Vaccine 29 (2011) 71007106

Contents lists available at ScienceDirect

Vaccine
journal homepage: www.elsevier.com/locate/vaccine

Effectiveness of a vaccination programme for an epidemic of meningococcal B in

New Zealand
Richard Arnold a, , Yvonne Galloway b , Anne McNicholas b , Jane OHallahan c
a
School of Mathematics, Statistics and Operations Research, Victoria University of Wellington, PO Box 600, Wellington 6140, New Zealand
b
Ministry of Health, Wellington, New Zealand
c
Royal New Zealand College of General Practitioners, Wellington, New Zealand

a r t i c l e i n f o a b s t r a c t

Article history: New Zealand has experienced a prolonged epidemic of meningococcal B disease since 1991. The epi-
Received 16 January 2011 demic has waned signicantly since its most recent peak in 2001. A strain-specic vaccine, MeNZB, was
Received in revised form 30 May 2011 introduced to control the epidemic in 2004, achieving 81% coverage of people under the age of 20. The
Accepted 29 June 2011
vaccine was rolled out in a staged manner allowing the comparison of disease rates in vaccinated and
Available online 29 July 2011
unvaccinated individuals in each year.
Vaccine effectiveness in people aged under 20 years is estimated using a Poisson regression model in
Keywords:
the years 20012008, including adjustments for year, season, age, ethnicity, region and socioeconomic
Meningococcal vaccine
Meningococcal disease
status. Further analyses investigate the dose response relationship, waning of the vaccine effect after one
Vaccine effectiveness year, and cross-protection against other strains of meningococcal disease.
Observational study The primary analysis estimates MeNZB vaccine effectiveness to be 77% (95% CI 6285) after 3 doses
and a mean follow-up time of 3.2 years. There is evidence for a protective effect after 2 doses 47% (95% CI
1667), and no evidence for a waning of effectiveness after one year. Simultaneous modelling of invasive
pneumococcal disease and epidemic strain meningococcal B suggests a degree of residual confounding
that reduces the effectiveness estimate to 68%. There is evidence for some cross-protection of MeNZB
against non-epidemic strains.
The MeNZB vaccine was effective against the New Zealand epidemic strain of meningococcal B disease.
Between July 2004 and December 2008 an estimated 210 epidemic strain cases (95% CI 100380), six
deaths and 1530 cases of severe sequelae were avoided in New Zealand due to the introduction of the
MeNZB vaccine.
2011 Elsevier Ltd. All rights reserved.

1. Introduction was to increase the proportion of infants attaining an increase in

serum bactericidal antibodies (SBAs) to a proportion similar to that
New Zealand experienced a prolonged epidemic of meningococ- achieved in older age groups with three doses [48].
cal B disease starting in 1991. Its most recent peak was in 2001 with Similar to other meningococcal vaccines [911] pre-licensure
signicant waning since then [1,2]. In response, a strain-specic Phase III trials were not carried out, so there was no direct measure
outer membrane vesicle vaccine, MeNZB, was delivered through a of vaccine efcacy prior to the use of MeNZB in the population.
national programme, staggered by region and age group, from July However, comprehensive administrative data sources allowed the
2004 to June 2006. MeNZB remained in the routine infant immu- impact of the programme at a population level to be readily mon-
nisation schedule until 1 June 2008 [3]. School aged children (517 itored. Vaccination status was available from an immunisation
years) were vaccinated by public health nurses in schools and all register and case details were sourced from disease surveillance
others were vaccinated by primary care providers. data [12]. A cohort study in children under 5 years found an
A three-dose schedule was offered to all under the age of 20 effectiveness of 80% (95% CI 5392) [13]. The staggered vaccine
years, and achieved a coverage of 81%. From January 2006 a fourth delivery allowed comparison of disease rates in vaccinated and
dose, administered at 10 months of age, was introduced for chil- unvaccinated individuals in each year, thus avoiding potential con-
dren who received their rst dose before 6 months of age. This founding through changing overall disease rates as the epidemic
progressed [14]. Using data from 2001 to 2006, a Poisson regression
analysis model estimated vaccine programme effectiveness at 73%
Corresponding author. Tel.: +64 4 463 5668; fax: +64 4 4635045. (95% CI 5285) [15]. In this paper we update this estimate using
E-mail address: Richard.Arnold@vuw.ac.nz (R. Arnold). data to December 2008. We also test for (i) waning effect of the

0264-410X/$ see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.vaccine.2011.06.120
 R. Arnold et al. / Vaccine 29 (2011) 71007106 7101

vaccine after 12 months, (ii) possible protective effects of fewer

400
than 3 doses, (iii) cross-protection against other meningococci
Epidemic Group B
(such as has been found elsewhere by Tappero et al. [16]), and (iv)

350
Non epidemic Group B
possible residual confounding. Non Group B
Unknown Group or Subtype

300
2. Methods

250
Number of Cases
The study population was New Zealand residents aged under
20 years between January 2001 and December 2008. Demographic

200
population strata were dened by four regions, age (up to 6 months,
611 months, 14 years and 519 years), ethnicity (Maori, Pacic

