7
Statistical Sampling Issues in the Control of Seafood Hazards

ABSTRACT

A statistical evaluation of Food and Drug Administration acceptance sampling plans was made by the committee. This chapter contains that evaluation for each of the published plans.

Based on this evaluation, the committee concludes that seafood safety should be controlled by instituting requirements for the suppliers, rather than by more frequent testing or larger sample sizes. Suppliers should be required to employ a Hazard Analysis Critical Control Point system that takes into consideration the source and condition of live animals and focuses on public health issues in handling and processing, rather than on quality control concerns.

INTRODUCTION

The committee reviewed the Food and Drug Administration (FDA) sampling inspection procedures for seafood to evaluate the statistical methods used and their appropriateness for public health. Because more than 50% of the seafood consumed in the United States is imported, it likely comes from many parts of the world and therefore is very heterogeneous in nature. This poses a number of problems in terms of sample selection and inspection. The purpose of this chapter is to address these problems and to make recommendations regarding their solution.

Because of the large volume of seafood consumed, any inspection system would have to involve sampling. Even if each boatload of fish were inspected, it would still not be reasonable to inspect each fish. Furthermore, much of the inspection is of a destructive nature. Therefore, it is important to investigate the statistical properties of the sampling procedures being used, as well as of any proposed procedures.

Development of an appropriate safety/quality control plan requires at the outset a set of difficult decisions. The planner must acknowledge that no sampling plan



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 267
Seafood Safety 7 Statistical Sampling Issues in the Control of Seafood Hazards ABSTRACT A statistical evaluation of Food and Drug Administration acceptance sampling plans was made by the committee. This chapter contains that evaluation for each of the published plans. Based on this evaluation, the committee concludes that seafood safety should be controlled by instituting requirements for the suppliers, rather than by more frequent testing or larger sample sizes. Suppliers should be required to employ a Hazard Analysis Critical Control Point system that takes into consideration the source and condition of live animals and focuses on public health issues in handling and processing, rather than on quality control concerns. INTRODUCTION The committee reviewed the Food and Drug Administration (FDA) sampling inspection procedures for seafood to evaluate the statistical methods used and their appropriateness for public health. Because more than 50% of the seafood consumed in the United States is imported, it likely comes from many parts of the world and therefore is very heterogeneous in nature. This poses a number of problems in terms of sample selection and inspection. The purpose of this chapter is to address these problems and to make recommendations regarding their solution. Because of the large volume of seafood consumed, any inspection system would have to involve sampling. Even if each boatload of fish were inspected, it would still not be reasonable to inspect each fish. Furthermore, much of the inspection is of a destructive nature. Therefore, it is important to investigate the statistical properties of the sampling procedures being used, as well as of any proposed procedures. Development of an appropriate safety/quality control plan requires at the outset a set of difficult decisions. The planner must acknowledge that no sampling plan

OCR for page 267
Seafood Safety requiring less than 100% inspection can provide 100% assurance that no nonconforming items will pass the inspection process. The definitions of nonconformance and appropriate quality/safety goals involve careful balancing of the costs and utility of two types of errors: A type I error occurs if a lot that is satisfactory is erroneously rejected, causing an economic loss to the supplier. A type II error occurs if a lot that is unsatisfactory is erroneously accepted, causing economic and safety risks for both the supplier and the consumer. As indicated by this phrasing, the two types of errors are not of equivalent value. A decision on their relative values and its consequences for sampling involve serious questions of social policy. The failure to develop formal sampling plans based on these risks may, in some cases, result from the reluctance to face these questions explicitly. SURVEILLANCE AND COMPLIANCE SAMPLING Two types of sampling procedures, which are named for their application, are currently used by the FDA in its inspection of fishery products: surveillance samples and compliance samples. The purpose of surveillance samples is to ascertain the overall characteristics of seafood, whereas the purpose of compliance samples is to determine whether a suspected lot meets regulatory requirements. For either type of sample, there is a danger that many contaminated fish will pass unnoticed because most lots are not inspected at all, and for those that are, the sample is extremely small. Large samples are not practical because of the destructive nature of most testing. In many tests, the samples must be sent to a laboratory for evaluation. In these cases, unless the product is frozen, the lot may be consumed before the results are obtained. Thus, a possible health hazard may be discovered only in time to prevent the distribution of future lots from the same source. If sufficient control could be maintained on the suppliers of seafood, there would be little need to inspect their products. Such control is not feasible, however, because of the diversity of producers and raw products. In spite of this, the committee recommends that the responsible agency develop a procedure that will enable it to gain enough confidence in certain suppliers for inspection of their products to be virtually eliminated, or at least made with reduced frequency. This would enable compliance inspection to concentrate on a smaller population of seafood producers. Some efforts are currently underway to develop such a procedure. For surveillance inspection, which is performed to obtain an estimate of the overall safety of fishery products, samples of incoming lots must first be made. To be representative, these samples should be random. A stratified random sampling procedure is recommended to select lots for sampling. Stratification would be based on harvesting areas and would ensure that lots selected are representative. Once the lots have been selected, random samples of fish should be chosen from each lot for inspection. Details of this procedure may be found in Cochran (1977) or Eberhardt (1990).

