Pneumococcal Diseases Caused by Serotype Replacement in Colombia: A Cost-Effectiveness Analysis of Pneumococcal Conjugate Vaccines in Infants and Its Herd Protection in Older Adults

Background. The recent and available evidence on the distribution of pneumococcal serotypes, which affects the effectiveness of pneumococcal conjugate vaccines (PCV), suggest that additional health economic studies will be important for better understanding of potential economic benets of pneumococcal vaccination. Methods A cohort simulation model was used for new births in Colombia between 2019-2022 and adults over 65 years, from social perspective (direct and indirect costs). The time horizon was a life expectancy and discount rate for costs and benets of 5%. The outcomes were presented in terms of avoided pneumococcal diseases ─ Invasive Pneumococcal Disease (IPD), Community Acquired Pneumonia (CAP), Acute Otitis Media (AOM), and sequelae ─ , years of life gained (AVG) and herd effect in older adults. Results. Based on data from National Data of the serotype distribution between 2017-2019, PCV10 covers 5% of serotypes while PCV13 63%. The additional cases that PCV13 would prevent are in children: 1,205 cases of IPD, 60,274 of CAP, 26,619 of AOM, 3,980 deaths, 28 cases of neuromotor disability and 1,251 cochlear implants. In older adults, PCV13 would prevent 818 additional cases of IPD and 29,983 of CAP, generating 149,186 additional LYGs to those of PCV10 with a saving to the health system and patients of US $ 169,261 thousand. The model shows robustness in the sensibility analysis. Conclusion. PCV13 is a cost-saving strategy versus PCV10 to prevent pneumococcal diseases. administration cost of USD 1 per dose incorporated. 2+1 dose vaccination schedule at 3, 5 and 12 months of age.

analysis that allows to determine the incremental cost-effectiveness ratio, both in the prevention of IPD in terms of LYG as in pneumococcal pneumonia and AOM in terms of avoided cases in children under ve years old between 2019 and 2022, as well as the herd effect in adults, expressed in terms of cases and deaths prevented by pneumococcal diseases.

Target population
Neonates projected to be born in Colombia between 2019 and 2022, who will be vaccinated, and adults older than 18 in the country or who turn 18 between 2019 and 2022, in which the herd effect will be evaluated.

Comparators
The strategies evaluated are the two pneumococcal conjugate vaccines that currently licensed in the scheme 2+1: PCV13 and PCV10. The rst one is produced by P zer® and contains thirteen serotypes of pneumococcus (1,3,4,5 The Serum Institute of India Pvt. Ltd.'s PCV candidate (SIIPL-PCV), pre-quali ed by WHO and with immunogenicity data (20), was not included in the analysis, due to the lack e cacy trial, it is not approved by National Institute for Surveillance on Drugs an Foods (INVIMA -Spanish acronym), and the price of this vaccine is not available, taking in consideration that for the national program, the purchase of vaccines from producer is via tender, the PAHO.

Time horizon
The relevant impact on public health of pneumococcal conjugate vaccines (PCVs) has been their high clinical e cacy, but also their protective effect during lifetime, for this reason, the time horizon for this cost-effectiveness analysis was the life expectancy.

Study perspective
This analysis has a social perspective, this implies identifying direct costs of resources associated with the use of these technologies and pneumococcal diseases treatment' costs and their aftermath, included or not in the Plan of Bene ts of Health of Colombia (21). In the same way, indirect costs by loss of productivity in people with neurological sequelae caused by pneumococcal meningitis; and the health outcomes perceived by patients.

Discount rate
A discount rate of 5 % was used for both costs and bene ts. This is considered the discount rate that best ts the Colombian context and is also the discount rate recommended by the Institute of Technological Evaluation in Health in Colombia (22).

Choice of health outcomes
Outcomes were avoided cases of IPD (pneumococcal invasive pneumonia, pneumococcal sepsis and meningitis); community-acquired pneumonia (CAP), deaths by IPD, both children and adults; and pneumococcal AOM in infants. Similarly, sequels derived from pneumococcal disease such as neuromotor disability and cochlear implants; and LYG in children and adults. Prevented deaths and IPD cases in unvaccinated adults were estimated from herd effect in this population (23).

