Introduction  

The safety profile of selective serotonin reuptake inhibitors (SSRIs) in adolescents has been extensively debated. Several systematic reviews have analysed what is known about the risk of suicidal behaviour1–3 as well as other psychiatric and somatic adverse risks and the perceived benefit/risk balance. The reviews have highlighted considerable variations in assessment, definitions and reporting of adverse effects in the clinical trials.

The Norwegian Regional Medicines Information and Pharmacovigilance Centres and the Center for Psychopharmacology at Diakonhjemmet Hospital regularly receive queries from hospital doctors and general practitioners regarding the safety of fluoxetine (FLX) and other SSRIs in adolescent patients.

One of the major clinical studies of efficacy and safety of FLX in adolescents is the ‘Treatment for Adolescents With Depression Study (TADS)’, which is often referred to in textbooks and reviews.

In 1998, the US National Institute of Mental Health (NIMH) issued a request for proposals (RFP-NIH-NIMH 98-DS-0008) with the objective of launching a clinical trial to address the effectiveness of treatment for adolescents with major depression.4 The subsequent study, ‘TADS’ was coordinated by the Department of Psychiatry and Behavioral Sciences and the Duke Clinical Research Institute, both at Duke University Medical Center, collaborating with and funded by NIMH,5 and carried out in the period 2000–2003.6 The study included 439 youths who were randomised to one of four treatment groups; (1) FLX, (2) cognitive–behavioural therapy (CBT), (3) cognitive–behavioural therapy plus fluoxetine (COMB) or (4) placebo (PBO) for 12 weeks (stage I).6 Double-blind treatment was performed among patients treated with FLX and PBO only, while patients treated with CBT with or without FLX received open treatment. Stage II and III were maintenance phases for the active treatment groups, with the option of intensifying treatment for partial responders. Patients in the PBO group were offered open active treatment of FLX, CBT or both. Stage IV consisted of an additional year of open follow-up.5

The two primary outcome measures in the TADS were Children’s Depression Rating Scale-Revised (CDRS-R) total scores, and responder status on the Clinical Global Impressions-Improvement scale. According to protocol, all analyses would be performed by intention to treat (ITT), regardless of later events.

Adverse events during the acute and maintenance phases were defined as secondary outcomes.7 Patients were monitored for safety regarding affective disorders, need for mental health treatment, need for concomitant medications, occurrence of adverse events and serious adverse events and use of adjunctive services and attrition prevention. Most assessments were based on both patient and parent information.8

The TADS has been described as the largest and arguably the highest quality acute-phase randomised PBO controlled trial of an antidepressant drug for adolescent depression.9 We understand from the protocol and monitoring procedures that the TADS team intended to evaluate the tolerability of treatment, and that the study was expected to provide improved insight into the potential adverse effects of antidepressant treatment in this age group, due to its study size and duration. Several publications from the TADS have addressed risks of adverse effects. Despite this, concerns have been raised regarding under-reporting of suicidal risk,10 study size and an increased risk of psychiatric adverse effects.11

In the TADS, adverse events were defined as an unfavourable medical change that occurred after beginning or during the study that might or might not be related to or caused by the study drug or CBT treatment. This was further specified as any medical event that caused clinically significant interference with functioning (eg, headache that caused school absence or otherwise caused clinically significant activity restriction), any event that required medical attention, and any medical event associated with impairment in functioning and induced the patient to take a concomitant medication. Conditions that did not lead to clinically significant interference with functioning or did not require medical attention were not defined as adverse events.7 8 The protocol specified that new-onset psychiatric symptoms, such as emerging mania or panic attacks, would be recorded if they caused clinically significant interference with functioning.8 It follows that such conditions would not be recorded unless a certain severity threshold was reached.

Harm-related adverse events were defined as involving harm to self, which could include a non-suicidal event. Examples given are cutting, worsening of suicidal ideation, suicide attempt or harm to others. Suicide-related adverse events were defined as worsening suicidal ideation and/or suicide attempt. Adverse event forms were to be used throughout the study and it must be assumed that such data were collected, as well as clinical scoring data for possible psychiatric adverse events.

Our objective in the present study was to identify all publications from the TADS and assess the findings regarding occurrence of any adverse effects in the treatment groups both for the short-term and long-term study stages. The TADS was chosen because of the non-industrial funding and because it is considered as a high-quality study.9

Methods

Results

We identified 48 publications that reported on the study protocol and/or various outcomes in the TADS population. The selection process and publication characteristics are described in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Selection and characteristics for publications from the TADS. NIMH, National Institute of Mental Health; TADS, Treatment for Adolescents With Depression Study.

Eight publications were assessed as providing at least some data on adverse effects,6 12–18 of which four publications reported possible adverse effects for subgroups of patients only; patients who responded to treatment,13 patients originally assigned to PBO treatment,16 patients who had at least one suicidal event17 and patients using attrition prevention services,14 respectively. Reporting of adverse effects was most detailed in the two initial results publications from stage I (0–12 weeks),6 12 and included a wide range of adverse effects, including several psychiatric and gastrointestinal reactions. One stage I publication did not address adverse effects explicitly; however, symptoms that may be associated with adverse effects were described as residual symptoms of depression.13

The publications that reported on adverse effects in the later study stages II–IV listed few adverse effects except suicidal behaviour (table 1). The publication that purported to report on long-term effectiveness and safety outcomes only included reporting of suicide-related adverse events.15

