Abstract
The long-term efficacy of hepatitis B vaccination among high-risk infants was determined in 805 vaccine responders, immunized at birth in Taiwan during 1981–1984 and followed to age 10 years, via life table survival and Cox multivariate analyses. At 10 years, cumulative persistence of antibody to hepatitis B surface antigen (anti-HBs) was 85%, and cumulative incidence of hepatitis B virus (HBV) infection was 15%. Three children became carriers. Twelve-month anti-HBs titer was the strongest predictor of efficacy. The higher the initial titer, the lower the risk of anti-HBs loss (relative risk [RR], 0.26 for titer of 100–999 mIU/mL; RR, 0.08 for titer > 1000 mIU/mL; P < .001) and HBV infection (RR, 0.55 and 0.27; P < .05). Maternal hepatitis B e antigen positivity but not hepatitis B immunoglobulin dose or gender predicted greater antibody persistence to age 10 years. Because the level of antibody persistence remained high and few became carriers, booster revaccination within 10 years seems unnecessary.
The World Health Organization (WHO) estimates that the hepatitis B virus (HBV) has infected 2 billion persons alive today, and 350 million of them are chronically infected HBV carriers. Without immunization, an increase to 400 million carriers is expected by the year 2000 [1, 2], nearly 20% of whom will die of liver disease caused by the infection [1, 3]. Elevated carrier rates in HBV-endemic areas are largely due to mother-to-infant transmission. Over 40% of infants born to hepatitis B surface antigen (HBsAg) carrier mothers will be infected with HBV, and infection acquired during infancy and early childhood has a ⩾90% chance of becoming chronic [4, 5]. If their mothers are hepatitis B e antigen (HBeAg)—positive, children have a 66%–93% risk of becoming carriers [6–10].
Plasma-derived hepatitis B vaccine is highly immunogenic and effective in preventing infection and clinical hepatitis in susceptible infants and adults, with an overall efficacy of 80%–95% [3, 4, 7–26]. Vaccine recipients who develop an antibody response of ⩾10 mIU/mL are nearly 100% protected against clinical illness [27, 28]. Recombinant DNA vaccine has replaced the plasma-derived vaccine since our prior study. The safety, immunogenicity, and protective efficacy of the recombinant vaccine have been equally satisfactory to those of the plasma-derived vaccine, even though the chemical composition of the two are not identical [29–34]. The Viral Hepatitis Prevention Board and the WHO had worked toward implementing universal hepatitis B vaccination of all children in all countries by 1997. As of mid-1998, 90 countries had added the hepatitis B vaccine into their routine immunization schedule [35]. The ultimate success of such immunization programs will largely depend on the long-term protection offered by vaccination and the proper assessment of the need and timing of booster doses [16].
The results of vaccine protective efficacy in the first 12 months and the first 5 years have been reported [10, 25, 26]. This study examined the effects of various factors on the loss of antibody to hepatitis B surface antigen (anti-HBs) and risk of HBV infection between ages 5–10 years in a cohort of children who were born to HBsAg-positive mothers, responded to vaccine, and were not infected by age 12 months. The potential predictors examined were vaccine dosage, initial immune response to vaccine, maternal HBeAg status, hepatitis B immunoglobulin (HBIG) administration, and gender.
Subjects and Methods
Vaccinees in original clinical trials
During 1981–1984, several randomized controlled vaccine trials using a 3-dose series of plasma-derived hepatitis B vaccine were conducted in Taiwan with 972 neonates born to HBsAg-positive mothers. Eligible infants were enrolled from Taipei Municipal Women and Children's Hospital, MacKay Memorial Hospital, and National Taiwan University Hospital, where all women seeking prenatal care were routinely screened for HBsAg by reverse passive hemagglutination with commercial kits (Organon Teknika, Durham, NC; Ortho Diagnostics, Raritan, NJ; Abbott, Abbott Park, IL; Wellcome Diagnostics, Research Triangle Park, NC). Trained project nurses gave both a written description and oral explanation of the study to each HBsAg-positive woman. Infants of parents who gave consent were enrolled in the trials if they were delivered in the hospital of recruitment, their birth weight was ⩾3000 g, and their 1-min Apgar score was ⩾9 [10, 25, 26].
Vaccine and vaccination
Infants were randomized at birth to receive one of a number of dosages (20, 10, 5, 2.5 μg) of vaccine plus HBIG or vaccine alone in various clinical vaccine trials during 1981–1984. The vaccine was administered according to a standard 3-dose schedule at 0, 1, and 6 months. Vaccines used were lots 773/801 and 751/800 from Merck Sharp and Dohme (West Point, PA), identical to the commercially available preparation. HBIG was 0.5 mL (290 IU/mL) of commercial lot C50241 from Abbott [10, 25, 26].
