The complexity of glycemic management in type 2 diabetes mellitus (T2DM) has increased dramatically in the past 15 years. In 1995, the drugs available for treatment of T2DM were insulin and sulfonylureas. Since then, 9 new drug classes have become available, significantly increasing the number of clinical options for physicians and patients. The expanded treatment options currently available, in turn, have produced more opportunities for individualized, patient-centered treatment approaches, while creating additional challenges. For example, among T2DM patients, there is substantial heterogeneity in clinical outcomes and patient preferences regarding which health outcomes and treatment effectsmatter to them most. For the physician who seeks an approach that maximizes an individual patient’s likelihood of responding favorably to treatment while optimizing other considerations (eg, quality of life, functional ability, healthcare spending), the challenge is made greater by an insufficient evidence base to inform clinical decision making.
At a minimum, such an evidence base would include data on the comparative effectiveness of various treatment options—both overall and for specific subgroups of patients—as well as data on patient preferences that drive treatment decisions and, often, health outcomes. Comparative effectiveness research (CER) plays an important role in generating evidence for patients, physicians, and payers; it is increasingly conspicuous in discussions about optimizing patient-centered care for T2DM. CER compares the benefits and harms of alternative treatment options to determine “what works best for which patients under what circumstances.”1 By also assessing utilization and costs, CER can provide a foundation for cost-effectiveness analysis,2 an important approach for identifying high-value health care.3
The hallmark of CER is the comparison of clinically relevant alternative diagnostic or management strategies in representative clinical practice populations. With its multiple treatment alternatives and heterogeneity of patient outcomes, T2DM management is well-suited to this type of research. Accordingly, CER is increasingly used within the diabetes arena. For example, a form of CER was used to evaluate available drug therapies for T2DM.4 However, without an economic evaluation or measurement of comparative clinical effectiveness in real-world settings, the findings are limited. In another example, a major pharmaceutical company developed its phase III clinical trials program using CER. Drawing on insights of an expert panel, the company developed a clinical research approach to provide clinical and economic data once the trials were completed, with a particular focus on enhancing liraglutide’s entry into the market and integration into formularies.5
Finally, a recent comprehensive review of randomized control trials (RCTs) and observational studies by the Agency for Healthcare Research and Quality (AHRQ) identified several gaps in the
evidence on the effectiveness of oral agents for T2DM. These gaps will limit clinicians in providing patient-centered care.6,7
Comparative Effectiveness Research Working Group
To better understand how CER may be used to help improve patient-centered T2DM care, we convened a multidisciplinary working group that included patient representatives as well as a range
of experts in: diabetes care, technology assessment, pharmacology, health economics, evidence synthesis, systematic reviews, clinical decision making, guideline development, epidemiology,
clinical trials, and public policy. The group considered the following questions:
1. What are the limitations in the available evidence for patient-centered T2DM care in diabetes?
2. What outcomes are important to patients and, therefore, should be included in studies of diabetes management?
3. How should RCTs be modified to improve the evidence base for patientcentered care?
4. How should observational studies be designed to improve the evidence base for patient-centered care?
The working group made recommendations, by consensus, for how CER could be used to improve the evidence base for patient-centered diabetes care in order to make results of future diabetes management studies more useful. The final recommendations are summarized here.
Limitations of the Evidence Base for Patient-Centered Diabetes Care
The working group highlighted 5 gaps in the evidence base for T2DM patient-centered care: (1) limited evidence on long-term and patient-reported outcomes; (2) the nonrepresentativeness of patient populations and clinical settings— particularly in clinical trials; (3) the dearth of systematic data on patient subgroups; (4) the insufficient attention paid to social, cultural, and economic factors that influence care; and (5) the comparatively few direct comparisons among alternative treatment strategies. We discuss each in turn.
