Diabetes and Depression Linked With Adverse Outcomes in HF
Diabetes and Depression Linked With Adverse Outcomes in HF
Participants (n=663) were primarily white (76%), males (69%), and aged 61±13 years (Table 1). A majority of participants were in NYHA functional classes III and IV (58%) with an LVEF of 30±14%. The mean PHQ-9 score for all participants was 7.51±6; one-third of participants (33%) were classified as having depressive symptoms based on the cut point. However, more than half of the participants were prescribed antidepressant medication (56%). Forty per cent were diagnosed with diabetes, and nearly three-fourths of these participants reported being a current smoker (70%). Half of the participants had a history of CAD; a majority had a history of HTN (68%).
When categorized based on the cut point for the depressive symptoms measure and the presence or absence of diabetes, the largest proportion of participants had neither (40%). Twenty per cent of participants reported depressive symptoms, but were without a diabetes diagnosis. Twenty-seven per cent were with diabetes, but did not have depressive symptoms. Fourteen per cent of participants had both diabetes and depressive symptoms. We compared demographical and clinical variables for these four groups (Table 2). Although the overall p value for age was significant (p=0.03), post hoc testing with a Bonferroni comparison revealed no significant differences among the groups. BMI in those with both diabetes and depressive symptoms, and those with diabetes only, was significantly higher than in those in the group with neither diabetes nor depressive symptoms (p<0.001); BMI in those who had diabetes only was also significantly higher than in those who had depressive symptoms only (p≤0.001). LVEF in those participants who had diabetes only was significantly greater than in those with only depressive symptoms (p=0.005). There were fewer women than expected in the group with diabetes only, and more women than expected in the group with both diabetes and depressive symptoms (p=0.006). The group without diabetes or depression had a greater proportion of patients classified as NYHA class I/II than predicted, while those in the diabetes alone group and the group with both diabetes and depressive symptoms had fewer participants than expected in NYHA class I/II.
During the study follow-up period, 47 (7%) participants died. Diabetes and depression were treated as categorical variables, and the four categories previously defined were entered in a hierarchical Cox regression model to predict all-cause mortality (Table 3). Variables were entered into the regression in three blocks to determine predictors of mortality during the follow-up period. In the first block, demographic variables (age, gender, and ethnicity) were entered. In the second block, NYHA functional class, smoking status, CAD and HTN diagnosis, BMI, LVEF and antidepressant prescription were entered. In the final block, the categories (no diabetes or depressive symptoms, depressive symptoms alone, diabetes alone, both diabetes and depressive symptoms) were entered. The final regression model was significant (χ=82.10, p<0.001). Independent predictors of increased mortality included the presence of both diabetes and depression (p=0.005), the presence of depression alone (p=0.05), white ethnicity (p=0.02), NYHA functional class (p=0.01) and prescription of antidepressant medication (p<0.001). Patients with HF with both diabetes and depression were 3.7 times more likely to die compared to those participants without either condition; patients with HF with depression alone were 2.3 times more likely to die compared to those with neither diabetes nor depression (figure 1). Those with white ethnicity were six times more likely to die compared to African-Americans. Those participants in NYHA functional classes III and IV were 2.7 times more likely to die compared to those in functional classes I and II; those prescribed antidepressant medication had a 7.8 times greater likelihood of mortality compared to those not prescribed these medications. Predictors of reduced mortality in the model included absence of a CAD diagnosis (p=0.001), increased BMI (p=0.02) and greater LVEF (p=0.002). Those participants without a CAD diagnosis were 70% less likely to die during the follow-up period. A 1 unit increase in BMI was associated with a 6% reduction in likelihood of mortality; a 1% increase in LVEF was associated with a 4% reduction in likelihood of mortality.
(Enlarge Image)
Figure 1.
All-cause mortality according to comorbid symptoms of diabetes and depression.
In a sensitivity analysis of these data to confirm these findings, the prior four groups were removed from the analysis and depression was entered as a continuous variable using the total score of PHQ-9; diabetes was entered as a categorical variable. All other variables were entered unchanged. These results demonstrated that diabetes (HR 0.93; 95% CI 0.54 to 1.75; p=0.929) was not a significant predictor of mortality. However, depressive symptoms (HR 1.12; 95% CI 1.06 to 1.18; p<0.001) were a significant predictor of mortality; for every 1 unit increase in the PHQ-9 score, the likelihood of mortality increased by 12%. In further analyses, we entered β-blocker use, ACE blocker use, diuretic use and other comorbidities. These analyses did not alter the results.
