Long-term immune responses to vaccination in HIV-infected patients: a systematic review and meta-analysis
Download the PDF here
Clinical Infectious Diseases Advance Access published January 10, 2014
Solen Kerneis1,2,3,4,5, Odile Launay1,2,3, Clement Turbelin4,5, Frederic Batteux1,6, Thomas Hanslik7,8, Pierre-Yves Boelle4,5
1Paris Descartes University, Paris, France 2CIC BT505, Inserm, Paris, France 3CICBT505, Assistance Publique-Hopitaux de Paris, Paris, France 4UMR S 707, Pierre et Marie Curie University, Paris, France 5UMR S 707, Inserm, Paris, France 6Laboratoire d'immunologie biologique, Cochin Hospital, Assistance Publique-Hopitaux de Paris, Paris, France 7Department of Internal Medecine, Ambroise Pare Hospital, Assistance Publique-Hopitaux de Paris, Boulogne Billancourt, France 8Versailles Saint-Quentin-en-Yvelines University, Versailles, France
Vaccine-induced antibodies may wane more quickly in persons living with HIV than in healthy individuals. Here, we reviewed the literature on vaccines routinely recommended in HIV-infected patients to estimate how seroprotection decreases over time in those who initially responded to immunization. For each study retrieved from the literature, the decrease of seroprotection was modeled with a log binomial generalized linear model, and data were pooled in a meta-analysis in order to provide estimates of seroprotection two and five years after last vaccine administration. Our analyses confirmed that duration of seroprotection was shorter in HIV-infected patients, and that with current guidelines, a substantial proportion of patients would have lost protective antibodies before being proposed a booster. We therefore discuss the implications on the monitoring of antibody levels and timing of revaccination in these patients.
"Indeed, loss of antibodies does not necessarily imply loss of clinical protection and immune memory can persist, even in the individuals with low antibody concentrations ......More generally, patients immunized while on HAART show prolonged seroprotection for most vaccine antigens, as illustrated in a recent review . Similarly, HIV viral load at the time of immunization is an important independent predictor of persistence of seroprotection, as illustrated for hepatitis A [21, 22], yellow fever  and Pneumococcus vaccines . On the other hand, whereas increasing the dose of antigen is an interesting strategy to improve responses in immuncompromised hosts , there was no clear trend in our meta-analysis indicating that this strategy would extend duration of seroprotection among primary responders." see Discussion below
Immune responses to most vaccines are known to be impaired in HIV patients [1,2].However, besides primary response, long-term persistence of protection has been poorly documented. As of today, recommendations on the timing of booster injections were based on data collected in healthy persons although antibody decay patterns may be different. In this respect, an important question is to estimate, among patients who initially responded to immunization, how seroprotection decreases over time. Here, we reviewed data on long-term persistence of antibody concentrations after vaccination in HIV-infected patients. This choice was supported by three main reasons: (i) antibody concentrations are reported in most vaccine trials, providing enough data to allow meta-analysis, (ii) correlates of protections have been defined for most vaccines and (iii) antibody levels can be routinely assessed for most antigens with standardized methods. For some vaccines (i.e. measles, varicella, yellow fever), cell mediated immunity is the critical determinant of protection, however methods of evaluation of cellular responses are not easily comparable between studies and correlates of protection not yet established. Our goal here was to provide a summary of available data to guide recommendations on revaccination in HIV-infected persons.
Of the 159 potentially relevant studies, 54 were selected for the qualitative review (figure 1).
For each vaccine, percentages of seroprotection retrieved from the literature are presented in figure 2. Median follow-up ranged from nine months to nine years after the last vaccine dose. The median number of patients included per study was 40 (Interquartile range: 18; 63). Data were available for 13 vaccine antigens: Streptococcus pneumoniae (n=14), hepatitis B (n=12), measles (n=12), hepatitis A (n=5), tetanus (n=8), yellow fever (n=3), Haemophilus influenzae type b (n=3), rubella (n=2), varicella, (n=1), pertussis (n=1), polio virus (n=1), mumps (n=1), and Japanese encephalitis (n=1). Of the 54 studies included in the review, 19 fitted the eligibility criteria for meta-analysis. Others were excluded because they were on pneumococcal vaccine (n=14), were retrospective (n=13), did not differentiate outcome of primary responders and non-responders during follow-up (n=4), or because only one study was available for the vaccine (n=4: pertussis , Haemophilus influenzae type b , varicella , and Japanese encephalitis .
