Scholarly article on topic 'An inter-machine comparison of tobacco smoke particle deposition in vitro from six independent smoke exposure systems'

An inter-machine comparison of tobacco smoke particle deposition in vitro from six independent smoke exposure systems Academic research paper on "Medical engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Toxicology in Vitro
OECD Field of science
Keywords
{"Particle deposition" / "Quartz crystal microbalance" / "Tobacco smoke" / Dosimetry / " In vitro exposure system" / Vitrocell}

Abstract of research paper on Medical engineering, author of scientific article — J. Adamson, D. Thorne, G. Errington, W. Fields, X. Li, et al.

Abstract There are several whole smoke exposure systems used to assess the biological and toxicological impact of tobacco smoke in vitro. One such system is the Vitrocell® VC 10 Smoking Robot and exposure module. Using quartz crystal microbalances (QCMs) installed into the module, we were able to assess tobacco smoke particle deposition in real-time. We compared regional deposition across the module positions and doses delivered by six VC 10s in four independent laboratories: two in the UK, one in Germany and one in China. Gauge R&r analysis was applied to the total data package from the six VC 10s. As a percentage of the total, reproducibility (between all six VC 10s) and repeatability (error within an individual VC 10) accounted for 0.3% and 7.4% respectively. Thus Gauge R&r was 7.7%, less than 10% overall and considered statistically fit for purpose. The dose–responses obtained from the six machines across the four different locations demonstrated excellent agreement. There were little to no positional differences across the module at all airflows as determined by ANOVA (except for one machine and at three airflows only). These results support the on-going characterisation of the VC 10 exposure system and suitability for tobacco smoke exposure in vitro.

Academic research paper on topic "An inter-machine comparison of tobacco smoke particle deposition in vitro from six independent smoke exposure systems"

ELSEVIER

Contents lists available at ScienceDirect

Toxicology in Vitro

journal homepage: www.elsevier.com/locate/toxinvit

An inter-machine comparison of tobacco smoke particle deposition in vitro from six independent smoke exposure systems

J. Adamsona'*, D. Thornea, G. Errington a, W. Fields b, X. Lic, R. Payne d, T. Krebs e, A. Dalrymplea K. Fowlerb, D. Dillon a, F. Xiec, C. Meredith a

aBritish American Tobacco, Group R&D, Southampton SO15 8TL, UK bR.J. Reynolds Tobacco Co., P.O. Box 1487, Winston-Salem, NC 27102, USA

cZhengzhou Tobacco Research Institute of China National Tobacco Corporation, No.2 Fengyang Street, High-Tech Zone, Zhengzhou, PR China d Covance Laboratories Ltd., Otley Road, Harrogate HG3 1PY, UK e Vitrocell® Systems GmbH, Fabrik Sonntag 3, 79183 Waldkirch, Germany

CrossMark

ARTICLE INFO

ABSTRACT

Article history: Received 3 April 2014 Accepted 23 June 2014 Available online 2 July 2014

Keywords: Particle deposition Quartz crystal microbalance Tobacco smoke Dosimetry

In vitro exposure system Vitrocell

There are several whole smoke exposure systems used to assess the biological and toxicological impact of tobacco smoke in vitro. One such system is the Vitrocell® VC 10 Smoking Robot and exposure module. Using quartz crystal microbalances (QCMs) installed into the module, we were able to assess tobacco smoke particle deposition in real-time. We compared regional deposition across the module positions and doses delivered by six VC 10s in four independent laboratories: two in the UK, one in Germany and one in China.

Gauge R&r analysis was applied to the total data package from the six VC 10s. As a percentage of the total, reproducibility (between all six VC 10s) and repeatability (error within an individual VC 10) accounted for 0.3% and 7.4% respectively. Thus Gauge R&r was 7.7%, less than 10% overall and considered statistically fit for purpose.

The dose-responses obtained from the six machines across the four different locations demonstrated excellent agreement. There were little to no positional differences across the module at all airflows as determined by ANOVA (except for one machine and at three airflows only). These results support the on-going characterisation of the VC 10 exposure system and suitability for tobacco smoke exposure in vitro.

© 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CCBY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/3XI/).

1. Background

Tobacco smoke generated by machine smoking is commonly used for tobacco product assessment in vitro, for modelling disease processes and for toxicological assessment. The World Health Organisation (WHO) upholds that machine testing of combustible

Abbreviations: ALI, air-liquid interface; ANOVA, analysis of variance; ISO, International Organisation for Standardisation; QC, quality control; QCM, quartz crystal microbalance; R&r, reproducibility (R) and repeatability (r); SD, standard deviation; VC 10, Vitrocell® VC 10 Smoking Robot; WHO, World Health Organisation.

* Corresponding author. Tel.: +44 02380 588220; fax: +44 02380 588856.

E-mail addresses: jason_adamson@bat.com (J. Adamson), david_thorne@bat.com (D. Thorne), graham_errington@bat.com (G. Errington), FIELDSW@RJRT.com (W. Fields), lixiang79ben@hotmail.com (X. Li), Rebecca.payne@covance.com (R Payne), t.krebs@vitrocell.com (T. Krebs), annette_dalrymple@bat.com (A. Dalrymple), FOWLERK2@RJRT.com (K. Fowler), debbie_dillon@bat.com (D. Dillon), xiefuwei@ sina.com (F. Xie), clive_meredith@bat.com (C. Meredith).

tobacco products cannot accurately estimate human exposure, and should not be used to support claims of reduced exposure or risk (World Health Organisation, 2007). However, the WHO does support the use of machine smoke emission data for product hazard assessment, to characterise cigarette emissions for product design and regulatory purposes (World Health Organisation, 2007).

Smoking machines generate, dilute and deliver mainstream tobacco smoke (also known as whole smoke) to an exposure chamber/module containing a biological system, usually supported at the air-liquid interface (ALI). There are many types of smoking machines and exposure chambers available for the testing of whole smoke at the ALI. Some are small bespoke laboratory set-ups whereas others are commercially available systems utilised by the tobacco industry and other well-known inhalation toxicology research groups (Thorne and Adamson, 2013). Whichever system is utilised, there is a clear need to characterise the capabilities, limitations, dilution principles, smoke losses and exact dose delivery

http://dx.doi.org/10.1016/j.tiv.2014.06.012 0887-2333/© 2014 The Authors. Published by Elsevier Ltd.

