National Institute of Arthritis and Musculoskeletal and Skin Diseases
6701 Democracy Blvd., Bethesda, Maryland
Welcome and Introductions
Gayle Lester, Ph.D.
National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH, Bethesda, Maryland
Many of you remember the Stepping Away Meeting in 1999, which had both a private and a public component. This is a small workshop version of that meeting to cover the types of issues that were addressed there.
To start us off, I am pleased to have Steve Katz, the director of NIAMS, here this morning, and I've asked him to make a few remarks of welcome.
Stephen I. Katz, M.D, Ph.D.
National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH, Bethesda, Maryland
I start by saying that my only two contributions to this meeting are, one, to encourage Joan [McGowan, Ph.D.] and Gayle and, two, to say: Don't miss this opportunity to get a picture of all these people involved in OA [osteoarthritis] together. We have the real experts here in this room, so don't miss an opportunity to record that, even though it won't accompany your paper in the OA journal.
On behalf of Joan and Gayle, I want to thank everyone for being here, Stefan [Lohmander, M.D., Ph.D.] in particular, who made a long trip. He has been extraordinarily helpful and positive at almost every session where we have talked about this difficult problem. Tim [McAlindon, M.D.] and MaryFran [Sowers, Ph.D.], thanks for being here on the phone. Even though you can't be here, we can add in your pictures. When I say "we," I mean "we" on behalf of all of the NIH. Chhanda Dutta [Ph.D.] is here from the National Institute on Aging, and, fortunately, I can tell you that many groups at the NIH are interested in moving this field forward, most notably the National Institute of Biomedical Imaging and Bioengineering, which has a keen interest in the Osteoarthritis Initiative, and also the National Center on Minority Health and Health Disparities, the National Center for Complementary and Alternative Medicine, and the NIH Office of Research on Women's Health.Vivian Pinn [M.D.] has been tremendously supportive in moving this field forward. This is a collective effort at the NIH to get the things done that we need to do to address this problem. We are also pleased that the CDC [Centers for Disease Control and Prevention] is represented. Chad Helmick [M.D.] is always a positive force in these discussions. I said to Jeff Sacks [M.D.] a little while ago that people used to think there was a schism between the CDC and us, but I am proud to say that two of my children work at the CDC. Both the NIH and the CDC can be proud when we sit before the Congress and talk about how the agencies interact. I don't think anyone can accuse us of being silos in the musculoskeletal diseases.
Patience White [M.D.] is here representing the Arthritis Foundation. I went to a NIAMS Council dinner 13 years ago, and to my right was Debra Lappin, and she has always represented the Foundation, so I make the mistake all the time of saying that she is the representative, but she always does represent the Arthritis Foundation. I also want to thank Kent [Kwoh, M.D.] for providing at least a base from which to start.
Unfortunately, there are no agents; we know that now more than we did a few weeks ago. Everyone here knows that OA is a big problem. And this has come to better light due to work from some of you. David's [David Felson, M.D.] NEJM1 paper brings to light some complexities not appreciated until recently, so we as an Institute, and across the NIH, are trying to improve the lives of patients and people who have OA. You are well aware of the value of bringing these groups together and the importance of partnerships in the clinical arena.
1 Englund M, et al. Incidental meniscal findings on knee MRI in middle-aged and elderly persons. NEJM 2008;359:1108-15.
I've got long-held plans, so I can't be here this afternoon, but I look forward to learning from your discussion today. Before I go, you know that Steve Hausman [Ph.D.] was our Deputy Director, and now, as of yesterday, Bob Carter [M.D.] — who has played many professional roles nationally and internationally — comes to us direct from the chairmanship of rheumatology at the University of Alabama to become Deputy Director of NIAMS. He is a rheumatologist and a B-cell biologist. Please welcome him.
Presentation on the Limits of Discussion
Marc C. Hochberg, M.D.
University of Maryland School of Medicine, Baltimore
I want to thank Steve [Katz, M.D., Ph.D.] and Gayle [Lester, Ph.D.] for inviting me to come today and for their support of the Osteoarthritis Initiative, and I want to acknowledge Joan [McGowan, Ph.D.] for her support of me over the years in our work on bone. I'm going to introduce the subject here. It's nice to see Bob Carter [M.D.] here. We both spent time in Baltimore years ago, and, Bob, you'll recognize the tie since it is Friday.
I have three brief introductory slides that I will run through quickly because everyone here knows about OA. Particularly in our economic situation now, which is front-page news, and which has overwhelmed the political debate in our country, we should not lose site of the fact that OA has a massive impact on the economy of the United States.
We are not going to talk about pathophysiology, but we are going to talk about risk factors and changes in bone and cartilage that precede the changes we recognize on radiographs. And we'll discuss the role of radiographs in case definition and how important it is to obtain a radiograph in primary care and case management.
Rather than look at management of the problem in terms of patient care, we are going to focus on what we know about the risk factors. I was particularly impressed by the presentation by Ewa Roos [Ph.D.] at the OARSI [Osteoarthritis Research Society International] meeting a few weeks ago, and I hope Stefan [Lohmander, M.D., Ph.D.] will comment.
One of the things that Gayle asked me to focus on was symptoms and whether pain was intermittent. I went back to the 1995 definition for OA disorders. This definition involved NIAMS, the AAOS [American Association of Orthopedic Surgeons], the Arthritis Foundation, and, I think, the National Institute on Aging, and other professional societies. That definition states, "When clinically evident, OA diseases are characterized by joint pain, tenderness, limitation of movement, crepitus, occasional stiffness, and variable degrees of inflammation without systemic effects." We'll talk a little bit about joint pain, frequent pain vs. infrequent pain, and also persistent pain.
