INTRODUCTION
Major depressive disorder (MDD) is one of the most burdensome diseases worldwide. For treating MDD, an antidepressant is the first choice of pharmacological intervention, but STAR*D research showed that only 27% reached remission in the first step of antidepressant medication [1]. Considering the social and personal cost of MDD, clinicians need to predict non-responders of antidepressants.
Childhood abuse history is one of the well-known risk factors of MDD and is also known as a predictor of non-responders to antidepressants. Childhood abuse is positively associated with early onset and recurrence of MDD [2], and MDD patients with a childhood abuse history showed poorer antidepressant treatment response than patients without a childhood abuse history [3]. Dealing with childhood abuse is inevitable in MDD treatment, due to its relatively high prevalence in MDD patients and its effect on treatment outcome [3].
Along with childhood abuse, suicidality is also an essential factor in MDD treatment. Even though much research on suicidality considers suicidality as an outcome of MDD treatment, suicidality can be a predictor of treatment response. In a naturalistic cohort study, MDD patients with severe suicidal ideation revealed significantly less remission after 6 weeks of antidepressant treatment [4]. In another 6-week citalopram treatment research, depression severity was significantly lower in suicidal patients than in non-suicidal patients, and suicidality was the only significant predictor of poor response, after controlling covariates including baseline depression severity [5].
Other research, however, suggests that the link between suicidality and depression treatment response is not that simple. For example, there was no significant difference in antidepressant treatment response between MDD patients with and without suicidal attempt history in 12 weeks of treatment trial [6]. In elderly patients, the suicidal ideation group and non-suicidal ideation group showed almost identical remission rates [7]. A recent scoping review reported that suicidal ideation was not a significant predictor in multivariate analysis in one study, and non-suicidal self-injury, which is one of the suicidal behaviors, did not significantly predict treatment response in another one [8].
It is necessary to elaborate on the model precisely predicting antidepressant response. Though childhood abuse and suicidality are both essential and related to MDD treatment, there was little research examining the interaction of childhood abuse and suicidality to investigate antidepressant response. In brain imaging research, the interaction effects of childhood abuse and suicidality in MDD were explored [9]. The results showed that the gray matter volume of the left caudate nucleus, which is related to reward-seeking behaviors, was significantly different by the participants’ childhood abuse history and level of suicidal ideas. It suggests effects of childhood abuse on MDD may be differentiated by the level of suicidality, but its effect on antidepressant response was not explored in the research. Therefore, we aimed to examine whether the interaction of childhood abuse history and suicidality is associated with antidepressant response in MDD patients.
METHODS
Study outline and design
This study was performed as a part of the MAKE Biomarker discovery for Enhancing anTidepressant Treatment Effect and Response (MAKE BETTER) project. Detailed information can be found in the published design paper [10] and Clinical Research Information Service (cris.nih.go.kr; identifier KCT0001332). Participants were registered regardless of depression subtype or physical comorbidity to reflect real-world treatment circumstances. Treatment interventions were also carried out in a naturalistic fashion. The clinician’s opinion and patient’s preference were reflected in determining the type, dose, and regimen of antidepressants and other medications, but it was guided by pre-planned measurements and time points. Assessments were scheduled at baseline, at 1, and 2 weeks, and then every 3 weeks in the acute treatment phase (3, 6, 9, and 12 weeks), and at every 3 months in the continuation treatment phase (6, 9, and 12 months). Clinical research coordinators who were blinded to treatment modalities obtained all data on sociodemographic and clinical characteristics at baseline, and treatment-related variables at follow-ups using a structured clinical report form (CRF). These staff were trained in CRF implementation and data collection methods by the research psychiatrists. After the data acquisition, participants’ data were registered on the website of the MAKE BETTER study (http://icreat.nih.go.kr/icreat/webapps/com/hismainweb/jsp/cdc_n2.live) within 3 days, and monitored by data management center staff. The Chonnam National University Hospital Institutional Review Board approved this study (approval number CNUH 2012-014).
Participants
Outpatients with depressive disorders who visited the psychiatric department of Chonnam National University Hospital were recruited for this research. 1,262 individuals had consecutively signed up for the study from March 2012 to April 2017. All the participants were taking newly initiated antidepressant treatment. They were enrolled regardless of whether their depressive episodes were first-onset or recurrent. Inclusion criteria were as follows: age >7 years; diagnosis of MDD, dysthymic disorder, or depressive disorder not otherwise specified (NOS) by the Mini-International Neuropsychiatric Interview (MINI) [11], a structured diagnostic psychiatric interview based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria [12]; Hamilton Depression Rating Scale (HAMD) [13] score ≥14; ability to complete questionnaires, to understand the objective of the study, and to sign the informed consent form. Exclusion criteria were as follows: unstable and uncontrolled medical condition; unable to complete the psychiatric assessment or comply with the medication regimen due to severe physical illness; current or lifetime DSM-IV [12] diagnosis of bipolar disorder, schizophrenia, psychotic disorder NOS, or other psychotic disorder; history of organic psychosis, epilepsy, or seizure disorder; any psychiatric diagnosis other than depressive disorder (e.g., alcohol/drug dependence); electroconvulsive therapy received for the current depressive episode; and pregnant or breastfeeding. All participants received study information and written informed consent was obtained before participation. For those aged <16, informed consent was obtained from their parents or legal guardians.