This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

of these machines, which will lend credence to biological data obtained from smoke exposures. In addition, good understanding and characterisation of the machine, and ultimately full validation, should facilitate the endorsement of the machine for the generation of biological data.

Method validation is the process of demonstrating that an analytical method is suitable for its intended use, and involves conducting a variety of studies to evaluate method performance under defined criteria (Thompson et al., 1999). Method validation studies may involve a single laboratory (intra-laboratory) or multiple laboratories (inter-laboratory). Organisations such as the Association of Analytical Communities (AOAC) and US Environmental Protection Agency (EPA) provide methods that are validated through inter-laboratory studies, and parameters which may be assessed in method validation include precision defined by reproducibility (R) and repeatability (r) and bias (Ellison et al., 2009). Thus, inter-laboratory studies/cross-machine comparisons enable confidence to be gained in a machine or laboratory set-up and can facilitate the standardisation of experimental testing protocols. As such, data generated from method validation studies or standardised protocols could provide information for future regulation or testing standards.

Currently, there are no defined regulatory protocols for tobacco whole smoke exposure systems, but product testing protocols for assays such as Ames bacterial mutagenicity and Neutral Red Uptake (NRU) cytotoxicity are being developed to support in vitro toxicity testing and disease model development. One such smoking machine/exposure system used for the testing of tobacco whole smoke is the Vitrocell® VC 10 Smoking Robot and mammalian exposure module (Vitrocell® 6 CF Stainless) both of which have been previously described (Adamson et al., 2013; Klein et al., 2013; Nara et al., 2013; Okuwa et al., 2010). Additionally, quartz crystal microbalance (QCM) technology can be employed to accurately assess deposited particle mass within the exposure module. QCMs enable a greater understanding of particle dose as mass per surface area, rather than simply a diluting airflow and sampling vacuum flow rate applied to the exposure module (Adamson et al., 2013; Paur et al., 2011; Bakand and Hayes, 2010; Lenz et al., 2009). To assess deposition, QCMs are installed in the exposure module in place of the biological system, giving real-time, in situ gravimetric data on particle deposition, in the nanogram range (Fig. 1).

In this study, six Vitrocell® VC 10 Smoking Robots were assessed for their ability to generate a consistent smoke dose, using QCMs to quantify deposited particle matter within the exposure module (Fig. 1d). The QCMs took readings from each position in the module, the first position being proximal to smoke entering the dilution bar, the last position being distal to smoke entry (Fig. 1b). It is important to consider this arrangement, as in some instances the linear direction of smoke entry may have an effect on regional/positional deposition differences across the QCMs left to right (Deschl et al., 2011). The aim of this study was to enable the comparison of multiple VC 10s, in four independent laboratories/geographical locations, tested with an identical diluting airflow range of 0.25, 0.5, 1.0, 2.0 and 4.0 L/min. These airflows were selected for testing based on a previously published study (Adamson et al., 2013) which represented reliably detectable levels for the QCMs during a 24 min exposure. To preserve anonymity of machine, lab, group and operator, all were coded (Table 1.) as is common practice for comparison studies. The six VC 10s were located in four laboratories: two laboratories were in the United Kingdom, one was in China and one was in Germany. Regional deposition patterns across the exposure module were assessed independently at each airflow for each machine. R&r analysis was estimated for all six instruments which were collectively termed the 'measurement system'. R&r analysis determines the precision of a measurement system and is often employed to compare multiple systems in different locations or with different operators; more specifically it calculates the degree to which repeated measurements taken under the same (unchanged) experimental conditions show the same result (Measurement Systems Analysis reference manual, 2002). Reproducibility (R) is the closeness of agreement between measurements or observations conducted on replicate specimens (machines) in different locations by different people; it assesses the ability of the experiment or measurement to be reproduced independently. Repeatability (r) looks at test-retest variability; it assesses the variation in measurements made within the same system by the same operator (Kaur et al., 2010). Thus, data were compared within each machine and across all six machines. Additionally, two important variables were assessed by comparing data from VC 10s which had a significant change. The first was laboratory geography/environment, where data were acquired on the same VC 10 (serial VC 10/300412) in two different laboratories after it was moved from one to another. The second

Fig. 1. A schematic cross-section of the Vitrocell® exposure system set-up (not to scale). (a) VC 10 Smoking Robot including the single piston/syringe and delivery tubing to the dilution bar; (b) smoke entry (dark arrow) to a single dilution bar where diluting air is added (white arrows). Multiple parallel bars make the dilution system; (c) each dilution bar has one smoke jet (ci) which is always 2.0 mm 0 (in this study), and 2 identical air jets above and below the dilution bar (cii) which are either both 1.0 mm 0 or both 0.8 mm 0, depending on dilution airflow; and (d) mammalian 6/4 CF Stainless exposure module with QCMs installed into each of the four separate wells. Image adapted from Adamson et al. (2013).

Table 1

The VC 10s used in this study, serial number identifier, laboratory location and their operators.

VC 10 name Serial number Laboratory Operator(s)

A1 VC 10/141209 A a

B1 VC 10/090610 B b, c, d

B2 VC 10/221211 B b, c, d, e

B3 VC 10/210311 B c, e

C1 VC 10/300412 C a

D1 VC 10/200410 D a, f

was service status of the machine, where data were acquired on the same VC 10 (serial VC 10/141209) in the same location before and after an annual service (which was 6 month overdue), and the results compared.

