Concordance of baseline pain measures (across two reporting instruments) influences treatment effect: post hoc analysis of a phase 3 randomized controlled trial of triamcinolone acetonide extended-release in patients with knee osteoarthritis