Antibody Levels and Protection after Hepatitis B Vaccination: Results
of a 15-Year Follow-up
Brian J. McMahon, MD; Dana L. Bruden, MS; Kenneth M. Petersen, MD; Lisa R. Bulkow, MS; Alan J. Parkinson, PhD; Omana Nainan, PhD;
Marina Khristova, PhD, DSc; Carolyn Zanis, BS; Helen Peters, BS; and Harold S. Margolis, MD
Background: The duration of protection afforded by hepatitis B
vaccination is unknown.
Objective: To determine antibody persistence and protection
from hepatitis B virus (HBV) infection.
Design: Prospective cohort study.
Setting: 15 villages in southwest Alaska.
Participants: 1578 Alaska Natives vaccinated at age 6 months
Intervention: During 1981–1982, participants received 3 doses
of plasma-derived hepatitis B vaccine. This cohort was followed
annually over the first 11 years, and 841 (53%) persons were
tested at 15 years.
Measurements: Antibody to hepatitis B surface antigen (anti-
HBs), markers of HBV infection, and testing to identify HBV
Results: Levels of anti-HBs in the cohort decreased from a geo-
metric mean concentration of 822 mIU/mL after vaccination to 27
mIU/mL at 15 years. Initial anti-HBs level, older age at vaccina-
tion, and male sex were associated with persistence of higher
anti-HBs levels at 15 years when analyzed by a longitudinal linear
mixed model. After adjustment for initial anti-HBs level and sex,
those vaccinated at age 6 months to 4 years had the lowest
anti-HBs level at 15 years. Asymptomatic breakthrough infections
were detected in 16 participants and occurred more frequently in
persons who did not respond to vaccination than those who
responded (P ? 0.01). Among infected persons with viremia, 2
were infected with wild-type HBV and 4 had HBV surface glyco-
protein variants, generally accompanied by wild-type HBV.
Limitations: The loss of participants to follow-up at 15 years
was 47%. However, characteristics of persons tested were similar
to those of persons lost to follow-up.
Conclusions: Hepatitis B vaccination strongly protected against
infection for at least 15 years in all age groups. Antibody levels
decreased the most among persons immunized at 4 years of age
Ann Intern Med. 2005;142:333-341.
For author affiliations, see end of text.
countries and has been shown to be effective in reducing
the rate of chronic hepatitis B virus (HBV) infection (1, 2).
Protection has been demonstrated in persons and popula-
tions vaccinated for 5 to 10 years, and rates of asymptom-
atic breakthrough HBV infection have been extremely low
(3–9). However, the duration of protection afforded by
hepatitis B vaccination beyond 10 years and the possible
need for booster doses of this vaccine are unknown.
Alaska Natives have a high prevalence of chronic HBV
infection, primarily acquired during early childhood (10).
Between November 1981 and May 1982, Alaska Natives
residing in 15 villages in southwest Alaska were enrolled in
a cohort study to ascertain the immunogenicity and long-
term effectiveness of hepatitis B vaccination (11–14). We
report data on the persistence of antibodies to hepatitis B
surface antigen (anti-HBs), incidence of HBV infection,
and the genetic characteristics of the HBV isolates in per-
sons with breakthrough infections 15 years after initial vac-
cination of this cohort.
niversal vaccination of infants with hepatitis B vaccine
is included in the immunization programs of most
Participants and Data Collection
A total of 1578 Alaska Natives who were serologically
negative for hepatitis B surface antigen (HBsAg) and anti-
body to hepatitis B core antigen (anti-HBc) were vacci-
nated on a 0-, 1-, and 6-month schedule with 3 doses of
plasma-derived hepatitis B vaccine (Heptavax, Merck &
Co., Inc., West Point, Pennsylvania) beginning in 1981
(11). Persons younger than 20 years of age received the
10-?g dose, and adults received the 20-?g dose. Of the
1578 persons vaccinated, 1436 (91%) were tested for an
anti-HBs response 6 months after the last vaccine dose.
From 1982 to 1992, serum specimens were obtained
annually and once again during 1996 from as many of the
1578 consenting participants as possible. The Institutional
Review Boards of the Alaska Area Native Health Service,
the Indian Health Service, the Centers for Disease Control
and Prevention, and the Yukon-Kuskokwim Health Cor-
poration and the Norton Sound Regional Alaska Native
health boards approved this study. All participants 18 years
of age and older and parents of children younger than 18
years of age had provided signed informed consent to par-
Editors’ Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
Editorial comment . . . . . . . . . . . . . . . . . . . . . . . . . 384
Summary for Patients . . . . . . . . . . . . . . . . . . . . . . I-34
Conversion of tables into slides
1 March 2005 Annals of Internal Medicine Volume 142 • Number 5 333
ticipate in the study; children older than 7 years of age gave
verbal assent. The number of HBsAg-positive persons in
each village was obtained from a registry used to follow
patients with chronic HBV infection (15).
