Predicting Worsening Suicidal Ideation With Clinical Features and Peripheral Expression of Messenger RNA and MicroRNA During Antidepressant Treatment

Predicting Worsening Suicidal Ideation With Clinical Features and Peripheral Expression of Messenger RNA and MicroRNA During Antidepressant Treatment



Objective: To investigate how the combination of clinical and molecular biomarkers can predict worsening of suicidal ideation during antidepressant treatment.

Methods: Samples were obtained from 237 patients with major depressive disorder (DSM-IV criteria) treated with either duloxetine or placebo in an 8-week randomized controlled trial. Data were collected between 2007 and 2011. The relationship between treatment-worsening suicidal ideation (TWSI) and a number of clinical variables, as well as peripheral expression of messenger RNA (mRNA) and microRNA (miRNA), was assessed at baseline. We generated 4 predictive models for TWSI: clinical, mRNA, miRNA, and a combined model comprising the best predictive variables from clinical, mRNA, and miRNA data.

Results: Eleven patients (9.8%) presented with TWSI in the duloxetine group. Among the clinical variables, only baseline depressive severity was found to be mildly predictive of TWSI. Two mRNAs (stathmin 1 [STMN1] and protein phosphatase 1 regulatory subunit 9B [PPP1R9B]) and 2 miRNAs (miR-3688 and miR-5695) were identified that were significantly predictive of TWSI when mRNA and miRNA were assessed separately (P = .002, .044, .004, and .005, respectively). The best model included baseline depression severity and expression of STMN1 and miR-5695 and predicted TWSI with area under the curve = 0.94 (P < .001). Additionally, the combined model did not significantly predict TWSI in the placebo group.

Conclusions: This study generated a predictive tool for TWSI that combines both biological and clinical variables. These biological variables can be easily quantified in peripheral tissues, thus rendering them viable targets to be used in both clinical practice and future studies of suicidal behaviors.

Trial Registration: identifiers: NCT00635219, NCT00599911, and NCT01140906

J Clin Psychiatry 2019;80(3):18m12556

To cite: Belzeaux R, Fiori LM, Lopez JP, et al. Predicting worsening suicidal ideation with clinical features and peripheral expression of messenger RNA and microRNA during antidepressant treatment. J Clin Psychiatry. 2019;80(3):18m12556.

To share:

aMcGill Group for Suicide Studies, Douglas Mental Health University Institute, Department of Psychiatry, McGill University, Montreal, Quebec, Canada

bDepartment of Public Health, EA 3279 Research Unit, University Hospital, Assistance Publique-H×´pitaux de Marseille, Marseille, France

cUniversity of Ottawa Institute of Mental Health Research, Ottawa, Ontario, Canada

dCentre for Addiction and Mental Health, Toronto, Ontario, Canada

eMcMaster University and St Joseph’s Healthcare Hamilton, Hamilton, Ontario, Canada

fDepartment of Psychiatry, University Health Network, Krembil Research Institute, University of Toronto, Toronto, Ontario, Canada

gDepartment of Psychiatry, University of British Columbia, Vancouver, British Columbia, Canada

hDepartment of Psychology, University of Guelph, Guelph, Ontario, Canada

iUniversity of Calgary Hotchkiss Brain Institute, Calgary, Alberta, Canada

jProvidence Care Hospital, Kingston, Ontario, Canada

kDepartment of Psychiatry, University of Michigan, Ann Arbor, Michigan

lSt Michael’s Hospital, Li Ka Shing Knowledge Institute, Centre for Depression and Suicide Studies, Toronto, Ontario, Canada

mDepartment of Psychiatry, Queen’s University, Kingston, Ontario, Canada

nDepartment of Psychiatry, Dalhousie University, Halifax, Nova Scotia, Canada

‘ ¡These authors contributed equally to this work.

