Systematic review of prognostic value of cardiorespiratory fitness in patients with cirrhosis

Abstract

Background

Cirrhosis is associated with significant risk of comorbidity and early mortality. Low physical function is common in patients with cirrhosis and could predict prognosis. Cardiorespiratory fitness (CRF), determined by maximal exercise with gas exchange measurement, has proven to predict the risk of mortality and disability in other chronic diseases. In patients with cirrhosis, it could help to inform prognostic stratification to improve care and management in clinical practice.

This systematic review aims to determine the association between CRF (VO2Peak, percentage of predicted VO2Peak, Anaerobic Threshold (AT)) and mortality prediction, as well as morbidity prediction in cirrhosis.

Methods

We reviewed the main electronic databases (PubMed, Scopus, Embase, Google Scholar) for all relevant literature running up to April 2024. Two independent researchers applied predefined inclusion criteria to assess articles for eligibility, and ultimately included 15 studies (seven studied mortality alone, six morbidity alone and two both).

Results

Eight out of nine studies reported CRF variables as a predictor of mortality, and eight studies found that CRF predicted occurrence of events associated with morbidity (sepsis, length of hospital stay and length of stay in critical care). VO2Peak below 17 mL.min-1.kg-1 (5 METS) is an independent predictor of mortality and morbidity. AT can be reached by the large majority of patients with cirrhosis after a moderate-intensity physical exercise, lasts 15 minutes, can be repeated and could be the best choice. AT inferior to 9 ml.min-1.kg-1 or 2.5 METS (metabolic-equivalent tasks defined as the amount of oxygen consumed while sitting at rest) can predict mortality risk and 10 ml.min-1.kg-1 (3 METS) the sepsis risk.

Conclusions

VO2Peak below 17 mL.min-1.kg-1 (5 METS) or an AT under 9 mL.min-1.kg-1 (2.5METS), consistently indicate worse outcomes. Assessing CRF could help to improve mortality and morbidity prediction and AT seems to be the best tool to predict the prognosis in patients with cirrhosis.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12876-026-04715-7.

Background

Cirrhosis is an end-stage liver disease characterized by fibrosis and regenerating nodules of hepatocytes and disorganized hepatic architecture resulting from prolonged liver inflammation, leading to diffuse hepatic fibrosis [1]. Cirrhosis affects more than 122 million individuals worldwide, and the incidence has increased in the last 30 years [2]. Furthermore, patients with cirrhosis are at high risk for developing hepatocellular carcinoma [3] and harbors a poor prognosis, with multiple comorbidities and complications [4]. To date, the only effective treatment for cirrhosis is liver transplant (LT), and therapeutic strategies currently aim to manage causes and complications. Given the prevailing context of graft shortage [5], selection for transplantation hinges on estimations of patient risk (morbidity and mortality), for which clinicians critically need good prognostic markers.

A cardiopulmonary exercise testing (CPET) involves incremental exercise to volitional exhaustion, typically lasting between 6 and 15 min, realized commonly on a cycle ergometer or a treadmill. The continuous measurement of respiratory gas exchange (oxygen uptake and carbon dioxide production) with a breathing mask during the test allows for the determination of cardiorespiratory fitness (CRF). CRF provides an integrated measure of multi-organ system function, as exercise requires coordination between multiple systems to efficiently provide oxygen and substrates to skeletal muscle [6]. Indeed, the main physiological systems contributing to cardiorespiratory fitness include the pulmonary system (ventilation and alveolar gas exchange), the cardiac system (cardiac output and ejection fraction), the hematologic system (hemoglobin concentration and oxygen delivery), the vascular system (vasoregulation and oxygen diffusion), and the muscular/mitochondrial system (oxygen extraction and ATP production) [6]. A recent meta-analysis demonstrated that patients with cirrhosis exhibit markedly reduced cardiorespiratory fitness, which further deteriorates as disease severity increases [7]. All of the physiological systems mentioned above are impaired in cirrhosis, as reflected by conditions such as hepatopulmonary syndrome, cirrhotic cardiomyopathy, anemia, and reduced muscle mass and function [8]. Some CRF variables, such as, ventilatory anaerobic threshold (AT) (oxygen uptake when minute ventilation increases more than oxygen uptake) or peak oxygen uptake (VO_2Peak) (oxygen uptake at the maximal exercise intensity) are widely used to assess mortality and morbidity in patients with chronic diseases such as heart failure [9] or cancer [10]. Moreover, Ney et al. demonstrated that CRF is an objective and independent predictor of pre-and post-transplant mortality in patients with cirrhosis [11]. A recent study reported that liver transplant outcome measures could be optimized by using a pre-assessment that includes both cardiac evaluation and frailty score, with CRF potentially serving as the critical link between these two factors [12].

