Potential overtreatment of diabetes mellitus in older adults with tight glycemic control.

Methods

We analyzed data from NHANES years 2001–2002, 2003–2004, 2005–2006, 2007–2008, and 2009–2010. The NHANES uses stratified, multistage, probability-cluster techniques to ensure that sample populations are representative of the nation’s noninstitutionalized civilians. Data are collected from household interviews and standardized medical examinations and blood sample collections are performed in mobile examination centers.

This study was deemed exempt from further review by the University of California, San Francisco, Institutional Review Board because it used only deidentified secondary data.

We included adults from the NHANES who were 65 years or older, reported a diagnosis of diabetes from a health professional, and had an HbA1c measurement. We used interview responses to classify participants in terms of age, sex, and race or ethnic group.

We classified older adults with diabetes into 3 health status categories endorsed by the ADA/AGS framework for considering treatment goals for glycemia.13 The 3 categories include those who are relatively healthy, those with complex medical histories for whom self-care may be difficult, and those with a very significant comorbid illness and functional impairments, many of whom may have limited life expectancy.15,16 Participants were categorized as having very complex/poor health status if they were receiving dialysis or had 2 or more activities of daily living (ADL) impairments. We could not determine the presence of other indicators of very complex/poor health status per the ADA/AGS framework, including end-stage (stage III–IV) congestive heart failure, oxygen-dependent lung disease, uncontrolled metastatic cancer, or severe cognitive impairment. Participants were categorized as having complex/intermediate health status if they had 3 or more chronic conditions or 2 or more instrumental ADL impairments. We included chronic illnesses identified by the ADA/AGS framework, including arthritis, congestive heart failure, lung disease, chronic kidney disease, coronary heart disease, stroke, or urinary incontinence, but we did not have information on active cancer, clinical diagnosis of depression, or falls. We also did not include hypertension because it is highly prevalent and usually not considered a serious chronic illness. Finally, we categorized participants as relatively healthy if they did not meet these criteria.

We used interview responses to identify chronic conditions (congestive heart disease, lung disease [emphysema, chronic bronchitis, or asthma], coronary heart disease [myocardial infarction or angina pectoris], stroke, or arthritis). Urinary incontinence status was based on a series of questions about leakage of urine with or without activity like coughing, lifting, or exercise and with or without preceding urge or pressure to urinate. We considered urinary incontinence to be a chronic condition if it occurred at least a few times a week. Chronic kidney disease was identified based on an estimated glomerular filtration rate of less than 60 mL/min/1.73cm2, calculated using the Chronic Kidney Disease Epidemiology Collaboration equation (http://www.ncbi.nlm.nih.gov/pubmed/19414839).

Functional limitations were assessed based on a series of questions designed to measure participants’ functional status. These questions were phrased to assess the individual’s level of difficulty in performing the task without using any special equipment. Patients who reported some or much difficulty or were unable to perform ADL were categorized as having ADL impairment. For ADLs, we used questions about dressing, feeding, walking from room to room, and getting in or out of bed. For instrumental ADLs, we used questions about preparing one’s own meals, managing money, and housework chores. Other ADLs and instrumental ADLs were not assessed in the NHANES.

We categorized glycemic control, based on the measured HbA1c level, as tight (HbA1c level, <7%), moderate (7%–8.9%), and poor control (≥9%). In additional analyses, we examined very tight control (HbA1c level, <6.5%). During 2001–2010, there were 3 laboratory instruments for measuring HbA1c and 2 laboratories used in the NHANES. Measurement of HbA1c was performed in the first laboratory using the Primus CLC330 and Primus CLC385 (Primus Corp) (from 2001–2004) and then in the second laboratory on the Tosoh A1C G7 (Tosoh Medics, Inc) (from 2005–2006). From 2007–2010, HbA1c testing was performed in the second laboratory on the Tosoh A1C G7. Laboratory method crossover studies were conducted at the time of each of the laboratory instrument changes. Both laboratories that analyzed NHANES HbA1c data from 2001–2010 were standardized by participating in the National Glycohemoglobin Standardization Program, and no adjustment across years for the HbA1c assay has been recommended.