Twelve studies were included, with follow-up times ranging from 12 to 115 months [8-19]. As illustrated on figure 2A, seroprotection typically decreased over time: after 3 doses of 40μg HBsAg , 71% of primary responders maintained protective antibody titers at year one , 33%-61% at year two [8,10], and 40% at year five . There was no clear trend of longer persistence of seroprotection with high-dose vaccine schemes [8,10]. Three retrospective studies reported data beyond five years after immunization [11,15,19] in HIV-infected children born to Ag HBs+ HIV-infected mothers, and maintenance of seroprotection was particularly poor: 24% after 5.5 years , 45% after 8 years , to only 1% after 9.6 years  after a three 10μg-doses scheme.
According to the meta-analysis, less than one half of primary responders would maintain protective antibody titers two years after immunization (38% (CI95% = 23%; 54%) in adults and 61% (27%; 90%) in children), and only 17% (CI95%: 3%; 36%) after five years (figures 3 and 4).Vaccine schemes with double-dosed schemes did not improve maintenance of seroprotection (compared to those with standard doses of vaccine (respectively 41% (CI95% = 14%; 71%), and 50% (CI95% = 24%; 77%) after two years, P-value=0.65).
One retrospective  and four prospective studies [21-24] were identified on hepatitis A. In adults, a slight decrease was observed over time (figure 2B): 100% of responders were still seroprotected after one year , 90% after three years , and 85% after four years . No data were reported beyond four years. In children, persistence of responses was lower: 85% (n=84) after 9 months  and 92% after 18 months  . Percentages of seroprotection were not greater with three-dose schedules [22,23], as compared to standard two-dose schedule [21-24]. The pooled percentages of seroprotection against hepatitis A were of 92%
(CI95% = 89%; 95%) two years after vaccination, and 82% (CI95% = 76%; 88%) after five years (figures 3 and 4).
Measles, Mumps and Rubella
Data on long-term persistence of measles immunity were reported in twelve studies, either in children (n=10) [25-34] or adults (n=2) [35,36]. As illustrated on figure 2C, percentages of seroprotection varied widely between studies, reflecting the great heterogeneity in the populations studied. In HIV vertically-infected Sub-saharian African children not on HAART and vaccinated in the first months of life, percentages of seroprotection rapidly dropped during follow-up: 62%-73% of primary responders were still protected after 15 months [29,31], 41%-49% after 28 months [25,31],and 30% after 50 months . HAART improved seroprotection, either for children started on HAART after immunization (69% after six months of HAART  40-60% after four to seven years of HAART [26, 34]) or those who were vaccinated or revaccinated after the start of HAART (73% after one year , 40-82% after four years ). In this last study, higher percentages of seroprotection were reported four years after vaccination in children started on HAART in their first year of life (82%) compared to others (40%) .
In adults, data were scarce and showed conflicting results. In adults vaccinated during childhood before HIV infection, at the median age of 35 years, 95% still had protective antibody titers (n=59) , whereas after revaccination of 26 HIV-infected adults at the age of 31 years, only 43% were still above the detection level after one year .
Because of great heterogeneity, the meta-analysis was restricted to five studies focusing on HIV vertically-infected children [26,27,29-31], immunized at the age of 6-42 months, with either one  or two doses [26,27,29,30] of the MMR vaccine. The pooled analysis estimated that in this particular population, protective antibody levels would persist in 68% (CI95% = 45%; 88%) of primary responders after two years, and 40% (CI95% = 10%; 73%) after five years (figure 3C).
Only three studies reported data on rubella [26,34] and mumps . After four years on HAART, antibodies persisted in 62% of children against mumps and 27-89% against rubella, the highest seroprotection rates being reported for children with virological response.
Tetanus, Poliomyelitis and Pertussis
Studies on tetanus included 241 children followed from 12 to 53 months after immunization (figure 2E) [28,30,32-34,37-39]. Retrospective studies accounted for 76% of participants (n=182). Percentages of seroprotection were high: 67-90% within the first two years [30,33,34,37-39], 78% after 4 years . Meta-analysis of the four prospective studies estimated overall percentages of seroprotection (figure 3D) of 74% (CI95% = 57%; 87%) and 43% (CI95% = 21%; 66%) two and five years after immunization, respectively. One study reported data on immunization against polio virus , one against pertussis , and one against diphtheria . In all reports, duration of seroprotection was shorter than in healthy children.
