The impact of pregnancy on the pharmacokinetics of antidepressants: a systematic critical review and meta-analysis

Abstract

Introduction:

Pregnancy-related physiological changes exert a crucial impact on the pharmacokinetics of antidepressants, however the current evidence presents inconsistencies. A clearer understanding of pregnancy-related effects on antidepressant disposition may facilitate the development of guidelines for appropriate dose adjustments during the course of pregnancy based on therapeutic drug monitoring.

Areas covered:

We systematically reviewed studies comparing antidepressant levels in the same individuals during pregnant and non-pregnant states. Using dose-adjusted plasma concentration measurements, we estimated alteration ratios between the 3^rd trimester and baseline (before or after pregnancy). Additionally, we performed a meta-analysis for changes in dose-adjusted concentrations to estimate mean differences.

Expert opinion:

Data for several antidepressants display clear alteration patterns during pregnancy. On the basis of the alteration ratios trimipramine, fluvoxamine and nortriptyline show a prominent decrease in dose-adjusted levels especially in the 3^rd trimester. Clomipramine, imipramine, citalopram and paroxetine show smaller decreases in dose-adjusted concentrations in the third trimester. For escitalopram, venlafaxine and fluoxetine, changes are considered negligible. For sertraline, there was a tendency towards increased dose-adjusted concentrations in pregnancy. Available evidence suffers from major limitations and factors affecting pharmacokinetics have been insufficiently addressed. Further research is required to promote knowledge on pregnancy effects on antidepressant pharmacokinetics.

1. Introduction

During pregnancy a series of physiological changes take place in different organ systems [1]. These changes can alter the disposition of drugs by affecting their absorption, distribution, metabolism and excretion [2]. Among factors with an impact on pharmacokinetics are changes in protein binding, activity of hepatic cytochrome P450 (CYP) and other enzymes, renal function, volume of distribution as well as efflux transporters, such as P-glycoprotein (P-gp) [3, 4]. Of these, we consider the elimination route of the medication as the most decisive in the context of pregnancy-related effects on drug disposition [1, 5]. For example, the clearance of medications that are mainly excreted by the kidneys is expected to increase up to 50% during pregnancy [1]. On the other hand, for medications undergoing hepatic metabolism, different patterns of alterations may be observed based on the implicated CYP isoenzymes, as alterations in their activity may vary not only in magnitude, but also in direction, i.e. increase or decrease [1]. These alterations typically progress throughout pregnancy and peak within the third trimester [2]. Thus, it is also in the third trimester the most extensive changes regarding drug disposition compared to non-pregnant women have been reported [6, 7].

Antidepressants have been increasingly used in women during pregnancy as well as in the postpartum period [8], ranking among the most prescribed medications in pregnant women [2]. Accordingly, evidence on pregnancy-related changes in pharmacokinetics of several antidepressants is available [5]. Various pharmacokinetic variables, such as trough (C_min) and peak (C_max) plasma levels as well as the area under the concentration-time curve (AUC), have been investigated when assessing pregnancy effects on the metabolic rate of antidepressants [6, 9]. These effects have been linked to higher dose requirements in order to offset alterations in antidepressant drug disposition and to maintain therapeutic effect during pregnancy and specifically the third trimester [6, 10], which is unanimously considered a high-risk period for psychiatric disorder relapse [11]. Therefore, therapeutic drug monitoring (TDM), i.e. the quantification of antidepressant drug levels in maternal serum or plasma, is a valuable tool to guide dosing during pregnancy, instead of just extrapolating doses from studies of non-pregnant subjects [3, 5].

The aim of this study was to systematically review and meta-analyze studies measuring plasma or serum concentrations of antidepressants in the 3^rd trimester and before or after pregnancy, in order to determine the effects of pregnancy on antidepressant disposition.

2. Methods

The study was conducted according to the PRISMA guidelines [12]. Two authors (GS and MP) independently searched in Embase and Medline for studies including serum/plasma antidepressant concentration measurements during pregnancy, using the search terms “antidepressant” AND “pregnancy” AND “blood OR plasma OR serum”. Databases were searched in August 2019 for publication since data inception. References from identified studies were scanned for additional reports.

