Defining and ensuring the microbial quality of agricultural water used to grow, harvest and pack fresh produce is undoubtedly the most complex and contentious issue associated with the Food Safety Modernization Act (FSMA) Produce Safety rule, so much so that the U.S. Food and Drug Administration (FDA) has expressed its openness to simplify agricultural water standards (see “Regulatory Updates”).[1,2]

Agricultural water can serve as a vehicle for pathogens—viral, parasitic and/or bacterial—if the water has been contaminated with feces carrying these pathogens. Feces can be introduced to water systems through runoff (e.g., rain in a pasture that drains into a water system) or direct animal activity (drinking, bird/reptile activity, etc.). Intuitively, you don’t want to apply “dirty” water to fresh produce. But how does one scientifically define “dirty” and demonstrate the health impact of using water of varying quality?

What Is Agricultural Water?
What makes any discussion of agricultural water’s microbial quality complex is that agricultural water can come from multiple sources (e.g., wells or surface water such as ponds, lakes, rivers or reservoirs), and it can be used for multiple purposes (irrigation, application of crop-protection chemicals, frost protection, etc.). Adding to this complexity is the fact that growers typically have only limited options for sourcing agricultural water and often limited or no flexibility regarding alternative agricultural water sources. The FSMA Produce Safety rule doesn’t require testing when water is provided by a public water supply (as long as a certificate of compliance is available), but this option is not available to many farms.[1]

So how does one define and ensure that the agricultural water being used on a farm is “safe and of adequate microbial quality for its intended use,” as required by the Produce Safety rule, subpart E (section 112.41)?[1] In layperson’s terms, this means it is safe to use in a particular manner for a particular fresh produce item; however, how the water is used and the particular item should be considered. Preventing water from becoming contaminated would be ideal, but oftentimes, this is out of a grower’s control. At the heart of the issue is that growers really don’t have a robust preventive control they can use to ensure agricultural water microbial quality, except for water treatment, which is expensive and impractical for the huge volumes of water that may be used on produce farms. This means that produce farmers must rely on monitoring their agricultural water sources and distribution systems to ensure that the agricultural water they use is safe and adequate for its intended purpose. Monitoring in this case is a very weak and reactive preventive control; on top of that, FDA prescribes very specific testing criteria in the Produce Safety rule.[1]

Monitoring and Testing the Microbial Quality of Agricultural Water
Three key variables need to be considered when monitoring and evaluating the microbial quality of agricultural water:

•    The analyte: What are you going to monitor/test for?

•    The test method: How are you going to measure the analyte?

•    Sampling: How will you gather a representative sample? 

The Analyte
Feces are a microbiological melting pot, with myriad organisms that could theoretically serve as indicators of fecal contamination. A brief summary of some terms commonly associated with water testing and fecal contamination is provided for reference (Table 1).

A U.S. Environmental Protection Agency (EPA) report entitled Ambient Water Quality Criteria for Bacteria, published in 1986, estimated the incidence of gastrointestinal illness (hypothesized to result from swallowing water) and correlated these findings with the number of E. coli and Enterococcus found in various bodies of recreational water.[3] A correlation between illness and the levels of fecal coliforms was not observed. This is not surprising, given that the term “fecal coliforms” is a misnomer: When used for environmental testing, common plant shoot and root colonizers (e.g., Enterobacter, Pantoea), plant pathogens (e.g., Erwinia, Pectobacterium) and many other soil and plant residents are likely to be present in the environment and appear as thermotolerant coliforms. This 1986 publication by EPA estimated that a geometric mean of 126 CFU/100 mL corresponded to an illness rate of 8 to 19 highly credible gastrointestinal illnesses per 1,000 persons. How is the likelihood of illness from swallowing contaminated swimming water related to the likelihood of illness associated with consuming fresh produce exposed to water of comparable microbial quality? The answer is unknown, but in the absence of better data, FDA relied on this study when crafting the agricultural water requirements of the Produce Safety rule. FDA also acknowledged that “a formal risk assessment would need to be done to determine an estimated disease burden for consumption of produce exposed to directly applied agricultural water during growing.”[4]

