Metabolomics is the youngest member of the ‘omics’ family, joining genomics, proteomics and transcriptomics as tools in global systems biology( Reference Rochfort 1 ). Metabolomics studies the small molecular weight molecules or metabolites that are present in biological samples with an aim to identify perturbations in metabolism under different conditions( Reference Brennan 2 ). It complements other ‘omic’ technologies such as trancriptomics and proteomics and is considered to best reflect activities at a functional level( Reference Putri, Nakayama and Matsuda 3 ). The metabolome responds to nutrients, stress or disease long before the transcriptome or proteome making it an attractive approach for multiple fields, with metabolite alterations now implicated in the development of a number of human diseases( Reference Wang, Larson and Vasan 4 – Reference Oresic, Seppanen-Laakso and Sun 6 ).
The metabolomics pipeline is composed of a number of steps. In general, these steps involve: (i) experimental design; (ii) sample preparation; (iii) data acquisition; (iv) data processing; (v) statistical analysis( Reference Brennan 2 , Reference Dunn, Bailey and Johnson 7 ), an overview of which is illustrated in Fig. 1. All stages should be carefully designed and executed in order to provide valid datasets and ultimately valid experimental conclusions and hypotheses( Reference Dunn, Bailey and Johnson 7 ). Numerous comprehensive reviews on the experimental strategies in metabolomics are available elsewhere( Reference Goodacre, Vaidyanathan and Dunn 8 – Reference Fiehn 11 ). The focus of this review will provide an overview of current technologies and applications of metabolomics in the determination of new dietary biomarkers. In addition, it will also address challenges in the field and future outlooks.
Fig. 1. (Colour online) An overview of the metabolomics pipeline. LC/MS, liquid chromatography–MS.
Metabolomic technologies
A number of analytical technologies( Reference Dunn and Ellis 12 ) are employed in metabolomics with the ultimate goal of analysing a large fraction or all of the metabolites present. Due to metabolite diversity and the range of concentrations in which they are present (pm–mm), a range of these technologies is often used as at present no single technique has the capability to measure the whole metabolome( Reference Goodacre, Vaidyanathan and Dunn 8 , Reference Sumner, Mendes and Dixon 10 , Reference Fiehn 11 ). An overview of metabolites captured using different technologies is depicted in Table 1. In general, metabolomic analyses have been classified into targeted or non-targeted approaches and the type of approach used for a particular study will be dependent on the research question/design. The two major metabolomic platforms used in these approaches are NMR spectroscopy and MS. Non-targeted metabolomics involves measuring as many metabolites as possible in a biological sample simultaneously, therefore providing a broader coverage of metabolites( Reference Alonso, Marsal and Julia 13 ). This approach offers the opportunity for novel target discovery; however, the challenges lie in the time required to process the extensive amounts of raw data produced, difficulties in relation to characterising unknowns and bias towards high-abundance molecules( Reference Roberts, Souza and Gerszten 14 , Reference Vinayavekhin and Saghatelian 15 ). In contrast, targeted approaches are taken when specific classes of metabolites are to be measured. Through the use of internal standards, analysis can be carried out in a quantitative manner( Reference Roberts, Souza and Gerszten 14 ). The major limitation of this approach is that it requires the compounds of interest to be known a priori, which need to be available commercially in purified form as standards in order to be quantified( Reference Shulaev 16 ).
Table 1. Overview of the metabolite coverage achievable by the different metabolomic technologies
LC–MS, liquid chromatography–MS.
NMR spectroscopy
NMR spectroscopy has been utilised extensively in the field of metabolomics research and has played a key role in our understanding of metabolism for many decades( Reference Malet-Martino and Holzgrabe 17 , Reference Sebedio and Brennan 18 ). It benefits from being perhaps the most selective analytical technique, with its ability to provide unambiguous information about a molecule, an important aspect in terms of characterising components of complex mixtures( Reference Dunn, Bailey and Johnson 7 , Reference Malet-Martino and Holzgrabe 17 ). NMR does not require extensive sample preparation time, has high reproducibility and little inter-laboratory variability( Reference Sebedio and Brennan 18 ). In addition, the analysis is non-destructive and does not require pre-selection of the analysis conditions, such as ion source conditions for MS or chromatographic operating conditions( Reference Dunn, Bailey and Johnson 7 ). However, a major drawback of NMR in comparison with MS-based methods is its limited sensitivity( Reference Ramautar, Demirci and de Jong 19 ). Another barrier delaying more prevalent use of NMR as a metabolomic tool is the need for manual spectral profiling( Reference Ravanbakhsh, Liu and Bjorndahl 20 ).