Assessment
Childhood abuse
Childhood abuse was retrospectively assessed by a semi-structured interview developed for the Netherlands Mental Health Survey and Incidence Study (NEMESIS-1) [14] at baseline. This childhood trauma interview covers three areas of interpersonal traumatic experience before the age of 16; emotional or psychological, physical, and sexual abuse. Emotional abuse was evaluated by asking, “Were you emotionally or psychologically abused, meaning being yelled at, falsely punished, subordinated to your siblings or being blackmailed?”; physical abuse was evaluated by asking, “Were you physically abused, meaning being hit, kicked, beaten up or other types of physical abuse?”; and sexual abuse was evaluated by asking “Were you being abused sexually, meaning being touched or having to touch someone sexually or pressured into sexual contact against your will?”. In our statistical analysis, a broad definition of childhood abuse was employed; participants experiencing at least one of the three types of childhood abuse were classified within the ‘present’ group.
Suicidal ideation
The level of suicidal ideation was rated at baseline with the suicidality item of the Brief Psychiatric Rating Scale [15]. Participants were asked ‘Have you felt that life wasn’t worth living? Have you thought about harming or killing yourself? Have you felt tired of living or as though you would be better off dead? Have you ever felt like ending it all?’. When they reported suicidal thoughts, interviewers questioned the frequency of suicidal ideation and the specificity of the suicidal plan. Based on the information, interviewers rated the individual’s severity of suicidal ideas on a 7-point Likert scale ranging from 1 (“not present”) to 7 (“extremely severe”). For data analysis, the scores 1 (not present) to 3 (mild) were categorized as the “mild or less” group, and the scores 4 (moderate) to 7 (extremely severe) as the “severe or more” group.
Socio-demographic characteristics
Socio-demographic data was collected at baseline. It includes age, gender, year of formal education, marital status (currently married or not), living arrangement (living alone or not), religion (religious observances or not), occupation (currently employed or not), and monthly income (above or below 2,000 USD).
Clinical characteristics
The diagnosis of MDD was established by MINI [11], which is based on DSM-IV [12] criteria, and specifiers such as melancholic and atypical features were also determined by DSM-IV [12] MDD criteria. Information of age at the first onset of depression, duration of illness, number of depressive episodes, recurrent depressive episodes, duration of the present episode, family history of depression, and number of concurrent physical disorders were additionally obtained.
Assessment scales
Trained research staff administered HAMD-17 [13] to evaluate the severity of depressive symptoms at every assessment point including baseline. HAMD-17 [13] score ranges from 0 to 52, and a higher score indicates higher severity. The subjective symptoms of depression and anxiety were assessed by the Hospital Anxiety Depression Scale (HADS) [16]; quality of life by EuroQol-5D [17]; daily functioning level by the Social and Occupational Functioning Assessment Scale (SOFAS) [18]; stressful life events by the Life Experiences Survey [19]; level of stress perception by the Perceived Stress Scale (PSS) [20]; resilience by the Connor-Davidson Resilience Scale [21]; perceived social support by the Multidimensional Scale of Perceived Social Support [22]; hazardous alcohol usage by Korean version of Alcohol Use Disorder Identification Test (AUDIT-K) [23].
Pharmacologic treatment and outcomes
Before treatment initiation, participants underwent a comprehensive assessment of their clinical symptoms, physical comorbidities, medical histories, and past treatments. Initially, participants were prescribed antidepressants for three weeks based on established treatment guidelines [24] and individual characteristics. Treatment effectiveness and tolerability were assessed before proceeding to subsequent measurement-based adjustments. If improvement was inadequate or adverse events were intolerable, participants could choose to continue with the current treatment or switch to alternative strategies, including switching antidepressants, adding other medications, or combining multiple antidepressants. The main outcome was the participants’ remission status at 12 months, and the remission was defined as HAMD-17 [13] score ≤7. The detailed treatment methodology of this study has been previously published [10,25].
