Measured and Modeled Nicotine Emissions Predict Systemic Nicotine Exposure Across Sub-Ohm and Pod-Based ENDS

Abstract

Background:

Electronic nicotine delivery systems (ENDS) can deliver nicotine at levels that rival or exceed cigarettes, shaping both smoking cessation outcomes and the emergence of nicotine dependence in previously nicotine-naïve users. We evaluated whether systemic nicotine exposure is associated with emitted nicotine dose measured in the laboratory and estimated from ENDS device physics.

Methods:

Across two clinical studies, we compared plasma nicotine boost following ad libitum ENDS puffing sessions with nicotine doses derived by two complementary approaches: (1) analysis of nicotine content in puffing-machine–generated aerosols, and (2) physics-based computer simulations.

Results:

Across both clinical datasets and three device types, plasma nicotine boost was associated with emitted nicotine dose measured in the laboratory (~40% of variance explained) and with that predicted by a first-principles model of coil heating and liquid vaporization (~33%), despite not incorporating individual-participant physiological differences.

Conclusions:

These results support a mechanistic link between ENDS device parameters and systemic nicotine exposure. Physics-based models may support prospective population-level evaluation of nicotine delivery, including for hypothetical products for which emissions data are not yet available.

INTRODUCTION

Electronic nicotine delivery systems (ENDS) have transformed how nicotine is consumed. Compared to combustible tobacco products, they allow great flexibility in the place and manner of nicotine consumption throughout the day and provide people with an opportunity to inhale nicotine discreetly where tobacco use is not permitted. Although advertised as alternatives for current cigarette smokers, two-thirds of current ENDS consumers in the USA aged 18-24 have never smoked [1]. ENDS therefore occupy a dual role in the nicotine landscape: they may help smokers transition away from cigarettes but can also lead nicotine-naïve users toward dependence and, for some, to later cigarette smoking [2]. The public health impact of ENDS depends both on exposure to nicotine, the primary addictive constituent of tobacco products, and on exposure to toxicants generated during aerosol formation. While the long-term health consequences of ENDS remain uncertain, an important challenge for product regulation is understanding how device design and user behavior influence nicotine delivery and whether systemic nicotine exposure can be predicted from product emissions.

ENDS generate an inhalable aerosol by heating and vaporizing a liquid composed of propylene glycol, vegetable glycerin, flavorants, and nicotine, the primary dependence-producing constituent of the aerosol and the principal target of ENDS regulatory frameworks, including proposals to limit nicotine concentration or flux. ENDS differ widely in atomizer design, electrical power, and liquid formulation. As a result, some products deliver nicotine ineffectively [3, 4], while others match and sometimes exceed the nicotine delivery profile of a tobacco cigarette [5, 6]. This variability may influence both the degree to which ENDS help people who smoke to quit smoking and the extent to which ENDS recruit nicotine-naïve individuals into a lifetime of nicotine dependence [7, 8]. Despite extensive analytical and clinical testing, the relationship between emitted nicotine mass, or “extracted dose”, and systemic nicotine exposure has not been demonstrated empirically across devices and conditions. Clarifying the relationship between emitted dose and systemic exposure is essential for regulatory science, but it also motivates a broader question: can systemic nicotine delivery be predicted from first principles, without direct physical measurement? In theory, a physics-based model of ENDS operation could estimate emitted and inhaled nicotine doses directly from design and liquid parameters. Such a framework would extend regulatory assessment beyond existing products, enabling exploration of regulation-driven design modifications and emerging technologies before they reach the market. Although theoretical models have successfully reproduced laboratory measurements of nicotine emissions, their capacity to predict human nicotine exposure has not been evaluated empirically.

Here, we address both challenges. First, we examine whether measured nicotine emissions predict observed blood nicotine concentrations in clinical studies spanning ENDS devices, liquids, and user populations. Second, we test whether first-principles predictions of emitted nicotine reproduce the same relationship—assessing the feasibility of predicting systemic nicotine exposure directly from device-level physics. Demonstrating such correspondence would establish a mechanistic link between ENDS engineering parameters, emitted dose, and human exposure, providing a foundation for predictive evaluation of nicotine delivery systems.

In the present work, we compared observed plasma nicotine concentrations following ad libitum ENDS in the clinical lab with estimated nicotine doses derived by two complementary approaches: (1) chemical analysis of nicotine content in puffing-machine–generated aerosols, and (2) first-principles computer simulations of the same ENDS devices operated under the observed puffing behavior. Figure 1 summarizes the overall study framework, showing the clinical, analytical, and virtual laboratory components and their corresponding outputs—measured exposure, measured dose, and computed dose.

