Knee osteoarthritis (OA) is the most common joint disorder among adults in the US, and severity of radiographic knee OA is strongly associated with knee pain (1). Pain from knee OA accounts for a large proportion of limitations in common activities of daily living (2), and it is also the main reason for total knee replacement (3). Because of the impact of symptoms from knee OA, people with symptomatic knee OA may be more motivated to take steps to prevent disease progression than those without radiographic OA. Thus, investigators and persons with knee OA are particularly interested in risk factors that are associated with radiographic OA progression because such knowledge will provide insightful guidance for secondary prevention (4).
Over the past few decades, many observational studies have examined risk factors for the occurrence (i.e., incidence) and worsening (i.e., progression) of knee OA. Several risk factors (i.e., female sex, obesity, high bone mineral density [BMD], joint injury, repetitive occupational stress on joints, and certain sports) have been found to be strongly associated with an increased risk for incident radiographic knee OA (5, 6). In contrast, findings on risk factors for radiographic OA progression have been inconclusive. Except for the level of serum hyaluronic acid and generalized OA, no other risk factor has been consistently associated with the risk of radiographic OA progression (7). Interestingly, some risk factors (e.g., high BMD, low vitamin C) increase the risk of incident radiographic knee OA but are not associated with, or even decrease, the risk of radiographic OA progression (8–11).
This paradoxical phenomenon is not limited to studies of radiographic OA. Many studies have shown that being overweight and obese increases the risk of cardiovascular disease; however, a number of studies have also reported that among subjects with preexisting cardiovascular disease, those who are overweight or moderately obese have an improved survival and lower risk of major cardiovascular events compared with subjects with normal weight (12). These findings have been termed the obesity paradox. Similarly, although low body mass index (BMI) is associated with increased incident chronic obstructive pulmonary disease (13), it does not increase the risk of recurrent exacerbations of the disease (14). Although the risk factors for incident events may be biologically different from those for secondary events, there are also compelling methodologic explanations behind such discrepancies.
In this article, we provide several explanations that may underlie the discrepancy between findings for radiographic knee OA progression and those for radiographic knee OA incidence. To be consistent with most studies, we consider knees eligible for studies of radiographic OA progression as those that have preexisting mild (Kellgren/Lawrence [K/L] grade 2) or moderate (K/L grade 3) radiographic OA at baseline. We discuss how study design, study implementation, and outcome measures in studies of radiographic OA progression can potentially bias the effect estimates of risk factors of interest using causal diagrams (15). We also present findings from several large prospective studies of radiographic OA progression to facilitate our explanations wherever applicable.
Randomized controlled trials to assess the relation of a risk factor to radiographic OA progression
We begin with the premise that an ideal design to evaluate the effect of a risk factor on progression of radiographic knee OA is a randomized clinical trial. Specifically, to test whether an intervention (e.g., a drug or weight loss) may decrease the risk of radiographic OA progression, subjects who have mild or moderate radiographic knee OA would be randomly allocated into appropriate intervention groups. At the end of followup, subjects' knee radiographs would be scored blinded to the intervention assigned. The effect of the intervention on the risk of radiographic OA progression would be estimated by comparing the treatment groups.
Such a trial can be illustrated using a causal diagram that consists of knees that have preexisting radiographic OA at baseline (i.e., study sample), the intervention (i.e., exposure), a set of covariates (i.e., potential confounders), and radiographic OA progression (i.e., outcome) (Figure 1). A causal diagram consists of a set of relevant variables (exposure, potential confounders, and outcomes) and arrows to indicate the flow of causation. An arrow placed with its base at a variable, such as exposure (e.g., intervention), and its head at another variable, such as outcome (e.g., radiographic OA progression), indicates that the exposure causes the outcome. In this causal diagram, the fact that the study sample is restricted to (i.e., conditioned on) knees that have mild or moderate radiographic OA at baseline is denoted by a box around the words preexisting radiographic OA. Because the intervention in the trial is randomly assigned to the study participants, the exposure is associated with neither potential confounders (either known or unknown) nor preexisting radiographic OA as indicated by the absence of arrows directed toward the exposure. Under such a scenario, one can obtain an unbiased effect estimate of the intervention on radiographic OA progression using appropriate statistical methods.