*Corresponding author: Gustavo Turecki, MD, PhD, McGill Group for Suicide Studies, Douglas Mental Health University Institute, Frank B Common Pavilion, Room F-3125, 6875 LaSalle Blvd, Montreal, Quebec, H4H 1R3 (

Suicidal behavior is the most serious complication of mood disorders, and suicidal ideation is an important correlate of suicidal behavior.1 Although there has been a debate over the last decade about the occurrence of treatment-induced suicidal behavior, antidepressant treatment is one of the most effective strategies for suicide prevention, and most patients have a reduction of suicide ideation during antidepressant treatment. However, increased suicidal ideation during antidepressant treatment may occur in up to 30% of patients.2-4 This increased ideation may be characterized as either treatment-worsening suicidal ideation (TWSI) or, in patients who do not display suicidal ideation initially, as treatment-emergent suicidal ideation (TESI). Both TWSI and TESI, along with suicidal behavior, represent a major concern in daily patient care.

Suicidal behavior and variation in suicidal ideation during antidepressant treatment are difficult to predict. A complete clinical interview is the gold standard to evaluate suicide risk in daily clinical practice, but such interviews have been of limited value when it comes to predicting TWSI or TESI because clinical variables explain very little of the variance of suicidal ideation occurring during antidepressant treatment.5 In addition, combining a large set of clinical factors does not seem to improve the ability to predict suicidal behavior.6 Similarly, while the addition of psychometric scales to the clinical assessment renders the prediction more structured, it does not significantly increase specificity and is not part of routine clinical evaluation.7 Moreover, prediction is especially difficult because of variability across time in suicidal intent, which may be sudden and impulsive.8

The assessment of biomarkers may improve the identification of patients most likely to experience TWSI and TESI. Pharmacogenomic studies have described several genetic polymorphisms associated with a greater risk of suicide events during antidepressant treatment.4 The study of messenger RNA (mRNA) and microRNA (miRNA) expression is also a promising approach because these are dynamic molecular markers that may change with treatment.9 Numerous studies have shown altered expression of mRNA, RNA molecules that are transcribed into proteins, in major depressive disorder (MDD) and in relation to suicidal behavior.10,11 Although less well investigated, miRNAs, small noncoding RNA molecules that regulate gene expression, have also been implicated in MDD and suicide.12,13 Moreover, high-throughput technologies investigating mRNA and miRNA allow for hypothesis-free biomarker discovery, but to our knowledge there are no reports to date of prediction of TESI or TWSI using these markers. Finally, only a few studies14-16 have combined biomarkers and clinical data to improve the accuracy of biomarker prediction. Thus, their application to suicide prevention or the prediction of suicidality in the context of antidepressant treatment has yet to be determined.

In this study, we aimed to investigate the utility of molecular biomarkers, specifically peripheral mRNA and miRNA measures, to predict worsening of suicidal ideation during antidepressant treatment. We investigated depressed patients who were treated with duloxetine or placebo as part of a double-blind, randomized, placebo-controlled trial and used high-throughput technologies to study mRNA and miRNA as a function of TESI or TWSI.


Patients and Clinical Data

We included data from 237 participants with MDD who were part of double-blind, randomized, placebo-controlled trials of duloxetine ( identifiers: NCT00635219, NCT00599911, and NCT01140906). Participants, aged 18-75 years, were recruited based on a primary diagnosis of MDD and a major depressive episode lasting at least 3 months, with a severity score on the Montgomery-Asberg Depression Rating Scale (MADRS)17 at baseline of ≥ 22. Participants resistant to at least 2 previous antidepressant treatments or who had received electroconvulsive therapy in the 6 weeks prior to study beginning were excluded. Other exclusion criteria were major depressive episode in bipolar disorder, presence of psychotic features, and recent substance use disorder. Psychiatric diagnoses were made using the Mini-International Neuropsychiatric Interview according to DSM-IV criteria.18 All patients included in this analysis were free of medication at study inclusion and then received treatment with duloxetine (60 mg) or placebo for 6-8 weeks. A total of 112 patients were treated with duloxetine and 125 were treated with placebo. This study was approved by the local institutional review board, and written informed consent was obtained from all subjects.

