Background and aims The associations between long-term risk of hepatocellular carcinoma (HCC) and spontaneous seroclearance of HBV e antigen (HBeAg), HBV DNA and HBV surface antigen (HBsAg) have never been examined by a prospective study using serially measured seromarkers. This study aimed to assess the importance of spontaneous HBeAg, HBV DNA and HBsAg seroclearance in the prediction of HCC risk.
Methods This study included 2946 HBsAg seropositive individuals who were seronegative for antibodies against HCV and free of liver cirrhosis. Serial serum samples collected at study entry and follow-up health examinations were tested for HBeAg, HBV DNA and HBsAg. Cox proportional hazards models were used to calculate the HRs of developing HCC after seroclearance of HBV markers.
Results The HR (95% CI) of developing HCC after seroclearance of HBeAg, HBV DNA and HBsAg during follow-up was 0.63 (0.38 to 1.05), 0.24 (0.11 to 0.57) and 0.18 (0.09 to 0.38), respectively, after adjustment for age, gender and serum level of alanine aminotransferase at study entry. High HBV DNA levels at the seroclearance of HBeAg (mean±SD, 4.35±1.64 log10 IU/mL) may explain the non-significant association between HBeAg seroclearance and HCC risk. Among HBeAg seronegative participants with detectable serum HBV DNA at study entry, the lifetime (30–75-years-old) cumulative incidence of HCC was 4.0%, 6.6% and 14.2%, respectively, for those with seroclearance of both HBV DNA and HBsAg, seroclearance of HBV DNA only, and seroclearance of neither.
Conclusions Spontaneous seroclearance of HBV DNA and HBsAg are important predictors of reduced HCC risk.
- HEPATOCELLULAR CARCINOMA
- HEPATITIS B
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
There is a dose–response relation between serum HBV DNA level and hepatocellular carcinoma (HCC) risk.
Significant decrease in serum HBV DNA level lowers HCC risk among individuals with high viral loads.
Baseline seroclearance of HBV e antigen (HBeAg), HBV DNA or HBV surface antigen (HBsAg) is associated with a decreased HCC risk.
What are the new findings?
In this long-term follow-up study with serially measurements of HBV seromarkers, HBV DNA seroclearance is essential for significantly lowering HCC risk.
HBsAg seroclearance is essential for ensuring complete immune clearance of HBV.
HBeAg seroclearance is not as important as previously described because individuals with newly-serocleared HBeAg still have high serum HBV DNA levels. Decreased HCC risk associated with HBeAg seroclearance observed in previous studies may be due to its association with subsequent seroclearance of HBV DNA rather than HBeAg seroclearance itself.
How might it impact on clinical practice in the foreseeable future?
HBeAg seroclearance itself may not be adequate to reduce HCC risk, but will lead to the seroclearance of HBV DNA and HBsAg, which should be the goal of antiviral treatment.
Lowering serum HBV DNA levels will maximise chances of HBsAg seroclearance and significantly decrease risk of HCC.
