Abstract

   Background/Objectives: This study investigates the metabolic profile of
   a single dose of etodolac in healthy volunteers, focusing on
   pharmacokinetics, clinical parameters, and metabolomic variations to
   identify biomarkers and pathways linked to drug response, efficacy, and
   safety. Methods: Thirty-seven healthy volunteers, enrolled after
   rigorous health assessments, received a single dose of etodolac
   (Flancox^® 500 mg). Pharmacokinetic profiles were determined using
   tandem mass spectrometry analysis, and the metabolomic profiling was
   conducted using baseline samples (pre-dose) and samples at maximum drug
   concentration (post-dose) via liquid chromatography coupled with a
   quadrupole time-of-flight mass spectrometer. Network analysis was
   employed to interpret the data. Results: Correlations were observed
   between metabolomic profiles and pharmacokinetic parameters as well as
   clinical characteristics. Notably, metabolites derived from arachidonic
   acid, such as prostaglandins and leukotrienes, were linked to
   etodolac’s pharmacokinetics. Other metabolites involved in pathways
   like cholesterol biosynthesis, bile salts, riboflavin, and retinoic
   acid signaling were correlated with hematological and liver function
   parameters. These findings are consistent with the infrequent adverse
   events reported by participants, including hematological and
   biochemical changes in liver function. Conclusions: A set of
   metabolites was identified in possible associations between specific
   pathways and unusual side effects, comparing the metabolic profiles
   before and after doses of etodolac. Our results highlight the
   importance of optimizing drug therapy and minimizing adverse events by
   taking into account individual metabolic profile information.

   Keywords: etodolac, metabolomics profile, metabolic pathways,
   pharmacokinetics

1. Introduction

   The World Health Organization (WHO) recommends using non-steroidal
   anti-inflammatory drugs (NSAIDs) to treat mild, moderate, and severe
   inflammation/pain [[42]1]. Etodolac (ETO) is the gold standard for
   osteoarthritis management and does not inhibit gastric prostaglandins
   [[43]2]. In Brazil, ETO, a NSAID, remains of clinical and commercial
   interest. The website of the Brazilian regulatory authority [[44]3]
   reports that bioequivalence studies of new formulations are currently
   underway. Therefore, adequate knowledge of the effects of drugs used in
   inflammation and pain management, from physiology to pharmacology, is
   essential for appropriate and optimized inflammation/pain management
   [[45]4].

   ETO is a NSAID classified by the Biopharmaceutics Classification System
   as BCS class II (dissolution rate-dependent absorption). ETO is highly
   safe and effective in relieving pain and controlling inflammation. ETO
   has poor solubility and a dissolution rate that could result in
   variable oral absorption and inconsistent analgesic responses [[46]5].
   NSAIDs inhibit the synthesis of prostaglandin (PG) by cyclooxygenase
   (COX), an enzyme of which two isoforms (COX-1 and COX-2) with different
   functions have recently been identified. Selective cyclooxygenase-2
   (COX-2) inhibitors differ from traditional NSAIDs in their preferential
   inhibition of the COX-2 receptor over the COX-1 receptor. COX1 is
   essential for the maintenance of gastrointestinal and renal function,
   while COX-2 converts arachidonic acid into prostaglandins.
   Prostaglandins are potent mediators of inflammation, pain, and fever.
   ETO exerts its therapeutic effects primarily by inhibiting the enzyme
   cyclooxygenase 2 (COX-2) [[47]5,[48]6,[49]7,[50]8]. So far, the results
   obtained in different clinical trials indicate that these selective
   COX-2 inhibitors are as effective in combating inflammation and
   hyperalgesia as conventional NSAIDs, with the added benefit of
   producing fewer gastrointestinal lesions [[51]9]. In this context, it
   is also interesting to evaluate the impact of selective COX2 inhibitors
   and check their possible influences from a metabolic approach.

   Pharmacokinetics, the ability to predict the concentration of a drug in
   the body over time, including absorption, distribution, metabolism, and
   excretion—ADME—is a traditional method of predicting pharmacokinetics
   based on a drug’s chemical structure [[52]10]. However, these methods
   have limitations. They are not always able to predict the behavior of a
   drug in a given individual. Pharmacometabolomics offers several
   advantages in drug development by integrating metabolomics with
   pharmacokinetics. Analysis of the metabolome improves the accuracy and
   sensitivity of the identification of factors that influence drug
   absorption, distribution, and/or biotransformation, as well as
   potential safety events [[53]11]. It is often used with other drugs,
   and it is important to know its metabolomics in single doses [[54]10].
   Prolonged or high-dose use of ETO is associated with an increased risk
   of peptic ulcer and gastrointestinal bleeding; in addition, ETO may
   affect renal function, particularly in patients with pre-existing risk
   factors, and may increase the risk of cardiovascular events such as
   myocardial infarction and stroke [[55]12], although the exact pathways
   by which this occurs remain to be elucidated. Song et al., 2017, have
   primarily focused on traditional pharmacokinetic (PK) parameters and
   bioequivalence assessments. While these studies provide valuable
   insights into drug ADME, they have not incorporated a metabolomic
   approach [[56]13].

