 19% to 18% - while we would expect this to be lower with significant GLP-1 and SGLT-2 use, we see it as questionable for the authors to make any assessment on this given 5% (!) of patients took GLP-1 in this study and zero took SGLT-2s. Meanwhile, the authors assert that glycemic control appears to have worsened somewhat in US adults with type 2 diabetes over this time period – the proportion of patients with A1c >9% rose from 10% to 12% (p<0.001), the proportion of patients with A1c between 8%-9% rose from 10% to 11% (p<0.001), and the proportion of patients meeting a target A1c <7% fell from 56% to 54% (p<0.001). It’s not particularly surprising to see these changes – while the persistence of poor glycemic control and debilitating hypoglycemia is depressing, we hardly think this was an environment in which patients could get any medicine they wanted. Yale seems to have decided to put lots of patients on DPP-4 inhibitors and these drugs aren’t known for their potency for patients late in disease progression – we would probably question who they put on DPP-4 inhibitors and the idea that this study says anything about the value of GLP-1 use when they had only 5% of the total on this class is highly questionable. The idea that this study would be pursued is valuable, of course, but there are many limitations.  Despite the many advances in the diabetes field over the last decade, too many patients with diabetes are still struggling – that said, with only 20% of this population on “new drugs” and 75% of that population on the least potent drugs, we think it a bit untenable to create headlines implying “new drugs have little value”.