OCR for page 267
Seafood Safety The FDA in the past has done very little surveillance sampling except for organic compounds, as discussed in Chapter 5. Estimates of overall seafood safety that are based primarily on compliance samples, made only when a problem is suspected, are usually biased. ATTRIBUTES AND VARIABLES SAMPLING PLANS Sampling plans are either attributes plans or variables plans. Attributes plans are those for which the item inspected is classified acceptable or nonacceptable and the statistics tallied are the number of unacceptable items in the sample. Fish may be classified unacceptable for any number of reasons, not all of which are linked to seafood safety. However, the basis for classifying a fish unacceptable must be stated prior to inspection. Usually, the reasons for the unacceptable conditions are listed along with the tally of the number of unacceptable fish in the inspection reports. Variables plans are those for which a quality characteristic is measured on each item inspected and the average measurement is used for the acceptability decision. Such sampling plans have a much greater ability to distinguish between good and bad lots. However, only a single characteristic may be measured with each sampling plan. Also, some characteristics are not measurable on a continuous scale, such as appearance, texture of meat, color, or odor. Virtually all FDA plans are attributes plans of either the two-class or the three-class type. The committee did not find any other types of sampling plans in use. Two-Class Attributes Plans Two-class plans are discussed first. Each fish inspected falls into one of two classes—conforming or nonconforming. The statistical properties of such plans are usually described by means of an operating characteristic curve. This shows the probability of acceptance of a lot versus its quality in terms of fraction or percent nonconforming. Acceptability decisions with two-class plans may be based on a single sampling plan, a double plan, or a multiple plan. A single sampling plan specifies a sample size and an acceptability criterion. A typical single sampling plan might be n = 13, c = 2. This means that the inspector takes a sample of 13 fish and accepts the lot or lots if 2 or fewer do not conform to the specifications for safety. A double plan that has approximately the same properties is sample size n1 = 8, n2 = 8; acceptance numbers c1 = 0, c2 = 3; rejection numbers r1 = 3, r2 = 4. This means the inspector takes an initial sample of 8 items; the lot is accepted if there are no nonconforming items and rejected if there are 3 or more. If there are 1 or 2, a second sample of 8 items is taken. If from the entire 16 fish, there are 3 or fewer nonconforming fish, the lot is accepted; otherwise, it is not. Such a plan would be feasible only if additional fish are available to the inspector. Often this is not true. A multiple plan that has the same statistical properties as these two plans is presented in Table 7-1. For this plan, the inspector inspects 3 items from a lot. If 2 or more of them are nonconforming the lot is rejected; otherwise, another sample of 3 items is selected. If none of the 6 items is nonconforming the lot is accepted; if 3 or more are nonconforming it is rejected; otherwise, another 3 are selected. The

