OFF-LABEL UTILIZATION OF ANTIDEPRESSANTS

Summary: While antidepressant prescription rules are established for approved indications by large scale studies, off-label utilization naturally often lacks the validation by large scientific databases, and is at its best based on expert consensus. The aim of the present survey was to study the prescription habits of hospital psychiatrists with regard to antidepressants, comparing patients treated for depressions and anxiety disorder with patients receiving off-label antidepressant treatment. Methods Data on drug use for this study were based on 6 reference days from April 1999 to November 2001 in the 98-bed psychiatrie hospital of the University of Lausanne, Switzerland. The drug prescriptions of 17 4 patients were assessed. Results Whereas the diagnosis did not influence the choice between newer or aider antidepressants, patients presenting an anxiety disorder were 4.5 times more likely (p<0.05) and patients with other diagnoses 8 times more likely (p<0.001) to receive an antipsychotic comedication compared to patients whose primary diagnosis was a depressive disorder. Also, patients receiving concomitantly a nonbenzodiazepine hypnotic were less likely to be prescribed an eider antidepressant (p<0.05). While patients with anxiety disorder and those with major depression received their antidepressants at comparable doses, patients with an off-label indication were treated preferentially with lower doses. The results of this survey suggest, that the prescribing hospital psychiatrists developed preferences with regard to the choice of the antidepressant class, which they then used for both registered and off label indications. They then seemed to adapt the dose and the comedication according to the diagnosis, confirming the initial study hypothesis. showing shorter treatment duration for off-label utilization, data comparing other aspects of antidepressant treatment, such as doses, comedications etc., is stiil lacking. The aim of the present survey is to study the prescription habits of hospital psychiatrists with regard to antidepressants, comparing patients treated for depression and anxiety disorder with patients receiving off-label antidepressant trèatment, examining particularly the administered doses and comedications. The underlying hypothesis was that physicians would choose the same class of antidepressants in off-label indications as they would for approved indications, but that they would adjust doses and comedication in response to the clinical situation. numerical ones exploratory differences between with chi-square (for nominal variables) and analyses of variance for numerical ones. models with multivariate logistic regression analysis. Binary stepwise logistic regression analysis was used to determine factors predicting the prescription of older antidepressants, of antipsychotics, of benzodiazepines and nonbenzodiazepine hypnotics. A further model predicting the antidepressant dose range, used was analyzed by multinominal stepwise logistic regression . Multinominal logistic regression broke the regression up into a series of binary regressions comparing each group to a baseline group, which we determined to be the low dose range group. The backward Wald method was applied.

is at its best based on expert consensus. The aim of the present survey was to study the prescription habits of hospital psychiatrists with regard to antidepressants, comparing patients treated for depressions and anxiety disorder with patients receiving off-label antidepressant treatment.

Methods
Data on drug use for this study were based on 6 reference days from April 1999 to November 2001 in the 98-bed psychiatrie hospital of the University of Lausanne, Switzerland. The drug prescriptions of 17 4 patients were assessed.

Results
Whereas the diagnosis did not influence the choice between newer or aider antidepressants, patients presenting an anxiety disorder were 4.5 times more likely (p<0.05) and patients with other diagnoses 8 times more likely (p<0.001) to receive an antipsychotic comedication compared to patients whose primary diagnosis was a depressive disorder. Also, patients receiving concomitantly a nonbenzodiazepine hypnotic were less likely to be prescribed an eider antidepressant (p<0.05).
While patients with anxiety disorder and those with major depression received their antidepressants at comparable doses, patients with an off-label indication were treated preferentially with lower doses.

Conclusions
The results of this survey suggest, that the prescribing hospital psychiatrists developed preferences with regard to the choice of the antidepressant class, which they then used for both registered and offlabel indications. They then seemed to adapt the dose and the comedication according to the diagnosis, confirming the initial study hypothesis.
showing shorter treatment duration for off-label utilization, data comparing other aspects of antidepressant treatment, such as doses, comedications etc., is stiil lacking.
The aim of the present survey is to study the prescription habits of hospital psychiatrists with regard to antidepressants, comparing patients treated for depression and anxiety disorder with patients receiving off-label antidepressant trèatment, examining particularly the administered doses and comedications. The underlying hypothesis was that physicians would choose the same class of antidepressants in off-label indications as they would for approved indications, but that they would adjust doses and comedication in response to the clinical situation.

