Pet Food Quality Assurance and Safety and Quality Assurance Survey within the Costa Rican Pet Food Industry

Simple Summary The health status of pets may be affected by infectious agents transmitted through the feed. The close coexistence between humans and animals makes the safety assurance of pet food paramount. Different unit operations are applied at the industry level to prevent and reduce the presence of physical, chemical, and biological hazards in pet food; through these operations and adequate implementation of good manufacturing practices, the safety of the finished product can be preserved. Abstract Costa Rican animal feed production is continually growing, with approximately 1,238,243 metric tons produced in 2018. Production-wise, pet cat and dog food are in fifth place (about 41,635 metric tons per year) amongst animal feeds, and it supplies up to 90% of the national market. Pet food production has increased as a response to the increase in the population of dogs and cats in Costa Rica, where 50.5% of households own at least one dog and indicates more responsible ownership in terms of feeding pets. Part of the process of making dry pet food involves a thermal process called extrusion, which is capable of eliminating the microbial load. However, extrusion can compromise nutritional quality to some extent by denaturing proteins, oxidizing lipids, and reducing digestibility. The objective of this study was to evaluate the quality and safety of dry pet food and to assess the effect of the extrusion process on digestibility and the quality of proteins, amino acids, and fatty acids. Pet food samples were collected before and after extrusion and were used to evaluate Good Manufacturing Practices (GMP), based on Central American Technical Regulation (RTCA 65.05.63:11). In general terms, weaknesses in infrastructure, documentary evidence, and post-process practices were observed in two Costa Rican feed manufactories. Feed safety was surveyed through the analysis of Salmonella spp., Escherichia coli, Listeria spp., Staphylococcus aureus, aerobic mesophilic microorganisms, fungi, and yeasts counts. The extrusion process effectively reduced pathogenic microorganisms, and showed no effect on the digestibility of dog food (p = 0.347), however, it could reduce the availability of some nutrients (e.g., amino acids, fatty acids). Furthermore, a retrospective diagnosis was made for puppy food (n = 68), dog food (n = 158), and cat food (n = 25), to evaluate the history of nutritional quality and safety. Finally, it can be confirmed that the correct implementation of GMP allows feed manufacturers to deliver a product of optimum texture, smell, nutritional composition, and safety.


Food Safety: Heavy Metal Analysis
The heavy metal analysis was performed using a methodology described by Granados-Chinchilla and coworkers [23]. Briefly, 300 mg of the sample was placed into the digestion vessel (DAP-60+, Berghof Products + Instruments GmbH, Baden-Wurtemberg, Eningen, Germany) and was mixed with 10.0 mL of HNO 3 (65 mL/100 mL, puriss, Sigma-Aldrich 84378, St. Louis, MO, USA). The mixture was microwave-digested using a 3 step temperature program.  [%]). Limit of detection and quantification for arsenic, cadmium, mercury, and lead in dog food were 1.00 and 3.03, 0.01 and 0.03, 0.02 and 0.06, and 0.60 and 1.82 µg kg −1 , respectively.

Nutritional Quality: Proximate Analysis
Dry matter (loss on drying/moisture), crude protein, fiber, and ash, as well as calcium, phosphorus, and pepsin digestibility of animal protein assays, were performed to assess the nutritional quality of the pet food. All tests were performed using ISO 17025 accredited methods based on AOAC  24, 935.13, and 971.09, respectively. Crude fat and water activity (a w ) were calculated in extruded pet foods by acid hydrolysis AOAC OMA SM 954.02, and Aqualab chilled mirror methods (measurement performed at 24.50 ± 0.24 • C, Aqualab 4TE, Decagon Devices, Pullman, WA, USA), respectively.

