The basic process of cost-benefit analysis is to think about all of the main effects and side effects of the regulation, and figure out how to put a dollar value on them all so we can compare everything and make sure we are doing more good than harm.
Valuing life and health in dollars sometimes sounds wrong to people, but it has to be done in order to govern intelligently and there is a large and well-researched literature showing how to do it right.
When a regulatory agency passes a regulation that costs money, that makes people poorer (and possibly unemployed). They react by eating less healthy foods, living in more dangerous areas, buying less safe products, working in harder and more dangerous jobs, etc. The value of a statistical life is about $10 million. That means if we pass a regulation that costs $10 million, we are basically killing someone, so we need to make sure that we are saving a life to make up for it.
Of course, quality of life matters as well. If we make enough people live longer and/or healthier lives, it is as good as saving a life. The measurement that is used to tie everything together is a year of healthy life. Giving someone a year of healthy life is valued at about $200,000. We count up the number of years of healthy life that the regulation will give to people, both by preventing premature death and improving their quality of life, and multiply that by $200,000 to find the dollar benefits of the rule.
We could count all of the costs and benefits in years of healthy life, and in a sense we are, but using dollars as the metric makes things easier.
Doing an analysis of food regulation follows the same basic process as any cost-benefit analysis. The only difference is the detailed knowledge of what to measure and how to measure it. We keep data on how many years of healthy life are typically lost due to various kinds of food-related illnesses and disease, and we have data on the size of the food industry and how much it would cost for them to respond to various types of regulation.
The data is never as good as we would like, of course, and sometimes we have to scramble for estimates of something that we do not know about. The important thing is to make sure that nothing big is being left out, and that we have the right order of magnitude for everything we do measure.