Cost-effectiveness analysis of a universal mass vaccination program with a PHiD-CV 2+1 schedule in Malaysia

Background Currently, two pediatric pneumococcal conjugate vaccines are available in the private market of Malaysia—13-valent pneumococcal conjugate vaccine (PCV13) and pneumococcal polysaccharide and non-typeable Haemophilus influenzae protein D conjugate vaccine (PHiD-CV). This study aimed to evaluate the cost-effectiveness of a universal mass vaccination program with a PHiD-CV 2+1 schedule versus no vaccination or with a PCV13 2+1 schedule in Malaysia. Methods A published Markov cohort model was adapted to evaluate the epidemiological and economic consequences of programs with no vaccination, a PHiD-CV 2+1 schedule or a PCV13 2+1 schedule over a 10-year time horizon. Disease cases, deaths, direct medical costs, quality-adjusted life-years (QALYs) and incremental cost-effectiveness ratios (ICERs) were estimated. Locally published epidemiology and cost data were used whenever possible. Vaccine effectiveness and disutility data were based on the best available published data. All data inputs and assumptions were validated by local clinical and health economics experts. Analyses were conducted from the perspective of the Malaysian government for a birth cohort of 508,774. Costs and QALYs were discounted at 3% per annum. One-way and probabilistic sensitivity analyses were performed. Results Compared with no vaccination, a PHiD-CV 2+1 program was projected to prevent 1109 invasive pneumococcal disease (IPD), 24,679 pneumonia and 72,940 acute otitis media (AOM) cases and 103 IPD/pneumonia deaths over 10 years, with additional costs and QALYs of United States dollars (USD) 30.9 million and 1084 QALYs, respectively, at an ICER of USD 28,497/QALY. Compared with a PCV13 2+1 program, PHiD-CV 2+1 was projected to result in similar reductions in IPD cases (40 cases more) but significantly fewer AOM cases (30,001 cases less), with cost savings and additional QALYs gained of USD 5.2 million and 116 QALYs, respectively, demonstrating dominance over PCV13. Results were robust to variations in one-way and probabilistic sensitivity analyses. Conclusions A PHiD-CV 2+1 universal mass vaccination program could substantially reduce pneumococcal disease burden versus no vaccination, and was expected to be cost-effective in Malaysia. A PHiD-CV 2+1 program was also expected to be a dominant choice over a PCV13 2+1 program in Malaysia. Electronic supplementary material The online version of this article (doi:10.1186/s12962-017-0079-2) contains supplementary material, which is available to authorized users.


Background
Streptococcus pneumoniae can result in a range of diseases, from invasive pneumococcal diseases (IPDs) (e.g. meningitis and bacteremia) to pneumonia and acute otitis media (AOM) [1]. Data from the World Health Organization (WHO) suggest that pneumococcal diseases caused around 480,000 deaths among children under 5 years of age in 2008 [2], making it the leading cause of vaccine-preventable death among children globally [3]. In addition, pneumonia has been reported as the fourth leading cause of death in 2009 among the Ministry of Health hospitals in Malaysia, accounting for 10.4%

