The impact of dementia on antidiabetic drug use in Medicare beneficiaries with diabetes: findings post-Medicare part D

Participants of the MCBS from 2006 to 2010 comprised the population of the study. In the study, we included survey participants if they were 65 years of age or older and had been diagnosed with diabetes. We excluded survey participants if they were institutionalized during the time of the interview or were eligible for Medicare because of end-stage renal disease.

Information regarding the overall and different classes of antidiabetic drugs was measured based on prescribed medicine events. Users of antidiabetic drugs in a given year were defined as having at least one prescribed medicine event with any antidiabetic drug during that year. Antidiabetic drugs were classified into eight categories in the study: alpha-glucosidase inhibitors, biguanides, DPP-4 inhibitors, GLP-1 receptor agonists, insulins, meglitinides, sulfonylureas and thiazolidinediones. Due to the low utilization, other classes of antidiabetic drugs (e.g., bile acid sequestrants, dopamine-2 agonists, sodium-glucose cotransporter-2 inhibitors and amylin mimetics) were not included in the study. Definitions of users of different classes of antidiabetic drugs were similar to users of overall antidiabetic drugs. Since in a given year a diabetic patient may use more than one class of antidiabetic drugs, the classification of antidiabetic drugs was not mutually exclusive in the study.

Based on the Chronic Condition Data Warehouse algorithms along with the relevant The International Classification of Diseases, Ninth Revision, Clinical Modification codes from Medicare part A and B claims, we identified Medicare beneficiaries who had been diagnosed with diabetes, dementia, Alzheimer's disease or vascular dementia. Characteristics of the study population, including age, gender, race, education, marital status, income level, residence status, census region, BMI and smoking status were reported by survey or proxy respondents. Based on Medicare part A and B claims, we also calculated a modified Charlson comorbidity index (CCI) excluding diabetes and dementia to adjust for possible comorbidity confounding.

This study was a pooled cross-sectional study from 2006 to 2010. Since each survey respondent might be interviewed in multiple years, the unit of analysis of the study is person–year. Survey sampling weights were used in the analysis to get national estimates of elderly Medicare beneficiaries with diabetes.

Rao–Scott Chi-square tests, to deal with the complex survey design, were performed to compare characteristics of Medicare beneficiaries with and without dementia and its subtypes. Three weighted logistic regression models (model 1, model 2 and model 3) were conducted in the study to identify predictors of antidiabetic drug use. Dependent variables (antidiabetic drug use) and covariates (used to control for confounding) were the same in all three models. To estimate the impact of dementia and its subtypes on antidiabetic drug use, main independent variables were different in three models. Dementia, Alzheimer's disease and vascular dementia served as main independent variables in model 1, model 2 and model 3, respectively. We further performed subgroup analyses to evaluate the impact of dementia and its subtypes on different classes of antidiabetic drug use. For each class of antidiabetic drugs, three weighted logistic regression models were performed. Results for certain classes of antidiabetic drugs with very low utilization rates were not reported.

Institutional Review Board approval was obtained for this study. SAS software (version 9.4; Statistical Analysis Systems, NC, USA) was used to perform all statistical analyses.

The study identified 8539 person–years observations for elderly Medicare beneficiaries with diabetes. Among them, 683 (6.3% weighted percentage), 296 (2.5%) and 59 (0.5%) had been diagnosed with dementia, Alzheimer's disease and vascular dementia, respectively. In addition, 22 had been diagnosed with both Alzheimer's disease and vascular dementia. Some characteristics of Medicare beneficiaries differed statistically significantly (p < 0.05) between individuals with and without dementia and its subtypes. Compared with elderly Medicare beneficiaries without dementia, dementia patients were more likely to be 75 years of age or older, be non-Hispanic black, have less than high school education, be widowed, have income equal or less than US$20,000 per year, live in other census regions (e.g., Puerto Rico), be underweight (BMI: <18.5) or normal weight (BMI: 18.5–24.9), be a nonsmoker or have one or more comorbidities in addition to dementia (Table 1). Similar types of differences were observed when elderly Medicare beneficiaries with and without subtypes of dementia were compared.