150
peoples or Other) and socioeconomic status (New Zealand Index
of Deprivation quintiles, 2001 [17]). The data were further divided
in time into years and seasonal quarters. JulySeptember approx-

100
imately covers New Zealands winter. Population estimates were
taken from Statistics New Zealand medium growth projections

50
[18].
Vaccination status came from National Immunisation Register

0
(NIR) reports at the midpoint of each quarter in each year, by dose,
region, age, and ethnicity, and was dened as follows: those fully 2001 2002 2003 2004 2005 2006 2007 2008
vaccinated had received 3 or more doses, unless the rst dose
was received below 6 months of age, in which case 4 doses were Fig. 1. Counts of meningococcal disease cases, classied into 4 types. The category
Unknown Group of Subtype may include unidentied epidemic group B cases.
required; unvaccinated had received no doses; partially vaccinated
were those not fully vaccinated. For testing the possible waning
effect of the vaccine, fully vaccinated individuals were further clas- identied through this system. Invasive pneumococcal disease only
sied as recently vaccinated (if their nal dose was within the became a notiable disease in October 2008, following the addition
previous 12 months), or earlier vaccinated otherwise. Deprivation of the pneumococcal conjugate vaccine to the routine infant sched-
quintile was not available on the NIR so we assumed that the dis- ule in June 2008. The laboratory-identied pneumococcal cases
tribution of individuals across deprivation quintiles was the same were matched with NIR records to determine MeNZB vaccination
among vaccinated people as in the general population (in each status.
demographic by time cell). A Poisson regression model was tted including all available
The unvaccinated population was calculated as the difference explanatory variables (year, quarter, region, age, ethnicity, depri-
between the population estimates and the number of vaccinated vation quintile) and vaccination status. All explanatory variables
individuals in each demographic by time cell. In cells where the were categorical, with the exception of deprivation quintile which
number of vaccinated individuals exceeded 98% of the population was introduced in quantitative terms (linear and quadratic). Using
estimate, the number of unvaccinated individuals was set to 2% the generalised estimating equation (GEE) methodology [19] we
of the total population. This assumption increased the estimated allowed for possible over or underdispersion, and an autoregres-
population under 20 by 1.3% in 2008, but ensured that no cells had sive AR(1) correlation structure was assumed to model correlation
nearly zero unvaccinated populations, thereby increasing the size between successive time periods (quarters). Model tting consid-
of the unvaccinated population and reducing the estimated disease ered all main effects, and all two-way interactions.
rate in that group. It is therefore conservative in its impact on the We report our results as relative risks (RR, ratio of rates of dis-
vaccine effectiveness estimate. ease in the unvaccinated compared to the vaccinated) and vaccine
Cases of meningococcal disease were derived from national dis- effectiveness (1 1/RR, the proportion of cases avoided).
ease notications made to the Institute of Environmental Science
and Research (ESR), where patient specimens, meningococci and 3. Results
meningococcal DNA from cases are also sent for strain characteri-
sation. We group cases as: (1) epidemic group B, (2) non-epidemic The counts and crude rates of conrmed cases of epidemic strain
group B, (3) non group B and (4) unknown group or subtype. Counts meningococcal B disease for people aged 019 years are shown by
in these four groupings are shown in Fig. 1. vaccination status in Table 1. Throughout the campaign there was
To test for residual confounding, ESR provided details of labora- a steady increase in vaccinated people, a decrease in unvaccinated
tory conrmed invasive pneumococcal disease cases from isolates people, with a small number in the transitory partially vaccinated
or clinical specimens referred over the same period. ESR esti- state. At the end of 2008 the mean length of time per person in the
mates that 75% of all laboratory conrmed pneumococcal cases are partially vaccinated state was 0.7 years, and 3.2 years in the fully

Table 1
Crude epidemic strain meningococcal rates for conrmed cases aged 019 years. Rates are cases per 100,000 persons per year.

Year Unvaccinated Partially vaccinated Fully vaccinated

Cases Population Rate Cases Population Rate Cases Population Rate

2001 296 1,154,102 25.6 0 0 0.0 0 0 0.0

2002 218 1,167,867 18.7 0 0 0.0 0 0 0.0
2003 189 1,181,667 16.0 0 0 0.0 0 0 0.0
2004 138 1,140,790 12.1 7 36,683 19.1 0 12,081 0.0
2005 50 530,728 9.4 22 265,924 8.3 9 395,151 2.3
2006 13 164,410 7.9 15 110,083 13.6 19 918,380 2.1
2007 11 133,337 8.2 10 104,762 9.5 17 956,883 1.8
2008 6 126,305 4.8 10 97,483 10.3 15 971,680 1.5
7102 R. Arnold et al. / Vaccine 29 (2011) 71007106

Table 2 Table 2 (Continued)

Parameter estimates for a Poisson regression model for epidemic strain meningo-
coccal rates for conrmed cases aged 019 years. Parameter Relative risk 95% conf. interval p-Value

