Psoas muscle index predicts time to rehospitalization in liver cirrhosis: An observational study

Abstract

Sarcopenia is frequent in liver cirrhosis (LC) where it is associated with morbidity and mortality. However, prognostic scores such as model for end-stage liver disease (MELD), MELD-sodium (MELD-Na), or Child–Turcotte–Pugh (CTP) do not contain sarcopenia as a variable. For this study, we utilized psoas muscle index (PMI) to objectively determine sarcopenia in hospitalized LC patients, and evaluated it as a predictor of time between discharge and readmission in LC. Abdominal computed tomography and magnetic resonance imaging scans of 65 consecutive LC patients were retrospectively examined to determine PMI. MELD, MELD-Na, and CTP were calculated from clinical data. PMI was then combined with CTP to form an experimental score: CTP sarcopenia (CTPS). For PMI alone and for each score, correlation with time between discharge and readmission for liver-related complications was calculated. PMI was also tested for correlation with sex, body mass index (BMI), MELD, MELD-Na, and CTP. CTPS was most closely correlated with time to readmission (R = 0.730; P < .001), followed by CTP (R = 0.696; P < .001), MELD-Na (R = 0.405; P = .009), and PMI alone (R = 0.388; P = .01). Correlation with MELD (R = 0.354; P = .05) was lowest. Additionally, there were significant differences in PMI between male and female individuals (5.16 vs 4.54 cm²/m²; P = .04) and in BMI between sarcopenic and nonsarcopenic individuals (29.63 vs 25.88 kg/m²; P = .009). Sarcopenia is an independent short-term prognostic factor in LC. By combining data on sarcopenia with CTP, we created an experimental score that predicts time to readmission better than MELD, MELD-Na, or CTP.

1. Introduction

Sarcopenia is a syndrome defined by progressive loss of skeletal muscle mass and muscle function.^[1,2] It is a common phenomenon in later stages of liver cirrhosis (LC).^[3,4] In these patients, it is closely associated with mortality before and after liver transplantation (LT), as well as with infections and other complications of LC.^[5–7]

Still, sarcopenia is not routinely used as a predictive factor in LC patients. This may be due to the fact that it is difficult to quantify, especially in cirrhosis. Questionnaires like SARC-F (“Strength, Assistance with walking, Rise from a chair, Climb stairs and Falls”) or U-TEST (“UnderweighT, Elderly, Strength, Thinner”) are only validated in geriatrics, orthopedics, or oncology,^[8–10] as are functional measurements like handgrip strength or chair stand-up test.^[11,12] Bioelectrical impedance is easy to measure but difficult to standardize.^[13] Because all of these methods are influenced by encephalopathy, ascites, tremor, or edema, none has been successfully adapted for LC patients so far.^[14,15] Instead, clinical scores are standard for outcome prediction in LC today.

One of the oldest clinical scores is Child–Turcotte–Pugh (CTP). Developed in the 1960s, it was initially meant to predict complications in open bypass surgery for portal hypertension.^[16] Therefore, it includes ordinal variables like ascites volume and hepatic encephalopathy (Table 1). It has since proven its value as a general tool for outcome prediction in LC and is still in use today, even though the surgical procedure has long been abandoned.^[17–19]

Table 1.

Calculation of traditional CTP score and experimental CTPS score.

Parameter	1	Points assigned	3
		2
Hepatic encephalopathy	None	Grade 1–2	Grade 3–4
Ascites	None	Slight to moderate	Severe
Total bilirubin (mg/dL)	<2.0	2–3	> 3.0
Serum albumin (g/dL)	>3.5	2.8–3.5	< 2.8
INR	<1.7	1.7–2.3	> 2.3
Sarcopenia	No		Yes

Range for CTP is 5 to 15 points; range for CTPS is 6 to 18 points. CTP is usually grouped into “Child-A” (5–6 points), “Child-B” (7–9 points), and “Child-C” (10–15 points), which correlates to overall survival. We did not divide CTPS into corresponding groups, because long-term data on survival are not available yet.

CTP = Child–Turcotte–Pugh, CTPS = CTP sarcopenia, INR = international normalized ratio.