All serum specimens were tested for HBsAg, anti-HBs,
and anti-HBc by radioimmunoassay using commercial test
kits (Abbott Laboratories, Abbott Park, Illinois). At the
initial testing of the cohort for anti-HBs, results were re-
ported in sample ratio units. However, subsequent anti-
HBs results were reported in milli international units
(mIU) per mL using a World Health Organization refer-
ence standard (12–14). To ensure comparability of results
over time, all serum specimens from each participant with
sufficient volume (99.9% of all specimens collected during
the study) were retested to determine anti-HBs levels in
Detection of HBV DNA and Nucleic Acid Sequencing
Hepatitis B virus DNA was extracted from serum
specimens (50 ?L) of participants with serologic markers
of HBV infection by using commercially available reagents
(MasterPure Complete DNA and RNA Purification Kit,
Epicentre Technologies, Madison, Wisconsin), as described
previously (16, 17). The HBsAg genomic region was then
amplified by dilution cloning polymerase chain reaction by
using previously described primers and methods to identify
circulating variants of HBsAg (16, 17). The polymerase
chain reaction products were purified and the nucleic acid
sequence of the amplified region were determined by using
prism dye or dRhodamine terminator cycle reactions (Ap-
plied Biosystems, Foster City, California) and automated
sequencing (ABI Model 373 or 377, Applied Biosystems)
(18). Sequence data were further analyzed by Sequence
Navigator (ABI) and GCG software (19).
The initial anti-HBs level was measured 6 months af-
ter the third dose of vaccine and 1 year after the first dose
of vaccine. Participants with an initial anti-HBs level of at
least 10 mIU/mL were considered vaccine responders. An
anti-HBs level of 2 mIU/mL or greater was considered a
positive result on subsequent specimens. A booster re-
sponse was defined as a 2-fold or greater increase in anti-
HBs levels between serologic test results.
A definite HBV infection in a participant was defined
as 1) 2 or more consecutive serum specimens positive for
anti-HBc more than 1 year after the initial vaccine dose, 2)
a single positive anti-HBc result with a positive HBV DNA
result, or 3) any HBsAg-positive test result. A “possible”
HBV infection was defined as a single positive or 2 non-
consecutive positive anti-HBc results. Participants who de-
veloped anti-HBc were interviewed for history of icterus or
other clinical signs or symptoms compatible with acute
hepatitis, and village and hospital medical records were
reviewed for evidence of an illness compatible with viral
Among persons who inadvertently received additional
doses of hepatitis B vaccine during the follow-up period,
anti-HBs results after the additional vaccine dose or doses
were excluded from the analyses. Results for anti-HBs
among persons with definite HBV infections were ex-
cluded from analyses after anti-HBc appeared.
The primary outcomes in this study were the cumula-
tive number of persons with a definite HBV infection dur-
ing all follow-up years and the anti-HBs levels at the 15-
year follow-up. The definitions of age classes (0 to 4 years,
5 to 19 years, 20 to 49 years, ?50 years) were similar to
those in a previously published analysis of this cohort (14).
Although these data have been presented previously (11–
14), we have provided the proportion of persons initially
responding to vaccination. Quantitative anti-HBs levels are
presented as geometric mean concentrations (GMCs). In
bivariate analyses, analysis of variance was used to test the
15-year anti-HBs concentrations (log-transformed). Inci-
dence rates of definite HBV infection were compared by
using the Fisher exact test.
We analyzed factors associated with anti-HBs levels
over the 15 years after the first vaccine dose by using a
linear mixed model (PROC MIXED in SAS version 9.1,
SAS Institute, Inc., Cary, North Carolina) (19). We chose
a longitudinal mixed linear model because it makes infer-
ences by using information from the entire cohort collected
at all follow-up time points. Levels of anti-HBs were log-
transformed before analysis, and concentrations of 0
mIU/mL were assigned a value of 1.0 mIU/mL. Factors
considered in the model were time (entered as a continu-
Although administration of hepatitis B vaccine for infants
is routine practice in many countries, we do not know
whether the protection that this vaccine offers lasts be-
yond 10 years. Such information is essential to develop
policies about booster vaccination.
