Plasma pharmacokinetics of intravenous and intranasal oxytocin in nonpregnant adults

Abstract

Background

The development of oxytocin as a therapeutic agent outside of obstetrics has been hampered by antibody-based assays that lack specificity, leading to inconsistent and incompletely reported pharmacokinetic models to guide drug dosing. This study describes the population plasma pharmacokinetics of intravenous and intranasal oxytocin using a sensitive and specific liquid chromatography-tandem mass spectroscopy (LC/MS) assay.

Methods

Two studies in healthy adult men and nonpregnant women were performed, the first with intravenous oxytocin 16.7 μg over 1 or 10 min and the second with intravenous oxytocin 13.7 μg over 30 min and, on a separate day, intranasal oxytocin 100 μg (n=24). Venous plasma oxytocin concentration was measured using LC/MS and enzyme-linked immunosorbent assay. Pharmacokinetic parameters were estimated using NONMEM.

Results

The pharmacokinetics of intravenous oxytocin were well described by a two-compartment model (0% bias, 18% median inaccuracy). The two-compartment model for intranasal oxytocin was characterised by substantial subject-to-subject variability (9% median bias, 47% median inaccuracy). Nasal oxytocin bioavailability was 0.7%. Oxytocin samples assayed with LC/MS were systematically higher than simultaneous samples assayed with enzyme-linked immunosorbent assay.

Conclusions

The pharmacokinetics of intravenous oxytocin are well described by a two-compartment model. The low bioavailability (<1%) and large intersubject variability in plasma oxytocin after intranasal dosing could partially explain the inconsistent reports of oxytocin efficacy in the clinical literature with this delivery method. A publicly available simulator was created to guide oxytocin dosing in future studies.

Clinical trial registration

NCT 03929367 (Study 1) and NCT 05672667 (Study 2).

Editor’s key points.

• •Oxytocin, often administered intranasally, has beneficial effects on neuroprotection, anxiety, sleep disorders, autism, addiction, and pain, but shows mixed results in clinical studies.
• •This study describes the population plasma pharmacokinetics of intravenous and intranasal oxytocin using an improved and sensitive assay.
• •The pharmacokinetics of intravenous and intranasal oxytocin were well described by a two-compartment model; the model for intranasal oxytocin exhibited substantial subject-to-subject variability with low bioavailability.
• •These findings could explain the inconsistent reports of oxytocin efficacy with intranasal delivery, and will be helpful in future studies of oxytocin efficacy.

Preclinical studies suggest that oxytocin circuits and signalling exert beneficial effects for neuroprotection, anxiety, sleep disorders, autism, addiction, and pain.¹ Oxytocin itself is the only oxytocin receptor agonist clinically available for translational studies in the USA. Most clinical research with oxytocin uses intranasal (i.n.) delivery. Investigators commonly assume that the effects of oxytocin by this route are attributable to penetration into the central nervous system from the nasal cavity, although some data support a primary mechanism of absorption into the circulation and distribution to the brain.² The scientific rigour of this literature has been criticised for multiple deficits in study design and statistical analysis, as well as reliance on single doses and times of outcome assessment,^3,4 resulting in wide variability in efficacy or lack thereof.5, 6, 7, 8

A key barrier to the study and use of oxytocin outside of obstetrics is poor understanding of its pharmacokinetics (PK) and PK/pharmacodynamic (PD) relationships.⁹ Oxytocin is traditionally measured using antibody-based methods (enzyme-linked immunosorbent assay [ELISA] and radio-immunoassay [RIA]). These assays fail to correlate with each other in duplicate samples in plasma or cerebrospinal fluid¹⁰ and demonstrate widely varying plasma concentrations before and after oxytocin administration.11, 12, 13, 14 Liquid chromatography-tandem mass spectrometry (LC/MS) assays unambiguously measure intact cyclised oxytocin rather than oxytocin metabolites.15, 16, 17, 18 The population PK of intravenous (i.v.) oxytocin measured by LC/MS have not been reported. The PK of i.n. oxytocin by LC/MS have only been reported in a single subject.¹⁷ The investigators detected oxytocin for less than 60 min after a 100 μg dose (60 IU), at variance with antibody-based methods.^19,20 Plasma oxytocin after i.n. administration is relevant to its analgesic effects, as oxytocin acts directly on primary sensory afferents in rodents21, 22, 23 and surgical wound infiltration of oxytocin produces analgesia in humans.²⁴