Nutritional Quality: Fatty Acid Profiling
A 250 g pet food sample was milled and sieved to 1 mm (using a ZM 200 ultracentrifuge mill, Retsch GmbH, Haan, Germany). After that, a subsample of ca. 1 g of each feed sample was placed in a 50 mL glass beaker, where 5 mL of diethyl ether was added and mixed using an ultrasonic shaker (USC200TH, VWR International, Center Valley, PA, USA) for 5 minutes. Afterward, a 200 µL aliquot was transferred to a GC 2 mL vial (Agilent Technologies, Santa Clara, CA, USA). Then, 800 µL of diethyl ether and 1000 µL of, a previously prepared, 0.25 g/100 mL TMHA solution in methanol were added to the same vial. An aliquot of 2 µL of the resulting mixture was injected into the GC system. Qualitative analyses of the volatile compounds were carried out using an Agilent gas chromatograph (7820, Agilent Technologies, Santa Clara, CA, USA) equipped with an Agilent Technologies J&W DBWAX microbore column of 10 m length, 0.1 mm diameter, 0.1 µm film thickness and Agilent 5977E mass spectrometer (MSD). The carrier gas was helium at a constant flow rate of 0.3 mL min −1 . The GC oven temperature was kept at 50 • C for 0.34 min and increased to 200 • C at a rate of 72.51 • C min −1 . The temperature was kept at 200 • C constant for 0.17 min, increased to 230 • C at a rate of 8.7 • C min −1 , and remained constant for 7.9 min for a total run time of 13.93 min. The split ratio was adjusted at 30:1. The injector, transfer line, ion source, and quadrupole temperatures were set at 250, 250, 230, and 150 • C, respectively. The mass range and the electron energy was set at 50-450 m/z and 70 eV, respectively. Constituents were identified by matching their spectra with those in NIST library 14. Only hits with a match factor above 80% were considered. FAME mixtures GLC-486 (n = 40 analytes) and GLC-860 (n = 60 analytes, Nu-Chek Prep, Inc., Elysian, MN, USA) were used as quality control comparing retention times and mass spectra with those found in the analyzed samples. Compounds used to check mass tuning included tetradecanoic (6.16  Louis, MO, USA) was used as an internal standard and 9c11t-C 18:2 , 10c12t-C 18:2 , C 12:0 , 4c7c10c13c16c19c-C 22:6 , 11t-C 18:1 were concurrently monitored by simultaneous ion monitoring (SIM) mode (total dwell time 100 ms and cycles 8.3 Hz. For compounds with no standard, identification should be considered as tentative.

Nutritional Quality: Amino Acid Profiling and Furosine Analysis
Sample digestion was carried out as previously reported for furosine [24]. Briefly, feed samples were sieved to 1 mm, and 200 ± 1 mg subsamples were used for digestion. The sample was transferred into a 40 mL glass vial (27184 SUPELCO, St. Louis, MO, USA), mixed with 2 mL water (to achieve dispersion), and 6 mL of an HCl 10.6 mol L −1 aqueous solution. Immediately, the vial was capped with a septum (tan PTFE/silicone, 27188-U, SUPELCO, St. Louis, MO, USA), and nitrogen was bubbled for 1 min into the solution through a needle. The resulting mixture was heated adiabatically for 23 h at 110 • C. The resulting hydrolysate was filtered by gravity through a grade 4 qualitative filter paper (Whatman, GE Healthcare Life Sciences Pittsburgh, PA, USA). Then, 130 µL of a borate buffer (Merck Millipore, Burlington, Massachusetts, USA, B0394, 50 mmol L −1 , pH = 10), 10 µL sample digest, 20 µL NaOH (Merck Millipore, 1064679010, 3.6 mol L −1 ), 10 µL OPA freshly prepared solution (Merck Millipore, P0657, 10 mg o-phtalaldehyde is dissolved in 500 µL chromatographic grade ethanol and made up to volume with borate buffer in a 10 mL volumetric flask), 10 µL FMOC (Merck Millipore, 160512, 10 mg 9-fluorenylmethoxycarbonyl chloride is dissolved in 500 µL chromatographic grade acetonitrile and made up in borate buffer in a 10 mL volumetric flask), 320 µL ultrapure water were Animals 2019, 9, 980 5 of 25 sequentially mixed together in a HPLC ready-to-inject vial. One microliter of the resulting mixture was injected into the LC system. The solvent system consisted of a 40 mmol L −1 NaH 2 PO 4 buffer adjusted at 7.8 pH (Merck Millipore, S9638, ACS, 98% pure, solvent A) and acetonitrile, methanol, and water (45:45:10, solvent B). Gradient mode was as follows: 0% B at 0 min, 0%B at 1.9 min, 57% B at 18.1 min, 100% B at 18.6 min, 100% B at 22.3 min, 0% B at 23.2 min, and 0% B at 26 min. Chromatographic separations were performed using an Agilent Technologies LC system equipped high resolution column (150 × 4.6 mm, 3.5 µm, PN 963400-901260), a 1260 infinity quaternary pump (61311C), column compartment (G1316A), an automatic liquid sampler module (ALS, G7129A) and a fluorescence detector (G1321A) (Agilent Technologies, Santa Clara, CA, USA). A detection system was set at 340 (excitation) and 450 nm (emission) for all amino acids except proline for which 266 and 305 nm wavelengths were used, respectively. Ultrapure water [type I, 0.055 µS cm −1 at 25 • C, 5 µg L −1 TOC] was obtained using an A10 Milli-Q Advantage system and an Elix 35 system (Merck KgaA, Darmstadt, Germany). The identity of each amino acid was assessed, and each compound quantified using NIST ® SRM ® 2389a amino acids in 0.1 mol L −1 .

