- Literature retrieval, characteristics and quality assessment
The literature search yielded 3956 records. Using Endnote X9, duplicate records (n = 295) and non-target article types (n = 636) were excluded. During the preliminary screening process (title and abstract screening), we eliminated irrelevant articles (n = 988) and experimental studies (n = 1967). During the second screening process (full-text review), we included nine articles that met the eligibility criteria [20-28]. A flow diagram of the literature selection process is presented in Fig 1.
The meta-analysis was conducted on 27 trials included in nine articles. We extracted the necessary characteristics (e.g., geographic region, target miRNA, the mode of miRNA regulation, disease type, sample size, BMI, proportion of males and diagnostic sensitivity and specificity) of each trial. Trials were categorized by these factors as follows: (1) geographic region: Asian (n = 11), non-Asian (n = 16); (2) regulation mode: miRNA in upregulation mode (n = 21), in downregulation mode (n = 6); (3) disease type: NASH (n = 13), NAFLD (including NAFL and NASH) (n = 14); (4) average BMI: ≥ 30 kg/m2 (n = 15), < 30 kg/m2 (n = 9); and (5) proportion of males: ≥ 50% (n = 12), < 50% (n = 8). Detailed information of these trials is presented in Additional file 1: Table S1.
The overall sample size (including controls and cases) in this meta-analysis was 4036, of which 2764 were in NAFLD trials and 1272 in NASH trials (Fig 2-A). Moreover, we summarized the target miRNAs among these 27 trials, and the top 4 in terms of total sample size were miR-122 (n = 1107), miRNA-99a (n = 792), miRNA-34a (n = 642) and miRNA panel (n = 368) (Fig 2-B). We conducted a Cochrane bias graph to assess the quality of each included article according to the QUADAS-2 tool (Additional file 1: Fig S1).
- Diagnostic value of serum miRNAs in NAFLD cases
2.1 Pooling all trials to evaluate the diagnostic efficacy of serum miRNA for total NAFLD
Significant heterogeneity existed among the trials based on the sensitivity and specificity values (pooled I2 for sensitivity and specificity was 94.82% and 88.37%, respectively). Hence, we used the random-effects model in our study. The pooled values were as follows: sensitivity = 0.72 [95% confidence interval (CI): 0.64–0.79], specificity = 0.81 (95% CI: 0.75–0.86), PLR = 3.77 (95% CI: 2.86–4.96), NLR = 0.34 (95% CI: 0.27–0.44), DOR = 10.95 (95% CI: 7.05–17.01), and AUROC = 0.84 (95% CI: 0.80–0.87) (Fig 3). These results indicated that serum miRNA had moderate diagnostic accuracy for total NAFLD.
Next, we evaluated efficacy of the most studied serum miRNAs, namely, miRNA-122, miRNA-99a and miRNA-34a, for the diagnosis of total NAFLD. (1) The pooled values of miRNA-122 were as follows: sensitivity = 0.84 (95% CI: 0.77–0.90) I 2= 80.62%; specificity = 0.72 (95% CI: 0.61–0.81) I2 = 85.44%; PLR = 3.01 (95% CI: 2.12–4.27); NLR = 0.22 (95% CI: 0.14–0.33); DOR = 13.79 (95% CI: 7.29–26.06); and AUROC = 0.86 (95% CI: 0.82–0.89) (Additional file 1: Fig S2). (2) The pooled values of miRNA-99a were as follows: sensitivity = 0.82 (95% CI: 0.71–0.89) I2 = 93.46%; specificity = 0.82 (95% CI: 0.53–0.95) I2 = 96.90%; PLR = 4.58 (95% CI: 1.30–16.12); NLR = 0.22 (95% CI: 0.11–0.47); DOR = 20.42 (95% CI: 2.86–146.00); and AUROC = 0.87 (95% CI: 0.84–0.90) (Additional file 1:Fig S3). (3) The pooled values of miRNA-34a were as follows: sensitivity = 0.81 (95% CI: 0.76–0.85) I2 = 5.73%; specificity = 0.83 (95% CI: 0.77–0.87) I2 = 33.16%; PLR = 4.70 (95% CI: 3.51–6.30); NLR = 0.23 (95% CI: 0.18–0.29); DOR = 20.34 (95% CI: 13.08–31.60); and AUROC = 0.85 (95% CI: 0.82–0.88) (Additional file 1: Fig S4). These results indicated all three miRNAs had a moderate diagnostic accuracy for total NAFLD. It should be noted that, miRNA-34a showed the lowest heterogeneity; thus, miRNA-34a was more available to diagnose total NAFLD.
