This article reports on a project in which 10 laboratories that conduct forensic glass analysis participated in three inter-laboratory exercises to evaluate the use of a standard method (ASTM 2927-16e1) for the analysis and comparison of glass evidence using LA-ICP-MS.
This study evaluated the rate of misleading evidence (ROME) when blind glass samples were distributed to the participants and asked to compare the glass samples (K vs. Q) and report their findings as they would in a case. Three different databases were used as background populations to calculate likelihood ratios (LRs) and frequency of elemental profile. The first database was composed of 420 vehicle windshield samples, and the analytical data and application of this database is reported in this article for the first time. The second database was provided by the BKA laboratory in Germany, representing 385 casework samples that included an assortment of float glass, container glass, and specialty glasses. The third background database was a combination of both databases. In the first inter-laboratory exercise, the likelihood ratio (LR) calculations result in 34/36 (94.4 percent) correct associations and no false inclusions for all labs. LRs in the second and third inter-laboratory exercises resulted in all participating laboratories correctly associating glass samples originating from the same source (57 comparisons), and all laboratories correctly discriminated glass samples from different sources (167 comparisons). The random match probability of glass samples known to originate from different glass sources was found to be 0.1 percent and in agreement with previously reported values by other researchers. (Publisher abstract modified)