Clinical parameters
A search strategy was carried out to identify clinical e cacy of pneumococcal vaccines, including only phase III and IV randomized controlled trials (RCTs) in infants, and excluding those that have included polysaccharide vaccine. Finally, we looked for RCTs that met the Grading of Recommendations Assessment, Development and Evaluation (GRADE) bias control criteria (24). The effectiveness was adjusted for the national distribution of pneumococcal serotypes in Colombia between 2017 and 2019 (66% PCV13 and 5% PCV10), based on the information reported in SIREVA II (14,15). These is the most commonly used approach to calculate the direct effect of each vaccine (25).
The vaccination coverage was 89% in each birth cohort (26), which in turn was adjusted among infants as a herd effect in this group (27), 93.6%. Also during this time the adults will be at risk of invasive pneumococcal disease, herd immunity is accounted in the model as 22.14% (28), and serotypes coverage in this group was adjusted (14) ( Table 1).
For children above 5 years old, the bacterial CAP, and pneumococcal AOM were taken from control arm in the RCT of PCV10 (18) and observational studies. The incidence and lethality of meningitis due to S. pneumoniae were estimated from average in the reports from 2017 -2019 years from Health National Institute. The incidence of sepsis due to S. pneumoniae was estimated from Colombian Health Services Delivery Records (RIPS) and the proportion of cases by S. pneumoniae reported in epidemiological studies. For the rest age groups the incidence rates were derived from RIPS during vaccinal period with PCV10. The mortality for all the causes was implemented an age-dependent according with the Colombia's National Administrative Department of Statistics (DANE) (31) ( Table 2).
Currency, price data, and conversion Costs were calculated in Colombian pesos (COP) and later were converted into USD, based on an average exchange rate reported from the Central Colombian Bank on 2019, COP 3,281 for USD 1 (38).

Estimating resources and costs
A top-down methodology to calculate costs of hospital care for children with IPD and bacterial CAP, the outpatient healthcare for children with neuromotor disability, and the adults with IPD was made, by mean a search of individual costs of all interventions, procedures and medicines required from the database of a Health Maintenance Organization (HMO) of Colombia with national presence and costs. A bottom-up methodology for calculation costs of outpatient treatment in children with pneumococcal AOM and adults with CAP was made, taken the costs from taken the costs from the Drug Price Information System (SISMED, acronym in Spanish) (39). Cochlear implant total costs for the treatment of profound sensorineural hearing loss in children were taken from an economic analysis published in Colombia in 2012 (40), and the costs were adjusted for the year 2019 based on the accumulated consumer-price index for the period 2012-2019 (Table 3).
Direct costs of healthcare for children with neuromotor disability were calculated at American dollars (USD) $ 294.14 per year, including outpatient healthcare by medical specialists, physiotherapists and medicines, during lifetime horizon. Indirect costs of these patients were calculated under the assumption that they need 24/7 on-site care, usually by a family member (most of the time, the mother) who cannot produce income at home or paid a caregiver because this person must constantly take care of the patient with neuromotor disability. In either case, the salary was calculated as the mean salary in Colombia during 2019, which must be paid during lifetime horizon of patient. Finally, costs of productivity because neuromotor disability in children were calculated based on Colombia mean salary, assuming most people work between 25 and 65 years old (age of economic productivity) (41); and adjusted by the national unemployment rate during 2019 (42).
The cost of the vaccines were taken from 2019 price list of the expanded immunization program published on PAHO website (43). PCV10 price was USD 12.85 per dose and PCV13 USD 14.50 per dose (43). The administration cost of USD 1 per dose was incorporated. The study assumed implementation of a 2+1 dose vaccination schedule at 3, 5 and 12 months of age.
Choice of model A dynamic population model was designed to estimate costs and long-term health outcomes ( Figure 1). Model structure was based on natural history of pneumococcal diseases (AOM, CAP and IPD) and their complications in infants and children (neuromotor disability and cochlear implant) considering the period PCV10 post-vaccinal. Clinical outcomes included on model were taken from PCVs clinical trials and after-effects were identi ed as most severe related with pneumococcal infections (44). Model includes vaccinated children (< 5 years) and unvaccinated adults (> 18 years), who bene t of PCVs herd effect (23).
Given that model time horizon is population lifetime, analysis was made with cycles based on annual probability of developing pneumococcal diseases.
The model censures every year those patients who developed a pneumococcal disease or died and starts calculating the probabilities, and so on until lifetime horizon of each cohort.
Population were divided into three groups, according to risk level of becoming ill and dying from pneumococcal diseases: children vaccinated under 5 years of age; adults not vaccinated between 18-64 years; and unvaccinated adults > 64 years old. This economic model follows the newborns cohort born between 2019-2022 to expectancy of life of rst cohort (80 years), which integrated the unvaccinated population in order to measure indirect vaccination effect.

