Depression, still today one of the most common psychiatric disorders encountered worldwide, produces a substantial burden on public health, representing one of the major causes of disability and generating a significant impact in terms of both healthcare costs and suffering of those affected (1). Estimated lifetime prevalence rates reported for the disease are approximately 15% in high-income countries, including Italy, and 11% in low to middle-income countries, with mean age at onset ranging from 24 to 26 years, and a 2-fold higher risk in females (1). The high burden produced by depression is correlated in part with an increased risk amongst depressed subjects of onset, persistence, and increased severity of a wide range of chronic conditions, including obesity, resulting in a further rise in levels of disability, premature death, and deterioration of quality of life (2–7).

The interaction between depression and medical conditions is highly complex and bidirectional, with each illness producing an impact on onset, course, prognosis, and treatment of the other (8). If, on the one hand, depression is approximately 2 to 3-fold more common in patients affected by a chronic physical condition (i.e., cancer, diabetes, heart conditions, and stroke), being manifested in approximately 20% of this patient population, at the same time, depressed subjects present with an increased incidence of chronic medical conditions compared to the general population (9, 10). In all cases, a condition of comorbidity between a clinically relevant depressive disorder and a somatic disorder is linked to a worse outcome of both the depressive and the somatic disorder (9). The latter finding underlines the importance of systematic assessment and recognition of levels of depression in people affected by somatic disorders, together with a correct management and appropriate treatment of any associated depression disorders.

Amongst the chronic physical conditions associated with depression, diabetes is of particular interest and relevance in view of evidence pointing to an exponential increase in cases of diabetes mellitus worldwide, which has been likened to a global epidemic of diabetes: a prevalence rate in the global population of 9.3% (463 million people) was estimated in 2019, which is likely to reach 10.2% (578 million people) by 2030 and 10.9% (700 million people) by 2045 (11). Based on the growing burden of this condition, even in developing countries, the World Health Organization has included diabetes amongst the chronic disorders requiring major investment in terms of prevention. Type 1 Diabetes Mellitus (T1DM), although manifested less frequently than Type 2 Diabetes Mellitus (T2DM) and representing 5–10% of the total burden of diabetes mellitus, is the most predominant endocrine disorder in pediatric and adolescent populations, and one of the most frequent chronic childhood disorders. From an epidemiological point of view, Sardinia, the second largest Italian island in the Mediterranean, has the second highest incidence of T1DM in the world after Finland (12).

Although both depression and anxiety are commonly observed disorders amongst patients affected by diabetes mellitus, prevalence data obtained for these conditions in the specific population, in particular T1DM, are scarce and somewhat lacking from a methodological viewpoint, and the direction of the association between diabetes and the psychopathological domain remains unclear due to the transversal design of the majority of studies conducted to date (13–15), while there are few studies with a longitudinal design (16, 17). Prevalence rates for depression were more than 3-fold higher in people affected by T1DM (12%, range 5.8–43.3% vs. 3.2%, range 2.7–11.4%) and up to 2-fold higher in those with T2DM (19.1%, range 6.5–33% vs. 10.7%, range 3.8–19.4%) vs. healthy controls; in all cases (T1DM, T2DM, and healthy controls), higher rates of depression were observed in women compared to men (14). Moreover, other studies focusing on the comorbidity between depression and T1DM reported an inconsistency in prevalence rates, although confirming the close relationship between depression and poor glycemic control and lack of self-care in diabetic patients (18, 19). Fisher et al. suggested that the cause of these discrepancies may lie in a methodological issue associated with a difficulty to distinguish between a depressive state and emotional distress linked to management of the chronic disease (20). The link between diabetes and depression has been examined in literature regarding both the concomitant psychological dimensions and potential reciprocal influencing effects of the two disorders. Indeed, depression manifested in subjects affected by diabetes tends to adhere to a recurrent or chronic course (18). The concomitant presence of mood disorders is correlated with a worse prognosis for diabetes (21), a less healthy lifestyle and a greater lack of diabetic self-care (22, 23), lower treatment compliance (24, 25), higher HbA₁c (Glycated Hemoglobin A1c) (19, 24, 26–29), additional complications of diabetes (30, 31), a worsening of quality of life (24, 32) and higher mortality rates (31–33). In addition, the continual self-care required in the treatment of diabetes (blood sugar measurements, multiple doses of insulin, diet, exercise, etc.), is much harder to ensure in patients affected by concomitant depression.