OCR for page 267
Seafood Safety acceptance (Ac) and rejection (Re) numbers for the next stage (9 items) are also 0 and 3. The process continues in this fashion until either an acceptability decision is made or the seventh stage is reached. At the seventh stage, the lot is either accepted or rejected. TABLE 7-1 Multiple Sampling Plan Stage ni Σni Ac Re 1 3 3 -a 2 2 3 6 0 3 3 3 9 0 3 4 3 12 1 4 5 3 15 2 4 6 3 18 3 5 7 3 21 4 5 a Cannot accept the first stage. For the above single sampling plan, n = 13, c = 2; if 5% of the lot is nonconforming, the probability of acceptance is 0.972. That is, 97.2% of such lots would be passed by this procedure. Similarly, if 10% are nonconforming the probability of acceptance is 0.857 or 85.7% If 20% are nonconforming, the probability of acceptance is 0.518 or 51.8%. If 50% of the fish in the lot are nonconforming, there is a 4.4% chance of acceptance. The double and multiple plans have operating characteristics that are approximately the same as those of the single plan. The advantage of the double and multiple plans is that, on the average, smaller numbers of items need be inspected before a decision is made. Thus, for example, if a lot is 20% nonconforming, the single plan requires 13 items, and the double plan requires 11.5 items on the average. The multiple plan requires an average of approximately 9 items. The disadvantage is that the inspector does not know in advance how many fish to select for inspection. Multiple plans are thus much more difficult to administer and are frequently avoided for that reason. The above single sampling plan (n = 13, c = 2) could be changed by reducing the acceptance number. This, however, would also affect the risk to the supplier of acceptable lots being rejected. The effects of such changes are indicated in Table 7-2. Assume that a 2% nonconforming lot should be accepted, whereas a 10% nonconforming lot should be rejected. Table 7-2 shows the probabilities of acceptance for acceptance numbers 0, 1, 2, and 3 for samples of size 13 and various percentages nonconforming. Table 7-2 indicates that, by using an acceptance number of 2, virtually all 2% nonconforming lots would be passed, but most (86%) of the 10% nonconforming ones would also be passed. To reduce the latter probability, the acceptance number could be reduced to 0. Then only 27% of such lots would be accepted. However, 13% of the good (2%) lots would be rejected. By increasing the sample size while keeping the acceptance number greater than zero, this problem can be solved. Probabilities for two-class plans are calculated by using the binomial distribution. Similar probabilities for three-class plans, discussed below, are based on the multinomial distribution.

OCR for page 267
Seafood Safety TABLE 7-2 Acceptance Probabilities for Various Acceptance Numbers and Samples of Size 13 Nonconforming (%) c = 0 c = 1 c = 2 c = 3 2 0.77 0.97 0.998 1.000 5 0.52 0.86 0.97 0.996 10 0.27 0.63 0.86 0.96 20 0.074 0.27 0.52 0.74 Three-Class Attributes Plans The three-class plan is used chiefly for compliance with microbiological standards. A three-class plan has two specification values (m and M): m represents a target value, or limit of microorganisms present for products manufactured under good manufacturing practices (GMPs); M represents the limit of the same microorganisms considered to be acceptable. That is, any number more than M is unacceptable, and a lot should be rejected if any sample unit contains M or more of the microorganisms. As an example of such a plan, for Staphylococcus aureus a sample of 5 is used. No sample unit may exceed 1,000 per gram (g) and 1 of the 5 items may exceed 500/g but not 1,000/g. Thus, m=500/g and M=1,000/g. This plan would be described as n = 5, c = 1, m = 500/g, and M = 1,000/g. Description of the performance of this type of plan would have to be done by means of a two-way table, where P represents the actual percentage of all items in the lot exceeding 500/g (m) but not exceeding 1,000/g (M), and Q represents the actual percentage of all items in the lot exceeding 1,000/g. The probabilities of such lots being accepted are shown in Table 7-3. Thus, if a lot of 40,000 pounds (lb) of seafood is submitted to this inspection containing P = 20% (8,000 pounds) exceeding 500/g but not 1,000/g, and Q = 10% (4,000 pounds) exceeding 1,000/g, respectively, the probability of acceptance is 41%. Similar two-way tables can be prepared that describe all of the three-class attributes plans. TABLE 7-3 Percentages of Lots Accepted   Q (%) P (%) 0 10 20 30 40 0 100 59 33 17 8 10 92 53 29 14 6 20 74 41 21 9 4 30 53 27 12 5 1 40 34 16 6 2 <1

OCR for page 267
Seafood Safety A list of FDA sampling plans follows, which indicates acceptance probabilities for various qualities for comparison purposes. Such qualities are usually stated in terms of acceptable quality level (AQL), which is defined as the quality that has a 95% chance of acceptance. Another pertinent level is the indifference quality level (IQL), which has a 50% chance of acceptance. A third is the limiting quality level (LQL), which has only a 10% chance of acceptance. These three quality levels are given for the two-class FDA plans to be presented. An organization responsible for compliance sampling procedures should make a decision regarding appropriate values of these quality levels before setting a plan. These decisions are policy decisions and should be based on risk assessment as described in Chapter 6. Most Probable Number (MPN) Some FDA plans make use of the most probable number (MPN) as a means of estimating bacterial densities. This number is based on the dilution method of estimation, which is a method for estimating, without directly counting, the number of organisms in a liquid. The dilution method consists of taking samples from the liquid, incubating each in a suitable culture medium, and observing any growth in the number of organisms present. The MPN estimation of density is based on two assumptions: (1) that the organisms are distributed randomly throughout the liquid medium, and (2) that the incubated medium is certain to show growth whenever the sample contains an organism. If the second assumption is not met, the MPN method will underestimate the density. If there are k organisms in a volume of liquid V, and if a sample of volume v is taken from V, the probability that none of the organisms will be found in the sample is (1 -v/V)k = ak. If n such samples are taken, the probability that s of them will contain no organisms is where n! = n(n -1) … (2)(1). For example, if 10 samples of 2 milliliters (mL) each are taken from a total volume of 100 mL, the probability that 6 of the 10 samples will contain no organisms (and thus no growth) is Recall that the purpose of this analysis is to determine k, the number of organisms in the total liquid. The MPN procedure determines the value of k that gives the largest probability of obtaining s sterile samples. For the present example, s = 6, and if k = 40, the probability of 6 sterile samples is 210 (0.9840)6 (1-0.9840)4 = 0.155. Similarly, if k = 50, the probability p of 6 sterile samples is 0.080. Table 7-4 indicates the values of