FDA has chosen to use the indicator microorganism generic E. coli as the requisite analyte for agricultural water testing. Unfortunately, generic E. coli is an indicator organism that provides information regarding potential overt fecal contamination, but it is not an index organism. So although it may indicate fecal contamination, its presence and concentration do not correlate well with the presence or absence of human pathogens in agricultural water. In fact, a large body of research in recent years has confirmed that quantitative testing for generic E. coli in agricultural water often has little predictive value regarding the presence or absence of human pathogens for many agricultural surface water sources. Recent research has demonstrated that the relationship between fecal indicator bacteria (generic E. coli and fecal coliforms) and Salmonella is complex and may have limited predictive value. Specifically, only weak linear relationships were shown to exist between biological indicators (E. coli/coliforms) and Salmonella levels (R2 < 0.1) in central Florida surface waters used for agriculture.[5]

When agricultural water will directly contact produce during or after harvest, FDA requires that the water have no detectable levels of generic E. coli. When the water will directly contact produce during growing, FDA uses established microbial water quality criteria with a statistical threshold value (STV) of 410 or less CFU of generic E. coli per 100 mL of water.[1] The rule explains that STV is a measure of variability of water quality distribution, derived as a model-based calculation, approximating the 90th percentile using the lognormal distribution and a geometric mean (GM) of 126 or less CFU of generic E. coli per 100 mL of water. GM is a measure of the central tendency of your water quality distribution.    

FDA chose this analyte in the absence of a better option, as described in response to comment 176 in the preamble of the rule.[1] The agency relied on EPA recreational water standards (e.g., for swimming).3 It’s a stretch to suggest that the likelihood of illness associated with swallowing pool or lake water is the same as the likelihood of illness associated with eating fresh produce irrigated with water of swimming quality.

The Test Method
In the FSMA Produce Safety rule, FDA has mandated provisions for the use of EPA method 1603 for quantifying generic E. coli in agricultural water. This membrane filtration method is highly precise, accurate and sensitive. In fact, this method is probably overly precise, accurate and sensitive, given the huge variability in sample results due to how sampling is done. This method is currently not readily available at contract laboratories commonly used by the produce industry, and the method requires that sample preparation and analysis commence within 6–8 hours of sample collection. This is very problematic for produce growers who live and work in remote locations and cannot get their agricultural water samples analyzed by shipping them overnight to an accredited laboratory.

Especially problematic is that FDA has been slow to recognize other rapid test methods, including those recognized by EPA, as equivalent in precision, accuracy and sensitivity to method 1603. In 2003, when EPA finalized its rule “Guidelines Establishing Test Procedures for the Analysis of Pollutants; Analytical Methods of Biological Pollutants in Ambient Water,” the agency included a table (IA: List of Approved Biological Methods) that provides several options for determining the level of E. coli per 100 mL of water.[6] This includes EPA methods 1603 and 1604, IDEXX Colilert™ and Colilert-18™, and AOAC method 991.15. Some of the methods directly measure CFU, whereas others rely on a well-established statistical determination of the most probable number (MPN) of organisms.

While FDA has not said that one cannot use alternative methods, the agency has not provided a mechanism for farmers and laboratories to assess them. The preamble of the rule lays out guidelines for evaluating an alternative method to establish that it meets the “same level of public health protection” as the rule (bear in mind that quantifying the level of public health protection using the E. coli-based approach is difficult at best, with FDA stating, “We acknowledge the difficulty of associating specific indicator concentrations with specific produce-related health risks.” [p. 74,4281]). An alternative approach must:[1]

•    Rely on peer-reviewed scientific material

•    Detect measurable levels of fecal contamination

•    Be as sensitive to the presence and level of fecal pollution as generic E. coli

•    Be supported by an equally robust and rigorous scientific analysis

•    Be quantitatively demonstrated as equivalent to the FDA-established criteria

•    Not increase the likelihood that the affected produce would be adulterated

It is unclear why FDA will not explicitly approve the use of alternative methods deemed equivalent by another federal agency (EPA). If FDA won’t give the green light to methods vetted by a sister agency, how can a farmer feel comfortable taking advantage of the purported flexibility in the rule?