2.1. Inclusion & exclusion criteria

We included studies having a within-subject study design, with multiple measures of antidepressant drug levels in plasma or serum (from now on commonly referred to as plasma, for simplicity) in pregnant and non-pregnant states (before or after pregnancy, from now on commonly referred to as “baseline”). The pregnancy sample(s) were obtained in the third trimester (from week 26 and onwards). Baseline samples were obtained prior to pregnancy, or more than four weeks following delivery, a time frame after which pregnancy-related physiological changes affecting pharmacokinetics are unlikely to be sustained [13, 14]). There were no restrictions with regard to diagnosis or dosage of antidepressant treatment. All samples were taken (or presumed to be taken) as trough samples (drug fasting), but there were no restrictions with regard to how this was defined in the studies (i.e. min/max hours from last dose to sampling). Regarding duration of antidepressant treatment, we only included patients with treatment duration warranting steady-state (i.e. approximately five times the elimination half-life of the antidepressant in question [15]). We also searched in a large TDM database [7] for previously unpublished cases.

2.3. Data extraction

Two authors (GS and MP) independently extracted data. Another author (AAW) provided raw data from the largest previously published cohort [7]. Sample sizes, number of patients, daily doses of antidepressants, concentrations of antidepressants (means and ranges or standard deviations depending on how data were provided) in maternal plasma before pregnancy, in the 3^rd trimester, and in the postpartum period (specified by the number of weeks after parturition) were extracted. When levels for multiple pregnancies per patient were provided, we used mean values in order to include each patient only once in the analysis. Mean values were also used when multiple measurements per patient in the 3^rd trimester or at baseline were provided. When data were provided in nmol/L or μmol/L values were converted to ng/mL using previously published conversion factors [16]. When antidepressant levels were below the limit of quantification of the method (LOQ), we used the half of LOQ in our estimations [17]. When additional information was necessary to interpret data, authors of the original studies were contacted.

2.4. Outcomes & statistical analysis

The primary outcome variable was the “alteration ratio”, defined as the ratio between the dose-adjusted plasma concentration of the antidepressant in the 3^rd trimester and the dose-adjusted plasma concentrations at baseline (pregravid and/or postpartum). Thus, an alteration ratio less than one indicates that the 3^rd trimester concentration is lower than at baseline, and vice versa. Ratios from each study were pooled to provide a combined alteration ratio; these estimations were performed by pooling ratios of individual patients as in an individual participant data meta-analysis. We used means of ratios instead of ratios of means as the research question seeks to assess alteration patterns using patients as their own reference. When cohort information, but not individual patient data, were available we calculated combined ratios by weighting the alteration ratio of each study by the study’s sample size. This analysis builds upon the theoretical framework that has been used in previous secondary analyses of TDM data [18-21].

We also used a classical meta-analytic approach to estimate mean differences (MDs) for changes in dose-adjusted concentrations (i.e. concentrations given in ng/mL per mg/day) between baseline and 3^rd trimester. When data were provided in AUCs, the AUC values were divided by the number of hours in the time interval to provide mean concentrations, which we consider justified to use as a proxy for the trough concentrations. Data provided as concentrations adjusted to defined daily dose (DDD) were converted to dose-adjusted levels (i.e. the concentration per mg/day) by dividing by the DDD. We used a random-effects model given the potential heterogeneity related to the analytical methods and the inherently large variability of the pharmacokinetic variables. We examined the robustness of the measures by checking the results under a fixed-effect model. The results were summarized using MDs and 95% confidence intervals (CIs). The DerSimonian-Laird estimator was used to calculate the heterogeneity variance parameter (τ²) [22]. Further, we calculated the I-square (I²) statistic, which indicates the proportion of variability that can be attributed to heterogeneity rather than random error. All analyses were performed using the meta package in the statistical software R [23]. We only performed meta-analyses when data from at least three studies were available.

2.5. Quality assessment

The methodological quality of the included studies was assessed using the ClinPK guidelines [24], which is an established tool for the evaluation of transparent and accurate reporting of pharmacokinetic studies.

3. Results

The search strategy identified 169 studies (see Figure 1). After screening abstracts 14 articles were reviewed in full text. Two articles described data in a form that could not be entered in the analyses [25, 26]; authors were therefore contacted, but no additional data were provided. One article provided paroxetine levels during pregnancy, but no baseline values [27]. These three articles were therefore excluded. Thus, eleven articles were finally included in the analysis. For citalopram, fluoxetine and norfluoxetine, data for S- and R-enantiomers were provided separately [6, 28] and total levels were estimated by summing the two enantiomers. One previously unpublished case report of a trimipramine-treated patient out of a large TDM database was also included [7], for which approval from the regulatory authorities was obtained accordingly. Table 1 contains all relevant data retrieved from the eleven studies finally included [6, 7, 9, 10, 14, 28-33]. The studies covered a total of eleven antidepressants.