While knowing that liver transplantation is the only curative option in patients with cirrhosis, the ability to predict mortality and morbidity in the waiting list and after the surgery remains crucial. The primary aim of this systematic review was to determine whether CRF can predict mortality and most represented morbidity-related factors (sepsis, length of hospital stay and length of stay in critical care) in patients with cirrhosis. Secondary objective is to review the different cutoffs that have been used to predict morbidity and mortality.

Patients and methods

This systematic review and meta-analysis was performed in accordance with the PRISMA statement guidelines [13] (Table S1 and S2, Supplementary File), and the protocol was prospectively registered in the International Prospective Register of Systematic Reviews database (PROSPERO registration number: CRD42023414651).

Literature search strategy

The PubMed-Medline, Scopus, Embase and Google Scholar electronic bibliographic databases were initially searched in March 2025 then completed with a final update search conducted in August 2025. Online searches for eligible studies included a combination of medical subject headings (MeSH terms) and text keywords (title and abstract) as search terms, with adaptations made for use in each database.

Search terms used for this literature review are available in Supplementary Files (Table S3). The search was followed by careful selection of eligible studies according to the criteria outlined below. Reference lists from relevant narrative reviews and eligible publications were also screened to identify any additional relevant publications.

Inclusion and exclusion criteria

Human studies written in English or French were included. Conference abstracts were considered but subjected to the same quality assessment procedures as full-text articles.

Inclusion criteria were: studies with patients suffering from cirrhosis regardless of age, gender, or clinical status, CRF assessed with maximal CPET and gas exchange measurement during the test then studies must report an analysis of the association with mortality or morbidity. Respiratory gas exchange (with breathing mask) during exercise should be based on measures of oxygen uptake (VO₂), carbon dioxide production (VCO₂). These measures should be performed using minute ventilation and rapid-response gas analyzers that determine the concentrations of oxygen (O₂) and carbon dioxide (CO₂) in inspiratory and exhalatory air on a breath-by-breath basis.

Exclusion criteria were: patients with hepatocellular carcinoma or other cancers, aerobic capacity assessed through submaximal test (for example, 6-minute walking distance test) without gas exchange measurement.

The search results were exported to a Microsoft Excel spreadsheet, and duplicates were removed. A first selection screen was performed based on study titles and abstracts, and then the remaining full-texts articles were screened to assess eligibility based on our pre-established inclusion and exclusion criteria. The selection process was conducted independently by two reviewers (AC and GE), and any divergences were discussed together between the two reviewers to reach a consensus. A third reviewer was consulted when necessary (FR). Included studies were then synthesized based on study design and outcome, i.e. all studies examining CRF as a predictor for mortality and morbidity.

Data extraction and synthesis

The lead author extracted the quantitative data, then a second member of the research team checked the extracted data for accuracy. The data extracted was as follows: (1) citation details; (2) study design; (3) population characteristics; (4) etiology of cirrhosis; (5) Child-Pugh (CP) score and groups (A, B and C) and MELD score; (6) Outcome used to measure CRF. When key study data was missing from the included studies, the corresponding authors were contacted via email to ask for the raw data and additional information. We made at least three attempts contact authors via email, followed by a subsequent attempt via ResearchGate. When necessary, WebPlotDigitizer was employed to extract data from graphs [14]. Oxygen uptake were converted in metabolic-equivalent tasks (METS) (defined as the amount of oxygen consumed while sitting at rest and is equal to 3.5 ml O2 per kg body weight/min).

Quality assessment

The quality of the included studies was assessed independently by two reviewers (AC and GE) using the ‘Quality Assessment Tool for Quantitative Studies’ developed by the Effective Public Health Practice Project [15]. The assessment encompassed six criteria: selection bias, study design, confounding factors, blinding, data collection methodology, and withdrawals/dropouts. Each criterion was classified as either ‘strong’, ‘moderate’, or ‘weak’. Overall methodological quality was then rated as strong if there were no ‘weak’ criteria, moderate if there was one ‘weak’ criterion, and weak if there were two or more ‘weak’ criteria [16]. Any divergencies were discussed between the two reviewers (AC and GE) to reach a consensus, and a third reviewer (FR) was consulted when necessary. This quality assessment process was presented individually for each study, and then represented across the whole corpus using a bar plot summarizing (in %) each criterion and overall quality.

