Performance of common imaging techniques vs serum biomarkers in assessing fibrosis in patients with chronic hepatitis B: A systematic review and meta-analysis.

BACKGROUND: Noninvasive biomarkers have been developed to predict hepatitis B virus (HBV) related fibrosis owing to the significant limitations of liver biopsy. Both serum biomarkers and imaging techniques have shown promising results and may improve the evaluation of liver fibrosis. However, most of the previous studies focused on the diagnostic effects of various imaging techniques on fibrosis in all chronic liver diseases. AIM: To compare the performance of common imaging methods and serum biomarkers for prediction of significant fibrosis caused only by HBV infection.
METHODS: A systematic review was conducted on the records available in PubMed, EMBASE, and the Cochrane Library electronic databases until December 2018. We systematically assessed the effectiveness of two serum biomarkers and three imagine techniques in predicting significant fibrosis solely caused by HBV infection. The serum biomarkers included aspartate aminotransferase-to-platelet ratio index (APRI) and fibrosis index based on the 4 factors (FIB-4). The three imaging techniques included acoustic radiation force impulse (ARFI), FibroScan, and magnetic resonance elastography (MRE). Three parameters, the area under the summary receiver operating characteristic curve (AUSROC), the summary diagnostic odds ratio, and the summary sensitivity and specificity, were used to examine the accuracy of all tests for liver fibrosis.
RESULTS: Out of 2831 articles evaluated for eligibility, 204 satisfied the predetermined inclusion criteria for this current meta-analysis. Eventually, our final data contained 81 studies. The AUSROCs of serum biomarkers of APRI and FIB-4 were both 0.75. For imaging techniques (ARFI, FibroScan, and MRE), the areas were 0.89, 0.83, and 0.97, respectively. The heterogeneities of ARFI and FibroScan were statistically significant (I 2 > 50%). The publication bias was not observed in any of the serum biomarkers or imaging methods.
CONCLUSION: These five methods have attained an acceptable level of diagnostic accuracy. Imaging techniques, MRE in particular, demonstrate significant advantages in accurately predicting HBV-related significant fibrosis, while serum biomarkers are admissible methods.

Core tip: Many researchers compared the diagnostic effects for liver fibrosis by new techniques within the domain of imaging techniques separately or focused on fibrosis in all chronic liver diseases. We perform a meta-analysis to compare the effectiveness of both some common imaging methods and serum biomarkers for prediction of significant fibrosis among hepatitis B virus (HBV)-monoinfected patients. The results reveal that imaging techniques have significant advantages in prediction of HBV-related significant fibrosis.

INTRODUCTION

Liver fibrosis is a consequence of the accumulation of extracellular matrix compone-nts, caused by persistent liver damage and consequent wound healing reaction[1,2]. It can develop to cirrhosis, even hepatic failure and hepatocellular carcinoma[3]. Approximately one-third of cirrhosis cases worldwide are caused by hepatitis B virus (HBV) infection[2]. It is estimated that about 450 million people are chronically infected with HBV around the world[1], resulting in 600000 to 1000000 deaths annually[4]. Therefore, the accurate diagnosis of the degree of liver fibrosis is essential for the management of chronic hepatitis B (CHB) patients and making clinical decision for doctors.

Liver biopsy is the current reference standard for the evaluation and staging of fibrosis, which is an invasive technique. However, it has several disadvantages, such as patients’ reluctance, pain, hemoperitoneum[5], intra- and interobserver variations, and sampling errors[6]. Therefore, several noninvasive methods, including serum biomarkers combining indices/scores and imaging techniques like magnetic resonance imaging (MRI), have been investigated for their potential in assessment of liver fibrosis.

There has been a dramatic increase in developing various noninvasive tests since 1991[7]. There are direct serum noninvasive tests such as hepascore and hyaluronic acid, indirect serum markers which consist of the combination of routine biochemical or haematological tests such as fibrosis index based on the 4 factors (FIB-4), aspartate aminotransferase-to-platelet ratio index (APRI), or Forns index and Fibrotest combined direct and indirect tests[8]. Among all of these tests, APRI and FIB-4, being extensively investigated in numerous studies, demonstrated the greatest potential.

