                                       
                              DRAFT FINAL REPORT
                                       
                                       
                                       
           EVALUATION OF THREE NEW CONDENSATION NUCLEI COUNTER (CNC)
                             FIT TESTING PROTOCOLS
                                       
                                       
                                       
                           LISA M BROSSEAU, ScD, CIH
                             RACHAEL M JONES, PhD
                                       
     This work was prepared for the Directorate of Standards and Guidance,
Occupational Safety and Health Administration, under contract with Project Enhancement Corporation
                                       
                                       
                                       
                               February 22, 2015
                                       

Contents
1.	INTRODUCTION AND PURPOSE	3
2.	STATEMENT OF WORK	4
3.	OVERVIEW OF ANSI/AIHA Z88.10-2010 ANNEX A2	4
4.	OVERVIEW OF PROPOSED PROTOCOLS	5
4.1	Fast-Half Method	6
4.2	Fast-Full Method	8
4.3	Fast-FFR Method	10
5.	DISCUSSION OF SELECTED EXERCISES	12
6.	DISCUSSION OF RESPIRATORS SELECTED	16
7.1	Normal Breathing Fit Factor Criterion	16
7.2	Definition of the Exclusion Zone	17
7.3	Potential Sources of Bias	18
8.	REVIEW OF ANSI/AIHA z88.10-2010 ANNEX A2 CRITERIA FOR HALF-FACEPIECE RESPIRATOR  -  INCLUDING REPRODUCTION OF DATA ANALYSIS	18
9.	REVIEW OF ANSI/AIHA Z88.10-2010 ANNEX A2 CRITERIA FOR FULL-FACEPIECE RESPIRATOR  -  INCLUDING REPRODUCTION OF DATA ANALYSIS	22
10.	REVIEW OF ANSI/AIHA z88.10-2010 ANNEX A2 CRITERIA FOR FFR  -  INCLUDING REPRODUCTION OF DATA ANALYSIS	26
11.	COMPARISON WITH EXISTING OSHA CNC FIT TEST PROTOCOL AND REFERENCE METHOD WITH RESPECT TO POORLY-FITTING RESPIRATORs	30
12.	CONCLUSIONS AND RECOMMENDATIONS	31
13.	REFERENCES	32

INTRODUCTION AND PURPOSE

The purpose of this task is to evaluate three new Condensation Nuclei Counter (CNC) fit testing protocols and the supporting studies submitted to OSHA by a fit testing instrument manufacturer to determine whether the studies submitted meet the criteria contained in the American National Standard for Respirator Fit Testing Methods, ANSI/AIHA Z88.10 - 2010, Annex A2  -  Criteria for Evaluating New Fit Test Methods. 

Appendix A of the OSHA Respiratory Protection Standard at 29 CFR 1910.134 currently includes three types of quantitative fit test (QNFT) methods: 1) generated aerosol using a non-hazardous material (corn oil, DEHS, or sodium chloride) in a test chamber measured with appropriate instrumentation; 2) ambient aerosol measured with a condensation nuclei counter (CNC); and 3) controlled negative pressure (CNP) measuring facepiece leak rate. 

Appendix A of the OSHA Respiratory Protection Standard specifies the procedure for adding new fit test protocols: (1) Submitting a test report prepared by an independent government research laboratory (e.g., Lawrence Livermore National Laboratory, Los Alamos National Laboratory, the National Institute for Standards and Technology) stating that the laboratory tested the protocol and found it to be accurate and reliable; or (2) Publishing an article(s) in a peer reviewed industrial hygiene journal describing the protocol and explaining how the test data support the protocol's accuracy and reliability.

In July, 2006, TSI Incorporated proposed two new CNC PortaCount(R) QNFT protocols for consideration by OSHA for inclusion in Appendix A of 29 CFR 1910.134. OSHA published a Notice of Proposed Rulemaking (NPRM) on January 21, 2009 (74 FR 3526-01) to solicit public comment on the two proposed CNC PortaCount(R) QNFT protocols, and received 26 comments in response to the NPRM. OSHA reviewed the comments and performed its own internal review of the two proposed protocols and concluded that the two proposed CNC PortaCount(R) QNFT protocols were not sufficiently accurate or reliable to include them among the QNFT methods in Appendix A. OSHA withdrew the proposed rule without prejudice on January 27, 2010 (75 FR 4323-01) and invited resubmission of the protocols after the developers of the protocols addressed the issues raised in the withdrawal notice and performed additional testing.

TSI Incorporated has subsequently submitted three new CNC PortaCount(R) QNFT protocols for inclusion in Appendix A. The TSI Incorporated submission included studies published in a peer reviewed industrial hygiene journal that described each of the proposed protocols and explained how test data support the protocols' accuracy and reliability. The application meets OSHA's criteria for initiating a rulemaking under section 6(b)(7) of the OSH Act. OSHA is planning to proceed with a NPRM seeking public comment on whether to amend the accepted fit testing procedures of the Respiratory Protection Standard to include the three proposed new protocols.


1. STATEMENT OF WORK

As subcontractors, we agreed to provide support at the direction of the OSHA Task Monitor including review of the literature, developing information, and performing statistical analysis to determine whether the three new fit testing protocols and their supporting studies submitted to OSHA meet the criteria contained in the American National Standard for Respirator Fit Testing Methods, ANSI/AIHA Z88.10  -  2010, Annex A2  -  Criteria for Evaluating New Fit Test Methods.

Our report is expected to include a review/analysis of the studies submitted to OSHA demonstrating the performance of the new CNC PortaCount(R) QNFT protocols to determine whether they meet the criteria as described in ANSI/AIHA Z88.10  -  2010, Annex A2  -  Criteria for Evaluating New Fit Test Methods.  The three new protocols are for filtering facepiece respirators, elastomeric half-facepiece respirators and full-facepiece respirators, and are characterized by shorter test duration and currently approved CNC PortaCount(R) QNFT protocols.  Specifically, we were asked to
   * review the elements of the new fit test protocols and how the elimination of certain fit test exercises normally required by the existing OSHA-approved protocol impact the acceptability of the new fit test protocol,
   * identify issues related to the new fit test protocols about which OSHA may wish to solicit public comment as part of the NPRM under development, and
   * perform other analyses, as directed.

We agreed that all data sources used to develop the reports under this Task Order shall be documented in the bibliography and made available to OSHA. All documents and analyses prepared for this task order will be submitted in both hard copy and an electronic format that is acceptable to OSHA (e.g., Microsoft Word or Excel).
2. OVERVIEW OF ANSI/AIHA Z88.10-2010 ANNEX A2

The ANSI/AIHA Z88.10-2010 Respirator Fit Testing Methods standard includes Annex A2 - Criteria for Evaluating New Fit Test Methods, which describes a method for comparing a new QNFT protocol to a currently-accepted QNFT protocol.  The following requirements must be met to establish that a new QNFT is similar to a current QNFT:

   1. The procedure involves performing sequential paired fit tests with the new fit test method and the reference fit test method during the same respirator donning.  The order of sequential fit tests should be randomized with respect to test method.
   2. A detailed description of the new fit test method/protocol is documented.
   3. A detailed description of the study materials and methods used is documented.
   4. Respirators used for testing shall represent a variety of different sizes and models.  The primary purpose of selecting different model respirators is to achieve a variety of different airflow and leak patterns.
   5. Reference method fit factors within one coefficient of variation of the required fit factor should be excluded. 
   6. After exclusion criteria have been applied there must be:
         a. A minimum of 100 sequential paired tests from at least 25 different subjects in the analysis data set.
         b. At least fifty of the paired tests must have reference method fit factors greater than 5% of the required fit factor and less than the required fit factor.
   7. Any reference fit factor below 10% of the required fit factor accepted by the new fit test method shall disqualify the method.
   8. Reference fit factors less than the required fit factor should be evenly distributed (i.e., not weighted towards the lower fit factors).
   9. The acceptance of a new fit test method is limited to the measurement or detection range over which the method was compared against the reference method.
   10. A cumulative distribution plot or histogram to visually confirm that the fit factors obtained from the reference method brackets the required fit factor.
   11. A table that provides respirator make, model, style, size, individuals tested and paired results of the new test and the reference test should be provided.  Similar information should be provided for any test results that were excluded from analysis.
   12. A description of the method and results used to calculate the coefficient of variation for the reference fit test method is documented.
   13. Test results should be summarized in a 2 x 2 contingency table:
         a. Test sensitivity must be at least 0.95
         b. Predictive value of a pass must be at least 0.95
         c. Test specificity must be at least 0.50
         d. Predictive value of a fail must be at least 0.50
         e. The Kappa statistic must be at least 0.70

Other considerations that might influence sampling bias of the Generated Aerosol Quantitative Fit Test Procedure, in particular, include:
   * Location and depth of probe
   * Position of face-seal leak
   * Interaction of breathing pattern with position of face-seal leak
   * Design of the facepiece
   * Measurement sample rate
   * When the measurement was taken (only inhalation, only exhalation or continuously)
   * Aerosol size distribution
3. OVERVIEW OF PROPOSED PROTOCOLS

OSHA's Respiratory Protection Standard, 29 C.F.R. § 134, contains mandatory qualitative and quantitative fit testing procedures in Appendix A, Part I.  In response to stakeholder recommendations for OSHA to consider the adoption of new fit test methods that might be developed at a later date, OSHA included Appendix A, Part II of the standard, which contains procedures that allow individuals to submit new fit test protocols for a notice-and-comment rulemaking under section 6(b)(7) of the OSH Act.

