Assessment of potential public health impact of a quadrivalent inactivated influenza vaccine in Thailand

Background Each year, an influenza B strain representing only one influenza B lineage is included in the trivalent inactivated influenza vaccine (IIV3); a mismatch between the selected lineage and circulating viruses can result in suboptimal vaccine effectiveness. We modeled the added potential public health impact of a quadrivalent inactivated influenza vaccine (IIV4) that includes strains from both influenza B lineages compared to IIV3 on influenza‐associated morbidity and mortality in Thailand. Methods Using data on the incidence of influenza‐associated hospitalizations and deaths, vaccine effectiveness, and vaccine coverage from the 2007–2012 influenza seasons in Thailand, we estimated rates of influenza‐associated outcomes that might be averted using IIV4 instead of IIV3. We then applied these rates to national population estimates to calculate averted illnesses, hospitalizations, and deaths for each season. We assumed that the influenza B lineage included in IIV3 would provide a relative vaccine effectiveness of 75% against the other B lineage. Results Compared to use of IIV3, use of IIV4 might have led to an additional reduction ranging from 0·4 to 14·3 influenza‐associated illnesses per 100 000 population/year, <0·1 to 0·5 hospitalizations per 100 000/year, and <0·1 to 0·4 deaths per 1000/year. Based on extrapolation to national population estimates, replacement of IIV3 with IIV4 might have averted an additional 267–9784 influenza‐associated illnesses, 9–320 hospitalizations, and 0–3 deaths. Conclusion Compared to use of IIV3, IIV4 has the potential to further reduce the burden of influenza‐associated morbidity and mortality in Thailand.

Background Each year, an influenza B strain representing only one influenza B lineage is included in the trivalent inactivated influenza vaccine (IIV3); a mismatch between the selected lineage and circulating viruses can result in suboptimal vaccine effectiveness. We modeled the added potential public health impact of a quadrivalent inactivated influenza vaccine (IIV4) that includes strains from both influenza B lineages compared to IIV3 on influenza-associated morbidity and mortality in Thailand.
Methods Using data on the incidence of influenza-associated hospitalizations and deaths, vaccine effectiveness, and vaccine coverage from the 2007-2012 influenza seasons in Thailand, we estimated rates of influenza-associated outcomes that might be averted using IIV4 instead of IIV3. We then applied these rates to national population estimates to calculate averted illnesses, hospitalizations, and deaths for each season. We assumed that the influenza B lineage included in IIV3 would provide a relative vaccine effectiveness of 75% against the other B lineage.

Introduction
The trivalent inactivated influenza vaccine (IIV3) is composed of antigens from three influenza viruses: influenza A (H1N1), influenza A(H3N2), and one of two co-circulating lineages of influenza B virus (Victoria and Yamagata). In five of the ten Northern Hemisphere influenza seasons in the United States (US) from 1999 to 2009, the predominant B virus was not included in the seasonal vaccine. 1,2 Because vaccination with a strain from one B lineage provides variable cross-protection against the other, a mismatch between the virus selected for the vaccine and the predominant circulating virus may reduce vaccine effectiveness. 2 In 2012, recognizing that vaccine manufacturers were developing a quadrivalent inactivated influenza vaccine (IIV4), the World Health Organization began identifying a second influenza B strain that could be included in a seasonal influenza vaccine. However, there is currently no recommendation preferring either the trivalent or quadrivalent vaccine. 3 Vaccine manufacturers began distributing IIV4 during the 2012-2013 Northern Hemisphere influenza season.
Many countries have modeled the potential public health or economic impact of incorporating IIV4 into the national vaccine program. [4][5][6][7][8][9][10] Most studies have found a reduction in influenza-associated morbidity and mortality from the inclusion of a second influenza B lineage. [4][5][6][7][8][9][10] In the US, the potential net impact of the IIV4 on influenza-associated outcomes was demonstrated to vary substantially between seasons, depending on the percentage of influenza virus infections caused by each of the two influenza B lineages, vaccine coverage, and vaccine effectiveness. 4 In Thailand, annual seasonal influenza vaccination is recommended by the Ministry of Public Health for various high risk groups, including persons aged ≥65 years, persons with chronic underlying medical conditions, children aged 6 months to 2 years, pregnant women after the first trimester, healthcare personnel, mentally ill persons unable to care for themselves, and persons weighing >100 kg. Currently, IIV3 is provided free of charge in the national vaccine program (through the Ministry of Public Health and National Health Security Office) to the first five risk groups, but supply is sufficient to cover only 25% of the target population (T. Thantithaveewat personal communication) and coverage does not exceed 20% in any risk group. 11 We conducted a model-based analysis to estimate the potential public health impact of the IIV4 introduction on influenza-associated morbidity and mortality in Thailand based on a model developed for the US. 4 This model used data from Thailand from 2007 to 2012, including annual rates of influenza-associated outcomes, the proportion of illness due to the two influenza B lineages, vaccine coverage, and the vaccine effectiveness of the Southern Hemisphere IIV3, the principal formulation used in Thailand. Given the uncertainty over the eventual price of IIV4 in Thailand, a range of price assumptions was incorporated into the model to take into account the potential for a higher IIV4 cost compared to IIV3, and a fixed national vaccine program budget.

