Context Reports of outbreaks of varicella in highly immunized groups have increased
concern about the effectiveness of varicella vaccine.
Objective To assess whether the effectiveness of varicella vaccine is affected
either by time since vaccination or by age at the time of vaccination.
Design Case-control study conducted from March 1997 through June 2003.
Setting Twenty different group practices in southern Connecticut.
Participants Case subjects, identified by active surveillance of all practices, consisted
of 339 eligible children 13 months or older who were clinically diagnosed
as having chickenpox and who also had a polymerase chain reaction (PCR) test
result that was positive for varicella-zoster virus DNA. For each case subject,
2 controls were selected, matched by both age and pediatric practice.
Main Outcome Measures The effectiveness of the vaccine, especially the effects of time since
vaccination and age at the time of vaccination, adjusted for possible confounders.
Results Although the adjusted overall effectiveness of the vaccine was 87% (95%
confidence interval, 81%-91%; P<.001), there was
a substantial difference in the vaccine's effectiveness in the first year
after vaccination (97%) and in years 2 to 8 after vaccination (84%, P = .003). The vaccine's effectiveness in year 1 was substantially
lower if the vaccine was administered at younger than 15 months (73%) than
if it was administered at 15 months or older (99%, P =
.01), although the difference in effectiveness overall for children immunized
at younger than 15 months was not statistically significantly different than
for those immunized at 15 months or older (81% vs 88%, P = .17). Most cases of chickenpox in vaccinees were mild.
Conclusions Although varicella vaccine is effective, its effectiveness decreases
significantly after 1 year, although most cases of breakthrough disease are
mild. If administered at younger than 15 months, the vaccine's effectiveness
was lower in the first year after vaccination, but the difference in effectiveness
was not statistically significant for subsequent years.
The live, attenuated varicella vaccine developed by Takahashi in 1974
was approved in the United States in 1995 and is recommended for routine administration
to healthy children at 12 to 18 months of age and to older children who have
not yet had chickenpox.1,2 Previously,
we reported that the overall effectiveness of the vaccine in clinical practice
was good (85%), at least during the first few years after vaccination.3 However, recent reports4,5 of
outbreaks of chickenpox in groups with substantial (73% and 80%) rates of
immunization, as well as studies6 of immunized
children with breakthrough infections, have increased concern about the current
recommendations for administration of the vaccine.
We now report additional results from an ongoing case-control study
on the influence of age at the time of vaccination and the time since vaccination
on the vaccine's effectiveness. This study includes, with additional analyses,
the 202 polymerase chain reaction (PCR)–positive cases and the 40 PCR-negative
cases and their matched controls from the previous report3 of
the vaccine's effectiveness.
A complete description of the methods has been published previously.3 Subjects were children 13 months to 16 years of age
with no contraindications to vaccination with varicella vaccine. Both potential
cases and potential controls who previously had chickenpox (determined by
both interview and review of medical records) were excluded, since varicella
vaccine is not recommended for such children. Both potential cases and potential
controls who had received the vaccine in the preceding 4 weeks were excluded
from the study. In this population, 99% of families in the practices had a
telephone and no potential case or control was excluded because he/she did
not have a telephone.
The case group consisted of children with chickenpox, identified by
active surveillance, who received medical care at the 20 participating pediatric
practices in southern Connecticut. Investigators were notified of all patients
in each practice who either called the practice because of symptoms and signs
presumed to be chickenpox or came to the physicians' offices because of an
illness thought to be chickenpox.
On approximately the third to fifth day of the illness, a research assistant
visited the home of each patient with chickenpox and conducted a brief interview
with the parent to ascertain demographic information, pertinent medical history,
risk factors, and whether the child attended school or day care. The severity
of the illness was assessed based on a modified version of a clinical scale
used in previous clinical trials of varicella vaccine that takes into account
the number and type of lesions (eg, vesicular, hemorrhagic), the height of
the fever, the presence of systemic signs, and a subjective assessment of
how ill the child was.3 A score of 7 or lower
was considered mild disease, scores of 8 to 15 were considered moderately
severe disease, and scores of 16 or higher were considered severe disease.
