Safeguarding Transactions: Best Practices for Sampling Quick-Frozen Meat

To understand why sampling and testing is crucial, it’s important to consider the potential risks if quality control measures are not strictly followed. this is especially important in petfood production. Whereas humans often have a wide variety of food options to choose from, pets are typically confined to eating what is prepared for them. Contaminated or nutritionally imbalanced petfood can pose serious health risks to pets, including malnutrition, illness, and even death. Thus, accurate sampling and testing are essential to verify that the food meets established regulatory safety and nutritional standards.

To understand why sampling and testing is crucial, it’s important to consider the potential risks if quality control measures are not strictly followed. this is especially important in petfood production. Whereas humans often have a wide variety of food options to choose from, pets are typically confined to eating what is prepared for them. Contaminated or nutritionally imbalanced petfood can pose serious health risks to pets, including malnutrition, illness, and even death. Thus, accurate sampling and testing are essential to verify that the food meets established regulatory safety and nutritional standards. 

Along with nutrition, there are basic business considerations, such as product consistency and confirmation that the product received is aligned with the expected quality. Checking what was bought has been a concern reflected in some of the oldest surviving writing tablets recording grain measures, and it continues to be important today in frozen meat products. 

Of course, the frozen state of the meat helps preserve its freshness and nutritional value until it reaches the processing plant. At this point, thorough testing and quality control processes are conducted to ensure that the ingredients meet the required standards. This includes checking for any signs of spoilage, contamination, or discrepancies in nutritional content. 

Consumers rely heavily on the integrity of these processes to provide themselves and their pets with safe, high-quality food. Understanding the steps involved in sampling and quality control can help food processors make informed decisions and trust the products they purchase. By demystifying the complexities of frozen meat sampling, we aim to highlight the importance of stringent quality assurance practices in protecting the health and well-being of pets and people. 

The purpose of this paper is to underscore the critical importance of proper sampling techniques in the petfood industry, and to advocate for best practices that ensure consistent results for both suppliers and receivers. Many professionals in this field do not receive formal training in sampling frozen blocks during their university education, making it essential to acquire this expertise later in their careers. A key consideration in the sampling process is maintaining the correct temperature of the sample, which is particularly crucial when employing near infrared (NIR) spectroscopy to analyze nutritional content. NIR is a preferred method due to its rapid analysis capabilities, which is vital when expedience is required, such as when trucks are waiting to unload. By following these recommended practices, the industry can achieve greater accuracy, consistency, and reliability in quality control, ultimately safeguarding the proper receipt, processing, and payment for frozen meat products. 

Why Sampling Matters

Sampling is a critical process in ensuring the safety and quality of frozen ingredients for petfood. It involves selecting a representative portion of the whole product to analyze, which allows producers and regulators to assess the overall quality and safety of the batch. Proper sampling helps detect potential issues such as unacceptable ingredient concentrations, contamination, nutritional imbalances, and other defects that could harm the final food products. Without accurate sampling, these issues might go unnoticed, leading to adverse effects on pet health and undermining consumer trust.

Proper sampling is also important because it is used to determine the value of the payment or even possible rejection of a load of product. Making sure the sample accurately reflects the concentrations in an entire pallet, truck load or container is no simple task, and an error of just a few percent can cause significant financial disruption that ripples through the entire system. 

Sampling for fresh/frozen meats is made more challenging because of the scope and volume of material involved. For example, Figure 2 shows a pallet of processed and frozen poultry bound for inclusion in petfood. Each of those blocks are typically about 50 lbs. When frozen meat is edible for bipeds, it is generally packaged in plastic and/or boxes, which changes the look but not the magnitude. Figure 2 gives a good idea of the scope of the problem in a single pallet. A typical truckload might contain 20 to 22 pallets. This is a substantial amount of material, and as such, sampling must be approached with the same level of rigor.

The Theory of Sampling (TOS), as articulated by Kim Esbensen, provides a comprehensive framework for understanding and implementing effective sampling methods. Esbensen emphasizes that the primary goal of sampling is to obtain a representative sample that reflects the entire batch's characteristics. This is crucial in the meat processing and pet food industry, where the quality and safety of products can vary significantly due to factors such as ingredient sourcing, processing methods, and storage conditions. By adhering to the principles of TOS, processors can minimize sampling errors and ensure more reliable quality assessments and accurate payments. In this case, the batch might be a truck or a train load of frozen meat blocks arriving and requiring proper sampling.

Esbensen’s work highlights several key principles that are particularly relevant to frozen meat sampling. One of the central tenets is the importance of correct sample extraction methods. When looking at these frozen blocks of meat, one has already passed by some of the best sampling opportunities. It is better to sample when things are continuously moving, rather than in a stratified sample. Once these are frozen, the differences are also frozen in place. While the product is moving, it is possible to collect a little bit here and there, every so often, and blend that for a representative sampling. Conversely, it is hard to randomly sample a little bit from many of those frozen blocks. 