Epidemiological model
To calculate the estimated net difference in rates of influenza-associated morbidity and mortality that would be expected from use of IIV4 in general population, observed rates of influenza-associated illnesses, hospitalizations, and deaths during years when IIV3 was used were compared to rates modeled with use of IIV4 ( Figure 1). 4 We calculated the hypothetical rates at which influenzaassociated illnesses, hospitalizations, and deaths would have occurred in the absence of IIV3 for each type, subtype, and lineage of influenza virus using the following formula: Expected rate without vaccination where VC is the vaccine coverage, and VE is the vaccine effectiveness. In base case analysis, we assumed that the influenza B lineage included in IIV3 would provide a relative vaccine effectiveness of 75% against the other lineage not included in the IIV3 formulation. 12 We also varied this value in sensitivity analyses to accommodate different levels of cross-protection against the other lineage that was not included. To calculate the expected rates of influenzaassociated outcomes if IIV4 had been used instead, we used the following formula: Expected rate with IIV4 ¼ expected rate without vaccination Â ð1 À ðVC Â VE IIV4 ÞÞ: We then calculated the annual net difference in the number of cases of influenza-associated illness, hospitalization, and death that would be expected using IIV4 rather than IIV3 by applying the net rate differences to national population estimates. We reported estimate of each season and also ranges based on these point estimates. Table 1 shows the range of data input values and respective sources. All data were available in aggregated form and IIV3, trivalent inactivated influenza vaccine; IIV4, quadrivalent inactivated influenza vaccine gathered from published and unpublished sources, except the observed incidence of influenza-associated hospitalizations and deaths and influenza virologic surveillance data, for which individual-level, anonymized data were retrieved. We consulted with 14 subject-matter experts from a vaccine company and Thailand's Ministry of Public Health to ensure that all assumptions made and data input values (Table S1) were best suited for Thailand. To account for uncertainty of some input parameters, we considered a possible range of each input, suggested by the experts, in sensitivity analyses.

Population data
We obtained population data from the National Economic and Social Development Board of Thailand. 13 Because this population source reports population figures by 5-year age increments, we used data from the Ministry of Interior, which reports the population in finer age intervals (<1 and ≥1 years), to calculate the proportion of children aged <6 months (by dividing the proportion of children aged <1 year by two). 14 We applied this proportion to the total population to get an estimated number of children aged <6 months. To calculate the population eligible for influenza vaccination, we subtracted the number of children aged <6 months from the total population.

Influenza season and influenza vaccine
In Thailand, influenza viruses circulate year-round with a peak during June-November (when approximately 60% of the cases occur) 15 and a smaller peak during January-March. 16 Because seasonal influenza vaccination in Thailand is recommended to start in May, we defined an influenza season as June through May of the following year (e.g. the 2012 influenza season is June 2012-May 2013). 15 Evidence indicates that protection against viruses that are antigenically similar to those contained in the vaccine declines after 6-8 months. 3 All seasonal influenza vaccines licensed and used in Thailand from 2007 to 2012 seasons were IIV3. Although both Northern and Southern Hemisphere vaccines were available in Thailand, all vaccines purchased by the government and provided free of charge (approximately half the supply), as well as the majority of the privately purchased vaccines, were the Southern Hemisphere formulation. 17 Therefore, estimates of vaccine effectiveness used in the model were from the Southern Hemisphere IIV3.

