Header graphic for print
Marler Blog Providing Commentary on Food Poisoning Outbreaks & Litigation

Continued Positive Tends in E. coli O157:H7 Prevalence in Federal Sampling – Is the Testing Sufficient and Frequent Enough?

The American Meat Institute (AMI) reported today that new federal data show that E. coli O157:H7 is found in less than one quarter of one percent of ground beef samples, a 72 percent decline since 2000 in ground beef samples testing positive for the pathogen.

n60.jpgThe data are from Food Safety and Inspection Service (FSIS) 2010-year end results of microbiological samples for raw ground beef products analyzed for E. coli O157:H7.  In 2010, there were 29 positive samples out of 11,616 taken in federal plants, no positives of 905 samples taken in retail stores and one positive out of 29 imported samples.  The overall prevalence rate was 0.24 percent.  The year-end data is available here.

According to the FSIS, criteria for sampling and analytical methodology for E. coli O157:H7 testing can be drawn from Agency guidance published in August 2008 Draft Compliance Guideline for Sampling Beef Trimmings for E. coli O157:H7 and Draft Guidance for Small and Very Small Establishments on Sampling Beef Products for E. coli O157:H7, and the Federal Register Notice on FSIS testing methodology [Docket No. FSIS-2008-0007] Revised FSIS Testing Methodology for Escherichia coli (E. coli) O157:H7 The sampling and testing comparison standards for ground beef product and beef manufacturing trimmings are N60 sampling for trim, testing most of the sample (or at least 325 grams) in the lab, and applying a screening or culture method designed and validated for detecting the lowest possible levels of sub-lethally injured E. coli O157:H7 cells.

The real question is if greater sampling amounts (greater than N60) and testing more often would continue to show this positive trend?

  • doc raymond

    testing methodologies used in 2000 are far different than methods for sampling used in 2010. This is comparing apples and oranges, and it fails to mention that industry began testing its own products at its own costs (a very good thing) and that positives are diverted to cooking or rendering. Those positives are not included in the 0.24% AMI is so proud of. And FSIS for years did not test that lot if industry had already found a positive, producing a falsely low number for prevalence. AMI, what is the incidence of O157 if all positives are counted, not just FSIS positives after product has been diverted? That is the true incidence. And, AMI, if you are so proud of 0.24% positive for O157, how can you continue to ignore pleas to make non-O157 STECs an adulterant in ground beef when they test positive in ground beef samples 2.0% of the time?

  • Every year when FSIS releases its microbiological testing data, the meat industry tries to use that data to say that the prevalence of pathogens in meat and poultry products has gone down. I have criticized this inappropriate use of statistics every time for many of the reasons that Dr. Raymond pointed out. These programs were not statistically designed to estimate prevalence and making year-to-year comparisons are like comparing apples and oranges. Considering the relatively insignificant number of samples collected (compared to the millions and millions of product produced) and the uneven distribution of pathogens in meat and poultry products, getting a positive is like finding the needle in the haystack or hitting the lottery. Either you were very lucky to find the hotspot or, more likely, there were so many hotspots that it was easy to find one. Basically, the lower the prevalence rate, the more you have to look to confidently say it isn’t there. Remember, absence of evidence is not evidence of absence. At the current number of samples, the testing program lacks any real power to estimate the prevalence – even assuming it is 1% or 2%. So, either FSIS is consistently very, very lucky or the testing program is biased to underestimate the prevalence. Odds are it is the latter.

  • John Munsell

    The two foundational legs upon which FSIS-style HACCP was built in the mid-90’s was (a) Prevention and (b) Corrective Actions. Prevention was to be accomplished via Hazard Analyses, Good Manufacturing Practices (GMP’s), SSOP’s, etc, which would at least theoretically Prevent pathogenic entrance into our food chain. Then, in those rare instances when pathogens somehow do slip through our abundance of preventive actions and surface in the food chain, meat plants could then implement Corrective actions to prevent recurrences. Both are superlative ideas, and were widely advertised and promoted when we were educated about HACCP in the 90’s. What does this have to do with this blog?
    As Dr. Raymond elucidated above, when preliminary lab reports indicate that the sampled lot is “potential” or “presumptive” positive, the lots are regularly diverted to plants which fully cook the product, providing the necessary Kill Step to sanitize the impacted meat. Such diversion directly benefits public health, as potentially harmful meat does not enter the food chain until all pathogens are killed. However, all preliminary lab test results (potential and/or presumptive) are NOT included in the FSIS or AMI summaries. Realizing that the preliminary results are only “Potential” or “Presumptive” positives, it is true that such preliminary prognoses should not be classified as a “Confirmed” positive, a designation assigned only AFTER the lab has performed final, confirmatory testing. This unwillingness to provide a confirmed, final test result is the underlying beauty of prematurely truncating testing activities! So, if the agency only finds a 0.24% prevalence in E.coli O157:H7, my guess would be that the figure would balloon to over 2.0% if all premature conclusions of lab tests were to be included.
    And think about it: when a plant unilaterally and prematurely concludes lab protocol, the plant is insulated from the potential risk involved with experiencing a “Confirmed Positive” lab result. Thus, the plant’s record is kept clean, circumventing inevitable FSIS enforcement actions if the true incidence of positives were consistently detected at the plant. Now, if all positives were revealed to the agency and to the public, then (AND ONLY THEN) would the source slaughter plant be forced to implement truly effective Corrective Actions, which is one of HACCP’s foundational cornerstones to which I referred in my original sentence above. Since the “hot” meat has been dirverted to a plant which fully cooks, FSIS enthusiastically pats the diversionary plant on the back, stating (1) their HACCP Plan is working because of the detection of unsafe meat, and (2) praises the plant for its proactive willingness to protect public health. Both (1) & (2) are correct, but woefully incomplete. In the midst of all these possibly contaminated lots, the slaughter plant which was the true SOURCE of contaminated meat is totally off the hook, off the FSIS radar, and does NOT utilize any corrective actions. Why should it? The sampled lot was NOT a confirmed positive, mind you. This sly circumvention of revealing the whole truth (by intentional design) enables the source originating slaughter plants to maintain kill floor dressing procedures in a status quo, that is, no need for any changes. Since no corrective actions are initiated at the SOURCE, we are virtually guaranteed ongoing outbreaks, all with full FSIS endorsement.
    I hope this explanation adequately translates meat industry jargon to the ordinary consumer, who otherwise wouldn’t understand or perceive the deception hidden behind the industry’s premature termination of lab test results, and one drawback of diverting meat to a cooking plant.
    I challenge AMI to strongly advise its members to fully conclude all lab testing protocol, and to notify FSIS with all test results. If AMI disagrees, I challenge AMI to explain its opposition to this pro-public health idea.
    John Munsell

