So  50/50  technically is a random guess, the zero point in one's measurement. The question is whether something in the environment is present which has some results occur more less often or more often.

In principle, one could also use exactly the same test to determine if some environmental chemical caused fish to live longer.   

50/50 =  no effect.    Calling 60/40 as zero simply means those doing the testing are adding a bias in their measurement and doing a correction mathematically instead of experimentally removing the bias that had been added in the first place.

LD50 is arbitrary; it's the "median lethal dose". These days it's more common to refer to ALDxx = approximate lethal dose; this is to emphasise that it's not necessary to sacrifice many dozens or even hundreds of animals to get an exact estimate, as was done until the 1990s. You could use lower (e.g. LD01) but that would probably require more animals:

"The choice of 50% lethality as a benchmark avoids the potential for ambiguity of making measurements in the extremes and reduces the amount of testing required....Measures such as "LD1" and "LD99" (dosage required to kill 1% or 99%, respectively, of the test population) are occasionally used for specific purposes." https://en.wikipedia.org/wiki/Lethal_dose

I concur.  To be statistically correct, one would /could include variance in the equation.  So if the answer were   LD50 +/-  10,  that could be okay;  LD50 +/- 25  less so, but that would still be in the right ball park.

If it were  LD60 +/-  30,   statistically, that would not be experimentally distinguishable from  LD50 +/- 25 unless quite large numbers of lethal doses were being counted.