Bias introduced by conditioning on preexisting radiographic knee OA in observational studies
In contrast, observational studies of radiographic OA progression can produce biased estimates of the effect of an exposure on the outcome given the very fact that the study sample is restricted to the knees that have preexisting mild or moderate radiographic knee OA. To illustrate this, we go back to a causal diagram, this time to depict the relationship of a risk factor (e.g., obesity) to radiographic knee OA progression in a hypothetical observational study (Figure 2). To simplify our discussion throughout this article we considered only 1 potential confounder (although this can be readily extended into more complex scenarios with more potential confounders), in this case a genetic factor that increases the risk of both incident and progressive radiographic OA. Note that the genetic factor is not associated with obesity before knees develop radiographic OA. Thus, it would not be a confounder in a study of obesity and incident radiographic OA.
As shown in Figure 2, knees that have preexisting radiographic OA at baseline developed radiographic OA because of either obesity or the genetic factor; thus, preexisting radiographic OA at baseline is a “common effect” of obesity and the genetic factor (i.e., obesity → preexisting radiographic OA ← genetic factor). The fact that there is no association between the obesity and genetic factor before knees develop mild or moderate radiographic OA is indicated by the lack of an arrow between obesity and the genetic factor. The only paths between these two variables are blocked by colliding arrow heads at their common effects nodes, i.e., preexisting radiographic OA (i.e., obesity → preexisting radiographic OA ← genetic factor) and radiographic OA progression (i.e., obesity → radiographic OA progression ← genetic factor).
However, as a result of conditioning on a common effect, in this case preexisting radiographic OA at baseline, obesity and the genetic factor are no longer independent. Such conditioning opens an alternative noncausal path between obesity and the genetic factor (i.e., obesity - - - genetic factor → radiographic OA progression). This alternative path biases the effect estimate of obesity on radiographic OA progression, unless the effect of the genetic factor is appropriately adjusted for. However, in many instances, not all confounders are measured or are known, therefore leading to confounded effect estimates.
Here we provide an intuitive example to illustrate the logic behind potential spurious associations created by conditioning on a common effect (16). Suppose one has 2 fair coins and 1 bell. The bell rings whenever either coin comes up heads on a toss of the 2 coins. Thus, the bell ringing is a common effect of heads appearing on the toss of either coin A or coin B (i.e., coin A → bell ringing ← coin B). Obviously, heads appearing as a result of 1 coin toss is independent of heads appearing as a result of the other coin toss; thus, the correlation coefficient between heads appearing from coin A and from coin B equals 0. However, suppose that we only examined the relationship between heads appearing with the 2 coin tosses in those instances where the bell did ring (analogous to only examining the effect of obesity on radiographic OA progression in people with preexisting radiographic OA). By conditioning on the status of the bell ringing (i.e., conditioning on a common effect), heads appearing from coin A and heads appearing from coin B are no longer independent. For example, if the bell rings and coin A came up tails, then that must mean that coin B came up heads (and vice versa if coin B came up tails). As a consequence, conditioning on the status of the bell ringing induces a negative correlation between heads appearing from coin A and from coin B.
Because we assumed that there are only 2 causal factors (e.g., obesity and the genetic factor) for the development of mild or moderate radiographic OA in our scenario, some cases of preexisting radiographic OA at baseline were caused by obesity, and others by the genetic factor. Although the genetic factor is not associated with obesity before knees developed radiographic OA, among knees that have radiographic OA at baseline, obesity and the genetic factor are no longer independent. This is because if the cause of preexisting radiographic OA is not obesity, it must then be the genetic factor, or vice versa. Subsequently, when evaluating the relation of obesity to the risk of radiographic OA progression, such a negative correlation between obesity and the genetic factor would bias the effect of obesity on radiographic OA progression downward unless the genetic factor is appropriately controlled for.
