To quote one of the founding fathers, “In this world nothing can be said to be certain, except death and taxes.” Though not a microbiologist by trade, Benjamin Franklin’s wise words resonate all the way to the interpretation of your microbiology test results. It is academically and universally recognized that no microbiological measurement is perfect due to statistical and practical uncertainty. In fact, acknowledging the uncertainty of a measurement is as important as the measurement itself.
Uncertainty is even more complex in food microbiology due to the particulate nature of bacteria and their ability to reproduce by binary fission. This results in localized pockets of higher concentrations of bacteria where each individual represents a unique variable entity. Consequently, there is an uneven distribution of microbes even in well-mixed samples that create problems not only for test methods but sampling in order to get a meaningful result for the batch. The working group of the International Laboratory Accreditation Cooperation states “it is virtually impossible to know the exact microbial concentration in any sample, natural or artificial.”
The vagaries of microbial measurement are often conveniently forgotten, resulting in unreasonable expectations of both laboratories and the methods deployed. So what do microbiological test results actually mean? What can be expected and do expectations apply equally to both product and environmental samples?
Food products are generally well controlled and manufactured to a consistency where microbial specifications are established. Conversely, there are no agreed standards for microbes for environmental surface samples that are less controlled and more variable. Each facility is expected to do “the best it can” for monitoring cleaning processes due the uniqueness of each manufacturing facility. Thus food manufacturers strive for high hygienic standards to protect their products, brands, and ultimately consumers.
Sources of Variation and Considerations
The unit of measurement for the enumeration of microbes is a colony forming unit (CFU) derived from plate count methods. This technique has remained largely unchanged since the pioneering days of Pasteur and Koch in the 19th century. It is defined as “a rough estimate of the number of viable bacteria or fungal cells in a sample” because it relies on the false assumption that each colony is derived from a single bacterium. Microbes exist as clumps or chains and are often difficult to separate into single cells. Hence, there is a large natural variation in CFU results from plate counts particularly if single replicate samples are used and single tests are conducted.
There are several steps in the plate count method where additional variation can be introduced. To obtain the optimum number of colonies for counting (30 to 300 CFU), dilutions of the sample have to be prepared. Since the distribution of microbes in the sample is not uniform, each series may produce different numbers of CFUs. More variation occurs if there are fewer than 30 colonies per plate. The normal expected variation from plate counts is typically 0.2 to 0.5 Log units, so for a target 1,000 CFU (Log 3.0), an actual result can be anywhere between 300 to 3,000 CFU and still be considered microbiologically equivalent.
Such variation is well known and regularly examined among accredited testing laboratories. Under the proficiency testing scheme, laboratories using standard methods are provided with several replicates of stable, homogenous samples. Results are expected to show a 10 fold (1 Log) variation between laboratories. Sometimes this variation is exceeded by >2 Logs for plate counts such as coliforms or Enterobacteriaceae.
Mathematical models can be applied to statistically assess confidence of results. Measurement Uncertainty is used to calculate the dispersion of the values attributed to a measured quantity. The uncertainty reflects the doubt in the result of the measurement. In the case of a standard method for total bacterial count in milk, this has been calculated as 39.6 percent, i.e. the “true value” of the obtained result (within 95 percent confidence limits) can be expected in a range ±39.6 percent of the result. This means the actual value is not known for certain, and for a sample expected to contain 10,000 CFU, the true value lies somewhere within the range 6,000 to 14,000 CFU on 95 percent of occasions but can also be outside this range 5 percent of the time.