1 Introduction
Tobacco use is a major public health concern. Cigarette smoking is the leading cause of preventable death, estimated at over 7 million avoidable deaths per year worldwide and trending upwards [
1‐
3]. Many of these deaths are directly or indirectly caused by lung cancer, chronic obstructive pulmonary disease, or cardiovascular diseases, and the total health burden of tobacco use is even larger. Although nicotine is the substance in tobacco responsible for smoking addiction, mostly other constituents of combusted tobacco are linked to health problems [
4]. Nicotine replacement therapy (NRT) substitutes nicotine from tobacco products through other means of administration. Most clinically used NRT treatments are designed for oromucosal (i.e. buccal and sublingual) absorption. Nicorette gum, the first NRT product, was first registered in Switzerland in 1978. Since then, a number of different dosage forms for oromucosal absorption have been introduced, among them mouth spray, lozenges, sublingual tablets and inhalers. Nicotine replacement therapy has been shown to increase the relative smoking abstinence rate compared with placebo with 50–70% [
5] and is an important tool in achieving reduction in tobacco use and lung cancer incidence [
6].
Knowledge of nicotine pharmacokinetics is central to understanding its role in addiction. After intravenous (iv) administration, nicotine is extensively distributed to body tissues, including the brain, with a steady-state volume of distribution averaging 136–213 L (1.8–3.3 L/kg), exceeding corporal water volume [
7‐
12]. After inhalation (e.g. smoking), nicotine is very rapidly distributed to the brain. In contrast to inhalation from smoking, the transdermal, gastrointestinal, oromucosal and nasal routes of absorption result in a more gradual increase of nicotine concentrations in the brain and other peripheral tissues. This results in a lower addictive potential and facilitates subjects to quit smoking [
13].
The systemic clearance (CL) of nicotine is relatively high, around 0.8–1.5 L/min and depends on liver blood flow [
7‐
11]. Nicotine is extensively metabolised in the liver to its major metabolite, cotinine. To a lesser extent, nicotine is also metabolised in the lung and kidney. Renal CL of unchanged nicotine (approximately 35–90 mL/min) accounts for 5–10% of total elimination. The reported elimination half-life of nicotine is on average 2 h [
7], with substantial inter-individual variability (IIV).
Nicotine absorption is pH dependent [
7]. In acidic environments, nicotine (p
Ka 7.9) is ionised and does not readily cross membranes whereas in basic environments nicotine is readily absorbed through oral, nasal and pulmonary mucous membranes. At physiological blood pH (7.4), only about 31% of nicotine is non-ionised. For this reason, to enhance absorption, some buccally administered dosage forms are buffered to a basic pH [
7]. As a result of swallowing part of the administered dose, all buccal formulations also partially deliver the nicotine dose to the gastrointestinal tract. Swallowed nicotine is absorbed in the small intestine but undergoes extensive pre-systemic metabolism by the liver and has a relatively low (30–40%) bioavailability (
F) [
8‐
10,
12].
Nicotine is rapidly and efficiently absorbed through the skin, with a relatively high
F of 68–82% [
14,
15]. The nicotine concentration–time profiles after application of a nicotine patch appear to vary widely, depending on the type of patch used [
16,
17]. When applying a patch daily for a 16-h period, nicotine stored in a shallow skin compartment is absorbed during the non-application night-time period, and a repeated-dose study over 6 days showed no accumulation of plasma nicotine [
18].
Though several studies of intravenously administered nicotine have been reported, only a few model-based pharmacokinetic (PK) analyses were carried out [
9,
11,
19,
20], in 10, 11, 24 and 40 subjects, respectively. In these studies, a two-stage approach was used to characterise nicotine pharmacokinetics through a two-compartment model in most subjects. Reports on the pharmacokinetics of orally ingested nicotine have mainly been based on model independent analyses. Levi et al. [
21] conducted a population PK (PPK) analysis of data from 66 healthy subjects, where the well-stirred model for liver CL was incorporated, to characterise nicotine pharmacokinetics after oral administration. The PPK analysis of buccally administered nicotine in 702 healthy adult smokers was described by Marchand et al. [
22], using a zero-order absorption process.
The PK properties of nicotine when administered buccally have been described previously for the sublingual tablet [
23], inhaler [
24], gum [
25] and mouth spray [
26]. However, no model-based PK analyses were presented for these studies.
Several PK studies with transdermal nicotine have been published [
14‐
18], with limited model-based PK analyses. A population-based analysis was recently published; however, using only data during the washout after patch application [
27].
During the development of different NRT formulations, a large volume of single- and repeated-dose nicotine PK data in healthy smoking volunteers was collected across a wide range of doses, routes of administration and formulations. We integrated data across a broad range of nicotine formulations and studies by developing predictive PPK models. This further increased our understanding of nicotine pharmacokinetics and its link with craving through exposure–response analyses, aiming to better understand and optimise the use of NRT [
28]. Together, these models allow simulations of various dosing scenarios, e.g. for NRT combination treatment, and simulations to inform the design and development of new improved NRT products.
