Minimal Inhibitory Concentrations of Thymol and Carvacrol: Toward a Unified Statistical Approach to Find Common Trends

Thymol and carvacrol are some of the most important and used components of Essential oils (EOs); they are widely studied, and there are much data available in the literature. Their Minimal Inhibitory Concentration (MIC) values found in the literature from 2005 to present were used to assess the bioactivity toward yeasts, molds, Gram-positive bacteria, and Gram-negative bacteria, as well as on some bacterial species/serotypes (Salmonella sp., Escherichia coli, E. coli O157:H7, lactic acid bacteria, Listeria monocytogenes, Staphylococcus aureus, S. epidermidis, etc.) to find possible common trends or differences between the two compounds and among the tested species. The results were quite interesting and pointed out that there is a common range for the MIC of thymol and carvacrol for some bacterial species (150–400 mg/L), with some exceptions to this generalized statement. In addition, the statistics pointed out that bacteria could experience homogeneous (S. epidermidis, E. coli O157:H7) or heterogeneous trends (for example, Salmonella sp.) depending on the existence of possible sub-species or different experimental set-ups. Moreover, this paper suggests that there are some drawbacks and issues that should be solved for the effective use of EOs, which are the strong variability among the microorganisms and the lack of standard protocols and reference strains.


Introduction
Essential oils (EOs) are aromatic and liquid extracts from plant materials, such as flowers, roots, barks, seeds, peels, and fruits [1], that are acquired through steam distillation [2], enfleurage, or expression [3]. They are mixtures of several compounds; some of them are labeled as major if they are present at higher concentrations, ranging from 20% to 70%, while others are considered minor because they are present as traces [4,5]. There are at least 3000 aromatic plants widely distributed in Europe, China, India, Asia, Africa, and South America [6], mainly belonging to Apiaceae, Asteraceae, Lamiaceae, Lauraceae, Myrtaceae, Poaceae, Rutaceae, and Zingiberaceae [7].
The major components of EOs belong to three different classes: terpenes (limonene, α-pinene, p-cymene, etc.), terpenoids (carvacrol, thymol, borneol, and camphor, among others), and phenylpropenes (eugenol, vanillin, and cinnamaldehyde) [7]. It has been reported that the presence of a phenolic ring is a key element for the antimicrobial effect of active components of EOs, as well as methoxy and olefin bonds, as they play a major role in the lipophilicity of the molecules and in their ability to consume proton motive force, affect the intracellular pH value, and disrupt the oxidative phosphorylation of bacteria [8].
The antibacterial effect of EOs has been studied by many authors in the past, and several mechanisms have been demonstrated or postulated, including significant changes in fatty acid profiles, injuries to the cytoplasmatic membrane, reductions in ATP production, among the tested species, as preliminary steps for the development and validation of an internationally recognized protocol and a set of guidelines and reference breakpoints to study their antimicrobial effect.

Data
The data used in this manuscript are MIC values of thymol [5-methyl-2-(methylethyl) phenol] and carvacrol [2-methyl-5-(methylethyl)phenol], available in the literature in research papers from international databases, Scopus and PubMed. The keywords for search were "thymol", "carvacrol", "antimicrobial effect", "Minimal Inhibitory Concentration", and other similar words. After this first step, ca. 4000 articles were found; to select the most relevant ones, only research papers that fit the following inclusion criteria were used: (a) Papers from 2005 to present. (b) Use of collection isolates or isolates that had been previously identified through genotyping; papers without a clear identification of the strains or with strains identified only through phenotyping were excluded. (c) Assays performed in the optimal broth for each microorganism, not containing restrictive ingredients or conditions (antibiotics, salt, or sub-optimal pH), which could affect the response of the test strains to carvacrol and thymol. (d) Use of thymol and carvacrol with a degree of purity of at least 96%. (e) Only research papers with a clear distinction between MIC and MBC (minimum bactericidal concentration) were used. The articles where the threshold values of thymol and carvacrol could be either MIC or MBC were excluded. (f) Use of both positive (for example, chloramphenicol, cycloheximide, or other antibiotics able to exert inhibition against the tested microorganism) and negative controls (distilled water and the solvent used to prepare thymol and carvacrol solution). Research papers not containing negative controls were excluded. (g) Experiments conducted using the micro-dilution approach; it was not possible to include papers that used CLSI protocols, as they are generally used for antibiotics, and only in recent years has a proposal for a protocol for EO S been released. (h) For the analyses, some papers with an agar diffusion approach were retained if the experiments had been conducted with at least 10 different concentrations of thymol and carvacrol and positive and negative controls were obtained, as reported for point "f". (i) Only species (bacteria or fungi) for which there were at least three independent evaluations of MICs were retained for this study.
The dataset produced and used for statistics is in supplementary Table S1.