On the other side, with the development of imaging technology, several imaging devices have been developed for the diagnosis and staging of liver fibrosis, which are truly noninvasive methods when compared with serum biomarkers. Historically, radiologists and clinicians have relied on the assessment of morphological changes associated with liver fibrosis. Other imaging methods depend on changes in physical properties that can be assessed by quantification techniques. These include mechanical properties, texture, T1ρ lengthening, perfusion, diffusion, and hepatocellular function[9]. Ultrasound (US)-based elastography techniques are portable, relatively inexpensive, fast to acquire, and do not require postprocessing[10]. In particular, 1D transient elastography, commercialized as FibroScan (Echosens, Paris, France), has been widely validated in clinical trials, adopted clinically, and used by clinicians at point of service[9]. Moreover, focal point shear-wave elastography, commercialized as acoustic radiation force impulse (ARFI) (Siemens Medical Solutions, Mountain View, CA), is mapped in 2D using US tracking pulses, and the resulting tissue displacement is measured[11]. It is integrated in a conventional US machine that can be performed with conventional US probes during an abdominal US scan[12,13]. In addition, many MRI techniques for imaging of liver fibrosis are being developed. They typically cover a larger liver volume than US elastography techni-ques, which may reduce sampling variability and technical failure rate.

These methods have exhibited promising results and may improve the manage-ment of liver fibrosis. However, most previous studies compared the diagnostic effects of these new imaging techniques alone and focused on fibrosis in all chronic liver diseases, which might underestimate or overestimate the role of the biomarkers. Therefore, we performed a meta-analysis to compare the pooled performance of some common imaging methods and serum biomarkers for prediction of significant fibrosis among HBV-monoinfected patients.

MATERIALS AND METHODS

Search strategy and selection criteria

Online database search was performed on EMBASE, PubMed, and the Cochrane Library (01/2008-12/2018) for the following terms: Diagnosis test, liver stiffness measurement, LSM, transient elastography, FibroScan, MR elastography, MRE, ARFI, acoustic radiation force impulse, point shear-wave elastography, imaging techniques, hepatitis B virus, HBV, chronic hepatitis B, CHB, and fibrosis. Other studies were identified by a manual search for referenced studies or review articles. EndNote X9 software was used to manage the references.

Studies should be included if they conformed all of the following five criteria: (a) The study evaluated the performance of the APRI and/or ARFI and/or FIB-4 and/or FibroScan and/or MRE for the prediction of fibrosis in HBV infected patients. Studies on patients with other etiologies of liver diseases were also included if data for HBV-infected patients could be independently extracted. Special populations of HBV-infected patients [e.g., HBV/HIV, HBV/ hepatitis C virus (HCV), or HBV/ hepatitis D virus (HDV) coinfection] were excluded; (b) Liver biopsy was used to diagnose liver fibrosis as a golden standard; (c) Data could be extracted to construct at least one 2 × 2 table of test performance, based on some cutoff points of the five biomarkers for the significant fibrosis stage; (d) The study assessed the diagnostic accuracy for fibrosis stage F ≥ 2 according to METAVIR or a comparable staging system; and (e) The study included at least 50 patients. Studies of smaller sample sizes were excluded due to concerns on their applicability.

Data extraction and quality assessment

Two reviewers (WWS and XYX) evaluated study eligibility, extracted data from the study, and graded the study quality independently. Any disagreements between the reviewers were resolved by detailed discussions together with a third reviewer (LHB). The parameters in our literature search included author, year of publication, region, age, patient gender , body mass index (BMI), histological scoring system, number of patients, average length of liver specimen, time interval between biopsy and other diagnostic tests, diagnostic method, prevalence of the fibrosis stage, as well as cutoff values to identify the fibrosis stage[14].

XYX and ZQM independently appraised the quality of included studies using the quality assessment of diagnostic accuracy studies questionnaire (QUADAS-2)[15]. It could estimate the internal and external validity of diagnostic accuracy of studies used in systematic reviews.