Chronic HBV infection is a global public health problem due to its widespread geographical prevalence and potential to cause serious clinical sequelae such as cirrhosis, hepatic decompensation and hepatocellular carcinoma (HCC).1–4 In highly endemic countries such as Taiwan, where hepatitis B transmission is mostly perinatal, the prevalence of chronic hepatitis B infection in the adult population is at least 8% and as high as 15%–20%.4 ,5
The natural course of chronic HBV infection is typically divided into three main phases and is characterised by interactions among the seromarkers ALT, HBV e antigen (HBeAg) and HBV surface antigen (HBsAg), with serum HBV DNA levels as the main driver of disease progression.3 ,6 ,7 Risk factors for HCC can be divided into host, viral and environmental factors.8 Viral factors can include genotype, viral mutants, co-infection with other diseases and, last, HBsAg, HBeAg serostatus and serum HBV DNA levels.8–11
Studies showed that the relative risk of developing HCC among patients who were HBsAg seropositive was much higher than uninfected individuals.12 In another study, HBeAg seropositivity was shown to be an important risk factor for HCC.13 ,14 In addition, the REVEAL study has shown that serum HBV DNA levels are the predominant drivers of disease progression, with increasing viral loads indicating higher risk for HCC, liver cirrhosis and death.7 ,11 ,15–18 Thus, serum HBV DNA levels are currently the main tool for monitoring clinical CHB.19–21 Last, HBsAg seroclearance, although very rare, is the most important milestone, representing immunity to HBV and an improved prognosis.22 ,23
Elucidating the determinants of HCC and finding ways to reduce HCC risk is one of the main goals of clinical research and antiviral therapy. Thus, the seroclearance of HBeAg, HBV DNA and HBsAg have all been listed as important treatment endpoints in hepatitis B therapy.19–21 Although the seroclearance of all three has been shown to individually reduce HCC risk, they have not been considered together in predicting risk for HCC. Moreover, previous studies of marker seroclearance and HCC risk have not fully taken into account that inherent differences in baseline profiles exist between those with and without seroclearance. Therefore, it is important to control for these differences when examining the effects of biomarker seroclearance.
A previous study that examined long-term trajectories of HBV DNA and HCC risk showed that spontaneous HBV DNA decrease over time reduces HCC incidence compared with individuals with persistently high viral loads.24 However, this study did not examine the seroclearance of other seromarkers, did not consider quantitative serum HBsAg levels and was not able to address individuals with viral loads <10 000 copies/mL.
In addition, previous studies have not considered the dynamic nature of the seroclearance of HBV markers, as well as the differing natural histories of each individual. Therefore, the aim of this study is to assess the importance of the seroclearance HBeAg, HBV DNA and HBsAg in the prediction of subsequent reduction in risk for HCC in a natural history cohort of previously untreated individuals with chronic hepatitis B infection. This study includes analysis of other seromarkers and takes into consideration differing individual natural histories, including all levels of HBV DNA, as well as HBsAg levels.
From 1991 to 1992, a total of 23 820 individuals aged 30–65 years were enrolled from seven townships in Taiwan in the Risk Evaluation of Viral Load Elevation and Associated Liver Disease/Cancer—Hepatitis B Virus (R.E.V.E.A.L.-HBV) study. Each participant consented to a questionnaire interview, physical examination, and blood collection for serological testing and biochemical assays at study entry and follow-up examinations. The consent also included follow-up through 30 June 2004, and medical record review and computerised data linkage with national health insurance, cancer registry and death certification databases until 31 December 2008.
Out of all enrolled individuals, 4155 participants who tested positive for HBsAg were included in this analysis. To examine the effect of biomarker seroclearance, only those participants with complete HBsAg seroclearance data, who were anti-HCV seronegative, did not have liver cirrhosis at study entry and who had at least two HBV DNA measurements were included, for a total of 2946 individuals. Cirrhosis at study entry was excluded by high-resolution real-time abdominal ultrasonography. This study was approved by the Institutional Review Board of the College of Public Health, National Taiwan University, Taipei, Taiwan.
Definition of seroclearance and ascertainment of HCC
Seroclearance of HBeAg and HBsAg was defined as the first instance at which an individual tested seronegative, and remained seronegative thereafter, for HBeAg and HBsAg. Further conversion to anti-HBe or anti-HBs was not determined. However, the anti-HBe seroconversion was found as high as 92% and 95% in a previous nested case-control study of the REVEAL-HBV cohort and a similar cohort, respectively.13 ,25 HBV DNA seroclearance was the first instance that an individual was undetectable and remained undetectable thereafter. Cases of HCC were detected through follow-up examinations, which included both ultrasound and AFP testing, and through computerised linkages with the National Cancer Registry and National Death Certification databases. Identified cases were confirmed through medical chart reviews by gastroenterologists according to the following criteria: histopathological confirmation; positive lesions detected by at least two different imaging techniques (such as abdominal ultrasonography, angiogram or CT); or positive lesions detected by one imaging technique combined with a serum α-fetoprotein level greater than 400 ng/mL.