Incidence of influenza-associated outcomes
The observed incidence of influenza-associated hospitalizations and deaths was measured from active population-based surveillance for hospitalized acute lower respiratory tract infections in two provinces in Thailand. 18,19 We chose to use the data from this source instead of the national passive surveillance system (based on discharge diagnoses classified according to the International Classification of Diseases 10) as data from the latter source are rarely laboratory-confirmed and likely to underestimate true disease rates. An influenzaassociated hospitalization was defined as a respiratory illness requiring hospitalization with a specimen positive for an influenza virus by polymerase chain reaction. The influenza-associated mortality rate was estimated from inhospital deaths among patients with influenza-associated hospitalizations.
The incidence of influenza-associated illnesses was not measured directly but was back-calculated from the influenza-associated mortality rate by applying the case fatality ratio (0Á03%) reported through the national passive surveillance system, averaged across 14 years. 8,20 Distribution of influenza type, subtype, and lineage The annual rates of influenza-associated outcomes by type, subtype, and lineage of influenza were calculated based on the distribution of influenza viruses identified through the national influenza virologic surveillance, recalculated by influenza season. 16 These proportions were applied to the incidence of influenza-associated illnesses, hospitalizations, and deaths to calculate rates by type and subtype or lineage.

Vaccine effectiveness
The vaccine effectiveness of the Southern Hemisphere IIV3 was estimated from published data from countries that used Southern Hemisphere vaccines. [21][22][23][24][25][26][27] The estimated vaccine effectiveness by influenza season was calculated by averaging all vaccine effectiveness data from a given season. Because there are no published data on the effectiveness of the Southern Hemisphere IIV4, we used data on the effectiveness of Southern Hemisphere IIV3 and assumed this level of effectiveness would apply to both lineages of influenza B included the IIV4. 4,5 Doses of vaccine available, administered, and vaccination coverage Data on the number of doses of influenza vaccine imported into the country were obtained from the Food and Drug Administration, Ministry of Public Health. 17 Influenza vaccine wastage rates in the public setting from 2010 to 2012 were from the published literature. 11 Because our study covered a wider range of influenza seasons, we used an average wastage rate of 9Á5% for the seasons with unavailable wastage data. To calculate the number of doses administered, we multiplied the number of IIV3 doses sold by 1-wastage rate. The wastage rates of vaccine used in private setting were not available; in consultation with subject-matter experts, we used a lower wastage rate of 3Á0% based on the assumption that the private vaccine providers who paid for the vaccines were unlikely to waste them. We applied this 3Á0% wastage rate in the private setting to the number of IIV3 doses sold privately to estimate the number of doses administered in this setting. We then divided the number of the IIV3 doses administered (both in public and in private settings) by the population aged ≥6 months to estimate the vaccine coverage in the general population. 13

Sensitivity analyses
We altered some model inputs in sensitivity analyses in which we (i) used a case fatality ratio of 0Á05% to calculate the incidence of influenza-associated illnesses, 20 (ii) used incidence of influenza-associated illness of 5% and 10%, 28,29 (iii) used estimates of vaccine effectiveness to prevent influenza-associated outcomes from the influenza B lineage not included in the IIV3 formulation of 0%, 10%, 20%, and 30%, and (iv) used vaccine wastage rates of 15% and 20% in public sector and of 10% and 15% in private sector, while keeping the values of other parameters fixed.

Potential price increase and budget
As the cost of IIV4 in Thailand was not available at the time of analysis, we calculated the reduction in vaccine coverage that would result from a 0% to 30% increase in the price of IIV4 over IIV3, assuming a fixed national budget for influenza vaccine purchase. Based on a comparison of the net difference in influenza-associated illnesses, hospitalizations, and deaths among various price assumptions, we estimated the price increase threshold below which IIV4 would continue to provide public health impact. We reported the threshold for which use of IIV4 would continue to avert at least one additional influenza-associated outcome compared to use of IIV3. The currency exchange rate used in the analysis was 35 Thai Baht to 1 USD.