One of the OSHA-accepted QNFT protocols listed in Appendix A is the ambient aerosol CNC QNFT, which typically involves the use of a PortaCount(R) instrument.  On July 10, 2014, TSI Incorporated (TSI) requested that OSHA approve three additional CNC QNFT methods that use the PortaCount(R) instrument to be added to to Part I.C of Appendix A.  The new CNC QNFT protocols are referred to as "Fast-Full method" for elastomeric full-facepiece respirators, "Fast-Half method" for elastomeric half-mask respirators, and "Fast-FFR method" for filtering facepiece respirators.

The experiments used the same materials and methods for all three respirator classes, differing with respect to the respirator class-specific exercise set, and the target ambient particle concentration.   A description of the protocol and methods used for the Fast-Half, Fast-Full, and Fast-FFR methods follows.
0.1 Fast-Half Method

A new Fast-Half fit test method described by Richardson et al. (2014a) used three models of NIOSH-approved, half-facepiece air-purifying respirators from "leading U.S. mask manufacturers" equipped with P100 filters.  Each model was available in three sizes.  Respirators were probed with a flush sampling probe located between the nose and mouth.  Twenty-five participants (9 female; 16 male) were included in the study; face sizes were predominantly in the smaller and central cells (1, 2, 3, 4, 5, 7, 8) of the NIOSH bivariate panel; no subjects were in cells 6, 9 or 10 (those with longer  -  nose to chin  -  face sizes).

Subjects donned the respirator for a 5-minute comfort assessment and then performed two sets of fit test exercises, either the seven standard OSHA exercises (minus the grimace) performed for 60 seconds each [Reference method; see below] or the Fast-Half test exercises for 30 seconds each [Fast-Half method; see below].  The order of the two sets of fit test exercises was randomized. 

The seven OSHA exercises were performed in the order described in the Respiratory Protection Standard, 29 C.F.R. § 134: normal breathing, deep breathing, head side-to-side, head up-and-down, talking, bending over, normal breathing.

The Fast-Half method includes four exercises: bending, jogging in place, head side-to-side and head up-and-down.  [Note: These are the same exercises as in the Fast-Full fit test.]  Two breaths were taken at each extreme of the head side-to-side and head up-and-down exercises and at the bottom of the bend in the bending exercise.

A single CPC instrument, PortaCount Model 8030 (TSI Incorporated, Shoreview MN), was used throughout the experiments.  The instrument was connected to two equal-length sampling tubes for sampling inside-facepiece and ambient particle concentrations.  TSI software was used to switch between sampling lines and record concentration data.  The experiments were conducted in a large chamber to which a NaCl aerosol was added to augment particle concentrations, which were expected to range between 5,000 and 20,000 particles/cm[3] (target = 10,000 p/cm[3]).

During the Reference Method, for each exercise, the ambient sampling tube was first purged for 4 seconds before an ambient sample was taken for 5 seconds, followed by an 11-second purge of the in-facepiece sampling tube and a 40-ssecond in-facepiece sample.  The Reference Method takes a total of 429 seconds (7 minutes 9 seconds) to complete.

During the first exercise of the Fast-Half Method (bending over) the ambient sampling tube was first purged for 4 seconds before an ambient sample was taken for 5 seconds; the in-facepiece sampling tube was then purged for 11 seconds and a sample was then taken from inside the mask for 30 seconds.   No ambient sample was taken during the next two exercises (jogging and head side-to-side)  -  just one 30-second in-facepiece sample was collected for each exercise.  For the last exercise (head up-and-down) a 30-sececond in-facepiece sample was taken, after which a 4-second ambient purge and 5-second ambient sample were conducted.  The Fast-Full Method takes a total of 149 seconds (2 minutes 29 seconds) to complete (80% faster than the Reference Method without the grimace exercise).

For the Reference Method, a fit factor was calculated for each exercise by dividing the in-facepiece concentration taken during that exercise by the mean ambient concentration for that exercise (average of the ambient measurements pre- and post-exercise).  The overall fit factor was determined by taking a harmonic mean of the seven exercise fit factors.

For the Fast-Half Method, the ambient concentration was calculated by taking the mean of two measurements  -  one before the first exercise and one after the last exercise.   Fit factors were calculated for each exercise by dividing the in-facepiece concentration taken during that exercise by the mean ambient concentration.  As with the Reference Method, the harmonic mean of the four exercise fit factors represents the overall fit factor.

To ensure that respirator fit was not significantly altered between the two sets of exercises, a 5-second normal breathing fit factor assessment was included before the first exercise set, between the two sets of exercises and at the completion of the second exercise set.  If the ratio of the maximum to minimum of these three fit factors was greater than 100, this experimental trial was excluded from data analysis.

An exclusion zone  -  the range of Reference fit factors around the expected fit factor of 100 inside which a measured fit factor cannot be confirmed be greater than 100 or less than 100 with confidence  -  was determined by calculating the log difference between 29 pairs of fit factors for Reference Method:Reference Method from 16 participants (6 female; 10 male), and taking the standard deviation, .  The exclusion zone was calculated as 10^(log10(100)  ).  Six pairs of fit factors were omitted because the normal breathing fit factor ratio exceeded 100 and 5 pairs of fit factors were omitted because they were identified as outliers (> 3 standard deviations from the mean of the remaining data points).  The exclusion zone ranged from 82-123.

The final dataset included 132 pairs of fit factors for the Reference Method:Fast-Half Method from 25 participants; equivalent fractions of each respirator and model were included.  Eleven pairs were omitted because the ratio of maximum to minimum normal breathing fit factors was greater than 100; 1 pair was omitted due to a methodological error; 122 pairs were included in the data analysis (see Section 8).

0.2 Fast-Full Method

A new Fast-Full fit test method described by Richardson et al. (2013) used three models of NIOSH-approved, full-facepiece air-purifying respirators from "leading U.S. mask manufacturers" equipped with P100 filters.  Each model was available in three sizes.  Respirators were probed with a non-flush sampling probe inside the nose cup, extending 0.6 into the breathing zone.  Twenty-seven participants (11 female; 16 male) were included in the study; face sizes were predominantly in the central cells (2, 3, 4, 5, 7, 8 and 9) of the NIOSH bivariate panel; 1 subject had a face size in cell 6 and none were in cells 1 (very small) or 10 (very large).

Subjects donned the respirator for a 5-minute comfort assessment and then performed two sets of fit test exercises, either the seven standard OSHA exercises (minus the grimace) performed for 60 seconds each [Reference method; see below] or the fast-fit test exercises for 30 seconds each [Fast-Full method; see below], in random order.

The seven OSHA exercises were performed in the order described in the Respiratory Protection Standard, 29 C.F.R. § 134: normal breathing, deep breathing, head side-to-side, head up-and-down, talking, bending over, normal breathing.

The Fast-Full method includes four exercises: bending, jogging in place, head side-to-side and head up-and-down.  Two breaths were taken at each extreme of the head side-to-side and head up-and-down exercises and at the bottom of the bend in the bending exercise.

A single CPC instrument, PortaCount(R) Model 8030 (TSI Incorporated, Shoreview MN), was used throughout the experiments. The instrument was connected to two equal-length sampling tubes for sampling inside-facepiece and ambient particle concentrations.  TSI software was used to switch between sampling lines and record concentration data.  The experiments were conducted in a large chamber to which a NaCl aerosol was added to augment particle concentrations, which were expected to range between 5,000 and 20,000 particles/cm[3] (target = 10,000 p/cm[3]).

During the Reference Method, for each exercise, the ambient sampling tube was first purged for 4 seconds before an ambient sample was taken for 5 seconds, followed by an 11-second purge of the in-facepiece sampling tube and a 40-second in-facepiece sample.  The Reference Method takes a total of 429 seconds (7 minutes 9 seconds) to complete.

During the first exercise of the Fast-Full Method (bending over) the ambient sampling tube was first purged for 4 seconds before an ambient sample was taken for 5 seconds; the in-facepiece sampling tube was then purged for 11 seconds and a sample was then taken from inside the mask for 30 seconds.   No ambient sample was taken during the next two exercises (jogging and head side-to-side)  -  just one 30-second in-facepiece sample for each exercise.  For the last exercise (head up-and-down) a 30-second in-facepiece sample was taken, after which a 4-second ambient purge and 5-second ambient sample were conducted.  The Fast-Full Method takes a total of 149 seconds (2 minutes 29 seconds) to complete (80% faster than the Reference Method without the grimace exercise).