In the 1990s, a new score, model of end-stage liver disease (MELD), was created, its main purpose being a fair and objective allocation of donor organs for LT. Because hepatorenal syndrome was a common cause of waiting list mortality then, MELD included serum creatinine.^[20,21]

However, neither CTP nor MELD contains sarcopenia as a variable. This is astonishing because sarcopenia is long known to be an independent risk factor for overall morbidity in LC patients. Clinically, it has been proven to be associated with waiting list mortality as well as with complications after LT.^[22–26]

Therefore, in this study we propose a method of predicting short-term course in LC patients that does include sarcopenia. For this, we utilize an objective technique of muscle mass quantification that is suitable for LC patients: psoas muscle index (PMI), defined as the cross-section area of psoas muscle at the level of the third lumbar vertebra (L3) divided by height squared. Unlike skeletal muscle index (SMI), which takes into account all muscle area at L3, PMI is relatively resistant to abdominal distension and thus suited for patients with and without ascites alike.^[27,28] Additionally, it is more precise than dual-energy X-ray absorptiometry or ultrasound,^[29–31] and it does not depend on additional commercial software, which makes it suitable for clinical use.^[32,33]

The data derived from PMI measurements were then combined with an existing scoring system in order to develop a new score with more predictive power (Table 1).

2. Methods

2.1. Study population

Electronic charts of our center were retrospectively searched for the time span between December 2015 and November 2020. Included were all adult LC patients who, for any reason, underwent computed tomography (CT) or magnetic resonance imaging (MRI) scans involving the L3 area (abdominal pain, suspected malignancy, infection of unknown focus). This returned 108 potential cases. Excluded from these were all patients with concomitant malignant disease, previous bariatric surgery, psychiatric disease, or eating disorders. For practical reasons, we also limited the time period between CT/MRI and laboratory analysis to 6 months. This left 65 patients who were included in the final analysis. For 13 of these, data were collected at 2 points of time, for 1 patient at 3 points of time. For further analysis, we only included 1 data set per patient. Etiology of cirrhosis was alcoholic in 30 patients (46%), viral hepatitis in 3 patients (5%), autoimmune liver disease in 2 patients (3%), and cryptogenic in 30 patients (46%).

2.2. Clinical and laboratory assessments

For each patient, we collected information on sex, age, height, weight, etiology of cirrhosis, presence and degree of ascites, presence and degree of hepatic encephalopathy, serum bilirubin, serum albumin, serum creatinine, international normalized ratio, serum sodium, and, if present, renal replacement therapy. Additionally, histologic findings from liver biopsies were available for 21 patients; in these cases, the modified hepatitis activity index was included.^[34]

Follow-up periods of all patients started with the date of MRI/CT. Endpoint was defined as liver-related readmission (i.e., readmission because of variceal bleeding, hepatic encephalopathy, hepatorenal syndrome, hydropic decompensation, spontaneous bacterial peritonitis, or for LT).

2.3. Scores

CTP and MELD scores were derived from the collected data. CTP for each patient was calculated according to known standards and summed up into 3 categories A, B, and C (Table 1).^[21]

MELD scores were calculated using the formula:

$${\displaystyle \begin{array}{l}\text{MELD}=11.2\times \text{ln}(\text{INR})+9.57\times \text{ln}(\text{creatinine})+3.78\times \text{ln}(\text{bilirubin})+6.43\end{array}}$$ (1)

A lower limit of 1 was set for all variables, creatinine and bilirubin were given in mg/dL. Values for creatinine were set to 4 if either they exceeded 4 mg/dL or if a patient was receiving renal replacement therapy.

Additionally, MELD-Na score, a variation of the MELD, which considers serum sodium levels, was calculated as follows:

$${\displaystyle \begin{array}{l}\text{MELD-Na}=\text{MELD }\text{ − }\text{ sodium }-[0.025\times \text{MELD}\times (140\text{sodium})]+140\end{array}}$$ (2)

Serum sodium was provided in mmol/L; values <125 or >140 were counted as 125 or 140, respectively..^[35]

2.4. Psoas muscle index

PMI was assessed using stored images from CT or MRI scans according to known standards.^[36] In brief, transverse images at L3, close to the intervertebral disc between third and fourth lumbar vertebra, were selected for each scan. Cross-sectional areas of right and left psoas muscle were each identified and manually traced by 1 single examiner (Fig. 1). The size of the areas in cm² was then computed by open source software summing tissue pixels and multiplying by pixel surface area (Horos 3.3.6; horosproject.org). Finally, cross-sectional area of both psoas muscles was normalized to body stature by division by height squared, and expressed in cm²/m². Sarcopenia was defined according to FLEXIT consortium criteria as a PMI ≤ 5.0 cm²/m² in men and < 4.2 cm²/m² in women.^[37]

Figure 1.