As such, the supplier is in the best position of representatively sampling the concentration before the product is frozen and loaded. Sampling after freezing leads to more possibilities of error of misrepresentation, and can affect results both on the high side and the low side. Those blocks all look identical, but when you consider that meat comes from individual parts of individual animals, you can easily imagine there is quite a bit of local variability that goes into the entire average for the pallet. For better or worse, sampling frozen blocks is typically done by grab sampling because larger samples can’t be mechanically blended down to smaller ones because of time and product consistency, as well as sanitary concerns. 

The act of freezing changes not only the ease of representative sampling, but also the sample itself. Freezing and rethawing is used to separate materials in a process called zone refining or zone melting. Mixtures tend to freeze and thaw at slightly different temperatures analogous to how they boil at different temperatures. When it comes to meat, the protein molecules are generally very structured and interact with each other. Protein molecules stay static, whereas the water and fat molecules move. So, it can be seen that in meat, the moisture and/or fat migrates a bit towards the top of the block as it is being frozen. This effect is shown in Figure 3. Notice that the top rough edges turned towards the middle of the stacks in Figure 2. The smooth corners are the bottom of the plate freezer. The top of the block is the rough side. The sides are not identical in texture, and they are not identical in composition. Out of the freezer, the sharp lines at the edge of the pallet might become rounded because of melting when the temperature for shipping requirements have not been followed.

In the freezing process, gravity has allowed the fat to migrate up, along with some of the moisture. And the freezing process allows frost on an exposed top of the block. Together, these factors create an unequal sampling situation. It might be very convenient to drill in from a random edge of a block stacked on the pallet, but this has a 1 in 4 chance of being a bad, unrepresentative sample. It is much better practice to pull a block and pick a spot on the face (large side) of the block and sample there. Freezing changes how we need to sample. Again, it is easier to sample representatively before the freezing process. But, once it is frozen, it is essential to use appropriate tools and techniques to avoid introducing biases or contaminating the sample. The top edge while freezing is different than the other ones. 

The way products change and migrate in food processing has been the subject of study and computer simulation. Appendix I contains references to Finite Element Methods (FEM) for evaluating freezing and thawing effects. Of course, one can also just look and see the effect in a home freezer too. It also happens on this industrial scale. The Appendix also contains a reference to moisture changes in meat freezing and thawing. This is a well-known effect, and it impacts the sampling of a fresh, then frozen, then rethawed, sample. 

Moreover, Esbensen’s Theory of Sampling underscores the necessity of adequate sample size and proper homogenization. A sample that is too small, or not properly homogenized, may not accurately represent the entire batch. In the context of petfood ingredients, this could mean missing out on detecting harmful contaminants or misjudging the economic or nutritional profile of the product. By following TOS guidelines, manufacturers can ensure that the sample size is adequate and that the sample is mixed thoroughly, leading to more accurate and reliable testing outcomes. The consistency of the product will dictate the number of sampling points that should be considered to achieve the desired accuracy. 

In conclusion, sampling is a fundamental process in the production and quality control of frozen meat products. By applying the principles of Esbensen’s Theory of Sampling, manufacturers can achieve more accurate and reliable assessments of their products. This not only helps in maintaining high standards of food safety and quality, but also builds consumer confidence in the products they choose for themselves or their pets. To simplify it down to a few key elements, every part of the shipment must have equal probability of being sampled. Random selection is key. To avoid biased sampling, don’t just pick from the edge, where frost, moisture, or fat may have segregated. Aim for a true cross-section that reflects the sample’s overall composition. Some references to Esbensen’s work are detailed in Appendix I. These methods are highly recommended as improper sampling cannot be corrected by using more accurate instrumental techniques later. The sampling bakes in errors and misrepresentations, if done incorrectly.

How To Extract a Sample 

How does one sample a frozen block? Most often a drill and drill bit are used to drill into the frozen block and make a measurement on the cuttings. But there are some practices related to the selection of the drill, the speed of the drill, and the spot to drill that are better than others. First, pick the block to be sampled randomly rather than just taking the top one. Also, don’t just drill into the side, no matter how tempting it might be because it is easily accessible. Edge effects during freezing require a sample to be taken on the large side - the face of the block. Both the top and bottom large sides are considered faces. Either is acceptable for accurate sampling. The edges, however, are not ideal. 

Once the sample is ready for the drill, the drilling spot should be moved about the face, rather than always being directly in the center. One potential course of action would be to divide the block visually into 9 sub blocks, with 3 across and three down. Randomly select one or more of the sub blocks to drill. 

The drill speed and the drill bit are important to ensuring appropriate sampling, with some being better than others. Exhibit 4 displays examples of typical drill bits that could be utilized.