The majority of metabolomic applications employ (one-dimensional) 1H (proton) NMR as the majority of known metabolites contain hydrogen atoms. In comparison with other methods, NMR is non-biased to particular metabolites, i.e. all metabolites will be detected once they are present in concentrations above the limit of detection. Two-dimensional (2D) NMR experiments such as TOCSY, 1H J-RES and 1H–13C HSQC are important to use in conjunction with one-dimensional spectra for metabolite identification( Reference Brennan 21 ) and to compare with reference databases such as the Human Metabolome Database( Reference Wishart, Knox and Guo 22 ) and Biological Magnetic Resonance Databank( Reference Kobayashi, Harano and Tochio 23 ). In addition, one-dimensional and 2D NMR experiments offer the potential to identify previously unknown metabolites( Reference Pudakalakatti, Dubey and Jaipuria 24 ).
NMR has the capability of not only measuring solution states (e.g. urine, plasma and serum), but can also measure tissue samples directly through a technique called magic angle spinning( Reference Beckonert, Coen and Keun 25 ). Its application has been particularly utilised in the area of tumour monitoring( Reference Beger 26 ). For example, it has been used to robustly determine the differences between benign and malignant tissue from patients with breast and colon cancer with a high degree of sensitivity and specificity( Reference Chan, Koh and Mal 27 ). This technique can also be used for real-time monitoring as it is conceivable that magic angle spinning–NMR spectroscopy can be performed within a 10–20 min time frame, highlighting the translational potential as a clinical resource for rapid diagnostics( Reference Nicholson, Holmes and Kinross 5 ).
As mentioned previously there are some disadvantages associated with NMR, the major one being its limited sensitivity. To address this issue a number of recent developments have been seen. For example, the use of labelled compounds can lead to significant improvements in sensitivity (e.g. using 13C or 15N NMR)( Reference Putri, Yamamoto and Tsugawa 28 ) and instrumental advances such as the use of cryogenically cooled probes and microcoil probes have also enhanced sensitivity, allowing the detection of metabolites at low abundance( Reference Brennan 21 , Reference Keun, Beckonert and Griffin 29 ).
In recent years, a number of software packages have been developed that support semi-automatic NMR spectral profiling of one-dimensional and 2D 1H NMR spectra( Reference Mercier, Lewis and Chang 30 , Reference Hao, Astle and De Iorio 31 ). Of note is an open-source software called BAYESIL (http://www.bayesil.ca), which provides a fully automated system, allowing analysis of complex mixtures quickly and accurately. The development of such tools provides quantitative and accurate NMR profiling effectively, without the need for trained experts, enabling a wealth of new applications of NMR in clinical settings( Reference Ravanbakhsh, Liu and Bjorndahl 20 ).
MS-based technologies
MS-based techniques are coupled with a chromatographic step, most commonly GC or liquid chromatography. This chromatographic step allows the separation of metabolites prior to detection, which reduces the complexity of the mass spectra and enhances resolution, sensitivity and selectivity( Reference Wang, Liu and Hu 32 , Reference Dettmer, Aronov and Hammock 33 ).
GC–MS has long been used in metabolomics for the comprehensive analysis of metabolites due to its high selectivity and sensitivity. It is most suitable for volatile metabolites and those that can become volatile following chemical derivatisation. Although, chemical derivatisation can improve significantly the GC separation of compounds, it is also a disadvantage as it can introduce unwanted artefacts( Reference Shulaev 16 , Reference Halket, Waterman and Przyborowska 34 ). The development of comprehensive 2D GC further enhances separation performance by coupling two columns coated with different stationary phases, which greatly increases separation capacity( Reference Wang, Liu and Hu 32 ), chemical selectivity and sensitivity, thus providing more accurate information about metabolite retention times and mass spectra( Reference Castillo, Mattila and Miettinen 35 ). A recent study by Rocha et al.( Reference Rocha, Caldeira and Carrola 36 ) used a 2D-GC–MS untargeted approach to analyse human urine. They identified 700 compounds from a diverse number of chemical families (e.g. ketones, alcohols, aldehydes, thiols, amines, etc.), providing the most complete information available on the volatile components of human urine.