OCR for page 267
Seafood Safety p for various values of k. As indicated in the table, if 6 of the 10 samples are to be sterile, the most probable number of organisms in the total volume is 25, with a probability equal to 0.251. Even though 25 is the most probable value, the probability is only 0.251, not very high. The procedure could be improved by using a dilution series, for example, three 1:10 dilutions of 5 samples per dilution. TABLE 7-4 Probabilities for Various Sizes of k k p 20 0.227 22 0.242 24 0.250 25 0.251 26 0.250 27 0.249 28 0.246 30 0.236 40 0.155 50 0.080 The MPN is a very imprecise estimate of the actual number of organisms present. To illustrate this imprecision, 95% confidence limits are listed in Table 7-5. That is, if the MPN is as shown and 5 samples per dilution for a series of three 1:10 dilutions are taken, the confidence is 95% that the true number of organisms lies between the two limits. As indicated in Table 7-5, the MPN becomes less precise as it becomes larger. TABLE 7-5 95% Confidence Limits MPN/100 mL Lower Upper 14 6 35 70 30 210 140 60 360 220 100 580 540 220 2,000

OCR for page 267
Seafood Safety SURVEY OF CURRENT SAMPLING PLANS Salmonella Sampling Plans In sampling plans for Salmonella, three categories of food are identified by the FDA Inspection Operations Manual, chart 1 (FDA, 1980b): Foods that would normally be in category II, except they are intended for consumption by the aged, infirm, or infants Foods that would not normally be subjected to a process lethal to Salmonella between the time of sampling and consumption Foods that would normally be subjected to a process lethal to Salmonella between the time of sampling and consumption (most seafood, except molluscan shellfish, is in food category III) For all categories, a sample unit consists of a minimum of 100 g selected at random from a lot. A 25-g analytical unit is analyzed for Salmonella from each 100-g sample unit. The analytical units may be condensed with the maximum composite size of 375 g or 15 analytical units. Table 7-6 describes the sampling plans for each of the food categories. The plans are all two-class plans. In the table, n refers to the number of sample units to be selected from each lot. If the analytical units are composited, 60 sample units would mean 4 composite units. It is assumed here that if one or more analytical units contain Salmonella , the composite will test positive. The c values refer to the acceptable number of positive tests. For these plans, no positive tests are allowed. TABLE 7-6 Sampling Plan Characteristics for Salmonella Food Category n c AQL (%) IQL (%) LQL (%) I 60 0 0.086 1.16 3.84 II 30 0 0.171 2.31 7.68 III 15 0 0.342 4.62 15.35 The AQL is the percent nonconforming that has a 95% probability of acceptance. That is, for category II, 95% of the lots for which 0.171% of the units contain Salmonella will be accepted in the long run. In a 40,000-pound lot, 68 pounds would contain Salmonella. Similarly, the IQL is the percent nonconforming that has a 50% probability of acceptance. For category II, this means that 50% of the lots for which 2.31% of the meat contains Salmonella will be accepted. For a 40,000-pound lot, this would be 924 pounds.

