5020 QUALITY ASSURANCE/QUALITY CONTROL*

5020 A. Introduction

Without quality control (QC), there is no confidence in sample Each method typically includes acceptance-criteria guidance
results. As described in Part 1000, essential QC measures in- for precision and bias of test results. If not, the laboratory should
clude calibration, reagent standardization, assessment of each determine its own criteria (e.g., using control-charting tech-
analyst’s demonstration of capabilities, analysis of blind check niques). For some Part 5000 procedures (e.g., the BOD proce-
samples, determination of the method’s sensitivity [method de- dure) the traditional determination of accuracy—adding a known
tection level (MDL) or minimum reporting level (MRL)], and amount of analyte to either a sample or a blank—is not practical.
regular evaluation of bias, precision, and the presence of labor- This does not, however, relieve analysts of the responsibility for
atory contamination or other analytical interference. The details evaluating test accuracy. Instead, obtain certified ready-made
of these procedures, their performance frequency, and expected analytes for such tests.
ranges of results should be formalized in a written Quality Evaluate precision by analyzing duplicate or spiked duplicate
Assurance Manual and standard operating procedures. samples.
Some of the methods in Part 5000 include specific QC pro- To help verify the accuracy of calibration standards and over-
cedures, frequencies, and acceptance criteria. These are consid- all method performance, participate in an annual or preferably
ered to be the minimum quality controls needed to perform the more frequent program of analysis of single-blind QC check
method successfully. Additional QC procedures can and should samples (QCS)—ideally provided by an external entity. Such
be used. Some regulatory programs may require additional QC programs are sometimes called proficiency testing (PT)/perfor-
or have alternative acceptance limits. mance evaluation (PE) studies. An unacceptable result on a PT
sample is often a strong indication that a test protocol is not
being followed successfully. Investigate circumstances fully to
* Reviewed by Standard Methods Committee, 2010. find the cause. In many jurisdictions, participation in PT studies
Roy-Keith Smith, Rodger B. Baird, Andrew D, Eaton, Robin Parnell. is a required part of laboratory certification/accreditation.

5020 B. Quality Control Practices

1. Initial Quality Control provide quantitative data at the reporting limit is verified. If MDL is
determined, verify MDL at least annually for each analyte in a
a. Initial demonstration of capability (IDC): Before analysts method and major matrix category. The laboratory should define all
run any samples, verify their capability with the method. Run a matrix categories in its QA plan. Review MDL requirements as per
laboratory-fortified blank (LFB) (5020B.2e) at least four times Section 1020. Analyze samples for MDL determinations over at
and compare to the limits listed in the method. If no limit is least a 3-d period to generate a realistic value. Include all sample-
specified, use the following procedure to establish limits: preparation steps in the MDL determination. Using data from the
Calculate the standard deviation of the four samples. The low-level LFBs included in each analytical run is an economical
LFB’s recovery limits are way to calculate MDL.
Ideally, use pooled data from several analysts rather than data
LFB’s initial recovery limits ⫽ Mean ⫾ (5.84 ⫻ Standard Deviation) from one analyst. (For specific information on MDLs and pool-
ing data, see Section 1020B.) To verify the MDL annually on
where each instrument used in the laboratory, analyze a QC sample
(subjected to all sample-preparation steps) spiked at a level 1 to
5.84 ⫽ the two-sided Student’s t factor for 99% confidence limit for 4 times the MDL. A successful verification is one that meets all
three degrees of freedom.1 the method’s detection criteria.
Also, verify that the method is sensitive enough to meet c. Operational range: Before using a new method or instru-
measurement objectives for detection and quantitation by deter- ment, determine its operational range (upper and lower limits) or
mining the lower limit of the operational range. (For basic at least verify that the intended range of use is within the
guidance on demonstrating capability, see Section 1020B.) operational range. For each analyte, use standard concentrations
b. Method detection level (MDL): If data will be reported below that provide increasing instrument response. The MRL is set to
the calibrated range, then before analyzing samples, determine the a concentration at or above the lowest standard used in the
MDL for each analyte via Section 1020 or other applicable proce- analysis. Verify quantitation at the MRL initially and at least
dures.2 MDL determination is not required if 1) data are not re- quarterly (preferably daily) by analyzing a QC sample (subjected
ported below the instrument’s calibrated range, and 2) the ability to to all sample-preparation steps) spiked at a level 1 to 2 times the

