Background

Cervical cancer poses one of the greatest challenges to women’s health. It is estimated that over a million women worldwide currently have cervical cancer, most of whom have not been diagnosed nor have access to treatment that could cure them or extend their survival [1]. The burden of cervical cancer is disproportionately borne by poorer countries. In 2012, 528,000 new cases of cervical cancer were diagnosed, and 266,000 women died of the disease, nearly 90 % of them in low- and middle-income countries (LMICs) [1]. It is expected that these numbers will double in the next 20 years due to aging and population growth [2].

The human papillomavirus (HPV) is the primary cause of cervical cancer and cervical intraepithelial neoplasia (CIN) [3]. HPV is typically transmitted in the cervix through microabrasions that may occur as a result of sexual intercourse [4]. Persistent infection with oncogenic strains of HPV causes cervical cancer [5]. Of the 40 HPV strains that affect the genital area, 15 of them are known to be oncogenic (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73, and 82) [6]. HPV infection has also been linked with other genital cancers (e.g., vaginal, vulvar, anal, and penile) as well as non-life-threatening diseases, such as genital warts [7]. HPV types 16 and 18 have been found to be responsible for about 70 % of all cervical cancer cases worldwide [8, 9]. In the remaining 30 % of cervical cancers, HPV types 31, 33, 35, 45, and 58 have all been implicated [10].

Cervical cancer can be prevented both by primary prevention through HPV vaccination and by secondary prevention through cervical cancer screening. Conventional cytology-based screening, combined with colposcopy and treatment of precursor lesions, has been the gold standard for the secondary prevention of cervical cancer. It has greatly reduced cervical cancer incidence and mortality from cervical cancer in many developed countries. Cytology-based screening is not as widespread in developing countries [10]. For instance, it is estimated that less than 5 % of women at risk of cervical cancer in sub-Saharan Africa have ever been screened [11].

Alternatives to cytology-based screening include visual inspection with acetic acid (VIA) and newer molecular testing for infection of the cervix with high-risk types of HPV DNA. VIA and HPV DNA testing have proven to be effective screening methods [12, 13]. HPV DNA testing has shown to be significantly more effective than VIA or cytology in reducing both cervical cancer precursors and cervical cancer [12, 13].

There currently are two HPV vaccines in widespread use. Both Cervarix® and Gardasil® offer protection against HPV 16 and 18, respectively, the two most oncogenic types [7]. Gardasil® also offers protection against HPV 6 and 11, which cause 90 % of genital warts [14]. The quadrivalent vaccine has also shown to protect against cancers of the anus, vagina, and vulva [14]. There is evidence indicating that the immune response against type 16 and 18 provides some cross-protection against types 45 and 31, both important in the aetiology of cervical cancer, thus potentially increasing the projected protection from vaccination to 75–80 % [10]. However, since prophylactic vaccination is not effective against infection from all 15 oncogenic HPV types, regular screening is still recommended among women that have received the vaccination [14].

Challenges of prevention and control strategies in low- and middle-income countries

The World Health Organization’s (WHO) recommendation for comprehensive cervical cancer prevention and control strategy includes primary, secondary, and tertiary prevention activities [1]. Primary prevention involves HPV vaccination of girls (and boys if affordable) between 9 to 13 years. In secondary prevention, women 30 years of age or older are to be “screened-and-treated” with low-cost technology, e.g., VIA followed by cryotherapy or HPV testing for high-risk HPV type. For tertiary prevention, all women with invasive cancer at any age are to be treated with ablative surgery, radiotherapy, or chemotherapy with palliative care as necessary. The recommendation suggests that the three prevention components be planned and implemented in combination with a structured national approach to community education and mobilization strategies and a national monitoring and evaluation system [1].

LMICs potentially face a number of challenges in implementing such a cervical cancer prevention and control program. Existing health services may not have the capacity to accommodate additional interventions, and thus, extra cost will be incurred to set up such a program. Resource constraints may necessitate implementation of a program in phases rather than immediate implementation. Resource constraints could also require the introduction of preventive and treatment services only for certain regions rather than country-wide. Sections of the population may systematically avoid participation in such a program. Given these potential challenges, it is therefore important to examine if and how they are accounted for by decision analytic models of HPV vaccines; for example, do analyses adopt a different modelling structure or just change parameters for the constraints mentioned above?

Data challenges in low- and middle-income countries

The data needed for modelling might be less readily available in LMICs or measured with much larger uncertainty. HPV-related outcome data in many LMICs may have poor resolution and are also often highly aggregated [15]. For example, incidence data of cervical cancer available from the WHO/Institut Català d’Oncologia Information Center on HPV and Cervical Cancer is stratified into the age groups 0–14, 15–44, 45–54, 55–64, and ≥65 years, and thus, simple natural history models may erroneously predict that 15-year-old women have the same cancer incidence as 44-year-old women, and this overestimates the proportion of cancers occurring in younger women and thus potentially biases estimates of vaccination cost-effectiveness [15]. These data challenges prompt the questions of how well they are recognized and overcome by the existing literature and which of these uncertain parameters have the greatest impact on the model outcomes. These are pertinent questions, and it is worthwhile examining how they have been handled in modelling studies of HPV vaccination.

