SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

Objective

Few methods exist to measure the progression of osteoarthritis (OA) or to identify people at high risk of developing OA. Striking radiographic changes include deformation of the femoral head and osteophyte growth, which are usually measured semiquantitatively following visual assessment. In this study, an active shape model (ASM) of the proximal femur was used to determine whether morphologic changes to the bone could be quantified and used as a marker of hip OA.

Methods

One hundred ten subjects who had no signs of radiographic hip OA at baseline (Kellgren/Lawrence [K/L] scores 0–1) were selected from the Rotterdam Study cohort of subjects ages ≥55 years. To measure the progression of OA, subjects were followed up with radiographic assessment after 6 years. At the 6-year followup, 55 subjects had established OA (K/L score 3), and in 12 of these OA subjects, the progression of the disease required a total hip replacement (THR). Age- and sex-matched control subjects had a K/L score of 0 at followup. Using the ASM, subjects were assessed for shape changes in the femoral head and neck before, during, and after the development of radiographic OA. Scores of shape variance, or mode scores, were assigned for 10 modes of variation in each subject, and differences in mode scores were determined.

Results

During followup, significant changes in shape of the proximal femur occurred within the OA group from baseline to followup (P < 0.0001 for mode 1 and P = 0.002 for mode 6) but not within the control group. At baseline (all subjects having K/L scores 0–1), there were significant differences in mode 6 between the OA group and the control group (P = 0.020), and in modes 3 and 6 between the OA subjects who underwent THR and the remaining OA subjects (P = 0.012 and P = 0.019, respectively).

Conclusion

Compared with traditional scoring methods, the ASM can be used more precisely to quantify the deforming effect of OA on the proximal femur and to identify, at an earlier stage of disease, those subjects at highest risk of developing radiographic OA or needing a THR. The ASM may therefore be useful as an imaging biomarker in the assessment of patients with hip OA.

Osteoarthritis (OA) is one of the most common disorders in the elderly. It is estimated that by age 75 years, 85% of individuals show either clinical or radiologic evidence of OA (1). As OA of the hip progresses, changes in the shape of the femoral head develop, with flattened and irregular features becoming apparent. These changes can be observed on standard radiographs but are hard to quantify. The severity of OA is generally assessed using semiquantitative methods based on visual evaluation of a radiograph; for example, the Kellgren/Lawrence (K/L) scoring method (2) is used to assess a number of typical features of OA, including joint space narrowing (representing the loss of cartilage), development of osteophytes, presence of subchondral cysts, and subchondral sclerosis, on a 5-point scale. Quantitative measures of OA from medical images have previously focused on features of cartilage, whether indirectly by measurement of the joint space width on radiographs (3–6) or directly by measurement of cartilage thickness or volume from magnetic resonance imaging (MRI) scans (7–10). Investigations of the trabecular structure of bone have also been performed, again using either standard radiographs (11–13) or 3-dimensional MRI (14).

In addition to the deformation of bone and the growth of cysts and osteophytes, changes to OA bone include thickening of the subchondral bone plate, increased stiffness, increased bone mineral density and content, and alterations in the trabecular structure and size of the bone (13, 15–17). In OA, a proliferation of bone is observed, along with alterations in the material properties of the bone. Reduced mineralization of the bone has been observed in subjects with OA as compared with normal subjects and those with osteoporosis (18, 19). This occurs in conjunction with an increase in water content (18) and a reduction in the hardness or elastic modulus of the material as compared with that in the bone of subjects with osteoporosis (20, 21). Other differences in OA bone include an increase in metabolic activity (22) and an increase in both the amount and types of fat (23). This accumulation of evidence of changes occurring in OA bone at all levels has prompted the need for more detailed investigations of the bone, whole joint, and systemic factors in OA (24, 25).

