Clinical characteristics associated with the prescribing of SSRI medication in adolescents with major unipolar depression

IMPACT is a pragmatic randomised control trial which aims to determine if one of two specialist treatments, cognitive behavioural therapy (CBT) or short term psychoanalytic psychotherapy (STPP) are superior to  a reference practice-based brief psychosocial intervention (BPI), at preventing recurrence of clinically meaningful levels of depressive symptoms indicating potential relapse in adolescents who enter the trial with MD [6].

The study was conducted over three regions within the UK, East Anglia, North London and the North West of England incorporating 16 routine Child and Adolescent Mental Health (CAMH) clinics. Individuals were recruited to the study with moderate to severe depression and aged between 11 and 18 years. Exclusion criteria included generalised learning difficulties or a pervasive developmental disorder, pregnancy, currently taking another medication that may interact with an SSRI and being unable to stop this medication, substance misuse, primary diagnosis of Bipolar Type I disorder, schizophrenia or eating disorder. Of the 561 individuals who were referred and had baseline assessments, 470 were considered eligible with 5 later withdrawing consent, resulting in the recruitment of 465 overall. As part of the screening process prior to enrolment, individuals were asked if their current depressive illness was a first episode or a relapse. At this point, demographic information including their ethnicity was collected. Ethical approval was by the Cambridgeshire 2 research Ethics Committee, Addenbrooke’s Hospital Cambridge, UK and is therefore, in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. Follow up was undertaken with repeated reassessments at up to 86 weeks after randomisation to evaluate recurrence of clinical level depressive symptoms. The further characterisation of the sample will be available with the publication of the IMPACT trial during 2016 [7].

Antidepressant prescribing prior to entering the study was determined using the Child and Adolescent Service Use Schedule (CA-SUS) [8]. The CA-SUS is an interview-based measure to collect data on service use and was designed for use within mental health populations, including young people with depression [8]. As well as ascertaining the current prescribing of psychotropic medication, it includes information on the individual’s accommodation, education, use of hospital and community based health and social services, and any criminal activity or criminal justice sector contacts. Baseline prescribing information was available for 457 individuals of the 465 recruited to the study (98 %). Subsequent analyses presented here excluded participants for whom this information is not available.

We hypothesised that SSRI prescribing may be associated with depression severity, psychiatric comorbidity or chronicity. This was investigated using the K-SADS-PL. With regard to depressive symptoms, this refers to the 22 items used to establish the diagnosis of current depression in the K-SADS-PL. A score of 3 or ‘threshold’ within the K-SADS-PL was used to define a symptom being present. The comorbid diagnoses considered, according to DSM-IV criteria, were panic with and without agoraphobia, separation anxiety, avoidant disorder, specific phobias, general anxiety disorder, obsessive compulsive disorder, post-traumatic stress disorder, anorexia nervosa, bulimia nervosa, attention deficit disorder, oppositional defiant disorder, conduct disorder, alcohol abuse and substance and alcohol dependence. Psychotic symptoms were also recorded but individuals with a diagnosis of mania or schizophrenia were excluded from the study. These data were available for all the trial participants. Using the K-SADS-PL, the age of onset of depressive illness was also compared between groups [data available for 282 individuals (62 %)]. The K-SADS-PL has been shown in numerous studies to have high inter-rater reliability and construct validity [9].

The MFQ was also used to investigate if depression severity was associated with SSRI prescribing prior to randomisation in the IMPACT study. It is a 33 item self-report measure of depressive symptoms consisting of a list of descriptive phrases with subjects being asked to rate their symptoms over the past 2 weeks. Re-test reliability and criterion validity are reported to be high [11]. Cronbach’s alpha within our study was 0.92. Items are scored from 0 to 2 (never = 0, sometimes = 1, mostly or always = 2), yielding a maximum score of 66. These data were available for 462 out of the 465 study participants (99 %).

The HONOSCA is a measure of a young person’s total mental health problems. It measures 15 items including psychiatric symptoms, behaviours, family, social and school functioning. Each item is scored from 0 to 4 where 0 indicates no problems and 4 very severe problems. It is known to be reliable, sensitive to change and correlates well with the clinician’s judgement of the young person’s outcome [13]. HONOSCA data were available for 432 of the study participants (93 %).

