Efficacy and safety of oral branched-chain amino acid supplementation in patients undergoing interventions for hepatocellular carcinoma: a meta-analysis

Potentially relevant studies regarding the efficacy of BCAAs on HCC were identified by searching the PubMed, Embase, Web of Science, and the Cochrane Libraries, using the key words “branched-chain amino acids,” “hepatocellular carcinoma,” and their abbreviations. Multiple synonyms were also used. The search was restricted to “human” and “English”. Reference lists of all retrieved documents were manually searched for potentially relevant reports missed by the intelligent retrieval systems mentioned above. The search was carried out in September 2014, and the whole selection process was completed independently by two investigators (LC and YQC). Inconsistent search results were resolved with the assistance of an arbiter (HDH), where necessary.

Published studies comparing the efficacy of oral BCAA supplementation to no additional BCAA supplementation in HCC were used for our meta-analysis, if the study was eligible and the data available. Inclusion criteria were as follows: a) randomized controlled trials (RCTs), with retrospective and prospective cohort study designs, b) All patients were diagnosed with HCC and underwent therapies for HCC, whether surgical or non-surgical, c) Studies directly comparing additional oral BCAA supplementation to usual meals with no BCAA supplementation with a follow-up duration equal to or more than 24 weeks and d) All patients demonstrated good compliance in taking the supplements and were monitored by a dietician. Studies were excluded if they featured: a) non-comparative data or observational methodologies, b) No available outcome measures and a follow-up duration of less than 24 weeks, c) Patients with serious impairment of organ function due to respiratory, renal, or heart disease or d) Patients with serious hepatic disease but without HCC.

The interesting outcome measures studied were as follows: a) Mortality and HCC recurrence were used as primary efficacy measures and b) Liver biochemical and clinical symbols were used as secondary efficacy measures. Data on mortality was studied at the time-points of 1 and 3 years, and HCC recurrence at the time-points of 1, 2, and 3 years. Liver biochemical data included serum albumin, total bilirubin, alanine aminotransferase (ALT), and aspartate aminotransferase (AST) at the time-points of 6 and 12 months. Clinical signs included instances of ascites and the occurrence of edema shortly after anti-HCC therapies. The side effects caused by BCAAs also received special attention.

All data were independently extracted from the included studies by two investigators (LC and YQC), and, where possible, calculated and confirmed. Any dispute between investigators was resolved by discussion or by arbitration from an arbiter (HDH), when necessary. If useful data were presented indirectly by figures and graphs or through different metrics, they were translated into correlative patterns by using get-data software or relevant formulae. If mean values or standard deviation (SD) for analysis were unavailable, they were calculated from medians and ranges using relevant formulae [31].

The quality of all included RCTs was assessed using the revised Jadad quality scale, which graded the quality of a study from 0 (lowest) to 7 (highest) by examining randomization, blinding, allocation concealment, and drop-out. For cohort designs, the quality of studies was assessed by the Newcastle-Ottawa Scale (NOS) based on several standards including selection of cohorts, comparability of cohorts, and assessment of the outcomes.

Data analysis was carried out with the use of Software Stata version 12.0 (Stata Corporation, College Station, TX, USA) and was based on an intent-to-treat principle. Both dichotomous and continuous variables were extracted in this analysis. For dichotomous outcomes, the results were presented as relative risk (RR) with a 95 % confidence interval (95 % CI), while continuous results were presented as standardized mean difference (SMD) with a 95 % confidence interval (95 % CI). The overall effects were measured using a z-score with a significance set at P < 0.05. An RR of approximately 1 (p > 0.05) indicated no statistical significance between BCAA and control groups while an RR value deviated from 1 (p < 0.05) represented statistical significance. A SMD of approximately 0 (p > 0.05) indicated no statistical significance between BCAA groups and control groups, while SMD values with a deviation from 0 (p < 0.05) were judged to be of statistical significance. Statistical heterogeneity was evaluated by using chi-square and I-square (I²) tests with a significance set at p < 0.1. P < 0.1 and I² > 50 % were considered to be significant heterogeneity. The random-effect method was used to combine results if confirmed significant heterogeneity was observed; otherwise, the fixed-effect method was used. To assess sources of potential bias, sensitivity analyses were performed where required. Funnel plots were used to assess the publication bias of selected articles and any potential bias was judged by Begg’s test.

As the search strategy shows (Fig. 1), 624 records in total were identified, and 311 records were removed as duplicate documents retrieved from two or more databases. The remaining studies received further screening by scanning the title or abstract, which resulted in the exclusion of a further 290 studies. As a result, 23 full-text articles were subjected to detailed evaluation, of which two were excluded because they analyzed the same patient groups, and a further ten were excluded due to lack of available data. Eventually, 11 eligible articles relating to a total of 974 subjects (450 in BCAA groups and 524 in control groups) were chosen for this meta-analysis. Of the 11 eligible studies, six were RCTs [19‐24] and five were cohorts (of which three were retrospective and two prospective designs) [25‐28, 32]. Out of the six RCTs, four received a Jadad score of 4 or 5 and were considered high-quality, while the other two were deemed low quality because of scores lower than 4 (Additional file 1: Table. S1). All five cohort studies received a NOS score of 7 or 8 (Additional file 2: Table S2). No publication bias was found among these included studies (Additional file 3: Figure S1). The detailed characteristics of the included studies are summarized in Table 1. All pooled results are summarized in Additional file 4: Table S3 and the results of each study reviewed are summarized in Additional file 5: Table S4.