4 Discussion
Capitalising on extensive research with multiple Nicorette formulations, large datasets of rich PK data from many single- and repeated-dose studies were included in our analyses to build PPK models for nicotine administered as iv infusions, orally ingested microtablets, transdermal patches, and mouth spray, chewing gum, lozenge and inhaler formulations designed for oromucosal absorption. Because of the large sample sizes, estimates of fixed and random effects in the models were initially expected, and then confirmed, to be more robust and precise compared to single-subject/single-study analyses mostly employed thus far in researching the pharmacokinetics of nicotine.
To our knowledge, only limited PPK modelling has been performed to date for oral nicotine (one study) [
21], for nasal spray, gum, combustion cigarettes, and tobacco heating systems (eight studies) [
22] and for transdermal nicotine (one study) using data after patch removal only [
27]. There was a frequent occurrence of quantifiable, sometimes significant, pre-dose residual nicotine concentrations across all studies and formulations, likely from prior smoking as subjects were regular smokers instructed, but not forced, to be abstinent only for a limited period (12–36 h) prior to dosing. Handling this by using a hypothetical nicotine dose at the start of washout, and scaling the exposure using a hypothetical
F appeared to describe this reasonably well and allowed for profiles with high pre-dose exposure to still be included in the analysis. The estimated hypothetical dose of pre-dose nicotine will not be an unbiased estimate, as it assumes that the nicotine was delivered as a bolus at the start of washout. However, as it is only a nuisance parameter, this is of little consequence. However, some of the estimated hypothetical nicotine doses at the start of the washout period were very high (over 100 mg), which likely indicates non-compliance with the requirement not to smoke during washout in these subjects. For reference, the typical dose of nicotine delivered from cigarette smoking is in the order of 0.3–2 mg per cigarette [
2,
38,
39].
For iv nicotine, owing to the richness of data and the enhanced sensitivity of a population mixed-effects analysis combining data from all subjects across multiple studies, the data were significantly better (
p < 0.001) described with a three-compartment rather than a two-compartment model as done by others [
9,
11,
19,
20]. Consequently, a typical terminal elimination half-life of nicotine of 4.5 h was estimated, somewhat longer than the 1.7–3.4 h estimated half-life in previous studies [
9,
12]. Our estimate is likely to be unbiased as we did not exclude the BQL observations from the analysis.
The typical systemic plasma CL of nicotine for a 70-kg individual was estimated at 1.1 L/min, which is approaching hepatic blood flow and in the 0.8–1.8 L/min range reported in previous studies [
7‐
11]. The typical volume of distribution at steady state of nicotine was 4.2 L/kg, which is somewhat higher than the 1.8–3.3 L/kg range reported previously [
7‐
12]. This is compatible with our ability to describe a slower phase in the systemic disposition of nicotine, which accounts for 61% of the area under the curve.
For oral nicotine, the
F estimate was 40% in the repeated-dose study 92NNBT005 in 20 subjects. This estimate is in the middle of the range of previously published estimates, 17–69% [
9,
10,
12,
40]. The
F estimate of 22% from the oral single-dose study, 93NNBT007, was not used in developing the PPK models for buccal NRT formulations as it is based on only six subjects, the absorption was delayed and
F was at the lower end of the published estimates. While the explanation for this difference is unknown, it may well be due to the tablets, which were designed to be taken sublingually, in this study being ingested whole, while they were chewed before swallowing in the repeated-dose study.
For the repeated-dose studies with oral nicotine, but also for the buccal formulations, many individual profiles, and even some mean profiles (e.g. for orally administered nicotine), showed that accumulation stopped, or concentrations even decreased, for part of the day. This appeared to start within 1–2 h (oral) and 3–5 h (buccal) of the first dose and lasted up to after the last observations. This phenomenon was modelled as a function of time, either on CL (time-dependent increase; for mouth spray, lozenge and inhaler) or on F (time-dependent decrease; for chewing gum). As these changes mainly affect pre-dose trough concentrations in the repeated-dose data, an effect on F or on CL has similar impact on overall concentration–time profiles. It is possible that the effect with the chewing gum data is also caused by a change in CL, but no such model was found that could describe the data well.
Diurnal variation in CL and an increased CL after meals have previously been observed by Gries et al. [
11]. They found a diurnal variation in CL of ± 14% and a meal effect of +40%. This is in the same order of magnitude as what was observed in the current analysis. As nicotine is a high extraction ratio drug, CL can be expected to change with hepatic blood flow changes, such as can be caused by changes in physical activity, and by food intake.
For buccal nicotine, the absorption of nicotine from mouth spray, gum, lozenge and inhaler is rapid with peaks occurring shortly after the end of dosing. However, exploring individual profiles, absorption appears to be variable, with many subjects displaying a distinct double peak. The second peak is likely due to intestinal absorption of swallowed nicotine.
