An Economic Analysis of Different Strategies of ImmunizationAgainst Hepatitis A Virus in Developed Countries

ANANYA DAS

Acute hepatitis A is a major public health problem indeveloped countries, and because a large proportion ofpatients with acute hepatitis A do not have any identifiablerisk factors, current practice of targeting the high-riskgroups for vaccination against hepatitis A virus (HAV) isunlikely to have a significant impact on the overall inci-dence of acute hepatitis A. No economic analysis of strate-gies of mass immunization against HAV is available. Threedifferent strategies of immunization against HAV usingcommercially available inactivated vaccine were comparedin a Markov model analysis of a cohort of 2-year-old healthychildren in a developed country. In strategy I, universalvaccination was pursued. In strategy II, children wereinitially screened for antibody and, if susceptible, they werevaccinated. In strategy III, no vaccination was offered. Costper person and quality adjusted life-years (QALY) gained ineach strategy were the outcome measures compared. Thebaseline analysis showed that strategy II is more cost-effective than strategy I, with marginal cost-effectivenessratios of $7,267.67 and $12,833.34, respectively, comparedwith a strategy of no intervention. Sensitivity analysisshowed that if the cost of the two-dose vaccine could bereduced to less than $57, the strategy of universal vaccina-tion would be the preferred immunization strategy. Differ-ent strategies of mass immunization against HAV in thedeveloped countries are cost-effective by current standardsof health care interventions and should be considered forincorporation into current childhood immunization pro-grams. (HEPATOLOGY 1999;29:548-552.)

Hepatitis A virus (HAV) continues to be a major publichealth problem in developed countries, including the UnitedStates. In 1994, a total of 26,796 cases were reported to theCenter for Disease Control’s National Notifiable DiseasesSurveillance System, and after correcting for under-reportingand asymptomatic infections, an estimated 80,000 cases and134,000 infections occurred in 1994 in the United States.1

Among older children and adults, infection is usually symp-tomatic, and severe diseases related to HAV tend to occurmore frequently in the older age groups.1,2

With improved standard of living in developed countries,

the prevalence of natural immunity among the older popula-tion has been steadily decreasing, as corroborated by sero-epidemiological studies in the United States and otherdeveloped countries.3-5 As a result, more and more of theolder population in the developed countries are becomingsusceptible to HAV infection and potentially life-threateningdisease.

With the development of an inactivated hepatitis A vaccineand demonstration of its effectiveness and safety in large,controlled trials,6-8 hepatitis A vaccination for pre-exposureprophylaxis is recommended for groups at high risk forhepatitis A infection, namely travelers to endemic areas,homosexuals, and intravenous drug abusers, persons who areat risk for occupational exposure to HAV, food handlers, andpatients with chronic liver disease and clotting disorders.1,9

Taking a lesson from the experience with hepatitis Bvaccination programs along with the fact that 35% to 50% ofpatients with acute hepatitis A do not have any identifiablerisk factor,10 it is unlikely that the vaccination strategy oftargeting the high-risk groups alone will have any significantimpact on the overall incidence of acute hepatitis A, and astrategy of universal vaccination will be needed to preventepidemics and ultimately eradicate this preventable butpotentially serious disease.2,11

However, before such a strategy of universal vaccination isadopted, it is important to study the cost-effectiveness of suchan approach and compare it with other strategies of vaccina-tion.

MATERIALS AND METHODS

To study the cost-effectiveness of various strategies of widespreadvaccination of an appropriate susceptible population against hepati-tis A in developed countries from a societal perspective, a Markovmodel was constructed using decision-analysis software (DATA 3.0,TreeAge Softwares, Inc., Williamstown, MA).