Although gross morphologic features of the proximal femur have been rarely studied in patients with OA, geometric features have been identified as potential risk factors. The relationship between OA and morphologic changes is often attributed to abnormal loading patterns caused by unusual geometry. For example, the neck-shaft angle may lie outside the normal range (e.g., a varus or valgus hip) (26, 27), the femoral head could be uncovered or incongruent with the acetabulum as measured by the Wiberg center edge angle, or the morphologic appearance of the femoral head may be nonspherical (28–30). It is not known at which stage of the disease process these morphologic changes appear and how they might be related to disease progression.

Changes in the density and shape of the bone are traditionally believed to be a secondary effect of earlier cartilage degeneration. This leads to changed biomechanics and subsequent bone adaptation in late-stage OA. However, there is increasing recognition that the bone may be affected from the onset of the disease as part of a systemic process (24, 25), concurrent with or maybe even before cartilage degeneration has occurred. In this model, changes to bone become an important player in the etiology of OA, and thus more focus on the bone in the treatment of the disease is warranted. The proximal femur is a complex structure that cannot be fully described by 2 or 3 geometric measures alone. In this study we built a model of the shape of the femoral head and neck, i.e., the sites of the biggest shape changes that occur in OA, in order to investigate the relationship between morphologic features and radiographic OA.

The radiographic images in the present study came from the Rotterdam Study, a prospective cohort study involving men and women ages ≥55 years (31). To assess the development of OA, we selected subjects who had no radiologic signs of OA at baseline but then developed OA during the 6-year period before the followup visit. This allowed comparison of the shape of the femur in control subjects with that in OA subjects before, during, and after the development of radiographic disease. The method used for shape analysis in this study was the active shape model (ASM) (32, 33). This is a method for building a statistical model of shape variation on the basis of a data set of digital images.

The aim of the present study was to discover whether statistical analysis could be used to model the differences in gross morphologic features of the femoral head and neck between healthy subjects and subjects with hip OA as visualized on plain radiographs, and to model the changes in shape that occur in these subjects over time. This would allow quantification of the deformation of the femoral head, and also allow identification of characteristic morphologic features that may act as risk factors in the development of OA.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

Subject selection.

One hundred ten subjects were selected from the Rotterdam Study (31) for this investigation. We obtained standard pelvic radiographs from each subject at baseline and at followup ∼6 years later. Both radiographs were assessed for signs of OA by a radiologist and a scientist, using the K/L scale (2). The K/L scale ranges from 0 to 4, with scores of 0 and 1 denoting no radiographic evidence of OA, and scores of 2, 3, and 4 denoting the presence of radiographic OA. If both observers were in agreement about the presence or absence of radiographic OA and the interobserver difference in score was within 1 point, then the highest score was used; otherwise the observers reached a consensus on the score (31). If a total hip replacement (THR) had been performed as a result of OA, a K/L score of 5 was recorded.

To allow analysis of the development of OA, we identified subjects in whom there were no radiographic signs of OA at baseline or in whom the diagnosis was doubtful at baseline (K/L score of 0 or 1) and in whom the K/L score increased by at least 3 points by the 6-year followup (scores increasing from 0 to ≥3, or from 1 to ≥4). Fifty-five subjects (41 female and 14 male) who met these criteria were identified from the records, to comprise the OA group. An age- and sex-matched control group was then identified. The control group comprised 55 individuals (42 female and 13 male) who had a K/L score of 0 for both hips at both the baseline and the followup time points.

THRs were performed between baseline and followup in 12 of the OA subjects (10 female and 2 male). In these individuals the followup radiograph was not assigned a regular K/L score but, rather, was assigned an artificial K/L score of 5 (to indicate surgical intervention). This split the OA group into 2 subgroups, the group of OA subjects who had undergone THR during the course of the study (the THR group) and the group who developed OA but were not treated surgically (the non-THR group; 31 female and 12 male).

Among all subjects with OA, the hip that fit the criteria set for inclusion (K/L score of 0 or 1 increasing by at least 3 points) was used in the analysis. In the control group, both the left hip and the right hip had a K/L score of 0 at baseline and at followup, and therefore the right hip of each control subject was used in the analysis. Table 1 summarizes the characteristics of the different groups and the inclusion criteria used.