OCR for page 267
Seafood Safety The LQL is the percent nonconforming that has a 10% probability of acceptance. For category II, this means that 10% of the lots for which 7.68% of the meat contains Salmonella will be accepted. For a 40,000-pound lot, this represents 3,072 pounds. The Food and Agriculture Organization/World Health Organization (FAO/WHO, 1969) sampling plan for Salmonella calls for a sample of 5 units with acceptance number 0. For this plan, the AQL is 1%, the IQL is 13%, and the LQL is 37%. The FDA plans are considerably tighter than the FAO/WHO plan. Staphylococcus aureus Plans The FAO/WHO (1969) plan for Staphylococcus aureus is a three-class plan. The sample size is still 5 with c = 2, m = 500/g, and M = 5,000/g. That is, for 5 sample units analyzed, no unit should exceed 5,000/g and no more than 2 units should exceed 500/g for the lot to be accepted. Because this plan is a three-class plan, simple descriptions such as AQL, IQL, and LQL do not apply. Instead, a two-way table describing the plan has been developed by the National Marine Fisheries Service. In Table 7-7, P is the percentage of units in the lot exceeding m (500/g) but not exceeding M (5,000/g) and Q is the percentage of units exceeding M. The percentage of lots that will be accepted is found in Table 7-7 for each P-Q combination. TABLE 7-7 Percentages of Lots Accepted for Several P and Q Combinations   Q (%) P (%) 0 10 20 30 40 0 100 59 33 17 8 10 99 58 32 16 7 20 94 54 29 14 6 30 83 47 24 11 4 40 68 36 16 6 2 50 50 23 9 2 <1 Plans for Fish, Fresh or Frozen The following FDA (1980a,b) sampling procedure applies to fish, fresh or frozen, for adulteration involving decomposition and can be found in both the Inspection Operation Manual 616.12 and the Compliance Policy Guides (CPG) 7108.05. The procedure is based on a lot of fish, each weighing up to 3 pounds. The sample

OCR for page 267
Seafood Safety size is 50 fish. A lot will be cited if it has decomposed fish in 2 or more cases and if any of the following occurs: 5% or more fish show class 3 decomposition over 25% or more of their bodies,1 20% or more show class 2 decomposition over 25% or more of their bodies,2 or % class 2 + 4(% class 3) ≥20%. (See Table 7-8.) TABLE 7-8 Chance of Acceptance (%)   Class 3 (%) Class 2 (%) 1 3 5 10 1 99 95 83 32 3 98 84 62 21 5 94 75 48 14 10 81 47 22 3 Fish–Adulteration by Parasites Action levels for parasites in fish (CPG 7108.06, Inspection Operation Manual chart 5) are as follows: Tullibees, ciscoes, inconnus, chubs, and whitefish: 50 cysts/100 pounds and 20% of fish are infested. Bluefin and other freshwater herring: Averaging 1 pound or less: 60 cysts/100 fish and 20% of fish examined are infested. Over 1 pound: 60 cysts/100 pounds of fish and 20% of fish examined are infested. Rosefish (redfish and ocean perch): 3% of the fillets contain one or more copepods with pus pockets. The sample sizes are as follows: 1. Single plan (Table 7-9)

OCR for page 267
Seafood Safety TABLE 7-9 Number of Pounds in a Single Sampling Plan   Sample Weight (lb) No. of Boxes in Lot Jumbo or Large Medium Small 5-19 28 23 16 20-100 53 45 33 100 or over 70 56 39   SOURCE: CPG 7108.06 (FDA, 1980a). 2. Double plan (for lots of 20-100 boxes; Table 7-10) TABLE 7-10 Number of Pounds in a Double Sampling Plan   Cysts/100 Cysts/100 lb (combined samples) Size of Fish n1 (lb) Ac (1) Re (1) n2 (lb) Ac Re Large or jumbo 35 30 70 63 49 50 Medium 27 26 67 43 49 50 Small 18 38 61 26 49 50   SOURCE: CPG 7108.06 (FDA, 1980a). 3. Sequential plans (for lots over 100 boxes; also listed in Inspection Operation Manual chart 5). For imported fish the FDA may use any of these plans; for domestic fish the double plan based on 25 fish is to be used. See Tables 7-9 and 7-10. Crabmeat-Adulteration with Filth Containing Escherichia coli Section 7108.02 of the CPG requires a two-class plan as follows: For a sample of size 6, take legal action if any fish indicates adulteration. Seize the lot if 2 or more contain Escherichia coli. The performance of these plans is indicated in Table 7-11.

