DESCRIPTION OF PROBLEM
Near-infrared reflectance spectroscopy (NIRS) is a rapid analysis method that enables a multipurpose analyzer to be calibrated to predict nutritional component values after analyzing the reflectance of a feed ingredient in the near-infrared spectrum. The reliability of a NIRS prediction for a nutritional component depends on the accuracy of the determined wet chemistry laboratory values used to make the initial NIRS prediction equation. As reviewed by Rahman et al. , NIRS can be very effective at predicting nutrient components such as protein, fat, vitamin, mineral, amino acid, and dry matter content in diets and feed ingredients. More specifically, the use of NIRS has already proven very effective in predicting dry matter, crude protein, and fat content in feed samples such as wheat . Correlation coefficients for protein calibrations range from 0.89 to 0.99 [3–7]. When comparing NIRS techniques to nitrogen-based regression analysis to predict amino acid concentrations, it was found that NIRS calibrations were able to explain 21 to 58% more variation within a feedstuff category .
The use of NIRS has the potential to allow animal nutritionists to formulate diets based on known nutrient specifications of their ingredients and help reduce the need for over formulation [9, 10], which is done to ensure diets are not marginal in nutrient requirements because of the variation in nutrient content of a given feed ingredient due to things such as the cultivar variety, the agronomic conditions during the production of the crop, harvesting conditions, storage protocol, and differences in further processing procedures. Fortunately, many components of feed absorb in the near-infrared reflectance region of the electromagnetic spectrum . Unlike other analysis techniques, with NIRS, no reagents are used and no waste products are created apart from dust during sample preparation . After the initial cost of purchasing a NIRS spectrometer, the cost to perform a sample analysis is extremely low, and NIRS does not require special laboratory technician training after calibrations have been developed. Samples can be analyzed in a matter of minutes, and a sample can undergo analysis for multiple nutrient components simultaneously . Thus, with NIRS, analysis of individual batches of feed ingredients is entirely possible as they are delivered to a feed mill.
Having real-time NIRS analyses of feed ingredients delivered to a feed mill allows the use of modeling programs for cost-effective diet formulations relative to bird performance parameters. NutriOpt  is a modular precision-feeding system, designed to improve financial and animal performance objectives through advanced, real-time nutritional analysis and modeling, to small- and medium-sized poultry integrations throughout the world. Although the NutriOpt technologies have been developed through years of ongoing research and development, its application in poultry trials to demonstrate the cumulative added value of the various NutriOpt technologies is limited. Therefore, the goal of the present research was to evaluate broiler performance and economic impacts of selected NutriOpt nutritional formulation solutions under typical US broiler-production conditions.
MATERIALS AND METHODS
After all the bulk feed ingredients were chemically analyzed for crude protein, fat, fiber, ash, moisture, and dry mater as well as analyzed by NIRS , 6 dietary treatments were created for this broiler experiment (Table 1). The same sequestered feed ingredients were used to make the starter, grower, finisher, and withdrawal diets to reduce nutrient variation. The nutritional consultant formulated control diets (Trt.1, Table 1) were formulated to reflect a standard US commercial formulation. The consultant’s diets (Trt. 1) were then shadow formulated into the Nutreco/Trouw Nutrition formulation program, and diets thereafter were formulated to meet the identical nutrient constraints and nutrient values (Trt. 2). Treatments thereafter (Trt. 3–6, Table 1) were modified to meet the objective of each treatment. For the NutriOpt modeling, it was assumed that the control treatment would have a carcass yield of 66% and breast meat yield would be 25%. In addition, the following revenue-generating parameters were assumed: $0.84/kg live weight; $2.27/kg breast meat; $1.50/kg carcass; and $1.0/kg remaining carcass. The starter diets (Table 2) were fed from days 1 to 14 of age, the grower diets (Table 3) were fed from days 14 to 28 of age, the finisher diets (Table 4) were fed from 28 to 42 D of age, and the withdrawal diets (Table 5) were fed from 42 to 49 D of age. The starter diets were in a crumble form, while the grower, finisher, and withdrawal diets were in pellet form.
This experiment was conducted in a facility with 2 identical, but separate, rooms. Each room was equipped with 48 (1.524 m by 1.22 m) floor pens. All pens were equipped with 5 nipple drinkers originating from a common water line and 1 pan feeder (0.09 m2 ). Before chick placement, new pine shavings were placed in the pens. A continuous lighting program was implemented with a light intensity of 20 lux for 24 hs (0–4 D), 20 lux for 20 h (5–7 D), 10 lux for 16 h (8–14 D), 2 lux for 16 h (15–42 D), and 2 lux for 20 h (43–49 D). Light intensity was verified by placing a Light ProbeMeter  into the pens.
