SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Authorship
  8. Acknowledgement
  9. References
  10. Supporting Information

Background

Malnutrition is a common and clinically significant problem in patients with cirrhosis. The impact of nutritional therapy remains unclear.

Aim

To provide an up-to-date systematic review and meta-analysis of RCTs of oral or enteral nutritional supplementation (ONS or ENS) on nutritional and clinical outcomes in adult patients with cirrhosis.

Methods

The primary outcome measure was survival. Included: full-text English language RCTs investigating ONS or ENS vs. a standard nonsupplemented diet in patients with cirrhosis. Excluded: parenteral or branched chain amino acids intervention; treatment duration ≤7 days, exclusive evaluation of posttransplant, postsurgical or quality of life outcomes.

Results

Six trials (4 ONS/2 ENS) and 470 patients were included with 71% males and median age 53 years. When all studies were combined, there was no reduction in mortality [Relative risk (RR): 0.75 (0.42, 1.32), P = 0.31]. Subgroup analysis of 3 of the 4 ONS studies did demonstrate a mortality reduction [RR: 0.40 (0.18, 0.90), P = 0.03]. Of the 2 ENS studies, one included the sickest patients in the meta-analysis (82% Child Pugh C) and the other had the shortest mean intervention duration (8.6 days), possibly impacting the potential for benefit. Study quality was suboptimal (median Jadad = 2).

Conclusions

Although there is insufficient evidence to definitively state that oro-enteral nutritional supplementation impacts clinical outcomes, on the basis of this analysis, one can be cautiously optimistic that there is the potential for benefit without an increase in adverse events. Adequately powered, Child Pugh stratified studies of at least 1 month in duration are needed to clarify the impact on relevant clinical outcomes.


Background

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Authorship
  8. Acknowledgement
  9. References
  10. Supporting Information

Malnutrition is a common problem in patients with cirrhosis, diagnosed in one-third of Child Pugh A (CP-A) patients and 50–90% of CP-B/C patients.[1-4] Malnourished cirrhotic patients develop more infections, have a higher prevalence of portal hypertension-related complications and require longer stays in hospital.[5-7] Moreover, malnutrition is an independent predictor of death.[1, 2, 5, 8, 9] With such a clear association between poor nutritional status and negative medical outcomes, several studies have attempted to clarify whether oral or enteral supplementation of macro and micronutrients can improve outcomes.[10-15]

When analysed individually, the conclusions of these cirrhosis-specific nutritional interventional trials have been discordant. A recent attempt to meta-analyse the data[16] was limited by the exclusion of trials where additional calories were provided in the form of food alone (as opposed to an oral nutritional supplement) and by the mixture of both branched chain amino acid (BCAA) and non-BCAA trials. The latter is considered a limitation as BCAA supplements may have unique pathophysiological properties and are not routinely available.[17]

Therefore, the practising clinician is left without a clear strategy for the management of malnutrition in cirrhosis. In keeping with this uncertainty, cirrhosis-specific guidelines for supplementation via the oral route or by an enteral feeding tube (herein also termed oro-enteral supplementation), are predominantly based on grade C evidence (expert opinions and/or clinical experience).[18]

Accordingly, the objective of the current study was to provide an up-to-date systematic review and meta-analysis of randomised controlled clinical trials of non-BCAA oro-enteral nutritional supplementation on nutritional and clinical outcomes in patients with cirrhosis.

Methods

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Authorship
  8. Acknowledgement
  9. References
  10. Supporting Information

Inclusion/exclusion criteria

Studies were considered if they met the following inclusion criteria: (i) English language article, (ii) full-text, (iii) all trial participants had liver cirrhosis as described by the authors, (iv) all trial participants were ≥18 years of age, (v) the intervention was an oral or enteral nutritional supplement and the control was ad libitum feeds, (vi) treatment duration of >7 days, (vii) randomised controlled trial design (crossover studies allowed), (viii) reported data for at least one of the prespecified outcomes [survival, change in liver function, nutritional outcomes, complications of cirrhosis]. Exclusion criteria were: (i) the exclusive assessment of posttransplantation, postsurgical or quality of life outcomes, (iii) >25% of patients had hepatocellular carcinoma (HCC) (25% was an arbitrary cut-off derived by author consensus) and (iv) that the intervention included BCAA supplementation or was given by the parenteral route.