Results

We identified 4094 records through database searching, of which 184 were eligible for assessment. Of these 184 eligible records, 123 were excluded, leaving 61 studies who did CRF evaluation and ultimately 15 studies (seven analyzed mortality alone, six morbidity alone and two with both analyses) with prognosis analysis included in this systematic review (Fig. 1). Overall quality of the studies was deemed to be good, with 10 studies identified as having low risk of bias [17–26] and 3 having moderate risk of bias [27–29]. Note that one study had high risk of bias [30] and one study could not be classified [31] (Fig. S1).

Fig. 1.

PRISMA flowchart for selection process

Mortality prediction

Nine studies [17–19, 21, 22, 24, 25, 27, 31], comprising 1,351 patients, measured the ability of CRF to predict mortality in multivariate analysis based on oxygen uptake values (VO₂) (Table 1).

Table 1.

Studies with CRF assessment for mortality prediction

Studies	Survival analysis	CRF variables	Multivariate analysis	Parameters controlled	Cuts-offs
Bernal 2014 [17]	Non-transplant (1 year) Transplant (1year)	AT AT	HR: 0.90 [0.83–0.93]* HR: 0.88 [0.69–0.98]*	Age, MELD, Sodium, Albumin, Ascites Donor risk index, Listing to LT	NA NA
Dharancy 2008 [18]	Transplant (1year)	VO2Peak	Regression coefficient: 2 [0-4.1]*	Non defined	< 60% of predicted VO2Peak
Epstein 2004 [27]	Transplant (100 days)	VO2Peak and AT	OR: 14.1 *	CP, MELD, time to LT	< 60% of predicted VO2Peak and < 50% for AT
Faustini-Peireira 2016 [19]	Non-transplant (3 years)	VO2Peak	RR: 4.1 [2.1–6.1]* HR: 0.70 [0.55–0.86]*	Age, CP, MELD, MEP, MIP, 6MWT	17 ml.min− 1.kg− 1 (5 METS)
Galant 2013 [31]	Transplant (3 years)	VO2Peak	OR: 3.29 [1.44–5.25]*	Non defined	14 ml.min− 1.kg− 1 (4 METS)
Moody 2021 [21]	Transplant (5 years)	VO2Peak	HR: 2.73 [0.60-12.54] ns	Age, sex, diabetes, smoking, VO2Peak and abnormal perfusion	14 ml.min− 1.kg− 1 (4 METS)
Neviere 2014 [22]	Transplant (1 year)	VO2Peak	Mortality HR: 0.94 [0.893–0.997]	Age, MELD, O2 pulse, VE/VO2	NA
Ow 2014 [24]	Non-transplant (90 days)	VO2Peak	Mortality OR: 0.75 [0.59–0.95]*	Age, MELD and UKELD score, AT, and VO2Peak	17.6 ml.min− 1.kg− 1 (5 METS)
Prentis 2012 [25]	Transplant (90 days)	AT	OR: 3.84 [1.17–12.58]***	Donor age, blood transfusion, FFP transfusion	9.0 ml.min− 1.kg− 1 (2.5 METS)

Data are presented as Mean [95%CI]

ns non-significant, MELD Model for End-Stage Liver Disease, LT Liver Transplantation, CP Child-Pugh, MEP Maximal Expiratory Pressure, MIP Maximal Inspiratory Pressure, 6MWT 6-Minute Walk Test, VO_₂peak Peak Oxygen Uptake, VE/VO₂ Ventilatory Equivalent for Oxygen, UKELD United Kingdom Model for End-Stage Liver Disease, AT Anaerobic Threshold, FFP Fresh Frozen Plasma

*,**,*** for p < 0.05,p < 0.01 and p < 0.001, respectively

Eight out of these 9 studies, comprising 1,193 patients, reported CRF as independent predictor of all-cause mortality. The sole exception was Moody et al. study [21], which included 158 patients and did not find a statistically significant association between CRF and all-cause mortality. Several studies [17, 18, 21, 22, 25, 27, 31] were interested in transplant-related mortality or not specifically related [19, 24]. Follow-up duration was very heterogenous, varying between 90 and 100 days [24, 25, 27] with 283 patients, 1 year [17, 18, 22] with 797 patients, 3 years [19, 31] with 113 patients or 5 years [21] with 158 patients. Four studies regrouping 5 arms reported CRF levels between survivors and non-survivors [17, 22, 24, 25]. Table 2 reported values of these 2 groups.

