Technical Report Puffing Topography Inter-lab Study · 2017. 2. 21. · 3 ISO 5725-1:1994 Accuracy...

45
Smoking Behaviour Sub-Group Technical Report Puffing Topography Inter-lab Study January 2017 Study Co-ordinators: Krishna Prasad and Sandy Slayford British American Tobacco, United Kingdom Statistical Support: Thomas Verron and Rémi Julien SEITA Imperial Tobacco, France

Transcript of Technical Report Puffing Topography Inter-lab Study · 2017. 2. 21. · 3 ISO 5725-1:1994 Accuracy...

  • Smoking Behaviour Sub-Group

    Technical Report

    Puffing Topography Inter-lab Study

    January 2017

    Study Co-ordinators:

    Krishna Prasad and Sandy Slayford

    British American Tobacco, United Kingdom

    Statistical Support:

    Thomas Verron and Rémi Julien

    SEITA – Imperial Tobacco, France

  • Table of Contents

    1. INTRODUCTION ............................................................................................................... 3

    2. EXPERIMENTAL DESIGN ............................................................................................... 3

    2.1 Part 1: Intra and inter-lab-device variability ....................................................................... 4

    2.2 Part 2: Time effect ................................................................................................................ 5

    3. STATISTICAL EVALUATION ......................................................................................... 6

    3.1. Uncertainty evaluation ......................................................................................................... 6

    3.2. Accuracy profile ................................................................................................................... 8

    3.3. Limitations ............................................................................................................................ 9

    4. RESULTS .......................................................................................................................... 10

    4.1. Outlier results ...................................................................................................................... 10

    4.1.1. Puff duration ................................................................................................. 11

    4.1.2. Puff volume .................................................................................................. 13

    4.2. Intra and inter-lab-device variability estimations ............................................................. 15

    4.2.1 Puff duration ................................................................................................. 15

    4.2.2. Puff volume .................................................................................................. 16

    4.3. Intra and inter-lab-device variability modeling ................................................................ 17

    4.3.1. Puff duration ................................................................................................. 17

    4.3.2. Puff volume .................................................................................................. 18

    4.4. Desirability indices ............................................................................................................. 19

    4.4.1. Puff duration ................................................................................................. 20

    4.4.2. Puff volume .................................................................................................. 21

    5. COMMENTS AND CONCLUSIONS .............................................................................. 22

    5.1. General comments .............................................................................................................. 22

    5.2. Puff duration ....................................................................................................................... 22

    5.3. Puff volume......................................................................................................................... 22

    APPENDIX A – List of participating laboratories and device manufacturers............................ 23

    APPENDIX B – Part 2 – Intra and inter-lab-device variability estimations ............................... 24

    APPENDIX C – Data representation .......................................................................................... 25

    APPENDIX D – Accuracy profile results ................................................................................... 36

    APPENDIX E – Experimental protocol ...................................................................................... 38

    APPENDIX F – Departures from experimental protocol ........................................................... 44

    APPENDIX G – Definitions ....................................................................................................... 45

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 3/45

    1. INTRODUCTION

    In 2012, the CORESTA Sub-Group, Smoking Behaviour (SG-TSB), agreed to conduct a

    study to determine if various human puffing topography devices deliver comparable results

    for puff volume and puff duration. In order to assess this, the uncertainty of both puffing

    topography devices and laboratories has been evaluated based on the experimental design

    described below.

    2. EXPERIMENTAL DESIGN

    Four laboratories took part in this study and four puffing topography devices (see Figure 1)

    were evaluated by four different SG-TSB member organizations. Appendix A lists the

    laboratories and device manufacturers involved in the study. Note that the listings of

    laboratories and device manufacturers are in alphabetical order, while the results are presented

    by code (1-4 for laboratories; A-D for devices). Labs and devices were both randomized for

    data presentation, so nothing can (or should) be inferred about the identity of any lab or

    device based on the order of presentation in this report.

    CME–P4

    CReSSmicro

    SA7

    SPA-M

    Figure 1: Presentation of the four evaluated devices

    The study was designed to evaluate the intra-lab-device variability (r*) and the inter-lab-

    device variability (R*) of the puffing topography devices (Part 1) and to evaluate the time

    effect (Part 2).

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 4/45

    2.1 Part 1: Intra and inter-lab-device variability

    In this part, three different units of each type of device were tested. The same device set was

    circulated to each of the four participating laboratories over the course of two years. Each

    participating laboratory was asked to test the set of 12 puffing topography devices (four

    brands and three units of each) on two samples of king size cigarettes: Camel Blue (CB) and

    West White (WW). These corresponded to two different open pressure drops: 45 mmWG and

    130.5 mmWG, respectively. Two different cigarette modes were considered as well: lit and

    unlit. Whichever mode, a total of 10 puffs per sample was taken at each smoking regime (see

    Table 1).

    The data structure with the key figures is illustrated in Figure 2. A total of 9600 lines of data

    was generated for each of the three parameters (puff volume, puff duration and flow rate).

    Figure 2: Data structure of Part 1

    Table 1 gives the details of the five smoking regimes considered in Part1. All puff profiles

    were square with inter-puff intervals of 30 s.

    Table 1: Set of five smoking regimes

    Smoking Regime Puff Volume (ml) Puff Duration (s) Average Flow (ml/s)

    1 35 3.5 10

    2 80 4.0 20

    3 90 3.0 30

    4 80 2.0 40

    5 50 1.0 50

    Puffing Topography Study

    Lab 1

    A

    B

    C

    D

    Rep. 1

    Rep. 2

    ….

    Rep. 10

    Terms

    Lab 4 Lab 2 Participating laboratories (4)

    Evaluated devices (4)

    Puffing regimes (5)

    Replicates (10)

    Individual results (9600)

    Lab 3

    Evaluated units (3)

    Tested configurations (4) CB Lit CB Unlit WW Lit WW Unlit Unlit

    #1 #2 #3

    SR1 SR2 SR3 SR4 SR5

    Equipment

    Protocol

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 5/45

    The objective was to determine the maximum difference expected between given puff

    volumes or durations measured: 1) using a given device in a given lab and 2) using different

    devices in different labs.

    2.2 Part 2: Time effect

    In order to evaluate the time effect, each participating laboratory was asked to test only one

    unit per device, with only one puffing regime (3) on five separate days (see Figure 3). A total

    of 3200 lines of data was generated for each of the three parameters (puff volume, puff

    duration and flow rate).

    Figure 3: Data structure Part 2

    Puffing Topography Study

    Lab 1

    A

    B

    C

    D

    SR3

    Rep. 1

    Rep. 2

    ….

    Rep. 10

    Terms

    Lab 4 Lab 2 Participating laboratories (4)

    Evaluated devices (4)

    Puffing regime (1)

    Replicates (10)

    Individual results (3200)

    Lab 3

    Evaluated unit (1)

    Tested configurations (4)

    #1

    Testing days (5) D1 D2 D3 D4 D5

    CB Lit CB Unlit WW Lit WW Unlit Unlit

    Equipment

    Protocol

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 6/45

    3. STATISTICAL EVALUATION

    3.1. Uncertainty evaluation

    The statistical evaluation of data for this collaborative study followed the methods provided

    by ISO 5725-21. For outlier testing, the Grubbs and Cochran methods were used.

    ISO 5725-2 Tests Consistency

    Cochran’s test Within-laboratory variability: Suitable for detecting whether or not the highest value in a set of laboratory standard deviations is an outlier (test statistic greater than 1% critical value) or a straggler (test statistic greater than 5% and less than 1% critical value).

    Grubbs’ test Between-laboratory variability: Suitable for detecting whether or not the highest (or lowest) laboratory’s average is an outlier (test statistic greater than 1% critical value) or a straggler (test statistic greater than 5% and less than 1% critical value).

    Both statistical tests allowed the identification of stragglers and outliers, however only the

    outlier values were discarded in the statistical evaluation.

    The protocol “Harmonized Statistical Procedure” defined by IUPAC (International Union of

    Pure and Applied Chemistry) was applied. This involved sequential applications of the

    Cochran and Grubbs tests until no further outliers were detected or until the data dropped

    amounted to 22.2% (for this study = 2/9) of the total data collected (see Figure 4).

