long-term morbidity and mortality in chinese insurance … · 2016-12-21 · journal of insurance...

JOURNAL OF INSURANCE MEDICINECopyright © 2001 Journal of Insurance MedicineJ Insur Med 2001 ;33:143-164

ORIGINAL ARTICLE

Long-Term Morbidity and Mortality in ChineseInsurance Applicants Infected With theHepatitis B Virus

Robert J. Pokorski, MD, FACP; Ulrike Ohlmer, DIPL. MATH., DR. RER. POL.

Background.--Worldwide, there are approximately 350 millioncarriers of the hepatitis B virus (HBV). The protracted course ofHBV infection makes it difficult to estimate morbidity and mortalityrisk in an insured lives population that is chronically infected withHBV because most studies on this topic have been based on olderpatients with advanced disease who were treated at tertiary centersthat specialize in care of patients with liver disease. Data from thesereports bias risk estimates toward severe cases and are not appro-priate indicators of what might be expected in an insurance context.This article discusses use of a Markov model to estimate long-termmorbidity and mortality risk associated with chronic HBV infectionin otherwise healthy Chinese insurance applicants.

Results.--The model was validated by comparing results to pop-ulation data published in Taiwan, Hong Kong, Shanghai, Singapore,and Korea. For males, mortality ratios were in the range of 150-175% for underwriting ages 20, 30, and 40 and slightly lower forage 50. For females, mortality ratios were in the range of 125-150%and slightly higher for age 50. Higher mortality ratios in males wererelated to the fourfold higher hepatocellular carcinoma (HCC) inci-dence rate. Mortality ratios varied with the extent of the underwrit-ing evaluation. Liver-related morbidity incidence increased with ageat underwriting for males and females. HBeAg (hepatitis B "e" an-tigen)/anti-HBe status was not a major factor for differentiating riskin an insurance context.

Conclusion.--Morbidity and mortality are within the insurablerange for the majority of HBV-infected Chinese applicants. Risk varieswith the extent of the underwriting evaluation and the percentage ofapplicants with significant liver fibrosis or early cirrhosis that are de-tected during the underwriting process. HBeAg/anti-HBe status is nota major factor for differentiating risk in an insurance context. Morbidityand mortality estimates provided by the model can be generalized toother populations and individuals where HBV infection occurs at birthor during early childhood, although some modification in insurancerisk might be required in non-Asian markets.

Address: Worldwide Medical Re-search and Development, General-Cologne Re, Financial Centre, POBox 300, 695 East Main Street,Stamford, CT 06904-0300.

Correspondent: Robert J. Pokorski,MD, FACP, Vice President; [email protected].

Key words: Hepatitis B, Markovmodels, mortality, morbidity.

Received: May 10, 2000.

Accepted: August 3, 2000.

W orldwide, there are approximately 350million carriers of the hepatitis B virus

(HBV). As many as 1 million of these people

develop hepatocellular carcinoma (HCC) each

year, and countless others develop end-stageliver disease.1 Prevalence of HBV is highest inChina, Southeast Asia, Indonesia, and sub-Saharan Africa, where more than 8% of the

143

JOURNAL OF INSURANCE MEDICINE

Table 1. Annual Transition Probabilities, Male and Female Unless Otherwise Stated*

Transition Transition Rate (Range) References

Expected deathHBeAg/HBsAg to anti-HBe/HBsAgAnti-HBe/HBsAg to HBsAg negative

Cirrhosis data

Overall cirrhosis incidence rate (all ages combined)

MaleFemale

Age-adjusted cirrhosis incidence multiplierHBeAg multiplier

Anti-HBe/HBsAg to compensated cirrhosis

HBeAg/HBsAg to compensated cirrhosis

HBsAg negative to compensated cirrhosis

Compensated cirrhosis to decompensated cirrhosis

Males

Ages 20-54Ages 55-70

Females, all ages

Decompensated cirrhosis to cirrhosis death

Compensated/decompensated cirrhosis to HCC

Males

Ages 20-60Ages 61-70

Females, all ages

HCC mortality rates

Year 1Year 2Year 3Year 4Year 5Year 6Year 7Year 8Year 9Year 10

Per mortality table0.034 (0.031-0.034)0.005 (0.001-0.01)

0.0074 (0.0074-0.021)0.25 × male overall cirrhosis incidence rate

Per derived table1.73 (1.73-1.8)

Overall cirrhosis incidence rate xage-adjusted cirrhosis incidence multiplierHBeAg multiplier × anti-HBePrlBsAg to

compensated cirrhosis transition rate0.01 × anti-HBe/HBsAg to compensated

cirrhosis transition rate

0.023 (0.023-0.05)0.7 × age 20-54 transition rate

0.023 (0.023-0.05)

0.19 (0.19-0.33)

0.028 (0.0083-0.05)Per derived values

0.028 (0.0083-0.05)

0.4580.2190.2130.2060.1990.0640.0640.0640.0640.064

7,89-11

1, 43, 51-55

13, 19

13, 1913, 19

19

22

3, 16

3,16

24, 56, 57

16, 58-64

* HBeAg, hepatitis B "e" antigen; HBsAg, hepatitis B surface antigen; HCC, hepatocellular carcinoma.

population are chronic carriers, as defined bythe presence of hepatitis B surface antigen(HBsAg). In these countries, HBV infectionalmost always occurs at birth or during earlychildhood,2 and sequelae of long-term chronicinfection are uncommon until decades later.

The protracted course of HBV infectionmakes it difficult to estimate morbidity andmortality risk in an insured lives populationthat is chronically infected with HBV. Moststudies on this topic have been based on olderpatients with advanced disease who were

144

POKORSKI--MORBIDITY AND MORTALITY IN CHINESE WITH HEPATITIS B

HBV Model

HBeAg/HBsAg

Anti-HBe/Hl~sAg

HBsAg negative

Compensated cirrhosis

HCC

HCC cure

Death. expected

Die, expected-<~ Death, expected

Convert to anti-HBe/HBsAg

Survive~ Develop compensated ckThosis

Stay in Markov state

Die, expected

~,~ <~ Death, expected

Convert to HBsAg negative

Survive ~ Develop compensated cirrhosis

Stay in Marker state

Die. expected

Survive

Death. expected

Develop compensated cirrhosis


Anti-HBe/HBaAg


HBeAg/HBsAg

HBsAg negative

Compensated cin’hosis

Anti-HBe/HBsAg


HBsAg negative

Die, expected<~ Death, expected

~ Develop decompensated cird~osis <~ Decompensated cirrhosis

Survive

~ Develop HCC~ HCC

Stay in Ma~ov state~ Compensated cirrhosis

Die, expected~ Death, expected

Develop HCCHCC

Survive Die, c|rrhos~s

~<~ Death, cirrhosis

Stay in Markov state<:~ Decompensated cirrhosis

Die. expectedDeath, expected

Die, HCC

~<~ Death, HCC

Exit Markov state after 10 years~ HCC cure

Stay in Markov state for 10 years<~ HCC

Death, expected

HCC cure

Survive

Die, expected


Death, HCC

Death, cirrhosis

Figure 1. Markov model used to estimate long-term morbidity and mortality in Chinese insurance applicants infected withthe hepatitis B virus.

treated at tertiary centers that specialize incare of patients with liver disease.3 Data fromthese reports bias risk estimates toward se-vere cases and are not appropriate indicatorsof what might be expected in an insurancecontext.