Limited Evidence Regarding Long- Term and Patient-Reported Outcomes
The comparatively little evidence on long-term outcomes is striking: many outcomes important to clinicians and patients are not tracked or reported. The AHRQ review, for example, found insufficient evidence to conclude that alternative T2DM treatments result in improvements in total mortality, cardiovascular mortality or morbidity.6,7 No study in the AHRQ analysis addressed retinopathy. Only 3 studies evaluated neuropathy, and these had significant methodological flaws.6,7 The working group also noted a lack of postmarketing surveillance, which limits the likelihood of identifying adverse events. Another limitation of the available evidence identified by the working group is the comparatively little attention paid to results that patients find most important, which we call “patientcentered outcomes.” The working group highlighted the importance of outcomes such as satisfaction with care, functional ability, and quality of life. Other outcomes that may be significant to patients include therapeutic side effects (such as weight gain and hypoglycemia), convenience, and cost. Patient- centered outcomes are important because they can influence adherence to care, among other things. Adherence is particularly challenging when patients may not fully believe in the value of the prescribed medications or if they find the regimens difficult to follow. To prevent long-term complications, diabetes care also often includes treating patients who are asymptomatic.
Nonrepresentativeness of Patient Populations and Clinical Settings
It is well understood that for purposes of methodological rigor, statistical power, and regulatory requirements, randomized controlled trials frequently are conducted with highly selective patient samples. For example, patients with a variety of comorbidities, poor adherence, or limited access to healthcare often are ineligible for clinical trials. Yet these are the patients most likely to pose management challenges in a clinical setting. Among the 166 studies examined by AHRQ, information was insufficient on patients with varying levels of cardiovascular and renal risk, with comorbid conditions, and the elderly.7 Finally, few studies report the recruitment methods used, making it impossible to judge how representative the trial population is likely to be. For example, trials that recruit from large urban teaching hospitals might cover different patient populations than trials based in community clinics. In addition, trials that recruit through physicians might draw different patients than those recruiting using more direct methods to access patients. It is often difficult to know how these various approaches affect the sample-frame of the study.
There is also wide agreement regarding the comparative “artificiality” of clinical trial settings. In terms of a more specific gap in the evidence base for T2DM care, a concern raised by the working group is the vigilant monitoring, support, and follow-up patients receive in a clinical trial compared with the reality of the “real-world” setting, which may contribute to differences in the effectiveness of a given therapeutic intervention. Furthermore, few trials report the study settings, which makes it difficult to assess how the results apply to individual practices.7
Dearth of Systematic Data on Patient Subgroups
Clinical trial participants always vary in terms of demographics, comorbidities, disease states, and other potentially significant dimensions. When the variation in treatment response is substantial in a trial, the overall result might not be applicable to all enrolled patients. Yet trials often are not large enough to permit meaningful analyses of subgroups.8
Pooling data from subgroup analyses is one way to overcome this limitation.9 Such research underscores the importance of assessing and reporting results of therapies in clinically important subgroups, in part to enable pooling of subgroup results from different studies.
Increasingly, large health plan databases allow for rudimentary comparisons of outcomes and costs.10-12 These analyses can serve as hypothesis-generating tools for more detailed economic and clinical analyses of best practices for patient care. For example, Onur and colleagues used a large commercial US healthcare data source to study the effectiveness of adding rapid-acting insulin to basal insulin therapy (with or without concomitant oral agent therapy).13 Both overall and diabetes-related healthcare costs were reduced. These results suggest that rapid-acting insulin in this population can improve glycemia and perhaps health, but meaningful numbers are small, rates of hypoglycemia unknown, and optimal patterns for dosing and administering rapid-acting insulin unknown.
Insufficient Attention Paid to Social, Cultural, and Economic Factors That Influence Care
Differences in education, patient-physician relationships, income, and cultural norms can all influence patient management, but are not often addressed in RCTs or observational studies. Treatment plans, for example, must account for such cultural and social factors as food insecurity, economic hardship, or even the celebratory role of food in many cultures. Beliefs about alternative approaches to health (use of herbal products and supplements) also should be considered.