The primary objective of this study was to define the plasma PK of oxytocin by LC/MS in humans after i.v. and i.n. administration. The secondary objectives were to create a publicly available tool using these models to guide dosing regimens in clinical studies. Lastly, we compared oxytocin PK from LC/MS and ELISA assays to understand the effect of oxytocin assay on our understanding of oxytocin pharmacology.

Methods

Two studies were performed at Atrium Health Wake Forest Baptist Clinical Research Unit after approval by the Atrium Wake Forest Health Sciences Institutional Review Board (IRB). Written informed consent was obtained from all subjects and both studies were overseen by a Data Safety Monitoring Board. Both studies were registered at https://clinicaltrials.gov before subject enrolment (Study 1: NCT 03929367; Study 2: NCT 05672667). Study 1 estimated intersubject variability in i.v. oxytocin plasma PK, compared PK models between ELISA and LC/MS assays, and explored PD endpoints for future studies. Study 2 defined population plasma oxytocin PK after i.v. and i.n. administration.

Oxytocin for i.v. (oxytocin injection, USP [16.7 μg ml⁻¹], Fresenius Kabi, Lake Zurich, IL, USA) or i.n. (TNX-1900 [25 μg 100 μl⁻¹ per spray actuation], Tonix Pharmaceuticals, Inc., Chatham, NJ, USA) administration was dispensed by the Atrium Wake Forest Health Sciences Research Pharmacy. Studies were performed under Investigational New Drug (IND) application 152732 (sponsor: JCE) for i.v. oxytocin and i.n. oxytocin (cross-referencing IND 128288 [sponsor: Tonix Pharmaceuticals, Inc.] for i.n. oxytocin) from the US Food and Drug Administration. Tonix Pharmaceuticals, Inc., provided TNX-1900 (i.n. oxytocin) and matching placebo (same excipients minus oxytocin) for i.n. administration and training purposes, in metred spray glass vials, without charge. TNX-1900 is an investigational new drug and has not been approved for any indication.

Study population

Healthy (ASA physical status 1 or 2) adult men and women of BMI <40 kg m⁻² and aged 18–60 yr in Study 1 and 18–75 yr in Study 2 were recruited from an IRB-approved database of volunteers and from advertising. We only recruited women of child-bearing potential or less than 1 yr after menopause if they were practicing highly effective methods of birth control. A negative urine pregnancy test was required before oxytocin administration.

We excluded individuals with a significant reaction to any ingredient of i.v. oxytocin and those with latex allergy. We also excluded women in whom increased myometrial tone by oxytocin could be detrimental; those with neuropathy, chronic pain, or diabetes mellitus or taking daily benzodiazepines or pain medications; and those with current or history of ventricular tachycardia, atrial fibrillation, prolonged QT interval, or hyponatraemia. For Study 2, we also excluded individuals with use of products administered intranasally or with chronic nasal obstruction or local pathology in nostril pathways which would prevent i.n. administration.

Sample size for Study 1 was a convenience sample of 10 subjects to estimate PK parameters and their variance to inform the design of Study 2. Sample size for Study 2 was determined using bootstrap analyses, which indicated 20 subjects would estimate parameters with a corrected 95% confidence interval (CI) range within 20% of the mean estimated value. Sample size was increased to 24 to allow stratification by age, weight, and sex.