Statistical Analysis
Descriptive statistics were performed on the GMP data for each facility. The same data set was used to detail the prevalence of each of the microorganisms, and the average digestibility of the product, by sampling area and by the facility. The total number of coliforms, aerobic mesophilic microorganisms, and fungal count in the finished product analyzed using a One-Way ANOVA Tukey's honestly significant difference post hoc test. The same analysis was used for the digestibility of the product before and after going through the extrusion process. A value of α = 0.05 was used as the significance level for all hypothesis tests. InfoStat (version 12, Córdoba, Argentina) was used to conduct all statistical analyses.

GMP
The results of the evaluation carried out on each feed mill are summarized in Figure 1. A color code was used to identify specific points with nonconformities (NC).
For Table A1 (2. Documentation), none of the facilities complied with points 2.2, 2.5, and 2.10. In facility A, the GMP manual is outdated, and in establishment B, it has not yet been enforced. Then, establishment A breaches point 2.3 and 2.4.
In Table A1 (3. Facilities), which evaluates the facilities, both establishments do not follow points 3.10 and 3.13. Infrastructure in both facilities is old and has not been remodeled in recent years. Windows and doors allow the entrance of birds and other animals, which increases the risk of contamination of raw materials and the finished product. The first defense against biological hazards in a production system relies on the facilities [25]; therefore, facilities represent a critical point. Moreover, neither facility had a protocol for people or external vehicles access, nor there was a clear written indication for the correct flow of workers and visitors through the feed plant. Additionally, establishment A exhibited non-conformities for points 3.4, 3.11, 31.6, and 3.17. The first point relates to the lack of a vehicle access procedure. The second point is associated with the infrastructure condition; and the last two with the lack of written procedures and records that function as evidence of the activities that are carried out. For example, garbage dumps (a source of pollution) are not well identified, and the frequency of waste disposal is not stated, as it should be [10]. Establishment B failed to comply with points 3.12 and 3.15. Failing to comply with 3.12 may indicate a risk of cross-contamination in the finished product (Figure 2A,B). In the case of point 3.15, the design of the facilities does not allow for effective cleaning. A diagram of the design of each of the facilities and process flows is represented (Figure 2A,B). In facility B, walls separate each area of the food plant, however, the process flow should be corrected as the transit lines between the raw materials and the finished product area should never cross (red steps, Figure 2A,B). Any contamination, especially biological hazards, can occur through employee traffic or the relocation of equipment [25]. of the product before and after going through the extrusion process. A value of α = 0.05 was used as the significance level for all hypothesis tests. InfoStat (version 12, Córdoba, Argentina) was used to conduct all statistical analyses.