Table 2.

Comparison of CRF between survivors and non-survivors

Studies	Sample characteristics	Vo2Peak Survivors	Vo2Peak Non-Survivors
Bernal 2014 [17] Transplant	Survivors: n = 21227% female, 51.7 ± 16.1years, MELD: 14 ± 5.1, 27% ascites, 27% alcohol, 36% viral, 36% other	Vo2Peak 16.2 ± 4.3	Vo2Peak 16.7 ± 5.4
	Non-Survivors: n = 11 33% female, 55.3 ± 9.0years, MELD: 14.3 ± 5.2, 35% ascites, 28% alcohol, 31% viral, 41% other	AT: 11.7 ± 3.7	AT: 10.0 ± 3.3*
Bernal 2014 [17] Non-Transplant	Survivors: n = 10038% female, 54.3 ± 11.3years, MELD: 13 ± 4.5, 23% ascites, 32% alcohol, 33% viral, 35% other	Vo2Peak: 17.6 ± 5.1	Vo2Peak 14.0 ± 4.7***
	Non-Survivors: n = 53 28% female, 56.7 ± 12.2years, MELD: 17.3 ± 5.3, 58% ascites, 38% alcohol, 30% viral, 32% other	AT: 10.3 ± 3.2	AT: 9.3 ± 3.0**
Neviere 2014 [22]	Survivors: n = 24333% female, 58.5 ± 8.6years, MELD: 14.9 ± 2.0, 70% alcohol, 19% viral,	Vo2Peak 18.6 ± 2.8	Vo2Peak 17.1 ± 3.3*
	Non-Survivors: n = 20 25% female, 60.7 ± 8.2years, MELD: 14.4 ± 2.0, 75% alcohol, 20% viral, 24.8 ± 5.5 kg/m²	AT: 12.4 ± 3.5	AT: 11.9 ± 2.7
Ow 2014 [24]	Survivors: n = 15433.5% female, 53.6 ± 10.5years, MELD: 13.6 ± 4.1, 31.8% alcohol, 29.9% viral, 20.8% autoimmune/biliary, 17.5% other	Vo2Peak 21.2 ± 5.3	Vo2Peak 15.2 ± 3.3***
	Non-Survivors: n = 10 50% female, 61.3 ± 6.9years, MELD: 17.9 ± 3.1, 40% alcohol, 10% viral, 40% autoimmune/biliary, 10%	AT: 12.5 ± 3.2	AT: 10.4 ± 2.5*
Prentis 2012 [25]	Survivors: n = 5453.8 ± 10.4years, MELD: 15.7 ± 6.0, 38.9% alcohol, 9.3% viral, 18.5% HCC, 22.2% cholangitis + biliary, 11.1% other, 26.2 ± 5.5 kg/m²	Vo2Peak 14.8 ± 4.0	Vo2Peak 12.5 ± 2.4
	Non-Survivors: n = 6 49.2 ± 12.5years, MELD: 17.8 ± 10.1, 50% primary biliary, 16.7% viral, 16.7% cholangitis, 16.7% autoimmune, 26.7 ± 6.9 kg/m²	AT: 12.0 ± 2.4	AT: 8.4 ± 1.3***

Data are presented as Mean±Standard Deviation

MELD Model for End-Stage Liver Disease, Vo_2Peak Peak Oxygen Uptake, AT Anaerobic Threshold

*,** and ***: p < 0.05, p < 0.01 and p < 0.001 between survivors and non-survivors respectively

No studies reported proportion of decompensated cirrhosis. However, studies reported proportions of ascites (51.9% in Dharancy et al., 2008 [18], 35% in Bernal et al., 2014 [17], 50% of non-survivors and 1.3% of survivors in Ow et al. [24]. Ascites did not differ between survivors and non-survivors in Neviere et al., 2014 [22].

Morbidity prediction

Eight studies [17, 20, 23, 25, 26, 28–30] comprising 1,044 patients assessed the value of CRF as a predictor of morbidity (sepsis, length of hospital stay and length of stay in critical care) in multivariate analysis depending on oxygen uptake values (VO₂) (Table 3).

Table 3.

