How many familial relationship testing results could be wrong?

Relationship testing was the first application of DNA-based technologies used for forensic purposes [1].Indeed, DNA testing has been the gold standard of the forensics applications and greatly impacts society and individuals alike.According to American Association of Blood Banks (AABB) annual reports [2], about 400,000 relationship testing cases are performed each year within the AABB accredited laboratories residing mostly in the United States.This number is likely an underestimate; some laboratories did not report their numbers to AABB, and some laboratories provide services without accreditation.China has a similar number of relationship testing cases as that of the US, and the number of cases has been increasing due to recent national policy change [3].DNA testings also are applied globally to address issues of fraud and abuse that arise with international migration and refugee family reunions [4].Recently, US Immigration and Customs Enforcement established DNA testing sites at the southwest border to identify fraudulent familial relationships claimed by immigrants, which may test tens of thousands of cases per year [5].In addition, many criminal cases and missing person identification cases rely on relationship testing.As of today, it is reasonable to estimate that more than 15 million relationship cases have been tested globally.With so many cases, it is not uncommon to observe some cases that may have either included unrelated individuals as a certain relationship (i.e., false positive) or excluded truly related individuals as unrelated (i.e., false negative) [6].A kinship relationship is usually evaluated by comparing the likelihoods of observing the genetic data given 2 alternative hypotheses (i.e., likelihood ratio [LR]): an individual is related to another individual in a defined relationship versus the 2 individuals not related.The higher the LR, the more supported is the proposed relationship.Additionally, the lower the LR (typically <1), the more support there is for the unrelated hypothesis.To render a conclusion about the purported relationship, some laboratories assign a minimum threshold to the LR to support an "inclusion" or "related."The threshold can vary among laboratories.According to the AABB relationship testing annual report in 2013 [2], 65% of AABB accredited laboratories used an LR threshold of 100 for standard trio cases (i.e., biological mother, child, and alleged father), and 75% of the laboratories used the same LR threshold for parent-child cases (i.e., duo case).Most laboratories report only the LR obtained without a threshold for full-sibling, half-sibling, and other reconstruction cases, although about 20% of the laboratories use an LR threshold of 100 for such cases.Other thresholds (e.g., 10 and 1,000) may be used depending on the proposed relationships [2].Different laboratories also may implement different conclusion policies for "inclusion," "exclusion," and "inconclusive."Unfortunately, there are no readily available surveys or widely accepted guidelines as to what is deemed "inconclusive" and whether decisions are impacted differently as they approach thresholds, especially for cases with a relatively high LR in an inconclusive range (e.g., 0.01 < LR < 100) but lower than the upper threshold.Although some