It has been hypothesized that the etiopathogenesis of diabetes and depression may share several relevant aspects, i.e., stress and inflammation, with an increased propensity for T2DM (34), and that the presence of one of the two conditions may increase the risk of onset of the other (13, 35, 36). Moreover, other factors indicated as potentially capable of increasing the risk of depression in diabetic patients include a personal or family history of depression, exposure to stressful or traumatic events such as domestic violence, stress, and the burden, particularly in the case of T1DM, of being exposed to an early onset chronic disease, the presence of physical illnesses and other clinical factors (35, 36). Other authors suggest a series of physiopathological mechanisms common to both conditions, inducing us to consider the existence of a biological link between the two conditions: ranging from the effects produced by the circulating cytokines associated with autoimmune diabetes to the effects of a lack of insulin on neurogenesis and the neurotransmitter metabolism, to chronic hyperglycemia, iatrogenic hyperglycemia, and basal hyperactivity of the hypothalamic–pituitary–adrenal axis (37).

An additional confounding factor has been detected in the relationship between diabetes and psychiatric morbidity determined by the concomitant presence of depression and Eating Disorders (ED) (38, 39). To control diabetes, the patient is required to place particular emphasis on his or her nutrition, constantly monitoring the intake of food which correlates with the insulin dose prescribed. Overall, this persistent focus on food, fundamental in the management and regulation of diabetes, predisposes diabetic patients to developing ED, which are, in turn, more difficult to diagnose in these patients compared to the general population. There is evidence of ED, Disturbed Eating Behaviors (DEB), binge eating and symptoms of bulimia in both T1DM (40–47) and T2DM patients (47, 48), which generally exceed those observed in the general population (38, 40, 43, 45, 49). It has been demonstrated how both ED and DEB are capable of seriously impinging on physical health, particularly in diabetic subjects, and that the long-term complications of diabetes may be exacerbated by DEB and by an improper use of insulin which, by compromising metabolic regulation, result in increased mortality rates (38). Studies have shown an association between poor glycemic control and both ED (50) and subclinical disordered eating (51) in diabetic patients.

Longitudinal data on non-diabetic populations support a bidirectional relationship between depression and eating pathology where both conditions represent a risk factor for the other (52). Comorbidity is very common: more than 40% of individuals with ED reported comorbid mood disorders according to a recent narrative review on European studies (53). Identifying ED in subjects with depressive disorders is critical given the high rate of antidepressants resistance in this population (54), yet most people with ED go undetected even among psychiatric outpatients as they often seek care for comorbid mental disorders (55). Other authors have advocated the need to screen for disordered eating in subjects with Major Depressive Disorder (56).

