The scheme for THIN to obtain and provide anonymous patient data to researchers was approved by the National Health Service South-East Multicenter Research Ethics Committee (MREC) in 2002 and scientific approval for this study was obtained from CMD Medical Research's Scientific Review Committee in March 2011.

Approximately 98% of the population in the UK is registered with a general practitioner [15]. The Health Improvement Network (THIN) database is one of the largest national collections of primary care data and is broadly representative of the general practice (GP) population in terms of demographics and consultation behaviour [16]. Participating general practitioners from 497 practices enter clinical information on patients, including demographics data, diagnoses, and prescriptions so as to offer a longitudinal medical record for each patient which is available to researchers as anonymised data [17]. Clinical diagnoses recorded by GPs on THIN have been shown to be accurate compared with other reliable sources [18]. The database provides the Townsend score measure of deprivation, a composite measure of social deprivation in quintiles (owner occupation, overcrowding, car ownership, and unemployment) [19]. It is based on patient postal code and linked to UK census data from 2001 for approximately 150 households in that postal area. We analysed data from 1995 to 2009.

We identified a cohort of children aged up to 18 years who were registered with a General Practice which was a part of THIN for at least six months between January 1995 and December 2009. Children entered the cohort when they registered with a General Practice, or, the date when their practice joined the THIN scheme and met standards for acceptable levels of data recording [20]. Children remained in the cohort until aged 18 years, transfer out of the practice, date of death or date of last data collection from the practice.

The Jointpoint analysis suggested for SSRIs as a group, there were two time points were prescription rates changed: 2002 and 2005. Up to 2002 prescription rates for SSRIs had been significantly increasing nearly 16% from 1995–2002 (Table 3, Figures 4A & 4B). However, between 2002 and 2005 the rates were stagnant, followed by a significant increase of nearly 11% from 2005 to 2009 (Table 3).

Individual SSRIs followed a similar pattern: fluoxetine, citalopram, paroxetine and sertraline rates all were increasing from 1995 to the early 2000s before showing a temporary decrease, or stall, in prescription rates. Paroxetine was the only SSRI which showed a statistically significant decrease in prescription rates. Rates for citalopram and sertraline started increasing again in 2005, while rates for fluoxetine and paroxetine remained stable (Table 3).

In contrast, rates for TCAs as a group showed a significant decrease between 1995 and 2006, after which there was no significant change. Rates for amitriptyline prescriptions showed a moderate decrease between 1995 and 2006, but started to increase after this point. Imipramine prescription rates showed a steady decline over the entire period.

To our knowledge, this is the first large paediatric database study to compare depression diagnoses, symptoms and antidepressant prescriptions and the effects the CSM advice over a longer time period in the UK. We have found that prescription rates of SSRIs as a group decreased from 3.2 per 1,000 person-years in 2002 to 1.7 per 1,000 person-years in 2005. More specifically, rates for contra-indicated SSRIs, i.e. citalopram, paroxetine and sertraline, went down during that period, while rates for fluoxetine remained stable and rates for TCAs were not affected. The decline in prescription rates was sharpest for paroxetine. Similar to SSRI prescription rates, rates for depression diagnoses entries decreased from 3.0 per 1,000 person-years in 2002 to 2.0 per 1,000 person years in 2005. Depression symptom recording saw a steady increase over the study period, increasing from 1.0 per 1,000 person-years in 1995 to 4.7 per 1,000 person-years in 2009. Finally, rates for SSRIs as group and citalopram in particular, were increasing after 2005.

The decrease in recording of both depression diagnoses and antidepressant prescriptions after 2002 could indicate caution on the part of GPs in diagnosing depression and prescribing antidepressants following the CSM advice. Moreover, GPs might prefer to record depression symptoms rather than diagnose a child as depressed. The decrease in contra-indicated SSRIs as opposed to fluoxetine strengthens the possibility of a link with the CSM advice.

Although the Jointpoint program points to 2002 as the time point where SSRI rates changed, the observed data shows that rates did not decrease until after 2003, the year when the CSM advice was issued. However, prescription rates could have started decreasing prior to the CSM advice as information regarding the safety and effectiveness of SSRIs was circulating in the specialist community before the advice was issued and could have influenced changes in prescription recommendations. Similarly, the program does not qualify the small dip in fluoxetine rates around 2005 as statistically significant. This dip might be related to the requirement of the US Food and Drug Administration (FDA) to add a black box warning to all antidepressants, including fluoxetine and TCAs, about an increased risk of suicidal behaviour in 2004 [12].