Searches

A comprehensive search strategy including terms for cirrhosis, malnutrition and nutritional supplements was used. The following databases were searched: MEDLINE (1966–2011), SCOPUS (1966–2011), EMBASE (1988–2011) and PubMed (restricted to the last 180 days to obtain the ‘in process’ citations missed by MEDLINE). Bibliographies of review articles were hand-searched to identify other relevant studies. The search strategies are included in Supplementary Appendix S1. The study protocol of this systematic review was not published online. The original search was carried out 7 February 2011 and updated on 31 March 2012.

Trial selection and data extraction

Quality assessment and study selection were carried out by two independent reviewers (MN, PT). A third reviewer was available to adjudicate on discrepancies between the two reviewers (SVZ). Study quality was assessed using the Jadad scale and by assessment of allocation concealment, loss to follow-up and intention-to-treat. The primary outcome was survival. A priori specified secondary outcomes were: (i) change in liver function (bilirubin, MELD, Child Pugh score), (ii) complications of cirrhosis (infections, ascites control) or (iii) nutritional outcomes [serum albumin, body weight, body mass index (BMI), triceps skin fold thickness (TSFT), mid-arm muscle circumference (MAMC)].

Subgroup and sensitivity analyses

Pre hoc subgroup analyses were planned for two outcomes: mortality and MAMC. The following subgroups were chosen because of their potential relevance to the response to nutritional therapy: (i) the mode of nutritional supplementation (oral vs. enteral), (ii) the severity of liver dysfunction (compensated vs. decompensated cirrhosis or CP-A vs. CP-B/C), (iii) the aetiology of cirrhosis (alcohol vs. non-alcohol-related), (iv) the dose of the intervention, (v) the duration of the intervention (<30 days vs. ≥30 days) and (vi) baseline nutritional status (malnourished vs. not).

In addition, pre hoc sensitivity analyses were planned dividing patients by (i) study quality (Jadad score <3 vs. ≥3) and (ii) random vs. fixed-effects modelling.

Statistical analysis

Review Manager Version 5.0 (The Cochrane Collaboration, Copenhagen, Denmark) was used to perform meta-analyses. For the continuous outcome variables, bilirubin, albumin, and Child Pugh score, mean differences (MD) were calculated. As studies reported the continuous outcomes MAMC and TSFT using different scales (change from baseline and per cent change from baseline), these were reported as Standard mean differences. For dichotomous outcomes, risk ratios (RR) were computed to estimate between-group differences. If no events were reported in one treatment arm, a correction factor of 0.5 was added to each cell of the two-by-two table to obtain estimates of the RR. We pooled studies using a DerSimonian and Laird random effects model. All results are reported with 95% CI. We quantified statistical heterogeneity using the I-squared (I2) statistic and associated P-value. Publication bias was not analysed, as there were insufficient studies.

We imputed means with medians and standard deviations using ranges or inter-quartile ranges when these quantities were not reported.[13, 14] For one study,[11] we assumed that quantities given as standard errors were in fact standard deviations, as, for all variables, the given values were simply too large to make sense as standard errors. In the single crossover study,[14] a correlation of 0.5 was used to estimate the within patient correlation when computing differences between interventions for continuous outcomes.

Results

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Authorship
  8. Acknowledgement
  9. References
  10. Supporting Information

Search

The initial search was carried out on 7 February 2011 and identified 4136 abstracts. Using predefined inclusion and exclusion criteria, after review of the title and abstracts 146 potentially relevant abstracts were identified. These full-text articles were reviewed in detail. One hundred and forty-one full-text articles were excluded, leaving 5 included articles from the first search.[10-14] The second search was limited to articles published between the end date of the original search and 31 March 2012 and retrieved 193 abstracts. From this search, 1 full-text article was identified as being relevant for review and for inclusion,[15] bringing the final total of included studies to 6. Figure 1 shows the combined trial flow data for the two searches.

image

Figure 1. Flow of studies through the selection process.

Download figure to PowerPoint

Patient demographics

In total, the sample size for the six studies was 470 with 231 patients allocated to the intervention arm and 239 patients to the control arm. The 45 patients in the crossover trial by Manguso et al.[14] were counted in both the intervention and the control arms. All patients had cirrhosis without a history of hepatocellular carcinoma. Seventy-one per cent of patients were male with a median age of 53 years (IQR: 49–57). Alcohol was the most common aetiology of hepatic dysfunction, with 4 of the studies including only patients with alcohol-related cirrhosis[10-12, 15] and a single study including only hepatitis C patients.[14] Supplementary Table S1 provides additional study characteristics.