Data collection and laboratory methods
All participants were interviewed with a structured questionnaire by trained public health nurses at enrolment. Information was collected on sociodemographic characteristics, family history of major diseases, as well as dietary and lifestyle habits. Blood collection and health examinations were performed at study entry and follow-ups every 6–12 months. Cirrhosis at study entry and follow-up examinations were determined using high-resolution real-time abdominal ultrasonography, and also confirmed through computerised linkages with National Health Insurance profiles until 30 June 2004. Serostatus of HBsAg, HBeAg, and anti-HCV and serum levels of HBV DNA and ALT were tested using commercial kits: HBsAg and HBeAg by radioimmunoassay (Abbott Laboratories, North Chicago, Illinois, USA), anti-HCV by enzyme immunoassay using second-generation test kits (Abbott Laboratories), ALT by serum chemistry autoanalyzer (model 736; Hitachi Co., Tokyo, Japan) using commercial reagents and serum HBV DNA levels by PCR (COBAS Amplicor, Roche Diagnostics, Indianapolis, Indianapolis, USA) for baseline samples and by real-time PCR (COBAS TaqMan, Roche Diagnostics) for follow-up samples. Serum HBsAg levels were quantified using the Roche Elecsys HBsAg II Quant assay. HBV genotype was determined in those with detectable serum levels of HBV DNA by melting curve analysis, and precore 1896 and basal core promoter 1762/1764 mutations were assayed in those with baseline HBV DNA >2000 IU/mL using direct DNA sequence analysis. Detailed laboratory methods have been previously described.10
Person-years of follow-up were calculated as the time from enrolment to the date of HCC diagnosis, death or 31 December 2008, whichever came first. Cox proportional hazards models were used to analyse the crude and multivariate-adjusted HRs (with 95% CIs) of HCC incidence. Since genotype data were only available in patients with detectable viral loads, subset analyses were performed using only patients with available genotype data. Lifetime cumulative incidence curves were also calculated with Cox proportional hazards models. Statistical significance was determined by two-tailed tests (p<0.05). Statistical analyses were performed with SAS software (V.9.2; SAS Institute, Cary, North Carolina, USA).
Among 2946 individuals, the majority were HBeAg negative (85%), men (64%), were non-smokers (66%), did not drink alcohol (88%) and had normal ALT <45 U/L (94%). During a total of 48 149.1 person-years of follow-up, there were 192 events of HBeAg seroclearance, 438 individuals who reached undetectable serum HBV DNA levels and 529 individuals who reached HBsAg seroclearance. Last, 153 cases of HCC occurred, for an incidence rate of 320.76 per 100 000 person-years (table 1). Among 2946 HBsAg seropositive and anti-HCV seronegative individuals, univariate analyses showed that baseline increasing age, male gender, cigarette smoking, alcohol consumption, history of diabetes, high baseline ALT levels, genotype C, precore and BCP mutations, HBeAg seropositivity, increasing HBV DNA and HBsAg levels, and subsequent HBsAg and HBV DNA seroclearance during follow-up were all significant predictors of HCC (table 1).
Incidence rates according to HBV marker seroclearance
To control for the different natural history profiles of each patient, individuals were first stratified by baseline HBV DNA levels, then classified according to whether or not they reached HBeAg, HBV DNA or HBsAg seroclearance during follow-up, as shown in figure 1. Seven individuals with HBV DNA seroclearance shortly after HBsAg seroclearance were not included for this analysis. Incidence rates of HCC varied greatly according to disease phase. Among those with HBV DNA levels >2000 IU/mL and who were HBeAg seropositive at study entry, reaching HBeAg, HBV DNA and HBsAg seroclearance during follow-up reduced incidence rates of HCC to zero, whereas those who remained HBeAg seropositive, HBV DNA detectable and HBsAg seropositive had incidence rates of HCC of 1225.91 per 100 000 person-years (figure 1). Patients who reached HBeAg seroclearance, but remained HBV DNA detectable and HBsAg seropositive still had incidence rates of HCC of 903.87 per 100 000 person-years (p=0.11).
