Article Text

Original article
Evaluating societal preferences for human papillomavirus vaccine and cervical smear test screening programme
1. Bridgette Oteng1,
2. Fawziah Marra1,2,
3. Larry D Lynd1,3,
4. Gina Ogilvie1,2,
5. David Patrick1,2,
6. Carlo A Marra1,3
1. 1University of British Columbia, Vancouver, British Columbia, Canada
2. 2British Columbia Centre for Disease Control, Vancouver, British Columbia, Canada
3. 3Centre for Health Evaluation and Outcome Sciences, St Paul's Hospital, Vancouver, British Columbia, Canada
1. Correspondence to Dr Carlo Marra, Faculty of Pharmaceutical Sciences, University of British Columbia, 2146 East Mall, Vancouver V5Z163, British Columbia, Canada; carlo.marra{at}bccdc.ca

## Abstract

Background Cervical cancer and genital warts are diseases associated with human papillomavirus (HPV) infection. Cervical smear testing is used as a cervical cancer screening tool in most countries worldwide. The newly introduced vaccines that prevent HPV infections are the quadrivalent vaccine (Gardasil), which prevents genital warts and cervical cancer, and the bivalent vaccine (Cervarix), which prevents cervical cancer only. Public preferences for HPV vaccines and smear test screening were determined using a discrete choice experiment.

Methods Participants from across Canada completed a choice-based questionnaire to measure preferences from which willingness to pay (WTP) was calculated for the following: (1) lifetime risk of cervical cancer and genital warts, (2) frequency of smear testing, (3) need for vaccine booster, (4) target group to vaccinate, (5) frequency of side effects and (6) cost of the vaccine (from 2008). A mixed effect logistic model was used to analyse the data.

Results Of the 1157 participants, the mean age was 44 years (SD 15) and 49% were women. Respondents preferred a vaccine that gave lifelong immunity, a vaccination programme that targeted boys and girls and a vaccine that gave protection from genital warts and cervical cancer. Respondents were averse to yearly smear testing. On average, respondents were willing to pay $C53 and$C22 to avoid a 1% increase in the risk of cervical cancer and genital warts, respectively.

Conclusions Society agrees with the introduction of the HPV vaccination programme, but would prefer a programme that targets boys and girls with the quadrivalent vaccine.

• Health Outcomes
• HPV
• human papilloma virus
• vaccines
• discrete choice experiment
• societal preferences
• cervical cancer
• vaccination
• viral vaccines

## Introduction

In Canada, cervical cancer is the second most commonly diagnosed cancer in women aged 20–44 years.1 Each year, approximately 1400 women are diagnosed as having cervical cancer and 400 die from this disease.1 The primary strategy for preventing cervical cancer is screening of asymptomatic women via the cervical smear test.2 Over the past 50 years, this approach has resulted in significant decreases in the incidence and death rate.3 4 However, in the past decade the rates of cervical cancer have reached a plateau at an incidence level of approximately 7.7 per 100 000 population,3 which has been attributed to decreased compliance with the smear test screening programme in those who are less educated or from low socioeconomic status backgrounds,5 living in rural or remote areas,6 and being a recent immigrant, aboriginal, African–American or Hispanic.7–9 In addition, since the sensitivity and specificity of the smear test is not 100%, women can develop cervical cancer despite compliance with the screening programme.6

Human papillomavirus (HPV) represents a large family of viruses with approximately 40 subtypes, which are acquired through sexual contact. The human papillomavirus is highly transmissible and the majority of women acquire HPV within 2–5 years of initiating sexual activity10; however, 90% of those women will then clear the virus within 2 years of acquisition.11 12 Women who do not clear HPV are at risk for developing persistent infections, potentially precancerous cervical lesions and cervical cancer.13 Of the 40 mucosal HPV types, 15 are classified as ‘high risk’ and are a major risk factor for cervical cancer (ie, HPV DNA has been detected in up to 99.7% of all cervical cancers).14 Of the high risk HPV subtypes types, infection with one of two subtypes (types 16 and 18) accounts for roughly 70% of cervical cancers diagnosed each year.15

Gardasil (Merck Frosst, Kirkland, Quebec, Canada) is a quadrivalent vaccine containing HPV types 6, 11, 16 and 18. In most parts of the world, it is licensed for girls/women 9–26 years of age for the prevention of cervical cancer caused by HPV types 16 and 18 and genital warts, which are caused by HPV types 6 and 11.16 In the phase II/III clinical trials, this vaccine was shown to have 100% efficacy against the development of cervical intraepithelial neoplasia 2/3, vulval intraepithelial neoplasia 2/3, vaginal intraepithelial neoplasia (VaIN) 2/3 and adenocarcinoma in situ.17–19 A second vaccine against HPV, Cervarix (GlaxoSmithKline, Mississauga, Ontario, Canada), has also been approved in many parts of the world. In contrast to Gardasil, this is a bivalent vaccine containing HPV types 16 and 18 only and thus only gives protection against cervical cancer and not genital warts.