Studies with morbidity prediction according to CRF

Studies	Morbidity analysis	CRF variables	Multivariate analysis	Parameters controlled	Cuts-offs
Neviere 2016 [23]	Sepsis risk (2 weeks)	VO2Peak	β : -0.214 (SE: 0.09)* OR : 0.807 [0.677–0.963]	Age, MELD, Hemoglobin, KCO (transfer coefficient for carbon monoxide); FeNO (fraction of exhaled nitric oxide)	17 ml.min− 1.kg− 1 (5 METS)
Wallen 2019 [30]	Sepsis risk (from the time of transplant listing until death, delisting, or LT)	AT	OR : 0.65 [0.471–0.965]*	Age, MELD-Na, HCC, ascites, β-blocker	10 ml.min− 1.kg− 1 (3 METS)
Bernal 2014 [17]	Hospital length of stay	AT	AT < 9.2: 21days (IQR: 14–30) AT > 9.2: 15days (IQR: 13–23)*	NA	AT: <9.2 ml.min− 1.kg− 1 (3 METS)
Kimber 2022 [28]	Hospital length of stay	VO2Peak AT	Multivariate analysis: HR: 1.20 [1.03–1.41]* HR: 1.55 [1.10–2.17]* No significant with more adjustments (CHILD, MELD and etiologies)	Age and gender	NA
Mancuzo 2015 [29]	Hospital length of stay	VO2Peak	Multivariate analysis: β-0.20 [-0.35 ; -0.04]*	AT, MELD, CHILD	< 20.0 ml.min− 1.kg− 1 (6 METS)
Wallen 2017 [26]	Hospital length of stay	VO2Peak	Univariate analysis: r=-0.21 Not remain in multivariate analysis	NA	NA
Bernal 2014 [17]	Intensive care unit length of stay	VO2Peak	VO2Peak <13.4: 4days [3–8] VO2Peak >13.4: 3days [2–5]**	VO2Peak	13.4 ml.min− 1.kg− 1 (4 METS)
Kimber 2022 [28]	Intensive care unit length of stay	VO2Peak	Multivariate Analysis: HR: 1.83 [1.06–3.16]*	Age, gender, MELD, CHILD, etiology, Charlson Comorbidity Index	NA
Miarka 2021 [20]	Intensive care unit length of stay	VO2Peak	Association between VO2Peak and ICU: -0.217 (non significant) Patients unable to perform CPET: longer stay**	NA	NA
Prentis 2012 [25]	Intensive care unit length of stay	AT	Significant predictor: AT < 11: 8.1 ± 10.1days *** AT > 11: 2.8 ± 2.9days	NA	11 ml.min− 1.kg− 1 (3 METS)
Wallen 2017 [26]	Intensive care unit length of stay	VO2Peak	Non associated	NA	NA

CRF Cardiorespiratory fitness, VO_₂peakPeak Oxygen Uptake, β Beta Coefficient (statistical regression coefficient), SE Standard Error, OR Odds Ratio, MELD Model for End-Stage Liver Disease, KCO Transfer Coefficient for Carbon Monoxide, FeNO Fraction of Exhaled Nitric Oxide, AT Anaerobic Threshold, MELD-Na MELD-sodium, HCC Hepatocellular Carcinoma, IQR Interquartile Range, HR Hazard Ratio, CHILD Child-Pugh Score, r Correlation Coefficient, ICU Intensive Care Unit, CPET Cardiopulmonary Exercise Testing

*, **, *** for p < 0.05, p < 0.01 and p < 0.001, respectively

Only Wallen et al. [26] failed to find VO_2Peak as an independent predictor of morbidity. Two studies [23, 30], comprising 160 patients, found that CRF was a predictive factor for development of sepsis and one didn’t run multivariate analysis [28]. AT is significantly inferior in patients who developed sepsis (median of 9.5 ml.min^− 1.kg^− 1 (IQR: 7.8–11.9)) compared with patients who did not (median of 11.8 ml.min^− 1.kg^− 1 (IQR: 10.5–13.8)) [30]. VO_2Peak was also significantly reduced with a 17 ± 4 ml.min^− 1.kg^− 1 value for patients with sepsis compared with 21 ± 4 ml.min^− 1.kg^− 1 for patients without (1 MET difference) [23]. Three studies [17, 28, 29], comprising 466 patients, found that CRF was a predictor of length of hospital stay and one did not find a relation in multivariate analysis [26]. Three studies [17, 25, 28], comprising 479 patients, demonstrated a link between CRF and length of stay in critical care while Wallen et al. did not find this link [26]. However, Miarka et al. [20], in a study comprising 98 patients, found that inability to perform CPET was associated with longer time spent in critical care.