Among HBeAg seronegatives with high viral loads at study entry, HBV DNA seroclearance greatly reduced incidence rates of HCC from 471.94 to 115.85 per 100 000 person-years (p=0.02). Among those who reached HBV DNA seroclearance, HBsAg seroclearance did not further significantly decrease HCC risk.
Among HBeAg seronegatives with low viral loads, those with persistently detectable HBV DNA had incidence rates of HCC of 126.78 per 100 000 person-years. Last, among those with undetectable viral loads at study entry, reaching HBsAg seroclearance did not further significantly reduce risk for HCC (figure 1).
A separate analysis was performed starting at the point of HBV DNA seroclearance among the 438 individuals who reached HBV DNA seroclearance during follow-up. Compared with those who did not yet reach HBsAg seroclearance, those who already cleared HBsAg had a multivariate HR (95% CI) of HCC of 0.34 (0.05 to 2.00).
Multivariate models for HCC prediction
To further examine the effect of each biomarker's seroclearance on HCC risk, three separate multivariate-adjusted subset models were examined. Examination of HBsAg seroclearance was limited to those with undetectable viral loads, and examination of HBV DNA was limited to HBeAg seronegatives. In this way, the seroclearance of each marker was only examined, conditioned on each individual's previous natural history. Among 748 individuals with undetectable viral loads, HBsAg seroclearance no longer significantly decreased risk for HCC (table 2). When further stratified among patients with low (<1000 IU/mL) HBsAg levels, HBsAg seroclearance at any time still did not show decreased risk for HCC (data not shown).
Among 1754 HBeAg seronegative individuals with detectable serum HBV DNA levels, reaching HBV DNA seroclearance was significantly associated with a reduced HCC risk, showing a multivariate-adjusted HR (95% CI) of 0.22 (0.08 to 0.63), compared with individuals with persistently detectable HBV DNA. Last, among 444 HBeAg seropositive individuals, HBeAg seroclearance was not significantly associated with HCC risk after adjusting for other baseline predictors (table 2). In a subset analysis of patients with complete data on precore and BCP mutations, reaching serum HBV DNA seroclearance was still associated significantly with a reduced HCC risk, showing an multivariate-adjusted HR (95% CI) of 0.25 (0.08 to 0.85). HBeAg seroclearance remained non-significantly associated with HCC risk, even after further adjustment for HBV mutations, showing an adjusted HR (95% CI) of 0.81 (0.45 to 1.45) (see online supplementary table 1).
In a subset analysis to further examine the role of baseline serum HBsAg levels, a significant association between HCC risk and elevated serum HBsAg levels (≥1000 IU/mL) was observed for those with undetectable serum HBV DNA levels, with an adjusted rate ratio (95% CI) of 3.86 (1.14 to 13.09). For those with low viral loads, the effect of HBsAg levels in stratifying HCC risk was not as strong. HBsAg levels were unable to stratify risk among those with high viral loads (table 3).
HRs of HCC according to disease phase
HRs of HCC were further examined according to disease phase, as shown in table 4. Among those who had undetectable HBV DNA at baseline or during follow-up, HBsAg seroclearance resulted in similar rates of HCC as those who remained HBsAg positive (p=0.37). However, HBeAg seronegative individuals who did not reach HBV DNA seroclearance did have higher rates of HCC, with an adjusted HR (95% CI) of 3.99 (1.78 to 8.99). The highest rates occurred among both groups who were HBeAg positive at baseline. The group with HBeAg seroclearance during follow-up had a HR (95% CI) of 15.13 (5.95 to 38.51), while those who were persistently HBeAg seropositive had an adjusted HR (95% CI) of 20.39 (8.56 to 48.57). Pairwise comparisons did not show any difference between these two groups (table 4 and figure 2). Cumulative lifetime incidences of HCC ranged from 4.0%, to 80.1% and 64% in the two groups who were HBeAg seropositive at baseline (p>0.05) (figure 2).