2.2 Pooling trials NO.1–13 and trials NO.14–27 separately to evaluate the diagnostic efficacy of serum miRNA for NASH and NAFLD
(1) The pooled values of NASH trials were as follows: sensitivity = 0.74 (95% CI: 0.66–0.81) I2 = 74.94%; specificity = 0.85 (95% CI: 0.77–0.91) I2 = 79.60%; PLR = 5.01 (95% CI: 3.11–8.05); NLR = 0.31 (95% CI: 0.23–0.42); DOR = 16.24 (95% CI: 8.17–32.28); and AUROC = 0.86 (95% CI: 0.83–0.89) (Additional file 1: Fig S5).
(2) The pooled values of NAFLD trials were as follows: sensitivity = 0.71 (95% CI: 0.58–0.81) I2 = 96.88%; specificity = 0.76 (95% CI: 0.68–0.83) I2 = 90.57%; PLR = 2.99 (95% CI: 2.24–3.99); NLR = 0.38 (95% CI: 0.26–0.55); DOR = 7.93 (95% CI: 4.66–13.49); and AUROC = 0.80 (95% CI: 0.77–0.84) (Additional file 1: Fig S6).
These results revealed that serum miRNA had moderate diagnostic accuracy in both NASH and NAFLD trials. Moreover, serum miRNA showed a better diagnostic efficacy for NASH than that for NAFLD, as indicated by the higher DOR, higher AUROC, and lower heterogeneity.
2.3 Pooling trials NO.9–13 to evaluate the diagnostic efficacy of serum miRNA for distinguishing NASH from NAFL
The pooled values for these trials were as follows: sensitivity = 0.83 (95% CI: 0.70–0.91) I2 = 86.22%; specificity = 0.85 (95% CI: 0.74–0.92) I2 = 85.11%; and AUROC = 0.91 (95% CI: 0.88–0.93). These results suggested that serum miRNA had high accuracy for discriminating NASH from NAFL (Additional file 1: Fig S7).
- Subgroup analysis
To investigate the source of heterogeneity, we conducted subgroup analysis (Qualitative Research). Trials NO.1-27 were divided into subgroups based on five study factors: geographic region, type of disease, regulation mode, proportion of males and BMI. Subsequently, we calculated pooled values for each subgroup (Table 1). The results as follows.
(1) Geographic region: compared with Asian trials, non-Asian trials had higher sensitivity (0.77 vs. 0.64) and specificity (0.83 vs. 0.76) and significantly lower heterogeneity (sensitivity I2 81.33% vs. 96.08%, specificity I2 76.03% vs. 92.08%). Non-Asian trials had a higher DOR (17 vs. 6) and AUROC (0.87 vs. 0.77).
(2) Type of disease: These results were described in Section 2.2.
(3) Regulation mode: compared with upregulation trials, downregulation trials had lower sensitivity (0.66 vs. 0.74), higher specificity (0.89 vs. 0.78), and lower heterogeneity (sensitivity I2 59.82% vs. 95.92%, specificity I2 58.57% vs. 90.07%). Apart from that, downregulation trials had a higher DOR (16 vs. 10) but equivalent AUROC values (0.83 vs. 0.83).
(4) Proportion of males: compared with trials that had a proportion of males ≥ 50%, trials with a proportion of males < 50% had higher sensitivity (0.77 vs.0.64) and specificity (0.87 vs. 0.74) and significantly lower heterogeneity (sensitivity I2 81.17% vs. 95.58%, specificity I2 58.46% vs. 90.78%). Trials with a proportion of males < 50% had a higher DOR (21 vs. 5) and AUROC (0.89 vs. 0.75).
