Depression is a leading cause of disability. About one- third of patients achieve remission with their initial antidepressant within about 7 weeks, while other patients fail to achieve remission [1,2]. To improve treatment outcomes, it is important to identify biomarkers that can be used to predict remission according to the antidepressant administered. Easily accessible peripheral biomarkers for predicting the treatment response in depression include serum levels of inflammatory cytokines, brain-derived neurotrophic factor (BDNF), serotonin, and cortisol [3-6]. However, the results have been inconsistent due to heterogeneity in study designs and the antidepressant type, and duration of use [3,7].

We recently demonstrated the value of baseline high- sensitivity C-reactive protein (hsCRP), interleukin (IL)-1β, IL-6, and leptin levels for predicting remission at 12 weeks in response to stepwise pharmacotherapy [8]. The results provided a basis for identifying serum biomarkers for predicting remission according to antidepressant type in the current naturalistic prospective study.

This study used data from the MAKE Biomarker Discov-ery for Enhancing anTidepressant Treatment Effect and Response study [9]. This study was approved by the Chon-nam National University Hospital Institutional Review Board (CNUH 2012-014). Written informed consent was obtained from all participants. A detailed description of the study methods is provided in the Supplementary Methods (available online).

Participants and Treatment

Outpatients with depressive disorders, who were eligible for the study were treated by stepwise pharmaco-therapy. Three-week antidepressant monotherapy was administered, followed by alternating switch, combination, and/or augmentation strategies during the acute treat-ment phase (3−12 weeks; 3-week interval) for patients exhibiting insufficient improvement (a reduction of the Hamilton Depression Rating Scale [HAMD] [10] score < 30% from baseline) or intolerable side effects.

The main antidepressants were identified considering dosage, and duration, and classified as selective serotonin reuptake inhibitors (SSRIs; escitalopram, fluoxetine, paroxetine, sertraline), serotonin norepinephrine reuptake inhibitors (SNRIs; desvenlafaxine, duloxetine, venlafaxine), mirtazapine, and other antidepressants (bupropion, vortioxetine). Considering the frequency and mechanism of action, four antidepressant types were selected for analysis: escitalopram, other SSRIs, SNRIs, and mirtazapine. Some antidepressants were excluded due to the rarity of their use and/or heterogeneous mechanism of action.

Serum Biomarkers

Serum biomarkers in fasting morning blood samples were evaluated at baseline by Global Clinical Central Lab (Yongin, Korea). The analyses were conducted without knowledge of the patients’ status. Fourteen biomarkers related to the antidepressant response, as identified by our literature search/meta-review [11], were measured: immune: hsCRP, tumor necrosis factor-alpha (TNF-α), IL-1β, IL-6, IL-4, and IL-10; metabolic: leptin, total ghrelin and total cholesterol; neuroplastic: BDNF; neurotransmitter: serotonin; endocrine: cortisol; and nutritional: folate and homocysteine.

Outcomes

Remission was determined as a HAMD [10] score ≤ 7 at each assessment point (3, 6, 9, and 12 weeks). Remission at 12 weeks was defined as maintenance of remission up to the 12-week assessment point.

Covariates and Statistical Analyses

For adjusted analyses, covariates were selected from among baseline sociodemographic and clinical characteristics significantly associated with remission at 12 weeks (p < 0.05; t test or χ² test), and other previously identified potential covariates [12]; collinearity was considered during covariate selection. The optimal cut-offs for individual biomarkers used to predict remission at 12 weeks were estimated by area under the receiver operating curve analysis. Odds ratios and 95% confidence intervals for remission at 12 weeks were calculated according to antidepressant type by pair-wise logistic regression analyses (adjusted for covariates) using the optimal cutoffs for the individual biomarkers.

Among the 1,094 patients who agreed to participate and provide blood samples, 1,024 were assessed at least once during the 12-week acute treatment phase and subsequently included in the analyses.

Associations between individual serum biomarkers and remission status according to the antidepressant used are summarized in Table 1. At 12 weeks, 448 (41.3%) patients were taking escitalopram, 206 (18.9%) were taking other SSRIs, 103 (9.5%) were taking SNRIs, and 267 (24.6%) were taking mirtazapine. Achievement of remission was independently associated with hsCRP, IL-1β, IL-6, and leptin levels below the cut-offs in escitalopram patients; hsCRP, TNF-α, IL-1β, and leptin levels below the cut-offs in other SSRI patients; hsCRP, IL-4, and IL-10 levels below the cut-offs in SNRI patients; and hsCRP and leptin levels below the cut-offs, and serotonin and folate levels above the cut-offs, in mirtazapine patients.

We investigated the associations between serum markers and remission by antidepressant type in a naturalistic study reflecting real-world clinical practice. The hsCRP levels were associated with remission at 12 weeks for all antidepressant types. The levels of proinflammatory cytokines such as IL-1β, TNF-α, and IL-6 were linked to remission in association with the use of SSRIs, including escitalopram. The levels of anti-inflammatory cytokines such as IL-4 and IL-10 were correlated with remission in association with SNRIs. Leptin levels were associated with remission in association with antidepressants other than SNRIs. Additionally, remission with mirtazapine was related to serotonin and folate levels.