Net difference in rates of influenza-associated outcomes
The observed annual incidence of influenza-associated illnesses, hospitalizations, and deaths when IIV3 was used from 2007 to 2012 ranged from 1541Á06 to 5487Á39 per 100 000 population, 59Á41 to 118Á42 per 100 000 population, and 46Á23 to 164Á62 per 1000 population, respectively ( Table 3). The modeled annual incidence of influenzaassociated illnesses, hospitalizations, and deaths if IIV4 had been used instead ranged from 1540Á11 to 5483Á10 per 100 000 population, 59Á37 to 118Á32 per 100 000 population, and 46Á20 to 164Á49 per 1000 population, respectively.
In Thailand, assuming that the influenza B lineage included in IIV3 would provide a relative vaccine effectiveness of 75% against the other B lineage, a dose for dose replacement of IIV3 with IIV4 might result in further reduction in cases of influenza-associated illness (annual additional cases averted ranged from 267 to 9784), hospitalization (annual additional hospitalizations averted ranged from 9 to 320), and death (annual additional deaths averted ranged from 0 to 3) compared to when IIV3 was used ( Table 3). The net impact of IIV4 was highest in years when the predominant circulating influenza B strain was not represented in the IIV3. The additional reduction in annual rates of influenza-associated outcomes ranged from 0Á40 to 14Á35 illnesses per 100 000 population, 0Á01 to 0Á47 hospitalizations per 100 000 population, and 0Á01 to 0Á43 deaths per 1000 population. Overall, we estimated that IIV4 could have averted an additional 21 974 cases of illness (average, 5Á41 per 100 000 population; 95% confidence interval [95% CI], 1Á37 to 9Á46 per 100 000 population), 698 influenzaassociated hospitalizations (average, 0Á17 per 100 000 population; 95% CI, 0Á04 to 0Á31 per 100 000 population), and seven influenza-associated deaths (average, <0Á01 per 1000 population; 95% CI, 0 to <0Á01 per 1000 population) compared to when IIV3 was used.
Compared to IIV3, there was a consistent positive net benefit from use of IIV4 across a range of assumptions for model parameters. The magnitude of the impact was increased when the parameters were varied in all sensitivity analyses (Table S2).

IIV4 price increase and budget scenarios
The cost of IIV3 is 3Á71 USD per dose in the public setting with bulk purchase and 11Á43 USD in the private setting. Assuming a fixed budget and that the influenza B lineage included in IIV3 provided 75% effective cross-protection against the other B lineage that was not included, the maximum IIV4 price increase threshold for which the vaccine would continue to provide additional public health impact over IIV3 was 9% (4Á05 USD per dose through bulk purchase in the public setting and 12Á54 USD in the private setting) ( Table 4). When 0-30% absolute cross-protection was assumed, the maximum IIV4 price increase threshold was estimated to be 13-22%.

