Exploratory analysis of osteoarthritis progression among medication users: data from the Osteoarthritis Initiative

Abstract

Background: We conducted an exploratory analysis of osteoarthritis progression among medication users in the Osteoarthritis Initiative to identify interventions or pathways that may be associated with disease modification and therefore of interest for future clinical trials.

Methods: We used participants from the Osteoarthritis Initiative with annual medication inventory data between the baseline and 36-month follow-up visit (n = 2938). Consistent medication users were defined for each medication classification as a participant reporting at all four annual visits that they were regularly using an oral prescription medication at the time of the visit. The exploratory analysis focused on medication classes with 40 or more users. The primary outcome measures were medial tibiofemoral joint space width change and the Western Ontario and McMaster Universities Arthritis Index (WOMAC) knee pain score change (12-36-month visits). Within each knee, we explored eight comparisons between users and matched or unmatched nonusers (defined two ways). An effect size of each comparison was calculated. Medication classes had potential signals if (a) both knees had less progression among users compared with nonusers, or (b) there was less progression based on structure and symptoms in one knee.

Results: We screened 28 medication classes. Six medication classes had signals for fewer structural changes and better knee pain changes: alpha-adrenergic blockers, antilipemic (excluding statins and fibric acid), anticoagulants, selective serotonin reuptake inhibitors, antihistamines, and antineoplastic agents. Four medication classes had signals for structural changes alone: anti-estrogen (median effect size = 0.28; range = -0.41-0.64), angiotensin-converting enzyme inhibitors (median effect size = 0.13; range = -0.08-0.28), beta-adrenergic blockers (median effect size = 0.09; range = 0.01-0.30), and thyroid agents (median effect size = 0.04; range = -0.05-0.14). Thiazide diuretics had evidence for symptom modification (median effect size = -0.12; range = -0.24-0.04).

Conclusions: Users of neurovascular, antilipemic, or hormonal interventions may have less disease progression compared with nonusers.

Keywords: knee; pain; radiographs.

Figure 1.

Plots demonstrating the distribution of standardized effect sizes for joint space width change. The plots demonstrate the standardized effect sizes from eight comparisons (represented by different colors) per knee (left knee = pluses, right knee = circles). *Antidepressants (not including selective serotonin reuptake inhibitors).

Figure 2.

Plots demonstrating the distribution of standardized effect sizes for Western Ontario and McMaster Universities Arthritis Index (WOMAC) pain change. The plots demonstrate the standardized effect sizes from eight comparisons (represented by different colors) per knee (left knee = pluses, right knee = circles). *Antidepressants (not including selective serotonin reuptake inhibitors).

Figure 3.

Q-Q plots of standardized effect sizes for joint space width at x = 0.250 (JSW250) change among users and matched nonusers. Q-Q plots of plots, (a) right and (b) left knee, of standardized effect sizes for JSW250 change among users and matched definite nonusers when restricted to only participants with JSW250 data (y-axis). The 28 medication classifications are represented by dots or triangles and the line represents expected values under no real effect due to any of the medication classes (based on the 70% of classes with the smallest absolute standardized effect sizes). The red triangles indicate potential deviations from a normal distribution. Among right knees (a) the medication classes marked are antineoplastic agents, anticonvulsants, and anti-estrogen. Among left knees (b) the medication classes indicated are antineoplastic agents, anticoagulants, alpha-adrenergic blockers, and fibric acid (right to left in the figure).

Figure 4.

Q-Q plots of standardized effect sizes for WOMAC pain change among users and matched nonusers. Q-Q plots, (a) right and (b) left knee) of standardized effect sizes for WOMAC pain change among users and matched definite nonusers (y-axis). The 28 medication classifications are represented by dots or triangles, and the line represents expected values under no real effect due to any of the medication classes (based on the 70% of classes with the smallest absolute standardized effect sizes). Among right knees (a) none of the medication classes deviated from a normal distribution. Among left knees (b) the standardized effect size for antineoplastic agents deviated from a normal distribution (red triangle).