Laboratory procedures
All HBV markers (HBsAg, anti-HBs, antibody to hepatitis B core antigen [anti-HBc], HBeAg, and antibody to hepatitis B e antigen) were tested blindly by RIA kits (Abbott) and according to the manufacturer's instructions. Quantitative levels of anti-HBs were expressed in international units [10, 25].
Follow-up
Follow-up visits for the 972 infants were scheduled every 3 months until age 24 months and then every 6 months until age 5 years to determine health status and HBV markers. A final follow-up visit was made at age 10 years to evaluate the long-term efficacy of the vaccine. These 972 infants consisted of 805 who developed anti-HBs after vaccination, 116 HBV carriers who were not protected by immunoprophylaxis, 22 who were infected but had not become carriers by age 12 months, 11 who were revaccinated because they either did not develop antibody or lost antibody soon after age 12 months, and 18 who were missing baseline data. Subjects selected for 10-year analysis of antibody levels and evidence of infection were the 805 infants who were free of infection and positive for anti-HBs at 12 months of age.
Definitions
Initial antibody titer refers to the level of anti-HBs measured at 12 months of age. Antibody loss and antibody persistence was defined as testing negative or positive, respectively, for anti-HBs at follow-up according to commercial test kit results. HBV infection refers to persons who had serologic conversion to HBsAg or became anti-HBc-positive after age 12 months. Chronic carriage was defined as being HBsAg-positive ⩾6 months. A natural booster episode refers to a spontaneous sharp rise in the serum anti-HBs titer in the absence of anti-HBc and HBsAg.
Statistical analysis
Incidence densities of antibody loss and HBV infection were calculated from person-time data accumulated beginning at age 12 months. Cumulative incidences were calculated by life-table analysis using the Kaplan-Meier method. Incidence density ratios adjusted for study factors of interest were estimated from Cox proportional hazards models. Because colinearity problems arose when vaccine dosage and initial antibody response were modeled categorically, the initial titer was modeled in continuous form. For ease of interpretation, however, the effect of initial titer is presented in the tables categorically. Statistics were calculated using Epi Info (version 6.04b; CDC, Atlanta) and Stata 5.0 (Computing Resource Center, Santa Monica, CA). P ⩽ .05, 2-tailed, was considered statistically significant for measures of association.
Results
Follow-up compliance among the original 972 infants was 98% at age 2 years, 97% at age 3 years, 93% at age 4 years, 84% at age 5 years, and 66% (646/972) at age 10 years. Compliance among the 805 infants selected for analysis was almost identical: 99% at age 2 years, 97% at age 3 years, 94% at age 4 years, 89% at age 5 years, and 67% (539/805) at age 10 years. These 805 infants contributed ∼5800 person-years of follow-up from 12 months of age. Those who were available for the 10-year serologic assay did not differ significantly from those who were lost to follow-up with respect to 12-month antibody titer, vaccine dosage, maternal HBeAg status, HBIG administration, or gender.
Of these 805 infants, 113 had initial antibody titer < 100 mIU/mL, 351 had a titer of 100–999 mIU/mL, and 284 had a titer > 1000 mIU/mL. The proportion with low antibody titer (< 100 mIU/mL) was 9.3%, 15.1%, 27.4%, and 30.0% for those who received vaccine dosages of 20, 10, 5, and 2.5 μg, respectively. The initial geometric mean titer (GMT) was 669.2, 520.4, 241.8, and 202.1 mIU/mL for vaccine dosages of 20, 10, 5, and 2.5 μg, respectively.
Antibody loss and persistence
Of the 805 children who responded to vaccine and were free of infection at 12 months of age, 49 lost detectable anti-HBs by age 5 years and an additional 63 children did so by age 10 years, for a total of 112 cases. The cumulative incidence of antibody loss between ages 1 and 10 years was 15.1 cases per 100 children (cumulative incidence of antibody persistence was 84.9%); the incidence density for antibody loss was 1.9 cases per 100 person-years (figure 1, table 1). Of the 49 children who lost antibody by age 5 years, 6 had a natural booster episode between ages 5 and 10 years.
Antibody to hepatitis B surface antigen (anti-HBs) persistence by 12-month anti-HBs titer (Kaplan-Meier survival curve). P < .001 by log-rank test of homogeneity.
Antibody to hepatitis B surface antigen (anti-HBs) persistence by 12-month anti-HBs titer (Kaplan-Meier survival curve). P < .001 by log-rank test of homogeneity.