Computed tomography images of 2 LC patients taken at L3 level. Psoas muscle area is marked in red. Panel (A) and (B) show a female patient with a psoas area of 15.5 cm² and a relatively high PMI of 6.2 cm²/m². Panel (C) and (D) show a male patient with a psoas area of 8.4 cm² and a relatively low PMI of 2.9 cm²/m². LC = liver cirrhosis, PMI = psoas muscle index.

2.5. a new score: CTP sarcopenia

With these data, a simple experimental score was created by adding 3 points to CTP if the patient did fall into the “sarcopenia” definition according to sex and PMI value, and adding 1 point if he or she did not (“CTP sarcopenia,” or CTPS see Table 1).

$${\displaystyle \begin{array}{l}\text{CTPS = }\text{ CTP+X }\text{for X = 1V3}\end{array}}$$ (3)

2.6. Statistical analysis

Differences in means of continuous parametric variables were analyzed by T test, applying Welch correction if variances were significantly different (P value of F test < .05). Differences in means of nonparametric variables were analyzed by Mann–Whitney U test. Differences in categorical variables were analyzed using Fisher exact test.

Relationship between 2 variables was assessed by Pearson’s correlation coefficient (r) analysis for parametric variables, and Spearman’s rank correlation for nonparametric variables. Differences in frequencies in 1 or more categories were compared by χ² test.

Time span between CT/MRI and readmission for sarcopenic and nonsarcopenic groups was compared by log-rank test. Simple linear regression was used in order to assess the variables best-predicting differences in time to readmission.

Significance was defined at P < .05. Tests were conducted 2-tailed.

Statistical data were computed using Prism 9.0.0 (GraphPad Software, Inc., La Jolla, CA).

All procedures were in accordance with the standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent or substitute was obtained from all patients for the purpose of publication. This study was approved by the ethics committee of Medizinische Hochschule Brandenburg (File No. E-01-20200714).

3. Results

Out of 65 patients, 23 (35%) were female, 42 (65%) were male. Mean age in years (mean ± SD) was 46 ± 13. Height and weight were significantly lower in women than in men, but there was no difference in BMI. Age was significantly higher in female than in male patients. There were no significant sex differences in CTP, MELD, and MELD-Na, or in any of the individual parameters these scores were calculated from (Table 2). Five male patients and 1 female patient died, the difference was not significant.

Table 2.

Demographic, clinical, and biochemical characteristics of patients included.