2. Materials and methods

2.1. Machine smoking

Six independent but comparable VC 10 Smoking Robots (Vitro-cell® Systems, Waldkirch, Germany) (Fig. 1) were tested. The six machines were located across four laboratories in Southampton, UK; Harrogate, UK; Zhengzhou, China; and Waldkirch, Germany. For anonymity, each machine was given a coded name based on its location (Table 1). Apart from machines B3 and D1, all VC 10s had a 4-port dilution bar with an exposure module (model 6/4 CF Stainless) (Vitrocell® Systems) containing 4 QCMs docked within it to detect deposited particle mass, as shown in Fig. 1. The dilution bars of machines B3 and D1 were the shorter variant with three instead of four smoke ports. Therefore, the module used with VC 10 B3 and D1 was the 6/3 CF Stainless and contained 3 QCMs for measurements taken in positions 1-3. Both the 3 and 4-port dilution bars and their respective modules are commercially available. Consequently, equipment and experimental variables were comparable across the four laboratories/six machines, including smoke transit tubing lengths to the dilution bar (34-45 cm) and module trumpet heights (of approximately 2 mm from the QCM surface), as per manufacturer's specifications. In this study, laboratories A, B and D had controlled environmental conditions where temperature and humidity were set and maintained at 20 ± 5 °C and 55 ± 15% relative humidity. The VC 10 in laboratory C was placed inside an open cabinet, sat within a test room with uncontrolled environmental conditions. Data generation for an individual machine took between two and six (nonconsecutive) days; all six machines were assessed and data collected between July 2012 and July 2013. The main reason that multiple machines were not tested in parallel was feasibility; there were limited QCM devices to make parallel geographic investigations.

The generation, dilution and delivery of whole smoke to biological systems by the VC 10 has been previously described (Adamson et al., 2013). Cigarette smoke can be delivered by three operational modes (single, serial-asynchronous and serial-synchronous) with the VC 10 as detailed by the manufacturer (Vitrocell® Systems, Waldkirch, Germany). For the purposes of this study, cigarette smoke was generated via the rotary carousel, where one cigarette is loaded and smoked at a time (VC 10 single mode). A single syringe located under the rotary carousel puffs on the cigarette to the ISO 3308:2000 smoking regime: 35 ml puff over 2 s, once a minute. Although not representative of human smoking behaviour, the ISO smoking regime was selected as an industry standard, allowing cross-comparisons of different machines with the same regime, cigarette and dilution range. Three reference cigarettes (3R4F, 9.4 mg tar) obtained from the University of Kentucky, USA, were smoked to a defined puff number of 8 puffs per cigarette, resulting

in 24 min duration per run. In all cases, the same dilution bar and module containing the QCMs were used for all airflows tested on an individual VC 10. In the dilution bar of all machines and at all airflows the 2.0 mm 0 smoke jet was installed (Fig. 1ci). For machines A1, C1 and D1, two 1.0 mm 0 diluting air jets were installed for all airflows tested (Fig. 1cii). For the B machines the two diluting air jets were both 0.8 mm 0 for 0.25 and 0.5 L/min airflows only (the other airflows p1.0 L/min had the 1.0 mm 0 diluting air jets), as per the supplier's original recommendations. However, full comparison studies of the jet combinations in laboratories A and D showed there were no statistically significant differences in particle dose determined by QCMs when using the 2.0 mm 0 smoke jet and 1.0 mm 0 airflow jets combination for airflows less than 1 L/min (p = 0.961 as determined by one-way ANOVA (at 0.5 L/min in laboratory D, n = 5/QCM position/jet combination)). Thus for experimental simplicity, the same smoke and air jet combination (2.0 mm 0 and 1.0 mm 0 respectively) were maintained for all airflows for VC 10 A1, C1 and D1. Five airflows (0.25, 0.5, 1.0, 2.0 and 4.0 L/min) were tested and smoke dilutions were obtained by adjusting the airflow delivered to the dilution system. All other programmed smoking settings for the smoke run were the same across the six machines and are described in Table 2.

2.2. Dosimetry measurements

For real-time assessment of deposited particle dose in machines A1, B1, B2 and C1, four identical quartz crystal microbalance (QCM) units were installed into each separate well of the 6/4 CF Stainless exposure module (Fig. 1d) (Adamson et al., 2013). As previously described, machines B3 and D1 were equipped with the 3-port dilution bar and 6/3 CF Stainless exposure module, therefore three QCMs were installed to assess smoke particle deposition. QCMs were not reordered between runs but always kept in the same positions for the entirety of each study. In previous investigations, we have discovered that randomising QCMs in the module does not affect dose in that specific position. Additionally, measurement stability and repeatability is greatly improved when the QCM housing units are left and not constantly removed and replaced (although the crystal surfaces are still cleaned in situ, between each run). Individual seals between lid and base for each QCM position were not leak tested prior to each exposure. Smoke leakage could be a possible cause for deposition variability; however, it would be unlikely that large smoke leakage was occurring as such an event would have been detected (visual and odor) by the opera-tor(s). Exposure module seal integrity and leak testing is conducted as part of the service agreement, but VC 10 users should also be

Table 2

Smoking settings used for the VC 10 on all machines. For machines A1 and C1, diluting airflows were set and maintained during the duration of the experiment by mass flow controllers (Analyt-MTC GmbH, Mülheim, Germany); for machines B1-B3 and D1, diluting airflows were set manually with valves integral to the Vitrocell® dilution system and checked using a mass flow meter prior to each run.

Machine parameters User input

Inhale Curve Number3 1

Piston size (ml) 100

Puff duration time (s) 2

Puff duration hold time (s) 0

Puff duration exhaust time (s) 8

Puff volume (ml) 35

Puff frequency (s) 60

Number of clear cycles after run 0

Max. number of inhalation cycles per cigarette 8

Number of cigarettes per run 3

a Inhale Curve Number 1 represents a bell-shaped puffing profile.

vigilant for such occurrences. Once docked to the Vitrocell® dilution bar (Fig. 1b) smoke was sampled from the stream of diluted smoke at a flow rate of 5 ml/min/well negative pressure through the module, as previously described (Adamson et al., 2013).