For the Reference Method, a fit factor was calculated for each exercise by dividing the in-facepiece concentration taken during that exercise by the mean ambient concentration for that exercise (average of the ambient measurements pre- and post-exercise).  The overall fit factor was determined by taking a harmonic mean of the seven exercise fit factors.

For the Fast-Full Method, the mean ambient concentration was calculated by taking the mean of two measurements  -  one before the first exercise and one after the last exercise.   Fit factors were calculated for each exercise by dividing the in-facepiece concentration taken during that exercise by the mean ambient concentration.  As with the Reference Method, the harmonic mean of the four exercise fit factors represents the overall fit factor.

To ensure that respirator fit was not significantly altered between the two sets of exercises, a 5-second normal breathing fit factor assessment was included before the first set of exercises, between the two sets of exercises and at the completion of the second set of exercises.  If the ratio of the maximum to minimum of these three fit factors was greater than 100, this experimental trial point was excluded from data analysis.

An exclusion zone  -  the range of Reference fit factors around the expected fit factor of 500 inside which a measured fit factor cannot be confirmed to be greater than 500 or less than 500 with confidence  -  was determined by calculating the log difference between 29 pairs of fit factors for Reference Method:Reference Method from 17 participants (7 female; 10 male), and taking the standard deviation, .  The exclusion zone was calculated as 10^(log10(100)  ). Five pairs of fit factors were omitted because the normal breathing fit factor ratio exceeded 100, and 3 pairs of fit factors were omitted because they were identified as outliers (> 3 standard deviations from the mean of the remaining data points).  The exclusion zone ranged from 345-726.

The final dataset included 148 pairs of Reference:Fast-Full fit factors from 27 participants; equivalent fractions of each respirator and model were included.  Eleven pairs were omitted because the ratio of maximum to minimum normal breathing fit factors was greater than 100; 1 pair was omitted due to an observational anomaly; 136 pairs were included in the data analysis (see Section 9).

0.3 Fast-FFR Method

A new Fast-Filtering Facepiece Respirator (Fast-FFR) fit test method described by Richardson et al. (2014b) used ten models of NIOSH-approved N95 FFRs from six "leading U.S. mask manufacturers."  The different models were selected to represent a range of styles  -  6 cup-shaped, 2 horizontal flat-fold and 2 vertical flat-fold models. Based on the respirator coding in the supplied data, it appears that each respirator model was available in one size; and different sizes of the same respirator design had unique model numbers (e.g., Kimberly Clark 46727 and 46867 appear to be horizontal fold FFR of size R and S, respectively).  No information was provided in the publication about whether models were available in different sizes.

Respirators were probed with a flush sampling probe located between the nose and mouth.  Lightweight sample tubing and neck straps were used to ensure the tubing did not interfere with respirator fit.

Twenty-nine participants (11 female; 18 male) were included in the study; face sizes were predominantly in the smaller and central cells (1, 2, 3, 4, 5, 7, 8) of the NIOSH bivariate panel; 1 subject was in cell 6 and no subjects were in cells 9 or 10 (those with longer  -  nose to chin  -  face sizes).

Subjects donned the respirator for a 5-minute comfort assessment and then performed two sets of fit test exercises, either the seven standard OSHA exercises (minus the grimace) performed for 60 seconds each [Reference method; see below] or the Fast-FFR test exercises for 30 seconds each [Fast-FFR method; see below], in random order.

The seven OSHA exercises were performed in the order described in the Respiratory Protection Standard, 29 C.F.R. § 134: normal breathing, deep breathing, head side-to-side, head up-and-down, talking, bending over, normal breathing.

The Fast-FFR method includes four exercises: bending, talking, head side-to-side and head up-and-down.  [Note: The talking exercise replaces the jogging exercise used in the Fast-Full and Fast-Half methods.]  Two breaths were taken at each extreme of the head side-to-side and head up-and-down exercises and at the bottom of the bend in the bending exercise.

A single CPC instrument, PortaCount Model 8038 operated in the N95 Mode (TSI Incorporated, Shoreview MN), was used throughout the experiments. The instrument was connected to two equal-length sampling tubes for sampling inside-facepiece and ambient particle concentrations.  TSI software was used to switch between sampling lines and record concentration data.  The experiments were conducted in a large chamber to which a NaCl aerosol was added to augment particle concentrations, which were expected to be greater than 400 p/cm[3].

During the Reference Method, for each exercise, the ambient sampling tube was first purged for 4 seconds before an ambient sample was taken for 5 seconds, followed by an 11-second purge of the in-facepiece sampling tube and a 40-second in-facepiece sample.  The Reference Method takes a total of 429 seconds (7 minute 9 second) to complete.

During the first exercise of the Fast-FFR Method (bending over) the ambient sampling tube was first purged for 4 seconds before an ambient sample was taken for 5 seconds; the in-facepiece sampling tube was then purged for 11 seconds and a sample was then taken from inside the mask for 30 seconds.  No ambient sample was taken during the next two exercises (jogging and head side-to-side)  -  just one 30-seccond in-facepiece sample for each exercise.  For the last exercise (head up-and-down) a 30-second in-facepiece sample was taken, after which a 4-second ambient purge and 5-second ambient sample were conducted.  The Fast-FFR Method takes a total of 149 seconds (2 minutes 29 seconds) to complete (80% faster than the Reference Method without the grimace exercise).

For the Reference Method, a fit factor was calculated for each exercise by dividing the in-facepiece concentration taken during that exercise by the mean ambient concentration for that exercise (average of the ambient measurements pre- and post-exercise).  The overall fit factor was determined by taking a harmonic mean of the seven exercise fit factors.

For the Fast-FFR Method, the mean ambient concentration was calculated by taking the mean of the two measurements taken before the first exercise and after the last exercise.  Fit factors were calculated for each exercise by dividing the in-facepiece concentration taken during that exercise by the mean ambient concentration.  As with the Reference Method, the harmonic mean of the four exercise fit factors represents the overall fit factor.

To ensure that respirator fit was not significantly altered between the two sets of exercises, a 5-second normal breathing fit factor assessment was included before the exercises, between the two sets of exercises and at the completion of the two sets.  If the ratio of the maximum to minimum of these three fit factors was greater than 100, this data point was excluded from data analysis.

An exclusion zone  -  the range of Reference fit factors around the expected fit factor of 100
within which a measured fit factor cannot be confirmed to be greater than 100 or less than 100 with confidence  -  was determined by calculating the log difference between 63 pairs of Reference Method:Reference Method data from 14 participants (5 female; 9 male), and taking the standard deviation, .  The exclusion zone was calculated as 10^(log10(100)  ). Two pairs of fit factors were omitted because the normal breathing fit factor ratio exceeded 100, and 6 pairs of fit factors were omitted because they were identified as outliers (> 3 standard deviations from the mean of the remaining data points).  The exclusion zone ranged from 78-128.

The final dataset included 114 pairs of Reference:Fast-FFR fit factors from 29 participants; equivalent fractions of each respirator and model were included.  Two pairs were omitted because the ratio of maximum to minimum normal breathing fit factors was greater than 100; 112 pairs were included in the data analysis (see Section 10).
1. DISCUSSION OF SELECTED EXERCISES