Suicidal ideation was assessed 6 times: at baseline (before treatment) and 1, 2, 4, 6, and 8 weeks after treatment initiation (NCT00635219 and NCT01140906) or at baseline and 1, 2, 3, 4, and 6 weeks after treatment initiation (NCT00599911). Ideation was assessed using item 10 from the MADRS, “suicidal thoughts” with scores ranging from 0 (enjoys life or takes it as it comes) to 6 (explicit plans for suicide when there is an opportunity; active preparations for suicide). Although not ideal, the use of a single suicide item from a depression scale has been included in several previous studies and has also been validated as an appropriate tool to evaluate suicidal ideation in prospective studies.19 Treatment-worsening suicidal ideation was defined as an increase of at least 1 point on MADRS item 10 at any time during the follow-up. TWSI was the primary outcome measure. Treatment-emergent suicidal ideation was defined as an increase of at least 1 point on MADRS item 10 at any time during the follow-up in patients who did not present with suicidal ideation at baseline (ie, baseline MADRS item 10 score = 0).

Severity of depressive symptoms was measured using the MADRS total score at each visit. Anxiety severity at baseline was evaluated using the Hamilton Anxiety Rating Scale.20 Family history was evaluated using a standardized questionnaire to assess potential occurrence of psychiatric disorders or suicide in parents and grandparents. Due to small sample sizes in each category of diagnosis, we combined positive family history in a unique variable by considering the presence of at least 1 major psychiatric diagnosis (major depression, bipolar disorder, schizophrenia, suicide, substance use disorder) in at least 1 family member.

clinical points
  • Treatment-worsening suicidal ideation is an important concern during antidepressant treatment of patients with major depressive disorder.
  • A combination of biomarkers, including expression of stathmin 1 (STMN1) and miR-5695, along with baseline Montgomery-Asberg Depression Rating Scale severity scores may be used to predict worsening of suicidal ideation during antidepressant treatment.

Biological Assessments

Whole blood samples were collected at baseline using PAXgene Blood RNA Tubes (PreAnalytiX GmbH; Hombrechtikon, Switzerland). Total RNA was extracted using the miRNeasy Micro Kit (QIAGEN; Hilden, Germany) with DNase treatment. RNA integrity was evaluated using an Agilent Bioanalyzer (Agilent Technologies; Santa Clara, California). All samples had an RNA integrity number (RIN) > 6.

To study mRNA, RNA was hybridized to the Illumina HumanHT-12 v4 microarray (Illumina, Inc; San Diego, California). Samples were randomized to avoid batch effects. All array probes and samples were subjected to quality control using Flexarray (NetApp, Inc; Sunnyvale, California). Data were normalized using background adjustment and log2 transformation, variance stabilization transformation (VST) correction, and quantile normalization. In total, 47,323 probes were present in the microarray. All probes were filtered using a detection P value < .01 in at least 10% of the samples, resulting in available expression data for 16,674 remaining probes.

miRNA were analyzed using the Illumina TruSeq Small-RNA protocol (Illumina, Inc; San Diego, California) as previously described.21,22 Samples were sequenced at the McGill University and Genome Quebec Innovation Centre (Montreal, Canada) using the HiSeq2500 Illumina sequencer with 50-nucleotide single-end reads (Illumina, Inc; San Diego, California). We used a detection threshold of 10 counts per miRNA (present in at least 80% of libraries tested). A total of 281 miRNAs survived our criteria and were included in the analysis. All small RNA-sequencing data were normalized with the Bioconductor DESeq2 package using the variance-stabilizing transformation method.23

Statistical Analyses

Data were expressed as proportions and frequency for categorical variables and as means and standard deviations for continuous variables. We first conducted our analysis on patients treated with an active antidepressant (duloxetine).