The Markov model is a decision-analysis tool widely used tostudy the impact of different health management strategies, particu-larly in the context of a protracted time horizon and recurring risk ofa particular disease, and is useful in studying the cost-effectivenessof different vaccination programs.12-14 Briefly, in a Markov model,different states of health (with allocated reward values for being inthat state) associated with a particular disease are defined. Transi-tional probabilities from one state to another are obtained fromavailable information. The model is run for several cycles tosimulate the natural history of the problem, and at the end of eachcycle, cumulative reward values in terms of outcome measures arecalculated.15

In this cost analysis, two different strategies of vaccination againsthepatitis A using the commercially available hepatitis A vaccine (aformalin-inactivated HAV vaccine, Havrix, manufactured by Smith-Kline Beecham Biologicals, Philadelphia, PA) was compared with astrategy of no vaccination in a Markov cycle tree simulation model

Abbreviations: HAV, hepatitis A virus; QALY, quality adjusted life-years.From the Division of Gastroenterology, Case Western Reserve University, University

Hospitals of Cleveland, Cleveland, OH.Received June 24, 1998; accepted September 23, 1998.Address reprint requests to: Ananya Das, M.D., 80 West Center Street, Apt. 402,

Akron, OH 44308. E-mail: msinha@ix.netcom.com; fax: (330) 374-7714.Copyright r 1999 by the American Association for the Study of Liver Diseases.0270-9139/99/2902-0031$3.00/0

548

(Fig. 1). The simulation model followed a cohort of 10,0002-year-old healthy children over their lifetime in cycles of 1-yearduration. The time horizon of the analysis was based on the averagelife expectancy of a healthy 2-year-old child, which is 74.5 years (allraces, both sexes, life expectancy obtained from U.S. life table,1992). During each cycle, subjects could occupy one of severalhealth states associated with hepatitis A infection, including death,acute hepatitis, well but susceptible to hepatitis A infection, wellwith natural immunity, and well with vaccine-induced immunityfrom hepatitis A. If allowed in the model, the subjects could maketransition from one state of health to another based on time-dependent transitional probabilities (Fig. 2). Death is considered anabsorbing state from which no transition is allowed to other states.At the end of each cycle, the simulation model accrued units ofutility and cost for each subject in a given state of health. The cohortof subjects accumulated utility and cost units until either deathoccurred or the total number of cycles was completed. Utility foreach health state was adjusted by the quality of life associated withthat state of health. In the present model, quality of life associatedwith the states of well but susceptible, well with natural immunity,and well with vaccine-induced immunity were taken as unity; inestimating the quality of life for the state of acute hepatitis, in thebaseline analysis, a dysutility of 30 days was attributed to an episodeof acute hepatitis A.1,16,17 A sensitivity analysis was also performedby varying hepatitis A–related disutility from 10 to 40 days.

In strategy I, universal vaccination of all subjects withoutconsideration of the status of their natural immunity was pursued.Each subject received two doses of the vaccine at 0 and 6 to 12months. Most subjects who were compliant with the vaccinationprogram entered the Markov model in the state of vaccine immunity,although a minority did not undergo seroconversion after vaccina-tion and entered the Markov cycle in the state of well but susceptibleto hepatitis A infection. In strategy II, all subjects were initiallyscreened for the presence of protective antibody by a commercialimmunoassay (HAVAB EIA, Abbott Laboratories, Chicago, IL), andif susceptible, they were vaccinated in the two-dose vaccinationschedule. In this strategy, most subjects who were compliant withthe screening and vaccination schedules started the Markov cycleeither in the vaccine-induced or natural immunity state. Noncompli-ant subjects either in strategies I and II and all subjects in strategyIII, in which no vaccination was offered, depending on if they hadnatural immunity or not, started the Markov cycle either in the stateof natural immunity or in the well-but-susceptible state, respec-tively. All subjects with natural immunity either continued to

occupy the same state or died in subsequent cycles, depending ontheir age-adjusted baseline mortality rates.

The transitional probabilities under different strategies wereobtained from literature or expert opinion (Table 1). In this model, itwas assumed that all subjects after being in a state of acute hepatitiswould acquire natural immunity and would be immune from furtherinfection with hepatitis A. In addition, hepatitis A vaccination isassociated with no major adverse side effects,7 and the medical costof minor side effects of vaccination are negligible. It was assumedthat 80% of the target population would comply with vaccination instrategy I or the initial screening test in strategy II. It was furtherassumed that 80% of all subjects who complied with the screeningtest in strategy II would also complete the two-dose vaccinationschedules. A level of anti-HAV antibody .20 mIU/mL by themodified HAVAB assay was considered protective.8 For the purposeof simplicity, the sensitivity and specificity of the screening test wastaken as 100%.18