Participants were asked to report prescription medications they had taken in the past 30 days and to bring medication bottles to the examination, where the information was documented. Receipt and type of oral glucose-lowering treatment was based on review of medications brought in to the examination; insulin use was based on review of medications brought in or self-reported use of insulin, because participants may not always bring vials or pens to the examination.

We calculated the weighted proportions of survey participants with glycemia that was poorly (HbA1c level, ≥9%), moderately (7%–8.9%), or tightly (<7%) controlled across health status categories. In addition, we calculated the weighted proportions of survey participants whose glycemia was tightly controlled and were treated with either insulin or sulfonylureas across health status categories. We conducted logistic regression analyses to assess linear trends in proportions of participants with tightly controlled glycemia, their health status, and patterns of treatment during the 5 NHANES surveys. To preserve statistical power in these trend analyses, we combined participants with complex/intermediate and very complex/poor health status into 1 category. All analyses incorporated a complex survey design using NHANES-recommended methods to produce nationally representative estimates. All data, except where otherwise noted, show annualized estimates of the number of US adults with the outcome of interest based on the mean of values across the 10 study years. Analyses were performed using SAS, version 9.3 (SAS Institute, Inc) and Stata SE, version 12 (StataCorp). We considered 2-sided P < .05 to be statistically significant.

Results

During the 10-year study period, we identified 6667 adults 65 years or older, of whom 1373 (20.6%) reported a diagnosis of diabetes (eFigure in the Supplement). For this analysis, we included 1288 participants who had an HbA1c measurement during the survey period (94.7 of participants with diabetes, representing 4.4 million-7.5 million older adults during each 2-year survey period). The characteristics of the study sample are presented in Table 1. The mean (SD) age was 73.2 (5.7) years, and 20.8% were 80 years or older. More than one-third of older adults reported at least 1 ADL impairment; similarly, more than one-third reported at least 1 instrumental ADL impairment. In the study sample of older adults with diabetes, 50.7% (95% CI, 46.6%–54.8%), representing 3.1 million (95% CI, 2.7–3.5) US adults, were relatively healthy, 28.1% (95% CI, 24.8%–31.5%) (1.7 million; 95% CI, 1.4–2.0) had complex/intermediate health, and 21.2% (95% CI, 18.3%–24.4%) (1.3 million; 95% CI, 1.1–1.5) had very complex/poor health. About one-fourth and one-third of older adults with complex/intermediate and very complex/poor health status, respectively, were 80 years or older compared with only 13.6% of those who were relatively healthy.

Among older adults with diabetes, 61.5% (95% CI, 57.5%–65.3%) had an HbA1c level of less than 7% (3.8 million; 95% CI, 3.4–4.2), 32.2% (95% CI, 28.7%–35.9%) had an HbA1c level of 7% to 8.9% (2.0 million; 95% CI, 1.6–2.3), and 6% (95% CI, 5%–8%) had an HbA1c level of 9% or greater (0.4 million; 95% CI, 0.3–0.5). A total of 41.9% (95% CI, 37.9%–46.1%) of older adults had an HbA1c level of less than 6.5% (2.6 million; 95% CI, 2.2–2.9) and 19.6% (95% CI, 17.0%–22.6%) had an HbA1c level of 6.5% to 7% or less (1.2 million; 95% CI, 1.0–1.4) There were no significant differences in the proportions of patients who attained tight (HbA1c level, <7%), moderate (HbA1c level, 7%–8.9%), or poor (HbA1c level, ≥9%) glycemic control across health status categories (P = .43) (Figure 1). Specifically, 62.8% (95% CI, 56.9%–68.3%) of adults who were relatively healthy, 63.0% (95% CI, 57.0%–68.6%) of those with complex/intermediate health, and 56.4% (95% CI, 49.7%–62.9%) of those with very complex/poor health had an HbA1c level of less than 7% (P = .26). Notably, 44.9% (95% CI, 38.7%–51.3%) and 37.9% (95% CI, 32.2%–44.0%) of patients with complex/intermediate and very complex/poor health had an HbA1c level of less than 6.5%, respectively.