GMP
The results of the evaluation carried out on each feed mill are summarized in Figure 1. A color code was used to identify specific points with nonconformities (NC).   63:11). The evaluation points were as follows: (2) Documentation; (3) facilities; (4) equipment; (5) personnel; (6) pest control; (7) production flow; (8) raw materials; (9) storage of risk ingredients; (10) water; (11) formulation; (12) grinding; (13) adding ingredients; (14) mixing; (15)  , which evaluates the facilities, both establishments do not follow points structure in both facilities is old and has not been remodeled in recent years. s allow the entrance of birds and other animals, which increases the risk of materials and the finished product. The first defense against biological hazards tem relies on the facilities [25]; therefore, facilities represent a critical point. cility had a protocol for people or external vehicles access, nor there was a clear r the correct flow of workers and visitors through the feed plant. Additionally, ibited non-conformities for points 3.4, 3.11, 31.6, and 3.17. The first point relates icle access procedure. The second point is associated with the infrastructure st two with the lack of written procedures and records that function as evidence are carried out. For example, garbage dumps (a source of pollution) are not well requency of waste disposal is not stated, as it should be [10]. Establishment B h points 3.12 and 3.15. Failing to comply with 3.12 may indicate a risk of crosse finished product (Figure 2A,B). In the case of point 3.15, the design of the low for effective cleaning. A diagram of the design of each of the facilities and resented (Figure 2A,B). In facility B, walls separate each area of the food plant, s flow should be corrected as the transit lines between the raw materials and the ea should never cross (red steps, Figure 2A,B). Any contamination, especially an occur through employee traffic or the relocation of equipment [25]. acilities), which evaluates the facilities, both establishments do not follow points ructure in both facilities is old and has not been remodeled in recent years. allow the entrance of birds and other animals, which increases the risk of materials and the finished product. The first defense against biological hazards em relies on the facilities [25]; therefore, facilities represent a critical point. ility had a protocol for people or external vehicles access, nor there was a clear the correct flow of workers and visitors through the feed plant. Additionally, bited non-conformities for points 3.4, 3.11, 31.6, and 3.17. The first point relates cle access procedure. The second point is associated with the infrastructure t two with the lack of written procedures and records that function as evidence re carried out. For example, garbage dumps (a source of pollution) are not well equency of waste disposal is not stated, as it should be [10]. Establishment B points 3.12 and 3.15. Failing to comply with 3.12 may indicate a risk of crossfinished product (Figure 2A,B). In the case of point 3.15, the design of the ow for effective cleaning. A diagram of the design of each of the facilities and esented (Figure 2A,B). In facility B, walls separate each area of the food plant, flow should be corrected as the transit lines between the raw materials and the a should never cross (red steps, Figure 2A,B). Any contamination, especially n occur through employee traffic or the relocation of equipment [25]. cilities), which evaluates the facilities, both establishments do not follow points ructure in both facilities is old and has not been remodeled in recent years. allow the entrance of birds and other animals, which increases the risk of materials and the finished product. The first defense against biological hazards m relies on the facilities [25]; therefore, facilities represent a critical point. ility had a protocol for people or external vehicles access, nor there was a clear the correct flow of workers and visitors through the feed plant. Additionally, ited non-conformities for points 3.4, 3.11, 31.6, and 3.17. The first point relates le access procedure. The second point is associated with the infrastructure t two with the lack of written procedures and records that function as evidence e carried out. For example, garbage dumps (a source of pollution) are not well quency of waste disposal is not stated, as it should be [10]. Establishment B points 3.12 and 3.15. Failing to comply with 3.12 may indicate a risk of crossfinished product (Figure 2A,B). In the case of point 3.15, the design of the w for effective cleaning. A diagram of the design of each of the facilities and sented (Figure 2A,B). In facility B, walls separate each area of the food plant, flow should be corrected as the transit lines between the raw materials and the should never cross (red steps, Figure 2A,B). Any contamination, especially n occur through employee traffic or the relocation of equipment [25]. For Table A1 (2. Documentation), none of the facilities complie facility A, the GMP manual is outdated, and in establishment B, it establishment A breaches point 2.3 and 2.4.
In Table A1 (3. Facilities), which evaluates the facilities, both es 3.10 and 3.13. Infrastructure in both facilities is old and has not Windows and doors allow the entrance of birds and other anim contamination of raw materials and the finished product. The first d in a production system relies on the facilities [25]; therefore, fa Moreover, neither facility had a protocol for people or external veh written indication for the correct flow of workers and visitors thro establishment A exhibited non-conformities for points 3.4, 3.11, 31. to the lack of a vehicle access procedure. The second point is a condition; and the last two with the lack of written procedures and of the activities that are carried out. For example, garbage dumps ( identified, and the frequency of waste disposal is not stated, as it failed to comply with points 3.12 and 3.15. Failing to comply with contamination in the finished product (Figure 2A,B). In the case facilities does not allow for effective cleaning. A diagram of the d process flows is represented (Figure 2A,B). In facility B, walls sep however, the process flow should be corrected as the transit lines b finished product area should never cross (red steps, Figure 2A,B biological hazards, can occur through employee traffic or the reloca critical non-conformity; ies), which evaluates the facilities, both establishments do not follow points re in both facilities is old and has not been remodeled in recent years. w the entrance of birds and other animals, which increases the risk of erials and the finished product. The first defense against biological hazards elies on the facilities [25]; therefore, facilities represent a critical point. had a protocol for people or external vehicles access, nor there was a clear correct flow of workers and visitors through the feed plant. Additionally, non-conformities for points 3.4, 3.11, 31.6, and 3.17. The first point relates ccess procedure. The second point is associated with the infrastructure with the lack of written procedures and records that function as evidence rried out. For example, garbage dumps (a source of pollution) are not well ncy of waste disposal is not stated, as it should be [10]. Establishment B ts 3.12 and 3.15. Failing to comply with 3.12 may indicate a risk of crossshed product (Figure 2A,B). In the case of point 3.15, the design of the r effective cleaning. A diagram of the design of each of the facilities and ed (Figure 2A,B). In facility B, walls separate each area of the food plant, should be corrected as the transit lines between the raw materials and the uld never cross (red steps, Figure 2A,B). Any contamination, especially cur through employee traffic or the relocation of equipment [25]. hich evaluates the facilities, both establishments do not follow points both facilities is old and has not been remodeled in recent years. e entrance of birds and other animals, which increases the risk of and the finished product. The first defense against biological hazards on the facilities [25]; therefore, facilities represent a critical point. a protocol for people or external vehicles access, nor there was a clear ct flow of workers and visitors through the feed plant. Additionally, -conformities for points 3.4, 3.11, 31.6, and 3.17. The first point relates procedure. The second point is associated with the infrastructure the lack of written procedures and records that function as evidence out. For example, garbage dumps (a source of pollution) are not well f waste disposal is not stated, as it should be [10]. Establishment B 12 and 3.15. Failing to comply with 3.12 may indicate a risk of crossproduct ( Figure 2A,B). In the case of point 3.15, the design of the ective cleaning. A diagram of the design of each of the facilities and igure 2A,B). In facility B, walls separate each area of the food plant, ld be corrected as the transit lines between the raw materials and the never cross (red steps, Figure 2A,B). Any contamination, especially rough employee traffic or the relocation of equipment [25].
point not evaluated.
In Table A1 (4. Equipment), both facilities breach points 4.3 and 4.6, sensitive issues. None of the facilities showed evidence that their equipment guarantees accuracy [26] or that batch-to-batch cross-contamination does not occur [25,27]. Establishment A also has NC in 4.2, where it failed to collect evidence that shows magnets and screens are routinely checked, a critical point, in physical hazards avoidance [10,25].
In Table A1 (5. Personnel), facility A failed to comply with the three points that make up this tier, while establishment B, only failed points 5.2 and 5.3. The collaborating staff, being directly in contact with the food, must ensure their health status and their training in food handling [26,28].
Table A1 (6. Pest control) and Table A1 (7. Production flow) are fully compliant, and these sections are decisive to maintain the safety of the final product. However, it is difficult to evaluate the pest control system in practice and the production flow beyond the design. On the other hand, pest control should include the management of birds, that are recognized as carriers of Salmonella spp. and insects [27].
In Table A1   In Table A1 (4. Equipment), both facilities breach points 4.3 and 4.6, sensitive issues. None of the facilities showed evidence that their equipment guarantees accuracy [26] or that batch-to-batch crosscontamination does not occur [25,27]. Establishment A also has NC in 4.2, where it failed to collect evidence that shows magnets and screens are routinely checked, a critical point, in physical hazards avoidance [10,25].
In Table A1 (5. Personnel), facility A failed to comply with the three points that make up this tier, while establishment B, only failed points 5.2 and 5.3. The collaborating staff, being directly in contact with the food, must ensure their health status and their training in food handling [26,28].
Table A1 (6. Pest control) and Table A1 (7. production flow) are fully compliant, and these sections are decisive to maintain the safety of the final product. However, it is difficult to evaluate the pest control system in practice and the production flow beyond the design. On the other hand,  is, raw materials are used without evaluation of suppliers. Furthermore, a procedure for entering and handling raw materials is not available. The issue above represents another critical point since the reception of raw materials is an entrance for physical, chemical, and biological hazards [27]. Though keeping control of each batch of raw material may be taxing, an inspection of suppliers, considering the requirements of the ingredients to be acquired, will reduce potential risks [25]. On the other hand, it is not guaranteed that the finished product possesses the nutritional formulation that was previously calculated, and there are no routine safety controls. Therefore, there are no records of the complete product analysis, which is a violation of the confidence of the consumer on what is stated in the product label [10], and breach of official regulation [29]. In Table A1 (20. Traceability), facility A breaches the two points, 20.1 and 20.2, while establishment B only breaches 20.2; that is, there are no procedures or records to keep traceability. The procedures and logs allow controlling the process of production with quality and safety standards that enable its commercialization [29,30] and are an essential part of a GMP system.
Facility A failed to comply with