Ideally, to assess the effect of obesity on the risk of radiographic OA progression one should compare the risk of radiographic OA progression among persons with obesity with those without obesity, with all else being equal. However, in an observational study, persons who have baseline radiographic knee OA but are not obese must have been exposed to other risk factors for radiographic OA. Thus, the 2 groups (i.e., obese versus nonobese) are not comparable in terms of distribution of potential confounders. Such a study can almost be considered to be a study that compares the risk of radiographic OA progression among persons who are exposed to obesity with those who are exposed to the genetic factor as well as other risk factors for radiographic OA. Of course, one should acknowledge that some of the knees with preexisting radiographic OA at baseline may have been exposed to obesity, the genetic factor, and other risk factors. Nonetheless, conditioning on preexisting radiographic OA at baseline would tend to bias the effect of obesity on progression toward the null unless the analysis adjusts for genetic and other risk factors. Unfortunately, not all factors are always known or measured, and therefore cannot be adjusted for, leading to bias.
Using data from the Multicenter Osteoarthritis Study (MOST), we explored this possibility when evaluating the effect of obesity as an example of a chronic risk factor on the progression of radiographic knee OA to illustrate this issue. Among knees eligible for progressive radiographic knee OA, all known risk factors for radiographic OA (e.g., female sex, knee injury, high BMD, and knee malalignment) were more prevalent among persons who were obese than those who were not obese. Because obesity is a strong risk factor for radiographic knee OA at baseline, as are many of the other factors examined, one would speculate that there must be other risk factors, not yet identified, that contribute to radiographic OA development in nonobese people at baseline. If these unknown risk factors are also associated with radiographic OA progression, it would bias the effect of obesity toward the null.
Bias introduced by loss to followup in observational studies
For logistical reasons, few large-scale observational studies of radiographic OA progression have been able to obtain knee radiographs at multiple time points from each participant within a relatively short period of time. For example, the average followup time between repeated knee radiographs was 9 years in the Framingham Osteoarthritis Study, 6 years in the Rotterdam Study, 4 years in the Chingford Study, and 2.5 years in the MOST study. When the followup time is long, a substantial proportion of subjects may be lost to followup, leading to potential selection bias. The rate of loss to followup is high in studies of radiographic OA. Of 1,473 subjects in the Framingham Osteoarthritis Study who had knee radiographs taken at baseline, 40% of them did not have knee radiographs at the followup visit (17). A similar proportion of loss to followup (40%) was also reported in a study in the UK (18).
Loss to followup in observational studies could bias the effect estimate of the risk factor of interest. The status of complete followup is a common effect of radiographic OA progression and obesity (i.e., obesity → complete followup ← radiographic OA progression) (Figure 3). For example, more obese subjects are less likely to complete followup due to poor health, and subjects whose radiographic OA gets worse are less likely to return for the last study visit due to loss of mobility. Because investigators are only able to examine the relationship of obesity to the risk of radiographic OA progression among those who have followup knee radiographs, conditioning on a common effect (i.e., complete followup), as denoted by a box around “complete followup” in Figure 3, creates a spurious association between its causes, denoted by a dotted line between obesity and radiographic OA progression, leading to a biased effect estimate. In this example, as in many instances in real life, radiographic OA progression status is unknown among subjects who are lost to followup; thus, the magnitude and direction of bias due to loss to followup is difficult to predict. One approach often used when loss to followup does occur is to compare the baseline characteristics between subjects who completed the followup with those who were lost to followup. However, such similarities do not guard against selection bias because reasons for loss to followup may not be the same between the exposed and unexposed groups, and therefore could still lead to selection bias.
In the MOST study, we found that the proportion of loss to followup was higher among subjects whose BMI was ≥30 kg/m2 (15.6%) than those whose BMI was <25 kg/m2 (11.3%), suggesting a positive association between obesity and proportion of loss to followup (as denoted by the path between obesity and complete followup in Figure 3). It is reasonable to assume that subjects who have radiographic OA progression, i.e., developing more severe or end-stage radiographic OA, have a higher proportion of loss to followup than those who did not have radiographic OA progression (as shown by the path between radiographic OA progression and complete followup in Figure 3). Thus, selection bias due to conditioning on those who have complete followup data on the outcome will introduce a spurious association between its causes (i.e., obesity - - - radiographic OA progression), which can dilute the effect of obesity.