The type of vaccine, age at which the vaccine is administered and who is vaccinated varies from country to country. In Canada, Gardasil is available through publicly funded, school-based programmes to girls only, while other countries have adopted programmes using Cervarix and/or programmes aimed at different target groups (for example, boys and girls). We chose to use discrete choice experiment (DCE), which is a stated preference technique used to establish consumer preference for our study.20 Although originally used in marketing, DCE has been widely adopted in health economics to determine preferences for interventions and health technologies and to characterise how people trade off risks and benefits among competing programmes.21 With limited resources but increasing availability of new programmes/technologies and services, DCE serves as a useful tool in ascertaining preferences in the allocation of these scarce resources. Through focus groups and extensive literature search, important attributes and levels used in DCE can be generated. The attributes are constructed as a set of hypothetical scenarios and respondents are asked to choose their best scenario or opt for the neither option. The objective of our study was to evaluate societal preferences for the different vaccination and screening strategies, identify the properties of the vaccine that are important and to determine the willingness to pay (using 2008 costs) for these attributes.

## Methods

### Questionnaire design

A DCE questionnaire was designed consisting of seven attributes, each with three or four levels (see supplementary material). Each questionnaire had 12 choice sets and each choice set consisted of 3 hypothetical options (option A, option B and neither) (see supplementary material). Although hypothetical options were used, the actual preference associated with a specific strategy was determined based on the levels associated with the strategy. Each attribute was selected based on the current vaccination and screening policy, literature reviews and a survey of parental intention to have their daughters receive the HPV vaccine.22 Policy and clinical experts in infectious diseases and immunisation also contributed to the selection of the attributes. Sawtooth CBC/SSI Web V.6.4.2 (Sawtooth Software, Sequim, Washington, USA) was used to design a choice-based fractional factorial experiment, where each choice set had three options (A, B or neither). The questionnaires were designed to ensure orthogonality (ie, minimal correlation between attributes), minimal overlap (each attribute level in a survey appears only once in a choice task) and level balance (attribute levels occur at an equal frequency within a questionnaire). In all, 10 versions of the questionnaire were generated with each version having 10 choice sets plus 2 choice sets as a consistency check. The consistency check choice sets had one option that was clearly dominant (greater likelihood of benefits and greater chance of risks) and were included to identify respondents who failed to understand the questions. The questionnaire was piloted on 300 participants to assess comprehension and ease of completion.

### Recruitment and study sample

Participants were recruited by Ipsos Reid (Vancouver, British Columbia, Canada) to ensure that they were representative of the Canadian population. Individuals were selected from a balanced sample in terms of sociodemographics such as age, gender, income, education and region. These participants were derived from the Ipsos I-Say panel of approximately 300 000 Canadian residents who have agreed to participate in surveys. Respondents who were 19 years or older, currently residing in Canada, and fluent in speaking and writing English were invited to visit a website where the DCE questionnaire was located. Apart from the DCE data, sociodemographics, information on vaccine practices and personal or relatives' history of HPV-related diseases was also obtained from the respondents. The Behavioural Research Ethics Board of the University of British Columbia approved the study protocol.

### Statistical analysis

Descriptive analyses were performed to characterise the sample according to sociodemographic status and vaccination practices. A mixed effects logistic model was used to analyse the DCE responses. SAS (V9.1, SAS Institute, Cary, North Carolina, USA) and Matlab code23 were used to analyse the data. The preference for a specific vaccination and screening programme was calculated using a formula by Hall et al 24 and determined based on the sum of the preference weights specific to the level that pertains to that specific option. The willingness to pay (WTP: the average amount a respondent is willing to pay to avoid or get an attribute level) and willingness to trade (WTT: the point at which respondents are indifferent between two attributes) were also calculated.25 Segmentation analyses were carried out to evaluate preferences across the various subgroups. Statistical significance was defined as a p value of <0.05.

Cost, lifetime risk of cervical cancer and genital warts (as a result of the programme being chosen by the client), and frequency of side effects were considered as continuous variables and modelled as fixed parameters. The need for vaccine booster, target group to vaccinate and frequency of smear testing were considered as categorical variables and were modelled as random parameters with normal distributions.