All 14 studies on persistence of antibodies against S. pneumoniae were prospective, either in children (n=5) [40-44] or adults (n=9) [45-53] and either on a conjugate vaccine (PCV, n=6) [40-44,49] or the 23-valent polysaccharide vaccine (PPSV23, n=9) [45-48,50-53]. Follow-up ranged from 8 to 60 months after immunization (median: 12). The median number of patients was 33. The definition of response varied greatly between studies: either a two-fold and greater increase in antibody titers [47,49], or an antibody concentration above either 0.35 μg/mL [42,43,47], 0.5, μg/mL , or 1 μg/mL [40,41,48,49,52]) or on the GMTs alone [44-47, 49, 50]. After PPSV23 in adults, rates of decrease of antibody concentrations were either similar [45,48,52,53] or more rapid [46,50] to that observed in healthy individuals. Beyond five years, [47,48], antibody concentrations had dropped below the cut-off values for most serotypes. This was reportedly more frequent in patients with low CD4 cell counts at vaccination or those who failed to achieve virological suppression . Vaccination with PCV was not directly compared with PPSV23 in adults, but two doses of 7-valent PCV elicited more sustained responses than one during the one year-follow-up .
In HIV-infected children, the immunogenicity of PCV has been assessed from eight months to five years after primary immunization [40-44]. Here again, results were variable: HIV-infected children reportedly experienced a significantly greater decline of antibody levels during follow-up than HIV-uninfected children [42-44]; however with a vaccine scheme including two PCVs and one PPSV23 at 8-weeks intervals, Abzug et al. showed that 22 months after immunization, decline in antibody concentrations were similar to those in HIV-uninfected infants after PCV alone .
After immunization against Haemophilus influenzae b, estimates of seroprotection ranged from 16% to 78% [5,30,54] in vertically HIV-infected children at different time points after immunization. In the only report on varicella vaccine, less than 50% of 94 HIV-infected children (aged 1-8 years) still had detectable VZV IgG one year after the start of immunization . For yellow fever, three retrospective studies [55-57] suggested that immunogenicity of YF immunization waned more quickly in HIV patients: the proportion of patients with non-reactive titers increasing from 17% to 23% over the ten years-period postvaccination
. In the single study on an inactivated Japanese encephalitis vaccine in 50 HIV-infected children on HAART, only 4/38 primary responders lost protective antibodies during the three-year follow-up, suggesting long-term benefit of the vaccination in this population .
Duration of seroprotection, as assessed by quantitative measurement of antibody responses, is shorter in HIV-infected patients than in otherwise healthy persons for most licensed vaccines. As a comparison, 65-95% of healthy children and 80% of adults maintain protective anti-HBs concentrations ten years after vaccination against hepatitis B. With inactivated hepatitis A vaccine, mathematical models estimate that protective levels of anti-HAV could be present
for more than 25 years in adults and 14-20 years in children. Percentages of seroprotection reach 93-10% in healthy children ten years after tetanus vaccine and 95-100% after measles vaccine (see references in table 2 of the supplementary material).
Our analyses showed a rapid decrease in seroprotection after immunization in HIV-infected patients and have several implications:
----Because of high prevalence and severity of chronic hepatitis B in HIV-infected patients, it appears that anti-HBs antibodies should be measured yearly in adults and every 2-5 years in children at risk (those in close contact with Ag HBs+ persons or living in a high endemicity country). Closer monitoring of anti-HBs titers could be considered in those with the lowest antibody titers at the end of the vaccination course (10 to 100 IU).
---For persons at increased risk for hepatitis A (i.e. travelers, men who have sex with men, users of drugs, persons with occupational risk for HAV infection or chronic liver disease), anti-HAV antibodies should be monitored every five years since almost 20% of primary responders would have lost seroprotection by year five after immunization.
---For tetanus, the meta-analysis of prospective studies found that around 75% of HIV-infected children would maintain protective concentrations after two years, and retrospective studies reported percentages of seroprotection around 70-80% after five years. In this respect, the interval of ten years between boosters seems reasonable.