Figure 1.

Prisma Checklist Study Flow Diagram

Table 1.

Dose-adjusted antidepressant blood levels, alteration ratios (pregnant state values divided by non-pregnant state values) for each cohort and combined alteration ratios for dose-adjusted levels of antidepressants. Numbers in parentheses refer to the range of ratios from individual patients. When dose-adjusted were not provided by the authors, estimations were performed by us.

Antidepressant	n	3rd trimester	Baseline (pooled pregravid & postpartum)	Alteration ratio	Combined alteration ratio	Quality score	Reference
Citalopram	11	1.23±0.411	2.14±0.81	0.58	0.70 (0.23-2.15)	13	[14]
	2	1.79±0.241	1.81±1.04	1.23±0.83		15	[6]
	12	1.14±0.51	1.90±0.75	0.73±0.51		16	[7]
Clomipramine	12	0.98	1.71	0.57	0.52 (0.23-0.76 )	13	[10]
	63	1.02±0.94	1.73±1.06	0.51±0.18		16	[33]
Escitalopram	2	2.29±1.61	3.15	1.09	1.10 (0.58-1.91)	15	[6]
	12	1.04±0.34	1.05±0.48	1.10±0.38		16	[7]
Fluoxetine	84	7.27±1.74	10.68±3.36	0.68	0.86 (0.09-2.61)	14	[30]
	8	10.11±6.951	11.33±4.99	1.03±0.78		15	[28]
	8	8.26±4.07	9.92±2.73	0.88±0.39		16	[7]
Fluvoxamine	2	0.50±0.30	1.28±0.84	0.40 (0.38-0.41)	NA	16	[7]
Imipramine	1	0.92	1.36	0.68	0.65 (0.63-0.68)	13	[10]
	1	0.67	1.06	0.63		9	[29]
Nortriptyline	6	0.71±0.48	1.32±0.63	0.51±0.17	0.50 (0.32-0.70)	13	[10]
	1	0.521	1.11	0.47		9	[29]
Paroxetine	4	1.70±1.33	2.00±0.84	0.82 (0.26-1.38)	NA	16	[7]
Sertraline	8	0.33±0.02	0.56±0.36	0.77±0.50	1.38 (0.34-3.04)	17	[9]
	6	0.36±0.161	0.44±0.34	1.37±1.08		15	[6]
	12	0.36±0.15	0.27±0.15	1.53±0.71		16	[7]
Trimipramine	1	0.111	0.421	0.27	NA	NA (unpublished)	[7]
Venlafaxine	15	0.30	0.89	0.34	1.00 (0.32-3.17)	8	[31]
	7	1.92±0.99	2.81±1.50	0.72±0.25		15	[32]
	5	0.96±0.51	1.05±0.69	1.52±1.34		16	[7]

n: number of women, NA: not applicable. For clomipramine, imipramine, fluoxetine and venlafaxine provided levels refer to active moiety, i.e. the sum of parent compound and active metabolite. The reason for differences between the reported size of the cohorts and the number of included patients in the ratio estimations are missing values at different time points. Dose-adjusted antidepressant blood levels are provided in ng/mL per mg/day. The quality score column represents the quality scores assigned to the study using the ClinPK checklist (for details, see Supplementary table 1).

¹Mean of two measures

²Co-medicated with propylthiouracil

³The original sample size was 12 patients, but data for both timepoints were available for 6 patients.

⁴Values were converted using a conversion factor (CF) of 3.30, fairly close to 3.23 (CF for fluoxetine) and 3.39 (CF for N-desmethyl-fluoxetine)

⁵Co-medicated with quetiapine (300mg/day) and trazodone (150mg/day)

3.1. Citalopram

Alteration ratios were estimated in three cohorts (n=25) [6, 7, 14] yielding mean values for cohorts between 0.58 and 1.23 (Table 1), with the combined ratio being 0.70. The meta-analytic approach revealed an overall MD between baseline and 3^rd trimester of 0.78 ng/mL per mg/day, CI: 0.42 to 1.14, p<0.001 for both random- and fixed- effect model; C/D levels were lower in the 3^rd trimester (figure 2). Observed heterogeneity was minimal (I²=0%, p=0.54).

