Performancevs. SMPR
Precision
Table1 lists the6-daymean results (twoanalysts, one instrument) for the testmethodonall 11SPIFAN
matrices. The yellow-shaded cells indicate%RSDs that exceed5%. A fewof thesewere flagged for the
elementsCr,Moor Se, but theseelementsarenot the subject of this call formethods.OnlyMnandCa
in theAdultHighFatRTFwere flagged for an intermediateprecision>5%RSD, and thisproduct has since
been shown tobeunstableandperhapsunfit for SLVorMLTwork.Most elements hadRSDsof 1-3%
over 6days. Thismeets theSMPRRepeatability criterionof 5%over the rangeof theassay, given that
therearemore sourcesof varianceat play in inter-daydeterminations. It also suggests that the
Reproducibilityof 8% formajor elementsand10% forminor/traceelementsmaybe readilyachieved
duringanMLT study. ThepresentMLT for 2011.19 (Cr,Mo, Se) isunderwayand some labs arealready
collecting12-element data to check thishypothesis.
The cases for Fe, Cu, andMnarediscussedunder LOQ. Therewereno SPIFANmatrices that really
challenged themethodanywherenear the required lower analytical range for theseelements, andno
low level spikeswereperformed.
SRM1849awas runas a control sampleduringall theseSLVexperiments. Theoverall precision for >16
datapointswas typically1-2% for all 12elements (datanot shown, butmeansare inTable3).
Recovery (Accuracy)
TheSMPR requires a recoveryof 90-110%over the rangeof theassay, and80-115% for low levelsof
Mn. Table2 shows the recoveryof eachelement ineachSPIFANmatrixmeasured in triplicateover each
of threedays. The spikeswereaddedat approximately100%of thenominal element concentration, and
the triplicatemeansoneachdaywereaveraged toone result before taking the (N=3) statistics shown in
Table2. All elements inallmatrices hadaverage spike recoveries in the90-110% range (with rounding),
and so theSMPR requirementsweremet for Recovery. Again, themethodwasnot challenged for the
low levelsof Cu, Fe, orMn in this regard.
Table3 shows theaccuracyof averagevalues for eachelement determined inSRM1849aduring this
study.OnlyKandCawerehigh, outside the certified range, butwehavenoticedother laboratories
producing similar resultsasours,whichareonly2-3%higher than theSRMmean. Theseobservations
point to theaccuracyof thismethod.
Table7provides a thirdmeasureof theaccuracyof thismethod. All 11SPIFANmatriceswerealsoput
throughanSLV for themicrowavedigestion, ICP-AESmethod runat theauthors’ laboratory for
commercial testing (separatemicrowavedigested solutionsbecause theyare indifferent concentration
ranges). The6-daymeans andRSDs are shown for both ICP-AESand ICP-MSmethods. Percent
differences>5%are flagged ina red color,while ICP-AESmethodprecision thatexceeds 5%RSD is
flagged inblue. The two independentmethods generallyagree very closely.Only5 casesof >6%
differenceoccurredout of 90 comparisons (excludes theAdultHighFat,whichagain causedproblems).
MTE-02
FORWORKINGGROUP/ERPUSEONLY
DONOTDISTRIBUTE
