WHAT MAKES AN EXPERIMENTAL FEEDING TRIAL SUCCESSFUL IN RESEARCH IN AQUACULTURE NUTRITION ?
Experimental feeding trials are the foundation of research to advance aquaculture nutrition science and the aquafeed industry. In vivo feeding trials are resource-intensive activities and therefore must be carefully designed to ensure that results are scientifically valid and contribute to the overall body of knowledge in the aquaculture nutrition field. Nevertheless, many fish nutrition studies published in recent years lack the central element that defines a scientifically valid study – careful and concise articulation of a scientific question followed by development of a testable hypothesis and a study design appropriate to test it. The researchers must couple their analytical skills and capacity with study designs that make these skills work for a solid hypothesis rather than developing a hypothesis that is dictated by analytical capacity. In this opinion article the authors review the steps that make experimental feeding trials successful and describe the pitfalls and traps that must be avoided to ensure that resources devoted to experimental feeding trials yield information that justifies their investment.
1. Introduction
Feeding farmed fish is central to the potential environmental sustainability, economic viability, and social acceptability of aquaculture. Achieving these constantly evolving goals, which are directly or indirectly affected by various external forces such as markets, resource availability, global changes, including climatic and socio-political, biotechnological innovations, knowledge gains and competition, requires the involvement and evolution of discipline commonly referred to as “aquaculture nutrition.” This field has progressed from an artisanal practice in the early days of aquaculture to a pioneering discipline in the mid-twentieth century, and has now matured into a highly developed scientific and technological discipline (Hardy et al., 2022).
Research activities in aquaculture nutrition aim to acquire knowledge and develop feasible solutions to a variety of technical problems, ultimately improving current practices. Researchers in this field have access to several tools, with in vivo feeding trials being one of the most powerful and informative options. Briefly, in in vivo feeding trials, individuals from a study species are fed experimental diets. The results obtained from these trials are then utilized to answer specific research questions and may lead to the discovery of new knowledge and insights that can be applied to improve aquaculture nutrition (Glencross et al., 2007). This approach has served the aquaculture industry well and characterized the history and the rapid evolution of fish nutrition science (Jobling, 2016).
Despite their effectiveness and the fact they are the most commonly used investigative tools, feeding trials also come with significant challenges. They are relatively expensive, time-consuming, labor-intensive, and complicated (Glencross, 2020), requiring the use of live animals and relative ethical considerations and constraints (Sloman et al., 2019), and, more often than not, yielding inconclusive results.
Recognizing the limitations as well as the potential for misuse of this analytical tool, the authors of this opinion paper believe it would be beneficial to the sector, particularly to graduate students and junior colleagues beginning their research journey, to take a moment for reflection on what strategies can enhance the success of an experimental feeding trial in aquaculture nutrition. As experimental feeding trials represent a substantial investment of time, effort and research funds, it is tempting to design studies to answer as many research questions as possible in a single feeding trial. The potential pitfall of this approach is that such a study may not clearly answer any questions.
Before analyzing the dos and don’ts toward maximizing the chances of achieving a successful in vivo trial, we thought that taking a step backwards was required. We believe it is essential to start with an emphasis on the importance of articulating a valid scientific question before entering into reasoning about experimental design, performance targets and objectives to achieve meaningful outcomes. Often published studies utilizing in vivo feeding trial seem to be missing an explicitly presented scientific justification and clearly phrased hypothesis.
2. The prerequisite: a valid scientific question
A valid and clearly stated scientific question is a prerequisite for any meaningful scientific experiment, and this principle holds true in the field of aquaculture nutrition.
Considering the aforementioned challenges of implementing in vivo feeding trials, it should be obvious that a well-defined, justifiable scientific question must be formulated before embarking on such a demanding exercise. Nevertheless, we frequently observe that certain research activities in the broad discipline of aquaculture nutrition adopt a “feed and see” approach or justify experiments with the claim that “it hasn’t been done before.” Neither of these approaches is a scientific question or hypothesis, and therefore neither should be an acceptable starting point to design nor undertake a feeding trial. This is not a simple semantic lucubration, but rather a pivotal step in the design of the experiment that determines the likely achievable final results of the research effort.
