This article is mentioned in:

Introduction

Liver cancer is the sixth-most commonly diagnosed cancer and the fourth leading cause of cancer death worldwide (1). Hepatocellular carcinoma (HCC) accounts for 90% of primary liver cancers, and hepatitis B virus (HBV) and hepatitis C virus (HCV) infection as well as alcohol consumption and non-alcoholic steatohepatitis (NASH) are known risk factors (2-4).

The early detection of HCC by regular surveillance may lead to curative treatment and improve the prognosis (5). Several methods developed for the diagnosis of HCC, including the evaluation of serum markers, ultrasonography (US), computed tomography (CT), and magnetic resonance imaging (MRI), have been tested clinically. Among these methods, US is simple with a low invasiveness, but its accuracy depends on the skill of the examiner. Contrast-enhanced CT and MRI are useful for the diagnosis but are invasive. The most widely used markers are α-fetoprotein (AFP) and des-γ carboxy prothrombin (DCP), serum proteins that are elevated in HCC. Although routine screening offers the best chance for early tumor detection, the reported sensitivities and specificities of elevated serum AFP and DCP levels vary significantly (6-12).

In 2009, the highly sensitive Lens culinaris agglutinin-reactive fraction of AFP (hs-AFP-L3) assay was developed, and AFP-L3 measurement became possible, even in cases with AFP <20 ng/ml (13). However, biomarkers that contribute to hepatocarcinogenesis during the long-term observation of patients with chronic liver disease are still unclear. We previously examined the clinical utility of hs-AFP-L3 in patients with chronic liver disease (9). Seven years have passed since that study, so we examined the clinical utility of hs-AFP-L3 in a long-term observation.

Materials and methods

Study population

Frozen serum samples were collected from 117 patients with chronic liver disease without HCC who visited our hospital between December 1, 2006, and March 31, 2011. The analysis was performed on 106 patients, excluding those who had been treated for HCC, those who had been under observation for less than 12 months, and those who were taking warfarin tablets (Fig. 1). In most patients with chronic hepatitis, liver imaging with US was performed every 6 to 12 months, and in patients with cirrhosis, CT, MRI, or US was performed every 3 to 6 months. The definitive diagnosis of HCC that occurred during follow-up was made by interventional radiology CT (IVR-CT). HBV was defined as Hepatitis B surface antigen (HBsAg) positivity. HCV was defined as anti-HCV antibody positivity.

Figure 1 Schematic representation of the patient selection process.

Measurement of serum AFP and hs-AFP-L3

AFP and hs-AFP-L3 levels were measured using the cryopreserved serum. Hs-AFP-L3 was measured by microchip capillary electrophoresis and a liquid-phase binding assay on a µ-TASWako i30 automatic analyzer (Wako Pure Chemical Industries, Ltd.) (13). When hs-AFP-L3 was not detectable, the percentage of hs-AFP-L3 was defined as 0%.

Statistical analysis

First, we considered the correlation between the hs-AFP-L3 level and other clinical data. In addition, the presence of HCC as of March 31, 2018, was confirmed, and the ability to predict hepatocarcinogenesis using liver tumor markers was compared using a receiver operating characteristic (ROC) curve. Finally, we investigated the factors contributing to hepatocarcinogenesis using univariate and multivariate analyses. The cut-off value was set to an optimal value using the Youden index (sensitivity + specificity-1) (14).

Statistical analyses were performed using the SPSS statistical software program, version 25 (IBM Corp.). Categorical data were compared using the chi-squared test and Fisher's exact test, as appropriate. Continuous variables were analyzed using the Mann-Whitney U test. The correlation coefficient was tested using Spearman's rank correlation coefficient or Pearson's correlation coefficient. The Kaplan-Meier method was used to estimate cumulative incidence rate of HCC, and its distribution curves were compared using the log-rank test. P-values of <0.05 were considered to indicate statistical significance. Factors contributing to hepatocarcinogenesis were determined using Cox's proportional hazards model with forward selection using P<0.10 as a cut-off for inclusion in the model.

Results

Clinical feature of patients

The clinical characteristics of the population analyzed are shown (Table I). The causes of chronic liver disease were HBV in 23 cases, HCV in 60 cases, and non-HBV and non-HCV in 23 cases, of which 17 were liver cirrhosis. The median observation period was 88 months (15-132 months). The AFP value in the analysis subject population was 27.1±119.3 ng/ml, and the hs-AFP-L3 value was 2.9±5.3%.