1
 QUALITY ASSURANCE/QUALITY CONTROL (5020)/Quality Control Practices

MRL. A successful verification meets the method’s or labor- TABLE 5020:I. MINIMUM QUALITY CONTROL FOR METHODS IN PART 5000
atory’s accuracy requirements at the MRL. Laboratories must
Method LFM† &
define acceptance criteria for the operational range, including the Section Analyte Blank LFB* LFMD‡ Other
MRL, in their QA documentation.
5210B BOD - - - 1,2,3
2. Ongoing Quality Control 5210C - - - 1,2,3
5210D - - - 1,2,3
a. Calibration: Calibrate the method using the directions in the
5220B COD ⫻ ⫻ ⫻ 1,2,3
procedure. Appropriate calibrations may be linear, weighted or
5220C ⫻ ⫻ ⫻ 1,2,3
second order. (For basic calibration guidance, see Section 1020B.) 5220D ⫻ ⫻ ⫻ 1,2,3
b. Calibration verification: Verify calibration by periodically
analyzing a calibration standard and calibration blank during a 5310B TOC ⫻ ⫻ ⫻ 1,2,3
run—typically, after each batch of ten samples and at the end of 5310C ⫻ ⫻ ⫻ 1,2,3
the run. The analyte concentration in calibration-verification 5310D ⫻ ⫻ ⫻ 1,2,3
standards should be varied over the calibration range to deter-
mine detector response. 5320B Dissolved Organic Halogen ⫻ ⫻ ⫻ 1,2,3
For the calibration verification to be valid, (unless the method
specifies otherwise) check standard results must not exceed 5510B Aquatic Humic Substances ⫻ ⫻ ⫻ 1,2,3
5510C ⫻ ⫻ ⫻ 1,2,3
⫾10% of its true value, and calibration blank results must not be
greater than one-half the reporting level. 5520B Oil and Grease ⫻ ⫻ ⫻ 1,2,3
If a calibration verification fails, immediately cease analyzing 5520C ⫻ ⫻ ⫻ 1,2,3
samples and initiate corrective action. Then, re-analyze the cal- 5520D ⫻ ⫻ ⫻ 1,2,3
ibration standard and blank. If the calibration verification passes, 5520E ⫻ ⫻ ⫻ 1,2,3
continue the analysis. Otherwise, repeat initial calibration and 5520F ⫻ ⫻ ⫻ 1,2,3
re-analyze samples run since the last acceptable calibration ver- 5520G ⫻ ⫻ ⫻ 1,2,3
ification.
If the LFB is not prepared from a second source to confirm 5530B§ Phenols ⫻ ⫻ ⫻ 1,2,3
method accuracy, (unless the method specifies otherwise) the 5530C ⫻ ⫻ ⫻ 1,2,3
5530D ⫻ ⫻ ⫻ 1,2,3
laboratory must also verify the accuracy of its standard prepa-
ration by analyzing a mid-level second-source calibration stan- 5540B§ Surfactants ⫻ ⫻ ⫻ 1,2,3
dard whenever a new initial calibration curve is prepared. Results 5540C ⫻ ⫻ ⫻ 1,2,3
must agree within 15% unless otherwise specified in a method. 5540D ⫻ ⫻ ⫻ 1,2,3
c. Quality control sample (QCS): Analyze an externally gen-
erated, blind QCS (unknown concentration) at least annually 5550B Tannin and Lignin ⫻ ⫻ ⫻ 1,2,3
(preferably semi-annually or quarterly). Obtain this sample from
a source external to the laboratory, and compare results to that 5560B§ Organic/Volatile Acids ⫻ ⫻ ⫻ 1,2,3
source’s acceptance results. If testing results do not pass accep- 5560C§ ⫻ ⫻ ⫻ 1,2,3
tance criteria, investigate why, take corrective action, and ana- 5560 D ⫻ ⫻ ⫻ 1,2,3
lyze a new QCS. Repeat this process until results meet the
5710B THMs and DBPs - - - 1,2,3
acceptance criteria. 5710C - - - 1,2,3
d. Method blank (MB): When appropriate (Table 5020:I), 5710D - - - 1,2,3
include at least one MB daily or with each batch of 20 or fewer
samples, whichever is more frequent. Any constituent(s) recov- 5910B UV-Absorbing Organic ⫻ - - 1,2,3
ered must generally be less than or equal to one-half the report- Constituents
ing level (unless the method specifies otherwise). If any MB * Laboratory-fortified blank.
measurements are at or above the reporting level, take immediate † Laboratory-fortified matrix.
corrective action as outlined in Section 1020B. This may include ‡ Laboratory-fortified matrix duplicate.
re-analyzing the sample batch or qualifying the reported data. § A sample preparation technique that is normally combined with a subsequent
determinative technique
Sample results less than MRL are considered valid even if the
⫻ indicates that a QC type is mandatory for the method.
MB has a positive result, but should be flagged. - indicates that a QC type is not mandatory for the method.
e. Laboratory-fortified blank (LFB): Section 1020 currently 1. Additional QC guidelines in method.
specifies that LFBs and LFMs be made from a second source. 2. Duplicates or LFMD of the sample will be run.
However, as long as each initial calibration solution is verified 3. Refer to 5020B for further QC requirements.
This table is not comprehensive; refer to the specific method and 5020B for further
via a second source (5020B.2b), the LFB/LFM need not be from details.
a second source (unless the method specifies otherwise).
Using stock solutions (preferably prepared with a second
source), prepare fortified concentrations so they are within the
calibration curve. Ideally, vary LFB concentrations to cover the Calculate percent recovery, plot control charts, and determine
range from the midpoint to the lower part of calibration curve, control limits (Section 1020B) for these measurements. Use the
including the reporting limit. control limits to determine ongoing demonstration of capability

