All over the world about 300 million people suffer from major depressive disorder (MDD) (1). The World Health Organization (WHO) has identified MDD as the primary cause of disability burden, leading to reduced productivity, heightened healthcare expenses, and, most significantly, hindrances in achieving a fulfilling and enriching life (2). The advent of antidepressant medications has brought about a transformative shift in the treatment of major depression. Unfortunately, however, about 60% of patients do not show an adequate response to first line pharmacological treatments and 30% respond poorly to different trials with various antidepressants (3). The extreme variability in antidepressant treatment response is likely due to neurobiological and environmental factors (4).

Treatment-resistant depression (TRD) is commonly defined by the lack of positive response to at least two types of antidepressant medication, administered at the correct dosage and for a suitable duration (5). However, experts still do not agree on the definition of appropriate dose and appropriate treatment duration (6) and a consensus definition of TRD has not yet been reached. There is also little consensus about the best tools to diagnose TRD and measure its outcomes. These limitations hampered the possibility to compare and summarize study results, thus limiting the possibility to define clinical guidelines (7).

Several studies have reported that TRD could be associated with increased mortality (8, 9), although the sample sizes were small and follow-up times have been relatively short. A Swedish population-based study considering 118,774 individuals diagnosed with depression reported an overall mortality 1.35 times higher among patients with TRD compared to individuals with MDD (10). The increased rate was mainly attributed to external causes, including suicide and accidents.

A systematic review of suicidality in TRD found an overall incidence of completed suicides of 0.47 per 100 patients/year and of attempted suicides of 4.66 per 100 patients/year (95% CI: 3.53-6.23) (11). These are respectively twice and ten times greater than those found in non-resistant patients: 0.22 completed and 0.43 attempted suicides per 100 patients/year (12). In general, several studies pointed out that 30% of patients with TRD had one or more suicide attempts (13). Another recent study (14) dealing with suicidality in the context of major depression found that individuals with TRD had higher suicide rates compared to those who were diagnosed with MDD. Previous studies also highlighted that suicide related mortality in TRD was higher than in MDD even when depressive symptoms were classified as “mild” (15, 16). Furthermore, most authors underlined that the type of suicide attempt, which can be classified in impulsive, frequent or well-planned (17), is almost never reported. This hampers the study of underlying moderators of the high suicide risk observed in TRD. For example, suicide attempts classified as impulsive may indicate a decreased impulse control in TRD patients or an increase of impulsiveness that might be responsive to a different treatment. Another possible interpretation is that TRD patients could be aware of the limited therapeutic options for future improvement, which could lead to a higher proportion of well-planned suicide attempts, compared to non-resistant patients.

Patients with TRD often have comorbid personality disorders and this comorbidity could represent the underlying moderator of frequent suicide attempts in this subset of TRD patients. In a recent review article (18) a significant impact to development of TRD is determined by the co-occurring diagnosis of a personality disorder, as personality disorders in general respond poorly to pharmacological treatment. In particular, borderline personality disorder is characterised by high levels of impulsivity, unstable self-image, feeling of emptiness and extreme mood instability. It has been reported that depressive disorder with comorbid borderline personality disorder (BPD) shows greater treatment resistance and worse functional impairment (19).

Based on these premises, our study aims to identify which socio-demographic and psychopathological data are associated with suicidal attempts in a sample of TRD outpatients.

Clinical characteristics of the sample are presented in Table 1 . 77.8% of the sample (N=49) were females, and the mean age was 49.2 years (SD 15.9). 47.6% of the sample (N=30) were in treatment with Esketamine. In our sample, 21 (33.3%) patients attempted suicide. 32 patients (50.8%) had a psychiatric comorbidity. Of these, 23 patients (36.5% of the sample) had a personality disorder. Detailed treatment regimens of each patient included in the study are described in Supplementary Table 1 .

Univariate analyses are described in Table 2 . Substance use, personality disorder, psychiatric comorbidity and higher HAM-D scores were significantly more frequent among suicide attempters. All other variables were not significantly different between the two groups.

We then constructed a regression model using suicidal attempts as the dependent variable and all other variables as independent predictors ( Table 3 ). The resulting model explained 34.8% of the variance (R²N=0.348, AIC=70.0, p<0.001). The independent predictors were the presence of a psychiatric comorbidity, substance use and HAM-D scores ( Table 3 ). We also performed a regression model with the same variables and gender as a controlling factor ( Table 4 ). In this model, the explained variance was also slightly increased and substance use and HAM-D were the only predictive variables for suicidal attempts (R²N=0.354, AIC 71.7 p<0.001).