Effect of selected exposures on the incidence rate of antibody loss among successfully vaccinated infants in Taiwan by age 10 years during 1981–1994.
Effect of selected exposures on the incidence rate of antibody loss among successfully vaccinated infants in Taiwan by age 10 years during 1981–1994.
Adjusting for losses to follow-up, the cumulative persistence of antibody at 10 years of age was 85% overall, 54% for children whose initial titer was < 100 mIU/mL, 85% for children whose initial titer was 100–999 mIU/mL, and 96% for children whose initial titer was ⩾1000 mIU/mL (figure 1). Thus, antibody loss was inversely correlated with the initial, or 12-month, titer achieved. Twelve-month anti-HBs titer was the strongest predictor of antibody loss after adjusting other variables. The higher the initial titer, the lower the risk of anti-HBs loss (adjusted relative risk [RR], 0.26; 95% confidence interval [CI], 0.17–0.38) for titer of 100–999 mIU/mL; RR of 0.08 (95% CI, 0.04–0.15) for titer of > 1000 mIU/mL (P for trend < .001; table 1). Compared with those who had an initial antibody titer of < 100 mIU/mL, children who had initial titers of 100–999 mIU/mL were 4 times less likely to have lost detectable levels of antibody by 10 years, and children whose initial titers exceeded 1000 mIU/mL were > 12 times less likely to have done so.
Children who lost antibody by age 5 years had an initial GMT of 55.4 mIU/mL; those losing antibody by age 10 years had a GMT of 112.6 mIU/mL. In contrast, children with persistent antibody at age 10 years had an initial GMT of 631.1 mIU/mL.
Vaccine dosage appeared to play an important role in predicting antibody loss. At age 5 years, a significant protective effect was observed among groups given 10-μg and 20-μg dosages when compared with the 2.5-μg group. At age 10 years, this effect was observed in univariate analysis but disappeared when adjusted for gender, HBIG administration, initial titer, and maternal HBeAg status (table 1). Because the initial titer of anti-HBs was significantly related to the dose of vaccine (χ2 = 46.96, P < .001), it seems logical that the protective effect of dose would operate primarily by influencing the initial titer as an intermediate variable. A multivariate model excluding initial titer was examined to observe the effect of dose without removing its influence on titer. In this model, the protective effect among children who received 10 or 20 μg of vaccine persisted when compared with children who received lower doses.
Infants whose mothers were HBeAg-positive were nearly 3 times less likely (RR = 0.37; 95% CI, 0.25–0.53) to lose detectable anti-HBs levels by age 10 years than those whose mothers were HBeAg-negative. This association held true even after excluding the 113 children who had HBV infections by age 10 years and adjusting for initial titer, vaccine dosage, gender, and HBIG administration (adjusted RR = 0.35; 95 % CI, 0.19–0.63). Ten-year antibody persistence was not related to gender or administration of HBIG at birth.
Prevention of HBV infection
Of the original 972 vaccinated infants, 14.3% (138) were infected before 12 months of age. A large proportion (116 [84.1%] of 138) of those infected during this period were confirmed carriers by age 12 months. Of the 805 children who had not become infected by 1 year of age, 52 became infected before 5 years of age and an additional 61 developed HBV infections by age 10 years, for a total of 113 cases of infection during the entire study period between ages 1 and 10 years. The cumulative incidence between ages 1 and 10 years was 14.8 cases per 100 children (figure 2); the incidence density of infection was 2.0 cases per 100 person-years (table 2).
Cumulative attack rate of HBV infection by 12-month antibody to hepatitis B surface antigen titer (Kaplan-Meier failure curve). P < .001 by log-rank test of homogeneity.
Cumulative attack rate of HBV infection by 12-month antibody to hepatitis B surface antigen titer (Kaplan-Meier failure curve). P < .001 by log-rank test of homogeneity.
Effect of various exposures on the incidence rate of HBV infection among successfully vaccinated infants in Taiwan by age 10 years during 1981–1994.
Effect of various exposures on the incidence rate of HBV infection among successfully vaccinated infants in Taiwan by age 10 years during 1981–1994.
Of the children infected after age 12 months, only 2.7% (3/113) became chronic: 1 girl by age 18 months, 1 boy by age 24 months, and 1 girl by age 30 months. A fourth child, a boy, became HBsAg-positive between 5 and 10 years of age, but no follow-up specimen was drawn to confirm chronic carriage (table 3). All 4 of these children had HBeAg-positive mothers. The rate of HBsAg positivity among HBV infections was 5.8% (3/52) from 12 months to 5 years of age and 1.6% (1/61) from 5 to 10 years of age. The cumulative incidence of confirmed chronic HBV infection from 1 to 10 years of age in this cohort was 0.4% (3/805) and only 2.7% (3/113) among newly infected children.