Figure 1. Study framework linking ENDS design, emissions, and human exposure.

Clinical laboratory studies at Virginia Commonwealth University (VCU) and Florida International University (FIU) provided plasma nicotine concentrations and puff topography records from ad libitum ENDS use. These puffing records were reproduced at the American University of Beirut (AUB) using a digitally controlled puffing robot (ALVIN) to generate aerosols for chemical analysis of nicotine content, representing the measured emitted dose. The same records and device parameters were used as inputs to a first-principles computational model (CALVIN) to simulate nicotine emissions. Measured and simulated nicotine doses were compared with observed plasma nicotine boosts to evaluate the predictive value of computer-simulated dose to systemic nicotine exposure.

METHODS

Clinical studies were conducted at Virginia Commonwealth University (VCU) and Florida International University (FIU), and the analytical and computational work was conducted at the American University of Beirut (AUB). Each clinical site provided plasma nicotine measurements and puff topography recordings from ad libitum ENDS use sessions. At AUB, the puff topography records were subsequently reproduced, or “played back,” on a puffing machine (the Aerosol Lab Vaping Instrument, ALVIN) and on a virtual puffing machine—a physics-based computational model, CALVIN—to measure and simulate emitted nicotine dose, respectively. At VCU, all conditions used a sub-ohm tank ENDS, whereas at FIU, disposable pod-based products were used. For each ad libitum use session, emitted nicotine dose was computed cumulatively from the observed puffing behavior over that participant’s analyzed session window. Plasma nicotine boost was computed as the post-session minus pre-session nicotine concentration for that same session, so both dose and boost refer to the same session-specific exposure interval.

Clinical laboratory

The de-identified plasma nicotine and puff topography data were obtained from two previously published clinical studies conducted at VCU [9] and FIU [10], each approved by the institutional review board of its respective institution. Those studies were designed to examine the effects of ENDS use conditions, such as nicotine concentration and device power, on nicotine delivery, puff topography, and subjective effects. The present work is a distinct secondary analysis, testing whether nicotine dose (measured via ALVIN and simulated via CALVIN) predicts plasma nicotine boost. Each study measured pre- and post-use plasma nicotine concentrations and recorded puff topography via a flow sensor attached to the ENDS mouthpiece during ad libitum use sessions. The resulting datasets provided the observed nicotine boost (“boost” = post–pre plasma nicotine, ng/ml) and the puffing behavior used to estimate emitted nicotine dose.

VCU study design and data processing

The VCU study enrolled 27 participants who each completed six laboratory sessions arranged in a Latin square design varying by liquid nicotine concentration (10, 15, and 30 mg/mL) and power setting (15 and 30 W). Participants used a Subox Mini-C tank (SSOC nichrome coil with cotton wick) paired with a variable-power Kangertech battery. The tank was filled with 3.5 mL of liquid composed of 30/70 propylene glycol (PG)/vegetable glycerin (VG) by volume and protonated nicotine (nicotine benzoate).

Participants were instructed to abstain from tobacco- and nicotine-containing products for ≥12 hours before each session. Each visit comprised a 10-puff directed bout followed by a 1.5-hour ad libitum bout, separated by a 30-minute rest. Blood samples were collected before and after each bout, centrifuged, and analyzed for plasma nicotine content as previously described [9].

For 15 randomly selected participants, the recorded puff topography from each of the six ad libitum bouts of each participant was processed to extract quantitative puffing parameters and subsequently used for both machine playback and computational simulation of emitted nicotine dose, as described further below.

FIU Study Design and Data Processing

The FIU study enrolled 50 adults who used ENDS and completed two sessions using the same pod-based device, either JUUL (n = 26) or NJOY Ace (n = 24), under two nicotine concentrations: 3% and 5% for JUUL, and 2.4% and 5% for NJOY Ace, in randomized order. JUUL is a temperature-regulated coil-and-wick device, while NJOY Ace employs a porous ceramic brick with a thin-film heating element embedded in the surface [11].

In each session, participants used the assigned device ad libitum following at least 12 hours of nicotine abstinence; total session time varied across participants (median ~62 minutes, IQR ~31–75) [10]. Blood was drawn prior to the first puff and within ten minutes of the final puff, then centrifuged and analyzed for plasma nicotine content as detailed in Ferdous et al. (2024) [10].

