Effect of medication nonadherence in diabetes mellitus

The efficacy of HMG-CoA reductase inhibitors (statins), angiotensin-converting enzyme inhibitors, and oral hypoglycemic medications in patients with diabetes mellitus has been shown in clinical trials.¹ In clinical practice, a large percentage of patients with diabetes, however, have not reached guideline-recommended targets for blood pressure, low-density lipoprotein (LDL) cholesterol levels, and glycosylated hemoglobin (A1C).² This may be the result of nonadherence to medication regimens.³ Therefore, we assessed the relationship between clinical outcomes and medication nonadherence among patients with diabetes. In particular, we analyzed the association between nonadherence and achievement of treatment targets (blood pressure, LDL cholesterol levels, and A1C), all-cause hospitalization rates, and all-cause mortality rates.

Subjects and methods

We evaluated adherence to medications among subjects who had been continuously enrolled in the Kaiser Permanente of Colorado diabetes registry from September 17, 2002, to December 31, 2003. Adherence was determined by calculating the proportion of days covered (PDC), defined by the total number of days supplied for filled prescriptions over the observation period (calendar year). We calculated a PDC for each of the types of medications considered important for patients with diabetes (statins, antihypertensives, and oral hypoglycemics). For subjects who were prescribed medications from several categories, a summary adherence measure was computed as the averaged PDC of any 1 or more categories of medications. Subjects with a PDC < 0.80 were considered nonadherent.

The primary end points of the study were all-cause hospitalization and all-cause mortality between January 1, 2004, and April 30, 2005. Attainment of treatment targets for LDL cholesterol level, blood pressure, and A1C level were the secondary end points. Low-density lipoprotein cholesterol and A1C levels were taken from the most recent measurement in calendar year 2004, while blood pressure measurements were based on the average of the 2 most recent 2004 readings. Multivariable regression analyses were used to assess the association between medication nonadherence and the outcomes of interest, adjusting for demographics, comorbidities, and baseline LDL cholesterol levels, blood pressure, and A1C levels.

Follow-up continued until subjects disenrolled from the health plan or died. As of April 30, 2005, 96.6% of subjects were still enrolled or had died. The mean follow-up period was 474 days.

Results

Among the cohort of subjects with diabetes, 13.0% required both insulin and oral hypoglycemic agents, 57.9% required hypoglycemic agents only, 9.7% required insulin therapy only, and 20.4% were diet controlled. Based on the summary adherence measure, 21.3% of subjects were considered nonadherent. As shown in Table 1, these subjects had fewer comorbidities and were younger than adherent subjects.

P

P

P

At baseline, the percentage of subjects at goal for blood pressure, A1C, and LDL cholesterol levels was not significantly different between adherent and nonadherent subjects. But as shown in Table 2, nonadherent subjects had higher systolic and diastolic blood pressures, higher LDL cholesterol levels, and higher A1C levels than adherent patients during the follow-up period. Unadjusted analysis showed that nonadherent subjects had higher all-cause mortality (5.9% vs 4.0%; < .01) and higher all-cause hospitalizations (23.2% vs 19.2%; < .01) than adherent subjects. As shown in Table 3, nonadherence to medication remained significantly associated with a greater risk of all-cause mortality (odds ratio [OR] = 1.58; 95% confidence interval [CI], 1.38-1.81; < .01) in multivariable analysis. The findings of an increased mortality risk were consistent when nonadherence was based on the individual categories of medication (ie, statins, oral hypoglycemics, and antihypertensives) and when different cut-off values were used to define nonadherence, ranging from summary PDC < 0.50 to PDC < 1.00.

P

P

We then evaluated the association between incremental increases in medication adherence and the primary and secondary end points. A decrease in systolic (—1.0 mm Hg; 95% CI, –1.5 to –0.6 mm Hg) and diastolic (–1.2 mm Hg; 95% CI, –1.4 to –0.9 mm Hg) blood pressure was shown for each 25% increase in antihypertensive medication adherence. A reduction of –3.8 mg/dL (95% CI, –4.5 to –3.0 mg/dL) in LDL cholesterol and –0.05% (95% CI, –0.08 to –0.01%) in A1C levels was also shown with each 25% increase in adherence to statin and oral hypoglycemic medications, respectively. Lower all-cause mortality (OR = 0.75; 95% CI, 0.68-0.83; < .01) and all-cause hospitalization (OR = 0.83; 95% CI, 0.79-0.88; < .01) were also associated with each 25% increase in adherence.