Of 841 Alaska Natives who received 3 doses of hepatitis B
vaccination during 1981–1982 and were followed for 15
years, 84% had protective levels of antibody to hepatitis B
surface antigen that indicated continued protection. The
greatest decline in antibody levels occurred in people who
received vaccine before 4 years of age. Definite asymp-
tomatic breakthrough infections occurred in 16 partici-
Only about half of the initial cohort of 1578 was available
for testing at 15 years.
Antibody Levels and Protection after Hepatitis B Vaccination
334 1 March 2005 Annals of Internal Medicine Volume 142 • Number 5
ous covariate; linear and quadratic term were considered),
age class at initial vaccination, sex, the log of the initial
anti-HBs level, presence of an HBsAg-infected person in
the household at the start of the study, and the proportion
of village residents with chronic HBV infection at the end
of follow-up, along with interaction terms involving time.
Significance of 1 factor alone, such as age or sex, is called a
main effect and is not of primary interest for this presen-
tation. Of primary interest are the interaction terms be-
tween time and other factors. A significant interaction of
time with another variable, such as sex, indicates that the
decline in anti-HBs level differed between males and fe-
males. We obtained parameter estimates by using restricted
maximum likelihood. We used an unstructured covariance
matrix to account for dependence of observations across
time within individuals. Backward elimination of statisti-
cally nonsignificant terms yielded a final model of main
effects and time interaction terms. If the time interaction
term was statistically significant, the main effect term re-
mained in the model regardless of statistical significance.
We used the Wald chi-square statistic to test covariates.
Contrast tests were 2-sided, and an ? level of 0.05 was
required. We used residual plots to evaluate model fit.
A secondary outcome was a boost in anti-HBs level at
the 11- or 15-year follow-up among persons without addi-
tional doses of vaccine. We used the chi-square test or the
Cochran–Mantel–Haenszel test to compare age and sex of
persons with a booster response to those of persons without
a booster response at both follow-up years. All P values
were exact where appropriate and were 2-sided; results
were considered statistically significant at the 0.05 level.
We conducted analyses by using StatXact4 (Cytel Software
Corp., Cambridge, Massachusetts) and SAS software, ver-
sion 9.1 (SAS Institute, Inc.).
Throughout the entire study, anti-HBs determinations
were observed for 68% of all potential observational time
points (Appendix Table 1, available at www.annals.org).
The 15 remote rural Alaskan study villages are accessible
only by airplane. During each year, study personnel flew
into each village for 1 to 2 days and attempted to recruit all
available participants. Persons not available or out of the
village for the day were considered missed for that time
point. Persons who moved out of 1 of the 15 original
villages were considered lost to follow-up unless they relo-
cated to Anchorage, Alaska, or the regional hub city, Be-
thel. Primary analysis used a linear mixed model, which
assumed that the missing data were missing at random
(dependent on observed data but independent of the anti-
HBs level at the time of the missed observation). Appendix
Table 1 shows results on the extent of missing data.
Characteristics of the 783 persons tested at 15 years
and their respective anti-HBs levels were compared with
those of persons not participating (Appendix Table 2,
available at www.annals.org). In univariable analysis, we
used the chi-square test and the Wilcoxon rank-sum test to
compare categorical and continuous covariates, respec-
tively, between persons participating and not participating
in the 15-year follow-up study. We used logistic regression
to test for differences in initial anti-HBs levels between
participants and nonparticipants after adjustment for age
class and sex. For each person, we calculated the propor-
tion of follow-up years (before dropout) with intermittent
observations for comparisons, and we compared these pro-
portions among groups by use of the Kruskal–Wallis test.
Sex and age class at time of initial vaccination signifi-
cantly differed between the groups (Appendix Table 2).
The groups did not differ for median initial anti-HBs level
or the percentage of persons initially responding to the
vaccine. After adjustment for the differences in sex and age
class, median initial anti-HBs level and the percentage of
persons initially responding to the vaccine did not differ.
Characteristics of persons with more intermittent missing
values were similar to those of persons not participating in
the 15-year follow-up. Persons with more intermittent
missing values differed by age class and sex but not by
initial anti-HBs level (before and after adjustment for sex
and age class).
Role of the Funding Source
The Centers for Disease Control and Prevention and
the Indian Health Service funded this study. Merck & Co.,
Inc., Vaccine Division, provided hepatitis B vaccine and
partially supported travel to the villages during the first 10
years of follow-up. However, Merck had no role in the
design, conduct, or analysis of the study or in the prepara-
tion or review of the manuscript.