Standardization and Data Analysis
The data reported in Table S1 were preliminary converted into mg/L and then into natural logarithm. A first classification was conducted for macro-categories, as MIC values were divided into four classes (Gram-positive bacteria, Gram-negative bacteria, yeasts, and molds), analyzed using the non-parametric tests of Mann-Whitney and Kruskal-Wallis (critical P set at 0.05) and the outlier test, and graphically reported through a box-whisker plot, where the box represents the interquartile range and the whisker represents the range from minimum to maximum.
A second classification was conducted for Gram-positive and Gram-negative bacteria, focusing on species/serovar level. The classes for Gram-positive bacteria were Staphylococcus aureus, Staphylococcus epidermidis, Listeria monocytogenes, Bacillus cereus, Bacillus subtilis, other Bacillus spp., enterococci, lactobacilli, and streptococci, while the groups for Gram-negative were Escherichia coli, E. coli O157:H7, Salmonella sp., other enterobacteria, Pseudomonas spp., and Proteus spp. The differences amongst the different groups of Grampositive or Gram-negative bacteria were analyzed using Kruskal-Wallis test, while data were presented in box-whisker plots.
As a second step, data for some species (S. aureus, S. epidermidis, Salmonella sp., E. coli non O157:H7, and E. coli O157:H7) for both thymol and carvacrol were analyzed through PERMANOVA (permutational analysis of variance), using the antimicrobial compounds and the species as categorical predictors, with critical P set to 0.05; the post hoc multiple comparison was conducted using the Mann-Whitney pairwise test. On the same data, a violin graph was built for each combination microorganism/antimicrobial compound, thus gaining 10 different probability distributions.
Statistics were obtained using the software Statistica for Windows ver. 7.0 (Statsoft, Tulsa, OK, USA) or PAST free software ver. 4.2 (Natural History Museum, University of Oslo, Oslo, Norway).

Results and Discussion
EOs, as well as their active compounds, were tested toward a wide range of microorganisms, both in laboratory media and in foods [8,21]; however, the lack of a standardized protocol could be an issue for the definition of sensitive and resistant species, as well as for the definition of a susceptibility hit [22]. Several variables could strongly affect the bioactivity of EOs [1,4,23]; among others, the protocol for their extraction is an issue. In addition, the effects on microorganisms could rely on both the main components and minor ones, as some moieties, when present in traces, could strengthen the bioactivity of an oil [4,7,8].
As a rationale, this paper focuses on two representative compounds, thymol and carvacrol, with a well-defined chemical structure, and whose chemistry is well known to avoid any confounding effect related to EO composition. They have a core phenolic ring in their moieties, and an OH-secondary group, although in different positions, similar densities, and boiling points (0.969 vs. 0.973 mg/L for the density and 233 • C vs. 237 • C for the boiling points) [11,14]. The choice of isomers with similar properties also avoids the cofounding effect related to chemistry. In the following sections, each compound is separately analyzed using non-parametric tests and box plot graphs; namely, the MIC values were analyzed according to different categories. First, the analysis was conducted using Gram-positive or Gram-negative bacteria, yeasts, and molds as predictors. Then, a focus was conducted within Gram-positive or Gram-negative bacteria to point out similarities or dissimilarities. For each category, some representative species were chosen, that is, species for which a high number of MIC values were found in the literature. For this approach, box plot figures were built using the interquartile range for the box and the median as the central point, thus showing the MIC range of thymol and carvacrol for 50% of the strains/species analyzed for each group; moreover, whisker was conducted with theoretical minimum and maximum values, that is, excluding the outliers, thus pointing out the possible range inside which MIC values could be considered typical of the species.
In the last step, a comparative analysis of thymol and carvacrol was conducted, along with a focus on the probability distribution of the data, to point out differences between the two compounds and the characteristic trend (if any) for some species. Figure 1 shows a box plot diagram for all the microorganisms, classified into four main categories: molds, yeasts, Gram-negative, and Gram-positive microorganisms; all data were preliminary standardized as ln (concentration). Median values were 5.76 ln (mg/L) for molds, 6.24 ln (mg/L) for yeasts, 5.99 ln (mg/L) for Gram-negative microorganisms, and 5.76 ln (mg/L) for Gram-positive microorganisms, corresponding to actual values of 317, 513, 400, and 317 mg/L; the differences amongst the categories were not significant (P, 0.66 for Kruskal-Wallis test). The figure also points out the strong variability for the dataset of bacteria, as suggested by the quartiles and minimum and maximum values. could lead to a systematic bias.

