Estimating numbers of Salmonella cells in chicken rinse samples is expensive, so samples are usually tested only for the presence of Salmonella, with results reported as prevalence, or the overall percentage of positive samples. Several baseline studies have included estimates of numbers of cells in chicken carcass rinses, however, and those results can be used to examine the relationship between Salmonella prevalence and numbers of Salmonella cells in chicken HACCP samples.
The 1994-95 Food Safety and Inspection Service chicken baseline sampled almost 1,300 chilled chicken carcass rinses, with a 30 ml aliquot of the total rinse tested for Salmonella and an additional 33.3 ml most probable number assay performed on positive samples only. A histogram of the numbers of Salmonella in positive samples in that baseline is shown in Figure 1.
Risk of illness ignored in system of sampling
Approximately 87 percent of positive rinses contained fewer than 120 cells (estimated cells per ml of rinse multiplied by 400 to get the number of cells in the entire 400 ml rinse used to sample the carcasses). The most contaminated carcass rinse contained an estimated 280 cells per ml, or 112,000 cells in the entire rinse. The lowest range in the histogram is samples containing less than 12 cells, but all samples in that group (41.9 percent of positive samples in Figure 1) actually tested completely negative in the most probable number assay. In this system, any rinse that contained only one cell and tested positive carried regulatory weight equivalent to the rinse that contained 112,000 cells, so risk of illness is ignored.
Carcass samples with high numbers of cells in the 400 ml rinse are likely to test positive in a 30 ml subsample, but probability enters the picture when the total rinse contains relatively small numbers of Salmonella. A 30 ml subsample of a 400 ml rinse with only a single cell has a 7.5 percent probability of containing that cell. There have to be nine cells in the total rinse before a 30 ml aliquot has a 50 percent likelihood of containing at least one cell, and perhaps surprisingly, there have to be 39 Salmonella cells in the 400 ml rinse to reach a 95 percent probability that a 30 ml subsample contains at least one cell.
Probability can also be calculated for the 33.3 ml most probable number assay performed on positive samples. If the original 400 ml rinse contained four Salmonella cells, for example, the 30 ml subsample that was tested could have contained 4, 3, 2, 1 or 0 cells. The associated probabilities can be calculated along with the odds of positive results in the most probable number assay. Cumulatively, a 400 ml rinse containing four Salmonella cells has a 0.0636 percent probability of a positive most probable number result. A spreadsheet was constructed to repeat these calculations for rinses containing up to 50 cells.
Salmonella prevalence versus numbers
Figure 2 shows the relationship between Salmonella prevalence test results and Salmonella numbers in the original rinse sample for the first two sections of Figure 1 (samples containing from 1 to 120 cells in the original 400 ml). To approximate the distribution of results in Figure 1, a high number of samples containing low numbers of Salmonella had to be assumed, agreeing with numerous studies that have reported testing a larger portion of rinse liquid finds many more positive samples. The red line is the frequency of samples containing numbers of Salmonella cells in 400 ml as indicated on the horizontal axis. The blue line shows the frequency of positive 30 ml subsamples, and the green line shows the frequency of 33.3 ml most probable number assays with at least one Salmonella-positive result.
The sections labeled “A” in Figure 2 are false negative results that occur when the 30 ml subsample does not contain any Salmonella despite the presence of Salmonella in the total rinse. The sections labeled “B” and “C” are Salmonella positive by the 30 ml test, but have a negative result in the most probable number assay and are indicated in the “less than 12” category in Figure 1. “D” and “E” are most probable number assays with a positive Salmonella result, included in the “12-120” category in Figure 1. The “C” samples contain more than 12 Salmonella cells, but test negative in the most probable number assay and are included in the “less than 12” category, while the “D” samples contain fewer than 12 cells, but test positive in the most probable number assay and are included in the “12-120” category.
Salmonella prevalence testing does not reflect risk of illness
The lesson from this calculation is that prevalence testing is more complicated than is usually assumed. Testing 30 out of 400 ml produces many false negative results, but most probable number assays done on the positives produce skewed results as well. The mean number of cells in the 400 ml rinse was 56 according to the 1994-95 baseline, but that is an overestimate because of all of the false negative and “less than 12” results (negative most probable number tests are not included in the average). Rinse samples also do not include Salmonella attached to carcasses. The overall result is that HACCP Salmonella prevalence testing has no relationship to the actual risk of human illness associated with Salmonella in chicken.
