Friend of the barfblog.com Don Schaffner of Rutgers University writes in this guest post:
It’s every scientist’s nightmare to get scooped. You have an idea, find the funding, do the research and write it up, scan the current literature hoping that no one else has that same idea and beats you to publication. I don’t worry about this too much, as there aren’t many people that do what I do, and even fewer that do it the way I do. It’s like Jerry Garcia said: “You do not merely want to be considered just the best of the best of the best, you want to be considered the only ones that do what you do.”
But getting scooped on this particular project was on my mind. Jensen farms had just made a bunch of people sick with listeriosis from cantaloupe, so when Larry Goodridge sent the word out to a bunch of us that we should all submit interlocking proposals to our respective state Departments of Agriculture through state administered Special Crops Research Initiative (SCRI) money for research on Listeria in cantaloupe, I was in. Larry’s been a good friend, and his nose for grant dollars is second to none. Alas, this time is was not to be, and only a few of the states funded the interlocking proposals; New Jersey was not one of them. The good news was that Michelle Danyluk at the University of Florida was one who did get funding, and I’m Michelle’s go-to collaborator for all things mathematical.
Prior to the Jensen Farms outbreak, Listeria in cantaloupe was a theoretical risk, but the outbreak underscored the point that L. monocytogenes was a real risk in these foods. Michelle proposed that her lab would collect data on L. monocytogenes growth in cut melons, and that my lab would build the models. As we suspected from research with Salmonella in cut melons, the organism would multiply rapidly at elevated temperatures but leave the visual appearance of the melons largely unchanged. Listeria’s ability to grow in the fridge made the potential risk even greater.
At this point, I know it’s a race against time, because some other modeler somewhere is thinking the same thing. Uber-technician Lorrie Friedrich got to work collecting data, and I sharpened my spreadsheets (or what ever it is that modelers do when waiting for data). Lorrie soon had the data, and I began making the models, and Michelle started the hard work of writing the manuscript first draft. All three of us work fast in general. In this particular case, we worked even faster. We had a rough draft together when I saw the bad news. My modeling colleagues at the U.S. Department of Agriculture, Agricultural Research Service, in Wyndmoor, PA, near Philadelphia, led by Dr. Lihan Huang published a paper entitled “Growth kinetics of Listeria monocytogenes and spoilage microorganisms in fresh-cut cantaloupe” in Food Microbiology. A quick read of the paper didn’t leave much doubt; we’d been scooped.
Undaunted, we pressed on. The first step was to check their model against ours. It was then that I notice something weird. Their growth curves showed extraordinarily high bacterial concentrations. A quick check showed they were reporting bacterial counts in some cases as ln CFU/g instead of log CFU/g. This explained the high counts in their figures, but even after correcting for this, their model was still giving strange predictions. I couldn’t even get their model to match their data. It was then that I noticed that one parameter in their model was about an order of magnitude different from the same parameter in our model. It might be that slightly different datasets will give model parameters that vary by 25% in some case, but an order of magnitude difference means an error. Once I corrected their parameter (multiplied by 10), their model and our model fell almost on top of one another (See Figure 2 in Danyluk et al.)
In the end, the reviewers were kind to us, and since we not only corrected, then corroborated Fang et al, (2013), we also validated both models against original data for L. monocytogenes growth in watermelon, and honeydew as well as predictions from ComBase Predictor for various assumed water activities and pH values. A contour plot of the model (Figure 3 in the paper) also shows that extended storage of contaminated cut melon slices at even slightly elevated temperatures (in home refrigerators for example) would result in significant risk amplification.
Modeling the growth of Listeria monocytogenes on cut cantaloupe, honeydew and watermelon
Food Microbiology Volume 38, April 2014, Pages 52–55
A recent outbreak linked to whole cantaloupes underscores the importance of understanding growth kinetics of Listeria monocytogenes in cut melons at different temperatures. Whole cantaloupe, watermelon, and honeydew purchased from a local supermarket were cut into 10, 1 g cubes. A four-strain cocktail of L. monocytogenes from food related outbreaks was used to inoculate fruit, resulting in ~10^3 CFU/10 g. Samples were stored at 4, 10, 15, 20, or 25 °C and L. monocytogenes were enumerated at appropriate time intervals. The square root model was used to describe L. monocytogenes growth rate as a function of temperature. The model was compared to prior models for Salmonella and Escherichia coli O157:H7 growth on cut melon, as well as models for L. monocytogenes on cantaloupe and L. monocytogenes ComBase models. The current model predicts faster growth of L. monocytogenes vs. Salmonella and E. coli O157:H7 at temperatures below 20 °C, and agrees with estimates from ComBase Predictor, and a corrected published model for L. monocytogenes on cut cantaloupe. The model predicts ~4 log CFU increase following 15 days at 5 °C, and ~1 log CFU increase following 6 days at 4 °C. The model can also be used in subsequent quantitative microbial risk assessments.