Table 1. Summary of subject groups*
 ControlOA
AllNon-THRTHR
  • *

    Values are the distributions of age and sex and Kellgren/Lawrence (K/L) scores for the osteoarthritis (OA) and control groups. The OA group was also assessed in 2 subgroups, those in whom a total hip replacement (THR) was performed during the course of the study, and those in whom surgical intervention was not performed (non-THR) and could therefore be assessed for changes on the followup scan. No significant differences were found in the distributions of age and sex between any of these groups. K/L scores ranged from 0 (no OA) to 4 (severe OA). An artificial level of 5 was used to indicate a THR.

  • Necessitating a THR.

Sex
 No. male1314122
 No. female42413110
 % female76757280
Age, years    
 Range55–7855–8055–8055–78
 Mean ± SD68.4 ± 5.869.0 ± 6.069.1 ± 6.268.4 ± 5.6
K/L score    
 Baseline00–10–10–1
 Followup03–53–55

Radiographic assessment.

Weight-bearing anteroposterior radiographs of the hip were obtained at 70 KV (focus 1.8, focus-to-film distance 120 cm, High Resolution G 35 × 43–cm radiographic film [Fujifilm Medical Systems, Stamford, CT]). Radiographs of the pelvis were obtained with both feet in 100° internal rotation and the x-ray beam centered on the umbilicus. One trained reader (MR), who was unaware of the clinical status of the patients, evaluated the radiographs of the hip obtained at baseline and at followup. All radiographs of the hip were grouped by patient and read in pairs in known chronologic order (chronologically ordered reading procedure). The interrater reliability (kappa statistic) for K/L scoring of radiographic hip OA was 0.68, as reported earlier (34).

Design of the ASM.

The shape of the femoral head and that of the femoral neck were assessed using the ASM (32, 33, 35, 36). The model was built using an ASM tool kit (Manchester University, Manchester, UK). The ASM template is a set of landmark points that define the shape to be identified. Each point is always placed on the same feature of the outline, to aid comparison between shapes. Key points are placed at easily identifiable features (for example, the place at which the femoral neck meets the greater trochanter), while the remaining points are spaced approximately evenly between features (for example, around the spherical femoral head). The template used in this study was based on the 29-point model, which was previously developed by members of our group for examining the relationship between femoral shape and hip fracture in osteoporosis (37). Because the proximal femur was generally positioned in a bottom corner of the radiograph, the edge of the greater trochanter and the base of the lesser trochanter were frequently missing from the radiographs, and therefore the full 29-point model incorporating both of these features could not be used in the present study. Accordingly, the template used in this study contained 16 points from the 29-point model, as shown in Figure 1A. The model starts at the point at which the femoral head meets the femoral neck above the lesser trochanter, covers the outline of the femoral head, and terminates at the point at which the superior femoral neck meets the greater trochanter.

thumbnail image

Figure 1. Outlines of the active shape model of the proximal femur of subjects with hip osteoarthritis (OA). A, Template used for the 16-point active shape model, with points (numbered 0–15) representing each mode of variation. B, Average outline of the mode points (averagepts) for all OA subjects in the study. C, Points marking the outline for a representative subject with OA at baseline and at followup, illustrating how the shape of the femur changes with the development of OA.

Download figure to PowerPoint

In ASMs, the shape is described by a series of orthogonal modes of variation. Each mode is independent of all the others, and therefore each is an independent descriptor of the shape. For each mode in the model, the mean and SD value for the whole radiographic data set of each group was calculated. This value was scaled to zero to serve as the zero mean and unit standard deviation. The score for each mode was then calculated for each subject at each time point, in which the mode score was expressed in terms of how many standard deviations it lay from the mean referent value (zero) of that mode. The overall size variance was extracted as a scaling factor. In the present study the radiographs from the Rotterdam Study (as described above) were used to generate the scores in the ASM. Only the first 10 modes were analyzed in this study, since they describe the majority (94.6%) of the variance in shape within the data set.