During the follow-up period, there were 115 (17%) hospitalizations for cardiac reasons. Diabetes and depression were again treated as categorical variables, and the four categories previously defined were entered in a hierarchical Cox regression in three blocks to determine predictors of cardiac rehospitalization (Table 4). In the first block, demographic variables (age, gender and ethnicity) were entered. In the second block, NYHA smoking history, CAD history, HTN history, BMI, LVEF and antidepressant use were entered. The final block contained the four categories of patients (neither diabetes nor depressive symptoms, diabetes only, depressive symptoms only, both diabetes and depressive symptoms). The final regression model was significant (χ=42.23, p<0.001). Independent predictors of increased risk for cardiac rehospitalization were the presence of both diabetes and depressive symptoms, and prescription of antidepressant medication. Those participants with both diabetes and depressive symptoms were 2.4 times more likely to be rehospitalized compared to those without either diabetes or depression (figure 2). Patients who were prescribed antidepressant medications were 1.7 times more likely to be rehospitalized compared to those without a prescription. Again, sensitivity analysis was carried out with removal of the four categories of patients and depressive symptoms entered as a continuous variable and diabetes as a categorical variable. Diabetes (HR 1.28; 95% CI 0.89 to 1.86; p=0.189) was not a significant predictor of rehospitalization in these patients. However, depressive symptoms (HR 1.06; 95% CI 1.03 to 1.10; p<0.001) were a significant predictor of rehospitalization; for each 1 unit increase in the PHQ-9 score, the likelihood of cardiac rehospitalization increased by 6%. In further analyses, we entered β-blocker use, ACE blocker use, diuretic use and other comorbidities. These analyses did not alter the results.
(Enlarge Image)
Figure 2.
Cardiac re-hospitalization according to comorbid symptoms of diabetes and depression.
Results
Characteristics of the Participants
Participants (n=663) were primarily white (76%), males (69%), and aged 61±13 years (Table 1). A majority of participants were in NYHA functional classes III and IV (58%) with an LVEF of 30±14%. The mean PHQ-9 score for all participants was 7.51±6; one-third of participants (33%) were classified as having depressive symptoms based on the cut point. However, more than half of the participants were prescribed antidepressant medication (56%). Forty per cent were diagnosed with diabetes, and nearly three-fourths of these participants reported being a current smoker (70%). Half of the participants had a history of CAD; a majority had a history of HTN (68%).
When categorized based on the cut point for the depressive symptoms measure and the presence or absence of diabetes, the largest proportion of participants had neither (40%). Twenty per cent of participants reported depressive symptoms, but were without a diabetes diagnosis. Twenty-seven per cent were with diabetes, but did not have depressive symptoms. Fourteen per cent of participants had both diabetes and depressive symptoms. We compared demographical and clinical variables for these four groups (Table 2). Although the overall p value for age was significant (p=0.03), post hoc testing with a Bonferroni comparison revealed no significant differences among the groups. BMI in those with both diabetes and depressive symptoms, and those with diabetes only, was significantly higher than in those in the group with neither diabetes nor depressive symptoms (p<0.001); BMI in those who had diabetes only was also significantly higher than in those who had depressive symptoms only (p≤0.001). LVEF in those participants who had diabetes only was significantly greater than in those with only depressive symptoms (p=0.005). There were fewer women than expected in the group with diabetes only, and more women than expected in the group with both diabetes and depressive symptoms (p=0.006). The group without diabetes or depression had a greater proportion of patients classified as NYHA class I/II than predicted, while those in the diabetes alone group and the group with both diabetes and depressive symptoms had fewer participants than expected in NYHA class I/II.
Predictors of All-cause Mortality
During the study follow-up period, 47 (7%) participants died. Diabetes and depression were treated as categorical variables, and the four categories previously defined were entered in a hierarchical Cox regression model to predict all-cause mortality (Table 3). Variables were entered into the regression in three blocks to determine predictors of mortality during the follow-up period. In the first block, demographic variables (age, gender, and ethnicity) were entered. In the second block, NYHA functional class, smoking status, CAD and HTN diagnosis, BMI, LVEF and antidepressant prescription were entered. In the final block, the categories (no diabetes or depressive symptoms, depressive symptoms alone, diabetes alone, both diabetes and depressive symptoms) were entered. The final regression model was significant (χ=82.10, p<0.001). Independent predictors of increased mortality included the presence of both diabetes and depression (p=0.005), the presence of depression alone (p=0.05), white ethnicity (p=0.02), NYHA functional class (p=0.01) and prescription of antidepressant medication (p<0.001). Patients with HF with both diabetes and depression were 3.7 times more likely to die compared to those participants without either condition; patients with HF with depression alone were 2.3 times more likely to die compared to those with neither diabetes nor depression (figure 1). Those with white ethnicity were six times more likely to die compared to African-Americans. Those participants in NYHA functional classes III and IV were 2.7 times more likely to die compared to those in functional classes I and II; those prescribed antidepressant medication had a 7.8 times greater likelihood of mortality compared to those not prescribed these medications. Predictors of reduced mortality in the model included absence of a CAD diagnosis (p=0.001), increased BMI (p=0.02) and greater LVEF (p=0.002). Those participants without a CAD diagnosis were 70% less likely to die during the follow-up period. A 1 unit increase in BMI was associated with a 6% reduction in likelihood of mortality; a 1% increase in LVEF was associated with a 4% reduction in likelihood of mortality.
(Enlarge Image)
Figure 1.
All-cause mortality according to comorbid symptoms of diabetes and depression.
In a sensitivity analysis of these data to confirm these findings, the prior four groups were removed from the analysis and depression was entered as a continuous variable using the total score of PHQ-9; diabetes was entered as a categorical variable. All other variables were entered unchanged. These results demonstrated that diabetes (HR 0.93; 95% CI 0.54 to 1.75; p=0.929) was not a significant predictor of mortality. However, depressive symptoms (HR 1.12; 95% CI 1.06 to 1.18; p<0.001) were a significant predictor of mortality; for every 1 unit increase in the PHQ-9 score, the likelihood of mortality increased by 12%. In further analyses, we entered β-blocker use, ACE blocker use, diuretic use and other comorbidities. These analyses did not alter the results.
Predictors of Cardiac Rehospitalization
During the follow-up period, there were 115 (17%) hospitalizations for cardiac reasons. Diabetes and depression were again treated as categorical variables, and the four categories previously defined were entered in a hierarchical Cox regression in three blocks to determine predictors of cardiac rehospitalization (Table 4). In the first block, demographic variables (age, gender and ethnicity) were entered. In the second block, NYHA smoking history, CAD history, HTN history, BMI, LVEF and antidepressant use were entered. The final block contained the four categories of patients (neither diabetes nor depressive symptoms, diabetes only, depressive symptoms only, both diabetes and depressive symptoms). The final regression model was significant (χ=42.23, p<0.001). Independent predictors of increased risk for cardiac rehospitalization were the presence of both diabetes and depressive symptoms, and prescription of antidepressant medication. Those participants with both diabetes and depressive symptoms were 2.4 times more likely to be rehospitalized compared to those without either diabetes or depression (figure 2). Patients who were prescribed antidepressant medications were 1.7 times more likely to be rehospitalized compared to those without a prescription. Again, sensitivity analysis was carried out with removal of the four categories of patients and depressive symptoms entered as a continuous variable and diabetes as a categorical variable. Diabetes (HR 1.28; 95% CI 0.89 to 1.86; p=0.189) was not a significant predictor of rehospitalization in these patients. However, depressive symptoms (HR 1.06; 95% CI 1.03 to 1.10; p<0.001) were a significant predictor of rehospitalization; for each 1 unit increase in the PHQ-9 score, the likelihood of cardiac rehospitalization increased by 6%. In further analyses, we entered β-blocker use, ACE blocker use, diuretic use and other comorbidities. These analyses did not alter the results.
(Enlarge Image)
Figure 2.
Cardiac re-hospitalization according to comorbid symptoms of diabetes and depression.