The trueness and precision (components of accuracy) of the method
were defined by assays with equine muscle, bovine kidney and swine
kidney, and poultry muscle blanks and spiked at 62 μg Kg−1 (14 ng
Hg) and 333 μg Kg−1 (75 ng Hg) in six independent replicates for
each analyzed level. Trueness was investigated through mean recovery
obtained for the six replicates of spiked samples at each level. The
minimum trueness criteria ranged from −20% to +10% for the levels
studied [24]. In addition, the trueness was also verified by the analysis
of the certified reference material of protein fish (DORM-3), where Hg
has certified the value of 382 ± 60 μg Kg−1
. This CRM was used due
to intralaboratory unavailability of CRM similar to matrices studied in
this work. The intralaboratory precision of the method was determined
through repeatability and within-laboratory reproducibility. For repeatability,
the six replicates of spiked samples at each level were prepared
two times (once a day) on 2 different days by the same analyst using the
same experimental conditions (same sample, reagents and standards).
Repeatability was expressed as a relative standard deviation obtained
for the replicates of spiked samples at each level [25]. For the withinlaboratory
reproducibility, another two analysts performed the six
replicates of spiked samples at each level on different days, once each,
by different analysts. The within-laboratory reproducibility was also
expressed as relative standard deviation using the same concentration
range applied to the recovery test. According to 2011/836/EU Regulation,
the maximum limit of precision is given by the HORRATr value
for repeatability and HORRATR for within-laboratory reproducibility.
For inorganic contaminants the maximum HORRATr and HORRATR
values are 2 [25]. The daily calibration was carried out with a high and
low concentration standard for each working range using the analytical
parameters as recommended by the manufacturer to check each curve.
The working calibration standards must be measured within 10% of
their true value for the curve to be considered valid [20].