Quarter Vacc. status

Parameter Relative risk 95% conf. interval p-Value
JanuaryMarch
b
Region Full 2.04 (0.82,5.06) 0.1246
Northern 0.84 (0.64,1.11) 0.2321 Partial 0.45 (0.20,1.02) 0.0551
Midland 1.64 (1.20,2.24) 0.0021 Unvaccinated 1.00
Central 0.69 (0.48,0.98) 0.0410 AprilJune
Southern 1.00 Full 1.94 (0.76,4.91) 0.1638
Age group Partial 0.53 (0.26,1.09) 0.0830
<6 months 1.45 (0.78,2.72) 0.2412 Unvaccinated 1.00
6 months1 years 1.50 (0.81,2.81) 0.1999 JulySeptember
14 years 1.45 (0.99,2.13) 0.0589 Full 3.40 (1.44,8.04) 0.0054
519 years 1.00 Partial 0.30 (0.16,0.58) 0.0003
Ethnicityb Unvaccinated 1.00
Maori 0.70 (0.40,1.21) 0.2006 OctoberDecember
Pacic 2.33 (1.25,4.34) 0.0080 Full 1.00
Other 1.00 Partial 1.00
Deprivationa 1.17 (1.08,1.26) 0.0002 Unvaccinated 1.00
Year a
Deprivation is included as a continuous variable: the effect size is the change in
2001 4.21 (2.80,6.35) <0.0001
disease rate per additional quintile of deprivation.
2002 3.11 (2.01,4.83) <0.0001 b
Relative risks for main effects refer to the reference category, and do not include
2003 2.75 (1.78,4.25) <0.0001
interaction effects which modify disease rates in other groups.
2004 2.14 (1.42,3.24) 0.0003
2005 1.76 (1.14,2.70) 0.0101
2006 1.43 (0.90,2.28) 0.1304
2007 1.17 (0.75,1.84) 0.4917 vaccinated state. 91% of individuals who started the vaccination
2008 1.00 programme became fully vaccinated.
Quarterb In the Poisson regression model main effects for all of the
JanuaryMarch 0.91 (0.73,1.14) 0.3990
covariates were signicant, along with interactions between
AprilJune 1.25 (1.03,1.52) 0.0233
JulySeptember 1.91 (1.56,2.34) <0.0001 region ethnicity, age ethnicity, deprivation age and quar-
OctoberDecember 1.00 ter vaccination status. The model would not converge for certain
Vacc. statusb two-way interactions involving year (with quarter, age, ethnicity,
Full 0.12 (0.05,0.28) <0.0001
vaccination status) due to small numbers of cases. Parameter esti-
Partial 1.40 (0.90,2.16) 0.1308
Unvaccinated 1.00
mates with condence intervals are displayed in Table 2. Strong
Region ethnicity effects were found for year (the waning of the epidemic), quarter
Northern (rates were highest in the JulySeptember quarter), age (highest
Maori 2.05 (1.19,3.53) 0.0097 rates in the youngest age groups), deprivation (higher rates among
Pacic 0.83 (0.45,1.55) 0.5619
the more deprived) and ethnicity (highest rates among Pacic peo-
Other 1.00
Midland ples, lower rates in Maori, and lowest for the Other ethnic group).
Maori 1.28 (0.74,2.21) 0.3846 As seen in Table 2, the interaction between vaccination status and
Pacic 0.49 (0.22,1.11) 0.0864 quarter only signicantly modies the vaccine effectiveness in the
Other 1.00
winter months. No other interactions with vaccine effectiveness
Central
Maori 2.47 (1.35,4.51) 0.0034
were found, in particular there was no evidence for changing effec-
Pacic 1.45 (0.75,2.80) 0.2721 tiveness with age. (A model including the interaction between
Other 1.00 vaccination status and age converged only if children younger than
Southern 6 months were excluded, but this interaction was not signicant.)
Maori 1.00
The overall vaccine effectiveness estimates are shown in Table 3.
Pacic 1.00
Other 1.00 Where vaccination status is classied into three levels we nd a
Age group ethnicity strong protective effect of full vaccination, 77% (95% CI 6285), and
<6 months a relative risk of 4.3 (95% CI 2.66.9) for unvaccinated vs. vacci-
Maori 2.71 (1.54,4.75) 0.0005 nated. No signicant protective effect of partial vaccination was
Pacic 2.57 (1.42,4.66) 0.0019
Other 1.00
found (p-value 0.08). The coverage of the vaccination programme
6 months1 years was very uniform across deprivation quintiles (in June 2006 cover-
Maori 3.07 (1.65,5.72) 0.0004 ages were 83%, 80%, 79%, 79%, 81% from the least to most deprived
Pacic 2.45 (1.25,4.78) 0.0087 quintiles for ages 019 years). In a sensitivity analysis we changed
Other 1.00
our assumption of uniform coverage across deprivation quintiles to
14 years
Maori 2.91 (2.01,4.22) <0.0001 match the June 2006 distribution. This changed our vaccine effec-
Pacic 2.67 (1.77,4.01) <0.0001 tiveness estimate by only 0.1% (increasing it to 76.6%). We also
Other 1.00 tested a signicantly more conservative scenario where coverage
519 years was assumed to decline linearly with increasing deprivation (with
Maori 1.00
Pacic 1.00
coverage rates in the most deprived being 20% less than the least
Other 1.00 deprived). This also led to only a small change in our estimates,
Deprivation age group reducing the vaccine effectiveness by 0.7% (from 76.5% to 75.8%).
<6 months 1.30 (1.06,1.59) 0.0105 Applying the results from this model in Table 2 we estimated the
6 months1 years 1.26 (1.03,1.54) 0.0268
number of cases avoided by the vaccination programme (Table 4).
14 years 1.00 (0.88,1.13) 0.9568
519 years 1.00 For each year we calculated the proportion of laboratory-conrmed
cases that were epidemic strain and thereby estimated the number
of undetected epidemic strain cases among cases with unknown
serotypewe then added these to the total cases. We estimate
 R. Arnold et al. / Vaccine 29 (2011) 71007106 7103