Cost estimates for an acute episode of hepatitis A were obtainedfrom available literature, and average direct medical costs based onhospital charges for an episode of acute hepatitis A varied from$1,070 to $2,460.19,20 Because this analysis was performed from asocietal perspective, it was important also to consider all relevantdirect and indirect costs to the society. Dalton et al. calculated thetotal cost incurred in a food-borne outbreak of 43 secondary cases ofhepatitis A associated with a catering facility in Denver, CO.19 Thetotal costs including cost of disease control in the form of state andlocal health-department personnel costs, provision and administra-tion of immune globulin, costs of additional hepatitis A serologicaltests performed, business losses, loss of work, and medical costsamounted to $809,706 ($18,830 per case of hepatitis A). However,this study included costs such as business losses from discardingpotentially contaminated food (from the catering service), and also,a much higher number of contacts per case of hepatitis A receivedimmune globulin prophylaxis. In the current model, for baselineanalysis, a more modest sum of $2,500 was taken as the total cost ofacute hepatitis A from the societal perspective, and a sensitivityanalysis was performed by varying the cost over a wide range. Costestimates for antibody testing and cost of vaccination (along withcost of administration) were obtained from an ambulatory practicein a university hospital setting and were compared with informationavailable in the literature.2,21 It was assumed that the hepatitis Avaccine schedule would be completely integrated into the currentchildhood immunization schedule, thus obviating the need for anyextra visits by children and their parents to the health care provider.Thus, no extra cost, either direct or indirect resulting from extravisits for hepatitis A vaccination, was considered in the presentmodel.

FIG. 1. Decision tree representing different strategies of vaccinationagainst hepatitis A. (j), The decision node where a particular strategy ischosen; the circles denote chance nodes representing compliance with theimmunization program and probability of presence of natural immunityagainst hepatitis A. In all the three strategies, the cohorts enter respectiveMarkov nodes (represented by ‘‘M’’) at the extreme right of the decision tree.

FIG. 2. Different Markov states of health associated with HAV infection,with the arrows denoting possible transitions from one state to another asallowed in the model in a single Markov cycle.

HEPATOLOGY Vol. 29, No. 2, 1999 DAS 549

All future costs were discounted at 3%. In the baseline analysis,nonmonetary utilities such as life-years gained were also discountedat 3%. Sensitivity analysis was performed by varying the discountrate from 0% to 7%. Because in a Markov process it is assumed thatall transitions occur at the end of a cycle while transitions actuallyoccur throughout the cycle, half-cycle corrections were performedfor both cost and utility values. Population-based mortality ratesadjusted for age, sex, and race were obtained from published UnitedStates vital statistics.22 Rates to probability transformations wereperformed using standard formulas.15 Sensitivity analyses wereperformed by varying different variables either one at a time orsimultaneously over a wide range to test the robustness of theconclusions of the baseline analysis.

RESULTS

Results of the baseline analysis in the present model areshown in Table 2. No strategies were clearly dominated byany other. The strategy of no vaccination was the cheapest,and the strategy of universal vaccination was the mosteffective strategy in terms of quality adjusted life-years(QALY) gained; the strategy of screening followed by vaccina-tion of the susceptible population was intermediate incost-effectiveness. While the marginal cost-effectiveness ratio

for the strategy of universal vaccination was higher at$12,833.34, the marginal cost-effectiveness ratio for thestrategy of screening and vaccination compared with thestrategy of no vaccination was $7,266.67.

When the baseline analysis was repeated by discountingthe cost and future utilities at 5%, the cost, effectiveness, andmarginal cost-effective ratio of strategies I and II were $57.28,19.59960 QALY, $11,000.00 and $54.20, 19.59932 QALY,and $5,739.08, respectively. In one-way sensitivity analysiswith the discount rate varying from 0% to 7%, no change inthe conclusions of the baseline analysis were seen. Inone-way sensitivity analysis performed with the probability ofseroconversion after vaccination within the range reported inliterature did not alter the conclusions of the baselineanalysis. Similarly, one-way sensitivity analysis performedafter varying hepatitis A–related disutility from 10 to 40 daysdid not change the results of the baseline analysis. One-waysensitivity analysis with the probability of initial compliancewith the vaccination program showed that when the probabil-ity of compliance falls below 65%, the strategy of universalvaccination was preferred to the strategy of screening andvaccination.

