The growers, 13 producers or co-packers of broccoli (BroccoSprouts^®) and other sprouts for the company Brassica Protection Products (Baltimore, MD, USA) during the 2001 calendar year, were selected for this review. These growers were selected because they are established companies that grow many different kinds of green sprouts, and they had previously agreed to comply with rigid standards of sanitation. Geographically, they are distributed around the United States with respect to both plant siting and product distribution. In particular, all surveyed growers follow all of the steps specified in the FDA Guidance for Industry 16. Specifically, they surface-disinfect all seeds and perform microbial testing of the spent irrigation water from each batch of green sprouts produced, and they are subject to announced and unannounced third party audits and state and federal inspections of these procedures.

Briefly, surface disinfection of seeds is accomplished by exposing seeds to 20,000 ppm of calcium hypochlorite for 15 min, followed by extensive rinsing to remove residual chlorine. Sprouts are grown in trays or drums which are provided with only light (fluorescent, incandescent and/or filtered sunlight), heat (constant temperature), and clean water (filtered, well, or municipal-chlorinated). A sample of the spent irrigation water, typically 1 L, is collected after 48 hours of sprout growth. Since the sprouts typically are grown for 72 – 120 h, a rapid microbial test permits the sprouters to abort contaminated batches of sprouts prior to packaging, or to shipping and distribution. All sprouts were held until test results were obtained. In cases where presumptive positives were obtained, the refrigerated, unused portion of the 48 hour spent irrigation water sample was used for re-testing as specified by the FDA guidance 16. Typically, the confirmation analysis was performed using a different, more specific method. In at least one situation, the grower was collecting individual samples, and the testing lab was preparing a composite sample for testing. In this case, a confirmation was performed on each individual sample (one water sample per drum).

All growers who contributed data for this study were requested to directly forward their hold-and-release testing results for the calendar year 2001 to Quality Associates Incorporated (QAI; Columbia, MD, USA). Many growers included a description of the green sprouts grown in each batch with these results. In addition to broccoli sprouts, the growers reported growing the following green sprouts: alfalfa, clover, radish, onion, pea, sunflower, and a variety of mixes. Data that was exclusively derived from batches of bean sprouts (e.g. mungbean and soybean) were not used herein. However, in some cases pooled water samples that were tested included water from both green sprouts and bean sprouts, and these test results are included in the analyses.

The raw data utilized were copies of the analytical reports, as they were received by the growers from the contract analytical laboratories that performed the microbial testing. All growers but one used the services of an external microbiological testing laboratory, to which samples could be delivered on the day of collection. One grower had instead set up an in-house laboratory for these analyses and all laboratory notes were provided to us by this grower.

The analytical laboratories generally used test kits designed for simple and quick screening of the samples for contamination by specific bacteria. All samples were screened for Escherichia coli O157:H7 and for Salmonella spp. Most of the testing for E. coli O157:H7 (over 75%) was performed using a test kit identified in the FDA Guidance for Industry 16. Only about 10% of the testing for Salmonella was performed using a test kit identified in this FDA Guidance. Most kits were Association of Official Analytical Chemists (AOAC) Official Methods; one kit was an AOAC Performance Tested Method. Instead of a kit, one laboratory used the official FDA Bacterial Analytical Method 23 for Salmonella testing. The methods used by two laboratories (both for E. coli and Salmonella) and for a third (only for Salmonella) were not specified. The specific methods are listed and referenced in Table 1. All results were entered into a database and data entries were independently verified by an auditor at QAI.

Hold and release testing results representing a total of 3216 samples were obtained from 13 growers. The majority of growers collected at least some of their samples as a composite from several drums. A few growers collected all samples from a single drum. Twenty five of the samples provided were derived from an unknown number of drums. None of these 25 samples resulted in presumptive positives and they are omitted from further reporting of the data, thus leaving 3191 samples from 6839 drums. The distribution of samples based on the number of drums composited is presented in Table 2.

Whereas all growers produced broccoli sprouts, many of the composite samples taken included samples from non-broccoli sprouts (alfalfa in most cases). Presumptive detection of Salmonella spp. or E. coli O157:H7 occurred in 24 of 3191 samples (0.75%) from a total of 6839 drums. As described in the FDA Guidance 16, the detections were considered presumptive and in general, the manufacturers claims for sensitivity of these tests ranged from 70% to 100% and the specificity ranged from 78% to 100% (Table 1). The majority of the tests are reported to have a rate of false positive responses below 2%. In most cases of presumptive positives, the water sample was then reanalyzed. In two of these cases, a follow-up analysis was again positive for Salmonella spp., and the sprouts were destroyed. In a single case a presumptive detection of E. coli O157:H7 was assumed by the grower to be real without re-testing; the sprouts were destroyed. None of these 3 samples included broccoli sprouts. Additional prophylactic measures were taken as outlined in reference 16. Sprouts were held at the growing facilities until microbial testing results were confirmed. There were no instances in which contaminated sprouts were released for distribution.

In order to determine whether detection of presumptive positives was affected by the number of drums pooled for assay, the 24 presumptive positive samples have been examined, relative to the number of drums sampled and pooled in a single assay (Table 3). Assuming that a presumptive positive was caused by a single drum, the frequency of occasions when a presumptive contamination was observed was calculated for each sample class. It thus appears that there is no loss of sensitivity when a single sample represents as many as seven drums although the number of presumptive positives observed in this review was limited. Based upon the Mann-Kendall Trend Test for Small Sample Size 24, there is no trend for the percent of presumptive positives per drum (p ≤ 0.01), between one (single) and 7 (pooled) drums, at the 95% confidence level.