Now, one of the ways we define OA for the purposes of epidemiology studies has been based on the radiograph. I've listed a few of the scales used [pointed to slide] — this list is not inclusive — but we're going to focus on symptomatic OA rather than asymptomatic radiograph OA.
What about the ACR classification criteria? What about a definition of OA that requires symptoms? In NHANES I [National Health and Nutrition Examination Survey] — a number of people in this room analyzed those data [pointed to slides] in terms of epidemiology and prevalence of radiographic OA and factors associated with OA. This is in combination with radiographic features.
There are people who don't have symptoms on most days, so if this is our definition of frequent knee pain, then we recognize that there are people who have infrequent knee pain, who may not have pain on most days or not during one month during the prior 3 months. And then we recognize that there are people without pain who show radiographic features. And this impacts our discussion of risk factors.
Population studies and clinical studies often use a definition that requires frequent knee pain. NHANES III data are the most frequently used data on symptomatic knee OA. I thought I would go through two studies. One is a paper analyzing NHANES III data, by Dillon, in the Journal of Rheumatology in 20062 . So, 3,128 adults were interviewed. There was oversampling of those age 60 and above. They had radiographs of the tibiofemoral joint and Gayle [Lester, Ph.D.] wanted to bring up, "Do we rely solely on the tibiofemoral joint or of more of the knee?" — so this was a single non-weightbearing AP [anteroposterior] radiograph. We can't look at other views with fluoroscopy or without fluoroscopy. It is not an optimal view. The duration was 6 weeks, which was the Rheumatology Society's old definition for the classification of rheumatoid arthritis. Persistent knee pain was defined as a positive response to "Have you ever had pain in your knees on most days for at least 6 weeks? This also includes aching and stiffness." It does not look at current pain; it looks at "ever" pain. So, the overall prevalence, unadjusted for population weights, was 30 percent. The overall prevalence of symptomatic OA was about 12 percent. Surprisingly, there was no prevalence difference by gender. It is not higher in women than in men, which is what those of us who teach would have expected. This was mid-90s data, wave two of NHANES III. The prevalence increased with increasing age. The prevalence in non-Hispanic blacks was higher than in non-Hispanic whites. The prevalence in Mexican Americans was intermediate between these other two. There was a higher prevalence in those with high body mass index. There was a significant sort of dose response relationship. There was lower prevalence in current smokers, which had also been reported in NHANES I. Also, on the impact in terms of six function questions, the results were consistent with reporting in terms of difficulty. There was no difference between persons with asymptomatic knee OA and those with no radiographic OA. Those with symptomatic OA compared with those with no OA had more difficulty on all six functions. The only drug for which there was no difference was over-the-counter ibuprofen.
The other paper was by Peat in Arthritis Care and Research3. There are two articles on the methodology online. This was a North England study of people registered in clinics who reported persistent knee pain. The question addressed was: Which adults with knee pain had radiographic OA? The study included a total of 3,106 adults age 50 and older who reported persistent knee pain in the prior 12 months. They were invited to come into one or more of these clinics for evaluation. A little more than one-third came in for evaluation. There were differences between those who came in and those who did not that might influence the results, and those were described in the paper. They had a weightbearing, semiflexed PA [posteroanterior] view, so not fluoroscopic, and also a supine lateral view with the knee bent at 45 degrees, as well as a skyline view. There were 650 who had complete x-ray data and indicator data examined for association with radiographic knee OA. The definition of radiographic knee OA here was a K/L [Kellgren and Lawrence] scale grade 2 or higher or the presence of osteophytes on either the lateral view or on the skyline view at the inferior or superior margin of the patella. Seventy percent had radiographic knee OA, which I would consider to be high for old data based on one view. In other words, the more views that are obtained, the more likely one is to find osteophytes. What investigators did then is to compare the 452 people with radiographic knee OA with the 198 who did not have radiographic knee OA to identify indicator variables of knee OA. Of the 57 variables, 17 were not associated with radiographic knee OA, 11 were excluded because of high collinearity, and of the remaining 29 indicators with a P value less than 0.10, 10 were independently associated with the presence of radiographic knee OA in the final multivariable analysis. I have listed the 10 retained right here. They are age, sex, BMI [body mass index], pain that started after an injury, difficulty descending stairs, palpable effusion, fixed-flexion deformity, knee flexion ROM [range of motion], coarse crepitus, and pain in the whole leg. The presence of pain in the whole leg reduced the likelihood of knee OA.
2 Dillon CF, et al. Prevalence of knee osteoarthritis in the United States: arthritis data from the Third National Health and Nutrition Examination Survey 1991-94. J Rheumatol 2006;33:2271-9.
3 Peat G, et al. Estimating the probability of radiographic osteoarthritis in the older patient with knee pain. Arthritis Rheum 2007;57(5):794-802]Arthritis Rheum 2007;57:794-802.
The authors identified cutpoints for each of these and came up with a scoring system. From the scoring system they estimated the likelihood of someone having radiographic knee OA, and this was one of the limitations because the area under the curve was 0.8, which is pretty good. So if you had a calculated probability above 79 percent, it turned out that 94 percent did have radiographic knee OA. With a probability of less than 20 percent, only 10 percent had knee OA. So what you might infer is that if you applied this model in primary care, you could then avoid getting radiographs in some people, save money, and manage people appropriately for symptomatic knee OA or look for other causes of knee pain, whereas for the uncertain group in the middle, you would get a radiograph. So that is what I wanted to comment on as an introductory discussion about persistent knee pain and frequent vs. infrequent pain. The point is to look at the entire joint as opposed to just the tibiofemoral joint because you will misclassify people.