   This study aims to investigate the relationship between variations in
   the metabolic profile and the pharmacokinetics as well as laboratorial
   and clinical effects of a single dose of etodolac in healthy subjects.
   It seeks to evaluate the potential of metabolomics to identify specific
   pathways and metabolites of drug response in the assessment of drug
   efficacy and safety. To predict key pharmacokinetic parameters and
   potential safety issues, pharmacometabolomics strategies using
   ultra-performance liquid chromatography-quadrupole time-of-flight mass
   spectrometry (UPLC-QToF) were applied to study healthy human plasma
   samples. The results of this study may help to better elucidate the
   involved metabolites and possible biological mechanisms underlying the
   use of a single dose of etodolac through metabolomic approaches, thus
   advancing personalized strategies.

2. Results

2.1. Volunteers’ Clinical and Laboratory Characteristics

   The study was completed with 37 healthy adult participants, 18 female
   and 19 male, aged between 18 and 41 years old (mean 28.03 ± SD 6.8) and
   with a body mass index between 18.7 and. 29.1 kg/m^2 (mean 24.2 ± SD
   2.7). Details of the experiment are shown in [57]Figure 1 and methods,
   as well as in [58]Table S1. In this study, there were 19 adverse events
   recorded -that is, individuals with results outside the reference
   values used to define individuals as healthy at baseline- (in its
   entirety of predictability: not expected and intensity: mild), which
   are presented in the Summary of Adverse Events in [59]Table S2.

Figure 1.

   [60]Figure 1
   [61]Open in a new tab

   Representative diagram of sample/data collection points used in this
   study.

2.2. Determination of Etodolac Pharmacokinetics

   [62]Figure 2 plots individual pharmacokinetics curves (concentration
   vs. time). In brief, the overall pharmacokinetic parameters were as
   follows: area under the curve from time zero to time t (AUC0-t)
   (h·μg/mL) = 202.83 (72.05), maximum concentration (Cmax) (μg/mL) =
   35.54 (9.14), time to maximum concentration (T[max]) (h) = 2.24 (1.11),
   elimination rate constant (Kel) (1/h) = 0.09 (0.02), and half-life
   (T1/2) (h) = 8.2 (2.73). [63]Table S3 shows the pharmacokinetics
   parameters for each participant.

Figure 2.

   [64]Figure 2
   [65]Open in a new tab

   Plot of all individual pharmacokinetic curves.

2.3. Metabolomic Analysis

   The untargeted analysis detected 2466 features in the positive mode and
   4777 in the negative mode. Our feature selection model detected 1390
   features in negative ion mode and 290 in positive ion mode as
   differences at the pre- and post-dose time points. Of these 1680
   differential features, 70 were putatively identified as metabolites
   ([66]Table S4). [67]Figure 3 shows the dispersion of participants and
   QC samples using principal component analysis and the heat maps for the
   relative abundance of selected and identified features.

Figure 3.

   [68]Figure 3
   [69]Open in a new tab

   PCA plots for negative and positive ionization modes and normalized
   relative abundance heatmaps for selected compounds, detected in
   negative (left) and positive (right) ionization modes.

2.4. Enrichment Analysis

   The pathway enrichment analysis ([70]Table S4) used the 70 selected and
   putatively identified features as input. With an adjusted p-value by
   FDR ≤ 0.05, there were 131 pathways. [71]Table S5 shows the results
   generated by the RaMP package. The Reactome database was used for the
   highly related pathways in the results. Some Wikipath pathways are
   their own versions of the Reactome pathways. In this way, the most
   important functional or structural features between the enriched
   pathways can be understood and discussed, including the following:
   digestion and absorption = R-HSA-8963743 (FDR p-value = 0.0011);
   digestion of dietary lipid = R-HSA-192456 (FDR p-value = 0.00015);
   transport of vitamins, nucleosides, and related molecules=R-HSA-425397
   (FDR p-value = 0.00015); synthesis of bile acids and bile salts =
   R-HSA-192105 (FDR p-value = 0.026); alpha-linolenic acid metabolism =
   map00592 (FDR p-value = 0.026); metabolism of alpha-linolenic acid
   (WP4586) (FDR p-value = 0.042); and prostaglandin and leukotriene
   metabolism in senescence (WP5122) (FDR p-value = 0.044).