Complete plasma data and puff topography records were available for 93 of the 100 clinical sessions at FIU. These records were processed using the same pipeline as for the VCU study and provided the input data for both ALVIN playback and CALVIN simulations.

Puff Topography Processing

Puff topography recordings collected in both clinical studies were analyzed to quantify individual puffing behavior. Individual puffs were identified as sequences of non-zero flow-rate values lasting at least 0.5s. For each puff, duration and average flow rate were computed, and session-level means were calculated across the ad libitum bouts. Summary statistics for puffing behavior and plasma nicotine boost for each device, power setting, and liquid nicotine concentration are presented in Table 1.

Table 1.

Mean (SD) of puff topography variables and plasma nicotine exposure for each condition of the clinical studies.

Study site /product		Condition	Nicotine (mg/ ml)	Power (W)	Puffs drawn	Puff duration (s)	Total puff time (s)	Flow rate (lpm)	Puff vol (ml)	Plasmanic boost (ng/mL)
VCU	Subox Mini-C	1	10	15	40(23)	3.64(1.78)	151(128)	8.37(3.88)	549(365)	10.0(10.8)
		2	10	30	35(14)	2.24(1.12)	79.4(54.7)	8.62(3.41)	348(226)	9.01(10.5)
		3	15	15	40(23)	3.39(1.44)	141(124)	8.88(3.55)	543(354)	10.2(17.5)
		4	15	30	34(17)	2(0.638)	70.5(44.3)	7.93(4.23)	286(202)	12.1(14.8)
		5	30	15	31(13)	2.97(1.24)	98.8(65.6)	8.74(4.55)	478(320)	12.8(13.3)
		6	30	30	21(16)	1.75(1.02)	43.8(55.4)	8.30(4.10)	269(211)	7.5(11.7)
FIU	NJOY Ace	1a	28	4	43(51)	2.49(1.26)	76(44.7)	1.04(0.33)	45(27)	7.1(8.4)
		2a	57	4	30(23)	2.41(1.05)	64.7(40.1)	1.05(0.28)	43(26)	12.5(10.3)
	JUUL	1b	36	1.6	86(137)	3.09(1.68)	173(148)	1.90(0.60)	93(56)	6.8(5.1)
		2b	60	1.6	60(55)	2.94(1.76)	130(82.4)	1.87(0.72)	85(46)	8.3(5.9)

Note: n = number of sessions; N = number of participants. VCU: n = 90 (puff topography and nicotine boost), N = 15. FIU: n = 99 (puff topography), n = 93 (nicotine boost), N = 50.

Measured dose: machine playback and nicotine quantification

Puffing behavior recorded during ad libitum use sessions at VCU and FIU was reproduced using ALVIN, a digitally controlled puffing machine capable of replicating detailed puff velocity records with 0.1-s resolution. Each playback session was programmed to reproduce the recorded puff-by-puff sequence (the full time-series, not a session average) from an individual clinical use bout under the same device, liquid, and power conditions used in the corresponding clinical session.

Device conditioning and setup

To ensure consistent device performance, each ENDS product underwent a standardized conditioning protocol before playback. For the Subox Mini-C, the procedure consisted of 75 four-second puffs at 1 L/min with power off, followed by three puffs at 15 W. For the JUUL and NJOY Ace devices, conditioning consisted of three four-second puffs at a flow rate of 1.5 L/min with the devices powered.

A new coil or pod from the same lot or retail package used in the clinical studies was installed for each playback session. The battery was fully charged between sessions and reused across samples.

Aerosol collection

As illustrated in Figure. 2, a glass fiber filter pad (47 mm Pall Type A/E) housed in a polycarbonate holder (Pall 1119) was positioned immediately downstream of the ENDS mouthpiece to collect the total particulate matter (TPM) generated during each playback session. Filter pads used during the conditioning protocol were discarded, and a new pad was installed before each playback.

Figure 2. ALVIN puffing machine sampling setup.

ALVIN reproduces the ad lib puffing sequences of individuals who participated in the clinical studies. The ENDS device for each session is connected to a filter trap containing a high efficiency glass fiber filter, where >99.99% of the particulate matter mass exiting the ENDS mouthpiece is collected for off-line nicotine analysis.