Statistical analysis and data synthesis

We extracted and tabulated the data in a series of 2 × 2 tables that included sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) at each threshold value. The primary target was to identify significant fibrosis, defined by METAVIR[16], Batts and Ludwig[17], Scheuer[18], and Chinese Hospital system[19], for stages F2 to F4, and Ishak[20] for stages F3 to F6. This gauge was chosen because significant fibrosis is often considered a threshold for the initiation of antiviral therapy[21]. The method of diagnostic test accuracy was examined in order to provide clinically meaningful results. We examined the area under the summary receiver operating characteristic curve (AUSROC) curve, the summary diagnostic odds ratio (DOR), and the summary sensitivity and specificity to further examine the accuracy of all tests for liver fibrosis.

The heterogeneity of the results between studies was assessed statistically using the Cochran-Q and the quantity I2. I2 value can describe the percentage of total variation across studies that is attributable to heterogeneity rather than chance[22]. If there was significant heterogeneity, a meta-regression was conducted to explore the covariates that may induce heterogeneity, according to the following predefined characteristics: (1) Study design (prospective or retrospective); (2) Length of liver specimen (≥ 10 mm, ≥ 15 mm, ≥ 20 mm, or not); (3) Liver biopsy scoring system; (4) Number of centers (single or multicenter); (5) Sample size; (6) Publication year (2014-2018 or 2008-2013); (7) Time interval between biopsy and tests (same day or not); (8) Prevalence of signifi-cant fibrosis; (9) BMI; and (10) Location of study.

The potential publication bias was assessed using the Deeks funnel plots[23]. All analyses were performed with Meta-Disc software (v. 1.4) and Stata15.0.

RESULT

Search results

The study selection process is shown in Figure 1. A total of 2831 studies were searched by the described search strategies, of which 2627 were excluded following title and abstract screening. Of those, 81 papers were included in the review following full-text screening (Text S5); 47 studies were related to the APRI[24-70], 11 to the ARFI[29,42,69-77], 32 to the FIB-4[24,25,27,30,35,36,39,40,43-46,48-55,57,58,60,62,63,65-69,78-79], 29 to the FibroScan[28,38,42,52,53,58,60,61,71,73,80-98], and 6 to the MRE[99-104].

Figure 1

Flow diagram of article selection.

Characteristics of the included studies

In the 47 studies of APRI, there were 13725 patients (median age, 37 years; 68.9% of males). The total prevalence of significant fibrosis was 55.9% (range: 16.7%-83.7%). Most studies were conducted in Asia (42 studies). Three studies were from Europe, one from North America, and another from Africa. We did not restrict study participants to any age, but only one study was conducted on children and adolescents exclusively[59]. Twenty-nine studies used the METAVIR score, ten used the Scheuer score, six utilized the Ishak score, and two used the Chinese Hospital System. There were 11179 patients (median age, 37 years; 68.7% of males) in 32 studies evaluating the performance of FIB-4. The total prevalence of significant fibrosis was 57.5% (range: 16.7%-83.7%). Twenty-nine studies were conducted in Asia, two in Europe, and one in North America. All studies included adults only. Seventeen studies used the METAVIR score, nine used the Scheuer score, four utilized the Ishak score, and two used the Chinese Hospital System.

A total of 1527 patients (median age, 36 years; 67.5% of males) were contained in 11 studies on ARFI. The total prevalence of F2 was 54.2% (range: 23.9%-74.3%). Ten studies were conducted in Asia and one in Germany. All the subjects were adults. Six studies used the METAVIR score, two used the Chinese Hospital System, and one used each of the Scheuer score, the Ishak score, and the Batts and Ludwig score.

In the 29 FibroScan studies, a total of 5035 patients (median age, 39 years; 71.4% of males) were included. The overall prevalence of significant fibrosis was 49.4% (range: 14.8%-85%). Twenty-one studies were from Asia, six from Europe, one from Africa, and another from North America. Twenty studies used the METAVIR score, five utilized the Batts and Ludwig score, three utilized the Scheuer score, and one used the Chinese Hospital System.