Determination of cut-off values to predict mortality

Some studies have proposed cut-off values for CRF to predict mortality. Galant et al. [31] determined a VO_2Peak cutoff value at 14 mL.min^− 1.kg^− 1 to predict survival, whereas Moody et al. [24] found that this value was not a significant predictor of mortality. Faustini-Pereira et al. [19] proposed a cut-off value of 17 mL.min^− 1.kg^− 1 (5 METS) to predict survival, and Ow et al. [24] found a close value of 17.6 mL.min^− 1.kg^− 1. Concerning anaerobic threshold (AT), Prentis et al. [25] found an oxygen consumption lower than 9 mL.min^− 1.kg^− 1 (2.5 METS) associated with increased mortality while Bernal et al. [16] found that AT was independently associated with non-survival at a cut-off of 8.5 mL.min^− 1.kg^− 1 in both transplant patients (Hazard Ratio (HR): 0.88, p < 0.05) and non-transplant patients (HR: 0.91, p = 0.02).

Age can be adjusted for by using the percentage of predicted VO_2Peak. Dharancy et al. [18] showed that a cut-off at 60% was a predictive factor for 1-year survival in 135 patients. Epstein [27] did not find significant differences between survivors and non-survivors at this 60% cut-off, but the combination of VO_2Peak under 60% of predicted VO_2Peak and under 50% for AT cut-off was significant for 100 days survival [27].

Only 3 studies reported AUROC curve with specificity and sensibility data of used cut-offs [19, 24, 25]. Ow and Prentis did analysis at 90 days with VO_2Peak and AT, respectively. They found high degree of accuracy with AUROC > 0.8 with excellent sensitivity (> 0.9) and good specificity (> 0.74). Faustini-Pereira et al., studied VO_2Peak and followed their patients for 3 years follow-up [19]. They found a good accuracy (AUROC = 0.78) with a high sensitivity (0.87) and a low specifity (0.41).

Determination of cut-off values to predict morbidity

Nevière et al. [23] identified a 17 mL.min^− 1.kg^− 1 (5 METS) VO_2Peak cut-off as optimal to predict post-operative sepsis. Concerning AT cut-off values Wallen et al. [30] used AUROC curve analysis to predict sepsis and identified the optimal cutoff as 10 mL.min^− 1.kg^− 1 (± 3 METS). Prentis et al. [25] determined that oxygen consumption lower than 11 mL.min^− 1.kg^− 1 at AT was associated with increased risk of longer hospital and critical care length of stay.

Concerning sepsis prediction, Neviere and Wallen [23, 30] showed good accuracy (AUROC > 0.73) while Wallen et al. found a low sensitivity (0.69) and a good specificity (0.81).

Discussion

The purpose of this study was to conduct a systematic review to ascertain the clinical relevance of CRF in individuals with cirrhosis. As expected, CRF variables can predict mortality and morbidity in patients with cirrhosis but surprisingly not at 5 years.

Eight studies found CRF as independent variable to predict mortality in patients with cirrhosis at short term (90–100 days) [24, 25, 27], mid-term (1 year) [17, 18, 22] or long term (3 years) [19, 31] but one study didn’t show CRF as predictor in very long term (5 years) [21]. Used cutoffs for mortality prediction were 17 mL.min^− 1.kg^− 1 (5 METS) for VO_2Peak [19, 24, 31], 9 mL.min^− 1.kg^− 1 (2.5 METS) for AT [25] and 60% for Predicted VO_2Peak [18, 27]. Eight studies found CRF as independent predictor of morbidity (hospital and critical care length of stay and sepsis). Concerning sepsis risk, a VO_2Peak under to 17 mL.min^− 1.kg^− 1 [17, 23, 29] and 10 mL.min^− 1.kg^− 1 for AT increase the risk [17, 25, 30].

