Treatment of Depression: Are Psychotropic Drugs Overdosed in Women and in the Elderly? Dosages of Psychotropic Drugs by Sex and Age in Clinical Routine

Psychotropic drugs show higher concentration-dose ratios in women and in the elderly. Furthermore, women partially exhibit more adverse drug reactions than men. Lower dosages in most psychotropic drugs may be recommended in women and in the elderly. This study describes prescribed dosages of psychotropic drugs in depressive patients across sexes and age groups in clinical routine. Method: In 32’082 inpatients with depressive disorders (data acquired by the European drug safety program AMSP), the inuence of sex and age on prescribed dosages are analysed for the 10 most commonly prescribed drugs in our dataset (descending in the frequency of prescription: mirtazapine, venlafaxine, lorazepam, quetiapine, citalopram, escitalopram, olanzapine, duloxetine, zopiclone and sertraline) and additionally zolpidem. Confounding variables for sex differences, such as severity of depression and age distribution are examined. Dose decline in patients over 65 years is calculated. The observed sex and age differences in prescriptions are compared to differences in pharmacokinetic literature.

Although women are the main users of psychotropic drugs [30], they are underrepresented in treatment studies, and therefore treatment guidelines are mainly based on ndings in male patients [31]. According to Sørup et al. [32], the in uence of patient's sex on ADRs is an understudied factor.
In 2013, the U.S. Food and Drug Administration (FDA) approved label changes specifying new maximal dosing recommendations for zolpidem, being 5 mg/day for females and 10 mg/day for males [33,34]. According to the FDA, higher dosages could lead to next-morning impairments, especially driving impairment [33,34]. However, this guidance is controversial. It has been argued that a disturbed sleep by underdosed zolpidem in females may impair driving even more than the intake of the 10 mg/day dose [18]. This controversy motivated our inclusion of zolpidem in our study.
In this study, a large data set from a pharmacovigilance project is used to test whether in clinical routine the dosage of psychotropic drugs in the treatment of depression is lower in females than in males. We selected the 10 most prescribed drugs in our dataset and additionally zolpidem. Furthermore, we also analyse to which extent the dosages of psychotropic drugs are in uenced by age in general and within the group of elderly.

Data source
The analysed prescription data of the present study was gathered in the project "Arzneimittelsicherheit in der Psychiatrie" (AMSP; drug safety program in psychiatry). AMSP is an ongoing European multi-center drug safety program which collects data on psychopharmacotherapy and severe adverse drug reactions from psychiatric hospitals in a naturalistic setting since 1993. AMSP´s pharmacovigilance methods have been described in detail previously [35][36][37]. Brie y, AMSP consists of two principal data collections (prescription data and severe adverse drug reactions) from 116 hospitals so far in Germany, Switzerland and Austria, as well as temporarily from one hospital each in Belgium and Hungary. The number of participating hospitals increased from nine in 1994 to 63 in 2015. In a cross-sectional approach, all participating hospitals record drug prescriptions for all inpatients under surveillance on two reference days per year. All drugs administered on these days are assessed along with the patients' age, sex and leading psychiatric diagnosis. Evaluations of the AMSP database have been approved by the Ethics Committee of the University of Munich and the Ethics Committee of the Hannover Medical School (Nr. 8100 BO S 2018). This study adheres to the Declaration of Helsinki and its later amendments. The AMSP program is a continuous observational post marketing drug surveillance program and does not interfere with the ongoing clinical treatment of patients under surveillance. Sample description 32'082 patients older than 17 and younger than 90 years of age, hospitalized between 2001 and 2015, with a leading admission diagnosis of major depressive disorder (MDD) are investigated. The dosages of the 10 most prescribed drugs and zolpidem are analysed (descending in the frequency of prescription): mirtazapine (n=10'431), venlafaxine (n=8'072), lorazepam (n=7'757), quetiapine (n=6'993), citalopram (n=3'909), escitalopram (n=3'842), olanzapine (n=3'441), duloxetine (n=3'070), zopiclone (n=2'936), sertraline (n=2'909) and zolpidem (n=1'873). The sample consists of 11´887 (37.1%) males and 20´195 (62.9%) females. Note that a patient may have had more than one psychotropic drug prescribed. Table 1 shows a summary of the sample composition. When analysing prescribed dosage differences between sexes, it is important to identify possible confounding variables that could account for distortions in the analysis. In this study, we examined the following possible confounding variables: severity of MDD, differences along calendar years and differences in age distribution.
In gure 1, the number of male and female patients along calendar years are depicted. We do not see any substantial difference in the ratio between male and female patients over time. Table 2 shows no substantial differences in the severity of cases for male and female patients. Table 3 gives the severity of cases for young and elderly patients. In the small group of psychotic MDD (13.4%), there is a higher percentage in the elderly than in younger patients (18.0% versus 12.2%). Nevertheless, we assume that severity and differences over years cannot distort our analysis to a signi cant extent. On the other hand, the age distributions (depicted in gure 2) differ substantially. In particular, there are proportionally more elderly women than elderly men. Therefore, we need to adjust our analysis in order to take into account differences in age distributions.