Discussion

Our goal in this study was to assess the relationship between clinical outcomes and nonadherence to medications in a large community cohort of subjects with diabetes. Results showed that approximately 20% of subjects were nonadherent, which correlated with higher blood pressure, LDL cholesterol, and A1C, and was associated with greater risk of all-cause hospitalization and all-cause mortality.

Other studies examining medication adherence in patients with diabetes have shown adherence rates of 36% to 93%, depending on the method used to assess adherence, the sample studied, and the medication evaluated.⁴ These studies have predominantly concentrated on intermediate outcomes. Improvement in LDL cholesterol and glycemic levels have been shown with increasing levels of adherence to statin and oral hypoglycemic medications, respectively.⁵ Higher medication adherence has also been correlated with lower medical care costs.⁶ The relationship between medication adherence and subsequent hospitalization and mortality shown in our study adds to current knowledge and highlights the consequences of nonadherence to medications.

Our findings indicate that the correlation of medication adherence with hospitalization and mortality is mediated partially through improvements in intermediary goals. A 25% increase in adherence was associated with lower blood pressure, LDL cholesterol, and A1C levels. However, the extent of the decrease in hospitalization and mortality rates was higher than expected, given the changes in the intermediate end points compared with the clinical trials. This may indicate that adherence may also be associated with other healthy behaviors that are related to patient outcomes.⁷ Regardless of drug assignment in randomized controlled trials, for example, adherent patients had better outcomes.⁷ This may be because adherent patients may also engage in other healthy lifestyle behaviors. Furthermore, nonadherence has be linked to missed medical appointments, cognitive impairment, and depression, which may lead to unfavorable outcomes.³ These results indicate that direct and indirect mechanisms most likely mediate the link between outcomes and medication adherence. Patients who are adherent to therapy are more likely to achieve treatment targets, such as LDL cholesterol levels, and are also more likely to practice healthy lifestyle behaviors, such as exercising regularly, both of which have been associated with reduced mortality risk.

Although nonadherence can have negative consequences,⁸ it is often not evaluated by physicians. Nonadherence may be one of the reasons that patients with diabetes have inadequate glycemic control. But physicians may increase medication dosage, thinking that poor glycemic control is the result of medication ineffectiveness, and this can lead to hypoglycemia if patients suddenly take all of their medications as prescribed. All patients should be evaluated for medication nonadherence, and if nonadherence is found clinicians should engage patients in a discussion of ways to improve adherence. Additional studies are required to further the understanding of nonadherence and to help develop strategies to prevent it.

One of the possible limitations of our study is that we used administrative and clinical records from a single managed care organization. The study was population-based, however, and involved a large, integrated health care delivery group. Another limitation is that the timing of the medication doses was not known and medication use was assumed.⁹ Pharmacy refill records, however, have been correlated with electronic adherence monitoring and are associated with a variety of clinical outcomes.⁹ In addition, refilling a prescription indicates a patient’s intention to continue taking the medication.⁹ A third limitation is that we were not able to measure other adverse health characteristics, such as lower socioeconomic status and depression, which may be characteristics of nonadherent patients and may also contribute to unfavorable outcomes. This study, however, which showed an association between negative outcomes and nonadherence, should provide the stimulus to further investigate the reasons for nonadherence.

Conclusions

We evaluated the association between medication nonadherence and adverse outcomes among subjects with diabetes. Nonadherent subjects had higher blood pressure, A1C, and LDL cholesterol levels. In addition, medication nonadherence was associated with higher risk of all-cause hospitalization and all-cause mortality. All patients should be evaluated for medication adherence, and if nonadherence is determined to be a problem, it should be addressed so that patients can receive the full benefit of medications.¹⁰