Our research focused on identifying the risk factors associated with suicide attempts among individuals diagnosed with Treatment-Resistant Depression (TRD) who were receiving care at an outpatient mental health facility in Italy. We observed that substance use was associated with a higher rate of suicide attempts. Additionally, psychiatric comorbidities, namely borderline personality disorder, were also associated with a higher rate of suicide attempts.

Notably, about 30% of individuals with TRD make suicide attempts at least once in their lifetime, as reported by several studies, twice as much as in non-resistant depression (31, 32). This datum is in line with the percentage registered in our sample (33%, n=21). This underscores the need of vigilant clinical monitoring of TRD patients, considering that a close psychiatric follow-up following a suicide attempt has demonstrated an antisuicidal protective effect (33).

In our sample, 77.8% were female patients. This prevalence is higher than in previous studies, but in line with generally higher prevalence of TRD in females (34), as well as in MDD in general (35, 36). For example, Herlein and colleagues identified a 62.3% prevalence of females in their sample, while another study recorded a female prevalence rate of 52.6% (37). In our sample, patients who have attempted suicide are on average younger (44.2 vs 51.6 years old), as previous research has already outlined (38).

The association between substance use and suicidal attempts that we observed is in line with previous research: one nested case-control study (39) based on a Swedish nation-wide register of TRD patients observed a correlation of substance use disorders, personality disorders, and anxiety disorders with attempted suicides. Our hypothesis is that substance use disorder may be a proxy of impulsiveness which could result in suicide attempts. Our findings emphasise the importance of implementing primary and secondary prevention strategies regarding psychoactive substance use, as well as to enhance cooperation between general psychiatry and addiction mental health services (40).

In our study, psychiatric comorbidities have been associated with suicide attempts. In a recent study, patients with TRD, compared to patients with MDD have a higher rate of psychiatric comorbidities, a longer duration of depressive episodes and three times the number of inpatient bed-days (41). These findings stress the importance of early identification of patients with MDD and high risk of TRD, in order to target health care efforts (42). Taking into account the HAM-D rating scale, we found that higher scores (i.e. worse depressive symptomatology) were associated with suicide attempts. This datum has been already highlighted by previous studies (43, 44). In the same way, it has been demonstrated that patients with TRD have on average higher HAM-D scores than patients with major depression, as well as a longer duration of illness (45).

Furthermore, we found an association between personality disorders and suicidal attempts. This finding has been already outlined by Reutfors et al. (39), who found as independent risk factors for suicidal attempts a history of suicide attempts, substance abuse, personality disorders, and somatic comorbidity. This finding implies a careful assessment of personality disorders through evidence-based instruments, as well as an adequate treatment with a targeted psychotherapy (46, 47). As far as our sample is concerned, the most represented among personality disorders was borderline personality disorder (BPD, n=10, 15.9% of the total sample and 43.5% of personality disorders, respectively). Depressive disorder and borderline personality disorder are often comorbid. The high impulsivity and the poor mentalizing skills that are core feature of BPD often lead to self-injurious behaviours and suicide attempts. As well as TRD, borderline personality disorder responds poorly to conventional pharmacological treatments. However, we didn’t find a specifical association between BPD and suicidal attempts, since patients with BPD were nearly equally distributed in the two categories. This may also be related to the relatively small sample size of our study, and study with a bigger sample could better define this association.

The topic of suicidality among patients with a primary diagnosis of treatment-resistant depression necessitates meticulous investigation, given the heightened prevalence of suicidal thoughts, attempts, and completed suicides in individuals with TRD as opposed to those with Major Depressive Disorder (MDD) (48). TRD represents a clinical challenge in psychiatry, since it is associated with a loss of quality of life, a lower productivity, more hospitalizations and higher healthcare costs (49, 50).