A vesicular lesion was gently unroofed with a capillary tube that was
also used to collect vesicular fluid. Material also was obtained by swabbing
the underlying skin with a cotton-tipped swab. A PCR assay was performed on
all specimens to detect the presence of DNA of varicella-zoster virus (VZV).7 Specimens were coded so that the technicians and the
investigators who performed and interpreted the PCR tests were blind to whether
the subject had received varicella vaccine. For PCR, assay results were considered
positive if the specimen was positive for DNA of VZV and all negative controls
in that batch were negative. The test results were considered negative if
the specimen was negative for DNA of VZV, all positive controls in that batch
were positive, and the specimen was positive for β-globin (indicating
the presence of amplifiable human DNA in the specimen). However, if the specimen
was negative for both DNA of VZV and β-globin, it was considered an inadequate
sample.
We selected 2 controls who had not had chickenpox for each case subject,
matched by both date of birth (±1 month) and source of primary care.
Controls were selected from a list of potential controls by using a table
of random numbers to select the order in which potential controls were contacted.
The medical records of the subjects (both cases and controls) were reviewed,
and all information about previous immunizations and about significant medical
illnesses was recorded. Records of all health care practitioners (including
previous practitioners) were checked. Antecedent vaccination was defined as
written documentation that varicella vaccine had been received at least 4
weeks before focal time (date of onset of varicella for each case, used for
both cases and their matched controls). Only written documentation of receipt
of vaccines was accepted as evidence of prior immunization.
The effectiveness of a vaccine, defined as the proportionate reduction
in the risk of infection among vaccinees that was attributable to vaccination,
is calculated with data from clinical trials as follows: (1 − relative
risk [RR]).8 In matched case-control studies
in which the controls are matched individually to the cases, the standard
measure of association is the matched odds ratio (OR). Since for this type
of study the matched OR closely approximates the RR that would be observed
in a clinical trial,9 the matched OR can be
substituted for the RR in the above equation and the vaccine's effectiveness
is estimated as follows: (1 − matched OR).
Data were analyzed primarily with SAS/STAT statistical software version
8.2 for Windows.10 Matched ORs, with both their
associated statistical significance (assessed with the maximum-likelihood χ2 test for matched triplets) and their 95% confidence intervals (CIs)
were calculated with the use of conventional techniques.11 The
vaccine's effectiveness was estimated directly from the above equation. Conditional
logistic regression was used to calculate matched ORs for the effects of time
since vaccination and age at the time of vaccination, as well as to adjust
for effects of possible confounders, including sex, race, attendance at group
day care, asthma, use of steroids, and receipt of varicella vaccine within
28 days after receiving the measles-mumps-rubella (MMR) vaccine.12 In
all calculations of ORs in the multivariable models, the unvaccinated group
was the reference group. Three separate conditional logistic regression models
were run: one to assess the effect of time since vaccination (a dummy-coded
variable), one to assess the effect of age at the time of vaccination (also
a dummy-coded variable), and one to assess the interaction between age and
time since vaccination. A t test was used to assess
the statistical significance of differences between groups in continuous variables
such as age, whereas the χ2 test was used to assess the statistical
differences between categorical values. All P values
are 2-sided. Results were considered statistically significant if the 2-tailed P value was <.05.
As a strategy to assess whether there might have been bias introduced
in the selection of the controls, we compared the proportion of subjects who
had received the MMR vaccine among both the cases and the controls.3 Since the MMR vaccine should have been administered
at approximately the same age as the varicella vaccine and it should have
no effect on the risk of developing varicella, we expected that there would
be no significant difference between the cases and the controls in the proportions
who had received the MMR vaccine. A significant difference could indicate
that selection bias may have been a problem (since this may be a marker for
use of medical care).