The OSHA Respiratory Protection Standard requires that all QNFT methods  -  except the Controlled Negative Pressure Method (see below)  -  utilize the same set of 8 exercises for the same time periods:
   1. Normal breathing (1 minute)  -  normal standing position, no talking
   2. Deep breathing (1 minute)  -  breathe slowly and deeply
   3. Turning head side-to-side (1 minute)  -  stand and turn head slowly, hold head briefly and inhale at each extreme
   4. Head up-and-down (1 minute)  -  stand and move head up and down, inhale at the up position
   5. Talking (1 minute)  -  talk slowly out loud
   6. Grimace (15 seconds)  -  smile or frown
   7. Bending over (1 minute)  -  bend at waist as if to touch toes.  May be replaced by jogging if fit test equipment does not allow bending.
   8. Normal breathing (1 minute)  -  normal standing position, no talking
While the writers of the original OSHA Respiratory Protection Standard could easily imagine the development of new respirator fit test methods, it is clear they never expected that such methods might involve fewer exercises.  However, the Standard neither precludes nor prohibits this possibility.
In fact, there is some precedent for allowing deviations from the required exercises in the standard, as the Controlled Negative Pressure (CNP) fit test method has a separate and different set of exercises from all other fit test methods.  The main differences between the CNP method and standard protocol are:
   * Because the CNP method requires the subject to hold his/her breath for 10 seconds while fit is measured, exercises are performed for 1 minute (as required by the standard protocol) followed by 10 seconds of breath holding.  Some exercises (see below) require two 10-second measurements.
   * The head side-to-side and head up-and down exercises require measures of fit during 10 seconds of breath holding for each head position (right/left and up/down).
   * No measurement of fit is made immediately after the 15-second grimace exercise. 
   * After the bending over exercise, the subject removes and re-dons the respirator within a 1-minute period and then holds his/her breath for 10 seconds during fit measurement.
In addition, in 2004 OSHA accepted a new CNP REDON Fit Test Method that uses only three exercises  -  normal breathing (30 seconds), bending over (30 seconds) and head shaking while shouting (3 seconds)  -  each followed by a 10-second breath holding leak measurement period (OSHA, 2004).  In this test, the subject dons the respirator three separate times, removing the respirator completely and loosening the facepiece straps each time.  OSHA received comments that this shorter set eliminated important exercises known to detect leaks  -  head side-to-side, head up-and-down and talking.  OSHA concluded that the proposed CNP REDON method was both valid and accurate, however, because fit factors measured with the new protocol were always lower (more conservative) than those of comparative methods.
For the purposes of comparing one fit test method to another, we agree with TSI that the grimace exercise introduces an unpredictable source of variability.  Seitsema et al. (2015) found that, while the grimace dislodges the respirator from the face, the manner in which the respirator re-seats is not consistent from one grimace to the next.  Thus, including the grimace complicates the comparison of one fit test method to another, even if test order is randomized.
We agree with Zhuang et al. (2004) that fit test methods using shorter or fewer exercises should use more aggressive exercises to ensure they are representative of head motions that may cause the fit to change while the respirator is worn during work activities.  And we agree that bending over, moving the head up-and-down or side-to-side, and jogging represent more aggressive types of exercises.  We are not convinced that talking represents a rigorous exercise, however.  Nor are we convinced that filtering facepiece respirators should be tested with a less rigorous set of exercises than elastomeric respirators.  
Richardson et al. (2013; 2014a; 2014b) discuss the selection of the most strenuous exercises for each of the new Fast Fit methods, based on recommendations in Zhuang et al. (2004) that talking, bending, and moving head up-and-down should be used when shorter or fewer exercises are selected.  Richardson et al. (2013; 2014a) indicate that they selected bending and head up-and-down based on these recommendations and added head side-to-side and jogging, as well, for their effect on respirator fit, but do not provide any data to support the latter two exercises.  Richardson et al. (2014b) note they selected the same three exercises recommended by Zhuang et al. (2004) for the Fast-FFR method, but again do not provide any data supporting the selection of the head side-to-side exercise.
Additional data provided by TSI Incorporated (TSI, 2015a) describe the process for identifying fit test exercises for each respirator.  TSI collected information about the most rigorous exercises from published literature (Zhuang et al., 2004) and undocumented anecdotal input from experienced industry fit test experts.  They also conducted a series of in-house pilot tests with naïve subjects performing two consecutive sets of seven OSHA exercises (without the grimace).  Methods are described in a White Paper (TSI, 2014).  They examined the percent of times an exercise yielded the lowest fit factor for all fit tests, fit tests for poor-fitting respirators (< pass/fail criterion) and fit tests for good-fitting respirators (>= pass/fail criterion).
Tables from the TSI White Paper with the results for each respirator are reproduced below.  Table FF1 shows that the bending over, head side-to-side, head up-and-down and second normal breathing exercises most often result in the lowest fit factors (44, 19, 16 and 19% of tests, respectively) for poor-fitting full-facepiece respirators. The first normal breathing, deep breathing and talking exercises were the least likely to yield the lowest fit factors (0, 0, and 3% of poor-fitting tests, respectively).
Table HF1 shows that bending over had the lowest fit factor most often (19% of the time) for poor-fitting elastomeric half facepiece respirators; the remaining exercises were all similar in their frequency of poor fit (14% of the time) with the exception of talking (10% of the time).
Table FFR1 shows that bending over most frequently produced the lowest fit factor (37% of the time) for poor-fitting N95 FFRs.  Head side-to-side and head up-and-down were the second most likely to have low fit factors (14% of poor fitting respirators).  Talking and deep breathing were the least likely to result in poor fits (2% of the time).
These data suggest that normal breathing and talking exercises are the least likely to yield useful information about fit performance when a respirator does not fit well.  This is not surprising, as these exercises are the least likely to impact the placement of the respirator facepiece on the face.  




The TSI rationale for selecting the bending over, head side-to-side and head up-and-down exercises for the Fast-Full test is supported by the TSI pilot study results.  A jogging exercise was added because an additional rigorous exercise was thought to be necessary; as well, jogging is listed as an alternative exercise in the, Annex A2  -  Criteria for Evaluating New Fit Test Methods and OSHA 29 CFR 1910.134 standards.

The same rationale was applied to the selection of exercises for the Fast-Half test, although only the bending over exercise appeared to be more rigorous than any of the other exercises in the TSI pilot test results with poor-fitting half-facepiece respirators.  TSI suggests that there are no data indicating that half-facepiece respirator fit differs from that of full-facepiece respirators.

Finally, TSI decided not to include the jogging exercise for the Fast-FFR test, because workers are generally not performing strenuous activities while wearing N95 FFRs.  Instead, they decided to include talking because this is a more common activity for these wearers.

We believe their selection of the four exercises for the Fast-Full test is supported by both data and practice; the TSI rationale is appropriate.  We believe the selection of the same four exercises for the Fast-Half test by analogy to the Fast-Full test is reasonable and supported by the same data and practice information.  Elastomeric respirators in both configurations do share many similar design features that impact facepiece fit.  Full-facepiece respirators have fewer leak points because the forehead is an easier facial feature to fit than the nose.  Thus, a set of exercises considered rigorous for a full-facepiece elastomeric respirator should also be rigorous for a half-facepiece respirator.

On the other hand, we are less convinced by the rationale for the Fast-FFR exercises.  The talking and deep breathing exercises were not as likely to produce low fits as either the first or second normal breathing exercises in their pilot tests with N95 FFRs.  While it is true that N95 FFRs are worn in many low-workrate jobs, they are also worn in more demanding job settings with high workrates.  The selection of a FFR is not governed by workrate or type of job as much as it is by the level of exposure.  There are many demanding jobs (e.g. construction) where an N95 FFR might be appropriate for protecting workers from relatively low dust exposure levels.  The fact that people talk more while wearing an N95 FFR rather than an elastomeric half-facepiece respirator has little to do with whether talking is a rigorous exercise for examining fit.  The data suggest it is not.  The TSI White Paper does not support the selection of talking as an exercise for the Fast-FFR test.
2. DISCUSSION OF RESPIRATORS SELECTED

Richardson et al. (2013; 2014a) describe their selection of full- and half-facepiece respirators as those from "leading U.S. mask manufacturers," but do not explain how these models are representative of different designs or features commonly encountered for these types of respirators.  Elastomeric respirators can differ in flange design, material, strap design, and, in the case of full-facepiece respirators, nose cup design.  Any or all of these can play a significant role in respirator fit.
It was probably important to identify respirator designs that would yield consistent miss-fits, to ensure an adequate number of fit factors below the pass/fail criterion.  This may have played a role in which respirator models were selected, although this point was not raised in any of the publications.  Certain designs (e.g. straps or materials) may make it easier to establish a poor but consistent fit than others.  We know of no published studies, however, where this has been explored in any detail.
Based on our knowledge of N95 FFRs, we believe that TSI identified the most common designs.  Richardson et al. (2014b) mention selecting respirators with different sizes, but provide no information about this feature of the selected respirator.  It appears the N95 FFR size was conflated with N95 FFR model (e.g., each N95 FFR model came in one size).
The rationale for selecting the different respirators for each method, and whether the ability to establish a poor fit played a role in this process, could be better explained.
3. REVIEW OF FAST FIT TEST METHODOLOGY
0.1 Normal Breathing Fit Factor Criterion

TSI developed the normal breathing fit factor criterion to identify pairs of sequential fit tests in which the respirator fit was stable or unstable. This criterion is not specified in the ANSI/AIHA Z88.10  -  2010 standard.  TSI suggests that the ANSI approach of comparing the two fit testing methods sequentially during the same donning implies that the baseline quality of the respirator fit is stable for the duration of the experiment, and propose to use the normal breathing fit factor criterion to evaluate this condition.
The normal breathing fit factor criterion requires that the ratio of the maximum and minimum fit factors obtained during normal breathing  -  measured prior to the first fit test protocol, between protocols and after the second fit test protocol  -  must not exceed 100.
We believe this is an appropriate method for assessing the stability of fit from one repetition of a fit test to the next.  That fit will not change in some radical fashion between the two tests is an important underlying assumption of the experimental protocol.  We can think of no other approach to test for and exclude data points where fit was not consistent.
0.2 Definition of the Exclusion Zone

An exclusion zone must be defined around the reference fit factor, FFR, within which there is no statistical certainty that the fit factor measured in an experimental trial is greater than or less than the FFR.  ANSI/AIHA Z88.10-2010 describes the use of a coefficient of variation, CV, to define the exclusion zone.  The CV is a relative standard deviation expressed as the ratio of the standard deviation over the mean value, CV = /. ANSI does not specify the mathematical expression for calculation of the exclusion zone with the CV. 