Table 3
Vaccine effectiveness estimates from Poisson regression models for epidemic strain meningococcal B disease for ages 019 years, years 20012008 for various levels of
vaccine protection. Relative risks (RR) are stated as ratios of rates in the less protected population to the more protected population. Vaccine effectiveness is 1 1/RR, and
the p-values are for test that the relative risks differ from 1.0 (0% vaccine effectiveness).

Vaccination status Relative risk Effectiveness p-Value

Estimate 95% CI Estimate 95% CI

3-Level vaccination statusa

Full vs. unvaccinated 4.3 (2.6,6.9) 76.5% (62,85) <0.0001
Partial vs. unvaccinated 1.4 (1.0,2.0) 27.4% (4,49) 0.0770
Full vs. partial 3.1 (1.9,5.0) 67.6% (48,80) <0.0001
4-Level vaccination statusa
Recently vs. unvaccinated 5.4 (2.6,11.4) 81.5% (61,91) <0.0001
Earlier vs. unvaccinated 3.9 (2.3,6.4) 74.2% (57,84) <0.0001
Partial vs. unvaccinated 1.4 (1.0,1.9) 26.7% (5,47) 0.0862
Recently vs. partial 4.0 (1.8,8.5) 74.8% (46,88) 0.0004
Recently vs. earlier 1.4 (0.6,3.2) 28.4% (63,69) 0.4260
Earlier vs. partial 2.8 (1.7,4.7) 64.8% (42,79) <0.0001
Dosesb
3+ vs. none 3.3 (2.2,4.7) 69.2% (55,79) <0.0001
2 vs. none 1.9 (1.2,3.0) 47.2% (16,67) 0.0076
1 vs. none 1.2 (0.7,1.8) 14.0% (35,45) 0.5087
3+ vs. 2 1.7 (1.0,2.9) 41.9% (3,65) 0.0365
2 vs. 1 1.6 (0.9,2.9) 38.5% (9,65) 0.0968
a
Terms in this model are those listed in Table 2.
b
Terms in this model are the same as Table 2 excluding the quarter by vaccination status interaction.

208 epidemic strain meningococcal B cases were avoided to the We tested for a possible protective effect of MeNZB against
end of 2008. We applied the observed (age dependent) case fatal- other meningococci strains. Poisson regression models were t-
ity rates for the epidemic strain to the numbers of cases avoided, ted as before using disease counts from each of six groupings, the
and estimate that 5.6 deaths (95% CI 2.910.0) were avoided. Based data were restricted to ages 6 months19 years, and to the years
on published data on sequelae of meningococcal disease [2026], 20022008. The resulting vaccine effectiveness estimates for the
including a chart review of 303 serogroup B cases by Healy et al. three level vaccination status are shown in Table 5.
[24], we estimate that 21 cases with serious sequelae were avoided The effect of full vaccination was highest for the epidemic
(approximately 10% of cases, interval estimate 1627 cases). group B and unknown group or subtype outcomes. This indicates
When vaccination status is classied into four levels (Table 3), specic activity against the epidemic strain since the unknown
dividing those fully vaccinated in the previous 12 months from group or subtype category is expected to contain unconrmed epi-
those vaccinated earlier, the same strong effect of full vaccination demic strain cases. All of the remaining outcomes have a lower
was found, again with no evidence of a protective effect of par- but signicant vaccine effectiveness which is possible evidence of
tial vaccination. There was no detectable waning of the vaccine cross-protection with lesser effectiveness. The condence intervals
effectiveness after 12 months (p-value 0.43). are, however, very wide due to low numbers of cases.
The partially vaccinated group is made up of those with 1 or 2
doses (and 3 doses in the case of children receiving their rst dose 4. Pneumococcal disease and MeNZB
under 6 months of age). We retted models with dose level (0, 1, 2
and 3+) as an explanatory variable to test whether there was some In order to test for residual confounding we investigated
protection from 2 doses (Table 3). The tted model included the whether MeNZB was effective against pneumococcal disease, since
same terms as shown in Table 2, except that there was no interac- there is no biological mechanism for the vaccine to confer any such
tion between quarter and vaccination status. We found a signicant protection. There were 1298 cases of laboratory conrmed inva-
relationship with vaccine effectiveness of 47% for 2 vs. no doses sive pneumococcal disease in the period 20022008, with the same
compared to 69% for 3+ doses. The 4th dose for infants receiving covariate information available as for meningococcal disease cases.
their rst dose under 6 months of age was added only towards the Vaccine effectiveness results are shown in Table 6. With high
end of the vaccination programme. We tted a separate Poisson signicance, individuals fully vaccinated with MeNZB have lower
regression model for children aged under 5 years who were eligi- rates of pneumococcal disease than unvaccinated individuals (vac-
ble to receive a 4th dose, and estimated a separate effect for all 4 cine effectiveness 56%, with 95% CI 4466). Moreover there is a
dose levels (1, 2, 3, 4). There were 47 conrmed epidemic strain signicant, but lesser, effect of partial MeNZB vaccination (26%,
cases in 380,000 eligible person years, of which 15 were in children 95% CI 343). These results suggest some degree of residual con-
who had had 3 doses and 10 who had had 4 doses. No signicant founding. To quantify its effect we tted further Poisson regression
additional protection of the 4th dose was found (RR for 4 vs. 3 doses models, where counts of both meningococcal and pneumococcal
was 1.9 with 95% CI 0.84.8 and p-value 0.16). disease cases were included simultaneously with disease type as