n = 65	all (n = 65)	female (n = 23; 35.4%)	male (n = 42; 64.6%)	P value
Age (yr)	60.43 ± 12.8	65.87 ± 11.9	57.45 ± 12.42	.01*
Height (m)	1.71 ± 0.09	1.62 ± 0.07	1.76 ± 0.07	<.0001*
Weight (kg)	81.15 ± 20.88	69.74 ± 15.32	87.4 ± 21.02	<.0001*
BMI (kg/m2)	27.76 ± 6.48	26.48 ± 4.96	28.47 ± 7.14	.3
Psoas area (cm2)	14.46 ± 4.32	11.85 ± 2.13	15.89 ± 4.56	<.0001*
PMI (cm2/m2)	4.94 ± 1.3	4.54 ± 0.87	5.16 ± 1.45	.04*
Sarcopenia	30 (46%)	8 (35%)	22 (52%)	.2
Etiology of cirrhosis
Alcoholic	30 (46%)	11 (48%)	19 (45%)
Autoimmune	2 (3%)	2 (9%)	–
Viral hepatitis	3 (5%)	1 (4%)	2 (5%)
Unknown	30 (46%)	9 (39%)	21 (50%)
Albumin (g/dL)	2.97 ± 0.67	2.97 ± 0.6	2.97 ± 0.71	.72
Bilirubin (mg/dL)	6.02 ± 4.02	7.49 ± 4.59	5.12 ± 3.71	.69
Creatinine (mg/dL)	1.28 ± 0.97	1.55 ± 1.45	1.14 ± 0.53	.64
INR	1.56 ± 0.4	1.64 ± 0.42	1.51 ± 0.38	.18
Sodium (mmol/L)	134 ± 6.01	135.2 ± 5.05	133.8 ± 6.47	.51
Ascites
None	24 (37%)	10 (44%)	14 (33%)
Moderate	20 (31%)	7 (30%)	13 (31%)
Massive	21 (32%)	6 (26%)	15 (36%)
Encephalopathy
None	62 (95%)	23 (100%)	39 (93%)
I	–	–	–
II	2 (3%)	–	2 (5%)
III	1 (2%)	–	1 (2%)
IV	–	–	–
Child–Turcotte–Pugh mean	9.25 ± 2.51	9.22 ± 2.45	9.26 ± 2.57	.91
Grade A (5–6)	13 (20%)	4 (17%)	9 (21%)	.0025*
Grade B (7–9)	19 (29%)	8 (35%)	11 (26%)
Grade C (10–15)	33 (51%)	11 (48%)	22 (52%)
MELD	16.36 ± 6.27	17.81 ± 7.54	15.56 ± 5.39	.22
MELD-Na	19.45 ± 6.92	20.1 ± 7.23	19.09 ± 6.8	.56
Outcomes
Death	6 (9%)	1 (4%)	5 (12%)
Readmission	34 (52%)	14 (61%)	20 (48%)
LT	1 (2%)	–	1 (2%)
Follow-up halted	24 (37%)	8 (35%)	16 (38%)

For categorical variables, total and relative values are provided. For continuous variables mean ± SD and P values are provided.

BMI = body mass index, INR = international normalized ratio, LT = liver transplantation, MELD = model on end-stage liver disease, PMI = psoas muscle index.

^*Significant P values <.05.

Regarding CTP score, 13 patients (20%) were classified as grade A, 19 patients (29%) as grade B, and 33 patients (51%) as grade C. Mean MELD score was 16 ± 6; mean MELD-Na was 19 ± 8 (Table 2).

Overall mean of PMI was 4.94 ± 1.3 cm²/m² (range 2.57–9.65). Male patients had a significantly higher PMI than female patients (5.16 vs 4.54 cm²/m², P = .04). No significant difference of PMI was found between age categories.

Sarcopenia was found in 30 patients (46%). It was more present in the male than in the female group, but the difference was not significant (52% vs 35%, P = .20). Age was similar in sarcopenic and nonsarcopenic patients. In sarcopenic patients, weight (77.5 vs 84.28, P = .07) and BMI were lower than in nonsarcopenic patients (25.88 vs 6.22 kg/m2, P = .009; Table 3).

Table 3.

Xxxx

n = 65	Sarcopenic (n = 30; 46%)	Nonsarcopenic (n = 35, 54%)	P value
Age (yr)	60.4 ± 14.43	60.46 ± 12.43	.99
Sex			.2
Female	8 (26.7%)	15 (42.9%)
Male	22 (73.3%)	20 (57.1%)
Weight (kg)	77.5 ± 20.9	84.28 ± 20.65	.0726†
BMI (kg/m2)	25.88 ± 6.22	29.38 ± 6.35	.0095*
Mean time to readmission (d)	82	258	.013*
Next visit within 6 mo	16 (53%)	9 (26%)	.04*
Time to death (d)	157 ± 238	252 ± 214	.28
PMI (cm2/m2)	3.945 ± 0.68	5.785 ± 1.09	<.0001*
Albumin (g/dL)	2.825 ± 0.52	3.079 ± 0.77	.4
Bilirubin (mg/dL)	6.39 ± 4.26	5.77 ± 3.82	.8
Creatinine (mg/dL)	1.056 ± 0.512	1.473 ± 1.212	.087†
INR	1.55 ± 0.4	1.55 ± 0.39	.99
Sodium (mmol/L)	134.0 ± 5.44	134.5 ± 6.52	.43
Child–Turcotte–Pugh	10.03 ± 1.92	8.57 ± 2.77	.0187*
A (5–6)	1 (3 %)	12 (34%)
B (7–9)	10 (33 %)	9 (26%)
C (10–15)	19 (63 %)	14 (40 %)
MELD	15.73 ± 6.08	16.86 ± 6.55	.37
MELD-Na	19.13 ± 6.64	19.66 ± 7.219	.76
Outcomes			.087†
Death	5 (17%)	1 (3%)
Readmission	14 (47%)	20 (57%)
LT	–	1 (3%)
Follow-up halted	11 (37%)	13 (37%)