Open Access
Cost Effectiveness and Resource Allocation *Correspondence: sharon.x.zhang@gsk.com 2 GSK, 150 Beach Road, #22-00 Gateway West, Singapore 189720, Singapore Full list of author information is available at the end of the article of inpatient deaths [4]. Although AOM is a much milder disease than pneumonia or IPDs, it is very common and is therefore associated with substantial impacts on healthcare costs and quality of life [5,6]. Invasive disease, pneumonia and AOM can also be caused by microorganisms other than S. pneumoniae, including non-typeable Haemophilus influenzae (NTHi) [7,8].
Pneumococcal conjugate vaccines (PCVs) were recommended as a priority for inclusion in national childhood immunization programs in all countries by the WHO in 2007 [9]. Although they are not currently included in the national immunization program in Malaysia, two PCVs are available in the private market of Malaysia: a 13-valent pneumococcal conjugate vaccine (PCV13; Prevenar 13) and a pneumococcal polysaccharide and NTHi protein D conjugate vaccine (PHiD-CV; Synflorix).
Two prior cost-effectiveness analyses have compared a 3+1 schedule using either PHiD-CV or PCV13 in Malaysia [10,11], but the results are inconsistent. Aljunid et al. [10] predicted that PHiD-CV would be more cost-effective than PCV13, but Wu et al. [11] predicted the opposite. Unlike Aljunid et al. [10], Wu et al. [11] did not account for the protective effects of PHiD-CV in a number of key areas: (1) cross-protection against serotypes 6A and 19A, which has been demonstrated in a number of recent studies [12][13][14]; (2) indirect (herd) protection against IPD, which has been shown in surveillance data from Finland [15] and New Zealand [16]; and (3) protection against NTHi AOM, which has been shown in both the randomized controlled Clinical Otitis Media and Pneumonia Study (COMPAS) study [17] and the randomized controlled Pneumococcal Otitis Efficacy Trial (POET) study of PHiD-CV's 11-valent precursor [18]. Wu et al. 's [11] methodology has recently been critiqued by Varghese et al. [19]. Excluding the protective effects of PHiD-CV in these key areas is against the current evidence and could have had a significant impact on the costeffectiveness of PHiD-CV versus PCV13. It is therefore necessary to conduct another cost-effectiveness analysis, taking into account the latest evidence of vaccine effectiveness.
The goal of this economic evaluation was to assess the cost-effectiveness of universal mass pneumococcal vaccination with a PHiD-CV 2+1 vaccination strategy versus no vaccination or a PCV13 2+1 vaccination strategy from the perspective of the Malaysian government. The results from this study can provide scientific evidence for Malaysian healthcare policymakers to support their decision making on the introduction of PCV into the national immunization program.

