Background In osteoarthritis (OA) multiple sets of criteria are available to distinguish between patients who clinically improve and those who do not. However, in addition to knowing who will improve after treatment, knowing who will deteriorate could help patients and doctors to make informed treatment choices. In contrast to important improvement, important worsening in OA is not well defined.
Objectives 1. To estimate the minimal clinically important worsening (MCIW) values for pain, function, stiffness and patient global assessment (PGA) in patients with knee and hip OA
2. To examine the sensitivity and specificity of several sets of previous described and proposed worsening criteria for determining worsening
Methods Data were used from a cohort of knee and hip OA outpatients visiting our department who received standardised evidence-based tailored conservative treatment in a stepped-care format for 3 months. Sub cohort I comprised 250 patients (mean age 56 years, 67% female, 83% knee OA) with three-month follow up and sub cohort II consisted of 400 patients (mean age 56 years, 66% female, 88% knee OA) with two-year follow up. For this study baseline and three month data were used.
First we estimated MCIW values for pain (numeric rating scale 0-10 (NRS) and WOMAC pain scale 0-100), WOMAC function (scale 0-100), WOMAC stiffness (scale 0-100) and PGA (scale 0-10) in cohort I using an anchor-based 7-point Likert transition scale. Second, on the basis of those MCIW values and on an inventory of definitions of worsening in the literature an expert group selected and defined different potential sets of worsening, varying in the number of outcome measures incorporated and the amount of change in relative and absolute MCIW values determined in the previous step. These sets included the inverse of the OMERACT-OARSI responder criteria, the inverse of the improvement criteria by Tubach, MCIW values of OA after rehabilitation (Angst) and several sets based on the inverse of these criteria with different relative and absolute changes. Third, in sub cohort ll we evaluated the sensitivity and specificity of the 5 MCIW values generated from sub cohort I and the different sets of worsening using the same anchor-based transition scale.
Results The estimated MCIW values for absolute and relative (compared to baseline) worsening at 3 months were: NRS pain 1.7, 13%; WOMAC pain 2.7, 8%; WOMAC function 4.6, 2%; WOMAC stiffness 5.6, 25% and PGA 0.76, 14%. The sensitivity of previously used definitions for worsening and inverse of improvement criteria was low (range 17-40%) while specificity was good (range 70-92%). Using our estimated MCIW values, sensitivity above 50% was observed. Sensitivity of 60-70% was observed in sets proposed by the expert group in which less stringent criteria were used compared to improvement criteria.
Conclusions Sensitivity of previously described definitions for worsening and criteria being the inverse of improvement criteria was low, while specificity was good. Our results suggest that, compared with improvement criteria, criteria for worsening should incorporate relatively small absolute and relative changes and that sensitivity is better in less stringent sets.
Pham et al. Osteoarthritis Cart 2004;12;389-99
Tubach et al. Arthritis Care & Research 2012;64;1699-1707
Angst et al. J Rheumatol 2002;29;131-138
Disclosure of Interest : None declared