Study 1

Subjects came to the Clinical Research Unit twice separated by at least 1 day. On the first day, we obtained a detailed medical history. We also tested subjects using vibratory and heat stimuli. On the second day, we inserted a catheter into a vein in an upper extremity. After obtaining baseline sensory testing measures, we administered oxytocin, 16.7 μg i.v. over 60 s (first three subjects) or 10 min (remaining eight subjects). We sampled venous blood (5 ml) before and at intervals to 240 min after completing the oxytocin infusion. We assessed the response to vibration and heat sensory stimuli at intervals during this period.

We measured peripheral oxyhaemoglobin saturation, blood pressure, and heart rate noninvasively before and at intervals to 240 min after starting oxytocin. We queried subjects for any adverse events at regular intervals and asked about any adverse events between queries. The queries were conducted in person on the study day and by phone for the next 5 days.

Study 2

Subjects came to the Clinical Research Unit twice separated by at least 5 days. On the first day, we inserted two i.v. catheters in upper extremities, one for oxytocin infusion and the other for blood sampling. After baseline measures, we infused oxytocin, 13.7 μg i.v. over 30 min. We obtained venous blood samples at intervals to 90 min after starting oxytocin.

On the second visit, we inserted an i.v. catheter for blood sampling. After baseline measures, subjects self-administered i.n. oxytocin 100 μg. At time 0, the subjects administered one 25 μg spray into each nostril, and at 5 min later, subjects administered another 25 μg spray into each nostril, for a total dose of 100 μg. We drew venous blood samples at intervals to 60 min after the initiation of i.n. oxytocin. We examined the nasal cavity for signs of irritation, oedema, or inflammation before and at the end of the study.

We recorded peripheral oxyhaemoglobin saturation, blood pressure, and heart rate noninvasively before and at intervals to 120 min after beginning oxytocin administration. We queried subjects for adverse events during the study day and over the phone for the next 7 days.

Plasma sample preparation and oxytocin assays

Blood samples were acquired in glass vacutainers containing ethylenediaminetetraacetic acid (EDTA) and immediately placed on ice before centrifugation. Supernatant plasma was stored at −80ºC and transported on dry ice to the laboratory (Syneos Health, inVentiv Health Clinique, Inc., Québec, QC, Canada). Oxytocin was quantified using an analytical method based on automated solid phase extraction, followed by LC/MS. The limit of sensitivity was 2 pg ml⁻¹ using a 350 μl aliquot of human plasma.

Duplicate samples from Study 1 were assayed by ELISA using a commercially available ELISA kit (#ADI-900-153A, Enzo Life Sciences, Farmingdale, NY, USA) according to the manufacturer's instructions. The limit of sensitivity of the assay was 15 pg ml⁻¹, and the intra-assay coefficient of variation was 11.6%.

Pharmacokinetic analysis

The parameters of one- and two-compartment pharmacokinetic models were fit to the observed plasma concentrations using NONMEM Version 7.5 (Icon, Dublin, Ireland) with first-order conditional estimation. The model was parameterised in central and (if necessary) peripheral volumes and clearances. Intersubject variability was estimated assuming a lognormal distribution. Intrasubject (residual) variability was estimated using an additive plus proportional variance model.

As described in Supplementary material, multiple models examined the influence of age, weight, height, BMI, and sex on oxytocin volumes and clearances. The contribution of patient covariates to the model was estimated using the improvement (decrease) in the NONMEM objective function (−2 log likelihood) and any decrease in model inaccuracy. Model inaccuracy was defined as the median absolute prediction error (MDAPE): median (absolute (measured−population predicted)/population predicted).

The PK after i.v. administration were used to inform the i.n. pharmacokinetic model. In six subjects, the first sample after i.n. delivery was below the level of quantitation. These samples were imputed to be half of the limit of quantitation (i.e. 1 pg ml⁻¹).

