Principal findings
Physicians initiated pharmacotherapy with fluoxetine less than 20 % of the time, even after publication of the guidelines for youth in 2009. The percentage of first prescriptions for paroxetine decreased sharply after 2003, a trend which is most likely due to its particularly prominent association with suicidality in young people. Our results suggest that prescriptions for paroxetine were not replaced with fluoxetine, as the guidelines suggest, but with citalopram, which became the most popular antidepressant by 2004–2009. Although citalopram is effective for depression in adults [
22], it has not been shown to be effective in children and adolescents, in contrast to fluoxetine [
5], which is the only second-generation antidepressant registered for the treatment of depression in young people in the Netherlands and many other countries. Antidepressants may also be prescribed for anxiety, particularly in younger children, but no randomized placebo-controlled trial of citalopram for that purpose appears to have been conducted in children and adolescents, although fluoxetine has been found effective [
23,
24]. Among the SSRIs, citalopram has also been most strongly associated with QT interval prolongation (particularly at higher doses), which may increase the risk for torsade de pointes and sudden cardiac death [
25,
26] and which may be an additional safety-related reason, apart from treatment-emergent suicidality, to prefer fluoxetine as a first-line treatment.
The starting dose of antidepressants was generally higher than recommended. In particular, teens were usually prescribed an adult starting dose and were only rarely prescribed according to guidelines. Young children were prescribed according to the guidelines much more frequently (58 %), but 10 % of children were actually prescribed the adult starting dose, which is two to four times higher than the recommended dose. Few differences between prescribers were apparent, although specialists prescribed some SSRIs in slightly lower starting doses than GPs. This may be due to the slightly lower mean age of children receiving SSRIs from specialists compared to GPs.
Sertraline and citalopram were more likely to be prescribed according to the guidelines than other antidepressants. For sertraline, this is likely because the recommended starting dose is higher than that of other antidepressants, especially for older children. If we had used the stricter guideline rather than the more lenient guideline, adherence would have been markedly lower (23 % overall). For citalopram, the higher adherence to guidelines may be due to the availability of a liquid solution for citalopram, which facilitates low starting doses. In contrast, for fluoxetine, the tablet with the lowest dose currently available in the Netherlands contains 20 mg (1 DDD), which makes it difficult to provide the recommended dose of 5 mg. Although liquid fluoxetine was previously available, it is not currently on the Dutch market. The difficulty of providing low doses of fluoxetine may be one reason for physicians’ preference for citalopram.
Several positive findings were also apparent. While GPs prescribed the majority of antidepressants in 1994–2003, prescriptions shifted to specialists over time, as recommended by guidelines. We also found that the starting doses of some antidepressants, particularly citalopram, decreased over the study period, suggesting increasing awareness among physicians of the importance of low starting doses in young people. This finding agrees with a previous study in the USA showing increased prescription of low doses after the FDA warning in 2004 [
18]. Finally, maintenance doses were nearly always in agreement with the guidelines; where they were not, this was usually because the dose was lower than recommended. In general, maintenance doses were very similar to starting doses. Up-titration from a low starting dose is recommended in the guidelines, but titration occurred in a minority of cases, probably because the starting dose was already within the maintenance range. Up-titration was more likely for second-generation antidepressants like citalopram and venlafaxine, for which a relatively low starting dose was also more likely.
The number of young people initiating antidepressant treatment decreased in the early 2000s, followed by a return to the level of 2001. Such a trend was also found in countries like the UK [
13], but only to a slight extent or not at all in other countries, such as Canada [
27] or Denmark [
28]. The decrease in antidepressant initiation in young people was likely related to media coverage of the potential for treatment-emergent suicidality with antidepressant treatment [
15], but this effect appears to have been transient.
Improving guideline adherence
Adherence to guidelines is often poor [
29], and physicians’ prescription choices are influenced by a multitude of other factors besides guidelines and continuing medical education. These influences may include the mass media (which may have been especially important with regard to the reduction in prescriptions for paroxetine) [
15] and promotion by pharmaceutical companies [
30]. A large body of research has examined barriers and facilitators to the implementation of guidelines in clinical practice [
31‐
33]. Adherence is more likely when recommendations are specific and concrete rather than vague, when few additional resources are required for implementation, and when the evidence is strong and straightforward [
33,
34]. While the recommendation to initiate antidepressant treatment in children with fluoxetine is highly specific and does not require any additional resources, the evidence base for the use of fluoxetine in young people is relatively limited, although stronger than that for other antidepressants [
5], which may affect physicians’ confidence in the recommendation. Dedicated effort, for example implementation interventions [
35], may be needed to improve adherence to guidelines. A variety of interventions have been found to increase guideline adherence, including provision of educational materials, audit and feedback, and reminders, but effects are modest [
36]. Educational meetings, which are a common form of continuing medical education, also have small effects on improving guideline adherence [
37]. A better understanding of the reasons behind physicians’ preference for citalopram may help clarify how guideline adherence could be improved.
Strengths and limitations
This study has several strengths. First, use of a general population prescription database excludes the possibility of recall bias and selection bias. Another important strength is that we specifically examined first prescriptions, in contrast to many previous studies. Furthermore, we included a long time period of 21 years, which allowed us to examine time trends and the possible influence of major events, such as the recognition of a link between antidepressants and suicidality in young people in 2003–2004. This long time period also included very recent data (up to and including 2014).
Some limitations must also be acknowledged. An important limitation is that we did not have information about the indication for a prescription. As the guideline recommending fluoxetine is a guideline for the treatment of depression in young people, it may not apply to all prescriptions included herein. In particular, amitriptyline was frequently prescribed in children and adolescents (approximately 15 % of all prescriptions), even though tricyclic antidepressants are not recommended for the treatment of depression. Although we attempted to remove prescriptions for bed-wetting and pain, the remaining patients may still have been treated with amitriptyline for complaints other than depression. A study among Dutch GPs suggested that SSRIs were usually prescribed for depression or anxiety, but tricyclics were often prescribed for bed-wetting, hyperactivity, tension headache or non-specific disease, and only rarely for depression [
38]. Consequently, without information on the indication for amitriptyline prescriptions, it is difficult to determine whether these prescriptions were appropriate (although bed-wetting is the only approved indication for children and adolescents in the Netherlands). However, as the majority of SSRI prescriptions to children and adolescents are for the purpose of treating depression [
38], this limitation does not invalidate our finding that citalopram is preferred over fluoxetine, in contrast to the guideline. Furthermore, low starting doses are important regardless of the indication and might even be of greater importance if antidepressants are prescribed for the treatment of anxiety, the most probable alternative indication for SSRIs, given the potential for increased anxiety early in treatment [
39].
A second limitation of our study is that inpatient prescriptions are not included in the database. Consequently, some ‘first prescriptions’ may actually have been repeat prescriptions after treatment initiation during hospitalization. However, only 3–4 % of all children who are treated in specialist mental health care are hospitalized in a year [
40].