Figure 2.

Forest plot for differences in citalopram dose-adjusted concentrations between baseline and 3^rd trimester (n=25 from 3 studies). Values are provided in ng/mL per mg/day. CI: confidence interval; MD: mean difference; SD: standard deviation.

3.2. Escitalopram

Evidence for escitalopram derived from two studies [6, 7], yielding means of alteration ratios of 1.09 and 1.10 respectively; the combined alteration ratio was 1.10 (range of individual patient ratios 0.58-1.91).

3.3. Clomipramine

Evidence for clomipramine derived from one study (n=12) [33] and one single patient of a small cohort treated with tricyclic antidepressants [10]. Both reported a reduction in 3^rd trimester dose-adjusted levels of clomipramine plus its active metabolite desmethylclomipramine compared to the baseline levels yielding alteration ratios of 0.51 and 0.57, respectively, and a combined ratio of 0.52 (range for individual patient ratios 0.23-0.76).

3.4. Fluoxetine

Means of alteration ratios for fluoxetine plus its active metabolite norfluoxetine in three cohorts (n=24) ranged between 0.68-1.03 (Table 1) and the combined ratio was 0.86 [7, 28, 30]. The meta-analysis revealed an MD between baseline and 3^rd trimester of 2.59 ng/mL per mg/day, CI: 0.63 to 4.55, p=0.009 for both random- and fixed-effect model; C/D levels were lower in the 3^rd trimester (figure 3). Observed heterogeneity was minimal (I²=0%, p=0.65).

Figure 3.

Forest plot for differences in fluoxetine dose-adjusted concentrations between baseline and 3^rd trimester (n=24 from 3 studies). Values are provided in ng/mL per mg/day. CI: confidence interval; MD: mean difference; SD: standard deviation.

3.5. Fluvoxamine

Evidence for fluvoxamine derived from two patients in one study [7], displaying a reduction in 3^rd trimester dose-adjusted fluvoxamine levels to 40% of the baseline (alteration ratio 0.40, range of individual patient ratios 0.38-0.41).

3.6. Imipramine

Two case reports for imipramine described decreased 3^rd trimester C/D levels of imipramine plus its active metabolite desipramine, compared to baseline [10, 29], with alteration ratios being 0.68 and 0.63 respectively and combined ratio being 0.65.

3.7. Nortriptyline

Data for nortriptyline derived from two studies [10, 29], both reporting a decrease of the dose-adjusted nortriptyline levels in the 3^rd trimester, with mean alteration ratios being 0.47 and 0.51 respectively and a combined ratio of 0.50.

3.8. Paroxetine

In four patients from an SSRI-treated cohort [7], a mean alteration ratio of 0.82 (range for individual patient ratios 0.26-1.38) was estimated.

3.9. Sertraline

Mean alteration ratios in three cohorts (n=18) [6, 7, 9] ranged between 0.77-1.53 with the combined ratio of 1.38 (individual patient ratio range 0.34-3.04) (Table 1). The meta-analysis revealed an MD between baseline and 3^rd trimester of −0.01 ng/mL per mg/day, CI: −0.19 to 0.17, p=0.90 for the random-effects model and −0.05 ng/ml per mg/day, CI: −0.17 to 0.06, p=0.34 for the fixed-effect model (figure 4; negative values imply higher concentrations in the third trimester). Observed heterogeneity was not substantial (I²=25%, p=0.26).

Figure 4.

Forest plot for differences in sertraline dose-adjusted concentrations between baseline and 3^rd trimester (n=18 from 3 studies). Values are provided in ng/mL per mg/day. CI: confidence interval; MD: mean difference; SD: standard deviation.

3.10. Trimipramine

There are no published data for trimipramine. Here we present previously unpublished data from a patient out of a large TDM database [7], which yielded an alteration ratio of 0.27.

3.11. Venlafaxine

Mean alteration ratio for venlafaxine plus its active metabolite O-desmethylvenlafaxine in three cohorts (n=13 patients) [7, 31, 32] ranged between 0.34 and 1.52 with a combined ratio of 1.00 (range for individual patient ratios 0.32-3.17). The meta-analysis revealed an MD between baseline and 3^rd trimester of 0.29 ng/mL per mg/day, CI: −0.39 to 0.98, p=0.40 for the random-effects model and 0.28 ng/mL per mg/day, CI: −0.37 to 0.94, p=0.40 for the fixed-effect model (figure 5). Observed heterogeneity was minimal (I²=5%, p=0.31).