Farms should use scientifically valid test methods and make decisions around water quality that protect public health. In a study by Kinzelman et al.[7] that evaluated IDEXX Colilert-18™ and Quanti-Tray/2000 against an EPA membrane filtration method (similar to method 1603), the researchers found that out of 234 paired samples, the results were different enough to warrant different decisions on acceptance 10 times. Out of these 10 instances, the IDEXX method would have resulted in a beach closure or water advisory seven times. Thus, in this study, the alternative method was seemingly more protective of public health.

Another challenge in meeting FDA’s requirements for an alternate method lies in the way bacteria are measured. The rule expresses bacteria in CFU. If you remember your microbiology class, these are the individual dots on a Petri plate. If you don’t have colonies growing on a surface, you can’t express bacteria in the CFU/mL unit. However, there are other ways to “count” bacteria, such as using MPN, a well-accepted approach to estimating the number of bacteria present in a population. It is a statistical calculation based on the number of broth tubes, each with different dilutions of a sample, that become turbid due to bacterial growth. Because results are based on statistics, they should be expressed with a confidence interval (generally 95%). Gronewald and Wolpert[8] found that MPN generally overestimated counts compared with direct colony counts (CFU), so that an MPN-based method was generally more conservative. Given the multitude of factors that can influence water testing results, arguing that the precision of CFU is needed is like arguing that you need to weigh an elephant to the fourth decimal point.

Sampling of dynamic agricultural water sources and distribution systems undoubtedly represents the most variability in any agricultural-water monitoring program. Where to sample, when to sample and how much to sample will all affect sample test results.

The rule requires farms to take a specific number of samples over a specific time frame, depending on the source of the water (Table 2). FDA states that these minimum testing frequencies are based on its statistical analysis of the average variability among surface water sources based on published literature and EPA’s use of an estimated standard deviation to establish the recreational water quality criteria.

FDA further stipulates that samples of agricultural water must be representative of the use of the water and must be collected as close in time as practicable to, but prior to, harvest.

When several farms are sourcing from the same water system, must they all follow the prescribed sampling and testing method? Where exactly should the sample be taken and at what time of day (UV light may kill bacteria, and samples taken in the morning versus afternoon may yield vastly different results[9])?

Given the complexity and dynamic nature of agricultural water systems, a lot of research has been done to try to understand how to effectively and efficiently sample agricultural water sources to ensure that produce growers can make informed, risk-based decisions. In general, more sampling (greater volume and number of samples) provides better information to make decisions, but given that the mandated analyte, generic E. coli, has low predictive power regarding the presence or absence of human pathogens, one has to question how much value this monitoring actually does provide.

So What Do You Do with All These Water Data?
The current body of scientific knowledge clearly tells us that the use of quantitative metrics for generic E. coli as an indicator organism to assess the microbial quality of agricultural water is often inappropriate or inadequate. Additionally, as new scientific knowledge becomes available, growers must be able to easily utilize updated and improved testing and sampling methods that can better assess the safety of the agricultural water they use.

What should really matter is whether an event has occurred that could compromise water quality. Because of the way the math in the rule works, it’s possible that a farm could have a long history of favorable test results, have an aberrant spike in detected E. coli and not need to take action because the rolling average would still put the overall program in compliance. Any reasonable farmer would take action because it’s the right thing to do, but technically, the farm could go on using this water. We must therefore question how protective of public health the water testing requirements actually are. We must also question why the test method is so prescriptive and the results so precise (to the point of requiring CFU and not MPN) when there are major fudge factors that obviate any semblance of precision.

Current Industry Practices
The fatal flaw in FDA’s agricultural water testing requirements is the agency’s failure to recognize that, by and large, farms are already testing water but with a different method. Most audits that evaluate compliance with Good Agricultural Practices require some form of water testing. While audits are not required by the regulation, they are often required by customers. For example, the Harmonized Audit, which has recently been benchmarked by GLOBALG.A.P. (the HPSS audit), requires that “water testing shall be part of the water management plan, as directed by the water risk assessment and current industry standards or prevailing regulations for the commodities being grown.”[10] This means that “as required, there shall be a written procedure for water testing during the production and harvest season, which includes frequency of sampling, who is taking the samples, where sample is taken, how the sample is collected, type of test and acceptance criteria. If all water is sourced from a municipal source, then municipal testing shall suffice. The frequency of testing and point of water sampling shall be determined based on the risk assessment and current industry standards or prevailing regulations for commodities being produced [italics added].” Other popular produce audits contain similar requirements, so it’s unclear how FDA’s new requirement will substantially change the end result: Will agricultural water be that much safer?