This article discusses use of a Markovmodel to estimate long-term morbidity and

mortality risk associated with chronic HBVinfection in otherwise healthy Chinese insur-ance applicants. The topic was chosen be-cause chronic HBV infection is one of themost common medical impairments in Asiaand there are extensive data in the medicalliterature regarding the incidence of compli-cations during decades of follow-up. The

145


180

160

140

120

100

80

4o20

20 30 40 50 60 70

Age

~ Markov model, Taiwan ~ Korea ~ Taiwan ~ Hong Kong ~ Shanghai --e-- Singapore

Figure 2. Male hepatocellular carcinoma (HCC) incidence rates calculated by Markov model versus 70% of actual HCCincidence rates, by country (assume 15% hepatitis B virus prevalence rate).

60.

i5O

40

30

~0

020 30 40 50 60 70

Age

._+_ Mankov model, Taiwan --a-- Korea ~ Taiwan ~ Hong Kong ~ Shanghai --e-- Singapore

Figure 3. Female hepatocellular carcinoma (HCC) incidence rates calculated by Markov model versus 70% of actual HCC

incidence rates, by country (assume 15% hepatitis B virus prevalence rate).

model is validated by comparing results topopulation data published in Taiwan, HongKong, Shanghai, Singapore, and Korea.

DESCRIPTION OF THEMARKOV MODEL

A Markov model was created to estimaterisk associated with chronic HBV infection(Figure 1). Simplifying assumptions weremade that caused the model structure to di-verge from a strict interpretation of the HBV

pathophysiologic disease process. For exam-ple, since HCC arises in a cirrhotic liver inthe vast majority of cases,4,5 the model as-sumes that cirrhosis must precede HCC.Though not totally correct from a pathophys-iologic standpoint (transitions directly fromchronic HBV infection to HCC can occur), er-rors would be small since cirrhosis usuallyprecedes HCC.

The 10 Markov states are described below.HBeAg/HBsAg.--For each Markov state in

the model, subjects are first exposed to age-

146


180

160

1140 -

120

20 30 40 50 60 70

Age

I_.~_ Markov model, Taiwan ._~_Taiwan __n__ Hong Kong _g._Korea

Figure 4. Male hepatocellular carcinoma (HCC) mortality rates calculated by Markov model versus 70% of actual HCCmortality rates, by country (assume 15% hepatitis B virus prevalence rate).

90

80

70

60

5O

40 ,,

30

2O

o ............20 30 40 50 60 70

Age

l..._~.__ Mankov model, Taiwan ~ Taiwan ---x-- Hong Kong ._=._. Korea ~

Figure 5. Female hepatocellular carcinoma (HCC) mortality rates calculated by Markov model versus 70% of actual HCCmortality rates, by country (assume 15% hepatitis B virus prevalence rate).

and gender-specific risk of expected deathper the chosen mortality table. For the stateHBeAg (hepatitis B "e" antigen)/HBsAg,those who survive expected death move toanti-HBeAg/HBsAg or to compensated cir-rhosis or stay in HBeAg/HBsAg and face thesame risks during the next cycle. Each cycleof the model represents 1 year.

Anti-HBe/HBsAg.--Subjects who surviveexpected death move to HBsAg negative orto compensated cirrhosis or stay in anti-HBe/

HBsAg and face the same risks during thenext cycle.

HBsAg Negative.--Subjects who survive ex-pected death move to compensated cirrhosisor stay in HBsAg negative and face the samerisks during the next cycle.

Compensated Cirrhosis.--Subjects who sur-vive expected death move to decompensatedcirrhosis (obvious subjective symptoms ofliver failure, jaundice, ascites, hepatic enceph-alopathy, or blood coagulation disorders3) or

147


Table 2. Percentage of Subjects in Each Markov State at Start of Model, by Age and Percentage of CirrhosisCases Detected During Underwriting*,1"

Percentage of Cirrhosis Cases Detected During Underwriting

Markov State atStart of Model Age 20 Age 30 Age 40 Age 50

Calculations -- 20% 50% 80% 20% 50% 80% 20% 50% 80%

HBeAg/HBsAg 0.80 0.54 0.55 0.55 0.37 0.38 0.38 0.22 0.24 0.25Anti-HBe/HBsAg 0.20 0.44 0.44 0.44 0.57 0.58 0.60 0.62 0.66 0.71Compensated cirrhosis 0.00 0.02 0.01 0.01 0.06 0.04 0.02 0.16 0.10 0.04

* HBeAg, hepatitis B "e" antigen; HBsAg, hepatitis B surface antigen; HBV, hepatitis B virus.-~ For example, for a cohort aged 40, if 50% of HBV-infected applicants with compensated cirrhosis are detected

and declined, the remaining subjects are distributed in these Markov states at the start of the model calculations:HBeAg/HBsAg, 0.38; anti-HBe/HBsAg, 0.58; and compensated cirrhosis, 0.04. For age 20, the assumption is that noapplicants have cirrhosis.

60

. 50

40

¯ 30

20

10

020 30 40 50 60 70

Age

._._. Markov model, Taiwan -.-.m-- Taiwan J

Figure 6. Male cirrhosis mortality rates calculated by Markov model versus 50% of actual cirrhosis mortality rates, Taiwan(assume 15% hepatitis B virus prevalence rate).

to HCC or stay in compensated cirrhosis andface the same risks during the next cycle.

Decompensated Cirrhosis.--Subjects whosurvive expected death move to HCC or todeath (cirrhosis) or stay in decompensatedcirrhosis and face the same risks during thenext cycle.

HCC.--This is a 10-year tunnel state, an ar-ray of temporary Markov states that can bevisited only in a fixed sequence.6 During eachof the 10 years that subjects stay in this state,those who survive expected death move (1)to death (HCC), (2) back to HCC to face the

same risks during the next cycle, or after 10years, (3) exit the tunnel state to HCC cure.

HCC Cure.--Subjects who survive expecteddeath stay in HCC cure and face the samerisk during the next cycle.

Death (Expected); Death (HCC); Death (Cir-rhosis).--There are no transitions from theseMarkov states.

ANNUAL TRANSITION RATES

Transition rates described in this sectionare summarized in Table 1.

148


50¸

’~ 20

o20 30 40 50 60 70

Age

L...~_ Markov model, Taiwan ._=_ Taiwan J

Figure 7. Female cirrhosis mortality rates calculated by Markov model versus 50% o, factual cirrhosis mortality rates, Taiwan(assume 15% hepatitis B virus prevalence rate).

Expected Death

Taiwan mortality tables were used to deter-mine age- and gender-specific population7

and insured lives8 mortality rates as de-scribed later in this article.