Figure 5.

Forest plot for differences in venlafaxine dose-adjusted concentrations between baseline and 3^rd trimester (n=13 from two studies and one case report). Values are provided in ng/mL per mg/day. CI: confidence interval; MD: mean difference; SD: standard deviation.

3.12. Quality of included studies

The quality of the included studies was acceptable with an average rating score of 13.7 (supplementary table 1). Some variation is mainly explained by the lower quality in case reports, where authors provided less detail on the items outlined by the ClinPK checklist. Specifically, two case reports did not specify if included levels were trough (2/11).

4. Discussion

The present systematic review and meta-analysis provides evidence for alterations of dose-adjusted antidepressant levels between baseline and 3^rd trimester for a number of widely prescribed antidepressants. The amount of available data varies between antidepressants, but clear patterns of significant alterations could be observed for several antidepressants.

Citalopram disposition seems to be largely affected by the pregnancy-related physiological changes; the combined alteration ratio of 0.70 means that 3^rd trimester values are on average 70% of the baseline. The meta-analysis indicated approximately the same difference between baseline and 3^rd trimester. The MD of 0.78 ng/mL per mg/day between baseline and 3^rd trimester means that with a dose of 20 mg/d, 3^rd trimester concentrations would, on average, decrease by approximately 15.6 ng/mL. However, total drug concentrations may provide a somewhat misleading pharmacokinetic picture for citalopram. Citalopram is a chiral drug subject to stereoselective metabolism; the pharmacologically active S-citalopram (escitalopram) is metabolized primarily by CYP2C19, whereas the inactive R-citalopram is metabolized primarily by CYP2D6. Pharmacokinetic data from TDM during pregnancy suggest that the overall activity of CYP2C19 metabolism in humans decreases, whereas the activity of CYP2D6 increases [7]. The mechanism of these changes observed in humans remain to be clarified. It may be a hormone interaction with CYP enzymes as shown in animals [34], or it may be contribution of fetal metabolism and elimination into amniotic fluid and placental transfer. Data for both enantiomers were available from two patients in a small cohort [6]; the alteration ratio was lower for R-citalopram (i.e. a more profound decrease in concentrations during pregnancy) than for S-citalopram (supplementary table 2). This change in chiral composition implies that the increased CYP2D6 activity has a greater influence on the net total plasma citalopram concentrations than the decreased CYP2C19 activity. This is clinically relevant, since it implies that declining total citalopram plasma concentrations should not necessarily translate into need for higher doses; the decline could be caused – primarily or only – by a decline in the inactive enantiomer concentration. Nevertheless, one study has provided clinical data linking decrease in dose-adjusted citalopram levels and changes in antidepressant efficacy [6]. We suggest use of TDM in treatment with citalopram during pregnancy in order to early detect patterns of changes and guide dose-adjustment when necessary. However, as it is suspected that most of the reduction in concentration would reside in the inactive R-citalopram, it would be advisable to perform a stereoselective analysis if available, and base dose adjustments on changes in the S-citalopram concentrations.

Escitalopram is also available as a single-enatiomer drug, for which our study has separate data. Evidence for escitalopram derived from 13 patients [6, 7] and yielded a combined ratio of 1.10. This 10% increase in drug concentrations in 3^rd trimester could be caused by CYP2C19 inhibition, as described above. The change is not likely to have clinical implications.

Evidence for changes in fluoxetine metabolism implies moderate changes resulting to a combined ratio of 0.86. The meta-analytic approach also supported a clear, but not large, difference between baseline and 3^rd trimester. The MD of 2.59 ng/mL per mg/day between baseline and 3^rd trimester means that with a fluoxetine daily dose of 20 mg/d the active moiety 3^rd trimester concentrations would decrease by 51.8 ng/mL. The therapeutic reference range for the active moiety of fluoxetine is 120-500 ng/mL [16]. Although the data regarding the pharmacokinetic correlates of treatment response to fluoxetine are not very strong [35], the clinical impact of the pregnancy-related changes in fluoxetine metabolism received attention in one cohort, with changes in fluoxetine, but not norfluoxetine concentrations being associated with depression scores [28].

