Table 1.

Observational studies included in this review.

Author (year)	Primary aim	Metabolomic profiling platform	Biosample	Study population	Assessment of PA intensity/duration or fitness level	Sample collection timepoint	Primary findings
Kujala et al. (2013)	To explore whether persistent PA has a global effect on the serum metabolome that reduces cardiometabolic disease risk.	NMR	Serum.	Discovery: 16 same-sex twin pairs discordant for PA; Men and Women; 50–74 years; Follow-up: 1037 age- and sex-matched pairs that were either persistently active or persistently inactive; Men and Women; 41–62 years.	Validated questionnaires and MET index.	One sample concurrent with estimate of active/inactive status.	Persistently physically active individuals have a coherently healthier circulating metabolite profile than their inactive counterparts. Larger differences were noted with more discordant activity profiles and with increased length of discordant activity.
Morris et al. (2013)	To explore differences between adults by fitness level.	GC-MS	Urine and plasma.	65 Men and Women; 18–60 years.	Cycle Ergometer based fitness test to assess maximal oxygen consumption levels (mL/kg/min).	One sample concurrent to fitness assessment.	Metabolomic profiles differed between fitness groups, specifically with respect to amino acids. Higher levels of branched chain amino acids were associated with lower fitness levels and higher insulin resistance. Most pronounced in females.
Floegal et al. (2014)	To investigate associations between diet, PA, cardiorespiratory fitness and obesity with serum metabolite networks.	MS	Serum.	100 Men and Women; 35–65 years.	Combined heart rate and movement sensor.	Two samples 4 months apart.	Cardiorespiratory fitness was more strongly correlated with metabolic networks than measures of physical activity.
Fukai et al. (2016)	To investigate relationship between daily PA and metabolomic profile.	MS	Plasma.	1193 Men; 35–74 years.	Self-reported questionnaires.	One sample concurrent to questionnaire.	PA is related to changes to the plasma metabolome, including known biomarkers for future insulin resistance and type 2 diabetes.
Xiao et al. (2016)	To identify a metabolic signature of PA.	LC-MS and GC-MS	Plasma.	277 Men and Women, aged 40–65 +	Accelerometer-measured PA for 7 days on four occasions over 1 year.	Two samples; beginning and end of 1-year study period.	Differing patterns were observed for light versus moderate to vigorous PA. Overall volume of PA was the biggest driver of metabolomic profile.
Barton et al. (2017)	To explore the impact of extreme PA.	NMR, UPLC-MS, GC-MS	Urine and feces.	46 male professional international rugby union players (mean (SD) age 29 (4) years) and 46 controls (mean (SD) age 29 (6)).	Adapted version of the EPIC-Norfolk physical activity questionnaire.	One sample concurrent to questionnaire.	Strong metabolomic differences observed between athletes and controls in urine, and to a lesser extent in fecal samples.
Al-Khelaifi et al. (2018a)	To compare blood metabolic profiles between moderate- and high-power and endurance elite athletes.	LC-MS	Serum.	191 elite athletes (171 males and 20 females) from different sporting disciplines (anonymized data, age not reported).	Categorized according to sports which are classified by level of power, endurance and whether they are dynamic or static.	One sample concurrent to categorization.	High-power and high-endurance athletes exhibit a distinct metabolic profile that reflects steroid biosynthesis, fatty acid metabolism, oxidative stress, and energy-related metabolites.
Al-Khelaifi et al. (2018b)	To analyze the presence of various xenobiotics in serum samples from elite athletes of different sports.	LC-MS	Serum.	478 elite athletes (anonymized data, age not reported).	Classified according to sport; football, athletics, cycling, rugby, swimming, boxing and rowing.	One sample concurrent to categorization.	Athletes from different sports exhibit a distinct xenobiotic profile that may reflect their drug/supplement use, diet and exposure to various chemicals.
Bell et al. (2018)	To determine metabolomic associations with total activity, moderate-to-vigorous PA and sedentary time.	NMR	Blood.	1826 boys and girls; mean (SD) age 15.4 years (0.2 years).	Accelerometry assessed PA average over three days at age ~12 years, ~14 years and ~15 years.	One sample at age 15 years visit.	Higher total activity associates with a range of metabolites, and associations are strongest for moderate-to-vigorous activity. Associations of current activity with most metabolic traits do not differ by previous activity.
Ding et al. (2019)	To identify metabolites associated with habitual PA.	LC-MS	Plasma	5197 Men and women; mean (SD) age ranged from 44.7 (4.5) to 64.0 (8.2) years across the cohorts.	Average of two questionnaires within two years of blood draw.	One sample within two years of both questionnaires.	10 habitual PA associated metabolites were replicated in two large scale cohort studies.
