You are on page 1of 3

comment

Progress and challenges in analyzing rodent


energy expenditure
Whole-body energy expenditure is the summed metabolic activities of tissues and, to remove the influence
of body size, ratios of energy expenditure to body mass are often applied but can generate spurious differences.
In 2011, a group of experts proposed adoption of ANCOVA for the analysis of metabolic rate but, seven years
later, analyses based on ratios remain the most frequent. We discuss some of the barriers to adopting better
analytical procedures.

Rodrigo Fernández-Verdejo, Eric Ravussin, John R. Speakman and Jose E. Galgani

I
n humans, average body mass index has a The extent to which these molecules are
increased over the last few decades1. This 2.0 1.5 kcal d–1 s.d. retained in a living organism is a function
between groups (kcal d–1)

is mostly explained by a shift towards the 1.0 kcal d–1 s.d. of the energy flux12. Such flux is dynamic,
Detectable difference

right in the distribution of individuals with 1.5 0.5 kcal d–1 s.d. with a constant exchange between the
larger body mass, while body mass of leaner macromolecules entering the body through
individuals has remained unaltered2. Such 1.0 feeding, and the molecules leaving the body
heterogeneous propensity to weight-gain through oxidation. A mismatch of this flux
prompts the search for factors that alter the 0.5 (in versus out) modifies body weight and
energy balance. composition, leading to obesity in the case
Technologies to measure energy 0 of chronic positive energy balance. Accurate,
0 6 16 32 48 64
expenditure in humans and rodents have precise and continuous determination of
made remarkable progress3,4. Additionally, Mice per group (n) energy expenditure and intake is thus crucial
genetic manipulation of rodent models b for understanding body weight control.
ANCOVA, ratios and no normalization (n = 3)
allows discernment of the influence In the best scenario, we should isolate the
ANCOVA and no normalization (n = 3)
of genetics on metabolic rate with an ANCOVA and ratios (n = 2)
primary event driving energy imbalance,
unprecedented sophistication. However, ANCOVA only (n = 6) that is, higher energy intake or lower energy
this has not been mirrored by improved expenditure, before changes in body mass
No normalization (n = 10)
methodologies to analyze metabolic occur. Living organisms are, however, almost
rates. Metabolic rate is often divided by Ratios and no
normalization (n = 8)
never in neutral energy balance, because
body mass to “normalize” data for size organisms acquire and dissipate energy at
differences between animals, which is variable rates. Therefore, distinguishing
mathematically wrong. normal hour-to-hour energy disparity from
Knowledge of this problem, and the Ratios only (n = 45) the imbalance that will lead to obesity is
approach to overcome it, dates back to challenging. Thus, energy balance is often
the 1950s5,6. However, compared with compared in animals of different body
calculating ratios, manual calculation of the Total = 77 articles mass and/or composition. Nevertheless,
analysis of covariance (ANCOVA) was time as body mass and composition influence
consuming. In the 1980s, computational Fig. 1 | Analytical aspects in rodent metabolic energy metabolism, these variables must
analysis eased calculations and increased studies. a, Detectable difference in energy be controlled to determine the influence of
interest in the appropriate analyses in expenditure according to the number of mice other variables (for example, genotype).
the field of human metabolism7 and in per group. The mean energy expenditure of
ecological studies of small mammals8. the control group was assumed as 10 kcal d–1, Comparison of energy expenditure
ANCOVA thus became the common with standard deviations (s.d) for control and in mice
approach to adjust for body size. In contrast, experimental groups of 0.5, 1.0 and 1.5 kcal d–1. Several studies have assessed the effect of
studies of small laboratory rodents persisted SAS 9.2 (SAS Institute) was used for calculations, factors, such as diets and genes, on energy
in using ratios. This erroneous practice has using two-sided Student’s t test with α = 0.05 expenditure of animals that differ in body
continued despite papers pointing out the and β = 0.20. b, Frequency of the method used to mass and/or composition. Such differences
potential for spurious results9,10. compare energy expenditure in mice among the may alter metabolic rate. Therefore, a highly
To reach a consensus on how mouse studies retrieved. debated issue is how to normalize energy
energy metabolism should be analyzed, a expenditure for differences in body size.
large group of experts in mouse and human how successful the application of ANCOVA An extensively used normalization method
metabolism discussed conceptual, analytical has been in mouse metabolism studies. is dividing metabolic rate by body mass.
and practical aspects involved in measuring However, this approach is inappropriate
energy metabolism in mice11. They proposed Energy balance and weight gain because the regression line between body
ANCOVA as a way to conduct, analyze and Body weight represents the combined mass mass and metabolic rate does not have a
interpret metabolic studies. Here, we assess of water, protein, fat and carbohydrate. zero intercept, and is therefore not linearly
Nature Methods | www.nature.com/naturemethods
comment