For each room, a computerized controller– regulated 2 gas-fired furnaces, an exterior evaporative cooling system on both sides of the room for air intake, 4 45.7-cm ceiling circulation fans, and 2 91.4-cm exhaust fans and one 61-cm exhaust fan for air clearance at the end of each room. Ambient temperature was set to 34ºC on day 1 and decreased by 0.28ºC until 24ºC was reached and then maintained. No significant differences in temperature and humidity were noted throughout the studies between the 2 rooms.
Before placing chicks, 96 pens were assigned to one of the 6 dietary treatments in a random block design (16 replicates per treatment [8 replicates per room]). The day of hatch information of Cobb 500 male fast-feathering broilers was obtained from a primary breeder hatchery . The chicks were sorted, and those that did not weigh between 38 and 46 g were discarded before the remaining birds were assigned to the 96 pens (23 birds per pen).
Feed and water were provided ad libitum throughout the duration of the experiment. All diets were formulated to meet or exceed NRC  requirements. All animal procedures were approved by the University of Georgia Animal Care and Use Committee, Athens, GA.
For each room, humidity, temperature, water consumption, and pen mortality were recorded twice daily. Dead birds were weighed on the day of death, so their weight could be included in the pen gain in the calculation of the feed to-gain ratio. Birds and feed were weighed on days 0, 14, 28, 42, and 49 to determine body weight, feed intake, body weight gain, and feed to-gain values on pen basis. On day 49, the mean bird weight for each pen was determined, and 8 birds per pen within 300 g above or below the mean weight of their pen were selected for processing. Individual weights for the selected birds were recorded, and each bird was leg banded before placement in a coop for an overnight feed withdrawal before processing. On day 50, birds were weighed and processed at the University of Georgia’s Pilot Processing Plant as previously described . Subsequently, eviscerated hot carcass weights were recorded for each bird before static chilling in an ice bath for 4 h. After a 4-h chill, chilled carcasses were drained before being cut up and deboned. Weights were recorded for drained chilled carcass, pectoralis major, pectoralis minor, wings, and leg quarters of each bird. Percent yield calculations were based on the fasted, live weight of the bird.
Body weight, body weight gain, feed-to-gain ratio, feed intake, and carcass yields were subjected to ANOVA according to the General Linear Model , with the statistical model including treatment and room as factors. Tukey’s multiple-comparison procedure was used to detect significant differences among dietary treatments . For all outcomes, a P value # 0.05 was used to determine significance among dietary treatments.
RESULTS AND DISCUSSION
From days 0 to 14 of age, body weight gain for the broilers fed with dietary Trt. 4 and Trt. 6 was significantly lower than that of the broilers fed Trt. 1 (Table 6), but feed-to-gain ratio did not differ between the control treatment (Trt. 1) and any of the other treatments (Table 6). During the grower phase (days 14–28) and for the overall period from 0 to 28 D of age, there were no differences in body weight gain between the broilers fed in the control treatment and those fed in any of the other dietary treatments (Tables 7 and 8). From 0 to 28 D of age, the broilers fed with Trt. 6, designed to maximize breast meat yield, had a significantly lower feed to-gain ratio than the broilers fed with the control diet (Table 8). The lower feed-to-gain ratio of broilers fed with Trt. 6 compared with broilers fed with the control diet and all other treatments was maintained in the finisher period (days 28–42 of age) and from days 0 to 42 of age (Table 9 and 10). During the withdrawal period (42–49 D of age), body weight gain, and feed-to-gain values did not differ among the dietary treatments (Table 11). For the overall experimental period (days 0–49 of age), body weight gain did not differ between broilers fed with the control dietary treatment and any of the other dietary treatments formulated using various formulation strategies (Table 12). From 0 to 49 D of age, the feed-to-gain ratio of the broilers fed with the dietary treatment (Trt. 6) to maximize breast meat yield was decreased compared with broilers fed with any of the other dietary treatments (Table 12).
The live body weights of the broilers selected for processing statistically mirrored the live body weights of the entire population of broilers for each treatment (Table 13) and loss of weight during fasting before processing did not differ for the broilers from the different dietary treatments (Table 13). Hot carcass, cold carcass, and leg quarter yields did not differ among the dietary treatments (Table 14). Total weight meat yield did not differ between broilers fed with the control dietary treatment and those fed with any of the other model-based dietary treatments (Table 15).