There were 1228 patients (median age, 49 years; 71.0% of males) used to evaluate the performance of MRE in six studies. The total prevalence of significant fibrosis was 61.4% (range: 45.2%-77.4%). One MRE study was from America and the other five were conducted in Asia. The results of methodological quality assessment according to the QUADAS-2 scale are described for all studies (Figure 2, Text S5).

Figure 2

Summary of methodological quality of studies according to QUADAS-2.

Performance of serum biomarkers for prediction of significant liver fibrosis

In the 47 studies evaluating the APRI, the area under the ROC curve (AUROC) ranged from 0.54 to 0.87. The AUSROC and the pooled sensitivity, specificity, and DOR are presented in Table 1 and Figure 3. The Cochran-Q was 1199.4 and I2 > 50%, which indicated significant heterogeneity across the included studies (P < 0.001). Meta-regression analysis was used to explore the heterogeneity of the APRI accuracy for predicting significant fibrosis, which was affected by BMI and prevalence of signifi-cant fibrosis with regard to sensitivity and only prevalence of significant fibrosis with regard to specificity (Figure S1). In the 32 studies evaluating the FIB-4 for the prediction of fibrosis stage of F2, the AUROC ranged from 0.56 to 0.85. The AUSROC was 0.75 (0.71-0.78) (Figure 3). The summary sensitivity, specificity, and DOR are shown in Table 1. The score of Cochran-Q was 488.9 (P < 0.05) and heterogeneity was statistically significant. The heterogeneity of FIB-4 accuracy was mainly affected by prevalence of significant fibrosis with regard to sensitivity, and prevalence of significant fibrosis and study design with regard to specificity according to the meta-regression analysis (Figure S2). The summary sensitivity and specificity of serum indicators by BMI and prevalence of significant fibrosis are listed in Tables 2 and 3.

Table 1

Meta-analysis of hepatitis B-related significant fibrosis

	SEN (95%CI)	SPE (95%CI)	DOR (95%CI)
APRI	0.69 (0.63-0.73)	0.71 (0.66-0.75)	5 (4-6)
FIB-4	0.62 (0.57-0.67)	0.75 (0.71-0.78)	5 (4-5)
ARFI	0.77 (0.70-0.83)	0.87 (0.81-0.92)	23 (15-38)
FibroScan	0.72 (0.68-0.76)	0.82 (0.77-0.86)	12 (9-16)
MRE	0.94 (0.91-0.96)	0.96 (0.93-0.97)	348 (185-656)

SEN: Sensitivity; SPE: Specificity; DOR: Diagnostic odds ratio; CI: Confidence interval; APRI: Aminotransferase-to-platelet ratio index; FIB-4: Fibrosis index based on the 4 factors; ARFI: Acoustic radiation force impulse.

Figure 3

Meta-analysis of hepatitis B-related significant fibrosis. A: Summary receiver operating characteristic (SROC) curve of the aminotransferase-to-platelet ratio index; B: SROC curve of the fibrosis index based on the 4 factors; C: SROC curve of the acoustic radiation force impulse; D: SROC curve of the FibroScan; E: SROC curve of the MRE. SROC: Summary receiver operating characteristic; APRI: Aminotransferase-to-platelet ratio index; FIB-4: Fibrosis index based on the 4 factors; ARFI: Acoustic radiation force impulse.