One-way sensitivity analyses with the total cost of twodoses of vaccine and the cost of the serological test for theantibody to HAV critically influenced the results of theanalysis. If the cost of the two-dose vaccine was less than $57or the cost of the serological test was more than $21, then thestrategy of universal vaccination was preferred to the screen-and-vaccination strategy. Similarly, when the total cost of anacute episode of hepatitis A exceeds $11,215, the strategy ofuniversal vaccination becomes the strategy of choice.

As expected, the annual incidence of acute hepatitis A inthe members of the cohort who were not immune to hepatitisA strongly influenced the results of the analysis (Fig. 3). Forbaseline analysis, an incidence rate of 0.01% was used, whichis characteristic of a low endemic area like the United States.1

In the one-way sensitivity analysis, when the annual inci-dence rate was approximately triple that rate, the strategy ofuniversal vaccination was favored between the two immuni-zation strategies.

When the age of the cohort population in the model wasvaried over a wide range for the purpose of a one-waysensitivity analysis between the two immunization strategies,the strategy of screening and vaccination continued to be themore cost-effective strategy throughout the strategy.

Figure 4 shows the result of a two-way sensitivity analysisin which the two variables of cost of vaccination (cost of thetwo-dose vaccine and administration costs) and cost of theserological test for anti-HAV antibodies were varied simulta-neously to determine the choice between the two immuniza-tion strategies. Within the range used for this analysis, with

TABLE 1. Clinical and Cost Variables Used in the Analysis

Variable Baseline (Range) References

Probability of compli-ance with vaccina-tion program

0.8 (0.5-1.0) Roberts33

Annual rate of declineof vaccine inducedantibody

12 (0-15)% Van Damme31

Probability of naturalimmunity againstHAV in differentage groups

0-10 years11-20 years21-30 years31-40 years41-50 years51-70 years.70 years

10%15%18%35%45%70%75%

Lemon,20 data from thethird National Healthand Nutrition Exami-nation Survey(NHANES)

Annual incidence ofhepatitis A infec-tion in the nonim-munized popula-tion in the modelcohort

0.0001 (0-0.001) ACIP1

Case fatality rate foracute hepatitis Ainfection

ACIP,1 Bader2

,50 years of age 0.3 (0.1-0.5)%.50 years of age 2.5 (1.0-3.0)%

Cost estimates ($)Total cost of an acute

episode of hepatitisA

2,500 (1,000-20,000) Dalton,19 Lemon20

Total cost of twodoses of hepatitis Avaccination(including cost ofadministration)

70 (25-150) Bader,2 Smith,16 Bryan21

Cost of serologicaltest for hepatitis Aantibody

15 (5-50) Bader,2 Bryan21

TABLE 2. Results of the Baseline Analysis

StrategyCost($)

Effectiveness(QALY)

C/E Ratio($/QALY)

MarginalC/E Ratio($/QALY)

Strategy III (no vaccination) $ 6.52 29.42402 0.22 —Strategy II (screen and vacci-

nation) $55.57 29.43077 1.89 $ 7,267.67Strategy I (universal vaccina-

tion) $59.42 29.43107 2.02 $12,833.34

Abbreviation: C/E, cost-effectiveness.

550 DAS HEPATOLOGY February 1999

higher cost of vaccination and lower cost of the screeningtest, the strategy of screening and vaccination is preferred tothe strategy of universal vaccination.

DISCUSSION

Although HAV infection is a self-limited disease, withsevere illness and death occurring in a minority of patients, itis a significant cause of morbidity and mortality, and conse-quently economic losses (an estimated annual cost of morethan $200 million in the United States alone), in developedcountries around the world.23 With sustained improvementin sanitation and hygiene in developed countries, the peakrate of infection has shifted from early childhood, wheninfection is largely asymptomatic, to older age groups withmore severe clinical disease expression, thus paradoxicallyincreasing the burden of clinical disease. With the availabilityof a safe and effective vaccine against HAV infection, allmortality and morbidity related to HAV infection should beconsidered preventable. Most cases of hepatitis A in devel-oped countries occur in community-wide epidemics ratherthan point-source outbreaks, and maintenance of an appropri-

ate level of herd immunity would stop virus transmission andeffectively eliminate epidemics.24 Current recommendationsfor hepatitis A vaccination are aimed at the high-risk groups,and because a significant proportion of patients with acutehepatitis A do not have any identifiable risk factors, it islogical to infer that strategies of vaccination targeting a largersection of the susceptible population will be more effective indecreasing and ultimately eradicating this disease.

