Technical Challenges with PCB Analysis
Michael Gershenzon November 18, 2010
Alcoa Technical Center, PA
- Current PCB Sources
- Background PCB Levels in the Environment
- Analytical Methods for PCB Detection
- Historical Notes on Method 1668
Three Case Studies – Importance of Site-Specific
- Case Study 1: Site-Specific MDL Study for EPA
- Case Study 2: Inter-laboratory Comparison of MDLs
for Method 1668A
- Case Study 3: Comparison of PCB Analysis Results
Obtained by Methods 608 and 1668A
Introduction – What are PCBs?
Polychlorinated Biphenyls (PCBs):
209 individual congeners
Man-made chemicals started to be released into the
environment in 1900’s
• Estimated that > 1.5 bn lbs. manufactured in the US before 1977
Formerly used in a variety of industrial applications (electric
components, hydraulics, heat transfer systems, flame
Production banned in late ‘70s due to toxicity, but still
ubiquitous in environment due to persistence (thermal and
What are the Current PCB Sources?
PCBs used many years after production ban
Dielectric oils considered hazardous waste
PCB > 50 ppm
• Contaminated sites
• Leaking transformers
Sediment “hot spots”
• Contaminated sites off-gas
• Combustion of contaminated recycled oil
• Rainwater contribution to “background” PCBs => surface waters
Example: PCB Concentrations in Rainwater (ng/L, ppt)
Source:Polychlorinated Biphenyls in Chicago Precipitation, Env. Sci. Techn., 1997
Example: PCB Concentrations in Human Blood Plasma
Geometric mean PCB concentration in g/L (ppb)
Source: AMAP Assessment Report: Arctic Pollution Issues 1998
Analytical Methods for PCB Detection
EPA Method 608 - Currently in use for SPDES compliance
GC/electron capture detector instrument
Measures Aroclors (mixtures of congeners, e.g. Ar. 1248 – 48%
The only EPA-approved method for use in NPDES permit program
Method Detection Limit (MDL) range 10 – 100 ng/L (ppt)
Cost per sample ~ $50/sample
EPA Method 1668 – Proposed for use in SPDES compliance
High Resolution GC/MS Method
Capable of detecting individual congeners
EPA-estimated Method Detection Limit Range: 0.004 – 0.45 ng/L
Cost per sample ~ $1,000/sample
How Many Grains of Salt are in 1668?
ppm (mg/L) ppb (g/L) ppt (ng/L) ppq (pg/L)
3,000,000,000 3,000,000 3,000 3
grains salt grains salt grains salt grains salt
Vol. ~277 million gallons
Can You Find a Cowboy in Texas?
1 ppq area of a cowboy hat
relative to the territory of Texas
Method 1668 Historical Notes
Published by EPA in 1999 (Revision A)
Inter-laboratory Validation Study conducted by EPA in 2003 – 2004 (14
• 11 labs submitted data
• 5 labs submitted data deemed unusable by EPA; Data from six labs was accepted
• Evidenced lab to lab variability in method implementation (utilizing smaller sample
volumes, diluted final extracts, misinterpretations of QA requirements)
• Final report published in 2008 (Revision B issued)
Inter-laboratory study conducted by GE and Env. Standards in 2008:
• Spiked (0.3 – 45 ppt) natural water samples submitted to three labs
• Significant lab-to-lab data variability (~order of magnitude)
EPA’s Lower Passaic River Oversight Program (2008):
• Two labs
• Matrix – River Sediments
• Only samples with PCBs > 5X quantitation limit were used in the statistical analysis
• High variability in results - 320 out of 1500 split samples had congener specific
result RPD’s >30%
• This amount of variability is concerning when reporting results from different labs at
or near compliance goals (which numbers do you use?)
Method 1668 Historical Notes (cont’d)
Proposed for approval by EPA (1668C, 2010)
Once approved, regulated parties may be required to use
1668 (instead of 608) under the “sufficiently sensitive” method
What would the data quality be like?