2.5. Community Network

   After module discovery using community network analysis, 12 modules
   representing sets of metabolites associated with PK parameters and
   clinical/laboratory participant characteristics ([72]Table S6). The
   pre-dose metabolome dataset formed five modules with 19 edges, where 8
   metabolites were significantly associated with systolic pressure, 9
   characteristics in serum, and 1 in urine-related. In the post-dose
   metabolome dataset, 35 metabolites were significantly associated with 5
   PK parameters, and 13 characteristics in serum by 67 edges that formed
   seven modules.

   [73]Figure 4A shows the community networks in the pre-dose period, and
   [74]Figure 4B shows the community network in the post-dose period.
   [75]Table S6 shows the complete lists of members for the pre-dose and
   post-dose communities.

Figure 4.

   [76]Figure 4
   [77]Open in a new tab

   Community networks were constructed using PK parameters, clinical and
   laboratory traits, and identified metabolites at (A) pre-dose and (B)
   post-dose. A red label above each node indicates the number of
   communities or modules.

2.6. Integrating Pathway and Network Results

   The set of metabolites associated with the PK parameters and/or
   laboratorial and clinical characteristics in each module derived from
   the network analysis was integrated, according to possible
   functional/biological pathway correspondence, with the results obtained
   in the enrichment pathway derived from the differential metabolites
   between the pre and post-dose time points, as shown in [78]Table 1.
   These combined results showed the association between metabolism at
   pre-dose with systolic blood pressure and laboratory tests such as
   renal function. Correlations were found between the relative abundance
   of certain metabolites after ETO administration with ADME parameters,
   laboratory-related liver and kidney function, and white blood cell
   counts.

Table 1.

   Integrated pathways and community network analysis results.
   Parameters Molecule
   (ID HMDB/PubChem) Biological/Chemical Class/Function/Pathway
   Predictive T0 (Baseline metabolism pre-dose)
   Laboratory/Clinical characteristics
   Systolic pressure dTDP-D-galactose
   (HMDB0006876) Cholesterol biosynthesis = WP1795
   Cholesterol synthesis disorders = WP5193.
   Mevalonic acid (HMDB0000227) Cholesterol metabolism (WP5304)
   Aspartate aminotransferase seric L-Palmitoylcarnitine
   (HMDB0000222) Carnitine metabolism (R-HSA-200425) An ester of carnitine
   that facilitates the transfer of fatty acids from the cytosol to the
   mitochondria during fatty acid oxidation.
   Total bilurubin
   Direct bilirubin
   Indirect bilirrubin
   Leukocites urine Dityrosine
   (HMDB0006045) Dityrosine has also been found to be involved in
   oxidative/nitrative stress under a wide range of conditions and in a
   variety of biological systems.
   Albumin seric
   Hemoglobin seric
   Erythrocytes seric
   Hematocrit seric Palmitoleic acid
   (HMDB0003229) Omega-9 fatty acid synthesis (WP4724)
   Effect Tmax (Metabolism post-dose)
   Pharmacokinetic parameters
   AUC0-t (h·μg/mL)
   Cmax (μg/mL) PS(PGF1alpha/20:5)
   (HMDB0282870) Phospholipids with anti-inflammatory properties
   PS(PGF2alpha/18:1)
   (HMDB0281557) Phospholipids with anti-inflammatory properties
   Kel (1/h), T1/2 (h) 15-Epi-lipoxin A4
   (HMDB0012587) Tryptophan metabolism (WP465)
   Riboflavin
   (HMDB0000244) Riboflavin Metabolism = map00740
   6-Ketoprostaglandin E1
   (HMDB0004241) Eicosanoid metabolism via cyclooxygenases (COX)
   Laboratory/Clinicals characteristics
   Creatinine seric
   Fasting glucose seric
   Hemoglobin seric hematocrit seric
   Erythrocytes seric L-Palmitoylcarnitine
   (HMDB0000222) L-palmitoylcarnitine is an acylcarnitine. The acyl
   groups, organic acids and fatty acids are transported from the cytosol
   to the mitochondria for degradation to produce energy.
   Linoleic acid
   (HMDB0000673) Belongs to the class of organic compounds known as
   linoleic acids and derivatives.
   Palmitoleic acid
   (HMDB0003229) Peptide hormone metabolism = R-HSA-2980736, sSynthesis,
   secretion, and inactivation of Glucagon-like Peptide-1 (GLP-1) =
   R-HSA-381771/WP2728; free fatty acid receptors = R-HSA-444209, Omega-9
   fatty acid synthesis = WP4724
   3-Oxooctadecanoic acid (HMDB0010736) It belongs to the class of organic
   compounds called long-chain fatty acids.
   Transaminase oxaloacetic acid (TGO) seric Etodolac
   (HMDB0014887) Drug
   Flavin mononucleotide
   (HMDB0001520) Metabolism of nitric oxide: NOS3 activation and
   regulation (WP1850), riboflavin metabolism = map00740, cytosolic
   iron-sulfur cluster assembly = WP2690, metabolism of cofactors =
   WP4990, riboflavin and CoQ disorders = WP5037, vitamin B12 disorders =
   WP4271
   Total bilirubin seric
   Indirect bilirubin seric
   Leukocytes seric
   Segmented neutrophils seric Glucuronic acid
   (HMDB0000127) Phosphatidylinositol phosphate metabolism = map00562
   Etodolac acyl glucuronide
   (HMDB0060916) Metabolite drug
   Ascorbic acid
   (HMDB0000044) Carnitine synthesis = R-HSA-71262/WP4996, catecholamine
   biosynthesis = R-HSA-209905, vitamin C (ascorbate) metabolism =
   R-HSA-196836, effects of nitric oxide = WP1995, metabolism of
   amine-derived hormones = WP3548, vitamin B12 metabolism = WP1533,
   pluripotent stem cell differentiation pathway = WP2848
   Mean corpuscular volume Vitamin A
   (HMDB0000305) Metabolism of fat-soluble vitamins = R-HSA-6806667,
   signaling by retinoic acid = R-HSA-5362517, retinoid metabolism and
   transport = R-HSA-975634, triglyceride metabolism = R-HSA-8979227,
   digestion and absorption = R-HSA-8963743, vitamins A and D—action
   mechanisms = WP4342
   Red cell distribution width Arg-gly-asp-ser
   (HMDB0248573-
   PubChem 107775) Tetrapeptide may play an important role in the cell
   attachment domain of fibronectin
   [79]Open in a new tab