Characteristics of patients, grouped into sarcopenic and nonsarcopenic. For categorical variables total and relative values are provided. For continuous variables, mean ± SD and P values are provided.

^*Significant P values <.05.

^†Significant P value <.1.

BMI = body mass index, PMI = psoas muscle index, INR = international normalized ratio, MELD = model on end-stage liver disease, LT = liver transplantation.

CTP grades differed significantly between sarcopenic and nonsarcopenic patients. In sarcopenic patients, 1 (3%) fell into category A, 10 (33%) fell into category B, and 19 (63%) fell into category C; values for nonsarcopenic patients were category A = 12 patients (34%), category B = 9 patients (26%), and category C = 14 patients (40%), respectively (P = .003). Sarcopenic patients were readmitted to the hospital within a significantly shorter time compared to nonsarcopenic patients (82 ± 63 vs 258 ± 276 days, P = .01). There was a trend of creatinine being lower in sarcopenic patients that was not significant (1.056 ± 0.512 vs 1.473 ± 1.212 mg/dL, P = .09; Table 3). Death was more prevalent in the sarcopenic group as well, but this also was not significant (5 vs 1, P = .09). Additionally, a survival analysis was conducted comparing sarcopenic and nonsarcopenic group; it yielded no difference between curves.

When further splitting sarcopenic and nonsarcopenic patients into groups by sex, 8 (35%) out of 23 women were sarcopenic with PMI < 4.3 cm²/m². There were no significant differences found between sarcopenic and nonsarcopenic women. Out of 42 men, 22 (52%) were sarcopenic with PMI ≤ 5.1 cm²/m². For sarcopenic men, time until readmission was significantly shorter (221 ± 323 days vs 484 ± 475, P = .04) compared to nonsarcopenic men. Also, the number of deaths was significantly higher in sarcopenic men (5 vs 0; P < .05).

Modified hepatitis activity index scores had no significant association with any other variables.

Regarding correlation of PMI with other variables, weight (R = 0.35) and BMI (r = 0.365) were moderately correlated with PMI, implying that patients with lower weight also had a lower PMI. For albumin (r = 0.342) and CTP (r = –0.341), there was a moderately strong correlation with PMI suggesting that patients with low PMI had lower levels of albumin and higher CTP scores.

Finally, CTPS, CTP, PMI, MELD-Na, and MELD were correlated with time to readmission (Fig. 2). Correlation for CTPS score (formula 3) was strongest (r = 0.7299; P < .001), followed by original CTP (r = 0.696; P < .001). Correlation with MELD-Na and PMI was moderate (r = 0.405; P = .009 and 0.388; P = .01, respectively). Correlation with MELD was weakest (r = 0.354; P = .02).

Figure 2.

Scatter graphs and nonlinear regression curves of correlations between time to readmission and MELD, MELD-Na, CTP, PMI, and CTPS. Pearson’s correlation coefficient r indicates the effect of correlation with values of R < 0.3 indicating weak, 0.3 < R < 0.7 indicating moderately strong and R > 0.7 indicating strong correlation. Negative values of r indicate negative correlation. P values indicate significance. “Time to readmission” refers to timespan between score calculation and readmission for liver-related causes. CTP = Child–Turcotte–Pugh, CTPS = CTP sarcopenia, MELD = model of end-stage liver disease, PMI = Psoas Muscle Index.