Statistical analyses
For descriptive statistics, baseline socio-demographic and clinical variables were compared by the childhood abuse history and the level of suicidal ideation, resulting in four distinct groups: those without childhood abuse history and low level of suicidal ideation, those without childhood abuse history and high level of suicidal ideation, those with childhood abuse history and low level of suicidal ideation, and those with childhood abuse history and high level of suicidal ideation. Analysis of variance or χ2 tests were conducted as appropriate, with post-hoc Scheffe’s tests or individual pairwise comparisons applying Bonferroni correction, respectively. Among these variables, statistically or theoretically significant variables were selected as covariates for the main analysis; age, gender, unemployment status, living arrangement, monthly income, recurrent depressive episodes, and baseline score of HAMD-17 [13] and AUDIT-K [23]. For primary analysis, logistic regression analysis was conducted to explore individual association of childhood abuse and suicidality with 12-month remission status with and without covariates. Next, multinomial logistic regressions were conducted to test the hypothesized interactive effects. The covariates were entered in the first step, childhood abuse and suicidal ideation were included in the next step, and at the last step, the interaction term of childhood abuse and suicidality variables was inserted. All statistical tests were two-tailed with a significance level of p<0.05, and SPSS version 21 (IBM Corp., Armonk, NY, USA) was used for analyses.
RESULTS
Recruitment and treatment flow
The patient recruitment flow chart is described in Figure 1. Of 1,262 patients evaluated at baseline, 1,015 (81.5%) were followed up at least once during the 12-month treatment period, and they comprised the sample analyzed for treatment outcome. Participants who were not included in this analysis significantly had more chance of unemployment and higher EuroQol-5D [17] scores at baseline. However, the prevalence of childhood abuse history and the level of suicidal ideation were not significantly different between the dropout and non-dropout groups.
Baseline characteristics
As Table 1 presents, participants were on average 56.76 (SD=15.12) years of age, 69.49% were female, 15.06% were living alone, 59.43% had less than $2,000 of monthly income, 29.16% were not employed, 43.66% had no religion. Clinically, 85.50% of baseline participants met the criteria of MDD, had depression for 4.89 years on average, and the mean duration of the episode at baseline was 7.41 days. At baseline, the mean score of the HADS [16] depression subscale was 13.59, and the anxiety subscale was 11.71, which are both over the ‘definite cases’ criteria [16]. The mean score of EuroQol-5D [17] was 8.86, SOFAS [18] was 55.90, life experiences survey [19] was 2.04, PSS [20] was 27.04, Connor-Davidson resilience scale [21] was 42.96, Multidimensional scale of perceived social support [22] was 39.14, and AUDIT-K [23] was 5.37. All sociodemographic and clinical variables were compared by childhood abuse history and the level of suicidal ideation. The results are presented in Table 1. The group differences were significant in most variables except gender, living arrangement, unemployed status, melancholic feature, and family history of depression. The individuals with a childhood abuse history had higher levels of suicidal ideation than the others, and in general, the group without a history of childhood abuse and with low levels of suicidal ideation (Group 1) exhibited a more favorable clinical presentation than the other groups (Group 2, 3, and 4). In the post-hoc analysis, individuals in Group 1 were significantly older than those in Groups 2, 3, and 4, and Group 1 had a significantly higher proportion of married individuals compared to Groups 3 and 4. Clinically, Group 1 had a significantly lower prevalence of MDD compared to Group 2, and had older ages at the first onset of depression than those in Groups 2, 3, and 4. The duration of illness and episode was also significantly shorter in Group 1 than in Group 4 and Group 3, respectively. In assessment scales, Group 1 generally demonstrated less pathologic results on HAMD-17 [13], HADS [16], PSS [20], and AUDIT-K [23] compared to the other groups.
Individual effects of childhood abuse on non-remission status at 12 months
Logistic regression was performed to test the single effect of childhood abuse on the non-remission of depression at 12 months. Without covariate variables, childhood abuse history was significantly associated with a non-remission state at 12 months (odds ratio [OR]=1.767; 95% confidence interval [CI], 1.230–2.539). After covariate variables were entered into the model, the effect of childhood abuse on the non-remission of depression at 12 months was also significant (OR=1.626; 95% CI, 1.097–2.410).
Individual effects of suicidality on non-remission status at 12 months
To evaluate the effect of suicidal ideation on the non-remission of depression at 12 months, logistic regression was executed and the results are presented in Table 2. Without control, suicidal ideation significantly predicted non-remission of depression at 12 months (OR=1.433; 95% CI, 1.078–1.903). Even after covariate variables are put into the model, the effect of suicidal ideation on non-remission of depression at 12 months remained significant (OR=1.382; 95% CI, 1.019–1.873).
