demonstrate the need for test methods to be as rapid as possible. However, it’s been said that
a result can be fast or it can be right. Methods that are developed to be fast at the expense of
test sensitivity and type I (false negative) and type II (false positive) errors may be less desirable
than methods that take longer but with greater accuracy.
2.4. Field contamination patterns
2.4.1. Non-homogeneous even with homogeneous contamination event -
Homogeneous contamination of a field is most likely associated with compost or water source
contamination, as these are the only two of the five most likely routes of contamination that are
reasonably likely to contaminate large numbers of plants in the field. Other field inputs, such as
chemical applications, can also lead to this type of contamination if not managed properly.
Even so, such contamination is expected to be at relatively low levels, and pathogens may die
off on the plant surface over time as a result of weather (e.g., heat and desiccation), resulting in
non-homogenous, apparently sporadic distribution.
2.4.2. Sporadic without detectable source –
Likely a more common scenario with typically low level contamination. It may go undetected as
contamination levels are non-homogeneous and potentially impacted by the sampling
scheme/program, loss of sensitivity due to sample compositing prior to enrichment and the
sensitivity of the detection method. However, as noted above, there is likely a non-zero level of
contamination that does not pose a health risk to consumers.
2.4.3. Directional because of known source –
This is an investigative type sampling for root cause analysis as a result of initial finding at a
perceived contamination source. Contamination may still be non-homogeneous and while
subsequent testing may lead to some higher degree of detection, this increase may be
incremental. In this case the product will likely not enter commerce.
2.5. Current industry practices and limitations
2.5.1. Z-patterns, multiple Z-patterns, serpentine
There have been multiple approaches and suggestions to the sampling patterns, including
sampling only the outer perimeters or upon receipt at the processing facility. It remains to be
determined what/if any pattern provides an advantage over others. Differences on what
constitutes a sampling lot could further complicate comparison of results. In one approach, 1
acre sub-lots are sampled and treated independently from the other 1 acre sub-samples. One
positive detection may only lead to disposition of that sub-lot, whereas multiple positives may
lead to disposition of the whole field. In another approach, a single z-pattern may constitute an
entire 5-10 acre field and the result would apply to the disposition of the entire field. While it
seems plausible that narrowing the size of the sampling lot would increase the likelihood of
detection, factors such as those discussed earlier complicate the data. Further statistical
analysis of historical data may help uncover more information. Regardless, the sampling plan
must define a sampling lot and the decision must take in to account the field observations, GAP
assessments and historical data.
Comment [DG2]:
unclear