Bias due to measurement of outcome and ceiling effect
One of the main motivations for studying risk factors for radiographic OA progression is that the effect of a specific risk factor or factors on incident radiographic OA may differ from that on progressive disease. If that is the case, it is also reasonable to speculate that the effect of a specific risk factor or factors on the risk of progression from mild radiographic OA to either moderate or severe radiographic OA may differ from that on progression from moderate to severe disease. However, the current definition of radiographic OA progression assumes that the effect of a specific risk factor is the same regardless of the severity of radiographic OA at baseline, which runs counter to the initial intention of evaluating whether risk factors have different effects depending on the stage of disease. If the effects of risk factors on radiographic OA progression vary with stage of radiographic OA severity, mixing data from various transitions of radiographic OA severity may mask the true causal association. In reality, there must be a continuum of radiographic changes in radiographic OA, and any threshold, including the current, widely used one, is inherently arbitrary (18).
Radiographs for such studies are typically acquired at fixed time intervals. In most cases, the time interval between two consecutive radiographs is relatively long. Knees that have the same severity of radiographic OA, indicated by either K/L grade or joint space narrowing (JSN) score, on the followup radiographs are treated the same regardless of how rapidly they actually reached that stage. Although it is difficult to measure the severity of radiographic OA with frequent, short time intervals to capture the exact date of change in severity of radiographic OA, long intervals between two consecutive assessments of the severity of radiographic OA have substantial drawbacks.
For example, assume that knee radiographs in a study are acquired every 24 months. Radiographic OA in one knee progressed from mild to moderate disease at 16 months and in another at 22 months. Neither of them progressed further by 24 months. At the 24-month followup radiograph, both knees would be considered to have experienced radiographic OA progression. Although one knee progressed more rapidly than the other, this information is not captured owing to the relatively long interval between the two consecutive assessments. Furthermore, some measures used to assess radiographic OA progression, such as semiquantitative JSN scores or K/L grades, cannot accurately estimate the rate of progression when the disease has reached its end stage according to that scale (i.e., JSN score of 3 or K/L grade 4), reflecting a ceiling effect in these scoring systems. Under such circumstances, even if we can obtain the risk of radiographic OA progression and estimate the risk ratio, that effect estimate will be smaller than the rate ratio and will be closer to the null. The degree of underestimation depends on the level of the risk, being slight for small risks and greater for large risks (19). This issue can be illustrated with the following example. Assume we recruit 10,000 male smokers and 10,000 male nonsmokers at age 50 years and follow them for a certain time period. We would expect that smokers would experience a higher mortality than nonsmokers over a 5-year period. However, if we followed those subjects for 50 years, almost all of the participants would die by the end of the study; thus, the mortality ratio from smoking would be close to 1.
Alternative approaches to studying risk factors for radiographic OA worsening
In an ideal epidemiologic study of the association between risk factors and the risk of radiographic OA progression, it is preferable to study a risk factor that either occurs at the time of diagnosis of mild or moderate radiographic OA or that changes over time. Such a study could be considered as a natural experiment, allowing one to assess whether such a risk factor or its change are associated with the risk of radiographic OA progression. However, most risk factors under study are typically chronic ones (e.g., BMI, nutritional factors) that either are likely to exist long before the diagnosis of mild or moderate radiographic OA or do not change substantially over a short period of time unless under special circumstances, for example, studying the effect of weight loss on the risk of knee OA progression among persons who underwent a surgical operation to induce weight loss.