When testing a hypothesis, we are defining the null hypothesis (there is no treatment effect) and the alternative hypothesis (there is a detectable treatment effect). By doing so, we are, first of all, reducing the number of questions and possible answers: either one or the other. This gives already an incredible power to the meaningfulness of our study compared to leaving the door open to testing several questions which typically results in numerous unlikely and inclusive answers. In fact, when testing a hypothesis, at the end of a study, one should be able to make one of two conclusions: accept the null hypothesis or accept the alternative hypothesis.
In contrast, the “feed and see” method involves recording outcomes of a given treatment, typically including a feed ingredient in aquafeed, without a hypothesis based on any solid justification, such as cost-effectiveness, nutritional content, or metabolic impact. The stereotypical “feed and see” experiment is commonly justified as testing an apparently “novel raw material X” to replace a currently utilized raw material such as fish meal or fish oil, which have been successfully utilized in commercial aquafeed for many years, but are in limited supply and becoming increasingly expensive (Turchini et al., 2019). Often, such experiments achieve inconclusive or confounded results. Nevertheless, results will be analyzed, presented, and discussed with authors trying to justify why the “novel raw material X” can or cannot be successfully included in aquafeed formulations. Very commonly authors fail to revisit their original rationale for conducting the study, if it was ever explicitly expressed, in their discussion and conclusion, resulting in tediously long discussions of each measured dependent variable, how they confirm or contradict published findings of other authors, and, in the absence of clear results, what trends are observed. No certain answer is achieved, and no usable information is generated. This kind of study is unlikely to be widely read, even less likely to be cited and almost certainly not going to be of any use for the industry. In essence, these studies are implausible to be of any practical or scientific value that advances the body of aquaculture nutrition knowledge.
If a valid scientific question, or hypothesis, had been formulated before the trial, things would have likely progressed in a different direction, and the same study could have been steered toward a successful outcome. Whilst the readers of this opinion might, by now, get the impression that the authors are “two grumpy old men” expressing their criticisms toward the research of others, our intentions are that of providing our suggestions toward increasing the likelihood of success of feeding trials in aquaculture, to the benefit of authors, researchers, the aquaculture industry and, importantly, to advance the discipline of aquaculture nutrition.
3. The advantages of moving from “feed and see” to a hypothesis, and how to avoid the “it hasn’t been done before” trap
Following on from the example described above, the hypothesis could have been that the “novel raw material X” can substitute for dietary fish oil without any effect on fish growth, health and final quality (null hypothesis), or it cannot, as it will affect growth and final fatty acid composition (alternative hypothesis). The experimental trial could have been based on a single experimental diet compared to the control, fish oil based diet, but run over a sufficiently long period of time to reliably assess possible effects on growth and whole body and tissues’ fatty acid turnover. At the end of such a trial, growth performance and feed efficiency parameters should be measured, as well as tissues fatty acid composition. The results of such a trial will either prove the null or the alternative hypothesis, providing very useful information to anyone interested in dietary lipids in aquafeed.
Alternatively, another hypothesis could have been that the “novel raw material X,” which in this second example has a protein content and an amino acid profile similar to that of fish meal, can substitute for only a fraction of fish meal because it has negative palatability attributes and limited digestibility (null hypothesis), or it cannot, as its partial inclusion will result in a measurable effect (alternative hypothesis). In this case, then, this can be combined with a secondary hypothesis, that the degree of inclusion of the “novel raw material X” substituting fish meal will not have any effect on feed intake and digestibility (null secondary hypothesis), or it has an effect (alternative secondary hypothesis). Consequently, the experiment should have been designed with gradual inclusion levels of “novel raw material X” replacing fish meal and could be conducted for a relatively short period of time, sufficient to measure palatability and feed intake and to collect enough feces for digestibility estimation. In contrast to the first example, running the trial over a long period of time and measuring growth indices and fatty acid composition of tissue would be meaningless, because it does not address the questions articulated by the two hypotheses. In fact, in this case, reliable and conclusive evidence testing the primary and secondary hypothesis will be achieved by the feeding trial followed by the analysis of voluntary feed intake and nutrient digestibility, and whatever the result will be, it will be conclusive and will positively contribute to the aquaculture nutrition field.