Leveraging the crossover design, we started by assuming that each individual's PK parameters were the same for the i.v. and i.n. sessions. With the volumes and clearances known from the i.v. arm, the only parameters estimated from the i.n. arm of the study were absolute bioavailability, the first-order absorption rate, and a zero-order lag (phase shift) between dose and initial increase in concentration. ADVAN4 from the NONMEM ADVAN library was used for this analysis.

This approach yielded poor fits to the data. We therefore followed the analysis described earlier with a second NONMEM analysis, also with ADVAN4. We fixed the absolute bioavailability that was estimated in the first step described in the previous paragraph to the population estimate (i.e. the estimate at ETA=0 in NONMEM parlance). With the bioavailability fixed at the population estimate, we estimated the first-order absorption rate constant, as well as the parameters of the two-compartment model of systemic drug disposition, again using ADVAN4.

Goodness of fit for the PK models was assessed with the NONMEM objective function, model inaccuracy (MDAPE), log likelihood profiles of all model parameters, and bootstrap re-estimation²⁵ of the volumes, clearances, absorption rate constant, and bioavailability. The bootstrap analysis was based on 3000 iterations for each model. Model bias was calculated as the median (measured−predicted)/predicted. Bias and inaccuracy were also calculated for ‘out of sample’ observations using a round-robin jackknife approach that fit the data from all but one individual, and then predicted the oxytocin concentration in the excluded individual. (Note: in a round-robin tournament, each team competes against every other team once in sequence. A round-robin jackknife subsets the data and sequentially tests the prediction of the excluded observations using the fit from the remaining data exactly once.²⁶)

The NONMEM Control Files and original data are provided in the Supplementary material.

Results

Study populations

Study 1 occurred between May 21 and July 18 in 2019. Eleven subjects were studied because of a technical failure in sensory testing in one subject, but PK samples were obtained from all subjects. Characteristics for participants in Study 1 were mean age 38 (range, 22–58) yr, 55% female, mean height 173 (sd 6.5) cm, mean weight 86 (sd 11) kg, baseline mean heart rate 69 (sd 19) beats min⁻¹, and baseline mean arterial pressure 90 (sd 6) mm Hg.

Study 2 occurred between February 27 and August 14 in 2023. Twenty-five subjects consented to the study. One subject withdrew before study, for 24 evaluable subjects. Characteristics for participants in Study 2 were age 40 (range, 21–68) yr, 50% female, height 171 (sd 7.9) cm, weight 82 (sd 10) kg, baseline heart rate 64 (sd 9) beats min⁻¹, and baseline mean arterial pressure 82 (sd 10) mm Hg.

The maximum BMI was 33.6 kg m⁻². There were five subjects with Class 1 obesity and none with Class 2 or 3 obesity. All subjects were White and non-Hispanic.

Pharmacokinetic analysis

Figure 1 shows the raw data after i.v. delivery (blue dots) and i.n. delivery (green dots). Despite a far higher dose (100 vs 16.7 μg), oxytocin concentrations after i.n. delivery were typically an order of magnitude less than the concentrations after i.v. delivery. The concentrations after i.n. delivery of oxytocin were also more variable than after i.v. delivery.

Fig 1.

Oxytocin observations after intravenous (shades of blue) or intranasal (green) administration of oxytocin. The lines show the LOESS average of the predictions. As documented in the results, intranasal administration resulted in concentrations one to two orders of magnitude less than intravenous administration, reflecting the exceptionally low bioavailability of intranasal administration. i.n., intranasal; i.v., intravenous; LOESS, locally estimated scatterplot smoothing.

The PK of both i.v. and i.n. oxytocin using the LC/MS assay were best described by two-compartment models (Table 1). As described in the Supplementary material, the model with no covariates had 22% inaccuracy and 0% bias. Model fit was improved by the incorporation of patient covariates. Model D in the Supplementary material was chosen for presentation in this manuscript. This model scales the clearances and volumes to patient weight. It also includes the influence of age on clearance. Bias and inaccuracy for Model D were 0% and 18%, respectively. Jackknife analysis demonstrated that the ‘out of sample’ bias and inaccuracy were also 0% and 18%, respectively.

