Introduction

CC is a significant health issue in Colombia. According to the National Cancer Institute every year there are 5,500 new cases of CC in the country [1]. For 2002, 2,045 deaths due to malignant cervical tumors were reported, making CC a leading cause of death in women, above breast and gastric cancer [2].

The screening program implemented in 1991 has contributed to a decrease in mortality from this cause over the past 7 years. However, this decrease is less than originally expected.

Epidemiological studies conducted in the 1990's provided evidence on the role of the HPV in the etiology of CC[3,4]. This led to the introduction of two HPV vaccines in recent years: a bivalent vaccine that prevents infection by types 16 and 18, and a quadrivalent vaccine against HPV 06/11/16/18.

According to the World Health Organization (WHO), it is estimated that developing countries, with a high incidence of CC and difficulty implementing screening programs, are the ones that may derive the greatest benefit from vaccination against the human papillomavirus [5]. For Colombia, local studies have detected quality problems in screening[6], and even though coverage is around 70%[7], technical and economic resources are not being correctly allocated to ensure quality of the testing or appropriate follow-up of abnormal results [8].

Hence, an in-depth analysis of the issue is necessary, in order to evaluate the convenience of vaccination with one of the two types of vaccines available. Such analysis must be based on the evidence available so far and consider the unique characteristics of the national context such as the prevalence of infections by the different HPV types, the coverage of the vaccination programs, and the cost of the vaccine in the domestic market, among others.

Consequently, the goal of this work is to compare the cost and the effectiveness of HPV vaccines in the Colombian context.

Materials and Methods

Model Design

A Markov model was developed in order to represent the natural history of CC and genital warts, based on prior models of the disease accessed through a systematic search in Pubmed-Medline, Embase and Lilacs. The model and the validity of the assumptions were then reviewed by gynecologists, gynecological oncologists and experts in decision-making models. The specialized TreeAge Pro 2011® software was used to build the model.

The model was designed as described in Figure 1. It consists of 13 mutually exclusive health states that represent the different events throughout the natural history of the disease. It was refined using models found in the relevant literature [9-13]. In the analysis, the cohort would start at 12 years of age, the age at which the women would be vaccinated. This age was chosen in accordance with the WHO recommendation for starting vaccination [5]. Using yearly cycles, the women in the cohort would go from one state to another in accordance with the transition probabilities that govern the model. The model was set to run until the cohort was 76 years old, according to life expectancy for women in Colombia.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Simplified model of the disease.

Markov model of disease is represented. Circles correspond to the states and arrows represent the allowed transitions. HPV: Human papillomavirus; LSIL: Low-grade Squamous Intraepithelial Lesion; HSIL: High-grade Squamous Intraepithelial Lesion; CC: Cervical Cancer. Source: Author.

https://doi.org/10.1371/journal.pone.0080639.g001

The following assumptions were used for the sake of simplification:

1. - All cases of genital warts are diagnosed and treated within the same year, and treatment efficacy is 100%.
2. - Only healthy women may become infected with HPV 6/11 and, consequently, acquire genital warts.
3. - The screening approach for all women who have access to it is called 1-1-3 program, which means that after two negative cytology test results, they wait three years for testing. Otherwise, they should come back the following year for another cytology test [14].
4. - For cancer stages I and II, treatment will be given to those women who are identified by screening or who develop symptoms that prompt them to seek medical care. For cancer stages III and IV, considering that the probability of symptoms appearing is 0.6 and 0.9, the model was simplified by bypassing screening as a tool for diagnosing the disease.

This model will measure costs in US Dollars and effectiveness in averted DALYS (Disability-Adjusted Life Years), in accordance with the WHO recommendation[15]. DALYS are defined only for CC stages. For other diseases such as Genital Warts, Low-grade Squamous Intraepithelial Lesion (LSIL) or High-grade Squamous Intraepithelial Lesion (HSIL) DALYS have not been defined.

Model Calibration

To ensure that model was actually representing the natural history of disease in Colombia a calibration procedure was performed. This calibration was expected to minimize any possible bias introduced when selecting probabilities used in models for other countries and when modifying those probabilities to be adjusted to the Colombian population, as discussed above.

For calibration, the model was run without any intervention, and the number of the resulting cancer cases, by age, was compared with the Cancer Population Registry of Cali, Colombia, between 1962 and 1966 [41]. Since in the preliminary evaluation of the model no effect of any of the interventions analyzed in the study (vaccination or screening) would be considered, a registry of incidence dating back before 1990 was used, considering that this was the first year of the National Program for the Control and Early Detection of Cervical Cancer [8]. This comparison is showed in Figure 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Comparison of the incidence calculated in the Model and the incidence reported in the National Register.

