13
Routine Analytical Chemistry Sub-Group Technical Report 2016 Collaborative Study of CORESTA Ignition Propensity Monitor Test Piece CM IP 2 for the Determination of Ignition Propensity March 2017 Project Leader: Guy Jaccard, Philip Morris International, Switzerland Authors: Donatien Tafin Djoko and Guy Jaccard, Philip Morris International, Switzerland

Technical Report 2016 Collaborative Study of CORESTA

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Technical Report 2016 Collaborative Study of CORESTA

Routine Analytical Chemistry Sub-Group

Technical Report

2016 Collaborative Study of

CORESTA Ignition Propensity Monitor Test Piece CM IP 2 for the

Determination of Ignition Propensity

March 2017

Project Leader:

Guy Jaccard, Philip Morris International, Switzerland

Authors:

Donatien Tafin Djoko and Guy Jaccard, Philip Morris International, Switzerland

Page 2: Technical Report 2016 Collaborative Study of CORESTA

Table of Contents

Table of Contents ............................................................................................................................. 2

1. Introduction ............................................................................................................................. 3

2. Organisation ............................................................................................................................ 3

2.1 Participants ................................................................................................................... 3

2.2 Protocol ........................................................................................................................ 4

3. Raw data ................................................................................................................................. 4

4. Statistical Analysis .................................................................................................................. 5

5. Results ..................................................................................................................................... 7

6. References ............................................................................................................................... 8

Appendix A: Descriptive Statistics .................................................................................................. 9

Appendix B: Test protocol ............................................................................................................. 12

Page 3: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 3/13

1. Introduction

The CORESTA Routine Analytical Chemistry Sub-Group has been given the responsibility to

provide a monitor test piece specific for ignition propensity testing, according to ISO

12863:2010.

A candidate CORESTA monitor test piece had been manufactured in the Philip Morris

International Neuchâtel (Switzerland) factory in January 2014, with product specifications

established on the basis of prototype cigarettes previously analyzed.

It has been qualified as CORESTA Monitor Test Piece CM IP 2 in 2015, on the basis of a

2014 collaborative study and re-confirmed a year later (see CORESTA RAC Technical

Report, 2014 collaborative study of CORESTA Ignition Propensity Monitor Test piece CM IP

2 for the determination of ignition propensity and the corresponding 2015 collaborative study

report, edited in March 2015 and April 2016 respectively).

The need to continue checking the stability of CM IP 2 monitor test piece and using the

results to support the laboratories proficiency and maintain their accreditation on the basis of

a collaborative study of CM IP 2 and of NIST standard reference material 1082 for

comparison purpose has been agreed during the CORESTA RAC Lausanne meeting in April

2016. The project has been approved by the CORESTA Scientific Commission and registered

as project RAC-113.

The goal of this report is to provide the statistical assessment of the results of the

collaborative study mentioned above.

2. Organisation

2.1 Participants

In total 14 laboratories participated in the collaborative study, according to the below list:

British American Tobacco, Germany

delfortgroup, Austria

Essentra Scientific Services, UK

Global Laboratory Services, USA

Heintz van Landewyck, Luxemburg

ITL Reemtsma, Germany

ITL Seita, France

Japan Tobacco, Japan

Karelia, Greece

Philip Morris International, Germany

Philip Morris International, Poland

Philip Morris International, Switzerland

Sampoerna, Indonesia

SWM, France

Page 4: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 4/13

2.2 Protocol

Participants were requested to follow the method ISO 12863:2010 with 10 layers of the

required substrate. The number of replicates per brand (CM IP 2 and NIST Monitor) and

laboratory was set to 5 on 5 different days. Each ignition propensity test consisted of a sample

of 40 cigarettes. The protocol sent to participants is provided in Annex B (pp. 11-12).

CM IP 2 Monitor Test Piece and NIST standard reference material 1082 were to be obtained

directly from respective suppliers: Borgwaldt and Cerulean for CM IP 2; NIST for NIST

reference material.

3. Raw data

The table below lists the results obtained by individual laboratories for the number (n) and the

percentage (%) of cigarettes with full length burn (FLB).