I have two slides on risk factors, and people will recognize them. One is from David's [David Felson, M.D.] Annals of Internal Medicine paper4 , which summarizes Stepping Away. We have the systemic factors here on the left [gestured to slide], with the nonmodifiable and the systemic modifiable, and the ones on the right, which are modifiable, and which, working together, lead to the development of OA. And Joanne [Jordan, M.D.] will recognize the other slide from her chapter in Rheumatology, showing systemic factors on the far left. Most are nonmodifiable, except for nutrition. Intrinsic joint vulnerabilities would be modifiable, and certainly loading factors, which apply to the joint, would be modifiable. Our discussion will focus primarily on modifiable risk factors. I apologize that I left the NIAMS slide out.
4 Felson DT, et al. Osteoarthritis: new insights. Part 1: the disease and its risk factors. Ann Intern Med 2000;133(8):635-46.
Presentation of White Paper
C. Kent Kwoh, M.D.
University of Pittsburgh, Pennsylvania
First of all, I want to thank Gayle [Lester, Ph.D.] for asking me to prepare a draft of the white paper. I feel like I'm bringing coals to Newcastle since most of the studies that we are talking about were done by the people here or the ones on the phone.
With the white paper, I tried to summarize risk factors for the development of knee OA since the conference in 1999 that Marc [Hochberg, M.D.] mentioned, so David [Felson, M.D.]'s paper forms a starting point, and we'll focus on modifiable risk factors. I tried to include observational, epidemiologic studies. I tried to include mainly those with radiographs, not to ignore symptoms, but the idea was to summarize the data, and doing so across different radiographs was hard enough, let alone across symptom definitions.
We can discuss what is an optimal radiograph, but, really, most of these studies included weightbearing, AP, fully extended views. Very few studies included lateral or skyline views. I have a large spreadsheet that talked about how a radiograph was defined, but I did not include it in the already busy paper. Most studies defined OA as having a K/L grade 2 or greater, although there is variability in the studies. I tried to include studies that reported a strength of association for knee OA, and most of these were odds ratios, and most of the time there were adjustments. There was variability in the studies of what was adjusted for, so it is hard to compare odds ratios across studies.
I did not include studies based on interventions. I did not focus on the newer genetic studies, or biochemical markers or animal studies. We searched from July 2000 through August 2008. We did a full text review of 221 citations. We summarized data from 38 papers.
We looked at demographics, and we saw an increased risk with age, with being a woman, and with African Americans. And, in Johnston County, there was more OA in blacks than in whites, particularly in women.
Looking at joint involvement, we saw an increased risk with Heberden's nodes, with radiographic hand OA.
There were a lot of studies on BMI and body composition. Increasing BMI was a greater risk. Waist circumference and waist-to-hip ratio was a higher risk. Also, data from Sweden showed that obesity at a younger age was important.
In terms of BMD, what we saw confirms what was reported in the earlier papers: there was an increased risk of knee OA with higher bone mass and also an increased risk of knee OA with decreased bone density in the femoral neck or the lumbosacral spine.
Malalignment is an important risk factor in progression. Not many studies have looked at it as a risk factor for development. The Framingham data didn't show a relationship between various alignment factors and knee OA. The Rotterdam study showed that varus malalignment increased risk.
It wasn't a modifiable risk factor, but I thought it was important to include the study from Johnston County that showed that limb length inequality increased knee OA risk. The Beijing study showed that greater knee height was associated with both radiographic and symptomatic knee OA. Also, finger length — having the second digit shorter than the fourth — and a round back also increased risk.
We know knee injury is an important risk factor in men and women. Menisectomy increased risk. Degenerative meniscal tears increased the risk of knee OA.
Looking at occupational exposure, manual labor increased the risk of knee OA. Building construction increased risk. Lifting more than 10 kilograms 10 times in a week increased risk. Climbing more than 15 flights a day increased risk. Squatting, in work done with the Beijing OA cohort and a Thailand study, increased the risk, and there was also increased risk with side knee-bending. You might use this position in yoga and tai chi; you might need to modify the regimen.
There was no increased risk with recreational walking, but it was hard to interpret data about increased risk with regular exercise. The risk was not so much participating in sports but in the co-occurrence of playing sports with knee injury.
Hormone replacement did not demonstrate increased risk.
Quite a number of modifiable risk factors have been associated with development. In terms of targeting, we certainly could look at higher BMI at younger ages, also certain occupations, and the co-occurrence of obesity, knee injury, and hand OA.
Discussion of Risk Factors, Part I
Opening the Discussion
National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH, Bethesda, Maryland
Some of the questions we need to consider are: What do we know with regard to each of the proposed risk factors? What is the level of evidence for each of these? Are there interactions between risk factors? And can the risk factor be easily measured? Are data being collected or even available on some of these large epidemiologic studies in Japan and Australia, as well?
I had made a list and grouped together things pretty grossly. Kent's [Kwoh, M.D.] list was broad but had nice subcategories, but from a gross picture, working down — the slide shows the non-modifiable ones.
Most risk factors will be in the modifiable category, with obesity and hormonal factors and biomechanical ones, including those listed on this slide and muscle strength, and also activity level. This includes work from the Beijing study and work that comes out of Stefan's [Lohmander, M.D., Ph.D.] laboratory. Also posttraumatic OA will be covered.
Mary Fran [Sowers, Ph.D.] will talk about hormonal factors, Tim [McAlindon, M.D.] will talk about BMD, and Marc [Hochberg, M.D.] will start with obesity.
Marc C. Hochberg, M.D.