The estimated fraction of swallowed nicotine was lowest for chewing gum (55%) followed by mouth spray (61%), inhaler (67%) and lozenge (69%). Estimation of the fraction swallowed is influenced by two assumptions made when modelling buccal formulations: fixing the oral F to 40% and assuming a F of the oromucosally absorbed fraction to be complete. The rank order and relative proportions of the fractions swallowed, however, do not depend on these assumptions. The differences in fraction swallowed between mouth spray, inhaler and lozenge remain small. This leads to a similar exposure after administration of chewing gum as after the other formulations, even though only part of the dose (on average 64–79%) is released from the chewing gum. When dose normalising for the average dose released, the predicted relative exposure is highest for chewing gum.
For several of the buccal formulations, the fraction of nicotine that is swallowed appeared to increase with dose. This leads to a lower overall F for higher doses, as the F of intestinally absorbed nicotine is lower than the oromucosally absorbed nicotine. One possible reason for this is the potentially irritating effects of nicotine in the oral cavity, which may lead to increasing saliva production at higher doses.
The
Ka was fastest for Nicorette chewing gum, followed by mouth spray, lozenge, and Freshmint/Freshfruit-coated chewing gum, and, with a much slower
Ka, for the inhaler. In Fig.
11, it can clearly be seen that nicotine concentrations increase most rapidly after the mouth spray, followed by chewing gum, lozenge and finally inhaler. This apparent discrepancy between
Ka and the rate of increase in concentrations is owing to the absorption of nicotine from chewing gum being rate limited by the nicotine release from the formulation as the release rate constant, 2.8 h
−1, is much slower than the
Ka of 9–26 h
−1.
For transdermal nicotine, the data were best described by a model with two parallel pathways for release of nicotine from the patch, zero-order release until patch removal and first-order release for an estimated fraction of the time of patch application, followed by absorption of nicotine though three serial transit compartments with a transit rate constant
Ktr of 3.6 h
−1, corresponding to a mean residence time of 1.1 h. The release part of the model resembles that of the model of Gabrielsson and Weiner [
41]. While using a different functional form, the transit compartment model used for the absorption part can be expected to give similar predictions as the Weibull model used by Linakis et al. [
27].
The
F of nicotine released from both patch types (i.e. Nicorette patch and Invisipatch) was estimated at 76%, which is in the range of 68–82% previously reported [
9,
15]. To account for a lower than expected accumulation upon repeated dosing, CL was modelled to increase by 12% from 24 h after the first dose onwards. A plausible biological explanation for this is missing at this time. Additionally, for buccal formulations, an apparent increase in CL over time was observed, albeit earlier.
Two patch products were included in the dataset: Nicorette patch and Invisipatch. The estimated fraction of the dose released through the first-order process differed between products, 40% for Nicorette patch and 72% for Invisipatch, and the estimated model parameters for nicotine release (Fr1, lag time, Frdur1) were fairly different. These differences could be related to design differences of the two patch formulations: for the Nicorette patch, the adhesive layer in contact with skin contains only a small amount of nicotine at product release whereas for Invisipatch the layer in contact with the skin already has a maximum nicotine concentration. However, comparing the simulated profiles of the two patches (Fig.
12), nicotine concentrations overlap to a large extent and no meaningful differences in efficacy or safety should be expected. Previously, the two patch formulations were demonstrated to be bioequivalent [
42].
The PPK models developed in this research were used to conduct nicotine exposure–response analyses linking nicotine exposure to momentary craving and smoking cessation, as published elsewhere [
28,
43]. The PPK and the exposure–response models were used to build a simulation platform that is being used to (i) maximise the value of the existing products by understanding the effect of different dosing scenarios, including combinations treatments, in momentary craving and (ii) streamline the development of new NRT products by leveraging the nicotine PKPD knowledge in identifying the appropriate nicotine dose and delivery rate of a new product to achieve a desired momentary craving effect, and optimising the PKPD information that needs to be collected from the clinical studies of new products or combinations.
Despite the wealth of available data, there were data-related limitations to our analyses. The studies included in the analyses were not designed to be included in any integrated modelling or other meta-analyses. These studies were conducted over a 19-year period. The designs of the studies, however, were remarkably similar and for each formulation the studies were carried out at the same clinical unit and bioanalytical laboratory. The populations included were homogeneous as to demographic and other subject characteristics, in particular regarding age and body size characteristics. Furthermore, there were very few non-Caucasian individuals in the dataset. Therefore, the results may not be representative for all smoking populations globally.
Finally, we tried to simultaneously model iv and extravascular data, without success in terms of numerical convergence and stability of the model runs. This may be due to the use of the LAPLACIAN method in NONMEM, which can be unstable, but which was necessary for fitting the iv data, owing to the large number of observations BQL. Therefore, separate models were developed for oral, buccal and transdermal nicotine for which the CL and disposition parameters, including their IIVs, were fixed to the estimates of the final iv model. This is considered a reasonable approach as the parameters of the iv model are based on a large sample. When doing this, fits of the extravascular data generally appeared to be unbiased, and F estimates were reasonable.