Elsevier

The Lancet Oncology

Volume 16, Issue 7, July 2015, Pages 804-815
The Lancet Oncology

Articles
A serum microRNA classifier for early detection of hepatocellular carcinoma: a multicentre, retrospective, longitudinal biomarker identification study with a nested case-control study

https://doi.org/10.1016/S1470-2045(15)00048-0Get rights and content

Summary

Background

The ability of circulating microRNAs (miRNAs) to detect preclinical hepatocellular carcinoma has not yet been reported. We aimed to identify and assess a serum miRNA combination that could detect the presence of clinical and preclinical hepatocellular carcinoma in at-risk patients.

Methods

We did a three-stage study that included healthy controls, inactive HBsAg carriers, individuals with chronic hepatitis B, individuals with hepatitis B-induced liver cirrhosis, and patients with diagnosed hepatocellular carcinoma from four hospitals in China. We used array analysis and quantitative PCR to identify 19 candidate serum miRNAs that were increased in six patients with hepatocellular carcinoma compared with eight control patients with chronic hepatitis B. Using a training cohort of patients with hepatocellular carcinoma and controls, we built a serum miRNA classifier to detect hepatocellular carcinoma. We then validated the classifiers' ability in two independent cohorts of patients and controls. We also established the classifiers' ability to predict preclinical hepatocellular carcinoma in a nested case-control study with sera prospectively collected from patients with hepatocellular carcinoma before clinical diagnosis and from matched individuals with hepatitis B who did not develop cancer from the same surveillance programme. We used the sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) to evaluate diagnostic performance, and compared the miRNA classifier with α-fetoprotein at a cutoff of 20 ng/mL (AFP20).

Findings

Between Aug 1, 2009, and Aug 31, 2013, we recruited 257 participants to the training cohort, and 352 and 139 participants to the two independent validation cohorts. In the third validation cohort, 27 patients with hepatocellular carcinoma and 135 matched controls were included in the nested case-control study, which ran from Aug 1, 2009, to Aug 31, 2014. We identified a miRNA classifier (Cmi) containing seven differentially expressed miRNAs (miR-29a, miR-29c, miR-133a, miR-143, miR-145, miR-192, and miR-505) that could detect hepatocellular carcinoma. Cmi showed higher accuracy than AFP20 to distinguish individuals with hepatocellular carcinoma from controls in the validation cohorts, but not in the training cohort (AUC 0·826 [95% CI 0·771–0·880] vs 0·814 [0·756–0·872], p=0·72 in the training cohort; 0·817 [0·769–0·865] vs 0·709 [0·653–0·765], p=0·00076 in validation cohort 1; and 0·884 [0·818–0·951] vs 0·796 [0·706–0·886], p=0·042 for validation cohort 2). In all four cohorts, Cmi had higher sensitivity (range 70·4–85·7%) than did AFP20 (40·7–69·4%) to detect hepatocellular carcinoma at the time of diagnosis, whereas its specificity (80·0–91·1%) was similar to that of AFP20 (84·9–100%). In the nested case-control study, sensitivity of Cmi to detect hepatocellular carcinoma was 29·6% (eight of 27 cases) 12 months before clinical diagnosis, 48·1% (n=13) 9 months before clinical diagnosis, 48·1% (n=13) 6 months before clinical diagnosis, and 55·6% (n=15) 3 months before clinical diagnosis, whereas sensitivity of AFP20 was only 7·4% (n=2), 11·1% (n=3), 18·5% (n=5), and 22·2% (n=6) at the corresponding timepoints (p=0·036, p=0·0030, p=0·021, p=0·012, respectively). Cmi had a larger AUC than did AFP20 to identify small-size (AUC 0·833 [0·782–0·883] vs 0·727 [0·664–0·792], p=0·0018) and early-stage (AUC 0·824 [0·781–0·868] vs 0·754 [0·702–0·806], p=0·015) hepatocellular carcinoma and could also detect α-fetoprotein-negative (AUC 0·825 [0·779–0·871]) hepatocellular carcinoma.

Interpretation

Cmi is a potential biomarker for hepatocellular carcinoma, and can identify small-size, early-stage, and α-fetoprotein-negative hepatocellular carcinoma in patients at risk. The miRNA classifier could be valuable to detect preclinical hepatocellular carcinoma, providing patients with a chance of curative resection and longer survival.

Funding

National Key Basic Research Program, National Science and Technology Major Project, National Natural Science Foundation of China.

Introduction

Hepatocellular carcinoma accounts for more than 90% of primary liver cancers. It is the sixth most common malignancy worldwide and the third leading cause of cancer death. Patients with hepatocellular carcinoma that is detected at an early stage have a good chance of a successful curative operation, and 5-year overall survival can reach 50–74%.1, 2 However, 5-year overall survival for hepatocellular carcinoma is still lower than 10% worldwide.1 Liver cirrhosis of any cause is the major risk factor for hepatocellular carcinoma, and chronic hepatitis B virus (HBV) infection is the leading cause of hepatic cirrhosis.3 Therefore, the discovery of effective and reliable strategies to monitor at-risk populations (eg, those with chronic hepatitis and liver cirrhosis) and to detect hepatocellular carcinoma at an early stage could improve the survival of patients with this disease.

