Type 2 Diabetes in the Real World: The Elusive Nature of Glycemic Control

Large randomized controlled trials (RCTs) have clearly demonstrated that achieving and sustaining optimal glycemic control prevents or delays the development of microvascular and macrovascular disease (1–3). Although the risk of developing diabetes-related complications rises steadily when glycated hemoglobin (HbA_1c) values are in excess of 6.5% (48 mmol/mol) (4,5), an HbA_1c of <7% (<53 mmol/mol) is generally considered a target goal for diabetes management (6). In truth, it is now widely recognized that individualizing the HbA_1c goal for each patient is critically important, especially when the presence of comorbidities necessitates less stringent targets (6). However, most government and private insurer reports continue to use 7% as a point of reference; therefore, we focus herein on this marker of glycemic control.

Despite our growing understanding of diabetes and the availability of new medications and technologies, a substantial number of individuals are not at their glycemic goal. Of note, recent data indicate that 85.6% of adults with diagnosed diabetes are treated with diabetes medication (7). Results from the National Health and Nutrition Examination Survey (NHANES) indicate that only about 50% of American adults with diabetes are achieving HbA_1c <7.0% (<53 mmol/mol) (8), and it is estimated that only 64% are reaching individualized glycemic goals (9). These findings are noteworthy because they provide nationally representative snapshots with which to assess care outcomes realized over time in real-world settings. After an initial increase in the percentage of patients attaining HbA_1c <7.0% (<53 mmol/mol) between the survey periods 1988–1994 (44%) and 2003–2006 (57%), performance stagnated, with a new wave of data showing percentages for the 2007–2010 and 2011–2014 periods at 52% and 51%, respectively (9) (Fig. 1A).

The Healthcare Effectiveness Data and Information Set (HEDIS) results, which include data from over 1,000 health plans covering over 171 million people in 2014, are even more troubling. In 2014, approximately 40% of commercially insured HMO patients and 30% of government-insured patients achieved HbA_1c <7.0 (<53 mmol/mol), again with no change over the past decade (10) (Fig. 1B).

While these worrisome findings are certainly skewed, at least to some extent, by the unknown number of individuals whose individualized HbA_1c targets are >7%, it remains evident that a considerable percentage of U.S. adults with diabetes are in persistent poor glycemic control. The question of why these population trends stand in such sharp contrast to the generally favorable results emerging from RCTs has not been fully explored. This Perspective calls out the distressing lack of improved glycemic control in this country over the past decade, as well as the notable gap between clinical trial and real-world results, and provides an urgent challenge to consider new approaches to achieving meaningful and sustained outcomes.

The number of diagnosed cases of diabetes has increased fourfold from 1980 through 2014 (from 5.5 million to 22.0 million), with type 2 diabetes making up the vast majority of cases (11). While a growing number of options are available to treat type 2 diabetes—over 40 new drugs, including combination products, for type 2 diabetes have been approved since 2005 (12)—the most recent analysis of NHANES data currently being prepared for publication suggests that this explosion in available options has not contributed to a meaningful improvement in glycemic control (8,9). This has profound implications, considering that poorly controlled diabetes is a recognized cause of severe microvascular and macrovascular complications (7). If the present trends in incidence and prevalence continue without change, it is estimated that one-third of Americans will have diabetes by 2050, portending even higher rates of morbidity and mortality as a result of poor glycemic control (13).

RCTs have been the “gold standard” of study design because they tightly control the setting and delivery of interventions, minimize the effect of external factors on outcomes, and lead to a random distribution of unmeasured confounders. Yet the degree to which results obtained under RCT conditions can be extrapolated to real life remains an open question (14,15). Indeed, it is important to be aware of the ways in which clinical trial results might inflate expectations of treatment efficacy. First, therapy interventions are often more focused in the unusual setting of the clinical trial, where patients may benefit from more frequent face-to-face visits, convenient access to therapy, closer monitoring, and wider availability of educational resources and support services (14,16–18). Second, trial participants are often more concerned with their health and treatment and thus more motivated to actively participate in their own care. Third, committing to a protocol over a defined period of time, sometimes with the benefit of financial incentives, may make it easier for patients to follow medication instructions appropriately (19,20).