University of Maryland School of Medicine, Baltimore
So, one word says it all: overweight. If you look at the 2000 article that really summarizes the data going back, and then you look at Kent's [Kwoh, M.D.] white paper, and you use this best evidence synthesis technique, I conclude that there is strong evidence that higher weight or BMI is associated in a causal manner with the development of radiographic knee OA and symptomatic knee OA. The association is clearly highest in those that fit our definition of obesity.
There are two main papers here. One is the Felson paper5 from the Framingham data that shows that weight loss reduced the odds of developing symptomatic radiographic OA. The other is Christensen's meta-analysis6 from four trials — but most of the data comes from just one study, the Messier study, called ADAPT [The Arthritis, Diet and Activity Promotion Trial], that found a small effect of weight loss, in conjunction with exercise, on both pain and function in patients with symptomatic radiographic knee OA. The difference between the control group and the intervention group was only in those with both interventions, weight loss through nutritional intervention and activity.
5 Felson DT, et al. Weight loss reduces the risk for symptomatic knee osteoarthritis in women: the Framingham Study. Ann Intern Med1992;116(7):535-9.
6 Christensen R, et al. Effect of weight reduction in obese patients diagnosed with knee osteoarthritis: a systematic review and meta-analysis. Ann Rheum Dis 2007;66(4):433-9.
So, I added a slide for discussion. Obesity is really migrated in all of the figures into a biomechanical local factor that influences the development of OA radiographic progression. There always has been an issue of whether there is a systemic effect, what used to be considered fat distribution and is now considered metabolic syndrome. NHANES III data, when stratified for BMI distribution, still shows an effect of waist circumference, which is a surrogate for distribution of visceral fat. Also, is there a role for leptin and the adipokines for the development of OA? It's a small literature and there is conflicting data.
A group in France has done work on the contribution of white adipose tissue in joint disease. There was also work presented at the meeting in Chicago from Duke, in a study of mice. Tim Griffin [Ph.D.] has some unpublished data, looking at the development of OA in those mice.
[Dr. Lohmander took the podium.]
Stefan Lohmander, M.D., Ph.D. (Univ. of Lund, Sweden): I will show three slides on obesity. We have data from 28,000 people from southern Sweden for a diet and cancer study that never really panned out, but we used that cohort for other things. We collected BMI. We merged it with the Swedish National Registry for knee replacement. So this is knee survival [pointed at slide], the proportion of knees that did not have total replacement, after about 11 years. The colors represent the BMI quartiles. In quartile 1, the medial BMI was 23/21. Clearly, things are not changing in Sweden as they are in the U.S. The graphic shows that the risk for total knee replacement is higher for high BMI, and is already quite operative at these BMI [pointing at slide]. Also, we looked at some other factors of body mass, and weight and waist circumference were major risk factors. An increased incidence of severe knee OA in women with metabolic syndrome was explained by BMI. We also, in a subset, did an analysis of the risk association between metabolic syndrome and CRP [C-reactive protein], and we did not find an independent association as soon as we corrected for BMI. These data are limited by the fact that we simply corrected for BMI, with the limitations that that has, and also that the data came from 15 years before the actual knee replacement took place, which limits the conclusions you can make. But they do show BMI's effect on the risk for knee replacement.
Discussion of Risk Factors, Part II
Biomechanical Risk Factors
Joel Block, M.D.
Rush Presbyterian Hospital, Chicago, Illinois
As the nonepidemiologist at the table, I am going to ask you to put your experimentalist hats on as I talk about aberrant loading and risk factors. It is clear that the peak knee adduction moment is higher in subjects with knee OA than in age-, gender-, and BMI-matched normals and that load is directly related to the severity of pain. Magnitude of load is intimately related to the progression of lower extremity OA.
Does aberrant loading precede OA, in which case it is a risk factor, or does it result from knee OA, in which case it might be a biomarker? Most animal studies that show that you can induce OA with load are more a trauma model than a loading model. Najia Shakoor [M.D.] looked at it in unilateral hip replacement and OA, and these people often go on to get knee OA. Three-quarters of them, when they progress to knee OA, get it in the contralateral knee and not the ipsalateral knee. Preoperatively, sure enough, they overloaded their knee. Two years later, they had clearly had a gait adaptation. They thought they were walking normally, but the overloaded knee was still there. It sets up a paradigm of high-risk people for contralateral knee OA.
The relevant issue is the asymmetry. We have this paradigm of a population that is going to progress in a contralateral vs. an ipsalateral state, and it allows us to ask the question: What is the contribution of aberrant or asymmetric load to OA? And we are in the midst of an ongoing study with clinically normal knees, but that are at high risk for knee OA. The subjects have moderate or severe hip osteoarthritis. These data will be presented next month in San Francisco. Clinically normal people have gait asymmetry already and contralateral loading vs. the ipsalateral knee.
I want to talk for a moment about neurosensory deficits. We've known for centuries that there is a neurosensory deficit in OA. Leena [Sharma, M.D.] has done a lot of the work on proprioception, but it is hard to do. Najia Shakoor started looking at vibratory pathways as a surrogate for proprioceptive, and it's easy and cheap. Sure enough, people with knee OA have a much higher vibratory threshold than age-matched normals. Compelling animals models and observational studies show it is a compelling risk factor for the progression of OA. It is harder to tell for new onset.
Leena Sharma, M.D.