Table 1: Results of the individual laboratories

Laboratory CM IP 2 (n FLB) CM IP 2 (% FLB) NIST (n FLB) NIST (% FLB)

1 2; 1; 2; 5; 1 5; 2.5; 5; 12.5; 2.5 4; 3; 7; 3; 9 10; 7.5; 17.5; 7.5; 22.5

2 1; 0; 3; 0; 0 2.5; 0; 7.5; 0; 0 0; 2; 1; 1; 0 0; 5; 2.5; 2.5, 0

3 1; 1; 0; 0; 1 2.5; 2.5; 0; 0; 2.5 2; 2; 1; 2; 1 5; 5; 2.5; 5; 2.5

4 1; 0; 0; 0; 0 2.5; 0; 0; 0; 0 0; 0; 0; 1; 0 0; 0, 0; 2.5; 0

5 4; 2; 3; 3; 4 10; 5; 7.5; 7.5; 10 4; 4; 4; 4; 5 10; 10; 10; 10; 12.5

6 4; 1; 1; 2; 2 10; 2.5; 2.5; 5; 5 3; 0; 3; 0; 3 7.5; 0; 7.5; 0; 7.5

7 2; 3; 2; 3; 0 5; 7.5; 5; 7.5; 0 5; 6; 5; 5; 5 12.5; 15; 12.5; 12.5; 12.5

8 1; 0; 1; 1; 2 2.5; 0; 2.5; 2.5; 5 5; 3; 2; 2; 5 12.5; 7.5; 5; 5; 12.5

9 0; 2; 0; 0; 0 0; 5; 0; 0; 0 3; 5; 10; 13; 5 7.5; 12.5; 25; 32.5; 12.5

10 1; 1; 0; 2; 1 2.5; 2.5; 0; 5; 2.5 No data No data

11 3; 6; 1; 3; 3 7.5; 15; 2.5; 7.5; 7.5 6; 13; 9; 6; 8 15, 32.5; 22.5; 15; 20

12 3; 2; 1; 2; 0 7.5; 5; 2.5; 5; 0 5; 3; 8; 6; 3 12.5; 7.5; 20; 15; 7.5

13 1; 2; 0; 0; 1 2.5; 5; 0; 0; 2.5 4; 3; 4; 5; 5 10; 7.5; 10; 12.5; 12.5

14 5; 3; 3; 3; 2 12.5; 7.5; 7.5; 7.5; 5 No data No data

Two laboratories did not report results for the NIST reference material (laboratories 10 and

14).

Page 5: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 5/13

4. Statistical Analysis

Ignition propensity tests are typical types of proportion-based analysis with a binary response

variable: Full Length Burning or not. Statistical analysis is performed assuming that the data

generating process is binomial.

Within each laboratory, over a short period, each cigarette is a realization of a Bernoulli

random variable with parameter 𝑝. If a random sample of 𝑛 cigarettes is selected, and if 𝐵 is

the number of units fully burned, then 𝐵 has binomial distribution with parameters 𝑛 and 𝑝:

𝑃{𝐵 = 𝑥} = (𝑛

𝑥) 𝑝𝑥 (1 − 𝑝)𝑛−𝑥 𝑥 = 0, 1, ⋯ , 𝑛

The sample fraction of Full Length Burn is defined as the ratio of the number of cigarettes

fully burned 𝐵 to the sample size 𝑛:

�̂� = 𝐵

𝑛

And the distribution of the random variable �̂� can be obtained from the binomial with

𝑚𝑒𝑎𝑛, 𝜇𝑝 = 𝑝, 𝑎𝑛𝑑 𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒, 𝜎𝑝2 =

𝑝 (1 − 𝑝)

𝑛

In our case, for each replicate, we have:

n = 40: Number of cigarettes in each replicate

r = 5: Number of replicates per laboratory

Having 𝐵𝑗,𝑖 fully burned cigarettes for laboratory 𝑗 in sample 𝑖, fraction of individual

determination of fully burned cigarettes in the 𝑖th sample is computed as

�̂�𝑗,𝑖 =𝐵𝑗,𝑖

𝑛 𝑖 = 1, 2, ⋯ , 𝑟

The average (amongst individual determination) within each laboratory is given by

�̅�𝑗 =∑ 𝐵𝑗,𝑖

𝑟𝑖=1

𝑟 ∗ 𝑛=

∑ �̂�𝑗,𝑖𝑟𝑖=1

𝑟

Page 6: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 6/13

Therefore the within standard deviation for laboratory 𝑗 is computed as

𝜎𝑗 = √�̅�𝑗 (1 − �̅�𝑗)

