Author + information
- aDuke University School of Medicine, Durham, North Carolina
- bDuke University Clinical Research Institute, Durham, North Carolina
- ↵∗Address for correspondence:
Dr. Eric D. Peterson, Duke University School of Medicine, 2400 Pratt Street, Durham, North Carolina 27705.
- behavioral science
- comparative effectiveness
- medication adherence
- patient engagement
- secondary prevention
“We have, in short, somehow become convinced that we need to tackle the whole problem, all at once. But the truth is that we don’t. We only need to find the stickiness Tipping Points.”
—Malcolm Gladwell (1)
In his book The Tipping Point, Malcolm Gladwell (1) investigates factors necessary to create sustained change in people's beliefs and behaviors. One of the 3 key elements he emphasizes is “stickiness”—or the degree to which a message or idea “sticks” with individuals over time. Stickiness is what makes children return to Sesame Street over and over again, or why their parents instinctively buy the same brand of cereal every time. Marketers understand that stickiness is critical to their success—and spend great efforts to ensure that their messaging is more impactful, memorable, and thus more sticky. Health care professionals, similarly, need to facilitate sustained behavioral change in their patients. However, unlike marketers, clinicians do not traditionally think about whether their messages are sufficiently sticky or compelling to their patients. As a result, medication adherence represents one of medicine’s greatest unsolved challenges.
In this issue of the Journal, Korhonen et al. (2) used national Medicare claims data to assess adherence to 3 critical therapies for secondary prevention of myocardial infarction (MI): statins, beta-blockers, and angiotensin-converting enzyme inhibitor/angiotensin II receptor blocker (ACE inhibitor/ARB) therapy. Using a standard (albeit arbitrary) definition of adherence based on the proportion of days covered (PDC), the authors classified patients with PDC ≥80% as “adherent” and those with <80% PDC as “nonadherent.” Over the first 6 months following an MI, adherence to any of these individual medications ranged from 70% to 80%, but fewer than one-half (48.5%) of all patients were adherent to all 3. Though this paints a grim picture of adherence, these estimates likely represent an optimistic view of medication adherence for multiple reasons. First, the study was limited to older adults (who typically have better adherence than younger patients). Second, only patients with pharmacy benefits were eligible, and those with insurance have improved adherence relative to with those without). Third, the study included only those who filled ≥1 prescriptions—those who never filled any were not included. Fourth, using PDC to track adherence presupposes that medications that were dispensed were actually taken, yet automated pharmacy renewal programs often result in patient medicine cabinets overflowing with unutilized medication stockpiles. Finally, the authors examined adherence over a relatively short 6-month window, and far greater attrition would be expected over longer time horizons. Combined, it is safe to say that adherence to medications post-MI is in dire need of an infusion of stickiness.
Merriam-Webster notes that “sticky” can also mean “difficult or challenging,” as in a “sticky problem.” Beyond describing adherence, Korhonen et al. (2) also attempt to use their data to determine whether it is time to retire certain guideline-recommended medications. The investigators first demonstrated that those who were adherent to all 3 secondary medications had the best outcomes, whereas those who were completely nonadherent did the poorest. The results were more interesting among those who were partially adherent. Patients adherent to ACE inhibitors/ARBs and statins alone (i.e., nonadherent only to beta-blockers) had similar outcomes as those who were adherent to all 3 medications. This was true for those who were nonadherent to either their ACE inhibitor/ARB or their statin: these groups had worse outcomes than those who were fully adherent. The authors interpreted this as demonstrating that beta-blockers may not provide incremental effectiveness when taken long term.