We now have the opportunity to use administrative claims data in conjunction with clinical trial data to guide clinical decision-making. In the case of type 2 diabetes, evidence from a new claims data study (21) may help to explain the discouraging NHANES and HEDIS findings mentioned above and, in so doing, validate what we intuitively know to be true—that clinical trial outcomes and actual clinical care often tell different stories. Using a large electronic medical record–administrative claims database (Optum Humedica SmartFile, 2007–2014), this retrospective analysis was undertaken to assess HbA_1c reductions in patients initiating a glucagon-like peptide 1 receptor agonist (GLP-1 RA) or a dipeptidyl peptidase 4 (DPP-4) inhibitor, to quantify the gap between real-world (i.e., usual care) and RCT efficacy, and to calculate the relative contribution of various biobehavioral factors in explaining this gap (21).

Specifically, investigators identified adult patients with type 2 diabetes who initiated a GLP-1 RA (n = 221) or DPP-4 inhibitor (n = 652) and then compared their real-world HbA_1c change over a 12-month period with that of participants in 11 pivotal RCTs, seven for the GLP-1 RA class and four for the DPP-4 inhibitor class (22–32). Of note is that although HbA_1c change was measured over a 12-month period for the real-world samples, only three of the 11 comparator RCTs were of similar length (27,29,32). This may somewhat reduce the gap between real-world HbA_1c change and the changes seen in the comparator RCTs. Additionally, the data set does not provide data regarding dosages in the real-world samples, which may differ from the dosages used in the comparator RCTs. Nevertheless, the study (21) revealed that in patients with similar inclusion criteria characteristics, HbA_1c reductions in the usual care group fell far short of those reported in the RCTs: HbA_1c reductions in the real-world group were −0.52% for the GLP-1 RAs and −0.51% for the DPP-4 inhibitors, whereas in the RCTs the changes ranged from −0.84% to −1.60% for the GLP-1 RA groups and from −0.47% to −0.90% for the DPP-4 inhibitor groups (Fig. 2). The finding that HbA_1c reductions were similar between the real-world treatment groups for GLP-1 RAs and DPP-4 inhibitors was surprising considering that clinical trial comparisons of incretin-based therapies consistently show greater HbA_1c reductions from baseline with GLP-1 RAs than with DPP-4 inhibitors (33).

To better understand the differences between usual care and clinical trial HbA_1c results, multivariate regression analysis assessed the relative contributions of key biobehavioral factors, including baseline patient characteristics, drug therapy, and medication adherence (21). Significantly, the key driver was poor medication adherence, accounting for 75% of the gap (Fig. 3). Adherence was defined, following the Centers for Medicare & Medicaid Services recommendation, as the filling of one’s diabetes prescription often enough to cover ≥80% of the time one was recommended to be taking the medication (34). By this metric, proportion of days covered (PDC) ≥80%, only 29% of patients were adherent to GLP-1 RA treatment and 37% to DPP-4 inhibitor treatment.

These data are consistent with previous real-world studies, which have demonstrated that poor medication adherence to both oral and injectable antidiabetes agents is very common (35–37). For example, a retrospective analysis of the MarketScan Commercial and Medicare Supplemental databases targeting adults initiating oral agents in the DPP-4 inhibitor (n = 61,399), sulfonylurea (n = 134,961), and thiazolidinedione (n = 42,012) classes found that adherence rates, as measured by PDC ≥80% at the 1-year mark after the initial prescription, were below 50% for all three classes, at 47.3%, 41.2%, and 36.7%, respectively (36). Rates dropped even lower at the 2-year follow-up (36) (Fig. 4A).