HBeAg/HBsAg to Anti-HBe/HBsAg

Subjects in HBeAg/HBsAg test positive forHBeAg and HBsAg. Those in anti-HBe/HBsAg have developed antibodies to HBe butremain HBsAg positive. For Chinese popu-lations age 20-60, this annual transition prob-ability was reported by Guan and Yu9 (0.034),Liaw et al1° (0.031), and Lok et a111 (0.031).The value used in the model was 0.034. Tran-sitions from HBeAg/HBsAg to anti-HBe/HBsAg were not allowed after age 60 becausedata indicate that people who are HBeAgpositive at older ages remain HBeAg posi-tive.9-12

Anti-HBe/HBsAg to HBsAg Negative

Transition to HBsAg-negative status (es-sentially, total clearing of HBV) is rare in peo-ple infected at birth or during early child-hood, with estimates varying from 0.001 to0.01 per year. The approximate median valueof 0.005 was chosen.

Cirrhosis Data

Overall Cirrhosis Incidence Rate (All AgesCombined)

This category refers to the annual rate forthe transition from chronic HBV to cirrhosis.Rates quoted in the literature vary dependingon characteristics of the study and the pop-ulation (age, severity of infection, prospectiveversus retrospective study, etc). The valuechosen for this transition (0.0074 per year)was based on data by Yu et al.13 Investigatorsin this very large prospective study (1506 pa-tients) used liver ultrasound examinations toexclude patients with preexisting cirrhosis,thereby providing a more accurate estimateof the rate at which cirrhosis develops, andthen followed the cohort to determine the in-cidence of cirrhosis.

Male~Female Differentiation

Population data indicate that fibrosis andcirrhosis are more common in men.14 Themechanism is related to favorable influencesof estrogen (women) on the rate of hepaticfibrosis,15 a higher frequency of acute exac-erbations in males (which accelerate the fi-brotic process),16 and perhaps gender differ-ences in lifestyle and occupational factors27

149


175

150

125

lOO20 30 40 50 60 70

Age

I~ HBeAgl00% ._~_Anti-HBel00% ]

Figure 8. Estimated mortality ratios for hepatitis virus B-infected Chinese males age 20 at underwriting, by hepatitis B ’~"antigen (HBeAg) and anti-HBe status.

Because cirrhosis (the final stage of the fi-brotic process) is the primary risk factor forHCC, the effect of a higher rate for develop-ment of fibrosis in males is to increase theincidence of HCC. Estimates of the male-to-female ratio of HCC incidence in Chinesepopulations vary from 3.214,18 to 4.1.17 In thismodel, the causes of excess morbidity andmortality are HCC and decompensated cir-rhosis, and transitions to these Markov statesdepend on prior transitions from HBeAg/HBsAg, anti-HBe/HBsAg, and HBsAg neg-ative to compensated cirrhosis. Thus, controlof transitions to compensated cirrhosis (andeffectively, transitions to HCC and decom-pensated cirrhosis) is handled in the modelby using different male and female transi-tions to compensated cirrhosis. The assump-tion was that HCC incidence would be four-fold higher in Chinese males. For females,this adjustment was accomplished by multi-plying 0.25 times male transition rates fromHBeAg/HBsAg to compensated cirrhosis,anti-HBe/HBsAg to compensated cirrhosis,and HBsAg negative to compensated cirrho-sis.

Age-Adjusted Cirrhosis Incidence Multiplier

Yu et aP3 and Liaw et a119 published similarage factors that modify the overall transition

rate (all ages combined, above) from chronichepatitis to compensated cirrhosis. Data byYu et a113 were fitted to a general linear modelto create age-adjusted cirrhosis multipliersfor ages 20-70, that is, values that multiplythe overall cirrhosis incidence rate to produceage-adjusted transitions to compensated cir-rhosis. The importance of this adjustment isthat transitions to compensated cirrhosis (andsubsequently to HCC and decompensatedcirrhosis) become a function of duration ofinfection, and the model conforms in this re-gard to clinical observations that indicate thatsequelae of HBV are related to age and du-ration of infection.2°,21

HBeAg Multiplier

HBeAg-positive patients progress to com-pensated cirrhosis at a faster rapid rate thananti-HBe patients. The magnitude of this ef-fect has been estimated at 1.7313 and 1.8.19 Yuet a113 reported that HBeAg was associatedwith an increased risk of cirrhosis, while therelation of HBeAg to HCC was not statisti-cally significanU3 The model assumes thatthe sole effect of HBeAg is to increase (by afactor of 1.73) the transition rate fromHBeAg/HBsAg to compensated cirrhosis.

150


150

~ ~25

~oo20 30 40 50 60 70

Age

l--e~ HBeAgl00% ._~._Anti-HBel00% t

Figure 9. Estimated mortality ratios for HBV-infected Chinese females age 20 at underwriting, by hepatitis B "e" antigen(HBeAg) and anti-HBe status.

Anti-HBe/HBsAg to CompensatedCirrhosis

This transition rate is calculated as theoverall cirrhosis incidence rate (0.0074) mul-tiplied by the age-adjusted cirrhosis incidencemultiplier.

HBeAg/HBsAg to Compensated Cirrhosis

This transition rate is calculated as theHBeAg multiplier (1.73) multiplied by theanti-HBe/HBsAg to compensated cirrhosistransition rate.

HBsAg Negative to Compensated Cirrhosis

This is a rare event. The transition rate wasbased on data by Wong et a122 and was cal-culated as 0.01 multiplied by the anti-HBe/HBsAg to compensated cirrhosis transitionrate.

Compensated Cirrhosis to DecompensatedCirrhosis

Males, Ages 20-54, and Females, All Ages

This transition rate (0.023) was based ondata by Liaw et al.3

Males, Ages 55-70

For older males, a transition rate of 0.023per year yielded a cirrhosis mortality rate(Markov state death (cirrhosis)) that exceededestimates of HBV-related cirrhosis mortalityin Taiwan23 (below). The fit of the model im-proved by calculating this transition as 0.7multiplied by age 20-54 transition rate(0.023).

Decompensated Cirrhosis to CirrhosisDeath

This transition rate (0.19) is the only modelvalue that was not based on data from a Chi-nese population.24

Compensated and DecompensatedCirrhosis to HCC

Males, Age 20-60, and Females, All Ages

The transition rate for compensated or de-compensated cirrhosis to HCC (0.028 peryear) was based on a prospective study byLiaw et al3 in which a cohort with recentlydiagnosed cirrhosis was followed to deter-mine the annual incidence of HCC.

151


175

~ 150

~125

10020 30 40 50 60 70

Age

._,_ HBeAgl00% _~_Anti-HBel00% ~ HBeAgS0%._anti-HBe20% I

Figure 10. Estimated mortality ratios for hepatitis B virus-infected Chinese males age 20 at underwriting, by hepatitis B"e" antigen (HBeAg) and anti-HBe status.