;Using a small diameter drill bit designed for drilling metal (type A) is not ideal, as it tends to rotate faster, generating more heat, and only cutting into a small diameter. Some meat products are simply frozen in larger pieces and the drill bit should cover something wider than an individual piece of an individual animal. One piece of one animal is not going to be representative of the entire pallet, let alone a truckload. The more mixed the components are during sampling, the more representative the sample will be. A larger diameter drill bit, such as type B helps to mitigate some of the problems mentioned above, however, as the drill bit is not being utilized to simply cut a hole in the block, this Forstner-type drill bit, or a hole saw, are not ideal either. Instead, utilizing the drill should help blend the material, while removing a larger diameter.

A spade bit drill, such as type C in Figure 4, is recommended. A commonly used diameter is a 2-inch-wide bit (50 mm) to sample a large area. That is important when you have products that are not already minced into small fragments. In the case of frozen fish, organ meats and other frozen block meats with larger pieces, it is important to cover more area, so the sampling is more of an average of the material. The spade drill should be used rotating at a speed slightly lower than type A or B bits, so it doesn’t heat the product too much. Looking back at the idea of zone refining and moisture migration, the hot drill is going to preferentially melt moisture and fat off the walls of the hole, so our sampling should be kept slow and wide to give everything equal chance. 

Meat products are generally high in moisture content, and bad drilling with a narrow high-speed drill bit will make the moisture content appear even higher, because the water and fat will separate out with temperature changes. With high-speed drilling, there is inevitably a hole that might have moisture and fat coming in from the surrounding wall area, but only protein from the hole itself. This is going to lead to errors in the final analytical data, no matter how good the analytical instrument. The side wall effect is the worst for small holes drilled at high speed, and it is least impactful for large holes drilled more slowly. Picking the right drill bit is part of creating a representative sample, as larger bits create bigger flakes of meat with less heat. 

An alternative to using a wide spade drill is to select a wood auger drill that is still a significant diameter, such as ¾ inch or greater, and use that to go through multiple blocks. As before, the point is to remove the material without heating the walls of the hole. Go slow and deeper with this type of drilling than you would with a wider spade bit. Removing more meat than you need and blending it is preferred to removing an inadequate amount.

Sample Temperature and NIR 

Once a representative sample has been created, it is necessary to measure the moisture, protein, fat, ash, and possibly cartilage and/or additives like carrageenan. Near infrared instruments, such as the DA 7250 shown in Figure 5 can measure these things simultaneously and rapidly, and do so without requiring hazardous chemicals. A general reference to NIR instruments as used in food measurements is listed in Appendix I.

Experts might disagree on if it is better to measure meat by diffuse transmission or diffuse reflection of the NIR radiation. Experts might also disagree on the effects of temperature on an NIR spectrum. With these types of products, our experience with NIR shows that frozen water looks different than room temperature water because the NIR spectrum shows not only the molecules, but also their motion. Liquid water shows a lot of hydrogen bonding. Protein is also affected by hydrogen bonding, but less so than water. Fat is the least affected by hydrogen bonding, and therefore least affected by temperature, but the changing temperature does cause some modest thermal expansion and contraction for fat. Thus, the apparent amount of fat in a fixed volume can appear to change slightly with temperature.

Figure 6 illustrates the appearance of NIR spectra of ground chicken as it is allowed to warm from a firm freeze of 10°F to ambient temperature, which was about 65°F for this measurement.

The frozen chicken has lower overall NIR absorbance, and the peaks shift to longer wavelengths. As the sample warms up and hydrogen bonding starts to come into play, the absorbances shift to shorter wavelengths and become stronger and more pronounced. These are very large effects, and they will change the NIR readings quite a lot. It is generally recommended that you sample either completely frozen or completely thawed meats consistently, and stay away from the in-between temperature range where the sample is part frozen, and part thawed. 

These spectral differences between frozen and thawed samples are so large it is hard to create one NIR calibration that spans the whole range between frozen and thawed. Ideally, the NIR calibration should be rated for roughly the same temperature as the sample. Also, the NIR calibration should have some calibration processing techniques applied to make it more robust to modest temperature changes. Using the same NIR calibration made for samples presented at 70°F on samples presented at 40°F and at 10°F will normally create problems with getting accurate results. NIR calibration sensitivity to temperature can be reduced, but the molecules dramatically change their motions and forms between 10°F and 70°F. A practice is to pick a target temperature and hold the sample to that temperature range plus or minus 10º F. 

How much does the apparent moisture change from 10°F to 70°F in a meat product? For chicken, using one example of an NIR instrument and calibration, the moisture read from 85% moisture hard frozen to 75% at room temperature. The actual product itself did not change. The temperature change alone distorted the NIR readings. Your specific calibration may differ in its sensitivity, or even in the direction of change, but the changing NIR spectrum of meat at different temperatures always happens. It is good to collect NIR spectra at a reasonably consistent temperature. 