Study intervention and control groups

Four studies evaluated oral nutritional supplementation [10, 11, 13, 14] and 2 studies enteral nutritional supplementation.[12, 15] In 3 studies, the intervention was provided in the inpatient setting.[10, 12] The dose of nutritional therapy is provided in Supplementary Table S2, divided into the nutrition prescribed (dose prescribed to the patient by the investigator) and the nutrition received (dose the investigators estimated that the patient had consumed). The median (IQR) dose of nutritional therapy received in the interventional arm was 2444 (2097–2853) kcal/day and in the control arm was 1801 (1646–2134) kcal/day. Of these calories, the median protein dose received was 80 (75–107) g/day in the intervention arm and 53.5 (36–80) g/day in the control arm.

Outcome measures

Mortality

With the exception of the Manguso study,[14] the remaining 5 studies reported mortality outcomes in a total of 141 patients in the intervention arm and 149 patients in the control arm (Figure 2). When combined, there was no statistically significant reduction in mortality, with a risk ratio (RR) of 0.75 (0.42, 1.32), P = 0.31. The heterogeneity between these studies was moderate with an I2 of 23%, (I2 P = 0.27).

image

Figure 2. Pooled analysis of the five oro-enteral nutritional supplementation studies – intervention vs. ad libitum feeds. There was no statistically significant reduction in mortality when all studies were considered together [RR of 0.75 (0.42, 1.32), P = 0.31].

Download figure to PowerPoint

The analysis of secondary outcome measures is provided in Supplementary Table S3, and showed no significant differences between the nutritional intervention and control groups.

Methodological quality of included studies

The overall quality of the included studies was poor as reflected by a median Jadad score of 2 (Supplementary Table S4).

Pre hoc subgroup analyses: Mortality

Mode of nutritional supplementation (oral vs. enteral)

When subdivided into enteral and oral nutritional supplementation, in the 3 oral nutrition supplement studies,[10, 11, 13] there was a significant reduction in mortality with a RR of 0.40 (0.18, 0.90, P = 0.03) (Figure 3). This is interpreted as a 60% reduction in mortality in the intervention group as compared with the control group and was associated with no heterogeneity (I2 P = 0.84). In the 2 enteral nutrition supplement studies,[12, 15] the effect on mortality was not significant, with an RR of 1.13 (0.69, 1.85), P = 0.63 and no heterogeneity (I2 P = 0.90).

image

Figure 3. Subgroup analysis of two enteral and three oral nutritional supplementation studies – intervention vs. ad libitum feeds. There was a significant reduction in mortality [RR of 0.40 (0.18, 0.90), P = 0.03] in the oral nutritional supplementation subgroup.

Download figure to PowerPoint

The severity of liver dysfunction

Excluding the study by Hirsch et al.[11] where insufficient data were available, in one study,[14] 66% of the patients and in 3 studies,[12, 13, 15] 80–90% of the patients had CP-B/C cirrhosis. In the study by Bunout et al.,[10] although CP data were not provided, 89% had ascites at baseline. Therefore, as the majority of patients had advanced disease, we were unable to formally subgroup studies by the severity of liver dysfunction.

Aetiology of cirrhosis – Alcohol-related liver disease

In four studies, all patients had alcohol as the aetiology of cirrhosis.[10-12, 15] In the two remaining studies, none[14] or 22%[13] of the patients had alcohol-related cirrhosis. As only one study provided mortality data in the low-alcohol group,[13] we were unable to perform this subgroup analysis.

Dose of the intervention

Using an estimated average body weight of 70 kg (data not shown), and the guideline recommendation of 35–40 kcal/kg body weight/day of caloric supplementation,[18] we subgrouped studies by whether or not the intervention group received ≥2400 calories per day[10, 11, 13, 15] or less.[12, 14] As only one study in the low-dose group provided mortality data,[12] this subgroup analysis was not possible.

Duration of the intervention (≤30 days vs. ≥30 days)

The shortest duration of nutritional therapy was in the De Ledinghen study, providing enteral feeding for a mean of 8.6 days. Four studies provided the nutritional intervention for ≥30 days[11, 13-15] ranging from a duration of 77 days in the study by Le Cornu et al., 3 months in the studies by Dupont et al. and Manguso et al. and 1 year in the study by Hirsch et al. When these studies were separated from studies providing the intervention for <30 days,[10, 12] there was no significant difference in mortality noted in either subgroup. The longer duration studies had a RR for mortality of 0.65 (0.27, 1.58), P = 0.34, I2 = 53%, (I2 P = 0.12) and the shorter duration studies a RR for mortality of 0.73 (0.25, 2.17), P = 0.57, I2 = 0%, (I2 P = 0.35).