OCR for page 267
Seafood Safety TABLE 7-11 Sampling Plan Characteristics of Crabmeat n Ac Re AQL (%) IQL (%) LQL (%) 6 0 1 0.85 11 32 6 1 2 6.3 26 51 Langostinos-Adulteration by Bacterial Contamination A sample size of 10 is required. The criteria for seizure are the existence of any of the following (CPG 7108.09): Coliform density greater than 20/g (by MPN method) in 20% of the samples E. coli density greater than 3.6/g (MPN) in 20% of the samples Coagulase-positive staphylococci density greater than 3.6/g (MPN) in 20% of the samples Aerobic plate count (at 35°C) greater than 100,000/g as a geometric average of all the subsamples For this plan, the sample size is 10 and the acceptance number is 0. It is understood, however, that a change is expected that will increase the acceptance number to 1. The statistics in Table 7-12 apply to these two sampling procedures. TABLE 7-12 Sampling Plan Characteristics for Langostinos n Ac Re AQL (%) IQL (%) LQL (%) 10 0 1 0.5 6.7 21 10 1 2 3.7 16 34 Canned Salmon-Adulteration Involving Decomposition For canned salmon adulteration involving decomposition, seizure is authorized if examination by two analysts in accordance with the sampling procedures shows either (CPG 7108.10) number of defective (class II or III) cans equal to or exceeding the action numbers, or two or more class III cans in either the first or the total sample. The plans in this Compliance Policy Guide are double three-class plans. Different plans are required for different lot sizes. The various plans for 1/4-pound through 1-pound

OCR for page 267
Seafood Safety cans are shown in Table 7-13, with acceptance probabilities for various combinations of class II and class III defective cans in Table 7-14. Lot sizes are for cases of 48 cans each. TABLE 7-13 Sampling Plans for Canned Salmon for ¼-lb Through 1-lb Cans Lot Size (cases) n1 c1 r1 n2 n1+n2 c2 r2 < 100 18 0 3 32 50 4 5 100-199 20 0 3 48 68 8 9 200-499 24 1 4 64 88 10 11 500-799 30 1 4 80 110 16 17 800-999 36 1 5 96 132 22 23 1,000-1,499 42 1 6 112 154 25 26 1,500 and up 48 1 6 128 170 29 30   SOURCE: CPG 7108.10 (FDA, 1980a). TABLE 7-14 Percentage of Lots Accepted (< 100 cans/lot)   Class II Cans in Lot (%) Class III Cans in Lot (%) 0 1 3 5 10 15 20 25 0 100 100 98.4 90.8 50.2 16.9 4.1 0.5 1 94.9 93.7 89.2 78.5 38.5 12.0 2.9 0.6 3 70.2 67.2 60.3 49.7 21.0 6.0 1.4 0.3 5 47.1 43.5 36.5 28.6 11.0 3.0 0.7 0.2 10 16.0 13.7 9.9 6.9 2.3 0.7 0.2 0.04 15 5.4 4.4 2.9 1.9 0.6 0.2 0.04 0.01 20 1.8 1.4 0.9 0.6 0.2 0.04 0.01 0.00 25 0.6 0.4 0.3 0.2 0.0 0.01 0.00 0.00 TABLE 7-15 Sampling Plans for Canned Sampling for ¼-lb Cans Lot Size (cases) n1 c1 r1 n2 n1+n2 c2 r2 < 100 6 0 2 12 8 1 2 100-199 7 0 2 16 23 1 2 200-499 8 0 2 22 30 1 2 500-799 10 0 2 27 37 2 3 800-999 11 0 3 30 41 3 4 1,000-1,499 12 0 3 32 44 4 5 1,500 and up 16 0 3 38 54 5 6   SOURCE: CPG 7108.10 (FDA, 1980a).

OCR for page 267
Seafood Safety For 1- to 4-pound cans, the sampling plans in (Table 7-15) apply. Several comparisons of these plans are shown in Tables 7-16 through 7-21. TABLE 7-16 Percentage of Lots Accepted for 1% Class II and 1% Class III Cans (1/4-1 lb) Lot Size Percentage of Lots Accepted Lot Size Percentage of Lots Accepted < 100 93.7 800-999 78.4 100-199 90.5 1,000-1,499 73.2 200-499 91.0 1,500 and up 68.1 500-799 83.4     TABLE 7-17 Percentage of Lots Accepted for 1% Class II and 1% Class III Cans (1-4 lb) Lot Size Percentage of Lots Accepted Lot Size Percentage of Lots Accepted < 100 97.1 800-999 96.3 100-199 95.8 1,000-1,499 95.9 200-499 94.0 1,500 and up 93.7 500-799 95.4     TABLE 7-18 Percentage of Lots Accepted for 5% Class II and 5% Class III Cans (1/4-1 lb) Lot Size Percentage of Lots Accepted Lot Size Percentage of Lots Accepted < 100 28.6 800-999 7.31 100-199 20.6 1,000-1,499 4.23 200-499 21.4 1,500 and up 2.42 500-799 12.5     TABLE 7-19 Percentage of Lots Accepted for 5% Class II and 5% Class III Cans (1-4 lb) Lot size Percentage of Lots Accepted Lot Size Percentage of Lots Accepted < 100 63.1 800-999 43.6 100-199 54.7 1,000-1,499 43.4 200-499 46.8 1,500 and up 31.1 500-799 42.2    