(5) BMI: compared with trials with BMI < 30 kg/m2, trials with BMI ≥ 30 kg/m2 had higher sensitivity (0.77 vs. 0.63) and specificity (0.84 vs. 0.75) and significantly lower heterogeneity (sensitivity I2 82.34% vs. 96.69%, specificity I2 76.67% vs. 93.11%). Trials with BMI ≥ 30 kg/m2 had a higher DOR (17 vs. 5) and AUROC (0.87 vs. 0.76).
Collectively, serum miRNA showed more accurate diagnosis of total NAFLD in these conditions: non-Asian, presence of NASH, female predominance and BMI ≥ 30 kg/m2. In terms of heterogeneity, all of the above five study factors could be the potential sources.
- Meta-regression
To determine the source of heterogeneity, we performed meta-regression (Quantitative Research). Geographic region, type of disease, miRNA regulation mode and miRNA profiling were considered as covariates. Due to lack of some data, proportion of males and BMI were not included in this meta-regression. We made the assignment as follows: disease (Yes = NASH, No = NAFLD), regulation mode (Yes = upregulation, No = downregulation), region (Yes = Asian, No = Non-Asian), and miRNA profiling (Yes = single miRNA, No = miRNA panel). The reason for heterogeneity might be related to geographic region (Asian) (sensitivity P < 0.01, specificity P < 0.001), but was unrelated to type of disease, miRNA regulation mode and miRNA profiling (Fig 4). Thus, Asian trials could be the significant factor to induce heterogeneity.
The major difference between Asian and non-Asian trials was in BMI. Of the 16 Non-Asian trials, 15 had a BMI ≥ 30 kg/m2 and one did not provide the BMI. Of the 11 Asian trials, nine had a BMI < 30 kg/m2 and two did not provide the BMI. Moreover, the pooled values of subgroups categorized by geographic region or BMI were similar. Apart from that, no significant differences in other aspects were seen between Asian and non-Asian trials. Therefore, based on these findings, we speculated that BMI < 30 kg/m2 could be a potential predominant factor to induce heterogeneity rather than Asian region. After omitting trials with study factors BMI < 30 kg/m2, proportion of males ≥ 50%, NAFLD, and miRNA in upregulation mode, the heterogeneity of sensitivity and specificity showed a downward trend: sensitivity I2 94.82% vs. 81.28% vs. 80.39% vs. 0%, specificity I2 88.37% vs. 72.92% vs. 76.92% vs. 25.99% (Additional file 1: Table S2).
- Clinical utility
We used Fagan’s nomogram to examine NASH trials (NO. 1–13) and NAFLD trials (NO. 14–27 (Fig 5 A-B). Pre-test probability was set at 50%. The results were as follows. (1) Among NASH trials, the PLR was 5 accompanied by post-test probability of 83% and NLR was 0.31 accompanied by post-test probability of 24%; and (2) Among NAFLD trials, the PLR was 3 accompanied by post-test probability of 75% and NLR was 0.38 accompanied by post-test probability of 27%. These findings revealed that serum miRNA exhibited higher positive diagnostic value for NASH than that for NAFLD. Furthermore, a likelihood ratio scattergram was constructed for NASH trials (Fig 5 C). The summary likelihood ratio for serum miRNA test was located in the lower right quadrant, which meant serum miRNA test did not reach the pathological standard for exclusion and confirmation, and hence, its clinical utility was limited.
- Publication bias
Based on Deeks’ funnel plot (Fig 6 A-C), publication bias was not observed in the trials in which serum miRNA was used to diagnose total NAFLD (P = 0.77), NAFLD (P = 0.84) and NASH (P = 0.29). In addition, publication bias was not observed that when serum miRNA-34a was used to diagnose total NAFLD (P = 0.46) (Fig 6D).