Nutritional Quality: Crude Protein and Digestibility before and after the Extrusion Process
Pet feeding is aimed to provide a better quality of life and to maintain optimal health status. Therefore, each serving offered to a cat or dog must ensure to cover the basic nutritional requirements based on its physiological stage and physical activity [31,32]. In the formulation of animal feed, both the quality of the raw materials and the manufacturing process are involved in the nutritional quality of the finished product [32][33][34]. Since 95% of dog food is extruded, it is characterized mainly by its low moisture content (i.e., below 10 g/100 g) [35], hardness, and durability [36].
As stated before, extrusion transforms, mix, and sterilizes a wide variety of ingredients, to produce a harmless, stable, and balanced food [34,37]. The same process that renders safe food can denature proteins, which consequently affects amino acid bioavailability [37]. It can also cause lipid oxidation, which may decrease the content of essential fatty acids, such as linoleic and linolenic acid [37].
Since not all nutrients are harnessed similarly, a standard indicator of food quality is digestibility, which is, measuring the proportion of nutrients available for absorption [38]. Table 1 shows the data obtained from the analysis of pepsin digestibility, moisture, crude protein, and fat performed on adult dog food samples. Moisture continually changes during the process and reduced to less than 10 g/100 g, one of the goals of the extrusion (Table 1) [36]. Crude protein values are below recommended may be associated with errors during formulation. Crude fat also varies, and, in some cases, it increases after exiting the extruder as flavoring and fat are added in a post-extrusion step (Table 1). Finally, no significant differences for protein digestibility are observed between the sampling areas, especially before and after the thermal process, which coincides with the results obtained by van Rooijen and collaborators [37], and Tran and collaborators [34] (Table 1). On the other hand, digestibility values equal to or greater than 80 g/100 g are recommended, since absorption by the animal is prominently improved for these values [35,38,39]. None of the feed samples tested met this requirement, which is related to the quality of raw materials [32] (Table 1).

