The Osteoarthritis Initiative (OAI) is a publicly and privately funded prospective 4-year longitudinal cohort study of 4796 persons aged 45 to 79 years with knee OA or at risk for knee OA. Subjects received no treatment as part of the study nor were they solicited for treatment by the investigators. Patients assigned to the incidence and progression subcohorts who subsequently underwent knee arthroplasty were included in the study. The incidence subcohort comprised 3285 persons who either had knee pain without OA or no knee pain but with OA or were at risk for having knee OA develop. The progression subcohort had 1389 persons with symptomatic knee OA in one or both knees. Symptomatic knee OA was defined in the OAI as (1) a self-report of pain, aching, or stiffness in or around one or both knees on most days for at least 1 month during the past 12 months and (2) radiographic tibiofemoral knee OA defined as Osteoarthritis Research Society International (OARSI) Grades 1 to 3 [1
] or the approximate equivalent, modified KL grades of 2 or higher [1
]. Subjects were recruited from the communities near the following institutions: (1) University of Maryland in Baltimore, MD, USA; (2) The Ohio State University in Columbus, OH, USA; (3) University of Pittsburgh in Pittsburgh, PA, USA; and (4) Memorial Hospital of Rhode Island, in Pawtucket, RI, USA. Key exclusion criteria for entry into the study were (1) the presence of rheumatoid arthritis, (2) bilateral knee arthroplasties before study entry or preexisting plans to undergo bilateral (not unilateral) knee arthroplasties in the next 3 years, (3) bilateral end-stage (modified KL Grade 4) knee OA, (4) use of ambulatory aids other than one straight cane for greater than 50% of the time, (5) comorbid conditions that might interfere with 4-year participation, and (6) likelihood of not residing in the clinic area for at least 3 years. In addition, because the study measured MRI-based changes, men weighing greater than 130 kg and women weighing greater than 114 kg (1595 persons) were excluded because they were unable to undergo 3.0-T MRI. The study was approved by the institutional review board of the OAI Coordinating Center at the University of California, San Francisco (San Francisco, CA, USA).
During the course of 3 years of followup, 116 persons had knee arthroplasties and this was the sample we included in the study (Table ). A total of 27 persons had knee arthroplasties during Year 1, 38 during Year 2, and 51 during Year 3 of the study. Nineteen of the 116 patients had more than one knee arthroplasty, with 13 undergoing bilateral arthroplasties in the same year and six with either repeated unilateral arthroplasties in different years or contralateral arthroplasties in different years. We selected only one knee per patient. For the patients with bilateral arthroplasties, we randomly selected either the right or left knee. For the unilateral arthroplasties, we selected the first surgical episode.
Demographic and clinical data for the 116 persons undergoing knee arthroplasties
Radiographs obtained less than 1 year before surgery were available for 97 patients. Radiographic examinations were obtained an average of 163 days (SD = 101.2) before surgery. Of the 97 patients, one had a unicompartmental arthroplasty and the remaining 96 underwent TKAs. We compared the demographic characteristics of the persons with (n = 97) and without (n = 19) radiographic data available within 1 year of surgery and no differences in age, sex, or comorbidity score were found (p values of t-tests for age and comorbidity = 0.77 and 0.70 respectively and chi-square for gender = 0.60).
We conducted a power analysis to assess the adequacy of our sample size. With a one-group chi-square test and a two-sided significance level set at α = 0.05 and a sample size of 90, our study had a power of 92% to detect a difference of 0.10 in the prevalence of KL grade of 3 or higher between the null hypothesis population (0.95) and the preknee arthroplasty OAI sample (≤ 0.85). We reasoned, if the proportion of patients with KL grades of 3 or greater was at least 0.10 lower than previously published estimates, the clinical implications potentially would be important; for example, if 15% of persons undergoing TKAs had KL grades less than 3, this would indicate a substantial number of patients potentially would be at risk for worse pain and function relative to persons with KL grades of 3 or greater [8
OAI investigators used a 20° fixed-flexion bilateral weightbearing posteroanterior radiographic approach to optimally capture the tibiofemoral joint space [6
]. Radiographs were obtained using a SynaFlexer™ frame (Synarc, San Francisco, CA, USA) to reproducibly position the subject’s knees and feet and to equally distribute body weight while the x-ray beam was angled caudally at 10° from a preset distance [16
]. The radiologic staff at the institutions underwent extensive training to enhance consistency. Lateral or skyline view radiographs were not obtained as part of the OAI study.
Radiographs were obtained at baseline and at the 1-year and 2-year followups. OAI site investigators assessed baseline radiographs and used modified KL and OARSI grading scales to quantify the extent of OA of both knees in each radiograph (Table ). At followups, there were no radiographic readings provided for the sample in the public use datasets. OAI researchers used an extensive training approach to maximize reliability and validity of baseline radiograph readings. Site investigators were experienced rheumatologists and all received standardization training in the KL and OARSI grading scales using teleconferences and a Web-based program. Before the study, each site investigator read 30 knee radiographs and provided modified KL and OARSI grades. They then were given feedback and the correct answers with annotated images and interpretive comments by an experienced musculoskeletal radiologist. During the validation process, investigators read different sets of 30 knee radiographs and their results were compared against the gold standard readings made by an experienced musculoskeletal radiologist. All site investigators had to pass the validation test before reading OAI radiographs.
Because OAI investigators only provided modified KL and medial and lateral joint space readings for baseline data and not for the followup radiographs, we recruited four orthopaedic surgeons (WAJ, RSN, DW, JRH) to make KL, modified KL, and OARSI readings on a total of 200 baseline and 1- and 2-year followup radiographs of patients who subsequently had knee arthroplasties. Surgeons were blinded to TKA status or involved side. Surgeons also were unaware of the baseline OAI KL or OARSI readings for each radiograph. All surgeons were provided with the KL, modified KL, and OARSI scales and radiographic examples [1
]. The KL and modified KL systems are semiquantitative and define the extent of knee OA by the presence and severity of osteophytes and tibiofemoral joint space narrowing. The OARSI system defines the extent of either medial or lateral tibiofemoral joint space narrowing based on a percentage estimate relative to normal. The surgeons practiced individually on 10 radiographs until they were prepared to use the scales to grade the radiographs. We used the weighted kappa to determine the extent of agreement between each surgeon’s readings of the baseline radiographs and the readings made by the OAI site investigators. The weighted kappa is the chance-corrected agreement coefficient when using ordinal scales and when differences in agreement become more serious as disagreements become greater [17
]. One surgeon (JRH) read 225 radiographs from the OAI dataset and had weighted kappas for interobserver variability of 0.67 to 0.84 when compared with publically available baseline radiographic readings by OAI investigators of the same subjects (Table ). Weighted kappa coefficients are commonly used agreement indices that account for chance agreement in ordinal scale measures like the radiographic assessments conducted in this study [17
]. Because of strong concordance with OAI readings, we considered this surgeon’s radiographic readings to be valid and we used the surgeon’s data from baseline and followup readings to examine the extent of medial and lateral joint space narrowing and KL grades.
Extent of agreement between orthopaedic surgeon and OAI site investigators using bilateral baseline OAI knee radiographic data
We determined differences in the proportion of patients with modified KL grades of 3 or higher, indicating at least moderate knee OA, between our sample of persons with TKAs and a fixed proportion of 0.95 based on evidence-based estimates [3
] using the Z
-test. Descriptive statistics were used to summarize the extent of medial versus lateral joint space narrowing. All data were analyzed with SPSS®
Statistics (Version 19; IBM, Armonk, NY, USA).