Few cost-analysis studies establishing the cost-effective-ness of hepatitis A vaccination in the high-risk groupsexist,16,25-27 but none of these studies attempt to study thecost-effectiveness of vaccination strategies that pursue wide-spread vaccination of the population in developed countriesincluding a strategy of universal vaccination. The presentstudy was performed with the purpose of assessing if massimmunization of young children against hepatitis A wouldalso be cost-effective, because such an approach seems to beepidemiologically sound for achieving effective control andultimate eradication of hepatitis A.1,11 Because the presentstudy deals primarily with a cohort of healthy 2-year-oldchildren and routine immunization of healthy children againsthepatitis A is not currently practiced, the strategy of ‘‘novaccination’’ was considered as the appropriate baselinestrategy against which other immunization strategies werecompared.

In the current cost-utility analysis, comparing differentstrategies of vaccination in a cohort of 2-year-old children in adeveloped country, the strategy of no vaccination was thecheapest, but least effective, and the strategy of universalvaccination was the costliest, but most effective, strategy. Thestrategy of screening and vaccination had a marginal cost-effectiveness ratio (compared with strategy of no vaccination)of $7,267.67. Although hepatitis A vaccination strategies arenot as cost-effective as current childhood immunizationprograms like universal vaccination against hepatitis B virus,the marginal cost-effectiveness of these strategies of vaccina-tion against hepatitis A are well within the range currentlyaccepted to be cost-effective in the arena of health careinterventions.28 In situations in which the cost of the serologi-cal tests was high, the cost of the vaccine was low, or the totalcost of an acute episode of hepatitis A was high, the strategyof universal vaccination was the clearly favored strategy.

Several points must be clarified in interpreting the resultsof the current analysis. Because of their critical role in HAVtransmission, children should be the primary focus of immu-nization strategies to lower disease incidence.1 However,because the data regarding the safety and effectiveness ofhepatitis A vaccine in children less than 2 years of age arelacking, this age group was not considered in the currentanalysis. Because the older age groups tend to have moresevere disease, and infection in the children tends to belargely asymptomatic, it is attractive to consider the strategyof immunizing the older adults (catch-up immunization)only. In our model, when a sensitivity analysis was performedby varying the age of the model cohort at which immuniza-tion was administered, the conclusions of the baselineanalysis were unchanged. However, the shifting of the age ofvaccination toward the older age groups may have adverseepidemiological implications, because the highest rate ofinfection occurs in the children, and they act as the mostimportant reservoir of infection in the community. It is alsoimportant to note that the current model does not take intoaccount the dynamic effect of a program of vaccination on the

FIG. 3. One-way sensitivity analysis with the annual incidence rate ofhepatitis A infection plotted on the x-axis and the cost-effectiveness ratioalong the y-axis. In the baseline analysis, with an annual rate of incidence at0.0001, the strategy of screening and vaccination is more cost-effective, butas the annual incidence rate crosses the threshold value of 0.00027, universalvaccination becomes the preferred strategy. (d), Universal vaccination; (j),screen and vaccinate.

FIG. 4. The result of a two-way sensitivity analysis in which the twovariables of cost of vaccination and cost of the serological test for anti–hepatitis A antibodies were varied simultaneously to determine the choicebetween the two immunization strategies. Within the range used for thisanalysis, with higher cost of vacccination and lower cost of the screening test,the strategy of screening and vaccination is preferred to the strategy ofuniversal vaccination. (i), Universal vaccination; (°), screen and vaccinate.

HEPATOLOGY Vol. 29, No. 2, 1999 DAS 551

‘‘herd immunity’’ of successive birth cohorts. Understandably,as the vaccination program proceeds, the background level ofinfection in the community declines, so that the incrementalbenefits from the continued vaccination program experi-enced by successive cohorts tend to decline. Althoughchronic disease does not occur with hepatitis A infection, inup to 6% to 10% of cases, a relapse of acute hepatitis may beseen.29 This has not been considered in the current model.