Methods for the early detection of hepatocellular carcinoma include serological tests and imaging examinations. Serological tests that have been investigated, alone or in combination, including those for α-fetoprotein, des-gamma-carboxy prothrombin (DCP), and AFP-L3. However, available data have shown that these tests are suboptimum for routine surveillance of hepatocellular carcinoma.4 α-fetoprotein is the most widely used serological biomarker for hepatocellular carcinoma worldwide.4, 5 However, not all hepatocellular carcinomas secrete α-fetoprotein, and so performance of α-fetoprotein to detect disease can be inadequate. Serum α-fetoprotein assessment at a cutoff of 20 ng/mL has a sensitivity of 40–65% for clinically diagnosed hepatocellular carcinoma and of only 14–40% for preclinical disease.6 The sensitivity and specificity of assessment of DCP for clinically diagnosed hepatocellular carcinoma are 28–89% and 87–96%, respectively, which are similar to those of AFP-L3.6

Guidelines from the European Association for the Study of the Liver (EASL) and American Association for the Study of Liver Diseases (AASLD) recommended hepatic ultrasound, a low-cost and widely available imaging method, for surveillance of hepatocellular carcinoma.4, 7 However, the interpretation of ultrasound is dependent on tumour size and the skills of the operator, and can be difficult in patients who are obese or have underlying cirrhosis.8 A meta-analysis showed that hepatic ultrasound had a sensitivity of only 63% to detect early-stage hepatocellular carcinoma, as defined by Milan criteria (one nodule <5 cm or three nodules each <3 cm in diameter without gross vascular invasion).9 New biomarkers with high accuracy to complement hepatic ultrasound are greatly needed to improve detection and surveillance of hepatocellular carcinoma.

MicroRNAs (miRNAs) are a class of endogenous small non-coding RNAs that regulate gene expression post-transcriptionally. The dysregulation of miRNAs is thought to have an important role in oncogenesis. Emerging evidence shows that circulating miRNAs could be diagnostic biomarkers for diseases, including cancer.10 Although a few investigations11, 12, 13, 14, 15, 16 have aimed to identify the circulating miRNAs that distinguish individuals with hepatocellular carcinoma from those who are cancer free, most studies have had limitations, including too few miRNAs examined, a small study population, no at-risk controls, and no independent validation. So far, only three studies have done high-throughput screening and verified the diagnostic performance of circulating miRNAs in independent cohorts with more than 100 patients with hepatocellular carcinoma.11, 12, 13 All three studies were case-control studies of diagnosed hepatocellular carcinoma, but none assessed whether circulating miRNAs could detect preclinical early-stage disease. 11, 12, 13

We did a large-scale, multicentre validation to identify a serum miRNA classifier that could differentiate individuals with hepatocellular carcinoma from healthy individuals and at-risk controls. We then did a nested case-control study, using prospectively collected sera from patients with hepatocellular carcinoma and at-risk controls, to assess the value of the classifier to identify preclinical hepatocellular carcinoma.

Section snippets

Study design and participants

In total, we collected 1416 serum samples from five groups of participants: healthy controls, inactive HBsAg carriers, patients with chronic hepatitis B, patients with HBV-induced liver cirrhosis, and patients with diagnosed hepatocellular carcinoma. Participants were enrolled from Third Affiliated Hospital of Sun Yat-sen University (SYSU), the Cancer Centre of SYSU, Sun Yat-sen Memorial Hospital of SYSU, and Guangdong Provincial Hospital of Chinese Medicine (figure 1). The sera used in the

Results

We collected 1416 serum samples from the five groups of participants (figure 1, table 1, appendix). For each group, the age and sex were well matched, and the concentrations of serum ALT and α-fetoprotein did not significantly differ among those in the training cohort and validation cohorts 1 and 2. BCLC stages differed among patients with hepatocellular carcinoma in the training and validation cohorts, with just under half the participants in the training cohort, just over half of participants

Discussion

In this study, the levels of serum miR-29a, miR-29c, miR-133a, miR-143, miR-145, miR-192, and miR-505 were significantly increased in patients with hepatocellular carcinoma compared with healthy controls, inactive HBsAg carriers, individuals with chronic hepatitis B, and those with HBV-induced liver cirrhosis. The classifier (Cmi) composed of these seven miRNAs had significantly higher sensitivity than did α-fetoprotein to distinguish individuals with hepatocellular carcinoma from the combined

References (38)

  • J Bruix et al.

    Management of hepatocellular carcinoma: an update

    Hepatology

    (2011)
  • CJ Chen et al.

    Early diagnosis of hepatocellular carcinoma by multiple microRNAs: validity, efficacy, and cost-effectiveness

    J Clin Oncol

    (2011)
  • A Singal et al.

    Meta-analysis: surveillance with ultrasound for early-stage hepatocellular carcinoma in patients with cirrhosis

    Aliment Pharmacol Ther

    (2009)
  • N Kosaka et al.

    Circulating microRNA in body fluid: a new potential biomarker for cancer diagnosis and prognosis

    Cancer Sci

    (2010)
  • LM Li et al.

    Serum microRNA profiles serve as novel biomarkers for HBV infection and diagnosis of HBV-positive hepatocarcinoma

    Cancer Res

    (2010)
  • J Zhou et al.

    Plasma microRNA panel to diagnose hepatitis B virus-related hepatocellular carcinoma

    J Clin Oncol

    (2011)
  • Y Tan et al.

    A serum microRNA panel as potential biomarkers for hepatocellular carcinoma related with hepatitis B virus

    PLoS One

    (2014)
  • J Xu et al.

    Circulating microRNAs, miR-21, miR-122, and miR-223, in patients with hepatocellular carcinoma or chronic hepatitis

    Mol Carcinog

    (2011)
  • ZQ Zhang et al.

    Serum microRNA 143 and microRNA 215 as potential biomarkers for the diagnosis of chronic hepatitis and hepatocellular carcinoma

    Diagn Pathol

    (2014)
  • Cited by (0)

    Authors contributed equally.

    View full text