Table 1.

Pharmacokinetic parameter estimates for intravenous and intranasal oxytocin. The intravenous model is Model D from the Supplementary material. Omega² is the estimated interindividual variance of the parameters. The 95% CIs were determined by bootstrap resampling as described in the methods. Wt is the weight in kilograms. ∗Not available: NONMEM was unable to determine interindividual variability for these terms. 95% CI, 95% confidence interval.

As shown in Fig. 2, the measured vs predicted concentrations for the i.v. pharmacokinetic model were unbiased over 2 orders of magnitude (left upper graph). The measured/predicted concentrations were also unbiased over time (left lower figure). As shown in Fig. 3, the log likelihood profiles for the i.v. pharmacokinetic model were smoothly parabolic. Both the likelihood profiles and the cumulative distribution (dotted green lines) fell within a narrow range of the parameter estimate, indicating that NONMEM determined the pharmacokinetic answers with reasonable confidence.

Fig 2.

Figures demonstrating the goodness of fit for the intravenous (left) and intranasal (right) pharmacokinetic models. The upper graphs show the measured (Y-axis) vs population predicted (X-axis) concentrations, with the line of identity shown in red and a LOESS smoother shown in blue. The lower graphs show measured and population predicted concentrations vs time. The red line in the lower graphs at 1.0 shows the predictions of a model in which the measured and predicted concentrations agreed perfectly. The intravenous pharmacokinetics are unbiased (bias=0%) and accurately predict the concentrations over time (median inaccuracy=18%). The intranasal pharmacokinetics show more bias (bias=+9%) and variability (median inaccuracy=47%). None of the graphs suggest model misspecification. LOESS, locally estimated scatterplot smoothing

Fig 3.

Log likelihood profiles and bootstrap results for the intravenous oxytocin pharmacokinetic model. The smooth parabolic shapes of the log likelihood profile for each parameter demonstrate that NONMEM was readily able to determine an optimum fit for each parameter. The dotted red line is the change in log likelihood associated with P<0.05. The dotted blue line shows the change in log likelihood associated with P<0.01. The dotted green line shows the cumulative distribution function for 3000 bootstrap replications. The green line with arrows shows the 95% CI calculated from the cumulative distribution of the bootstraps. The graphs indicate that NONMEM was able to estimate each parameter with reasonable confidence. Specific −2 log likelihood determinations appear as black dots on the log likelihood profile. 95% CI, 95% confidence interval; I.V., intravenous.

The pharmacokinetic model for i.n. delivery described the data less well. The absolute bioavailability of i.n. oxytocin was 0.66% (Table 1). Oxytocin concentrations after 100 μg of i.n. oxytocin were an order of magnitude below the concentrations after 16.7 μg of i.v. oxytocin (Fig 1).

As shown in Fig. 2, the measured vs predicted concentrations were unbiased (right upper graph), as were the measured/predicted concentrations over time (right lower graph). However, the model was not accurate. Model bias and inaccuracy were +9% and 47%, respectively.²⁷ The high inaccuracy reflects the variability in observed concentrations after i.n. delivery (Fig 1, green dots).

As shown in Fig. 4, the log likelihood profiles were parabolic for systemic clearance, intercompartmental clearance, and bioavailability. The log likelihood profiles exhibited odd shapes for central and peripheral volumes, as well as the absorption rate constant. These odd shapes reflect difficulty estimating the volume parameters because of the underlying variability in the observations and the limited duration of observation. NONMEM similarly had difficulty estimating the absorption rate constant. However, the bootstrap results in Fig. 4 suggest that despite the difficulties creating parabolic log likelihood profiles, the estimates of systemic clearance, intercompartmental clearance, and bioavailability were readily reproduced by the 3000 bootstrap iterations. None of the subject covariates improved the i.n. pharmacokinetic model.