Statistical analysis
The statistical analysis was performed using R version 3.6.2. To assess the in uence of age in the decision of prescribed dosages, the slope of a linear t among patients older than 65 and younger than 90 years of age is computed. To estimate its signi cance, a t-test is performed on the coe cient of the slope under the hypothesis to be zero. Along with the slope, also the percentage change in relation to the previous year is computed, which is calculated from the coe cient of an exponential t. Since in pharmacokinetic literature the age differences are often compared between age groups of up to 65 years old and over 65 years old, we also compared the mean dosage differences between these age groups in a table, with the elderly group chosen as reference. We compute the p-value from a Wilcoxon test.
In order to account for the sex differences in dosages, we adjust for age by comparing prescriptions in patients with similar age. We divide the patients by sex and assign them to 24 age groups of three consecutive years, and we calculate the mean dosage in each group for both genders separately. Then, we calculate the sex difference within each age group, where female patients are chosen as reference. Finally, the total sex difference is calculated as the weighted mean of all percentage differences, by weighting with the number of patients (male and female) of each age group. We depicted the analysis by plotting the daily prescribed doses in relation to patients' age in bars of six consecutive years. To calculate the signi cance, we run a paired Wilcoxon signed-rank test, in which the input are the mean values of male and female of the 24 age groups.

Age
The results of the in uence of age in dosage prescription are depicted in gure 3 with antidepressants, gure 4 with antipsychotics and gure 5 with hypnotics and tranquilizers. Several bar plots can be seen, in which the height of the green bars gives the mean daily prescribed doses in mg including both sexes, and the x-axis gives the age of the patients. Further, a red line is depicting a linear t in patients over 65 years old. The slope of the t is computed in the upper left corner, along with the percentage change. The percentage change is the average percentage increase (or decrease) of daily dosage in relation to the previous year of age. Along with the slope and the percentage change, the p-value under the hypothesis of zero slope is computed. Further, in table 4 the mean differences of dosages between patients up to 65 years and older than 65 years is shown. This description holds for each of the following three sections. In the six analysed antidepressants (Fig. 3), we found for each additional year of age a slight increase of daily dose in young patients, a plateau in middle aged patients and a decrease in elderly patients. Among patients older than 65 years, a decrease between 0.65% for the noradrenergic and speci c serotonergic antidepressant drug (NaSSA) mirtazapine and 1.32% for the selective serotonin reuptake inhibitor (SSRI) sertraline for every increasing year can be seen. For the SSRI citalopram and escitalopram, we found a decrease of 0.68% and 1.13%, respectively, and for the serotonin and norepinephrine reuptake inhibitors (SNRI) venlafaxine and duloxetine a decrease of 1% and 0.76%, respectively. The results of all these antidepressants are statistically signi cant at a level of 0.01. Comparing the mean prescribed dosages between patients up to 65 years old and older than 65 years old, we found statistically signi cant differences except for mirtazapine, varying from 7% for citalopram to 14% for venlafaxine (Table 4).

Age: Antipsychotics
In gure 4 for both antipsychotics, olanzapine and quetiapine, a plateau can be seen in young and middle age patients. However, a more pronounced slope than in antidepressants can be seen in elderly patients. Within patients over 65 years old, we found a decrease of 2.48% and 2.83% in daily prescribed dosages for each increasing year for olanzapine and quetiapine respectively (p<0.0001). Comparing the mean prescribed dosages between patients up to 65 years old and older than 65 years old, we found statistically signi cant differences of 24.4% for olanzapine and 21% for quetiapine (Table 4).

Age: Hypnotics and Tranquilizers
Between the two hypnotic drugs (zopiclone and zolpidem) and the tranquilizer drug lorazepam depicted in gure 5, the distribution of the daily prescribed dosages differs substantially. In the case of zolpidem (Figure 5a), a slight increase can be seen in young patients, a plateau in middle aged patients and a decrease in elderly patients. This distribution is similar to the distribution of antidepressants. For zopiclone (Figure 5b), a constant dosage around 7.2 mg per day can be seen for all ages. Therefore, age does not seem to in uence prescribed dosages in the case of zopiclone. On the other hand, for lorazepam ( Figure 5c), a steady reduction of prescribed dosages can be seen across all ages.
Within patients over 65 years old, we found a decrease of 1.17% and 1.26% in daily prescribed doses for each increasing year of age for zolpidem and lorazepam respectively (p< 0.001). In the case of zopiclone, we did not nd statistically signi cant differences with increasing age. When comparing the mean difference in prescribed dosages between patients up to 65 years old and older than 65 years old, statistically signi cant differences were found for zolpidem (10.4%) and lorazepam (29.7%), but not for zopiclone (Table 4).