Nutritional Quality: Water Activity and Moisture in Pet Foods
Water activity obtained for a dry extruded adult dog, puppy, and cat foods were 0.5356 ± 0.0961, 0.5837 ± 0.0682, and 0.5477 ± 0.0505, respectively (Table 2). These values are relatively higher than those reported elsewhere for cat food (0.30-0.50) and dog food (0.30-0.54) [40]. As a cost management strategy, Costa Rican feed industry usually maintains moisture contents between 8 and 10 g/100 g ( Table 2). Increased water activity may have a severe impact on the pet food shelf life [41]. Water activity has demonstrated to be a functional alternative to moisture content analysis [42] as a w is related to lipid quality and has proved to influence lipid oxidation [43], lipid modification [44], mycoflora, and fumonisin B 1 accumulation [45].

Nutritional Quality: Crude Protein, Digestibility and Amino Acid Profile
Some feed ingredients used commonly in pet food formulations have been described elsewhere [46,47]. In Costa Rica, the basis for pet food formulations is corn and soybean meal. Additionally, fish, poultry by-products, meat and bone meals (i.e., swine and cattle and other ruminants) are regularly used to supply protein and fat, which is especially practical for animals that are strictly carnivore [6]. Our data for crude protein and amino acid profiles are in agreement with other reports [48]. Taurine is a dietary essential sulfur-containing amino acid for felines [49][50][51][52]. From the diets tested, n = 1 sample was below the taurine limits recommended for adult cats, which is of great concern as the endogenous synthesis of this compound by cats is limited and putative precursors (cysteic and cysteinesulfinic acids) cannot substitute its presence [51] (Table 3). Crude protein values were above the recommended threshold except for n = 1 and n = 3 samples of puppy and adult dog food (Table 3). Although no recommendation is given by AAFCO in regard of taurine for dogs, recent reports have established a connection among cardiomyopathy in golden retrievers, fed with commercial diets, and taurine deficiency [53]. Hence, average values of 0.17 ± 0.03 g taurine/100 g sample found during our survey maybe not enough for certain dog breeds; the amount that must be, then, more strictly monitored (Table 3). However, Tjernsbekk and coworkers [54] have already demonstrated that crude protein alone cannot be used to assess nutritional and protein quality, where digestibility and amino acid profile is crucial. Values below thresholds can be explained by the effect that the extrusion process has on the feed ingredients, including their protein quality and amino acid profile [55][56][57][58]. Fortunately, no puppy food samples were below AAFCO suggested thresholds. Mean values of apparent protein digestibility (75.68 ± 7.75 g/100 g protein using in vitro method) seem to be in accordance with those reported elsewhere [59] for puppy food, albeit on the low range of that stated (Table 3). However, this is not the case for adult dry dog food, where mean values (68.61 ± 3.76 g/100 g protein) are below those declared therein [59] (Table 3).

Nutritional Quality: Furosine Content in Pet Foods
Recently, cell lines have been reported to be sensitive to furosine, and that this compound reduced weight gain, and affected the functions of liver and kidney in animals [60]. Besides being widely used as a marker of thermal treatment and nutritional quality of food, furosine detection may have now safety implications. As dry pet food is extruded to improve the digestibility of nutrients and increase shelf life and food safety [14], pet food ingredients can suffer considerable thermal processing, such as the case of soy. Nevertheless, extrusion tends to increase furosine content in soy. At least two research groups have evaluated furosine generation in soybean-based feed products obtained under severe thermal treatment conditions and reported values of 66.55 ± 0.37 and 108.01 ± 8.97 mg/100 g protein [61][62][63]. Our data (Table 3) is in agreement with the results by Chiang [63,64] in dry dog food stored for 12 weeks at 37.8 • C. Increased concentrations of furosine are undesirable in pet food as they indicate impaired lysine utilization by the animal and compromised weight gain [65].