Another claims database study of 1,321 patients with type 2 diabetes who were treated with once-daily liraglutide underscores the particular challenge posed by the injectable GLP-1 RA class, as only 34% (n = 454) were classified as adherent as measured by PDC ≥80% (37). Furthermore, as one might expect, patients in the adherent versus poorly adherent group were more likely to achieve HbA_1c goals of <7.0% (<53 mmol/mol) (50% vs. 39%, P < 0.001) and to have at least a 1.0% reduction in their HbA_1c (38% vs. 32%, P = 0.022).

Data exploring persistence, defined as the continuation of treatment for the prescribed duration but allowing for a strictly defined permissible gap (e.g., days or months) after the last expected refill date, show even more alarming trends. A 2013 utilization study conducted in the Truven Health Analytics MarketScan commercial health insurance database revealed that only 18% of DPP-4 inhibitor users (n = 208,683) and 11% of GLP-1 RA users (n = 124,925) were persistent with therapy (nonpersistence in this study was defined as a gap of over 30 days after the last expected refill date) during the first year of treatment (38) (Fig. 4B).

Our current ability to assess adherence and persistence is based primarily on review of pharmacy records, which may underestimate the extent of the problem. For example, using the definition of adherence of the Centers for Medicare & Medicaid Services—PDC ≥80%—a patient could miss up to 20% of days covered and still be considered adherent. In retrospective studies of persistence, the permissible gap after the last expected refill date often extends up to 90 days (39,40). Thus, a patient may have a gap of up to 90 days and still be considered persistent.

Additionally, one must also consider the issue of primary nonadherence; adherence and persistence studies typically only include patients who have completed a first refill. A recent study of e-prescription data among 75,589 insured patients found that nearly one-third of new e-prescriptions for diabetes medications were never filled (41). Finally, none of these measures take into account if the patient is actually ingesting or injecting the medication after acquiring his or her refills.

Acknowledging and addressing the problem of poor medication adherence is pivotal because of the well-documented dire consequences: a greater likelihood of long-term complications, more frequent hospitalizations, higher health care costs, and elevated mortality rates (42–45). In patients younger than 65, hospitalization risk in one study (n = 137,277) was found to be 30% at the lowest level of adherence to antidiabetes medications (1–19%) versus 13% at the highest adherence quintile (80–100%); hospitalization risk was defined as the probability of one or more all-cause hospitalizations during a 12-month period (43). In patients over 65, a separate study (n = 123,235) found that all-cause hospitalization risk was 37.4% in adherent cohorts (PDC ≥80%) versus 56.2% in poorly adherent cohorts (PDC <20%) (45). This latter study also found better adherence to be associated with significantly lower all-cause outpatient costs, acute care costs, and total medical costs when compared with poorer adherence thresholds (all P < 0.001) over 3 years. Better adherence was also linked to fewer emergency room visits, shorter hospital length of stay, and lower risk of acute complications (all P < 0.001) (45). Furthermore, for every 1,000 patients who increased adherence to their antidiabetes medications by just 1%, the total medical cost savings was estimated to be $65,464 over 3 years (45).

The cumulative impact of poor adherence is also supported by an earlier retrospective cohort study of 11,532 patients with diabetes in the Kaiser Permanente of Colorado registry. There was a 39% increased risk for all-cause mortality associated with poor adherence to oral hypoglycemic agents (42). Additionally, an overlooked result of the recently completed Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results (LEADER) trial provides further insight regarding the influence of adherence. While the primary analysis showed that liraglutide lowered cardiovascular deaths by 22% compared with placebo in the global findings, this reduction in risk was not observed in the North American sample; according to the study sponsor, “for reasons that are still unclear, the N.A. [North American] patient groups tend to have lower compliance and adherence compared to global rates during large cardiovascular studies” (46,47).