140

130

12o

110

10020 30 40 50 60 70

Age

[._~_ HB eAg 100% ._.=.._ Anti-HBe 100% --a-- HB eAg80%_anti-HBe20% I

Figure 11. Estimated mortality ratios for hepatitis B virus-infected Chinese females age 20 at underwriting, by hepatitis B"e" antigen (HBeAg) and anti-HBe status.

Males, Ages 61-70

For older males, a transition rate of 0.028per year yielded HCC incidence and mortal-ity rates (Markov state death (HCC)) thatwere less than HBV-related estimates of HCCincidence and mortality in Taiwan (below).The fit of the model improved by multiplyingthe age 20-60 transition rate (0.028) by a fac-tor, thereby increasing HCC incidence at old-er ages. These factors are as follows: age 61,

1.05; age 62, 1.10; age 63, 1.15; age 64, 1.20;age 65, 1.25; age 66, 1.30; age 67, 1.35; age 68,1.40; age 69, 1.45; and age 70, 1.50.

HCC Mortality Rates

Determination of these values was prob-lematic. Many sources in the Chinese litera-ture list mortality following diagnosis ofHCC, but data generally have one or moredeficiencies:

152


175

150

"125

"10030 40 50 60 70

Age

[._~_ Detect 20% ---B-- Detect 50% ]

Figure 12. Estimated mortality ratios for hepatitis B virus-infected Chinese males age 30 at underwriting, by percentdetection of existing cirrhosis.

175

150

125

"10030 40 50 60 70

Age

[ ~ Detect 20% + Detect 50% J

Figure 13. Estimated mortality ratios .for hepatitis B virus-infected Chinese females age 30 at underwriting, by percentdetection of existing cirrhosis.

¯ Data do not reflect modern surgical andnonsurgical treatments. In these reports,survival after diagnosis of HCC is mea-sured in weeks and months.25,26

¯ Survival rates are based on operated cases,thereby biasing the results to show favor-able results, that is, patients with diseasethat is so advanced that it cannot be treatedsurgically are not included in the statistics.

¯ Data do not include recent favorable trends

due to early detection via screening withannual or biannual liver ultrasound exam-inations and/or alpha-fetoprotein (AFP)levels.27

HCC survival rates were estimated as fol-lows:

¯ Screening programs are often used in high-risk populations to detect early, asympto-

153


200

A 175

~ 150

~ 125

100. ,40 50 60 70

Age

[._~_ Detect 20%


175

~ 150

~ 125

10040 50 60 70

Age

[~ Detect 20% .._=_ Detect 50% ~ Detect 80% l

Figure 15. Estimated mortality ratios for hepatitis B virus-infected Chinese females age 40 at underzoriting, by percentdetection of existing cirrhosis.

matic HCC, but even aggressive screeningprograms often fail to detect early HCC.28,29

¯ The extent of screening varies with the sizeof the at-risk population, the amount of re-sources committed to the project, and theurban/rural distribution of the population.For example, screening would generally beless successful in rural areas with less so-phisticated medical facilities and a lowerpopulation density.

¯ The most favorable outcomes that can beachieved with current technologies and an

aggressive screening program that targetsmost of the at-risk population result in de-tection of (1) one quarter of HCC cases atan early, potentially curable stage; (2) onehalf at an intermediate stage, where cure isunlikely but possible in a minority of cases;and (3) one quarter at end-stage disease,where survival is less than 1 year.3° How-ever, studies of Asian populations indicatethat a maximum of 15% of patients pre-senting with HCC are curable with sur-gery.31

154


175

=~ 150o

~ 125i

50 60 70Age

l--e-- Detect 20% Detect 50% Detect 80% 1


175

125

IO050 60 70

Age

f+ Detect 20% + Detect 50% + Detect 80% ]

Figure 17. Estimated mortality ratios for hepatitis B virus-infected Chinese females age 50 at underwriting, by percentdetection of existing cirrhosis.

¯ Cumulative 5- and 10-year survival ratesfor early, potentially curable HCC cases are60 and 42.7%, respectively, and for inter-mediate stage disease are 22.2 and 16.1%,respectively. For end-stage disease, surviv-al at I year after HCC diagnosis is 0%. An-nual transition probabilities from HCC todeath (HCC) in a market with an HCCscreening program were derived fromthese cumulative survival rates by weight-ing the likely stages at diagnosis as follows:early diagnosis, 15% 3~; intermediate diag-

nosis, 55%; and late (end-stage) diagnosis,30% (Table 1).

¯ Sensitivity testing was performed usingdifferent weightings for the percentage ofcases that would be diagnosed in early, in-termediate, and late HCC stages (data notshown). For all scenarios (including a mostfavorable scenario that detected 25% ofHCC cases at an early, potentially curablestage), mortality rates were so high thatoutcomes calculated by the model did notvary significantly.

155


VALIDATION

Description of the Validation Process

Two methods were used to validate themodel. First, transition probabilities werebased exclusively on Chinese data (with thesingle exception noted earlier24). In cases ofconflicting data, preference was given tolong-term prospective studies in cohorts thatwere well characterized with respect to riskfactors used in the model. For example, tran-sition probabilities from chronic hepatitis tocirrhosis were based on a cohort examined byliver ultrasound to exclude cases with pre-existing cirrhosis.13

Second, results of the model were com-pared to real-life population data sets: (1)HCC incidence rates in Taiwan,32 HongKong,33 Shanghai,18 Singapore,34 and Korea35;(2) HCC mortality rates in Taiwan,36 HongKong,37 and Koreag8; and (3) cirrhosis mortal-ity rates in Taiwan.39 The validation processwas designed as follows:

¯ Expected death rates (data for all nodesidentified in Figure 1 as death (expected))for the validation phase of the model werebased on the 1996 Taiwan Life Tables7 forthe general population. Taiwan mortalityrates were used because most of the tran-sition probabilities in the model were de-rived from Taiwanese cohorts.

¯ All subjects were age 20 at the beginningof the validation phase.

¯ A simplifying assumption was made thatcirrhosis prevalence at age 20 was 0%.Studies indicate that 3-5% of 20 year oldswith chronic HBV infection already havecirrhosis.4°,41 These estimates probablyoverestimate cirrhosis prevalence becauseof referral bias, that is, clinics specializingin care of HBV-infected children treat pa-tients with more severe disease.

¯ Prevalence of HBeAg at age 20 has beenestimated at 80-85%.9,42 The model as-sumes an HBeAg prevalence rate at age 20of 80%.

¯ At the beginning of the model run for age20, subjects are distributed among the

Markov states as follows: HBeAg/HBsAg,80%; anti-HBe/HBsAg, 20%; HBsAg neg-ative, 0%; compensated cirrhosis, 0%; de-compensated cirrhosis, 0%; HCC, 0%; HCCcure, 0%; death (expected), 0%, death(HCC), 0%; and death (cirrhosis), 0%. Thisdistribution indicates that at age 20 all sub-jects are in Markov states HBeAg/HBsAg(80%) or anti-HBe/HBsAg (20%) and nonehave cirrhosis or HCC.

