This study used a previously published static Markov model that reproduced the natural history of oncogenic HPV with a one year cycle length [30
]. Previous versions of this model simulated CIN1, CIN2/3, and cancer associated with oncogenic HPV types. This updated version includes non-oncogenic (low-risk) HPV infections, CIN1 disease due to non-oncogenic HPV infections, and genital warts (Figure ) [40
]. The model vaccine efficacy calculations, detailed by Debicki and colleagues [32
], allow specification of the proportion of HPV types −16 or −18, non-vaccine oncogenic HPV types and HPV types −6 and −11 within all lesions, as well as vaccine efficacy by infection and lesion type.
Figure 1 Overview of the cohort model. NoHPVonc: Women in this health state have no oncogenic HPV infection. HPVlr: Women in this health state have a low risk (non-oncogenic) HPV infection. HPVonc: Women in this health state have an oncogenic HPV infection. CIN1onc: (more ...)
The model was developed in Microsoft®
Excel 2007 and simulates the effect of adding vaccination to the current screening program, where two cohorts of 100,000 12-year-old females were followed over a lifetime, one cohort vaccinated with the HPV-16/18 AS04-adjuvanted vaccine and the other with the HPV-6/11/16/18 vaccine. The model was previously calibrated to reproduce Canadian cervical cancer incidence and mortality, while keeping transition probabilities within pre-determined ranges [30
]. The incidence of cervical cancer has been decreasing in older age groups, so the model calibration was updated as shown in Figure to reflect the latest Canadian cervical cancer incidence data [43
] and published genital warts incidence data [44
Figure 2 Validation of the model. Panel A shows the model predicted cervical cancer incidence per 100,000 women compared to reported data from Canada in 2007 . Panel B shows the model predicted genital warts incidence compared to data reported by Marra et (more ...)
Table summarizes the base case model inputs. The model was parameterized using Canadian-specific screening, economic, and epidemiological data where available, as well as expert opinion. All events were costed from the perspective of the health care system in 2006 dollars. The incidence and cost per case of genital warts was obtained from analysis of an administrative databases in British Columbia, Canada [44
]. The remaining data comes from the previously published Canadian analysis [30
], with the exception of the distribution of HPV types within the health states displayed in Table . These distributions were updated to reflect a recent World Health Organization (WHO) review of Canadian data [45
]. As vaccine coverage rates do not impact the cost-effectiveness ratios estimated with a static model, a coverage rate of 100% was assumed.
Key base case cost-effectiveness model inputs
Assumed HPV distributions for each model health state for the base case and sensitivity analyses
A 98% vaccine efficacy against HPV types 16 and 18 was used based on the latest results from each vaccine’s clinical trials [25
]. For the HPV-16/18 AS04-adjuvanted vaccine, 47.7% (96.1% CI: 28.9- 61.9) and 68.4% (96.1% CI: 45.7-82.4) cross-protective efficacy against CIN1+ and CIN2+, respectively, was demonstrated for the 10 most frequent non-vaccine oncogenic HPV types −31, -33, -35, -39, -45, -51, -52, -56, -58, and −59 in clinical trials [25
]. For the HPV-6/11/16/18 vaccine, 23.4% (95% CI: 7.8 to 36.4) and 32.5% (95% CI: 6.0-51.9) cross-protective efficacy against CIN1+ and CIN2+ respectively for the same 10 non-vaccine oncogenic HPV types [27
] was demonstrated in clinical trials. These efficacy values come from two independent studies that report efficacy values for an HPV-naïve population in a similar manner for the same HPV types. In this analysis, the observed reduction in CIN1+ was input as efficacy against CIN1, while the observed reduction in CIN2+ was input as efficacy against CIN2/3 and cervical cancer outcomes. A 98% vaccine efficacy against HPV types −6 and −11 was assumed for the HPV-6/11/16/18 vaccine based on clinical trial data [49
]. This analysis assumed life-long protection against all HPV types, including cross-protection against non-vaccine types. Both of the vaccines were assumed to cost $100 per dose plus an administration fee of $10.97 [67
], and it was assumed that 3 doses were given for both [28
]. For the cost-effectiveness analysis, it was assumed that the entire cohort received all doses (100% coverage). Since this model is a static cohort model, it does not estimate the impact of reduced transmission of virus from women to men and any accompanying indirect benefit to men.
The lifetime number of CIN lesions, cervical cancer cases, cervical cancer related deaths, genital warts cases, QALYs, and costs were determined for each cohort. An ICER (cost per QALY gained) was calculated to compare the costs and outcomes of the two vaccines. A discount rate of 3% was applied to both costs and outcomes.