To select the best mRNAs and miRNAs to be used as predictive biomarkers, expression levels were first analyzed at the transcriptome-wide level. Mean differences of RNA expression for each mRNA or miRNA were assessed between patients with or without TWSI during the follow-up using t tests and correction for multiple testing. The best potential mRNA and miRNA predictive biomarkers were selected according to a false discovery rate (FDR) threshold below 10%.24

We aimed to build 4 different models to predict occurrence of TWSI: best clinical variables (“clinical model”), best predictive mRNAs (“mRNA model”), best predictive miRNAs (“miRNA model”), and a model that combined the best predictive variables from clinical, mRNA, and miRNA data (“combined model”). For each model, variables were selected according to univariate logistic regression with occurrence of TWSI as dependent variable. Variables were included in subsequent multivariate analysis if P < .1 in the univariate analyses. For clinical variables, due to missing data, we used a multiple imputation strategy,25 and we computed both univariate logistic regressions before and after imputation of missing data. We included in the clinical model only variables that achieved statistical threshold in both analyses. For mRNA, unlog value of expression was used to avoid large odds ratios.

Models were built using a stepwise multiple logistic regression, and predictive scores were calculated for each model. A binormal smoothing was performed for building receiver operating characteristic (ROC) curves. A resampling method with a bootstrap strategy was used to calculate confidence intervals of area under the curve (AUC), sensitivity, and specificity. We then compared the AUC of the 3 first models (clinical, mRNA, and miRNA) to the final combined model using pROC package (R software; R Foundation). Finally, we tested the AUC of the combined model in the cohort of patients treated with placebo.


A total of 237 participants were included in the study (69.6% female; mean ± SD age = 46.8 ± 12.8 years): 112 treated with duloxetine and 125 treated with placebo (Table 1). Mean ± SD baseline MADRS score was 31.2 ± 3.7. During the follow-up, TWSI was identified in 32 participants (13.5%): 11 (9.8%) treated with duloxetine and 21 (16.8%) treated with placebo. TESI was identified in only 2 participants (0.8% of total), both treated with placebo. No suicide attempts were reported in this sample.

Table 1

Click figure to enlarge

Clinical Predictive Model

In the duloxetine group, before and after imputation of missing data, only depression severity at baseline, as measured by MADRS total score, predicted TWSI in univariate logistic regressions (P < .1). We found no association between TWSI and age, sex, level of anxiety symptoms, melancholic features, total number of major depressive episodes, familial history of psychiatric disorders, and history of antidepressant treatment (Supplementary Table 1). As a result, we built a predictive model using only the baseline MADRS total score for our clinical model. MADRS at baseline demonstrated a trend to be predictive for TWSI with an AUC = 0.66 (95% CI, 0.52-0.80; P = .08), sensitivity = 90.9%, and specificity = 42.6% (Table 2 and Figure 1).

Table 2

Click figure to enlarge

Figure 1

Click figure to enlarge

mRNA Predictive Model

In patients treated with duloxetine, after the analysis of the expression of 16,674 probesets, followed by correction for multiple testing with FDR < 10%, we found the expression of 2 probesets to be significantly associated with TWSI. Protein phosphatase 1 regulatory subunit 9B (PPP1R9B) was underexpressed before treatment in patients who presented with TWSI (fold change [FC] = -1.21, P = 1.13 ×— 10-5). Stathmin 1 (STMN1) was overexpressed before treatment in patients who presented with TWSI (FC = 1.20, P = 5.99 ×— 10-6). The expression levels of these 2 mRNAs were predictive of suicidal ideation in logistic regressions (Supplementary Table 2). We built a multivariate regression model using these 2 mRNA biomarkers. When combined, they predicted suicidal ideation with an AUC = 0.86 (95% CI, 0.75-0.97; P < .001), sensitivity = 81.8%, and specificity = 79.2% (Table 2 and Figure 1).

miRNA Predictive Model

After small-RNA sequencing, we found 2 miRNAs associated with TWSI in patients treated with duloxetine (FDR < 10%): both miR-3688 (FC = 2.10, P = 6.1 ×— 10-4) and miR-5695 (FC = 1.84, P = 1.72 ×— 10-4) were overexpressed in patients who presented with TWSI. Levels of these 2 miRNAs were predictive of suicidal ideation in logistic regression (Supplementary Table 3). We built a multivariate regression model including these 2 miRNA biomarkers. When combined, they predicted suicidal ideation with an AUC = 0.83 (95% CI, 0.73-0.97; P < .001), sensitivity = 100%, and specificity = 57.4% (Table 2 and Figure 1).