Underlying mechanisms may include increased demethylation reflected as increased norfluoxetine/fluoxetine (or decreased inverse) ratios during pregnancy [7, 30], although another study did not describe such changes (see supplementary table 3) [28]. The induction of CYP2D6 may cause the decrease in fluoxetine concentrations [5]. Moreover, as norfluoxetine is excreted by the kidneys, some of the total effect may be related to pregnancy-related alterations in kidney function.

Fluvoxamine TDM data from two patients showed a large decrease in 3^rd trimester (alteration ratio 0.4). Given the relatively narrow reference range of fluvoxamine (60-230 ng/mL [16]), concentration changes down to just 40% of the baseline may have a prominent effect on treatment response. The two main isoenzymes involved in fluvoxamine pharmacokinetics, CYP2D6 and CYP1A2 [16], are affected in opposite directions during pregnancy; CYP2D6 activity increases, while CYP1A2 activity decreases [7]. Therefore, we may hypothesize that the increased overall activity of CYP2D6 overcomes the inhibition of CYP1A2 and drives the reduction in dose-adjusted fluvoxamine levels in 3^rd trimester. This reduction may affect treatment response and clinicians are advised to use TDM in fluvoxamine-treated pregnant women.

Conclusions on imipramine pharmacokinetic changes during pregnancy are solely based on two case reports [10, 29], with a combined ratio of 0.65 indicating a moderate decrease in the active moiety in 3^rd trimester values compared to baseline. In light of the therapeutic reference range of imipramine, which is 175-300 ng/mL [16], this drop could cause changes in therapeutic effect. Multiple liver enzymes are involved in imipramine metabolism including CYP1A2, CYP2C19 and CYP2D6 [16]; as pregnancy-related changes in activity of each of these enzymes follow contrasting patterns, the overall high activity of CYP2D6 observed during pregnancy, seems to dominate. In both cases described in the literature pregnancy-related changes in imipramine active moiety concentrations were followed by large imipramine dose increases [10, 29]. Given the major impact of pregnancy on imipramine disposition and as dose adjustments may be required, measuring imipramine active moiety concentrations regularly during pregnancy may provide a valuable overview guiding clinicians in the appropriate dose selection.

Pregnancy-related effects are additionally prominent on nortriptyline pharmacokinetics. In a total of 7 patients a combined alteration ratio of 0.5 was estimated. The reduction in 3^rd trimester levels to half that of baseline implies a large change as nortriptyline has a relatively narrow reference range of 70-170 ng/mL [16]. As nortriptyline is primarily catalyzed by CYP2D6 [16], the increased activity of this enzyme can explain the decrease in dose-adjusted levels in the 3^rd trimester. Clinicians can reduce the impact of these changes by regularly measuring nortriptyline levels during pregnancy and adjusting the daily dose as clinically indicated.

Changes in 3^rd trimester paroxetine pharmacokinetics are based on data from one cohort of 4 patients [7]. The alteration ratio of 0.82 implies a rather small impact taking into consideration the reference range of 20-65 ng/mL [16]. However, the estimated effects have been larger in our previous analysis using a different statistical approach [7]. Moreover, a pharmacogenetic study including CYP2D6 genotyping reported that whereas paroxetine levels decreased in ultrarapid and extensive (normal) metabolizers, concentrations increased in intermediate and poor metabolizers [27]. The reason could be that the induction effect of CYP2D6 will not translate into increased enzyme activity in those with genes coding for non-functional enzymes. Unfortunately, we were unable to include these pharmacogenetic data for paroxetine in our analysis due to lack of baseline measures outside the pregnancy period. Thus, for paroxetine-treated pregnant women, apart from TDM, clinicians may also consider CYP2D6 genotyping.

For sertraline, increased dose-adjusted levels were observed in the 3^rd trimester. As the reference range of sertraline is very wide (10-150 ng/mL [16]), it is unlikely that these changes may affect treatment outcomes. Regarding the underlying mechanisms, the complex metabolic pathway of sertraline involving multiple liver enzymes does not allow any simple hypothesis. These enzymes include CYP2B6, CYP2C19, CYP2C9 and CYP2D6 [16], which are affected by pregnancy in a divergent fashion [7]. The interplay of these alterations and the resulting effects are difficult to predict. Nevertheless, increased levels of the major sertraline metabolite during pregnancy [7, 9] may imply some stronger role for CYP2C19 [36] compared to other isoenzymes, although data are not always consistent [6]. Inhibition of CYP2C19 could account for the increase in dose-adjusted sertraline levels in 3^rd trimester. The extent of this inhibition might have been affected by changes in activity of P-gp activity [37]. No pharmacogenetic data for the pregnancy effects on sertraline are available to our knowledge. However, using the paroxetine evidence as an example [27], we would expect different patterns of pregnancy effects on sertraline in patients with specific CYP2C19 variants. For example, in CYP2C19 poor metabolizers decreased C/D levels could be expected due to induction of CYP3A4 and other CYP enzymes involved. So, even though the average effect of pregnancy on sertraline is small, it might well have clinical relevance in some subgroups.

