OBJECTIVES : Human papillomavirus (HPV) vaccination rate in Japan is almost zero after 2013 when the government suspended active recommendation of the vaccination as a routine vaccination. One of reasons of such low vaccination rate is the overestimation of adverse events after HPV vaccination and the underestimation of the benefit by preventing the vaccine preventable diseases in the future. In this study, we created a Markov model to properly assess benefit-risk of HPV vaccination incorporating adverse events immediately after the vaccination and disease prevention long after.

METHODS : We defined nine mutually exclusive health states so as to explain the natural history of HPV infection including HPV relevant diseases and the adverse events after vaccination. Each person belongs to only one health state at a time. The transition probabilities between states were extracted from publicly available data. We calculated cervical cancer incident and mortality in five year increments in 2007 and 2012, and analyzed correlation of histograms of model and epidemiological data.

RESULTS : The nine health states were 1) susceptible female individuals, 2) female individuals who received HPV vaccine, 3) female individuals who have adverse events after HPV vaccination, 4) female individuals with HPV infection, 5) female individuals with low-grade squamous intraepithelial lesion, 6) female individuals with high-grade squamous intraepithelial lesion, 7) female individuals with cervical cancer, 8) female individuals who undergo hysterectomy, and 9) female individuals who die of cervical cancer. The deviation of the histogram estimated with mathematical model from that generated with epidemiological data was less than 1%. The correlation of histograms for incident and mortality were 0.989 and 0.987 in 2007, and 0.992 and 0.986 in 2012, respectively (the correlation is 1.0 when two histograms match perfectly).

CONCLUSIONS : Our Markov model was well fit to the epidemiological data. We can simulate various scenarios using this model.