    The Cochran outlier test was applied first, and if an outlying laboratory was identified, then a

    single-value Grubbs test was performed on the individual values of this outlying laboratory

    (Individual Grubbs). If no individual value was identified as an outlier, the outlying

    laboratory is removed. If an individual value was identified as an outlier, only that value was

    removed. This was followed by the application of single-value Grubbs test and outlying

    laboratories were removed. If no laboratory was identified as an outlier, the pair-value test

    was applied (two values at the same end). Any laboratory(ies) flagged by these tests were

    removed until the data dropped amounted to 22.2% (for this study = 2/9) of the total data

    collected.

    1 “ISO 5725-2:1994: Accuracy (trueness and precision) of measurement methods and results – Part 2: Basic

    method for the determination of repeatability and reproducibility of a standard measurement method.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 7/45

    Figure 4: Process of outlier detection

    At the end of the loop of outlier detection, if any laboratory(ies) had been removed, then the

    process was stopped. On the remaining data set, a nested ANOVA was performed in order to

    evaluate the intra-lab-device variability (r*) and the inter-lab-device variability (R*).

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 8/45

    3.2. Accuracy profile

    In addition for Part 1, the accuracy profile approach was used to test the ‘fitness for purpose’

    of the devices when used by one laboratory. The accuracy profile is a graphical decision-

    making tool (see Figure 5).

    A Tolerance Interval is computed considering the puff measurements (10 replicates) and

    intermediate precision (three units) and Acceptability Interval (λ) defined by the

    collaborative study are determined for each device type. These intervals give information

    about the expected proportion of future measurements that will lie at a certain level of risk (1

    – β) in the specified range and whether or not the device results are acceptable.

    Figure 5: Accuracy profile as a graphical decision-making tool

    The ‘fitness-for-purpose’ of the device is represented by the Desirability Index (ID), which is

    computed as the cube root of the product of the three indices: the Range Index (IR), the

    Precision Index (IP) and the Trueness Index (IT). All indices vary from 0 to 1. For further

    details on computation of the Desirability Index (ID), refer to E. Rozet et al2 .

    2 E. Rozet, V. Wascotte, N. Lecouturier, V. Préat, W. Dewé, B. Boulanger, P. Hubert, Improvement of

    the decision efficiency of the accuracy profile by means of a desirability function for analytical

    methods validation: Application to a diacetyl-monoxime colorimetric assay used for the determination

    of urea in transdermal iontophoretic extracts, Analytica Chimica Acta, Volume 591, Issue 2, 22 May

    2007, Pages 239-247.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 9/45

    Finally, the Desirability Index gives the ‘fitness for purpose’ of the device within the limits of

    inter-lab-device variability (estimated in the collaborative study) when it is used within a

    given laboratory. Figure 6 shows some examples of the range of information provided by the

    accuracy profiles at different levels of the Desirability Index.

    Figure 6: Examples of desirability index

    3.3. Limitations

    Despite having a well-balanced design, there are some limitations to this collaborative study.

    First of all, the small number of participating laboratories (only four) cannot be representative

    of inter-laboratories variability, and can impact the interpretation of the statistical tests of

    outlier detection. Although the four participating laboratories were experienced in the use of

    puffing topography devices, it would have been better to have at least eight laboratories as

    recommended by ISO 5725-13 and ASTM E2480

    4.

    Secondly, the small number of participating laboratories makes evaluation per device less

    robust. Consequently, we had to combine the inter-laboratories variability with the inter-

    3 ISO 5725-1:1994 Accuracy (trueness and precision) of measurement methods and results -- Part 1: General

    principles and definitions.

    4 ASTM E2480 - 12 Standard Practice for Conducting an Interlaboratory Study to Determine the Precision of a

    Test Method with Multi-Valued Measurands.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 10/45

    devices variability, to calculate global lab-device performance evaluation. With such

    evaluation, the inter-laboratories variability and inter-devices variability are confounded. For

    this reason, we decided to designate the intra-lab-device variability and the inter-lab-device

    variability as r* and R*, respectively, in order to avoid any confusion with the usual notations

    r and R for repeatability and reproducibility respectively.

    Thirdly, although global indices allow a comparison of devices, it is important to note that a

    bad index is not necessarily the consequence of a bad device; it can also reflect a bad use of

    the device, or a cigarette sampling issue.

    4. RESULTS

    This section relates only to the results obtained within Part 1 of the study. Appendix B

    provides the intra-lab-device variability and the inter-lab-device variability estimations for

    Part 2 for information purposes only. Indeed, the evaluations conducted within this study lead

    to the conclusion that the estimations of the intra-lab-device variability and the inter-lab-

    device variability are similar between both parts. This means that the major part of the study

    variability has been captured in Part 1.

    4.1. Outlier results

    In order to accurately estimate the intra-lab-device and the inter-lab-device variabilities, the

    outlier detection procedure, as described previously, was applied to 160 datasets

    corresponding to five smoking regimes, two lighting modes and 16 combinations of

    laboratories-devices (four laboratories and four devices). Each dataset is a set of 60 individual

    results (three units, two samples and 10 puffs).

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 11/45

    4.1.1. Puff duration

    Table 2 provides the summary of outliers and stragglers detected throughout the study for

    each combination for puff duration. Tables 3 and 4 then link this information to the related

    smoking regime and lighting mode.

    Table 2: Summary of outliers and stragglers for puff duration

    Single test Double test Single test Double test

    Dev.A_Lab.1 10 600 2 (20%)

    Dev.A_Lab.2 10 600 6 (60%)

    Dev.A_Lab.3 10 600 5 (50%)

    Dev.A_Lab.4 10 600 6 (60%)

    Dev.B_Lab.1 10 600 4 (40%)

    Dev.B_Lab.2 10 600

    Dev.B_Lab.3 10 600

    Dev.B_Lab.4 10 600 1 (10%)

    Dev.C_Lab.1 10 600 1 (10%)

    Dev.C_Lab.2 10 600 2 (20%)

    Dev.C_Lab.3 10 600

    Dev.C_Lab.4 10 600

    Dev.D_Lab.1 10 600 2 (20%)

    Dev.D_Lab.2 10 600 4 (40%)

    Dev.D_Lab.3 10 600

    Dev.D_Lab.4 10 600 5 (50%)

    Total 160 9600 1 (0.6%) 37 (23%)

    Combination

    Labs-Devices

    Number of

    datasets

    analysed

    STRAGGLER DATASET OUTLIER DATASET

    GRUBBSCOCHRAN

    GRUBBSCOCHRAN

    Total number

    of datapoints

    analysed

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 12/45

    Expressing the number of outliers from Table 2 as a percentage of the total number of 160

    datasets, almost 1 out 4 (23%) datasets was detected as an outlier, as related to the within-

    device laboratory variability (Cochran’s test). Details on how outliers are distributed across

    the different smoking regimes and lighting modes are given in the two following tables.

    Table 3: Distribution of stragglers for puff duration

    Table 4: Distribution of outliers for puff duration

    C: Cochran’s test, G: Grubbs’ simple test (_h:high and _l:low),

    G2: Grubbs’ double test (_h:high and _l:low), Gi: Grubbs’ test on individual values.

    In some cases, where the “Gi” label occurs in the table, it means that a few of the individual

    results were removed because they were identified as outliers. As the rest of the dataset was

    still considered in the estimation of the intra-lab-device variability and the inter-lab-device

    variability, the “Gi” status is not mentioned in the summary table (Table 2).

