Software

The Simcyp (version 21) population-based PBPK simulator (Simcyp, Sheffield, UK) was used to develop venlafaxine and ODV PBPK models and perform all the simulations. The PBPK models for quinidine, ketoconazole, and cimetidine used for model verification against clinical DDI studies were published previously [18,19,20]. Clinical study data from the literature were digitised with GetData Graph Digitizer version 2.22.

Venlafaxine and ODV PBPK models

The workflow for the PBPK model development and verification of venlafaxine and ODV is described in Fig. 2. Venlafaxine primarily undergoes metabolism via CYP2D6, with lesser contributions from CYP2C9 and CYP2C19, resulting in the formation of its active metabolite, ODV. The latter is further metabolised primarily by CYP3A4, undergoes UGT-mediated conjugation and is excreted renally. The elimination pathways for venlafaxine and ODV, including the estimated fraction metabolised by CYP enzymes (fm) and fraction excreted renally (fe), are illustrated in Fig. 2. Several UGT isoforms including UGT1A1, 1A3, 2B4, 2B15, 2B17 contribute to ODV O-glucuronide formation, but the exact contribution of each UGT isoform is unknown [8]. Thus, UGT-mediated elimination of ODV is represented by an additional clearance in human liver microsomes in our model. Distribution for both venlafaxine and ODV was described by the whole body full PBPK distribution model. The advance dissolution, absorption and metabolism (ADAM) model [21] was used to describe the absorption of venlafaxine IR and ER formulations, as well as the intake of venlafaxine and ODV from breast milk. The input parameters of venlafaxine and ODV PBPK models are detailed in Table S1.

Fig. 2

Flow chart of the development and verification venlafaxine and ODV PBPK models. PM: poor metaboliser; EM: extensive metaboliser; TDM: therapeutic drug monitor; fm: fraction metabolised by CYP/UGT enzymes; fe: fraction excreted renally

Clinical data

Individual PK data for breastfeeding mothers and their infants from two clinical studies were used.

(i)

Study 1: PK data were reported for two out of three mothers aged 26 and 35 years, who were on stable twice-daily (BID) doses of venlafaxine IR (150 mg BID and 225 mg BID) for 0.5 and 5 months respectively. Their infants, aged 0.37 and 1.3 months (38 weeks at delivery), were breast-fed according to their normal pattern during the study. Venlafaxine concentrations were not detectable in all infant blood samples [22].

(ii)

Six breastfeeding women aged 30–41 years and their infants, including one pair of twins, aged 2.7 to 10.3 months participated [12]. The individual age of the mothers and their infants were not reported. Mother subjects 1–5 received venlafaxine IR (Subject 1 [likely a CYP2D6 PM], 112.5 mg BID; Subject 2, 150 mg BID; Subjects 3 and 4, 150 mg in the morning and 112.5 mg at night; Subject 5, 150 mg in the morning and 75 mg at night), while Subject 6 received 225 mg venlafaxine ER once per day in the morning. Infants were breastfed 5–6 times daily. Infants of mother Subjects 2 and 6 received supplementary solid food 2–3 times per day. Infant of mother Subject 3 received supplementary formula milk (50–120 mL per feed). Maternal PK profiles of both venlafaxine and ODV were reported. Infant blood samples were collected at 6.2 to 6.8 h (average 6.5 h) post maternal dose. Venlafaxine concentrations were undetectable in two infants.

In addition, PK data from healthy adult subjects were collated from 3 clinical studies to verify the model, including single intravenous (IV) dose of venlafaxine and desvenlafaxine (ODV), as well as a single oral dose of venlafaxine IR and ER [23,24,25]. Clinical DDI studies, including those with quinidine (CYP2D6 inhibitor), ketoconazole (CYP3A4, CYP2C9 and CYP2C19 inhibitor), and cimetidine (CYP2D6 inhibitor) [26,27,28] were collated, to verify each CYP component of the venlafaxine and ODV models. Furthermore, therapeutic drug monitoring (TDM) data from one clinical study involving children and adolescents and one retrospective analysis that included adolescents were utilised [29, 30]. The details of these clinical studies are provided in the supplementary materials.