𝑛

The aggregated within laboratory standard deviation (repeatability standard deviation) is

defined as

𝑆𝑟 = √∑ 𝜎𝑗

2𝐿𝑗=1

𝐿, 𝑗 = 1, ⋯ , 𝐿 is the number of laboratory

The between-laboratory standard deviation 𝑆𝐿,

𝑆𝐿 = √∑ (�̅�𝑗 − �̿�)2𝐿

𝑗=1

𝐿 − 1−

𝑆𝑟2

𝑟

Where

�̿� = ∑ �̅�𝑗

𝐿𝑗=1

𝐿 , 𝑗 = 1, 2, ⋯ , 𝐿

The reproducibility standard deviation,

𝑆𝑅 = √𝑆𝐿2 + 𝑆𝑟

2

The repeatability coefficient of variation,

𝐶𝑉𝑟[%] = 𝑆𝑟

�̿�∗ 100

The reproducibility coefficient of variation,

𝐶𝑉𝑅[%] = 𝑆𝑅

�̿�∗ 100

Page 7: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 7/13

The 95% confidence repeatability limit

𝑟 = 2 ∗ √2 ∗ 𝑆𝑟

The 95% confidence reproducibility limit,

𝑅 = 2 ∗ √2 ∗ 𝑆𝑅

Mandel’s h between consistency statistics is computed as

ℎ𝑗 = �̅�𝑗 − �̿�

𝑆𝐿

The within laboratory consistency statistics

𝑘𝑗 = 𝜎𝑗

𝑆𝑟

The approach implemented here is well documented in the statistical and scientific literature.

The reader may refer to NIST Technical Note 1436. Additional and advanced details on

proportion data analysis are deeply covered by Joseph L. Fleiss, et al. in “Statistical Methods

for Rates and Proportions, 3rd

Edition, Wiley”.

5. Results

The average proportions of FLB, standard deviations and the Mandel h and k statistics by

participant and brand are reported in Table 3 and in Table 4 in Appendix A. The statistics h

and k are illustrated in figures 1 and 2 in Appendix A.

Table 2 summarizes the overall statistics of the two test items in proportion of FLB for the

collaborative test organized in 2016 in comparison with the previous collaborative studies.

The CM IP 2 results are well in line with the results obtained in 2014 and 2015, in terms of

global mean, repeatability and reproducibility. Results for CM IP 2 are stable in terms of

ignition propensity test.

If used as a monitor test piece with one replicate of 40 test pieces, and assuming that a range

of mean value of ± 2SR is acceptable, one can expect values between 0 and 10.8 % (or 0 and 4

cigarettes with FLB). For the NIST 1082 standard reference material, values between 0 and

24.8% (or 0 and 10 cigarettes) would be acceptable.

Page 8: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 8/13

Table 2: Global summary statistics by test item

Global mean (%)

Sr SLc SR CVr (%)

CVR (%)

r R Test item

3.96 3.06 2.66 3.82 77.15 96.31 8.65 11.66 CM IP 2, coll. test

2016

4.10 3.12 1.39 3.42 76.11 83.34 8.83 9.66 CM IP 2, coll. test

2015

3.24 2.78 1.75 3.28 85.81 101.38 7.86 9.29 CM IP 2, coll. test

2014

9.83 4.61 6.27 7.50 46.90 76.31 13.04 23.36 NIST 1082, 2016

10.34 4.74 5.06 6.93 45.83 67.05 13.41 19.61 NIST 1082, 2015

8.73 4.44 2.58 5.13 50.81 58.78 12.55 14.51 NIST 1082, 2014

6. References

• Gann, R.G., Steckler, K.D., Ruitberg, S., Guthrie, W.F., and Levenson, M.S., "Relative

Ignition Propensity of Test Market Cigarettes" NIST Technical Note 1436, National

Institute of Standards and Technology, Gaithersburg, MD, 34 pages (2001).

• Joseph L. Fleiss, Bruce Levin, Myunghee Cho Paik, "Statistical Methods for Rates and

Proportions" 3rd

Edition, Wiley, 800 p. (2003).

Page 9: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 9/13

Appendix A: Descriptive Statistics

Table 3: Descriptive statistics for CM IP 2 monitor test piece per laboratory. h and k stands for

Mandel h (between laboratories) and k (within deviation) consistency statistics.

Laboratory P (%) SD Mandel h Mandel k

1 5.5 3.60 0.577 1.179

2 2.0 2.21 -0.737 0.724

3 1.5 1.92 -0.925 0.628

4 0.5 1.12 -1.301 0.365

5 8.0 4.29 1.515 1.403

6 5.0 3.45 0.389 1.127

7 5.0 3.45 0.389 1.127

8 2.5 2.47 -0.550 0.807

9 1.0 1.57 -1.113 0.514

10 2.5 2.47 -0.550 0.807

11 8.0 4.29 1.515 1.403

12 4.0 3.10 0.013 1.013

13 2.0 2.21 -0.737 0.724

14 8.0 4.29 1.515 1.403

Table 4: Descriptive statistics for NIST 1082 standard reference material per laboratory. h and

k stands for Mandel h (between laboratories) and k (within deviation) consistency statistics.