Conclusions
Using a simple model, we explored the potential public health impact of IIV4 versus IIV3 for preventing influenzaassociated illnesses, hospitalizations, and deaths in Thailand. Our findings indicate that the influenza disease burden for all seasons from 2007 to 2012 would have been decreased by use of IIV4 if vaccine coverage had been maintained at the same level as IIV3; the impact was greatest for influenza-associated illnesses and hospitalizations but minimal against influenzaassociated deaths likely due to an underestimation of in-hospital mortality data. When considering various price assumptions and a fixed national vaccine program budget, we estimated that the maximum price increase that could be supported for IIV4 that still would have averted at least  *Assuming no loss in purchasing power (dose for dose replacement of IIV3 with IIV4) and influenza B lineage included in IIV3 would provide a relative vaccine effectiveness of 75% against the other lineage not included in the IIV3 formulation. **Observed incidence when IIV3 was in use. ***Modeled incidence if IIV4 had been used instead. †Negative numbers indicate net influenza-associated outcomes averted with IIV4 compared to IIV3.
one additional influenza-associated outcome was 9-22% depending on assumptions regarding the level of influenza B cross-protection. The potential public health impact of the IIV4 was estimated previously in the United Kingdom, US, Germany, China, and Canada. [4][5][6][7][8][9][10] Our findings are consistent with other reports that estimated a reduction in influenza Bassociated outcomes by the IIV4 in years when the predominant circulating influenza B lineage was not included in the vaccine formulation. In our analysis, the estimated potential impact of IIV4 was substantial in the years when a significant mismatch between the vaccine and circulating strains occurred. We also found that a dose for dose replacement of IIV3 with IIV4 appeared to be beneficial for all influenza seasons from 2007 to 2012 in contrast to the US where the benefit of IIV4 varied by season. 4 One explanation for this finding could be the perennial transmission of influenza in Thailand compared to the seasonal transmission in the US.
Estimates of the incidence of influenza-associated outcomes (illness, hospitalization and death) were subject to some limitations. Incidence rates for hospitalizations and deaths were measured in just two provinces in Thailand and generalized to the entire country, but may not accurately reflect disease burden in other provinces. Nonetheless, we believe these estimates, derived from active population-based surveillance, are more accurate than the national data collected from passive surveillance system which, by nature, are limited by variability and incompleteness in reporting. 30 The incidence rates for influenza-associated illnesses were not directly measured but estimated from observed hospitalization rates and a case fatality ratio measured through passive surveillance. While the estimated rates for Thailand (range, 1541-5487 per 100 000 population) appeared to be comparable with those reported from the US (range, 3000-10 000 per 100 000 population), 4 biases in the measurement of a case fatality ratio for influenza through passive surveillance could have affected the accuracy of our estimation of incidence rates. We found that the calculated incidence of influenza-associated illnesses in our study were slightly lower (range, 1-5%) than what was reported in a recent meta-analysis and the World Health Organization's (WHO's) estimation. 31,32 Our sensitivity analyses showed a greater impact on the burden of influenza-associated illnesses for a dose for dose replacement of IIV3 with IIV4 when the WHO's incidence of 5% and 10% were used. 32 Further, we believe that a mortality rate based on hospital-based surveillance is likely to underestimate mortality because it will not include out-of-hospital mortality and mortality that might occur from influenza-associated complications after virus is no longer detectable. The mortality rates derived from the surveillance were indeed lower than recent reports from Thailand by Cooper et al. 33  Reviewing the published literature to gather input parameters for the model highlighted important gaps in influenza knowledge in Thailand. In particular, we found few estimates of vaccine effectiveness for countries in the Southern Hemisphere in general, and Thailand in particular. There were only two published articles and one abstract that reported the effectiveness of Southern Hemisphere IIV3 used in Thailand, and these reports were limited to key risk groups. 22,25,35 Population-based estimates of influenza incidence in Asia are also limited. 36 Our model has a few limitations. First, we could only investigate the public health impact of IIV4 for six influenza seasons due to the lack of high-quality data for other seasons. Second, the model did not take into account the impact of herd immunity, although the estimated vaccination coverage for all target groups is low in Thailand, making it likely that herd immunity effects are minimal. Further, the vaccine effectiveness estimates used in the model were against laboratory-confirmed medically attended influenza virus infection and we did not apply vaccine effectiveness against mortality when we considered the benefit of IIV4 for such outcome. It is likely that the effectiveness of the influenza vaccine in preventing mortality might differ and, therefore, we may inaccurately estimate the potential impact of IIV4 on mortality. Nonetheless, the base case analysis using 75% relative vaccine effectiveness assumption addressed recent findings that suggested vaccine effectiveness for the opposite B lineage may be better than previously expected. 12,37 The adapted model in this study was aimed to be simple yet informative to permit updates to include other scenarios (e.g. risk group specific, budget circumstances) should more data become available or future influenza seasons. Our findings support a net positive public health impact from IIV4 implementation in Thailand. Assuming limited flexibility in the available budget for influenza vaccine purchase, policymakers could make use of the pricing threshold data to inform decisions about an acceptable IIV4 price range.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1. Data input values used in the assessment of potential public health impact of a quadrivalent inactivated influenza vaccine in Thailand. Table S2. Difference in influenza-associated outcomes between that expected with a quadrivalent inactivated influenza vaccine (IIV4) and that observed with trivalent inactivated influenza vaccine (IIV3) in Thailand in various scenarios.