To summarize there is evidence that: (i) the prevalence of both depression and DEB is higher in patients with diabetes compared to healthy controls (14, 40); (ii) both conditions have been associated with poor metabolic compensation (24, 51); (iii) there is a substantial and clinically relevant relationship between ED and depression in non-diabetic subjects, given the high rates of comorbidity (53), the difficulty to detect ED in patients with depression (56) and its impact on antidepressants response (54). Despite the aforementioned existing evidences, very few studies to date have examined the association between the depressive dimension and DEB in T1DM adult patients (40). Most of the literature has focused on adolescents and women (57), despite the fact the over 60% of depressive disorders has the onset after the age of 25 (58). Moreover, even if ED are more frequent in women (59), men presentations tend to be of similar severity (60), with a recent study finding males with ED more likely to have comorbid depression compared to females (61). Interestingly, a recent longitudinal study on adolescents and young adults with T1DM was conducted with the objective to investigate the directionality of effect linking DEB to depressive symptoms (62). To our knowledge, this was the only study assessing the directionality of disordered eating and depression in T1DM patients. Using cross-lagged analysis, the authors found that DEB predicted increases in depressive symptoms over time, while depressive symptoms at baseline did not predict DEB at follow-up (62). Other published literature mostly explored the association between DEB and depression in T1DM patients using cross-sectional design and/or bivariate analysis, without taking into account potential covariates and confounders (40, 57).

In the present study conducted on a sample of adult outpatients, males and females, affected by T1DM, we aimed to: (i) investigate the current prevalence of DEB and depression and the lifetime prevalence of ED; (ii) explore the relationship between DEB and depression; (iii) test the association of both depression and DEB with metabolic compensation. We hypothesized that:

To our knowledge, this is the first study to examine, in a sample comprehensive of adolescents and adults, males and females, the independent association between depression and DEB in T1DM by means of a detailed psychodiagnostic assessment tailored to the diabetic population.

This was a cross-sectional study in a sample of insulin-treated diabetic outpatients with T1DM aged from 17 to 55 years old. An unselected sample of consecutive patients attending two specialist centers for the diagnosis and treatment of diabetes over a 4-month period (from June to September 2017) was assessed. These centers routinely provide a global medical, psychological, and psychopathological evaluation with a multidisciplinary approach. When necessary, psychological support or psychiatric assessment and intervention are made available. During the 4-month period of the study, the total number of patients registered to the participant centers, including all ages, types of diabetes mellitus (1/2) and hypoglycemic treatment (oral/insulin), was about 2000 (10% with T1DM). The primary study population was made up of 211 insulin-treated diabetic patients aged 17–55 years, 192 patients with a type 1, 19 patients with a type 2 diabetes, 108 males and 103 females. For the purposes of this study, only patients with T1DM were included in the final sample. Patients with a type 2 diabetes (n = 19) and who did not complete the psychometric assessment (n = 20) were excluded. The final sample included 172 participants, 86 males and 86 females. The data used for the purposes of this study derived from the routine assessment of psychopathology in the diabetic patients attending the specialist centers by means of questionnaires, specifically created to assess depression, anxiety, and DEB, and a structured clinical interview used to assess ED. After signing an informed consent form, socio demographic data, medical history and clinical data were collected by means of a specific data form. Lifetime prevalence of ED according to DSM-5 criteria was assessed by means of the Module H modified, according to DSM-5 criteria, of the Structured Clinical Interview for DSM-IV Axis I Disorder (SCID-I, Research Version, Non-Patient Edition) (63, 64). Structured Clinical Interviews (Module H modified) were conducted by residents in psychiatry trained in the use of the instrument by a senior psychiatrist (FP). The inter-rater reliability for module H, assessed using Cohen’s K before starting the study, remained substantial (K > 0.80) among the 8 clinicians who performed the assessments. The rationale for performing diagnosis with DSM-5 criteria was related to the increased ability of DSM-5 criteria to identify the presence of an ED and to formulate a “specified ED diagnosis” (65). The following questionnaires were used for the purposes of this study: Beck Depression Inventory–IA version (BDI-IA) (66) and Diabetes Eating Problems Survey—Revised (DEPS-R) (67).

The present study was aimed at investigating a sample of 172 outpatients affected by T1DM (age: 17–55 years; mean duration of illness: 17.69 years) attending a diabetology clinic to assess the following: the current prevalence and severity of depression and DEB, lifetime prevalence of ED according to DSM-5 criteria, association between DEB and depression and impact of DEB and depression on diabetes control.