Assumptions
This decision model considers following assumptions with respect to comparators: Herd effect on unvaccinated children and adults is same for both PCVs, although between the comparators, only PCV13 has shown protects unvaccinated individuals against vaccine-type pneumococcal infections across all age groups(45-48), but not PCV10 (49-51).
The great migration of Venezuelans that Colombia has received in recent years (more than one million people in 2017) would imply a change in frequency distribution of country pneumococcal serotypes. Since SIREVA II records (15) do not identify patients' nationality, model assumes that such migration would not affect effectiveness of either of two PCVs. De nition of older adult may differ among published studies, some include adults > 60 years (29) or > 50 years (52). Model assumes older adult de nition as those > 65 years of age, because most manuscripts have assumed such de nition (28). For this reason, model has two different age groups to determine herd effect: 18-64 years and > 65 years.
Life expectancy of 80 years in Colombia, which is slightly higher than what country currently has. Given that model evaluates a dynamic cohort of newborns until 2027, a life expectancy of 80 years was assumed because it can be achieved in next ten years, since it is a life expectancy like that currently reported by countries of the region, as Chile (53).
Direct e cacy of vaccination is attenuated in both vaccines. This assumption is based on data from an exploratory analysis of The Community-Acquired Pneumonia Immunization Trial in Adults (CAPiTA) trial (e cacy of PCV13 in older adults), in which they observed that e cacy had a 3 % e cacy attenuation annual rate without statistical signi cance (54). For this reason, sensitivity analysis of this model will evaluate results with different values of effectiveness attenuation.

Sensitivity Analyses
Sensitivity analyses were performed to consider the uncertainties associated with different parameters, incidence rates, outcomes and costs. First, a sensitivity analysis was conducted to determine the in uence of a speci c parameter had on the incremental cost-effectiveness value, and the second, sensitivity analysis was a probabilistic analysis with the objective of captured the uncertainties of all parameters, it was conducted using second-order Monte Carlo simulations. The rates were adjusted to beta distributions and the costs to gamma distributions.

Base case analysis
The vaccination in the rst year of the vaccination with PCV13 compared with PCV10 could potentially prevent 2 cases of pneumococcal meningitis, 7 cases of pneumococcal sepsis, 1,083 cases of pneumonia, 897 cases of AOM, and 66 fatalities. Considering the herd effect for population non vaccinated, 41 episodes of pneumococcal meningitis, 777 cases of pneumococcal sepsis and 29,983 cases of pneumonia among adults could potentially be avoided. PCV13 yields 150,351 LYG in vaccinated cohorts (Table 4).
Regard to costs, the incremental cost of vaccination with PCV13 of full cohorts was estimated to USD 15,676 thousand, and due the reduced burden of pneumococcal disease, the potential cost saving for all cohorts were calculated to USD 169,261 thousand, direct costs were USD 125,025 thousand and indirect costs were USD 6,210 thousand. Taking herd effect into account, the cost saving for the cohorts were estimated to USD 53,702 thousand ( Table   5).
As result, PCV13 shows as cost-saving strategy versus PCV10, with an incremental 149,186 LYG more than PCV10 and saves more than USD 169 million. PCV13 is a cost-saving strategy even when herd protection was not incorporated.