Repeatability 

The foundation of science is a repeatable measurement. When sampling something as variable and as economically important as a truckload of frozen meat, it is good practice to measure more than one place, and ideally, there should be consistency in the measurements. Making three measurements is a good minimum to use to see a consistent measurement. With three measurements, you can measure a standard deviation of the readings. Also, with three measurements, you can tell if one is an outlier from the other two. If only two measurements were taken, it would not be possible to tell which is more likely to be right if they disagree significantly. There can be time pressures on measurements, but especially if the measurement has significant financial consequences, it is good to repeat the sampling and measuring process at least three times. Both sampling and the instrument/analysis contribute, and both need to be repeated. One sample run three times is not the same as three separate valid samples. This is absolutely the case if the first reading is outside of the expected range. It is proper to confirm that with at least two other sampling points and measurements.

Conclusion 

In Leo Tolstoy’s novel Anna Karenina, he writes “All happy families are alike; each unhappy family is unhappy in its own way." The same is true for accurate and representative measurements of frozen meats. The happy measurements are random, representative, and consistent in their NIR spectrum, and collected at a consistent temperature. There are a lot of ways to do it wrong by taking shortcuts or not paying attention to details. Sampling is important. Temperature is important. Repeatability is important. 

Summary Tips for Frozen Block Customers 

• Randomly select pallets and blocks for sampling. A random number generator could be utilized to select pallets and blocks, or a list could be prepared of these random selections in advance, and simply mark them off as you use them for an incoming truckload.
• When drilling a sample out, go slow. Start by cleaning the drill bit with an alcohol wipe to keep it sanitary and avoid sample carryover. Use a 2” wide spade drill bit or a comparable wood drill bit and drill slowly. Get a bit more sample than needed for the analysis. Remember, the point is to drill out a sample, not to make the hole. Drill on the face of a block. Proper drilling with tend to create big flakes. When you drill slowly, you obtain a sample from where the bit is and not from heat melting the wall of the hole.
• Gently and consistently warm the sample to your target NIR temperature. We do not want a mix of frozen and tempered parts in the sample. A good target is approximately 50°F +/-10. There should not be any frozen parts or any ice left. A consistent temperature is good for NIR spectra. Using a microwave is a difficult way to get a consistent temperature. A refrigerator or water bath is more appropriate. Take care not to get water in the sample.
• Temperature affects NIR spectra. Use your NIR instrument and calibration with the proper equation made and validated for that temperature range. Fill the cup to the appropriate level, and use standard operating procedures. Level off the sample, and keep the instrument clean. If there is a window surface, ensure that that is also kept clean. Water or fat tends to smear on windows, and it will affect your measurement. One of the most observed issues with bad spectra happens when operators do not fill the sample cup. If you didn’t get enough sample, go get more rather than compromising on the procedure.
• Repeat multiple samples for initial method validation and for making important decisions. Rather than run the same sample three times and averaging, average three different samples. This helps average out the sampling, as well as the temperature and NIR part of the reading.
• Regularly review the USDA Food Safety and Inspection service for professional guidelines and updates on sampling meat products. It is located at https://www.fsis.usda.gov. Look under the sections on inspection, compliance guidance and safe food handling practices. You can also search directly for meat sampling mentions. 

Appendix I – References 

1. Esbensen, K. H. (2004). "Theory of Sampling (TOS) – the Missing Link in Bulk Materials Sampling." Journal of Chemometrics, 18(6), 567-587.
2. Esbensen, K. H. (2010). "Sampling errors and sampling quality assurance." Spectroscopy Europe, 22(1), 8-15.
3. Esbensen, K. H., Paasch-Mortensen, P. (2016). "Introduction to the Theory and Practice of Sampling." International Journal of Food Science & Technology, 51(9), 1939-1948.
4. Tobi Fadiji, Seyed-Hassan Miraei Ashtiani, Daniel I. Onwude, Zhiguo Li, Umezuruike Linus Opara (2021) “Finite Element Method for Freezing and Thawing Industrial Food Processes.” Foods 2021, 10(4), 869; https://doi.org/10.3390/ foods10040869.
5. Coleen Leygonie, Trevor J. Britz , Louwrens C. Hoffman (2012). “Impact of freezing and thawing on the quality of meat: Review.” Meat Science, Volume 91, Issue 2, June 2012, Pages 93-98.
6. Practical NIR Spectroscopy With Applications in Food and Beverage Analysis Subsequent Edition by B. G. Osborne, T. Fearn, P. T. Hindle (Authors). ISBN-13 978-0582099463, Addison-Wesley Longman Ltd. September 6, 1993.