Both the D and D+E groups receive the same dietary intervention. Our weight loss goal is a group mean of at least 10% of baseline weight, with a desired 15% weight loss. The dietary plan is based on partial meal replacements, including up to 2 meal-replacement shakes per day (provided by General Nutrition Centers, Inc., Pittsburgh, PA.). All participants are closely monitored, and their daily caloric intake adjusted according to their rate of weight change.

Initial diet plans provide an energy-intake deficit of 800–1000 kcals/day as predicted by energy expenditure (estimated resting metabolism × 1.2 activity factor). The lowest calorie level permitted is 1100 kcals for women and 1200 kcals for men. The calorie distribution goal is 15–20% from protein, < 30% from fat, and 45–60% from carbohydrates, consistent with the Dietary Reference Intakes for Energy and Macronutrients [21] and successful weight-loss programs [22]. For the third meal, participants follow a weekly menu plan and recipes provided by the study composed of traditional foods. It contains 500–750 kcals and is low in fat and high in vegetables. Snacks include a snack bar, fruit, or vegetable providing approximately 100–120 kcals each. The food plan is tailored to individual preferences and energy needs. Meal plans are developed by the intervention staff to provide the prescribed macronutrient-balanced energy intake as fewer meal replacements are consumed during the trial.

Body weight is monitored in weekly nutrition education and behavioral sessions. Nutrition interventionists, trained in behavioral therapy and experienced in working with older adults, run all group and individual sessions. There is one individual session and 3 group sessions per month for the first 6 months. The behavioral sessions focus on awareness of changing eating habits to lower caloric intake. Information on what food changes to make, how to make them, and why they are important is clearly explained and discussed with participants and significant others. Group sessions include problem solving, review of a specific food topic, and tasting of several well-balanced, low-fat, nutritious foods prepared with easily available ingredients. During the individual sessions, the nutrition interventionist reviews individual progress, solves problems, answers questions, and sets goals. From months 7 to 18, participants attend biweekly group sessions, with an individual appointment every 2 months.

Once they reach the study's desired weight-loss goal of 15% of baseline body weight, participants either begin weight maintenance or continue to lose weight using safe, healthy nutritional practices. Weight loss can continue throughout the intervention, provided the individual wants to and has not reached a level that may be associated with health hazards in this population; i.e., a >20% body weight loss at 6 months or > 30% at 12 months. Weight maintenance uses relapse-prevention techniques that include managing the environment, adjusting appropriate goal-based daily energy needs, and continuing self-regulatory skills.

For participants who have difficulty achieving the weight-loss goal, we use a behavioral "toolbox" that includes frequent individual behavioral counseling sessions, incentives, home-assessments, or other items deemed appropriate. Diet plans can be altered to restrict one of the snacks and/or the size of the evening meal. Participants record their food and beverage intake in daily logs that are collected monthly and used as a self-monitoring tool for dietary intake. Nutrition interventionists use them to design rescue procedures, using behavioral and educational strategies.

Weight is monitored weekly. A loss of >20% body weight after 6 months or >30% after 12 months triggers a safety alert.

The exercise intervention is identical for both exercise groups. Sixty-minute sessions are conducted 3 d·wk^-1 for 18 months, and each group (E and D+E) exercises at a different time to prevent crossover effects from their interactions. The first 6 months are facility-based. Anytime after 6-month follow-up testing is completed, participants that would like to add home based exercise to their training routine will alternate between the facility (2 d·wk^-1) and home (1 d·wk^-1) during a 2-week transition phase. Subsequently, participants can decide to remain in the facility program; opt for a home-based program; or combine the facility and home-based programs.

The 3 d·wk^-1 exercise program consists of an aerobic phase (15 min), a strength training phase (20 min), a second aerobic phase (15 min), and a cool-down phase (10 min). The strength-training phase is particularly relevant to help offset any possible loss of muscle and bone mass resulting from dietary weight loss.

Walking is the primary mode of aerobic training. Participants who meet the inclusion criteria but cannot walk for extended periods use upright and recumbent stationary bicycles to supplement walking. Exercise intensity ranges between 50–75% of the heart-rate reserve, using the symptom-limited maximum heart rate obtained from a graded exercise test (GXT). The strength training session is approximately 20 minutes and targets lower extremity muscles and, to a lesser extent, upper body muscles. This phase includes 1–2 sets of 10–12 repetitions on leg extension, leg press, seated leg curl, seated calf raise, compound row, and vertical chest or incline press.

During months 7–18, home-based participants are telephoned every other week in the first 2 months, every third week during the next 2 months, and every month thereafter. Monthly exercise and attendance logs are gathered to monitor progress. The home-based participant exercises at home 3 d·wk^-1 and comes to the facility once a month for an individual meeting with the exercise interventionist. Home-based strength training utilizes a Thera-Band exercise program.