   The table lists the molecules that appear simultaneously in the results
   of pathway enrichment and network analysis. For each metabolite, the
   first column indicates the associated pharmacometric parameter and
   participant traits, either baseline or post-ETO. The last columns
   indicate the biological function and/or pathway defined by the
   enrichment analysis. Identificator Human Metabolome Database (HMDB),
   Reactome Homo Sapiens pathway (R-HSA), WikiPathways (WP), Public
   Chemical Databases (PubChem).

3. Discussion

   We used metabolite set enrichment and community network analysis to
   detect and understand the associations of differential metabolic
   profiles with pharmacokinetics, clinical, and laboratory parameters in
   healthy volunteers before and after receiving a single dose of ETO.
   Significant associations were found between the pre and post-dose ETO
   metabolomic profiles.

   The baseline demographics did not show any significant differences in
   terms of gender, age, and body mass index. Infrequent but mild adverse
   events were reported, as discussed below. The values of the
   pharmacokinetic parameters obtained in this study are comparable to
   those reported so far in the literature for ETO [[80]14,[81]15].

3.1. Mapping Predictive Associations

   In this context, metabolic changes observed in individuals after a
   single dose of ETO may result from primary activity or secondary
   physiological disturbance. The metabolism of prostaglandin and
   leukotriene metabolism were biological pathways that represented
   significant changes in our study. Our results support previous reports
   by previous studies that NSAID drugs potentiate the release of
   anaphylactic slow-reactants, which are now known to be a mixture of the
   leukotrienes C4, D4, and E4 [[82]6,[83]15].

   The association of metabolite sets at pre-dose with pharmacokinetic
   parameters may explain or even predict biological pathways related to
   the drug’s ADME. Although the pathways obtained in the enrichment
   analyses do not explain the module network functional activity, they do
   explain functionally where the metabolite associated with the module
   may act.

3.1.1. Association with Pharmacokinetic Parameters

   At pre-dose, we found no association between metabolite networks and
   pharmacokinetic parameters. That is, none of the differential
   metabolites between pre-dose and post-dose were able to ’predict’
   parameters related to the pharmacokinetics of ETO in this study.

3.1.2. Laboratory/Clinical Parameters

   The baseline metabolomic assessment shows metabolites mainly involved
   in cholesterol biosynthetic pathways, cholesterol synthesis disorders,
   and omega-9 fatty acid synthesis are correlated with clinical (systolic
   blood pressure) and laboratory signs plasma blood tests (total
   cholesterol level, renal function, biliary pigment, and liver function)
   and urine with white blood cell count during the post-trial evaluation
   of participants.

   The production of mevalonic acid by the enzyme
   3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase is the
   rate-limiting step in the biosynthesis of cholesterol. Also, the
   dTDP-D-galactose belongs to the class of organic compounds known as
   cholesterols and derivatives. Cholesterol fulfills a number of
   biological functions and is necessary for the successful maintenance of
   cellular homeostasis. It plays a central role in maintaining the
   rigidity and fluidity of cell membranes, acts as a precursor to bile
   acids, and aids in the synthesis of steroids and vitamin D [[84]16].
   Aldosterone is synthesized in the adrenal cortex. It acts mainly on the
   renal tubules, stimulating potassium excretion and the uptake of sodium
   and water. The ultimate effect is an increase in blood pressure
   [[85]17]. Cholesterol also serves as a precursor for bile acids and
   influences blood pressure through the synthesis of aldosterone.