Overall, most of the sample studied was characterized by poor glycemic control: indeed, at the last assessment, 69.8% of participants displayed HbA₁c levels (parameter measuring blood sugar levels over the previous 2–3 months) exceeding 7%, indicative of poor glycemic control. The presence of diabetic complications (neurological, ocular, or renal) was reported in 21.5% of cases, whilst poor or inadequate treatment compliance was reported, based on clinical observations, in 41.3% of patients. High clinically relevant levels of depression were detected in 35.5% of the sample (26.7% of males and 44.2% of females). This finding is in line with those reported by other authors who highlighted how depression is a singularly common psychopathological dimension amongst patients with T1DM, implicating a 2.4 to 3.8-fold higher risk compared to a non-diabetic population (75). Likewise, the finding of higher rates of depression amongst females adheres perfectly to the data in literature reporting overall a 2-fold increase in risk of depression in the female gender compared to males, both in the general population (1) and in the population of diabetic patients (14, 15). In the most recent meta-analysis Farooqi et al. (76) reported a prevalence of depression of 22% in T1DM patients which is below the prevalence in our sample. This could be explained by the fact that most of the studies to date were conducted on samples of adolescents and youth while depressive disorders tend to occur more often later in life (58). Moreover, our study was conducted in a specialist care setting which was associated with higher prevalence rates in the aforementioned meta-analysis (76). Finally, the use of self-report questionnaires to assess depression in patients with diabetes have been found to be associated with higher rates compared to standardized diagnostic interviews (15), possibly underlying an overestimation of the depressive dimension due to the difficulty of distinguishing between a true depressive state and a condition of distress linked to diabetes management. To this regard, Fisher et al. raised the issue of the complex interpretation of data deriving from self-administered questionnaires to assess for depression, comparing the data deriving from these questionnaires with findings obtained by means of the Mood Disorders Module of Structured Clinical Interview, considered the diagnostic gold standard, thus identifying a high rate of false positives (20). It should however be considered that issues relating to reliability and validity were also raised about the use of structured interviews (77). Wisting et al. (40) who conducted a study on a comparable sample of T1DM adults, found a much lower prevalence of depression (6.2%) using the Hospital Anxiety and Depression Scale (HADS) (78). Even considering only moderate-to-severe cases, therefore using a cut-off of 19 in the BDI-IA [which is more conservative than the cut-off of 16 identified by Lustman et al. (79) in their validation study on diabetic outpatients], the prevalence of depression in our sample is still more than twice (14.5%) the prevalence found by Wisting et al. We think this finding is of interest, especially because both studies were conducted in a specialist care setting and both samples showed comparable rates of DEB. Part of this discrepancy is probably due to the different self-report questionnaire used. HADS was consistently more specific and less sensitive compared to BDI-IA in studies on patients with Human Immunodeficiency Virus (80) and Chronic Obstructive Pulmonary Disease (81). Although Wisting et al., did not report socio-demographic data, we can also hypothesize that different socio-demographic characteristics might have driven the higher prevalence of depression in our sample (i.e., less education).

According to DSM-5 criteria, the overall lifetime prevalence of ED was 20.9%, 7% in males and 34.9% in females. A total of 19.2% of the sample, 10.5% of males and 27.9% of females, achieved or exceeded a score of ≥20 at DEPS-R, a cut-off used to identify a risk of ED amongst diabetic patients. On examining prevalence data of DEB assessed by means of DEPS-R in greater detail, 56.4% of the sample had a history of “not taking enough insulin when overeating”; 6.4% (n = 11) of “skipping insulin dose after overeating”; 14.5% replied in the affirmative to “feeling fat when taking the whole prescribed insulin dose.” These findings are in line with literature data relating to the prevalence of ED and DEB in diabetic patients vs. the general population (38, 42, 45, 46, 49). Wisting et al. (40) found a comparable prevalence of DEB in their T1DM adult outpatients’ sample (20.3%), using the same assessment instrument. Compared to males, the female members in our sample displayed a significantly higher prevalence of lifetime ED according to DSM-5 criteria and higher scores at DEPS-R, indicative of an increased severity of DEB. Data present in literature support the finding of a generalized higher prevalence and an increased risk of ED in female adolescents and adults, although a relatively high prevalence is also observed amongst males (82, 83). Indeed, the finding in our study of higher scores in the female gender at DEPS-R is in line with data reported in the context of validation studies conducted on the rating scale (67, 69, 71, 72).