Participants randomized to the D+E intervention receive both interventions described above. Dietary counseling is scheduled before or after the exercise sessions to minimize the number of visits.

We conservatively estimate 80% retention and 65% adherence rates over this 18-month intervention. Adherence to the nutrition intervention is defined as attending the nutrition classes; to the exercise sessions, the number of sessions completed divided by the number scheduled (72 weeks × 3 d·wk^-1 = 216 sessions).

The health psychologist trains the exercise interventionists in behavioral techniques that include: (1) tailoring training to the participant's ability; (2) frequent contact during all intervention phases; (3) clear behavioral goals and feedback on achievements; (4) training participants to self-monitor their heart rate and to complete activity logs; (5) establishing personal commitment to the project through the exercise leader; and (6) targeting educational materials and field trips to prevent relapse and to develop prompts to exercise. Primary interventionists regularly review adherence data and identify any participants who need additional reminders and/or counseling.

The 30-month recruitment period is divided into 10 waves of 45 participants each, entering the study at 3-month intervals. Recruitment strategies include mass mailings, newspaper advertisements, and presentations at local aging service networks, senior centers, and churches. Our Claude D. Pepper Older Americans Independence Center recruitment core also has access to a large database of older adults who have consented to be contacted about participation in clinical trials. Selected physicians in orthopedics, geriatrics, and rheumatology receive information on the study and encourage qualified patients to contact IDEA recruitment personnel for further information. Specific strategies aim to maximize the number of African Americans who qualify for, and are enrolled in, the study. At biweekly meetings, all recruitment activities and the number of participants randomized are reviewed.

Figure 2 shows eligibility and screening measurements for the IDEA trial. All participants are tested at baseline and 6 and 18 months post randomization. A subset of 225 participants receive supplemental performance tests, x-ray, and MRI (Figure 2).

The Eligibility Questionnaire addresses joint pain, physical function, activity level, co-morbid diseases, willingness to participate in an 18-month study, height and weight (to determine BMI), caregiver status, status of significant others, and distance of home from the center (within a 50-mile radius). Demographic information includes data on occupation, income, and education. The medical form is used to determine co-morbidities and to analyze medication data gathered by self-report. A computerized system places individual drugs in major therapeutic categories to assess changes in nonsteroidal anti-inflammatory drug use. Participants undergo an initial symptom-limited, maximum exercise stress test to exclude anyone who has severe manifestations of coronary heart disease using a modified Naughton protocol that has 10 stages. Volunteers are screened for cognitive deficiencies using the Modified Mini-Mental State Exam (3MSE) [23]. A score of < 70 justifies exclusion because we feel that cognitively impaired persons could not adhere to the protocol, especially the dietary requirements for the weight-loss groups.

We plan to assess biological markers of inflammation before and after weight-loss and exercise treatments to determine whether they significantly lower inflammation levels. Our primary measure is IL-6. This cytokine is consistently implicated in OA pathogenesis and showed significant improvement with weight loss in ADAPT. We will also measure CRP as an overall marker of chronic inflammation [10] and TNFα and TNFα soluble receptor 1 (sTNFR1) because these cytokines are also implicated in OA pathogenesis [24]. Leptin, an adipokine that increases synthesis of TGFβ, a known stimulator of osteophyte formation, is also measured.

Blood samples (approximately 40 ml per visit) for assessing biomarkers are collected via venipuncture in the early morning, after a 12-hour fast and at baseline, 6-month, and 18-month assessment visits. Urine samples (20 ml per visit) are collected in 250 ml specimen cups by each participant for analysis of new and emerging OA biomarkers. In the event of acute respiratory, urinary tract, or other infection, we postpone blood and urine sampling for 1–2 weeks after recovery from symptoms.

Our primary measures of knee joint loads are tibiofemoral compressive forces and the internal abduction moment. We obtain 3-D kinematic gait data using a 25-reflective marker set, arranged in the Cleveland Clinic configuration, a 6-camera Motion Analysis System sampling at 60 Hz and EVaRT 4.4 software. Ground reaction forces and moments are obtained with a 6-channel force platform (Advanced Mechanical Technologies, Inc., Newton, MA) operating at 480 Hz. Six successful trials are collected on each participant, and 3 chosen for subsequent analysis. A successful trial is defined as one in which the participant's entire foot is placed on the surface area of the force platform while walking within ± 3.5% of freely chosen speed.