We also performed an analysis in which we assessed the vaccine's effectiveness
for potential cases whose PCR test results were negative (ie, children who
did not have chickenpox by our definition). If, using the same methods, the
study showed that the vaccine's effectiveness in preventing PCR-positive varicella
was good but the vaccine was not effective in preventing PCR-negative potential
cases of varicella, it would be strong evidence that the results were not
attributable to bias (since all potential cases and controls were selected
in the same manner).3 The study was approved
by Yale's Human Investigation Committee; written informed consent was obtained
from case parents and oral informed consent was obtained from control parents
(written assent was obtained from the case subject when appropriate).
From March 1997 through June 2003, of the 634 potential case subjects
contacted who were eligible for the study, 530 (84%) were enrolled. Of the
others, a sample for PCR could not be obtained from 30 (5%) and 74 (12%) refused
to participate. For the potential case subjects who were enrolled, the results
of the PCR assay were positive in 364 (69%), negative in 124 (23%), and indeterminate
in 42 (8%). Of the 1164 potentially eligible controls who were reached, 80
(7%) refused to participate. Information was complete for the case and at
least 1 matched control for 339 matched case-control groups in which the results
of the PCR assays in the potential case subjects were positive. Data from
these subjects formed the basis for the analyses of the effectiveness of the
vaccine.
Characteristics of the subjects included in the analyses of the vaccine's
effectiveness are shown in Table 1.
Cases and controls were similar in demographic characteristics but differed
in receipt of varicella vaccine and in the proportions vaccinated at younger
than 15 months and at 12 months or less before the onset of varicella in the
case subjects. The numbers of cases who were enrolled each year (1997-2003)
who were included in the analyses were 75, 71, 68, 48, 33, 43, and 1, respectively.
The results of the unadjusted estimate of the overall effectiveness of the
vaccine are shown in Table 2.
Of the 339 case-control groups, 330 had 2 matched controls and 9 had 1 matched
control. The overall effectiveness of the vaccine was 87% (OR, 0.13 [95% CI,
0.09-0.20]; P<.001). The effectiveness was virtually
unchanged after controlling for potential confounders (sex, race, location
of care during the day, history of asthma, use of corticosteroids, and receipt
of varicella vaccine within 28 days after being immunized with the MMR vaccine).
Chickenpox was significantly more severe in unvaccinated children (mean [SD]
and median [interquartile range] severity scores, 7.3 [3.1] and 8 [4-9], respectively)
than in vaccinated children (mean [SD] and median [interquartile range] severity
scores, 4.5 [2.2] and 3 [3-5], respectively) (P<.001).
Of the 122 vaccinated case subjects, 106 (87%) had mild varicella compared
with 98 (45%) of the 217 unvaccinated case subjects (P<.001).
The rash was mostly vesicular in 37 (30%) of the vaccinated cases compared
with 126 (58%) of the unvaccinated cases (P<.001).
The vaccine's overall effectiveness against moderate or severe disease was
98% (95% CI, 93%-99%; P<.001) and was not significantly
different if the child was vaccinated at 15 months or younger.
The vaccine's effectiveness in the first year after vaccination was
97%, which decreased to 86% in the second year after vaccination and to 81%
in years 7 to 8 after vaccination (Table
3). The difference between the effectiveness in year 1 and year
2 was statistically significant (P = .007), as it
was between year 1 and each of the subsequent years. However, the differences
between the effectiveness in year 2 and in each subsequent year (including
years 7-8) were not statistically significantly different (P = .63). The trend for the decrease in the vaccine's effectiveness
with time was not linear. Consequently, we combined the estimates of the vaccine's
effectiveness for years 2 through 8 (Table
3).
The vaccine's effectiveness in the first year after vaccination was
substantially lower if the vaccine was administered when the child was younger
than 15 months than if the child was 15 months or older at the time of vaccination
(73% vs 99%, P = .01), although the difference in
the effectiveness for these age groups was not statistically significant either
for years 2 to 8 or overall (Table 4).
For comparison, the results for all children in the study (vaccinated at ≥12
months) are also included in Table 4.