---To improve maintenance of seroprotection against measles, the initial vaccination scheme should include two doses, ideally administered after the start of HAART, in children with undetectable viral load. Children vaccinated before the start of HAART and/or with detectable HIV viral load at the time of immunization could be proposed a third dose 2-5 years after primary vaccination, when CD4-cell count is >200 or >15%.
---The WHO recently stated that a single dose of yellow fever vaccine was sufficient to confer life-long immunity against the disease in immunocompetent persons. Data show that immunity wane more quickly in HIV-infected persons. In these patients, determination of antibody titers should be considered in case of potential exposure, and revaccination proposed to those with negative titers.
---For now, data on immunogenicity of PCV or strategies combining PCV and PPSV23 are too preliminary to state on the optimal timing of booster injections in adults.
Several methodological issues warrant discussion. First, it is remarkable that the median number of patients per study was only 40. Data on immunogenicity of vaccines in the immunocompromised host are scarce and the small samples in each category gave little power for comparisons between age classes or vaccine schemes. Second, only prospective studies were included in the meta-analysis, which may have lead to a certain degree of underestimation of seroprotection. Indeed, as illustrated for tetanus, studies pooled in the meta-analysis were those with the shortest duration of follow-up and lower percentages of seroprotection. This choice was however essential to guarantee the quality of data analyzed. Conversely, relying on the exponential decrease could lead to overestimate the decrease in seroprotection. The choice of the model was supported by the good fit in studies where at least two follow-up points were available (see supplementary material). This model however leads by structure to an always decreasing trend in seroprotection with time, implying that all vaccinees will loose seroprotection at some point. An alternative assumption would be that some individuals would remain seroprotected permanently over time, as it has been hypothesized for example with hepatitis A vaccine.
There was also great heterogeneity in some important variables that complicated comparisons across studies: immunologic status at the time of vaccination, use of antiretroviral therapy and whether treatment was started before or after vaccination, age, vaccine strain, assays used to measure antibody concentrations, antibody levels considered as protective, exposure to wild type pathogens that could confound antibody measurements. The small samples gave little
power for comparisons but several variables seemed to be associated with persistence of seroprotection. For example, like for immunocompetent persons, certain vaccine antigens such as rubella, tetanus and hepatitis A induce more sustained responses than others . The viro-immunologic status and use of antiretroviral therapy at the time of immunization is also of critical importance in maintenance of seroprotection. As illustrated for measles, persistence of antibody concentrations was improved in children vaccinated or revaccinated while on HAART [30, 32], as compared to others [25, 26, 27, 28, 29, 31]. More generally, patients immunized while on HAART show prolonged seroprotection for most vaccine antigens, as illustrated in a recent review . Similarly, HIV viral load at the time of immunization is an important independent predictor of persistence of seroprotection, as illustrated for hepatitis A [21, 22], yellow fever  and Pneumococcus vaccines . On the other hand, whereas increasing the dose of antigen is an interesting strategy to improve responses in immuncompromised hosts , there was no clear trend in our meta-analysis indicating that this strategy would extend duration of seroprotection among primary responders.
Finally, antibody response is only one component of the immune response to vaccines. For some antigens (i.e. measles, varicella, yellow fever), cell-mediated immunity is the critical determinant of protection, and excluding quantification of CMI may lead to a mistaken conclusion that boosters are required for all vaccines. Indeed, loss of antibodies does not necessarily imply loss of clinical protection and immune memory can persist, even in the individuals with low antibody concentrations .
As recently underlined in a recent review: "a large variety of immune cells and functions are involved in controlling infections and any assignment of one as the mechanistic immune functions that contribute to protection (mCoP) must recognize that others may be involved as supplements or co-correlates of protection"
. In this respect, cellular responses seem to play a crucial role in the protection conferred by certain vaccines and our choice to exclude reports on CMI lead to a certain degree of imprecision in the evaluation of immune responses. Evaluation of CMI provide important data on the degree of protection at an individual level but methods of measurement are not routinely available and correlates of protection need to be established.
Clinical implications of this work need to be further explored on large prospective cohorts. In the future, the evaluation of new vaccines that specifically target persons with impaired immunity (inactivated or subunit zoster vaccine, cytomegalovirus vaccine) should confront clinical effectiveness with precise evaluation of the both humoral and cellular responses, in an attempt to establish reliable correlates of protection in these populations.