We adapted a previously developed HPV dynamic mathematical model to Estonia (Elbasha & Dasbach, 2010: [10]). Details of the model structure and equations have been published previously [10]. Individuals enter the model as they are born, move between successive age groups at an age- and gender-specific rate per year, and exit the model as they die. The model estimates health benefits and costs in a dynamic population. The model also estimates the impact of vaccination on vaccinees and their contacts (via herd immunity impact).
Demographic and epidemiological model
The model simulated aging and all-cause mortality over time within the Estonian population. The model simulated the transmission of HPV infection within the population as determined by the course of sexual mixing, a feature which allows for estimating both the direct and indirect (i.e. herd immunity) benefits of vaccination. Hence, the model required inputs on sexual activity risk groups in the population.
Vaccination and screening strategies
In the model, it is assumed that vaccination occurs prior to sexual debut and would consist of the three recommended doses, and the vaccination would confer type-specific protection. Other parameters subjected to greater uncertainty such as vaccination coverage and duration of protection were further explored in the sensitivity analysis. The model incorporates vaccine efficacy from the most recent clinical trials. The prophylactic vaccine efficacy against transient HPV 6, 11, 16 and 18 infections was assumed to be 76.1%, 76.1%, 76.0%, and 96.3%, respectively (Merck & Co., Inc. Unpublished data 2009). The vaccine efficacy against persistent HPV 16 and 18 infections was assumed to be 98.8% and 98.4%, respectively.
In both the vaccinated and unvaccinated cohorts, screening was modelled according to the actual current/routine Estonian cervical cancer screening practice.
Model inputs: values and sources
The model requires input values for demographic, behavioural, epidemiological, screening, treatment, vaccine, and economic parameters (see Additional file 1: Tables A1-A13).
When available, we used data from Estonia as inputs, otherwise we used the default data in the model published for the US [10] (inputs not presented in this report are available in the online supplement of Elbasha & Dasbach, 2010 [10]).
The Estonian Health Insurance Fund (HIF) database was used as the source for health care utilization data, HPV disease occurrence and health care costs (see Additional file 1: Tables A1-A13). Health insurance in Estonia is funded through a compulsory scheme under which employers are obliged by law to pay social and health insurance taxes for their employees. Self-employed people pay a social tax based on their income. Individuals whose social/health insurance tax is paid by their employer or who pay it themselves are considered to be covered by health insurance ("the insured") and are members of the HIF. As of 31 December 2008, 1,281,718 people were registered as insured by the HIF representing 95.6% of the Estonian population [11]. The HIF database is a “reimbursement database” containing information on ambulatory and in-patient/hospital care as well as reimbursed pharmaceuticals. As HIF reimburses health care providers on a fee-for-service basis and, the database is considered to be relatively complete.
Data sources:
-
1)
Demographic data (population size and composition, age and gender specific mortality) was taken from Statistics Estonia (Statistics Estonia, [12]) (Table A1);
-
2)
Sexual behaviour data was derived from the Estonian Health Interview Survey 2006 [13]. Estimates on the mean number of sexual partners per year by gender and age group were used and categorised as: low (0–1 per year), medium (2–4 per year), and high (5+ per year). The parameter was modified for model calibration purposes. The data on mean partners by age categories were changed so that the model projected a close estimate of the observed cervical cancer incidence and deaths in the Estonian population (Tables A3, A4 and section on Validation Analyses for more details);
-
3)
The age and stage-specific cervical cancer mortality was assumed to be the same as in the US [14] (Table A2);
-
4)
Screening data (cervical cancer screening rate by age group, % per year, proportion of women with a follow-up screening following an abnormal PAP) (Table A5);
-
5)
Treatment variables (Table A5):
-
(i)
proportion of women with cervical cancer who develop symptoms and seek care, by cancer stage – Estonian cancer registry;
-
(ii)
proportion of CIN/carcinoma in situ (CIS) treated, by stage – expert opinion and literature [10]. Based on local data (expert opinion), it was assumed that 50% of diagnosed CIN1 cases and 100% of diagnosed CIN2/CIN3 cases receive treatment; and
-
(iii)
hysterectomy rates by age group, % per year – based on the data from HIF;
-
6)
Economic variables:
-
(i)
direct medical costs of interventions were estimated using the national tariffs of the HIF for 2011 were reported in euros. Euro is the national currency in Estonia.
