In the September 2010 issue of The Carlat Child Psychiatry Report, Dr. Fisher neatly explains the meaning of Number Needed to Treat (NNT) and Number Needed to Harm (NNH), which can be a shorthand way of assessing whether to prescribe or not, and then applies this to the question of SSRI antidepressants in pediatrics. She illustrates the point that compared to other specialties, SSRIs are no worse than other commonly used medications (although there is some controversy over statins, too). However, to do so, she glosses over the rich controversy regarding SSRI use in children. I will elucidate that controversy here.
Unfortunately a key reason for a more pessimistic view is lack of trust in the published literature (Spielmans GI, Parry PI, Bioethical Inquiry 2010;7(1):13–29). In the January 2008 issue of the New England Journal of Medicine, Turner et al highlight the problem of publication bias with antidepressants (Turner et al, N Engl J Med 2008;358(3):252–260).
For instance, the Keller paper that Dr. Fisher refers to (Keller MB et al, JAACAP 2001;40(7):762–772), which gives paroxetine (Paxil) a NNT of six, has received strong criticism. In internal company documents (see http://bit.ly/avtAyP), the manufacturers of paroxetine acknowledge that data from two studies of paroxetine in adolescents (studies number 329 and 377), show no effect.
According to this document, they elected not to submit either study to regulatory bodies, but decided to publish the positive data on secondary outcome measures in study 329—the Keller study. Jureidini et al concluded, “Study 329 was negative for efficacy and positive for harm,” based on company documents that give a fuller picture of the data than the published paper (Jureidini JN et al, Int J Risk & Safety Med 2008;20(1–2):73–81).
Based on the previously unpublished data for study 329 (eg, “serious” events were 12% paroxetine vs 2.3% placebo, and “severe” events 27% paroxetine vs 17% placebo), the relatively high NNH rates for suicidality of up to 160 for SSRIs should be considered with some skepticism. Other SSRIs like fluoxetine (Prozac) appear less concerning in this regard than paroxetine; however another critical analysis looked at other published studies of SSRIs in the pediatric age group and came to similar concerns of overestimation of benefits and underestimation of harms (Jureidini JN et al, BMJ 2004;328 (7444):879–883).
Although NNT and NNH are easy shorthand ways of comparing efficacy and adverse event potential, they can be misleading depending on the outcome measures. The NNT derived from the randomized controlled trials (RCTs) apply to relatively modest changes on physician recorded rating scales. Young patient self-reports in RCTs of SSRIs are few and do not reflect benefit (Jureidini, ibid). Therefore, NNT reflects modest improvement, whereas NNH reflects serious or potentially fatal outcomes.
In considering NNH it is worth recognizing that serotonin has many functions throughout the body and there are other serious side effects from SSRIs in addition to suicidality. Sexual side-effects like delayed ejaculation can be problematic for adolescents’ developing sexuality. There is some evidence linking SSRIs with growth hormone suppression (Weintrob N et al, Arch Pediatr Adolesc Med 2002;156(7):696–701), risk of osteoporosis (Verdel BM et al, Bone 2010;47(3):604–609), disturbances with spermatogenesis (Tanrikut C et al, Fert Sterility 2010;94(3):1021–1026), and birth defects (Healy D et al, Int J of Risk Safety Med 2010;22(1):1–10). Additionally there are risks of drug interactions and more benign side effects such as gastrointestinal upset and headache.
A further problem that concerns me before prescribing is the risk of dependence states, which now do seem to be a factor in SSRIs (Healy D, Med Hypotheses 2010;74(5):764–768). There is some evidence to argue there are worryingly low NNHs for sexual dysfunction (NNH of 2 or 3), growth retardation (NNH of 2 or 3), and physical dependency states (NNH of 3 or 4) (Healy D, J Psychopharm 2007;21(6):668–669).
One criticism of RCTs in depression for SSRIs is that subjects often have milder major depression, dysthymia, or adjustment disorders and this may underestimate the benefit a group of those with severe major depression have from SSRIs. A recent meta-analysis of antidepressant efficacy in RCTs concludes that antidepressant efficacy “may be minimal or nonexistent, on average, in patients with mild or moderate symptoms. For patients with very severe depression, the benefit of medications over placebo is substantial” (Fournier JC et al, JAMA 2010;303(1):47–53). It should be noted, however, that this was in regard to adults. The Treatment for Adolescents with Depression Study (TADS) study was well powered and had adolescents who were more severely depressed (TADS team, Arch of Gen Psychiatry 2007;64(10):1132–1144). This may have led to the relatively low NNT of four.
However the reported benefit of fluoxetine over placebo derives from the clinical global impression (CGI) scale and the children’s depression rating scale–revised (CDRS-R) used categorically (benefit vs no benefit), rather than dimensionally (ie, how much benefit?). If the CDRS-R is used dimensionally (as it arguably should be), then fluoxetine fails to statistically differentiate itself from placebo. Apart from the issue of statistical significance, a three-point difference on a 96 point scale is not likely to be of great clinical significance.
The other two arms of the TADS study compared cognitive behavioral therapy (CBT) alone with CBT plus open-label fluoxetine, and were limited by lack of a CBT plus placebo arm. When the TADS was extended to 36 weeks, there was no differentiation in efficacy of fluoxetine from CBT (TADS team, Arch of Gen Psychiatry 2007;64(10):1132–1144).
The TADS team concluded that combination therapy was superior on the basis of more rapid response (in first 12 weeks), and because the increased suicidality seen in fluoxetine monotherapy was not seen in combination with CBT (fluoxetine alone 14.7%, combination therapy 8.4%, CBT 6.3%). Given the other factors of concern from SSRIs, a case can be made from TADS to use CBT rather than medication in the first instance.
My first job as a consultant child and adolescent psychiatrist was in a mood disorders unit for young people, where I prescribed SSRIs and other new antidepressants to almost all. Benefit often seemed underwhelming after an initial probable placebo response, and agitation/activation reactions were not uncommon. Today I reserve SSRIs for severe OCD, phobic anxiety states not responding to CBT, and severe depression with melancholic features like psychomotor retardation, and use them only very rarely for the typical adolescent depression that fails to improve.
However the “typical” depressed teenager seems to benefit quite well from an eclectic mix of a “behavioral activation” approach of exercise and socializing, improved sleep hygiene, diaphragmatic breathing relaxation practices, healthier diet, omega-3 supplementation, and reduced substance abuse—along with addressing school academic and bullying issues, and providing family and individual psychotherapy. Mental health services need to be structured and funded to support a holistic approach.
Editor’s Response: Dr. Parry is accurate and scholarly—however, the problems he notes are pharmaceutical industry practices, not psychopharmaceutical industry practices.
Acknowledgements: Thanks to Professor David Healy and Professor Jon Jureidini for their reading and helpful comments regarding this commentary.