Methods
A published Markov cohort model [20] was adapted to simulate the epidemiological and economic burden of pneumococcal and NTHi-related diseases (IPD, pneumonia and AOM) over a 10-year time horizon in Malaysia. In this study, infants could be vaccinated with PHiD-CV 2+1 or PCV13 2+1 or neither, as these are the vaccinations being considered for inclusion in the Malaysian universal mass vaccination program. The model has a number of mutually exclusive disease-related outcomes, namely pneumococcal meningitis, pneumococcal bacteremia, all-cause pneumonia, AOM and no pneumococcal infection (Fig. 1). Patients with AOM could be hospitalized (myringotomy), visit their general practitioner (GP) or be non-consulting. Patients with pneumonia could be treated as outpatients or inpatients. Those with meningitis, bacteremia or hospitalized pneumonia were at risk of death. Individuals of the birth cohort moved between the Markov states according to estimated transition probabilities. The reason for choosing 10 years' time horizon in the base case is because we assumed that the duration of vaccine protection to be 10 years for the 2+1 regimen. In addition, it is because the serotypes are changing over time due to the vaccination program [21].
The analyses were conducted from the perspective of the Malaysian government for a birth cohort of 508,774 in 2012 [22]. Costs and quality-adjusted life-years (QALYs) were discounted at 3% per annum according to Malaysian guidelines [23]. All costs are reported in 2014 United States dollars (USD), converted when necessary from Malaysian Ringgits (MYR) using the exchange rate as at 31 Dec 2014 (3.5 MYR = 1 USD) [24]. The model and all-related data inputs and assumptions were reviewed and validated by a group of local clinical and health economics experts to ensure the validity of the model adaptation in the local setting.
Epidemiological data (Table 1) Age-specific incidence data for hospitalized pneumococcal meningitis and bacteremia (Additional file 1: Table  S1) and hospitalized all-cause pneumonia (Additional file 1: Table S2) were sourced from a study that estimated disease burden from the hospital records of six tertiary hospitals across Malaysia during 2006-2007 [25]. Malaysian case fatality ratios (CFRs) were not available, so we referred to the CFRs from a retrospective populationbased National Health Insurance Reimbursement Database study from Taiwan [26] (Additional file 1: Tables  S1 and S2). This approach was conservative, considering Taiwan has a more advanced healthcare system and better healthcare access than Malaysia. GP consultation rates for all-cause pneumonia (Additional file 1: Table S2) were also lacking in Malaysia, but hospitalized pneumonia incidence rates were available. Therefore, an estimate was calculated, based on Malaysian hospitalization rates [25] and the ratio of Taiwanese GP consultation to hospitalization rates [26], considering similar healthcare access level within a country. Due to a lack of local/regional data on the long-term sequelae of IPD, these were not included.
Age-specific AOM GP consultation rates (Additional file 1: Table S3) were based on published data from the Philippines [27] due to the lack of Malaysian data. Myringotomy rates (Additional file 1: Table S3) were adapted  from data from the Philippines [27] that were based on insurance data from Taiwan [26] and consultation with local ear, nose and throat experts. In the base case, all cases of AOM were assumed to result in a GP consultation. Complications and long-term sequelae of AOM were very conservatively not taken into account, due to a lack of suitable data.
Vaccine effectiveness data ( Table 2) Vaccination was assumed to occur at ages 2, 4 and 13 months. In the base case, it was assumed that 100% of infants would be vaccinated, as WHO 2015 immunization coverage estimates for Malaysia for other childhood vaccines were close to 100% [28]. It was further assumed that all children would receive the defined number of doses at the recommended times, hence the effect of partial vaccination was not considered. Both vaccines were assumed to cost USD 34.25 per dose (hypothetical price). Due to a lack of local data, administration costs were not considered; and vaccine wastage was assumed to be 0%. Vaccine effectiveness was assumed to have a ramp-up increase from age 2-13 months (50% after dose 1; 90% after dose 2), have full effectiveness from age 13 months to 3 years, and have waning effectiveness (exponential decline) to age 10 years [20].
Vaccine effectiveness against all-cause pneumonia is lower than against IPD as it can be caused by a number of pathogens. VE against all-cause pneumonia has been reported in various trials, with PCV7 (18% [95% CI 5-29%] [ [17]) and PCV11 (16% [95% CI -7 to 34%] [40]) (all intention-to-treat analyses), with no relationship between valency and VE. As there are no published randomized controlled studies for PCV13 against all-cause pneumonia, we used the value of 21.8% for consolidated pneumonia from COMPAS (PHiD-CV) [17] as the VE against pneumonia hospitalizations for both vaccines. VE against all-cause suspected pneumonia (8.7%) was taken from the same study [17] and was used for pneumonia associated with a GP visit.
Overall vaccine effectiveness against AOM was estimated based on causative pathogens (S. pneumoniae, H. influenzae or other) and VE against vaccine and nonvaccine S. pneumoniae serotypes and NTHi. Based on a review paper by Leibovitz et al. [41] of 17 AOM etiology studies across the world, we assumed that 35.9% of AOM cases were attributable to S. pneumoniae and 32.3% to NTHi. Data from a multinational AOM study were used for the percentages of AOM cases caused by each serotype [42]. VE against vaccine-type S. pneumoniae (excluding serotype 3) for both vaccines was taken to be 69.9% based on COMPAS [17]. For PHiD-CV, vaccine efficacies of 63.7% [18] and 61% [14,17,31] for serotypes 6A and 19A, respectively, were used. For both vaccines, a negative VE of −33% was used against other nonvaccine serotypes based on PCV7 data [43]. This study also gave a VE for PCV7 against H. influenzae AOM of −11% [43], which was used for PCV13. Two studies have demonstrated efficacy of PHiD-CV [17] (or its 11-valent precursor [18]) against NTHi AOM. The lower value of 21.5% (95% CI−43.4 to 57.0%) from COMPAS [17] was used. Although the 95% CI spans zero, it should be noted that the COMPAS trial [17] was not powered to provide conclusive evidence of protection against NTHi AOM. However, the positive effect of PHiD-CV against NTHi AOM is consistent with the significant efficacy observed with the 11-valent predecessor protein D conjugate formulation used in the POET study (35.3% [95% CI 1.8-57.4%]) [18]. The overall estimated vaccine effectivenesses against AOM were 24.1% for PHiD-CV and 14.2% for PCV-13.
No vaccine effectiveness for PHiD-CV against NTHi invasive disease or pneumonia was assumed due to a lack of evidence. Indirect effects (herd protection and serotype replacement) were also not taken into account in the base case.