Combined Model

Using the same procedure, we included in a multiple logistic regression all biomarkers and clinical data that were predictive of TWSI in univariate analyses. Using a stepwise multivariate logistic regression, we found that miR-5695, STMN1 mRNA, and MADRS score at baseline significantly predicted TWSI during the follow-up (Supplementary Table 4). Based on multivariate logistic regression, we built a combined variable that allowed prediction of TWSI. Based on logistic regression β-coefficients, the probability was calculated as follows:

e(−9.522191 + 2,825924 ×— miR5695 + 0.359468 ×— STMN1 – 0.531583 ×— MADRS)/
(1+e(−9.522191 + 2.825924 ×— miR5695 + 0.359468 ×— STMN1 – 0.531583 ×— MADRS))

Using ROC-curve analysis, we found an AUC = 0.94 (95% CI, 0.88-0.99; P < .001), sensitivity = 100%, and specificity = 89.1% for the combined model (Table 2 and Figure 1).

Comparing the AUCs of these 4 different models, we found that the AUC of the combined model is significantly higher than the AUC of the clinical model (P < .001) and the AUC of the miRNA model (P = .005). The AUC of the combined model was higher than that of the mRNA model, although the difference was not significant (P = .23).

Combined Model in Patients Treated With Placebo

Finally, we tested our combined model among placebo-treated patients. We found no predictive value for placebo’s worsening suicidal ideation of our combined model with an AUC = 0.52 (P = .812).


In this study, we analyzed a large number of clinical and molecular variables and generated 3 models that significantly predicted TWSI. We identified 2 mRNAs and 2 miRNAs that were significantly predictive of TWSI when assessing mRNA or miRNA separately. Furthermore, we determined that combining clinical information with mRNA and miRNA expression values yielded the best ability to predict TWSI in our cohort. Our final combined model included the clinical variable of baseline depression severity (MADRS score) and the expression of the mRNA STMN1 and the miRNA miR-5695. The high accuracy and sensitivity of our tool allowed us to predict TWSI in the majority of patients. However, our model did not significantly predict TWSI in the placebo group, indicating that it is specific to antidepressant treatment. This model satisfies a number of key characteristics of biomarkers, namely clinical validity, clinical utility, clinical usefulness, and biological plausibility.26

Among all the clinical variables examined, only baseline MADRS was mildly predictive of TWSI, which is consistent with the findings that clinical variables explain very little of the variance of suicidal ideation occurring during antidepressant treatment.5

In the mRNA predictive model, our analyses identified 2 mRNAs, PPP1R9B and STMN1, whose expression values at baseline were predictive of TWSI. PPP1R9B, also known as spinophilin, is a scaffold protein involved in the development of dendritic spines and synapses.27 Additionally, this protein has been linked to (or associated with) the effects of environmental enrichment on neuronal plasticity in rodents,28 regulation of μ-opioid receptor function,29 and noradrenergic responses to the antidepressant desipramine.30 STMN1 encodes the protein stathmin, a neuronal growth-associated protein involved in regulating microtubule dynamics that has been associated with neuronal plasticity, fear, and anxiety responses.31,32 Although the relationship between the functions of these 2 mRNAs and their expression in the blood is unknown, their potential relevance to the pathophysiology of suicidal behavior indicates biological plausibility, an important characteristic for biomarkers.

Our miRNA predictive model identified 2miRNAs, miR-3688 and miR-5695. These 2 miRNAs have not been associated previously with any psychiatric phenotype. miRNAs are known to target a large number of mRNAs for either destruction or translational repression. However, the targets of these 2 miRNAs have yet to be established. We thus performed a target identification based on convergent results of 4 algorithms (miRWalk, miRanda, RNAhybrid, and Targetscan), and we conducted a gene ontology analysis using a set of shared predicted targets of these 2 miRNAs. We found that the most significantly enriched term was regulation of neurogenesis (GO:0050769, analysis not shown). As miRNAs can be packaged into extracellular vesicles and secreted into the blood from all bodily tissues, including the brain,33 it is possible that our findings here represent a plausible pathologic process occurring elsewhere in the body. Additional work will be necessary to identify their relevant gene targets in order to better understand their potential role in TWSI.