Liquid chromatography–MS requires no need for chemical derivatisation of metabolites, which is an advantage over GC–MS. Analysis of a wide range of metabolites ranging from low to high molecular weight and from hydrophilic to hydrophobic can be carried out through selection of the appropriate column and mobile phases. HPLC separations are best suited for the analysis of labile and non-volatile polar and non-polar compounds in their native form( Reference Zhang, Sun and Wang 37 ). Column technology has also improved greatly in terms of metabolite coverage and for reducing analysis time. The introduction of ultra-high pressure liquid chromatography, using porous particles with internal diameter smaller than 2 µm, in conjunction with MS( Reference Zhang, Sun and Wang 37 ) has substantially increased chromatographic resolution and peak capacity compared with conventional HPLC columns( Reference Lei, Huhman and Sumner 38 ). Overall, the ultra-high pressure liquid chromatography–MS technique shows great promise as a hyphenated micro-separation tool in metabolomics as the majority of primary metabolites are intrinsically polar( Reference Zhang, Sun and Wang 37 ).
During the last decade liquid chromatography–MS techniques have developed, which employ soft ionisation approaches such as electrospray ionisation and atmospheric pressure chemical ionisation, making MS more sophisticated and more robust for daily use( Reference Shulaev 16 ).
A number of different types of MS detectors exist, which vary in terms of cost, selectivity, sensitivity and accuracy. These include ion traps, single quadrupoles, triple quadrupoles, time-of-flight, quadrupole time-of-flight, orbitraps and Fourier transform ion cyclotron resonance. Triple quadrupoles are particularly good for quantification of biomarkers because of their high selectivity and robustness, whereas quadrupole time-of-flights are particularly suited for untargeted metabolomic analysis because of their higher mass accuracy (detect the difference between m/z 300·00 and 300·003) and scan rate. They are also increasingly used for quantification, much like the triple quadrupole instruments( Reference Lind, Savolainen and Ross 39 ).
MS-based metabolomics offers quantitative analyses with the ability to quantify very low concentrations of potential biomarkers( Reference Dettmer, Aronov and Hammock 33 ). Metabolites can be fully quantified through the use of internal standards, ideally using stable isotope labelled standards and spiking them into the sample or semi-quantitatively which is often the case when a range of metabolites are measured. Semi-quantification normalises the metabolite signal intensity to that of an internal standard or another relative metabolite, whereas full quantification or absolute quantification determines the absolute metabolite quantity. Stable isotopes are critical for absolute quantification; however, the availability of commercial isotope labelled standards is limited and costs can be prohibitive to large-scale use( Reference Lei, Huhman and Sumner 38 ).
A drawback of these conventional separation methods is the low sample throughput, so a number of complementary approaches are emerging in order to reduce analysis time( Reference Gonzalez-Dominguez, Garcia-Barrera and Gomez-Ariza 40 ). One such approach is direct infusion MS, which allows the direct introduction of a sample into the spectrophotometer, bypassing the conventional chromatographic separation step. Advantages include rapid analysis, reproducibility and consequently high-throughput screening ability( Reference Draper, Lloyd and Goodacre 41 ); however, it does suffer from metabolite interferences, particularly in complex matrices such as serum( Reference Gonzalez-Dominguez, Garcia-Barrera and Gomez-Ariza 40 ).
Overall, one of the main advantages of these MS techniques is the associated high sensitivity, allowing the detection of metabolites that are below the detection limit of 1H NMR spectroscopy. Exciting developments in automation and quantitation for NMR and MS-based metabolomics have been recently described( Reference Wishart 42 , Reference Gowda and Djukovic 43 ). Shifts towards commercial kits, automation and better standardisation will ultimately reduce costs, increase throughput, allow greater reproducibility and substantially reduce sample-handling errors( Reference Wishart 42 ).