QCM units and software were supplied by Vitrocell® Systems. The QCMs read at a resolution of 10 ng/cm2/s; mass readings were taken every 2 s during the smoke run and reported as mass per unit area in real-time (Mulhopt et al., 2009). In terms of their limit of quantification, QCMs have been shown to be able to detect particle mass in the nanogram range delivered from low tar delivery products (1 mg) at high dilutions (Adamson et al., 2012), and in the case of the VC 10 with 3R4F reference cigarettes, at very high diluting airflow rates up to 12 L/min (Thorne et al., 2013). In this study during data collection, QCMs were stabilised prior to exposure (zero point stability) and were allowed to plateau post-smoke particle deposition, until the mass had stopped increasing. This took an additional 60-120 s post-exposure and represented the maximal particle deposition recorded for the set airflow.

2.3. Graphics and statistics

For all six machines, experiments were repeated 5 times per airflow. Figs. 2-6 were produced using Minitab® version 16.1.0. The means of deposited mass ± standard deviation (Table 3) were calculated from the raw data in Microsoft Excel® 2010. Differences in regional deposition across/between module positions (Figs. 3 and 6, and Table 4.) were determined by one-way analysis of variance (ANOVA) with Tukey method using Minitab®. Statistical tests for R&r (Table 5) were conducted using JMP® Pro version 10 and Minitab®. Evaluating the Measurement Process (EMP) Gauge R&r was conducted in JMP® Pro version 10 as described by Wheeler (2006).

3. Results

The data generated from each VC 10 were analysed within itself (intra-machine) to assess regional deposition at the airflows tested, and analyses were made between VC 10s (inter-machine) at the same airflows and collectively in the diluting range 0.25-4.0 L/ min (inter-laboratory). Fig. 2 and Table 3. illustrate the deposited particle dose-response relationship for each machine at airflows 0.25-4.0 L/min, demonstrating good parity between the six VC 10s. In addition, Fig. 3 details the pattern of particle deposition across the module positions based on a data set of n = 5/QCM position/airflow/machine. From the data presented in Fig. 3, it is clear to see varied deposition patterns across QCM positions at the lower airflows 0.25 and 0.50 L/min compared to the higher airflows 1.0, 2.0, 4.0 L/min. Furthermore, despite the machines being the same

—•— Al

-o- ei

J>\ - 83

-Q-- CI

1 —0- D1

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Airflow (lymin)

Fig. 2. QCM particle deposition: an interaction plot of the data means for the six VC 10s within the airflow range tested.

model there are some differences in the varied distribution between VC 10s at the same airflows. For example, machine C1 shows a distinct distribution of deposition at the highest concentration of 0.25 L/min, where deposition ascends between positions 1-3 and then drops at position 4. For other machines, there is a total ascending gradient from first to last QCM position, such as machine B2 at 0.25 L/min or C1 at 0.50 L/min. In other cases, there are 'zigzagged' distributions where deposition ascends and descends between QCM positions, such as machine B1 at 0.25 L/min.

Particle deposition distribution across the modules for five of the six machines (all but VC 10 C1) had good agreement at all airflows, as there were no significant differences observed across the QCM positions as determined by one-way ANOVA (Table 4). There were significant differences across QCM positions for three of the five airflows on VC 10 C1.

In order to visualise the overall pattern of distribution across the QCM positions from each VC 10, multi-vari charts were produced to display the mean particle distribution at all airflows tested per machine (Fig. 4). For the complete airflow range tested (Fig. 4a), machine C1 demonstrated a very distinct pattern of particle deposition across the positions left to right, where mass ascended significantly from positions 1-3 and dropped at position 4. As indicated by the diamonds (the mean of the four QCM positions), machine C1 delivered the highest dose of the six machines. The B machines also showed slight variation in their patterns of distribution. However, the overall means, as indicated by the diamonds, were more consistent between the three B machines and machine D1. Machine A1 had the lowest mean delivery of all. The mean pattern of particle deposition when 0.25 L/min is excluded (as recommended by the supplier due to inherent limitations of low flow turbulence and inefficient mixing in the short transit length available) resulting in a range of 0.5-4.0 L/min was also assessed (Fig. 4b). The variation between QCM positions improved for every VC 10 and a more linear pattern was observed across all positions. The data means across the machines also showed more alignment in this range, as indicated by the proximity of the diamonds (Fig. 4b).

Evaluating the Measurement Process (EMP) Gauge R&R (Wheeler, 2006) was applied to the total data package, which is the data from all QCM positions at all airflows for all machines. As a percentage of the total, reproducibility (error between all six VC 10s) and repeatability (error within an individual VC 10) accounted for 0.3% and 7.4% respectively, the inter-machine variability (reproducibility) being especially low. Thus, total R&r was 7.7%; less than 10% overall and considered statistically fit for purpose (Barrentine, 1991). As expected, by the nature of the dilution process giving the different smoke doses, the main product variation in this evaluation was from airflow at 90.4%. The interaction variation of VC 10 and airflow was also low at 1.9%. Table 5 shows the R&r values for the individual airflows tested and the mean mass per airflow for all six machines. Values for reproducibility and repeatability decrease with a reduction in mean mass/increase in diluting airflow. Only reproducibility (R) at 0.25 L/min was greater than 10%. The R value for 0.25 L/min (20.30%) indicates poor inter-laboratory precision at this airflow, which confirms the supplier's recommendation not to operate at airflows less than 0.5 L/min (or at least to use with caution). However, within any individual VC 10, intra-machine variability (r) at 0.25 L/min was 7.71 which would be considered perfectly acceptable.

Main effects plots were used to assess variables associated with the inter-machine study: the individual machines (VC 10), laboratory and QCM position. Fig. 5a and b shows the data means for machine and lab respectively. Machine C1 in lab C delivered higher particle doses overall. Fig. 5c shows the data means for QCM position for all machines at all airflows and would suggest that within the dilution bar of any VC 10 there is a tendency for the mass to

Fig. 3. A multi-vari chart for deposited particulate mass, showing deposition gradients across the module for all airflows tested. The connected black dots represent QCM positions, left to right. Data sets were n = 5/QCM position/airflow/machine. Asterisks indicate statistically significant differences in deposition across/between module positions as determined by one-way ANOVA. Diamonds detail the mean deposited particle mass detected at each airflow. Circles indicate the total mean deposited mass for an individual VC 10 at all airflows tested, and are connected by a dashed line to show the relationship between machines.

ascend slightly from position 1 (proximal to smoke entry) to position 3-4 (distal to smoke entry).

