AOAC OMB Final Action Recommendation (December 2019)-2016.14

2016.14 (Jan. 2019) FOS-03 MLT Report FOR ERP USE ONLY DO NOT DISTRIBUTE

Table 6. Results per sample and laboratory (samples 9 to 16) a

Infant formula powder FOS/GOS based

Child formula powder

Toddler formula powder milk-based

Infant formula powder milk-based

Lab code

Sample 9

Sample 12

Sample 10

Sample 15

Sample 11 0.0346

Sample 14 0.0341

Sample 13

Sample 16

1 2 3 4 5 6 7 8 9

0.303 0.308 0.258 0.272 0.260 0.279 0.271 0.275 0.285 0.279 0.314 0.255

0.297 0.310 0.236 0.259 0.231 0.291 0.256 0.283 0.279 0.268 0.293 0.247

0.221 0.241 0.176 0.205 0.184 0.212 0.212 0.220 0.209 0.233 0.217 0.196

0.227 0.228 0.180 0.208 0.153 0.228 0.196 0.224 0.214 0.221 0.217 0.199

0.302 0.288 0.204 0.277 0.205 0.297 0.250 0.294 0.270 0.276 0.231 0.246

0.309 0.280 0.220 0.286 0.163 0.303 0.248 0.308 0.274 0.285 0.276 0.239

0.0488 b 0.0468 b

0.0258 0.0311

0.0250 0.0308

Unable to report results

0.0293 0.0277 0.0183 c 0.0356 0.0263 0.0316 0.0325 0.0297

0.0327 0.0291 0.0298 c 0.0371 0.0262 0.0327 0.0347 0.0284

10 11 12 13 14

Unable to report results

a Total Fructans results in g/100 g product as consumed (reconstituted or RTF) b Outlier according to Single Grubb’s Test c Outlier according to Cochran’s Test

Blank measurements According to the experience gained through the Single Lab Validation (SLV) and as described in the method, “a blank measurement is only necessary for some matrices containing low amounts of fructans and may be skipped if it has already been established that it is not needed for a given matrix”. As all samples in this study were unknown, participating laboratories made blank measurements for every sample, even if theoretically only infant formula powder FOS/GOS sample with a low fructan content of 0.031 g/100 g could require this measurement. The variations of the blank measurements between the participants were high and it was questionable if the signals in the blanks were of sufficient size to be above LoQ. The mean results with blank measurements subtracted were compared with the mean results obtained without subtracting blank measurements using an unpaired t-test. As can be seen from the p-values depicted in Table 7 , the mean results of the samples are not statistically different, whether the blank measurement is subtracted or not. This demonstrates that for these matrices a blank subtraction is probably not necessary. It may be worthwhile adding a statement in the method to ignore any signals in the blank if they are not greater than 10 x the S/N ratio.