Previous studies have shown that HBeAg seroclearance, HBV DNA seroclearance and HBsAg seroclearance are able to reduce the risk for HCC. However, this study is the first to examine the effects of all three biomarkers on the risk for HCC, while taking into account the dynamic nature of each biomarker's seroclearance. The natural history of CHB consists of three major milestones, beginning with HBeAg seroclearance, followed by HBV DNA seroclearance, and ending with HBsAg seroclearance. Since the seroclearance of all three biomarkers is highly correlated, analysis of the seroclearance of each biomarker must be conditioned on the seroclearance of the previous biomarker. This way, collinearity is controlled for, as are the baseline profiles of patients. This study contains sufficient sample size to parse out all different outcomes and is also able to consider individuals with low viral loads, which previous studies did not consider.24
In multivariate analyses, only HBV DNA seroclearance significantly reduced risk of HCC, when compared with those who had not yet cleared HBV DNA, suggesting significant implications for the clinical management of chronic hepatitis B. Most cases occurred before HBV DNA even had a chance to clear. Although HBsAg seroclearance did reduce incidence rates of HCC, these reductions were not statistically significant, even when limited to individuals with low levels of serum HBsAg.
The decreased incidence of HCC among HBeAg seropositive individuals with HBsAg levels >1000 IU/mL is similar to a recent study which found less severe fibrosis in patients with high HBsAg levels.26 Although the exact mechanism for this observation is unclear, we hypothesise that these individuals may be in the phase of immune tolerance. Better outcomes are to be expected, as immune-mediated inflammation leading to fibrosis and liver injury is not seen until the immune clearance phase, during which lower HBsAg levels signal better immune control.27 HBsAg has been widely found in immunohistological staining of liver histology during the immune tolerance phase.28 Indeed, in our study, >84% of HBeAg positive individuals with HBsAg levels >1000 IU/mL had normal ALT, and >90% had viral loads >106 copies/mL, most likely placing them in the immune tolerance phase, where low rates of HCC are expected. Moreover, 75% of the 34 individuals with HBeAg and HBV DNA seroclearance were younger than 56.8 years at the end of follow-up; thus, no newly developed HCC cases may be due to an inadequate follow-up time for HCC to occur and a small sample size (figure 1).
Interestingly, our analyses also showed that reaching HBeAg seroclearance did not significantly reduce risk (figures 1 and 2). Although a previous landmark study showed that HBeAg seropositivity predicted high risk for HCC,13 this study has shown that perhaps HBeAg seroclearance is no longer sufficient for reducing HCC risk. HBeAg is highly correlated with and is a surrogate marker for HBV DNA replication.6 ,29 ,30 However, data on HBV DNA were not yet available at the time of the study by Yang et al, and thus their study could not adjust for the effect of HBV viral load.13
We posit that the higher risk of HCC previously seen in HBeAg seropositives may have been due to the inherently higher HBV DNA viral loads in these individuals compared with individuals who were already HBeAg seronegative at baseline with left-truncated HBV DNA levels. Many were already HBeAg seronegative for many years, so their mean viral loads would be lower than those with newly incident HBeAg seroclearance, as we had in this study. To verify this, indeed, over 65% of viral loads in baseline HBeAg seronegatives were <2000 IU/mL, while among newly HBeAg-serocleared individuals, 61.7% had viral loads ≥2000 IU/mL, with a mean of 4.35±1.64 log10 IU/mL. Such high viral loads still place individuals at high risk for developing HCC.11
Indeed, those with viral loads >2000 IU/mL at the point of HBeAg seroclearance had an adjusted rate ratio (95% CI) of future HCC of 3.47 (1.23 to 9.77), even after adjustment for genotype, precore and basal core promoter mutations, and despite having cleared HBeAg. Thus, the previous decreased HCC risk seen with HBeAg seronegatives may be a secondary association, reflecting the underlying decreasing viral replication rather than the effect of HBeAg seroconversion. This is supported by a recent study which found decreasing HBV DNA levels to be the most important predictor of HBeAg seroclearance.31
In studies of treatment-induced HBeAg seroclearance, however, patients with HBeAg seroclearance have been shown to have complete viral load suppression either before or simultaneously with HBeAg seroclearance, which was quite different from the observations of this natural history study.32 ,33 Patients with interferon-induced HBeAg seroclearance and viral load suppression had significantly better clinical outcomes when compared with untreated individuals.32 ,33 Thus, it seems that in treated patients, HBeAg seroclearance may still play an important role, as it would ensure a low viral load and subsequently lower risk for HCC.