In addition to the variables associated with the inter-machine study (Fig. 5), two other conditional and important VC 10 variables were assessed. The variables were laboratory environment and service status (maintenance) of the machine. Fig. 6a shows how moving a VC 10 from one lab to another affected smoke delivery (deposition). As previously shown, VC 10 C1 demonstrated the most varied patterns of deposition across the module, especially at airflows 0.25, 0.5 and 2.0 L/min where there were statistically significant differences (Fig. 3). Consequently, when machine C1 was relocated to laboratory B and the data generation were replicated, the results were compared to the range from machines B1-B3 (shown in the light grey boxplots) (Fig. 6a). Not only did the data obtained with machine C1 in Lab B align with the other machines in the same location, but also there were no significant

differences between module positions observed - demonstrating a significant improvement. This could be due to laboratory conditions (where laboratory B had stable environmental conditions whereas laboratory C did not), or it may also have been due to a machine service prior to installation in laboratory B. A similar result was demonstrated on machine A1 when data were captured pre- and post-service (Fig. 6b). In this case, other than service status and date of the experiment, the location and set-up were the same. Pre-service, machine A1 gave extremely varied results (Adamson et al., 2013), confirmed by significant differences between QCM positions at all airflows tested (Fig. 6b). However, in the days directly after a 6 month overdue service (in which machine A1 was dismantled, cleaned and reassembled) the results obtained were fully aligned with the VC 10 data from the other laboratories, and deposition positional effects had improved significantly. Both of these circumstances (environment and machine

Fig. 5. A main effects plot for deposited mass. The charts show the data means for (a) the six different VC 10 machines; (b) the four different laboratories where the testing was conducted, and; (c) QCM positions (1-4, or 1-3 for B3 and D1). Mean reference line for each chart is 4.67 ig/cm2.

Fig. 6. Variables affecting smoke particle deposition. Multi-vari charts showing (a) the effect of geographical location/laboratory environment on the same VC 10, and (b) smoke particle deposition before and after VC 10 service. (a) VC 10 C1 was moved from laboratory C to laboratory B; the grey boxplots show the means of the deposition from machines B1-B3 putting VC 10 C1 data (black dots) into context of the B lab machines. (b) Data from machine A1 is shown pre- (June 2012) (Adamson et al., 2013) and postservice (July 2013). Data sets were n = 5/QCM position/airflow/machine. Asterisks indicate statistically significant differences in deposition across/between module positions as determined by one-way ANOVA.

service) have shown to have a significant effect on the performance of the VC 10, and will be discussed later.

4. Discussion

There are a variety of smoking machines used for the assessment of tobacco smoke in vitro. The system described in this study

is the commercially available Vitrocell® VC 10 Smoking Robot which dilutes smoke with continuous airflow at adjusted rates to obtain different smoke concentrations. Before the advent of particle dosimetry tools such as the QCM, there was no quick or simple way of measuring dose/exposure, and especially not with the ability to observe it in real-time and with the sensitivity the QCM delivers. Furthermore, biological response (from any in vitro sys-

Table 3

Mean deposited particulate mass values obtained from the QCM (± standard deviation) for the six machines at all airflows tested. Data sets were n = 5/QCM position/airflow/ machine; all machines except B3 and D1 had 4 QCM positions for each airflow tested (n = 20) whereas VC 10 B3 and D1 had 3 QCM positions for each airflow tested (n = 15).

Airflow, L/min Mean deposited mass, ig/cm2 ± SD

0.25 0.50 1.00 2.00 4.00

VC10A1 10.15 ±1.72 5.22 ±0.50 3.23 ±0.37 1.33 ±0.22 0.41 ± 0.06

VC10B1 10.16 ±1.02 6.24 ±0.46 3.88 ±0.21 1.79 ±0.12 0.70 ±0.12

VC10B2 12.58 ±2.68 6.24 ± 0.62 3.35 ±0.34 1.24 ±0.20 0.46 ±0.13

VC10B3 11.93 ± 3.42 6.02 ±0.30 3.60 ±0.27 1.82 ±0.12 0.66 ± 0.07

VC10C1 14.08 ±4.55 7.11 ±1.32 3.52±0.55 1.36±0.24 0.40±0.08

VC10D1 11.64 ±1.15 5.73 ±0.54 3.16 ±0.18 1.63 ±0.11 0.58 ±0.10

Table 4

Level of statistically significant difference across positions (1-4, or 1-3 for B3 and D1) for all machines at all airflows tested (n = 5/QCM position/airflow/machine) as determined by one-way ANOVA.

Airflow, L/min p-Value

0.25 0.50 1.00 2.00 4.00

VC 10 A1 0.353 0.611 0.865 0.994 0.418

VC 10 B1 0.507 0.284 0.431 0.802 0.979

VC 10 B2 0.330 0.418 0.646 0.967 0.916

VC 10 B3 0.829 0.335 0.178 0.360 0.868

VC 10 C1 0.012* 0.038* 0.244 0.013* 0.189

VC 10 D1 0.345 0.184 0.535 0.364 0.820

* Statistically significant difference between positions, p = 60.05.

Table 5

Mean mass and reproducibility and repeatability (R&r) results for the measurement

system (all six VC 10s).

Airflow, L/min Mean mass, lg/cm2 Reproducibility, R Repeatability, r

0.25 11.75 20.30 7.71

0.50 6.11 8.05 2.03

1.00 3.46 3.24 0.99

2.00 1.51 3.06 0.51

4.00 0.53 1.61 0.27

tem) could have been presented as a function of the machine's dilution and test article delivery mechanism. QCMs when used concurrently with a biological assay enable biological response to be presented as deposited particle mass per surface area. As demonstrated herein, data from QCMs or other dosimetry tools will facilitate cross-comparisons between different systems (where the systems use the same/similar dosimetry tools).