Both curves represent incidence by five-year groups. Source: Author.

https://doi.org/10.1371/journal.pone.0080639.g002

After comparing the results obtained with the official registry, options for modifying the model or the parameters were then considered. The model with the parameters derived from the systematic search was preserved to the greatest extent possible so as not to introduce arbitrary modifications that would undermine its veracity. An effort was made to keep modifications to a minimum.

A Chi-square goodness of fit test with 14 degrees of freedom on the Observed versus Expected Model frequencies was done and the difference was not significant (P<0.20). Consequently, it was deemed appropriate to use the health state transition probabilities found in the literature (see Table 1).

Results

Result of the Baseline Cost-Effectiveness Analysis

The baseline analysis using the main parameters revealed that the most costly strategy is the bivalent vaccine, followed by the quadrivalent vaccine, and lastly, screening with no vaccination. The same order was found for the effectiveness results, where the bivalent and quadrivalent vaccines were estimated to be more effective than screening without vaccination. The results are shown in Table 2 and Figure 3.

Name of the Strategy	Costs (US dollars)	Effectiveness (DALYS)	Average C-E Ratio¥	Incremental Costs	Incremental Effectiveness(Averted DALYS)	Incremental C-E Ratio ∞
No Vaccination	551	3.483	158
Quadrivalent Vaccine	688	3.478	198	137	0.00563	24241
Bivalent vaccine	713	3.478	205	162	0.00563	28765

Table 2. Results of the baseline cost-effectiveness analysis.

¥ Strategy costs divided by DALYS

∞ Strategy Costs - ‘No vaccination’ Costs / ‘No vaccination’ DALYS - Strategy DALYS

CSV

Download CSV

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Cost-Effectiveness baseline analysis.

Source: Author. This represents the relationship between cost and effectiveness for assessed alternatives.

https://doi.org/10.1371/journal.pone.0080639.g003

In accordance with the WHO CHOICE methodology, in the base case analysis it was determined that HPV vaccination, either with the bivalent or the quadrivalent vaccine, is a ‘non cost-effective’ strategy. The ICERs were estimated to be 24,241 USD/Averted-DALY for the Quadrivalent Vaccine and 28,765 USD/Averted-DALY for the Bivalent vaccine relative to screening; both values are higher than 3 times GDP per capita for Colombia (18,673 USD) which is often recommended as an appropriate maximum willingness to pay threshold value.

Sensitivity Analysis

Considering one-way sensitivity analyses, it was also noted that cost variations within the defined values did not result in significant modifications of the cost-effectiveness profiles of the alternatives evaluated. However, when variation in the cost of the vaccines was considered, it was noted that there was a considerable impact on the cost-effectiveness profile. Table 3 shows how at extreme price values, one vaccine can be dominated by the other one. Within reasonable sensitivity analysis intervals, neither vaccine is consistently more cost-effective.

Comparison	Parameter	Value	ICER	Comparison	Parameter	Value	ICER
QV Vs. NV	Efficacy Bivalent Vaccine	0.9	24210	BV Vs. NV	Cost Bivalent Vaccine	one hundred and thirty-three US Dollars	17258
BV Vs. NV			28727	QV Vs. NV			24210
BV Vs QV				BV Vs QV			QV dominated
QV Vs. NV		1	24210	QV Vs. NV		four hundred and eighty-seven US Dollars	24210
BV Vs. NV			25048	BV Vs. NV			67404
BV Vs QV			32548	BV Vs QV			BV dominated
BV Vs. NV	Efficacy Quadrivalent Vaccine	0.338	28727	QV Vs. NV	Cost Quadrivalent Vaccine	one hundred and fifty-seven US Dollars	19763
QV Vs. NV			80519	BV Vs. NV			28727
QV Vs BV			847	BV Vs QV			BV dominated
QV Vs. NV		0.984	21439	BV Vs. NV		two hundred and eighty-two US Dollars	28727
BV Vs. NV			28727	QV Vs. NV			37549
BV Vs QV			56118	BV Vs QV			QV dominated
QV Vs. NV	Discount Of Costs	0	9262	QV Vs. NV	Discount of Health Outcomes	0	6162
BV Vs. NV			14338	BV Vs. NV			7311
NV Vs. BV		6%	29042	BV Vs. NV		6%	80795
NV Vs. QV			33308	QV Vs. NV			95871
QV Vs. NV	Vaccination Coverage	0.5	24300	QV Vs. NV	Protection Duration	30 Years	40082
BV Vs. NV			29069	BV Vs. NV			46757
BV Vs. NV		1	24245	QV Vs. NV		70 Years	24210
QV Vs. NV			28629	BV Vs. NV			28727
QV Vs. NV	Screening Coverage	0	4976	QV Vs. NV	Cross Protection		13835
BV Vs. NV			5942	BV Vs. NV			16986
QV Vs. NV		1	32462	NV: Non Vaccination – Screening Alone; BV: Bivalent Vaccine; QV Quadrivalent Vaccine; A vs. B means that B is used as reference to calculate the ICER.
BV Vs. NV			38761

Table 3. One way sensitivity analysis and other analysis results.