Similarly, HBsAg seroclearance has also been found to be associated with reduced HCC risk.34 However, in this study, while decreases in incidence rates were seen for those with HBsAg seroclearance, the additional decrease was not significant when adjusted for HBV DNA levels, HBsAg levels and other predictors. In a separate subanalysis of 438 individuals with newly incident HBV DNA seroclearance, those with and without HBsAg seroclearance had incidence rates of HCC of 92.90 and 218.38 per 100 000 person-years, respectively. In multivariate analyses, this difference was not statistically significant, with a rate ratio (95% CI) of HCC of 0.34 (0.06 to 2.00), most likely due to insufficient cases. It must be noted, however, that the effect of HBsAg seroclearance is still inconclusive, as a simple sample size calculation determined we did not have sufficient power to detect differences between the two groups. A longer term follow-up of individuals who already have undetectable viral loads is needed in the future.
However, despite insignificant results for HBsAg seroclearance, we examined the issue using serum HBsAg levels instead. Serum HBsAg levels were able to predict HCC risk among those with undetectable viral loads, with the lowest HCC risk among those with low HBsAg levels (<1000 IU/mL), and the highest risk among those with high HBsAg levels. Similar to other studies, the effect of serum HBsAg levels was weak or insignificant among those with low and high viral loads.35 ,36 While the exact mechanisms underlying this observation are yet unclear, these and other results suggest possible different roles for seromarkers among different phases of disease, with serum HBsAg levels as a marker of host immune control.26 ,35 ,36 Nonetheless, a previous study showed that lower HBsAg levels had a very high probability of HBsAg loss, and this and another study by Lee et al have shown that HCC risk could be further refined among those with low viral loads by using quantitative serum HBsAg levels, with lower HBsAg levels conferring lower HCC risk.35 ,36
Therefore, HBsAg clearance should still be considered as an important endpoint for antiviral therapy, and is especially important among those with undetectable viral loads. Although the results from this study suggest that HBV DNA seroclearance will greatly reduce risk for HCC, these individuals may still be at risk for future reactivation. Thus, further HBsAg seroclearance is the best sign of immune clearance of HBV, ensures no chance of future reactivation, confers a reduced risk for HCC and negates the need for further antiviral therapy.37
The mechanisms by which HBV viral loads cause HCC are still unknown. Indirect pathways may be through immune-mediated ‘flares’ that cause inflammation, liver injury and fibrosis, later leading to HCC.38 Viral loads may also directly influence hepatocarcinogenesis through direct integration into the genomes of infected hepatocytes, which was found in >85% of HBV-related HCC cases.39 Other mechanisms may include the HBx protein, which may play a role in activating relevant cellular signalling pathways that regulate gene expression, response to DNA damage or cell proliferation.38 However, exact mechanisms remain to be further elucidated.