Table 4
Observed numbers of cases and case fatalities, with estimated cases and case fatalities avoided by the MeNZB vaccination programme (ages 019 years, 20012008).

Age group Actual cases Case fatalities Case fatality rate Cases prevented Case fatalities prevented

<6 months 102 8 7.8% 5 0.4

6 months1 year 97 8 8.2% 11 0.9
14 years 337 10 2.9% 78 2.3
519 years 486 9 1.8% 114 2.0
Total 1022 54 3.9% 208 5.6
(95% CI) (105,375) (2.9,10.0)
7104 R. Arnold et al. / Vaccine 29 (2011) 71007106

Table 5
Vaccine effectiveness estimates from Poisson regression models for various groupings of meningococcal disesase cases, for ages 6 months19 years in the period 20022008.
Relative risks (RR) are stated as ratios of rates in the less protected population to the more protected population. Vaccine effectiveness is 1 1/RR, and the p-values are for
test that the relative risks differ from 1.0 (0% vaccine effectiveness).

OutcomeMeNZB vaccination status Relative risk Effectiveness p-Value

Estimate 95% CI Estimate 95% CI

Full vs. unvaccinated

(1) Epidemic group Ba 4.8 (2.9,7.8) 79.0% (66,87) <0.0001
(2) Non-epidemic group B 2.2 (1.1,4.1) 53.7% (12,76) 0.0194
(3) Non group B 2.3 (1.2,4.3) 56.2% (17,77) 0.0118
(4) Unknown group or subtype 3.4 (1.8,6.3) 70.5% (45,84) 0.0001
(5) Conrmed non-epidemicb 2.5 (1.6,3.8) 59.9% (38,74) <0.0001
(6) All meningococcalc 3.1 (2.3,4.1) 67.4% (57,76) <0.0001
Partial vs. unvaccinated
(1) Epidemic group Ba 1.7 (1.1,2.6) 41.8% (11,62) 0.0120
(2) Non-epidemic group B 1.5 (0.8,3.1) 35.3% (29,68) 0.2166
(3) Non group B 1.1 (0.5,2.7) 12.0% (107,63) 0.7706
(4) Unknown group or subtype 2.0 (1.0,3.8) 49.0% (1,74) 0.0464
(5) Conrmed non-epidemicb 1.3 (0.7,2.4) 25.2% (35,59) 0.3348
(6) All meningococcalc 1.6 (1.2,2.1) 37.3% (16,53) 0.0018
Full vs. partial
(1) Epidemic group Ba 2.8 (1.6,4.7) 64.0% (39,79) 0.0002
(2) Non-epidemic group B 1.4 (0.8,2.6) 28.4% (31,61) 0.2785
(3) Non group B 2.0 (1.0,4.1) 50.3% (2,76) 0.0580
(4) Unknown group or subtype 1.7 (0.8,3.6) 42.1% (22,73) 0.1496
(5) Conrmed non-epidemicb 1.9 (1.1,3.2) 46.4% (7,69) 0.0256
(6) All meningococcalc 1.9 (1.4,2.6) 48.1% (31,61) <0.0001
a
Estimates for epidemic group B differ slightly from those presented in Table 3 due to a different age range (6 months19 years) and time interval (20022008).
b
Conrmed non-epidemic combines groups (2) and (3).
c
All meningococcal is the sum of groups (1)(4).

an additional stratication. Pneumococcal disease was taken as Consistent signicant protective effects of full vaccination
the reference category, and we estimated the additional protective against pneumococcal disease are seen in each of the three models:
effect of MeNZB on meningococcal cases. The results for epidemic 53.0%, 54.1% and 52.3%. There is evidence for a dose response rela-
strain group B meningococcal disease, non-epidemic Group B, and tionship, with some protection from partial vaccination against
non Group B disease are shown in Table 6. pneumococcal disease, and a signicant difference between the
Table 6
Vaccine effectiveness estimates from Poisson regression models for pneumococcal disease, and combined models of pneumococcal and various groupings of meningococcal
disease cases (ages 6 months19 years in the period 20022008). In the combined models pneumococcal disease is the reference category, for each model the estimates
in the rst three rows are for relative risks of pneumococcal disease between the three MeNZB vaccination categories. The second three (shaded) rows are the additional
protective effect of the vaccine for meningococcal disease.

