Edgar Allan Poe’s fictional detective C. Auguste Dupin warns against tackling questions that are too complicated to test, but too fascinating to give up. Whether psychotherapy or medication can reduce suicidality is probably such a question. Particularly if we are really interested in whether treatments can reduce attempted suicides, not whether they change patients’ answers in an interview or on a questionnaire.
There is no doubt about the clinical and public health significance of the question. After all, psychotherapy and medication are treatments of choice for suicidal patients. The logic is that many, even if not all, suicidal persons are depressed; we know about effective treatments for depression; and so we can generalise from knowledge about what works for depression to what works for suicidality. However, we must hope for more definitive evidence, and a new study attempts to provide it.
The authors include suicide expert Ad Kerkhof, and Pim Cuijpers, who has done some of the most influential meta-analyses and systematic reviews on the treatment of depression. Together with doctoral student Erica Weitz and depression expert Steven Hollon, they analyzed data from the US National Institute of Mental Health Treatment for Depression Collaborative Research Project (TDCRP). Conducted in the 1980s, it was then the largest ever comparison of psychotherapy and medication for treatment of depression. Two hundred and fifty patients with major depression were randomized to cognitive therapy, interpersonal psychotherapy, antidepressant medication, or a pill-placebo plus clinical management as a control group.
The original study did not specifically target suicidality. It actually excluded patients with moderate to severe suicidality. However, the two primary depression outcome measures for the study, the self-report Beck Depression Inventory (BDI) and the interview-administered Hamilton Rating Scale for Depression (HRSD), each contained a single item inquiring about suicidal thoughts and behaviour:
Suicidal ideation/suicidality is rated on the HRSD on a 5-point scale:
1—feels life is not worth living,
2—wishes he were dead or any thoughts of possible death to self,
3—suicide ideas or gesture.
4—attempts at suicide (any serious attempt rates a 4).
The suicidality question on the BDI is measured on a 4-point scale:
0—I do not have any thoughts of killing myself,
1—I have thoughts of killing myself, but I would not carry them out,
2—I would like to kill myself,
3—I would like to kill myself if I had the chance.
The new analysis required that patients have at least some suicidal ideation on either measure. Of the 250 patients, 146 met this criterion. At the start of treatment, patients scored a mean of 1.15 on the HRSD suicide item and .74 on the BDI’s item. The sample included one person who had made a suicide attempt. This case proved to be an outlier and was removed from the analysis. Thus, this study captures mostly mild to moderate suicidal thoughts.
Based on measures taken pre- and post-treatment, the authors found that all treatments, including the pill placebo with clinical management, significantly reduced scores on both the interview and self-report measures of suicidality, with all having a medium effect size. According to the interview measure, interpersonal psychotherapy and antidepressant medication reduced suicidality more than the pill placebo with clinical management. No differences were found between treatments using the self-report measure.
The authors recognised that because the comparison-control group (pill placebo plus clinical management) significantly reduced suicidality, no conclusions could be drawn about specific components of the treatments being essential. It is important to note that pill placebo plus clinical management was not an inert control condition. Neither patients nor therapist knew that any antidepressant was not given, and there were considerable positive expectations, support and encouragement. I am sure that outcomes would have been better in this group than for a waiting list control condition, but there was none included the study.
Recall that the items measuring suicidal ideation were part of depression scales. Did these specific items decrease simply as a result of overall improvements in depression? The authors state they ruled that out with complex multivariate analyses, but I was left unconvinced.
Suicidal ideation is a surrogate outcome. That is, it serves as a proxy for the more interesting, but less frequent outcomes of suicide gestures and attempts and completed suicides. However, the problem with a proxy outcome is the treatment can have a positive effect that is insufficient to change the clinical variables of interest. There was a time when pharmaceutical companies relied on surrogate outcomes like reduction in blood pressure when rates of heart attack were the actual variable of interest. In that context, many treatments affected surrogate outcomes without changing the real variables of interest. The same could be happening here.
Overall, the study demonstrates a dilemma. Mild suicidal ideation is common among depressed patients seeking treatment, but overall is a poor predictor of suicide attempts, which are comparatively infrequent. While many of the patients who ultimately attempt suicide present with serious suicidal ideation, most of them start off with signs of only mild to moderate suicidality. The seeming paradox is due to having to predict later infrequent events from imperfect and nonspecific risk indicators.
We can certainly study treatment of patients at high risk because of a recent suicide attempt, but what we learn then is not readily generalisable to the more common clinical situation of patients expressing only mild to moderate suicidality when they enter treatment. On the other hand, if we study the treatment of this moderate suicidality seen in the clinic, we can’t measure the impact on actual attempts or death by suicide, because to do so would involve a prohibitively large sample.
We are left with the uncomfortable situation of attempting to address a clinical problem in studies with poor measures and inadequate sample size. Or simply having to settle for answering the question “Do depression treatments reduce suicidality?” with “Probably: they reduce depression.”
Weitz E, Hollon SD, Kerkhof A, & Cuijpers P (2014). Do depression treatments reduce suicidal ideation? The effects of CBT, IPT, pharmacotherapy, and placebo on suicidality. Journal of affective disorders, 167C, 98-103 PMID: 24953481
Post written by James Coyne (@Coyneoftherealm) for the BPS Research Digest. James Coyne, PhD is Professor of Health Psychology, University Medical Center, Groningen and the 2015 Carnegie Centenary Professor at University of Stirling.