Table I Clinical features of patients with benign liver disease (n=106).

Table I

Clinical features of patients with benign liver disease (n=106).

Variable	Value
Agea, years	57.5 (11-82)
Sex (male/female), n	38/68
CH/LC, n	89/17
Etiology (HBV/HCV/NBNC), n	23/60/23
Child-Pugh class (A/B/C/unknown), n	83/3/2/18
AFPb, ng/ml	27.1±119.3
hs-AFP-L3b, %	2.9±5.3
DCPb, mAU/ml	18±8
Platelet countb, x104/µl	16.4±7.2
ALTb, U/l	79±121
Total bilirubinb, mg/dl	1.0±0.6
Albuminb, g/dl	4.2±0.5
Hyaluronic acidb, ng/ml	175.7±444.6
Observation perioda, months	88 (15-132)

[i] ^aMedian (min-max).

[ii] ^bMean ± SD. CH, chronic hepatitis; LC, liver cirrhosis; HBV, hepatitis B virus; HCV, hepatitis C virus; NBNC, HBV(-) and HCV(-); AFP, α-fetoprotein; hs-AFP-L3, highly sensitive Lens culinaris agglutinin-reactive fraction of AFP; DCP, des-γ-carboxy prothrombin; ALT, alanine transaminase.

Correlation between hs-AFP-L3 and clinical data

We confirmed the correlation between hs-AFP-L3 and other clinical data. Hs-AFP-L3 showed a positive correlation with the age, alanine transaminase (ALT), hyaluronic acid, and AFP and a negative correlation with the platelet count and albumin (Table IIA). In a study of 90 patients with AFP <20 ng/ml, hs-AFP-L3 showed a positive correlation with the age, hyaluronic acid, and AFP and a negative correlation with the platelet count (Table IIB).

Table II Association between highly sensitive Lens culinaris agglutinin-reactive fraction of AFP and clinical data.

Table II

Association between highly sensitive Lens culinaris agglutinin-reactive fraction of AFP and clinical data.

A, Correlation with highly sensitive Lens culinaris agglutinin-reactive fraction of AFP (n=106)
Variable	Agea	ALTa	Hyaluronic acida	AFPa	DCPa	Pltb	Albumina	T-Bila
Correlation coefficient	0.232	0.262	0.479	0.724	-0.080	-0.256	-0.354	0.163
P-value	0.017	0.007	<0.001	<0.001	0.418	0.008	<0.001	0.096
n	106	106	106	106	106	106	101	105
B, Correlation with highly sensitive Lens culinaris agglutinin-reactive fraction of AFP (n=90; AFP <20 ng/ml)
Variable	Agea	ALTa	Hyaluronic acida	AFPa	DCPa	Pltb	Albumina	T-Bila
Correlation coefficient	0.269	0.164	0.352	0.666	-0.159	-0.282	-0.211	-0.011
P-value	0.010	0.123	0.001	<0.001	0.134	0.007	0.052	0.922
n	90	90	90	90	90	90	85	89

[i] ^aSpearman's rank correlation coefficient.

[ii] ^bPearson's correlation coefficient. AFP, α-fetoprotein; ALT, alanine transaminase; DCP, des-γ-carboxy prothrombin; Plt, platelet count; T-Bil, total bilirubin.

Cumulative incidence of HCC

The presence of HCC as of March 31, 2018, was confirmed, and 17 out of 106 patients (16.0%) were found to have developed HCC. Cumulative incidence of HCC development was 10.5% at 5 years and 19.6% at 10 years (Fig. 2). The clinical characteristics of hepatocarcinogenesis cases are shown in Tables III and IV. In the background comparison between the non-carcinogenic group and the carcinogenic group, the age (P=0.009), AFP (P<0.001), hs-AFP-L3 (P<0.001), platelet count (P<0.001), ALT (P=0.018), albumin (P=0.023), and hyaluronic acid (P<0.001) differed significantly (Table IV).

Figure 2 Cumulative incidence rate of HCC in the present study population (n=106). HCC, hepatocellular carcinoma.

Table III Characterization of 17 patients with benign liver disease who developed HCC.

Table III

Characterization of 17 patients with benign liver disease who developed HCC.