Sex
The results for sex in uence in prescribed dosages is given in gure 6 with antidepressants, gure 7 with antipsychotics and gure 8 with hypnotics and tranquilizers. Several plots can be seen, where the y-axis gives the prescribed daily dose and the x-axis the age. Doses of males are depicted in blue and of females in orange, the standard errors are depicted in red. On the upper left corner further statistics can be found: the p-value under the hypothesis, that there is no difference between sexes; furthermore, the difference between sexes computed as percentage change, where females are chosen as reference. This description holds for each of the following three sections.

Sex: Antidepressants
In the case of antidepressants ( gure 6), we found statistically signi cant differences in daily dose prescription between sexes for escitalopram and venlafaxine, in which males received respectively 5.13% and 4.59% more than females. For citalopram, sertraline, mirtazapine and duloxetine we did not nd statistically signi cant differences between sexes.

Sex: Antipsychotics
Regarding antipsychotics (Figure 7), we found statistically signi cant differences for olanzapine, in which males receive 4.94% more than females. For quetiapine, we did not nd statistically signi cant differences.

Sex: Hypnotics and Tranquilizers
Among the hypnotic drugs zolpidem and zopiclone, and the tranquilizer lorazepam (Figure 8), we did not nd statistically signi cant differences between males and females.

Age
In order to discuss our ndings, we present a literature review about age and sex in uence on plasma levels of patients taking psychotropic drugs and compare it to the in uence of age and sex on prescribed dosages in our ndings. A comprehensive For duloxetine, we found a signi cant decrease in dosage prescription within the elderly and signi cant differences between young and elderly. In comparison, a study of Lobo et al. [16] computed a t over all ages and found a slope of -0.33 for apparent total clearance. Using the same method for the description of dosages in our data, the analogue slope shows -0.146 (Additional le 2, Fig. 1 supplementary). The slopes need to be normalized in order to be compared, which yields -0.50 for the apparent clearance, and -0.17 for dosage prescription. Again, the in uence of age on prescription of duloxetine is substantially smaller than the in uence on the plasma level according to the literature.
The case of mirtazapine is particularly noteworthy, since we found no differences in the prescription between young and elderly patients, and literature shows an increase of plasma levels in elderly above 40%. Although we found statistically signi cant differences between prescribed dosages in young and elderly patients for the remaining antidepressants, the differences in pharmacokinetic literature are clearly more pronounced than in real clinical decisions.

Age: Antipsychotics
Regarding the antipsychotics olanzapine and quetiapine, we found statistically signi cant mean differences in prescribed dosages of 24.4% and 21% respectively (between patients up to 65 and over 65 years old). In literature, the differences in pharmacokinetics are more pronounced: Castberg et al. [11] found 75% higher olanzapine plasma levels in elderly analysing 6239 patients and 35% for quetiapine in 1949 patients. Bakken et al. [24] found 50% higher levels analysing 601 patients for quetiapine (see also Additonal File 2, Fig.1 and 2, suppl.).

Age: Hypnotics and Tranquilizers
For the analysed hypnotics and tranquilizers, we did not nd literature describing pharmacokinetic data at steady state, but for single dose. For zolpidem, dosage prescriptions in our dataset are only 10.4% higher in young patients, whereas Olubodun et al. [38] found more pronounced age differences: in males 133% in the maximum concentration (C max ), 80% in half-life and 264% in the Area Under the Curve (AUC); in females 80% in C max and 60% in AUC. No signi cant differences have been found by Olubodun et al. [38] in half-life between younger and elderly females [38].
In the case of zopiclone, Gaillot et al. [26] found increasing differences with increasing age. Elderly between 60 and 68 years old had 31% higher AUC, and elderly between 74 and 85 years old had 105% higher AUC. We did not nd any in uence of age in dosage prescriptions of zopiclone. In particular, we found a at distribution with daily doses around 7.2 mg/day.
For lorazepam, Greenblatt et al. [40] found no signi cant in uence of age in elimination half-life (n=30), but a total clearance of 22% less in elderly patients. We found a difference in dosage prescriptions of 30% between young and elderly patients.
In summary, for zolpidem and zopiclone physicians take into account the kinetic data only to a limited extent. In the case of lorazepam, there is not enough pharmacokinetic data to draw conclusions.

Sex
A comprehensive table describing bullet points of the different studies in literature about sex differences in plasma levels of psychotropic drugs is given in the supplementary material (Additional le 1; Table 2 supplementary).