This study is the first to consider the seroclearance of all three biomarkers together while also taking into account their dynamic nature. According to the results of this and other R.E.V.E.A.L.-HBV studies, suppression of HBV viral loads will maximise chances of HBsAg seroclearance and will significantly decrease risk of HCC, confirming results from previous studies.7 ,24
There are some limitations to be noted. Due to the nature of observational cohort studies, left truncation prevented us from determining the exact points of HBeAg seroclearance for those who were HBeAg seronegative at baseline. Participants were treatment naive chronic HBV carriers. Therefore, the applicability of these results to patients undergoing treatment is unclear. Participants were between 30- and 65-years-old mostly infected perinatally with HBV genotype B and/or C. These results need to be clarified in younger carriers, those infected in adulthood or in individuals infected with other HBV genotypes. Data on basal core and precore mutations were only available at baseline in those with serum HBV DNA levels ≥2000 IU/mL. Thus, the true effect of virus mutations on HBeAg seroclearance could not be determined. Last, the number of HCC cases was often small among certain categories. There may have been insufficient power to detect differences among some groups, a limitation that can be expected in cohort studies. These findings should be further validated in studies with larger case numbers and longer follow-up time. The small number of HCC cases also limited the ability to examine the role of incident cirrhosis in some groups.
Last, the lack of risk reduction in those with HBeAg and HBsAg seroclearance may also be a result of delayed seroclearance, which has been shown as ineffective for reducing HCC risk. The median age at HBeAg, HBV DNA and HBsAg seroclearance in this cohort was 45, 53 and 55 years of age, respectively. Therefore, significant liver injury may have already occurred by the time seroclearance occurred. It is thus important for future studies to determine at what age seroclearance is most effective. However, as the number of cases in this study was very limited, it was not possible to truly examine the effect of delayed seroclearance.
In conclusion, HBV DNA seroclearance during the course of chronic hepatitis B is the most significant factor in reducing risk for future HCC and should be a major goal during the management of chronic hepatitis B patients.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online tables
Collaborators Other members of REVEAL-HBV Study Group are as follows: Chang-Gung Memorial Hospital and Chang-Gung University: YF Liaw. National Taiwan University Hospital: DS Chen, PJ Chen, CY Hsieh, HS Lee, PM Yang, CH Chen, JD Chen, SP Huang, CF Jan. National Taiwan University: THH Chen. National Defense Medical Center: CA Sun. Taipei City Psychiatric Center: MH Wu. Tzu Chi University: SY Chen. Shin Kong Wu Ho-Su Memorial Hospital: KE Chu. Huhsi Health Center, Penghu County: SC Ho, TG Lu. Provincial Penghu Hospital: WP Wu, TY Ou. Sanchi Health Center, Taipei County: CG Lin. Provincial Chutung Hospital: KC Shih. Provincial Potzu Hospital: WS Chung, C Li. Kaohsu Health Center, Pingtung County: CC Chen. Paihsa Health Center, Penghu County: WC How.
Contributors C-JC had full access to all of the data in the study and takes responsibility for the integrity of the data as well as the accuracy of the data analysis. Study concept and design: C-JC, JL and H-IY. Acquisition of data: JL, H-IY, M-HL, C-LJ, S-NL, L-YW, S-LY, CKH, P-JC and C-JC. Analysis and interpretation of data: JL, H-IY and C-JC. Drafting of the manuscript: JL.
Funding This work was supported by The Department of Health, Taiwan; Bristol-Myers Squibb Co., USA; Roche Diagnostics, Switzerland; Academia Sinica, Taiwan; and the National Health Research Institutes (NHRI-EX98-9806PI), Taiwan.
Competing interests Dr Richard Batrla-Utermann is an employee of Roche Diagnostics. All authors report no potential conflict of interest.
Patient consent Obtained.
Ethics approval National Taiwan University.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The sponsors of this study had no role in study design, data collection, data analysis, data interpretation or writing of this report. Moreover, the principle investigator of this manuscript had full access to all data in the study and had final responsibility for the decision to submit the results of this study for publication.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.