In this study, six VC 10 Smoking Robots in four different locations were assessed using QCMs with the aim of conducting a method validation study involving multiple laboratories. Deposited particle mass was monitored in real-time and the final deposited mass recorded, regional deposition was assessed and dose comparisons were made within and between machines. Five airflows were selected for testing based on a previously published study (Adamson et al., 2013) which represented reliably detectable levels for the QCMs during a 24 min exposure. For all airflows a sample rate of 5 ml/min/well negative pressure was applied to the module. The selected airflows represent low to mid-range (0.25-4.0 L/min) airflows of which the VC 10 is capable of operating; however, this range does not delimit the working range of the VC 10 for biological exposure or the ability of the QCMs to detect mass. Indeed, the VC 10 and QCM technology is used at airflows up to 12 L/min with the Neutral Red Uptake assay (NRU) and Ames assay (Thorne et al., 2013), and any positional differences observed in QCM deposition across the module have not been reflected in the biological response. Fig. 4b showed the improvement in the consistency of dose delivery in the module when the 0.25 L/min data were removed from the rest of the airflows tested; this is

because 0.5 L/min is stated as the lower limit of recommended working range, whereas 0.25 L/min is out of the (supplier) recommended working range due to low mixing flow/low turbulence. Thus the variances in dose delivery from flows less than 0.5 L/ min can be explained by less turbulence and insufficient transit length/time for the whole smoke aerosol to achieve homogeneity prior to exposure. The Gauge R&r results from this study confirmed this with an R value of 20.30% for the 0.25 L/min airflow, indicating that inter-laboratory precision at 0.25 L/min is poor (amongst these six VC 10s).

All deposition measurements within a module were made from four QCM readings apart from machines B3 and D1 which had three QCM readings. The variables which changed were operator, the date when the data were collected and the dilution bar air jet diameter (for airflows 60.5 L/min). There was agreement in dose response relationship across the six VC 10s (Fig. 2); however, the distribution across the module positions (left-right) was irregular across certain VC 10 systems and airflows (Figs. 3 and 4). Notably, machine C1 was slightly separate from the other five machines when analysed statistically. However, when evaluating these results, other factors that could influence the variability must be considered. For example, the deposition data range/distribution may shift or change depending on how often the machine is used, its service status and most importantly its cleanliness (Fig. 6b). This has been demonstrated in this study where the data generation for machines commencing directly after a full machine service resulted in highly repeatable results per airflow. The service of the VC 10 provided by Vitrocell® (either annual or biannual, dependent on usage and user preference) consists of stripping down the machine to its component parts, a thorough cleaning of everything, and replacement of all dispensable parts such as labyrinths, O-rings, seals, filters and tubing. The service also involves an extensive checking (and re-calibration if out of specification) of parameters such as puff volume and duration, carousel turning time and speed, cigarette loading, lighting and removal.

With any laboratory equipment there is a requirement over its period of use to ensure that the apparatus/machinery is still functioning to manufacturers' specification, that it performs correctly and that results are consistent over time, especially if its use results in the machine requiring cleaning/maintenance. Of the VC 10s assessed in this study, many users have noted/observed slight changes in machine performance and delivery over time, namely with smoke dose, mixing and deposition. Control charts for the data obtained on machine A1 have also confirmed change over time. Fig. 6b clearly shows a marked improvement in smoke particle deposition post-service, not only with the range of doses per airflow aligning with the other VC 10s, but also an improvement in the pattern of deposition across the module positions where no positional differences were observed post-service. It should also be noted that the annual service was overdue at 18 months as opposed to 12 months at that time (and which is now scheduled every 6 months). In support of this, the data from machine D1 was also acquired directly after service and the standard deviation

for this machine at all airflows was amongst the smallest of the six VC 10s (Table 3). Accordingly, the other machines in the study were all within their annual service period, although the data were not generated directly after their service. Thus, we have observed in two separate instances how data acquisition in close proximity to machine service has delivered more robust results. This observation will clearly be of benefit to Vitrocell® users, who should be aware of the impact the frequency of use has on machine performance. The types of product which are being smoked (especially high tar delivery products where deposits can cover all parts of the robot which come into contact with smoke) and the cleaning frequency of the VC 10 should also be monitored by Vitrocell® users. Collectively, this information may inform and guide future users and not delimit them as they consider experimental design.

Another technical variable is the diameters of the different jets on the dilution bar (Fig. 1c) and the combinations in which they should be used. They are recommended by the supplier based on airflow range: the 2.0 mm 0 smoke jet with 1.0 mm 0 air jets are recommended for airflows equal to or greater than 1 L/min; the 2.0 mm 0 smoke jet with 0.8 mm 0 air jets are recommended only for airflows less than 1 L/min. In this study, the B machines were fitted with air jets according to the supplier's original recommendations. In the dilution bar of machines A1, C1 and D1, the two 1.0 mm 0 diluting air jets were used for all airflows tested. This was based on previously observed QCM data which suggested jet diameter had made no difference to particle dose. Thus, we conducted a comparison study of the jet combinations (in two machines only, one of which is reported here) which showed there was no difference in deposition if the 2.0 mm 0 smoke jet and 1.0 mm 0 airflow jets were installed for airflows less than 0.5 L/ min (p = 0.961) which is the supplier recommended lowest airflow. At 0.25 L/min, there was a statistically significant difference in deposition when comparing the two previously described jet combinations (p = 0.001, as determined by one-way ANOVA (at 0.25 L/ min in laboratory D, n = 5/QCM position/jet combination) however the data sets did overlap). When the recommended jet combination was used, the mean values were lower and so too were SDs when compared to the other data. However, this low airflow should still be used with caution. Thus for experimental simplicity, the same smoke and air jet combination (2.0 mm 0 and 1.0 mm 0 respectively) were maintained for all airflows for machines A1, C1 and D1. We have noted that there was a difference in the jets used between the B machines and the other VC 10s, but ultimately all machines performed similarly (with the exception of three airflows on machine C1) and collectively, R&r for the test system of all six VC 10s, gave acceptable values. The effect of jet combination on dose may require further investigation, but so far the results we have seen (at least in these six machines and from this data) suggest jet diameter does not play a significant role in particle deposition in the supplier's recommended range (i.e. when 0.25 L/min is excluded).