Methods
The present study was developed as part of the AMSP project (Arzneimittelsicherheit in der Psychiatrie = drug safety in psychiatry), which is a program for continuous assessment of adverse drug reactions in psychiatrie inpatients under naturalistic conditions of routine clinical treatment. The methodology has been described elsewhere (6,21,23,24). Currently, more than 35 German, Swiss, Austrian, Hungarian and Belgian sites are participating. Data on drug use in the participating hospitals are based on two reference days per year. Ali drugs given on a reference day are recorded along with age, sex, and diagnosis (ICD-10) for ail patients under surveillance. The daily dosage is also recorded.
Data of the present study were drawn out of 6 reference days from April 1999 to November 2001 in the 98-bed psychiatrie hospital of the University of Lausanne, Switzerland. Currently, the mean hospitalization duration is 14 days and the nurse/bed ratio is 0.95.

Definition of drug classes
Antidepressants were classified into tricyclic drugs (TCA) and drugs of the newer generations (DNG).
DNGs were defined as the selective serotonin re-uptake inhibitors paroxetine, sertraline, fluoxetine, citalopram and fluvoxamine as weil as mirtazapine, reboxetine, nefazodone and moclobemide. The class of TCAs included the tricyclic compounds clomipramine, trimipramine, imipramine, amitriptyline and dibenzepin.
..  Benzodiazepines were classified as one group, including sedative and hypnotic drugs, as sedative benzodiazepines were often also used as hypnotics, the different indications therefore being difficult to assess .. Nonbenzodiazepine hypnotics (NH) formed a further drug class, including zolpidem, zopiclone and zaleplon.

Dose ranges
ln order to compare the antidepressant prescription with regard to applied doses, three dose ranges were defined for each administered drug (Table 1 ). For patients receiving more than one antidepressant, the doses were summed up: twice the low dose range giving a middle dose range, a low dose and a middle dose or two middle doses giving a high dose.

Label and off-label indications
Based on the primary diagnosis, patients were grouped into 3 classes: (1) Depression, (2) Anxiety Disorder, and (3) Other Disorders.
As no patient had a manie or mixed episode as primary diagnosis, ail patients presenting an ICD-10 diagnosis of affective disorder were depressive.
Whereas ail used antidepressants are approved in Switzerland for depression, they differ with regard to the approval for the different anxiety disorders. Anxiety disorders were nevertheless grouped into one class, as knowledge of the diverse approved indications among the prescribing physicians cou Id not be assumed. On the other hahd, physicians are locally trained to consider newer antidepressants as the first line pharmacotherapy for anxiety disorders.
Analyses ln descriptive data analyses means and standard deviations were calculated for numerical variables while nominal ones frequency categories values and percentages are reported. ln exploratory analyses, the differences between groups were tested with chi-square tests (for nominal variables) and analyses of variance for numerical ones.
Predictive models were built with multivariate logistic regression analysis. Binary stepwise logistic regression analysis was used to determine factors predicting the prescription of older antidepressants, of antipsychotics, of benzodiazepines and nonbenzodiazepine hypnotics. A further model predicting the antidepressant dose range, used was analyzed by multinominal stepwise logistic regression .
Multinominal logistic regression broke the regression up into a series of binary regressions comparing each group to a baseline group, which we determined to be the low dose range group. The backward Wald method was applied.
The data were analyzed using the SPSS for Windows program, version 12.0.

Characteristics of the sample
The drug prescriptions of 17 4 patients were assessed. The mean age was 42.6 ± 11.6 (range 20 -66) . and the proportion of women 59.2 %. There were no differences between index days with regard to age and sex distribution.  Table 3 shows the result of the stepwise logistic regression, with the choice of an older antidepressant being the dependent variable and the index year, the diagnostic class, sex, age and the prescription of concomitant drugs (benzodiazepines, nonbenzodiazepine hypnotics, anticonvulsants, lithium, somatic drugs) as independent parameters. With increasing age the probability of receiving an older antidepressant increased (p<0.05). Furthermore, a clear secular trend appeared, as the probability of receiving an older antidepressant was significantly lower in 2000 (p<0.05) and in 2001 (p<0.05) compared with 1999. Also, patients receiving concomitantly a nonbenzodiazepine hypnotic were less likely t6 be prescribed an older antidepressant (p<0.05). The diagnostic class was not retained as a predictive parameter. The positive predictive value of the mode! was 85.1 %.

Prescription of concomitant antipsychotic drugs
The correspondent result of the stepwise logistic regression analysis is shown in Tab. 4. The following parameters were entered: Index year, sex, age, diagnostic class, antidepressant type (older vs. newer) and concomitant treatments. Only the diagnostic class was retained, patients with anxiE:;ty disorder or other non-depressive disorders presenting significantly higher odds-ratios compared to patients with the diagnosis of depression. The positive predictive value of the definitive mode! was 73.0 %.

Prescription of concomitant benzodiazepines
A stepwise logistic regression predicting the prescription of benzodiazepines including the following parameters was computed: Sex, index year, age, diagnostic class, antidepressant type (older vs. newer and sedative vs. non-sedative), and concomitant treatments. None of the parameters was retained.