Comparatively Few Direct Comparisons Among Alternative Treatment Strategies
Finally, the working group noted that, in view of the vast number of treatment alternatives now available, there are a number of important comparisons among alternative treatments that have not been systematically examined to date.7 For example, there are few good studies of comparative effectiveness and safety of 2 drug combinations or of monotherapy and combination therapy involving meglitinides, dipeptidyl peptidase- 4 (DPP-IV) inhibitors, and glucagon- like peptide-1 agonists with other first-line diabetes medications. There are also few comparisons with a basal or premixed insulin added to metformin or thiazolidinediones.7 The absence of some key treatment comparisons limits the ability of clinicians to determine the best treatment alternative for patients and to provide patient-centered care. However, the working group recognized that even if head-to-head trials were available, care for specific patients must be individualized.
The problems identified above present important challenges for the provision of patient-centered care for patients with diabetes. The following recommendations (Table 1) are viewed by this working group as essential for improving the relevance of RCTs and observational studies to the accumulating evidence base for patient-centered diabetes management. The recommendations are consistent with the Institute of Medicine’s report on comparative effectiveness.1 Furthermore, many of the areas identified by the group could be addressed by the newly created Patient Centered Outcomes Research Institute (PCORI), as well as by experts in academia and the private sector who conduct this type of research.
Recommendation 1. Outcome measures in research on the management of T2DM should include long-term health outcomes and other patient-reported outcomes. Many current trials focus on intermediate end points, primarily glucose control, and fail to provide direct evidence of clinical outcomes such as mortality, morbidity, complications, and adverse effects of treatment. The working group affirmed the importance of long-term outcomes, including mortality, cardiovascular mortality, stroke, cancer, and osteoporosis, while acknowledging the difficulty of ascertaining these long-term outcomes owing to the length of follow-up, sample sizes required, costs, and because the effect of therapy early in the course of disease will be confounded by the effect of therapy later in disease.
The working group also emphasized that, to the extent possible, patient-centered outcomes should be included both in randomized trials and observational studies. For example, anxiety about hypoglycemic episodes may be an important barrier for many patients. Trials should include measures of adherence and persistence with treatment, as these are key considerations in successful patient-centered care. Finally, studies should report the goals of therapy, given that success should be assessed relative both to the targets and the adverse events associated with therapy.
Recommendation 2. Randomized control trials used in comparative effective research should be designed with the key decision-makers and objectives in mind. Because the goal of comparative effectiveness is to help consumers, clinicians, purchasers, and policy makers make informed decisions, it is important that there be trials specifically designed to compare treatments and outcomes that are important to these groups. This includes, for example, trials which compare clinically relevant alternatives (eg, active comparators in appropriate doses) rather than an active treatment with a placebo or an alternative that is ineffective or unlikely to be used clinically.
The working group noted that trials should also include patients who are representative of those seen in clinical practice. Many trials are designed to minimize potential confounding factors (eg, comorbid conditions) through the use of strict inclusion/exclusion criteria and selective samples. Consider the result, however: those patients with T2DM and comorbidities who are most likely to be excluded from clinical trials are typically older and more likely to be socioeconomically disadvantaged. It is important that these patients be included in trials when feasible to increase the trials’ relevance to the treated population. 7 RCTs designed in accordance with these recommendations have been called pragmatic, or practical, clinical trials.14,15 Such trials are designed to show whether management strategies work in conditions that resemble, as much as possible, actual practice. In contrast, many current RCTs aim to determine the benefit of an intervention under “ideal” circumstances and often are performed to satisfy approval requirements of the US Food and Drug Administration (FDA).14 As Sox and Greenfield note, such trials often ask what works, rather than which therapy works best compared with other therapies.16 Clearly, there is a need for both types of studies.