OCR for page 267
Seafood Safety TABLE 7-20 Percentage of Lots Accepted for 15% Class II and 10% Class III Cans (¼-1 lb) Lot Size Percentage of Lots Accepted Lot Size Percentage of Lots Accepted <100 0.65 800-999 0.03 100-199 0.36 1,000-1,499 0.00 200-499 0.59 1,500 and up 0.00 500-799 0.13     TABLE 7-21 Percentage of Lots Accepted for 15% Class II and 10% Class III Cans (1-4 lb) Lot Size Percentage of Lots Accepted Lot Size Percentage of Lots Accepted <100 18.8 800-999 4.36 100-199 13.6 1,000-1,499 3.44 200-499 10.0 1,500 and up 1.10 500-799 5.69       SOURCE: CPG 7108.110 (FDA, 1980a). Considerable variation can be seen in the percentage of lots accepted among lot sizes for each of the can sizes. The percentage of lots accepted decreases with lot size. However, the decrease is more marked for smaller cans. Also, the percentage of lots accepted is higher for the larger cans than for the smaller ones for any lot size and quality. Shrimp – Adulteration Involving Decomposition Shrimp are classified as follows (CPG 7108.11): Class 1. Very fresh to fishy odor characteristic of the product Class 2. Slight odor persistent and perceptible to an experienced examiner as pertaining to decomposition Class 3. Strong odor of decomposition that is distinct and unmistakable In terms of the criteria for decomposition, a subsample of 100 shrimp is classed as decomposed if 5% or more shrimp are class 3, 0% or more shrimp are class 2, or % class 2 + 4 (% class 3) ≥20%.

OCR for page 267
Seafood Safety The sample sizes are listed in Table 7-22. TABLE 7-22 Single Sampling Plans for Shrimp Inspection   Lot Size Number of Cases n Ac Re 1-20 6 1 2 21-100 12 2 3 101 or more 18 3 4   SOURCE: CPG 7108.11 (FDA, 1980a). The acceptance and rejection numbers refer to number of subsamples of 100 shrimp each. That is, for the first plan (n = 6, Ac = 1), 6 subsamples of 100 shrimp each are selected. If 2 or more subsamples are classed as decomposed for any of the above three criteria, the lot will not be accepted. The percentage of lots accepted will depend on a combination of class 2 and class 3 decomposition. These percentages are given in Tables 7-23 through 7-25. TABLE 7-23 Percentage of Lots Accepted for Lot Size 1-20, n = 6, c = 1   Class 3 (%) Class 2 (%) 1 3 5 1 100 63 4 5 99 33 1 10 76 9 0 TABLE 7-24 Percentage of Lots Accepted for Lot Size 21-100, n = 12, c = 2   Class 3 (%) Class 2 (%) 1 3 5 1 100 52 0 5 100 16 0 10 71 1 0

OCR for page 267
Seafood Safety TABLE 7-25 Percentage of Lots Accepted for Lot Size 101 or more, n = 18, c = 3   Class 3 (%) Class 2 (%) 1 3 5 1 100 45 0 5 100 9 0 10 69 0 0 As can be seen, the probability of acceptance decreases with lot size for constant quality because it would be more serious to accept a large lot of poor quality than a small one. It should also be noted, for example, that in a lot having 3% class 3 decomposition and 5% class 2 decomposition, the percentage of lots accepted will range from 9 to 33. The numbers in these tables could be reduced by either increasing the sample sizes or decreasing the acceptance numbers. However, if this is done the probability of accepting reasonably good lots will also be reduced. As an example of this, suppose the acceptance numbers are changed from 1, 2, and 3 to 0, 1, and 2, but the same sample sizes are retained (6, 12, 18). The percentage of lots accepted for 5% class 2 and 3% class 3 lots will go from 33, 16, and 9 for the present plans to 8, 5, and 3; the percentage of lots that have 1% class 2 and 5% class 3 will go from 99, 100, and 100 to 86, 97, and 99, respectively. CONCLUSIONS AND RECOMMENDATIONS The above discussion of current sampling plans indicates that they provide relatively little protection to the public. Fortunately, most lots contain little contaminated seafood and most of the contaminants are removed by cooking. However, in view of the serious health hazards discussed in other chapters, it is necessary to improve seafood safety with regard to these hazards. Safety should be improved by means of controls on the process rather than by reliance on sampling inspection. As indicated in this section, unless the sample size increases significantly, very little consumer protection is provided. None of the sampling procedures described above provides much protection to the consumer, and increasing the sample size is not a reasonable solution. As indicated, the statistical uncertainties associated with lot sampling make this an unreliable method for ensuring the safety of food products even if testing methods for dangerous microorganisms, toxins, and contaminant chemicals were fully available and completely reliable. This is well recognized by scientists and administrators involved in food regulation, many of whom are strongly supportive of the Hazard Analysis Critical Control Point (HACCP) system. The HACCP system provides on-line, real-time control at the processing level and has worked well in the low-acid canned food industry. General procedures for this type of system have been discussed in a number of publications (ICMSF, 1988; NRC, 1985) and laid out in more detail by the Interagency Advisory Committee. Effective application of the HACCP system requires that both industry and regulatory agency personnel understand its various