Alternatively, observational studies should ideally be conducted using the measures for radiographic OA progression that are continuous or interval scaled, collecting and controlling for as many potential confounders as possible; assembling a large sample of knees with mild radiographic OA (when cartilage loss in most knees has not occurred yet) and having knee imaging obtained more frequently within a short period of time; and minimizing loss to followup. However, these alternatives will not allow investigators to eliminate the bias induced by conditioning on preexisting radiographic OA in observational studies when the risk factor under study is a cause of radiographic OA at baseline. Although a randomized clinical trial is an ideal study design that would avoid the bias due to conditioning on preexisting radiographic OA, logistical challenges, ethical issues, and potential of loss to followup limit its utility for studies of risk factors for radiographic OA progression. Furthermore, the findings from a randomized clinical trial may not be generalizable to the general population if one or some characteristics differ between participants in the trial and target population, and such characteristics would potentially modify the effect of the factor of interest.
Given the limitations of not having the ideal study parameters and capabilities available at present, an alternative approach is needed to assess the effect of a specific risk factor on the risk of the severity of radiographic OA in order to minimize the potential biases that we have discussed above. We explored using such an approach to assess the effect of BMI on the risk of OA worsening. Specifically, among persons without radiographic OA (K/L grade 0 or 1) at baseline, we evaluated the relation of BMI to incident mild (K/L grade 2) and incident moderate/severe radiographic OA (K/L grade 3 or 4). Of 3,313 knees without OA at baseline in the MOST study, 116 knees (3.5%) developed mild radiographic OA and 96 knees (2.9%) developed moderate/severe disease over the 30-month followup period. Higher BMI increased the risk of both mild and moderate/severe radiographic OA (Table 1). Compared with those with a BMI <25 kg/m2, subjects with a BMI ≥35 kg/m2 had a 3-fold increased risk for mild radiographic OA and a 4-fold increased risk for moderate/severe radiographic OA. These findings clearly suggest that obesity not only increases the risk of developing radiographic knee OA, but is also associated with more rapid worsening of radiographic OA once it has developed. However, using the same data, BMI was not associated with an increased risk of radiographic OA progression when the study sample was restricted to those with radiographic OA at baseline (20).
Table 1. BMI and risk of mild and moderate/severe radiographic OA*
Mild radiographic OA (K/L grade 2)
Moderate/severe radiographic OA (K/L grades 3 and 4)
BMI = body mass index; OA = osteoarthritis; K/L = Kellgren/Lawrence; RR = relative risk; 95% CI = 95% confidence interval.
For trend, P < 0.004.
For trend, P < 0.001.
This approach has some advantages. The risk factor is not a consequence of the disease process; thus, reversal causality is unlikely. Although unmeasured confounding is still an issue, it should have a much weaker effect compared with the usual approach of studying the risk of radiographic OA progression. However, there are drawbacks as well. First, although the study would answer the question of whether obesity increases the risk of mild, moderate, and severe radiographic OA among knees without preexisting disease, it does not directly test the hypothesis of whether obesity accelerates radiographic OA progression among the knees with preexisting disease. Second, the risk of moderate/severe radiographic OA among those without disease at baseline, especially severe disease, may not be high. Thus, unless the sample size is large enough or the followup time is long enough, the study may not have enough power to examine the relationship of a risk factor to the risk of moderate/severe disease.
Although the risk factors for progressive radiographic OA may be different from those for incident radiographic OA because of different underlying pathophysiology for various stages of radiographic OA, there are other compelling explanations that may underlie the discrepancy between study findings for radiographic OA progression and those for radiographic OA incidence. We contend that using an observational study design to assess risk factors for radiographic OA progression in joints with existing disease is subject to various biases that may account for such null findings. These biases are likely to be important threats to the validity of any observational study of risk factors for radiographic OA progression and will make it formidably challenging to study these risk factors outside of a randomized trial. We hypothesize that risk factors may actually exist for progressive radiographic OA, but that flaws in study design and our current measure for radiographic OA progression may have prevented us from identifying them.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Zhang had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Zhang, Niu, Felson, Choi, Nevitt, Neogi.
Acquisition of data. Zhang, Felson, Choi, Nevitt.
Analysis and interpretation of data. Zhang, Niu, Choi, Nevitt, Neogi.