I like to know how to sample a finished product or ingredient so that the sample to be tested is representative of the product as a whole so it will increase confidence in subsequent test result. This is needed to verify a particular finished product lot or incoming ingredient lot is allergen free.
Sampling is not a simple process of looking up a sample size in a table. There are many factors that influence how you develop a sampling plan. When I develop a QA program, I always try to develop the program to answer a specific question or develop a null hypothesis. Once I have framed the question, I can then develop a sampling plan to help develop the answer.
It appears that the question you would like to ask is the following:
• Is an allergen present in either a lot of finished product or in a lot of ingredient?
This question deals with an attribute issue.
In developing a plan, one needs to take into account a number of statistical assumptions including the following:
• Is the process relatively stable? In statistical process control terms, the process is rarely affected by special or assignable causes of variation. The following is an alternative way to describe a stable process. Is the allergen evenly distributed in the lot or can the allergen be concentrated in one part of the lot? Answering this question helps defines the unit.
• A random sampling plan must be used to select the units to be tested.
• A unit must be defined. The unit must either possess the characteristic or not possess the characteristic. The presence of the characteristic makes the unit defective. Many times in food sampling, a unit may be difficult to define.
• A test must be available that can determine if the unit contains the characteristic. It is permissible to test a portion of the unit as long as long as that portion of the unit correctly identifies whether the unit is or is not defective.
• A sampling plan must be developed in which the units will be collected. Every unit must have an equal chance of being selected for analysis (random sampling).
• The number of units that possess the characteristic must be small (less than 10%) as compared to the number of units that do not possess the characteristic. The removal of the number of units for analysis cannot affect the portion of defective units in the lot.
• The number of “units” in the lot does NOT affect the sensitivity of the attribute sampling plan.
• The portion of units that are defective is critical to the sampling plan. If the portion of defective units declines there needs to be an increase in the number of units sampled to ensure that the sensitivity or power of the sampling plan does not change.
• The total number of units sampled is critical for the sensitivity or power of the sampling plan. The power increases with an increase in number of units sampled and tested to determine if they are defective.
• If the portion of defective units can be estimated, it is possible to calculate the power of the sampling plan using the binomial distribution. Likewise, if a sampling plan is selected, it is possible to calculate the power of the sampling plan for a specific proportion of defective units.
• Need to define the confidence level that is desired to determine whether a lot contains or does not contain the allergen.
• Acceptance number. The smaller the acceptance number, the less of a risk the lot will contain units that are defective. The smaller the acceptance number the more sensitive the sampling plan.
The alternative method is to develop a QA system based on the concepts of process control. A classical approach is to use HACCP.
John G. Surak, PhD
– Providing food safety and quality solutions –
Additional ASQ resources:
Statistical Process Control for the FDA-Regulated Industry
by Manuel E. Peña-Rodríguez
Abstract: The focus of this book is to understand and apply the different SPC tools in a company regulated by the Food and Drug Administration (FDA): those that manufacture pharmaceutical products, biologics, medical devices, food, cosmetics, and so on. The book is not intended to provide an intensive course in statistics; instead, it is intended to provide a how-to guide about the application of the diverse array of statistical tools available to analyze and improve the processes in an organization regulated by FDA.
This book is aimed at engineers, scientists, analysts, technicians, managers, supervisors, and all other professionals responsible to measure and improve the quality of their processes. Although the examples and case studies presented throughout the book are based on situations found in an organization regulated by FDA, the book can also be used to understand the application of those tools in any type of industry.
Readers will obtain a better understanding of some of the statistical tools available to control their processes and be encouraged to study, with a greater level of detail, each of the statistical tools presented throughout the book. The content of this book is the result of the author’s almost 20 years of experience in the application of statistics in various industries, and his combined educational background of engineering and law that he has used to provide consulting services to dozens of FDA-regulated organizations.
The Certified HACCP Auditor Handbook, Third Edition
by ASQ Food, Drug, & Costmetic Division
Abstract: This handbook is intended to serve as a baseline of hazard analysis critical control point (HACCP) knowledge for quality auditors. HACCP is more than just failure mode and effect analysis (FMEA) for food: it is a product safety management system that evolved and matured in the commercial food processing industry allowing food processors to take a proactive approach to prevent foodborne diseases. Both the FDA and the USDA have embraced HACCP as the most effective method to ensure farm-to-table food safety in the United States.
This handbook also assists the certification candidate preparing for the ASQ Certified HACCP Auditor (CHA) examination. It includes chapters covering the HACCP audit, the HACCP auditor, and quality assurance analytical tools.