Evidence for venlafaxine led to an estimation of a combined alteration ratio of 1, while the meta-analytic approach implied minimal differences, although with wide CIs. Thus, venlafaxine seems to be less affected by the pregnancy-related physiological changes, although in some patients the effect may be more pronounced [31]. The reference range of the venlafaxine active moiety is relatively wide, 100-400 ng/mL [16]. As previously discussed for paroxetine, the observed interpatient variability in pregnancy effects on venlafaxine could be related to genetic differences or differences in post-dose intervals, i.e. when levels are not trough. However, we suspect that pregnancy effects on venlafaxine in patients with specific genetic variants may be smaller because the concentration of the active metabolite O-desmethylvenlafaxine, which contributes to the active moiety [16], would tend to counteract the changes in the concentration of the parent substance [7, 32]. Therefore, venlafaxine concentrations (i.e. concentrations of the active moiety) are not expected to be prominently altered during pregnancy.

Data for clomipramine and trimipramine were particularly scarce, although data from a small clomipramine-treated cohort of pregnant women [33] provided some valuable insight. Concentrations were low during pregnancy for both clomipramine and trimipramine, with alteration ratios of 0.57 for clomipramine active moiety and 0.27 for trimipramine. The reference ranges for clomipramine active moiety and trimipramine are narrow, 230-450 ng/mL and 150-300 mg/mL, respectively [16]. Therefore, the impact of the reported changes can be crucial. Increased activity of CYP2D6 and CYP3A4 appears to be the most important underlying mechanisms which surpass the decrease in activity of CYP2C19 and CYP1A2 in pregnancy. For both drugs, TDM is highly recommended during pregnancy.

4.1. Limitations

Despite the available studies, the current literature suffers from several limitations that may hamper conclusions on pregnancy-related effects on pharmacokinetics of antidepressants. First, for most newer antidepressants there is scarce evidence, and even for some older antidepressants data is lacking. Second, the majority of the available studies refer to single cases or small cohorts of antidepressant-treated pregnant women. Whenever data are slightly more robust, heterogeneity is prominent. There is also a risk that we were unable to unravel changes in the disposition of some medications due to Type II errors, which can be seen from the rather wide CIs. Larger cohorts would display even larger interindividual variability and most likely prove that it is rather impossible to tailor treatment to the group, but to the person using TDM and clinical judgement for each individual. Third, post-dose intervals for trough samples were not standardized between studies and in one study it varied from 8 up to 30 hours. Given the small number of subjects in our calculations, our study might be vulnerable to skewed single observations and missing clinical information, e.g. a sample taken shortly after last drug intake, but sent to the TDM laboratory as if it was a trough sample. Specifically, two case reports did not specify if levels were trough [29, 31]. Fourth, none of the included studies assessed adherence, which is reportedly low in pregnancy [38] and could have accounted for decreased C/D levels. Fifth, as included patients were not always on monotherapy [10, 31], co-medication with other drugs might have affected the results. And finally, only hypotheses can be formulated for the impact of genetic variants on pregnancy-related pharmacokinetic changes. Thus, there is an urgent call for further research in the field in order to successfully move from assumptions based on theoretical knowledge deriving from other populations to pharmacogenetic evidence for antidepressant-treated pregnant women.

5. Expert opinion

The severe, quantitative and qualitative limitations related to the amount of original data available hamper a comprehensive understanding of patterns for pharmacokinetic changes for antidepressants in pregnant women. Consequently, the current literature barely allows for practical recommendations for antidepressant dose adjustments during pregnancy, as knowledge of gestation-related physiological changes is the main contributor to our current conceptualization of antidepressant disposition. Dose-effect relationships for antidepressants have been investigated in non-pregnant patients only, and clinicians treating pregnant patients frequently have to proceed based on clinical intuition rather than any type of evidence. Therefore, there is an urgent call for data in order to create reliable algorithms for the appropriate dose selection in pregnant patients instead of encompassing off-label strategies linked to anecdotal experiences. In other disciplines, such as neurology, TDM is regularly used to guide dose selection [39], whereas psychopharmacological treatment of pregnant women mainly progresses using pharmacoepidemiological data that capture prescription patterns rather than efficacy or safety hierarchies.