The two examples above clearly show that having a simple, clear, and well-defined scientific question will enable researchers to achieve a robust and substantiated answer, but also how the experimental design, including number of treatments, duration of the trial and subsequent analysis, must be modified relative to the scientific question. In other words, the study design must match the hypothesis for which the study is intended to address, rather than what is often the case where the study design is determined by experimental facilities that are available, such as number and size of replicate tanks, or analytical capacity of the research laboratory.
As mentioned earlier, often the “feed and see” method is also accompanied by the “it hasn’t been done before” justification, and this approach should be promptly questioned by responsible researchers. First, in most cases it is not true, as, for example and still utilizing the “novel raw material X” example, a quick literature search will confirm that the “novel raw material X” is not that novel, someone had already tested it, and results were somewhat negative or inconclusive. Second, if it is true that this wasn’t previously tested, before embarking on uncharted research territory, one should inquire why it hasn’t been explored before. Maybe the supply of “novel raw material X” is extremely limited, or the material is very expensive, or contains high concentrations of antinutrients or toxic compounds, or any other reason that would severely limit its use in aquafeed formulations. This simple step can save valuable resources, including time, money, and, importantly, safeguard animal welfare.
A second common “it hasn’t been done before” justification is relative to the species being studied. The sentence “this is the first study on ingredient X in species α” is unfortunately turning into a mantra in aquaculture nutrition literature. This isn’t always a fatal flaw, but the burden on the author is to identify reasons why performance of species α might differ from species β, which has already been the subject of one, if not many, feeding trials with ingredient X. What is different about the physiology or feeding behavior of species α compared to species β? Why is this study justified?
4. The Maslow’s Hammer and “let’s measure it all”
In some instances, researchers seem to prioritize conducting feeding trials as their ultimate goal, and then seek justifications afterward. This is a good example of the law of instrument, also known as the Maslow’s Hammer. The law of instrument is a well-known cognitive bias resulting in the excessive dependency on a familiar tool, and it is exemplarily summarized by the famous sentence: “if the only tool you have is a hammer, you tend to see every problem as a nail” (Maslow, 1966).
Researchers, who have invested significant effort, time, money and resources in establishing an experimental aquatic animal house facility and possessing a particular array of analytical tools and capability, following the law of instrument, tend to see all aquaculture nutrition related problems as answerable by a in vivo feeding trial. This, of course, can lead to problematic research practices. Instead, researchers should start with a valid scientific question, and then, if a feeding trial is necessary to answer that question, it becomes a valuable tool in the research process, but running an in vivo feeding trial should never be the ultimate objective of the research process itself.
Consistently with the need of being able to justify a feeding trial, despite the lack of a hypothesis, and aiming at maximizing the chances of seeing something when running a “feed and see” experiment, an increasingly common practice that we have observed in many recent published studies is the expansion of dependent variables, with the over-reliance of a plethora of analytical tools and analysis. This approach, which we can call “let’s measure it all,” is a perfect match with the previous considerations, and the resulting research process becomes:
1. “it hasn’t been done before”
2. “feed and see”
3. “let’s measure it all”
Astronomically expensive feeding trials, using thousands of animals, over long period of times, and collecting hundreds of different samples, to be analyzed for anything possible, from advanced analytical chemical analyses to all the “omics” and molecular testing, are then implemented without having any hypothesis and often utilizing large numbers of postgraduate students working very long hours. Whilst there are educational and training benefits of such activities, with current postgraduate students becoming extremely proficient in running complicated analytical procedures, some might question the utility of such efforts. Further, it places the focus of education and training on laboratory competency. Although this is certainly important, it is equally or even more important to train students to think critically about scientific questions and formulate a valid scientific hypothesis before embarking on a research project.
