2015. J. Anim.Sci. 93(6):2826-2839
日糧凈能預測育肥豬生長性能模型的建立和驗證
S. Nitikanchana, S. S. Dritz, M. D.Tokach, J. M. DeRouchey, R. D. Goodband和B. J. White
本論文收集41個試驗報告（共285個能量水平處理），通過薈萃分析（Meta分析）建立運用日糧凈能預測生長性能的模型。日糧能量和營養水平由NRC原料數據庫計算得到。本論文研究了生長性能和影響因子（如凈能、體重、粗蛋白、SID賴氨酸、粗纖維、中性洗滌纖維NDF、酸性洗滌纖維ADF、脂肪、灰分）的相關關系（線性相關和二次線性相關關系），以及影響因子之間的互作關系，并且以AIC（Akaike information criteria）為準則，確定合適的預測模型。最初的預測模型中包括了凈能和粗蛋白的互作或凈能和SID賴氨酸的互作。但是當SID賴氨酸低于需要量的試驗數據移除后，這些互作對生長性能的影響就不再顯著。在含凈能和體重的模型中增加日糧脂肪這一影響因素后可以顯著改善模型G:F（增重耗料比）的預測效果，表明日糧凈能可能低估了脂肪對G:F的影響。Meta分析表明，只有在日糧中其它營養素（如賴氨酸）都能滿足動物需要后，日糧凈能才能準確預測不同日糧原料和環境條件下豬只的日增重。統計結果表明隨著日糧凈能和體重的增加，日增重隨之提高，但是在體重高于87kg時日增重隨之降低。G:F值隨日糧凈能和脂肪含量的提高而提高，但是隨體重的增加而降低。本論文還通過2個試驗、5個日糧處理、543頭育肥豬評估模型預測的生長性能與實際生長性能的差異。5個日糧處理包括3個不同水平的凈能（通過在玉米-豆粕型基礎日糧中添加次粉、大豆皮、含油8-9%的DDGS和精煉動物油，調節凈能水平）。具體處理信息如下：1）30%DDGS、20%次粉、4-5%大豆皮（低能處理組）；2）20%次粉、4-5%大豆皮（低能處理組二）；3）玉米豆粕型日糧（對照組，中等能量水平）；4）處理二日糧添加3.7%精煉動物油使其凈能水平與處理三相同；5）玉米豆粕型日糧添加3.7%精煉動物油（高能組）。試驗結果（日增重和G:F）與模型預測結果十分接近，但是纖維含量最高的低能量組（30%DDGS日糧）偏差較大，模型預測的日增重和G:F高估了3%和6%。因此，本模型可以準確預測不同凈能水平下生長育肥豬的生長速度和飼料效率，在超高纖維低能量日糧的極端情況下效果相對不理想。
譯者注：
薈萃分析為一種對不同研究結果進行收集、合并及統計分析的方法。薈萃分析的主要目的是將以往的研究結果更為客觀的綜合反映出來。研究者并不進行原始的研究，而是將研究已獲得的結果進行綜合分析。
AIC信息準則是衡量統計模型擬合優良性的一種標準，為日本統計學家赤池弘次創立和發展的，可以權衡所估計模型的復雜度和此模型擬合數據的優良性。
Regression analysis to predict growth performance from dietary net energy in growing-finishing pigs
S. Nitikanchana, S. S. Dritz, M. D.Tokach, J. M. DeRouchey, R. D. Goodband and B. J. White
Data from 41 trials with multiple energy levels (285 observations) were used in a meta-analysis to predict growth performance based on dietary NE concentration. Nutrient and energy concentrationsin all diets were estimated using the NRC ingredient library. Predictor variables examined for best fit models using Akaike information criteria included linear and quadratic terms of NE, BW, CP, standardized ileal digestible (SID) Lys, crude fiber, NDF, ADF, fat, ash, and their interactions.The initial best fit models included interactions between NE and CP or SID Lys.After removal of the observations that fed SID Lys below the suggested requirement, these terms were no longer significant. Including dietary fat in the model with NE and BW significantly improved the G:F prediction model, indicating that NE may underestimate the influence of fat on G:F. The meta-analysis indicated that, as long as diets are adequate for other nutrients(i.e., Lys), dietary NE is adequate to predict changes in ADG across different dietary ingredients and conditions. The analysis indicates that ADG increases with increasing dietary NE and BW but decreases when BW is above 87 kg. The G:F ratio improves with increasing dietary NE and fat but decreases with increasing BW. The regression equations were then evaluated by comparing the actual and predicted performance of 543 finishing pigs in 2 trials fed 5 dietary treatments, included 3 different levels of NE by adding wheat middlings,soybean hulls, dried distillers grains with solubles (DDGS; 8 to 9% oil), or choice white grease (CWG) to a corn-soybean meal-based diet. Diets were 1) 30%DDGS, 20% wheat middlings, and 4 to 5% soybean hulls (low energy);2) 20% wheat middlings and 4 to 5% soybean hulls (low energy); 3) acorn-soybean meal diet (medium energy); 4) diet 2 supplemented with 3.7% CWG to equalize the NE level to diet 3 (medium energy); and 5) a corn-soybean meal diet with 3.7% CWG (high energy). Only small differences were observed between predicted and observed values of ADG and G:F except for the low-energy diet containing the greatest fiber content (30% DDGS diet), where ADG and G:F were overpredicted by 3 to 6%. Therefore, the prediction equations provided a good estimation of the growth rate and feed efficiency of growing-finishing pigs fed different levels of dietary NE except for the pigs fed the low-energy diet containing the greatest fiber content.