   Omega-3, -6, and -9 fatty acids are unsaturated fatty acids that have
   multiple biological effects and health benefits. Different omega-9
   fatty acids have different pharmacological effects. These include
   modulation of inflammation, lipid, cardiovascular, and cancer disorders
   [[86]18]. Palmitoleic acid was identified as a lipokine that has
   important positive effects on metabolic disorders. [[87]19]. In our
   study, this metabolite was associated with albumin and hemoglobin
   levels at T0, which may be useful/important in assessing the
   nutritional status of the individual and identifying possible renal or
   hepatic problems prior to ETO administration.

   Total, direct, and indirect bilirubin and urinary leukocytes are
   biomarkers that can provide a more complete picture of an individual’s
   health status, particularly in the context of liver, inflammatory, and
   oxidative stress-related diseases. Recent studies suggest that human
   serum albumin, the major plasma protein, may have a direct
   vasculoprotective antioxidant effect [[88]20]. Dityrosine is a product
   of radical oxidation [[89]21]. Previous studies have shown that
   dityrosine production is approximately twofold increased in uremic
   human serum albumin from dialysis patients. Dityrosine may help
   phagocytes respond to inflammation [[90]20]. The process described here
   may also be associated with the enrichment of several oxidative
   stress-related pathways, such as the cellular response to hypoxia and
   the effects of nitric oxide, as shown in [91]Table S5.

3.2. Exploring Explanatory Parameters

   The results indicate a possible trend toward a cause and/or effect
   relationship of ETO in the body. In fact, correlations were obtained
   between certain metabolites after ETO administration and
   pharmacokinetic and/or clinical parameters.

3.2.1. Pharmacokinetic Parameters

   We found a correlation between AUC0-t/Cmax and PS(PGF1α/20:5) and
   PS(PGF2α/18:1) in our findings. The anionic phospholipids,
   phosphatidylserine (PS), and phosphatidylglycerol (PG) are naturally
   occurring molecules with anti-inflammatory properties. PS, a crucial
   component of cell membranes, undergoes oxidation [[92]22]. This
   oxidation endows PS with new properties that can affect various
   biological processes. These oxidized forms of PS may play a role in
   regulating the inflammatory response and immunomodulatory activity
   [[93]22]. Additionally, they have been linked to increased activity of
   certain drug transporters [[94]23].

   In addition, tryptophol was identified, showing a more than fourfold
   increase between T0 and Tmax, and correlating positively with Cmax. The
   relationship between tryptophol metabolism and etodolac
   pharmacokinetics may be indirect and complex. Tryptophol, a catabolite
   of tryptophan, participates in multiple metabolic pathways that can
   influence liver enzyme activity, such as tryptophan metabolism
   [[95]24,[96]25]. These enzymes are responsible for the
   biotransformation of etodolac. However, further research is needed to
   support this hypothesis.

   About the Kel, we obtained evidence for two metabolites associated with
   arachidonic acid metabolism, specifically the 15-epi-lipoxin A4
   [[97]26]. This metabolite increases in relative abundance at Tmax
   compared to T0. It is positively associated with the elimination
   constant and negatively associated with T1/2. The 6-ketoprostaglandin
   E1 [[98]27] increases twist time in relative abundance at Tmax compared
   to T0 and has been shown to be negatively associated with T1/2. Our
   results tend to support direct or indirect disturbances after ETO use
   [[99]6,[100]28].

3.2.2. Laboratory/Clinical Characteristics

   Our results on clinical traits networks indicate variation in plasma
   concentrations of metabolites involved in fatty acid metabolism. They
   are either substrates, products, or regulators. Directly involved in
   the energy-producing pathway are L-palmitoylcarnitine and
   3-oxooctadecanoic acid. Linoleic acid and palmitoleic acid are involved
   in several physiological processes. These include inflammation and
   insulin sensitivity. Alterations in their metabolism can affect
   cellular integrity, renal function (creatinine), and glycemic
   regulation (glucose). In this network, the metabolite
   palmitoylcarnitine decreased in abundance in the blood after ETO
   administration, correlating with hematocrit values. This metabolite may
   be involved in pathways associated with retinoic acid signaling, which
   is implicated in several cellular processes, including response in
   vascular endothelial cells, and could influence hematological
   parameters [[101]29].

   Etodolac abundance showed a negative correlation with serum oxaloacetic
   transaminase (TGO). ETO, like other NSAIDs, can affect hepatic
   metabolism, which may be reflected in changes in the levels of enzymes
   such as TGO [[102]15]. Although they are generally well tolerated by
   the liver, prolonged use or high doses may damage liver cells.
   Consequently, flavin mononucleotide increased relative abundance
   post-ETO dose and correlated negatively with serum TGO. Consistent with
   this finding, we also found a correlation between etodolac acyl
   glucuronide (an ETO metabolite) and serum bilirubin levels.