The readers of this opinion piece might have noticed that in the above sections we have not provided bibliographic references to back up our statements. This is not an oversight, but rather intentional, as we believe it is not productive to point fingers at any specific studies that have, in our opinion, exemplified these occurrences. But we are confident that many avid readers of scientific literature in the discipline of aquaculture nutrition would most likely recognize this issues and concur with us.
studies focusing on functional ingredients. The dependent variables in these cases often include any and all clinical chemistry measurements that the researchers have access to (often far too many, thanks to relatively easy institutional access to the nearby medical facilities). These measurements are implemented independently from their relevance to the functional ingredient’s supposed benefit. In similar fashion, studies focusing on novel raw materials also use a similar “let’s measure it all” approach, resulting in claims that such given raw material improves immune function. Surprisingly, however, the same raw material, supposedly beneficial to the immune system, rarely if ever shows any demonstrable effect on survival, either in disease challenge trials or in field trials with natural exposure to common pathogens.
Another, Maslow’s hammer afflicting aquaculture nutrition, as well as many other fields, is the use of gene transcription rate, more often than not, as the sole dependent variable. Without entering into further details, the authors simply like to draw the attention of the readers to the fundamental concept of gene transcription rate: it is a measurement of RNA, not proteins, nor their activities within any given pathway (Panserat and Kaushik Citation2010). Whilst the information achieved by gene transcription assessment could be highly beneficial to help understand physiological responses of dietary interventions, without additional chemical, enzymatic, or physiological confirmations, it is fundamentally only an indication of the cell internal communication signal of requesting increased synthesis of that given protein. Consistently, the fluctuating and relatively poor correspondence between gene expressions and protein levels has been showed in a variety of species (Kosti et al. Citation2016; Li et al. Citation2014). Furthermore, gene expression in samples taken during or at the end of a feeding trial measure RNA expression at a single point in time, and therefore it is a kind of a snapshot of activities at a single moment of a dynamic and extended process (Panserat and Kaushik Citation2010). As such, such measurements must be interpreted cautiously given that metabolic activity resulting from elevated gene expression is a dynamic process. In this manner, gene expression as a dependent variable differs from common measured or calculated dependent variables such as weight gain, thermal growth coefficient, feed intake and feed conversion ratio.
Summing up, aquaculture nutrition research should always begin with a valid scientific question. If that question requires a feeding trial to achieve the answer, then researchers should proceed accordingly toward developing an adequate experimental design, including the selection of appropriate dependent variable that will be measured.
5. The experimental design: dos and don’ts
Once a valid scientific question is established and researcher determine that a feeding trial is necessary to address it, careful consideration of experimental design becomes paramount. While avoiding an in-depth discussion of experimental intricacies, and consciously omitting all statistical relative considerations and suggesting readers to refer to the abundantly available specialized literature (Casler, 2015; Kaps and Lamberson, 2017; Quinn and Keough, 2023), in the context of aquaculture nutrition research and in vivo feeding trials, there is always a fine balancing act to be faced: the ideal statistical needs, in terms of independent experimental units, number of replicates, number of fish and samples, and the ultimate power of the statistical test, clashing with the logistic and practical constraints due to what is feasible in the available experimental aquatic animal housing facility. Because of this, key factors to contemplate and ponder upon include the definition of experimental treatments and the experimental diet formulations.
With a properly formulated and explicit scientific hypothesis, the resulting definition of the treatments is a relatively simple exercise. An often forgotten but very important principle is that the total number of treatments should be the fewest number required to unequivocally test the hypothesis. Adding a treatment, just because additional rearing units are available or in the hope doing so might provide additional data and some additional discovery might result is a big mistake. Not only will the addition of an additional treatment be unlikely capture much, most likely it will add noise and dilute the strength of the study and confound the original question when the authors write their research paper. If researchers have available tanks in their experimental system after assignment of tanks to the feeding trial, remaining tanks would be better used to increase the number of replicates rather than add a treatment or treatments, as increasing replication will increase the power of the statistical test. Also, whilst it might be an obvious point, the authors believe there is value in reminding all researchers that the inclusion of a control/reference treatment is fundamental to well-designed feeding trials.