Since the last report on this cohort (14), 967 (61%)
persons were tested at 11 years (November 1992) and 841
(53%) persons were tested at 15 years (November 1996)
after their first dose of vaccine (Appendix Table 1). Sixteen
persons had evidence of definite HBV infection, of whom
10 were tested at the 15-year follow-up. Seventy-eight per-
sons had received an additional dose of vaccine, of whom
48 were drawn at the 15-year follow-up. Levels of anti-
HBs were used in analysis from 783 (93%) of the 841
persons tested at 15 years (Appendix Table 1).
Anti-HBs Levels over Time
The GMC of the 1436 persons tested 6 months after
their third dose was 822 mIU/mL, and 1351 (94%) were
considered vaccine responders. Among participants 6
months to 19 years of age, 99% (990 of 1003) responded
to the vaccine series compared with 91% (263 of 289) of
those 20 to 49 years of age and 68% (98 of 144) of those
50 years of age and older; the GMCs for these 3 age groups
were 1748, 242, and 50 mIU/mL, respectively.
Fifteen years after their first dose of vaccine, 661
(84%) of the 783 participants tested had anti-HBs levels of
Antibody Levels and Protection after Hepatitis B Vaccination
1 March 2005 Annals of Internal Medicine Volume 142 • Number 5 335
2 mIU/mL or greater, a level considered positive, and 517
of 783 (66%) had anti-HBs levels of 10 mIU/mL or
greater. Of those, 144 (18%) had anti-HBs levels between
2 and 9.9 mIU/mL, 299 (38%) had levels between 10 and
99.9 mIU/mL, and 218 (28%) had levels greater than 100
mIU/mL. The GMC in this group was 27 mIU/mL. The
GMC at 15 years differed by age group at initial vaccina-
tion (P ? 0.001). The GMC in persons vaccinated at age
5 to 19 years was 42 mIU/mL 15 years later compared
with 23 mIU/mL in those vaccinated at age 6 months to 4
years, 19 mIU/mL for those vaccinated at age 20 to 49
years, and 9 mIU/mL for those vaccinated at 50 years of
age or older. Of the 677 participants who responded to
hepatitis B vaccine with an initial anti-HBs level of 10
mIU/mL or greater, 88% (599) had detectable HBs at 15
years; 127 (19%) had anti-HBs levels between 2 and 9.9
mIU/mL, 270 (40%) had levels between 10 and 99.9
mIU/mL, and 202 (30%) had levels greater than 100
mIU/mL at 15 years. The overall GMC among initial re-
sponders was 32 mIU/mL at 15 years and was 23 mIU/
mL, 46 mIU/mL, 24 mIU/mL, and 15 mIU/mL for those
vaccinated at age 6 months to 4 years, 5 to 19 years, 20 to
49 years, and 50 years of age or older, respectively.
Levels of anti-HBs 15 years after the first vaccine dose
varied by level of the initial anti-HBs response (P ? 0.001,
controlling for age class). Persons with an initial anti-HBs
Table 1. The Predicted Geometric Mean Concentrations of
Antibody to Hepatitis B Surface Antigen 15 Years after Initial
Hepatitis B Vaccination from Linear Mixed Model*
Sex and Age Class Predicted Anti-HBs Level, mIU/mL
Initial Anti-HBs Level
100 mIU/mL 1000 mIU/mL10 000 mIU/mL
* Anti-HBs ? antibody to hepatitis B surface antigen.
Table 2. Antibody Concentrations and Markers of Hepatitis B Virus Infection in 24 Study Participants with Evidence of Breakthrough
Hepatitis B during 15 Years after Hepatitis B Immunization*
Age at First
SexTime from First
Anti-HBs Level, mIU/mLHBV DNA
1 y before
At Time of First
* Anti-HBc ? antibody to hepatitis B core antigen; anti-HBs ? antibody to hepatitis B surface antigen; HBV ? hepatitis B virus; NA ? specimen not available.
† Definite: Conversion from anti-HBc negativity to anti-HBc positivity on at least 2 consecutive specimens or anti-HBc–positive specimen with coincident positivity for
HBV DNA; possible: anti-HBc positivity on 1 specimen with coincident boost in anti-HBs level, 2 nonconsecutive anti-HBc–positive specimens, or single hepatitis B surface
‡ Level in standard ratio units.
§ Hepatitis B surface antigen–positive 5 years after anti-HBc conversion date.
? Hepatitis B surface antigen–positive at time of first anti-HBc–positive result.
¶ Received a fourth dose of vaccine 4 years after the first vaccine dose.