In the present study DEB was also significantly associated with variables related to poor metabolic control: (i) we found a positive signification correlation with HbA₁C (although with a weak effect size); (ii) we found higher DEPS-R scores in patients with poor or less than sufficient treatment adherence compared to those with sufficient, good or excellent treatment adherence; (iii) we found higher DEPS-R scores in subjects with diabetes complications compared to those without. Eisenberg Colman et al., reported an association between DEB and higher HbA₁C and worse treatment adherence in a sample of adolescents with T1DM (84). Our study replicates this finding in a sample of adult outpatients, in line with the study of Wisting et al. (40). Contrary to our hypothesis, no difference in DEB was found when comparing patients on CSIT with patients on MDI regimen. Prior research has consistently shown an improvement in psychological wellbeing with CSIT (85). In a recent study, adolescents with T1DM on CSIT scored significantly lower on the Binge Eating Scale (BES) compared to those on basal-bolus regimen (86). In our sample, using an instrument specifically constructed to assess DEB in T1DM patients, this association was not replicated. We can hypothesize that BES (87), a scale designed to assess binge eating in obese patients, has only limited validity to assess DEB in T1DM patients as most of insulin-related disordered behavior is not included. Finally, BMI was significantly positively correlated with DEPS score. In other studies, higher BMI indexes were found to underlie an increased drive to lose weight, resulting in dieting, negative affect, and disordered eating (38, 39, 88). BMI should therefore be deemed a potential risk factor for disordered eating in patients affected by T1DM (88).

As hypothesized prevalence rates of both lifetime DSM-5 ED and current DEB were significantly higher in depressed patients compared to non-depressed (respectively 34.4% vs. 13.9% and 36.1% vs. 9.9%). Of note no patient was ever diagnosed or treated for ED prior to study inclusion. This finding supports the importance of screening for ED and DEB in T1DM patients with depressive symptoms, as other authors reported a worse course of the disease and poor response to treatment of the depressive disorder in case of comorbidity (54). Future research should focus on the longitudinal course and therapeutic targeted interventions for depression and ED comorbidity in T1DM patients.

Contrary to our hypothesis, we did not detect any significant association between the presence of depression and variables related to poor glycemic control. Specifically, no association was revealed with HbA₁c levels. There are conflicting data in literature regarding the link between indicators of glycemic compensation and depression, with some studies that highlight a poorer compensation in depressed subjects (19) and others that do not (40, 57). It is worthy of note that in comparison with other studies in which indicators of metabolic compensation and levels of depression were evaluated contextually (19), in the present study HbA₁c levels related to the last control undertaken over the previous 3 months, whilst the reference time frame for assessment of depression using the BDI-IA scale was the previous 2 weeks. The findings obtained in our study agree with those reported by Fisher et al. who raised the issue of the methodological limitation of a non-contextual evaluation to justify the lack of association between depression and Hb1Ac levels (20).