5.1. The power of the statistical test: more replicates, less treatments
If a well-developed hypothesis is formulated, often a test diet and a control diet will suffice. But feeding trials in aquaculture nutrition scientific literature that include only two treatments, a control/reference diet and a test diet, are as rare as hen’s teeth. That is because, we believe, most experimental aquatic animal houses have multiple tanks (typically from 12 to 24 and in multiple of threes). Researchers unfailingly want to use all of their available tanks and by doing so, add unnecessary additional treatments that will most likely result in diluting the meaningfulness of the trial itself. In fact, the number of available tanks dictates the replication of experimental treatment groups. Whilst, as mentioned before, we refer readers to the numerous scientific publications and scholarly books for comprehensive guidance on these statistical aspects and considerations, we must again remind readers about the power of the statistical test, and to plan for adequate replication of experimental units to enhance statistical validity. The most powerful experimental design does not necessarily require three replicate tanks per treatment group. Power testing often calls for more than three replicate tanks. In essence, our recommendation is to use as few treatments as possible, and to expand the number of replicates. This will increase the power of the statistical test and thus reduce the risk of a Type II error (false negative), which occurs when the test fails to reject the null hypothesis when it is actually false. A call for attention to the power of the statistical test in aquaculture research was made over thirty years ago by Searcy-Bernal (1994). An important call that unfortunately has thus far received relatively limited attention. In fact, in the literature it is possible to find an extraordinarily large number of studies where a “novel raw material X” was included at Y % inclusion level to replace a commonly used raw material, such as fish meal, fish oil or, more recently, soybean meal. In all these cases, the null hypothesis is that there is no treatment effect. As such, at the end of the trial, when authors do not observe a statically significant difference in fish performances, they typically conclude that the “novel raw material X” can be safely utilized up to Y% to replace that given commonly used raw material (i.e., the null hypothesis is confirmed). Interestingly, though, the test treatment almost always results in lower numerical values for dependent variables, such as weight gain, compared to the control, but the P value cannot be questioned, and thus “novel raw material X” is promoted as a viable alternative. Yet, the power of the statistical test is very seldom reported and if it were reported, a different conclusion may well be made.
We believe that given the normal variability observed among replicate tank mean values in in vivo feeding trials, the standard use of triplicates and the use of a threshold p-value of 0.05 are the cause of a low power of the statistical test, and most of these studies and their conclusions are likely to be Type II errors. As such, we believe that to confidently state that “novel raw material X” can be used without affecting animal performance, the power of the statistical test should be reported. Readers are at this point suggested to explore the growing literature pointing at the “power vacuum,” and limitations of p-value, with the review of Halsey (Citation2019) being highly recommended. Without the need to embark into complicated statistical discussions, researchers are recommended to increase the number of replicates, run trials for longer periods of time (discussed below), and possibly also increase the p-value to 0.1, if not more. Then, if no statistically significant difference is observed, it will be possible to confidentially state that “novel raw material X” can be used in aquafeed formulation for that given species, in those environmental conditions and at that physiological of life stage.
Counterintuitively, however, the concept of using as few treatments as possible to address the hypothesis at times should result in larger number of treatments compared to many published studies, when, for example, the hypothesis is that our previously introduced “novel raw material X” can be included in feed formulation up to a specific inclusion level without any negative effects, or we hypothesize that there is an ideal inclusion level of “nutrient Y” for maximal positive effect. In such cases, multiple treatments (and in fish nutrition we would like to suggest at least five, but the more the merrier), with a gradual inclusion of the “novel raw material X” or “nutrient Y” must be implemented, and results then analyzed by regression analysis, and not multiple comparison test, i.e., ANOVA, as previously recommended (Francis et al., 2007; Yossa and Verdegem,2015).
5.2. The experimental diets
Once the experimental question has been defined and a hypothesis has been formulated, the design of experimental treatments is the next fundamental step that directly affects the achievable results. In aquaculture nutrition it is commonly accepted that experimental treatments in a feeding trial with “novel feed ingredient X” should first of all be nutritionally consistent with ideal or commercial formulations, and that all tested treatments must be iso-energetic, iso-proteic and iso-lipidic to minimize confounding variables (NRC, 2011). Interestingly, studies focusing on protein sources are in most cases correctly designed to be iso-proteic on a digestible nutrient basis; whilst, to the best of authors knowledge, not a single trial focusing on lipid sources has been designed and implemented to be iso-lipidic and iso-energetic on a digestible nutrient basis, greatly reducing the appreciation of the potentials of energy sources rich in saturated fatty acids (Mock et al.,2021; Turchini et al.,2024).