Health outcomes (Table 3)
Due to a lack of published pneumococcal disease-related disutility weights in Malaysia, short-term disutility data  [44] and Canada [45] were applied to agespecific healthy utilities [46].

Treatment costs (Table 4)
Only direct medical costs were included. Direct medical costs for acute episodes were based on local published data from 2010 [10]. Costs were inflated using the consumer price index for heath for Malaysia [47] to 2014 values and then converted into USD [24].

Cost-effectiveness analysis
The

Sensitivity analyses
For comparisons of PHiD-CV 2+1 versus no vaccination or PCV13 2+1, extensive one-way sensitivity analyses were performed to evaluate the robustness of the results. These were mainly performed using values based on the upper and lower limits of 95% CIs for vaccine effectiveness and ±20 or ±50% of base-case values for most other parameters (see Tables 1, 2, 3, 4 for further details). Probabilistic sensitivity analyses (PSAs) were performed for PHiD-CV 2+1 versus no vaccination or PCV13 2+1, each using 1000 simulations.

Scenario analyses
A set of alternative scenario analyses on parameters of particular interest were performed for comparisons of PHiD-CV 2+1 versus no vaccination or PCV13 2+1: (1) discount rates of 0 and 5, as per Malaysian guidelines [23]; (2) a lower proportion of AOM cases due to NTHi (20% rather than 32.3%); (3) a different adjustment factor for estimating total AOM cases (i.e. consulting plus nonconsulting cases) (0.7 and 1.3 rather than 1.0); (4) a time horizon of 100 rather than 10 years; (5) inclusion of indirect effects (herd protection and serotype replacement) for all age groups and both vaccines (only applied to IPD [50]); it was assumed that this would reach a steady state of 30% (i.e. 30% reduction in disease incidence).

PHiD-CV 2+1 versus no vaccination
It was projected that vaccination with a PHiD-CV 2+1 program would prevent 1109 cases of IPD, 24,679 cases of all-cause pneumonia, 72,940 cases of AOM and 103 IPD/pneumonia deaths compared with no vaccination strategy for the birth cohort of 508,774 in Malaysia over 10 years (Table 5).
Undiscounted vaccination costs were estimated to be USD 52.1 million, but these would be partially offset by a reduction in direct medical costs of USD 22.4 million, for a total direct medical cost of USD 29.7 million ( Table 6). Discounted total cost and QALY increases were predicted to be USD 30.9 million and 1084 QALYs, respectively, demonstrating a cost-effective ICER of USD 28,497/QALY.

PHiD-CV 2+1 versus PCV13 2+1
It was projected that vaccination with a PHiD-CV 2+1 program would result in 40 more IPD cases than vaccination with a PCV13 2+1 program, but 30,001 fewer cases of AOM (Table 5). Similarly, undiscounted direct medical costs with PHiD-CV 2+1 were predicted to be somewhat higher for IPD, but substantially lower for AOM, resulting in an overall cost saving of USD 5.7 million (Table 6). Discounted total cost savings and QALYs gained were predicted to be 5.2 million and 116, respectively, meaning that PHiD-CV 2+1 was predicted to be dominant over PCV13 2+1 (lower cost and more QALYs gained).

Sensitivity analyses PHiD-CV 2+1 versus no vaccination
According to one-way sensitivity analyses, the most influential factors on the cost-effectiveness of PHiD-CV 2+1 versus no vaccination were the efficacy of PHiD-CV against NTHi AOM, the AOM GP consultation rate, and the incidence of hospitalized pneumococcal meningitis ( Fig. 2; Additional file 1: Table S4). In the PSA, the probability of PHiD-CV being cost-effective compared with no vaccination was 31.6% at the cost-effectiveness threshold for Malaysia (Fig. 3a).

Scenario analyses
Most of the scenario analyses predicted that PHiD-CV 2+1 would be cost-effective versus no vaccination; running the model over 100 years resulted in a prediction that PHiD-CV 2+1 would be highly cost-effective (Table 7). Scenario analyses predicted PHiD-CV 2+1 to be dominant over PCV13 2+1 for all scenarios tested.

