A new sampling algorithm for use with ATP testing

A new sampling algorithm for
use with ATP testing
Greg Whiteley
Managing Director – Whiteley Corporation
Declaration of conflict of interest
• None to declare: with respect to this presentation
• I have no association with any brand of ATP device, nor does my
company, nor do any of my affiliates
• I have no financial links to ATP testing, other than to use these
devices and conduct validation research through WSU
• All data referenced in regards to any ATP testing device has been
peer reviewed and internationally published except where noted
that writing for submission, or current peer review is still underway
Project Design features
Cleaning processes
Outcome:
Measurable
cleanliness
Lets change just
one item and use
a control group
Should be a
snack!
We looked for an ICU
The ICU manager wasn’t happy with the
micro results after the hygiene audit
Background on ATP testing
Advantages:
Easy to use; real time results; broad indicator of
cellular contamination
Problems:
Variability & imprecision; relative scaling; lack of
brand to brand interoperability; sampling error
Early Validation work on ATP testing devices
• Each device uses a unique
scale – all named ‘RLU’
[Relative Light Units]
• The Lower Limit of
Quantitation is different
from the lower level of
detection for some brands
• Several brands do not
read down to zero, so a
practical zero is required
leading to user confusion
• Imprecision is virtually
undetectable in use
Whiteley et al., Healthcare Infection:2012:17:91-97
Calibration testing of ATP testing devices compared to a
laboratory standard analytical tool - HPLC
HPLC Calibration curve using pure ATP
Concentration (ppm)
12
y = 6E-05x - 0.0242
R² = 1
10
8
6
4
2
0
0
20000
40000
60000
80000
100000
Mean Peak Area
120000
140000
160000
180000
200000
Variance measurements Cv = σ /
Coeffecient of Variance CoV for three portable ATP bioluminometers
0.7
0.6
0.5
0.4
CoV
Cleantrace 3M [n=57 (246 swabs)]
Kikkoman [n=49 (222 swabs)]
0.3
Hygiena [n=47(199 swabs)]
LCMS [n=22 (72 runs)]
0.2
n = number of separate
dilutions tested per brand
0.1
0
0
10
20
30
40
Number of dilutions tested for each unit
Whiteley et al., Infect Control Hosp Epidemiol:2013:34:538-540
50
60
Summary findings on variability
Whiteley et al., ICHE 2015
Whiteley et al., Infect Control Hosp Epidemiol:2015:36:658
Key Findings
1. The variability with bacteria is the same as the
variability with the pure source ATP
2. With a Cv of 0.4, any reading has a 20% chance to
be wrong by a factor of two: i.e. the error
potential on a reading of 100 RLU is from 50 RLU
to 200 RLU.
3. Readings using more than a single point are
required for statistical validity.
Finding the bad bugs in a busy ICU
Ref: Whiteley et al., Am J Infect Control:2015:43:1270-5
So where are the bad bugs?
A cross sectional ICU pilot survey
Data: pre-publication Knight, Whiteley, Jensen, Gosbel and others., 2016
We needed to improve certainty over the ATP readings
SO WE DEVELOPED A NEW
SAMPLING ALGORITHM
Key features of the new sampling algorithm
• Starts with duplicate sampling on proximal
surfaces of an HTO or RMD
• Uses a standardised sampling area – normally
2cm x 5cm = 10cm2
• Compares results to a predetermined initial
cleanliness threshold
• Uses a cleaning intervention step for internal
validation of cleanability & cleanliness
A new ATP testing Sampling Algorithm
2 samples both
RLU <100 RLU*
2 samples both
RLU > 100 RLU*
2+ samples:
one < 100 RLU*
& one > 100
RLU*
Cleaning
Intervention
Step
Resample aiming for
cleanliness at
< 50 RLU
Cleaning
Intervention
Step
Resample aiming for
cleanliness at <50RLU
Cleanliness
intervention step to
repeat and sample
repeat until cleanliness
is < 50 RLU
Continue sampling for
up to four samples to
indicate outlier effect
Cleaning
Intervention
Step
Cleanliness
intervention step to
repeat and sample
repeat until cleanliness
is <50RLU
* 100 RLU measure is specific only to Hygiena
Food Premises Cleanliness Study*
•
•
•
•
Conducted in regional location in NSW
8 Food Premises & 72 items examined
Statistics using Mann Whitney and Wilcoxon
Cleaning Intervention Step using disposable
detergent (anionic) wipe
• Cleaning Principle: “one wipe, used on one
surface, wiping in only one direction”**
[**Sattar & Maillard, 2013. Am J Infect Control:41:S97-S104]
*Paper: Whiteley Nolan & Fahey: J Env Health (USA) under peer review, 2016
Results: Spread of duplicate samples with
Line of best fit between duplicate samples
Paper: Whiteley Nolan & Fahey: J Env Health (USA) under peer review, 2016
Before and after Cleaning
Intervention Step results
Paper: Whiteley Nolan & Fahey: J Env Health (USA) under peer review, 2016
Key findings from the Food Premises survey
• Before and after findings – all statistically significant
(P>.001) for all categories except <25 RLU (P=0.136)
Classification
Before
After
Significance
Clean
2x<100 RLU
19
6
P = 0.001
Unclean
2x > 100RLU
922
10
P = 0.001
Mixed
± 100RLU
192
5
P = 0.001
Very Clean
Av < 25 RLU
8
5
P = 0.136
Paper: Whiteley Nolan & Fahey: J Env Health (USA) under peer review, 2016
The algorithm mapped – IDH 2016
Paper: Whiteley Glasbey Fahey, IDH 2016: pending DOI
Key comparative charts
Table 1 from the paper
Initial Cleanliness threshold
Secondary Cleanliness threshold
Tertiary Cleanliness threshold
Lower Limit of Quantitation
Cleanliness
Threshold
Hygiena
Cleantrace
Kikkoman
T C1
T C2
T C3
LLQ
100 RLU
50 RLU
25 RLU
0 RLU
500 RLU
250 RLU
125 RLU
100 RLU
460 RLU
230 RLU
115 RLU
90 RLU
Table 2 from the paper
Identified Cleanliness
Threshold
TC 1
TC 2
TC 3
Defining the Cleanliness Thresholds relationships
TC1 = TC2 x 2 = TC3 x 4
TC2 = TC1 ÷ 2 = TC3 x 2
TC3 = LLQ + 25 RLU (may be higher for some devices)
Paper: Whiteley Glasbey Fahey, IDH 2016: pending DOI
Medical Device Study
•
•
•
•
Survey is currently underway and on-going
Tested 258 individual surfaces and items so far
Over 1000 swabs so far
Cleaning intervention step using two different
wipes – testing difference in cleaning
outcomes between wipes
• Statistical analysis – standard methods
Conclusions
1. New sampling algorithm improves certainty
when using ATP testing devices;
2. Algorithm is simple to use in field hygiene
assessments with slightly increased costs;
3. New sampling algorithm requires further
study in well designed trials to test validity
and precision improvement of ATP testing…
Acknowledgements
• Dr Trevor Glasbey, Regulatory and Research Manager, Whiteley
Corporation
• Paul Fahey, Statistician, Western Sydney University
• Mark Nolan, Environmental Health Officer, National Parks and Wildlife
Service of NSW
• ASUM
• ASUM Staff: Lyndal Macpherson, Associate Professor Sue Westaway, Dr
Jocelyne Basseal, Ann-Marie Gibbons
• Jessica Knight, Professor Iain Gosbell, and Associate Professor Slade
Jensen – all based at Western Sydney University and the Ingham Research
Institute
• Thank you
Thank you
Questions?