Prescription of concomitant nonbenzodiazepine hypnotics
The following parameters were used to compute the corresponding stepwise logistic regression (Tab.  Furthermore, those receiving an older antidepressant had an odds ratio of 6.66 to receive their antidepressant treatment rather at a high than at low dose, compared to the patients having a · treatment with newer antidepressants only (p<0.01 ).

Discussion
The aim of ~he present study was to investigate if hospital psychiatrists would prescribe antidepressants and comedication differently for patients with a registered indication compared to those administered antidepressants on an off-label basis.
Whereas the diagnosis did not seem to influence the choice of a newer or and aider antidepressant, it was significantly associated with the prescription of antipsychotic comedication and antidepressant dose. Patients presenting an anxiety disorder were 4.5 times more likely and patients with other diagnoses 8 times more likely to receive an antipsychotic comedication compared to patients whose primary diagnosis was a depressive disorder. While patients with anxiety disorder and those with major depression received their antidepressants at comparable doses, patients with an off-label indication were treated preferentially with lower doses.
While antipsychotics, on one hand, high potency as well as low potency, have often been advocated as being valid alternatives to benzodiazepines as anxiolytic treatments, and, on the other hand, sedative antidepressants have been suggested to reduce the need of concomitant benzodiazepine use, our data did not confirm these assumptions. Neither the use of sedative antidepressants, nor the prescription of benzodiazepines had an impact on the prescription of concomitant antipsychotics.
Antipsychotics were only prescribed in the function of the diagnosis.
A secular trend was found with regard to the prescription of tricyclic antidepressants, these older drugs being less likely to be prescribed in 2000 and 2001 compared to 1999, even when Checking for confounding factors like diagnosis or comedication. As the prescription of a nonbenzodiazepine hypnotic was associated With a lowered odds of receiving an older antidepressant, one could hypothesize that the observed secular trend may be due to the replacement of tricyclics used for their sedative or sleep-enhancing properties by nonbenzodiazepine hypnotics.
As in previous studies investigating prescription habits of psychotropic drugs (1,9,11,12,17), polypharmacy was highly prevalent, the mean number of prescribed drugs being 4 with a range from 1 to 1 O drugs per patient. Whereas polypharmacy has long been considered a malpractice in earlier studies, it has become increasingly apparent nowadays that psychiatrie polypharmacy can have some advantages, i.e., to further improve sleep, have a more potent anxiolytic or sedative effect and to overcome therapy resistances (17,20,22).
One further intriguing finding was the observation that the risk of receiving a tricyclic antidepressant augmented with increasing age. This is at first sight counter-intuitive, as the risk of cardiovascular and other side effects may be more important in eider patients and should have discouraged the doctors from prescribing these drugs for those patients. On the other hand, eider patients may be often the more chronic patients and may have received the antidepressant for longer habituai treatment, which the physician could possibly not have had the courage to change anymore. As the present data are only collected on a crossover basis, this assumption can, however, not be tested out.
The results of this study need to be viewed against their methodological limitations. The method of index day assessments allows only cross sectional analyses for each patient. The course of the prescription, i.e., titration schemes, and the duration of the antidepressant treatment as well as of the comedièations cannot be evaluated. That not every patient can be considered as having been already stabilized on the medication at the index day, especially data on doses, but also all other data, has to 12 ..
• .. be interpreted with caution. Also, diagnoses were not based on a structured interview, which may have diminished their validity. However, as the ICD-10 diagnoses were derived post hoc from the medical records, they surely better reflect clinical practice. Furthermore, only the first diagnosis (hospitalization indication) was retained.
The two reference days were always fixed for the period of late spring and late autumn, allowing each center to choose a day within ± 7 days. The choice of seasonally fixed reference days may be a confounding factor in depressive disorders as well as anxiety disorders.
With regard to generalizability, one has to consider that the survey was limited to one site only, the hospital being University based and organized in specialized wards. The refuse, biases by local customs and conditions of practice cannot be excluded.
ln conclusion, the present results thus suggest that the prescribing hospital psychiatrists developed preferences with regard to the choice of antidepressant class, which they then used for both registered and off-label indications. They then seemed to adjust the dose and the comedication according to the diagnosis, confirming the initial study hypothesis .
Factors influencing this spreading out of prescription habits from labeled to off-label indications remain to be explored. The results furthermore generate various related important questions, which should be addressed through further research: Does the rapid spreading out of off-label prescription increase or decrease the risk of adverse drug reactions? Does it unduly increase or decrease costs? What is the impact of doctors' interaction with pharmaceutical companies? .,.