After completion of each session, the filter assembly was weighed using a digital microbalance (AND GR-120, 0.1 mg resolution) to determine TPM gravimetrically. Static charge was neutralized prior to weighing with an anti-static brush. The device was also weighed before and after each session to determine liquid mass consumption.

Nicotine quantification

Nicotine mass on the filter was extracted by placing the pad in a sealed glass vial containing 6 mL of ethyl acetate and shaking for 30 minutes. The resulting solution was diluted with ethyl acetate and spiked with hexadecane (5 μg/mL) as an internal standard before analysis by gas chromatography with flame ionization detection (GC-FID). Nicotine concentrations were quantified against a calibration curve (1–80 μg/mL) prepared from standard solutions containing the same internal standard. Extraction recovery from the PG/VG matrix exceeded 90%, and the limit of quantification was 0.4 μg/mL. Further analytical details are described in El-Hellani et al. (2018) [12].

The total nicotine mass collected from each playback session represented the measured extracted nicotine dose (mg).

Computed dose: mathematical model and computer simulation

Nicotine emitted during each clinical use session was simulated using CALVIN [13]. CALVIN represents the ENDS atomizer as a one-zone lumped system comprising a heating filament, a wick saturated with liquid, and the liquid contained within the wick. Inputs include device geometry, power, liquid composition, and puff velocity versus time. The model solves time-resolved conservation equations for energy and species to describe coupled heat and mass transfer from the heated coil–wick interface during each puff. A boundary-layer formulation is used to compute instantaneous heat and mass transfer, capturing the time evolution of vaporization as the puff proceeds.

The liquid in the heated zone is treated as an ideal multicomponent solution whose composition changes according to the species conservation equations. The vapor pressure of each constituent—propylene glycol (PG), vegetable glycerin (VG), water, and nicotine—is calculated assuming ideal solution behavior. The local boiling point of the mixture is determined at each time step from Raoult’s law based on instantaneous composition. The heated zone is assumed to be replenished with fresh liquid between puffs, which is perfectly mixed with the residual liquid remaining from the prior puff. The maximum liquid temperature is constrained by the computed boiling point of the evolving mixture.

At each time step, the instantaneous nicotine flux (mg/s) emitted from the heated surface is computed. Integration of this flux over the full simulated puff sequence yields the total predicted emitted dose (mg of nicotine per session). Additional details are given in Talih et al. (2017) [13].

For the present analyses, simulations were performed assuming a thermal efficiency of unity for all devices, treating each device as a hypothetical product in which all electrical energy is available to vaporize e-liquid. Device-specific thermal efficiencies can be estimated from a small number of empirical measurements, but were intentionally not applied here to mimic a use case in which the model is used to evaluate hypothetical design or regulatory scenarios without device-specific calibration.

Statistical analysis

Associations between emitted nicotine dose (laboratory-measured and model-simulated) and plasma nicotine boost were assessed by linear regression, overall and stratified by device (Subox Mini-C, JUUL, NJOY Ace). To account for repeated sessions contributed by the same participant, population-averaged linear models were fit using generalized estimating equations (GEE), with participant as the subject variable, an identity link, an independent working correlation structure, and robust standard errors. Regression coefficients are reported as the change in plasma nicotine boost (ng/mL) per 1 mg increase in emitted nicotine dose, with 95% confidence intervals and p-values. R² values are reported from the corresponding ordinary linear regression models as descriptive measures of explained variance. As a sensitivity analysis, models were refitted as random-intercept linear mixed-effects models with participant-level random intercepts. Analyses were performed in SPSS v29 (IBM Corp.).

RESULTS

Plasma nicotine boost was positively associated with both machine-measured and model-predicted nicotine dose. Across all three devices, 183 ad libitum sessions had both dose measures and plasma nicotine boost values available (individual values, Table S1). In this pooled dataset, measured emitted dose explained 41% of the variance in plasma nicotine boost (R²=0.41; p<.0001), and simulated dose explained 33% of the variance (R²=0.33; p<.0001). In GEE models with participants as the clustering unit, both measured and simulated emitted dose were significant predictors of plasma nicotine boost (Table 2). These pooled analyses place observations from heterogeneous devices and study conditions on a common emitted dose scale and test whether emitted nicotine dose has exposure relevance across a broader design and use space (Figure. 3A-B).

Table 2. Association between measured and simulated nicotine dose and plasma nicotine boost, stratified by ENDS device.