Among older adults with an HbA1c level of less than 7%, 54.9% (95% CI 50.4%–59.3%) were treated with either insulin or sulfonylureas (4.0% were treated with both; eTable in the Supplement). This proportion did not differ by health status: 50.8% (95% CI, 45.1%–56.5%), 58.7% (95% CI, 49.4%–67.5%), and 60.0% (95% CI, 51.4%–68.1%) of participants with relatively healthy, complex/intermediate, and very complex/poor health status received insulin or sulfonylurea, respectively (P = .14) (Figure 2). Similarly, large proportions of participants with HbA1c levels of less than 6.5% were treated with insulin or sulfonylureas across health status categories (43.9% [95% CI, 36.8%–51.3%] of relatively healthy, 52.3% [95% CI, 40.4%–64.0%] with complex/intermediate health, and 56.3% [95% CI, 44.1%–67.8%] with very complex/poor health).

During the 10 years, there were no significant trends in the proportion of older adults with diabetes who had an HbA1c level of less than 7% (P = .34) or the proportion with an HbA1c level of less than 7% who had complex/intermediate or very complex/poor health (P = .27). Among participants with an HbA1c level of less than 7% who were relatively healthy, the proportion treated with insulin or sulfonylureas decreased over time (P = .05). However, among participants with an HbA1c level of less than 7% who had complex/intermediate or very complex/poor health, treatment with insulin or sulfonylureas remained stable over time (P = .65) (Table 2).

Discussion

In a nationally representative sample of noninstitutionalized adults from 2001 through 2010, a total of 62% of older participants with diabetes had an HbA1c level of less than 7%, corresponding to 3.8 million US persons. Despite unproven benefits and potential harms of tight glycemic control in older persons with extensive comorbidities, most older adults with complex/intermediate and very complex/poor health had an HbA1c level of less than 7%, corresponding to 1.8 million persons. Moreover, approximately 60% (or 1.0 million) of adults with complex or very complex medical problems were treated with insulin or sulfonylureas to achieve tight glycemic targets, which may lead to severe hypoglycemia. We did not find significant changes in treatment patterns across health status over time. Our findings suggest that a substantial proportion of older adults with diabetes in the United States were potentially overtreated.

Consistent with our findings, studies based on national surveillance data from 1999 through 2010 that assessed trends in risk-factor control among US adults with diabetes suggest that most older adults met stringent glycemic control goals during this time.14,17 However, although individualized glycemic targets were considered to some extent in 1 of these studies,17 the authors did not take into account achieved glycemic targets that may be too low. Similarly, current performance metrics for glycemic control in diabetes focus exclusively on achieved glycemic targets below a certain threshold but do not provide a lower acceptable limit.

Studies from the Department of Veteran Affairs also suggest a high prevalence of potential overtreatment among adults with diabetes.18,19 In 1 study, the prevalence of potential over-treatment, defined as an HbA1c level of less than 7%, treatment with insulin or sulfonylurea medications, and coexisting risk factors for hypoglycemia (age ≥75 years; elevated creatinine level of ≥2 mg/dL [to convert to micromoles per liter, multiply by 88.4]; or cognitive impairment or dementia), was approximately 50%.19 These results indicate that a substantial proportion of veterans received intensive treatment despite a high risk for serious hypoglycemia. Our research supports and extends these findings in several ways. We used nationally representative samples (with equal numbers of men and women) served by different health systems that are more generalizable to the US population at large. In addition, we incorporated functional impairments, which are an important risk factor for adverse outcomes in diabetes treatment but are not captured in administrative records such as those used in the study by Tseng et al.19 Finally, we were able to detect changes in potential overtreatment over time.

