The following information should be cited as: 

Chang, B.P., Franklin, J.C., Ribeiro, J.D., Fox, K.R., Bentley, K.H., Kleiman, E., & Nock, M.K. (in press). Biological risk factors for suicidal behaviors: A meta-analysis. Nature: Translational Psychiatry.

Overview

Researchers have long believed that biological factors may play a crucial role in predicting future suicidal thoughts and behaviors. As with the literature on many other types of risk factors, however, the evidence has been mixed. For example, some studies have reported that non-suppression during the dexamethasone suppression test (an index of adrenal/cortisol reactivity) is a strong predictor of future suicidal behaviors (Jokinen & Nordstrom, 2008; Targun et al., 1983), but others studies have found that non-suppression predicts significantly fewer suicidal behaviors (Black et al., 2002). We undertook this meta-analysis in an effort to resolve such discrepancies and to identify potentially promising biological risk factors for suicidal thoughts and behaviors. Below we expand on these three major takeaways from the meta-analysis:

First, given the focus on this line of research, there have been surprisingly few longitudinal studies that used a biological factor to predict any kind of suicidal thought or behavior. In fact, there was only one qualifying instance of a biological factor predicting suicide ideation.

Second, overall and individually, biological factors are weak predictors of future suicidal behaviors. This finding applied to genes, hormones, neurotransmitter metabolites, and all of general classes of biological factors. 

Third, it is unlikely that any single biological factor will ever accurately predict suicidal thoughts or behaviors. However, when combined with a range of other biological and non-biological factors, a given biological factor may meaningfully improve risk detection. 

 


WHAT'S BEEN DONE SO FAR?

Answer: Through 2014, there were only 47 papers on this topic, producing 85 qualifying risk factor tests and 9 protective factor tests. The first qualifying study was published in 1976, with the number of qualifying risk/protective factor tests slowly increasing from pre-1985 (n = 5), to 1985-1994 (n = 10), to 1995-2004 (n = 37), to 2005-2014 (n = 42). Suicide death was the most common outcome examined, followed by suicide attempt. General population reference groups were most common, followed by clinical and self-injurious reference groups. Studies were generally small, with a median of 21 participants with any kind of suicide-related outcome, and follow-up intervals were typically long (M = 109.99 months, SD = 117.41 months) -- only two risk/protective factor tests included follow-ups of less than one year. 

Sample Severity

Outcome Types


WHAT ARE THE EFFECTS OF Biological Risk Factors?

Answer: Effects were universally small and most failed to reach statistical significance. Notably, serotonin-related genes and metabolites were commonly studied, but did not produce accurate prediction. As shown below, publication bias inflated effects -- with virtually no effects remaining statistically significant after these corrections. These findings suggest that small studies reporting relatively large effects tend to get published, but these reports are not reflective of the true (and very small) effects of biological factors on suicidal behaviors. Results were not moderated by factors (e.g., study length, reference group, etc.). 

Overall Prediction of Suicide Attempt and Death

Notes: wOR = weighted odds ratio; A wOR of 1.0 = chance-level prediction. Considering the low base rates of these outcomes, wORs would need to be in the 100s or 1000s to be clinically helpful. Both uncorrected effects reached statistical significance, but neither effect was significant when publication bias was factored in.

Prediction by Biological Factor Type

Notes: We only graphed categories that included data from more than one study. None of the above categories were significant after accounting for publication bias.

Where do we go from here?

Answer: Biological factors may still have potential, but there is much work to be done before discovering that potential. We propose the following recommendations:

  • Biological factors are poor predictors of future suicidal behaviors and should not be used in practice for the foreseeable future. 
  • Biological risk factors should be studied much more often, especially in relation to suicide ideation. An important caveat to this is that these studies must test novel risk factors (e.g., brain imaging, EEG, cardiovascular measures, etc. rather than serotonin and cortisol) and examine them more dynamically (e.g., changes physiological activity in the context of a suicide-related stimulus or situation).
  • Any biological factor in isolation is unlikely to accurately predict suicidal thoughts or behaviors. When combined with a large number of other factors, however, biological factors may be an important contributor to accurate risk identification.