   Variation plasmatic of flavin mononucleotide was found. Riboflavin is
   known to be involved in a variety of redox reactions central to human
   metabolism through the cofactors flavin mononucleotide (FMN) and flavin
   adenine dinucleotide (FAD) [[103]30]. Insufficient riboflavin intake
   would, therefore, be expected to lead to disruptions in intermediary
   metabolic steps, with functional implications [[104]31]. A recent
   publication investigating the impact of a biliary injury on the
   recurrence of biliary cancer and benign disease after liver
   transplantation analyzed possible underlying mechanisms associated with
   FMN [[105]32]. Therefore, it may be essential to consider both the
   direct effect of the drug and the patient’s nutritional status when
   evaluating liver test results during treatment with etodolac.

   Our study also showed a pre/post increase in vitamin A, considered an
   important micronutrient in the mammalian diet, which exists in three
   forms: retinal, retinol, and retinoic acid, the latter being the most
   metabolically active [[106]33], retinoic acid, which acts through
   nuclear retinoic acid receptors and is a potent regulator of pattern
   formation during embryonic development, which is required for the
   homeostasis of adult tissues [[107]34].

3.2.3. Adverse Event-Related Association

   Prolonged or high-dose use of ETO is associated with an increased risk
   of peptic ulcer, gastrointestinal bleeding, renal dysfunction,
   myocardial infarction, and stroke [[108]12]. Our unique dose study of
   ETO showed an infrequent adverse events profile that included changes
   in hematological profile and biochemical liver function. However, our
   results are limited in correlating these findings with possible
   associations between the observed adverse events and the properties of
   ETO as a selective COX-2 inhibitor. Our module and enrichment analysis
   consistently shows associations between metabolite sets and hemogram
   and liver function at both pre and post-dose. Network analysis showed
   consistent pre- and post-dose ETO associations between specific fatty
   acids (palmitoleic acid, linoleic acid) and erythrocyte levels, which
   may indicate an effect on membrane phospholipid synthesis and cell
   survival.

   In addition, consistent pre- and post-dose ETO associations were
   observed between arg-gly-asp-ser, L-palmitoylcarnitine, dityrosine
   ([109]Table S6), and biochemical liver function. It is plausible to
   think of this association since the liver plays a central role in the
   metabolism of amino acids and is responsible for their synthesis,
   degradation, and interconversion. Changes in the levels of these
   compounds may reflect liver dysfunction. Other associations were
   observed, although the underlying mechanisms remain unclear. To fully
   understand these associations and their clinical implications, further
   research is needed to evaluate changes in the metabolic profile after
   long-term use of ETO.

3.3. Limitations and Conclusions

   Several study limitations should be considered, such as the sample
   size, which limits the power of the study to detect small effects. The
   genetic conditions of the participants are important factors that may
   influence the pharmacokinetics of etodolac, such as variability in its
   metabolism, classically through genetic polymorphisms in UGT1A9 and
   CYP2C9 [[110]35], were not accessible, so their impact on drug
   metabolism was not evaluated. Unidentified compounds could also
   confound the results. However, the controlled trial design of
   bioequivalence studies minimizes confounding factors, and the
   untargeted metabolomic analysis offers a comprehensive view of
   metabolic changes in patients, with identification accuracy deemed
   acceptable based on established criteria.

   The complex interaction between ETO and the metabolic landscape of the
   human body is highlighted in this study. Pathways involved in the
   inflammatory process seem to be affected following ETO administration.
   Variations observed pre- and post-dose ETO dosing may explain rare
   adverse events. Our findings reinforce the significant contribution of
   a metabolomic approach to a comprehensive understanding of PK profiles.
   By providing a predictive, sensitive, and specific assessment of drug
   response and adverse effects, our results demonstrate the potential of
   metabolomics as a valuable tool for the identification of specific
   metabolites and pathways in the understanding of individual variability
   in response to a drug in clinical trials. The identification of
   specific pathways and metabolites associated with adverse events in
   healthy volunteers provides the basis for the development of predictive
   biomarkers in patient populations, for example, in the context of
   inflammation and/or pain. This could enable clinicians to proactively
   monitor patients for potential side effects and adjust treatment
   accordingly. For future perspectives, we suggest studies incorporating
   metabolomic analyses in studies of prolonged use or high doses of ETO
   to validate our initial findings. We also suggest the initiation of
   studies investigating the metabolites set listed in our study as
   possible relevant molecules for more specific biological tests related
   to the mechanism of action of ETO to investigate the clinical utility
   of these findings in patient populations and explore the potential for
   metabolomic profiling-based personalized medicine. We also suggest
   considering genetic factors that may influence these pharmacophenotypic
   relationships.