Although encouraging, our results should be interpreted with caution. Firstly, our analysis comprised a relatively small sample of patients with major depression. However, previous studies also included limited sample sizes and/or more heterogeneous phenotypes.15,16,31,34 Moreover, the relatively low frequency of TWSI (< 10% in our sample) reduces the ability to detect biomarkers of TWSI. In the same vein, we choose a more inclusive definition of TWSI (ie, increase of 1 point), while other studies have utilized a 2-point increase35 or used a composite score based on several tools.2 In our cohort, the majority of patients exhibiting TWSI displayed an increase of 1 point, with the exception of 2 patients in the duloxetine group with an increase of 2 points. The maximum severity (score) of MADRS item 10 was 3. As a consequence, our results may suffer from low specificity or overestimation. However, one previous study36 also demonstrated the accuracy of this definition as a biomarker of suicidal ideation. Moreover, at least theoretically, a stricter definition would also be associated with lower sensitivity, which is, based on a clinical point of view, the more important parameter when predicting severe complications such suicide ideation and its potential consequences. Overall, this is a common issue in randomized controlled trials, from which patients with higher suicide risk are often excluded. As such, larger samples including patients with higher risk for suicidal behaviors are needed to overcome this issue.

Secondly, we did not replicate previous findings investigating clinical variables as predictors of suicidal behavior, aside from a trend for depression severity to be associated with TWSI. However, it is worth noting that most of these previous findings suggested a low effect size for clinical predictors.5

The identification of biomarkers that allow for the diagnosis of psychiatric disorders, and prediction of clinical outcomes, is an important research goal. Numerous studies have identified both clinical and biological variables that may be used for these purposes; however, many of these findings have failed to be replicated. Ultimately, biomarkers must demonstrate replicability, accuracy, and cost-effectiveness to warrant their inclusion in clinical practice.

Although our predictive tool is promising, its clinical utility remains to be demonstrated. It is worth noting that the generalization of our findings to suicidal ideation in general is unclear, as we focused our analyses on suicidal ideation specifically associated with antidepressant treatment and not during placebo treatment. Our study design did not allow us to test if the use of our predictive tool may significantly improve the prognosis of patients treated with antidepressants. On the basis of our findings, a randomized controlled trial comparing treatment management with and without our predictive tool would be the only method to test the efficacy of these biomarkers. However, to date, the prediction of suicidal ideation and suicide behavior is based only on clinical interviews, and the accuracy of such practice has not been demonstrated. As such, we could speculate that the predictive tools described herein have a clinical utility that is at least comparable to current standard practice.

In summary, we report a predictive tool for TWSI during antidepressant treatment that combines both biological and clinical variables. These biological variables can be easily quantified in peripheral tissues, thus rendering them viable targets to be used in both clinical practice and future studies of suicidal behaviors.

Submitted: September 4, 2018; accepted December 3, 2018.

Published online: May 7, 2019.

Potential conflicts of interest: None.

Funding/support: This research was conducted with the support of the Canadian Biomarker Integration Network in Depression (CAN-BIND), an Integrated Discovery Program, with funding from the Ontario Brain Institute, an independent non-profit corporation, funded partially by the Ontario government. Additional funding for CAN-BIND is provided by the Canadian Institutes of Health Research (CIHR), Brain Canada, Lundbeck, Bristol-Myers Squibb, Pfizer and Servier. Dr Turecki holds a Canada Research Chair (Tier 1) and a NARSAD Distinguished Investigator Award. He is supported by grants from the Canadian Institute of Health Research (CIHR) (FDN148374 and EGM141899) and by the Fonds de recherche du Québec – Santé (FRQS) through the Quebec Network on Suicide, Mood Disorders and Related Disorders. Dr Belzeaux received FondaMental Servier Fellowship funding. Dr Lopez received a Frederick Banting and Charles Best Canada Graduate Scholarships doctoral funding award from CIHR. Dr Lam is supported by the BC Leading Edge Endowment Fund and VGH Foundation. This study was conducted on samples generously made available by Lundbeck.