Thymol
In addition, the difference in the bacterial resistance to EOs is a matter of debate, as other authors have not reported any differences [27,28]. As correctly stated by Angan et al. [29], the protocols for laboratory testing could strongly affect the results, and the choice between the agar-diffusion and micro-dilution method could lead to differences or not, depending on the ability of hydrophobic EOs to enter into contact with microorganisms. After the analysis on the data related to the macro-categories, MICs were separately analyzed for Gram-negative and Gram-positive bacteria to point out possible species For Gram-negative bacteria, the first and third quartiles were, respectively, 4.13 and 6.90 ln (mg/L), thus highlighting that 50% of Gram-negative microorganisms showed a MIC value for thymol ranging from 60 mg/L to 900 mg/L. The statistics also pointed out the theoretical minimum and maximum values of the distribution (0 and 10.8 ln (mg/L), i.e., 1 and 49,000 mg/L), which further highlight the strong variability of the dataset. In the case of Gram-positive bacteria, the first and third quartiles were 4.85 and 8.06 ln (mg/L) (127-3165 mg/L), while the theoretical minimum and maximum of the distribution were 2.76 and 12.12 ln (mg/L) (16-183,505 mg/L).
In the past, many authors have reported a higher sensitivity of Gram-negative microorganisms than Gram-positive bacteria, and this difference has been generally attributed to the peptidoglycan of Gram-positive microorganisms and the higher resistance that it exerts toward hydrophobic compounds [24][25][26]. The data of this meta-analysis do not confirm this trend, probably depending on the high variability inside the dataset but also on the use of a wide range of microorganisms rather than a single strain. This latter idea confirms the hypothesis that in some situations, the choice of the incorrect type or test strain could lead to a systematic bias.
In addition, the difference in the bacterial resistance to EOs is a matter of debate, as other authors have not reported any differences [27,28]. As correctly stated by Angan et al. [29], the protocols for laboratory testing could strongly affect the results, and the choice between the agar-diffusion and micro-dilution method could lead to differences or not, depending on the ability of hydrophobic EOs to enter into contact with microorganisms.
After the analysis on the data related to the macro-categories, MICs were separately analyzed for Gram-negative and Gram-positive bacteria to point out possible species differences. Figure 2 shows the box plot for Gram-negative species for which 6-10 different MIC values were found, that is, Escherichia coli (separately analyzed for O157:H7 serotype and other E. coli), Salmonella sp., Pseudomonas, Proteus, and other enterobacteria. In particular, the species (or the serotypes) with the highest number of input values were E. coli O157:H7, other E. coli, and Salmonella sp., with median values of 6.91, 6.21, and 5.76 ln (mg/L) (1000, 498, and 317 mg/L), thus pointing out some dissimilarities between the serotype O157:H7 and the other E. coli, as well as between E. coli species and Salmonella sp. However, the groups also showed a strong variability in their dataset. type and other E. coli), Salmonella sp., Pseudomonas, Proteus, and other enterobacteria. In particular, the species (or the serotypes) with the highest number of input values were E. coli O157:H7, other E. coli, and Salmonella sp., with median values of 6.91, 6.21, and 5.76 ln (mg/L) (1000, 498, and 317 mg/L), thus pointing out some dissimilarities between the serotype O157:H7 and the other E. coli, as well as between E. coli species and Salmonella sp. However, the groups also showed a strong variability in their dataset. For E. coli O157:H7, the first and third quartiles were 4.66 and 6.38 ln (mg/L) (106 and 590 mg/L), while the theoretical minimum and maximum MIC values were set by the statistics to 0 and 6.9 ln (mg/L), with some outliers below the minimum, which suggests that for some analyses, some authors used strains, which could not be considered a good test for this serotype, or the application of protocols, which could lead to bias. For the other E. coli strains, the statistics suggest a trend toward a MIC increase, as the first and third quartiles were 5.08 and 8.15 ln (mg/L) (161 and 3460 mg/L), while the minimum and maximum of the distribution were set by the statistics to 4.1 and 10.89 ln (mg/L) (60.34 and 53,100 mg/L).
A high variability was also found for Salmonella sp.; the first and third quartiles for this species were 0.69 and 5.99 ln (mg/L) (corresponding to actual values ranging from 0.5 to 400 ppm), while the extremes of the statistical distribution were set at −3.9 and 11.6 ln (mg/L).
Although the dataset had fewer values, the statistics also confirmed the high resistance of Pseudomonas spp., which showed a median MIC of 6.90 ln (mg/L) and quartiles at 6.21 and 8.99 ln (mg/L) (respectively, 992, 498, and 8000 mg/L).
The last analysis conducted for thymol was on Gram-positive bacteria ( Figure 3); most data were found for Staphylococcus spp. (S. aureus and S. epidermidis) and streptococci. The plot points out the different trends for the two species of the genus Staphylococcus, as S. epidermidis showed a lower variability in MIC distribution (median 5.73 ln (mg/L), and interquartile range 5.73-6.2 ln (mg/L), corresponding to actual values of 308-493 mg/L). In the case of S. aureus, although the median was similar (5.73 ln (mg/L)), the MIC distribution was characterized by higher variability, as one could infer from the interquartile range (4.85-7.60 ln (mg/L), that is, 128-2000 mg/L) and minimum and maximum of the distribution (3.2-10.7 ln (mg/L)). For E. coli O157:H7, the first and third quartiles were 4.66 and 6.38 ln (mg/L) (106 and 590 mg/L), while the theoretical minimum and maximum MIC values were set by the statistics to 0 and 6.9 ln (mg/L), with some outliers below the minimum, which suggests that for some analyses, some authors used strains, which could not be considered a good test for this serotype, or the application of protocols, which could lead to bias. For the other E. coli strains, the statistics suggest a trend toward a MIC increase, as the first and third quartiles were 5.08 and 8.15 ln (mg/L) (161 and 3460 mg/L), while the minimum and maximum of the distribution were set by the statistics to 4.1 and 10.89 ln (mg/L) (60.34 and 53,100 mg/L).
A high variability was also found for Salmonella sp.; the first and third quartiles for this species were 0.69 and 5.99 ln (mg/L) (corresponding to actual values ranging from 0.5 to 400 ppm), while the extremes of the statistical distribution were set at −3.9 and 11.6 ln (mg/L).
Although the dataset had fewer values, the statistics also confirmed the high resistance of Pseudomonas spp., which showed a median MIC of 6.90 ln (mg/L) and quartiles at 6.21 and 8.99 ln (mg/L) (respectively, 992, 498, and 8000 mg/L).
The last analysis conducted for thymol was on Gram-positive bacteria ( Figure 3); most data were found for Staphylococcus spp. (S. aureus and S. epidermidis) and streptococci. The plot points out the different trends for the two species of the genus Staphylococcus, as S. epidermidis showed a lower variability in MIC distribution (median 5.73 ln (mg/L), and interquartile range 5.73-6.2 ln (mg/L), corresponding to actual values of 308-493 mg/L). In the case of S. aureus, although the median was similar (5.73 ln (mg/L)), the MIC distribution was characterized by higher variability, as one could infer from the interquartile range (4.85-7.60 ln (mg/L), that is, 128-2000 mg/L) and minimum and maximum of the distribution (3.2-10.7 ln (mg/L)).