¯ Prevalence estimates of chronic HBV infec-tion in East and Southeast Asia among thepopulation age 20 and older range from 8-25%.2‘42-46 The model assumes a prevalencerate of 15%. Sensitivity testing (data notshown) indicated that HBV prevalencerates between 10 and 20% would approxi-mate the real-life population data sets (be-low), with the best fit achieved for esti-mates of 15-20%.

¯ In Taiwan, 87% of all primary liver cancersare HCC,32 that is, most but not all livercancers are HCC. Estimates in the Chineseliterature of the percentage of HCC casesattributable to HBV range from 67 to85%,5"10’13’17"43’47’48 with most estimates about80%. The model assumes that 70% of allliver cancers are due to HBV (0.87 × 0.80,ie, the percentage of liver cancers that areHCC multiplied by the percentage of HCCcases attributable to HBV). Sensitivity test-ing (data not shown) indicated that esti-mates (of the percentage of all liver cancersthat are due to HBV) between 50 and 80%would approximate the real-life data sets(below), with the best fit achieved for esti-mates of 70%.

¯ Data were not found for the percentage ofcirrhosis deaths attributable to HBV. Thisvalue would be less than the percentage ofHCC cases attributed to HBV, the reasonbeing that there are more disease processesthat might lead to cirrhosis and subsequentdeath due to liver failure, including heavyalcohol intake, certain liver and biliary tractdiseases, and genetic disorders. The modelassumes that 50% of cirrhosis deaths areattributable to HBV. Sensitivity testing(data not shown) indicated that estimates

156


14

~ ~o

2

020 30 40 50 60 70

Age

[._.~._ Decompensated cirrhosis --i--- HCC Both causes 1

Figure 18. Estimated incidence of total and permanent disability in hepatitis B virus-infected Chinese males age 20 atunderwriting, by cause of disability.

(of the percentage of cirrhosis deaths thatare due to HBV) between 40 and 60%would approximate the real-life data sets(below), with the best fit achieved for theestimate of 50%.

¯ Values for all other transition probabilitieswere set as described previously.

Outcome of the Validation Process

The output of the model based on thesedata and assumptions is represented in Fig-ures 2-7.

¯ Figures 2 and 3 display, respectively, maleand female HCC incidence rates calculatedby the model compared with 70% of actualHCC incidence rates in Taiwan,32 HongKong,33 Shanghai,18 Singapore,34 and Ko-rea.38 The calculated male and female in-cidence curves (Markov model, Taiwan)closely approximate population HCC inci-dence rates in these 5 markets.

¯ Figures 4 and 5 display, respectively, maleand female HCC mortality rates calculatedby the model compared with 70% of actualHCC mortality rates in Taiwan,36 HongKong,37 and Korea.38 The calculated maleand female mortality curves (Markov mod-el, Taiwan) closely approximate populationHCC mortality rates in these 3 markets.

Figures 6 and 7 display, respectively, maleand female cirrhosis mortality rates calcu-lated by the model compared with 50% ofactual cirrhosis mortality rates in Taiwan.39

The calculated male and female mortalitycurves (Markov model, Taiwan) closely ap-proximate population cirrhosis mortalityrates in Taiwan.

DISTRIBUTION AMONG MARKOVSTATES BY AGE AT UNDERWRITING

Having demonstrated that the model ap-proximates population HCC incidence rates,HCC mortality rates, and cirrhosis mortalityrates in East and Southeastern Asia, the nextstep is to estimate insured lives morbidityand mortality experience. Expected deathrates (data for all nodes identified in Figure1 as death (expected)) for this phase of themodeling process are based on 90% of the1989 Taiwan Standard Ordinary (TSO89) in-sured lives table.8 Values for other variableswere set as described previously.

Markov State Distribution, Age 20

The initial run of the model was for healthyapplicants age 20 at underwriting, with theassumption that 80% of subjects were in theHBeAg/HBsAg Markov state and 20% were

157


7

4

20 30 40 50 60 70

Age

I--e-- Decompensated cirrhosis _..=_ HCC ~ Both causes

Figure 19. Estimated incidence of total and permanent disability in hepatitis B virus-infected Chinese females age 20 atunderwriting, by cause of disability.

in the anti-HBe/HBsAg state (Table 2). Dur-ing each 1-year cycle of the model, subjectsage 1 year and gradually move into differentMarkov states as they convert to anti-HBe/HBsAg or to HBsAg negative, develop com-pensated cirrhosis, decompensated cirrhosis,or HCC, or die.

Markov State Distribution, Ages 30, 40,and 50

Data from the age 20 run of the modelwere used to estimate the likely Markov stateof subjects who might apply for insurance atolder ages after they had been redistributedamong the various states. Assumptions wereas follows regarding the Markov state inwhich applicants would be found at under-writing ages 30, 40, and 50 (Table 2):

¯ Applicants age 30, 40, or 50 would not besold coverage if there was a history of se-vere HBV-related disease, including de-compensated cirrhosis and current orcured HCC, that is, at the time of under-writing, no subjects would begin the modelin decompensated cirrhosis, HCC, or HCCcure.

¯ No subjects would be in any of the threedeath states at the time of application.

¯ No subjects would be in the HBsAg nega-tive Markov state, the reason being that thismodel is concerned with estimating risk inpeople chronically infected with HBV, notin those who have cleared the virus.

¯ Thus, all subjects at underwriting ages 30,40, and 50 would be in 1 of 3 Markovstates: HBeAg/HBsAg, anti-HBe/HBsAg,and compensated cirrhosis.

¯ Distribution among these three states wasdetermined by two factors:

the weighted probabilities (weighted sothe total sums to 100%) of being locatedin these states per attained age (30, 40, or50) as determined during the initial age20 run of the model, after eliminating sub-jects who had died, were HBsAg negative,or were in states indicative of severe dis-ease; andthe percentage of applicants with com-pensated cirrhosis that were detected andeliminated from the risk pool during theunderwriting process, that is, the effec-tiveness of the selection process. Whatpercentage of applicants with existing cir-rhosis would be detected during under-writing? The likelihood of identifyingthese higher risk applicants would de-

158


20.

20 30 40 50 60 70

Age

[_.~_ Age 20 + Age 30 .--a.-- Age 40 ---.x-- Age 50 ]

Figure 20. Estimated incidence of total and permanent disability due to both hepatocellular carcinoma and decompensatedcirrhosis in hepatitis B virus-infected Chinese males, by age at underwriting.

pend on the extent of the underwritingevaluation, for example, few or no labo-ratory tests versus a full blood profile andAFP level, insurance examination to de-tect signs and symptoms of more ad-vanced disease, a physician’s statement(very useful for this impairment, particu-larly if a liver biopsy has been performed),and liver ultrasound to detect cirrhosisand early HCC (especially in largeamount cases). Three scenarios were cre-ated, which assumed cirrhosis detectionrates of 20, 50, and 80% (ie, 20, 50, and80%, respectively, of applicants with ex-isting cirrhosis would be detected andeliminated from the risk pool during theunderwriting process). Table 2 summariz-es the calculated percentages (per theseassumptions) of subjects in each Markovstate at the beginning of the model run forages 20, 30, 40, and 50.