Statistical analysis.

In order to investigate differences between the OA and control groups, differences in mode scores were tested. At the baseline and followup time points, differences were investigated using age- and sex-adjusted logistic regression. To detect changes in mode values, independent of age or sex, over time, as a reflection of the progression of OA, a general linear model (GLM) for repeated measures was applied, using OA status and sex as fixed factors and age as a covariant. When no significant effects of age and sex were found, the GLM was repeated using only OA status as a fixed factor, to increase the power of the test. When a general change in mode value and an effect of OA were detected, the change in mode value was tested separately for the OA and control groups using a paired t-test. At baseline, radiographs from all subjects were available for investigation. Following testing for differences between the OA and control subjects, the OA subgroups (THR and non-THR) were compared to assess differences between OA subjects who did and those who did not require a THR during the study. Due to the surgical removal of the joint, the proximal femur was not available in the THR group at followup, and therefore comparisons at baseline and over time were made between the non-THR OA group and the control group.

Differences in the age and sex distribution between the groups were investigated using t-tests and chi-square or Fisher's exact tests, as appropriate. Relationships between mode score and age or sex were tested at baseline using Pearson's correlation coefficients and t-tests. Statistical tests were performed using SPSS and SigmaStat version 3.1 (SPSS, Chicago, IL). P values less than or equal to 0.05 were considered significant.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

Characteristics of the subjects.

The group of control subjects was selected to match the age and sex of the OA subjects as closely as possible (Table 1). Chi-square testing showed that there was no significant difference in the numbers of men and numbers of women between the OA group (75% female) and the control group (76% female) (P = 1.0). Moreover, analysis by t-test showed that there was no significant difference in age distribution between the OA group and control group (P = 0.60). The OA subgroups, THR and non-THR, were also investigated for differences in age and sex. Using Fisher's exact test, no significant differences in the numbers of male subjects and numbers of female subjects were found between the OA subgroups (P = 0.71), and a t-test showed no significant difference in age between the subgroups (P = 0.71).

At baseline, subjects were asked whether they had pain in either their left hip or their right hip. This information was available for all but 2 of the OA subjects. Two subjects in the THR group and 3 in the non-THR group reported having pain in the affected hip at baseline; however, there was no significant difference in the incidence of pain in the affected hip at baseline between the OA subjects who subsequently underwent THR and those who did not undergo THR (P = 0.46 by chi-square test). At followup, subjects were again asked whether they had pain in either hip. Of the 43 OA subjects in the non-THR group, 16 reported having hip pain at followup, including the 3 subjects who reported having pain in the observed hip at baseline. None of the 3 ASM modes of variation identified as having an association with OA were significantly related to self-reported hip pain either at baseline or at followup (P > 0.05). Significant differences were observed between men and women in only 2 modes of variation, mode 1 (P = 0.004) and mode 7 (P < 0.001). Age was significantly correlated with mode 3 only (P = 0.034).

Findings of the ASM.

Changes in the shape of the proximal femur in OA subjects over time.

The 10 modes of variation analyzed reflect the many different femoral shapes observed in the study group. Paired t-tests were performed to identify which modes were associated with changes in the shape of the femur in the OA group during the course of the study. The average outline of the model for all OA subjects is shown in Figure 1B, and the baseline and followup landmark points, which outline the change in shape of the femoral head over time, are shown in Figure 1C for a typical OA subject over the 6 years of followup.

When these shapes were described numerically by the modes of variation, a small number of modes were found to be descriptive of the changes. Shape changes that were found to be significant from baseline to followup are summarized in Table 2. No modes of variation changed significantly during the course of the study in the control group. In contrast, significant changes were observed over time in the OA (non-THR) group, in mode 1 and in mode 6 (P < 0.0001 and P = 0.002, respectively). The results of paired t-tests, reported as age and sex interactions, were not significant. The shape changes represented by modes 1, 3, and 6 are shown in Figures 2A–C.