Carvacrol
The same approach was used to gain details and information on the quantitative distribution of MIC values of carvacrol. Figure 4 shows the trend for the macro-categories (molds, yeasts, Gram-negative and Gram-positive bacteria). Bacteria showed a higher variability than fungi, but this result could also be due to the lower observations available for molds and yeasts; namely, for Gram-negative bacteria, the median of MIC values was 5.74 ln (mg/L) with an interquartile range of 5.01-7.82 ln (mg/L) (311 and 150-2490 mg/L). The statistics also predicted the minimum and the maximum of the MIC distribution at 3.9 and 11.5 ln (mg/L) (49.40-100,000 mg/L). In the case of Gram-positive bacteria, MIC values were predicted to be in the range 3.2-8.9 ln (mg/L) (minimum and maximum), with a median value of 5.73 ln (mg/L) (45-7330 and 308 mg/L, respectively); MIC median values for fungi were 5.74-5.77 ln (mg/L) (ca. 300 mg/L).

Carvacrol
The same approach was used to gain details and information on the quantitative distribution of MIC values of carvacrol. Figure 4 shows the trend for the macro-categories (molds, yeasts, Gram-negative and Gram-positive bacteria). Bacteria showed a higher variability than fungi, but this result could also be due to the lower observations available for molds and yeasts; namely, for Gram-negative bacteria, the median of MIC values was   A focus on the MIC distribution at the species level was also possible for some Gram-negative and Gram-positive microorganisms. Concerning Gram-negative bacteria ( Figure 5), the box plot suggests a similar trend for E. coli O157 and other E. coli (median MIC at 5.76 ln(mg/L)), although non-O157 strains showed a higher variability, with an interquartile range of 6.13-8.77 ln (mg/L) (459-6400 mg/L). A focus on the MIC distribution at the species level was also possible for some Gramnegative and Gram-positive microorganisms. Concerning Gram-negative bacteria ( Figure  5), the box plot suggests a similar trend for E. coli O157 and other E. coli (median MIC at 5.76 ln(mg/L)), although non-O157 strains showed a higher variability, with an interquartile range of 6.13-8.77 ln (mg/L) (459-6400 mg/L).

Comparison
As a second step, a PERMANOVA analysis was conducted on some representative species (S. aureus, S. epidermidis, E. coli, E. coli O157:H7, and Salmonella sp.) for which a high number of MIC values had been recovered. For this analysis, two predictors were used: the antimicrobial compound (thymol or carvacrol) and the bacterial species. PER-

Comparison
As a second step, a PERMANOVA analysis was conducted on some representative species (S. aureus, S. epidermidis, E. coli, E. coli O157:H7, and Salmonella sp.) for which a high number of MIC values had been recovered. For this analysis, two predictors were used: the antimicrobial compound (thymol or carvacrol) and the bacterial species. PERMANOVA is a non-parametric approach similar to multivariate ANOVA and is useful to point out the effect of two or more predictors on a dependent variable. Its outputs are the sum of squares, the mean square (that is, the sum of squares standardized to the degree of freedom), the Fisher test (standardized weight of each predictor), and the P-value.
The only significant predictor was the species, while the kind of antimicrobial and the interactive term (antimicrobial compound vs. bacterial species) were not significant ( Table 1). Significant differences were found between S. epidermidis and E. coli (Mann-Whitney test, p < 0.01) and between S. epidermidis or E. coli O157:H7, while the effect of the kind of antimicrobial compound was not significant for any of the tested species (p > 0.05). The lack of differences between thymol and carvacrol was an expected result, as the two compounds are isomers; they have a phenolic ring and a secondary -OH group, although in different positions. These elements could play a major role in the bioactivity of compounds of natural origin, although the position of the secondary group could also cause some differences in the bioactivity between carvacrol and similar compounds lacking the hydroxyl group, like carvacrol methyl ester, methanol, and cymene [23], as evidenced in the case of carvacrol and by the differences. There are conflicting data in the literature, as some authors reported similar MIC values for these two compounds, for example, against E. coli (200 mg/L), S. aures as planktonic cells or biofilms (0.01-0.03%), or S. epidermidis (0.03-0.12%) [31,32]; other authors reported a higher bioactivity of the carvacrol, for example, against Ps. aeruginosa [23]. However, it is worth mentioning that the cited authors performed the experiments on a single strain, and the results were probably affected by strain variability and protocol conditions. As a final step for these representative species, a violin plot was conducted (Figure 7). This approach combines the classical box plot with a probability distribution. It is a powerful tool for the visualization of data from both small-size and high-size samples; moreover, it enables the reader to understand whether the data distribution follows a normal trend or if it is unimodal (with a single value as the mode) or multimodal (with two or more values considered the mode). None of the studied species showed a normal distribution for MIC values of both carvacrol and thymol, and generally, the mode was in the range 5-6 ln (mg/L), thus suggesting that most strains had real MIC values of thymol and carvacrol ranging from 150 to 400 mg/L. In addition, the violin plot shows the existence of possible sub-species or of different sensitivity profiles depending on different experimental approaches, for example, in the case of S. aureus for thymol, with most strains with an MIC value of 5 ln (mg/L) (150 mg/L) and a few strains with a very sensitive trend and an MIC of −1 or −2 ln (mg/L) (0.13-0.37 mg/L). Possible sub-classes were also pointed out in the case of E. coli again for thymol: most strains had a resistant trend (MIC > 5 ln (mg/L)) and a few strains had a sensitive trend (MIC at −1 ln (mg/L)). The existence of classes or a strong variability amongst the strains of a single species could be an issue, as it suggests an unpredictable use for practical applications (foods or clinical uses, among others), as also suggested for other natural products [22]. and carvacrol ranging from 150 to 400 mg/L. In addition, the violin plot shows the ence of possible sub-species or of different sensitivity profiles depending on differe perimental approaches, for example, in the case of S. aureus for thymol, with most s with an MIC value of 5 ln (mg/L) (150 mg/L) and a few strains with a very sensitive and an MIC of −1 or −2 ln (mg/L) (0.13-0.37 mg/L). Possible sub-classes were also po out in the case of E. coli again for thymol: most strains had a resistant trend (MIC (mg/L)) and a few strains had a sensitive trend (MIC at −1 ln (mg/L)). The existe classes or a strong variability amongst the strains of a single species could be an iss it suggests an unpredictable use for practical applications (foods or clinical uses, a others), as also suggested for other natural products [22]. For other combinations of antimicrobial compounds/microorganisms, the s showed a homogeneous trend with a strict range of values for the MIC, along with abnormalities in the distribution, which are probably related to methodological iss to the use of sensitive strains; this kind of distribution was found for S. epidermidis coli O157:H7 for both thymol or carvacrol, thus suggesting the possibility of pointin well-defined ranges of concentrations of thymol or carvacrol that are useful for antim bial purposes. For other species (for example, Salmonella sp.), the distribution was n mogeneous, and the strains belonging to them could experience heterogeneous t with strong intra-species variability, suggesting the need for a preliminary screen point out the concentrations of thymol and carvacrol exerting an antimicrobial actio For other combinations of antimicrobial compounds/microorganisms, the species showed a homogeneous trend with a strict range of values for the MIC, along with some abnormalities in the distribution, which are probably related to methodological issues or to the use of sensitive strains; this kind of distribution was found for S. epidermidis or E. coli O157:H7 for both thymol or carvacrol, thus suggesting the possibility of pointing out well-defined ranges of concentrations of thymol or carvacrol that are useful for antimicrobial purposes. For other species (for example, Salmonella sp.), the distribution was not homogeneous, and the strains belonging to them could experience heterogeneous trends with strong intra-species variability, suggesting the need for a preliminary screening to point out the concentrations of thymol and carvacrol exerting an antimicrobial action.