Case no.	Age, years	Sex	CH/LC	Etiology	AFP, ng/ml	hs-AFP-L3, %	DCP, mAU/ml	ALT, U/l	Months until HCC detection
1	60	Male	LC	HCV	28.5	9.6	32	65	8
2	70	Female	LC	NBNC	10.9	8.4	15	39	12
3	65	Female	CH	HCV	3.3	<0.5	26	82	20
4	69	Male	CH	HCV	4.7	6.4	13	52	22
5	70	Female	CH	HCV	8.3	7.0	15	48	31
6	56	Female	CH	HCV	46.9	4.5	5	98	43
7	73	Female	CH	HCV	11.4	7.3	12	54	49
8	59	Female	CH	HCV	32.0	3.7	13	116	53
9	52	Male	CH	HBV	23.6	7.1	10	489	56
10	73	Female	LC	HCV	9.6	6.6	13	71	57
11	70	Female	CH	HBV	24.3	8.4	13	30	62
12	53	Male	CH	HCV	6.4	5.2	29	145	64
13	66	Female	CH	HCV	10.0	5.8	19	122	65
14	57	Female	CH	HCV	7.8	7.9	45	52	81
15	53	Male	LC	NBNC	3.7	<0.5	53	22	82
16	62	Male	CH	HCV	23.0	7.5	35	72	82
17	66	Female	CH	HCV	576.0	3.1	8	126	87

[i] HCC, hepatocellular carcinoma; CH, chronic hepatitis; LC, liver cirrhosis; HBV, hepatitis B virus; HCV, hepatitis C virus; NBNC, HBV(-) and HCV(-); AFP, α-fetoprotein; hs-AFP-L3, highly sensitive Lens culinaris agglutinin-reactive fraction of AFP; DCP, des-γ-carboxy prothrombin; ALT, alanine transaminase.

Table IV Clinical features of patients in the non-carcinogenic and carcinogenic groups.

Table IV

Clinical features of patients in the non-carcinogenic and carcinogenic groups.

Variable	Non-carcinogenic group (n=89)	Carcinogenic group (n=17)	P-value
Agea, years	54±15	63±7	0.009b
Sex (male/female), n	32/57	6/11	0.958c
CH/LC, n	76/13	13/4	0.468d
Etiology (HBV/HCV/NBNC), n	21/47/21	2/13/2	0.197c
Child-Pugh class (A/B/C/unknown), n	68/2/2/17	15/1/0/1	0.428c
AFPa, ng/ml	22.9±116.1	48.8±136.4	<0.001b
hs-AFP-L3a, %	2.4±5.5	5.8±2.8	<0.001b
DCPa, mAU/ml	17±6	21±14	0.928b
Platelet counta, x104/µl	17.5±7.1	10.7±4.9	<0.001b
ALTa, U/l	75±123	99±107	0.018b
Total bilirubina, mg/dl	1.0±0.6	0.9±0.4	0.585b
Albumina, g/dl	4.2±0.5	4.0±0.5	0.023b
Hyaluronic acida, ng/ml	172.6±483.2	191.6±110.9	<0.001b
Observation perioda, months	84±32	98±18	0.060b

[i] ^aMean ± SD.

[ii] ^bMann-Whitney U test.

[iii] ^cχ² test.

[iv] ^dFisher's exact test. CH, chronic hepatitis; LC, liver cirrhosis; HBV, hepatitis B virus; HCV, hepatitis C virus; NBNC, HBV(-) and HCV(-); AFP, α-fetoprotein; hs-AFP-L3, highly sensitive Lens culinaris agglutinin-reactive fraction of AFP; DCP, des-γ-carboxy prothrombin; ALT, alanine transaminase.

Predictive ability for hepatocarcinogenesis

On comparing the predictive ability for hepatocarcinogenesis using an ROC curve, the cut-off value of hs-AFP-L3 was 3.5%, and the sensitivity, specificity, and the area under the ROC curve (AUC) were 82.4, 73.0%, and 0.800, respectively. Similarly, the cut-off value of AFP was 6.3 ng/ml, and the sensitivity, specificity, and AUC were 82.4, 75.3%, and 0.833, respectively. The cut-off value of DCP was 25 mAU/ml, and the sensitivity, specificity, and AUC were 35.3, 93.3%, and 0.507, respectively (Fig. 3A). The ability of hs-AFP-L3 and AFP to predict HCC development was higher than that of DCP. In the analysis of 90 patients with AFP <20 ng/ml, the cut-off value of hs-AFP-L3 was 4.9%, and the sensitivity, specificity, and AUC were 80.0, 85.0%, and 0.812, respectively. Similarly, the cut-off value of AFP was 4.6 ng/ml, and the sensitivity, specificity, and AUC were 80.0, 75.0%, and 0.844, respectively. The cut-off value of DCP was 25 mAU/ml, and the sensitivity, specificity, and AUC were 40.0%, 95.0%, and 0.616, respectively (Fig. 3B). In patients with AFP <20 ng/ml, the ability of hs-AFP-L3 and AFP to predict HCC development was higher than that of DCP.