A number of sensitivity analyses were conducted by varying inputs assumed to impact the relative value of the two vaccines. The HPV-6/11/16/18 vaccine’s price per vaccine dose was varied to determine the point at which both vaccines would be predicted to have equivalent lifetime costs to the health care system. In a series of two-way sensitivity analyses, the efficacy of each vaccine against CIN2+ outcomes associated with non-vaccine oncogenic types was varied using the confidence intervals from the clinical trials. In addition, alternative WHO data for the continent of North America was used to populate the distribution of HPV types within cervical health states [68
]. The overall impact of genital warts was tested with a series of one way sensitivity analyses varying the costs of genital warts (±25%), the quality of life impact (measured as a utility decrement) of genital warts (±25%; QALY decrement increased from 0.02 in base case to 0.041 [Maximum decrement observed in recent publications] [70
]), and the incidence of genital warts (±10%; ±25%). The proportion of genital warts attributed to HPV types - 6 and - 11 was increased to 90% from the base case of 76%. Finally, simulation studies with transmission models have estimated that protecting females from HPV-6 / -11 may also reduce infection levels in males by as much as 90% over the next 70 years due to herd immunity [72
]. Consistent with those predictions, Donovan et al. report an observed decrease in genital warts of 39% (95% CI 33–46; p trend <0.0001) in males aged 12 – 26 years in Australia since implementation of the quadrivalent vaccine [73
]. As there has been no corresponding decrease in older males and very few males have received the vaccine, the authors attribute this decrease in young males to protection via herd immunity. In Sweden, however, Leval et al. reported decrease in genital warts amongst women but no decrease in men since implementation of an opportunistic program for females [74
]. Although the model used for the current assessment includes females only, the impact of herd immunity on males was simulated by increasing overall genital warts incidence by multiplier of 2.0 or 2.2 and assuming quadrivalent vaccine efficacy to range from 20% to 90% of the cases that would normally be seen in men.
Multivariate probabilistic sensitivity analyses were conducted to explore the combined effect of parameter uncertainty using @Risk software (Palisade Corporation, Ithaca, New York, USA). Distributions were assigned to transition probabilities, vaccine effectiveness, proportion of outcomes (genital warts, CIN and cancer) attributed to each HPV type, screening effectiveness, costs, utilities using normal distribution (limited from 0–1 for transition probabilities) when confidence intervals were available, otherwise, a uniform distribution was assigned ranging from 25% below and above base case value (Table ). In total, 10,000 samples were generated from the assigned distribution.
Distributions used for probabilistic sensitivity analyses
In order to customize the model to each of the Canadian provinces, a budget impact module that calculated the size of the target population and the cost of vaccination was constructed. It was linked to the cost-effectiveness model in order to predict lifetime costs and outcomes. The target population was based on the number of females in the assumed target age group for each province (Table ) in the 2011/12 school year based on age-specific population data from Statistics Canada [85
]. Two provinces, Alberta and Quebec, had catch-up programs in the 2011/12 school year and these populations were also modelled. In contrast to the cost-effectiveness analysis, coverage was not expected to be 100%: the number of females vaccinated within each of these age groups was calculated using expected coverage rates (Table ). It was assumed that provinces employed either a 3 dose or a 2 + 1 dosing strategy (Table ) based on current policies. In a 3 dose strategy, the aim is to deliver three doses during the school year, but it was assumed that vaccinees received an average of 2.7 doses due to imperfect coverage. In a 2 +1 strategy, the aim is to deliver two doses during the school year but it was assumed that vaccinees received an average of 1.95 doses. All vaccinated females were assumed to receive a follow-up booster dose 5 years later. The efficacy of a 2 + 1 dosing strategy is not officially approved for either vaccine and efficacy data for this strategy is not yet available from clinical trials. It was therefore assumed that the clinical efficacy of the 2 + 1 strategy was the same as that obtained with the 3 dose strategy.
The total annual cost from the perspective of the budget holder (i.e. the Department of Public Health) was calculated using the cost of vaccine purchase and administration. The net impact was calculated by subtracting the cost of providing all vaccinated females with the HPV-6/11/16/18 vaccine from the cost of providing all with the HPV-16/18 AS04-adjuvanted vaccine. The lifetime health care system costs associated with the HPV-16/18 AS04-adjuvanted vaccine plus cervical cancer screening were compared to those associated with the HPV-6/11/16/18 vaccine plus cervical cancer screening. The total cost of each strategy included the health care costs incurred by those receiving vaccination plus cervical cancer screening as well as those receiving cervical cancer screening only. The lifetime number of cervical cancer events experienced in the cohort of the assumed target population, including both vaccinated and unvaccinated individuals, was also calculated for each vaccine.