Model type and herd immunity

There are three types of models that have been employed in cost-effectiveness analyses of HPV vaccination: static models, transmission dynamic models, and hybrid models combining features of both static and dynamic models [7]. A static model typically tracks progression of HPV disease for a single cohort over an expected lifetime [16]. Transmission dynamic models have the advantage of accounting for both the direct and indirect (i.e., herd immunity) benefits of vaccination in the population [16]. Thus, dynamic models account for the immunity that occurs when HPV vaccination of a significant portion of the population (or herd) provides a measure of protection for individuals who have not been vaccinated. A hybrid model is a combination of a cohort model and a dynamic model. It corrects the invariant incidence probability in a cohort model to a dynamic probability and thus does not ignore the indirect benefits of herd immunity for the cohort being simulated [16]. However, herd immunity depends on the rate of vaccine coverage. As stated previously, this coverage may vary dramatically in LMICs, both across time or across populations. It is therefore relevant to ask how this is accounted for in HPV vaccine cost-effectiveness studies.

Study rationale and objectives

Appraising the introduction of HPV vaccines requires estimation of the avertable disease burden, cost-effectiveness of the vaccine compared with alternative uses of the resources, affordability of the vaccine, feasibility of achieving high coverage, likelihood of public acceptability, and political support for vaccination [17]. The lack of data on the long-term effectiveness of HPV vaccination has prompted the development of various decision analytic models to guide policy makers by projecting the long-term epidemiologic and economic consequences of alternative vaccination policies [16]. For such analyses to provide a reliable guide to policy development and implementation, they should reflect the LMIC-specific challenges described above.

Two systematic reviews of cost-effectiveness analyses of HPV vaccination in LMICs have been published previously [18, 19]. These reviews discussed the cost-effectiveness estimates and investigated how they are affected by model characteristics and underlying assumptions in general. The focus of our proposed systematic review is to examine how the modelling studies accounted for the challenges specific to low- and middle-income countries. The review seeks to answer the following questions:

  • Does the existing literature on cost-effectiveness modelling of HPV vaccine acknowledge the particular challenges facing LMICs?

  • How were particular challenges accommodated in the models, e.g., through a different model structure or just by varying parameters?

  • Is the uncertainty among the less readily available essential data/parameters regarding LMICs so large that the model-based recommendation is affected?

  • Does the choice of modelling herd immunity influence model-based recommendations, and in particular, does imperfect HPV vaccination coverage influence model-based recommendations?

Methods/design

Protocol

This protocol adheres to the Preferred Reporting Items in Systematic Reviews and Meta-analyses (PRISMA) statement [20]. The protocol is registered in the International Prospective Register of Systematic Reviews (PROSPERO) CRD42015017870.

Eligibility criteria

Inclusion criteria are as follows:

  • Studies based on decision analytical models of HPV vaccination;

  • Studies that considered the cost-effectiveness of HPV vaccination and reported the additional cost and additional health effects in terms of life years gained (LYGs), quality-adjusted life years (QALYs), or disability-adjusted life year (DALYs);

  • LMICs as identified by the World Bank classification of income groups [21];

  • Both single- and multi-country studies;

  • The review will include both original research papers and reviews (inclusion of the latter is to ensure that no original study was missed);

  • Studies to be included in the review could be published in any language;

  • Studies published since 2006.

Information sources

We will search MEDLINE (via PubMed), EMBASE, NHS Economic Evaluation Database (NHS EED), EconLit, Web of Science, and Tufts CEA Registry. For existing systematic reviews, we will search Cochrane Reviews, Cochrane Database of Abstracts of Reviews of Effects (DARE), and Cochrane Health Technology Assessment Databases. Reviews will be included to reduce the possibility of missing an individual study.

Search strategy

A search strategy will be developed for each of the databases. The “Appendix” section provides details of our planned bibliographic database search strategies for MEDLINE (via PubMed), EMBASE, CINAHL, Cochrane Reviews/Cochrane DARE/NHS EED, EconLit, Web of Science, and CEA Registry. The reference lists of all included and relevant articles identified during the search will be reviewed to identify further studies that were missed. Furthermore, we will use the PubMed “related articles” feature. Hand searching of a selection of relevant journals will be conducted as advised by experts in economic evaluation.

Study selection

Titles and abstracts will be screened for inclusion independently by two of the reviewers using the eligibility criteria. Opinion of a third reviewer will be sought to arrive at a consensus in case of disagreement on a study for inclusion.

Data extraction

Data will be extracted independently by two of the reviewers from included studies using a predefined data extraction spreadsheet (Table 1). Data to be extracted will be arranged into the following classes: model characteristics, base-case assumptions, results, sensitivity/uncertainty analyses, data sources, and miscellaneous (conflict of interests and factors not considered). Cost presented in different currencies will be adjusted to 2013 value using consumer price index. Afterwards, cost data will be converted to international dollar units using purchasing power parities (PPPs). Authors of various studies may be contacted to clarify methods and results if the need arises.

Table 1 Relevant data extraction information
Full size table

Risk of bias and data synthesis

One of the investigators will assess the validity of included studies using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement [22]. Details of the CHEERS statement are summarized in Table 2. Descriptive, narrative, and interpretative synthesis of data will be undertaken to address the study objectives [23]. The WHO Commission on Macroeconomics and Health will be used to determine the thresholds of cost-effectiveness such that an intervention will be considered “very cost effective” and “cost effective” when its incremental cost-effectiveness ratio is less than the gross domestic product (GDP) per capita and less than three times GDP per capita, respectively [23].

Table 2 CHEERS statement for checking the validity of included studies [22]
Full size table

Discussion

This protocol describes a systematic review for studies on cost-effectiveness of prophylactic HPV vaccination in LMICs. The aim is to assess how cost-effectiveness studies of HPV vaccine accounted for the individual challenges of low- and middle-income countries. The gaps identified in this systematic review will expose areas for additional research as well as challenges that need to be accounted for in future modelling studies. The review will also present current data on cost-effectiveness of HPV vaccines in LMICs as newer studies have been carried out since the publication of the last reviews.