Table 2

Subgroup analysis of body mass index in prediction of significant fibrosis

Diagnostic test	BMI	Number of studies (n)	AUSROC (95%CI)	SEN (95%CI)	SPE (95%CI)
APRI	Overweight	7 (985)	0.78 (0.74-0.81)	0.78 (0.59-0.90)	0.71 (0.61-0.79)
	Normal	13 (2598)	0.75 (0.71-0.79)	0.63 (0.55-0.71)	0.77 (0.68-0.84)
	NA	27 (10142)	0.74 (0.70-0.77)	0.69 (0.62-0.74)	0.68 (0.61-0.74)
FIB-4	Overweight	5 (717)	0.76 (0.72-0.79)	0.58 (0.47-0.70)	0.80 (0.67-0.89)
	Normal	6 (1367)	0.70 (0.66-0.74)	0.58 (0.46-0.69)	0.75 (0.60-0.85)
	NA	21 (9095)	0.75 (0.71-0.79)	0.64 (0.57-0.70)	0.74 (0.69-0.78)
ARFI	Overweight	5 (679)	0.85 (0.82-0.88)	0.76 (0.64-0.85)	0.86 (0.80-0.91)
	Normala	3 (481)	0.91 (0.88-0.95)	0.84 (0.80-0.88)	0.76 (0.64-0.85)
	NAa	3 (367)	0.93 (0.85 – 0.99)	0.72 (0.63-0.80)	0.95 (0.91-0.97)
Fibroscan	Overweight	12 (1927)	0.81 (0.77-0.84)	0.68 (0.62-0.74)	0.85 (0.76-0.91)
	Normal	11 (2103)	0.83 (0.79-0.86)	0.73 (0.66-0.79)	0.82 (0.73-0.88)
	NA	6 (1005)	0.85 (0.81-0.88)	0.79 (0.69-0.87)	0.78 (0.72-0.83)

The results were calculated with Meta-Disc software (v. 1.4). BMI: Body mass index; NA: Not assessable; SEN: Sensitivity; SPE: Specificity; CI: Confidence interval; AUSROC: The area under the summary receiver operating characteristic curve; APRI: Aminotransferase-to-platelet ratio index; FIB-4: Fibrosis index based on the 4 factors; ARFI: Acoustic radiation force impulse.

Table 3

Subgroup analysis of prevalence of F2 in prediction of significant fibrosis

Diagnostic test	F2%	Number of studies (n)	AUSROC (95%CI)	SEN (95%CI)	SPE (95%CI)
APRI	F2% < 50	16 (4566)	0.72 (0.68-0.76)	0.68 (0.59-0.76)	0.66 (0.57-0.75)
	F2% ≥ 50	31 (9159)	0.77 (0.73-0.81)	0.69 (0.62-0.75)	0.73 (0.68-0.78)
FIB-4	F2% < 50	9 (3234)	0.75 (0.71-0.79)	0.60 (0.47-0.72)	0.76 (0.68-0.83)
	F2% ≥ 50	23 (7945)	0.75 (0.71-0.78)	0.63 (0.57-0.68)	0.74 (0.69-0.78)
ARFI	F2% < 50	4 (554)	0.89 (0.85-0.91)	0.73 (0.62-0.82)	0.93 (0.68-0.97)
	F2% ≥ 50	7 (973)	0.88 (0.85-0.91)	0.80 (0.72-0.87)	0.82 (0.74-0.88)
Fibroscan	F2% < 50	12 (2272)	0.85 (0.81-0.88)	0.77 (0.70-0.83)	0.80 (0.71-0.86)
	F2% ≥ 50	17 (2763)	0.82 (0.78-0.85)	0.69 (0.64-0.74)	0.84 (0.78-0.88)

SEN: Sensitivity; SPE: Specificity; CI: Confidence interval; AUSROC: The area under the summary receiver operating characteristic curve; APRI: Aminotransferase-to-platelet ratio index; FIB-4: Fibrosis index based on the 4 factors; ARFI: Acoustic radiation force impulse.