Discussion
In this cost-effectiveness analysis, the adoption of pediatric mass vaccination with a PHiD-CV 2+1 program was predicted to prevent 1109 cases of IPD, 24,679 cases of pneumonia, 72,940 cases of AOM, and 103 IPD/pneumonia-related deaths over 10 years compared with no vaccination program in Malaysia. PHiD-CV 2+1 was predicted to result in 1084 QALYs gained at a cost of USD 30.9 million (discounted), resulting in a cost-effective ICER of USD 28,497/QALY. Our results are in line with various other health economic studies that have predicted that the introduction of routine infant vaccination with PHiD-CV would be cost-effective, including those in Georgia [51], Latin America [52][53][54][55] and Kenya [56]. Compared with a PCV13 2+1 program, a PHiD-CV 2+1 program was predicted to result in 40 more IPD cases, but 30,001 fewer cases of AOM. This was expected to result in more QALYs gained at a lower overall cost, i.e. PHiD-CV 2+1 was predicted to be dominant over PCV13 2+1. This dominance is in line with various other costeffectiveness analyses, from Malaysia [10], Japan [57], the Philippines [27], Turkey [58], Europe [20,[59][60][61], Canada [20] and Peru [53]. However, some studies have predicted the reverse: that PCV13 would be dominant over PHiD-CV (Malaysia and Hong Kong [11], Colombia [62], Europe [63] and Canada [64]) or more cost-effective than PHiD-CV (Peru [65]). These differences in predicted outcomes are largely due to the assumptions used. As noted in a recently published critical assessment of economic evaluations involving PHiD-CV and PCV13, "the pivotal assumptions and results of these analyses also depended on which manufacturer sponsored the study" [66].
We will focus our discussion on the key differences in assumptions of a prior Malaysian study by Wu et al. [11] (PCV13 dominance) compared with those used in the current study (PHiD-CV dominance): (1) inclusion of herd effects for PCV13 but not PHiD-CV; (2) no crossprotection for PHiD-CV against serotypes 6A and 19A; (3) higher effectiveness of PCV13 against all-cause pneumonia; and 4) no impact of PHiD-CV on NTHi AOM. In addition, recent evidences has shown lower VE for PCV13 against serotype 3 IPD [32 -35], which was not reflected in the previous study by Wu et al. [11].
In the current study, we conservatively chose not to include herd effect for PHiD-CV or PCV13 in the base case, as this would be counteracted by serotype replacement, resulting in an inconclusive overall effect. In one of the scenario analyses, we included an overall beneficial indirect effect. The same net value was used for PHiD-CV and PCV13, as expert advice was that both herd effect and serotype replacement would likely be slightly higher for PCV13, giving a similar net effect. However, it should be noted that inclusion of indirect effects for PCV13, but not for PHiD-CV is considered as unrealistic, given that a population-based observational study of PHiD-CV in Finland [13] reported a 48% (95% CI 18-69%) reduction in IPD among unvaccinated children aged 2-5 years. Additionally, surveillance data have demonstrated herd effects of PHiD-CV following the introduction of childhood vaccination programs in Finland [15] and New Zealand [16].
In the current study, cross-protection of PHiD-CV against non-vaccine serotypes 6A and 19A was accounted for, while no cross-protection was assumed for PHiD-CV against non-vaccine serotypes in the prior Malaysian study [11]. For serotype 19A, this was based on a number of robustly designed studies [12][13][14]. The available evidences have recently prompted the European Medicines Agency to include protection against 19A IPD in PHiD-CV's label [67]. A similar label update was approved in Canada [68] and is underway in many countries. Although the evidence for cross-protection against serotype 6A is less conclusive, the data generally indicate that cross-protection is likely [13,14] and would be similar to that seen with PCV7 [43,69].
We assumed that PHiD-CV and PCV13 would have equal effectivenesses against all-cause pneumonia, based on a number of studies of different valent vaccines with overlapping 95% CIs [17,[37][38][39][40]. It is inappropriate to assume that vaccine effectiveness against all-cause pneumonia would be related to the number of serotypes in the vaccine. This type of serotype-based approach has recently been criticized by Hausdorff et al. [70], because currently available pneumonia efficacy data provide no indication of a trend for greater protection with higher valence vaccines.
Lastly, cautions should be made when assuming that PHiD-CV had no effect against NTHi AOM, merely based on a few studies in which PHiD-CV was shown to have no effect on nasopharyngeal colonization [71][72][73] without looking into other available evidences. Several randomized controlled studies have reported a beneficial effect of PHiD-CV (or its 11-valent precursor) against NTHi AOM [17,18]; and Prymula et al. [18] also reported a 41.4% (95% CI −4.9 to 67.3%) VE against nasopharyngeal carriage. Furthermore, Australian studies have reported less NTHi-infected middle-ear discharge among those with AOM with perforation or chronic suppurative otitis media vaccinated with PHiD-CV versus PCV7 (35% vs.53% of ear discharge swabs; p = 0.03) [74] and with PHiD-CV versus PCV13 (36% vs 64% of swabs; p = 0.05) [75]. Also, a randomized controlled PCV7 trial [43] showed that the number of H. influenzae AOM cases was increased in the PCV7 group. In the absence of PCV13-specific data, we assumed that PCV13 would also increase NTHi AOM cases (vaccine efficacy −11%).
Guidelines from the International Society for Pharmacoeconomics and Outcomes Research recommend that all evidence-not selected sources-should be incorporated into health economic studies [76]. By incorporating the wider body of evidence available around these critical parameters and employing robust sensitivity analyses, we believe that we have addressed the associated uncertainties and attempted to present a more balanced result.  As mentioned previously, outcomes of health economic studies of PHiD-CV and PCV13 can be biased, depending on the sponsor [66]. Therefore, decision makers should be encouraged to rigorously evaluate the underlying assumptions in all cost-effectiveness analyses.