Recommendation 3. Enhanced patient registries could serve as a basis for observational studies. Observational studies follow patients as they are provided care in more typical clinical settings. They can play an important complementary role in CER in diabetes. They are often useful for identifying potential harms. The advantages of well-designed observational studies include the ability to assess the applicability of evidence derived through RCTs, assess how treatment is used in practice, study populations and subpopulations not studied in clinical trials, and provide long-term follow-up for large numbers of patients.17 Referencing these advantages, the working group recommended enhanced patient registries that could serve as the basis for observational studies. These registries would combine administrative data, laboratory data, relevant clinical data from medical records, and hemoglobin A1C targets, with patient-reported outcomes such as hypoglycemia, quality of life, and satisfaction with care. Such registries could build on the traditional strengths of observational studies while addressing some of their limitations.
Some designs and analytic approaches are more effective than others in controlling for confounds often found in observational studies. Relatively new statistical methods, such as marginal structural models, for example, are designed to reduce inferential errors that result from confounding.18 These methComparods assume that potential confounders are observable. In light of this, an important aspect of enhanced patient registries is that they include sufficient clinical detail to capture the patient characteristics (eg, comorbid conditions) that influence both the choice of therapy and the outcomes of interest. In addition, some observational study designs (quasi-experimental designs, instrumental variable approaches) can control for confounding even if the confounders are not observable.19
The use of comparative effectiveness observational studies based on enhanced registries would have several advantages. First, they would include patients more representative of clinical practice. Second, they would enable investigators to assess adherence to treatment regimens under real-world conditions in patients with varying socioeconomic, educational, and cultural backgrounds. Finally, these studies could track patient-centered outcomes over much longer time periods than is feasible in RCTs.
The ease of conducting observational studies will increase as the use of electronic medical records increases. For example, a recent study discovered a previously unknown drug interaction
between paroxetine and pravastatin that raised glucose levels. The researchers analyzed the FDA’s adverse event reporting system for side-effect profiles, and then assessed glucose levels in patients receiving both medications in three populations via electronic medical records.20 Such a study would have been impossible even in the recent past.
To provide patient-centered care, clinicians must identify the outcomes important to patients and understand how different therapies affect those outcomes. Despite many RCTs and observational studies that have been conducted on diabetes management, there are major gaps in the evidence that limit clinicians’ ability to provide patient-centered care. These gaps exist in part because RCTs are designed for regulatory purposes or to demonstrate causality and the safety and efficacy of an intervention under “ideal” circumstances.
As a consequence, important outcomes are not always addressed, the patient populations and settings may be nonrepresentative, subgroup differences on the outcome measures of interest are not well-reported, and direct comparisons among treatment alternatives are lacking. These gaps can be addressed by some minor modifications to the outcome measures utilized in RCTs, the use of enhanced patient registries that can serve as the foundation for high-quality observational studies, and reliance on quasi-experimental designs for observational studies.
A large trial funded by the National Institutes of Health (NIH), The Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE), should provide helpful information on how to best choose pharmacologic treatments for the management of T2DM. It is a multicenter RCT among patients with recent onset T2DM. It will compare glimepiride, sitagliptin, liraglutide, and glargine as add-on therapy to metformin. The trial has been designed following principles of CER and will seek to collect information on healthcare use, patient preferences, and quality of life. The results should help guide patients, clinicians, pharmacists, and health plan administrators in the best treatment approaches for T2DM.
The aim of our recommendations is to facilitate the development of evidence that can inform patient-centered decision making. We also highlight these issues because PCORI is setting an agenda for patient-centered outcomes research. Implementing the above recommendations would lead to improved representativeness of patients and care settings and better evidence about the real-world outcomes from alternative treatment choices. In turn, studies that more comprehensively capture patient-centered outcomes will better inform clinical guidelines for care. Acknowledgments:
All working group members were supported by Sanofi US.
We would like to acknowledge Kari Edwards for her helpful edits and helpful reviews on earlier versions of the manuscript.
Funding disclosure: The working group activities were funded by Sanofi US.
Authorship Forms are available at Evidence-Based Diabetes Management. Contact email@example.com.
1. Institute of Medicine. Initial national priorities for comparative effectiveness research. Vol 2010. Washington, DC: National Academies Press; 2009.