Virtual populations

Simulations were performed using virtual populations from the Simcyp population library, including North European healthy adult and paediatric populations. The demographic, physiological, and biochemical parameters used to construct the virtual adult population have been described previously [31]. The paediatric demography (age, height, weight, BSA [body surface area]), developmental physiology (tissue volume, tissue blood flow, renal function, gastrointestinal tract anatomy) and biochemistry (albumin, CYP ontogeny) are integrated. The algorithms describing these developmental changes are described in detail elsewhere [32,33,34,35,36]. To account for the ontogeny of UGTs– specifically UGT1A1, UGT1A3, UGT2B4, UGT2B15, and UGT2B17 - the additional HLM clearance of ODV was adjusted based on literature-reported UGT levels at birth when simulating drug exposure in neonates (< 1 month old). The average activity of these UGT isoforms at birth is approximately 60% of the adult level; UGT1A1 and UGT2B17 show the lowest activity at birth, which is around 30% of the adult level [37, 38].

Simulations

Simulations were performed using 10 trials of the number of subjects in the corresponding clinical studies. For individual PK data concerning lactating mothers and their infants, 100 trials per each subject were simulated. Where available, demographic information and dose regimens were aligned with those of corresponding clinical studies. When individual age data were unavailable, the reported age range was used. When sampling times are unspecified in the clinical study, steady-state concentrations of venlafaxine and ODV were simulated following multiple oral doses of venlafaxine for 7 days.

Venlafaxine and ODV are excreted into breast milk with observed mean (CV) of M/P AUC ratios of 2.78 (20%) and 2.82 (12%), respectively [12, 22]. These values were incorporated into the PBPK model along with the simulated average plasma concentration of venlafaxine and ODV to estimate the infant daily dose (IDD) using Eq. 1.

$$\eqalign}(}/}/}) \cr & = }(}/})\quad *}(1/}/}) \cr} $$

(1)

The daily milk intake was assumed to be 150 mL/kg/day for exclusively breastfed infants. In cases where infants also received supplementary formula milk or solid food [12], breastmilk intake was assumed to be halved (75 mL/kg/day).

The relative infant daily dose (RIDD) was calculated using Eq. 2 where IDD represents the combined IDD values of both active moieties.

$$\eqalign}(\% )} \cr & }(})}(}) \cr} $$

(2)

Further, sensitivity analyses were performed to assess the impact of variability in milk composition (fat content and pH) on the M/P ratio using the algorithms published previously [39].

To simulate drug exposures in breast-fed infants, the estimated IDDs for venlafaxine and ODV were divided based on feeding frequency. For infants (less than 1 month old), IDDs were divided into 12 feeds every 2 h. For infants older than 1 month, IDDs were divided into 6 feeds and administered every 4 h. Simulated plasma ODV levels in infants formed by the metabolic conversion of venlafaxine and direct ODV intake from milk were added together to represent ODV levels in infants. The time-varying physiology features were also used when simulating drug exposures in infants [34].

Furthermore, to investigate the effects of various parameters to the observed variability, we simulated the exposures of venlafaxine and ODV in infants from birth to 1 year old, considering composite maternal/infant CYP2D6/CYP2C9/CYP2C19 polymorphisms, infant age-related physiological and biochemical maturations, as well as natural changes in feeding patterns and infant food consumption. The estimated average and maximum IDDs for both venlafaxine and ODV, derived from mothers who are EMs of CYP2D6/CYP2C9/CYP2C19 (All EMs), CYP2D6 PM, and PMs of CYP2D6/CYP2C9/CYP2C19 (all PMs), were administered to infants carrying the same CYP phenotypes based on the feeding frequencies corresponding to four age groups (less than 1 week, 1 week to 1 month, 1 to 6 months, 6 to 12 months). To simulate the likely real-world conditions, the average IDDs were used; for infants aged less than one week, 60% of UGT activity was assumed, reflecting the average activity at birth of UGT1A1, UGT1A3, UGT2B4, UGT2B15, and UGT2B17. Moreover, as food consumption typically increases in infants aged 6 months, with potentially reduced milk intake, we halved the IDD in this group (> 6 months). When investigating the worst-case scenarios, maximum IDDs were used without milk intake reduction in the older group, and a 30% UGT activity level was assumed for the youngest group, representing the lowest birth activity among these UGT isoforms (UGT1A1 and UGT2B17).