Northwestern University, Chicago, Illinois
I'm taking a narrow approach, going over only what is known about incident radiographic OA and medial-lateral laxity. By laxity I mean abnormal rotation or displacement of the tibia with respect to the femur. Medial-lateral laxity could lead to abrupt motion and large displacements, altered congruence and contact regions of opposing surfaces, and an increase in regional shear and compression forces. We developed a way to measure medial-lateral laxity static, and we focus on this valgus rotation plane in our study, which has gone on since 1997. What we and others have found is: (1) that medial-lateral laxity increases with age, (2) that it is greater in women than men, (3) that it is associated with greater risk of OA after ligament injury, but these are smaller studies, (4) that it is present before full-blown OA and exacerbated by aspects of OA, (5) that it is associated with worse function, and also (6) that it alters the strength-function relationship. However, it does not seem to have a strong association with medial or tibial cartilage loss.
We looked at quantitative and semiquantitative MRI outcomes. It isn't possible to use radiographic methods to determine the effect of laxity on cartilage; there are too many reciprocal changes with laxity. You need a method like MRI. We waited all these years, we did it with MRI, and the results were not impressive. Now, this is just one measure of stability. It just means that static medial-lateral laxity is not a predictor of progression in our study. My gut feeling is that it will not be associated with incident OA either. In progressive OA, the milieu of the joint is much more vulnerable to instability. Alignment influences load distribution at the tibiofemoral and patellofemoral compartments, and several studies now show a relationship between malalignment and disease progression. If one reviews existing studies on this topic, they differ in their results and methodology, but the bottom line is that it is probably too early to say what the relationship is between malalignment and incident knee OA. We'll just see as time goes on how this risk pans out. It does seem that the relationship between malalignment and incident knee OA is probably going to be weaker than with progressive OA. It probably has an effect in and of itself, but it probably also affects other factors.
Activity: Recreational as Well as Occupational
Stefan Lohmander, M.D., Ph.D.
University of Lund, Sweden
I would say that my personal view is that even quite active activity is not strongly associated with the development of OA if you correct for injuries. Injuries are the key here. If you don't correct for injuries, I'd be careful using the data.
Stefan Lohmander, M.D., Ph.D.
University of Lund, Sweden
Ewa [Roos, Ph.D.] asked me to bring this slide along. The point is that joint injury, aging, obesity, and gender aspects can be translated into changes in muscle function — such as weakness, loss of endurance, neuromuscular dysfunctions, and microklutziness — which result in increased joint loads and lead to joint failure. Now, my caveat, not being in this particular area, is what is the consequence and what is cause?
Intuitively, it makes good sense to me, and I think this is something that we should look into a little bit better. Maybe the outcomes we have been looking at so far are not the ones to help us understand muscular function or neuromuscular function in OA. There are some studies, maybe not of the quality we wish, that are pointing in this direction.
Injury, Part 1
Stefan Lohmander, M.D., Ph.D
University of Lund, Sweden
We maybe just need to discuss the quantity a bit. The reference at the bottom of the slide is to some Swedish data. Roughly speaking, in the U.S. you are doing 150,000 cruciate ligament reconstructions in the knee annually and up to 1 million arthroscopic meniscus resections annually, and about half of these individuals get radiographic OA. But let's look at what is what here. You might believe that the numbers refer to injuries, trauma, etc., but I've plotted cruciate ligament reconstructions and arthroscopies to age, using population data from Sweden, and what you see is that these are two quite different things. Cruciate ligament repairs are the classic ones, with women having them at 18 to 20 and men a few years later, but the vast majority of meniscus surgery is in middle age. The average age in Sweden for it is about 55. I am willing to bet it is about the same in the U.S. We know that form of surgery is not really effective in that particular age group. What I'm seeing, and Jody [Buckwalter, M.D.], too, is that in the classic cruciate repair group, some of the meniscus interventions also result from injury — trauma to the normal joint — but the majority result from something quite different. Knee ligament lesions are common in unselected individuals over 50, but 20 to 60 percent and upwards have significant mensicus pathology, which increases with OA. On the one hand you have trauma to the normal joint, and on the other hand you have something which is quite likely associated with OA development. It influences our interventions and our thinking with regard to injury prevention. All of these surgeries are not done as a result of a trauma, as we tend to think of them.
This slide shows the prevalence of radiographic knee OA in men and women with ACL tears after 12 to14 years of followup. You see knees that have aged very prematurely as compared to the healthy population. These individuals have injuries around the age of 20, and 15 years later, 50 percent have symptomatic radiographic OA. And this is where prevention could have an effect. We know that prevention programs can bring down rates, and we also know that nothing we can do in the surgery prevents OA. In fact, if anything, the surgery increases the risk of OA.
Many of these subjects went back but quit fairly quickly thereafter. They cope for a couple of years, and then the problems start kicking in. These people in our study were all playing soccer. They were not elite professionals, but they were active athletes. A few were bilateral, but most were unilateral. So, surgery doesn't help OA, and injury prevention does help. Soccer injuries are different than any skiing injuries. Prevention programs would be aimed at getting individuals to not return to the activity. Obviously, that is an ethical dilemma for the sport. We do know that women and men injure themselves to about the same extent per sport hour, but women are more likely than men to injure their knees. This was a relatively small group as compared to the meniscal surgery tears. These individuals did not have radiographic OA in their mid 30s. You will note that those individuals in their 30s or 40s that have surgery for meniscal lesions have OA 20 years later. The number is approaching 100 percent, but it is not posttraumatic OA as we have been thinking about it. These individuals — mostly I am talking about those with degenerative lesions — have an increased prevalence of hand OA at a younger age, and we also know that they have an increased risk of OA in the contralateral, uninjured knee. And yet this is not traumatic OA, but they get labeled as post-menisectomy. They are actually just young people with old knees. It is likely that OA is not a disease of the tissue but of the joint. That damage in the cartilage also occurs at the same time in the meniscus, and the fact that orthopedic surgeons see these patients is because they come to them from primary care with suspected meniscal lesions, but they really have early OA. The OA becomes symptomatic at an earlier age than the average. It is not posttraumatic OA. This is something different. An arthroscopic resection is not a good treatment for early stage OA. So, consider that ligament lesions in healthy knees can be affected by prevention programs; consider OA of the meniscus as an integral part of the OA pathology.