Nutritional Quality: Fatty Acids, Fiber, and Minerals (Ca and P)
Fat provides energy, twice as much as proteins and carbohydrates, to intervene in the health and good condition of the skin and fur, and increases the palatability of food [35]. Although fat is not an essential nutrient, it must meet the requirements of essential fatty acids [35]. Fatty acids, such as linoleic acid, are precursors of other compounds, such as arachidonic acid, eicosapentaenoic acid (EPA) or docosahexaenoic acid (DHA), in puppies. Said molecules are essential for the proper development of the nervous system, and their deficiency is associated with vision problems and learning problems [39] ( Table 4).  Calcium and phosphorus are essential for bone development and are minerals vital throughout the life of the animal, especially during the growth phase to avoid rickets [35]. In addition, pet food should maintain calcium within the recommended minimum and maximum values, as well as the 2:1 (Ca:P) ratio. Inverting this ratio may result in inadequate calcium absorption [35,39].
Fiber, in food for dogs and cats, has a mechanical function, by intervening in the formation of the bolus and collaborating with mineral absorption and biliary metabolism [35,39] (Tables 5 and 6).

Food Safety: Microbiological Quality
Our microbiological survey included the sampling of strategic zones within each feed plant (i.e., after the mixer, the extruder, and finished product) (Table 7, Figure 2A,B). Sampling after the thermal process demonstrates its effectiveness in eliminating microorganisms (Table 7, Figure 2A,B). Sampling the finished product shows that the GMP implemented can maintain microbiological safety (Table 7, Figure 2A,B). However, testing the finished product alone is not an effective way to determine the absence of a microorganism. Routine control should include the critical stages in the production process [27]. Table 7 describes the number of samples per sampling area, in addition to the results for different microbiological analyzes. The conformity criterion was based on current international regulation, where the sternest limits have been taken as the threshold value.  Raw materials, and those of animal origin, in particular, carry a microbial load that was detected in the first stage of the production chain. EFSA [27], and Huss and collaborators [32] indicated that animal-based ingredients are potential sources of Salmonella spp. for compound feed, a fact which was also experimentally demonstrated by Leiva and collaborators [70]. The second production stage reflects the effectiveness of the thermal process to eliminate microorganisms [10,70]. During dry dog food production, the extrusion process requires high temperatures (from 80 to 150 • C) for at least two minutes to achieve the destruction of 10 3 CFU per 100 grams [25]. However, the eradication of the initial microbiological load does not exclude the possibility of subsequent contamination [25,70]. Our results indicated the finished product might have considerable mesophilic aerobic bacteria, fungi, and yeast counts, although not pathogenic, are indicators of quality and storage conditions [71].
Salmonella spp. and E. coli are pathogenic bacteria that can be transmitted to humans, either by contact with the animal or contaminated food [41]. The latter is relevant since dogs have a close relationship with humans [27,32]. The presence of pathogenic microorganisms in feed is evidence of a deviation from GMP, such as those mentioned above [10,28,32]. Once a biological hazard has entered the production chain, due to a failure in the GMP system, it is a challenge to eradicate it. Consequently, more expensive processes within the production are needed to reduce or eliminate the risk [25]. On the other hand, the total of the samples assayed were negative for both Listeria spp. and S. aureus.