## Results

A total of 1275 respondents completed the questionnaire; of them, 1157 (91%) chose a dominant option in 1 of the consistency check choice sets and thus were included in the final analysis (table 1). There were no significant differences in sociodemographic factors between those included and excluded from the analysis (data not shown). The average age of respondents was 44 years (SD=15), of whom 569 (49%) were women. More than three-quarters of the respondents (79%) had completed at least a high school education and 703 (61%) earned an individual income >$C55 000/year. The majority of respondents were from two-parent households (a family consisting of parents and children). In all, 532 (46%) had children, and the children of 458 (86%) of these had received all childhood vaccines. In total, 817 (70%) said they would vaccinate their children against HPV. Table 1 Baseline characteristics All the parameter estimates were found to be significant predictors of choice except need for vaccine booster every 10 years (p=0.85), yearly (p=0.51) and every 5 years (p=0.36), smear testing and never having a smear test (p=0.35) (table 2). In general, respondents were averse to requiring a vaccine booster every 5 years and preferred never requiring a vaccine booster, and they also preferred every 3 years of smear testing. The strong aversion was for vaccinating neither girls nor boys, indicating their desire to have an HPV vaccination programme, but they had a stronger preference for vaccinating girls and boys relative to just vaccinating girls. The four continuous variables were all statistically significant but had negative effects on preference. Thus, preference for vaccination and screening decreased as cost, the risk of cervical cancer, the risk of genital warts and the frequency of vaccine side effects increased. The model revealed heterogeneity in respondents' preferences for all levels of attributes except for the need for vaccine booster every 5 years (p=0.63) and never having smear testing (p=0.90). Table 2 Parameter estimates for mixed effect logistic model On average, the respondents were willing to pay$C53 and $C22 to avoid a 1% increase in the risk of cervical cancer and 1% increase in genital warts, respectively (table 3). Since vaccinating neither girls nor boys had the lowest negative preference weight, that is the respondents wanted an HPV programme, they would have to theoretically be compensated with$C463 (ie, WTP -$C463) if a jurisdiction chose not to have an HPV programme. Respondents had a mean WTP of$C303 to vaccinate girls and boys. The determination of WTT also revealed that, on average, respondents would (be willing to) likely accept a 2.4% and 4.6% increase in genital warts risk and adverse events, respectively in order to avoid a 1% increase in the risk of cervical cancer (table 4).

Table 3

Estimate of willingness to pay (WTP)

Table 4

Expected utilities for possible HPV vaccination and screening strategies

To determine the value society places on the quadrivalent vaccine (protection against genital warts and cervical cancer) versus bivalent vaccine (cervical cancer protection only), the following assumptions were made: smear testing will be required every 3 years, a vaccine booster is not required, and girls and boys will be vaccinated with a 6% frequency of side effects and no out-of-pocket costs, a 70% cervical cancer risk (lifetime) reduction with both vaccines and 90% genital warts risk (lifetime) reduction with the quadrivalent vaccine. Based on the above assumptions, the expected utility for the no vaccine option was −2.13, with 11% of all respondents choosing this option. The expected utilities were 0.80 and 0.18 for the quadrivalent and bivalent vaccinations, respectively, with a 0.69 chance of the former option been chosen and 0.54 chance of the latter option been chosen (table 4).

The secondary analyses looked at preferences across the different demographic variables (table 5). Across the subgroups shown in table 5, respondents had the strongest aversion to vaccinating neither girls nor boys. The lifetime risk of genital warts was also statistically significant across all subgroups, suggesting that respondents would prefer a vaccination programme that would prevent this infection. While women had a negative preference for never having a smear test, men had a positive preference for their female family member receiving smear testing every 3 years. Generally, respondents across the different subgroups were indifferent about frequency of smear testing, as the parameter estimates for all levels of this attribute were mostly insignificant. Those who considered the need for a vaccine booster in their decision making either disliked having vaccine booster every 5 years or had a significant positive preference for never having a vaccine booster. Respondents across the various groups were averse to increasing vaccine cost, increased risk of cervical cancer and genital warts, and more frequent side effects. Thus, preference for a vaccination and screening programme decreased as the cost and the risk of genital warts, cervical cancer and side effects increased. Respondents who would not vaccinate their children had a significantly negative preference for vaccinating girls only and a significantly positive preference for vaccinating neither girls nor boys. They also had an insignificant negative preference for vaccinating girls and boys (figure 1). The results from this subgroup serve as evidence of construct validity for the study.

Table 5

Overview of the subgroup analyses

Figure 1

Density plot of the distribution of target group to vaccinate for those who would (not) vaccinate their child against human papillomavirus (HPV).