OutcomeMeNZB vaccination status Relative risk Effectiveness p-Value

Estimate 95% CI Estimate 95% CI

Pneumococcal disease only

Full vs. unvaccinated 2.3 (1.8,2.9) 56.3% (44,66) <0.0001
Partial vs. unvaccinated 1.3 (1.0,1.8) 25.8% (3,43) 0.0297
Full vs. partial 1.7 (1.3,2.2) 41.1% (25,54) <0.0001
Pneumococcal disease and epidemic B
Baseline: pneumococcal
Full 2.1 (1.6,2.8) 53.0% (39,64) <0.0001
Partial 1.3 (1.0,1.7) 23.7% (1,41) 0.0452
Full-partial 1.6 (1.3,2.1) 38.4% (22,51) <0.0001
Contrast: epidemic B vs. pneumococcal
Full 2.1 (1.3,3.4) 53.3% (25,71) 0.0015
Partial 1.7 (1.1,2.6) 40.0% (5,62) 0.0301
Full-partial 1.3 (0.8,2.0) 22.2% (21,50) 0.2648
Pneumococcal disease and non-epidemic group B
Baseline: pneumococcal
Full 2.2 (1.7,2.8) 54.1% (42,64) <0.0001
Partial 1.3 (1.0,1.7) 22.1% (1,4) 0.0594
Full-partial 1.7 (1.3,2.1) 41.0% (26,53) <0.0001
Contrast: non-epidemic B vs. pneumococcal
Full 1.3 (0.8,2.0) 21.2% (23,50) 0.2976
Partial 1.1 (0.6,2.0) 7.8% (71,50) 0.7971
Full-partial 1.2 (0.7,2.0) 14.6% (47,51) 0.5712
Pneumococcal disease and non-group B
Baseline: pneumococcal
Full 2.1 (1.6,2.7) 52.3% (38,63) <0.0001
Partial 1.4 (1.1,1.9) 29.9% (9,46) 0.0074
Full-partial 1.5 (1.2,1.8) 31.9% (16,45) 0.0004
Contrast: non-group B vs. pneumococcal
Full 1.1 (0.6,2.3) 13.0% (76,57) 0.6980
Partial 1.4 (0.7,3.0) 29.3% (50,67) 0.3674
Full-partial 0.8 (0.3,2.0) 23.1% (203,50) 0.6514
 R. Arnold et al. / Vaccine 29 (2011) 71007106 7105