CSV

Download CSV

It was found that in order for the quadrivalent vaccine to become a ‘cost-effective’ alternative in comparison with screening, it must be priced at US$ 49 or less per dose. A price of US$ 19 per dose would be necessary for this alternative to become ‘Very cost-effective’.

Regarding the bivalent vaccine, to be ‘cost-effective’ when compared with screening, the maximum price per dose of bivalent vaccine should be US$  47, and to be ‘Very cost-effective’ a price of US$ 17 would be necessary.

Considering that the quadrivalent vaccine has additional protection against genital warts, if its price is the same reported in 2010 (USD188/ vaccinated girl), the bivalent vaccine should cost US$ 183 per vaccinated girl to be as cost-effective as the quadrivalent vaccine.

In terms of transition probabilities, it was noted that reasonable changes in parameter values do not have a high impact on the results. However, it was found that results may change when the risk of developing LSIL and moving on to develop HSIL is modified. In general, to the extent that risks are increased, vaccines become more cost-effective, with an ICER slightly below the 3 times per capita GDP threshold or higher with relatively small increases in disease progression risks (Results available from authors on request).

On the other hand, if the sensitivity of the cytology test is very low, vaccination becomes a ‘cost-effective’ alternative. By the same token, if the sensitivity of the cytology test increases, vaccines go further into the ‘non cost-effective’ range when compared with screening alone (Data available on request).

It was also observed that the quadrivalent vaccine dominates the bivalent vaccine when its effectiveness reaches the maximum value (0.984), although it does not become a ‘very cost-effective’ alternative. When the effectiveness of this vaccine drops below 0.5, it conversely becomes dominated by the bivalent vaccine.

Regarding the time discounting of costs it was found that the vaccination with both vaccines became ‘cost-effective’ when rate was 0%. It was also noted that when the discount rate of benefits is between 0 and 1%, averted DALYS increased to such an extent that the bivalent vaccine becomes a ‘very cost- effective’ alternative, just like the quadrivalent vaccine.

The results derived from one-way sensitivity analysis are shown in Table 3.

With probabilistic sensitivity analysis it was found that ‘No-vaccination’ can be the chosen alternative in 39.4% of iterations, with a threshold of 3 GDP per capita. Also, quadrivalent and bivalent vaccines resulted in ICER values below 3 GDP per capita in 33.4 and 27.2% of iterations, respectively. However, with a threshold of 1 GDP per capita, ‘No vaccination’ is the selected alternative in 87.2% of iterations. Acceptability curves are shown in Figure 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Probabilistic Sensitivity Analysis: Acceptability Curve.

Source: Author.

https://doi.org/10.1371/journal.pone.0080639.g004

Discussion

As a form of health intervention, vaccination has been shown to reduce medical costs substantially in different countries. Global Alliance for Vaccines and Immunization has determined that most vaccination campaigns cost less than US$ 50 per year of life saved, whereas providing treatment for diseases such as hypertension may create costs to a level ranging between US$ 4,000 and US$ 90,000 [47].

As far as vaccination in Colombia is concerned, other strategies classified as ‘very cost-effective’, with an ICER lower than the per capita GDP, have been found recently. Such is the case of the conjugated pneumococcal vaccine [48],the tetanus vaccine[49] and the Pneumovax 23 in adults over 28 years of age [50].

This study compared screening with two HPV vaccines available in the market, using a Markov model with probabilities adjusted to the local population. This way, the long term progression of this disease was represented in the model, which made it possible to assess the impact of vaccination, even though a decrease in disease incidence is expected to be observed 15 years after vaccination or later.

To our knowledge, this is the first published cost-effectiveness study with such a close approximation to Colombian specific population. The effect of introducing the HPV vaccination on mitigating the burden of the disease had already been evaluated, though not from the perspective of a cost-effectiveness study [51]. Also, although in 2008 Goldie et al. published a cost-effectiveness study of HPV vaccination which covered six countries including Colombia, their model may not represent the specific Colombian reality, considering that probabilities included data for the entire Latin American population and the screening programs proposed did not match the Colombian situation. Additionally, costs were derived by approximation, from studies conducted in other countries, using GDP and other indicators [52]. Consequently, ours is the first cost-effectiveness study that comes closer to recreating the current characteristics of the Colombian population.

