I'll also point out that accurate epidemiology gets harder and harder as the potency (lethality), mobility, incubation time and longevity of the toxic material increase.  in other words an oil spill into a harbour is nasty, but it's visible, the immediate kill effect is visible, it's visually detectible and remedial efforts are, if not perfect, at least feasible (booms, absorbent barriers, dispersal with surfactants, pump-n-filter etc).  and eventually -- maybe in years or decades -- that oil will settle or break down.  a plume of fine particulate or aerosol isotopes from a nuke plant is a far tougher nut to crack:  invisible for a start, and depending on the isotope, possibly toxic for millennia rather than a decade or two.  toxic  effect can occur from minimal inhalation or ingestion (i.e. high lethality) with a long delay (incubation), and it's highly mobile (can travel far and wide in a short time depending on wind conditions or river volume and speed).

containment is impossible, you can't put the genie back into the bottle, not even in a half-assed way like an oil spill.  figuring out who is exposed and who is not is nearly impossible, as no one may be immediately symptomatic and the amount of contaminant needed to do mortal harm may be too small to detect.  with a 20-30 year incubation and current standards of mobility, the exposed population may be anywhere by the time they are finally symptomatic.

dimethyl mercury is about as close as the non-radioactive world gets to this kind of bad scariness.  it is lethal in tiny quantities, incubation is fairly long (months/years though not decades, ahd a distinctive neurological trauma signature which makes it a bit easier to trace).  the response of the chemistry labs of the academic research world to a high profile case of death by dimethyl mercury was, in effect, to stop using it -- to phase it out -- because the properties of the substance mean that there is no safe way to use it.

one response to methods or substances which are highly lethal and present intractable epidemiology is to shrug and say "nothing can be proven, we are doing the best we can, it is unreasonable to expect more, no one can show conclusively that these deaths are really related."  another is to conclude that it's inherently unethical to use methods and substances which present so intractable an epidemiology problem that they create a de facto culture of impunity, in which "nothing can be proven."


The difference between theory and practise in practise ...