    LIT_SR1 LIT_SR2 LIT_SR3 LIT_SR4 LIT_SR5 UNLIT_SR1 UNLIT_SR2 UNLIT_SR3 UNLIT_SR4 UNLIT_SR5

    Dev.A_Lab.1

    Dev.A_Lab.2

    Dev.A_Lab.3

    Dev.A_Lab.4

    Dev.B_Lab.1

    Dev.B_Lab.2

    Dev.B_Lab.3

    Dev.B_Lab.4 C

    Dev.C_Lab.1

    Dev.C_Lab.2

    Dev.C_Lab.3

    Dev.C_Lab.4

    Dev.D_Lab.1

    Dev.D_Lab.2

    Dev.D_Lab.3

    Dev.D_Lab.4

    Combination

    Labs-Devices

    PUFF DURATION - STRAGGLER

    LIT_SR1 LIT_SR2 LIT_SR3 LIT_SR4 LIT_SR5 UNLIT_SR1 UNLIT_SR2 UNLIT_SR3 UNLIT_SR4 UNLIT_SR5

    Dev.A_Lab.1 Gi Gi Gi C Gi C Gi Gi

    Dev.A_Lab.2 C Gi Gi Gi C C C Gi C C

    Dev.A_Lab.3 C C C Gi C C

    Dev.A_Lab.4 C C Gi C C Gi C C

    Dev.B_Lab.1 C C C C

    Dev.B_Lab.2

    Dev.B_Lab.3

    Dev.B_Lab.4 Gi Gi Gi Gi Gi

    Dev.C_Lab.1 C

    Dev.C_Lab.2 C C

    Dev.C_Lab.3

    Dev.C_Lab.4

    Dev.D_Lab.1 C C

    Dev.D_Lab.2 C C Gi C C

    Dev.D_Lab.3 Gi

    Dev.D_Lab.4 C C C C C

    Combination

    Labs-Devices

    PUFF DURATION - OUTLIER

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 13/45

    4.1.2. Puff volume

    Table 5 provides the summary of outliers and stragglers detected throughout the study for

    each combination for the puff volume. Tables 6 and 7 then link this information to the related

    smoking regime and lighting mode.

    Table 5: Summary of outliers and stragglers for puff volume

    Single test Double test Single test Double test

    Dev.A_Lab.1 10 600

    Dev.A_Lab.2 10 600

    Dev.A_Lab.3 10 600

    Dev.A_Lab.4 10 600 2 (20%)

    Dev.B_Lab.1 10 600 10 (100%)

    Dev.B_Lab.2 10 600 1 (10%) 3 (30%) 5 (50%)

    Dev.B_Lab.3 10 600 5 (50%)

    Dev.B_Lab.4 10 600 1 (10%) 3 (30%) 6 (60%)

    Dev.C_Lab.1 10 600 2 (20%)

    Dev.C_Lab.2 10 600

    Dev.C_Lab.3 10 600

    Dev.C_Lab.4 10 600

    Dev.D_Lab.1 10 600 7 (70%)

    Dev.D_Lab.2 10 600

    Dev.D_Lab.3 10 600 5 (50%)

    Dev.D_Lab.4 10 600 1 (10%)

    Total 160 9600 41 (26%)

    Combination

    Labs-Devices

    Number of

    datasets

    analysed

    STRAGGLER DATASET OUTLIER DATASET

    GRUBBSCOCHRAN

    GRUBBSCOCHRAN

    Number of

    datapoints

    analysed

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 14/45

    Expressing the number of outliers from Table 5 as a percentage of the total number of 160

    datasets, more than 1 out 4 (26%) datasets was detected as an outlier related to the within-

    device laboratory variability (Cochran’s test). Details on how outliers are distributed across

    the different smoking regimes and lighting modes are given in the two following tables.

    Table 6: Distribution of stragglers for puff volume

    Table 7: Distribution of outliers for puff volume

    C: Cochran’s test, G: Grubbs’ simple test (_h:high and _l:low),

    G2: Grubbs’ double test (_h:high and _l:low), Gi: Grubbs’ test on individual values.

    In some cases, where the “Gi” label occurs in the table, it means that a few of the individual

    results were removed because they were identified as outliers. As the rest of the dataset was

    still considered in the estimation of the intra-lab-device variability and the inter-lab-device

    variability, the “Gi” status is not mentioned in the summary table (Table 2).

    LIT_SR1 LIT_SR2 LIT_SR3 LIT_SR4 LIT_SR5 UNLIT_SR1 UNLIT_SR2 UNLIT_SR3 UNLIT_SR4 UNLIT_SR5

    Dev.A_Lab.1

    Dev.A_Lab.2

    Dev.A_Lab.3

    Dev.A_Lab.4 G2_h G2_h

    Dev.B_Lab.1

    Dev.B_Lab.2 G2_h G2_h G_h, G2_h

    Dev.B_Lab.3

    Dev.B_Lab.4 G_h, G2_h G2_h G2_h

    Dev.C_Lab.1

    Dev.C_Lab.2

    Dev.C_Lab.3

    Dev.C_Lab.4

    Dev.D_Lab.1

    Dev.D_Lab.2

    Dev.D_Lab.3

    Dev.D_Lab.4

    Combination

    Labs-Devices

    PUFF VOLUME - STRAGGLER

    LIT_SR1 LIT_SR2 LIT_SR3 LIT_SR4 LIT_SR5 UNLIT_SR1 UNLIT_SR2 UNLIT_SR3 UNLIT_SR4 UNLIT_SR5

    Dev.A_Lab.1

    Dev.A_Lab.2

    Dev.A_Lab.3

    Dev.A_Lab.4

    Dev.B_Lab.1 C C C C C C C C C C

    Dev.B_Lab.2 C C C C C

    Dev.B_Lab.3 C C C C C

    Dev.B_Lab.4 C C C C C C

    Dev.C_Lab.1 C C

    Dev.C_Lab.2

    Dev.C_Lab.3

    Dev.C_Lab.4

    Dev.D_Lab.1 C C C C C C C

    Dev.D_Lab.2

    Dev.D_Lab.3 C C C C C

    Dev.D_Lab.4 C

    Combination

    Labs-Devices

    PUFF VOLUME - OUTLIER

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 15/45

    4.2. Intra and inter-lab-device variability estimations

    The intra-lab-device variability limits (r*) and the inter-lab-device variability limits (R*) are

    calculated for each smoking regime, for each mode (lit and unlit) and for each parameter (puff

    duration and puff volume).

    4.2.1 Puff duration

    Estimations for puff duration are summarized in Table 8 by keeping the outlier dataset

    (combination laboratory-device) and in Table 9 by removing the outlier dataset. The

    corresponding graphs named “individual values” are shown in Appendix C of this report.

    Table 8: r* and R* estimations for puff duration with outlier dataset

    Table 9: r* and R* estimations for puff duration without outlier dataset

    Both tables demonstrate the benefit of the outlier detection, since the range of CVr* changes

    from [17.2% ; 41.8%] to [3.9% ; 12.3%] and the range of CVR* changes from [23.9% ;

    57.4%] to [13.6% ; 32.3%].

    Considering the range of the inter-lab-device variability limits computed after outlier removal,

    the Acceptability Interval () was set at a conservative 30% to build the accuracy profile later.

    Mode Regime Number

    combination

    labs-devices

    Mean r* R* CVr* CVR*

    LIT SR1 16 3.19 0.549 0.760 17.2% 23.9%

    LIT SR2 16 3.76 1.361 1.531 36.2% 40.7%

    LIT SR3 16 2.93 0.521 0.691 17.8% 23.6%

    LIT SR4 16 2.09 0.688 0.855 33.0% 41.0%

    LIT SR5 16 1.07 0.318 0.499 29.6% 46.5%

    UNLIT SR1 16 3.18 0.695 0.869 21.9% 27.3%

    UNLIT SR2 16 3.74 1.035 1.171 27.7% 31.3%

    UNLIT SR3 16 2.92 0.905 1.035 31.0% 35.5%

    UNLIT SR4 16 2.04 0.436 0.601 21.3% 29.5%

    UNLIT SR5 16 1.08 0.453 0.621 41.8% 57.4%

    Puff duration [s]

    Mode Regime Number

    combination

    labs-devices

    Mean r* R* CVr* CVR*

    LIT SR1 12 3.15 0.126 0.553 4.0% 17.6%

    LIT SR2 12 3.74 0.298 0.565 8.0% 15.1%

    LIT SR3 12 2.90 0.342 0.529 11.8% 18.2%

    LIT SR4 15 2.04 0.245 0.420 12.0% 20.6%

    LIT SR5 12 1.01 0.124 0.267 12.3% 26.4%

    UNLIT SR1 12 3.17 0.123 0.599 3.9% 18.9%

    UNLIT SR2 12 3.72 0.179 0.507 4.8% 13.6%

    UNLIT SR3 12 2.91 0.215 0.500 7.4% 17.2%

    UNLIT SR4 12 2.01 0.124 0.363 6.2% 18.1%

    UNLIT SR5 12 1.03 0.108 0.332 10.5% 32.3%

    Puff duration [s]

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 16/45

    4.2.2. Puff volume

    Estimations for puff volume are summarized in Table 10 by keeping the outlier dataset

    (combination laboratory-device) and in Table 11 by removing the outlier dataset. The

    corresponding graphs named “individual values” are shown in Appendix C of this report.