4. Discussion

Cirrhosis is the common end stage of diverse liver diseases. While some patients can live with it for years or even decades, others will show rapid worsening with multiple complications, frequent clinical readmissions, and finally death or transplantation.^[38–40]

Thus, various tools have been tried to predict morbidity and mortality in LC. Since no single parameter was sufficient, scores had been developed that incorporate multiple clinical and laboratory variables. Surprisingly, the 2 scores most widely used today are 2 of the oldest: CTP from 1964,^[16] and MELD from 2002. Both utilize a combination of clinical and/or laboratory parameters, and both do not take into account sarcopenia.^[20,21]

Now it has become evident in the last decades that the progressive loss of muscle mass is an independent pathogenic factor in LC and that it is negatively correlated to survival.^[5–7]

In this study, we therefore tested a radiological tool of measuring sarcopenia for its usefulness in predicting outcome in LC. PMI was chosen here for several reasons. First, it is least influenced by ascites. It is therefore superior to bioelectrical impedance,^[13,41] which is heavily skewed by body water content.^[42–44] We also found it more practical than SMI, because SMI takes into account abdominal wall musculature, which will be distended and thinned by ascites. While this thinning will not alter the cross-sectional area of the muscle, it renders it very difficult to measure correctly..^[27,28] Compared to functional tests like handgrip strength or stand-up test, PMI is not influenced by neuropathy, encephalopathy, or tremor.^[11,12] Finally, for this study, we could retrospectively collect it out of existing data.

However, the relevance of PMI in hepatic disease is not undisputed. While several groups found it predictive for waiting list mortality or postoperative outcome in LT patients^[45–47] as well as in hepatocellular carcinoma therapy,^[48] or in response to l-carnitine,^[49] others reported only weak correlation with waiting list mortality or outcome.^[37,50]

In our own small patient collective, low PMI was unambiguously correlated with overall morbidity and with shorter time to readmission.

Moreover, if we take the grade of correlation as measure of predictive power, PMI as a single parameter outperformed a complex score like MELD and was roughly on par with MELD-Na; only CTP performed better. This may be due to the fact that, in an attempt to count in hepatorenal syndrome, MELD includes creatinine. However, serum creatinine is influenced not only by kidney function but conversely by muscle mass.^[51–53] Hence, its single-sided use in a predictive score for LC is at least questionable.^[54–56] This was further affirmed by our findings that low serum levels of creatinine correlated with low PMI and high CTP.

If one would stick to MELD in a clinical setting, it should at least be improved by either excluding creatinine or by leaving it in and complementing it with a measurement of sarcopenia.^[57]

Another option would be to take the best-performing score, CTP, and expand it with data on sarcopenia. We did this by adding a very simple factor: CTP plus 1 point for patients without sarcopenia, and CTP plus 3 points for patients with sarcopenia. The result was a new score, CTPS, that predicts outcome better than any other score including CTP alone.

The binary principle of using only “sarcopenia or no sarcopenia” instead of PMI has several advantages here: first, it is simple to calculate. Second, it is not affected by differences in weight and sex. Finally, by only using a “yes or no” statement instead of the original PMI values, future refinements of the score could incorporate other, simpler methods of detecting sarcopenia.

There are limitations to our study. First, it was conducted as a retrospective single-center study with a relatively small sample size. Therefore, observation periods were heterogeneous, and data were not suited to perform multivariate analysis. Additionally, the fact that not all patients were readmitted may introduce bias. However, the rate of readmittance was not significantly different in the sarcopenic and nonsarcopenic group, but the time to readmittance was. So we still think this data is valid. Also, we had no information on nutrition intake, amount of exercise, or muscle conception, so possible differences in muscle density due to fatty infiltration were not taken into account.^[58]

Still, even in this small preliminary study, the principle of CTPS appeared highly promising. We, therefore, propose larger prospective studies with larger sample size, longer observation periods, and multivariate analysis in order to fine-tune and validate the score and to evaluate its clinical usefulness in predicting short and long-term outcome in LC.

Author contributions

FH planned and supervised the study, researched literature and software, and wrote the manuscript. TS took PMI measurements, researched clinical data, conducted statistics, researched literature, and co-wrote the manuscript. AGS provided radiological data, researched literature and radiological software and edited the manuscript. SL cared for patients, supervised the study, researched literature, and edited the manuscript.