The gold-standard of endurance capacity is CRF, because it involves multi-organ system function (pulmonary, cardiac, hematologic, vascular, mitochondrial and muscular) [6]. Indeed, in general population, a meta-analysis of over 2 million individuals demonstrated that each 3.5 ml·min⁻¹·kg⁻¹ improvement in VO_2Peak (1MET) was associated with an 11% reduction in mortality, irrespective of sex [32]. Likewise, a longitudinal study assessing two exercise tests over an 11-year period reported that each 3.5 ml·min⁻¹·kg⁻¹ (1MET) increase in VO_2Peak was associated with a 29% reduction in mortality [33]. Moreover, according to a study conducted in the UK Biobank involving over seventy thousand participants, it was demonstrated that cardiorespiratory fitness (CRF) can predict mortality independently of muscle strength [34]. Between the lowest and highest tertiles, a 35% reduction in all-cause mortality and a 51% reduction in cardiovascular mortality was observed. Several meta-analyses or systematic review have shown that CRF can predict mortality and/or morbidity in diverse diseases such as cancer [35], cardiovascular disease [36], and pulmonary hypertension [37], and on liver cancer surgery [38]. These results are in accordance with our systematic review findings. Low physical function [39] and its decrease [40] can improve the prediction of mortality in cirrhosis independently of MELD score and could have a bigger influence in mortality than muscle mass [39, 41].

Note that Moody et al. [24] did not find that VO_2Peak was a predictor of mortality. This could be explained by the long duration between CRF measurement and mortality (5 years), which was the longest of all studies included in this systematic review.

In liver cancer surgery, AT (5 min of CPET) is sufficient to predict recovery after liver cancer surgery, particularly in patients with an AT < 10.5 mL.min^− 1.kg^− 1 [38]. In patients with hepatocellular carcinoma (HCC), maintenance of CRF following transplantation was associated with improved overall and recurrence-free survival [42]. The same group found an anaerobic threshold cutoff of 11.5 ml·min⁻¹·kg⁻¹ as an independent predictor of event-free survival [43]. The suggested cutoff values (9-11 ml·min⁻¹·kg⁻¹ or 2.5/3 METS) in these studies and in reviews about liver cancer surgery [38] or cardiovascular diseases [44] are consistent with our systematic review findings for AT. These results are useful when VO2Peak is not reachable by the patient due to exercise intolerance, as AT is easier to achieve.

Limitations

This systematic review has potential limitations that may affect the generalizability of our findings. A significant number of studies did not provide sufficient details on the diagnosis of cirrhosis, the use of elastography should improve the quality of the diagnosis in the future, at least in non-decompensated patients, especially those without ascites. Furthermore, to take account of sex-based differences, it was proposed few years ago a sex-based threshold of VO_2Peak below which a loss of independence is likely, it was 18 mL.min^− 1.kg^− 1 (5 METS) for men and 15 mL.min^− 1.kg^− 1 (4 METS) for women [45]. Further work is needed to refine the different thresholds to be used and specially to take the gender into account.

Morbidity outcomes in this systematic review didn’t represent overall morbidity and present some limits. Indeed, hospital length of stay is a questionable criteria as good prognosis marker due its dependance on numerous factors (difference between hospitals, countries, physicians).

The impossibility to do a meta-analysis, due to heterogeneity of follow-up time and different statistical methods (as odds ratios or hazard ratios or relative risks), and small sample size limit the reliability of our findings. Moreover, differences in liver function stage or etiologies could represent potential sources of heterogeneity and reduces our possibility to perform a meta-analysis.

Furthermore, a possible practical difficulty linked to cutoff relative to body-weight is the presence of water retention which can modify the real value of the weight of the patient [46]. EASL guidelines suggest subtracting a percentage of weight based upon the severity of ascites (mild 5%; moderate 10%; severe 15%) [47].

Methodological considerations

An important issue in the assessment of physical function to improve prognosis prediction is the time at which the assessment is carried out. For example, in the multicenter FrAILT Study, frailty increment was significantly associated with death/delisting and in the contrary, frailty improvement with a better survival [40]. It was clearly specified that the patients included were outpatients but there was no information on the delay between the last complication and the baseline evaluation.

In patients with HCC, a ≥ 10% decline in AT at 6 months after hepatectomy compared to pre-operative CRF is associated with significantly lower five-year survival rate compared to patients who do not lose cardio-respiratory capacity (9.9% vs. 39.9%) [42]. In this study CRF was evaluated only at 6 months and there is no comparison with other time points.

