Statistics are essential in scientific research for analyzing data and drawing meaningful conclusions. However, it is not uncommon to come across publications that reach incorrect conclusions due to an inadequate application of statistical methods. One such example is the recommendation made by various studies to feed pigs using low protein diets supplemented with amino acids (AA). A more adequate reinterpretation of these studies suggests that this practice may be doing more harm than good, leading to an increase in carbon footprint, reduced protein production efficiency, and higher production costs. In this post, we’ll explore some of the risks of relying on poorly-designed statistical studies and highlight the importance of applying appropriate statistical methods in animal nutrition research.
In two recent literature reviews [1, 2], it was concluded that 3 to 4 % reductions in crude protein (CP) relative to current requirements [3] result in no negative consequences on animal performance if diets are supplemented with essential AA that meet animal requirements. However, rather than solid evidence, these claims appear to be based on inadequate statistical methods and interpretations used by the reviewed experiments. The first issue identified in the reviewed papers is the acceptance of a null hypothesis in underpowered studies. The papers listed in Table 1 decreased a 3% CP while supplementing AA, and in average a similar reduction in weight gain to feed (G:F) ratio was observed. These differences in G:F ratio were ignored as the statistical test used in the papers did not show any statistically significant effect. Nevertheless, the studies did not have enough experimental units to detect these differences in performance (low statistical power). The power tests performed on the data from the papers in Table 1 revealed that these studies had only one-tenth of the experimental units required to reach adequate conclusions. As a result, concluding that no negative effects were observed between the control and low CP + AA groups was inadequate, because these studies lacked the means to detect these differences in the first place.
Table 1. Studies with a small sample size concluded that reduced CP + AA performed similarly to the control groups.
Another identified problem is that some studies evaluated the effects of low CP diets + AA at levels that cannot be considered as low. As shown in Table 2, different CP levels were considered low, despite the fact that these diets were not protein deficient according to National Research Council [3] requirements. In this case, the studies concluded that low protein diets result in comparable animal performance to higher protein levels if supplemented with AA; however, this claim does not appear to be accurate, as all experimental diets had adequate to excess levels of CP. The identified methodological problem in these studies is not related to the statistical method used, but rather to a flawed research hypothesis.
Table 2. Studies that concluded that low CP diets had similar effects to control diets but both diets provided protein in excess.
In two additional studies, both the low CP diet and the control diets exhibited a similar degree of insufficiency in meeting the necessary animal needs. This study [14], found that lowering CP from 23.1 to 18.9% while supplementing with AA resulted in comparable performance. In this case, both diets produced similar G:F ratios because they provided excesses and deficiencies in the same proportion to the requirement, as shown in Figure 1. Likewise, in another study [15], 17% CP diets resulted in a similar pig performance of 20.2% CP with a requirement of 18.8% CP; in this case, the two diets provided 1.3 and 1.8% CP levels above and below the requirement, respectively. As a result, comparable performance in these two previous examples cannot be attributed to AA supplementation, but rather to both diets being insufficient in the same proportion relative to the requirement. As illustrated in Figure 1, identifying this problem requires a simple data visualization exercise; however, traditional statistics focused on modeling rather than visualization.
Another important problem identified in the field of low CP diets supplemented with AA is the use of inadequate statistical models. The study cited as reference [16] incrementally reduced CP in diets while supplementing AAs, and their data show a linear decrease in G:F ratios, but they incorrectly concluded that the 20% and 17% CP groups performed similarly because they used an ANOVA test instead of linear regression (Figure 2). Similarly, in these studies [17-19] there were linear decreases in performance due to incremental reductions in CP regardless of AA supplementation, but because they were analyzed using ANOVA rather than linear regression, the differences were not detected and it was concluded that the performance of the different groups was comparable. In contrast, other studies [20, 21], provided decreased levels of CP supplemented with AA, but used linear regression and detected a linear decrease in performance. An ANOVA is an adequate statistical test when the independent variable (x-axis) is categorical, and the dependent variable (y-axis) is numerical. In these examples, both variables are numerical, for which an ANOVA test is an inadequate model to use.