Conclusions
EOs could be promising alternatives to inhibit a wide range of microorganisms, both for food preservation and for clinical uses. The survey of the literature and the analysis of the MIC values of thymol and carvacrol suggest that there are some issues that should be addressed before an effective scaling up of these compounds can take place. The most important could be the strong variability among the microorganisms and the lack of standard protocols and reference strains.
Despite the issues and limitations of using data produced by different researchers and authors with very different experimental approaches, this paper could offer a contribution to the debate on the bioactivity of EOs. First, it confirms the strong variability in the resistance/susceptibility profiles of bacteria and fungi and the existence of differences among strains of the same species; however, it also offers novel points, perspectives, and ideas for future research. The most important key-points could be summarized as follows: (a) There is no evidence of a different resistance between Gram-positive and Gramnegative bacteria and fungi, at least for thymol and carvacrol and at least for the data used in this paper. Some differences were found at the species level (for example, between S. epidermidis and E. coli). (b) The probability distribution found through the violin plot pointed out a possible homogeneous trend toward resistance to thymol and carvacrol for some species (S. epidermidis, E. coli O157:H7), while for other species (for example, Salmonella sp.), the trend was strongly variable, and this could be an issue to think for an effective industrial use of some Eos. (c) Some species could show different profiles of resistance/sensitivity, as suggested, for example, for E. coli in the case of thymol. An open question is whether these different profiles depend on the existence of sub-species or on different experiment set-ups. (d) There is a well-defined range for thymol and carvacrol concentrations (5-6 ln (mg/L); 150-400 mg/L), where the probability of finding MIC for many species is high. (e) Finally, the statistical treatment of data is crucial, as most of the evidence and new perspectives in this paper were found only after a standardization of the data as a natural logarithm and using both non-parametric tests and probability distributions (i.e., a violin graph).
There is a lot of work to do in the field of the antimicrobial use of EOs, mainly in terms of protocol standardization, definition of thresholds, and guidelines, which are fundamental steps to gain robust data and build robust sensitivity/resistance profiles. This paper wants to offer a contribution to this context, also proposing possible approaches for data treatment and their standardization.