In response to the New Performance Standards for Salmonella in poultry announced by Food Safety and Inspection Service in May 2010, several comments pointed out that Salmonella prevalence in chickens had been reduced by about 60 percent since the introduction of HACCP without any change in the rate of human salmonellosis. There was no reason to believe, according to critics of the new standards, that a further lowering of the Salmonella standard for chickens would produce any reduction in human illness.

In its May 2011 response to comments in the Federal Register, FSIS argued that it was not appropriate to compare FoodNet salmonellosis rates to Salmonella prevalence trends in individual commodities, because a reduction in salmonellosis due to one commodity might be obscured by an increase in another commodity. FSIS asserted it was possible, however, to analyze trends by comparing FoodNet serotype data to serotypes isolated from chicken HACCP verification samples. FSIS appears to discount the possibility that a decrease in a serotype from one commodity might be offset by an increase in the same serotype from another commodity.

Estimating Salmonella prevalence in raw products

In 2012, the Data Coordination Committee of FSIS ruled that it is not possible to estimate Salmonella prevalence in raw products from Salmonella HACCP verification testing because of methodological and sampling issues. Intact raw products failed three of eight requirements and raw ground products failed five of eight requirements for HACCP verification samples to yield a valid estimate of national prevalence. Serotype data suffer from the same statistical problems.           

The ruling by the committee was not completely unexpected because FSIS publications in recent years have cautioned against use of HACCP data to look for trends in Salmonella prevalence in raw chicken and other commodities. Changes in Salmonella sampling policies introduced in 2006 make comparisons to earlier data "inappropriate" and "prevent valid comparisons over time." The last 26 issues of the Quarterly Progress Reports on Salmonella Testing of Selected Raw Meat and Poultry say that the "percentage of positive samples is no longer useful as an indicator of trends."

FSIS uses the same Salmonella verification samples to identify serotypes found in chicken. The FSIS website presents quarterly serotyping results with statements such as, "Data reported here are not intended to be reflective of national trends in prevalence of serotypes," and, "The collection of serotype data is not designed for trend analysis."

Despite these valid cautionary statements about the limitations of Salmonella prevalence and serotype data, FSIS argues that a weight of evidence approach permits a comparison of FoodNet salmonellosis rates to Salmonella serotype prevalence trends in individual commodities. The FSIS website has an analysis comparing Salmonella serotype incidence in FSIS-regulated products and human salmonellosis cases. The document compares FoodNet salmonellosis numbers with broiler chicken HACCP samples for serotypes Enteritidis, Typhimurium, Heidelberg, and two others using data from HACCP verification samples from 2000 to 2009, thus crossing the 2006 trend barrier.


Enteritidis risk

For Enteritidis, FSIS reports a 0.81 correlation between relative risk indicated by FoodNet cases and isolates from broiler chicken HACCP samples. FSIS serotype reports from 2006 to 2011 show a mean percentage of Enteritidis-positive samples of 1.70 for the winter months and 0.94 for the summer months. The peak in FoodNet Enteritidis cases is consistently in the summer months. "The Agency recognizes that serotyping information alone does not provide definitive associations," according to some FSIS serotype reports. If serotype data are available by month, with further characterization by gel electrophoresis and antibiotic resistance patterns for both commodity and human isolates, why are gross yearly numbers being compared?

Typhimurium, Heidelberg risk

For Typhimurium and Heidelberg, FSIS compares relative risk from FoodNet against a two-year moving average of volume-weighted percent isolations of those serotypes from HACCP samples. The 0.57 correlation for chicken is not significant, but the correlation for ground turkey is 0.86. Using a two-year moving average likely gave higher correlations than using the actual numbers.

The original comment about HACCP not reducing the rate of human salmonellosis is still accurate, however. The accompanying graph shows the rate of Typhimurium and Heidelberg cases reported by state health departments for the last 25 years divided by national population estimates. Both serotypes show a variable but steady decline in cases since 1986, with no indication whatever of any change caused by introduction of HACCP in 1998. Enteritidis is more variable, but also shows no change caused by HACCP.

Despite all of the Salmonella data collected by different agencies, there is no clear evidence that HACCP has reduced the incidence of human salmonellosis or that a further tightening of HACCP standards will produce an improvement in the future.