Magnitude and Duration of SelectionProcess

In addition to detecting existing cirrhosiscases, the underwriting process also decreas-es the rate at which transitions occur betweencompensated cirrhosis and both decompen-sated cirrhosis and HCC, the reason beingthat severe cases of compensated cirrhosis

(which would be more likely to progress todecompensated disease and/or HCC in thefirst few years after underwriting) are detect-ed and eliminated from the risk pool. Thereare no HBV-specific data regarding the mag-nitude or the duration of this selection effect.The model assumes that transitions fromcompensated cirrhosis to decompensated cir-rhosis and from compensated cirrhosis toHCC would decrease (compared with whatwould be expected without underwriting)per this schedule: year 1, 10% of expected;year 2, 20% of expected; year 3, 33% of ex-pected; year 4, 50% of expected; year 5, 75%of expected, and year 6, 100% of expected (ie,no further effects from the selection process).Shorter or longer select periods would short-en or lengthen, respectively, the period oftime between underwriting and claims due toHCC and decompensated cirrhosis.

RESULTS

Mortality Results

Risk Associated With HBeAg Status

Medical studies report a less favorable out-come in patients with HBeAg (HBeAg/HBsAg Markov state), principally becauseHBeAg is associated with more rapid devel-opment of cirrhosis,13,19 but the magnitude of

159


8

5

4

3

2

1

20 30 40 50 60 70

Age

-.e--Age 20 ._=._Age 30 --a--Age 40 .--x.--Age 50]

Figure 21. Estimated incidence of total and permanent disability due to both hepatocellular carcinoma and decompensatedcirrhosis in hepatitis B virus-infected Chinese females, by age at underwriting.

this risk has not been quantified in an in-sured lives context. Figures 8 and 9, respec-tively, display estimated mortality ratios forHBV-infected Chinese males and females age20 at underwriting, by HBeAg and anti-HBestatus, based on comparison with 90% ofmortality rates in the TSO89 Insured LivesTable.8 HBeAgl00% means all subjects in themodel are HBeAg positive and anti-HBel00%means all subjects in the model are anti-HBepositive. As noted earlier, the assumption isthat all 20-year-old applicants are healthyand that none have cirrhosis. For males, Fig-ure 8 indicates that estimated mortality ratiosare relatively similar for ages 20-30 and 55-70 and are approximately 25 mortality per-centage points higher for HBeAg-positivemiddle-aged adults. A similar pattern is seenfor females (Figure 9), but the difference inmortality ratios is only about 10 mortalitypercentage points. When averaged over a 50-year follow-up (ages 20-70), estimated mor-tality ratios for HBeAg positive subjects areapproximately 15 mortality percentage pointshigher for males, and 5 mortality percentagepoints higher for females.

Age 20 at Underwriting

Insurers in Asia may or may not knowHBeAg and anti-HBe status at the time of un-

derwriting. The question then is what themortality risk of a healthy 20-year-old HBV-infected applicant (1) known to be HBeAgpositive (HBeAgl00%), (2) known to be anti-HBe positive (anti-HBel00%), and (3) whoseHBe status is unknown at the time of under-writing (HBeAg80%_anti-HBe20%, ie, theusual HBeAg and anti-HBe distribution for a20 year old9,42). Figures 10 and 11 display thisinformation for males and females, respec-tively.

¯ Mortality ratios for males age 20 at under-writing (Figure 10) increase gradually forthe first decade, plateau between ages 40and 55, and taper gradually at older ages.Peak mortality ratios are as follows:HBeAgl00%, 173%; HBeAg80%_anti-HBe20%, 168%; and anti-HBel00%, 150%.

¯ Mortality ratios for females age 20 at un-derwriting (Figure 11) demonstrate a sim-ilar pattern. Peak mortality ratios are as fol-lows: HBeAgl00%, 131%; HBeAg80%_anti-HBe20%, 129%; and anti-HBel00%, 121%.

¯ Depending on HBeAg/anti-HBe status,mortality ratios for males are 30-40 per-centage points higher than for females.

Age 30, 40, and 50 at Underwriting

Subjects that enter the model at age 20 areredistributed among the different Markov

160


states after each cycle (1 year) in a mannerthat fairly closely approximates real-life ex-perience in East and Southeastern Asia. Whenolder people with HBV infection apply for in-surance, coverage, underwriters usually iden-tify severe risks (decompensated cirrhosisand HCC) and some, but not all of the ap-plicants, with compensated cirrhosis. Thus,estimated mortality ratios for applicants olderthan age 20 show a steeper increase (com-pared with age 20) because applicants withexisting, undetected cirrhosis progress morerapidly to HCC and/or decompensated cir-rhosis.

Figures 12-16 display estimated mortalityexperience for HBV-infected applicants byage, gender, and assumed percent cirrhosisdetection rate. For a given age and gender,mortality ratios are inversely related to thepercent cirrhosis detection rate: lower detec-tion rates result in higher mortality ratios be-cause more claims occur in the early policyyears.

¯ Age 30. Mortality ratios for males (Figure12) increase rapidly during the first decade,plateau between ages 40 and 55 at valuesthat vary with percent cirrhosis detection,and taper gradually thereafter. Peak mor-tality ratios are detect 20% (172%) and de-tect 50% (170%). The 80% detection rate isnot shown because calculated values areidentical to the 50% detection rate per as-sumptions in Table 2. For females (Figure13), mortality ratios peak after the first de-cade, followed by a long plateau. Peak mor-tality ratios are detect 20% (154%) and de-tect 50% (137%) (80% detection rate notshown for reasons explained above).

¯ Age 40. Mortality ratios increase until ap-proximately ages 45-50. For males (Figure14), peak mortality ratios are detect 20%(176%), detect 50% (170%), and detect 80%(165%). Female peak mortality ratios (Fig-ure 15) are detect 20% (156%), detect 50%(143%), and detect 80% (131)%.

¯ Age 50. Mortality ratios increase until ap-proximately ages 55-60. For males (Figure16), peak mortality ratios are detect 20%

(160%), detect 50% (151%), and detect 80%(143%). Female peak mortality ratios (Fig-ure 17) are detect 20% (167%), detect 50%(149%), and detect 80% (131%).

Morbidity Results

An assumption was made that total andpermanent disability (TPD) would occurupon diagnosis of HCC or decompensatedcirrhosis. Figures 18 and 19 display estimatedincidence rates of TPD for males and females,respectively, age 20 at underwriting accord-ing to liver-related causes of disability. HCCis the most significant cause of disability.TPD incidence rates for age 20 at underwrit-ing (both liver-related causes) peak at age 70at 13.7 per 1000 for males and 6.1 per 1000for females.

Figures 20 and 21 summarize estimated in-cidence rates for both liver-related causes ofTPD in HBV-infected males and females, re-spectively, for ages 20-50 at underwriting. PeakTPD incidence rates occur at age 70 for males(18.4 per 1000) and females (7.8 per 1000).