    Table 10: r*and R* estimations for puff volume with outlier dataset

    Table 11: r* and R* estimations for puff volume without outlier dataset

    Both tables demonstrate the benefit of the outlier detection, since the range of CVr* changes

    from [11.0% ; 23.9%] to [4.3% ; 12.7%] and the range of CVR* changes from [21.0% ;

    32.8%] to [9.7% ; 18.2%].

    Considering the range of the inter-lab-device variability limits computed after outlier removal,

    the Acceptability Interval () was set at 20% to build the accuracy profile later.

    Mode Regime Number

    combination

    labs-devices

    Mean r* R* CVr* CVR*

    LIT SR1 16 36.9 4.80 10.23 13.0% 27.7%

    LIT SR2 16 82.9 18.61 23.38 22.5% 28.2%

    LIT SR3 16 93.9 14.07 19.74 15.0% 21.0%

    LIT SR4 16 82.2 14.22 18.20 17.3% 22.1%

    LIT SR5 16 51.5 11.63 16.86 22.6% 32.8%

    UNLIT SR1 16 36.3 4.02 9.05 11.0% 24.9%

    UNLIT SR2 16 80.7 15.78 20.74 19.6% 25.7%

    UNLIT SR3 16 89.7 18.92 23.53 21.1% 26.2%

    UNLIT SR4 16 79.9 19.06 24.08 23.9% 30.1%

    UNLIT SR5 16 50.9 9.45 14.25 18.6% 28.0%

    Puff volume [ml]

    Mode Regime Number

    combination

    labs-devices

    Mean r* R* CVr* CVR*

    LIT SR1 12 35.3 2.01 4.56 5.7% 12.9%

    LIT SR2 12 82.3 9.34 15.00 11.4% 18.2%

    LIT SR3 12 91.6 11.63 14.53 12.7% 15.9%

    LIT SR4 12 82.7 8.35 11.89 10.1% 14.4%

    LIT SR5 12 51.6 2.68 7.41 5.2% 14.4%

    UNLIT SR1 12 34.9 2.09 4.26 6.0% 12.2%

    UNLIT SR2 12 81.6 3.63 11.25 4.4% 13.8%

    UNLIT SR3 12 91.1 5.62 12.60 6.2% 13.8%

    UNLIT SR4 12 80.9 4.10 9.36 5.1% 11.6%

    UNLIT SR5 12 52.1 2.24 5.04 4.3% 9.7%

    Puff volume [ml]

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 17/45

    4.3. Intra and inter-lab-device variability modeling

    We have investigated whether there is a relationship between the average for results and the

    precision (r* and R*).

    4.3.1. Puff duration

    Figure 7 shows the power relationships established between the average puff duration and the

    associated standard deviations of precision (Sr* and SR*).

    Figure 7: Power relationships for puff duration and the associated Sr* and SR*

    Sr* and SR* increase more or less linearly when puff duration increases. However, the

    relationships between Sr* and puff duration are not good enough (R² lower than 0.5).

    Improvements of model quality are shown in Figure 8, where non-linear relationships are

    established between Sr* or SR* and the average flow.

    Puff duration average (s)

    Sr

    or

    SR

    (s)

    1.0 1.5 2.0 2.5 3.0 3.5 4.0

    0.00

    0.05

    0.10

    0.15

    0.20

    0.25 Sr LIT 0.0483x0.497 R² 0.294 Sr UNLIT 0.0369x0.3843 R² 0.465

    SR LIT 0.0951x0.6071

    R² 0.986 SR UNLIT 0.1109x0.4218

    R² 0.766

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 18/45

    Figure 8: Non-linear relationships for average flow and Sr* and SR* of puff duration

    Considering results shown in Figure 8, the intra and inter-lab-device variabilities are non-

    linearly dependent on average flow rate.

    4.3.2. Puff volume

    Figure 9 shows the power relationships established between the average puff volume and the

    associated standard deviations of precision (Sr* and SR*).

    Figure 9: Power relationships for puff volume and the associated Sr* and SR*

    Flow rate average (ml/s)

    Sr

    or

    SR

    (s)

    10 20 30 40 50 60

    0.00

    0.05

    0.10

    0.15

    0.20

    0.25 Sr LIT -0.0538 + 0.011x -2e-04x2 R² 0.964 Sr UNLIT 0.0112 + 0.0039x -1e-04x2 R² 0.708

    SR LIT 0.1663 + 0.0038x -1e-04x2

    R² 0.998 SR UNLIT 0.2372 -0.0023x -2e-06x2

    R² 0.932

    3.5s 4s 3s 2s 1s

    Puff volume average (ml)

    Sr

    or

    SR

    (m

    l)

    30 40 50 60 70 80 90 100

    0

    1

    2

    3

    4

    5

    6 Sr LIT 6e-04x1.9184

    R² 0.953 Sr UNLIT 0.0223x0.9509

    R² 0.844

    SR LIT 0.0199x1.2356

    R² 0.971 SR UNLIT 0.0214x1.1666

    R² 0.928

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 19/45

    Sr* and SR* increase more or less linearly when puff volume increases. Variability obtained

    on LIT cigarettes is higher than that obtained on UNLIT cigarettes. Improvements of model

    quality are shown in Figure 10, where non-linear relationships are established between Sr*

    and SR* and the average flow.

    Figure 10: Non-linear relationships for average flow and Sr* and SR* of puff volume

    Considering results shown in the Figure 10, the intra and inter-lab-device variabilities are

    non-linearly dependent on average flow rate.

    4.4. Desirability indices

    As described previously, the Desirability Index gives the device ‘fitness for purpose’ within

    the limits of inter-lab-device variability when it is used within a given laboratory.

    Consequently, for each of the two assessed parameters, 64 accuracy profiles were computed

    for the 64 tested configurations (four devices * four laboratories * two modes * two samples).

    Figures 11 and 12 summarize the computed Desirability Indices, distributed on a scale from

    0% to 100%. Indices are categorized into three classes: 1) “Unsatisfactory” for indices below

    60%, 2) “Questionable” for indices between 60% and 80%, and 3) “Fit for purpose” for

    indices greater than 80%.

    Additionally, the right margin provides the device code information (A-B-C-D) as well as the

    laboratory code information (1-2-3-4). The left margin provides the number of defective

    profiles.

    Appendix D provides the complete Table of Indices.

    Flow rate average (ml/s)

    Sr

    or

    SR

    (m

    l)

    10 20 30 40 50 60

    0

    1

    2

    3

    4

    5

    6 Sr LIT -3.7528 + 0.4915x -0.0079x2

    R² 0.985 Sr UNLIT -0.7841 + 0.1608x -0.0026x2

    R² 0.881

    SR LIT -2.777 + 0.5101x -0.008x2

    R² 0.902 SR UNLIT -2.2137 + 0.4241x -0.0069x2

    R² 0.964

    35ml 80ml 90ml 80ml 50ml

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 20/45

    4.4.1. Puff duration

    Considering the range of inter-lab-device variability limits computed after outlier removal, the

    Acceptability Interval () was set at 30% for puff duration. Desirability Indices are

    summarized in Figure 11.

    Figure 11: Desirability indices for puff duration

    Device A:

    Desirability Indices are spread between the “Questionable” and “Fit for purpose” areas.

    Indeed, indices vary from 66% to 89% and two datasets were defective (from laboratories 1

    and 4).