2. Garber A, Sox H. The role of costs in comparative effectiveness research. Health Aff. 2010;29(10):1805-1811.
3. Owens DK, Qaseem A, Chou R, Shekelle P. High-value, cost-conscious health care: concepts for clinicians to evaluate the benefits, harms, and costs of medical interventions. Ann Intern Med.
4. Ahmann A. Application of comparative effectiveness research in evaluation of treatments for T2DM. Am J Manag Care. 2011;17(2):S41-S51.
5. Oderda G, Sifford-Wilson S. Bringing liraglutide to market: a CER case study. J Manag Care Pharm. 2012;18(5 supp A):S12-S18.
6. Bennett W, Maruthur N, Singh S, et al. Comparodsative effectiveness and safety of medications for T2DM: an update including new drugs and 2-drug combinations. Ann Intern Med. 2011;154(9):602-613.
7. Bennett W, Wilson L, Bolen S, et al. Oral diabetes medications for adults with T2DM: an update. In: AHRQ, ed. Vol Comparative Effectiveness Review No. 27. Rockville, MD March 2011.
8. Bravata DM, Gienger AL, McDonald KM, et al. Systematic review: the comparative effectiveness of percutaneous coronary interventions and coronary artery bypass graft surgery. Ann Intern
9. Hlatky M, Boothroyd D, Bravata D, et al. Coronary artery bypass surgery compared with percutaneous coronary interventions for multivessel disease: a collaborative analysis of individual
patient data from ten randomised trials. Lancet. 2009;373(9670):1190-1197.
10. Best J, Romley J, Goldman D, Conrad R, Peters A. Expenditures for Medicaid patients treated with exenatide compared with other diabetes management regimens. Am J Manag Care.
11. Baser O, Wei W, Baser E, Xie L. Clinical and economic outcomes in patients with T2DM initiating insulin glargine disposable pen versus exenatide BID. 2011;14(6):673-680.
12. Aagren M, Luo W, Moës E. Healthcare utilization changes in relation to treatment intensification with insulin aspart in patietns with T2DM. Data from a large US managed-care organization.
J Med Econ. 2010;13(1):16-22.
13. Levin P, Wei W, Pan C, Douglas D, Baser O. Combination therapy with insulin glargine and exenatide: real-world outcomes in patients with T2DM. Current Medical Research and Opinion.
14. Luce B, Kramer J, Goodman S, et al. Rethinking randomized clinical trials for comparative effectiveness research: the need for transformational change. Ann Intern Med. 2009;151(3):206-
15. Tunis S, Stryer D, Clancy C. Practical clinical trials: increasing the value of clinical research for decision making in clinical and health policy. JAMA. 2003;290(12):1624-1632.
16. Sox HC, Greenfield S. Comparative effectiveness research: a report from the Institute of Medicine. Ann Intern Med. 2009;151(3):203-205.
17. Dreyer N, Tunis S, Berger M, Ollendorf D, Mattox P, Gliklich R. Why observational studies should be among the tools used in comparative effectiveness research. Health Aff. 2010;29(10):1818-1825.
18. Klungsøyr O, Sexton J, Sandanger I, Nygård J. Sensitivity analysis for unmeasured confounding in a marginal structural Cox proportional hazards model. Lifetime Data Analysis. 2009;15(2):278-294.
19. Marra G, Radice R. A flexible instrumental variable approach. Statistical Modeling. 2011 2011;11(6):581-603.20. Tatonetti N, Denny J, Murphy S, et al. Detecting drug interactions from adverse-event reports: interaction between paroxetine and pravastatin increases blood glucose levels. Clin Pharmacol Ther. 2011;90(1):133-142.
20. Tatonetti N, Denny J, Murphy S, et al. Detecting drug interactions from adverse-event reports: interaction between paroxetine and pravastatin increases blood glucose levels. Clin Pharmacol Ther. 2011;90(1):133-142.