## Discussion

Our study respondents were in favour of an HPV vaccination and screening programme. The target group to vaccinate was the most important attribute, with respondents preferring a vaccination programme that includes girls and boys, rather than girls only. These findings are in line with other studies using survey-based methods, which have indicated that parents would like to have an HPV vaccine administered to boys and girls since immunising boys against HPV will protect future partners and reduce disease transmission.26–28 While their reasons for wanting the vaccine for boys may also be equity related, most cost effectiveness analyses of HPV vaccine have shown that it is more cost effective to vaccinate girls only than to vaccinate both genders.29–32

Our study showed a significantly positive preference for never requiring a vaccine booster dose; that is, respondents preferred a vaccine that would give lifelong immunity. Respondents were more concerned about being protected against cervical cancer than genital warts, which is likely because genital warts are non-life threatening and less than 1% of those infected develop clinically obvious warts.33 We saw a positive preference for a vaccination strategy provided by the government and therefore provided without any out-of-pocket cost. However, the quadrivalent vaccination strategy was preferred to the bivalent vaccination strategy, which suggests that protection against genital warts and cervical cancer are important and that respondents were, on average, willing to pay for the additional protection. Since the quadrivalent vaccine comes at a higher cost, decision makers will need to decide if the extra preference obtained from this vaccine is worth this additional cost.

A limitation of a DCE is the concern that participants may not truly understand the question, given the hypothetical nature of the choices and the need to make a decision while considering multiple criteria. However, this limitation is minimised by ensuring that the instructions on how to answer the DCE questionnaire are clear and concise, and measures are put in place to test the understanding of the study participants with regards to the DCE methodology (ie, there is significant pilot testing in advance of releasing the questionnaire). For this study however, this did not appear to be a concern given that 91% of the dominant question were answered correctly, suggesting an understanding of the task, and all preference weights across levels of specific attributes were in the a priori hypothesised direction, which provides evidence of the face validity of the model and the preference weights. Another limitation of a DCE is that it is a ‘stated preference’ technique as opposed to a ‘revealed preference’ technique. Stated preference only requires respondents to make decisions based on how they think they would choose, whereas revealed preference studies actually observe the behaviour of individuals to determine exactly what they would choose when given a choice. This limitation is not specific to this particular study, but rather is a limitation of the DCE technique. Although evaluating societally revealed preferences would be preferable, this is much more difficult and not possible due to the high associated cost. As a result, a common assumption of stated preference techniques is that participants would actually choose the option that they state they would choose if presented with those options in reality.

Given that the study participants were recruited from a panel of respondents who actively participate in surveys makes the study results vulnerable to selection biases. For example, volunteer bias is an error that occurs as a result of low response rate because certain groups of people (usually healthier, younger and well educated) tend to have a high participation rate than others. This effect can likely compromise the interpretation and limit the generalisation of the research findings.34 Furthermore, the sample is likely to under-represent those women less likely to take up smear tests and may overestimate the acceptability of smear tests. Given these factors and the high level of support for vaccines (and specifically the HPV vaccine) indicated by the data in table 1, it is possible that we overestimated the population's preferences for the HPV vaccine.

### Conclusions

Our study sample revealed their preference for a vaccination and screening programme, they were willing to pay for it, and also value the additional protection against genital warts. This study also makes some important contributions to the current literature on application of discrete choice experiment in health and addresses the gap in the literature concerning the public's preference for the HPV vaccines and the aspects of the vaccine they consider important. Through DCE, this study has shown that the public generally has a positive preference for the HPV vaccination programmes and has a stronger preference for the quadrivalent vaccine than the bivalent vaccine. Furthermore, it has shown that sociodemographic information and previous vaccine practices can be used to identify subgroups in the population that respond differently to the various attributes and levels. This permits programmes to be targeted more specifically. Our study has demonstrated DCEs can predict relative preferences for a health technology and potentially predict the uptake of the HPV vaccines.

### Key messages

• The incidence and prevalence of anogenital warts in Canada appears to be similar to other jurisdictions.

• The costs associated with the diagnosis and treatment of anogenital warts in Canada are significant.

• Treatment of anogenital warts was mostly performed with ablative treatment rather than pharmaceutical products.

• These data may help inform decision makers regarding the cost effectiveness of vaccination and screening strategies for human papilloma virus.

## Acknowledgments

LDL and CM are Michael Smith Foundation for Health Research Scholars. LDL is a Canadian Institute of Health Research New Investigator. CM is a Government of Canada Research Chair in Pharmaceutical Outcomes.

• ## Web Only Data sti.2009.041392

Files in this Data Supplement:

• ## Web Only Data sti.2009.041392

Files in this Data Supplement:

## Footnotes

• Competing interests None declared.

• Ethics approval This study was conducted with the approval of the University of British Columbia Behavioral Ethics Committee.

• Provenance and peer review Not commissioned; externally peer reviewed.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.