    Device B:

    Desirability Indices are systematically spread in the “Fit for purpose” area. Indeed, indices

    vary from 86% to 97% and four datasets were defective (from laboratory 1).

    Device C:

    Desirability Indices are spread between the “Questionable” and “Fit for purpose” areas.

    Indeed, indices vary from 69% to 95% and one dataset was defective (from laboratory 1).

    Device D:

    Desirability Indices are systematically spread in the “Fit for purpose” area. Indeed, indices

    vary from 85% to 98%.

    Unsatisfactory Questionable Fit for purpose

    A

    B

    C

    D

    1

    1

    1

    1

    2

    2

    2

    2

    3

    3

    3

    3

    4

    4

    4

    4

    Nb=7 Nb=0 Nb=0 Nb=10 Nb=47

    # D

    efe

    ctive

    pro

    file

    #1

    #1

    #4

    #1

    0% 20% 40% 60% 80% 100%

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 21/45

    4.4.2. Puff volume

    Considering the range of inter-lab-device variability limits computed after outlier removal, the

    Acceptability Interval () was set at 20% for puff volume. Desirability Indices are

    summarized in Figure 12.

    Figure 12: Desirability indices for puff volume

    Device A:

    Desirability Indices are systematically spread in the “Fit for purpose” area. Indeed, indices

    vary from 83% to 95%.

    Device B:

    Desirability Indices are spread between the “Questionable” and “Fit for purpose” areas.

    Indeed, indices vary from 69% to 90% and eight datasets were defective (from laboratories 1,

    2, 3 and 4).

    Device C:

    Desirability Indices are spread between the “Questionable” and “Fit for purpose” areas.

    Indeed, indices vary from 76% to 96% and one dataset was defective (from laboratory 1).

    Device D:

    Desirability Indices are spread between the “Questionable” and “Fit for purpose” areas.

    Indeed, indices vary from 75% to 97% and four datasets were defective (from laboratory 1).

    Unsatisfactory Questionable Fit for purpose

    A

    B

    C

    D

    1

    1

    1

    1

    2

    2

    2

    2

    3

    3

    3

    3

    4

    4

    4

    4

    Nb=13 Nb=0 Nb=0 Nb=8 Nb=43

    # D

    efe

    ctive

    pro

    file

    #4#1#1#2

    #1

    #4

    0% 20% 40% 60% 80% 100%

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 22/45

    5. COMMENTS AND CONCLUSIONS

    This study resulted in estimations of the intra-lab-device variability limits (r*) and the inter-

    lab-device variability limits (R*) for the two parameters of puff duration and puff volume.

    The study has been completed by using an accuracy profile per device and laboratory.

    5.1. General comments

    For puff duration and volume, non-linear relationships were observed between the flow rate

    and the intra- and inter-lab-device variabilities.

    Laboratory 1 reported some data containing odd values, leading to outlier removal and to

    defective accuracy profiles.

    It should be noted that the collaborative study took much longer than expected to complete,

    starting in April 2013 and finishing in September 2015. As a reminder, during this period the

    same device set was circulated to each of the four participating laboratories. None of the

    devices or associated equipment were serviced for the duration of the study. This may have

    affected the performance of some devices for the later laboratories that participated.

    5.2. Puff duration

    Coefficients of variation of the intra-lab-device variability CVr* (r* expressed as a percentage

    of the overall mean) of puff duration ranged from 4% to 12% regardless of the cigarette

    sample (West White or Camel Blue) and the mode (Lit or Unlit). However, the coefficients of

    variation of the inter-lab-device variability ranged from 15% to 32%.

    The Desirability Indices of non-defective accuracy profiles (57 of 64) ranged from 66% to

    98%. Two devices, B and D, provided better Desirability Indices (from 85% to 98%) than A

    and C (from 66% to 95%).

    Most of the laboratories showed significantly higher puff duration variabilities for device A

    than for the other devices.

    Laboratory 1 reported data containing odd values, leading to outlier removal and to defective

    accuracy profiles.

    5.3. Puff volume

    Coefficients of variation of the intra-lab-device variability CVr* ranged from 4% to 13%,

    which is similar to those obtained for puff duration. However, the coefficients of variation of

    the inter-lab-device variability for puff volume ranged from 10% to 18%, which is

    significantly lower than the range obtained for puff duration.

    The Desirability Indices of non-defective accuracy profiles (51 of 64) ranged from 69% to

    97%. Device A provided the best Desirability Indices (from 83% to 95%), followed by C and

    D (from 75% to 97%), and then device B (from 69% to 90%, but with a high percentage

    [50%] of defective accuracy profiles).

    Most of the laboratories showed significantly higher puff volume variabilities for device B

    than for the other devices.

    The high variabilities reported for device D seemed to be the result of odd data obtained with

    one of the three units used.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 23/45

    APPENDIX A – List of participating laboratories and device

    manufacturers

    Laboratories and devices are listed below in alphabetic order. The codifications used in this

    report were randomly assigned to laboratories and devices.

    Laboratories Country

    BAT United Kingdom

    ITG France

    RJRT USA

    SODIM France

    Device Manufacturer

    CME-P4 RJRT

    Cressmicro BORGWALDT

    SA7 BAT

    SPA-M SODIM

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 24/45

    APPENDIX B – Part 2 - Intra and inter-lab-device variability

    estimations

    B.1 Puff duration

    Estimations for puff duration are summarized in Table B.1.1 by keeping outlier dataset

    (combination laboratory-device) and in Table B.1.2 by removing outlier dataset.

    Table B.1.1: r* and R* estimations for puff duration with outlier dataset

    Table B.1.2: r* and R* estimations for puff duration without outlier dataset

    B.2 Puff Volume

    Estimations for puff volume are summarized in Table B.2.1 by keeping outlier dataset

    (combination laboratory-device) and in Table B.2.2 by removing outlier dataset.

    Table B.2.1: r* and R* estimations for puff volume with outlier dataset

    Table B.2.2: r* and R* estimations for puff volume without outlier dataset

    Mode Regime Nb combination

    labo-device

    Mean r* R* CVr* CVR*

    LIT SR3 16 2.96 1.452 1.645 49.0% 55.5%

    UNLIT SR3 16 2.94 1.585 1.720 53.9% 58.5%

    Puff duration [s]

    Mode Regime Nb combination

    labo-device

    Mean r* R* CVr* CVR*

    LIT SR3 12 2.85 0.273 0.445 9.6% 15.6%

    UNLIT SR3 12 2.83 0.186 0.456 6.6% 16.1%

    Puff duration [s]

    Mode Regime Nb combination

    labo-device

    Mean r* R* CVr* CVR*

    LIT SR3 16 92.66 21.627 28.035 23.3% 30.3%

    UNLIT SR3 16 89.38 18.840 23.860 21.1% 26.7%

    Puff volume [ml]

    Mode Regime Nb combination

    labo-device

    Mean r* R* CVr* CVR*

    LIT SR3 12 92.72 11.039 16.647 11.9% 18.0%

    UNLIT SR3 12 90.47 5.508 11.597 6.1% 12.8%

    Puff volume [ml]

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 25/45

    APPENDIX C – Data representation

    The following figures show the raw data for the five smoking regimes, the two lighting modes

    and the two reported parameters. The plots indicate mean values with 95% confidence

    interval (CI) highlighted in green (for valid laboratories) or in red (for laboratories detected as

    an outlier). Laboratories excluded by outlier statistics are denoted by the name of the test

    rejecting the laboratory (G for Grubbs, G2 for double Grubbs or C for Cochran). The

    confidence limits at 95% and 99% for the difference between the reference value (average of

    all “valid” laboratories) and the average of one laboratory are plotted in green and red,

    respectively.