The EQ-5D was used to investigate whether SSRI prescribing prior to randomisation in the IMPACT study was associated with functional impairment or quality of life. It is a standardised measurement of health status and health-related quality of life. The EQ-5D-3L version was used which consists of five dimensions (mobility, self-care, usual activities, pain/discomfort and anxiety/depression), recorded on one of three levels (no problems = 1, some problems = 2, extreme problems = 3). This allowed individuals to be classified into one of 243 health states. For example, 11,111 representing no problems in any dimension and 33333 representing extreme problems in all five dimensions. A summary index value was derived by applying general population weights to each health state [15]. The summary scores range from 1 (perfect health) to 0 (death) with negative scores indicating states considered to be worse than death. The EQ-5D was available for 436 of the individuals (94 %).

The RTSHIA, a self-report instrument for assessing life-time self-harm and risk taking behaviour in adolescents, was used to investigate whether SSRIs are prescribed to adolescents based on hazardous or risk-taking behaviour. The version used differs from the original in that 3 questions were omitted regarding risky sexual behaviour. The version used contained 31 items answered on a Likert scale (3 = many times, 2 = more than once, 1 = once, 0 = never) and scored as a total for hazardous behaviour ranging from 0–93. Subscales for risk taking and self-harming were examined as well as the total score. Examples of questions that related to risk taking but not self-harm included drug and alcohol use, staying out late without parental knowledge and actively placing oneself in risky situations such as cheating at school and shop lifting. Cronbach alpha scores for each component of the RTSHIA within our study were; total = 0.90, risk taking = 0.80, self-harming = 0.90. This was available for 436 individuals of those entered into the trial (94 %).

We also considered if antisocial behaviour may be considered a hazardous behaviour associated with prescribing. This was investigated using the Behavioural Checklist, an 11-item screen for current symptoms of antisocial behaviour derived from DSM-IV criteria and converted to a self-report format. It is scored on a 4 point Likert scale (always = 3, mostly = 2, sometimes = 1, never = 0) with a range of 0–44. Within our study, Cronbach alpha for the Behavioural Checklist was 0.79. The Behavioural Checklist was completed by 450 of the study participants (97 %).

Data regarding antidepressant prescribing prior to randomisation were available for 457 out of the 465 individuals recruited to the IMPACT trial. Of these, 89 (19.5 %) had been prescribed an antidepressant before entering the trial. The only antidepressants that had been prescribed were SSRIs with the majority (83 %) taking fluoxetine, as per NICE guidelines. The remaining individuals were prescribed citalopram (10 %) or sertraline (6 %). One individual was documented as taking an SSRI but the type of SSRI was not reported.

Prior to entering the study, individuals were asked if this was their first episode of depression. This information was available for 396 (85 %) of the study’s participants. There was no difference in any of the demographic variables (age, gender, ethnicity, MFQ score, age of onset of depression, number of depressive symptoms and number of comorbid disorders) measured between those for whom this information was available and those for whom it was not. Of those with available information, 370 (93.4 %) reported their current depressive illness as a first episode, the rest reporting a recurrence. Of the 77 individuals prescribed antidepressants for whom screening information was available, 73 reported their current depressive illness as being the first episode (94.8 %).

As shown in Table 1, individuals did not differ with regard to age, gender or ethnicity (white vs non-white) regardless of whether or not they were prescribed SSRIs. There was, however, a main effect of treatment centre, which will be discussed in more detail below.

The severity of the young person’s current depressive illness was compared between those who had been prescribed antidepressants before entering the IMPACT study and those who had not been using both observer and self-rated measures (Table 2).

No significant difference between groups was observed in self-reported depression scores, the number of interview reported depressive symptoms (K-SADS PL), or the number of comorbid disorders associated with their depressive illness. There was also no difference in the age of onset of depressive illness obtained from the K-SADS PL.