DISCUSSION

Insurability

The model suggests that morbidity andmortality experience would be within insur-able ranges for the majority of HBV-infectedpersons. For males, mortality ratios are in therange of 150-175% for underwriting ages 20,30, and 40 and slightly lower for age 50. Forfemales, mortality ratios are in the range of125-150% and slightly higher for age 50.Higher mortality ratios in males are relatedto the fourfold higher HCC incidence rate.

Risk Varies With Extent of UnderwritingEvaluation

Some HBV-infected people have significantliver fibrosis or early cirrhosis when they ap-ply for insurance, and they are closer to thedate of HBV-related complications. Cirrhosisdetection rates in Figures 12-17 are theoreti-cal illustrations of the inverse relationship be-

161


tween rate of detection and estimated insuredlives experience. For example, a detection rateof 20% means that fewer high-risk applicantswould be detected and mortality ratios wouldbe higher; if the cirrhosis detection rate were50%, more high-risk applicants would be de-tected, with correspondingly lower mortalityratios.

What percentage of higher risk applicantswould actually be detected during the un-derwriting process? There are no specificdata that answer this question. In general, thelikelihood of identifying applicants at higherrisk would depend on the extent of the un-derwriting evaluation, including:

¯ few or no laboratory tests versus a fullblood profile (eg, a serum albumin levelbelow the lower limit of normal indicates apoor prognosis) and AFP level;

¯ insurance examination to detect signs andsymptoms of more advanced disease;

¯ a physician’s statement (very useful for thisimpairment, particularly if a liver biopsyhas been performed); and

¯ liver ultrasound to detect cirrhosis and ear-ly HCC (especially in large amount cases).

HBeAg/Anti-HBe Status Not a MajorFactor for Differentiating Risk

Authors have reported the frequency dis-tribution of HBeAg and anti-HBe by age andthe long-term impact of HBeAg versus anti-HBe status. These data were used in the mod-el to approximate real-life population data inTaiwan, Hong Kong, Shanghai, Singapore,and Korea. For subjects age 20 at underwrit-ing, the difference between peak mortality ra-tios for HBeAg positive versus anti-HBe pos-itive is approximately 25 (males) and 10 (fe-males) mortality percentage points (Figures 8and 9, respectively). When averaged over a50-year follow-up (ages 20-70), estimatedmortality ratios for HBeAg-positive subjectsare approximately 15 mortality percentagepoints higher for males and 5 mortality per-centage points higher for females. Thus,HBeAg/anti-HBe status is not a major factor

for differentiating risk in an insurance con-text.

HBV Mutations

HBV often mutates to a form in which ac-tive replication persists even though HBeAgis not detectable. These are known as precoreand core promoter mutations.49 Studies in-volving these mutations were not incorporat-ed in the model because the reported preva-lence of the mutation varies a great deal fromstudy to study, there are insufficient data re-garding the long-term impact of the core mu-tation,so and real-life data sets could be ap-proximated without further complicating themodel.

Generalization to Other Populations

The focus of this analysis is on Chinesepeople infected with HBV. Morbidity andmortality estimates provided by the modelcan be generalized to other populations andindividuals where HBV infection occurs atbirth or during early childhood. Some mod-ifications in insurance risk (usually higherrisk estimates) might be required in non-Asian markets where few or no HBV-relateddeaths would be included in insured livesmortality rates.

Serum Alanine Aminotransferase Levels

Serum alanine aminotransferase (ATL) lev-els were not included as a variable in theMarkov model because the literature indicat-ed that the correlation between ALT levelsand prognosis was weak. The natural historyof HBV acquired at birth or during earlychildhood is such that ALT levels fluctuateduring the course of HBV infection, increas-ing and decreasing depending on host andviral factors, and values at a single point intime are not strongly predictive of outcome.For example, an elevated ALT level at the timeof underwriting might indicate seroconver-sion or attempted seroconversion fromHBeAg to anti-HBe (a favorable develop-ment), while a normal ALT level could occur

162


in an applicant with progressive fibrosisand/or cirrhosis (both unfavorable develop-ments).

One situation where ALT levels mightprove useful is in the case of sustained ALTelevation. Yu et a113 observed more rapid pro-gression to cirrhosis in patients with sus-tained elevation of serum ALT for 6 monthsor longer. These cases could be identified inan insurance setting only via a physician’sstatement that listed results of prior enzymelevels.

CONCLUSIONS

Morbidity and mortality are within the in-surable range for the majority of HBV-infect-ed Chinese applicants, with higher male mor-tality ratios due to the fourfold higher HCCincidence rate. Risk varies with the extent ofthe underwriting evaluation and the percent-age of applicants with significant liver fibro-sis or early cirrhosis that are detected duringthe underwriting process. HBeAg/anti-HBestatus is not a major factor for differentiatingrisk in an insurance context. Morbidity andmortality estimates provided by the modelcan be generalized to other populations andindividuals where HBV infection occurs atbirth or during early childhood, althoughsome modification in insurance risk might berequired in non-Asian markets.

REFERENCES

1. McMahon BJ. Chronic carriers of hepatitis B viruswho clear hepatitis 13 surface antigen: are they re-ally "off the hook"? Hepatology. 1998;28:265-266.

2.Mahoney FJ. Update on diagnosis, management,and prevention of hepatitis B virus infection. ClinMicrobiol R~ 1999;12:351-366.

3. Liaw YF, Lin DY, Chen TJ, et al. Natural courseafter the development of cirrhosis in patients withchronic type B hepatitis. A prospective study. Liver1989;9:235-241.

4.Beasley RR Epidemiology of hepatocellular carci-noma. In: Vyas GN, Dienstag JL, Hoofnagle JH,eds. Viral Hepatitis and Liver Disease. Orlando, FL:Grune and Stratton; 1984;209-224.

5. Tang ZY. Liver cancer. In: Pollock RE, ed. Manualof Clinical Oncolog)4 7~ edition. Wiley-Liss, Inc.1999:407-424.

6. Sonnenberg FA, Beck JR. Markov models in med-ical decision making. Med Decis Making. 1993;13:322-338.

7. Ministry of Interior. Abridged Life Tables of the Tai-wan-Fuchien Area. Taipei, Republic of China: Statis-tics Bureau; 1996.

8. Life Insurance Association of The Republic of Chi-na. Research Report of Individual Life Insurance Mor-tality Experience of the Taiwan Life Insurance Industry:TS089 Insured Lives Table. Taipei, Republic of Chi-na: Life Insurance Association of The Republic ofChina, 1998.

9. Guan R, Yu HK. Hepatitis B: current strategies forprevention and management. Med Prog. 1997; Feb:21-28.

10. Liaw YE Chu CM, Lin DY, et al. Age-specific prev-alence and significance of hepatitis B e antigen andantibody in chronic hepatitis B virus infection inTaiwan: a comparison among asymptomatic car-riers, chronic hepatitis, liver cirrhosis, and hepa-tocellular carcinoma. J Med Virol. 1984;13:385-391.