Performance of imaging techniques for prediction of liver significant fibrosis

The AUROC ranged from 0.72 to 0.96 in 11 studies evaluating the ARFI. When combined, the AUSROC was 0.89 (0.86-0.91). The summary DOR was 23 (95%CI: 15-38), and the score of Cochran-Q was 27.8 and I2 > 50%, indicating significant heterogeneity across the included studies (P < 0.05). In meta-regression analysis, we found that the causes of the heterogeneity were publishing year and sample size for sensitivity, and prevalence of significant fibrosis and BMI for specificity (Figure S3). In the 29 studies evaluating the FibroScan, the AUROC ranged from 0.61 to 0.94. The AUSROC and the pooled sensitivity, specificity, and DOR are also shown in Table 1 and Figure 3. There was significant heterogeneity across the included studies (I2 > 50%, P < 0.001). The heterogeneity of the FibroScan was mainly affected by prevalence of significant fibrosis and BMI with regard to sensitivity or specificity (Figure S4). In the six studies evaluating the MRE for the prediction of significant fibrosis, the AUROC ranged from 0.98 to 0.99. The AUSROC was 0.97 (0.96-0.98) (Figure 3). The summary sensitivity, specificity, and DOR are listed in Table 1, and the Cochran-Q and I2 of this measure were 0.44 (P > 0.05) and 0, respectively. The result of the heterogeneity was statistically insignificant. The subgroup analysis of imaging methods by BMI and prevalence of significant fibrosis are listed in Tables 2 and 3, respectively.

Publication bias

Funnel plots of these markers are illustrated in Figure S5. Symmetry was noted in the funnel plots and publication bias was not observed in any of the serum biomarkers or the imaging methods.

DISCUSSION

Liver fibrosis progression is common in people infected with HBV. Patients with significant fibrosis should be considered for clinical therapy, which can potentially reduce and even reverse complications[105]. Considering the risks and limitations of biopsy, researchers should make persistent efforts in researching some noninvasive methods to more conveniently and securely identify patients with significant fibrosis. FIB-4 and APRI are such noninvasive serum biomarkers that they have been gaining increasing acceptance in clinical practice. FibroScan, ARFI, and MRE gradually reveal an advantage with the development of US-based or MRI elastography techniques. These methods may reduce the demand for liver biopsy and also help to monitor the efficacy of treatments[62].

Many scholars have researched the diagnostic effect of APRI and FIB-4 and usually compared them with other new methods. In addition, the World Health Organization (WHO) recommends APRI as the preferred noninvasive test to assess significant fibrosis or cirrhosis and FIB-4 to detect fibrosis stages ≥ F3, considering lower cost, routinely available methods, and untrained staff[106]. Therefore, among all serum biomarkers, only APRI and FIB-4 were singled out in our analysis. Based on the results from an increasing number of studies, the difference between APRI and FIB-4 is not significant in relative diagnostic outcomes, although some articles described that FIB-4 had a higher diagnostic accuracy for cirrhosis when compared with APRI [107]. If only APRI and FIB-4 are available, APRI based on AST, ULN, and platelet count is suggested to diagnose liver significant fibrosis considering resource-limited settings.

With the development of imaging technology, US imaging and MRI devices have been explored to diagnose and stage liver fibrosis, which are truly noninvasive methods. In our meta-analysis, the AUSROCs of imaging techniques are generally bigger than those of APRI and FIB-4. Of these tests, transient elastography performed with FibroScan (Echosens, Paris) has been evaluated widely and has a good performance in predicting cirrhosis, which is corroborated by the Guidelines Development Group. They considered it the most useful test for the assessment of cirrhosis in middle-income countries[106]. In addition, FibroScan acquires information from a much larger portion of the liver tissue compared to biopsy, and therefore, the risk of sampling error is significantly lowered. However, its performance is mediocre for diagnosing moderate liver fibrosis, and the AUSROC ranked in the middle of the five methods. Morbid obesity (BMI > 30), narrow intercostal spaces, and food intake may diminish the accuracy of FibroScan[107]. Because of the limitation of US technology, some researchers advocated that the accuracy of ARFI is similar if not superior to FibroScan for the diagnosis of liver fibrosis[108,109]. Our meta-analysis reveals that ARFI has a higher diagnostic accuracy than FibroScan in identifying HBV-related significant fibrosis. In contrast to FibroScan that has a fixed region of interest box at a fixed insertion depth, ARFI could be performed at variable depths. Furthermore, it can also be performed in obese patients and patients with ascites and integrated in a conventional US system[13]. MRE, the only MRI index, had the best result in prediction of significant fibrosis. The AUSROC of MRE reaches the standard of “best” and even closes to “1”[110]. And the summary sensitivity and specificity have reached 94% and 96%, respectively. It typically covers larger liver tissue than US elastography techniques, which may reduce sampling variability. While MRE has been incorporated into MRI devices, it might be more expensive and require more operator training and expertise. Therefore, it is not widely applied at present, and the number of research articles is limited.