Our study assessed inter-machine variability across six VC 10s and inter-laboratory variability across four independent labs. Gauge R&r assessment on the whole data package demonstrated good results, indicative that the systems collectively were fit for purpose: reproducibility (between all six VC 10s) and repeatability (error within an individual VC 10) accounted for 0.3% and 7.4%, respectively, where total R&r was 7.7%. The largest variation in this study came from the machine in laboratory C (Figs. 3, 5b and 6a, and Table 4.); thus there may be co-effects of different environmental conditions, the control and maintenance of such conditions, operator, differing cigarette shipments or cigarette conditioning that impacted the variability of the results. In this study, laboratories A, B and D had better control over environmental conditions where temperature and humidity were set and maintained at 20 ± 5 °C and 55 ± 15% relative humidity. The VC 10 in laboratory

C was placed inside an open cabinet, sat within a test room with uncontrolled environmental conditions; additionally the cigarettes were not conditioned for C1 data. Therefore it is likely that external and laboratory environment contributed to the difference in results observed from machine C1 when it was moved from an uncontrolled environment (laboratory C) to a controlled one (laboratory B) (Fig. 6a). On the other hand, during data collection in laboratory D the humidity reached up to 72% (note that ISO requirements specify humidity of 60 ± 5%) due to the ventilation system connecting directly with the external environment and yet this did not appear to affect the data, which were aligned with the other machines and with small SDs (Table 3). This further highlights the requirement for users to characterise and understand each individual VC 10 within its own environment, and to be aware of important issues such as usage and machine cleanliness. It would not be prudent to advise a service requirement for an individual machine based on this data alone; clearly this would be linked to each VC 10's specific usage. At the very least an annual service is required, while more frequent/daily use may require service at shorter intervals (e.g. every 6 months).

5. Conclusions

The results presented in this study demonstrate that the Vitro-cell® system is fit for its intended purpose, when particle deposition is assessed collectively using QCMs. This study has highlighted some slight differences between VC 10 machines in terms of dilution and particle deposition gradients, which serves as a foundation for understanding the working limits of the system. At present it is unclear why some VC 10s observe a concentration gradient and some do not, but what is clear is that this gradient is more pronounced at higher smoke concentrations/lower diluting airflows, especially 0.25 and 0.5 L/min. This observation may suggest that there are technical challenges with turbulent mixing or with the aerosol dynamics in the dilution system at the lower airflow rates. This does not mean that these airflows should be avoided; instead it could suggest a recommendation that biological data at these airflows should be carefully considered in cases where the lower airflow doses are having an effect on the biological response.

QCMs are proving to be effective tools in the determination of tobacco smoke particle dose in vitro. However, QCMs are only one dosimetry tool and others have been investigated. These include, but are not limited to: deposition quantification using chemical spectrofluorescence (Adamson et al., 2012); particle number, size and surface area determination using differential mobility spectrometry (Adamson et al., 2011); CO analysis (Thorne et al., 2013); and vapour phase assessment of compounds using a headspace stir bar-sorptive extraction technique (Kaur et al., 2011). The investigation of these and additional tools to qualify and quantify smoke dose remains an active area of research. QCMs determine deposition in real-time and with precision (Paur et al., 2011; Lenz et al., 2009), enabling the user to see puff-by-puff and cigarette-by-cigarette profiles (Adamson et al., 2012). Additionally, QCMs are being used concurrently with biological exposures and are proving beneficial as a quality control tool (Thorne et al., 2013). The QCM technology can be operated single-handedly and requires little or no analytical resource when used for simple quantification alone. Furthermore, there is scope to use microbalance technology for other engineered liquid droplet aerosols, aiding inhalation toxicologists in other fields. For in vitro applications, the microbalance technology can effectively discriminate between high (9.4 mg) and low (1 mg) tar delivery products, showing aligned results with traditional chemical methods of particle quantification (Adamson et al., 2012). The tool has also

enabled users of the VC 10 in a previous study to investigate the effect of varying sample flow rate applied to the module on particle deposition, where this information was previously unknown/ assumed (Adamson et al., 2013).

In terms of the Vitrocell® VC 10 Smoking Robot itself, this study has clearly demonstrated the importance of variables which may affect dose, and these should be assessed and monitored by users of the VC 10. Primarily, our results suggest the environmental conditions of the laboratory and machine service status may play an important role in acquiring robust and repeatable QCM data over time. Dilution bar jet diameter (smoke and air jets) appears to have no effect on QCM determined particle deposition (based on this data) but may need further investigation by individual users.

In summary, all of these additional factors aside, the dose-responses obtained from all machines across the four different locations demonstrated excellent agreement with Guage R&r for the entire test system of 7.7%, being considered statistically fit for purpose.

Authors' contributions

Jason Adamson, David Thorne and Annette Dalrymple conceived and designed the study. Jason Adamson performed the experimental work in GR&D Southampton and at Vitrocell® Systems Germany, conducted the data analysis from all six machines, and drafted the manuscript. Xiang Li and Jason Adamson performed the experimental work at ZTRI China. Rebecca Payne performed and managed the experimental work at Covance. Graham Errington conducted the statistical analysis for all machines. David Thorne, Fuwei Xie, Tobias Krebs and Wanda Fields reviewed the collected data and data analysis for quality control. Annette Dalrymple, Kathy Fowler, Fuwei Xie, Tobias Krebs, Wanda Fields, Debbie Dillon and Clive Meredith provided scientific support and reviewed the manuscript. All authors read and approved the final manuscript.

Conflict of Interest

The authors declare that there are no conflicts of interest. Jason Adamson, David Thorne, Graham Errington, Annette Dalrymple, Debbie Dillon and Clive Meredith are employees of British American Tobacco (BAT). The experimental work conducted at BAT was funded by BAT. Wanda Fields and Kathy Fowler are employees of R.J. Reynolds Tobacco (RJRT). Xiang Li and Fuwei Xie are employees of Zhengzhou Tobacco Research Institute (ZTRI). The experimental work conducted at ZTRI was funded by ZTRI. Rebecca Payne is an employee of Covance Laboratories Ltd. UK. All of the experimental work conducted at Covance was funded by BAT or RJRT. Tobias Krebs is Managing Partner of Vitrocell® Systems GmbH. All of the experimental work conducted in Germany was funded by Vitrocell®.