An often-overlooked aspect, when experimental diets are formulated, is their practicality. Authors should ensure that the proposed diets are technically feasible and that the resulting pellets can be realistically manufactured, possessing the essential chemical, physical, nutritional, and sensory characteristics suitable for the specific species. (Glencross et al. Citation2007, Turchini et al. Citation2019). This might be challenging for researchers. Typically, nutritionists possess sound biological and physiological knowledge and expertise, but have relatively little direct experience of the complexity of feed manufacturing and extrusion processes. Nevertheless, this assessment might be relatively simple for feed producers and formulators, and asking an experienced feed producer for advice is an easily implemented solution that might lead not only to more practically useful formulation of the experimental diets, but likely also to new partnerships and collaborative activities.
Once the experimental diets have been formulated to match the experimental treatments and the overlay experimental design, they are manufactured. Next, before commencing the in vivo trial, it is strongly suggested that the proximate and, in some cases, selected nutrient levels in the experimental diets should be analyzed to confirm that they match expected levels, and that no unexpected differences in nutritional characteristics can hamper the final results of the experiment. The authors have learned this lesson the hard way. Most fish nutritionists around the world undoubtedly have had similar experiences, where unexpected deviations from the proximate or nutrient contents of their experimental feed formulations were discovered after a feeding trial has ended and fish were already sampled. At this point, nothing can be done to rescue the experiment.
5.3. The feeding trial
After confirming that the proximate and nutrient contents of experimental diets match the calculated levels, the feeding trail can begin. Fish must be healthy and of the right size, often based not only on experimental design but also tank size and fish availability, and the rearing conditions should be the appropriate ones to match the experimental design. The feeding regime and method also must be carefully selected and implemented. In many studies, feeding the fish to apparent satiety is a common practice, but if we suspect, or observe, that “novel raw material X” has an effect on palatability and thus might affect feed intake, then we should also consider having a pair-fed, restricted treatment. Including pair-fed treatment groups will make it possible to separate effects of “raw material X” level in the experimental diets on feed intake from effects on fish growth and metabolism. Many studies that did not consider this have the value of their findings negatively impacted, as it is impossible to assess growth performance variation for animals with different feed intake.
Another important and often overlooked aspect of conducting feeding trials to evaluate the effects of the level of a feed ingredient on fish growth performance is that fish commonly reduce feed intake at the start of the study for a short period when their feed is switched from a standard, often commercial, feed they were fed before the study to experimental diets (NRC, 2011). This effect is particularly noticeable when the pre-study feed was fishmeal-based and the experimental feeds contain an oilseed protein or single-cell protein ingredient. In our observations, this initial decrease in feed intake disappears within a few days or weeks, and growth rates of fish fed the control and some of the experimental diet treatments becomes more-or-less identical. Importantly, however, this short-term early decrease in feed intake will reduce initial weight gain and affect the final weight of the fish, especially in short-term studies with fingerlings. For example, if fish fall behind in weight gain by 10% during the first two weeks of a 12-week study due to a decrease in feed intake before resuming normal intake equivalent to that in a control treatment group, this 10% difference in weight will persist until the end of the study and incorrectly be attributed to a difference in nutritional quality of “raw ingredient X,” rather than an initial feed intake difference caused by simply changing feed. In this case, implementing a pair-feeding approach will allow the researchers to detect and account for this phenomenon.
Throughout an in vivo trial, fish should be robust and active, free of signs of disease, and mortality should be absent, or extremely limited. At the conclusion of an in vivo feeding trial, the control treatment must display optimal growth performance and feed efficiency matching the expectations for the given species under those conditions.