Because a larger number of persons have been vaccinatedwith HAVRIX compared with the recently licensed VAQTA(manufactured by Merck & Company, Inc.), the currentmodel uses data mostly from the HAVRIX trials. However, thetwo vaccines seem to be comparable in their safety, immuno-genecity, and effectiveness.1

In the current model, only the two-dose vaccinationschedule (one primary and another booster dose at 6- to12-month intervals) has been considered. The addition of athird dose does not seem to significantly improve thealready-high seroconversion rates associated with the two-dose schedule,30 and will surely add to the overall cost of thevaccination program and may decrease the compliance rate.Sensitivity analysis done with the probability of seroconver-sion after vaccination within the range reported in theliterature did not alter the results of the analysis.

Although clinical data on long-term persistence of vaccine-induced hepatitis A antibody are not available, based onmathematical models that assume continuous logarithmicdecline of antibody over time, the predicted duration ofvaccine immunity is in the range of 25 years.31 In the currentmodel, an annual rate of decline in antibody titer was taken as12% for baseline analysis, and sensitivity analysis using arange varying from 0% to 15% did not alter the conclusions ofthe analysis.

Development of vaccines combining inactivated hepatitisA with other vaccines in a single formulation for readyintegration into the current childhood immunization pro-gram is likely in the near future.32 Such a vaccine would besignificantly advantageous in terms of administrative re-sources, parent and physician acceptance, and compliance,and surely would enhance the cost-effectiveness of strategiesof vaccination against hepatitis A.

In summary, in the current cost-utility analysis of differentstrategies of vaccination against HAV, more widespread immu-nization of the susceptible population in developed countrieswas found to be reasonably cost-effective and should beactively considered for incorporation into current immuniza-tion programs. While in the current analysis, a strategy ofinitial screening for protective antibody followed by vaccina-tion of the susceptible population turned out to be morecost-effective, universal vaccination with a lower-priced vac-cine will undoubtedly be the most effective way to controland ultimately eradicate this disease.

REFERENCES

1. Prevention of hepatitis A through active or passive immunization:recommendations of the advisory committee on immunization practices(ACIP). MMWR December 27, 1996;45(RR-15):1-30.

2. Bader TF. Hepatitis A vaccine. Am J Gastroenterol 1996;91:217-222.3. Shapiro CN, Coleman PJ, McQuillan GM, Alter MJ, Margolis HS.

Epidemiology of hepatitis A: seroepidemiology and risk groups in theUSA. Vaccine 1992;10(Suppl 1):59-62.

4. Frosner GG, Papaevangelou G, Butler R, Iwarson S, Lindholm A,Courouce-Pauty A, Haas H, et al. Antibody against hepatitis A in sevenEuropean countries. I. Comparison of prevalence data in different agegroups. Am J Epidemiol 1979;110:63-69.

5. Gust ID. Epidemiological patterns of hepatitis A in different parts of theworld. Vaccine 1992;10(Suppl 1):56-58.

6. Werzberger A, Mensch B, Kuter B, Brown L, Lewis J, Sitrin R, Miler W, etal. A controlled trial of a formalin-inactivated hepatitis A vaccine inhealthy children. N Engl J Med 1992;327:453-457.

7. Clemens R, Safary A, Hepburn A, Roche C, Stanbury WJ, Andre FE.Clinical experience with an inactivated hepatitis A vaccine. J Infect Dis1995;171(Suppl 1):S44-S49.

8. Innis BL, Snitbhan R, Kunasol P, Laorakpongse T, Poopatanakool W,Kozik C, Suntayakorn S, et al. Protection against hepatitis A by aninactivated vaccine. JAMA 1994;271:1328-1334.

9. Public health control of hepatitis A: memorandum from a WHOmeeting. Bull World Health Organ 1995;73(1):15-20.

10. CDC. Hepatitis surveillance report no. 56. Atlanta: US Department ofHealth and Human Services, Public Health Service, CDC, 1996.

11. Hollinger FB, Eickoff T, Gershon A, Jong EC, Koff RS. Discussion: whoshould receive hepatitis A vaccine? A strategy for controlling hepatitis Ain the United States. J Infect Dis 1995;171(Suppl 1):S73-S77.

