From March 2000 through April 2002, 5994 community-dwelling men at six clinical centers in the United States (Birmingham, AL; Minneapolis, MN; Palo Alto, CA; Monongahela Valley [near Pittsburgh], PA; Portland, OR; and San Diego, CA) participated in the baseline examination for the Osteoporotic Fractures in Men (MrOS) study. Eligible men were at least 65 years of age without bilateral hip replacements. Details of the MrOS design and cohort have been published . Subjects included in our current study were drawn from the population of 4215 MrOS participants who returned for the second clinic visit (2005–2006) and had pelvic radiographs digitized and archived . The majority of the study subjects were white men with a mean age of 77 ± 5 years, who were slightly overweight, with good self-rated health. Seventy-seven percent (3216 of 4140) had at least one comorbid condition and 17% (733 of 4140) had at least one mobility limitation. Twenty-seven percent (1134 of 4140) had hip pain and 16% (675 of 4140) were using nonsteroidal antiinflammatory drugs (Table 1).
The study protocol, amendments, and informed consent documentation, including analysis plans, were reviewed and approved by the local institutional review boards.
Imaging Techniques and Analyses
Standing pelvis radiographs were taken at the second clinic visit by positioning the patient using a standardized foot mat with toes internally rotated 15° and the x-ray beam directed 2 inches above the pubis symphysis. A standardized exposure technique was used at all sites. The technique required a Bucky imaging system (Agfa, Mortsel, Belgium), 40 inches of focus-to-film distance, 70 to 80 kVp (mAs dependent on the screen/film system), large focal spot, and full size of the film collimation. In addition, a cassette size of 14 inches to 17 inches was used .
All images were reviewed on picture archiving communication system workstations (Agfa, Ridgefield Park, NJ, USA). The pelvis radiographs were initially checked for image quality and postsurgical changes that obstructed assessment of FAI features. Of the 4215 (8430 hips), 75 (140 hips) radiographs were excluded from the study at the time of the reading as a result of poor image quality (over-/underexposure, beam-hardening artifacts, motion degradation, and presence of supraimposed metallic structures such as belts). FAI assessment was not done on 127 hips that had previously undergone THA or previous pelvic or femoral fractures; however, the contralateral hips were included, resulting in 8263 hips available for FAI assessment. Furthermore, there were 112 hips that had one or more FAI features that did not classify them as cam, pincer, or mixed FAI; thus, they were not included in the analysis assessing FAI type. The total number of hips available for hip-based analysis was 8151. A musculoskeletal radiologist (LN) with 3 years of experience in musculoskeletal imaging assessed the pelvic AP radiographs. During the readings, cases that did not satisfy study inclusion criteria (listed previously) were excluded and, in the remaining hip radiographs, FAI measurements. Different radiologic criteria based on pelvic radiographs have been evaluated for the diagnosis of FAI [37, 46]. Among the many criteria for FAI that have been studied, the following have been reported to have high degrees of intra- and interobserver agreement on AP-view pelvis radiographs and therefore were assessed: caput-collum-diaphyseal (CCD) angle, impingement angle, and lateral center-edge (CE) angle [4, 5, 9, 10, 14, 20, 24, 25, 45, 48–51, 53] (Figs. 1–3). The CE angle was assessed (Fig. 1) according to a previously described technique . A threshold of 39° was used as a sign of impingement . The impingement angle was measured according to a previously described approach (Fig. 2) . Cam morphology was assessed using the impingement angle and the CCD angle using < 70°  and < 125° as abnormal values. The other morphological sign for cam impingement is the CCD angle, defined as the angle between the longitudinal axes of the femoral neck and shaft (Fig. 3) .
The hip was determined to have a radiographic morphology of pincer- or cam-type FAI if it met the diagnostic criteria for at least one diagnostic measure. Mixed-type impingement was defined as having radiographic signs of both pincer and cam FAI types for at least one measure. A normal hip was defined as having all three angles measure within normal limits.