A Butterworth low-pass filter with a 6 Hz cut-off frequency is used to remove high frequency noise from the linear kinematic data. OrthoTrak 6.0 β4 clinical gait analysis software is used to generate lower extremity angular kinematic and kinetic data using inverse dynamics. Knee-joint forces are calculated using a model developed by DeVita et al[25,26]. We have successfully applied it to healthy and OA subjects[25,26], finding ranges (2.8 – 6.0 BW) similar to those in previous studies [27,28].

The reliability of the gait data was determined using 21 IDEA participants (4 males, 17 females) with a mean age of 65.7 ± 5.8 yrs, a mean mass 92.4 ± 13.0 kg, and a mean height of 165.6 ± 10.0 cm. Participants were tested twice, 1 week apart. Intra-class correlation coefficients (ICC) were used to compare the between day reliability of peak internal knee flexion/extension and internal knee abduction/adduction moments, peak knee compressive force, peak vertical ground reaction force, and stride length. Overall day-to-day reliability was excellent with ICC values ranging from 0.86–0.98 for the selected variables (Table 3). Knee flexion/extension moment and stride length were most affected by BMI, with lower reliability associated with the highest tertile of BMI (Table 3 and Figure 3).

The WOMAC is used to measure self-reported physical function and pain [29]. The LK version asks participants to indicate on a scale from 0 (none) to 4 (extreme) the degree of difficulty experienced performing activities of daily living in the last 48 hours due to knee OA. Individual scores for the 17 items are totaled to generate a summary score that can range from 0–68, with higher scores indicating poorer function.

The pain index assesses participants' pain on the same scale, ranging from 0 (none) to 4 (extreme). The pain subscale consists of 5 items and total scores can range from 0–20, with higher scores indicating greater pain. This instrument is recommended by the Osteoarthritis Research Society as the health status measure of choice for older adults with knee OA. It has been validated for use in orthopaedic and pharmacologic interventions [29,30].

Our mobility measure is 6-minute walk distance. Participants are told to walk as far as possible in 6 minutes on an established flat, indoor course. Six-minute walk distance is related to symptom-limited maximal oxygen consumption (r = 0.53) and the test has a 3-month test-retest reliability of 0.86 [31].

Body mass index is calculated as mass (measured in kilograms on a standard calibrated scale) divided by height squared (measured in meters). Whole body lean mass (LM), fat mass (FM), and bone mass are measured by DXA on all 450 participants. BMD is also measured at the posteroanterior (PA) lumbar spine and hip. DXA scans are obtained with a fan-beam scanner, Delphi A™, Hologic (Waltham, MA), using manufacturer's recommendations for patient positioning, scan protocols, and scan analysis. Coefficients of variation (%CV) are 1.2% for whole body FM; 0.5% for whole body LM; 0.9% for whole body BMD; 1.2% for PA spine BMD; and 0.9% for total hip BMD.

We measure bilateral knee concentric strength at baseline, FU6, and FU18 on a random sample of participants (n = 225) with equal numbers in each group using a Kin-Com 125E isokinetic dynamometer set at an angular velocity of 30 deg/sec. Grip strength, an excellent measure of overall body strength, is also measured using a standard grip strength dynamometer.

During SV1 and FU visits, participants are administered a medication questionnaire adapted from the Atherosclerosis Risk In Communities (ARIC) study [32] and widely used in field research and our prior studies. We obtain information about all prescription and over-the-counter medicines and supplements used during the 2 weeks prior to the interview.

MRI of the most symptomatic knee will be performed at baseline and FU18. MRI scans are performed with 1.5 Tesla General Electric scanners using an extremity coil. The following MR sequences were obtained: (1) Double oblique coronal three-dimensional spoiled gradient echo (SPGR) with fat suppression; (2) Axial T1-weighted spin-echo (SE); (3) Double oblique coronal T1-weighted SE; (4) Sagittal T1-weighted SE; (5) Sagittal T2-weighted fast spin echo (FSE) with fat suppression; (6) Double oblique coronal T2-weighted fast spin echo (FSE) with fat suppression.

MRI has the capability of visualizing all potentially relevant OA joint structures; therefore, it is not surprising that it is an important tool in improving our understanding of knee OA by providing a non-invasive tool for the study of healthy and diseased states, as well as in providing a means of assessing risk factors for and the effectiveness of therapeutic interventions for prevention of pain, dysfunction, and disability in OA. MRI scoring will be read paired and blinded to timepoint by a musculoskeletal radiologist using the BLOKS method [33]. The following MRI features are scored: (1) Bone marrow lesion (BML) size; (2) BML % area; (3) Percent BML rather than cyst; (4) Cartilage loss; (5) Extent of any cartilage loss at specified points;(6) Osteophytes; (7) Synovitis; (8) Effusion; (9) Meniscal extrusion; (10) Meniscal signal; (11) Meniscus tear; (12) Ligaments: presence or absence of tear in patella tendon, ACL or PCL; and (13) Periarticular features: presence or absence of abnormal features such as pes anserine bursa and iliotibial band.