    Codes for participating laboratories that did not provide data are marked with an asterisk.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 26/45

    C.1 Puff duration

    Smoking regime 1

    Figure 13: Individual observations of Puff Duration in LIT mode

    Figure 14: Individual observations of Puff Duration in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 27/45

    Smoking regime 2

    Figure 15: Individual observations of Puff Duration in LIT mode

    Figure 16: Individual observations of Puff Duration in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 28/45

    Smoking regime 3

    Figure 17: Individual observations of Puff Duration in LIT mode

    Figure 18: Individual observations of Puff Duration in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 29/45

    Smoking regime 4

    Figure 19: Individual observations of Puff Duration in LIT mode

    Figure 20: Individual observations of Puff Duration in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 30/45

    Smoking regime 5

    Figure 21: Individual observations of Puff Duration in LIT mode

    Figure 22: Individual observations of Puff Duration in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 31/45

    C.2 Puff Volume

    Smoking regime 1

    Figure 23: Individual observations of Puff Volume in LIT mode

    Figure 24: Individual observations of Puff Volume in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 32/45

    Smoking regime 2

    Figure 25: Individual observations of Puff Volume in LIT mode

    Figure 26: Individual observations of Puff Volume in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 33/45

    Smoking regime 3

    Figure 27: Individual observations of Puff Volume in LIT mode

    Figure 28: Individual observations of Puff Volume in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 34/45

    Smoking regime 4

    Figure 29: Individual observations of Puff Volume in LIT mode

    Figure 30: Individual observations of Puff Volume in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 35/45

    Smoking regime 5

    Figure 31: Individual observations of Puff Volume in LIT mode

    Figure 32: Individual observations of Puff Volume in UNLIT mode

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 36/45

    APPENDIX D – Accuracy profile results

    Puff Duration: complete results of the computed accuracy profiles.

    Device.code Labo.code Lighting Sample Beta Lambda Accuracy_Index Range_Index Precision_Index Trueness_Index

    A 1 LIT CB 90 30 0 0 0 0

    A 1 LIT WW 90 30 76.5 73.26 64.07 95.36

    A 1 UNLIT CB 90 30 74.31 69.3 61.36 96.48

    A 1 UNLIT WW 90 30 81.61 84.33 67.93 94.9

    A 2 LIT CB 90 30 79.79 78.79 68.34 94.33

    A 2 LIT WW 90 30 66.35 39.66 77.57 94.92

    A 2 UNLIT CB 90 30 86.85 87.17 77.6 96.84

    A 2 UNLIT WW 90 30 69.18 46.4 75.56 94.45

    A 3 LIT CB 90 30 74.66 61.86 68.67 97.98

    A 3 LIT WW 90 30 84.6 83.48 77.24 93.91

    A 3 UNLIT CB 90 30 80.82 72.1 78.95 92.74

    A 3 UNLIT WW 90 30 82.52 75.62 79.02 94.02

    A 4 LIT CB 90 30 0 0 0 0

    A 4 LIT WW 90 30 86.38 84.56 79.96 95.31

    A 4 UNLIT CB 90 30 87.84 94.85 73.43 97.31

    A 4 UNLIT WW 90 30 89.09 90.43 81.44 96

    B 1 LIT CB 90 30 0 0 0 0

    B 1 LIT WW 90 30 0 0 0 0

    B 1 UNLIT CB 90 30 0 0 0 0

    B 1 UNLIT WW 90 30 0 0 0 0

    B 2 LIT CB 90 30 86.51 100 71.68 90.34

    B 2 LIT WW 90 30 91.92 100 91.12 85.23

    B 2 UNLIT CB 90 30 86.47 100 73.26 88.27

    B 2 UNLIT WW 90 30 91.91 100 91.35 84.99

    B 3 LIT CB 90 30 97.05 100 95.32 95.91

    B 3 LIT WW 90 30 95.72 100 94.51 92.81

    B 3 UNLIT CB 90 30 96.77 100 94.97 95.41

    B 3 UNLIT WW 90 30 96.16 100 95.92 92.7

    B 4 LIT CB 90 30 93.7 100 90.92 90.48

    B 4 LIT WW 90 30 93.9 100 96.12 86.13

    B 4 UNLIT CB 90 30 93.7 100 92.43 89.02

    B 4 UNLIT WW 90 30 94.12 100 97.34 85.66

    C 1 LIT CB 90 30 79.68 100 52.32 96.68

    C 1 LIT WW 90 30 92.98 100 84.15 95.51

    C 1 UNLIT CB 90 30 0 0 0 0

    C 1 UNLIT WW 90 30 92.87 100 84.18 95.15

    C 2 LIT CB 90 30 69.39 100 36.42 91.74

    C 2 LIT WW 90 30 89.39 100 82.21 86.89

    C 2 UNLIT CB 90 30 74.65 100 46.24 89.95

    C 2 UNLIT WW 90 30 87.42 100 77.16 86.59

    C 3 LIT CB 90 30 91.33 100 78.98 96.44

    C 3 LIT WW 90 30 95.07 100 89.62 95.88

    C 3 UNLIT CB 90 30 93.11 100 84.02 96.09

    C 3 UNLIT WW 90 30 94.71 100 88.8 95.67

    C 4 LIT CB 90 30 75.19 100 46.91 90.62

    C 4 LIT WW 90 30 87.84 100 78.12 86.75

    C 4 UNLIT CB 90 30 82.66 100 63.31 89.2

    C 4 UNLIT WW 90 30 88.57 100 80.12 86.73

    D 1 LIT CB 90 30 88.21 100 71.24 96.35

    D 1 LIT WW 90 30 97.44 100 93.23 99.24

    D 1 UNLIT CB 90 30 96.45 100 91.02 98.58

    D 1 UNLIT WW 90 30 97.67 100 93.86 99.26

    D 2 LIT CB 90 30 85.8 100 72.27 87.41

    D 2 LIT WW 90 30 93.86 100 91.94 89.92

    D 2 UNLIT CB 90 30 92.43 100 88.49 89.23

    D 2 UNLIT WW 90 30 91.43 100 84.8 90.14

    D 3 LIT CB 90 30 89.03 100 72.51 97.32

    D 3 LIT WW 90 30 98.23 100 96.51 98.21

    D 3 UNLIT CB 90 30 97.36 100 94.07 98.1

    D 3 UNLIT WW 90 30 98.41 100 96.89 98.38

    D 4 LIT CB 90 30 85.39 100 73.43 84.79

    D 4 LIT WW 90 30 92.46 100 88.24 89.58

    D 4 UNLIT CB 90 30 93.95 100 93.84 88.38

    D 4 UNLIT WW 90 30 92.79 100 89.05 89.71

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 37/45

    Puff Volume: complete results of the computed accuracy profiles.

    Device.code Labo.code Lighting Sample Beta Lambda Accuracy_Index Range_Index Precision_Index Trueness_Index