Among vaccinees who developed chickenpox, the disease was mild in 88% of those
vaccinated at younger than 15 months and in 81% of those vaccinated at 15
months or older (P = .30). The vaccine's effectiveness
in year 1 vs years 2 to 8 was also significantly different if the child was
15 months or older at the time of vaccination but not if the child was younger
than 15 months at the time of vaccination.
Although there was a substantial difference in the proportions of cases
and controls who had received varicella vaccine, all but 1 case and 1 control
had received the MMR vaccine, initial administration of which is recommended
at approximately the same age as for the varicella vaccine (Table 1). There were 113 potential cases for whom the PCR test result
was negative and for whom information about the potential case and at least
1 matched control was complete. Of these, 98 (87%) of the 113 PCR-negative
potential cases and 182 (81%) of their 225 matched controls had received varicella
vaccine. The matched OR was 1.44. The effectiveness of the vaccine against
PCR-negative potential cases (1 − matched OR) was −56% and was
not significantly different than 0% (95% CI, −197% to 18%; P = .18). Both of these analyses suggest that bias did not have a substantial
effect on the results of this study.
This study indicates that at least through the first 8 years after vaccination,
the overall effectiveness of live, attenuated varicella vaccine remains good,
although breakthrough varicella is not rare. Most vaccinated children who
develop chickenpox have mild disease, regardless of their age at the time
of vaccination or the time since vaccination, at least up to 7 to 8 years
after vaccination (ie, the vaccine's effectiveness against moderate to severe
disease is excellent throughout the period of the study).
However, there is a substantial, statistically significant decrease
in the vaccine's overall effectiveness in the second year after vaccination,
after which the decrease in the vaccine's effectiveness is not statistically
significant, at least through years 7 to 8 after vaccination. We do not know
the explanation for this phenomenon, although it is consistent with observations
in other studies4-6 that
the risk of breakthrough infection increases over time. Presumably, this is
a result of waning immunity in a proportion of immunized children in addition
to occasional primary vaccine failure.13 Although
the breakthrough infections are usually mild, such infections nevertheless
may place the child at higher risk of subsequently developing zoster and may
result in spread of varicella to susceptible contacts.4,13
The vaccine's effectiveness in the first year after vaccination is substantially
lower in children who are vaccinated at younger than 15 months. This is consistent
with other reports4-6 that
have indicated that children vaccinated at younger than 15 months are at increased
risk of breakthrough infection. Changing the age at which immunization with
varicella vaccine is begun from 12 to 15 months might alleviate this problem.
However, the improved effectiveness of the vaccine would have to be balanced
against both the risk of leaving such children unvaccinated for these 3 months
and the risk that some children might not return for vaccination in a timely
manner. Alternatively, administering a second dose of the vaccine might also
solve both this problem and the problem of waning immunity.13,14 More
data are needed about the effect of a second dose of the vaccine on duration
of immunity to varicella.
Because this is a nonexperimental study, bias may have affected the
results. However, the analyses that showed both that there was no difference
in the proportions of cases and of controls who had received MMR vaccine and
that the vaccine was not effective against potential cases with a PCR result
that was negative (even though both these cases and their matched controls
were selected in the same manner as were the PCR-positive cases and their
matched controls) suggest that bias did not have an important effect. The
study was conducted only in private practices, although the racial distribution
of the population was similar to that of the entire state of Connecticut.
The vaccine had only been in routine use in this country for up to 7 to 8
years at the time the analyses were performed, so the effect of longer duration
since vaccination could not be assessed. In addition, during much of the study
period varicella was still circulating widely, and subclinical infection and
boosting of vaccine-induced immunity through natural exposure likely occurred
to some extent. As the incidence of varicella continues to diminish, boosting
of immunity through natural exposure will become increasingly rare.
It is clear that the incidence of varicella in the United States is
decreasing as a result of the widespread use of varicella vaccine.15,16 Nevertheless, in the United States,
deaths from varicella and other complications in immunocompetent persons still
occur and will continue to occur until the infection is eliminated.17 It is important to monitor closely the incidence
of varicella and the effectiveness of the vaccine over time to determine if
a booster dose is needed to improve its effectiveness.