Sensitivity results
The determinist sensibility analysis showed that variation in probabilities, clinical outcomes and costs were relatively insensitive to changes in the parameters, the parameters with the greatest impact in this sensitivity analysis was the "cost of inpatient pneumonia in adults", which also re ects that PCV13 is a cost-saving strategy. None of the parameters change these results (Figure 2). In the probabilistic analysis, PCV13 will be cost-effective strategy in all thresholds compared with PCV10.

Discussion
PCV13 is a cost-saving strategy to avoid pneumococcal diseases in children and adults in Colombia compared to PCV10. The PCV13 inclusion would avoid an important number of pneumococcal OMA in infants and deaths, sepsis, consolidated pneumonia and LYG both adults and children, In terms of economical results, this vaccine would bring approximately USD 125 million of savings associated to care of health outcomes and 6 million indirect costs, the attention of the main sequels of pneumococcal disease during the horizon time of the study.
The main factor of these results is explained to difference of pneumococcal serotypes covered by both technologies: PCV10 contained less than 5% of the isolated serotypes of the cases of IPD reported by SIREVAII network and PCV13 almost 66% which affecting directly the effectiveness of both vaccines. The 19A and 3 serotypes entirely responsible for the difference observed in the vaccinal serotypes. Other study reported the positive trend of these serotypes in the pneumococcal disease over time as described in other study (17). Raw data comparing 19A and 3 serotypes distribution in 2011-2014 with 2015-2019 period reported a notable increase especially 19A serotype (30-44% and 8-9%, respectively) (14).
One report highlights the emergence of Streptococcus pneumoniae serotype 19A in the pediatric population in Bogotá, Colombia as the main cause of invasive pneumococcal disease after the introduction of PCV10 with important association to multidrug resistance, in patients with complete vaccination schedule with 3 doses of PCV10, suggesting that the emergence of serotype 19A occurs in vaccinated children and that cross protection of PCV10 against this serotype is not generated (55). An increasing burden of serotype 19A, is a substantial change in the epidemiological where product switching is recommended according with the WHO position paper. (19) Other Latin-American countries have observed similar phenomenon, which non-vaccinal serotypes grow in isolated of IPD. A study carried up in Porto Alegre (Brazil) reported changes in percentage of 19A serotype from 1.2% in 2011 to 18.5% in 2014 (56). Other study in Brazil analyzing the laboratorybased surveillance reported the changes from 3.2% in the period early-post -PCV10to 9.3% ve years after (57). In Chile, the 19A serotype went up the relative frequency in general population from 3.3% in pre-PCV10 to 12% in 2014 and 13% in 2015. In infants the changes were more notables starting 5.8% reaching 25% and 19% in the 2014 and 2015 years (58).
This cost-effectiveness analysis has two features following the recommendation of the WHO (59): herd effect on adults and the indirect costs and loss of productivity of children who develop neuromotor disability. Tsaban and Ben-Shimol published a systematic review of the studies, based on which it was possible to calculate an indirect effect of PCV vaccination in adults which could vary according to uptake rates and time of PCV implementation (approximately 22% (range: 0% − 70%) (28). This study found that the population most bene t of herd effect is the adults > 65 years old. This means that an important proportion of population at risk of developing pneumococcal diseases, especially CAP, could be protected and reduce healthcare costs associated to attention of this disease.
Similarly, considering social perspective to make this economic evaluation requires us to calculate indirect costs of care for children who develop neuromotor disability as a complication of pneumococcal meningitis. These indirect costs are mainly produced by costs of the child's caregiver, which may be his mother or a third party (nurse) who cares the child. If caregiver is the mother, she must stop working to care for the child and therefore withdraws from labor market, in case the caregiver is a third party, the family must bear costs of this care. Likewise, costs of lost productivity of children with neuromotor disabilities should be included in economic evaluations of pneumococcal vaccines. Main raison d'être of the health systems is to avoid population poverty by two mechanisms: reduce out-of-pocket expenditure on health and offer good health to be economically productive. If a complication such as neuromotor disability cannot be avoided, child and his family will have greater economic vulnerability because both child and his mother or a family member cannot be economically productive for his family and for society.