In contrast with studies conducted previously for other vaccines in Colombia, the baseline analysis led to classifying HPV vaccines as ‘non cost-effective’ when compared with the existing screening approach. Moreover, the quadrivalent vaccine was more cost-effective than the bivalent vaccine, especially because of the money saved by reducing the number of genital warts cases to be treated.

This model turned out to be very sensitive to the cost and efficacy of the vaccines. Vaccines never dominated or were dominated by screening within the ranges studied. However, the ICER positions of the vaccines did change, when the parameters for cost and effectiveness were modified. In terms of cost, lower prices per dose of the quadrivalent vaccine would make it dominant over the bivalent vaccine, as is also true when its efficacy comes close to the maximum value reported in the literature. This points to the need for paying close attention to the dominance relationships between the two vaccines.

Although the cost of the vaccines was taken from official databases, considered the best sources available on the grounds of impartiality and ability to represent the range of medication costs, prices will always be subject to trade conditions and to agreements between suppliers and buyers. Differential negotiations with the suppliers of these vaccines may change their cost-effectiveness profile. Although they will probably continue to be more costly and more effective than screening, the price relationship between the two vaccines will vary.

As far as the impact of efficacy is concerned, the study shows an overlap between the two vaccines in terms of the confidence intervals for this measurement. In view of this, there is not sufficient evidence to assert that any vaccine is more effective than the other one. Given this uncertainty, the most appropriate thing would be to conduct a randomized clinical trial comparing the two vaccines in a similar population of 12 year-old girls, in order to measure efficacy in terms of the number of HPV infections prevented. This way, the two vaccines would be compared head-to-head in the same scenario, thus reducing uncertainty in the analysis.

There are previous cost-effectiveness studies in the scientific literature comparing the two vaccines. In Ireland, a study used the assumption of 90% efficacy for the two vaccines, and the price of the two vaccines is the same in that country. Although the two vaccines were shown to be ‘very cost-effective’ alternatives according to the threshold already established by the Irish government, the quadrivalent vaccine showed the best cost-effectiveness profile. In order to have the same ICER for the quadrivalent vaccine, the bivalent vaccine had to be 22% less costly [25]. Similar results were obtained by Brisson et-al in a cost-effectiveness analysis for Canada. By assuming efficacy of 95% and cost of Canadian Dollars (CAND) 400 for both vaccines, once again the quadrivalent vaccine was deemed more ‘cost-effective’ than the bivalent (ICER = CAND21000/QALY and CAND31000/QALY respectively) [53]. Also, Jit et al found that in order to be cost-effective the bivalent vaccine should be cheaper considering its lack of protection against genital warts [24]. In all these studies, genital warts were accounted for not only in terms of treatment savings but also in terms of QALYS, so the result was to be expected considering the advantage of the quadrivalent vaccine. For the local Colombian context, a result of this type is completely plausible, even more when considering the uncertainty already described.

In view of the above, this analysis enabled us to clearly identify the relationship between screening and vaccination in terms of cost as well as effectiveness. However, in the analysis between vaccines, some sources of uncertainty were identified and, consequently, the results should be interpreted with caution. The model was developed using the best scientific evidence available, but there are weaknesses in knowledge that preclude more reliable conclusions.

To become a ‘cost-effective’ alternative for Colombia we found that a vaccine price reduction would be necessary. The prices considered in this study (US$ 188 for quadrivalent and US$ 214 for bivalent per vaccinated girl) should be reduced about 32-36% to consider these vaccines as cost-effective alternatives in the local context.

The results found here are consistent with cost-effectiveness analysis of HPV vaccination performed for other countries, even when vaccines have been found cost-effective in other scenarios. In a Canadian study with similar assumptions to this one, Brisson et-al reported an ICER of US$  21,000- 31,000/QALY [53]]. For UK, Jit et-al calculated an ICER of USD44,321/QALY[23], and in a review of cost-effectiveness analysis for USA, it was found that ICERs can reach USD24,300/QALY[54]. Even though in all these cases vaccines were classified as ‘likely to be cost-effective alternatives’, this result is opposite to the classification obtained in this study because of the difference in considered societal willingness-to-pay thresholds. For such developed countries, thresholds can reach values of US$  150,000 whereas for Colombia three times GDP per capita is about US$ 18,700. This difference, added to other ones in the design of the other studies such as including herd protection and measuring health results using QALYs, can explain why these results differ from studies previously published.