    A 1 LIT CB 90 20 83.09 100 66.36 86.45

    A 1 LIT WW 90 20 86.7 100 69.22 94.15

    A 1 UNLIT CB 90 20 86.15 98.52 68.78 94.37

    A 1 UNLIT WW 90 20 86.63 100 68.05 95.53

    A 2 LIT CB 90 20 84.08 100 72.1 82.43

    A 2 LIT WW 90 20 93.41 100 86.36 94.38

    A 2 UNLIT CB 90 20 92.84 100 82.28 97.24

    A 2 UNLIT WW 90 20 91.89 100 80.85 95.98

    A 3 LIT CB 90 20 91 100 77.22 97.57

    A 3 LIT WW 90 20 90.86 100 77.31 97.02

    A 3 UNLIT CB 90 20 94.6 100 89.62 94.47

    A 3 UNLIT WW 90 20 91.52 100 79.38 96.57

    A 4 LIT CB 90 20 84.25 100 75.54 79.17

    A 4 LIT WW 90 20 91.26 100 81.78 92.94

    A 4 UNLIT CB 90 20 93.62 100 85.2 96.32

    A 4 UNLIT WW 90 20 90.78 100 79.34 94.31

    B 1 LIT CB 90 20 0 0 0 0

    B 1 LIT WW 90 20 0 0 0 0

    B 1 UNLIT CB 90 20 0 0 0 0

    B 1 UNLIT WW 90 20 0 0 0 0

    B 2 LIT CB 90 20 0 0 0 0

    B 2 LIT WW 90 20 90.35 100 89.84 82.09

    B 2 UNLIT CB 90 20 74.15 86.95 86.8 54.01

    B 2 UNLIT WW 90 20 89.02 100 84.31 83.68

    B 3 LIT CB 90 20 0 0 0 0

    B 3 LIT WW 90 20 87.27 100 74.34 89.4

    B 3 UNLIT CB 90 20 69.15 84.95 69.14 56.29

    B 3 UNLIT WW 90 20 85.87 100 70.9 89.3

    B 4 LIT CB 90 20 0 0 0 0

    B 4 LIT WW 90 20 0 0 0 0

    B 4 UNLIT CB 90 20 71.63 83.99 83.71 52.27

    B 4 UNLIT WW 90 20 87.01 100 76.66 85.92

    C 1 LIT CB 90 20 76.48 98.58 47.37 95.79

    C 1 LIT WW 90 20 79.99 100 52.56 97.36

    C 1 UNLIT CB 90 20 86.69 100 66.35 98.19

    C 1 UNLIT WW 90 20 0 100 0 97.36

    C 2 LIT CB 90 20 88.92 100 73.99 95.03

    C 2 LIT WW 90 20 94.74 100 86.21 98.64

    C 2 UNLIT CB 90 20 92.03 100 79.41 98.15

    C 2 UNLIT WW 90 20 92.11 100 79.67 98.08

    C 3 LIT CB 90 20 90.73 100 74.81 99.86

    C 3 LIT WW 90 20 94.23 100 84.25 99.31

    C 3 UNLIT CB 90 20 89.6 100 72.08 99.79

    C 3 UNLIT WW 90 20 94.77 100 85.61 99.41

    C 4 LIT CB 90 20 93.61 100 85.38 96.07

    C 4 LIT WW 90 20 95.54 100 88.54 98.51

    C 4 UNLIT CB 90 20 95.2 100 87.61 98.47

    C 4 UNLIT WW 90 20 94.86 100 86.95 98.18

    D 1 LIT CB 90 20 0 0 0 0

    D 1 LIT WW 90 20 0 0 0 0

    D 1 UNLIT CB 90 20 0 0 0 0

    D 1 UNLIT WW 90 20 0 0 0 0

    D 2 LIT CB 90 20 90.08 100 75.86 96.35

    D 2 LIT WW 90 20 97.31 100 92.26 99.87

    D 2 UNLIT CB 90 20 93.15 100 82.55 97.93

    D 2 UNLIT WW 90 20 97.04 100 91.85 99.48

    D 3 LIT CB 90 20 87.29 97.5 75.7 90.1

    D 3 LIT WW 90 20 97.41 100 95.82 96.45

    D 3 UNLIT CB 90 20 77.51 84.84 69.12 79.42

    D 3 UNLIT WW 90 20 96.77 100 95.35 95.05

    D 4 LIT CB 90 20 75.09 100 43.34 97.69

    D 4 LIT WW 90 20 91.88 100 77.87 99.6

    D 4 UNLIT CB 90 20 78.57 100 49.82 97.35

    D 4 UNLIT WW 90 20 91.3 100 76.76 99.15

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 38/45

    APPENDIX E – Experimental protocol

    Study title

    CORESTA collaborative study to evaluate the accuracy and precision of puffing topography

    devices.

    Background/Rationale

    Commercial and proprietary puffing topography equipment have been available and used for

    many years. However, there is limited information regarding validation of instruments used

    for puffing topography. The FDA’s recent Draft Guidance for Applications for Premarket

    Review of New Tobacco Products (PMTA) recommends studies to evaluate topography data

    from adult smokers. This study has the advantage of evaluating several topography devices

    under the same conditions, in different laboratories.

    Objectives

    This study will measure puff volume and puff duration with four different types of human

    topography devices, using two commercial cigarettes that represent the low and higher end

    pressure drops (PD) of commercial products. The cigarettes will be tested over various puff

    volumes and durations in both unlit and lit machine smokings. The data collected will be used

    to determine:

    Accuracy: How close a measured value is to the true value (calibrated puff volume).

    This will be done by taking 10 puffs on unlit cigarettes and 10 puffs on lit cigarettes

    (unless a set butt length is reached) at each of five puffing regimes (Table 1).

    Precision: How close the measured values of each unit are to each other (measured puff

    volume). This is measured in terms of Repeatability and Reproducibility.

    Repeatability: within device and within laboratory will be determined.

    Reproducibility: within and between device type and within and between laboratories

    will be determined.

    Design

    This study will assess four topography devices over a matrix of puff volumes and durations

    that cover the range of these parameters found in typical cigarette smoking behaviour. It will

    include measuring two different commercial cigarette brands with different pressure drops and

    the cigarettes will be measured both when unlit and lit.

    Each participating laboratory will be provided with five units of each of the four different types

    of topography devices, although only three units of each type will be used in the study. Serial

    numbers will be recorded to ensure that each laboratory uses the same subset of units.

    Each device will be calibrated prior to use following provided instructions. A smoking

    machine or puff generator (Borgwaldt A14 syringe driver) capable of drawing constant flow

    rates will be used to generate the five puffing regimes (Table 1). The smoking machine will be

    set up to produce the constant flow rates and duration conditions shown in Table 1. A bubble

    meter will be used to confirm the puff volumes after changing smoking machine settings and

    daily prior to testing. When confirming the puff volume, a standard pressure drop of

    approximately 100mmWG will be used. Bubble meter readings will be recorded.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 39/45

    Table 1: Puff Volume, Duration Matrix

    Regime Puff

    Profile

    Puff Volume (mL)

    Duration (s)

    Inter-Puff Interval(s)

    Flow Rate (mLs

    -1)

    1 Square 35 3.5 30 10

    2 Square 80 4.0 30 20

    3 Square 90 3.0 30 30

    4 Square 80 2.0 30 40

    5 Square 50 1.0 30 50

    Each puff in the matrix above will be drawn 10 times (10 total puffs) through each of the three

    units of the four different topography devices with the cigarette initially unlit. Following each

    set of 10 unlit puffs, the cigarette will be lit using an electric lighter and 10 puffs will be drawn,

    or if drawing 10 puffs would burn into the filter, puffs will be drawn through the cigarette until

    the burning coal reaches a pre-determined butt length (overwrap +3 mm). This process will be

    repeated for both commercial cigarette brands.

    In addition, the 30 mLs-1

    flow rate for 3.0 seconds duration, puff volume 90 mL smoking

    regime will be repeated (x 10 lit & unlit puffs) with one of each of the four units over five

    separate days, smoking both the West King Size and Camel King Size cigarettes. The same

    unit should be used for the additional smoking on each day. The original day of smoking on

    each device will be counted as the first of the five days.

    Settings

    This analysis will be performed in a laboratory at each of the participating companies.

    Laboratory conditions will be controlled according to ISO 3308:

    Temperature: 22°C +/- 2°C;

    Humidity: 60%RH + / - 5%RH

    The cigarettes will be conditioned prior to testing, according to ISO 4387:

    Temperature: 22°C +/- 1°C;

    Humidity: 60%RH + / - 3%RH

    Cigarettes will be conditioned for at least 48 hours but no more than 10 days.

    There is no requirement to record the airflow above the cigarettes.

    Study equipment and supplies

    Topography devices under investigation include:

    1. CReSSmicro (Borgwaldt)

    2. SPA-M (Sodim)

    3. SA7 (BAT)

    4. CME5-P4 (RJRT)

    Five units of each of the above devices will be shipped to each laboratory taking part in the

    study.

    5 Carolina Medical Electronics, Inc.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 40/45

    Cigarettes:

    A sufficient quantity of each of the two commercial cigarettes will be supplied along with the

    devices. The two commercial cigarettes are:

    West White King Size Box (ITG)

    Camel Blue King Size Box (RJRT)

    Physical Data for:

    West White King Size Box

    o Open pressure drop (draft) - 45 mm of H20 (SD = 1.4)

    o Closed pressure drop (draft) - 151 mm of H20 (SD = 11.0)

    o Filter pressure drop (draft) - NA

    Camel Blue King Size Box

    o Open pressure drop (draft) - 130.5 mm of H20 (SD=7.5)

    o Closed pressure drop (draft) - 157.1 mm of H20 (SD=9.3)

    o Filter pressure drop (draft) - 98.0 mm of H20 (SD=6.5)

    Smoking Machine :

    A Borgwaldt A14 syringe driver (single-port smoking machine) or a linear, multi-port smoking

    machine, with the capability of generating square puff profiles, will be used in each laboratory.