In total, these new data oblige those of us who treat people with diabetes to be even more mindful of the potential for poor adherence. There are many potential contributors to poor medication adherence, including depressive affect, negative treatment perceptions, lack of patient-physician trust, complexity of the medication regimen, tolerability, and cost (48). Although these factors have not been well studied, it is believed that many of them can be addressed with appropriate education, thorough instruction in medication administration, ongoing and respectful patient-physician interactions, and shared decision-making between patients and health care professionals (49–51). However, these results are often not easy to achieve, especially given the limited time available for a meaningful discussion during typically short clinic visits. Medication adherence is not a single behavior but rather a dynamic constellation of behaviors influenced by social, environmental, and individual circumstances that defy “one-size-fits-all” solutions (52). A recent review of interventions addressing problematic medication adherence in type 2 diabetes found that few strategies have been shown consistently to have a marked positive impact, particularly with respect to HbA_1c lowering, and no single intervention was identified that could be applied successfully to all patients with type 2 diabetes (53). Additional evidence indicates that improvements resulting from the few effective interventions, such as pharmacy-based counseling or nurse-managed home telemonitoring, often wane once the programs end (54,55).

We suspect that the efficacy of behavioral interventions to address medication adherence will continue to be limited until there are more focused efforts to address three common and often unappreciated patient obstacles. First, taking diabetes medications is a burdensome and often difficult activity for many of our patients. Rather than just encouraging patients to do a better job of tolerating this burden, more work is needed to make the process easier and more convenient. For example, once-weekly medications, combination pills, and new innovations in delivery methods may be beneficial.

Second, poor medication adherence often represents underlying attitudinal problems that may not be a strictly behavioral issue. Specifically, negative beliefs about prescribed medications are pervasive among patients, and behavioral interventions cannot be effective unless these beliefs are addressed directly (35). Therefore, improved communication between patients and clinicians regarding the benefits and risks of specific treatment options, as well as wider adoption of shared decision-making strategies, is warranted (35).

Third, the issue of access to medications remains a primary concern. A study by Kurlander et al. (51) found that patients selectively forgo medications because of cost; however, noncost factors, such as beliefs, satisfaction with medication-related information, and depression, are also influential.

Only about half of patients with type 2 diabetes are meeting glycemic goals (8–10), and there has been negligible change in the percentage of individuals achieving their target goals over the last decade. Despite the approval of 40 new treatment options for type 2 diabetes since 2005 (12), these therapies and approaches have not had a meaningful impact on the degree of glycemic control in a large subset of the population with diabetes (9). Although these treatment options have shown notable efficacy in RCTs, their impact on glycemic control in real-world clinical practice has been minimal.

Poor medication adherence and persistence continue at alarming rates. They are demonstrably key contributors to the disconnect between RCT and real-world results in lowering HbA_1c and are implicated in higher morbidity, mortality, and health care costs (43,44). Achieving meaningful and sustained glycemic control requires innovative approaches for the real world. In order to help our patients achieve meaningful and sustained HbA_1c reductions—and move the dial positively on the next NHANES and HEDIS HbA_1c analyses—we appeal to the diabetes community to drive even harder for innovative approaches designed with the real world in mind.

Acknowledgments. The authors thank Carol A. Verderese, The Diabetes Education Group, Lakeville, CT, and Christopher G. Parkin, CGParkin Communications, Inc., Boulder City, NV, for their editorial support.

Funding. Intarcia Therapeutics, Inc. provided funding for editorial support in the writing of this manuscript.

Duality of Interest. S.V.E. is a consultant for Intarcia, Sanofi, AstraZeneca, Eli Lilly, Dexcom, Bayer, Abbott, and Johnson & Johnson. W.H.P. is a consultant for Intarcia, Sanofi, Eli Lilly, Novo Nordisk, Dexcom, Roche, Abbott, and AstraZeneca. No other potential conflicts of interest relevant to this article were reported.