Research Article

*Correspondence to Graeme Jones, Menzies Research Institute, Private Bag 23, Hobart, Tasmania 7001, Australia

Funded by:
  National Health and Medical Research Council of Australia
  Masonic Centenary Medical Research Foundation

Received: 11 March 2005; Accepted: 3 June 2005

10.1002/art.21267  About DOI

Osteoarthritis (OA) is the most common form of arthritis, especially knee OA, and in most developed countries is a leading cause of musculoskeletal disability ([1]). A modest but significant genetic effect on knee OA has been well demonstrated using radiographs ([2]), which provide only a broad-brush view of joint pathology due to their 2-dimensional nature and semiquantitative grading scales. Magnetic resonance imaging (MRI) allows direct visualization of joint structures and provides accurate and reproducible quantitative estimates of cartilage volume and defects as well as bone area ([3][4]). In a previous cross-sectional study, we reported high heritability for knee cartilage volume, chondral defects, bone size, and lower limb muscle strength, which was largely independent of radiographic OA, suggesting that these characteristics are under strong genetic control but are of uncertain relevance to radiographic OA ([3][5]). However, all of these measures appear to have relevance to symptoms ([6]), OA progression ([4]), and arthroplasty ([7]). This suggests that adjusting for radiographic OA may not be the best method of assessing relevance to OA, especially because mild radiographic OA is associated with substantial reductions in cartilage volume and increases in joint surface area ([8]), signifying that much has happened at a structural level prior to the onset of radiographic OA.

With regard to genetic studies, results of an independent twin study have confirmed the cartilage volume estimates ([9]). However, in cross-sectional studies the genetic contribution may reflect both the effect of growth and subsequent loss. Certainly, the extent of cartilage-volume loss is high in persons with well-established OA ([10]), but the factors underlying this are uncertain. Longitudinal studies are required to estimate the genetic contribution to change in all of the above-mentioned factors. The aim of this study, therefore, was to use a sibpair design to estimate the heritability of longitudinal changes in knee cartilage volume, chondral defects, subchondral bone size, and lower limb muscle strength, and to assess whether these estimates are independent of radiographic OA.

SUBJECTS AND METHODS

Study subjects.

The study was carried out in southern Tasmania, as described previously ([3]). Briefly, subjects were the adult children of patients in whom a knee replacement had been performed for idiopathic knee OA. The Southern Tasmanian Health and Medical Human Research Ethics Committee approved the study, and written informed consent was obtained from all participants.

Anthropometric studies.

Weight, height, and muscle strength were measured at baseline and followup, as described previously ([3]). Each subject's medical history (e.g., knee pain and knee injury) was collected by questionnaire at baseline.

MRI.

An MRI scan of the right knee was obtained at baseline and followup, using the same machine and the same protocol as described previously ([3]). Knee cartilage volume, bone area, and chondral defects were assessed in a manner identical to that used in our previous studies, with excellent reproducibility ([3][11]). The longitudinal changes in cartilage volume and bone area were expressed as the percent change per year. The method used to assess bone area measured bone changes only and not cartilage changes. The difference in chondral defect scores between baseline and followup was computed by subtracting the baseline score from the followup score, with progression of chondral defects defined as any difference  1.

Radiography.

A standing anteroposterior semiflexed view of the right knee was obtained in all subjects at baseline and assessed using the Altman atlas, as previously described ([3]). Individual features, including joint space narrowing and osteophytes, were assessed using a 4-point scale.

Statistical analysis.

A variance components analysis was performed to estimate the heritability of various traits. Using the software package SOLAR ([12]), trait variance was modeled as a mixture of genetic variance (attributed to many genes with small, additive effects) and random variance (due to random environmental variations not correlated between subjects within families). The estimated heritability was then defined as the proportion of genetic variance in the model with the maximum likelihood.