The ANSI standard recommends defining the CV by measuring the fit factor on one participant multiple times during a single donning. This approach captures the variability arising from the PortaCount(R) instrument and within-person variability in exercise performance.  TSI used an alternative approach that took advantage of the fact that one participant may have participated in multiple experimental trials.  Thus, one participant may have had overall fit factors measured two or more times by the Reference Method.  This approach captures the variability arising from the PortaCount(R), within-person variability in exercise performance, and from respirator donning. 
It would be expected that adding respirator donning would increase variability.  The change in respirator fit introduced by the grimace exercise, however, is unpredictable and may be larger or smaller than the variability in respirator fit introduced by respirator donning.  If the grimace exercise was not performed in the sequential measurement of respirator fit used to determine the CV, then the approach of using overall fit factors from two respirator donnings would be expected to produce a higher value of CV.  If the grimace exercise was performed, the relative magnitude of the CVs determined by the two approaches is unknown.
TSI determined the exclusion zone as follows.  For participant i = {1, 2,..., n} with two overall fit factors, FFi,1 and FFi,2, the difference in the logarithms of the fit factors was calculated, Di = log10(FFi,1)  -  log10(FFi,2). Note, if FFi,1 = FFi,2, then Di = 0; and the paired differences should be approximately normally distributed. The standard deviation of the D1, D2, ..., Dn values was calculated, and denoted . The upper and lower bounds of the exclusion zone were equated with 10^(log10(100)  ).  This calculation method is appropriate.
The value of  was determined by TSI after exclusion of pairs of fit factors owing to failure of the normal breathing fit factor criterion or when Di was judged to be an outlier. Exclusion of pairs of fit factors owing to failure of the normal breathing fit factor criterion is appropriate prior to calculation of .  Outliers were defined by TSI as having Di greater than three standard deviations from the mean from the remaining data points.  While it is appropriate to identify outliers as being greater than three standard deviations from the mean, the standard deviation used to identify outliers is typically calculated with the outlier points included. After the outliers are excluded, the standard deviation can be recalculated for use in statistical analysis. The latter approach would increase the magnitude of , and increase the width of the exclusion zone.  We repeated the calculation of the Exclusion Zone and repeated the evaluation of the Fast-Full, Fast-Half and Fast-FFR methods using the latter approach (see Sections 8-10).

0.3 Potential Sources of Bias 

ANSI Z88.10-2010 recommends consideration of factors that might influence sampling bias of the Generated Aerosol Quantitative Fit Test Procedure.  Topics include: location and depth of probe, position of face-seal leaks, interaction of breathing pattern with position of face-seal leak, design of the facepiece, measurement sample rate, when the measurement was taken (only inhalation, only exhalation or continuously), and the aerosol size distribution.  Though the new protocols do not use generated aerosol, we identified information provided by TSI that is relevant to some of these considerations.
Regarding location and depth of probe, internally flush sample probes were used for FFRs, but placement varied by respirator model and is specified in the manuscript.  For elastomeric half-facepiece respirators, flush probes were placed between the nose and mouth. For full-facepiece respirators, non-flush probes were placed between the nose and mouth, and extended about 0.6 cm into the breathing zone.
Artificial leak paths were not used, so TSI could not describe the positions of face-seal leaks.
Participants were instructed to take two breaths at each extreme during the head side-to-side, at the bottom of the bend during the bending exercise, and when looking up during the head up-and-down exercise.  At other times in the exercise participants could choose when to take breaths.  TSI explained that taking two breaths at the extremes of the head side-to-side and head up-and-down exercises was expected to "increase the rigor of the exercises and to make sure everyone does them exactly the same."  Because leaks will occur only during inhalation and are more likely when the head is located at the extreme point, this protocol was expected to ensure that leaks would be detected for these exercises.  
In-mask measurements of particle concentrations were measured continuously during the exercise, but the frequency of measurement within the exercise was not specified. There was no specific coordination between in-mask sample collection, though the repeated breaths at the extremes of the movement likely ensured measurement of the in-mask particle concentration at the extremes of the movement.
There is some concern, however, that requiring two breaths at the extremes of head movements limits the number of movements undertaken in a 30-second period.   
The aerosol size distribution was not specified.
1. REVIEW OF ANSI/AIHA z88.10-2010 ANNEX A2 CRITERIA FOR HALF-FACEPIECE RESPIRATOR  -  INCLUDING REPRODUCTION OF DATA ANALYSIS

The evaluation of the protocol follows the numerical sequence of the ANSI/AIHA Z88.10-2010 criteria outlined in Section 3.
   1.    Participants performed the new and reference fit test methods in succession without pause and without adjustment of the respirator.  Randomization was indicated in the revised materials provided by TSI.  The reference method was tested first in approximately half of the experimental trials, indicating effective randomization. 
   
   2.    A detailed description of the new fit test method/protocol is documented.
   
   The new fit test method involves four exercises: bending, jogging, head side-to-side and head up-and-down.  Each exercise is performed for 30 seconds.  During the test, participants take 2 breaths at each extreme during the head side-to-side, at the bottom of the bend during the bending exercise, and when looking up during the head up-and-down exercise.  At other times in the exercise participants could choose to take breaths.
   
   Ambient aerosol is sampled by the PortaCount(R) instrument at the start and end of the fit test only.  In-mask aerosol is sampled by the PortaCount(R) instrument for 30 seconds during each exercise (the duration of the exercise).
   
   The overall fit factor is the harmonic mean of fit factors measured in each exercise.
   
   3.    A detailed description of the study materials and methods used is documented.  The materials and methods are described in Section 4.1.
   
   For the Fast-Half method, the target ambient particle concentration was 10,000 p/cm[3] (range 5,000 -20,000 p/cm[3]). In the experimental trials, the mean ambient particle concentrations were 9,073-41,916 p/cm[3], with overall mean 20,722 p/cm[3].  For the reference method, the mean ambient particle concentrations were 9,304-37,520 p/cm[3], with overall mean 20,889 p/cm[3], based on the first and last ambient particle concentration measurement in the protocol.  The ambient particle concentrations during the experiments tended to be greater than the target range, indicated by the overall mean values being near the high end of the target range.  
   
   A paired t-test indicated that there was no evidence to reject the null hypothesis that there was no difference between the mean ambient particle concentrations during the paired Fast-Half and reference methods (p-value = 0.43).
   
   The magnitude of variability in the ambient particle concentration in each protocol, indicated by the CV in the initial and final ambient particle concentration measurements for each protocol, was similar.  The mean coefficient of variation was 8.1% and 6.5% for the reference and Fast-Half methods, respectively.

   4.    Three models of NIOSH-approved elastomeric half-facepiece respirators were used. The specific models were:
         *       3M 6000
         *       North 5500
         *       Moldex 700
   Three sizes were tested for each respirator.  These respirators represent a variety of different sizes and models and seem sufficient to achieve a variety of different airflow and leak patterns.  The respirators also reflect commonly used brands and models. We cannot evaluate what proportion of the market share these respirators occupy.
   5.    TSI used pairs of fit tests performed on the same person with the same respirator model in two different experimental trials to calculate the CV for the reference fit test method, and define the exclusion zone. As described in Section 7.2, we have some concern with the method used to calculate the standard deviation.  
   
   Fifty-nine pairs of fit tests were available, but 11 were excluded by TSI, leaving 48 pairs of fit tests.  Six pairs were excluded because the normal breathing fit factor criterion. Five pairs were excluded because they were identified as outliers. The pairs of fit tests identified as outliers, with the difference in the logarithms of the fit factors, were:
   * FT01102013105341, D = -0.567
   * FT01102013102805, D = 0.655
   * FT24092013103845, D = 1.020
   * FT18092013141747, D =  -0.335
   * FT04092013134313, D = 0.390
   Based on this analysis, TSI defined the fit factor exclusion zone to be 82-123.  
   For the 48 pairs of fit-tests, the mean value of the difference was -0.0001 and  = 0.09, so three standard deviations span (-0.27, 0.27).  All of the outliers fall outside the three standard deviation range.
   For the 53 pairs of fit-test (excluding only six pairs for the normal breathing fit factor criterion), the mean value of the difference was 0.022 and  = 0.215, so three standard deviations span (-0.62, 0.67).  The differences for experimental trials FT01102013105341, FT01102013102805, FT18092013141747, and FT04092013134313 fall within this range; only FT2409201310384 is an outlier by this definition. Based on 52 pairs of fit tests,  = 0.16, which gives an exclusion zone of 68-146.
   The pairs of fit tests used to define the exclusion zone included all respirator models in all sizes. 
   6.    The exclusion range defined by TSI for the reference method was fit factors between 82-123.  A total of six tests had fit factors for the reference method in this range. These experimental trials, with overall fit factor, include:
   * FT15102013085836, FF = 83
   * FT15102013145433, FF = 108
   * FT16102013123954, FF = 111
   * FT16102013091742, FF = 114
   * FT09102013132229, FF = 115
   * FT23102013095103, FF =120
   With the wider exclusion zone, fit factors 69-144, three additional pairs of fit tests would be excluded: FT16102013104537 (FF =71), FT25102013082635 (FF = 127), and FT15102013081303 (FF = 135) 
   7.    Two exclusion criteria were applied: 1) the exclusion range for the fit factor of the reference method excluded six paired tests, and 2) the normal breathing fit factor criterion excluded one additional paired test.  After exclusions by TSI there were 122 paired tests from 25 different subjects. Using the wider exclusion zone, there are 119 paired tests from 25 different subjects. These meet the minimum of 100 sequential paired tests from at least 25 different subjects.
   