Laboratory P (%) SD Mandel h Mandel k

1 13.0 5.32 0.505 1.153

2 2.0 2.21 -1.250 0.480

3 4.0 3.10 -0.931 0.672

4 0.5 1.12 -1.489 0.242

5 10.5 4.85 0.106 1.051

6 4.5 3.28 -0.851 0.711

7 13.0 5.32 0.505 1.153

8 8.5 4.41 -0.213 0.956

9 18.0 6.07 1.303 1.317

10 No data

11 21.0 6.44 1.782 1.397

12 12.5 5.23 0.425 1.134

13 10.5 4.85 0.106 1.051

14 No data

Page 10: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 10/13

Figure 1: Mandel h (upper plot) and Mandel k (lower plot) statistics of FLB (%) for CM IP 2

monitor test piece (the orange and red lines are the 95 and 99% limits respectively)

-2.500

-1.500

-0.500

0.500

1.500

2.500

1 2 3 4 5 6 7 8 9 10 11 12 13 14

CM IP 2

h

FLB(%): CM IP 2 - Mandel h statistic by laboratory

0.0000.2000.4000.6000.8001.0001.2001.4001.6001.8002.0002.200

1 2 3 4 5 6 7 8 9 10 11 12 13 14

CM IP 2

k

FLB(%): CM IP 2 - Mandel k statistic by laboratory

Laboratory Codes

Laboratory Codes

Page 11: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 11/13

Figure 2: Mandel h (upper plot) and Mandel k (lower plot) statistics of FLB (%) for NIST 1082

standard reference material (the orange and red lines are the 95 and 99% limits respectively

-2.500

-1.500

-0.500

0.500

1.500

2.500

1 2 3 4 5 6 7 8 9 10 11 12 13 14

NIST

h

FLB(%): NIST Mandel h statistic by laboratory

0.0000.2000.4000.6000.8001.0001.2001.4001.6001.8002.0002.200

1 2 3 4 5 6 7 8 9 10 11 12 13 14

NIST

k

FLB(%): NIST Mandel k statistic by laboratory

Laboratory Codes

Laboratory Codes

Page 12: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 12/13

Appendix B: Test protocol

1. Objective

To assess the long term performance of the CORESTA Monitor test Piece CM IP 2,

according to ISO standard 12863, in terms of global mean value and variability (intra-

and inter-laboratory). The NIST standard is used for comparison purpose.

2. Test coordinator

Guy Jaccard

Principal Scientist

Philip Morris Products SA

Reduced-Risk Products. Product Stewardship

CH-2000 Neuchâtel

Switzerland

E-mail: [email protected]

Tel: +41 (58) 242 11 44

3. Parameter to be measured

The parameter to be measured is the ignition propensity, according to the ISO method

12863

Table 1 Parameter to be measured

Parameter

1 Ignition propensity

Number of cigarettes with full length burn

% of cigarettes with full length burn

4. Test methods

For cigarette testing, we recommend applying the following standard:

ISO 12863

Page 13: Technical Report 2016 Collaborative Study of CORESTA

RAC-113-CTR 2016 Collaborative Study CM IP 2 – March 2017 13/13

5. Test samples

Test samples consist of 1 machine-made CORESTA Monitor IP No 2 and the NIST

standard. CORESTA CM IP 2 Monitor test pieces shall be bought at Cerulean or

Borgwaldt, NIST Standard cigarettes shall be taken from each laboratory stock or

purchased.

Table 3 Test cigarettes for the Proficiency Test

Test cigarette

CORESTA Monitor Test Piece CM IP 2

NIST STANDARD

6. Schedule

For a given brand, 1 replicate of 40 cigarettes should be measured in a single test day,

for a total of 5 replicates over 5 independent days. There are no constraints concerning

the spacing between the days of experiment, however please keep them as close as

possible.

The study starts in June 2016. Each laboratory is free to organize at its will the

timeframe during which it performs the study. However, the test results should be sent

to the test coordinator no later than August 30th, 2016.

7. Reporting of test results

The test results should be reported using the Excel file Template for coll test IP_IP

No2.xls. See also the example provided in Example.xls to fill the excel template (Data

COR Day 1 sheet) with the Ignition Propensity tests results (only use the number 1 in

the appropriate column)

In addition to IP test specific results, the temperature and relative humidity during the

testing shall be reported as well.

Results should be sent electronically:

To: [email protected]