To assess whether the estimated heritabilities differed from zero, a null model with only the random variance term was also fitted. All models were fitted after first adjusting trait scores within SOLAR for various combinations of covariates, as follows: age, sex, weight, and height (step 1); previous covariates, knee pain, previous knee injury, and longitudinal changes in muscle strength (step 2); all previous covariates and longitudinal changes in cartilage volume (for studies of bone size)/bone size (for cartilage volume and chondral defects) (step 3); and all previous covariates and radiographic OA score (step 4). P values less than 0.05 were considered significant.

RESULTS

At baseline, the study group comprised 128 subjects (61 men and 67 women) representing 115 sibpairs, with an average age of 45 years. Ten subjects were lost to followup (followup rate 92%), and 3 families were excluded because only 1 sibling was available. The average followup time was 2.4 years (range 1.7-3.3 years). The structure of the families studied is presented in Table 1, and the general characteristics and study traits are shown in Table 2. The distribution of longitudinal changes in cartilage volume, bone size, and muscle strength approximated a normal distribution. Knee pain at baseline was common in this group, while radiographic OA was relatively uncommon and mild.

Table 3 presents the heritability estimates for the study traits. After adjustment for age, sex, height, and weight (step 1), changes in global cartilage volume, lateral bone size, and muscle strength all had significant heritability. After adjustment for knee pain, previous knee injury, and change in muscle strength (step 2), the heritability estimates increased by 8-50%, with the largest increase observed for global cartilage volume. In addition, the heritability estimates for changes in medial tibial cartilage volume and progression of chondral defects in the medial compartment became statistically significant. Further adjustment for bone size or cartilage volume (where appropriate) and radiographic OA (steps 3 and 4) led to small reductions in the heritability estimates for all study traits except chondral defects in the lateral compartment, for which a 75% decrease was observed.

DISCUSSION

To our knowledge, this is the first longitudinal evaluation of a genetic contribution to knee structure and lower limb muscle strength. We have documented significant and high heritability estimates, particularly for longitudinal changes in global cartilage volume, medial tibial cartilage volume, lateral tibial plateau size, muscle strength, and progression of chondral defects. These heritability estimates are higher than, but largely independent of, the heritability of radiographic OA. Furthermore, the heritability estimates of the study traits remained largely unchanged after adjustment for each other, suggesting that they are under independent genetic control, with at most a small shared genetic component.

In previous cross-sectional studies, we ([3]) and other investigators ([9]) demonstrated high heritability for both lateral and medial tibial cartilage volume. Results of this longitudinal study are consistent with the previous results, highlighting the strong genetic component to both knee cartilage volume and its rate of change. In contrast to previous studies, the results of which suggested that both lateral and medial tibial cartilage volume had a high and significant heritability ([3]), we demonstrated a stronger genetic influence on medial tibial cartilage volume compared with lateral tibial cartilage volume. This finding is surprising given the cross-sectional results and needs to be confirmed in further studies but most likely reflects the relatively greater effect of measurement error in longitudinal studies. However, it possibly explains why OA targets the medial compartment more commonly than the lateral compartment ([13]). Alternatively, cohort effects may bias results in a cross-sectional study.

Similarly, we observed that lateral tibial plateau size had a higher and significant heritability compared with medial tibial plateau size, although the longitudinal changes over 2 years in medial tibial plateau size were larger than those in lateral tibial plateau size. This also contrasts with the findings from a cross-sectional study ([3]), in which both lateral and medial tibial plateau size had low heritability after adjusting for body size. The observed significant increase in the medial but not the lateral tibial plateau probably reflects the OA disease process and/or subchondral bone remodeling. Indeed, in individuals with radiographic knee OA the tibial plateau size is larger than that in persons without radiographic OA, and this is more pronounced in the medial than the lateral tibial plateau ([14]). Adjustment for radiographic OA led to no changes in the heritability estimates of both cartilage volume and bone size, which casts doubt on the relevance of these MRI measures in OA. However, all of these measures have relevance to various facets of knee OA, and there are a number of reasons (as mentioned in the introduction) for questioning the value of adjusting for radiographic OA in samples from patients with earlier disease, because cartilage loss and bone expansion need to be substantial before radiographic OA is evident. Nevertheless, these results need to be confirmed in independent samples in which the subjects are of different races/ethnicities and in which the prevalence of both radiographic and symptomatic OA is higher.