Definition of Arthrosis
Hip radiographs were assessed for five individual radiographic features (IRFs) of arthrosis: (1) joint space narrowing (JSN; 0–4) laterally and medially; (2) osteophyte formation (0–3; femoral or acetabular and either inferiorly or superiorly in each location; (3) cysts (0–3); (4) subchondral sclerosis (0–3); and (5) femoral head deformity (0–3) [32, 44]. Atlas figures were consulted during the readings to improve reliability . These methods have been validated [14, 32, 44]. A summary grade for arthrosis severity of 0 to 4, modified from Croft [4, 32], was assigned to each hip based on individual radiographic features. Grade 0 had no IRFs or was defined as normal. Grade I required either the presence of possible JSN or osteophytes (severity grade = 1). Grade II required the presence of definite (severity grade ≥ 2) JSN or osteophytes plus at least one other feature (cysts or subchondral sclerosis). Grade III required definite JSN or osteophytes plus at least two other features. Grade IV met the criteria for Grade III and also had femoral head deformity. Minimum joint space was measured in millimeters using electronic calipers [4, 11, 14, 32]. In our study, no arthrosis was found as defined by Croft Grade 0 or I, and prevalent arthrosis was defined by Croft Grades II to IV. The prevalence of Croft scores for arthrosis among study patients was distributed as: 2604 (63%), 1128 (27%), 235 (6%), 126 (3%), and 46 (1%) for Grades 0, I, II, III and IV, respectively (Table 1).
After a calibration session in which 30 pelvic radiographs were evaluated by two trained musculoskeletal radiologists (LN, SL), 220 (5%) pelvic radiographs that were originally read were reread independently for inter- and intraobserver reproducibility. Reproducibility for radiographic endpoint assessments was calculated using the absolute intraclass correlation coefficient (ICC) . ICC values for intraobserver agreement for the impingement angle, CE angle, and CCD angle were 0.94 (95% confidence interval [CI], 0.92–0.95), 0.98 (95% CI, 0.97–0.981), and 0.99 (95% CI, 0.99–0.99), respectively. The ICC values for interobserver agreement for impingement angle, CE angle, and CCD angle were 0.98 (95% CI, 0.97–0.98), 0.98 (95% CI, 0.97–0.98), and 0.99 (95% CI, 0.99–0.99), respectively.
All covariates were assessed with validated questionnaires and objective measurements during Visit 2 with the exception of education, which was obtained at the baseline visit . Covariates included patient age, height (measured with a Harpenden stadiometer [Seritex, Tinton Falls, NJ, USA]), weight (measured with a calibrated balance beam or electronic scale), self-rated health status (excellent/good versus fair/poor), and self-reported presence of hip pain (yes/no) ascertained with the questions “in the past 30 days, have you experienced pain in your right hip?” and “in the past 30 days, have you experienced pain in your left hip?” Body mass index (BMI) was calculated (kg/m2). Participants were asked about coexisting morbidities, including stroke, myocardial infarction, cancer, chronic obstructive pulmonary disease, hypertension, congestive heart failure, diabetes, and Parkinson’s disease. Men were classified as having one or more comorbidities or none. Participants were asked to bring in all current medications used within the last 30 days, and a computerized medication-coding dictionary  was used to categorize the medications. All prescription medications recorded by the clinics were stored in an electronic medications inventory database (San Francisco Coordinating Center, San Francisco, CA, USA). Each medication was matched to its ingredient(s) based on the Iowa Drug Information Service Drug Vocabulary (College of Pharmacy, University of Iowa, Iowa City, IA, USA).
Subject characteristics for the whole population of men with valid radiographs at Visit 2 were reported as means and SDs for continuous variables and numbers and percentages for categorical variables.
The outcome variable in the analysis was arthrosis and was defined in two ways. The primary definition of arthrosis was a Croft Grade of II or higher. A secondary definition of arthrosis associated with hip symptoms was arthritis: Croft Grade ≥ II and hip pain in the same hip. Arthrosis was defined as Croft Grade ≥ II and no hip pain in the same hip. Hip pain was defined as hip pain only and Croft Grade 0 or I.
The exposure outcomes were each individual radiographic FAI measurement (impingement, CE and CCD angles) and FAI type defined as a three-category variable (cam, pincer, and mixed). The distribution of the individual FAI measurements was explored visually using histograms and normality plots.
Individual radiographic FAI measurements were categorized based on known clinical cut points : (1) impingement < 70°, CCD > 125°, and CE > 39°; (2) CCD < 125°, impingement ≥ 70°, and CE ≤ 40°; and (3) CE > 40°, impingement ≥ 70°, and CCD ≥ 125°. The reference group for the analyses was hips without any radiographic feature of impingement.
We analyzed the association of radiographic signs of radiographic features of FAI with arthrosis in men using a hip-based analysis. Generalized estimating equations were used to test the association between FAI type and arthrosis by accounting for the correlation between the two hips using an unstructured correlation matrix. When analyzing the clinical hip OA definition, we compared the three groups with clinical or radiographic abnormalities (subjects with arthrosis and/or pain) with the reference group (no pain, no arthrosis) using three separate regression models to determine the association of arthrosis and hip pain, independently, and combined with radiographic signs of FAI. Covariates known to be associated with FAI and arthrosis were included in multivariate models. Models were adjusted for age, race, clinic site, BMI, self-rated health status, and at least one medical condition.