Table 2. Summary of significant differences in mode scores between the groups
 Mode 1Mode 3Mode 6
  • *

    Values are the change within subjects over the total course of the study (percent variance), investigated by paired t-tests.

  • Values are P values, determined by age- and sex-adjusted logistic regression, comparing groups of control and osteoarthritis (OA) subjects at the same time point (baseline or followup) or from baseline to followup for those OA subjects who did not require total hip replacement (non-THR).

  • For OA vs. control, n = 110 each; for OA non-THR vs. OA THR, n = 55 each.

  • §

    For OA non-THR vs. OA non-THR at baseline, n = 86 each; for OA non-THR vs. control at followup, n = 98 each.

Percent variance from baseline to followup*28.616.24.3
Baseline comparisons   
 OA vs. control0.570.180.020
 OA non-THR vs. OA THR0.520.0120.019
Followup comparisons§   
 OA non-THR vs. OA non- THR at baseline<0.00010.880.002
 OA non-THR vs. control at  followup0.1430.990.030
thumbnail image

Figure 2. Shape variations in mode 1 (A), mode 3 (B), and mode 6 (C). Outlines reflect ±2 standard deviations (std) (encompassing >95% of the range) for each of the 3 modes. For comparison, the outlines are fixed at the point at which the upper femoral neck meets the greater trochanter. In each case the outline more commonly associated with osteoarthritis is shown by a solid line, while that more commonly associated with the unaffected age- and sex-matched control group is shown by a broken line.

Download figure to PowerPoint

Differences in mode scores between the OA group and the control group.

Interestingly, at baseline, the mode 6 scores of shape variation were found to be significantly lower in the OA subjects than in the control subjects, as determined using age- and sex-adjusted logistic regression (P = 0.020). The scores for mode 6 were also significantly lower in the OA group than in the control group at followup (P = 0.03). This finding appeared to be related to the overall severity of OA, with the highest scores being found at baseline in the control group and the lowest scores being found at baseline in the group of OA subjects who subsequently required THR (Figure 3B).

thumbnail image

Figure 3. Differences in mode scores between the osteoarthritis (OA) and control subjects. The mean mode scores for mode 3 (A) and mode 6 (B) were compared between groups. The highest scores were found in the control group at both the baseline and the followup time points, lower scores were found at baseline and at followup in the subjects who developed OA but did not undergo total hip replacement (No-THR), and the lowest scores were found in the baseline group of OA subjects who subsequently required THR. Results are the mean ± SEM for each group, scaled to zero (horizontal line) as the referent mean.

Download figure to PowerPoint

Differences within the OA subgroups.

Further distinctions could be made within the group of OA subjects when they were grouped according to whether they went on to have a THR within the study period (the THR group) or did not require a THR (the non-THR group). Significant differences were observed at baseline between the OA subgroups of THR and non-THR, in mode 3 (P = 0.012) and in mode 6 (P = 0.019). The mean ± SEM of the scores for modes 3 and 6 are shown in Figures 3A and B.

The odds ratios based on the results of the age- and sex-adjusted logistic regression analyses, comparing the groups at baseline and again at followup, showed clear differences between the groups in the likelihood of having a decreased mode score, as shown in Table 3. These odds ratios revealed differences between groups both at baseline and also, as expected, at followup (Table 3).

Table 3. Odds ratios associated with a significant difference in mode scores*
Mode, significant comparisonOR (95% CI)P
  • *

    Odds ratios (ORs) reflect a 1–standard deviation decrease in the mode score. Values are based on the results of age- and sex-adjusted logistic regression. 95% CI = 95% confidence interval; OA = osteoarthritis; THR = total hip replacement.