The heritability estimates for cartilage volume and bone size remained largely unchanged after adjustment for each other, suggesting that they are largely under independent genetic control, with a lesser shared genetic component. However, adjustment for knee pain and previous knee injury, surprisingly, led to an increase in the heritability estimates for both cartilage volume and bone size, with a maximum increase of 40% for medial tibial cartilage volume. This implies negative confounding, which seems unlikely given the variables in question, or may represent better estimates due to less environmental noise.

Similar to our previous cross-sectional report ([5]), we demonstrated high heritability for the progression of chondral defects. The heritability increased by 59-73% after adjusting for knee pain and previous knee injury. Again, this implies negative confounding or the effect of less environmental noise. However, a 75% reduction in heritability for progression of chondral defects in the lateral compartment, after adjustment for radiographic OA, supports direct relevance to OA. This was not the case for the medial compartment, the heritability of which remained unchanged even after adjusting for radiographic OA. The reason for this discrepancy remains unclear, but the higher standard error for the heritability estimates indicates that the results are not robust, possibly reflecting relative limitations of the program we used for dichotomous traits as compared with continuous traits ([15]). It is likely that for the medial compartment, the true heritability is substantially less than 98%.

Consistent with our cross-sectional study ([3]), in this longitudinal study we demonstrated a strong genetic component in the loss of lower limb muscle strength over time. Muscle weakness is well recognized as a risk factor for the development of OA ([16]). Results of the current study suggest that change in muscle strength is under strong genetic control. Identification of susceptibility gene(s) for muscle strength may help to provide a new approach for the prevention of OA.

The current study has a number of potential limitations. First, there is controversy about the ideal study design for estimating heritability of disease. The twin model is often used but has been criticized as overestimating heritability due to the assumption of similar shared environments between monozygotic and dizygotic twins. This has been documented for bone mineral density ([17]) but not for OA. Family studies such as the present one may be more likely to represent true heritability but make it more difficult to assess the contribution of shared environment. Using multiple sibpairs from the same family may bias heritability estimates upward. However, the heritability estimates from an independent sample (1 pair from each family) were very comparable with those from the whole sample (data not shown), indicating that this is not an issue in the current study and is consistent with our previous report ([3]).

Second, the choice of subjects who are at higher risk of disease may bias the heritability estimates and limit the generalizability of the results to the general population. However, it is most likely that this bias will act to decrease estimates by decreasing genetic heterogeneity in comparison with an unselected sample. Our data may partly support this, with some inconsistency in estimates between sites.

Third, measurement error in the assessment of both MRI results and muscle strength may have reduced the estimates. However, both assessment techniques have high reproducibility at our institution, suggesting that this is not of major concern.

Fourth, we did not assess meniscal degeneration or extrusion, each of which is thought to be a risk factor for OA, but it remains totally uncertain whether these factors influence heritability results or will be heritable themselves.

Last, the followup rate was 92%, suggesting that the loss of subjects to followup was not of major concern in this study. However, the followup period was relatively short, and longer studies may be required to accurately associate the clinical significance of the MRI changes.

In conclusion, early longitudinal changes in knee structures that are relevant in later OA, such as changes in medial tibial cartilage volume, lateral tibial bone size, progression of chondral defects, as well as muscle strength, have high heritability, most likely reflecting a strong genetic component and suggesting their potential to be studied in quantitative trait linkage and association analysis.

Acknowledgements

A special thanks goes to the subjects and orthopedic surgeons who made this study possible. The role of C. Boon in coordinating the study is gratefully acknowledged. We would like to thank Martin Rush who performed the MRI scans and Kevin Morris for technical support. We thank Drs. Leigh Blizzard and Stephen Quinn for statistical support.