Interpretation of observational treatment comparisons like those conducted by Korhonen et al. (2) can, however, be sticky. The reason(s) that patients adhere to or discontinue a medication are complex and influenced by factors that confound observed associations with clinical outcomes. As an example, even those who adhere to the placebo arm of randomized trials have significantly better outcomes than those not adherent to the sugar pill (3). In the current study, those that discontinue ACE inhibitors may do so in response to the development of renal insufficiency or hypotension (both risk factors for death). If sicker patients quit a medication more commonly than their healthy peers, then the benefits of adherence to that medication will be exaggerated. By contrast, the directionality of bias may be in the opposite direction for beta-blockers. Those with disease progression, recurrent events, or refractory angina may be more adherent to their beta-blocker, whereas younger, healthier, and sexually active individuals may be more susceptible to real or perceived beta-blocker side effects and thus less adherent. In this case, the effectiveness of beta-blockers could be missed if beta-blocker adherence was also acting as a surrogate for more advanced disease. Although the authors did attempt to adjust for certain confounding factors, their data were limited to billing data, which can lack the granularity and specificity to fully account for all potential confounders.
Although the conclusions reached by Korhonen et al. (2) should be considered preliminary, the authors should be credited with raising such a provocative hypothesis. The first beta-blocker trials in MI are over 3 decades old (4), and much has changed in the clinical management for MI patients in the interim. This includes better short-term management of MI, new concomitant drugs, and quicker and more durable revascularization options. These advances have halved MI mortality rates, and may have reduced the incremental effectiveness of beta-blockers in the modern era. But only a large randomized trial (RCT) could confirm whether or not this hypothesis is correct.
Although such a proposed trial idea is intriguing, we suspect that there is a low likelihood that such a study would ever be undertaken. The current paradigm of RCTs is not suited to trials of “less” medication. Research sponsors, focused on finding novel, effective therapies, have relatively little incentive to prove that dropping a given therapy is safe. Additionally, new therapies tend to be evaluated in RCTs in a sequential fashion, where new medications are tested as add-ons to existing standard practice. Nor are most trials able to truly determine when patients can safely “stop” therapy. Subjects in RCTs are monitored for a finite time period (typically until the treatment curves diverge) and usually are not followed for prolonged periods to see whether the treatment effects are sustained. As a result, long-term effects of therapy are often extrapolated, leading to guideline recommendations for indefinite therapy for beta-blockers, ACE inhibitors, and statins for nearly all patients following MI. Given that these medications are now generic and low cost, designing a trial to determine the safety of stopping one of these medications is unlikely.
The final sticky issue raised by Korhonen et al. (2) centers on what can be done to reduce medication nonadherence. A number of potentially useful tools to improve patient adherence have been proposed, including: patient education, behavioral counseling, electronic and mobile reminders, financial and nonfinancial incentives, and even family and peer engagement (5). Although it is unlikely that we will find a single “magic bullet” that works for all patients, as Malcolm Gladwell (1) points out, “[We] don’t need to tackle the whole problem, all at once.” Even modest improvements in adherence rates can result in clinically significant improvements in patient outcomes (6).
Despite this, investment in the evaluation of adherence interventions to date has been miniscule in comparison to the huge sums spent on RCTs of new therapies. Even when conducted, adherence studies tend to be unpowered, performed in selected patient groups and clinical settings, often lack control groups, and typically fail to follow patients long term. This underinvestment in implementation seems short-sighted, if not paradoxical—sponsors spend billions of dollars on the development of a new drug only to accept that a majority of patients will likely discontinue their therapy within 6 months.
We believe that data from the current study should be a call to action. The field of medicine should learn from their colleagues in marketing, behavioral economics, and social science to identify which levers are most effective for improving patient medication adherence, and then test the impact of moving those levers in large outcomes trials. How many more studies documenting poor adherence do we need to see before we reach the tipping point?
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Navar is supported by National Heart, Lung, and Blood Institute grant K01HL133416. Dr. Peterson has received research funding from AstraZeneca, Janssen Pharma, Merck & Co., Regeneron Pharm, Genentech, Eli Lilly, and Sanofi; and has received consulting support from Janssen, Merck and Co., Bayer, Daiichi-Sankyo, and Sanofi. Dr. Navar has been a consultant for Amgen, Sanofi, and Regeneron; and has received fees for research consulting from Amgen and Sanofi.
- 2017 American College of Cardiology Foundation
- Gladwell M.
- Korhonen M.J.,
- Robinson J.G.,
- Annis I.E.,
- et al.