11. Lok ASE Lai CL, Wu PC, et al. Spontaneous hep-atitis B e antigen to antibody seroconversion andreversion in Chinese patients with chronic hepati-tis B virus infection. Gastroenterology. 1987;92:1839-1843.

12. Evans AA, O’Connell AP, Pugh JC, et al. Geo-graphic variation in viral load among hepatitis Bcarriers with differing risks of hepatocellular car-cinoma. Cancer Epidemiol Biomarkers Prea. 1998;7:559-565.

13. Yu MW, Hsu FC, Sheen IS, et al. Prospective studyof hepatocellular carcinoma and liver cirrhosis inasymptomatic chronic hepatitis B virus carriers.Am J Epidemiol. 1997;145:1039-1047.

14. Bissell DM. Sex and hepatic fibrosis. Hepatology.1999;29:988-989.

15. Yasuda M. Suppressive effects of estradiol on di-methylnitrosamine-induced fibrosis of the liver inrats. Hepatology. 1999;29:719-727.

16. Chan HLY, Lok ASE Hepatitis B in adults: a clin-ical perspective. Clin Liver Dis. 1999;3:291-307.

17. Chen CJ, Yu MW, Liaw YE Epidemiological char-acteristics and risk factors of hepatocellular carci-noma. J Gastroenterol Hepatol. 1997;12(suppl):S294-$308.

18. Bosch FX, Ribes J, Borras J. Epidemiology of pri-mary liver cancer. Semin Liver Dis. 1999;19:271-285.

19. Liaw YE Tai DI, CHU CM, et al. The developmentof cirrhosis is patients with chronic type B hepa-titis: a prospective study. Hepatology. 1988;8:493-496.

20. Chen PJ, Chen DS. Hepatitis B virus infection andhepatocellular carcinoma: molecular genetics andclinical perspectives. Semin Liver Dis. 1999;19:253-262.

21. Chiaramonte M, Stroffolini T, Vian A, et al. Rate

163


of incidence of hepatocellular carcinoma in pa-tients with compensated viral cirrhosis. Cancer.1999;85:2132-2137.

22. Wong JB, Koff RS, Tin6 F, et al. Cost-effectivenessof interferon-(~2b treatment for hepatitis B e anti-gen-positive chronic hepatitis B. Ann Intern Med.1995;122:664-675.

23. Taiwan Dept of Health and Vital Statistics. Inci-dence Rate of Chronic Liver and Cirrhosis (ICD571).Republic of China; 1997. Taipei.

24. Dusheiko GM, Roberts JA. Treatment of chronictype B and C hepatitis with interferon alfa: an eco-nomic appraisal. Hepatology. 1995;22:1863-1873.

25. Lai CL, Lam KC, Wong KP, et al. Clinical featuresof hepatocellular carcinoma: review of 211 patientsin Hong Kong. Cancer. 1981;47:2746-2755.

26. Shiu W, Dewar G, Leung N, et al. Hepatocellularcarcinoma in Hong Kong: clinical study on 340cases. Oncology. 1990;47:241-245.

27. Lai ECS, Fan ST, Lo CM, et al. Hepatic resectionfor hepatocellular carcinoma: an audit of 343 pa-tients. Ann Surg. 1995;221:291-299.

28. Lau H, Fan ST, Ng IOL, et al. Long term prognosisafter hepatectomy for hepatocellular carcinoma: asurvival analysis of 204 consecutive patients. Can-cer. 1998;83:2302-2311.

29. Lai CL. Hepatocellular carcinoma: an overview.NZ Med J. 1999.

30. Llovet JM, Br~i C, Bruix J. Prognosis of hepatocel-lular carcinoma: the BCLC staging classification.Semin Liver Dis. 1999;19:329-338.

31. Bismuth H, Majno PE, Adam R. Liver transplan-tation for hepatocellular carcinoma. Semin LiverDis. 1999;19:311-322.

32. Incidence Rate of Hepatocellular Carcinoma. TaiwanCancer Registry Annual Report. 1995.

33. Cancer Incidence and Mortality in HK 1994, appendix2.2, HK Cancer Registry 1. 1998.

34. Age-Specific Incidence of Liver Cancer. Singapore Can-cer Registry Report 4, 1988-1992.

35. Survey Report on Cancer Patients in 1992, 1996, Korea.Ministry of Health and Welfare.

36. Incidence Rate of Malignant Neoplasm of the Liver. Tai-wan Health and Vital Statistics; 1997.

37. Cancer Incidence and Mortality in Hong Kong 1994,appendix 4.2, Hong Kong Cancer Registry 1. 1998.

38. Annual Report on the Cause of Death Statistics. Na-tional Statistical Office, Republic of Korea; 1988.

39. Incidence Rate of Chronic Liver Disease and Cirrhosis.Taiwan Health and Vital Statistics; 1997.

40. Bortolotti F, et al. Natural history of chronic viralhepatitis in childhood. Acta Gastro Enterol Belg.1998;61:198-201.

41. Bortolotti F, Jara P, Crivellaro C, et al. Outcome ofchronic hepatitis B in Caucasian children during a20-year observation period. J Hepatol. 1998;29:184-190.

42. Chang MH. Hepatitis B: long-term outcome andbenefits from mass vaccination in children. ActaGastro Enterol Belg. 1998;61:210-213.

43. Lee CM, Lu SN, Changchien CS, et al. Age, gender,and local geographic variations of viral etiology ofhepatocellular carcinoma in a hyperendemic areafor hepatitis B virus infection. Cancer. 1999;86:1143-1150.

44. London WT, Evans AA, McGlynn K, et al. Viral,host and environmental risk factors for hepatocel-lular carcinoma: a prospective study in HaimenCity, China. Intervirology. 1995;38:155-161.

45. Ince N, Wands JR. The increasing incidence of he-patocellular carcinoma. N Engl J Med. 1999;340:798-799.

46. Lemon SM, Layden TJ, Seeff L, et al. The 20th Unit-ed States-Japan Joint Hepatitis Panel Meeting. He-patology. 2000;31:800-806.

47. Shiu W, Dewar G, Leung N, et al. Hepatocellularcarcinoma in Hong Kong: clinical study on 340cases. Oncology. 1990;47:241-245.

48. Munoz N, et al. Epidemiology of hepatocellularcarcinoma. In: Okuda K, Ishak KA, eds. Neoplasmsof the Liver. New York, NY: Springer-Verlag; 1987.

49. Bl~ickberg J, Kidd-Ljunggren K. Genotypic differ-ences in the hepatitis B virus core promoter andprecore sequences during seroconversion fromHBeAg to anti-HBe. J Med Virol. 2000;60:107-112.

50. Chan HLY, Leung NWY, Hussain M, et al. Hepa-titis B e antigen-negative chronic hepatitis B inHong Kong. Hepatology. 2000;31:763-768.

164

long-term morbidity and mortality in chinese insurance … · 2016-12-21 · journal of insurance...

Documents