Generalized estimating equations with robust standard errors (participant as subject) were used to estimate associations; R² is from the corresponding linear regression. Results are presented separately for each device, and for all devices combined. β coefficients represent the change in plasma nicotine boost (ng/mL) per 1 mg increase in emitted nicotine dose, with 95% confidence intervals shown. All associations were statistically significant (p<.001).

Device	Dose	R2	β (ng/ml/mg)	95% CI for β
Subox Mini (N=90)	Measured	0.44	1.51	[1.04, 1.98]
	Simulated	0.45	0.94	[0.55, 1.33]
JUUL (N=50)	Measured	0.33	1.33	[0.75, 1.91]
	Simulated	0.39	0.41	[0.22, 0.61]
NJOY Ace (N=43)	Measured	0.39	2.36	[1.12, 3.6]
	Simulated	0.35	1.63	[0.8, 2.46]
All devices (N=183)	Measured	0.41	1.57	[1.14, 1.99]
	Simulated	0.33	0.75	[0.54, 0.95]

Figure 3. Plasma nicotine boost (ng/mL) versus laboratory-measured (A) and model-simulated (B) emitted nicotine dose (mg), for all 183 ad libitum sessions.

Each point is one session, colored by device: JUUL (purple circles, n=50), NJOY Ace (blue squares, n=43), and Subox Mini-C (teal triangles, n=90); orange lines show least-squares regression fits. Higher emitted nicotine dose was associated with higher plasma nicotine boost for both the laboratory measurement (R²=0.41) and the physics-based model (R²=0.33; both p<0.0001), and this association held across all three devices.

When analyses were stratified by device, both dose measures remained significant predictors of plasma nicotine boost (Table 2). Within each device the linear model explained a substantial proportion of the variance (R² values in the range of 0.3–0.5), and slopes were positive for all devices but differed in magnitude. The persistence of the dose-boost association within each device/protocol cell indicates that the pooled association was not driven solely by between-study or between-device differences. Accounting for repeated sessions per participant did not change this conclusion: in a random-intercept mixed-effects sensitivity analysis, emitted dose remained a significant predictor of plasma boost for both measured and simulated dose, overall and within each device (Table S2).

Negative plasma boost values occurred in about 7% of sessions, almost entirely in the VCU study, where the protocol involved a 10-puff directed puffing bout 30 minutes prior to the ad libitum bout. Because the pre–ad libitum blood sample was drawn after this directed bout, the baseline already contained residual nicotine; when ad libitum use did not raise plasma nicotine above that residual, the post-minus-pre boost was negative. The FIU sessions, which used a single ad libitum bout after at least 12 hours of abstinence, showed a single negative value among the 93 plasma boost values recorded.

DISCUSSION AND CONCLUSIONS

As ENDS use continues to rise globally, effective product regulation depends on understanding how device design and user behavior determine systemic nicotine exposure and exposure to other toxicants. Such understanding is essential to maximize potential benefits for people who use combustible cigarettes while minimizing risks for nicotine-naïve individuals, across products that differ widely in power output, atomizer design, liquid composition, nicotine concentration, and formulation. In the present work, we focus on nicotine because it is the primary addictive constituent of tobacco products and a key determinant of their abuse liability. Understanding how nicotine emissions translate into systemic nicotine exposure is therefore fundamental to evaluating the impact of product design and informing regulatory standards.

To address this challenge, we previously developed and validated a first-principles mathematical model that accurately predicts ENDS nicotine dose based on device physics and puffing behavior [13]. However, the extent to which measured or simulated dose predicts systemic nicotine exposure has not been empirically tested. In this study, we compared empirically measured and model-predicted nicotine dose with observed plasma nicotine boost from two clinical datasets spanning multiple devices and conditions. Across three ENDS products and 183 ad libitum sessions, measured dose explained 41% of the variance in plasma nicotine boost (R²=0.41), and simulated dose, computed from device characteristics and puff topography without using measured emissions, explained roughly one-third of the variance in nicotine boost (R²=0.33), regardless of the exposure range. This level of prediction is notable given that the analysis ignores inter-individual physiological characteristics such as body mass, blood volume, and metabolic rate, as well as differences in pulmonary behavior, such as inhalation duration, inhalation volume, and breath hold duration, that the puff-topography inputs do not capture. Both the measured and simulated quantities represent nicotine emitted by the device rather than nicotine absorbed systemically; the fraction reaching the bloodstream depends on these physiological and behavioral factors and is reflected in the dose–boost slope.