Metabolomics and dietary biomarkers
The diet is an important environmental exposure and therefore its measurement is a vital part of health-related research. Measuring habitual dietary intake should be both accurate and applicable to large numbers of free-living individuals, hence measuring dietary exposure is one of the greatest challenges in nutritional research( Reference Fave, Beckmann and Draper 44 ). Traditional tools for collecting information on dietary intake including FFQ, food diaries and 24-h recalls are often unreliable and are subject to possible underreporting and recall errors( Reference Bingham 45 , Reference Kipnis, Midthune and Freedman 46 ). Due to these well-documented problems, there has been a growing appreciation to improve methods to assess dietary intake( Reference Jenab, Slimani and Bictash 47 , Reference Penn, Boeing and Boushey 48 ). The use of dietary biomarkers provides a more objective and accurate measure of intake and if used in combination with traditional methods will improve the ability to assess dietary intake. Currently, ideal biomarkers exist for salt and protein intake (sodium/nitrogen measure in a 24-h urine) and energy expenditure (double-labelled water technique)( Reference Bingham 45 ). The further development of robust dietary biomarkers will improve the assessment of the relationship between diet and chronic disease( Reference Heinzmann, Brown and Chan 49 ). In recent years, metabolomics has emerged as a key tool in dietary biomarkers discovery.
Applications of metabolomics to identify novel biomarkers of dietary intake have in general taken three approaches: (1) specific acute intervention studies to identify specific biomarkers of intake; (2) searching for biomarkers in cohort studies by correlating to self-reported intake of a specific food/food group(s); (3) analysis of dietary patterns in conjunction with metabolomic profiles to identify nutritypes and biomarkers. These biomarker discovery studies have in general applied untargeted metabolomic approaches( Reference Beckmann, Lloyd and Haldar 50 ).
Use of specific acute intervention studies to identify specific biomarkers of intake
Dietary intervention studies generally involve participants consuming specific food(s) followed by the collection of biofluids either postprandially or following a short-term intervention( Reference Brennan 2 ). The most popular choice of biofluid for these types of studies is urine, due to it being non-invasive; however, plasma and serum are also possible and useful. This approach has resulted in the identification of a number of putative biomarkers of specific foods and drinks such as citrus fruit( Reference Heinzmann, Brown and Chan 49 , Reference Andersen, Kristensen and Manach 51 , Reference Atkinson, Downer and Lever 52 ), cruciferous vegetables( Reference Edmands, Beckonert and Stella 53 , Reference Andersen, Reinbach and Rinnan 54 ), red meat( Reference Stella, Beckwith-Hall and Cloarec 55 , Reference Cross, Major and Sinha 56 ), coffee( Reference Heinzmann, Holmes and Kochhar 57 , Reference Lang, Wahl and Stark 58 ), tea( Reference van Velzen, Westerhuis and van Duynhoven 59 , Reference Daykin, Van Duynhoven and Groenewegen 60 ), sugar-sweetened beverages( Reference Gibbons, McNulty and Nugent 61 ) and wine( Reference Jacobs, Fuhrmann and van Dorsten 62 ). A good example of a robust biomarker of citrus fruit is proline betaine, which was identified initially by Atkinson et al.( Reference Atkinson, Downer and Lever 52 ) and subsequently validated by independent research groups( Reference Heinzmann, Brown and Chan 49 , Reference Andersen, Kristensen and Manach 51 , Reference Pujos-Guillot, Hubert and Martin 63 ). The biofluid used in all of these studies was urine; however, different metabolomic analytical strategies were applied to measure proline betaine, which included NMR and MS. One such study was performed by Heinzmann et al.( Reference Heinzmann, Brown and Chan 49 ), which involved eight volunteers consuming standardised meals over a 3 d period. On the second day of the study, a mixed-fruit meal (apple, orange, grapefruit and grapes) was introduced. Urine samples were collected and subsequent analysis via NMR spectroscopy and partial least-squares-discriminant analysis identified proline betaine as a potential biomarker of citrus fruit intake. To confirm the findings, proline betaine was measured in citrus fruits and the urinary excretion kinetics were evaluated. Furthermore, the biomarker was further validated in a large cohort study (INTERMAP, UK), with excellent sensitivity and specificity (90·6 and 86·3 %, respectively) for discriminating between consumers and non-consumers of citrus fruit( Reference Heinzmann, Brown and Chan 49 ).
Although a number of biomarkers of specific foods have been reported, it is worth noting that few have been validated in large separate cohorts, making it difficult to translate these biomarkers into practice. It should also be noted that these acute biomarkers of intake are often short-term biomarkers that are rapidly excreted in urine, almost completely over a period of 24 h( Reference Gibbons and Brennan 64 ). Therefore, searching for longer-term biomarkers of habitual intake is required.