2
 QUALITY ASSURANCE/QUALITY CONTROL (5020)/Quality Control Practices

limits. Some methods may have specific limits to use in lieu of where:
plotting control charts; if so, control charts may still be useful in Cs ⫽ LFM concentration determined experimentally,
identifying potential problems. Ensure that the LFB meets the f ⫽ spike dilution correction,
method’s performance criteria when such criteria are specified. C ⫽ concentration of sample before spiking, and
Establish corrective actions to be taken if the LFB does not S ⫽ concentration of spike.
satisfy acceptance criteria.
When appropriate (Table 5020:I), include at least one LFB Note: f should be more than 0.95. Spiking that dilutes a sample
daily or per each batch of 20 or fewer samples. Some regulatory by more than 5% changes the matrix significantly. Ideally, keep
programs require a higher frequency of LFBs. f above 0.99 (equivalent to 1% dilution of sample due to spike
f. Duplicates: When appropriate (Table 5020:I), randomly addition) so f can be ignored and the equation simplified to
select routine samples to be analyzed twice. Process duplicate eliminate f.
sample independently through the entire sample preparation and
analysis procedure. Include at least one duplicate for each matrix b. LFB recovery:
type daily or with each batch of 20 or fewer samples. (Some
regulatory programs require more frequent use of duplicates.)
Cb
Calculate control limits for duplicates when method-specific ⫻ 100 ⫽ % Recovery LFB
limits are not provided. When appropriate (Table 5020:I), run I
either a sample duplicate or an LFMD per batch. It is not
necessary to perform both. (For basic guidance on duplicates, see where:
Section 1020B.) Cb ⫽ LFB concentration determined experimentally, and
g. Laboratory-fortified matrix (LFM)/Laboratory-fortified ma- I ⫽ initial concentration of analytes added to LFB.
trix duplicate (LFMD): When appropriate for the analyte (Table
5020:I), include at least one LFM/LFMD daily or with each c. Relative percent difference:
batch of 20 or fewer samples. (Some regulatory programs require
more frequent use of LFMs. For basic guidance on LFMs and

冢冉 冊冣
LFM ⫺ LFMD
LFMDs, see Section 1020B.) ⫻ 100 ⫽ %RPD
LFM ⫹ LFMD
To prepare an LFM, add a known concentration of analytes
(ideally from a second source) to a randomly selected routine 2
sample without increasing its volume by more than 5%. Ideally,
the new concentration should be at or below the midpoint of the or
calibration curve, and for maximum accuracy, the spike should

冢冉 冊冣
approximately double the sample’s original concentration. If D 1 ⫺ D 2
⫻ 100 ⫽ %RPD
necessary, dilute the spiked sample to bring the measurement D1 ⫹ D2
within the calibration curve. Also, rotate the range of spike 2
concentrations to verify performance at various levels.
Calculate percent recovery and relative percent difference, where:
plot control charts (unless the method specifies acceptance cri- LFM ⫽ concentration determined for LFM,
teria), and determine control limits for spikes at different con- LFMD ⫽ concentration determined for LFMD,
centrations (Section 1020B). Ensure that the method’s perfor- D1 ⫽ concentration determined for first duplicate, and
mance criteria are satisfied. D2 ⫽ concentration determined for second duplicate.
Process fortified samples independently through entire sample
preparation and analysis procedure.
4. References
3. Calculations
1. MEIER, P.C. & E.E. Zünd. 2000. Statistical Methods in Analytical
a. LFM recovery: Chemistry, 2nd ed. Wiley Interscience, New York, N.Y.
2. U.S. ENVIRONMENTAL PROTECTION AGENCY. 1995. Definition and pro-
共C s ⫻ f兲 ⫺ C cedure for the determination of the method detection limit, rev. 1.11,
⫻ 100 ⫽ % Recovery LFM or LFMD
S 40 CFR Part 136, Appendix B. Fed. Reg. 5:23703.