Role of the sponsor: Neither the funders nor Lundbeck had any role in the design, analysis, interpretation, or publication of this study.

Supplementary material: Available at PSYCHIATRIST.COM.


1. Turecki G, Brent DA. Suicide and suicidal behaviour. Lancet. 2016;387(10024):1227-1239. PubMed CrossRef

2. Perroud N, Uher R, Marusic A, et al. Suicidal ideation during treatment of depression with escitalopram and nortriptyline in genome-based therapeutic drugs for depression (GENDEP): a clinical trial. BMC Med. 2009;7(1):60. PubMed CrossRef

3. Zisook S, Trivedi MH, Warden D, et al. Clinical correlates of the worsening or emergence of suicidal ideation during SSRI treatment of depression: an examination of citalopram in the STAR*D study. J Affect Disord. 2009;117(1-2):63-73. PubMed CrossRef

4. Brent D, Melhem N, Turecki G. Pharmacogenomics of suicidal events. Pharmacogenomics. 2010;11(6):793-807. PubMed CrossRef

5. Coughlin CG, Jakubovski E, Bloch MH. Time course and predictors of suicidal ideation during citalopram treatment in the STAR*D trial. J Clin Psychiatry. 2016;77(10):e1262-e1269. PubMed CrossRef

6. Large M, Kaneson M, Myles N, et al. Meta-analysis of longitudinal cohort studies of suicide risk assessment among psychiatric patients: heterogeneity in results and lack of improvement over time. PLoS One. 2016;11(6):e0156322. PubMed CrossRef

7. Blasco-Fontecilla H, Delgado-Gomez D, Ruiz-Hernandez D, et al. Combining scales to assess suicide risk. J Psychiatr Res. 2012;46(10):1272-1277. PubMed CrossRef

8. Lopez-Castroman J, Blasco-Fontecilla H, Courtet P, et al. Are we studying the right populations to understand suicide? World Psychiatry. 2015;14(3):368-369. PubMed CrossRef

9. Turecki G. The molecular bases of the suicidal brain. Nat Rev Neurosci. 2014;15(12):802-816. PubMed CrossRef

10. Oquendo MA, Sullivan GM, Sudol K, et al. Toward a biosignature for suicide. Am J Psychiatry. 2014;171(12):1259-1277. PubMed CrossRef

11. Fiori LM, Turecki G. Broadening our horizons: gene expression profiling to help better understand the neurobiology of suicide and depression. Neurobiol Dis. 2012;45(1):14-22. PubMed CrossRef

12. Lin R, Turecki G. Noncoding RNAs in depression. Adv Exp Med Biol. 2017;978:197-210. PubMed CrossRef

13. Dwivedi Y. MicroRNAs in depression and suicide: recent insights and future perspectives. J Affect Disord. 2018;240:146-154. PubMed CrossRef

14. Guintivano J, Brown T, Newcomer A, et al. Identification and replication of a combined epigenetic and genetic biomarker predicting suicide and suicidal behaviors. Am J Psychiatry. 2014;171(12):1287-1296. PubMed CrossRef

15. Le-Niculescu H, Levey DF, Ayalew M, et al. Discovery and validation of blood biomarkers for suicidality. Mol Psychiatry. 2013;18(12):1249-1264. PubMed CrossRef

16. Niculescu AB, Levey DF, Phalen PL, et al. Understanding and predicting suicidality using a combined genomic and clinical risk assessment approach. Mol Psychiatry. 2015;20(11):1266-1285. PubMed CrossRef

17. Montgomery SA, Asberg M. A new depression scale designed to be sensitive to change. Br J Psychiatry. 1979;134(4):382-389. PubMed CrossRef

18. Sheehan DV, Lecrubier Y, Sheehan KH, et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry. 1998;59(suppl 20):22-33, quiz 34-57. PubMed

19. Courtet P, Jaussent I, Lopez-Castroman J, et al. Poor response to antidepressants predicts new suicidal ideas and behavior in depressed outpatients. Eur Neuropsychopharmacol. 2014;24(10):1650-1658. PubMed CrossRef