Fig 4.

Log likelihood profiles and bootstrap results for the intranasal oxytocin pharmacokinetic model. The smooth parabolic shape of the log likelihood profile for systemic clearance, intercompartmental clearance, and bioavailability indicates that NONMEM estimates these parameters with reasonable confidence. The non-parabolic distortions in the log likelihood profiles for central volume, peripheral volume, and the absorption rate constant indicate that NONMEM had difficulty fitting for these parameters. This difficulty reflects both the high intersubject variability and the relatively brief period of sampling. The dotted red line is the change in log likelihood associated with P<0.05. The dotted blue line shows the change in log likelihood associated with P<0.01. The dotted green line shows the cumulative distribution function for 3000 bootstrap replications. The green line with arrows shows the 95% confidence interval calculated from the cumulative distribution of the bootstraps. Specific −2 log likelihood determinations appear as black dots on the log likelihood profile. 95% CI, 95% confidence interval; I.N., intranasal.

Assay comparison

Figure 5 shows the differences between simultaneous measurements of oxytocin using LC/MS and ELISA in Study 1. Figure 5 demonstrates that oxytocin concentrations measured by ELISA were about 40% lower than the simultaneous measurements with LC/MS. Figure 5 also shows the difference in the duration of quantifiable oxytocin between the two assays.

Fig 5.

Differences between simultaneous measurements of oxytocin using LC/MS and ELISA in Study 1. The graph on the left shows simultaneous measurements. The black line is the line of identity. The red line is a linear regression through the simultaneous measurements. The figure on the upper right shows the limits of agreement, defined as the difference between assays (LC/MS−ELISA) divided by the mean of the paired measurements. The figure to the lower right shows simultaneous assays in Study 1. The orange and teal lines are a LOESS smoother. The LC/MS assay could measure oxytocin concentrations for well over 2 h. The less sensitive ELISA assay was generally unable to measure concentration 1 h after drug administration. ELISA, enzyme-linked immunosorbent assay; LC/MS, liquid chromatography-tandem mass spectroscopy; LOESS, locally estimated scatterplot smoothing.

Discussion

Animal studies and clinical genetic and epigenetic associations support a role for oxytocin signalling in salience and strength of social interactions and in ameliorating a variety of diseases, yet i.n. oxytocin shows subtle or no therapeutic efficacy in many studies. The key findings of this study using a highly selective and sensitive oxytocin assay suggest ways forward to address this discrepancy. The population PK of oxytocin in plasma after i.v. administration were accurate and reliable, leading to the generation of a publicly available tool (https://steveshafer.shinyapps.io/oxytocin_simulator) to guide dosing in future studies and to generate PK/PD models from previous publications, particularly with repeated PD measures that include estimation of time to peak effect.²⁸ In contrast, i.n. administration resulted in highly variable concentrations with low (0.66%) bioavailability and mild rhinorrhoea (one subject) and sternutation (two subjects). These adverse effects likely reflect administration site reactions. Systemic adverse events occurred only by the i.v. route, were present only in the first minutes of infusion despite increasing plasma concentrations thereafter, and were related to initial infusion rate.

Intravenous oxytocin pharmacokinetics

The PK model had a median absolute error of 18%. This is nearly as accurate as the PK model for remifentanil,²⁹ which is considered the i.v. anaesthetic with the most predictable PK. This accuracy attests to both the accuracy of the assay and the quality of the two-compartment fit. Importantly, PK parameters differed significantly between LC/MS and ELISA measurements from duplicate samples, with ELISA also demonstrating poorer assay accuracy and sensitivity. The low accuracy of the i.n. PK reflects the considerable variability introduced by spraying drug into the nasal cavity and the variability in subsequent absorption. Age, weight, and BMI modestly improved the fit of the i.v. pharmacokinetic model. None of the tested covariates improved the fit of the i.n. PK.