   The required fit factor is 100, and 5% of the required fit factor is 5.  
   
   The lower bound of the exclusion zone defined by TSI was 82.  Fifty-one of 116 pairs had reference fit factors  5 and less than 82.  Forty-nine of 116 pairs had reference method fit factors > 5 and less than 82.  The language in ANSI/AIHA Z88.10-2010 suggests that latter criteria (> 5 and < 82) should be applied, in which case the experiment does not meet this criterion as 49 pairs, not 50 pairs, fall in this range. TSI, however, appears to have applied the other criterion ( 5 and < 82), which is met (see Table VI in Richardson et al. (2014a)).
   
   The lower bound of the wider exclusion zone is 69. Fifty of 109 pairs had reference fit factors  5 and less than 69.  Forty-eight of 109 pairs had reference method fit factors > 5 and less than 69.  The language in ANSI/AIHA Z88.10-2010 suggests that latter criteria (> 5 and < 69) should be applied, in which case the experiment does not meet this criterion as 48 pairs, not 50 pairs, fall in this range. 
   
   8.    When the new fit test method identified an appropriate fit (e.g., fit factor >= 100), the reference fit factors were 10.  Experimental trial FT15102013094915 had reference fit factor equal to 10, though the new fit test method had fit factor 106. No reference fit factor was measured to be less than 10% of the required fit factor (e.g., < 10) when the fit was determined to be acceptable by the new fit test method. The method is not disqualified on this criterion.
   
   9.    Figure 4 of Richardson (2014a) demonstrates that the fit factors less than the required fit factors are evenly distributed, and are not primarily lower fit factors.

   10.    The two fit test methods were compared over a range a measurements that include the target fit factor for FFRs, FF  = 100, which is the intended target of the method. 

   11.    Figure 4 of Richardson (2014a) shows the cumulative distribution of the fit factors obtained for the reference method, which has fit factors ranging from 2 to 10,000.

   12.    Information about the respiratory make, model, style, size, individuals tested and paired results of the new test and the reference test were provided by TSI.  Similar information was provided for any test results excluded from analysis, along with reasons for exclusion.

   13.    The test results provided by TSI included the components of a 2  2 contingency table: A = 2, B = 58, C = 52, D = 4.  All calculations were accurate and align with the ANSI/AHIA Z88.10-2010 criteria (Table 1).  With the wider exclusion zone, A = 2, B = 56, C = 51 and D = 4, and the new fast fit test protocol still meets all of the criteria (Table 1).


                 Table 1. Performance of the Fast-Half method.

                                     ANSI
                                   Criterion
                                      TSI
                                  Calculation
                                     Wider
                                Exclusion Zone
Test Sensitivity
                                   >= 0.95
                                     0.96
                                     0.96
Predictive Value of Pass
                                   >= 0.95
                                     0.97
                                     0.97
Test Specificity
                                   >= 0.50
                                     0.97
                                     0.97
Predictive Value of Fail
                                   >= 0.50
                                     0.93
                                     0.93
Kappa Statistic
                                   >= 0.70
                                     0.90
                                     0.89

2. REVIEW OF ANSI/AIHA Z88.10-2010 ANNEX A2 CRITERIA FOR FULL-FACEPIECE RESPIRATOR  -  INCLUDING REPRODUCTION OF DATA ANALYSIS

Evaluation of the protocol follows the numerical sequence of the ANSI/AIHA Z88.10-2010 criteria outlined in Section 3.
   1.    Participants performed the new and reference fit test methods in succession without pause and without adjustment of the respirator.  Randomization was indicated in the revised materials provided by TSI.  The reference method was tested first in approximately half of the experimental trials, indicating effective randomization.
   
   2.    A detailed description of the new fit test method/protocol is documented.
   The protocol is the same as used for elastomeric half-facepiece respirators. 
   
   3.    A detailed description of the study materials and methods used is documented.  The materials and methods are described in Section 4.2.

   For the Fast-Full method, the target ambient particle concentration was 10,000 p/cm[3] (range 5,000 -20,000 p/cm[3]). In each experimental trial, the mean ambient particle concentrations were 11,268-36,604 p/cm[3], with overall mean 21,586 p/cm[3].  For the reference method, the mean ambient particle concentrations were 11,494-36,730 p/cm[3], with overall mean 21,741 p/cm[3], based on the first and last ambient particle concentration measurement in the protocol.  The ambient particle concentrations during the experiments tended to be greater than the target range, indicated by the overall mean values being near the high end of the target range. 
   
   A paired t-test indicated that there was no evidence to reject the null hypothesis that there was no difference between the mean ambient particle concentrations during the paired Fast-Full and reference protocols (p-value = 0.42).
   
   The magnitude of variability in the ambient particle concentration in each protocol, indicated by the coefficient in variation in the initial and final ambient particle concentration measurements for each protocol, was similar.  The mean coefficient of variation was 7.4% and 7.9% for the reference and Fast-Full methods, respectively.
   
   4.    Three models of NIOSH-approved elastomeric full-facepiece respirators were used. The specific models were:
         *       3M 6000
         *       Scott AV2000
         *       Moldex 9000
   Three sizes were tested for each respirator.  These respirators represent a variety of different sizes and models and seem sufficient to achieve a variety of different airflow and leak patterns.  The respirators also reflect commonly used brands and models. We cannot evaluate what proportion of the market share these respirators occupy. 
   5.    TSI used pairs of fit tests performed on the same person with the same respirator model in two different experimental trials to calculate the CV for the reference fit test method, and define the exclusion zone. As described in Section 7.2, we have some concern with the method used to calculate the standard deviation.  
   
   Sixty-two pairs of fit tests were available, but 8 were excluded by TSI, leaving 54 pairs of fit tests.  Five pairs were excluded because the normal breathing fit factor criterion. Three pairs were excluded because they were identified as outliers. The pairs of fit tests, and difference in the logarithm of the fit factors, identified as outliers were:
   * FT09072013141424, D = 0.648
   * FT24072013090908, D = -1.32
   * FT24072013084154, D = -0.500
   Based on this analysis, TSI defined the fit factor exclusion zone to be 344.6-725.5.  
   For the 54 pairs of fit-tests, the mean value of the difference was 0.026 and  = 0.16, so three standard deviations span (-0.45, 0.51).  All of the outliers fall outside the three standard deviation range.
   For the 57 pairs of fit-test (excluding only five pairs for the normal breathing fit factor criterion), the mean value of the difference was 0.004 and  = 0.26, so three standard deviations span (-0.78, 0.79).  The differences for experimental trails FT09072013141424 and FT24072013084154 fall within this range; only FT24072013090908 is an outlier by this definition. Based on 59 pairs of fit tests,  = 0.19, which gives an exclusion zone of 320.6-779.7.
   The pairs of fit tests included all respirator models in each of the three sizes.
   6.    The exclusion range defined by TSI for the reference method was fit factors between 344.6-725.5.  A total of twelve tests had fit factors for the reference method in this range. These experimental trials, with the observed fit factors, include:
   * FT27082013103356, FF = 352
   * FT31072013103331, FF = 355
   * FT31072013121048, FF = 365
   * FT06082013142411, FF = 388
   * FT01082013091359, FF = 419
   * FT20082013134438, FF = 442
   * FT20082013142756, FF = 450
   * FT31072013095052, FF = 474
   * FT06082013102257, FF = 498
   * FT06082013095906, FF = 568
   * FT14082013094147, FF = 603
   * FT07082013101524, FF = 653
   With the wider exclusion zone, fit factors 320.6-779.7, three additional pairs of fit tests would be excluded: FT07082013083847 (FF = 338), FT14082013125055 (FF = 341), and FT20082013092904 (FF = 341).
   7.    Two exclusion criteria were applied: 1) the exclusion range for the fit factor of the reference method excluded six paired tests, and 2) the normal breathing fit factor criterion excluded one additional paired test.  After exclusions by TSI there were 136 paired tests from 27 different subjects. Using the wider exclusion zone, there are 133 paired tests from 27 different subjects. These meet the minimum of 100 sequential paired tests from at least 25 different subjects.
   
   The required fit factor is 500, and 5% of the required fit factor is 25.  
   
   The lower bound of the exclusion zone defined by TSI was 344.6.  Fifty-four of 116 pairs had reference fit factors  25 and < 344.6.  Fifty-two of 136 pairs had reference method fit factors > 25 and < 344.6.  These sample sizes meet the ANSI/AIHA Z88.10-2010 criterion.
   
   The lower bound of the wider exclusion zone is 320.6. Fifty-one of 133 pairs had reference fit factors  25 and < 320.6.  Forty-nine of 133 pairs had reference method fit factors > 25 and < 320.6.  The language in ANSI/AIHA Z88.10-2010 suggests that latter criteria (> 25 and < 320.6) should be applied, in which case the experiment does not meet this criterion as it has 49, not 50, pairs in this range. 
   