12. Krahn M, Detsky AS. Should Canada and the United States universallyvaccinate infants against hepatitis B. Med Decis Making 1993;13:4-20.

13. Wiebe T, Fergusson P, Horne D, Shanahan M, Macdonald A, Heise L,Roos LL. Hepatitis B immunization in a low-incidence province ofCanada. Med Decis Making 1997;17:472-482.

14. Fenn P, Gray A, McGuire A. An economic evaluation of universalvaccination against hepatitis B virus. J Infect 1996;32:197-204.

15. Beck JR, Pauker SG. The Markov process in medical prognosis. MedDecis Making 1983;3:419-458.

16. Smith S, Weber S, Wiblin T, Nettleman M. Cost-effectiveness of hepatitisA vaccination in healthcare workers. Infect Control Hosp Epidemiol1997;18:688-691.

17. Stefen R, Rickenbach M, Wilhelm U, Helminger A, Schar M. Healthproblems after travelling to developing countries. J Infect Dis 1987;156:84-91.

18. Polesky H, Hanson M. Comparison of viral hepatitis marker testmethods based on AABP-CAP survey data. Am J Clin Pathol 1981;76:521-524.

19. Dalton CB, Haddix A, Hoffman RE, Mast EE. The costs of a food-borneoutbreak of hepatitis A in Denver, Colo. Arch Intern Med 1996;156:1013-1016.

20. Lemon SM, Shapiro CN. The value of immunization against hepatitis A.Infect Agents Dis 1994;1:38-49.

21. Bryan JP, Nelson M. Testing for antibody to hepatitis A to decrease thecost of hepatitis A prophylaxis with immune globulin or hepatitis Avaccines. Arch Intern Med 1994;154:663-668.

22. US Bureau of the Census: Statistical Abstract of the United States. 113thed. Washington, DC, 1993.

23. Haddler SC. Global impact of hepatitis A virus infection: changingpatterns. In: Hollinger FB, Lemon SM, Margolis HS, eds. Viral Hepatitisand Liver Disease. Baltimore: Williams & Wilkins, 1991:14-20.

24. Margolis HS, Alter MJ. Will hepatitis A become a vaccine preventabledisease? Ann Intern Med 1995;122:464-465.

25. Doorslaer EV, Tormans G, Van Damme P. Cost-effectiveness analysis ofvaccination against hepatitis A in travellers. J Med Virol 1994;44:463-469.

26. Gutherson T, Steffen R, Van Damme P, Holdener F, Beutels P. Hepatits Ainfection in aircrews: risk of infection and cost-benefit analysis ofhepatitis A vaccination. Aviat Space Environ Med 1996;67:153-156.

27. O’Connor JF, Imperiale TF, Singer ME. Is hepatitis A vaccination foradults cost-effective? [Abstract] Med Decis Making 1997;17:530A.

28. Tengs TO, Adams ME, Pliskin JS, Safran DG, Siegel JE, Weinstein MC,Graham JD. Five-hundred life-saving interventions and their cost-effectiveness. Risk Anal 1995;15:369-390.

29. Glikson M, Galun E, Oren R, Tur-Kaspa R, Shouval D. Relapsinghepatitis A. Review of 14 cases and literature survey. Medicine 1992;71:14-23.

30. Chen XQ, Bulbul M, De Gast GC, Van Loon AM, Nalin DR, Van Hattun J.Immunogenecity of two versus three injections of inactivated hepatitis Avaccine in adults. J Hepatol 1997;26(2):260-264.

31. Van Damme P, Thoelen S, Cramm M, De Groote K, Safary A, Meheus A.Inactivated hepatitis A vaccine: reactogenecity, immunogenecity andlong term antibody persistence. J Med Virol 1994;44:446-451.

32. Kane MA. Perspectives on the control of hepatitis A by vaccination.Vaccine 1992;10(Suppl 1):S93-S96.

33. Roberts JD, Poffernorth LA, Roos LL, Bebchuk JD, Carter AO. Monitor-ing childhood immunizations: a Canadian Approach. Am J Public Health1994;84:1666-1668.

552 DAS HEPATOLOGY February 1999