In another study [22], pigs were provided with CP levels ranging from 13.24 to 19.24% CP supplemented with AA and reported a quadratic effect in daily weight gain. However, the authors preferred to use a linear plateau model, and concluded that diets above 14.23% CP performed similarly (Figure 3). This instance demonstrates how analyzing the same data with varying models can lead to different conclusions. It also highlights the notion that if we torture the data enough, it will confess to anything.
The two previously mentioned reviews [1, 2], concluded that reductions in CP supplemented with crystalline AA result in no negative effects on animal performance, but it seems that mainly because there are more studies reaching into this conclusion than not. In these two reviews, a greater number of studies found no differences in performance when CP is reduced and AA is supplemented, not because it is necessarily a truly objective physiological response, but because more studies with methodological flaws have been published than studies with properly executed methods. Publishing scientific articles with methodological flaws can create a lot of noise in the animal sciences field. These flawed studies can lead to conclusions that are not supported by reliable evidence, potentially leading animal production down a path that is not beneficial for the sustainability of animal agriculture. For example, a common argument for feeding crystalline AA-supplemented low protein diets to pigs is that it reduces N excretion, which reduces the carbon footprint of animal production and thus improves animal production sustainability [1]. However, it appears to be the exact opposite (as discussed in this other post). According to previously reviewed data, a 3 to 4% decrease in CP results in a similar decrease in G:F efficiency, though this decrease is frequently overlooked. Neglecting this decline in productivity can have a negative impact on animal sustainability by increasing the amount of feed required to produce a unit of pork. This increases production costs and the carbon footprint of each pork unit, resulting in a less sustainable animal agriculture with a greater environmental impact and resource usage. Several studies, however, have shown that reducing protein intake, particularly during the nursery period, improves intestinal health [1]. This benefit may outweigh the decrease in performance; however, direct comparisons have yet to be determined.
Finally, extensive research has shown that insufficient essential AA intake results in decreased performance [3]. As a result, it appears critical to formulate pig diets that meet both the requirements of CP and essential AAs in order to avoid any performance declines, at least until more adequately designed studies show that low CP can be supplemented with AA with no associated negative consequences.
Final remarks:
The prevalence of false conclusions due to poor experimental designs and data interpretation in scientific studies is staggering. Animal science is no exception, as many unreliable conclusions are drawn from poorly conducted experiments in this field. This blog will look at different areas of animal science where these issues are prevalent. The goal is to not only identify these problematic studies, but also to thoroughly investigate each statistical method in order to avoid similar issues in the future. Future posts will delve into the appropriate use of each statistical test, the level of reliance that should be placed on p-values, the execution of power tests, and the use of data visualization and other analytic techniques.
Questions for the reader:
Flawed studies concluding that lowering CP while supplementing with AA have no negative effects on animal performance have been cited in the scientific literature hundreds of times. Is it possible for a statement with no scientific basis (hard data) to become true after being repeated so many times?
Do you have any additional information or data regarding the effects of low protein diets supplemented with AA on pig productivity that has been overlooked in this post? If you have any such information, kindly share it with us to enhance our comprehension.
Thanks for reading, and I hope you found this post helpful!
Christian R
References
1. Wang, Y., et al., Advances in low-protein diets for swine. Journal of animal science and biotechnology, 2018. 9(1): p. 1-14.
2. Rocha, G.C., M.E. Duarte, and S.W. Kim, Advances, Implications, and Limitations of Low-Crude-Protein Diets in Pig Production. Animals, 2022. 12(24): p. 3478.
3. NRC, Nutrient Requirements of Swine: Eleventh Revised Edition. 2012, Washington, DC: The National Academies Press. 420.
4. Kerr, B., et al., Influence of dietary protein level, amino acid supplementation, and dietary energy levels on growing-finishing pig performance and carcass composition. Journal of animal science, 2003. 81(12): p. 3075-3087.
5. Cho, J.H., et al., Effects of reducing dietary crude protein on growth performance, odor gas emission from manure and blood urea nitrogen and IGF‐1 concentrations of serum in nursery pigs. Animal Science Journal, 2008. 79(4): p. 453-459.
6. Nørgaard, J. and J. Fernández, Isoleucine and valine supplementation of crude protein-reduced diets for pigs aged 5–8 weeks. Animal Feed Science and Technology, 2009. 154(3-4): p. 248-253.