These results provide new interpretation to two recent studies which used the same LC/MS assay, both of which used a formulation of oxytocin that does not require refrigeration and can be administered via inhalation. The first compared four different inhaled doses with intramuscular injection in nonpregnant women.³⁰ Applying our population PK to their observed plasma oxytocin concentrations suggests that intramuscular oxytocin is 36% bioavailable and the inhaled oxytocin formulation is 2% bioavailable. The second compared inhaled and i.v. injection in nonpregnant women,³¹ observing a 2–3% bioavailability of the inhaled formulation and peak concentrations after 8.4 μg i.v. over 30 s or 5 min of 1000 and 700 pg ml⁻¹, similar to simulations from our population PK model (900 and 680 pg ml⁻¹).

Intranasal oxytocin pharmacokinetics

Oxytocin, like other peptides and small molecules, can reach the central nervous system, especially the olfactory bulb, trigeminal nucleus, and meninges, via transport from the nasal cavity.³² Alternatively, increased circulating oxytocin in blood after i.n. spray can cross the blood–brain barrier via specific receptors.³³ In support of this pathway, prevention of absorption by a preceding vasoconstrictor spray blocks EEG effects from i.n. oxytocin in humans.² The current study demonstrates that the bioavailability of oxytocin in plasma after i.n. administration is <1%. Our estimate of 0.66% bioavailability is considerably less than the 11% reported by Martins and colleagues,^27,34 who used ELISA to determine oxytocin concentration. The difference might be related to the use of the ELISA assay, or to a difference in the excipients of their formulation.

Limitations

Generalisability is limited by the single-centre design, lack of severely obese subjects, and, despite recruitment efforts, lack of race and ethnic diversity. Additionally, our results likely do not represent oxytocin PK in parturients. Using the same LC/MS assay as we used, Gajewska-Knapik and colleagues³¹ observed that plasma oxytocin concentrations after oxytocin 16.7 μg i.m. or 400 μg of inhaled powder administered immediately after delivery were near or below the limits of detection (2 pg ml⁻¹). In nonpregnant women, the same dose of inhaled oxytocin could be quantified for 3 h after administration. This is consistent with the known increase in the rate of oxytocin metabolism in the third trimester of pregnancy.³⁵

The difference between the LC/MS and ELISA assay results could be attributable to either errors in the standard curves used by each assay or differences in the moiety measured (e.g. parent vs parent plus metabolites). We were not able to determine the cause of the differences between the assays.

Conclusions

We created population pharmacokinetic models for i.v. and i.n. oxytocin using a highly specific and selective LC/MS assay. We created a free online simulator (https://steveshafer.shinyapps.io/oxytocin_simulator) that applies these models. As of this writing, the simulator uses Model G (see Supplementary material) for i.v. oxytocin, and the model in this paper for i.n. oxytocin. Oxytocin concentrations after i.n. administration of 100 μg are two orders of magnitude less than observed during and after i.v. administration of 16.7 μg. This suggests that the PD effects of oxytocin after i.n. administration are unlikely caused by systemic absorption.

Authors’ contributions

Study design, data interpretation, drafting of the manuscript, and critical revision of the manuscript: all authors

Data acquisition: RSC, JCE, DGR, SM

Data analysis: SLS, JCE

Funding

US National Institutes of Health (P01 NS119159 to JCE and SLS).

Data sharing

The raw data (Excel CSV format) and NONMEM control are available in the online supplementary material.

Declarations of interest

DTH and GMS are employees of Tonix Pharmaceuticals, Inc. SLS is a consultant to, has equity interest in, and has received research funding from Concentric Analgesics (San Francisco, CA, USA), related to development of vocacapsaicin, a nonopioid analgesic unrelated to oxytocin. DGR, SM, RSC, and JCE declare that they have no conflicts of interest.

Handling Editor: Hugh C Hemmings Jr

Appendix A. Supplementary data

The following are the Supplementary data to this article: