Both depression and poor diet quality have been independently associated with poor glycemic control in subjects with T2D, increasing diabetes related complications and decreasing quality of life [6–8]. In this particular sample, diet quality was not associated with symptoms of depression based on the regression analysis. Significant differences in symptoms of depression according to diabetes status and gender were found only among those with the lower diet quality.
Our results contradict findings of previous studies that overall dietary quality, rather than individual nutrients, be related to depression. Higher depression symptoms as measured by the Center for Epidemiologic Studies Depression scale (CES-D) score were associated with lower diet quality as defined by the Alternate Healthy Eating Index among Latinos at risk for T2D . Analysis in this study was not adjusted by energy intake.
Jacka et al.  examined the relationship between depression and habitual diet in 1046 Australian women using a diet quality score derived from a food frequency questionnaire concluding that a "traditional" dietary pattern characterized by whole grains, fish, meat, fruits and vegetables was associated with lower odds for major depression. However, this association was attenuated when analysis was adjusted by overall energy intake. The authors explained this phenomena stating that overall amount of "bad" food consumed in the diet may be more relevant to depression than the proportion it represents in the overall diet.
A study using the original HEI found that depression was associated with poorer diet quality in women with breast cancer . According to the researchers, data analysis was not adjusted by energy intake because several individual components of the original HEI were expressed as a function of energy. Therefore, the overall score controlled for differences in kilocalories ingested. However, the original HEI scores were positively and significantly correlated with energy intake.
In fact, one of the major differences between the original HEI and the HEI-05 is the inclusion of a density approach (amounts per 1000 kcal of intake) because the original HEI had a tendency to measure quantity rather than quality .
In our study, the use of the HEI-05 may be the reason why an association between depression symptoms and diet quality was not found. Subjects with T2D may have better HEI-05 scores because they were eating fewer kilocalories than subjects without T2D. This panorama may have changed if a score not based on 1000 kcal could have been used.
In addition, participants with T2D had a significantly higher BDI score compared to those without T2D, but also, a higher HEI-05 score, meaning that, even thought they had better nutritional habits, they still had more depressive symptoms. This may imply that other factors like gender and diabetes status may be better predictors of depression in this particular sample of Cuban-Americans.
Still our results showed that significant differences among gender and diabetes status are found in participants with the lowest overall diet quality, meaning that special emphasis must be placed in the diets of females and subjects with T2D with symptoms of depression. In addition, since differences are only significant below the median value (HEI-05 ≤ 55.6); this may be an adequate cut off point to divide poor and good diet quality in this population.
The BDI is a practical instrument that can be used in a clinical setting to detect T2D patients with symptoms of depression in order to warrant appropriate nutrition interventions. Nutritional habits are considered environmental factors, which could be changed with adequate nutrition counseling, education and providing related nutritional services, and may have a profound impact in both depression and T2D management.
Limitations of our study included the cross sectional design, which cannot establish causality. In addition, diagnosis of depression was based on a self-reported BDI score and there was no psychiatric or psychological diagnosis of depression. Finally, our low response rate (4%) may indicate that our sample was not representative of the general Cuban-American population and results cannot be generalized. However, this low response rate was expected. Due to their political history, Cuban-Americans are afraid of government control, may be less likely to share information with any organization .