protective effect of partial and full vaccination. Having controlled This would also lead to a reduction in transmission between sib-
for the residual confounding, the additional protective effect of the lings in households where all the children were vaccinated. Since
vaccine against the three groupings of meningococcal disease are meningococcal disease is only symptomatic in its invasive state,
shown in the second three rows for each model. There is a residual and presents rapidly and acutely, disease rates are unlikely to be
protective effect of 53.3% of full vaccination against the epidemic modied by increased contact with a health professional.
strain, with high statistical signicance (95% CI 2571) and weaker If there is a differential effect of the vaccine campaign on the
evidence for an effect of partial vaccination. This can be interpreted separate risks of pneumococcal and meningococcal disease, then
as clear evidence for a protective effect of MeNZB against the Epi- the reduced MeNZB vaccine effectiveness results in Table 6 are sig-
demic strain. nicant underestimates. However quantifying the true degree of
There is however no residual effect of any level of MeNZB vacci- residual confounding is not straightforward. One simple and con-
nation against non-epidemic B and non Group B. This suggests that servative method is to estimate the degree of residual confounding
at least some of the cross-protective effects for these groupings from the partially vaccinated group, since they have had the short-
found in Table 5 may be due to residual confounding. We discuss est exposure to the campaign. From Table 6 their disease rates are
the consequences of these ndings further below. modied by a factor of 1.3, reducing the relative risk for full vac-
cination with MeNZB to 4.1/1.3 = 3.2, and the vaccine effectiveness
estimate from 77% to 68%.
5. Discussion The epidemic was waning before and during the rollout of the
vaccination programme, however the staggered introduction of the
In an observational evaluation of vaccine effectiveness there vaccine enabled year-by-year comparison of rates in vaccinated
remains the possibility of residual confounding, caused by the vol- and unvaccinated populations to estimate the vaccine effective-
untary nature of participation in the vaccination programme. If this ness. Since our model ts separate effects for each year our results
is so, then (part of) the estimated vaccine effectiveness might be due cannot be used to predict the future progress of the epidemic, and
to those vaccinated individuals forming an intrinsically lower risk in particular it cannot be known whether the epidemic would have
group. We tested this using pneumococcal disease because of simi- continued to wane in the absence of the vaccination programme.
larities with meningococcal disease: it is a bacterial infection which We have found some evidence for (lesser) cross-protection against
causes invasive disease; is carried in the respiratory tract; spreads other strains, which is consistent with the ndings of Tappero et al.
through the air and droplets; has high rates in young children; is [16], who observed some cross protection for other strains as mea-
most common in winter; and is identied from invasive sites. sured by SBAs.
There is no biological means for MeNZB to confer true protection With the vaccination programme now concluded there are no
against pneumococcal disease, and therefore our results indicat- further data to add. No additional reference disease apart from
ing protection must derive from a true but unmeasured difference pneumococcal disease has been identied as suitable to repeat tests
between the vaccinated and unvaccinated populations. Two types for residual confounding.
of difference might exist: a pre-existing difference or a difference
which is created by the experience of the vaccination campaign.
6. Conclusion
One possible pre-existing difference may be due to levels of
socio-economic deprivation: where highly deprived children, who
Our ndings provide a further and comprehensive evaluation of
may come from higher risk overcrowded households, may have
the effectiveness of New Zealands MeNZB vaccination programme.
missed school and GP visits during the campaign. We have used
The data suggest that the vaccine is effective against epidemic strain
the NZ index of deprivation (which includes overcrowding) as an
meningococcal B disease, and to a lesser extent against other strains
explanatory variable in all our analyses, and while coverage of
of meningococcal disease. Our test for residual confounding indi-
the most deprived quintiles was found to be high, this index may
cates that the effectiveness is greater than 53% (95% CI 2571), and
not capture the spectrum of deprivation adequately. In sensitivity
may be as high as 77% (95% CI 6285), with 68% being a conservative
analyses where the coverage of the vaccination programme was
estimate.
assumed to decline with increasing deprivation only a minor effect
There was no evidence of a waning vaccine effect after one year.
on the vaccine effectiveness estimate was found, consistent with
There does appear to be some protective effect of partial vaccina-
our ndings in an earlier report [15]. Another pre-existing differ-
tion after 2 doses, although there was no evidence for protection
ence may be that a segment of the unvaccinated population might
after only 1 dose. There was no additional protective effect of a
be a strictly non-compliant group, and that this group is at higher
4th dose, though this may be due to the small numbers of eligi-
risk.
ble recipients. No dependence of vaccine effectiveness on age was
If the unknown confounder is pre-existing, as in these exam-
found.
ples, there should be no difference in pneumococcal disease rates
The New Zealand experience has demonstrated that a robust
between the partially and fully vaccinated groups, since nearly all
measure of vaccine effectiveness can be estimated by taking advan-
(91%) partially vaccinated individuals ultimately join the fully vac-
tage of a vaccine programme with a staggered rollout. Such a
cinated group. However this expectation is in contrast to what we
situation frequently arises where there are limitations imposed by
see: namely a clear dose response for pneumococcal disease in all
vaccine supply, or in the logistics of programme delivery. Reliable
four of the models reported.
population estimates and high quality administrative data on vac-
The dose response relationship suggests the difference between
cinations and cases are vital for the Poisson regression method,
vaccinated and unvaccinated individuals may be created by the
which allows for the inclusion of relevant explanatory factors, and
experience of the vaccination campaign. This is a possible expla-
for tests for residual confounding.
nation since, unlike meningococcal disease, pneumococcal disease
is not necessarily asymptomatic in its non-invasive state. Even
if it is not diagnosed as pneumococcal disease, the symptoms of Acknowledgements
the non-invasive disease (e.g. otitis media) can be identied and
treated. The increased number of contacts with a health profes- The authors are grateful to Diana Martin and staff at Environ-
sional may, through additional attention and treatment, reduce mental and Scientic Research Limited for their provision of the
the risk of pneumococcal disease among those being vaccinated. pneumococcal disease data.
7106 R. Arnold et al. / Vaccine 29 (2011) 71007106