Test-retest reliability was performed on 10% of scans (N = 22) 1 week apart. Prior data demonstrates the inter-reader reliability for the BLOKS items range from 0.51 for meniscal extrusion up to 0.79 for meniscal tear. The reliability for other key features is 0.72 for BML grade, 0.72 for cartilage morphology, and 0.62 for synovitis [33]. Our reliability exceeded previously reported results (Table 4). Reliability for meniscal extrusion was 0.95; meniscal tear, 0.99; BML size, 0.72, cartilage morphology, 0.93, and synovitis 0.66.

Measurements are confined to the medial femorotibial compartment, because (1) it is the most common site of knee OA [34]; (2) OA there is strongly related to bone marrow edema lesions and varus knee alignment [34,35]; and (3) we have additional, well-established measures of joint integrity using BLOKS [33]. The tibia (MT) and central weight-bearing medial femur (cMF) are segmented by manually tracing of the total area of the subchondral bone (tAB) and the area of the cartilaginous joint surface (AC) on a slice-by-slice basis [36,37]. Segmentation of tAB included all areas of cartilage-covered and denuded subchondral bone, but not osteophytes [36,37]. Likewise, AC segmentation excluded osteophyte cartilage. MT was segmented throughout all slices that displayed cartilage; analysis of the femoral condyle was confined to the portion of the condyles that is displayed without relevant partial volume effect in coronal images (cMF). All segmented data are quality controlled (QC'd) by one expert by checking all segmented slices of each data set, and corrections of the segmentation were performed if necessary (Figures 4 &5).

The same software used for segmentation and QC readings is used to calculate the size of the total area of the subchondral bone (tAB), the tAB covered with cartilage (cAB), the denuded bone area (dAB), the area of the cartilage surface (AC), the mean cartilage thickness across the entire tAB (ThCtAB), counting all denuded areas as 0 mm cartilage thickness the mean cartilage thickness across cAB (ThCcAB), and maximal cartilage thickness. In addition, subregional measurements are performed by analyzing ThCtAB in central, internal, external, anterior and posterior subregions of MT, and central internal and external subregions of cMF, using a recently developed approach by Wirth et al. [38]. Figure 6 represents a 3-D visualization of the segmented data.

All primary analyses will be based on the intention-to-treat method in which each participant is included in the initial randomization group regardless of adherence. IDEA stratification factors, BMI, and gender will be included in all statistical models, so the analysis will match the design, and the estimated variance will not be biased. Assumptions will be verified for all models, and appropriate transformations used when necessary.

The primary aim of this study is to test whether the intensive dietary intervention, the exercise-only intervention, and a combination affect inflammation (IL-6) and knee joint loads (knee maximum compressive force) differently. Analyses for the primary aim will test the effect of the 3 interventions on the primary outcome measures using repeated measures analysis of covariance (ANCOVA). These methods are robust-to-moderate departures from linearity when the covariates are not extremely different. Eighteen months after randomization, intervention effects will be estimated and tested, adjusting for prerandomization levels, BMI, and gender. Subsequent models will be adjusted for other variables, including age, education, long-term individual preferences for exercise, and specific factors showing group differences at baseline.

The effect of the interventions on WOMAC function and pain; inflammation markers, including CRP, TNF-α, TNF-sR1, and leptin; and maximal internal abduction moment will be determined with repeated measures ANCOVA and statistical tests and estimates as described above, verifying assumptions and using appropriate transformations. We will examine whether increased weight loss slows disease progression and how changes in the biomechanical and inflammatory pathways result in improved function, diminished pain, and slower disease progression. Multivariate and multiple regression methods will be used to study the associations between factors in the hypothesized pathway.

A total sample size of 450 (150/group) provides 80% statistical power to detect 20% differences in IL-6 and 15% differences in knee joint loads at the two-sided 0.008 significance level (adjusted for 3 pairwise intervention comparisons and 2 outcomes) with 80% retention. Our preliminary data suggest that a 20% difference in IL-6 is clinically important, and an additional 5% weight loss for a 200 lb patient would result in a 120 lb difference in knee joint loads. Estimates of mean squared error were obtained from the ADAPT study [44], which measured the same outcomes and used a similar patient population. Based on the assumptions above and the evaluation of secondary outcomes, the trial is designed to ensure that 120 subjects per group are evaluated at the end of 18 months.