Considering the past and the current development of the related research activity, which experienced a small boom approximately fifteen years ago, it is, unfortunately, unlikely that robust evidence for patterns of pregnancy-related alterations for antidepressants will be made available in the near future despite the increasing awareness for perinatal mental health and the need for better efficacy and safety outcomes. Apart from independent researchers, there is no history of any manufacturer of any antidepressant that previously collected and/or made this type of data available; thus, it would be a surprise if the pharmaceutical industry will provide such evidence. Moreover, there is currently no antidepressant holding a US Food & Drug Administration (FDA) or European Medicines Agency (EMA) approval specifically for pregnant patients and no regulatory authorities would require gestational pharmacokinetic data for drugs not having such an approval. Indeed, the manufacturers generally advise against using their product during pregnancy, although prescriptions for antidepressants during pregnancy are skyrocketing worldwide [2]. In summary, we are not particularly optimistic regarding the progress of understanding of pregnancy-associated pharmacokinetic changes for antidepressants in light of new evidence.

On the other hand, advocating the use of TDM as integral part of gestational and perinatal psychopharmacology may provide valuable guidance to clinicians [21], although we acknowledge that studies are required to validate the clinical utility of TDM in pregnant women [39]. It should also be emphasized that the utility of TDM assumes that the link between the peripheral (plasma) concentration of a drug and its concentration in the brain is the same during pregnancy as in non-pregnant women, and also that the link between the concentration at the site of action and the pharmacodynamic effect on the (in this case) serotonin or noradrenalin transporter is the same as outside pregnancy. Nevertheless, the contribution of TDM in personalizing dosing is evident in other fields of perinatal pharmacotherapy as well, such as for antiretroviral agents and anticonvulsants [2, 39]. Moreover, as TDM technologies are advancing and commercialization of TDM opens new horizons [40], interest for establishing TDM in treatment of several patient subgroups receiving psychopharmacotherapy including pregnant patients may be stimulated. If the availability of TDM increases, this may decrease TDM costs and therefore make it more accessible, while more promptly delivered results will enhance its clinical utility. To our knowledge, there are currently no other established tools to handle the dosing challenges in pregnant patients. Despite some early promises of pharmacogenetics in terms of personalized doses, pharmacogenetic testing in general moved from hope to hype within short time [41]. Nevertheless, the understanding of pharmacokinetic phenomena, such as enzyme induction or inhibition, within a pharmacogenetic context is a gradually emerging notion [42], that may receive more attention in the future. However, it is unclear if this will activate some potential for pharmacogenetic tests in personalized dosing of antidepressants for pregnant patients or remain limited in the research field. In fact, it is more likely that pharmacogenetics will not get established as integral part of perinatal psychopharmacology in the nearest future. Summing up, unless pharmacokinetic changes related to pregnancy receive more attention, it is unclear how appropriate dosing strategies for effective and safe treatment of the mother and for safety of the fetus will realistically improve on the middle- or long-term. Meanwhile, despite all the uncertainties, the best approach would be to use TDM combined with a thorough clinical follow-up to guide treatment with antidepressants during pregnancy. Our secondary evidence aims to capture patterns of pharmacokinetic changes for eleven widely prescribed antidepressants that may assist clinicians prescribing antidepressants in pregnant women.

Article highlights.

• Pregnancy-related physiological changes exert a crucial impact on the pharmacokinetics of several antidepressants.

• Due to pharmacokinetic changes during pregnancy dose adjustment may be required for some drugs.

• Limited data imply drops in dose-adjusted maternal plasma levels by 50% or even more in the 3^rd trimester for fluvoxamine, nortriptyline and trimipramine.

• For other antidepressants, changes in plasma concentrations during pregnancy are less pronounced or even show the contrary which may be a risk of high drug exposure to the infant.

• More quantitative and qualitative evidence is required for the further understanding of the pregnancy impact on pharmacokinetics of antidepressants.

• We recommend thorough clinical follow-up and use of therapeutic drug monitoring for all patients treated with antidepressants during pregnancy.