Model has two main features, rst a dynamic cohort and second the economic evaluation of herd effect in adults. An economic evaluation which incorporates dynamic cohort is closer to reality in mass vaccination technologies, allows to identify true economic outcomes of complications of low prevalence and high cost, usually minimized in xed cohorts cause this kind of models evaluated for time frame the same cohort but during this time they would be different newborn cohorts are not be considered, each year new population is exposed to pneumococcal diseases so they would be vaccinated.
Studies monitoring the effect of vaccination in real life had identify the effects of herd immunity in PCVs (60), showed the extend vaccines bene ts beyond the directly vaccinated population, which acts as a positive externality, decreases costs of healthcare and improves quality of life. Importance of herd effect lies in fact that pneumococcal diseases are common in older adults and their lethality is very high, in this way it is possible to indirectly prevent mortality from pneumococcal diseases. In addition, treatment of CAP and invasive pneumonia are expensive in older adults and complicate underlying diseases such as asthma or emphysema. Herd effects have been frequently quanti ed as part of economic evaluations of vaccines (61), the relevant results related with the incorporation of these indirect effects on unvaccinated people in the analysis is the decreased of Incremental Cost Effectiveness Ratio (ICER) and it has a minor impact on the ICER.
All these features allow to identify most of health outcomes in vaccinated and unvaccinated people and estimate in a better way total health resources used since vaccination programs are not just the vaccine acquisition costs, they are, as well, the potential additional disease burden over time and associated costs that may be incurred as a result of implementation. One of the limitations of this study was not included all the indirect effects, as the contributions to antimicrobial resistance, some studies have demonstrated that PCV13 prevents the transmission of S. pneumoniae (62)(63)(64)(65), thereby reducing infections and limiting their transmission, vaccines allow providers to reduce the prescription of antibiotics (66). Therefore, there is a broader bene t of vaccination with PCVs, which, would be worth incorporating it into future studies. Other limitations is related to low precision and lack of statistical power of the PCV10 e cacy in Pneumococcal OMA due to the number of cases isolated (56 episodes) according to report of clinical trial, however, it is the currently available evidence considering that the isolation of causal agent in OMA is not a common practice.
The cost-effectiveness analysis in Colombia have become a key component of decisions on whether to adopt a public health policy, the previous Colombian speci c cost-effectiveness analyses related to PCV have serve as input to the decision makers. One of them showed the cost effectiveness of PCV10 (4), with ICER of cost per LYG was US$ 1,837 with PCV10 and US$ 9,514 with PCV13 (67). However, up-date analysis that incorporates the serotypes during 2011 to 2014 period, found PCV13 was found to be good value for money and prevent additional burden of pneumococcal disease saving additional treatment costs, when compared with to keep-PCV10 in Colombia, with ICER per YLS was USD 2,319 for keep-PCV10 and USD 1,771 for switch-to PCV13 (17). These results showed ICER of PCV13 compared with PCV10 have changed and in this analysis PCV13 have showed a cost saving strategy. Although it was clear that these analyses are not comparable, because necessities used their own assumptions and estimates parameters, they are evidencing the changes in the in the results of ICERs, where PCV13 have becomes prefer option.

Conclusions
PCV13 is dominant versus PCV10 as a strategy to vaccinate newborns in Colombia because PCV13 avoided more IPD, consolidated CAP, pneumococcal AOM, complications, sequelae and deaths than PCV10. This is explained, mainly, because serotypes included in PCV13 represents 67% of those isolated in Colombia in children under 5 years between 2017-2019, while those included in PCV10 represents 5.2% isolated in the same period.