The sharp decline in paroxetine prescription rates could be related to the advice by the MHRA against the prescribing of paroxetine specifically. This advice was issued in June 2003, preceding the overall SSRI advice in December of the same year. It followed a review of randomised controlled trials that showed higher rates of suicidal thought and behaviour (but not completed suicides) in patient who took paroxetine (25 out of 738; 3%), compared to those who took placebo (8 out of 647; 1%; p for difference = 0.01) [8].

The BBC programme that initially started the controversy implied that paroxetine (brand name Seroxat) was addictive, had severe withdrawal symptoms and could increase the risk of suicidal behaviour [28]. This might have led to patients being biased against taking paroxetine as a first line of treatment, and making it difficult to determine whether the sharp decline in paroxetine prescriptions in primary care was due to a negative public opinion of the drug in response to the issues raised in the programme, or the advice issued by the CSM half a year later.

The increase in SSRI prescription rates after 2005 could indicate that concerns about a possible suicidality risk associated with SSRIs have waned. Several studies found no increased risk of suicidality for SSRIs [29], [30], or increases in suicide rates that coincided with decreases in SSRI prescription rates [31]. In the US, 2004 saw the largest single-year increase in suicide rates in adolescents aged 10–19 years [32]. From 1990 to 2003 suicide rates had been decreased by 28%, but in 2004 they had increased with 15% from 6.78 to 7.32 per 100,000 people. This might have led GPs to revaluate prescribing SSRIs to children and adolescents, although doubts continue to exist regarding the safety and effectiveness of SSRIs [33], [34].

The antidepressant prescription rates we found are similar in size and trend to those found by a General Practice Research Database study which studied prescription rates from 1992 to 2001 [1]. We also found similar age and gender effects. Our results also confirm findings by Murray et al. who found a decrease in SSRI prescribing in between 2003 and 2004, while prescribing rates for fluoxetine remained stable [13]. However, the study by Murray et al. did not assess data for individual drugs, apart from fluoxetine, whereas our study did take different SSRIs into account.

A study based on Australian data also found a decrease in antidepressant use in children, in particular of SSRIs [14]. They also saw a sharp rise of fluoxetine over time, which we did not find in the UK. This might be explained by sertraline being the most commonly prescribed antidepressant in Australia before the SSRI controversy started, whereas fluoxetine was already the drug of first choice in the UK before the CSM advice.

The main strength of this study is its sample size that enables examination of outcomes separately for girls and boys, and by drug. There is no clear reason to believe the results would differ for the entire population of UK children.

However, there are also limitations. In using data from general practices, few children might have been missed out if their depression was not severe enough to warrant a visit to a GP, or if they were diagnosed outside a general practice setting, e.g. by a child psychiatrist. However, a study on depression in adults found that although incidence rates in the THIN database are lower than depression rates found in epidemiological studies, associations with covariates such as gender and deprivation were similar [35]. Also, non-psychiatric physician's recognition of depression has been found to have a limited sensitivity, but a high specificity [36], [37]. Childhood depression rates might have been underestimated in this study, but trends and associations with other variables are likely to be representative of the general population. Moreover, as we were specifically interested in the effects of the CSM advice in primary care settings, this limitation will have only a minimal effect on our results.

While data on prescriptions is available in the THIN database, there is no information on dispensing and treatment compliance. Thus the antidepressant prescription rates we found might not reflect antidepressant use. However, we aimed to study prescription rates in primary care, so this does not affect our estimates.

After 2002, general practitioners decreased their prescribing of contra-indicated SSRIs, particularly paroxetine. Rates for fluoxetine, the only SSRI not to be contra-indicated, remained stable. Depression diagnoses mirrored prescription rates and decreased between 2002 and 2005, suggesting caution on the side of GPs. The timing and direction of these trends imply that GPs followed the CSM advice, although it cannot be ruled out that these trends resulted from the negative media attention SSRIs received around the same time. After 2005, rates for all antidepressants, except paroxetine, started recovering. This is in line with results from observational studies that found no increased risk of suicidal behaviour with SSRIs.