Figure 3 Comparison of the predictive ability for hepatocarcinogenesis using an ROC curve. (A) ROC curve of hs-AFP-L3, AFP and DCP in the analysis target population (n=106). (B) ROC curve of hs-AFP-L3, AFP and DCP in patients with AFP <20 ng/ml (n=90). ROC, receiver operating characteristic; AFP, α-fetoprotein; hs-AFP-L3, highly sensitive Lens culinaris agglutinin-reactive fraction of AFP; DCP, des-γ carboxy prothrombin; AUC, area under the curve.

Factors contributing to hepatocarcinogenesis

An examination of the factors contributing to hepatocarcinogenesis according to a univariate analysis showed that age ≥55 years old (P=0.016), platelet count ≤13.1x104/µl (P=0.001), hyaluronic acid ≥80.8 ng/ml (P<0.001), ALT ≥47 U/l (P=0.008), AFP ≥6.3 ng/ml (P<0.001), hs-AFP-L3 ≥3.5% (P<0.001), DCP ≥25 mAU/ml (P=0.002) were significant factors. In the multivariate analysis, the platelet count ≤13.1x104/µl (hazard ratio [HR]=4.966, 95% confidence interval [CI] 1.597-15.437, P=0.006) and hs-AFP-L3 ≥3.5% (HR=5.450, 95% CI 1.522-19.512, P=0.009) were extracted as significant factors contributing to hepatocarcinogenesis (Table V). In addition, in patients with AFP <20 ng/ml, the univariate analysis showed that age ≥64 years old (P=0.005), liver cirrhosis (P=0.047), platelet count ≤13.1x104/µl (P=0.002), hyaluronic acid ≥67.7 ng/ml (P=0.010), ALT ≥47 U/l (P=0.037), AFP ≥4.6 ng/ml (P=0.002), hs-AFP-L3 ≥4.9% (P<0.001), DCP ≥25 mAU/ml (P=0.003) were significant factors. In the multivariate analysis, hs-AFP-L3 ≥4.9% (HR=11.608, 95% CI 2.422-55.629, P=0.002) and DCP ≥25 mAU/ml (HR=3.936, 95% CI 1.088-14.231, P=0.037) were extracted as significant factors contributing to hepatocarcinogenesis (Table VI).

Table V Factors contributing to hepatocarcinogenesis (n=106).

Table V

Factors contributing to hepatocarcinogenesis (n=106).

	Univariate analysisa	Multivariate analysisb
Variable	P-value	Hazard ratio	95% CI	P-value
Age (<55 vs. ≥55 years)	0.016
Sex (female vs. male)	0.978
Background liver (CH vs. LC)	0.128
Total bilirubin (<0.6 vs. ≥0.6 mg/dl)	0.212
Albumin (>4.4 vs. ≤4.4 g/dl)	0.113
Platelet count (>13.1 vs. ≤13.1x104/µl)	0.001	4.966	1.597-15.437	0.006
Hyaluronic acid (<80.8 vs. ≥80.8 ng/ml)	<0.001
ALT (<47 vs. ≥47 U/l)	0.008	3.019	0.841-10.836	0.090
AFP (<6.3 vs. ≥6.3 ng/ml)	<0.001
hs-AFP-L3 (<3.5 vs. ≥3.5%)	<0.001	5.450	1.522-19.512	0.009
DCP (<25 vs. ≥25 mAU/ml)	0.002

[i] ^aLog-rank test.

[ii] ^bCox proportional hazards model. ALT, alanine transaminase; AFP, α-fetoprotein; hs-AFP-L3, highly sensitive Lens culinaris agglutinin-reactive fraction of AFP; DCP, des-γ-carboxy prothrombin; CH, chronic hepatitis; LC, liver cirrhosis; CI, confidence interval.