Food Safety: Heavy Metal Contaminants
Mean values for lead (508.52 ± 305.72 and 703.01 ± 705.40 mg kg −1 for puppy and adult dog food, respectively) were consistently higher than other heavy metals (Table 8). These values are low when compared with previous reports of cadmium and lead in dog food in concentrations of 0.20 ± 0.01 and 3.23 ± 0.08 mg kg −1 , respectively [72]. Low amounts of these contaminants are expected, as Costa Rica is a country without a mining industry; therefore, heavy metal contamination in the compound feed may come from mineral sources [23,73]. However, dogs have been documented to be resistant to relatively elevated dietary doses of lead (i.e., 10 mg kg −1 ; [74]). Diets with doses of heavy metals as the  (Table 8). Recently, red meat-based dog diets have been reported to contain higher values of lead [74]. Finally, pet food can be considered safe, and hazard notifications caused by arsenic and mercury, although scarce, have triggered stern actions, such as product recalls [75]. In the case of puppy food, the average value for crude protein (i.e., 23.70 g/100 g) is close to the recommended threshold of 22.50 g/100 g, which means that at least some of the food samples have crude protein below this point (Table 9). For example, the minimum value found for this type of food is 9.24 g/100 g, which is considered extremely low (Table 9). This result may be a consequence of pet treat mislabeling as compound food, when the actual formulation round 12 g crude protein/100 g. Low protein values lead to growth retardation [76]. In the case of minerals, calcium, phosphorus, and selenium were also found at values as low as 0.22, 0.48 g/100 g, and 0.21 mg kg −1 , respectively (Table 9). An opposite trend was observed in crude fat and iron supplementation (Table 9). In the case of microbiological analysis, only n = 1/68 samples (1.47% prevalence) were found to be above the legal threshold for Salmonella spp. (Table 9). As the thermal processing of extruded compound feed was found to be adequate, it is expected that contamination was acquired, afterward, by inadequate handling of the finished food or cross-contamination from high-risk raw materials previous to or during packaging (e.g., meat and bone meal) [20,70]. Similar studies have not reported Salmonella on dry extruded dog food [21,77]. Noteworthy, similar prevalence values as those stated herein have been reported in industrialized countries, where most of the Salmonella strains were recovered from the processing equipment [78]. However, an outstanding 71.43% of samples were positive for total coliforms, a general indicator of hygiene [20] (Table 9). Comparable results have been found in commercially available dog food sold in bulk and sealed packages [77]. The authors noted that 43.8% of the samples had mesophilic microorganisms in the range from 100 to 10,000 CFU g −1 , and 77% of the samples had mold and yeast counts from 0 to 100 CFU g −1 . Our data showed a low microorganism load (i.e., 382 ± 340 CFU g −1 , and 62.80 ± 47.57 MPN g −1 for the total mesophilic bacterial count and total coliforms, respectively) ( Table 9). In the current study, bulk feed samples (a practice in our country is a called "feed quartering and repacking") were not included, however, that market should be regulated since bulk sales can reduce food safety [77]. Interestingly, even when yeast and mold counts are relatively low, the presence of mycotoxins could indicate previous fungal colonization of the raw materials before formulation and thermal treatment. For example, aflatoxin B 1 , deoxynivalenol and fumonisin B 1 were found in several samples, though aflatoxin and deoxynivalenol levels were below their respective threshold for pet food (i.e., 20 and 2000 µg kg −1 ) (Table 9) [20]. Fumonisin B 1 had an average of 5.54 mg kg −1 , a contamination level above the recommended guideline for the sum of fumonisin B 1 and B 2 of 5000 µg kg −1 [20]. On the other hand, contamination of pet foods with life-threatening pathogens (e.g., Listeria spp., Salmonella spp., toxigenic E. coli strains) has been reported only in raw pet foods in a collaborative study from the US [21]. Regarding crude protein and fat, the results are similar to was found in puppy food. On the contrary, Ca and P were at recommended values; however, the Ca/P ratio in some cases exceed the maximum recommended of 2.5 (Table 10). Calcium and phosphorus intake should be strictly monitored in dogs as malnutrition is a risk factor for developmental orthopedic diseases and secondary nutritional hyperparathyroidism in large breed puppies and dogs [79,80]. Finally, n = 2/158 samples (1.26%) were found to be contaminated with Salmonella spp. (Table 10). Salmonella outbreaks associated with dry pet food and treats have become a concern for these products since they serve as a means of pathogen exposure for pets and their owners [81].

Cat Food
In the case of cat food, some samples were found to be below the recommended dietary minimum for both crude protein and fat (set at 26.0 and 9.0 g/100 g, respectively) ( Table 4). Interestingly, all samples were compliant for calcium and phosphorus (Table 11). Similarly to puppy and dog food, n = 1/25 sample (4.00% prevalence) was found to be contaminated with Salmonella spp. (Table 11). In all the cases, the samples with low nutrient values may not be able to supply the needs for maintenance or growth. Thus, the addition of dietary supplements or an increase in the amount of food intake may be required [3].

Conclusions
Good manufacturing practices are the basis of food safety, and their correct implementation is the foundation for risk management. The thermal process to which dry dog food is subjected to is useful in the elimination of pathogens. The prevalence of microorganisms, whether pathogens or indicators in the finished food, is associated with cross-contamination and deviance from the GMP program. Shortcomings hinder safe food production and seem to be reflected mostly in the lack of documentation, precise procedures that engender incorrect practices, and incorrect workflow within the manufacturing plant. The quality of the infrastructure, access to the facilities, the untethered transit of people and vehicles, the handling of raw materials and the finished product, pest control, and cleaning procedures are critical points to consider when product safety is to be guarded. The data also supports the fact that cross-contamination of feed is an issue. The thermal process to improve the safety of pet foods is of no use if the final product will be contaminated downstream. Although the digestibility of food is not affected by the extrusion process, it can reduce the availability of some nutrients (e.g., amino acids, fatty acids). Formulations should account for the loss of nutrients during thermal treatment since some of these nutrients are present in low concentrations and are prone to lability. Additionally, the monitoring of some compounds that are being produced during the thermal process (e.g., Maillard reaction) may be required as they can dictate the quality of the pet food. Costa Rican pet foods seem to be safe regarding heavy metals, though these analytes should be included in quality routine analysis.