Sex: Antidepressants
Although pharmacokinetic data of several antidepressants is surprisingly limited, we found overall much higher differences in plasma levels between sexes in pharmacokinetic literature than in real clinical decisions. In 2009, Reis et al. [1] analysed the plasma levels for citalopram and escitalopram at steady state of 2330 and 1470 subjects respectively, and reported that female patients have 25% and 9% higher plasma levels than men at the same dosage. In the prescribed dosages we only found a difference of For sertraline, we found a dosage difference of 1.4%. This case is consistent with the results of Reis et al. [1] in which they did not nd statistically signi cant differences in plasma levels for sertraline in 1071 patients. For duloxetine, Lobo et al. [16] estimated the pharmacokinetic sex difference to be 64% for the same dosage, which is in part explained by differences in smoking habits. In contrast, we only found a 2.8% dose difference in real clinical decisions.
Overall, the difference in dosage prescriptions between sexes for the six most prescribed antidepressants does not re ect pharmacokinetic data except for sertraline, since there was no difference found in prescribed dosages as well as in plasma levels.

Sex: Antipsychotics
For olanzapine, Castberg et al. [11] found a 26.1% higher dose adjusted plasma concentration in women than in men for 5826 patients on stable dose. In contrast, we found a prescribed dosage difference of only 4.9%.
For quetiapine, Bakken et al. [24] investigated 601 patients and did not nd evidence that sex in uences pharmacokinetics. Similarly, we did not nd evidence that sex in uences prescribed dosages. On the other hand, Castberg et al. [11] examined 2210 patients with quetiapine prescriptions and did not nd evidence of sex differences in plasma levels for young patients, but with increasing age, the pharmacokinetic differences become noticeable and signi cant. In our study, we did not nd evidence of increasing dosages differences with increasing age in real clinical conditions.
Similarly to antidepressants, for antipsychotics we found literature with evidence of much higher pharmacokinetic sex differences than naturalistic differences in dosage prescriptions.

Sex: Hypnotics and Tranquilizers
For the benzodiazepine lorazepam, we surprisingly found very limited literature about its pharmacokinetics. There is a study of Greenblatt et al. from 1979 [40], where 30 patients are analysed with a single intravenous dose. No apparent relation between sex and pharmacokinetics was found.
Similarly, we did not nd evidence of sex affecting dosage prescriptions of lorazepam.
Greenblatt et al. [40] also conducted a study in 2019 in which the clearance of zolpidem in 70 patients was analysed. Although signi cant differences between sexes in terms of pharmacokinetics was found, according to the study, no evidence of sex related effectiveness nor adverse reaction was found. In our study, we did not nd evidence for sex related prescription differences. For the dosage prescription of zolpidem, it is not clear whether sex should be considered. Nevertheless, the results of our study show that physicians in Germany, Switzerland and Austria tend to prescribe zolpidem without considering the FDA recommendations of a maximum dose of 5 mg for women (mean dose for both sexes about 10 mg per day).
Regarding zopiclone, Gaillot et al. 1983 [39] found that C max is 14% to 24% higher, and AUC is 16% to 20% higher in women than men. In contrast, we found no evidence of sex in uencing the decision of dosage prescription.
Overall, we found very old and limited pharmacokinetic literature about the two most prescribed hypnotics/tranquilizers and zolpidem. There is certainly room for research in this eld.

Conclusion
In conclusion, we found that patients´ age does in uence physician decisions, but to a lower extent than what pharmacological literature would suggest. On the other hand, we found that sex overall does not in uence physician decision-making. Further, the plasma level differences found in pharmacokinetic literature between sexes are overall much larger than the prescribed dosages differences. We recall, patients body weight is not present in the dataset we analysed. Therefore, the results are not adjusted for weight. According to the Federal Statistical O ce of Germany, males have an average body weight of 85 kg and women of 68.7 kg [41]. Men have in average 23.7% more body mass than women [41]. Hence, not only the sex differences in pharmacokinetics are not taken into account, but also sex differences in body mass do not seem to in uence clinical decisions. Overall, the results of our study imply two possibilities: either plasma levels are not as correlated to drug effectiveness and adverse effects as is generally believed, or there is an underestimation of biological sex differences on behalf of physicians. The latter interpretation may also explain the partially disproportional number of adverse reactions in women. Although there is a large overlap in biological parameters between sexes, mean plasma levels differ substantially. However, we do not see these sex differences in the mean dosages. There is an urgent need for studies to examine whether adjusting the dosage in women on the basis of blood levels leads to better e cacy and lower frequency of adverse effects than the "treatment as usual" with similar dosages for both sexes.

Declarations
Ethics approval and consent to participate  Age distributions On the top the female age distribution, on the bottom the male age distribution. In our sample are proportionately more elderly women then elderly men, and overall more women.    Daily dose by age and sex: Antidepressants Daily prescribed doses (mg) of antidepressants by age: males in blue and female in orange. The standard error is depicted in red.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.