Cancer poses danger because of its unregulated growth, development of resistant subclones, and metastatic spread to vital organs. Although the major transitions in cancer development are increasingly well understood, we lack quantitative theory for how preventive measures and post-excision ('reactive') treatments are predicted to affect risks of obtaining a life threatening cancer or relapse, respectively. We employ analytical and numerical models to evaluate how continuous measures such as life style changes, and certain non-targeted and targeted treatments affect both neoplastic growth and the frequency of resistant clones. We find that preventive measures can have a negligible impact on pre-cancerous lesions and yet achieve considerable reductions in risk of invasive cancer. Importantly, our model, based on realistic parameter estimates, predicts that daily cancer cell arrest levels of 0.2-0.3% produce optimal outcomes for prevention, whereas for reaction the level is 0.3-0.4%. For similar cancer cell populations, prevention outcomes are, on average, always better than reactive ones. This is because reactive measures are more likely to select for faster growing subclones with higher probabilities of resistance, highlighting the difficulty in countering relapse regardless of therapeutic impact on cancer cell populations. We discuss these results and other important mitigating factors that need to be taken into consideration in a comparative understanding of preventive versus reactive treatments.