1.Takahashi M, Otsuka T, Okuno Y, Asano Y, Yazaki T, Isomura S. Live vaccine used to prevent the spread of varicella in children in
hospital.
Lancet.1974;2:1288-1290.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=4139526&dopt=Abstract
Google Scholar 2. Prevention of varicella: recommendations of the Advisory Committee
on Immunization Practices (ACIP)
MMWR Recomm Rep.1996;45(RR-11):1-36.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=8668119&dopt=Abstract
Google Scholar 3.Vázquez M, LaRussa PS, Gershon AA, Steinberg SP, Freudigman K, Shapiro ED. The effectiveness of the varicella vaccine in clinical practice.
N Engl J Med.2001;344:955-960.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=11274621&dopt=Abstract
Google Scholar 4.Galil K, Lee B, Strine T.
et al. Outbreak of varicella at a day-care center despite vaccination.
N Engl J Med.2002;347:1909-1915.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=12477940&dopt=Abstract
Google Scholar 5.Galil, K, Fair E, Mountcastle N, Britz P, Seward J. Younger age at vaccination may increase risk of varicella vaccine failure.
J Infect Dis.2002;186:102-105.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=12089668&dopt=Abstract
Google Scholar 6.Verstraeten T, Jumaan AO, Mullooly JP.
et al. A retrospective cohort study of the association of varicella vaccine
failure with asthma, steroid use, age at vaccination, and measles-mumps-rubella
vaccination.
Pediatrics.2003;112:e98-e103.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=12897314&dopt=Abstract
Google Scholar 7.LaRussa P, Lungu O, Hardy I, Gershon A, Steinberg S, Silverstein S. Restriction fragment length polymorphism of polymerase chain reaction
products from vaccine and wild-type varicella-zoster virus isolates.
J Virol.1992;66:1016-1020.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=1346169&dopt=Abstract
Google Scholar 8.Orenstein WA, Bernier RH, Hinman AR. Assessing vaccine efficacy in the field: further observations.
Epidemiol Rev.1988;10:212-241.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=3066628&dopt=Abstract
Google Scholar 9.Cornfield J. A method of estimating comparative rates from clinical data: application
for cancers of the lung, breast, and cervix.
J Natl Cancer Inst.1950-51;11:1269-1275.Google Scholar 10.SAS Institute Inc. SAS OnlineDoc. Version 8. Cary, NC: SAS Institute Inc; 1999.
11.Abramson JH, Gahlinger PM. Computer Programs for Epidemiologists PEPE v. 4.0. Salt Lake City, Utah: Sagebrush Press; 2001:113-118.
12.Breslow NE, Day NE. Statistical Methods in Cancer Research, Vol. 1: The
Analysis of Case-Control Studies. Lyon, France: World Health Organization; 1980.
13.Gershon AA. Varicella vaccine—are two doses better than one?
N Engl J Med.2002;347:1962-1963.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=12477947&dopt=Abstract
Google Scholar 14.Watson B, Boardman C, Laufer D.
et al. Humoral and cell-mediated immune responses in healthy children after
one or two doses of varicella vaccine.
Clin Infect Dis.1995;20:316-319.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=7742436&dopt=Abstract
Google Scholar 15.Seward JF, Watson BM, Peterson CL.
et al. Varicella disease after introduction of varicella vaccine in the United
States, 1995-2000.
JAMA.2002;287:606-611.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=11829699&dopt=Abstract
Google Scholar 16. Decline in annual incidence of varicella—selected states, 1990-2001.
MMWR Morb Mortal Wkly Rep.2003;52:884-885.http://www.ncbi.nlm.nih.gov/htbin-post/Entrez/query?db=m&form=6&Dopt=r&uid=entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=13679791&dopt=Abstract
Google Scholar 17. Varicella-related deaths—United States, 2002.
MMWR Morb Mortal Wkly Rep.2003;52:545-547.Google Scholar