7. Zhang, S., et al., Supplementation with branched-chain amino acids to a low-protein diet regulates intestinal expression of amino acid and peptide transporters in weanling pigs. Amino acids, 2013. 45: p. 1191-1205.
8. Tang, W., et al., Effects of Bacillus subtilis DSM32315 supplementation and dietary crude protein level on performance, gut barrier function and microbiota profile in weaned piglets. Journal of Animal Science, 2019. 97(5): p. 2125-2138.
9. Rattigan, R., et al., Effects of reducing dietary crude protein concentration and supplementation with either laminarin or zinc oxide on the growth performance and intestinal health of newly weaned pigs. Animal Feed Science and Technology, 2020. 270: p. 114693.
10. Atakora, J.K., S. Moehn, and R.O. Ball, Enteric methane produced by finisher pigs is affected by dietary crude protein content of barley grain based, but not by corn based, diets. Animal feed science and technology, 2011. 166: p. 412-421.
11. Morales, A., et al., Low-protein amino acid–supplemented diets for growing pigs: Effect on expression of amino acid transporters, serum concentration, performance, and carcass composition. Journal of Animal Science, 2015. 93(5): p. 2154-2164.
12. Nguyen, D., et al., Influence of low-protein diets and protease and bromelain supplementation on growth performance, nutrient digestibility, blood urine nitrogen, creatinine, and faecal noxious gas in growing–finishing pigs. Canadian Journal of Animal Science, 2018. 98(3): p. 488-497.
13. Kim, Y.J., et al., Effect of low protein diets added with protease on growth performance, nutrient digestibility of weaned piglets and growing-finishing pigs. Journal of Animal Science and Technology, 2021. 63(3): p. 491.
14. Yue, L. and S. Qiao, Effects of low-protein diets supplemented with crystalline amino acids on performance and intestinal development in piglets over the first 2 weeks after weaning. Livestock Science, 2008. 115(2-3): p. 144-152.
15. Duan, Y., et al., Effects of supplementation with branched-chain amino acids to low-protein diets on expression of genes related to lipid metabolism in skeletal muscle of growing pigs. Amino Acids, 2016. 48: p. 2131-2144.
16. Gloaguen, M., et al., The use of free amino acids allows formulating very low crude protein diets for piglets. Journal of animal science, 2014. 92(2): p. 637-644.
17. Powell, S., et al., Growth performance of 20-to 50-kilogram pigs fed low-crude-protein diets supplemented with histidine, cystine, glycine, glutamic acid, or arginine. Journal of animal science, 2011. 89(11): p. 3643-3650.
18. Zhang, S., et al., Effects of dietary leucine supplementation in low crude protein diets on performance, nitrogen balance, whole‐body protein turnover, carcass characteristics and meat quality of finishing pigs. Animal Science Journal, 2016. 87(7): p. 911-920.
19. Peng, X., et al., Effects of low-protein diets supplemented with indispensable amino acids on growth performance, intestinal morphology and immunological parameters in 13 to 35 kg pigs. Animal, 2016. 10(11): p. 1812-1820.
20. Guay, F., S. Donovan, and N. Trottier, Biochemical and morphological developments are partially impaired in intestinal mucosa from growing pigs fed reduced-protein diets supplemented with crystalline amino acids. Journal of animal science, 2006. 84(7): p. 1749-1760.
21. Nyachoti, C., et al., Performance responses and indicators of gastrointestinal health in early-weaned pigs fed low-protein amino acid-supplemented diets. Journal of Animal Science, 2006. 84(1): p. 125-134.
22. Toledo, J.B., et al., Reduction of the crude protein content of diets supplemented with essential amino acids for piglets weighing 15 to 30 kilograms. Revista Brasileira de Zootecnia, 2014. 43: p. 301-309.
I can’t disagree with your critique of statistical methods in these studies. I can strongly disagree with your conclusion “low protein diets for pigs: the ultimate solution for making pork more expensive”. You presented zero economic data to support this conclusion. The reduced crude protein diets are substantially cheaper, so even if G;F is reduced (disputable) the cost per pound of gain can is usually improved.
