Evaluating the neonatal BCG vaccination programme in Ireland

Vaccine efficacy

Several systematic reviews to date conclude that BCG vaccination of infants is very
effective in preventing miliary TB and TB meningitis in children 10]. One meta-analysis conducted in the USA 11] provided evidence for a protective effect of the BCG vaccine against pulmonary TB,
TB deaths, TB meningitis, laboratory confirmed TB cases and disseminated TB. In 1995
Colditz and colleagues (13) reviewed the results of 5 randomised control trials and
11 case control studies. They estimated the protective effect of the BCG vaccine for
preventing pulmonary TB, TB deaths, TB meningitis, laboratory confirmed TB cases and
disseminated TB. Trunz and colleagues 12] re-evaluated BCG efficacy against childhood TB meningitis and miliary TB by adding
seven more published investigations to earlier meta-analyses of published case-control
studies 10], 11]. A total of 18 case-control studies provided revised estimates of efficacy for TB
meningitis and miliary TB, which are similar to earlier published estimates 10], 11], 13].

In Ireland, the Health Protection Surveillance Centre (HPSC) documents all cases of
notified TB. In addition to documenting cases, other information such as age, gender,
local health office/county, country of birth, BCG vaccination status, presence of
BCG scar, and treatment outcome is recorded. It therefore serves as a reliable source
to estimate the effectiveness of the BCG vaccination in the Irish population. To this
end, a retrospective evaluation of TB cases, stratified according to BCG status from
the years 2002 to 2011 was conducted. The analysis was fitted in R (version 2.15.2)
to combine this Irish data with data from Colditz 11] and Trunz 12]. Resulting risk ratios provide estimates of vaccine efficacy for the Irish population
aged 0-15 years strengthened by international estimates. A forest plot of the results
is shown (Fig. 1), illustrating the impact different data sources have on the revised combined estimates.

Fig. 1. Estimates of vaccine efficacy from meta-analysis and combined results used in the
model. Shown are risk ratios and 95 % confidence intervals. The size of the red squares
indicates the relative influence of the trial on the combined estimate; trials of
large sample size will have a higher impact compared to trials of smaller sample size.
The black diamond illustrates the combined estimate for each outcome; its width is
a measure of the associated uncertainty

Vaccine uptake

Not all local health offices report their BCG vaccine uptake rates to the HPSC. The
mode of delivery also varies across the country with approximately 75 % of all BCG
vaccinations being performed in local health offices and the remainder (25 %) being
done in maternity hospitals (National Immunisation Office (NIO), personal communication).
For the purposes of the current study, it was assumed that the vaccine uptake in the
base case analysis would be approximately 80 % for the universal strategy and was
varied by 75–95 % in the sensitivity analysis. These figures were based on expert
clinical opinion (from the NIO) and the latest HPSC immunization statistics 14].

For the selective strategy, it was assumed that the vaccine uptake in the base case
analysis would be approximately 44 % and was varied by 38–50 % in the sensitivity
analysis. These figures were based on experience from France 15], where, following a change in their vaccination schedule to a selective based strategy,
reported uptake rates of lower magnitude than those experienced for the universal
strategy. Experience from other low incidence countries, showed similar evidence for
a reduced uptake rate for a selective based strategy, e.g., Sweden reported a vaccine
uptake rate of less than 2 % in 1976 a year after changing to a selective based strategy
5].

Incidence of childhood TB

In the southern area of Ireland, a universal neonatal BCG vaccination was offered
in Cork up to 1972 when a decision was taken to discontinue the vaccine for a variety
of reasons. This small cohort, together with the information provided by the HPSC,
therefore provided a reliable source to estimate the incidence of childhood TB in
a non-vaccinated Irish population. Annualized incidence rates were calculated for
each TB health state, for each year between 0 and 15 years, and these were used as
the basis to build the model. In a separate scenario analysis, which examined the
effect of the vaccine up to 25 years, annualized incidence rates from those aged 16–25 years
were used (Table 1).

Table 1. Annualised incidence rate of TB clinical syndromes and mortality by age in a non-vaccinated
cohort of Irish children 2002-2010

Mortality data

There were no recorded deaths due to TB in the 0–15 year old population, so a case
fatality rate (CFR) of 0.8 % (based on UK Surveillance data from 2001 to 2010, compiled
by the Health Protection Agency, UK 16]) was applied.

Long term complications of TB meningitis

There is a wide range of possible long-term sequelae following meningitis and the
analysis is restricted to four main complications: hearing loss, developmental delay,
focal neurological deficits andepilepsy. The probability of these events occurring
was derived from published studies.

In 1964, Todd and Neville 17] estimated the incidence of hearing loss and epilepsy attacks. They combined the results
of their own study with those of Wasz-Hockert Donner 18], Lorber 19], Voljavec et al. 20], Pohitonova 21] and Lapides 22]. From these 6 studies they combined information on the incidence of sequelae in 855
children who had survived an attack of meningeal TB. The estimates for hearing loss
and epilepsy attacks were 7 and 7.6 % respectively. Lorber 19] estimated the incidence of cognitive impairment at 50 %. Schoeman 23] estimated that focal neurological deficits occurred in 25 % of cases analysed. The
probabilities of long-term complications of meningitis were varied by +/?20 % in the
sensitivity analysis.