Analyses stratified by device showed that these findings were not driven by any single product or by the pooled structure alone. For each of the three devices, both measured and simulated dose were significant predictors of nicotine boost, and linear dose–boost models explained a substantial proportion of within-device variability. The absolute slopes and R² values differed across devices, which is unsurprising given differences in hardware, liquid formulations, study protocols, and participant sample. Thus the pooled coefficient should not be interpreted as a universal conversion factor between emitted dose and plasma nicotine boost. Rather, the pooled analysis tests whether measured and simulated emitted dose retain exposure relevance across heterogeneous ENDS products and study conditions. Whether with measured or simulated dose, the consistency of the positive dose-boost association across variations in device, liquid, power, protocol, and participants supports the robustness of the underlying relationship, while the variation in slopes indicates that product- or protocol-specific calibration may be needed for quantitative prediction.

In practical terms, this approach provides a scalable framework for linking device-level engineering parameters and puffing topography to systemic nicotine exposure. Although individual behavior remains variable and may adapt to product constraints, device firmware can limit puff duration[14], and puff count can in principle be constrained over defined intervals. Thus, physics-based dose modeling may help evaluate how product-level design limits could bound delivered nicotine dose, while recognizing that users may circumvent controls by using multiple devices.

Several limitations should be considered. The study examined only three specific devices and a limited set of liquid–power combinations; additional products, formulations, and usage patterns will be needed to fully assess generalizability. The model does not account for individual physiological or inhalation factors such as inhalation depth, lung volume, blood volume, or metabolic rate, all of which contribute to nicotine pharmacokinetics. Incorporating such factors could in principle improve predictions but would require subject-level information that is not available when products are only hypothetical. Because the model uses only device parameters and puff topography, with no subject-specific inputs, it explains a moderate proportion of the variance in plasma nicotine boost; the remainder reflects individual physiological factors it does not attempt to capture. Its predictions are therefore best interpreted at the population level and as estimates with exposure relevance, rather than as individualized predictions of plasma nicotine exposure.

We also did not account for ENDS device-specific energy losses in the main analyses; adjusting for device thermal efficiency can compress between-device differences in dose-boost response and modestly improve pooled fits, but such efficiencies are not generally known for hypothetical products. The same assumption explains why the boost-per-mg slope is larger for measured than for simulated dose: because the model estimates an upper bound on emitted nicotine, a given plasma boost corresponds to a larger simulated dose and therefore a smaller slope.

An additional limitation is that the devices were evaluated in separate clinical studies. Device, study site, and some protocol features were linked by design: the sub-ohm device was tested in one study, whereas the two above-ohm devices were tested in the other. Their independent effects therefore cannot be fully disentagled statistically in this secondary analysis. We addressed this limitation by presenting device/protocol-specific analyses in addition to pooled models.

Finally, session duration differed between the two studies: a 1.5-hour ad libitum bout at VCU versus a median of approximately one hour at FIU. Emitted dose was computed from each session’s measured puff topography over the observed session window, defined as the period from the start of the first puff to the end of the last puff. Plasma nicotine boost was computed from the pre- and post-session blood samples bracketing that same window. Thus, dose and boost were aligned within each session. However, session duration remains a protocol-linked difference and should be considered when interpreting between-device or between-study differences in slope.

Despite these limitations, the results demonstrate that both measured emissions and a physics-based emission model provide useful quantitative links between ENDS engineering parameters and systemic nicotine exposure. The findings support the exposure relevance of emitted nicotine dose across heterogeneous products and study conditions, while indicating that product- or protocol-specific calibration may be needed for quantitative prediction. The demonstrated link between emitted nicotine dose and systemic exposure supports the feasibility of using physics-based models to prospectively evaluate nicotine delivery at the population level, including for hypothetical products or regulatory scenarios where detailed emissions data are not yet available. It also provides a quantitative foundation for device-centric regulatory strategies aimed at controlling delivered nicotine dose.

What is already known on this topic

Nicotine emissions and exposure from ENDS vary widely due to differences in design, operating conditions, and liquid compositions.

What this study adds

Plasma nicotine exposure is associated with emitted nicotine dose estimated by laboratory bench measurements and a physics-based model of coil heating and liquid vaporization.

How this study might affect research, practice or policy

This study supports the technical feasibility of regulating device-level parameters to influence nicotine exposure.

DATA AVAILABILITY STATEMENT

All data are available in the main text or the supplementary materials.