Meta-regression method is a convenient and reliable way for heterogeneity screening. In our study, BMI, prevalence of significant fibrosis, sample size, publishing year, and study design provide heterogeneity in summarizing test results. This is consistent with a range of previous studies[38,111,112]. Some researchers suggested that BMI may reduce the accuracy of ultrasonic technique for detecting significant fibrosis[9,38]. Our results reveal that BMI may affect predicting outcomes of either serum biomarkers or imaging techniques. Imaging technology shows a better predictive effect in the normal weight group, while serum indicators reflect their superiorities in overweight HBV patients (Table 2). The high proportion of fibrosis stage ≥ 2 will reduce the accuracy of evaluating fibrosis by all methods (Table 3). Furthermore, publishing year and sample size are the main contributors to the heterogeneity of ARFI. The AUSROC of ARFI was 0.89 when we excluded the only one article before 2014, the AUSROC of which was 0.75. Therefore, the technology of ARFI, a new diagnostic method, is progressing gradually, and the evaluation effect is increasingly improved. Besides, the AUSROCs in subgroups of sample size between 100 and 300 were much more than others in most cases (Text S6). Although other methods showed no significant differences in the AUSROC among subgroups of sample size, there was a general pattern that the areas were larger in the middle while smaller on both sides (Text S6). This phenomenon is more prominent if the group was changed into three subgroups (n < 100, 100 ≤ n ≤ 300, and n > 300) (Table 4). A few years ago, our team observed that when the sample size was larger than 150, the diagnostic accuracy of predicting fibrosis was noticeably enhanced[111]. Therefore, the sample size of 150-300 will be beneficial to the accuracy of diagnostic tests.

Table 4

Subgroup analysis of sample size in prediction of significant fibrosis

Diagnostic test	Sample size	Number of studies (N)	AUSROC (95%CI)	SEN (95%CI)	SPE (95%CI)
APRI	n < 100	8 (588)	0.73 (0.69-0.77)	0.67 (0.60-0.74)	0.68 (0.61-0.75)
	100 ≤ n < 300	27 (5286)	0.77 (0.73-0.80)	0.69 (0.62-0.76)	0.72 (0.65-0.78)
	n ≥ 300	12 (7851)	0.73 (0.69-0.77)	0.67 (0.57-0.76)	0.69 (0.58-0.78)
FIB-4	n < 100	6 (411)	0.77 (0.73-0.80)	0.64 (0.54-0.73)	0.77 (0.67-0.85)
	100 ≤ n < 300	14 (2917)	0.76 (0.72-0.80)	0.62 (0.53-0.71)	0.77 (0.69-0.83)
	n ≥ 300	12 (7851)	0.74 (0.70-0.77)	0.62 (0.53-0.69)	0.74 (0.67-0.79)
ARFI	n < 100	2 (173)	0.72/0.75	0.83/0.50	0.65/0.90
	100 ≤ n < 300	9 (1354)	0.90 (0.87-0.92)	0.78 (0.71-0.84)	0.88 (0.82-0.93)
	n ≥ 300	0	-	-	-
Fibroscan	n < 100	6 (445)	0.80 (0.76-0.83)	0.70 (0.57-0.80)	0.76 (0.68-0.83)
	100 ≤ n < 300	20 (3659)	0.84 (0.81-0.87)	0.74 (0.69-0.79)	0.83 (0.77-0.87)
	n ≥ 300a	3 (931)	0.82 (0.69-0.95)	0.62 (0.57-0.66)	0.88 (0.84-0.90)

The results were calculated with Meta-Disc software (v. 1.4). SEN: Sensitivity; SPE: Specificity; CI: Confidence interval; AUSROC: The area under the summary receiver operating characteristic curve; APRI: Aminotransferase-to-platelet ratio index; FIB-4: Fibrosis index based on the 4 factors; ARFI: Acoustic radiation force impulse.