Malnourished vs. not malnourished at study baseline

Only two studies provided information about the number of patients who were diagnosed with malnutrition at study baseline. In the study by Le Cornu et al.,[13] 100% had an MAMC <25th percentile and 52% of patients had an MAMC <5th percentile. In the study by Dupont et al.,[15] 62% had an MAMC in the lowest 10th percentile. As demonstrated in Table S1, the mean and median baseline MAMC results were quite similar across studies, ranging from a low of 21.4 cm in the intervention group of the Dupont study,[15] to a high of 25.4 cm in the intervention group of the study by De Ledinghen et al.[12] Due to insufficient data as well as the similarity in the baseline MAMC measurements between studies, we were unable to evaluate the impact of baseline nutritional status on outcomes.

MAMC

In the subgroups with sufficient information to allow analysis (mode of nutritional supplementation, aetiology of cirrhosis, dose and duration of the intervention), there was no significant difference in the MAMC in either subgroup.

Pre hoc sensitivity analyses

Study quality

As all but one study had a Jadad score of <3, a sensitivity analysis of study quality was not possible.

Random vs. fixed-effect modelling

A sensitivity analysis by fixed-effects modelling did not alter any of the study conclusions.

Other reported events including adverse events

Other events are presented in Supplementary Table S5. Data on infections were elaborated to include the types of infections if available. Of the studies which reported statistical differences, the only statistically significant difference between the intervention and control groups was in the study by Hirsch et al.,[11] in which the intervention group had a lower rate of infections and a lower rate of total admissions to hospital.

The two enteral nutrition studies commented on the adverse events that were potentially attributed to nasogastric tube placement. In the enteral study by De Ledinghen et al.,[12] although no p values were provided, the authors commented that the 3 lethal complications in the enteral nutrition group (pneumonia, severe encephalopathy, recurrent variceal bleeding) could theoretically be attributed to the nasogastric tube. In the study by Dupont et al., 17% (5/29) of patients randomised to the enteral nutrition arm had enteral nutrition for less than a week due to poor tube tolerance or placement failure. Overall, however, there was no significant difference between the rate of complications (infections, bleeding, hepatic encephalopathy etc.) between the intervention and control arms (P = 0.59).

Discussion

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Authorship
  8. Acknowledgement
  9. References
  10. Supporting Information

Malnutrition is a well-recognised[1-4] and clinically important[1, 2, 5, 8, 9] problem in patients with cirrhosis. It is clear that the insufficient intake of macro and micronutrients is a potentially reversible contributor to the malnourished state.[19] What has been unclear and variably interpreted[16, 18] is whether oro-enteral supplementation has any impact on malnutrition-associated morbidity and mortality. In keeping with the findings of Koretz et al.,[16] our aggregated results do not support a significant reduction in mortality or a change in nutritional outcomes with nutritional intervention. Despite this finding, the statistically significant reduction in mortality in the ONS subgroup carries with it at most cautious optimism for a true benefit and at a minimum the conclusion that existing studies are insufficient to rule out a benefit.

There are many factors that influence whether a nutritional intervention has a chance of having a beneficial effect. These include the characteristics of the intervention (dose, duration, type) and the characteristics of the patient receiving the intervention (nutritional state, disease activity). In the current study, multiple subgroup analyses were performed to explore these contributing factors. Most subanalyses either showed no difference or had insufficient information to carry out the analysis.

The subgroup analysis of ONS vs. ENS did, however, yield a statistically significant reduction in mortality in the ONS group. How can these results be rationalised? In the three ONS studies that included mortality data,[10, 11, 13] all of the individual studies trended towards favouring the nutritional intervention. The fourth ONS study by Manguso et al.[14] did not provide mortality data, but in this study as well, the nutritional intervention proved beneficial, resulting in a statistically significant increase in muscle mass. Neither ENS study[12, 20] showed a benefit of nutritional intervention over control, but both studies had properties that may have made a benefit less likely. The study by De Ledinghen et al. had the shortest intervention duration of the whole study group with a mean of 8.6 ± 2.6 days.[12] The ENS study by Dupont et al. included the sickest patients of the group (82% CP-C, all requiring a serum bilirubin of ≥51 μmol/L for enrolment).[15] Supported by the authors and by a letter to the editor accompanying the article, the lack of efficacy of ENS in these advanced cirrhotics may have been a case of ‘missing the boat’ by focusing an intervention at a stage when the disease was already too advanced.[21] Therefore, we would conclude that within recognised study limitations, the subgroup results are suggestive of the potential for clinical benefit with nutritional supplementation. This will need to be confirmed in future studies.