Some pharmacological therapies have shown promising results in treating TRD. For example, esketamine, the levo enantiomer of ketamine, has been approved for the treatment of TRD and can be administered as a nasal spray. This drug has proven generally safe and well tolerated and has provided meaningful and rapid impact on reducing depressive symptoms and suicide ideation (51). A possible beneficial effect on suicidal behaviour in TRD patients has also been suggested for lithium. In a meta-analysis by Cipriani and colleagues (52) authors found that lithium helped reducing suicide risk in patients with mood disorders. It was hypothesised that it may exert its antisuicidal effects by reducing relapse of mood disorder and also by decreasing aggression and possibly impulsivity, which might be another mechanism mediating the antisuicidal effect. Regarding TRD, evidence about lithium’s antisuicidal effects is still poor.

As far as non-pharmacological treatments are concerned, various studies have focused on Repetitive Transcranial Magnetic Stimulation (rTMS) that is a non-invasive brain stimulation technique used to treat mood disorders, including TRD, but also other mental illnesses such as obsessive-compulsive disorder (OCD) and borderline personality disorder (BPD). A recent meta-analysis by Chen et al. (53) found that rTMS significantly reduced suicidal ideation and improved depressive symptoms. Focusing on suicidal ideation, this was reduced after rTMS in patients with major depressive disorder but not in those with TRD.

Our study has several limitations. First of all, the sample size was quite small, thus limiting the generalizability of the results. In this regard, future studies with larger sample sizes and a prospective design could provide a more precise insight into our findings. Moreover, our study did not collect other relevant parameters related to suicidal attempts, such as the number of hospitalizations and the modality of attempted suicide, which is known to be a relevant diagnostic and prognostic factor (54). Secondly, we did not use a specific appropriate assessment instrument for suicidal ideation (48, 55). These limitations are mainly due to the research design which was carried out at a public mental healthcare facility, in a “real world” setting. Therefore, assessments were not specifically designed to investigate specific psychopathological domains. A larger sample size and a more careful assessment of these features could lead to more accurate results. Lastly, we did not perform a psychopathological assessment at follow-up with the rating scales performed at baseline. Such finding would have allowed us to highlight symptomatic changes at follow-up and any correlations with suicidal behaviors. In future research, it could be interesting, as we found an association between HAM-D scores and suicide attempts, to analyze separately every single item of this scale with respect to suicidality in a more dimensional approach. Indeed, the National Institute of Mental Health Collaborative Depression study highlighted three group of symptoms (1, anhedonia, hopelessness; 2, anxiety, agitation, panic; 3, aggression, impulsivity) as more predictive of suicide than either diagnoses or syndrome (56).

Our study supports the importance of performing a careful clinical evaluation of patients with treatment-resistant depression and of raising awareness among clinicians to prevent pseudo-resistance and to identify factors associated with increased severity of symptoms. In this regard, it would be useful to insert standardised evaluation such as the HAM-D scale (which is overall not time-consuming) in clinical practice. Additionally, assessment of personality disorders as well as potential dysfunctional personality traits deserve a place in clinical practice (57). On the other hand, clinicians should strictly adhere to pharmacological guidelines, following the correct doses of each medication and the adequate duration of the treatment in order to identify TRD correctly and perform a differential diagnosis between TRD and refractory depression, that is a form of depressive disorder that has not shown adequate response to any treatment (58). For a more personalised and targeted therapy, pharmacogenomic tests hold great promise for future routine clinical practice but now they are mostly limited to specialised services (59). In conclusion, clinical practice should incorporate personalized medicine principles in order to choose more effective pharmacological and psychosocial therapeutic strategies.

The data analyzed in this study is subject to the following licenses/restrictions: The dataset includes privacy-related information. Requests to access these datasets should be directed to Filippo Besana; filippo.besana01@universitadipavia.it.

Ethical approval was not required for the study involving humans in accordance with the local legislation and institutional requirements. Written informed consent to participate in this study was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and the institutional requirements.

SC: Formal analysis, Writing – original draft, Writing – review & editing, Data curation, Software. FB: Formal analysis, Writing – original draft, Writing – review & editing, Conceptualization. GM: Data curation, Writing – original draft, Writing – review & editing. FM: Conceptualization, Data curation, Writing – original draft, Writing – review & editing. VA: Data curation, Writing – original draft, Writing – review & editing. PP: Supervision, Visualization, Writing – original draft, Writing – review & editing. NB: Formal analysis, Supervision, Writing – original draft, Writing – review & editing. MO: Conceptualization, Methodology, Supervision, Writing – original draft, Writing – review & editing.