  • Dr. Lipska and her colleagues conclude that newer drug classes may not be cost-effective – this is highly misleading in our view. Their paper cites a doubling of diabetes-related healthcare spending between 1987 and 2011, over half of which is attributable to prescription drugs. Dr. Lipska points out that, despite these increases in drug spending and the increases in “newer” drug utilization she found in this study, there was no accompanying improvement in glucose-lowering or hypoglycemia avoidance. In our view, it would take a lot more than a tiny bit of GLP-1 use and a LOT less TZD use and much more DPP-4 and insulin use to sway hypoglycemia. It’s also unclear what proportion of patients on GLP-1 agonists or DPP-4 inhibitors were also taking sulfonylureas or insulin at the same time, which could exacerbate any hypoglycemia risk. We’d like to see the subgroup analysis for: those going onto DPP-4 inhibitors, those going onto insulin, those going on GLP-1, at minimum, and better understand changes in A1c or hypoglycemia rate related to specific drug class initiation over time. This should be easy analysis to make but it’s not in the paper.
  • When evaluating these findings, it’s important to keep in mind the substantial advances in diabetes medication that have occurred since 2013 and that are on the horizon. The time period of the study concluded before SGLT-2 inhibitors were available, before the launch of better next-generation basal insulins with longer, smoother profiles of action (Sanofi’s Toujeo [insulin glargine] in March 2015 and Novo Nordisk’s Tresiba [insulin degludec] in January 2016), and before outcomes trials such as EMPA-REG OUTCOME and LEADER demonstrated pronounced cardiovascular and renal benefits for an SGLT-2 inhibitor (Lilly/BI’s Jardiance [empagliflozin]) and a GLP-1 agonist (Novo Nordisk’s Victoza [liraglutide]), respectively. Prescriptions of GLP-1 agonists remained extremely low in 2013, even among this suitably-insured study population. An increase in GLP-1 agonist use from 3% to 5% would be highly unlikely to substantially affect the glycemic control and hypoglycemia incidence in the entire larger population, in our view and we’re extremely surprised that the authors would put forward a “cost-effectiveness” study based on the size of this population and the 0% of patients taking an SGLT-2. A substantial portion of patients – though still minority at 15% – were using DPP-4 inhibitors by the end of the study, but it’s widely acknowledged that DPP-4 inhibitors’ efficacy does not match that of GLP-1 agonists, which clearly contributed to the findings that were billed as disappointing by the authors. We reckon, by contrast, that it would be very interesting to study a population in which a majority of patients took GLP-1 or SGLT-2 or DPP-4 inhibitor combinations, but that is not close to being the case in this study. We expect the use of GLP-1 agonists will increase in coming years given the positive cardiovascular outcomes data from LEADER and from SUSTAIN 6 (for Novo Nordisk’s once-weekly semaglutide), the increasing availability of Lilly’s once-weekly, extremely patient-friendly option Trulicity (dulaglutide; launched in 2014), and the launch of new methods of administration of the drugs (such as Intarcia’s implantable exenatide mini-pump ITCA 650 and Novo Nordisk’s oral semaglutide formulation). GLP-1 agonists are also increasingly favored as a basal insulin intensification option, given their efficacy in addressing postprandial excursions with fewer injections and less hypoglycemia risk than rapid-acting insulins – basal insulin/GLP-1 agonist fixed-ratio combinations will likely only accelerate this trend. We also expect the increasing popularity of SGLT-2 inhibitors (launched in 2013) could help move the needle on both glycemic control and hypoglycemia reduction, given their non-insulin-dependent mechanism of action. Thus, the published findings are to some extent outdated and we believe the implications of these findings on clinical practice and guidelines should be limited – rather, we view these results as another piercing reminder that we are not doing well enough in diabetes care, even with access to more advanced therapies.
    • That said, Dr. Lipska emphasized to us that scope of the paper was limited by the available data sources and by the specific research question under investigation. For instance, severe hypoglycemia was defined by emergency room visit or hospital admission for hypoglycemia – a rather conservative definition compared to the definition of severe hypoglycemia requiring assistance used in clinical trials such as the SWITCH trials for Novo Nordisk’s basal insulin Tresiba (insulin degludec). Dr. Lipska pointed out that hospital visits are the only measurement of severe hypoglycemia that is possible to capture from the insurance claims data used in this study and acknowledged that she would be excited to expand the definition of severe hypoglycemia with more comprehensive data sources. Similarly, BMI and weight information was not available within the data set used, though we felt that the weight effects of various diabetes drugs could help drive adherence and would be valuable to consider. Overall, we would love to see sub-group analyses evaluating the impact of specific classes, such as GLP-1 agonists, on glycemic control and hypoglycemia over time. While Dr. Lipska acknowledged that such analyses are possible, she emphasized that the question at hand in the published paper was to investigate drug utilization and population-level changes in glycemic control – while this is a certainly an interesting question, we do feel that the topline conclusion tying these two measures together could be misleadingly interpreted, as we outlined above. Overall, we feel that a conclusion on the cost-effectiveness of new drug classes could not be adequately drawn from the data used in the study. We would love to see further cost-effectiveness studies that consider a variety of broader, potentially confounding factors including social and cultural determinants, time with healthcare provider, access to healthcare provider and ease of communication, stress at home, and diabetes complications prevention and adherence education.
  • Dr. Lipska has used cost-effectiveness arguments in the past to advocate for older insulins, such as in a New York Times op-ed published this past February. We felt then that she didn’t dedicate enough attention to the hypoglycemia risks associated with NPH or regular human insulin, and that she didn’t acknowledge the importance of patient-specific or individualized therapy. While she overlooked some evidence-based advantages to newer insulin products, we find the “in real life”. Here again, we worry about the lack of acknowledgement of data that shows superior A1c-lowering efficacy of GLP-1 agonists. While it’s most definitely valuable to see how drug use, A1c, and hypoglycemia have trended together over time in this specific Yale population – and while it’s somewhat disappointing to note lower-than-expected improvements in A1c and hypoglycemia avoidance but we hardly think that GLP-1 use by 5% of the population should be expected to make a major difference in this population– we believe this study should’ve been done when far more of the population were on GLP-1 or SGLT-2 inhibitors. Had the authors done more nuanced sub-group analyses (how did the 5% of patients on GLP-1 do?), that would’ve been far more valuable and longer-term investigations of new drug classes and cost-savings. See our list of questions below for more on this.
  • Trends for the various drug classes analyzed in this study also raise interesting questions:
    • The very notable substantial increase in insulin use from 17% in 2006 to 23% in 2013 reflects greater acceptance and adoption of basal and rapid-acting insulin analogs, presumably due to their improved reliability and safety profile over human insulin. Mount Sinai’s Dr. Zachary Bloomgarden highlighted the increase in insulin use as the most notable data point from the paper in his view. We would be interested in learning more about trends in the time from diagnosis to insulin initiation – is the growing popularity and greater availability of insulin analogs perhaps reducing clinical inertia and driving more timely initiation of insulin? In addition, we’re curious how the growing popularity of GLP-1 agonists since 2013 might affect insulin utilization rates. Dr.Bloomgarden underscored that insulin utilization rates will be even more important to watch as biosimilar insulins enter the market, sulfonylurea utilization continues to decline, and utilization of DPP-4 inhibitors and metformin continue to climb.
    • Declining popularity of TZDs (down 80% between 2006 and 2013) is unsurprising given the scare over increased risk for MI with rosiglitazone and the now-largely-debunked controversy over pioglitazone and bladder cancer risk. Given the durability of TZDs, we imagine the impact of so many patients going off this drug that worked well for A1c and hypoglcyemia (but not weight or fractures) didn’t bode well for the A1c and hypoglycemia results. As a side note, we’ve observed some renewed interest in TZDs from diabetes experts and thought leaders over the past few months. The IRIS trial led by Dr. Silvio Inzucchi (Yale University, New Haven, CT) found a reduced risk of stroke and MI with pioglitazone (Actos) treatment. During a panel dedicated to diabetes at the recent CMHC meeting in Boston, Drs. Robert Ratner (ADA, Arlington, VA), George Bakris (University of Chicago, IL), Robert Eckel (University of Colorado, Aurora, CO), and Jay Skyler (University of Miami, FL) endorsed low-dose pioglitazone and drew a distinction between pioglitazone and rosiglitazone. TZD prescriptions have trended down since 2013 and we don’t expect major increases over the next several years. While pioglitazone’s generic status and insulin-sensitizing properties are certainly attractive features, Dr. Nissen’s railing on Avandia put a downturn in the class that could likely never be reversed. While the drug could serve as an alternative to sulfonylureas for some patients who struggle with affordability and access, the “Nissen negativity” was hard for many to fight. Overall, we expect many patients and providers (primary care physicians especially) to be reluctant to embrace the class again.
    • Metformin use increased from 48% to 54% between 2006 and 2013, reflecting its steadfast place as first-line therapy. Again, low cost and generic status are hard to argue against when it comes to metformin, though many in the field have been pushing for earlier treatment intensification and acceleration to combination therapies. With use of GLP-1 agonists still low at only 5% in 2013, signs of clinical inertia are abundant. Furthermore, the study found that the use of oral monotherapy increased slightly from 2006 to 2013 (24% to 26%), while the use of multiple oral agents declined by a whopping 20% (from 33% to 27%). Maybe if providers weren’t delaying introduction of a second agent after metformin in type 2 diabetes patients, we’d see greater A1c-lowering over time – this is one area where a dedicated sub-analysis of the recent Diabetes Care paper might unveil valuable knowledge.
    • DPP-4 inhibitors showed the most dramatic growth between 2006 and 2013. Thought leaders are a bit divided on the position of DPP-4 inhibitors in diabetes care, with some like Dr. Jay Skyler highlighting their inferior A1c-lowering ability and lack of CV benefits compared to GLP-1 agonists and others like Dr. Robert Ratner pointing to their favorable safety profile – most seem to favor their use in just-diagnosed or fairly early-stage patients though our guess is that overall, it’s a real mix of patients who are taking them. Indeed, there is a good deal of familiarity with DPP-4 inhibitors in diabetes care, and our industry round-up for 1Q16 and 2Q16 also shows growth for the drug class in recent quarters. DPP-4 inhibitors are especially favored among primary care physicians, due to their benign clinical profile and easy oral administration (as opposed to agents that cause multiple side effects like weight gain, hypoglcyemia, nausea, etc). While it’s a newer diabetes agent, it’s arguably the least effective of the new diabetes medications for mid-stage or later-stage patients from an A1c efficacy standpoint, to say nothing of them to-date neutral impact on cardiovascular outcomes compared to the cardioprotection offered by some GLP-1 agonists and SGLT-2 inhibitors for some patients. We’d love to learn more about how relative use of DPP-4 inhibitors, GLP-1 agonists, and SGLT-2 inhibitors relates to trends in average A1c and severe hypoglycemia between 2013-2016.
    • Use of GLP-1 agonists increased substantially, but from a very low base of 3.3%. We were surprised to see such low use of these agents (and we wonder if patients felt entirely comfortable titrating the drugs), but we expect much of the growth in GLP-1 agonists has occurred more recently than 2013 (the class experienced record-high sales growth in the first half of 2016). The cardioprotection showed in LEADER for Novo Nordisk’s Victoza (liraglutide) and SUSTAIN 6 for Novo Nordisk’s semaglutide may eventually affect volume of GLP-1 agonist prescriptions going forward (with the growing influence of CVOTs, CV effects cannot and should not be ignored), although the label still hasn’t changed, and we’re curious as to how use of these agents changed between 2013-2016 as well.