However, there are several limitations in our systematic review. First of all, the cut-off value was not considered due to the absence of threshold effect in all tests (Text S7). Second, the ALT levels were not considered because only a limited number of studies calculated the results by the subgroup of ALT. Third, only the studies published in English and Chinese languages were included. Lastly, other methods, though we have retrieved them, were not considered in this meta-analysis, because the number of articles was not adequate.

In conclusion, these five methods have attained an acceptable level of diagnostic accuracy in our meta-analysis. Imaging techniques are generally better than serum biomarkers in prediction of HBV-related liver significant fibrosis. MRE is a promising indicator and will surprise us with the development of global economy and popu-larization of technology. If condition allows, ARFI could be a better choice, than FibroScan, for assessment of liver fibrosis in HBV-monoinfected patients with obesity or ascites. The prediction effects of serum markers are generally admissible. Their low cost and easiness in operation make them attractive especially in areas with limited resources and access to imaging technology.

ARTICLE HIGHLIGHTS

Research background

Liver fibrosis can develop to cirrhosis and even hepatic failure and hepatocellular carcinoma. Approximately one-third of cirrhosis cases worldwide are caused by HBV infection. Therefore, the accurate diagnosis of the extent of liver fibrosis for chronic hepatitis B patients is essential. Many serum biomarkers combining indices/scores and imaging or magnetic resonance imaging techniques have been undergoing dramatic development because of several drawbacks of liver biopsy. These methods have promising results and may improve the management of liver fibrosis. However, most of the previous studies compared the diagnostic effects of these new techniques within the domain of imaging techniques or focused on fibrosis in all chronic liver diseases, which might misestimate the role of these biomarkers. Therefore, we performed a meta-analysis to compare the pooled performance of some common imaging methods with serum biomarkers for prediction of significant fibrosis among HBV-monoinfected patients.

Research motivation

We aimed to assess the accuracy of diagnostic tests for predicting significant fibrosis among patients monoinfected with HBV. Most studies are centered on the domain of imaging techniques and serum biomarkers separately or focused on fibrosis in all chronic liver diseases. Therefore, the key points are that data for HBV infected patients could be extracted independently and we want to integrate and compare the performance of methods of different fields. With the development of medicine and technology, more innovative methods could be invented in the near future. It will provide a boost to precision medicine that chooses a more appropriate and effective method to evaluate liver fibrosis for different populations of patients.

Research objectives

We aimed to compare the pooled performance of some common imaging methods with serum biomarkers for prediction of significant fibrosis among HBV-monoinfected patients. Some serum biomarkers have been calculated and imaging techniques have been developed for the diagnosis of liver fibrosis, respectively. Integrating and comparing the performance of methods of different fields could provide a basis for future research and clinical application and a boost to precision medicine.

Research methods

We examined the areas under the summary receiver operating characteristic curves, the summary diagnostic odds ratios, as well as the summary sensitivities and specificities to further examine the accuracy of all tests for liver fibrosis. Then, we assessed the heterogeneity between studies using the Cochran-Q and I2. And the Deeks funnel plots were used to assess publication bias. Meta-regression was conducted to further accurately explore the covariates that may induce heterogeneity.

Research results

Our meta-analysis revealed that ARFI showed a higher diagnostic accuracy than FibroScan in identifying HBV-related significant fibrosis. Furthermore, it can also be performed in obese patients and in patients with ascites and be integrated in a conventional ultrasound system. MRE, the only MRI index, had the best result in prediction of significant fibrosis. The area under the SROC curve of MRE reaches the standard of “best” and even closes to “1”. The performances of APRI and FIB-4 are poorer than imaging techniques. However, the cut-off value should be considered in future studies.

Research perspectives

Imaging techniques are better than serum biomarkers in prediction of HBV-related liver significant fibrosis in general. MRE is a promising indicator and other serum biomarkers are general.

ACKNOWLEDGEMENTS

We thank Prof. Li-Jin Xu for comments that greatly improved the manuscript. And we also thank all participants for their contribution of time to this study.