Most remarkably, we found that lower hsCRP levels were associated with remission at 12 weeks for all antidepressant types (hsCRP < 0.61 mg/dl). Consistent with our findings, lower baseline CRP levels (< 1.00 mg/L) predicted a better response to escitalopram than nortriptyline [13], and higher CRP levels predicted a poor response to fluoxetine and venlafaxine [14]. Lower levels of CRP (< 0.8 mg/L) predicted a better response to escitalopram [15]. Although no significant association has been reported [16,17], hsCRP may be a potential marker for predicting remission at 12 weeks in association with the use of antidepressants, particularly those that act on serotonin pathways such as SSRIs, SNRIs, and mirtazapine.

The levels of pro-inflammatory cytokine including IL- 1b, TNF-α, and IL-6 predicted acute-phase remission in association with SSRIs in this study. Previous studies reported that the baseline levels of several pro-inflammatory cytokines can predict an acute treatment response up to 12 weeks. Higher baseline IL-1β, IL-6, and TNF-α levels predicted poor responses to escitalopram and fluoxetine [18,19], although inconsistent findings have also been reported (i.e., associations of higher levels with a good response, or no association) [16,17,20]. As shown in previous meta-analyses, non-responders to depression treatments tend to have higher baseline levels of inflammatory markers such as CRP, TNF-α, IL-1β, and IL-6 [3,21]. Our findings provide additional evidence supporting the potential role of inflammation in the response to antidepressants, particularly SSRIs.

Lower levels of anti-inflammatory cytokines including IL-4 and IL-10 were correlated with remission at 12 weeks in association with SNRI use in this study. Although baseline IL-4 and IL-10 levels were not associated with the response to SNRIs at 6−8 weeks [22,23], a relationship between anti-inflammatory markers and remission in association with SNRIs is intuitive given the anti-inflammatory properties of these drugs, where the levels of Th2 cytokines such as IL-4 were reduced more than those of other cytokines in association with venlafaxine use [23]. Future studies on the value of IL-4 and IL-10 for predicting remission at 12 weeks in the context of SNRI use are needed to replicate our findings; larger samples representing more ethnic groups should be included.

In this study, lower leptin levels were associated with remission at 12 weeks in association with antidepressant use, excluding SNRIs. Equivale findings have been reported regarding the predictive value of peripheral leptin levels in terms of the antidepressant response; lower levels were associated with a poor response in a study that did not stratify by antidepressants [24], and no significant association was reported elsewhere [25]. Based on previous findings of differential effects of leptin levels on the treatment response according to physical comorbidities [26], the significant association between low leptin levels and remission at 12 weeks seen in this study may have been mediated by other covariates, such as physical comorbidities. Therefore, further studies are needed on the association between leptin levels and treatment outcomes by antidepressant type considering these mediating factors.

For mirtazapine, remission at 12 weeks was associated with higher serotonin and folate levels in this study. Few studies have evaluated the associations of serotonin and folate levels with remission at 12 weeks in the context of mirtazapine use. Moreover, conflicting findings have been reported regarding the association between blood serotonin levels and antidepressant treatment outcomes; higher levels were reportedly associated with a better response to SSRIs, including sertraline [27,28], while there was no such association in other studies [29,30]. Similar studies have linked low folate levels with a poor response to SSRIs and nortriptyline [31,32]. Studies on the associations of serotonin and folate levels with mirtazapine use are needed, because cytokine and BDNF levels have rarely been used as predictive markers of the response to this agent [33,34].

This study had several limitations. The naturalistic study design permitted the inclusion of patients with various comorbidities and type I error was increased due to multiple comparisons. Strengths of our study included the high follow-up rate and lack of attrition. Our findings suggest that baseline levels of nine serum markers may help clinicians determine the most appropriate antidepressant to achieve remission in the acute phase of depression. Our findings are expected to improve outcomes by facilitating tailored antidepressant treatment based on the patient’s biological markers, although studies with larger populations are needed for validation.

Jae-Min Kim declares research support in the last 5 years from Janssen and Lundbeck. Sung-Wan Kim declares research support in the last 5 years from Janssen, Boehringer Ingelheim, Allergan and Otsuka. All other authors report no biomedical financial interests or potential conflicts of interest.

Jae-Min Kim, and Il-Seon Shin designed the study. Hee-Ju Kang and Ju-Wan Kim construct and performed study methodolology. Ju-Yeon Lee, and Sung-Wan Kim contributed project administration. Ju-Yeon Lee, Sung- Wan Kim, and Il-Seon Shin contributed to validation. Jae-Min Kim, Hee-Ju Kang, Ju-Yeon Lee, and Sung-Wan Kim acquired data and curated data. Jae-Min Kim, Ju-Wan Kim, Wonsuk Choi, and Il-Seon Shin contributed to formal analysis. Hee-Ju Kang and Jae-Min Kim contributed to writing original draft. Jae-Min Kim, Hee-Ju Kang, Ju-Wan Kim, Wonsuk Choi, and Il-Seon Shin contributed to reviewing and editing of draft.