Overall, treatments 2 through 5 performed as well as the control (Trt. 1) with regard to both live and processing performance. The diets for Trt. 2 were formulated using nutrient values for the bulk ingredients predicted by NIRS and the corresponding proprietary nutritional matrix of the NutriOpt system developed by Trouw Nutrition. It is important to note that all dietary treatments were formulated from the same batch of ingredients to reduce natural variations in product quality from batch to batch. Typically, most nutritional matrices used in commercial settings take a conservative approach which ultimately results in diets that may be slightly over formulated relative to minimum requirements. This is to ensure that diets are not marginal in nutrient requirements by taking into account the larger variation in product quality and nutrient values encountered in commercial production. In the present study, the equivalent performance of Trt. 2, which was shadow formulated to Trt. 1 nutrient specifications, demonstrates the potential cost savings of formulating diets on known NIRS nutrient specifications of the major ingredients. Utilization of NIRS at commercial feed mills confers the ability of the nutritionist to enhance their rolling averages for feed ingredients and reduce feed costs by applying known ingredient nutrient specifications in formulating diets to meet nutrient requirements.
Dietary Trt. 3 was similar to Trt. 2 in that the diets were formulated based on the NIRSpredicted nutrient values and NutriOpt nutritional database as in Trt. 2, and also used a Trouw Nutrition amino acid ratio matrix. These amino acid ratios are derived from previous work completed internally by Trouw Nutrition to develop specific amino acid recommendations relative to digestible lysine. Thus, compared with Trt. 1, there is an overall reduction in the protein and amino acid content of the starter, grower, and finisher diets while maintaining digestible lysine levels in Trt. 3. Interestingly, this reduction in amino acid content did not significantly impair performance and resulted in finished feed costs that were lower than those of Trt. 1 (Tables 2–5). These results align well with previous findings investigating the effect of different amino acid strategies in Cobb 500 broilers where statistically equivalent performance could be obtained at lower amino acid densities .
The diet formulation strategy of Trt. 4 mirrored that of Trt. 3, but also added Trouw Nutrition’s retainable phosphorus and phytase matrix. Retainable phosphorous is the amount of non-phytate and phytate phosphorous that can be digested and absorbed for use by the animal . Using retainable phosphorous for formulation allows for the total phosphorous content in the diet to be reduced, which has environmental benefits as well as live performance benefits . The use of Trouw Nutrition’s phytase matrix compared to the supplier specifications used in Trt. 1, compounded with using retainable phosphorous, permitted a reduction in feed costs for Trt. 4 compared with Trt. 1 (Tables 2–5) without reducing overall live performance (Table 12) or processing yields (Tables 14 and 15).
The diets for Trt. 5 and 6 were formulated based on Trt. 4 nutritional strategies combined with NutriOpt modeling technology. This modeling technology has the capability to estimate and predict the most appropriate nutritional strategy based on desired final outcomes. For Trt. 5, the desired final outcome was optimizing profits per kg of live weight, and for Trt. 6, it was optimizing profit per kg of breast meat yield. An obvious feature of Trt. 5’s approach to enhancing live weight profit is by reducing the amino acid density of the diet and increasing the AME value in respective feeding phases relative to Trt. 1 given the commodity costs and revenue generators used at the time of optimization (Tables 2–5). Even though the broilers in Trt. 5 were fed a diet that was lower in cost in 3 of the 4 feeding phases (Tables 2–5) relative to Trt. 1, body weight gain and feed-to-gain ratio from 0 to 49 D of age were equivalent between the 2 treatments (Table 12). This was interesting to observe because the broiler model for improved profit margin predicted a substantial improvement in cumulative feed conversion (1.700 vs. 1.489) and profit per kg ($23.39/kg). Total white meat yield (WMY) in Trt. 5 was also not significantly different compared with that in Trt. 1 despite the lysine content and amino acid density being reduced in Trt. 5 diets. As indicated previously for Trt. 3, Cobb 500 broilers demonstrate a resilience to small and moderate adjustments in amino acid density in relation to WMY .