Interactive effects of childhood abuse and suicidality on non-remission status
To explore the interaction effects of childhood abuse and suicidality, logistic regression was executed and the results are presented in Figure 2. For childhood abuse and suicidal ideation, their interaction term was significantly related to the remission of depression at 12 months (Wald=6.19, p=0.013). To be specific, within the low suicidal ideation group, individuals with a childhood abuse history were more likely to experience non-remission at 12 months of antidepressant treatment (OR=2.55; 95% CI, 1.55–4.21), but in the high suicidal ideation group, childhood abuse history was not associated with non-remission of depression at 12-month (OR=0.80; 95% CI, 0.41–1.56).
DISCUSSION
In this prospective antidepressant trial, childhood abuse and suicidal ideation respectively showed a significant relationship with non-remission status at 12 months, with and without control variables. In moderation effect analysis, the interaction of childhood abuse and suicidal ideation significantly predicted antidepressant non-response at the 12-month time point.
As expected, MDD patients with a childhood abuse history were more likely to be non-responsive to 12 months of antidepressant treatment. Lots of studies have suggested that exposure to an abusive environment in early life stages can interfere with antidepressant treatment efficacy even into adulthood. Childhood abuse leaves biological and psychological scars. Depressive adults with childhood abuse tend to have increased levels of biological allostatic load [26], which is reported to predict poor antidepressant response [27]. Psychologically, experiencing childhood abuse lowers the threshold of stress perception in adulthood depression [28]. Antidepressants cannot mitigate environmental stress factors, so the one sensitized to stress might have more chances of getting depressed than others without stress sensitivity in the medical treatment process. Since childhood abuse impedes the healthy development of both body and mind, the resulting biological and psychological factors in abused patients can reduce the effectiveness of antidepressants.
Along with prior research, higher suicidal ideation at baseline was significantly related to a non-remission state at 12 months [29]. The depression severity tends to be higher in suicidal patients, so their high level of depressive pathology might require longer or intensive treatment than non-suicidal depression. Additionally, some of the challenging characteristics observed in suicidal patients may be difficult to regulate through pharmacological treatment. For example, perceived burdensomeness and thwarted belongingness are known as risk factors for high suicidality [30] but they are intricately linked to the patient’s social circumstances and cognitive patterns, posing challenges for regulation through pharmacological treatment only.
Suicidal ideation also had a significant moderating effect on the relationship between childhood abuse history and antidepressant response. Specifically, in the low suicidal ideation group, depressive patients with a childhood abuse history were more likely to be in a non-remission state after 12 months of medication. It suggests that considering only childhood abuse history or suicidal ideation is not enough to predict antidepressant response. Clinicians must meticulously combine the information to predict treatment responses. Particularly, patients reporting a history of childhood abuse but no suicidal ideation are expected to exhibit poorer treatment responses compared to other groups, necessitating consideration for additional interventions even though they don’t report suicidal ideation. The clinicians may consider proactive interventions for those patients, such as quickly changing medication strategy (i.e., combination, augmentation) [25], or using antidepressants with higher serotonin transporter affinity [31] to improve antidepressant response. Combining evidence-based cognitive therapy can also be another option to improve antidepressant response [32] for those who are predicted to be less effective in antidepressants. By leveraging a range of patient characteristics to forecast treatment outcomes, clinicians can optimize the allocation of time and resources in facilitating recovery from depression, thereby enhancing their role as navigators in the patient’s therapeutic journey.
Despite the strengths of this research, there are several weaknesses. First, this study was designed to capture naturalistic clinic settings, so pharmacologic choices were minimally restricted by research protocol, and patients’ preferences and psychiatrists’ instruction were mainly engaged. It might have resulted in inter-therapist variability and the variability might affect results. However, the clinicians made decisions based on verified protocols such as CANMAT clinical guidelines [33], therefore it was not completely random. Moreover, natural design makes research outcomes more applicable to the real world. Second, childhood abuse types were not specified in this research. Each different types of traumatic experience can result in different antidepressant responses, interpersonal trauma, for example, has more adverse effects than non-interpersonal trauma [3]. However, covering the difference in abuse types was beyond the scope of this research, so the dynamic of abuse types, suicidality, and antidepressant response can be a future subject. Lastly, the data of potentially eligible populations (e.g., the number of depressive patients in the clinic) was not available. It may lead to the concern that this cohort data would not be a good representation of the population. Though all research staff did our best to enroll all eligible participants, it needs to be taken in mind when you interpret the results. Therefore, future research may need to verify the underlying biological mechanisms of childhood abuse affecting antidepressant treatment outcomes, with more detailed information on the eligible cohort population and strict research structure.
Despite some limitations, this research has several strengths. The majority of research on antidepressant response was carried out for acute treatment periods, such as 6 weeks or 12 weeks. In this research, participants were followed for a much longer period, 12 months, to verify the long-term response of antidepressants. Therefore, the long-term outcome of antidepressant treatment could be explored in contrast with prior research that has focused on the acute phase of treatment. The large sample size of over 1,200 is also one of the strong points of this study.