Injury, Part 2
Jody Buckwalter, M.D.
University of Iowa, Iowa City
I�m going to focus on once someone has a joint injury, what are the risk factors? We talked about age and injury. It was interesting to me that you have to be over 50 to meet some definitions of OA, and, as Stefan [Lohmander, M.D., Ph.D.] said, most of the patients under age 50 who present with disabling OA have an injury or joint dysplasia. After 50, if you have an injury, your risk of OA is nearly doubled, which fits in with what Stefan was saying.
There are differences among joints. Nearly 80 percent of those presenting with end-stage ankle OA have injuries. Most of them have had a fracture. The risk factor for OA in these end-stage ankle patients is severity of injury. How can we get at severity as a risk factor for subsequent OA and not do it crudely? With a fracture — here's a tibial pilon fracture — you can calculate the energy of injury by using a high-resolution CT scan to look at the liberated fracture space. If you calculate the severity of injury by the energy it took to injure it, we can target who is going to develop OA after a few years. We're looking for early interventions immediately after injury. Again, looking at a tibial pilon fracture, you can calculate the distribution of contact stress over time, and after 2 years there is a very striking threshold effect. Those that have less than 3.6 MPa-s [megapascal-seconds] have no OA, and those over 4.0 MPa-s all have OA. And we know that even skilled reconstruction of the joint can't put it back together perfectly.
So, which patients, because of individual incongruity, are at risk? We looked at the degree of instability vs. cartilage degeneration, and with increasing instability we get increasing cumulative contact stress within 12 weeks. Now, in humans this is harder to study, but we can measure increased contact stress at every part of the joint. We can model a stable or unstable joint, and if we make a 2 mm incongruity, we increase the stress, but we also increase the peak contact stress rate nearly fourfold. And once you have had an injury, the next 48 hours have a lot of effect on instability, so we can talk about what are the risk factors once you have an injury, and we can identify those people at risk and deal with residual factors. The knee acts as a hinge, and it is much more complex to model. The knee seems to be able to handle instability better than the ankle, but motor control seems to be more of an issue in the knee, so it is much more important. The ankle is very constrained; there is very little side-to-side motion. The hip is also very constrained. It almost never subluxes in normal patients. This shows that incongruity is a risk factor.
MaryFran Sowers, Ph.D.
University of Michigan, Ann Arbor
There has been an interest at looking at hormones because women have a greater risk for prevalent and incident OA. It appears to be an increased incidence in OA that extends beyond menopause, and that, it appears, is due to more than just age. It has been suggested that sex hormones have a role. We have some mixed messages on this. There is some rationale for why estrogen might be important. It has been associated with synoviocytes, chondrocytes, and bone receptor cells. We see the presence of estrogen receptors on chondroblasts, chondrocytes, and mandibular articular cartilage. But these studies do not address estrogen's role in the inflammatory process, and they don't address pain, so what is important to think about is the recognition that when you talk about sex steroids, you are really talking about a multiplex system. And in people, testosterone is aromatized to estrogen, which is estradiol (E2), which is in turn converted to estrone (E1).
And as we move from androgens to estrogens, what is important here is fat tissue. So if you look at types of estrogen [points to left column of slide], at premenopausal women with circulating normal levels of estradiol, they have OA, but postmenopausal women, well, one or two reports again suggest that estradiol has an effect, and again it is suggested by BMI. The third line down shows bioavailable estrogen. If you consider that only 1 to 2 percent of those sex steroids are biologically active because the rest are complexed to SHBG [sex hormone-binding globulin], my read of the literature is that people have not looked at the bioavailable fraction of either estrogen or testosterone. The bottom line of the table is HT [hormone therapy] use. There is a body of literature that says HT users have less OA. But the publications, I argue, are somewhat naive in their assumptions about what you are replacing. It is whether or not they have an action that is really important to discern. It also depends upon the amount of SHBG levels.
I would like to demonstrate how important obesity is to circulating estradiol levels. On the left slide, in white, I have placed circulating estradiol. The active hormone is on the vertical axis, and on the bottom axis are the years before FMP [final menstrual period]. The estradiol levels drop like a rock 2 years before the final menstrual period, and then they continue to drop. The right slide shows rate of change. On the vertical axis you see how much change happens longitudinally, and, again, on the bottom axis are the years around FMP. The maximum rate of change is reached at the last menstrual period, and then it slows down. Interestingly enough, we have now reported that there is a second drop, to even lower levels, in the 6 to 8 years after menopause. These patterns are different based on whether you are obese or not obese. There is a load drop at FMP, then it slows down, then it raises again, and the second decline is observed in those with low BMI. In contrast, it is a little later and a less defined drop, and then no secondary decline, in those with higher BMI. You must interpret these changes with respect to the body size of an individual.
When you look at HT without progestin or with progestin, you will see 2 to 3 times the amount of these estrone products that are actually appearing in the HT group, but their SHBGs are remarkably higher. It is going to make all the difference in the world as to whether there are circulating hormone levels. When you are looking at estradiol and HT, you cannot look at them without looking at obesity. This has been largely ignored, and it is one of the things contributing to why there is a mixed understanding concerning hormone levels.