The broilers from dietary Trt. 6 had significantly higher WMY than Trt. 5 as a percent of live fasted weight (Table 15). The improvement in WMY relative to Trt. 5 is not surprising as the digestible lysine content was 7, 16, 19, and 18% greater in Trt. 6 during the starter, grower, finisher, and withdrawal periods, respectively. Increasing the lysine content of the diet is known to increase breast meat deposition [28-30]. Through the breast meat yield model, Trt. 6’s formulation approach was predicted to increase breast meat yield by 0.7% compared with that of Trt.1 at 50 D of age. Relative to Trt. 1, the digestible lysine levels of the starter, grower, finisher, and withdrawal diets of Trt. 6 were 5, 11, 16, and 14% greater, respectively, but this only resulted in a corresponding numerical improvement in total weight meat yield (Table 15). Interestingly, the Trouw model for improved breast meat yield was optimized with results obtained with smaller-sized broilers grown for a 35-day market weight. Kidd et al.  reported that smaller broilers (35 D of age) were more sensitive to moderate alterations in amino acid densities than larger birds (55 D of age) when considering WMY. Similarly, Meloche et al.  reported that a 10% change in amino acid density influenced breast meat yield more significantly at 35 D of age than at 49 D of age when no significant differences in yield were observed. Intermediately, Corzo et al.  were able to demonstrate significant differences in breast meat yield when comparing various combinations of high, medium, and low amino acid densities at 42 D of age. Taken together, amino acid densities appear to have a more influential role on WMYin broilers grown to smaller market weights.
Dietary Trt. 6 used a non-traditional approach when considering the AME of each dietary phase (Tables 2–5). During the starter and finisher phases, the AME of Trt. 6 was similar to that of Trt. 1. However, during the grower phase, the AME of Trt. 6 was 37 kcal/kg lower than that of Trt. 1, and in the withdrawal phase, the AME of Trt. 6 was 23 kcal/kg higher than that of Trt. 1. Typical commercial diets implement a step-wise increase in AME as birds mature. Although other reports [32, 33] indicate that feed intake might have differed between the birds in these 2 treatments with the differences in dietary AME levels, this was not seen in the current research. Altering the ratio of dietary amino acid to AME is an approach that can increase the amino acid consumption and, therefore, muscle accretion . In the present research, Trt. 6 had a higher ratio of digestible lysine to AME than Trt. 1 in all dietary phases (Tables 2–5), but as indicated earlier, this did not significantly increase breast meat yield for the broilers in Trt. 6 compared with that for broilers in Trt. 1.
Feed accounts for over 60% of production costs in broiler production. Thus, considering live performance and processing performance in terms of economic impact is key to understanding the true effect of each treatment regimen in the present research. Across the world in the various broiler markets, the driving economic factors that influence operation decisions can vary considerably. For that reason, treatment performance was analyzed on an average feed cost/kg, feed cost/bird, live weight, carcass without giblets (WOG), and WMY basis.
Average feed cost per kg was calculated for birds from 0 to 42 and 0 to 49 D of age (Table 16). Producers who are primarily concerned with the cost of feed would benefit with the formulation strategies used for Trt. 2 to 5 as they cost less per kg of feed and on a per-bird basis than Trt. 1 through 0 to 42 and 0 to 49 D of age. Profit per kg of live weight was also analyzed at 42 and 49 D of age (Table 16). For broilers produced to both 42 and 49 D of age, dietary Trt. 2, 4, 5, and 6 provided a greater profit than Trt. 1.
Profit per kg of WOG was better for Trt. 2, 4, 5, and 6 than that for Trt. 1 (Table 16). Profit per kg of WMY at 50 D of age was more advantageous for Trt. 2, 4, and 6 than that for Trt. 1. Producers who are financially driven by WMY would benefit the most from a formulation strategy similar to Trt. 2, Trt. 4, or Trt 6.
CONCLUSIONS AND APPLICATIONS
The current research demonstrates the potential profitability of being able to formulate poultry diets using known nutrient specifications determined by NIRS with robust prediction equations.
Once known nutrient specifications are obtained by NIRS in real time, they can be used with dietary formulation–modeling systems that are tailored for such things as maximizing total WMY or minimizing feed cost. In the current research using Trouw Nutritional formulation technologies, producers who are driven by feed costs and feed cost per bird would benefit most from the formulation strategies used in Trt. 4 > 3 and 5 > 2. Stakeholders who are concerned about live weight profit would benefit from formulation strategies used in Trt. 2 > 6, 5, 4. Companies who primarily market WOG would profit the most from formulation strategies used in Trt. 2, 6, and 4, while those who debone for breast meat would profit the most from formulation strategies used in Trt. 2, 4, and 6.
This article was originally published in 2020 Journal of Applied Poultry Research. 29:314–327 https://doi.org/10.1016/j.japr.2019.11.006. This is an Open Access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).