Searching for biomarkers in cohort studies by correlating to self-reported intake of a specific food/food group(s)
Searching for biomarkers of specific food(s) can also be carried out through the use of cohort studies. In this approach, dietary data are collected using a traditional method (e.g. FFQ, food diaries) to identify low and high consumers or consumers and non-consumers of a specific food. Metabolomic profiles are then compared between these groups in order to identify potential biomarkers. These cohort studies tend to be larger in terms of study participants in comparison with acute intervention studies, but rely on self-reported dietary assessment methods which are prone to error. It also needs to be highlighted that biomarkers identified in cohort studies do not assess the direct relationships of food amounts consumed, they are simply correlations between the food and the metabolite(s) and therefore the relationship is only an association( Reference Brennan, Gibbons and O'Gorman 65 ). Confirmation of these associations would be required in an intervention study in order to validate the metabolite as a specific biomarker of intake.
To date, a number of biomarkers of food intake have been identified using cohort studies for e.g. fish( Reference Lloyd, Beckmann and Haldar 66 ), red meat( Reference Wittenbecher, Muhlenbruch and Kroger 67 , Reference Myint, Fraser and Lindsted 68 ), whole-grain bread( Reference Garcia-Aloy, Llorach and Urpi-Sarda 69 ) and walnuts( Reference Garcia-Aloy, Llorach and Urpi-Sarda 70 ). An important recent study applied metabolomics to serum samples to identify biomarkers of red meat intake and identified a relationship between a number of those identified biomarkers (ferritin, glycine, diacyl phosphatidylcholines 36 : 4 and 38 : 4, lysophosphatidylcholine 17 : 0 and hydroxy-sphingomyelin 14 : 1) with risk of type-2 diabetes. The authors found that high levels of ferritin, low glycine and altered hepatic derived lipids in the circulation were associated with both total red meat consumption and diabetes risk. The findings are consistent with the hypothesis that metabolic processes reflected in the circulating concentrations of these biomarkers take part in linking red meat consumption to type-2 diabetes risk( Reference Wittenbecher, Muhlenbruch and Kroger 67 ).
This was the first reported study to evaluate a large set of metabolites as potential mediators linking exposure and disease, which is an important next step in biomarker discovery. Identifying these links between diet and disease will provide an insight into which metabolites and metabolic pathways are potential disease mediators, which could then be targeted or modulated through dietary interventions to improve health outcomes. In order to achieve this much work is required, particularly in the area of biomarker validation.
Use of dietary patterns to identify biomarkers of intake
The concept of identifying biomarkers using dietary patterns in combination with metabolomic patterns was pioneered in our research group( Reference O'Sullivan, Gibney and Brennan 71 ). Since then, this approach has been used in a number of studies( Reference Pere-Trepat, Ross and Martin 72 – Reference Bouchard-Mercier, Rudkowska and Lemieux 76 ), where patterns of intake are related to metabolomic patterns( Reference Brennan 2 ). This approach generally involves applying a multivariate statistical strategy such as principal component (PC) analysis or k-means cluster analysis to dietary data to identify dietary patterns and then through the use of regression (or other statistical method) linking these to metabolomic profiles in order to identify dietary biomarkers and/or nutritypes.
A recent study used this novel approach to distinguish between two dietary patterns in an attempt to develop a compliance tool( Reference Andersen, Rinnan and Manach 77 ). An untargeted approach was applied using ultra-high pressure liquid chromatography/quadrupole time-of-flight/MS to analyse urine samples from 181 participants as part of a parallel intervention study, who were randomly assigned to follow either a New Nordic Diet or an Average Danish Diet for 6 months. Partial least-squares-discriminant analysis was applied to the urinary metabolomic data to develop a compliance model for the New Nordic Diet and Average Danish Diet based on the metabolites that were identified as being the most discriminatory between the two diets. This resulted in a model with a misclassification error rate of 19 %, showing good promise as a compliance measure for different dietary patterns, which could be used to identify non-compliant subjects or groups of individuals with certain dietary responses( Reference Andersen, Rinnan and Manach 77 ). This study demonstrates that metabolomics can be used to discover which metabolites are the strongest predictors of compliance to complex diets; however, these metabolites should be followed up by quantitative measurements to further enhance and validate the model.