Resource use data

Cost estimates

Cost estimates are explained in terms of healthcare resource use (i.e., vaccination
costs and direct medical costs) and unit cost data.

Vaccination costs

Vaccination cost estimates were obtained from the NIO of the HSE. The cost of vaccination
includes the cost of the vaccine as well as the cost of administration of the vaccine.
The vaccine ingredient costs are calculated as 80 or 50 % of the cost of distributed
vaccine in 2012 for the universal and selective strategies respectively (Personal
communication with NIO). The cost of administration is based on an administration
fee of €27.75 and €50 per infant, for the universal strategy and selective strategies
respectively. The higher fee for the selective strategy reflects the difficulty that
may be present in identifying those at high risk. Changes to the administration fee
are explored in the sensitivity analysis.

Direct medical costs

Due to the lack of published Irish cost data, and the time constraints with conducting
specific micro-costing studies, an assessment of resource use items associated with
diagnosis and treatment of each of the health states (pulmonary TB, extrapulmonary
TB and TB meningitis) were obtained by the evaluation team with local clinical experts.
The costs associated with contact tracing per primary case of TB and diagnosing and
treating 50 % of those contacts for latent TB were added to the direct medical costs
of pulmonary TB, extrapulmonary TB and TB meningitis. Costs of contact tracing and
latent TB are based on 9.4 contacts per TB case and are shown in Table 2. The costs of managing the long-term sequelae of meningitis were also included and
the overall cost shown in Table 2. A comprehensive list of resource utilisation and unit cost data used to estimate
direct medical costs of latent TB, pulmonary TB, extrapulmonary TB, TB meningitis
and long term sequelae of meningitis are provided in Additional files 1, 2, 3, 4 and 5 respectively.

Table 2. Summary of parameter estimates, with base case values, range, distributions and sources

Unit cost data

Unit costs for inpatient procedures were obtained from 2010 Diagnosis Related Group
(DRG) data provided by the National Casemix Unit of the HSE 24]. Unit costs for tuberculin skin tests were obtained from the Finance department of
a local university teaching hospital (St. James’s Hospital, Dublin). Costs of antimicrobial
therapy were obtained from the Pharmacy department of St. James’s Hospital Dublin.
Every effort was made to incorporate Irish unit cost data. However, where data were
not available it was adapted from the UK. UK costs were converted to euro using the
exchange rate (€1.18) published by the Central Bank of Ireland and all costs were
inflated to 2012 euro, using the consumer price index for health inflationary unit
(105.6).

Adverse events

The model uses the incidence rate of severe adverse events (1 in 1000) reported in
the summary of product characteristics for the BCG vaccine 25]. Following review of all local reports 26]–33] as well as international reports 34]–36] published in the literature, none of the reports indicate that the rate is higher
than this. The estimate of 1 in 1000 is applied only in the first year for the cohort
as the literature indicates that the median onset of adverse effects ranges from 30 days
to 4 months 32]–35]. This rate is varied in the sensitivity analysis from ?1/10,000 to ?1/100 which
is within the rates reported in the summary of product characteristics 25].

Time horizon

The analytic time frame of the study was fifteen years in the base case analysis given
the current paucity of data on the duration of vaccine protection beyond this time
37].

Outcome measure

The main outcome measure in the analysis was cost per LYG.

Discounting

An annual discount rate of 5 % was applied to both costs and consequences in the economic
model, consistent with Guidelines for the Conduct of Economic Evaluations in Ireland
38].

Sensitivity analyses

One-way

One-way sensitivity analysis was undertaken for the base case analysis. Ranges for
the parameter values used are shown in Table 2.

Probabilistic sensitivity analysis

A probabilistic sensitivity analysis was also conducted whereby all parameters were
varied simultaneously. One thousand simulated combinations of the parameters were
drawn and for each of these combinations a cost per LYG was estimated.

Scenario analyses

Long term effect of vaccination

A scenario analysis was conducted which examined the effect of vaccination up to
25 years. To do this, incidence data was gathered from those aged 16 to 25 years from
non-vaccinated and vaccinated cohorts respectively and rates of vaccine efficacy against
the TB health states (Protective effect: 0.55 95 % CI 0.31 to 0.77) and against TB
death (Protective effect: 0.44 95 % CI?0.22 to 0.70) were applied 39].

Vaccinating a birth cohort-the societal perspective

A scenario analysis was conducted whereby the indirect costs due to TB were included
in the analysis. This analysis included the value of work lost per week 40] for a parent of a child with TB or an adverse event due to vaccination. This was
based on the work of Roberts et al., 41], which stated that 42 % of parents would take time off work and an assumption that
the number of work days lost is four.