Comparison Among Conventional Multivariable Analysis, Proxy Exposure Analysis, and Instrumental Variable Analysis:...

A catch-up campaign for measles and rubella (MR) vaccinations was conducted among third-year high school (MR-4) students between 2008 and 2012 as part of governmental policy. We examined three analytical methodologies to assess the vaccination policies. We measured the antibody titers of the first-year students involved and not involved in the catch-up campaign and analyzed the levels of antibody titers after two-or-more-dose vaccination using conventional regression analysis, the individual's birth year using proxy exposure analysis, and the birth fiscal year using two-stage least square (2SLS) regression as instrumental variables. Of the 379 students invited to participate in the study, 220 (58.0%) vaccinated against measles and 213 (56.2%) vaccinated against rubella were included in the analysis. Conventional regression analysis revealed that two-or-more-dose vaccination produced 0.27 (95% confidence interval [CI], 0.04-0.49) greater log-antibody titers for measles. Proxy exposure analysis of the birth year revealed that two-or-more-dose vaccination produced 0.40 (95% CI, 0.18-0.63) greater log-scale antibody titers against measles. According to the two-stage least squares regression analysis, two-or-more-dose vaccination produced 0.72 (95% CI, 0.31-1.13) greater log-scale antibody titers against measles. All three analyses showed similar trends, but no findings were reported for rubella. These results were consistent across the three analytical methodologies used to assess the vaccination policies.