Implementing the Three Class Sampling Plan in Food Quality Control

When assessing the safety of a food batch, simply testing one sample isn’t enough. Microorganisms distribute unevenly within food lots, making it essential to use statistically sound sampling methods. The three-class sampling plan offers a structured, internationally recognized approach to evaluate whether a food lot should be accepted, investigated, or rejected based on microbial contamination levels.

What is a three-class sampling plan? 🔗

A three-class sampling plan is a systematic method for assessing the microbiological quality of food lots by classifying samples into three categories: acceptable, marginally acceptable, and unacceptable. Unlike the simpler two-class plan that only uses a single limit, the three-class approach incorporates two microbiological limits-designated as m and M-to create a more nuanced evaluation of food safety.

The International Commission on Microbiological Specifications for Foods (ICMSF) originally developed this approach in the 1970s, and it was adopted by the Codex Alimentarius Commission in 1981 as the standard framework for microbiological sampling in international food trade.

Understanding the key parameters 🔗

Every three-class sampling plan relies on four essential parameters that work together to determine lot acceptance:

The sample size (n) 🔗

The parameter n represents the number of independent sample units randomly selected from the lot for testing. The value of n typically ranges from 5 to 60, depending on the risk level associated with the product. Higher-risk foods require larger sample sizes to increase the probability of detecting contamination.

The acceptable limit (m) 🔗

The m value represents the microbiological concentration that separates good quality from marginally acceptable quality. According to Hong Kong’s Centre for Food Safety, this limit commonly reflects the upper boundary of what can be achieved through good manufacturing practice (GMP). Samples with counts at or below m are considered acceptable.

The unacceptable limit (M) 🔗

The M value marks the concentration beyond which the level of contamination becomes hazardous or unacceptable. Any sample exceeding this limit automatically renders the lot unacceptable, regardless of other test results. The M value is typically set based on expert judgment regarding safety thresholds rather than arbitrary statistical cutoffs.

The acceptance number (c) 🔗

The parameter c specifies the maximum number of sample units that can fall into the marginally acceptable range (between m and M) while still allowing the lot to be accepted. If more than c samples exceed m-even if none exceed M-the lot is rejected.

How lot acceptance decisions work 🔗

The three-class plan evaluates each sample unit and assigns it to one of three categories based on its microbial count:

Acceptable: Count is at or below m
Marginally acceptable: Count is above m but at or below M
Unacceptable: Count exceeds M

A lot passes inspection only when two conditions are met: no sample exceeds M, and no more than c samples fall between m and M. For instance, with a plan where n=5, c=2, m=10⁵/g, and M=10⁷/g, five samples are tested. The lot will be rejected if any sample exceeds 10⁷/g or if three or more samples exceed 10⁵/g.

Practical example in food quality control 🔗

Consider a ready-to-eat salad manufacturer testing for Staphylococcus aureus. Using ICMSF recommendations, they might apply a three-class plan with n=5, c=1, m=10² CFU/g, and M=10⁴ CFU/g.

If testing five sample units yields counts of 50, 80, 200, 60, and 75 CFU/g, one sample (200 CFU/g) falls in the marginally acceptable range. Since only one sample exceeds m and c=1, the lot passes. However, if two samples exceeded 10² CFU/g, or if any single sample exceeded 10⁴ CFU/g, the entire lot would be rejected.

Choosing between two-class and three-class plans 🔗

The selection of a sampling plan depends on the nature of the microorganism and the associated risk. The Centre for Food Safety guidance notes that two-class plans are preferred when the organism of concern should not be present in the food at all-such as Salmonella in ready-to-eat products. Three-class plans are more appropriate when the presence of some microorganisms is tolerable up to certain levels.

Three-class plans offer particular advantages for organisms where low levels pose minimal risk but higher concentrations become dangerous. Research published in Food Control suggests that three-class plans can effectively detect and reject lots with rare but high contamination levels while accepting those with only sporadic moderate contamination.

Operating characteristic curves and plan performance 🔗

The effectiveness of any sampling plan is evaluated through its operating characteristic (OC) curve, which shows the probability of accepting a lot based on the actual proportion of defective units. For three-class plans, this becomes an OC surface, plotting acceptance probability against both the proportion of marginally acceptable and defective units.

The stringency of a plan can be adjusted by modifying n and c values. Increasing the sample size while keeping the c/n ratio constant creates a steeper OC curve, improving the ability to discriminate between acceptable and unacceptable lots. The ICMSF publications provide detailed tables showing acceptance probabilities for various combinations of defective and marginally acceptable proportions.

ICMSF case classification system 🔗

The ICMSF developed a 15-case classification system that relates sampling plan stringency to the severity of the hazard. Cases range from 1 (no direct health hazard, only quality concerns) to 15 (severe direct health hazard with potential for increased risk during handling). Foods in higher case categories require more stringent sampling plans with larger sample sizes and lower acceptance numbers.

For example, case 8 plans with n=5, c=1 might apply to organisms like S. aureus in certain products, while case 11 plans with n=10, c=0 would apply to Salmonella testing where no positive samples can be tolerated.

Limitations and proper application 🔗

While three-class sampling plans provide valuable risk management tools, they have inherent limitations. The Codex Alimentarius guidelines emphasize that well-designed sampling plans define the probability of detecting microorganisms, but no plan can ensure the complete absence of a particular organism.

Several important considerations apply to proper use:

Random sampling is essential. The validity of any sampling plan depends on selecting sample units independently and randomly from the lot. Biased sampling invalidates the statistical properties of the plan.

Resampling changes plan characteristics. Taking additional samples after unfavorable results-sometimes called the “resampling syndrome”-changes the operating characteristics of the plan and increases the probability of accepting poor-quality lots.

Testing alone isn’t sufficient. The Codex Alimentarius recognizes that microbiological testing of finished products alone cannot guarantee food safety. Sampling plans should complement, not replace, preventive approaches like HACCP and good manufacturing practices.

Regulatory applications worldwide 🔗

Many national and international regulatory bodies have adopted three-class sampling plans into their food safety frameworks. The microbiological specifications used by food authorities often incorporate ICMSF principles, applying them to import inspections, domestic surveillance programs, and verification of food safety management systems.

To enhance food safety and improve quality, regulatory agencies may adopt more stringent microbiological limits by decreasing m and M values or by adjusting c and n parameters to increase sampling plan stringency. These adjustments allow authorities to tailor sampling approaches to local conditions and specific food safety objectives.

What do you think? How might the balance between sampling costs and food safety assurance affect a manufacturer’s choice of sampling plan parameters? And in your experience, what challenges arise when implementing statistical sampling plans in real-world food production environments?

References

1. https://www.ncbi.nlm.nih.gov/books/NBK216671/
2. https://www.fao.org/4/y1579e/y1579e04.htm
3. https://www.cfs.gov.hk/english/food_leg/food_leg_mgref_annex.html
4. https://www.sciencedirect.com/science/article/abs/pii/S0956713524002615
5. https://www.sciencedirect.com/topics/food-science/microbiological-criterion
6. https://www.eurofinsus.com/food-testing/resources/microbiological-specifications-in-food-operations/