FDA estimates that implementation of the Produce Safety rule will result in almost 332,000 fewer illnesses every year (p. 74,3571). It’s unclear how the agency calculated this number. The test methods most commonly used by farms of a commercial scale are included in Table IA in 40 C.F.R. part 136. This means that EPA has already evaluated their appropriateness for monitoring E. coli levels. How will switching to a comparable method yield such an increase in public health protection? (Note: Farms that are not of a commercial scale, with less than $25,000 in annual sales of produce, are exempt from the rule and therefore are not required by regulation to test agricultural water.[1])

Fixing FSMA Ag Water Provisions
The current FSMA Produce Safety rule agricultural water provisions have simply missed the mark. The produce industry and FDA have spent an inordinate amount of time contemplating, debating and discussing how to define what constitutes agricultural water that is of adequate microbial quality to use when growing, harvesting and packing fresh produce. The reason so much discussion and debate have occurred is because there simply is not a one-size-fits-all solution, and more data, situation-specific information and knowledge are needed. However, this is not to say that produce farmers should do nothing.

Regulatory Updates

On June 6, 2017, FDA announced that it intends to extend compliance dates for the agricultural water provision of the rule:

The Center for Produce Safety recently published an agricultural water testing colloquium report, focused on evaluating the water testing method:

Additionally, the current provisions need to be amended, because as currently formulated, they are overly prescriptive, cannot be practically implemented by most farms and provide limited value in protecting public health. A more flexible approach that incorporates new science and technology, such as the use of metagenomics, to better characterize agricultural water source and distribution system microbial quality is the only viable solution going forward. Everyone agrees that agricultural water needs to be of appropriate quality for its intended purpose. However, how that is measured does and will vary based on situation specifics of a particular water source and distribution system. Over time, additional FDA guidance to industry, using data derived from agricultural research, should provide the produce industry with situation-specific, safe-harbor guidance as to how one determines whether the microbial quality for agricultural water in a specific situation is or is not adequate for its intended purpose. It is highly recommended that FDA not only reevaluate but also reopen the FSMA Produce Safety rule agricultural water provisions, because as currently formulated, they offer little value to produce growers or consumers in the way of public health protection. FDA knows its agricultural water standards are too complex to implement as is. Therefore, we urge FDA to reevaluate and simplify the current FSMA Produce Safety rule agricultural water provisions to allow growers to put limited produce safety resources to work where they are most effective. 

Jennifer McEntire, Ph.D., is the vice president of food safety & technology for the United Fresh Produce Association.

Jim Gorny, Ph.D., is the vice president of food safety & technology for the Produce Marketing Association and a member of the Editorial Advisory Board of
Food Safety Magazine.

3. Environmental Protection Agency. 1986. Ambient Water Quality Criteria for Bacteria — 1986.
4. Ravaliya, K et al. 2014. “Review of Water Quality Standards in Development of Proposed Microbial Standard in §112.44(c) of the Proposed Standards for the Growing, Harvesting, Packing, and Holding of Produce for Human Consumption.” Memo to the File FR Produce Safety Rule References.
5. McEgan, R et al. 2013. “Predicting Salmonella Populations from Biological, Chemical and Physical Indicators in Florida Surface Waters.” Appl Environ Microbiol 79(13):4,094–4,105.
6. Environmental Protection Agency. 2003. 40 C.F.R. Part 136 of the Federal Register, pp. 43,272–43,283.
7. Kinzelman, JL et al. 2005. “Use of IDEXX Colilert-18® and Quanti-Tray/2000 as a Rapid and Simple Enumeration Method for the Implementation of Recreational Water Monitoring and Notification Programs.” Lake Reservoir Manage 21(1):73–77.
8. Gronewald, AD and RL Wolpert. 2008. “Modeling the Relationship Between Most Probable Number (MPN) and Colony Forming Unit (CFU) Estimates of Fecal Coliform Contamination.” Water Res 42:3,327–3,334.
9. Mudd, D et al. 2012. “Examination of Diurnal Variation at a Non-Sewage Impacted Beach via qPCR and Culture-Based Methods.” J Env Protect 3:1,310–1,317.