5.4. Let them grow
Last, one of the most frequently overlooked aspects in aquaculture nutrition trials that can compromise the credibility and scientific utility of many studies is the duration of the feeding/growth trial. This timeframe should not be solely based on customary practices observed in other published experiments, typically spanning 8 to 12 wk. Instead, it should be thoughtfully determined based on the expected growth rate of the fish during the designated period. If the study involves juvenile specimens of a fast-growing species, a shorter duration may suffice, while trials involving sub-adult specimens or slower-growing species may require a longer timeframe. A minimum expected growth of 300% has been recommended (NRC, Citation2011) and is often used by editors and reviewers as the first, starting point when assessing a scientific study. With rapidly growing juvenile fish, it is not uncommon for fish to reach 5–10 times their starting weight in a short-duration (12-week) study. But with larger specimens, or in some slow growing species, a 12-week study might result in weight gains as little as 50% of their initial body weight.
A growth of 300% means doubling the initial weight, twice. For example, if the initial body weight of the fish is 100 g the final body weight to reach 300% gain will be 400 g (100 × 2 x 2). This minimum expected growth should serve as the minimum guiding criterion to assess validity of the intervention trial when the hypothesis includes effect of dietary treatments on fish weight gain or tissue composition. It is obvious that fish need to be exposed to experimental diets for a sufficient period of time to biologically display possible effects.
Using fish oil replacement studies as an example, and testing the hypothesis that “novel raw material X,” in this instance an imaginary oil extracted by a very sustainable and economical underutilized biomass, can replace fish oil without affecting the final omega-3 fatty acid composition of the fish fillet, we can design an experiment with two diets: a control with 100% fish oil, and a test with 100% novel raw material X. If we then run the experiment for a short period of time, and fish grow from 100 g to 120 g (20% gain), when we will analyze the omega-3 content of the filled of the fish under either treatment, we will likely find very little if any differences. We might then incorrectly conclude that the novel raw material X can safely replace fish oil without affecting the final fatty acid composition of edible portion of the fish. Yet, this is a perfect example of Type II error, as we are simply unable to see the real effects of novel raw material X, because the final fatty acid composition of fish fed with this oil was fundamentally the same fatty acids that were present in the initial fish, only marginally diluted in a slightly larger body. Put another way, there was insufficient time for dietary fatty acids from raw material X to be deposited in the tissues at high enough amounts to significantly alter the final fatty acid composition.
6. Conclusions: the successful feeding trial
After addressing all the aforementioned considerations and executing an experimental feeding trial, researchers often overlook a crucial, initial question: Was this feeding trial successful? This is not only about success in addressing the original research query or whether positive or negative outcomes were achieved, nor is it solely about generating novel scientific insights. Instead, it pertains to whether the trial’s results are reliable and can contribute to answering the research question and advancing scientific knowledge more broadly. In this concise opinion piece, we have aimed to provide clarity on how to address this specific question. We believe that there are five steps required to achieve a successful feeding trial:
1.a valid hypothesis must be conceived;
2.an appropriate experimental design must be developed to address the hypothesis;
3.the experimental diets must match the design;
4.the fish must be healthy, readily consume feed and growing well; and
5.a minimum of 300% weight gain must be achieved.
When these five steps are followed, the experimental in vivo feeding trial is successful. Researchers can celebrate!
Therefore, when reporting their findings in a scientific article, authors should start their discussion by addressing this very point by including a short summary relative to the successfulness (or possibly limitations) of the trial. This statement is completely independent from the results achieved. Further, this will be very important for the readers as it is at this very point that they will make up their mind about the trustworthiness and usefulness of the study. The following discussion section is where the authors should clearly state whether the results of the research study support the null hypothesis (no statistical differences) or the alternate hypothesis (statistical differences between control and experimental treatment groups), before entering into more complicated and detailed reasonings about the findings.
In conclusion, we hope readers will find this short opinion piece useful, and that these considerations presented here might help in further supporting the expansion of the effectiveness of research efforts in aquaculture nutrition. It is our belief that by improving the quality of in vivo feeding trials, research efforts will effectively support the continuous expansion of an environmentally sustainable, economically viable and socially accepted aquaculture industry.
Source : Turchini, G. M., & Hardy, R. W. (2024). Research in Aquaculture Nutrition: What Makes an Experimental Feeding Trial Successful? Reviews in Fisheries Science & Aquaculture, 1–9. https://doi.org/10.1080/23308249.2024.2413672