Case Study 1 /Alcoa/:
Method 608 Site-Specific MDL
‣ Conducted at an Alcoa location
‣ Site‐specific water used in the study (stormwater)
‣ Study conducted as outlined in 40 CFR Part 136 Appendix B
grab (50 gal)
split into 7
(20, 50, 75,
100, 175, and 300 ppt)
Lab‐reported MDL using reagent‐grade water = 10 ppt
20 ppt Spike Results
5 out of 7 samples spiked at 20 ppt Aroclor were reported as below DL..
300 ppt Spike Results
MDL = 56 ppt, calculated according to 40 CFR 136 Appendix B
Method 608 Site-Specific MDL Study Results Summary
Site Specific MDL
3 out 7 < DL
52 out out of 7 7 < < DL DL
3 out 7 < DL
0 out of 7 < DL
0 out of 7 < DL
Original Method MDL
0 out of 7 < DL
Aroclor Spike Concentration (ng/L)
Original MDL = 10 ppt; Computed MDL = 82 ppt – 99% upper confidence limit
Case Study 1 Conclusions
The inability to detect PCBs spiked at 75 ppt and
below suggests an MDL < 75 ppt is inadequate
Each computed MDL (using site-specific water)
exceeds lab-reported MDL = 10 ppt (using reagent
Computed MDLs allow the plausibility of an MDL as
high as 82 ppt, the 99% confidence limit..
Relative variability would be accentuated for 1668
Case Study 2: Inter-laboratory
Comparison of MDLs for Method 1668A
608 MDL/S‐S MDL range
Lab A; Avg = 2.0 ppq;
Lab B; Avg = 3.8 ppq;
Lab C; Avg = 50.4 ppq;
EPA Estimated MDL; Avg = 178.5 ppq;
Method 608 MDL Range
Significant (up to ca. 2 orders of magnitude) discrepancy in lab-reported MDLs
What is this discrepancy due to? Method implementation, etc.? Which lab to use?
Case Study 2 Conclusions
Up to 2 orders of magnitude difference in lab-reported MDLs
Lab MDL studies can provide unrealistically low “optimized”
- one analyst - all samples run on one day
- one instrument - ideal conditions
looks good for competing laboratories, but does not
describe routine lab capability
Reporting of positive results based on “super low” MDL studies
can result in NPDES permit violations when, in reality, the
reported detection is instrument noise
Example of an unrealistically low MDL presented in Case Study 1
Case Study 3 /Alcoa/: Method 1668A vs
608 Results Comparison
Water samples collected at 8 different outfalls and internal sampling locations.
Submitted for parallel analysis by 608 and 1668A.
Example of 1668A Analysis Report
Oneway Analysis of Conc (ng/L) by Congener; Site = Alcoa; Homolog = 4
Actual reported data
zeros are neither reportable values nor a valid detection limit..
Potential Issues with Inadequate
The average value and variability for the observed concentrations in surface waters have reduced
“Earlier limitations were caused more by [sample] contamination than by inadequate detection limits..”
Source: W. Fitzgerald. Sci of Tot. Env., 87/88, p. 223, 1989 and N.S. Bloom, National Forum on Mercury in Fish.. 1995
Sample Collection and Handling
Requirements for ppq-Level Analysis
PCB concentration (pg/L)
What is a realistic MDL for 1668?
What is an upper bound on the noise level?
False Georgia positive/negative identifications?
How do you conduct a site‐specific MDL Study for 1668?
- spiking levels?
- background levels?
St Lawrence - blanks?
How high is the cost?
1668 Lab 608 MDL
Comparison of PCB water criteria, h.h. criteria, river, rainwater, and method MDL levels
Alcoa: D. Belleau, D. Coleman, T. Lightfoot, C. Dobbs, J. Morton, R.
Dworek, J. Lease, and J. R. Smith
Environmental Standards, Inc.: David Blye
Contact: Michael Gershenzon, firstname.lastname@example.org