Discussion

The TADS protocol included a threshold limit on what would be considered an adverse event, specifying that the event must cause clinically significant interference with functioning, require medical attention or cause a need to take medication.6 As an example, emerging mania was not recorded unless symptoms exceeded this threshold.7 It must be assumed that this reduced the number of reported adverse effects, which may not have been severe enough to reduce daily functioning or cause a need for additional treatment. We have not been able to find a published version of the questionnaires that were used and consequently do not have information as to which adverse effects were specifically asked for. The protocol does not define how the scoring parameters for adverse events should be analysed. The number of suicidal events is described, but other parameters, such as absolute or worsening scores on risk assessment scales, are not consistently reported. An example is the SIQ-Jr scores, where week 12 publications report mean scores and number of patients with score increase to ≥31,6 12 while the follow-up publication by week 36 reported per cent of patients with SIQ-Jr score ≥31.15 Scoring of mania symptoms is described as inconsistent and varying between clinicians.12 It is conceivable that some patients may have had worsening scores without passing the threshold score for suicidality or mania, respectively. Conversion into dichotomous scales, as was done for SIQ-Jr scores ≥31 and ADS Mania subscale score change increase ≥3 points, does not give insight into the magnitude in case of increased scores.

All analyses were planned as ITT, regardless of later events.7 Nine cases of suicidal behaviour were presented as occurring in the PBO group17 although the patients were using FLX at the time and the PBO period had ended. As pointed out by Högberg et al,10 the risk of suicidal behaviour will not appear to be increased for FLX compared with PBO if patients using FLX are assessed in the PBO group. ITT analyses of adverse events may be biased towards finding no differences between groups.20 This is especially relevant in studies with large drop-out rates and in study groups where patients received add-on treatment that differed from the assigned medication, as was the case in the TADS.19 Other authors have questioned whether the TADS may have under-reported adverse effects due to small numbers and patients leaving the study early.11 Use of ITT analyses will have led to underestimation of the frequency of psychiatric and other adverse events, a fact which has been little discussed.

Risk of suicidal behaviour was the only adverse effect that was addressed during all four treatment stages. Several psychiatric and physical adverse effects were reported during the first 12 weeks, but not mentioned in publications from the further treatment stages. Examples are sedation, insomnia, vomiting and upper abdominal pain, which occurred in more than 2% of patients in the first 12 weeks.12 The 2% occurrence is described as infrequent (≤5%), but should more correctly be classified as common.21 The risk of psychiatric adverse events such as mania, irritability, agitation and anxiety is given as 11% in the FLX group and 5.6% in the COMB group.12 In the review by Jane Garland et al, the occurrence of emotional/behavioural adverse effects is given as 10%–25%,3 but the numbers may not be comparable due to different inclusion criteria. Other adverse effects of SSRI treatment, such as appetite changes, weight changes and sexual problems, are not mentioned in any publication. Growth issues were not addressed. Changes in weight or appetite may have occurred without reaching the severity threshold. Sexual adverse effects may not have been relevant to many patients at the time due to their age, or may not have been forthcoming in interviews, especially as many patients were interviewed in the company of caregivers.12 Risk of sexual adverse effects was discussed in the adverse event monitoring protocol22 and procedures in case of pregnancies were established,23 so it is reasonable to assume a that certain proportion of patients were sexually active. Prolonged treatment into adulthood may well increase the relevance of such concerns.

To our knowledge, this is the first systematic assessment of adverse effects reporting in publications from the TADS. We conducted an extensive literature search and believe that all relevant studies have been identified, however, we cannot exclude the possibility that some publications may have been overlooked. Our findings regarding adverse effect reporting and potential for bias are based on analysis of only one study and do not give information on adverse effects reporting or bias in other studies of SSRIs in adolescents. However, discrepancies and weaknesses in the reporting of adverse events in such studies have also been noted by other authors.24 25 We have not had access to primary data.

A previous assessment of the adverse effects reporting in TADS focused on the occurrence of suicidal events and increased risk of suicidal behaviour10 and this is reflected in the most recent Cochrane review.1 Like Högberg et al,10 we have noted the misleading PBO group classification of patients with a suicidal event who were using FLX at the time. Our analysis encompasses all adverse events mentioned in publications from the TADS. Gaps and discrepancies in coding, transcription and reporting of harms in clinical trials have been reported, and the number of adverse events may differ between study reports and published papers.24 26 Several barriers to accurate harms reporting24 are relevant to the TADS, notably the severity threshold, conversions from continuous to dichotomous outcomes, individual judgements of association between event and medication, handling of adverse events in patients who discontinued treatment and the extensive use of concomitant medications. In future studies, the potential for bias may be substantially reduced by avoiding severity thresholds and defining a consistent method of describing adverse effects such as suicidal risk and mania score worsening. Occurrence or worsening of mania and other psychiatric adverse effects for individual patients should be reported in more detail. We would also suggest that if risk is presented as percentages, it should be calculated based on the number of patients who were receiving treatment at the time the adverse event occurred. This will be of particular importance in studies with large drop-out rates and treatment changes. The full spectrum of adverse effects should be reported for all study stages. A plan for data sharing should be in place to facilitate reanalysis and evaluation by other researchers, as practised by the BMJ.27

Due to its long duration (36 weeks) and follow-up (1 year), the TADS could have provided valuable information on the long-term occurrence of adverse effects both in frequency and severity. The adverse effects profile of FLX in the TADS has only been reported in detail for stage 1, where approximately 200 patients received FLX for 12 weeks. The raw data from the trial have been requested28 and planned for release into the public domain,29 but we have not been able to ascertain that these have been made publicly available. The incomplete reporting of adverse effects in a major study like TADS may lead to bias and erroneous conclusions regarding the safety profile of FLX when given to minors. The risk of suicidal behaviour has been the subject of many discussions and regulatory actions, but there has been considerably less focus on other clinically important adverse effects. This may have clinically important implications, since the benefit/risk estimations regarding FLX use in adolescents will be biased. If adverse effects are not acknowledged as such, there is a risk that symptoms may be misinterpreted and treated as more serious illnesses.