  • Just to add a bit more from Barb:
    Can we just replay the Larry King Live show? What was that – a year and a half or two years ago? They are still saying the same thing and it is still an inappropriate use of statistics. How’s this:
    “Every year FSIS releases the results from its microbiological testing programs and every year the meat industry tries to use that as evidence that the prevalence of foodborne pathogens like E. coli O157:H7 have gone down in meat and poultry products. The problem is that these testing programs were not statistically designed to estimate the prevalence of pathogens in meat and poultry products or to make year-to-year comparisons. They were designed to determine if a particular establishment was meeting a particular standard at a particular point in time – period. Pathogens are not distributed evenly in meat and poultry products so getting a positive is about the equivalent of finding a needle in the haystack or hitting the lottery. When a positive does occur, you were either very lucky to find the hotspot or, more likely, there were so many hotspots that it was easy to find one. If anything, these testing programs underestimate the prevalence and, due to the way the sampling is done, making year-to-year comparisons is completely inappropriate.”
    If you need something more concise, let me know. I’m working on a deadline and then traveling next week but I developed a sheet working out the stats the last time around and I’ll see if I can find it later in the week. Bill, are you still in Vegas? Assuming that the true prevalence rate of E. coli in ground beef is 1% or 2% (which I think industry has suggested), I showed you are more likely to hit the jackpot than actually getting the number of positives that FSIS gets; in other words, FSIS is consistently very, very lucky in their testing. I’ll get it updated and circulated as soon as possible.
    Barbara Kowalcyk
    CEO & Director of Research & Public Policy
    Center for Foodborne Illness Research & Prevention
    Office: 724-458-0767
    Cell: 608-347-1227

  • Walt Hill

    When a pathogen hits an incidence level of 0.24%, testing as a control measure is not feasible because even if one assumes (1) that the pathogen is randomly distributed in all products tested and (2) that the level of contamination is above the limit of detection, the probably of contamination still existing and not being detected when all N=60 tests are negative is still about 0.87. Check my math: 0.9976^60 = 0.866.

  • Minkpuppy

    The diversion of preumptive positives to cooking prior to obtaining confirmed results is the convenient little loophole the large plants use to avoid regulatory action. It clearly shows how HACCP has tied the hands of FSIS when it comes to enforcing zero tolerance at source plants.

    It’s ridiculous to assume FSIS results have anything to do with O157 prevalence and I’m skeptical that it’s really telling us how a particular plant is doing in its attempts to control it. The sampling method is too flawed.

    Too many plants rely on the certificates of Analysis they receive from their suppliers. COAs aren’t worth the paper they’re printed on. One lab can have a negative result on the affected lot and another could come back positive for the same lot of meat. FSIS needs to stop acting like a COA makes everything OK.

  • All, what would be the result if we increased testing for O157 to a number greater that N-60? What if FSIS tested more at all points in distribution – from slaughter to grocery? I assume there would be more recalls? There certainly would be more pressure from retailers upstream to provide a non-tainted product? What if all points of distribution where required to post positive test results online along with diverting to cooking?
    What really is the rationale for not labeling non-O157’s adulterants? We know they cause human illness. Several are as virulent as O157’s. AMI and its lackey, Shawn Stevens, best argument is that we need more data. We now have the tests (per FSIS), we just need to implement them in Health Department labs and in the distribution chain.

  • Walt Hill

    Increasing the number of samples past N=60 is statistically an uphill battle. For 4 times the increase, you might get a 2-fold reduction of the fraction lots that test negative but are actually positive. At the current level of 0.24% contamination, taking 240 samples (.9976^240 = 0.562) means that more than half of the contaminated lots would test negative.
    With respect to non-O157 adulterants, the scientific answer is they are adulterants. Politically and economically, just follow the money and the public can take their chances.