If you have tools to mitigate the risk, then you eliminate the correlation.

No, only if you eliminate the risk. Smoking is still negatively correlated with life expectancy in Japan. The correlation exists if you only look at the USA. It exists if you only look at Japan. It exists if you look at the global population.