OCR for page 267
Seafood Safety components and that the two groups work closely together in the development and implementation of HACCP plans. The focus must be set clearly on consumer safety in the assessment of hazards and the identification of critical control points (CCPs). The purpose of the HACCP system is to provide as high a level of assurance as possible that the food product reaching the consumer is safe and wholesome; it is not a quality control plan, though CCPs may coincide with quality control points in a process. There is a potential danger in attempting to incorporate quality control considerations, such as weight control, workmanship, or breading levels, into the HACCP plan because these may then become the focus of attention at CCPs, when they have no impact on the safety of the food. Cluttering up an HACCP plan with such considerations detracts from the essential simplicity and specificity of the system and unnecessarily complicates control and testing procedures without adding to safety. Indeed, confounding workmanship with safety controls can lead to a false sense of security by unsophisticated operators who confuse positive control signals on workmanship with safety indicators. Application of a well-designed HACCP plan will provide greater assurance of safety for the consumer and may be readily assessed by the regulatory agencies. It is essential for fisheries products that such plans take into account the quality and condition of the water from which animals are harvested and the intrinsic condition of the animal at harvest because natural toxins, contaminant chemicals, or sewage contamination are significant risks for certain fish and shellfish. NOTES 1.   The decomposed product has an odor that is distinct and unmistakable. 2.   First stage of identifiable decomposition. The product produces an odor, which while not intense, is persistent and perceptible by an experienced examiner. On the other hand, class 1 include fishery products ranging from fresh to those having odor characteristics of that product, that is not identifiable as decomposition. REFERENCES Cochran, W.G. 1977. Sampling Techniques, 3rd ed. John Wiley & Sons, New York. 428 pp. Eberhardt, K.R. 1990. Survey sampling methods. Pp. 91-940 in H.M. Wadsworth, ed. Handbook of Statistical Methods for Engineers and Scientists. McGraw-Hill, New York. FAO/WHO (Food and Agriculture Organization/World Health Organization). 1969. Codex Alimentarius. Sampling Plans for Packaged Foods, CAC/RM42-1969. Food and Agricultural Organization of the United Nations and World Health Organization, Geneva, Switzerland. FDA (Food and Drug Administration). 1980a. Fish and seafood. Compliance Policy Guides (CPG) 7108.02-7108.25, Chap. 8. Food and Drug Administration, Public Health Service, Washington, D.C. FDA (Food and Drug Administration). 1980b. Inspections Operations Manual 616.12. Food and Drug Administration, Public Health Service, Washington, D.C. ICMSF (International Commission on Microbiological Specifications for Foods). 1988. Microorganisms in Foods. 4. Application of the Hazard Analysis Critical Control Point (HACCP) System to Ensure Microbiological Safety and Quality. Blackwell Scientific Publishers, Oxford, England. 357 pp.

OCR for page 267
Seafood Safety NFI (National Fisheries Institute). 1989. Shrimp/Fish Statistical Sampling HACCP Industry Workshop Report, Atlanta, Ga., August 16. Prepared by NFI in collaboration with National Marine Fisheries Service, Arlington, Va. 82 pp. NRC (National Research Council). 1985. An Evaluation of the Role of Microbiological Criteria for Foods and Food Ingredients. Subcommittee on Microbiological Criteria in Foods and Food Ingredients, Food and Nutrition Board. National Academy Press, Washington, D.C. 436 pp. Schilling, E.G. 1982. Acceptance Sampling in Quality Control. Marcel Dekker, New York. 775 pp. Wadsworth, H.M., K.S. Stephens, and A.B. Godfrey. 1986. Modern Methods for Quality Control and Improvement, John Wiley & Sons, New York. 690 pp.