proportional13,14. Scaling body mass to a The concepts aforementioned also apply Unfortunately, data is often analyzed in a
power was intended to solve this problem. for energy intake. ANCOVA should thus way that negates all the previous investment
However, the power exponent commonly be used for comparison of energy intake in time, effort and money. Although the
utilized (that is, 0.75)15 derives from when body mass or composition differs search conducted by ourselves and by
between-species comparisons that may not between groups. Importantly, since energy Tschöp et al.11 use different search criteria
apply for within-species comparisons. To expenditure and intake are required to and are not directly comparable, our data
account for differences in body composition, understand weight gain, reporting them in suggest that while ANCOVA is used more
the metabolic-rate-to-lean-mass ratio is the same units is essential; using different often than before, its implementation still
calculated. This approach does not solve units blurs the significance of results. needs to be increased.
the problem, because the regression line
between lean mass and metabolic rate also How energy expenditure is analyzed Why does this issue still exist?
has no zero intercept. In 2011, Tschöp et al.11 found that only We notice potential barriers for adopting
Generalized linear modeling (GLM) 2% of studies used ANCOVA or GLM to appropriate methods. First, reviewers
and ANCOVA are appropriate approaches correct for body mass or composition, and are unfamiliar with the best practice.
for normalizing metabolic rate to body recommended more widespread use. Have Second, since most studies use the wrong
mass or composition. These approaches these recommendations been followed since method, new studies simply repeat this
allow combination of discrete (for example, then? We set an arbitrary search criterion wrong method. This is particularly true
genotype) and continuous (for example, to include a wide range of publications. if researchers are unfamiliar with the
body mass) traits as determinants of energy We searched in PubMed for original correct analysis, which may be perceived
expenditure. Lean and fat masses can also be articles published or available online from as complex and hard to interpret. But with
included to account for body composition. July 2012 until June 2018, using the key: the appropriate statistical packages (such as
Considering tissues and organs expend (“energy expenditure” OR “energy balance” STATA; https://www.stata.com), conducting
energy at different rates, ANCOVA could OR “heat production” OR “metabolic ANCOVA is as easy as calculating ratios.
also include the masses of individual rate” OR “oxygen consumption”) AND Mina et al.20 recently proposed a free web-
organs and tissues as covariates16,17. Thus, (“2012” [Date - Publication]: “3000” [Date based tool (CalR; https://CalRapp.org/) that
a predictive model of metabolic rate can - Publication]) AND (“mouse” OR “mice”). provides appropriate analyses of indirect
be generated from control mice based on We also searched papers that cited the guide calorimetry data in various experimental
the size of, among others, the liver, brain or by Tschöp et al11. The inclusion criteria designs. Third, using ANCOVA often
skeletal muscle. were: comparison of genetically altered reveals no effect of a given manipulation
Alternatively, metabolic rate can be mice versus control mice (mouse models (for example, genotype) making publication
compared before differences in body mass injected with anti-sense oligonucleotides or more difficult, particularly if it does
and composition are evident. In this case, adenoviruses were also included); mice fed not conform to a previous consensus.
metabolic rate can be simply reported ad libitum or pair-fed designs; mice awake Researchers could thus feel pressured to
as kcal d–1. However, the ability to detect and without movement restrictions at the continue with a particular method, even
‘statistically significant’ differences depends time of measurement; analysis at room if it is wrong. Fourth, some commercial
on the reliability of the instrument, the temperature or at thermoneutrality; mice equipment automatically expresses
statistical approach and the sample size. fed with chow, high-fat or western diet; and metabolic rate per kg of body mass,
With the sample size often included in journal impact factor ≥ 15.0 according to suggesting this is an appropriate analysis.
studies (∼6 per group), differences in the Journal Citation Reports 2016. Finally, some authors cite the guide for
body mass or composition may remain We identified 77 articles, of which analysis by Tschöp et al.11, but still use ratios.
undetectable, which may influence the 14 (18%) cited Tschöp et al. Eighteen percent This incongruence may lead some to assume
results by including or not including a of articles used ANCOVA (n = 14), of which that using ratios was recommended by
normalization of the data. True differences almost two-thirds included body mass as the Tschöp et al.
in energy expenditure may also be covariate (Fig. 1b and Supplementary Table).
undetectable, and the detectable differences Only five articles conducted ANCOVA Concluding remarks
in energy expenditure may be a function with lean mass as covariate (Supplementary Mouse models provide an invaluable
of the sample size (Fig. 1a). With six mice Table). As reported11, normalization using opportunity to compare metabolic
per group and a standard deviation of ratios with body mass or lean mass were the phenotypes to better understand weight-
0.5 kcal d–1, differences in energy most frequent methods (> 50%) (Fig. 1b). related diseases. These models are
expenditure of less than 10% cannot be Some studies analyzed data using more than accompanied by major progress in the
detected. On the one hand, such lack of one approach. For instance, three studies measurement of energy metabolism using
power may be considered an advantage as analyzed data without normalization, indirect calorimeter cages. Unfortunately,
it reduces the risk of making a type I error with ANCOVA and with ratios (Fig. 1b). such advances have not been associated with
and only genes with large effect size will be This multiple-analysis strategy appears appropriate data analysis, potentially leading
identified. On the other hand, genes with unnecessary and may lead to confusion if to biased interpretations.
small, but perhaps significant, long-term the approaches generate discordant results. Editorial boards and peer-reviewers
influence may be wrongly discarded Presenting individual data for the regression have a major responsibility in demanding
(an increased risk of a type II error). between energy expenditure and body appropriate analyses. Including specific
Notably, ANCOVA may be unnecessary mass (or lean mass), and comparing groups guidelines for metabolic data in the
and could generate spurious results when with ANCOVA, appears to be the most ‘Information for authors’ may promote
mice display no overlap of body weight appropriate analysis. better analyses. A straightforward strategy
between groups18,19. See ref. 11 for further Thousands of dollars are spent on genetic is to require energy expenditure plotted
details on the appropriate analysis of mouse models, and energy metabolism against body mass (or lean mass). This will
metabolic rates. is measured with expensive equipment. force consideration of mass effects, which