Table VI Factors contributing to hepatocarcinogenesis (n=90; AFP <20 ng/ml).

Table VI

Factors contributing to hepatocarcinogenesis (n=90; AFP <20 ng/ml).

	Univariate analysisa	Multivariate analysisb
Variable	P-value	Hazard ratio	95% CI	P-value
Age (<64 vs. ≥64 years)	0.005
Sex (female vs. male)	0.844
Background liver (CH vs. LC)	0.047
Total bilirubin (<1.2 vs. ≥1.2 mg/dl)	0.218
Albumin (>4.4 vs. ≤4.4 g/dl)	0.119
Platelet count (>13.1 vs. ≤13.1x104/µl)	0.002
Hyaluronic acid (<67.7 vs. ≥67.7 ng/ml)	0.010
ALT (<47 vs. ≥47 U/l)	0.037
AFP (<4.6 vs. ≥4.6 ng/ml)	0.002
hs-AFP-L3 (<4.9 vs. ≥4.9%)	<0.001	11.608	2.422-55.629	0.002
DCP (<25 vs. ≥25 mAU/ml)	0.003	3.936	1.088-14.231	0.037

[i] ^aLog-rank test.

[ii] ^bCox proportional hazards model. ALT, alanine transaminase; AFP, α-fetoprotein; hs-AFP-L3, highly sensitive Lens culinaris agglutinin-reactive fraction of AFP; DCP, des-γ-carboxy prothrombin; CH, chronic hepatitis; LC, liver cirrhosis; CI, confidence interval.

Comparison of cumulative incidence of HCC by hs-AFP-L3

The cumulative incidence of HCC was significantly higher in patients with hs-AFP-L3 ≥3.5% than in those with hs-AFP-L3 <3.5% (24.7% at 5 years, 40.0% at 10 years vs. 1.6% at 5 years, 6.8% at 10 years, P<0.001) (Fig. 4A). In addition, in cases of AFP <20 ng/ml, the cumulative incidence of HCC was significantly higher in patients with hs-AFP-L3 ≥4.9% than in those with hs-AFP-L3 <4.9% (24.6% at 5 years, 39.7% at 10 years vs. 1.5% at 5 years, 3.6% at 10 years, P<0.001) (Fig. 4B).

Figure 4 Comparison of the cumulative incidence rate of HCC by hs-AFP-L3. (A) In the analysis target population (n=106), the cumulative incidence of HCC was significantly higher in patients with hs-AFP-L3 ≥3.5% than in those with hs-AFP-L3 <3.5% (log-rank test P<0.001). (B) In patients with AFP <20 ng/ml (n=90), the cumulative incidence of HCC was significantly higher in patients with hs-AFP-L3 ≥4.9% than in those with hs-AFP-L3 <4.9% (log-rank test P<0.001). HCC, hepatocellular carcinoma; AFP, α-fetoprotein; hs-AFP-L3, highly sensitive Lens culinaris agglutinin-reactive fraction of AFP.

Discussion

Recent advances in imaging technology have enabled the early detection of HCC (15-17), so low AFP cases often result in a diagnosis of hepatocarcinogenesis. In the present study, it was found that 10 of 17 patients with AFP <20 ng/ml (58.8%) had hepatocarcinogenic on long-term follow-up. In our previous study, we found that hs-AFP-L3 was a useful marker for predicting hepatocarcinogenesis, but the observation period averaged 32.8 months, and the examination was made by a univariate analysis (9). In the present study, after a long-term observation, we performed a multivariate analysis of factors associated with the development of HCC and revealed that hs-AFP-L3 was the best predictive marker for hepatocarcinogenesis. It was found to be particularly useful in cases with AFP <20 ng/ml.

AFP is a glycoprotein with a molecular weight of 67 kDa that was first reported in human fetal serum by Bergstrand and Czar (18) in 1956. AFP is elevated in patients with HCC but also in the active phases of chronic hepatitis and cirrhosis as well as in AFP-producing tumors other than liver cancer (19). The sugar chains of AFP differ depending on the producing cell, and the L3 fraction is specific for HCC in terms of its affinity for lentil lectin (20-24). However, measurement of AFP-L3 has not always been reliable for serum samples with a low total AFP concentration, as determined by conventional lectin affinity system (LiBASys) (25). The highly sensitive AFP-L3 measurement method uses an on-chip electrokinetic reaction and separation by affinity electrophoresis (micro-total analysis system; µ-TAS) (26). This system has enabled the accurate measurement of AFP-L3 at very low AFP concentrations.