Time line of children who became hepatitis B carriers.
Time line of children who became hepatitis B carriers.
Adjusting for losses to follow-up, the overall cumulative attack rate of HBV infection over the 9 years of follow-up was 15%; by initial titer, it was 9% among those who had a 12-month titer ⩾1000 mIU/mL, 17% among those who had a titer of 100–999 mIU/mL, and 28% among those with titers < 100 mIU/mL (figure 2). Thus, the risk of HBV infection was inversely related to the initial titer achieved. The higher the initial titer, the higher the protective effect observed (adjusted RR, 0.55; 95% CI, 0.35–0.87; RR, 0.27; 95% CI, 0.15–0.48). Compared with those whose initial titer was < 100 mIU/mL, children who had an initial titer of 100–999 mIU/mL had half the incidence density of infection, and children whose titers exceeded 1000 mIU/mL had one-fourth the incidence density of infection over the entire 9 years of follow-up (table 2).
The presence of HBeAg in the mother significantly increased the risk of infection in the child before the fifth year of life (adjusted RR, 4.58; 95% CI, 1.37–15.26), but this RR reduced considerably and became statistically nonsignificant by age 10 years (adjusted RR, 1.74; 95% CI, 0.92–13.20). The risk of infection was not related to HBIG administration or gender (table 2).
In univariate analysis, there was a significant trend of protection against infection at age 10 years with increasing vaccine dosages (P for trend = .008), but the differences between dosage groups were not statistically significant. These differences were nonsignificant in the full multivariate model as well (table 2). The initial titer was removed as an intermediate variable from the multivariate model, to ascertain the full effect of vaccine dosage without removing its influence on initial titer. When this model was used, the differences between dosage groups still did not reach statistical significance, but there was a clearer protective dose-response effect. The effect of vaccine dosage was not different when analyzed with respect to the mother's HBeAg status.
Effect of antibody loss on HBV infection
Children who lost detectable anti-HBs levels by 5 years of age were> 2 times (RR, 2.42; 95% CI, 1.22–4.81) more likely to become subsequently infected by age 10 years (table 4). All of the infections among the 59 cases occurring between ages 5 and 10 years were sub-clinical, however, and indicated only by presence of anti-HBc. None became HBsAg-positive.
Effect of antibody loss on rate of hepatitis B virus infection.
Effect of antibody loss on rate of hepatitis B virus infection.
Discussion
In this cohort of 805 high-risk children who responded to vaccination at birth, the cumulative persistence of antibody at age 10 years was 85%, and the cumulative incidence of HBV infection was 15.0%. Only 3 (0.4%) children became carriers. Twelve-month antibody titer was found to be a strong predictor of both antibody persistence and HBV infection. This finding is consistent with those of other long-term follow-up studies [3, 11, 17, 19–23, 36, 37]. Children who had high levels of antibody at age 12 months retained detectable levels and were protected from HBV infection through their tenth year of life more frequently than those who had lower levels. Children whose mothers were HBeAg-positive were more likely to maintain detectable levels of anti-HBs over 10 years, but they were at a much higher risk of developing HBV infection before age 5 years than children whose mothers were HBeAg-negative. Vaccine dosage was an important predictor of antibody persistence, particularly before age 5 years, but it seemed to exert much of its influence through its effect on initial titer. Dosage was not strongly associated with risk of infection.
A plausible explanation for why children born to HBeAg-positive mothers tended to keep their antibodies longer is that HBeAg-positive persons are more infectious, increasing the likelihood that the persistence of anti-HBs represented booster effects from repeated exposure to virus from the mother. Marion et al. [9] found comparable results and made similar conclusions.
The fact that maternal HBeAg status played an important role in the risk of HBV infection before age 5 years but not thereafter suggests that the primary mode of transmission during the first 5 years of life was mother-to-child, including perinatal and postnatal intimate contact, and that the primary mode of transmission after age 5 years was horizontal in nature. Another possibility is that the HBeAg titer in the mothers waned over time, and they became less infectious than at the time their children were born. HBeAg levels are known to decrease more rapidly than HBsAg levels [38].