I am a PhD swine nutritionist with over 30 years experience in research, and application of research, in large and small scale pork production systems in the US and abroad. A very high percentage of production systems across the globe have been using reduced crude protein diets supplemented with synthetic amino acids for decades, and continue to do so today. They would not be doing this if it were not economically advantageous.
I encourage you to expand your literature review. There are dozens of papers published on this topic. You only cited a few. In my experience, studying the literature, and doing my own studies in commercial research units, G:F is improved when diets with reduced crude protein are fed. Theoretically this makes sense as some of the soybean meal is replaced by corn, resulting in higher NE in the diets with reduced crude protein. NE and G:F are highly correlated. So why would reduced crude protein diets have lower G:F?
As you study the literature it is important to critically assess the levels of secondarily required amino acids relative to lysine used in the experimental diets. In order for the reduced crude protein diets to work properly it is critical that sufficient levels of other essential amino acids are present. We continue to be challenged in this regard as more and more synthetic amino acids are commercially available and economically feasible.
Hello Chad! thank you for your comments. I believe that your comment captures the spirit of this post. We should question everything we read. In this post I just included the papers with methodological problems reported in the two literature reviews described. The other papers concluded that low CP diets + AA resulted in a similar reduction in performance, except when low CP was not really low, and in this case reducing CP results in a lot of savings. One important problem with the discussion of low protein diets, is that the term low is subjective. Low can be anything, that is why I use the NRC recommendations as a reference point. There are a lot of papers that show that low protein diets produce benefits, but are they really low in protein (below the requirement)? do we have accurate estimates of protein requirements?
According to my rough calculations, a 3% reduction in CP in 100 kg of feed (corn + soybean meal) would save $4.30, and a 3% reduction in G:F would result in a loss of only $3.50 in live weight. However, other costs should also be taken into account. Depending on the production system, the calculation of this cost may be complex; however, if we consider that feed accounts for 60 to 80% of the total cost, the producer would lose between $4.30 and $6.30 (opportunity cost and others). Depending on how prices change, this strategy could lead to gains, but based on my calculations, it’s more likely to lead to losses. My math could be wrong, though.
The data I reviewed clearly shows that AA supplementation improve diets that are deficient in CP. But not to the level of a diet that is adequate both in AA and CP. Certainly there are points where the reduction in CP may result in benefits, however do we have the capacity to quantify that in an experiment? do we have the capacity to quantify the opportunity costs of these strategies in commercial barns?
Again thank you for your comments!
CR
Good debate Christian,
“Low” is subjective. “Reduced” is objective. The terminology should be “reduced” crude protein, not “low” crude protein. When we had only synthetic lysine available commercially you could only reduce crude protein in a finishing diet by about 1%. Now that we have not only lysine, but methionine, threonine, tryptophan and valine that are commercially available we can drop crude protein about 4% in finishing diets.
Using today’s actual ingredient costs the cost, difference between a 20% protein mid-finishing diet and a diet with 16% protein formulated to the same essential amino acid requirements is approximately $15/ton in favor of the reduced crude protein diet.
The studies you used in your blog were cherry picked to show reduced G:F with “low” protein. As I pointed out above, this is flawed, and if you look at the entire body of data you will see that when protein is reduced and amino acids are formulated to the proper levels relative to lysine, energy, and stage of growth, reduced protein diets tend to have better G:F, due to increased NE. For sake of argument, let’s say that G:F is not affected by crude protein level. Using a 2.65 F/G from 50 to 280 lbs. body weight, feed cost for pigs on the reduced crude protein diets is $107.27/head and feed cost on the higher crude protein diets is $111.87/head. A difference of $4.60/head in favor of reduced crude protein diets.
If reduced crude protein diets were not superior economically, why are they so widely used across the global swine industry? Nutritionists in the feed and production industry are highly scrutinized and most have done their own research on this topic. They would not be using this technique if it did not work.
Christian,
You might want to go back and read the paper that is first in your bibliography (1. Wang, Y., et al., Advances in low-protein diets for swine. Journal of animal science and biotechnology, 2018. 9(1): p. 1-14.)
Yang states in his abstract:
“Recent years have witnessed the great advantages of reducing dietary crude protein (CP) with free amino acids (AA) supplementation for sustainable swine industry, including saving protein ingredients, reducing nitrogen excretion, feed costs and the risk of gut disorders without impairing growth performance compared to traditional diets.”