Information about 1-year mortality was reported in nine studies [19, 20, 22, 24‐28, 32] while 3-year mortality was reported in five studies [19, 22, 24, 25, 27]. A fixed-effect model was used to analyze mortality at the two time-points since there was no significant heterogeneity among these studies (P = 0.912, I² = 0.0 %; P = 0.129, I² = 43.9 %, respectively). This meta-analysis showed that both 1-year and 3-year mortality were lower in BCAA groups when compared with control groups (18 % vs. 21 %; 37 % vs. 44 %, respectively). 1-year mortality did not achieve statistical significance (RR = 0.856, 95 % CI: 0.669–1.094, P = 0.214), while the difference became statistically significant with prolonged duration up to three years (RR = 0.797, 95 % CI: 0.667–0.952, P = 0.012) (Fig. 2).

Kanekawa, et al. [28] reported that the mortality of the BCAA group was significantly lower than that of control group for Child-Pugh class B patients, but there was no significant difference for Child-Pugh class A patients. Another three studies [26, 27, 32], which also reported significantly lower mortality in the BCAA groups compared with the control groups, were found to include a relatively higher proportion of Child-Pugh class B patients at baseline than that of studies [19, 20, 22, 25] which reported no significant difference in mortality (Table 2).

The 1-year, 2-year, and 3-year recurrence of HCC were obtained in five [20‐23, 26], three [20, 22, 23], and three [19, 22, 23] studies respectively. No significant heterogeneity was found at the three time-points (p = 0.940, I² = 0.0 %; p = 0.308, I² = 15.2 %; p = 0.948, I² = 0.0 %, respectively). Similarly, a fixed-effect approach was used to estimate the overall effects. The rates of HCC recurrence were also lower in BCAA groups than control groups (11 % vs. 18 %; 30 % vs. 34 %; 54 % vs 56 %, respectively). However, at none of the three time-points did HCC recurrence achieve statistical significance (RR = 0.613, 95 % CI: 0.315–1.192, P = 0.149; RR = 0.882, 95 % CI: 0.544–1.431, P = 0.612; RR = 0.946, 95 % CI: 0.749–1.194, P = 0.638, respectively) (Additional file 6: Figure S2).

Liver biochemical parameters included serum levels of albumin, total bilirubin, ALT, and AST at the time-points of 6 and 12 months. The eligible articles showed no significant baseline difference among these parameters, so data at 6 and 12 months were suitable for direct comparison. Seven studies [19‐22, 24‐26] reported the 6-month albumin level and an equal amount of articles [19‐21, 23‐26] mentioned the 12-month albumin level. Significant heterogeneity was detected at 6 months (P = 0.035, I² = 55.7 %) but not detected at 12 months (P = 0.207, I² = 29.0 %). A random-effect model for 6 months and a fixed-effect model for 12 months showed that albumin levels were higher in both models for the BCAA groups compared with control groups (SMD = 0.515, 95 % CI: 0.217–0.812, P = 0.001; SMD = 0.234, 95 % CI: 0.033–0.435, P = 0.022, respectively) (Fig. 3). However, there were no significant differences between the two groups in terms of total bilirubin, ALT, and AST at 12 months (P = 0.454, P = 0.882, P = 0.235, respectively) on the basis of no significant heterogeneity among included studies of ALT and AST (P = 0.769, I² = 0.0 %; P = 0.680, I² = 0.0 %, respectively), but significant heterogeneity was found among the included studies in total bilirubin (p = 0.049, I² = 58.0 %), which was analyzed using a random-effect model (Additional file 7: Figure S3).

Five studies [19, 21, 22, 24, 25] provided data for ascites and three [19, 21, 24] reported the occurrences of edema a short time after therapies for HCC. No significant heterogeneity was found in ascites and edema (p = 0.420, I² = 0.0 %; p = 0.575, I² = 0.0 %, respectively). A fixed-effect analysis showed that the ascites rate in BCAA groups was distinctly lower when compared with control groups (9 % vs.16 %, P = 0.029), and a similar result was found in the edema rate (9 % vs. 16 %, P = 0.035) (Fig. 4).

Nagasue, et al. [19] reported that seven patients suffered side effects caused by BCAAs: nausea and vomiting in four, diarrhea in one, abdominal distension in one, and hypertension in one. Three of the reported patients could not continue BCAA supplementation. Meng, et al. [20] reported that two patients experienced occasional, diffuse abdominal pain after ingestion and one suffered from transient diarrhea, but all of these symptoms stopped spontaneously and none of the patients ended BCAA supplementation. Adverse reactions to BCAA supplementation were either not observed or not mentioned in the remaining articles. No serious events induced by BCAAs were reported in any of the included studies (Additional file 5: Table S4).