Multivariate analysis revealed higher scores at DEPS-R increased the odds of depression at BDI independently of metabolic covariates (HbA₁c, CSIT, duration of diabetes) and confounders (gender, age, education, employment, thyroid disorder, BMI, and physical activity). In particular, there was an increase in the odds of depression by 9% for every increase of DEPS-R by each one point. Lower education was also independently associated with the outcome. This finding highlights the independent association between depression and DEB in patients with T1DM, bearing in mind, however, the cross-sectional design of the study. Several authors had previously reported a link between levels of depression and DEB (38, 39). In the context of a study conducted on a sample of female adolescents with T1DM, Colton et al. highlighted how subjects manifesting a more significant depressive symptomatology also displayed a higher incidence of DEB (39). Wisting et al., found a medium correlation between DEPS-R and HADS Depression in their T1DM adults sample (40). In our study, we extend these findings by showing with multivariate analysis an independent association between DEB and depression in our sample. Interestingly, Bächle et al. (89) conducted a study on T1DM patients aged 18–21 years old. In their sample, a positive SCOFF questionnaire (a screening questionnaire for ED) increased depression severity independently of several confounders (age, diabetes duration, BMI, socioeconomic status, family structure). Our results strengthen their findings, given that in our study: (i) DEB was assessed with a psychodiagnostics questionnaire validated on diabetic outpatients; (ii) we included different covariates in multivariate analysis, in particular those related to glycemic control; (iii) we recruited adult subjects with a mean age of 36.76 years old. Finally, the finding observed in the present study of an association between lower education and levels of depression is in line with data reported in literature, thus confirming this association (90).

The present study should, however, be interpreted considering a series of methodological limitations. Firstly, the cross-sectional design of the study did not allow the causal relationship, and therefore the direction of relationship between the factors investigated, to be determined. Secondly, the limitation of assessing depression using a self-assessment tool which, as highlighted previously, particularly in populations affected by a chronic medical condition such as diabetes, may not be capable of discriminating between a depressive state and a condition of distress correlated with the underlying illness. Thirdly, the non-contextual assessment of the depressive dimension and indicators of metabolic compensation. Fourthly, the absence of a control group. Lastly, the variables included in the final model explained only 28.4% of the variance in depression in our sample. Future research should include other variables (i.e., family history of mood disorders, personality traits, life events, childhood trauma) to find other significant predictors of depression in T1DM adults outpatients. Despite the aforementioned limitations, to our knowledge, this is the first study to examine the independent association between depression and DEB in T1DM adult outpatients, taking into account a numerous set of covariates and using a validated psychodiagnostic tool tailored to the diabetic population to assess DEB.

This study highlights a statistically significant and independent association between DEB and depression in a population of adult patients affected by T1DM. Prevalence of lifetime ED and current DEB and depression were elevated in our sample. Bearing in mind the diagnostic, clinical, and therapeutic implications of comorbidity between diabetes and depression and diabetes and ED, and the singularity of the forms of presentation of depressive and ED in diabetic subjects, routine analysis of the depressive dimension and ED using purpose-developed screening tools of proven validity should be deemed essential in this patient population. Furthermore, the complex management of comorbidity between diabetes, depression and ED implies a need for timely and efficient structured multidisciplinary intervention programs aimed at managing the physical and mental health of patients and minimizing the risk of short- and long-term complications linked to the comorbid conditions. The definition of specific therapeutic interventions for management of DEB could have positive effects on the control of diabetes and on the levels of depression in this population of patients. In the light of the previously acknowledged methodological limitations, the significance of the findings reported in this study suggest the need for further studies to be carried out on a larger patient population using a longitudinal study design and accurate method of evaluation aimed at yielding a wider and more reliable investigation of the complex relationship between diabetes, depression, ED, and DEB.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethical review and approval was not required for this study on human participants in accordance with the local legislation and institutional requirements. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

FP planned the study and drafted and revised the manuscript. FS drafted the result section and contributed to analyzing and interpreting the findings. VD and ED contributed to the planning of the study and collected data. FF, MG, RC, MT, EC, and EL collected somatic data. EN and PG contributed to analyzing and interpreting the findings. AL, LL, MM, PP, GS, and AC reviewed the manuscript and contributed to the discussion. BC and FV contributed to the planning of the study, reviewed the manuscript, and contributed to the discussion. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.