Because widespread evaluation of plasma-derived hepatitis B vaccines began only in the early 1980s, the precise duration of protection after vaccination is still unknown; although data show that it may last 3–10 years or more [3, 4, 8, 9, 11–15, 17, 19–23, 25, 36, 39]. While some studies have found that protection wanes within this period [12, 19, 23], in many cases the 3- to 10-year range reflects the length of follow-up [3, 4, 8, 9, 13–15, 17, 21, 25, 36]. Typically, levels of anti-HBs decline rapidly over the first year after the initial vaccination series and decline more gradually in subsequent years [8, 11, 19, 21]. An anti-HBs titer of 10 mIU/mL is generally considered the minimal protective level [3, 4, 8].
In this study, persistence of antibody was directly related to the initial titer of anti-HBs, and loss of antibody was related to increased rates of infection, but antibody loss did not necessarily mean loss of immunity to HBV. In fact, this cohort of children showed excellent long-term protection from chronic carriage and clinical illness. Without vaccination, we would have expected 40% of the infants in this high-risk population to become infected with HBV and 90% of those infections to progress to the chronic carriage state [5]. The fact that the children remained protected against chronic carriage and clinical illness despite declining levels of antibodies corroborates the majority of the findings that are available in the current literature. Most of the evidence from long-term efficacy trials suggests that vaccine-induced protection from clinically significant infection, including acute hepatitis, viremia, and development of the carrier state, does not wane over time, despite declining antibody levels. Although some studies have found a correlation between decreasing residual antibody titers and increased risk of infection [12, 19, 23], it has been largely demonstrated that even after anti-HBs levels are no longer detectable by current assays, immunologic memory and subsequent protection seem to persist [3, 8, 11, 14, 17, 36, 40].
When infection occurs after vaccination, it is usually manifested only by anti-HBc seroconversion, without detectable HBsAg or clinical disease [36]. This information is especially useful in determining the need for booster doses after the initial series of vaccine. It has been recommended that the need for boosters should be determined not by measuring decline in antibody levels but by showing the failure to protect against the clinically significant aspects of HBV [39]. It is still conventional to use residual antibody titers as a measure of immune status, however, since immunologic memory cannot be measured easily and antibody levels are directly quantifiable.
Clinically significant aspects of HBV disease in vaccinated persons have been limited to persons with no response or a poor response to the vaccine [4, 19, 21, 23]. These persons are also unlikely to respond well to revaccination [19]. One of the most important predictors of long-term vaccine efficacy is the peak initial anti-HBs response. In numerous studies, there was a direct correlation between peak initial response to vaccine and persistence of antibody [3, 11, 17, 19–23, 36]. Peak anti-HBs response in turn appears to be closely related to the dose of the vaccine [11, 13, 25, 26], but results vary depending on the background risk of infection in the different study populations. The relationship between risk, vaccine dosage, and antibody response needs further investigation.
In this population of high-risk infants, vaccination was enormously successful in preventing HBV infection. The most significant result was that it greatly reduced the rate of chronic carriage of HBsAg, the outcome that is associated with chronic liver disease. Similar dramatic results were achieved in American Samoa, where HBV is endemic. After a vaccine intervention program that targeted the entire susceptible population, the prevalence of HBV infection declined by 50%, and the number of chronic carriers decreased by 80% [18]. A study in Taiwan between 1981 and 1994 showed that after the institution of a nationwide hepatitis B vaccination program in 1984, the incidence of hepatocellular carcinoma in children decreased significantly [41].
Administration of HBIG with vaccine also affects the long-term efficacy of HBV vaccination, with varying consistency [3, 4, 8, 25]. In this study, there was no significant correlation between HBIG administration and long-term efficacy.
Even if vaccination did not prevent all infections, it delayed the age of infection, and this delay greatly reduced the proportion of infections that became chronic. These findings are consistent with epidemiologic evidence that has shown that infants infected during their first 6 months of life have about a 90% probability of becoming carriers but that infections occurring later in life have a much reduced probability of becoming chronic [42]. In this population, among children who did not become infected during the first year of life, chronic carriage was extremely rare to age 10 years. Based on this information, it appears that booster revaccination within the first 10 years may not be necessary, particularly in endemic areas. However, a possible consideration for revaccination is that many of these children are about to enter another risk category, that of sexually active adolescents. Since they are still considered protected at age 10 years, it is difficult to conclude whether or not re-vaccination is indicated even after these risks are considered. Further follow-up of vaccinated infants into adolescence is necessary before any definitive proposal can be made.
Acknowledgments
We thank the National Science Council, Republic of China, Yuan-Chih Memorial Fund, and the Far East Foundation of Taipei for support of the clinical trails and acknowledge the valuable collaborative efforts of ChinYun Lee and the follow-up work of nurses Eileen Chen and Evelyn Leon.