Why do we look at testosterone and OA? Androgen receptors have been identified in chondrocytes, osteoblasts, and osteocytes. There is a very small body of literature on testosterone and OA. It has been examined in mouse models, and there was a favorable response found. A couple of reports have found that women with generalized OA had higher testosterone levels. Some studies in young women found that testosterone was not associated with knee cartilage volume or bone structures. I think that Kent [Kwoh, M.D.] is right in his conclusion about HT, but the reason he's right is that this is proven to be a more difficult system to study, primarily because of its association with obesity. It might modify the association between endogenous and exogenous OA. I would also like to argue that HT and its impact is an important thing to study when you have a population that has 25 percent exposure and when that population is a vulnerable group and taking it at a vulnerable time.
Tim McAlindon, M.D.
Tufts Medical Center, Boston, Massachusetts
Thirty percent or more of load is absorbed by bone, and cartilage absorbs less than 30 percent. The animal model shows that bone changes happen before detectable cartilage changes. When you have higher subchondral turnover, it is at the expense of material density. Macroscopic subchondral damage is common. All of these observations have led to the hypothesis that bone characteristics might influence expression.
There is a hypothesis that people with softer bone might have more erosive OA. I describe this theory not because I think it is resolved but because I think it offers a potential for intervention if we can figure out what is going on. There is a possibility we can do things for bone health. Similarly, vitamin D, if it is in fact helpful for OA, would be a relatively easy intervention.
Periarticular BMD seems to be associated with knee OA. There is a radiograph study, in the direction that higher density predicted more progression. In terms of epidemiologic studies, all of them show effects, but it is confusing to determine in what direction. Framingham showed a differential effect, but there appeared to be a strong inverse relationship so that those who lost BMD at the femoral neck seemed to have higher rates of progression. Similarly, in the Chingford study, those with lower BMD had more radiologic progression. So, we appear to see an effect, but we need to figure out what is going on here. Is it real or is it a statistical anomaly? Vitamin D did show a strong effect. That was corroborated in a hip study of osteoporotic fractures, and that definition of OA included incidence. On the other hand, SOF [the Study of Osteoporotic Fractures] and the BOKS [Boston Osteoarthritis of the Knee Study]/MOST study have been mostly negative. There are more data emerging, I believe, but those are not yet in the public domain. Finally, there are some tantalizing findings regarding VDR [vitamin D receptor] genotype, and some of them are positive. The definitions in terms of genotyping and of expression of OA are different in all studies.
Joanne Jordan, M.D. University of North Carolina, Chapel Hill
The Johnston County study started in the early 90s. We looked at the differences between African Americans and whites, added hand films and spine films, and have found racial differences in a lot of things. Amanda Nelson [M.D.] presented at OARSI some preliminary data. There are 800 people from our cohort. We looked to see if patterns of joint involvement were different in African Americans than in whites. The reason was mostly clinical. Many clinicians don't see the same hand patterns in African Americans as in whites.
We defined hand OA using the GOGO [the Genetics of Generalized Osteoarthritis study] definition. You needed three joints at a minimum across both hands. We looked at PFJ [patellofemoral joint], TFJ [tibiofibular joint], and hip, and we found clear differences by race. [Pointed to slide.] So, as you see on the left slide, African Americans had less of any phenotypes that involved hand OA. There was no difference whatsoever with hip OA. There was only statistical significance with PFJ. There was more large joint OA with African Americans, not only in the knee but in combination. So, we also looked at an adjusted model, and after we adjust, we see that hand OA — similar to what you found in China, where no hip osteoporosis was found — had a dramatically different odds ratio. When OA was more common, knee and hip were still more common after adjustment. We also see that none of these were more common in African Americans. We are suspicious that some of this [gestured to slide] is associated with genetic subsets, although some of it could be occupational. Certainly, we see it in a genetic subset in whites, so we're wondering. We were also surprised to see so few with hand OA, clearly a nonmodifiable risk factor. It is a good idea to think about the interactions of risk factors in these particular groups.
Status of Interventions for Risk Factors
Bracing and Osteotomy
David Hunter, M.B.B.S., Ph.D.
New England Baptist Hospital, Boston, Massachusetts
I am going to talk about braces and osteotomy. Probably it is best understood at the hip. With hip dysplasia and a predisposition to hip OA, there is a vogue now in certain centers to perform certain procedures to correct hip dysplasia abnormalities. There is some evidence that you can predict the success of these procedures, but it is not clear what would happen from a natural history perspective.
There are two other interventions, primarily in people with symptomatic OA, osteotomy and the use of braces. Bracing could have applications elsewhere if developed for people with certain risk factors that predispose them to knee OA. The OARSI guidelines were recommended by varus and valgus instability. With regard to osteotomy [pointed to slide], a review was done.7 These were obviously uncontrolled series and studies; most of the uncontrolled were not compared to conservative therapy. Probably most important for a research agenda, there is no evidence to suggest that osteotomy is better than conservative therapy alone.
7 Brouwer RW, et al. Osteotomy for treating knee osteoarthritis. Cochrane Database of Systematic Reviews 2007, Issue 3. Art.No.:CD004019.
There is good evidence to suggest that bracing can correct alignment by 2 to 2.5 degrees. An anecdotal study suggests you can open up medial joint space after application of the brace. In the unbraced, compared to 4-degree brace, the data suggests a meaningful magnitude of effect on mechanical forces to the knee. It can also have an influence on symptoms. A 1999 paper8 compared bracing with a sleeve, and it showed better improvement than many other interventions on their own.
8 Kirkley A, et al. The effect of bracing on varus gonarthrosis. J Bone Joint Surg Am 1999;81A(4):539-548.
Joel Block, M.D.