Despite the signals for a reduction in mortality, there was no signal to support an improvement in muscle mass with nutritional supplementation. As published studies[14, 22] have shown that cirrhotics are capable of increasing muscle mass with nutrition, why did this parameter not improve in all of the included studies? We hypothesise that the advanced CP stage of most of the patients in the study made it more challenging to alter muscle mass. Although more studies are needed, existing data would support this reasoning. The included study by Manguso et al. enrolled the most well-compensated patients[14] and was able to demonstrate an improvement in lean body mass with nutrition. The non-included nocturnal supplementation study by Plank et al.[22] also showed an increase in muscle mass, most evident in CP-A vs. CP-B/C patients. Secondly, it is possible that even if there was an increase in muscle mass with nutritional supplementation, the nutritional assessment tools utilised in the included studies may have been too insensitive to detect these subtle improvements. The most commonly utilised tools (MAMC, serum albumin and body weight) are not as sensitive as gold standard nutritional assessment tools such as in vivo neutron activation analysis or muscle mass quantification by cross-sectional imaging (Tandon et al. unpublished observations). Lastly, the dose of nutrition is an important variable. It is unlikely to have contributed significantly to the lack of change in muscle mass in these studies, however, as the nutritional supplementation received in the majority of studies[10, 11, 13, 15] was within ESPEN guidelines.

Several study limitations exist. Importantly, the sample sizes were small, the secondary outcome measures were heterogeneous and as evidenced by the low Jadad scores, the quality of the included studies was suboptimal. Moreover, study inclusion was limited to English language full-text articles and thereby some relevant information may have been excluded. As detailed in Table S2, the methods utilised to determine the amount of calories a patient received were prone to error. Lastly, although this could also be viewed as a strength, to reduce inter-study heterogeneity and to focus on readily available interventions where there was a difference in calories prescribed to the intervention and control groups, the use of BCAA and late-evening supplementation were excluded from this meta-analysis. The BCAA data in cirrhosis have been variably interpreted and recently reviewed.[16, 23-25] Late evening supplementation as well has been recently reviewed and has shown promise as being superior to daytime supplementation in patients with cirrhosis.[17, 19, 22]

In conclusion, although there is insufficient evidence to definitively state that the oro-enteral supplementation of macro and micronutrients impacts outcomes, the meta-analysed data are suggestive that there is the potential for clinical benefit without an increase in adverse events. Adequately powered and likely multicentre studies will be needed to clarify the impact on relevant clinical outcomes (mortality, hepatic function, quality of life and muscle mass and muscle function as measured by gold-standard techniques). Confirmation of these benefits and whether they may be augmented by the addition of BCAA supplements and/or by changing the timing of the nutritional supplementation to before bedtime will also require further study. In future studies, we would concur that the ESPEN recommendation of 35–40 kcal/kg body weight/day remains a reasonable target. Moreover, we would recommend a minimum intervention period of 1 month and preferably of at least 3–6 months. In clinical practice, as is the case with other lifestyle modifications, it is intuitive that continued exposure to the lifestyle intervention is required for effect. Future studies should be stratified by disease stage and baseline nutritional status to assess the individualised responses and predictors of response to nutritional supplementation in the cirrhotic patient.

Authorship

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Authorship
  8. Acknowledgement
  9. References
  10. Supporting Information

Guarantor of the article: Dr Puneeta Tandon.

Author contributions: Study design (M Ney, P Tandon, L Gramlich, M Ma). Data collection (M Ney, P Tandon, S Veldhuyzen van Zanten). Data analysis (M Ney, P Tandon, B Vandermeer). Manuscript preparation and review (all authors). All authors have approved the final version of the article, including the authorship list.

References

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Authorship
  8. Acknowledgement
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Authorship
  8. Acknowledgement
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
apt12252-sup-0001-TableS1-S5-AppendixS1.docWord document280K

Table S1. Individual study demographics.

Table S2. Study interventions.

Table S3. Meta-analysis results for secondary outcome measures.

Table S4. Quality analysis of included trials.

Table S5. Other reported events, including adverse events.

Appendix S1. Search criteria.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.