4. Materials and Methods

4.1. Study Design and Healthy Human Subjects

   This study was a nested, single-dose, open-label, randomized
   pharmaco-bioequivalence study conducted under biosafety and ethical
   standards ([111]Figure 1). The data used in this study were obtained
   from 37 healthy volunteers administered the reference ETO drug
   (Flancox® 500 mg, by Apsen Farmaceutica, São Paulo, Brazil) and
   selected for the study after assessment of their health status by
   clinical evaluation before and after ETO administration, by physical
   examination and electrocardiogram, and by a comprehensive set of
   laboratory tests (systolic pressure, diastolic pressure, heart rate,
   temperature clinical, total cholesterol seric, triglycerides seric,
   urea seric, creatinine seric, fasting glucose seric, uric acid seric,
   aspartate aminotransferase seric, alanine aminotransferase seric, total
   bilirubin seric, direct bilirubin seric, indirect bilirubin seric,
   alkaline phosphatase seric, total protein seric, albumin seric,
   hemoglobin seric, erythrocytes seric, hematocrit seric, mean
   corpuscular volume (VCM) seric, mean corpuscular hemoglobin (MCH)
   seric, corpuscular hemoglobin concentration (CHCM) seric, red cell
   distribution width (RDW) seric, leukocytes seric, segmented neutrophils
   seric, eosinophils seric, lymphocytes seric, monocytes seric, platelets
   seric, density urine, leukocytes urine, erythrocytes urine, pH urine),
   as listed in detail in the [112]Table S1. Venous blood (5 mL) was
   collected for development and validation analysis using tubes
   containing ethylenediaminetetraacetic acid (EDTA). After collection,
   the samples were centrifuged at 2000× g for 10 min at 4 °C.
   Subsequently, the plasma was carefully collected and stored at −80 °C
   for later analysis. The same parameters were used to monitor adverse
   events for safety and tolerability. None of the subjects were smokers,
   alcoholics, or had a history of drug abuse.

4.2. Ethics

   The protocol complied with Brazilian legislation on human clinical
   research and was duly approved by the Research Ethics Committee.
   Institutional Review Board (IRB) approval for the study was obtained
   under protocol number 58683422.6.0000.5514, dated 1 May 2022. All
   subjects gave their written informed consent and were free to withdraw
   from the study at any time.

4.3. Determination of Pharmacokinetic Profile

   The analytical method used to calculate the pharmacokinetic parameters
   was validated according to Resolution No. 27/2012 of the Agência
   Nacional de Vigilância Sanitária (ANVISA). In this study, an LC-MS/MS
   system was used to quantify ETO. Chromatographic separations were
   performed using an LC-20AD analytical pump (Shimadzu, Kyoto, Japan) and
   a SIL-20A HT autosampler (Shimadzu, Kyoto, Japan), coupled to a Quattro
   Micro (Micromass, Newcastle, UK) system for triple quadrupole mass
   spectrometry analyses with an electrospray ionization (ESI) source
   (Waters, Newcastle, UK). The targeted analysis for ETO was detected by
   a multiple reaction monitoring (MRM) transition of 286 > 242, and
   mefenamic acid (Internal Standard) was detected by an MRM transition of
   240 > 196 in negative acquisition mode. Data were collected using
   MassLynx 4.1 (Waters, Newcastle, UK). Further information on the
   validation method can be found in [113]Supplementary Materials and
   Methods. WinNonLin 8.3 software (Pharsight, New Jersey, USA) was used
   to calculate pharmacokinetic parameters using non-compartmental
   statistics.

4.4. Metabolomic Profile

   Samples were collected at two times: (i) baseline samples (time 0 h),
   termed the pre-dose group, and (ii) samples at maximum drug
   concentration (Cmax), termed the post-dose group. Samples were
   randomized, and 25 μL of each was collected to form a pooled sample for
   quality control (QC). An aliquot of plasma (200 μL) was extracted by
   adding cold methanol (400 μL) for protein precipitation. The samples
   were then vortexed for 30 s and centrifuged at 12,879× g for 10 min at
   4 °C. Finally, the supernatant (400 μL) was collected and dried over a
   stream of nitrogen gas (N[2]). Samples were resuspended with 200 µL of
   acetonitrile (ACN)/water (H[2]O) (1:1 v/v). Sample extraction and
   analysis were randomized to reduce instrumental and biological bias. QC
   samples were analyzed every 10 samples to check for variations in
   extraction and system stability.

