Using the pre-side-effect pain scores provided a statistical way to remove the bias created by forcing the dose of Neurontin up to twice the FDA-approved maximum and causing side effects that could have tipped off half the people treated with Neurontin that they had been assigned to the active-treatment group.
Except that dosing may have been what was effective. Perhaps the sode effects were not causing am unblinding but were signals of efficacy. The study seems inconclusive to me.