In a previous report on the prediction of hepatocarcinogenesis, Kumada et al (22) conducted a study of 104 patients with hepatocarcinogenesis and 104 controls matched by propensity scores in HCC surveillance involving 2,830 patients with chronic liver disease. One year before the diagnosis of HCC, the cut-off value of hs-AFP-L3 was 7%, and the sensitivity and specificity were 34.3 and 74.7%, respectively. Similarly, with cut-off values of AFP 20 ng/ml and DCP 40 mAU/ml, the respective sensitivity was 35.0 and 12.1%, and the respective specificity was 86.4 and 93.9% (22). In the present study, the best cut-off values of hs-AFP-L3, AFP, and DCP for predicting HCC were 3.5%, 6.3 ng/ml, and 25 mAU/ml, respectively, with respective sensitivities of 82.4, 82.4 and 35.3% and respective specificities of 73.0, 75.3, and 93.3%. Similarly, in patients with AFP <20 ng/ml, the best cut-off values of hs-AFP-L3, AFP, and DCP were 4.9%, 4.6 ng/ml and 25 mAU/ml, respective, with respective sensitivities of 80.0, 80.0 and 40.0% and respective specificities of 85.0, 75.0 and 95.0%. In this way, our findings differed from those of previous reports. This discrepancy is attributed to differences in the study design, as the previous report used stored sera collected annually for three years before the diagnosis of HCC. In addition, the median observation period in our study was 88 months (15 to 132 months), which was longer than in the previous study.

In this study, 11 out of 17 cases of hepatocarcinogenesis developed HCC more than 4 years after the test, and 9 out of 11 cases (81.8%) had hs-AFP-L3 ≥3.5%. The doubling time of HCC is reported to be 100 days, and it theoretically takes about 9 years for a 10-µm HCC to become a 10-mm lesion, which can be detected by diagnostic imaging (27). In other words, the involvement of hs-AFP-L3 in hepatocarcinogenesis several years later may indicate the presence of minute HCC.

However, in hepatitis C patients treated with interferon, the hepatocarcinogenesis rate decreases after achieving a sustained virologic response (SVR), and AFP values after antiviral therapy are known to be independent predictors of hepatocarcinogenesis (28,29). In addition, it has been reported that serum Wisteria floribunda agglutinin positive Mac-2 binding protein (WFA+M2BP), a liver fibrosis marker that has recently been clinically applied, becomes a risk factor for hepatocarcinogenesis after achieving an SVR of hepatitis C (30,31). Direct-acting antivirals (DAAs) have been developed for HCV, and virus elimination by DAA reportedly suppresses hepatocarcinogenesis (32,33). The usefulness of measuring the hs-AFP-L3 value before and after DAA therapy is unclear at present and needs to be clarified in the future.

The study is limited by its retrospective nature and the small number of cases.

In conclusion, hs-AFP-L3 is a useful marker for predicting hepatocarcinogenesis in the long-term observation of patients with chronic liver disease.

Acknowledgements

The authors would like to thank Ms. Hiromi Eguchi, Ms. Yuko Nakamura and Ms. Eriko Koreeda (all Digestive and Lifestyle Diseases, Department of Human and Environmental Sciences, Kagoshima University Graduate School of Medical and Dental Sciences, Kagoshima, Japan) for their technical assistance and data management.

Funding

No funding was received.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Authors' contributions

KT, SM, KO, KK and AI conceived the study. KT, SM, KO, OT, AT, SI, HS, KK and SK contributed the acquisition of data. KT and SM confirm the authenticity of all the raw data. KT and SM analyzed the data and prepared the manuscript. KT, SM and AI reviewed the manuscript. KT, SM, KO, OT, AT, SI, HS, KK, SK and AI have been involved in revising the manuscript critically for important intellectual content and agreed to be accountable for all aspects of the work. All authors read and approved the final manuscript.

Ethics approval and consent to participate

The study protocol conformed to the ethical guidelines of the Declaration of Helsinki and was approved by the Kagoshima University Ethics Committee on Epidemiological Studies (approval no. 180162; Kagoshima, Japan). Written informed consent was obtained from all patients for the use of stored serum samples.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References