Tight glycemic control, such as an HbA1c level of less than 7% (or even <6.5%) may be appropriate in patients who are relatively healthy, with a long life expectancy, in whom long-term benefits of glycemic control are more likely to materialize. Tight glycemic control may also be appropriate in patients with complex/intermediate or very complex/poor health, if it is consistent with patient goals of care and achieved through lifestyle modification or low-risk medications, such as metformin hydrochloride. However, intensive glycemic control strategies markedly increase the risk of hypoglycemia.20–22 In turn, hypoglycemia has been associated with poor outcomes, such as increased mortality, cardiovascular disease, falls and accidents, dementia, and low health-related quality of life.23–30 Therefore, intensive strategies to lower glucose levels may result in more harm than benefit, particularly among older, sicker patients.

Our study suggests that a substantial number of older adults with diabetes were potentially overtreated. We incorporated the presence of chronic comorbidities and functional impairments to determine participant health status, according to a framework developed jointly by the ADA and AGS, to guide glycemic treatment targets among older adults.13 Based on this framework, we estimated that approximately 1 million older adults with diabetes attained tight glycemic control with the use of insulin or sulfonylureas, despite complex/intermediate or very complex/poor health, indicating potential overtreatment. Given incomplete information on all comorbidities in our study, it is likely that we underestimated the complexity of health status, and thus, true estimates of adults who are potentially overtreated may be even higher.

To motivate improvements in quality, Pogach and Aron31 have proposed an overtreatment measure to minimize treatment that is unlikely to provide benefit and may result in harm. Based on electronic health records, this measure would identify patients with diabetes with an HbA1c level of less than 7% who are at high risk for hypoglycemia, including all persons 75 years or older. The measure would then prompt physicians and other health care professionals to reevaluate therapy.31 Our findings suggest that such a measure might identify a substantial proportion of older adults with diabetes whose therapy requires careful reassessment.

Our study has some limitations. We combined NHANES data from 2001 through 2010 to increase our sample size, but our study may have been underpowered to detect more subtle changes in treatment patterns over time. The ADA has only recently endorsed individualized glycemic targets, and it is possible that potential overtreatment may have declined since 2010; however, the Veteran Affairs and AGS guidelines have endorsed this approach for nearly a decade and we detected no signal to suggest a change. Use of medications was based on prescriptions brought in for examination; a small percentage of medications were classified in NHANES as combination therapy and were not otherwise subcategorized, although some may have included sulfonylureas. We categorized participants according to health status based on a limited number of questions about ADL and instrumental ADL impairments and did not take into account dementia or cognitive impairment; therefore, we may have overestimated the number of adults who were relatively healthy and underestimated potential overtreatment. Some combinations of comorbid conditions used to categorize health status may have stronger associations with hypoglycemia and life expectancy than others; however, to the extent possible, we have followed the ADA/AGS framework,13 which does not distinguish between these combinations. Since NHANES is a study of noninstitutionalized adults, our findings do not apply to older adults who live in nursing homes or other facilities. We were not able to determine whether potential overtreatment directly resulted in harm, such as hypoglycemia. Achieved HbA1c level is a poor predictor of self-reported serious hypoglycemia,32 although multiple randomized trials showed that intensive glycemic control substantially raises hypoglycemia risk.20–22 Finally, a recent study suggests a shift in the distribution of HbA1c levels toward higher values across NHANES cycles33; this would tend to result in underestimation of potential overtreatment.

Conclusions

Using a nationally representative sample of US adults, we showed that nearly two-thirds of older adults with diabetes who have complex/intermediate or very complex/poor health attained tight glycemic control. These vulnerable adults are unlikely to experience the benefits of intensive glycemic control and instead are likely to experience harms from treatment, such as hypoglycemia and other adverse effects. Recognition of both the harms and benefits of glycemic control is critical for patients and physicians and other health care professionals to make informed decisions about glucose-lowering treatment.