Utilization rates of overall and some classes of antidiabetic drugs were significantly different between elderly Medicare beneficiaries with and without concomitant dementia or Alzheimer's disease (Table 1). About half (51.5%) of dementia patients used antidiabetic drugs; while about three fifths (59.4%) of individuals without dementia used antidiabetic drugs (p = 0.002). Similarly, patients with Alzheimer's disease had lower utilization of antidiabetic drugs (47.8%), compared with those without Alzheimer's disease (59.2%; p = 0.003). For different classes of antidiabetic drugs, lower utilizations of biguanides (p < 0.001), DPP-4 inhibitors (p = 0.024) and thiazolidinediones (p = 0.002) were observed in dementia patients. Patients with Alzheimer's disease had lower utilization of biguanides, compared with individuals without Alzheimer's disease (p < 0.001). Overall and different classes of antidiabetic drug utilizations between individuals with and without vascular dementia were similar.

After controlling for demographic and socioeconomic characteristics of elderly Medicare beneficiaries, dementia patients were about 30% (odds ratio [OR]: 0.69; 95% CI: 0.56–0.84) less likely to take antidiabetic drugs compared with individuals without dementia (Table 2). The likelihood of taking antidiabetic drugs was about 40% (OR: 0.61; 95% CI: 0.44–0.84) lower in patients with Alzheimer's disease compared with individuals without Alzheimer's disease. The probability of taking antidiabetic drugs was not significantly different between individuals with and without vascular dementia. (OR: 0.59; 95% CI: 0.33–1.04) Other predictors of lower utilizations of antidiabetic drugs were younger age, higher level of education and lower BMI.

Same demographic and socioeconomic characteristics were controlled in the subgroup analyses. Compared with individual without dementia, dementia patients were about 40% (OR: 0.62; 95% CI: 0.49–0.79), 20% (OR: 0.78; 95% CI: 0.63–0.96) and 30% (OR: 0.70; 95% CI: 0.53–0.94) less likely to take biguanides, sulfonylureas and thiazolidinediones, respectively (Table 3). The likelihood of using DPP-4 inhibitors, insulins and meglitinides was similar between individuals with and without dementia. The probability to take biguanides and sulfonylureas were about 40% (OR: 0.59; 95% CI: 0.41–0.84) and 30% (OR: 0.69; 95% CI: 0.50–0.97) lower in patients with Alzheimer's disease, compared with individuals without Alzheimer's disease. The likelihood of using DPP-4 inhibitors, insulins, meglitinides and thiazolidinediones was similar between individuals with and without Alzheimer's disease. Utilization rates of biguanides, insulins, sulfonylureas and thiazolidinediones were not significantly different between individuals with and without vascular dementia.

This study provided national estimates on the prevalence of overall and different classes of antidiabetic drug use among elderly Medicare beneficiaries with diabetes. Overall, the prevalence of antidiabetic drug class utilization did not change much as compared with Margolis et al.'s report among diabetic Medicare beneficiaries [11]. However, this study further provided national estimates on the prevalence of overall and different classes of antidiabetic drug utilizations among diabetic patients with dementia and its subtypes. Among diabetic patients without dementia or its subtypes, biguanides were the most frequently used class of antidiabetic drugs; while sulfonylureas ranked as the second. Utilization patterns of biguanides and sulfonylureas were consistent among diabetic patients with vascular dementia. However, among diabetic patients with dementia or Alzheimer's disease, the utilization patterns were reversed, where sulfonylureas ranked as the first and biguanides ranked as the second most frequently used class of antidiabetic drugs. Healthcare providers should be aware of the special utilization patterns of biguanides and sulfonylureas among diabetic patients with dementia or Alzheimer's disease and modify prescription patterns when appropriate.

Dementia poses a significant challenge on the self-management in diabetes. Because of dementia, diabetic patients may have problems when communicating with healthcare providers and forget to fill prescriptions. Furthermore, diabetic patients with dementia may be unable to interpret blood glucose test results. Hence, they cannot adjust antidiabetic drug doses when necessary. In addition, dementia patients may forget to take antidiabetic drugs as prescribed. All of these may contribute to the lower utilizations of overall antidiabetic drugs, biguanides, sulfonylureas and thiazolidinediones among diabetic patients with dementia or Alzheimer's disease observed in the study. Insulin use was not associated with dementia or Alzheimer's disease in the study. The administration of insulins requires visual and motor skills. Diabetic patients with dementia may be unable to inject insulins properly by themselves and need caregivers for help. With the involvement of caregivers, having dementia or Alzheimer's disease should have minimal impact on insulins use.