   The analysis was adapted [[114]36] and performed on an ACQUITY® FTN H
   Class (Waters, Newcastle, UK) liquid chromatography (UPLC) coupled to a
   XEVO-G2XS (QToF) quadrupole time-of-flight mass spectrometer
   (UPLC-QToF) (Waters, Manchester, UK) equipped with an electrospray
   ionization (ESI) source operated in the positive and negative
   ionization mode, separately. An ACQUITY UPLC ® CSH C18 column (2.1 mm ×
   100 mm × 1.7 µm, Waters, Newcastle, UK) was employed as a stationary
   phase. The mass spectrometer was operated in MSE mode (DIA, data
   independent analysis) with an m/z range of 50–1200 Da and an
   acquisition time of 0.5 s per scan. MSE analysis was operated at 6 eV
   for low collision energy and a ramp of 20–50 eV for high collision
   energy. See [115]Supplementary Materials and Methods for more
   information on metabolomic analysis.

4.4.1. Data Processing and Selection of Features

   The LC-MS raw data files were processed using Progenesis™ QI v2.4
   software (Non-linear Dynamics, Newcastle, UK), which allows adduct
   selection, peak alignment, deconvolution, and compound annotation via
   MSE experiments. Progenesis QI generates an intensity table of features
   labeled according to their nominal masses and retention times as a
   function of intensity. The SERRF package [[116]37], a QC-based signal
   correction method implemented in R 4.2.3 statistical programming
   language [[117]38], was used to correct any potential signal drift.
   Zeros were then replaced by half of the smallest positive value in the
   original dataset. An interquartile range criterion (lowest 10% by IQR
   rank) was used to filter out features with constant or near-constant
   values. Features with a relative standard deviation (RSD) in the QC
   sample data greater than 20% were filtered out. The filtered data were
   transformed using a logarithmic scale and a Pareto scale.

   The limma method [[118]39], implemented in the MetaboAnalystR 3.0
   [[119]40] package in R, was used to select features with differential
   abundance between the pre-dose and post-dose groups. After the
   empirical Bayes fit and the covariate adjustment applied with the limma
   method, we selected features with an FDR -≤ 0.05. First, the covariates
   or confounding variables were selected according to the literature
   [[120]41] for known characteristics: sex, age, and body mass index
   (BMI). The bioequivalence period variable was also considered a
   covariate. The participant tag variable was considered a blocking
   factor to preserve the relationship between samples before and after
   etodolac administration.

4.4.2. Identification of Metabolites

   Through the low and high energy acquisition facilitated by MSE, we
   obtain data on precursor ions (low energy) and fragments (high energy)
   within a single spectrum. The characterization of molecules is based on
   mass precision (≤5 ppm), fragmentation profile (≤10 ppm), isotopic
   similarity (>70%), and biological significance. For Progenesis PQI data
   compatibility, the in-house software v1.0 ‘SDF2PQI’ was employed to
   improve fragment matching, as it converts files from external SDF-based
   spectral libraries, allowing their use in Progenesis PQI [[121]42].
   MassBank of North America (MoNA) [[122]43], the Human Metabolome
   Database (HMDB) [[123]44], and the LIPID MAPS structure database
   [[124]45] were used as libraries.

4.5. Enrichment Pathway Analysis

   A starting approach to interpreting the metabolite differential set was
   the knowledge-driven pathway enrichment. The package RaMP 2.0 was used
   [[125]46], developed in R. This tool queries three different databases
   to enrich sets: KEGG [[126]47], WikiPathways [[127]48], and Reactome
   [[128]49]. The RaMP analysis employed the following criteria: a minimum
   of one metabolite hit per pathway, pathway size ranging from five to
   150 metabolites, and FDR value < 0.05. The output comprised a
   collection of enriched pathways, along with their respective sources,
   unique identifiers, statistical significance values, and matched and
   complete metabolites.

4.6. Community Network Analysis

   Metabolic processes are interdependent. The scope of this association
   may encompass causal, biological, and chemical connections [[129]50].
   In this study, we evaluated the degree of association between
   individual profiling metabolites and PK parameters or patient
   characteristics observed after single-dose ETO administration,
   utilizing the distance correlation metric as a measure of dependence
   [[130]51]. We calculate the linear distance correlation (d) between
   variables, using Pearson’s correlation—to determine the sign or
   directionality of the distance correlation for each pair—but also a
   distance-based non-linear correlation with values between 0 and 1.
   Subsequent analyses were restricted to those comparisons exhibiting a
   d-value > 0.5 and an FDR p-value < 0.05.

   A similarity matrix based on these correlations was used to construct a
   tripartite network in which metabolites (previously determined,
   selecting the differential variables by the limma method when comparing
   T0 and Tmax) were the links between PK parameters and patient
   characteristics (as listed in detail in the [131]Table S1). Correlation
   values were used as weighted edges: metabolites, PK parameters, and
   patient characteristics as network nodes. Community structures within
   our graph were identified using the Leiden algorithm. The structure of
   the network and its constituent modules, or communities, were
   determined with the support of the igraph package [[132]52].

Acknowledgments