Viral load tests are imperfect.
There is very little viral RNA to to count it we use something called Polymerase Chain Reaction - this amplifies the quantity to something we can count.
The Lower Limit of Quantification (LLOQ) may be 30, 25, 15, 12 or 10 and virus that can be detected, but is at a lower level than this results in a <30 detected, <25 detected, ... <10 detected.
The Lower Limit of Detection (LLOD) is the smallest quantity that can be seen. A test with LLOQ <15 will have an LLOD of around <6 so RNA in the range 0-5 will come back as undetected.
When we see the rapid drop in HCV RNA this is all the easy stuff being killed. The problem is the mutants which are present in low levels. They are present in low levels because they don't grow as fast, but kill off all the faster growing stuff and you are left with the mutants.
Think of it like pulling all the plants out of a garden to leave raw earth. Pretty soon weeds take over the empty space.
So recurrence comes from hard to kill mutants that were present below the LLOD at the end of treatment. In other words the undetected was really "can't detect with the sensitivity required". For this reason we over treat everyone. 8 weeks Havoni for GT1 naive with low fibrosis gets at 121/123 SVR if the viral load is < 6 million to start with.
A high viral load is a negative predictor because it says 1) your body is not doing a good job killing it and 2) there are more virions, so more mutants.