Transparency Document

The Transparency document associated with this article can be found in the online version.

Acknowledgements

The authors would like to thank Betsy Bombick at RJRT for her scientific and technical review of the paper; Marco Hebestreit at

Vitrocell® Systems for his invaluable technical support with the VC 10 and QCMs; and Adam Seymour, Laura Jeffrey, Sally Forrest, Lara Walker, Joanne Kilford and Victoria Hargreaves at Covance Laboratories Ltd., UK. for supporting the experimental work conducted there.

References

Adamson, J., Azzopardi, D., Errington, G., Dickens, C., McAughey, J., Gaga, M.D., 2011. Assessment of an in vitro whole cigarette smoke exposure system: the Borgwaldt RM20S 8-syringe smoking machine. Chem. Cent. J. 5, 50-61.

Adamson, J., Hughes, S., Azzopardi, D., McAughey, J., Gaca, M.D., 2012. Real-time assessment of cigarette smoke particle deposition in vitro. Chem. Cent. J. 6, 98.

Adamson, J., Thorne, D., Dalrymple, A., Dillon, D., Meredith, C., 2013. Assessment of cigarette smoke particle deposition within the Vitrocell® exposure module using quartz crystal microbalances. Chem. Cent. J. 7, 50.

Bakand, S., Hayes, A., 2010. Troubleshooting methods for toxicity testing of airborne chemicals in vitro. J. Pharmacol. Toxicol. Methods 61 (2), 76-85.

Barrentine, L.B., 1991. Concepts for R&R Studies. ASQ Quality Press, Milwaukee, WI.

Deschl, U., Vogel, J., Aufderheide, M., 2011. Development of an in vitro exposure model for investigating the biological effects of therapeutic aerosols on human cells from the respiratory tract. Exp. Toxicol. Pathol. 63, 593-598.

Ellison, S.L.R., Barwick, V.J., Duguid Farrant, T.J., 2009. Practical Statistics for the Analytical Scientist: A Bench Guide, second ed. RSC Publishing, ISBN: 978-184755-955-5.

Kaur, N., Lacasse, M., Roy, J.P., Cabral, J.L., Adamson, J.A., Errington, G., Waldron, K.C., Gaca, M.D., Morin, A., 2010. Evaluation of precision and accuracy of the Borgwaldt RM20S® smoking machine designed for in vitro exposure. Inhal. Toxicol. 22 (14), 1174-1183.

Kaur, N., Cabral, J.L., Morin, A., Waldron, K., 2011. Headspace stir bar sorptive extraction-gas chromatography/mass spectrometry characterization of the diluted vapor phase of cigarette smoke delivered to an in vitro cell exposure chamber. J. Chromatogr. A 1218, 324-333.

Klein, S.G., Serchi, T., Hoffmann, L., Blömeke, B., Gutleb, A.C., 2013. An improved 3D tetraculture system mimicking the cellular organisation at the alveolar barrier to study the potential toxic effects of particles on the lung. Particle Fibre Toxicol. 10, 31.

Lenz, G.A., Karg, E., Lentner, B., Dittrich, V., Brandenberger, C., Rothen-Rutishauser, B., Schulz, H., Ferron, G.A., Schmid, O., 2009. A dose-controlled system for airliquid interface cell exposure and application to zinc oxide nanoparticles. Particle Fibre Toxicol. 6, 32.

Measurement Systems Analysis reference manual, 2002. Automotive Industry Action Group (AIAG). Daimler Chrysler Corporation, Ford Motor Company, General Motors Corporation: Supplier Quality Requirements Taskforce, Michigan, United States.

Mülhopt, S., Diabate, S., Krebs, T., Weiss, C., Paur, H.R., 2009. Lung toxicity determination by in vitro exposure at the air liquid interface with an integrated online dose measurement. J. Phys.: Conf. Ser. 170, 012008.

Nara, H., Fukano, Y., Nishino, T., Aufderheide, M., 2013. Detection of the cytotoxicity of water-insoluble fraction of cigarette smoke by direct exposure to cultured cells at an air-liquid interface. Exp. Toxic. Pathol. 65 (5), 683-688.

Okuwa, K., Tanaka, M., Fukano, Y., Nara, H., Nishijima, Y., Nishino, T., 2010. In vitro micronucleus assay for cigarette smoke using a whole smoke exposure system: a comparison of smoking regimens. Exp. Toxicol. Pathol. 62, 433-440.

Paur, H.-R., Cassee, F.R., Teeguarden, J., Fissan, H., Diabate, S., Aufderheide, M., Kreyling, W.G., Hänninen, O., Kasper, G., Riediker, M., Rothen-Rutishauser, B., Schmid, O., 2011. In-vitro cell exposure studies for the assessment of nanoparticle toxicity in the lung-A dialog between aerosol science and biology. J. Aerosol Sci. 42, 668-692.

Thompson, M., Ellison, S.L.R., Fajgelj, A., Willetts, P., Wood, R., 1999. Harmonised guidelines for the use of recovery information in analytical measurement (technical report). Pure Appl. Chem. 71, 337-348.

Thorne, D., Adamson, J., 2013. A review of in vitro cigarette smoke exposure systems. Exp. Toxicol. Pathol. 65,1183-1193.

Thorne, D., Kilford, J., Payne, R., Adamson, J., Scott, K., Dalrymple, A., Meredith, C., Dillon, D., 2013. Characterisation of a Vitrocell® VC 10 in vitro smoke exposure system using dose tools and biological analysis. Chem. Cent. J. 7,146.

Wheeler, D.J., 2006. EMP III. Evaluating the Measurement Process & Using Imperfect Data. SPC Press Statistical Process Controls, Inc., Knoxville, Tennesse, ISBN 9780-945320-67-8.

World Health Organisation, 2007. The Scientific Basis of Tobacco Product Regulation: Report of a WHO Study Group. WHO Technical Report Series 945. ISBN 978 92 4 120945 8.