** No anti-HBc–positive result, single hepatitis B surface antigen–positive result; participant received fourth dose of HBV vaccine 1 year before hepatitis B surface
Antibody Levels and Protection after Hepatitis B Vaccination
336 1 March 2005 Annals of Internal Medicine Volume 142 • Number 5
level of 4000 mIU/mL after vaccination had a GMC of
231 mIU/mL 15 years later, compared with 47 mIU/mL,
13 mIU/mL, and 3 mIU/mL among those whose initial
anti-HBs levels were 1000 to 4000, 250 to 1000, and 10 to
250 mIU/mL, respectively.
The anti-HBs GMC was 975 mIU/mL for women
and 722 mIU/mL for men after initial vaccination but had
decreased to 23 mIU/mL for women compared with 32
mIU/mL for men 15 years after the first hepatitis B vaccine
dose (P ? 0.01, linear model controlling for age class and
initial anti-HBs level).
At 11 and 15 years, 62 participants without documen-
tation of additional vaccine doses had a booster response. A
booster response was associated with age at the time of first
vaccination for those 19 years of age and younger (relative
risk, 1.94; P ? 0.01) versus those 20 years of age or older.
Among those older than 20 years of age, female sex was
associated with a booster response (relative risk, 3.29; P ?
0.01; P ? 0.02 for interaction between sex and age group).
Factors related to the persistence of anti-HBs (log-
transformed) over the 15-year follow-up period were deter-
mined in a mixed linear model (Appendix Table 3, avail-
able at www.annals.org). Persistence of anti-HBs levels
were not affected by the presence of an HBsAg-positive
member in the household at the start of the study or by the
proportion of persons in the village who were HBsAg-pos-
itive at the end of follow-up (main effect or in interaction
with time). After adjustment for sex and initial anti-HBs
level, those vaccinated at 20 to 49 years of age and those
vaccinated at 50 years of age or older did not significantly
differ (main effect and interaction with time). Persons with
higher initial responses to the vaccine had the greatest de-
clines over time in anti-HBs level (P ? 0.001, interaction
term between initial anti-HBs level and time). Females had
a greater anti-HBs decline than males (P ? 0.001, interac-
tion term between sex and time). The 15-year predicted
GMC among women who were 20 years of age or older at
the time of vaccination and who initially responded to the
vaccine to an anti-HBs level of 1000 mIU/mL was 53
mIU/mL compared with 76 mIU/mL for males (Table 1).
The greatest anti-HBs decline occurred among those first
vaccinated at 0 to 4 years of age, followed by those vacci-
nated at age 5 to 19 years of age and 20 years of age or
older, respectively (P ? 0.001, interaction term between
age class and time). Among females 0 to 4 years of age at
the time of vaccination who initially responded to the vac-
cine with an anti-HBs level of 1000 mIU/mL, the pre-
dicted anti-HBs GMC at 15 years was 12 mIU/mL, com-
pared with 27 mIU/mL among persons vaccinated at age 5
to 19 years and 53 mIU/mL among persons vaccinated at
20 years of age or older (Table 1). When we compared the
predicted level from the model to the observed anti-HBs
level, there were a number of outliers not well fit by the
model. These were related to increases in anti-HBs levels
during follow-up thought to be due to natural boosting.
The characteristics of these increases have been described
previously (20), and we did not remove these observations.
Breakthrough HBV Infections
During the 15 years after vaccination, 16 participants
were found to have a definite breakthrough infection and 8
were found to have a possible breakthrough infection; none
had clinical signs or symptoms of hepatitis (Table 2). The
incidence of new HBV infections during the 15-year
follow-up period was 0.84 (95% CI, 0.48 to 1.37) infec-
tion per 1000 persons per year when only the definite
Table 3. Number of Persons with Definite Hepatitis B Virus Breakthrough Infections and Rate per 1000 Persons per Year by Group at
Initial Vaccination, Sex, and Initial Response to 3 Doses of Plasma-Derived Hepatitis B Virus Vaccine*
CharacteristicPerson-Years at Risk
(Number of Persons)
Definite HBV Breakthrough Infections P Value
NumberRate (per 1000
Initial responder to vaccine
16 368 (1351)
Age at first vaccine dose
18 997 (1578) 16 0.84NA
* HBV ? hepatitis B virus; NA ? not applicable.
† P value compares rate in persons who initially responded to the vaccine to the rate in persons who did not respond to the vaccine.
‡ P value comparing rate for all 4 age groups.
Antibody Levels and Protection after Hepatitis B Vaccination
1 March 2005 Annals of Internal Medicine Volume 142 • Number 5 337