   8.    When the Fast-Full test method identified an appropriate fit (e.g., fit factor >= 500), the reference fit factors were  416.  No reference fit factor was measured to be less than 10% of the required fit factor (e.g., < 50) when the fit was determined to be acceptable by the new fit test method. The method is not disqualified on this criterion.
   
   9.    Figure 4 of Richardson et al. (2013) demonstrates that the fit factors less than the required fit factors are evenly distributed, and are not primarily lower fit factors.

   10.    The two fit test methods were compared over a range a measurements that include the target fit factor for FFRs, FF  = 500, which is the intended target of the method. 

   11.    Figure 4 of Richardson et al. (2013) shows the cumulative distribution of the fit factors obtained for the reference method, which has fit factors ranging from 10 to 70,000.
   
   12.    Information about the respiratory make, model, style, size, individuals tested and paired results of the new test and the reference test were provided by TSI.  Similar information was provided for any test results excluded from analysis, along with reasons for exclusion.

   13.    The test results provided by TSI included the components of a 2  2 contingency table: A = 1, B = 64, C = 58, D = 1.  All calculations were accurate and align with the ANSI/AIHA Z88.10-2010 criteria (Table 2).  With the wider exclusion zone, A = 1, B = 64, C = 55 and D = 1, and the new fast fit test protocol meets all of the criteria (Table 2).

                 Table 2. Performance of the Fast-Full method.

                                     ANSI
                                   Criterion
                                      TSI
                                  Calculation
                                     Wider
                                Exclusion Zone
Test Sensitivity
                                   >= 0.95
                                     0.98
                                     0.98
Predictive Value of Pass
                                   >= 0.95
                                     0.98
                                     0.98
Test Specificity
                                   >= 0.50
                                     0.98
                                     0.98
Predictive Value of Fail
                                   >= 0.50
                                     0.98
                                     0.98
Kappa Statistic
                                   >= 0.70
                                     0.97
                                     0.97

3. REVIEW OF ANSI/AIHA z88.10-2010 ANNEX A2 CRITERIA FOR FFR  -  INCLUDING REPRODUCTION OF DATA ANALYSIS

Evaluation of the protocol follows the numerical sequence of the ANSI/AIHA Z88.10-2010 criteria outlined in Section 3.
   1.    Participants performed the new and reference fit test methods in succession without pause and without adjustment of the respirator.  Randomization was indicated in the revised materials provided by TSI.  The reference method was tested first in approximately half of the experimental trials, indicating effective randomization.
   
   2.    A detailed description of the new fit test method/protocol is documented.

   The new fit test method involves four exercises: bending, talking, head side-to-side and head up-and-down.  Each exercise is performed for 30 seconds.  During the test, participants take two breaths at each extreme during the head side-to-side, at the bottom of the bend during the bending exercise, and when looking up during the head up-and-down exercise.  At other times in the exercise participants could choose to take breaths.
   
   Ambient aerosol is sampled by the PortaCount(R) instrument at the start and end of the fit test only.  In-mask aerosol is sampled by the PortaCount(R) instrument for 30 seconds during each exercise (the duration of the exercise).
   
   The overall fit factor is the harmonic mean of fit factors measured in each exercise.
   
   3.    A detailed description of the study materials and methods used is documented.  The materials and methods are described in Section 4.3.
   
   For the Fast-FFR method, the target ambient particle concentration expected to be greater than 400 p/cm[3]. In experimental trials, the mean ambient particle concentrations were 359-1,839 p/cm[3], with overall mean 808 p/cm[3].  For the reference method, the mean ambient particle concentrations were 415-1,991 p/cm[3], with overall mean 901 p/cm[3], based on the first and last ambient particle concentration measurement in the protocol.  The ambient particle concentrations during the experiments were in the target range. 
   
   A paired t-test indicated that there was evidence to reject the null hypothesis that there was no difference between the mean ambient particle concentrations during the paired Fast-FFR and reference protocols (p-value <0.001).  This difference is unlikely to impact interpretation of the results, however, because the ambient particle concentrations were still within the target range.
   
   The magnitude of variability in the ambient particle concentration in each protocol, indicated by the coefficient in variation in the initial and final ambient particle concentration measurements for each protocol, was higher for the Fast-FFR than for the reference method.  The mean coefficient of variation was 23.2% and 11.4% for the Fast-FFR and reference methods, respectively. This may be due to the shorter sampling time in the Fast-FFR in the context of the relatively low ambient particle concentrations used in the FFR fit test.  
   
   4.    Ten models of NIOSH-approved N95 FFRs from 6 manufacturers were used.  The specific models were:
   1. 3M 8000 (R-cup)
   2. 3M 9010 (R-vertical fold)
   3. MSA 10102581 (R-cup)
   4. Gerson 1730 (R-cup)
   5. US Safety AD2N95A (R-cup)
   6. Moldex 1712 (R-vertical fold)
   7. Moldex 2200 (M/L-cup)
   8. Moldex 2201 (S-cup)
   9. Kimberly Clark 46727 (R-horizontal fold)
   10. Kimberly Clark 46867 (S-horizontal fold)

   The respirators reflect three different FFR designs: vertical fold (two manufacturers), horizontal fold (one manufacturer) and molded cup (five manufacturers).  The parenthetical statement was assumed to denote the respirator size: R for regular, M for medium, L for large and S for small.  If this assumption is correct, the respirators represent a variety of different sizes and models, and seem sufficient to achieve a variety of different airflow and leak patterns.  The respirators also reflect commonly used brands and models.  We cannot evaluate what proportion of the market share these respirators occupy.
   1.    TSI used pairs of fit tests performed on the same person with the same respirator model in two different experimental trials to calculate the coefficient of variation for the reference fit test method, and define the exclusion zone. As described in Section 7.2, we have some concern with the method used to calculate the standard deviation.  
   
   Sixty-three pairs of fit tests were available, but 8 were excluded by TSI, leaving 55 pairs of fit tests.  Two pairs were excluded because the normal breathing fit factor criterion. Six pairs were excluded because they were identified as outliers. The pairs of fit tests, with the difference of the logarithms of the fit factors, identified as outliers were:
   * FT13112013130916, D = -1.127
   * FT19112013081625, D = -1.251
   * FT12112013144611, D = -0.612
   * FT 12112013142015, D =  -0.573
   * FT12112013101113, D = -0.350
   * FT06112013124410, D = 0.271 
   Based on this analysis, TSI defined the fit factor exclusion zone to be 78-128.  
   For the 55 pairs of fit-tests, the mean value of the difference was 0.0016 and  = 0.11, so three standard deviations span (-0.32, 0.32).  The difference for experimental trial FT06112013124410 falls within this range. If this experimental trial were not excluded,  would still equal 0.11, resulting in no change in the exclusion zone.
   For the 62 pairs of fit-test (excluding only two pairs for the normal breathing fit factor criterion), the mean value of the difference was -0.0583 and  = 0.262, so three standard deviations span (-0.85, 0.73).  The differences for experimental trails FT12112013144611, FT 12112013142015, FT12112013101113 and FT06112013124410 fall within this range and are not outliers.  Only FT13112013130916 and FT19112013081625 are outliers by this definition. Based on 59 pairs of fit tests,  = 0.16, which gives an exclusion zone of 69-144.
   All respirator models were represented in the experimental trials used to determine the exclusion zone.  
   2.    The exclusion zone defined by TSI for the reference method was fit factors between 78-128.  A total of seven tests had fit factors for the reference method in this range. These include:
   * FT10122013100830, FF = 105
   * FT18122013104730, FF = 82
   * FT20112013140605, FF = 84
   * FT20112013130114, FF = 109
   * FT18122013095959, FF = 110
   * FT26112013082004, FF =122
   * FT18122013123246, FF = 125
   Note that experimental trial FT10122013100830 was also excluded by the normal breathing fit factor criterion.  The wider exclusion zone, fit factors 69-144, would exclude four additional pairs of fit tests: FT20112013140818, FT20112013104650, FT10122013092202, and FT18112013103005. 
   3.    Two exclusion criteria were applied: 1) the exclusion range for the fit factor of the reference method excluded six paired tests, and 2) the normal breathing fit factor criterion excluded one additional paired test.  After exclusions by TSI there were 106 paired tests from 29 different subjects. Using the wider exclusion zone, there are 102 paired tests. These meet the minimum of 100 sequential paired tests from at least 25 different subjects.
   
   The required fit factor is 100, and 5% of the required fit factor is 5.  
   
   The lower bound of the exclusion zone defined by TSI was 78.  Sixty-one pairs had reference fit factors  5 and < 78.  Fifty-seven pairs had reference method fit factors > 5 and < 78.  The experimental trials meet this criterion.
   The lower bound of the wider exclusion zone is 69. Sixty pairs had reference fit factors  5 and < 69.  Fifty-six pairs had reference method fit factors > 5 and < 69.  Even with the wider exclusion zone, the experimental trials meet this criterion.
   4.    When the new fit test method identified an appropriate fit (e.g., fit factor >= 100), the reference fit factors were 110.  No reference fit factor was measured to be less than 10% of the required fit factor (e.g., < 10) when the fit was determined to be acceptable by the new fit test method. The method is not disqualified on this criterion.
   