This is why new studies on the assessment of physical functions must address this gap in the timing of assessment to improve prognostic evaluation. Moreover, evaluating CRF is difficult because it is time-consuming (30–45 min) and deadline for obtaining an appointment could be longer that one or two months in many centers. It might be simpler to move toward simpler physical tests that demonstrate a link with CRF. Additionally, a recent study by Hughes et al. in liver transplant candidates found strong correlations between CRF and 6-Minute Walking Test (6MWT) or Liver Frailty Index (LFI) [48]. To assess physical fitness, many centers quantify myopenia, using skeletal muscle mass at third lumbar vertebra (L3-SMI). However, it is an irradiating examination, with a specific software needed and, in our institution, the estimation of muscle area can only be done by radiologists. It measures muscle mass only and this technique should be supplemented by a functional assessment because muscle mass is not correlated to LFI in patients with cirrhosis [49]. Thus, it appears that CRF and physical frailty are not interchangeable with muscle mass evaluation. Actually, sarcopenia corresponds to low muscle strength and muscle mass [50] while physical frailty is defined as reduced endurance and strength [51], which can explain differences. To date, the assessment of cardiorespiratory fitness and physical frailty is recommended for patients with cirrhosis who are awaiting transplantation [52]. However, it should not, at this stage, be considered an eligibility criterion, particularly because the patient’s condition can deteriorate rapidly.

Perspectives

Finally, it could be much easier to select one of the three parameters used to evaluate CRF (AT, VO_2Peak or % of predicted VO_2Peak). We think that AT could be the best choice because this threshold can be reached by the large majority of patients with cirrhosis. However, this parameter should be compared to others well-known predictive factors such as Liver Frailty Index, 6MWT and L3-SMI to predict the prognosis. Ideally, the best predictor should be repeated over time. All worldwide guidelines suggest a physical evaluation of liver disease patients but tools need to be validated.

Physical function is a dynamic phenomenon, a recent meta-analysis showed that exercise in patients with cirrhosis improves CRF before or after transplantation [7]. Furthermore, prehabilitation has shown to be feasible and improves survival in patients in transplant candidates [53]. However, effects of liver transplantation are not consensual and need to be studied in prospective studies. In bedridden patients with anemia, blood transfusion could improve CRF [54] and this possibility should be studied in cirrhosis.

Conclusion

This systematic review highlights the prognostic value of CRF in patients with cirrhosis. The evidence strongly supports that lower CRF is associated with increased risk of mortality and morbidity, including sepsis and longer hospital or ICU stay. VO_2Peak below 17 mL.min^− 1.kg^− 1 (5 METS) or an AT under 9 mL.min^− 1.kg^− 1 (2.5METS), consistently indicate worse outcomes.

CRF measurement via CPET is not yet widely implemented in routine practice due to feasibility constraints, it offers substantial potential for improving risk stratification in this vulnerable population. Future studies should focus on refining cutoff values, considering sex-specific thresholds, and identifying simpler surrogate assessments for CRF in clinical settings. 6MWT and LFI have shown a correlation with CRF and prognosis ability and their clinical applicability, they could be considered as reliable surrogate of CRF. However, their response to exercise and prehabilitation need to be studied.

Abbreviations

6MWT

6-Minute Walk Test

AT

Anaerobic threshold

AUROC

Area under the receiver operating characteristic curve

CO₂

Carbon dioxide

CP

Child-pugh score

CPET

Cardiopulmonary exercise testing

CRF

Cardiorespiratory fitness

HCC

Hepatocellular carcinoma

HR

Hazard ratio

IQR

Interquartile range

LT

Liver transplantation / liver transplant

MELD

Model for End-Stage Liver Disease

METS

Metabolic equivalent of task

MeSH

Medical subject headings

O₂

Oxygen

OR

Odds ratio

PRISMA

Preferred reporting items for systematic reviews and meta-analyses

PROSPERO

International Prospective Register of Systematic Reviews

VE/VO₂

Ventilatory equivalent for oxygen

VO₂

Oxygen uptake

VO₂Peak

Peak oxygen uptake

Authors’ contributions

Couret A, Rannou F, Pereira B, Duclos M, King JA, Dharancy S, Nevière R, Weil-Verhoeven D were responsible for writing original draft, writing review, editing, data curation, and visualization; Ennequin G & Abergel A were responsible for conceptualization, methodology, software, formal analysis, resources, data curation, writing original draft, writing review, editing, project administration, and supervision; all of the authors read and approved the final version of the manuscript to be published

Funding

No funds, grants, or other support was received.

Data availability

All data are available in articles from this systematic review or by asking authors.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.