-
(ii)
costs of diagnosing and treating HPV disease (genital warts, cervical cancer screening and visit, colposcopy, biopsy, CIN 1,2,3 episode-of-care, cervical cancer (local, regional, distant) episode of care) – data from HIF (Table A6; A8—A13);
-
(iii)
vaccine cost: A cost of €59.00 per dose for the HPV4 vaccine was used, based on assumptions made in an earlier published analysis for Estonia [15] (Table A4)
-
7)
Vaccine strategy variables are based on the following assumptions and data sources:
-
(ii)
(i) HPV4 coverage of 85% of females age 12. This was set slightly below the 91.2% reported in 2010 for coverage with two doses of measles, mumps, and rubella vaccine (MMR), the second dose of which is administered to 13- to 14-year-old adolescents within the state calendar vaccination programme and via the school health system [16];
-
(ii)
100% adherence with the 3-dose regimen;
-
(iii)
Vaccine will be delivered through the existing school-based delivery system, similar to the current MMR regimen;
-
(iv)
Lifetime duration of protection was assumed in base scenarios. Effects of different lengths of duration were tested in a sensitivity analysis.
Cost effectiveness analysis
To calculate the cost-effectiveness of the vaccination strategy in preventing disease with respect to costs, we used the total discounted costs and effects (i.e. quality-adjusted life years, QALYs) accrued over a 100-year period with and without vaccination. In addition, the incremental costs incurred to achieve the incremental benefits by vaccination were calculated and the ratio of the incremental costs to incremental QALYs gained (i.e., the incremental cost-effectiveness ratio, ICER) are presented.
QALYs were estimated based on health utilities from the U.S. (Table A7). In addition to utilities for HPV disease states, age and gender-specific utility weights were also incorporated for individuals without HPV disease to account for the impact of co-morbid conditions (Table A7).
Both costs and medical outcomes were discounted at an annual rate of 3% [17, 18].
Model validation
As described above, the model was calibrated for Estonia using sexual activity and cancer detection rate parameters. We assessed the predictive validity of the model by comparing model predictions with observed data on the age specific incidence of cervical cancer and cervical cancer deaths in Estonia.
Sensitivity analyses
The cost effectiveness analysis is based on a number of assumptions. Due to uncertainty in some of these assumptions, several one-way sensitivity analyses were carried out. The parameters included duration of vaccine protection (20 years), vaccine coverage rates (70%, 95%), HPV disease diagnosis and treatment costs (+/−20%), vaccine cost per dose (+/−10%), discount rate (5%), and QALY weights. We also examined a scenario assuming no quality of life adjustments (cost per life years saved). Additionally, we examined the impact of HPV 6 and 11 protection in HPV4 on cost-effectiveness by running a scenario without the HPV 6 and 11 protection.
We also examined the cost-effectiveness of HPV vaccination under a hypothetical improved screening program. This hypothetical program was assumed to have immediate coverage of 95% (compared with the current/routine coverage of 72%) and the same age-specific annual rate as the base case scenario. Hypothetical improved screening independently gradually reduced the HPV16/18-related cervical cancer incidence to ~6 per 100,000 per year (age-standardized) at the steady-state in the model, about 100 years after screening had been initiated in the population. This steady-state cervical cancer incidence corresponded to ~ 8 per 100,000 per year of any HPV type (based on the assumption of 76% contribution of HPV 16/18 to cervical cancer; [6]) and mirrored the current cervical cancer incidence observed in the Western Europe [19].