   5.    Figure 4 of Richardson et al. (2014b) demonstrates that the fit factors less than the required fit factors are evenly distributed, and are not primarily lower fit factors.

   6.    The two fit test methods were compared over a range a measurements that include the target fit factor for FFRs, FF  = 100, which is the intended target of the method. 

   7.    Figure 4 of Richardson et al. (2014b) shows the cumulative distribution of the fit factors obtained for the reference method, which has fit factors ranging from 1 to 10,000.

   8.    Information about the respiratory make, model, style, size, individuals tested and paired results of the new test and the reference test were provided by TSI.  Similar information was provided for any test results excluded from analysis, along with reasons for exclusion.

   9.    The test results provided by TSI included the components of a 2  2 contingency table: A = 0, B = 31, C = 69, D = 6.  All calculations were accurate and align with the ANSI/AIHA Z88.10-2010 criteria (Table 3).  With the wider exclusion zone, A = 0, B = 29, C = 68, D = 5; the new fast fit test protocol still meets all of the criteria (Table 3).


                 Table 3. Performance of the Fast-FFR method.

                                     ANSI
                                   Criterion
                                      TSI
                                  Calculation
                                     Wider
                                Exclusion Zone
Test Sensitivity
                                   >= 0.95
                                     1.00
                                     1.00
Predictive Value of a Pass
                                   >= 0.95
                                     1.00
                                     1.00
Test Specificity
                                   >= 0.50
                                     0.84
                                     0.85
Predictive Value of Fail
                                   >= 0.50
                                     0.92
                                     0.93
Kappa Statistic
                                   >= 0.70
                                     0.87
                                     0.89


4. COMPARISON WITH EXISTING OSHA CNC FIT TEST PROTOCOL AND REFERENCE METHOD WITH RESPECT TO POORLY-FITTING RESPIRATORs

For the most part the three studies meet the ANSI/AIHA Z88.10 Annex A2 criteria.  We disagree with the method used to calculate the standard deviation and identify outliers in the determination of the exclusion zone.  Applying our recommended approach to determine the exclusion zone reduces the number of experimental trials that fall below the lower bound of the exclusion zone and greater than 5% of target fit factor, such that there were too few observations in this range for the Fast-Full method: ANSI/AIHA Z88.10-2010 specifies 50 paired observations of the new and reference method must fall in this range.  Relatedly, we interpret this ANSI/AIHA Z88.10-2010 criterion to mean that 50 observations must be > 5% of the target fit factor and less than the lower bound of the exclusion zone; while TSI interpreted the criterion to indicate that 50 observations must be >= 5% of the target fit factor and less than the lower bound of the exclusion zone.  Applying our interpretation, the Fast-Half method would not meet this criterion. However, we do not believe this is a serious issue, as all of the other ANSI criteria are easily met or exceeded by the remaining data for the Fast-Half and Fast-Full methods. 
The studies demonstrate that the three new fit test methods are equivalent to the current OSHA Ambient Aerosol Condensation Nuclei Counter Quantitative Fit Testing Protocol.
For all three respirator classes, it appears that the selected respirators reflect the range of designs for each type of respirator (elastomeric full- and half-facepiece and filtering facepiece), but the explanations for selection criteria and reasons for selecting these particular respirators are lacking in the three publications.
We conclude that the study methods were, in general, scientifically, technically and statistically sound and acceptable, with the small concerns described above.  The reference method applied, as described in the three publications (Richardson et al. 2013; 2014a; 2014b), appears to follow the protocol as described in the OSHA Respiratory Protection Standard, with slightly more strict direction for breath-taking at the extremes of the head up-and-down and head side-to-side exercises.
OSHA has asked us to consider whether the grimace exercise should be retained in these modified protocols.  This is a good question and one we are unable to answer.  This would be a reasonable question to include in the Notice of Proposed Rulemaking (NPRM).  It is unclear what or if the grimace adds to our understanding of fit under the current protocol.  While the grimace exercise is included in all QNFT protocols (including the Controlled Negative Pressure method), it is not required for qualitative fit tests.  This exercise certainly has the capacity to dislodge a respirator, but in an unpredictable manner.  It is not clear how or whether this type of dislodgement occurs during regular respirator wear.
OSHA has asked us to consider whether it was appropriate for TSI to eliminate the talking exercise from the Fast-Full and Fast-Half protocols.  Zhuang et al. (2004) identified talking, moving head up-and-down and bending over as important exercises to include when using shorter or fewer exercises.  Again, this would be a good question to include in the NPRM.  We also suggest a question in the NPRM that addresses the use of talking rather than jogging in the Fast-FFR test.
5. CONCLUSIONS AND RECOMMENDATIONS

Overall, we found the performance and reporting of the Fast-Full, Fast-Half and Fast-FFR methods were consistent with the ANSI/AIHA Z88.10-2010 standard.  We identified the following topics that should be included in a request for public comment:
Topic 1: The selection of respirators.  TSI has not described the specific criteria used to select the respirators for testing.  We believe that the respirators selected reflect a range of models and sizes in each respirator class.  We are unable to evaluate whether the respirators selected represent the majority of respirators used in the United States.
Topic 2: Determination of the exclusion zone.  TSI used a somewhat circular logic to identify outliers in paired tests of the reference method: Outliers were defined as pairs of reference method fit tests with differences in the logarithm of the fit factors greater than three standard deviations from the mean from the remaining data points. We suggest that outliers should be identified as pairs of reference method fit tests with differences in the logarithm of the fit factors greater than three standard deviations from the mean from all data points. This definition decreases the number of pairs of fit tests identified as outliers and increases the width of the exclusion zone. The larger exclusion zone does not affect the performance statistics of the fast fit test methods.  However, for the Fast-Full method, the wider exclusion zone decreases the number of tests where the reference method falls between the lower bound of the exclusion zone and 5% of the required fit factor (> 25 and < 320.6), and no longer meets the minimum criterion of ANSI/AIHA Z88.10-2010.
Topic 3: Selection of the exercises.  TSI has provided a reasonable and acceptable rationale for selecting four rigorous exercises for the Fast-Full and Fast-Half methods.  The rationale for replacing jogging with talking in the Fast-FFR method was not well-supported by data or reasoning.  We suggest including this as an issue to be addressed in the NPRM.
Topic 4: Exclusion of the grimace.  We understand and generally support the omission of the grimace from the experimental protocol comparing the traditional with new fast-fit methods.  Because the protocol requires two consecutive fit tests, the grimace introduces an unpredictable source of variability.  Whether the grimace should be eliminated completely from these new methods is more difficult to assess.  If the experimental protocol does not allow comparison of consecutive methods that include the grimace, then what other criteria could be used to assess its inclusion or exclusion?  Perhaps this issue should be addressed in the NPRM.
Topic 5: Breathing instructions. While the current OSHA methods include no instructions about the timing, frequency or depth of breathing during the head up-and-down or side-to-side exercises, the rationale provided by TSI is reasonable  -  that requiring two deep breaths at the extremes of the former and at the head down location in the latter makes these exercises both more rigorous and more reproducible from one subject to the next.  It is also reasonable given the short duration of these exercises.  We are not aware of any research or data that would support this vs. any other approach to these exercises, however.  We suggest including this as an issue to be addressed in the NPRM.
6. REFERENCES

OSHA (2004) Controlled Negative Pressure REDON Fit Testing Protocol. Federal Register 69(149):46986-94 (August 4, 2004).

Richardson AW, KC Hofacre, J Weed, R Holm and R Remiarz  (2013) Evaluation of a faster fit testing method for full-facepiece respirators based on the TSI Portacount(R). Journal of the International Society for Respiratory Protection. 30(2): 116-128.

Richardson AW, KC Hofacre, J Weed, R Holm and R Remiarz (2014a) Evaluation of a faster fit testing method for elastomeric half-mask respirators based on the TSI Portacount(R). Journal of the International Society for Respiratory Protection. 31(1): 9-22

Richardson AW, KC Hofacre, J Weed, R Holm and R Remiarz (2014b) Evaluation of a faster fit testing method for filtering facepiece respirators based on the TSI Portacount(R). Journal of the International Society for Respiratory Protection. 31(1): 43-56.

Sietsema M and LM Brosseau (2015) Comparison of two quantitative fit test methods. (In preparation.)

TSI Incorporated (2014) Analysis of the Talking Exercise Used for Respirator Fit Testing. White Paper. (7/10/2014)

TSI Incorporated (2015a) Exercises Selection for the Fast-Full, Fast-Half and Fast-FFR Fit Test Studies. (2/6/2015)

Zhuang Z, CC Coffey and RB Lawrence (2004) The effect of ambient aerosol concentration and exercise on Portacount(R) quantitative fit factors. Journal of the International Society for Respiratory Protection 21: 11-20.