Rush Presbyterian Hospital, Chicago, Illinois
Again, as I said before, loading is clearly important in the progression of prevalent OA and also of incident OA, although David Hunter [M.B.B.S., Ph.D.] disagrees. The entire lower extremity acts a single kinetic unit. If you alter one level, you can expect to have effects at different levels.
With that in mind, you can imagine targeting the foot, the hip, or the entire gait cycle itself. But, in contrast to therapeutic intervention, risk factor modifications would require no risk and minimal interruption in lifestyle. Now, inserts are the best studied of these. The rationale for lateral orthotic inserts is that you change the spatial position of the femur, tibia, and calcaneus; you make the mechanical axis more upright; and you decrease loading across the medial knee. The hope is that decreasing the load might relieve pain and retard disease progression.
There are lots of good lateral wedge studies, but few with placebo control. In a 2-year study of lateral wedge vs. neutral orthotics, wedge angle was customized for each individual. Even average angle was not reported. The primary outcome was the WOMAC, and they found no effect, although wedge-users may have used less nonsteroidals. [David] Felson [M.D.]'s group reported a short-term crossover, and they found no effect at all of wedges. [Yoshikata] Toda [M.D.]'s group in Japan think they are getting a pain advantage with a strapped wedge, but they always compare it to the regular wedge, not a neutral. When you study a wedge, pain is probably an irrelevant outcome. The more relevant parameter is biomechanics. And it appears the wedges do have an effect, maybe a 6 to 10 percent reduction in stress across the medial knee. What is relevant from a risk factor perspective is whether these are durable over time, but it looks like there is a mechanical effect of the wedges. We just finished a 36-month wedge study and found that the unloading of the medial knee is durable over 3 years. Compared to neutrals, we actually found a radiographic difference. The neutral group lost joint space, but the wedge group remained the same.
I mentioned that the proximal tibial plateau is a good long-term measure of history, and people given neutral orthotics continued to increase in BMD. The wedge people did not increase at all. There is evidence that wedges have a durable mechanical effect, and it might be mechanical, as well. They are, clearly, pretty well tolerated.
The second strategy that targets the feet is barefoot technology. Just taking someone's shoes off significantly unloads their knee. So simply altering the shoe while keeping it stylish could have a large-scale societal effect. We demonstrated a significant reduction in medial loading of the knees with the barefoot shoes.
Braces do unload knees, but trying to get an asymptomatic 30-year-old to wear braces for the next 30 years is unlikely. Again, canes and walkers could help, but healthy people will not use them.
Muscle strength provides joint stability and affects dynamic loading, but there is no evidence that quadriceps strengthening affects knee loading. Anyway, so 2 years ago we did a pilot study and exercised people's hips and got an overall decrease in medial knee loading of 13 percent.
Finally, there are a variety of gait strategies: reduce walking speed, increase the toe-out angle, and increase mediolateral trunk sway. As far as therapy goes, a variety of therapies are promising in the long run for structural modifications, but maybe not for pain modifications. Behavioral modification does not really seem to be as successful over the society, but that's your area more than mine. There is a single randomized, controlled trial that provides evidence that prolonged unloading modifies the progression of structural degeneration in established OA. Flip-flops replicate barefoot walking, so there's a behavior modification!
Status of Surveillance, Risk Assessment
Chad Helmick, M.D.
Centers for Disease Control and Prevention, Atlanta, Georgia
Public health. Here's the definition we use: What we as a society do collectively to assure the conditions in which people can be healthy. What we are doing at the CDC we got our direction from the National Arthritis Action Plan — is focusing on things not being done in a clinical setting, behavioral things. A short definition of surveillance is that it is ongoing. A conceptual approach to arthritis surveillance is to focus on it as a generic condition. Through 1995 we used the NHIS [National Health Interview Survey] condition list and ICD-9 [international classification of diseases]-based codes for arthritis and other rheumatic conditions. We also use a self-report approach.
The problem we have for specific conditions is self-report surveys. When you ask people what they have, 50 percent of people say "other" or that they're not sure. They don't know what they have. Since self-report is not useful for OA, we need some level of clinical involvement, whether that is radiographs or some level of pain remains to be seen.
The National Arthritis [Data] Workgroup put out some estimates, but they are not surveillance. There is no way to take into account temporal trends in obesity. In the same paper we looked at some small studies, but the issue becomes generalizability. Jeff Sacks [M.D.] in our group came up with a surveillance approach for pediatric arthritis that takes the NAMCS [National Ambulatory Medical Care Survey] survey into account. That is another way of doing that. By using ICD codes — OA specific codes — we can determine the number of visits, and from that come up with a prevalence estimate. It would be a minimal estimate that did not include people who don't come to the health care system or those who don't report it to their doctor. Another issue is: What is the research question we really want to answer? What definition of knee OA do we use? Those are just some of the things I wanted to talk about. I'll be happy to address comments.
Summary of Next Steps
National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH, Bethesda, Maryland
What is the future of MRI for the identification of new risk factors? I think David [Hunter, M.B.B.S., Ph.D.]'s point is a good one: How do we facilitate the next big steps? We need trials that are sufficiently long. We want to with this meeting today identify some of the areas where we need to move forward and what we should encourage the research community to think about when they apply for their grants.
I would like to tell everyone how much I appreciate their participation and the interactions we had today. It has been stimulating to hear everyone speak their minds openly and to shine lights on what's really been done well and what hasn't. I know a lot of you have traveled a long way, and myself, and Joan [McGowan, Ph.D.] and Chhanda [Dutta, Ph.D.], and the whole NIH community appreciates your effort.