In addition, WHO’s proposed price of US$ 10-25 per vaccinated girl supports the idea of reducing the price [5], along with an analysis of Goldie et-al about HPV vaccination in Latin America, which suggests that for these countries, the prices of vaccination currently set in Europe and North America are not appropriate [52]. These prices are near the ones considered in the base case analysis of this study. A reduced price would be also consistent with prices of other vaccines in Colombia, like pneumococcal and tetanus vaccines, which are deemed ‘very cost-effective’ at prices per dose between USD1.4 and US$ 14 [49,50].

As in previous studies, it was found that when screening coverage is reduced, the impact of vaccination is greater [55]. In Colombia, it is reported that for 2010 61% of women between 18 and 25 years went for a Pap cytology every year, and 6% every three years. However, for Caribbean region within the country, coverage can drop to 47%. In Guajira department, for example, only 37% of women go yearly for Pap Cytology [7]. By the same token, the cost-effectiveness profile of the vaccines is improved when the sensitivity of the Pap cytology testing is reduced. Therefore, this may suggest greater impact of vaccination in regions of the country where the health systems are not well developed and, consequently, the performance and coverage of the cytology tests is low.

Additionally, it is important to mention that the model included the percentage of diagnosed women who underwent treatment, obtained from previous studies [56]. According to this, 86.4% of diagnosed women were treated in the model, which adjusted the analysis to Colombian situation in a closer way and permits the consideration of health system characteristics that could impact HPV vaccination efficacy.

One weakness of this work is the fact that it is a predictive theoretical model based on data reported in the scientific literature and not on specific clinical findings pertaining to CC, the natural history of the disease, or the associated prevention and promotion programs. A pilot trial of vaccination would allow considering other variables that are outside the scope of this work.

Additionally, it would be important to collect cost data not on the basis of a typical case but, better still, from a real-world study designed to follow these patients and to determine how much they and their families spend for the treatment of the disease.

Finally, three additional analyses which were not performed for this study should be discussed. The first one is male vaccination. Previous analyses have found that vaccinating boys in addition to men is unlikely to be ‘cost-effective’ [23,57], and considering also that this has not yet been approved for Colombia, such an analysis is not currently considered of major relevance for this health technology.

Secondly, the effect of HPV vaccination in Human Immunodeficiency Virus (HIV) or Acquired Immunodeficiency Syndrome (AIDS) population was not included here. Even though epidemic in Colombia is considered as concentrated and low [58], this analysis was not possible because of lack of information By the moment of this analysis, no assessment of safety or efficacy in immunocompromised population had been published. An initial consideration of different variables suggest that vaccination would be less effective in HIV/AIDS population, considering that distribution of HPV types is different in HIV-infected population. Since HPV16 is present in lower rates than in the rest of the population, and other types incidence is increased, vaccine efficacy can be reduced [59]. Also, it has been found that HPV related lesions are 5 times less likely to regress in HIV- positive women than in those without HIV [60]. However, the available information about efficacy in this subgroup of population was deemed as ‘not-enough’ to propose further assumptions in the model that could result in reliable conclusions. For this reason it is suggested that in further assessments of HPV vaccination for Colombia this variable could be included, if information is available.

On the other hand, herd protection is suggested to be included in further analysis, since this consideration can influence the final results. For example, Elbasha et-al described that if herd protection is not taken into account, along with protection against genital warts, the ICER can increase from USD 3,000/QALY to USD 214,000/QALY [9]. So, by considering herd protection in the local context, vaccination strategies could be potentially favored and different results could be expected. Given this, assumptions about HPV infection transmission should be closely evaluated to avoid results that are not representative of the actual epidemiologic and demographic conditions, leading to over- or underestimation of the real vaccination costs and effects.

Conclusions

Since 2009, the WHO has stated its position regarding HPV vaccination. It recommends including vaccination whenever the prevention of CC is a public health priority, as long as programs are viable, funding is guaranteed, and the cost-effectiveness profile is taken into consideration in the country, or at least in the region [5]. This work is an attempt to come closer to fulfill the latter premise. The cost-effectiveness profile of the two vaccines is now available from a societal perspective, based on the best information available to date for recreating the local scenario as carefully as possible. Even though it was found that vaccination may be classified as ‘non cost-effective’, threshold vaccine prices were calculated that would allow these vaccination programs to become cost effective. Consequently, any future decisions will depend on price considerations as well as on the analysis of the program and on the budgetary impacts.

Acknowledgments

We thank Dr. Monica Molano of “Grupo de Investigación en Biología del Cancer – Instituto Nacional de Cancerología” who provided essential data about HPV distribution among Colombian women.