20. Hamilton M. The assessment of anxiety states by rating. Br J Med Psychol. 1959;32(1):50-55. PubMed CrossRef

21. Lopez JP, Diallo A, Cruceanu C, et al. Biomarker discovery: quantification of microRNAs and other small non-coding RNAs using next generation sequencing. BMC Med Genomics. 2015;8(1):35. PubMed CrossRef

22. Lopez JP, Fiori LM, Cruceanu C, et al. MicroRNAs 146a/b-5 and 425-3p and 24-3p are markers of antidepressant response and regulate MAPK/Wnt-system genes. Nat Commun. 2017;8:15497. PubMed CrossRef

23. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15(12):550. PubMed CrossRef

24. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B. 1995;57(1):289-300.

25. Little R, Rubin D. Statistical Analysis with Missing Data. 2nd ed. Hoboken, NJ: Wiley & Sons, Inc; 2014.

26. Belzeaux R, Lin R, Turecki G. Potential use of MicroRNA for monitoring therapeutic response to antidepressants. CNS Drugs. 2017;31(4):253-262. PubMed CrossRef

27. Evans JC, Robinson CM, Shi M, et al. The guanine nucleotide exchange factor (GEF) Asef2 promotes dendritic spine formation via Rac activation and spinophilin-dependent targeting. J Biol Chem. 2015;290(16):10295-10308. PubMed CrossRef

28. Hu XL, Bergström SA, Brink M, et al. Enriched environment increases spinophilin mRNA expression and spinophilin immunoreactive dendritic spines in hippocampus and cortex. Neurosci Lett. 2010;476(2):79-83. PubMed CrossRef

29. Charlton JJ, Allen PB, Psifogeorgou K, et al. Multiple actions of spinophilin regulate mu opioid receptor function. Neuron. 2008;58(2):238-247. PubMed CrossRef

30. Cottingham C, Li X, Wang Q. Noradrenergic antidepressant responses to desipramine in vivo are reciprocally regulated by arrestin3 and spinophilin. Neuropharmacology. 2012;62(7):2354-2362. PubMed CrossRef

31. Brocke B, Lesch KP, Armbruster D, et al. Stathmin, a gene regulating neural plasticity, affects fear and anxiety processing in humans. Am J Med Genet B Neuropsychiatr Genet. 2010;153B(1):243-251. PubMed

32. Ehlis AC, Bauernschmitt K, Dresler T, et al. Influence of a genetic variant of the neuronal growth associated protein Stathmin 1 on cognitive and affective control processes: an event-related potential study. Am J Med Genet B Neuropsychiatr Genet. 2011;156(3):291-302. PubMed CrossRef

33. Prada I, Gabrielli M, Turola E, et al. Glia-to-neuron transfer of miRNAs via extracellular vesicles: a new mechanism underlying inflammation-induced synaptic alterations. Acta Neuropathol. 2018;135(4):529-550. PubMed CrossRef

34. Mullins N, Hodgson K, Tansey KE, et al. Investigation of blood mRNA biomarkers for suicidality in an independent sample. Transl Psychiatry. 2014;4(10):e474. PubMed CrossRef

35. Voegeli G, Ramoz N, Shekhtman T, et al. Neurotrophin genes and antidepressant-worsening suicidal ideation: a prospective case-control study. Int J Neuropsychopharmacol. 2016;19(11):pyw059. PubMed CrossRef

36. Consoloni JL, Ibrahim EC, Lefebvre MN, et al. Serotonin transporter gene expression predicts the worsening of suicidal ideation and suicide attempts along a long-term follow-up of a Major Depressive Episode. Eur Neuropsychopharmacol. 2018;28(3):401-414. PubMed CrossRef

Editor’s Note: We encourage authors to submit papers for consideration as a part of our Focus on Suicide section. Please contact Philippe Courtet, MD, PhD, at

Related Articles

Volume: 80

Quick Links: Depression (MDD) , Suicide


Buy this Article as a PDF