Results
Literature Search and Study Characteristics
We identified 17 studies that met the inclusion criteria, including 7 cohort studies and 10 case–control studies, the search process was summarised in Figure 1. The main characteristics of each study were presented in Table 1, a total of 1 670 093 participants and 4826 HCC cases were involved after combining all the studies. Among these studies, three were carried out in US, six in Europe, and eight in Asia. It is should be noted that relative risks for different types of meat were reported in different publications for the cohort of the NIH-AARP Diet and Health Study in US and the National Cancer Institute study in Italy. Most studies adjusted for potential confounding factors including age, sex, HBV infection, alcohol drinking, etc. Overall, the quality score ranged from 4 to 9, with an average score 6.76, and no study of low quality was found.
(Enlarge Image)
Figure 1.
Flow chart of study selection.
Red Meat
Among all the studies, 9 studies reported the association between red meat and HCC risk, including three cohort studies and six case–control studies. After combining all the RRs, a pooled RR of 1.10 (95% CI: 0.85–1.42) was calculated with the random-effects model (Figure 2 and Table 2), indicating high consumption of red meat was not associated with increased HCC risk. The RRs were 0.97 (95% CI: 0.71–1.32) for case–control studies and 1.43 (95% CI: 1.08–1.90) for cohort studies, respectively.
(Enlarge Image)
Figure 2.
Forest plot of red meat consumption and HCC risk.
There was a moderate heterogeneity among studies (I=61.1%, Q = 20.59, P=0.008). Meta-regression analysis showed that 'publication year' eliminated the heterogeneity completely (P<0.05), with I reducing from 61.1% to 0.0%, while study design, location, quality score were found not to associate with the heterogeneity (P = 0.228, 0.363, 0.231, respectively). Besides, we conducted a sensitivity analysis by omitting one study each time and recalculating the pooled RRs, the results showed that the pooled risk estimates did not change significantly with a range from 1.01 (95% CI: 0.79–1.30 to 1.17 (95% CI: 0.93–1.47), indicating that the overall RR was not substantially influenced by any of the individual studies. No significant publication bias was found, either from Egger's test (P=0.59) or from Begg's test (P=0.84).
Processed Meat
Three cohort studies and two case–control studies were included in the meta-analysis concerning the association between processed meat intake and HCC risk. No substantial heterogeneity was observed among the studies (I = 42.90%, Q = 7.00, P=0.136), and the pooled RR was 1.01 (95% CI: 0.79–1.28) using the fixed-effects model (Table 2). Further subgroup analyses were not conducted since the number of included studies was small. We found no evidence of publication bias from the included studies, either from Egger's test (P=0.86) or from Begg's test (P=1.00).
White Meat
Our analysis of eight studies on white meat consumption and HCC risk yielded a summary RR of 0.69 (95% CI: 0.58–0.81) (Figure 3 and Table 2). The inverse association did not change across all subgroup analyses, regardless of study design, quality score, publication year and ethnicity (Table 2). We found no evidence of the heterogeneity among the included studies (Q = 5.77, P=0.57, I = 0.0%), either no publication bias was detected (P=0.461 by Egger's test, P=0.80 by Begg's test).
(Enlarge Image)
Figure 3.
Forest plot of white meat consumption and HCC risk.
Fish
The association between fish intake and HCC risk were evaluated in 10 studies, including six case-studies and four cohort studies, almost all the studies reported an inverse association, with two of them being statistically significant. We found a low heterogeneity (Q = 11.6, P=0.27, I = 18.6%) among all the studies, and the fixed-effects pooled analysis showed that high fish intake was associated with a reduced risk of HCC (RR: 0.78; 95% CI: 0.63–0.90)(Figure 4 and Table 2). Stratification by study design, quality score and ethnicity did not change the association, although some RRs in the subgroups became statistically insignificant (Table 2).
(Enlarge Image)
Figure 4.
Forest plot of fish consumption and HCC risk.
A small publication bias was detected among all these studies (P=0.07 by Egger's test, P=0.24 by Begg's test), however, with the trim and fill method, no studies were found to be filled and data remained unchanged, suggesting that the influence of publication bias could be negligible.
Total Meat
We found no association between the total meat consumption and HCC risk, the summary RR was 0.97 with its 95% CI (0.85–1.11) after combining the only four studies involved (Table 2). No heterogeneity was observed (Q = 2.0, P=0.57, I= 0.0%), the publication bias was not evaluated considering the limited number of studies included.