How does this fit with your narrative?
Thanks for your interesting comments Chad. I guess I did cherry picked the studies with methodological problems, but sadly, all the others show that low CP diets + AA below the requirements do not result in performance improvements. When you say that going from 20% to 15% CP saves money, it could be because 15% CP is not a low protein diet. How do we know 15% is a low protein? (NRC recommends even less than that). The fact that we supplement a 15% CP diet with crystalline Lysine, for example, is not to match a 20% CP diet, it is simply because that diet may be adequate in protein but inadequate in Lysine. So formulating diets adequate in protein, while supplementing AA to meet the requirements is a good practice. But the claim that we can reduce protein and replace it with AA despite the protein level is a claim that may causing more harm than good. Especially if we don’t know how our diets compare to the optimal CP level. Regarding the conclusion of the meta-analysis, it is very common to find meta-analysis that evaluate conclusions and not data, or include studies that reach incorrect conclusions because of poor methods. So the narrative of the first paper mentioned in this post does not fit my narrative because, in my opinion, that conclusion is based on inadequate analyses.
So your approach was to cherry pick, then nitpick the statistics to prove that low protein diets “make pork more expensive”. You didn’t prove that.
It is very likely that I can go back into the studies you cherry-picked and nitpick the diet formulas and analyses to show a nutritional reason why G:F was worse in the reduced protein diets. Here’s one example: In the Kerr study, analyzed lysine levels were lower in the reduced protein diets when the experimental design mandated that they be the same. Oops. Now we are comparing lysine levels, not the concept of reduced crude protein. Dr. Kerr is one of the best researchers I know. Stuff happens. Proper statistical analysis is critical to good research. So is experimental design, diet design and execution.
I encourage you to continue to critique statistical analysis of published research. Just be careful to look at the big picture and avoid making bold, unfounded statements like the title of this blog.
Thanks for your comments Chad! I appreciate your opinion. My goal is not to target researchers, but rather methodologies. I avoid reading who conducted the experiment in order to avoid falling victim to authority bias. I cannot trust the results of an experiment involving a variable with a statistical power of 20%, regardless of the author. I hope people don’t believe my analysis and conduct their own, as well as learn about data analysis to refute statements like the title of this post, preferably in a quantitative rather than qualitative manner.
Hi Christian, what will be your recommendation on how to set up a trial? First, what is your hypothesis about diets with reduced protein levels? how would you evaluate your hypothesis? How many reps would you need for each of your parameters of interest? which statistical analysis would you use? and how would you assess the economics?
Hi Santa Maria! Thanks for your comment. I believe, there are two distinct approaches to AA supplementation in swine nutrition: low protein diets with added AAs (Strategy 1) and adequate protein diets +AA, also called functional AAs (Strategy 2). In my opinion, Strategy 1 is based on mathematical efficiency calculations and a willingness to accept null hypotheses, with the conclusion typically being that adequate and low crude protein diets perform similarly. While this approach may save on costs, a major challenge facing the swine industry is that one-third of pigs do not reach the market. Thus, Strategy 2 is superior, as it has been shown to improve metabolic status, survival, reproduction, gut health, and other variables. Compared to Strategy 1, Strategy 2 provides more benefits in terms of animal welfare and economic returns. However, embracing this approach requires a paradigm shift, as it involves investing in more expensive diets that yield greater pork production. I’ve spoken with a few researchers who are still skeptical of the concept of functional AAs, so it is not just a matter of designing new experiments, but also implementing existing data. I hope we can find time later to discuss the execution of some experiments; I will also write more about this topic in future posts. Thank you once more for your comment.
What is the definition of an adecuate proteine diet?
That is what we must define before we can call a diet Low protein. A good start is the protein requirement of NRC (Nitrogen requirement); however many nutritionist believe there is not a protein requirement, but a minimum before noting negative effects. The issue is that, due to the low power of the studies, we can begin to notice negative effects when they are too large. In other words, if we want to notice an effect, we can either increase the number of animals or wait until the effect is too large to notice. Basically because of the low power we do not have the resolution to identify the problem.