Nature Methods | www.nature.com/naturemethods


comment

are obscured in histograms. A check-list Medicina, Pontificia Universidad Católica de Chile, for proper analysis when comparing individuals of different
body size.
verifying that energy expenditure was Santiago, Chile. 15. Kleiber, M. The Fire of Life: An introduction to animal energetics
analyzed with ANCOVA and plotted *e-mail: jgalgani@uc.cl (Wiley, 1961).
correctly should be mandatory. Analysis 16. Mitchell, S. E. et al. Oncotarget 8, 17453–17474 (2017).
17. Javed, F. et al. Am. J. Clin. Nutr. 91, 907–12 (2010).
of energy intake, along with the detailed Published: xx xx xxxx
18. Speakman, J. R. Front. Physiol. 4, 34 (2013).
reporting of the composition of diets, https://doi.org/10.1038/s41592-019-0513-9 19. Tracy, C. & Sugar, J. Physiol. Zool. 62, 993–997 (1989).
also deserves consideration. These efforts 20. Mina, A. I. et al. Cell Metab. 28, 656–666 (2018).

should increase the consistency, rigor and References


Acknowledgements
reproducibility required for comprehensive 1. NCD-RisC. Lancet 387, 1377–1396 (2016).
R.F.V. was supported by Fondecyt #11180361 (Conicyt,
2. Wardle, J. & Boniface, D. Int. J. Obes. 32, 527–32 (2008).
understanding of the role that biological and 3. Lam, Y. Y. & Ravussin, E. Mol. Metab. 5, 1057–1071 (2016). Chile). J.E.G was supported by Fondecyt #1170117
environmental traits have on obesity. ❐ 4. Lighton, J. R. B. Eur. J. Clin. Nutr. 71, 301–305 (2017). (Conicyt, Chile).
5. Tanner, J. M. J. Appl. Physiol. 2, 1–15 (1949).
First paper calling for attention regarding the fallacy of Author contributions
Rodrigo Fernández-Verdejo1, Eric Ravussin2, expressing physiological variables per unit of body size.
J.E.G., E.R. and J.R.S. conceived the work. R.F.V. and J.E.G.
John R. Speakman3,4 and Jose E. Galgani1,5* 6. Cochran, W. G. Biometrics 13, 261–281 (1957).
designed, acquired and analyzed the data, and drafted the
1
Departamento de Ciencias de la Salud, Carrera 7. Ravussin, E., Lillioja, S., Anderson, T. E., Christin, L. &
manuscript. All authors interpreted the data, substantively
Bogardus, C. J. Clin. Invest. 78, 1568–1578 (1986).
de Nutrición y Dietética, Facultad de Medicina, 8. Packard, G. C. & Boardman, T. J. Physiol. Zool. 61, 1–9 (1988). reviewed the manuscript, approved the submitted
Pontificia Universidad Católica de Chile, Santiago, 9. Arch, J. R. S., Hislop, D., Wang, S. J. Y. & Speakman, J. R. version and agreed to be personally accountable for their
Chile. 2Pennington Biomedical Research Center, Int. J. Obes. 30, 1322–1331 (2006). contributions.
10. Butler, A. A. & Kozak, L. P. Diabetes 59, 323–9 (2010).
Baton Rouge, LA, USA. 3Institute of Biological
11. Tschöp, M. H. et al. Nat. Methods 9, 57–63 (2011). Competing interests
and Environmental Sciences, Zoology Building, 12. Hall, K. D. et al. Lancet 378, 826–37 (2011). The authors declare no competing interests.
University of Aberdeen, Aberdeen, UK. 4Institute 13. Heymsfield, S. B. et al. Am. J. Physiol. Endocrinol. Metab. 282,
132–138 (2002).
of Genetics and Developmental Biology, Chinese Additional information
14. Ravussin, E. & Bogardus, C. Am. J. Clin. Nutr. 49, 968–975 (1989).
Academy of Sciences, Beijing, China. 5Departamento Remarkable analysis of the main determinants of energy Supplementary information is available for this paper at
de Nutrición, Diabetes y Metabolismo, Facultad de expenditure in humans. This paper also elaborates the need https://doi.org/10.1038/s41592-019-0513-9.

Nature Methods | www.nature.com/naturemethods

You might also like