Funding: This work was supported by the New Zealand Min- [12] McNicholas A, Galloway Y, Martin D, Sexton K, OHallahan J. Surveillance of vac-
istry of Health. Novartis Vaccines funded the New Zealand Ministry cine breakthrough cases following MeNZB vaccination. New Zealand Medical
Journal 2008;121(1272):1279.
of Health to establish a Data Management Group to manage and [13] Galloway Y, Stehr-Green P, McNicholas A, OHallahan J. Use of an observa-
analyse data collected on meningococcal B vaccine safety and effec- tional cohort study to estimate the effectiveness of the New Zealand group
tiveness. B meningococcal vaccine in children aged under 5 years. International Journal
of Epidemiology 2009;38:4138.
[14] Ameratunga S, Macmillan A, Stewart J, Scott D, Mulholland K, Crengle
References S. Evaluating the post-licensure effectiveness of a group B meningococ-
cal vaccine in New Zealand: a multi-faceted strategy. Vaccine 2005;23:
[1] Baker MG, Martin DR, Kieft CEM, Lennon D. A 10-year serogroup B meningo- 22314.
coccal disease epidemic in New Zealand: descriptive epidemiology, 19912000. [15] Kelly C, Arnold R, Galloway Y, OHallahan J. A prospective study of the effec-
Journal of Paediatric Child Health 2001;37:S139. tiveness of the New Zealand meningococcal B vaccine. American Journal of
[2] Martin D, Lopez L, Sexton K. The epidemiology of meningococcal disease in Epidemiology 2007;166:81723.
New Zealand in 2006. Ministry of Health; 2007. [16] Tappero JW, Lagos R, Ballesteros AM, Plikaytis B, Williams D, Dykes J, et al.
[3] Ministry of Health. National immunisation schedule. Health provider booklet. Immunogenicity of 2 serogroup B outer-membrane protein meningococcal vac-
Wellington: Ministry of Health; 2008. cines: a randomized controlled trial in Chile. Journal of the American Medical
[4] Feiring B, Fuglesang JE, Oster P, Naess LM, Oddveig S, Helland ST, et al. Persisting Association 1999;281:15207.
immune responses indicating long-term protection after booster dose with [17] Salmond C, Crampton P. The NZDep01 index of deprivation. Wellington, NZ:
meningococcal Group B outer membrane vesicle vaccine. Clinical and Vaccine Wellington School of Medicine and Health Sciences; 2002.
Immunology 2006;13:7906. [18] Statistics New Zealand. National Population Projections; 2010.
[5] Holst J, Feiring B, Fuglesang JE, Hoiby EA, Nokleby H, Aarberg IS, et al. Serum http://www.stats.govt.nz.
bactericidal correlates with the vaccine efcacy of membrane vesicle vaccines [19] Hardin JW, Hilbe JM. Generalized estimating equations. Boca Raton, FL: Chap-
against Neisseria Meningitidis serogroup B disease. Vaccine 2003;21:7347. man & Hall/CRC; 2003, 222 pp.
[6] Oster P, OHallahan J, Aberge O, Tilman S, Ypma E, Martin D. Immunogenicity [20] Baraff LJ, Lee SI, Shriger DL. Outcome of bacterial meningitis: a meta-analysis.
and safety of a strain-specic MenB OMV vaccine delivered to under 5-year Pediatric Infectious Disease Journal 1993;12:38994.
olds in New Zealand. Vaccine 2007;25:30759. [21] Barquet N, Domingo P, Cayla JA, Gonzalez J, Rodrigo C, Fernandez-
[7] Perkins BA, Jonsdottir K, Briem H, Grifths E, Plikaytis BD, Hoiby EA, et al. Viladrich P, et al. Meningococcal disease in a large urban population
Immunogenicity of two efcacious outer membrane protein-based serogroup (Barcelona, 19871992). Archives of Internal Medicine 1999;159:2329
B meningococcal vaccines among young adults in Iceland. Journal of Infectious 40.
Diseases 1998;177:68391. [22] Bohr V, Hansen B, Kjersem H, Rasmussen N, Johnsen N, Kristensen HS, et al.
[8] Wong S, Lennon D, Jackson C, Stewart J, Reid S, Ypma E, et al. Immunogenicity Sequelae from bacterial meningitis and their relation to the clinical condition
and tolerability in infants of a New Zealand epidemic strain meningococ- during acute illness, base on 667 questionnaire returns. Part II of a three part
cal B outer membrane vesicle vaccine. Pediatric Infectious Disease Journal series. Journal of Infection 1983;7:10210.
2009;28:38590. [23] Erickson L, Wals PD. Complications and sequelae of meningococcal disease in
[9] American Academy of Pediatrics Committee on Infectious Diseases (2005) Quebec, Canada, 19901994. Clinical Infectious Diseases 1998;26:115964.
Policy Statement. Prevention and control of meningococcal disease: recom- [24] Healy CM, Butler KM, Smith EOB, Hensey OP, Bate T, Moloney AC. Inuence of
mendations for use of meningocccal vaccines in pediatric patients. Pediatrics serogroup on the presentation, course, and outcome of invasive meningococcal
116:496505. disease. Clinical Infectious Diseases 2002;34:132330.
[10] Patel MM, Tate J, Cortese M, Payne DC, Armstrong G, Parashar UD, et al. The [25] Kirsch EA, Barton RP, Kitchen L, Giroir B. Pathophysiology, treatment and out-
impact of indirect benets of vaccination on postlicensure vaccine effectiveness come of menigococcemia: a review and recent experience. Pediatric Infectious
estimates: a scenario analysis. Vaccine 2010;28:798792. Disease Journal 1996;15:96779.
[11] Ramsay ME, Andrews N, Kaczmarski EB, Miller E. Efcacy of meningococcal [26] Wilson N, Baker M, Martin D, Lennon D, OHallahan J, Jones N, et al. Meningo-
serogroup C conjugate vaccine in teenagers and toddlers in England. The Lancet coccal disease epidemiology and control in New Zealand. New Zealand Medical
2001;357:1956. Journal 1995;108:43742.
Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.