Antidiabetic drugs can slow the progression of Alzheimer's disease and improve cognitive functions through different mechanisms [6,7]. Biguanides were found to play a neuroprotective role in preventing against apoptotic cell death in primary cortical neurons [12]. Sulfonylureas could inhibit mammalian target of rapamycin, which played a role in developing diabetes and Alzheimer's disease [13]. Thiazolidinediones had been shown to reduce soluble amyloid beta oligomers, which caused synapse deterioration underlying memory loss in Alzheimer's disease [14]. However, this study found that patients with dementia or Alzheimer's disease were less likely to use antidiabetic drugs, biguanides, sulfonylureas and thiazolidinediones. Healthcare providers should be aware of the low utilizations of antidiabetic drugs among diabetic patients with dementia or Alzheimer's disease and prescribe appropriate antidiabetic drugs to them if not contraindicated or tolerated. Strategies are warranted to address the underuse of antidiabetic drugs among Medicare diabetics with a concomitant dementia.

This study is the first of its kind to provide national estimates on the prevalence of overall and different classes of antidiabetic drug utilizations among diabetic patients with dementia, and to assess the impact of dementia and its subtypes on overall and different classes of antidiabetic drug use. Several factors should be considered when interpreting the data. First, disease/conditions were identified based on Medicare claims only. Dementia is frequently undiagnosed and not properly coded in the reimbursement systems [15–17]. The study might underestimate the number of dementia patients. Second, this study was unable to measure the disease severity, which may influence antidiabetic drug use. Third, because of small sample sizes, the impact of dementia and its subtypes on alpha-glucosidase inhibitors and GLP-1 receptor agonists use were unmeasurable. Similarly, the association between vascular dementia and the use of DPP-4 inhibitors and meglitinides were not assessed due to the low prevalence of vascular dementia. Fourth, characteristics of diabetic patients with and without dementia and its subtypes were statistically significant different. Whether they were clinically or epidemiologically different were unknown. Fifth, this study might be subject to endogeneity problems, however, as noted in Table 3, we did control for a large number of confounding factors, including age, gender, race, education, marital status, income, residence, census region, BMI, smoking and CCI. These variables combined should be adequate and the endogeneity problem should be minimal. Finally, due to the cross-sectional design, the study was unable to establish a causal relationship between dementia and antidiabetic drug use. Also, the duration of diabetes and dementia could not be measured in the study.

Despite of the limitations listed above, the findings of the study, which were based on real-world, nationally representative data, provide important information to patients, healthcare givers and/or policy makers in future clinical practice. First, special attention by healthcare providers or family care givers should be given to elderly diabetics with dementia, particularly those with Alzheimer's disease, as they are more likely to under use diabetes medications compared with their counter parts. Strategies such as intensive family care givers, memory aids, reminders, mutual goal setting or incentives and/or reinforcements in future clinical practice may be considered.

Second, certain types of antidiabetic drugs may be significantly under used by diabetics with dementia. These drugs include biguanides, sulfonylureas and thiazolidinediones. Strategies such as education sessions may be helpful by emphasizing the importance of being adherent to these drugs. Third, similar strategies may also be needed for patients with predictive characteristics (e.g., educational status, lower BMI).

In conclusion, about half of elderly diabetic Medicare beneficiaries with dementia used antidiabetic drugs from 2006 to 2010. Patients with dementia were about 30% less likely to use antidiabetic drugs. Sulfonylureas were the most frequently used class of antidiabetic drugs in diabetic patients with dementia. Given the benefits of antidiabetic drugs in treating dementia, healthcare providers should prescribe antidiabetic drugs in diabetic patients with dementia when appropriate. Strategies that focus on how to improve antidiabetic drug use among the elderly diabetics with a concomitant dementia are warranted.

The literature documents that patients with diabetes have a higher risk of developing dementia. Our study showed that diabetic patients with developed dementia significantly underuse antidiabetic drugs compared with their counterparts without dementia. Specifically, patients with dementia were about 30% less likely to use antidiabetic drugs, and thus posing a serious challenge for diabetes management among Medicare population. Healthcare providers should be aware of the potentially significant underuse of biguanides, sulfonylureas and thiazolidinediones when prescribing these drugs. Future studies on the impact of improved use of antidiabetics on outcomes of diabetic patients with comorbid dementia are warranted.

This study was supported by the ASPIRE-II grant (Grant No. 11150A031), ASPIRE-I grant (11150E403) and Social Science Grant (11150A032). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

The authors state that they have obtained appropriate institutional review board approval or have followed the principles outlined in the Declaration of Helsinki for all human or animal experimental investigations. In addition, for investigations involving human subjects, informed consent has been obtained from the participants involved.