Strengths and limitations
As discussed above, the results of cost-effectiveness analyses very much depend on the input parameters used. In the absence of a head-to-head PHiD-CV versus PCV13 trial, we based effectiveness estimates on serotype distribution and data from various clinical trials. Additionally, there are also no randomized controlled PCV13 efficacy studies, hence we extrapolated effectiveness (based on serotype distribution and PCV7 vaccine serotype VE) or assumed values to be the same as for PHiD-CV. We note that VE data were taken from studies of 3+1 schedules, whereas the current model used a 2+1 schedule. This approach was based on the results of a PHiD-CV trial that reported similar vaccine effectivenesses of 100% (95% CI 83-100%) for a 3+1 schedule and 92% (95% CI 58-100%) for a 2+1 schedule  against vaccine-type IPD [29]. Therefore, it is likely that we slightly overestimated the effectiveness of a PHiD-CV 2+1 schedule. However, the same approach was taken for PCV13, so this would have had little effect on the between-vaccine comparison.
Although Malaysian data were used where possible, this was not always feasible, so some data from other countries (Asian when available) had to be used; but this was approved by local clinical experts. However, for some inputs, it was not possible to find any suitable sources, so administration costs, other vaccine program-related costs (e.g. capital costs, logistics costs) and vaccine wastage were all assumed to be zero. Although this could overestimate the cost-effectiveness of PHiD-CV versus no vaccination, there would be no impact on the between-vaccine comparison. We also assumed that all children would receive the recommended number of doses, but in reality, it is likely that some children would not receive the full schedule. This would have resulted in a slightly lower efficacy but also slightly lower costs.
Lastly, the cost-effectiveness thresholds that we used were those recommended by the WHO at the time of the study [48]. These have recently been criticized [77], but in the absence of new recommendations, we have used the old thresholds.

Conclusions
In this cost-effectiveness analysis, a PHiD-CV 2+1 universal vaccination program could potentially prevent a substantial number of cases of pneumococcal diseases compared with no vaccination, and was projected to be a cost-effective strategy in Malaysia. A PHiD-CV 2+1 vaccination program was also predicted to be dominant (more QALYs gained at a reduced overall cost) over a PCV13 2+1 strategy.