Individuals who had been prescribed an SSRI reported a lower level of health-related quality of life than those who were not taking antidepressants, as measured using the EQ5D (Table 2). They also reported higher HONOSCA scores, suggesting an overall poorer mental health status (Table 2). Together, these findings suggest that individuals who had been prescribed SSRIs subjectively perceived themselves as more functionally impaired.

Individuals who had been prescribed antidepressants had higher self-reported lifetime histories of hazardous behaviour overall, as measured using the RTSHIA (Table 2).

When analysis of hazardous behaviour was split into risk taking and self-harm components, there was no difference between the groups with regard to risk taking behaviour, however, those who had been prescribed antidepressants had higher lifetime histories of self-harming behaviours.

This hazard-related prescribing led us to the question of whether other forms of risky but non-depressive behaviour were also associated with antidepressant prescribing (Table 2). Surprisingly, individuals who had been prescribed antidepressants before entering the IMPACT study had fewer symptoms of self-reported antisocial behaviour compared with those who had not been prescribed antidepressant medication. It should be noted, however, that the median scores for each group are low in both, given the scale range (0–33).

Prescribing rates of SSRI antidepressants differed significantly across the three UK research centres from which they were recruited (Table 1). Descriptively, 30 % of individuals recruited from East Anglia had been prescribed SSRIs prior to entering the study, compared with 11.1 and 13.9 % from London and the North West of England, respectively. Participants in East Anglia were, therefore, 2.7 (z = 3.4, p = 0.001) times more likely to be prescribed an SSRI compared with those recruited from the other two research centres.

These prescribing differences, however, did not explain the observed relationship between medication prescribing, quality of life/functional impairment and risky behaviour (Table 3). The pattern of quality of life, self-harming and antisocial behaviour scores were the same across research centres with the group who had been prescribed SSRIs prior to entering the study exhibiting lower EQ-5D scores (poorer health-related quality of life), higher HONOSCA scores (poorer overall mental health outcomes), higher self-harm/suicidality levels and lower antisocial behaviour levels, compared with those who had not been prescribed antidepressant medication.

We tested for potential effects of centre on sample characteristics using a multinomial logistic regression (Supplementary Table 1). This showed that individuals recruited from London and the North West displayed lower self-harm scores compared with those recruited from East Anglia [median score (IQR): East Anglia: 16.0 (9.0–26.0), London: 13.0 (5.0–21.0), North West: 10.0 (4.0–20.5)]. Individuals from London were also slightly older [median age (IQR): East Anglia: 15.5 (14.7–16.3) years, London: 16.1 (14.8–17.2) years, North West: 15.6 (14.6–16.7) years] and more likely to be of a non-Caucasian ethnicity (East Anglia: 90 % Caucasian, London: 53 % Caucasian, North West: 84 % Caucasian). Adolescents from the North West had significantly lower HONOSCA scores compared with the East Anglian population [median score (IQR): East Anglia 19.2 (16.0–24.9), London: 20.0 (14.4–25.8), North West: 17.0 (13.0–21.3)].

A logistic regression using gender as an interaction term allowed for any potential effects of gender on each of the univariates predicting SSRI prescribing to be investigated (Supplementary Table 2). Depression severity (as measured by MFQ) was added to the model as the difference between the two groups approached significance. Age was also added to the model in case this was shown to be a prescribing predictor when other interactions were controlled for. This showed that SSRI prescribing appeared to be more likely in girls with higher self-harm scores and boys with greater depressive symptoms (as measured by MFQ). This was confirmed by a further logistic regression analysis examining for main effects and interactions between SSRI prescribing and the possible predictors; quality of life, total mental health problems (as measured by the HONOSCA), self-harm, antisocial behaviour, depression severity, research centre and ethnicity. Given the findings from our first model, this analysis was undertaken for boys and girls separately and the results are presented in Table 4. They confirmed that the main predictor for prescribing differed based on gender. For girls, prescribing was associated with a history of self-harming behaviour. In contrast, prescribing for boys was associated with depression severity at randomisation. The area boys were recruited from also appeared to play a role as to whether they were prescribed SSRIs or not, they were more likely to be prescribed antidepressants if they came from East Anglia as opposed to London or the North West of the UK.