The identification of dietary patterns may also be important for studying relationships between diet and disease. For example, Bouchard-Mercier et al.( Reference Bouchard-Mercier, Rudkowska and Lemieux 76 ) investigated the metabolic signatures associated with the Western and Prudent dietary patterns using a targeted approach to profile participants (n 37) plasma. Applying PC analysis to the metabolic plasma profiles resulted in the identification of two PC. The first, PC1 was mainly composed of medium- to long-chain acylcarnitines, whereas PC2 was dominated by short-chain acylcarnitines and amino acids, including the branched-chain amino acids. The authors found that PC1 was not correlated to any food groups; however, PC2 was negatively correlated with fruit intake and positively associated with desserts (r = −0·38, P = 0·03; r = 0·37, P = 0·04, respectively). In addition, PC2 also had a significant positive correlation with saturated fat intakes (r = 0·39, P = 0·02). The Western dietary pattern had an inverse relationship with PC1 and a positive relationship with PC2, suggesting that people eating a Western diet are potentially at risk of increasing their long-term risk of cardiometabolic diseases( Reference Bouchard-Mercier, Rudkowska and Lemieux 76 ).
In summary, these studies indicate the potential of metabolomics as a tool for not only evaluating compliance to a dietary pattern, but also to identify and evaluate relationships between diet and disease.
Challenges and future outlooks
The application of metabolomics for dietary biomarker identification has grown significantly over the past 5 years( Reference Scalbert, Brennan and Manach 78 ) and in general this approach has produced a number of robust biomarkers of dietary intake. However, a number of challenges exist that need to be overcome in order to advance this field of research. One of the main challenges is in the area of biomarker validation, which is often omitted in study design. Following biomarker identification whether it is via an acute intervention study or cohort study, subsequent validation in an independent cohort(s) is critical. The independent cohort should not only be large, but also diverse (e.g. multicultural populations) and a number of considerations need to included such as age, sex, ethnicity, as well as sample processing, chromatographic separation and analytical instrumental settings. Another important aspect is the validation of the biomarkers performance across different laboratories. It can be a challenge to obtain identical metabolomic profiles as often laboratories use different protocols and therefore inter-laboratory validation is important( Reference Gowda and Djukovic 43 ).
Ideally, a well-validated biomarker should demonstrate a dose–response, which would confirm its suitability for use over a range of intakes( Reference Scalbert, Brennan and Manach 78 ); however, in the majority of studies this important step is often missing. Indeed biomarkers identified solely from cohort studies fail to assess the direct relationships of foods consumed and do not demonstrate responsiveness to intake, identifying merely an association between the biomarker and food( Reference Brennan, Gibbons and O'Gorman 65 ). A recent review by Gibbons highlighted the need to combine such studies with acute intervention studies to examine direct relationships and dose–response relationships( Reference Gibbons and Brennan 64 ).
Another challenge is in the area of metabolite identification, which is one of the main limiting factors, particularly when using MS-based metabolomics. The reason for this is the huge chemical diversity that is present in biological samples( Reference Sebedio, Martin and Pujos 79 ). To progress this field the metabolomics community are actively developing databases; examples include the Human Metabolome Database, Metlin and MassBank, which are publicly available. The Human Metabolome Database contains over 41 000 metabolites and is the most comprehensive collection of human metabolism data in the world( Reference Wishart, Tzur and Knox 80 ). A similar database containing food constituents and food additive metabolites has also been created and is updated regularly, which is called the FoodDB. The database contains over 28 000 metabolites, a valuable tool for food and nutrition researchers( Reference Scalbert, Brennan and Manach 78 ), although caution needs to be exerted as identifying metabolites originating from foods remains difficult.
Conclusion
Metabolomics has proven a powerful tool in the area of dietary biomarker discovery and its application has the potential to greatly enhance our ability to assess dietary intake. However, it is pertinent that validated biomarkers of intake are translated into practice. In order for them to meet their full potential cooperation across disciplines is necessary.
Acknowledgements
The authors kindly acknowledge our funding bodies.
Financial Support
The authors received the following funding: ERC (grant no. 647783) and SFI (grant no. JPI-HDHL/B3075, Foodball).
Conflict of Interest
None.
Authorship
A. O. G. conducted the literature search and drafted the manuscript. L. B. critically reviewed the manuscript. Both authors read and approved the final manuscript prior to admission.