Author (year)	Primary aim	Metabolomic profiling platform	Biosample	Study population	Exercise/physical activity intervention or test	Sample collection timepoint	Primary findings
Sun et al. (2017)	To identify changes in metabolomic pathways in urine after an 800-meter run.	NMR	Urine.	19 athletes, men mean (SD) age 19.2 (0.7) years.	800-meter run.	Two samples; before and 35 min after run.	Results indicated changes in the urinary metabolomic profile after exertion indicating increase in oxidative stress.
Davison et al. (2018)	To examine the role of exercise in hypoxia using metabolomics profiling.	LC-MS	Serum.	24 men mean (SD) age 28 (5).	1-hour exercise in hypoxia and in normoxia.	Three samples; pre-exercise and immediately and 3 h post exercise.	There are small differences in the metabolomic changes associated with exercise and recovery in hypoxia and normoxia, but both affect pathways of purine and pyrimidine metabolism.
Howe et al. (2018)	To investigate the physiological responses to prolonged physical exertion.	LC-MS	Plasma.	9 ultramarathon runners, men mean (SD) age; 34 (7) years.	80.5 km self-paced treadmill-based time trial.	Two samples pre and immediately post exercise.	The findings provide a potential explanation for the cardio-protective effects of ultramarathon running.
Stander et al. (2018)	To characterize the acute metabolic changes induced by a marathon.	GC-MS	Serum.	31 runners; men mean (SD) age 41 (12) years.	Marathon running.	Two samples: pre and post marathon.	Running a marathon places immense strain on the energy-producing pathways of the athlete, leading to extensive protein degradation, oxidative stress and autophagy.
Valerio et al. (2018)	To compare the metabolomic response to high load versus low load resistance exercise.	NMR	Serum.	9 well trained men, mean (SD) age 26.4 (4.4) years.	Resistance exercise: Leg presses at difference loads.	Two samples pre and 5 min post each trial.	Both high and low load resistance training elicited changes in the metabolome relative to controls; and there were additional differences by load.
Manaf et al. (2019)	To identify changes in the plasma-metabolome associated with the onset of fatigue during prolonged cycling.	LC-MS	Plasma.	18 men; aged 18–24 years.	Cycling to exhaustion.	Four samples; 10 min after exercise onset, pre-fatigue, post-fatigue and 20 min post fatigue.	There was a unique metabolomic trajectory associated with time to fatigue and recovery.
Manaf et al. (2019)	To identify changes in the plasma-metabolome associated with the onset of fatigue during prolonged cycling.	LC-MS	Plasma.	18 men; aged 18–24 years.	Cycling to exhaustion.	Four samples; 10 mins after exercise onset, pre-fatigue, post-fatigue and 20 min post fatigue.	There was a unique metabolomic trajectory associated with time to fatigue and recovery.
Pitti et al. (2019)	To compare the saliva metabolome before and after a soccer match.	NMR	Saliva.	17 professional team soccer players, women mean (SD) age across groups ranged from 18 (1) to 23 (5) years.	Soccer/Football match.	Two samples; pre and postgame.	PCA demonstrated a separation in the metabolomic profile of players before and after the game. Changes were driven by amino acids and energy metabolites.
Siopi et al. (2019)	To investigate whether the response of the human serum metabolic fingerprint to exercise depends on exercise mode.	LC-MS	Serum.	23 sedentary men; age not reported.	Four trials: resting, high-intensity interval exercise (HUE), continuous moderate-intensity exercise (CME), and resistance exercise (RE).	Three samples; pre-exercise, immediately post-exercise and 1-hour post-exercise.	The largest changes from baseline were found in the immediate post-exercise samples. RE caused the strongest responses overall, followed by HUE, while CME had minimal effect.
Schader et al. (2020)	To compare training induced metabolic shifts across fitness levels.	LC-MS	Plasma.	76 runners; men mean (SD) age between groups ranged from 32.8 (7.9) to 50.5 (7.9) years.	Marathon running.	Five samples; five weeks and one week before the race and immediately, 24 h, and 72 h after the race.	Prolonged intense exercise is associated with an extensive and prolonged perturbation in plasma metabolite concentrations that is greater in the slower, less aerobically fit runners.

All Individuals were ‘healthy’ at blood draw unless otherwise stated.

GC–MS – Gas Chromatography–Mass Spectrometry.

LC-MS - Liquid Chromatography-Mass Spectrometry.

MET - metabolic equivalent of task.

MS – Mass Spectrometry.

NMR - Nuclear Magnetic Resonance Spectroscopy.

PA – Physical Activity.

SD – Standard Deviation.

UPLC-MS – Ultra Performance Liquid Chromatography- Mass Spectrometry.

BM – Boston Marathon.

CE-TOFMS – Capillary Electrophoresis Time of flight mass spectrometry.

CME – Continuous moderate exercise.

ETT – Exercise Testing Protocol.

HIIE – High intensity interval exercise.

LC/TOF-MS – Liquid chromatography/Time of flight Mass spectrometry.

MS/MS – tandem Mass Spectrometry.

OPLS - Orthogonal partial least squares.

PCA – Principal components analysis.

RE – Resistance exercise.