    The Borgwaldt A14 syringe driver will be provided by BAT. The Cambridge filter pad will be

    changed after smoking each six cigarettes in the lit condition.

    Assessments

    Ten measurements of

    Puff volume (mL)

    Puff duration (s)

    Peak flow (mLs-1, where possible)

    Resistance to draw (mLs-1-mmH2O, where possible)

    Each puff in the matrix will be drawn through each of the three units of all four different

    topography devices and recorded with the cigarette unlit first, followed by lighting the cigarette

    with an electric lighter and puffing either 10 puffs or to a pre-determined butt length

    (overwrap+3mm) if 10 puffs would result in burning into the filter.

    This will be repeated sequentially using the second commercial cigarette.

    In addition, the 30 mLs-1

    flow rate for 3.0 seconds duration, puff volume 90 mL smoking

    regime will be repeated (x 10 lit puffs only) with one of each of the four devices over five

    separate days (same unit each day).

    Operation

    The laboratory operators will be trained on the usage of each type of topography device. A

    user guide will be available for each device type.

    Data collation

    A standard template (Excel spreadsheet) will be provided to report the measurements.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 41/45

    Data analysis

    The evaluation of instruments will be based on a comparison of accuracy profiles.

    Figure 1 gives an illustration of an accuracy profile with the key definitions.

    Figure. 1: Illustration of an accuracy profile.

    In this study, for each laboratory and each device, two accuracy profiles will be developed by

    estimating the precision and the “trueness” 1) for each flow rate and 2) for each puff duration.

    A specific device will meet the objective if the accuracy profile is within the acceptance limits.

    The acceptance limits, for each studied parameter, will be defined by the CORESTA subgroup

    members after a preliminary analysis of results.

    An example of accuracy profiles is shown in Figure 2 (accuracy profiles based on the flow rate

    to compare two instruments).

    Figure. 2: Accuracy profiles based on the flow rate of devices 1 and 2.

    To compare the accuracy profile of different devices, an indicator score will be calculated.

    This score is the sum of the combination of three quality criteria:

    Range Index

    Precision Index

    Trueness Index

    bias (%)

    parameter

    +

    -

    C1 C2 C3 C4

    RLinfRLsup

    RANGE

    mean relative bias

    accep

    tance

    limits

    0

    bias limits of confidence

    "Profil Exactitude : Modèle linéaire"

    Concentrations

    Re

    co

    uvre

    me

    nt

    (%)

    10 20 30 40 50

    70

    80

    90

    100

    110

    120

    130

    Device 1

    Flow rate

    Recovery

    (%

    )

    "Profil Exactitude : Modèle linéaire"

    Concentrations

    Re

    co

    uvre

    me

    nt

    (%)

    10 20 30 40 50

    70

    80

    90

    100

    110

    120

    130

    Device 1

    Flow rate

    Recovery

    (%

    )

    "Profil Exactitude : Modèle linéaire"

    Concentrations

    Re

    co

    uvre

    me

    nt

    (%)

    10 20 30 40 50

    60

    80

    100

    120

    Device 2

    Flow rate

    Re

    co

    ve

    ry (

    %)

    "Profil Exactitude : Modèle linéaire"

    Concentrations

    Re

    co

    uvre

    me

    nt

    (%)

    10 20 30 40 50

    60

    80

    100

    120

    Device 2

    Flow rate

    Re

    co

    ve

    ry (

    %)

    Device 2

    Flow rate

    Re

    co

    ve

    ry (

    %)

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 42/45

    A score closer to 100 indicates that the device better met the objective. For example, in Table

    2, Device 2 (Accuracy_Index = 78.67) is slightly better for this comparison than Device 1

    (Accuracy_Index = 77.95).

    Table 2: Accuracy index for devices 1 and 2

    This statistical evaluation is proposed for the unlit cigarettes by considering a puff as a

    replicate to estimate the repeatability (10 puffs), and an instrument unit as a replicate to

    estimate the intermediate precision (3 units). The same evaluation will be made with the lit

    cigarettes, with the exception that a variable number of puffs may be used (by smoking instead

    to a set butt length), as required. See Appendix 2 for an explanation of terms used in the Data

    Analysis section.

    Appendix 1

    Borgwaldt A14 Syringe driver entry guidelines

    Setting Entry / Format Entry Example

    Keyboard Control 2 2

    Smoking 2 2

    Puff Parameters 2 2

    Bell profile? No No

    Profile no. 1 1 = square puff

    Correct? Yes Yes

    Puff vol. (ml) ### 080

    Correct? Yes Yes

    Puff dur. (s) #,# 4,0 = 4.0

    Correct? Yes Yes

    Puff period (s) ## 30

    Correct? Yes Yes

    Puff vol. test? Yes Yes

    Tester o.k.? Connect bubble meter prior to

    selecting Yes

    Yes

    Puff volume test

    Volume o.k.? Yes Yes

    Accuracy_Index Range_Index Precision_Index Trueness_Index RL_Inf RL_Sup

    Device 1 77.95 99.26 48.68 98.01 10.3 50.0

    Device 2 78.67 93.1 54.23 96.46 12.8 50.0

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 43/45

    Appendix 2

    The relevance of method performance is defined by indicators6

    Accuracy: closeness of agreement between a test result or measurement result and the true

    value. Accuracy refers to a combination of trueness (common systematic error) and precision

    (random error).

    Trueness: closeness of agreement between the expected test result or measured result

    and a true value.

    Precision: closeness of agreement between independent test/measurement results

    obtained under stipulated conditions.

    -repeatability: precision under repeatability conditions.

    -repeatability conditions: observation conditions where independent

    test/measurement results are obtained with the same method on identical

    test/measurement items in the same test or measurement facility by the same

    operator using the same equipment within a short interval of time.

    -intermediate precision: precision under intermediate precision conditions.

    -intermediate precision conditions: conditions where test results or

    measurement results are obtained with the same method, on identical

    test/measurement items in the same test or measurement facility, under some

    different operating condition.

    6 ISO 15725-1

    TruenessTrueness

    Accuracy

    =

    Trueness + precision

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 44/45

    APPENDIX F – Departures from experimental protocol

    No departures from the experimental protocol were reported.

  • TSB-047-CTR Puffing Topography Inter-lab Study – January 2017 45/45

    APPENDIX G – Definitions

    Repeatability r:

    The difference between two single results found on matched cigarette samples by one operator

    using the same apparatus within the shortest feasible time interval will exceed the repeatability

    value ( r ) on average not more than once in 20 cases in the normal and correct operation of the

    method.

    ISO 10362-1:1991(E), page 3

    Reproducibility R:

    Single results on matched cigarette samples reported by two laboratories will differ by more

    than the reproducibility value ( R ) on average not more than once in 20 cases in the normal

    and correct operation of the method.

    ISO 10362-1:1991(E), page 3

    1. INTRODUCTION2. EXPERIMENTAL DESIGN2.1 Part 1: Intra and inter-lab-device variability2.2 Part 2: Time effect

    3. STATISTICAL EVALUATION3.1. Uncertainty evaluation3.2. Accuracy profile3.3. Limitations

    4. RESULTS4.1. Outlier results4.1.1. Puff duration4.1.2. Puff volume

    4.2. Intra and inter-lab-device variability estimations4.2.1 Puff duration4.2.2. Puff volume

    4.3. Intra and inter-lab-device variability modeling4.3.1. Puff duration4.3.2. Puff volume

    4.4. Desirability indices4.4.1. Puff duration4.4.2. Puff volume

    5. COMMENTS AND CONCLUSIONS5.1. General comments5.2. Puff duration5.3. Puff volume

    APPENDIX A – List of participating laboratories and device manufacturersAPPENDIX B – Part 2 - Intra and inter-lab-device variability estimationsAPPENDIX C – Data representationAPPENDIX D – Accuracy profile resultsAPPENDIX E – Experimental protocolAPPENDIX F – Departures from experimental protocolAPPENDIX G – Definitions