Knee osteoarthritis and the risk of premature mortality in the community: An international individual patient level meta-analysis in 9889 subjects

Abstract

Purpose: The rising prevalence of osteoarthritis (OA) is recognised has having an impact on the increased need for medical consultations due to pain and functional limitations; however, the effect of OA on mortality is less understood. A recent systematic review has shown that there is currently mixed evidence on the relationship between lower limb OA and all-cause mortality, largely due to heterogeneity in definitions of OA, confounders, and inclusion criteria. The aim of this research was to combine individual patient data (IPD) from major international population-based cohorts using harmonised OA data, and to assess whether subjects with knee OA have an increased association with premature mortality. Methods: General population OA cohort studies were identified by literature review and expert consultation, regardless of whether they had published on the relationship between OA and mortality. Six population-based cohorts and one enhanced risk factor cohort* were identified and available for analysis: three US cohorts (Framingham, Johnston County Osteoarthritis Project and Multicentre Osteoarthritis Study* (MOST)); one UK cohort (Chingford); a Dutch cohort (Rotterdam Study); and an Australian cohort (Tasmanian Older Adult Cohort (TasOAC)). The definition of knee OA was established via an expert consensus study, which selected the two OA variables used in this analysis. Radiographic OA (ROA) was defined as Kellgren and Lawrence (K/L) grade 2 ≥ or equivalent validated scoring methods (e.g. the OARSI atlas). Pain was defined by using either an NHANES-type question (i.e. ‘have you had pain for most days in the last month in your joint’), the WOMAC pain subscale, or alternative equivelent pain question. Subjects were divided into four categories: (1) No OA (ROA-/Pain-); (2) ROA only (ROA+/Pain-); (3) Symptomatic OA (ROA-/Pain+); (4) Symptomatic Radiographic OA (ROA+/Pain+). Included subjects were free from RA, aged 45–80 years old, had complete symptomatic and radiographic OA data, confounder data (age, sex, and race) and mortality data. Complete case analysis was undertaken for all cohorts due to the low level of missingness (<10%). The association between OA and the time to all-cause mortality was assessed using Cox proportional hazard regression models to estimate hazard ratios (HRs) and 95% confidence intervals. Symptomatic radiographic OA (SROA) was assessed by comparing subjects who were ROA+/Pain+ against subjects free from OA (ROA-/Pain-). Symptomatic OA (SOA) compared subjects who were ROA-/Pain+ against ROA-/Pain- subjects. IPD meta-analysis methods were utilised, using a two-stage approach. Individual studies were analysed separately using identical methods and then data was pooled in the second stage using random effects analysis, with Dersimonian and Laird estimation. Heterogeneity was assessed using I-squared and the tau-squared statistic, and uncertainty around tau was accounted for by producing confidence intervals around the pooled estimate using the Hartung-Knapp correction. Results: The number of baseline subjects with complete data ranged from 253 to 3384 in each cohort, totalling 9889 subjects ranging from a median follow-up of 5.6 to 19.8 years. Prevalence of knee SROA ranged from 4.4% in one of the youngest cohorts (Chingford) to 33.3% in the risk enriched MOST cohort. Subjects with SROA had an overall pooled estimate of HR 1.47 (95% CI 1.07, 2.04) in an unadjusted model, with a pooled estimate of HR 1.19 (95% CI 1.04, 1.37) when adjusted for potential confounding factors (figure 1). Subjects with SOA had an overall pooled estimate of HR 1.13 (95% CI 0.70, 1.81) in an unadjusted model, with a pooled estimate of HR 1.19 (95% CI 0.78, 1.81) in the adjusted model (figure 2). Conclusions: Subjects with knee SROA had a 19% increased association with premature death independent of age, sex and race compared to pain-free and ROA-free subjects. This increase was not found in subjects with pain alone (SOA). This international study is one of the largest analyses of well-phenotyped knee osteoarthritis and mortality in the general population.