Close Concerns Questions

Q: How would these same trends extend beyond 2013, until today? What would be the impact of SGLT-2 inhibitors coming to market and of the recent CVOTs (i.e. LEADER and EMPA-REG OUTCOME)?

Q: How did A1c, hypoglycemia, and hypoglycemia-related hospitalizations change specifically for patients started on GLP-1 agonist therapy, either alone or in combination? How does this compare to patients started on a DPP-4 inhibitor, a TZD, or a sulfonylurea?

Q: Does baseline A1c influence choice of agents in real-world practice within this data set?

Q: A private insurance or Medicare Advantage plan doesn’t automatically free individuals from struggles over payer coverage and reimbursement – how did formulary positioning and prior authorizations affect rates of use for different drug classes?

Q: What would the results be if we accounted for a broader population, beyond those on private and Medicare Advantage health plans?

Q: Does the data reveal an impact of private insurance vs. Medicare Advantage on drug use, A1c changes, and hypoglycemia rates?

Q: What is the effect of diabetes duration on choice of agent? Of those on DPP-4 inhibitors, what was the average duration of diabetes? We are curious mainly about DPP-4 inhibitors used in early-stage diabetes – if those in more advanced stages of diabetes are on a DPP-4 inhibitor, especially as monotherapy, that’s more likely to be a therapeutic decision mistake.

Q: How can we incorporate patient reported outcomes and quality of life measures into analyses of cost-effectiveness? Were outcomes beyond A1c considered like time in range or patient reported outcomes?

Q: Is the “severe hypo” rate that didn’t move average for other populations?

-- by Payal Marathe, Helen Gao, and Kelly Close