Mode 3  
 OA THR (baseline) vs. OA non-THR (baseline)3.71 (1.33–10.35)0.012
Mode 6  
 Control (baseline) vs. OA (baseline)1.62 (1.08–2.45)0.020
 Control (followup) vs. OA non-THR (followup)1.68 (1.05–2.69)0.030
 OA THR (baseline) vs. OA non-THR (baseline)2.35 (1.15–4.82)0.019

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

The ASM is a method for building a statistical model of shape variation within a data set of digital images. It has been used for many purposes, from face recognition to medical imaging, including a number of musculoskeletal applications for investigating bone (37–39), prostheses (40), and cartilage (9, 41). In this study it was applied to build a statistical model of the proximal femur from standard pelvic radiographs as a way to observe the changes in the shape of the femoral head and neck attributable to OA over a 6-year period. We have shown that the ASM provides a comprehensive method of quantifying those changes in the radiographic images. It appears to represent a powerful tool for identifying those subjects at high risk of developing OA, as well as those who may require surgical intervention in the near future, as shown, in this case, by the need for THR within 6 years from a starting point of no apparent incident disease.

As expected, we were able to apply the model to chart the progression of OA over the 6-year study period, with the results revealing significant differences between the OA and control subjects at the end of the study. Of particular interest, however, the ASM also succeeded in identifying significant differences at baseline, at a time before the OA could be detected clinically using traditional methods (i.e., the K/L score), between those who went on to develop OA and the control group. In addition, these differences appear to be related to the subsequent severity of the disease, and therefore offer the possibility of early identification of those subjects who may be at higher risk of rapid disease progression. These changes could be separated from age-related changes and sex differences, and were solely related to the joint changes induced by the disease, as reflected in the K/L scores at the second time point.

As OA progresses, the femoral head becomes flattened and deformed. Although this process can be seen clearly on standard pelvic radiographs in severe cases, it is difficult to quantify. The shape of the proximal femur and its relationship to OA have been investigated in a number of studies, although the focus has been mostly on geometric properties such as the neck-shaft angle (varus or valgus femurs) (27), the femoral head height or diameter, or other measures (34, 42). The position of the femoral head in relation to the pelvis and acetabulum (the Wiberg CE angle) and the size and angle of the weight-bearing surface of the femoral head have also been investigated (28, 43). In this study we took a different approach and used the ASM to build a statistical model of shape variation across the whole of the femoral head and neck.

This model showed clear patterns of relationships between the subject groups and the modes of variation, with modes 1, 3, and 6 being identified as the key modes, which together accounted for 49.2% of the total variance in the data set (Table 2). The development of OA (measured by a change in the K/L score over time) resulted in a significant change in the mode 1 score (Table 2). This mode score reveals the way in which the femoral head flattens and deforms in combination with a shortening of the femoral neck as OA progresses (Figure 2A). There was no significant difference in this score between the OA and control groups at baseline, since there was a large biologic variation in the shape of the proximal femur, and relatively flat heads were not uncommon at baseline, even in members of the control group who did not go on to develop OA. Men, in particular, had a significantly lower mode 1 score than women at baseline, indicating a flatter femoral head among male subjects in the study. The change in mode 1 therefore appeared to reflect the progression of OA and its effect on the shape of the hip, but could not be used as a predictive measure or a marker for early OA.

Mode 6, in contrast, was significantly linked to OA in every comparison; not only did mode 6 scores significantly decrease over time in the OA group, in parallel with the increase in K/L scores, but also there were significant differences in the mode 6 score between the OA and control groups both at baseline (before OA was identifiable by the K/L score) and at followup. The range of shapes described by mode 6 is shown in Figure 2C. Subjects with lower mode 6 scores have a less pronounced curve from the upper femoral neck into the head, together with a sharper transition from the femoral head to the lower part of the neck, which, in this study, was associated with more severe OA. One of the most interesting observations about mode 6 was the significantly lower scores in the THR OA group than in the non-THR OA group, indicating that it may be possible to identify subjects at highest risk of requiring a THR at an earlier time point, even before OA is diagnosed. In general, a lower mode 6 score was associated with more severe OA or a worse prognosis (Figure 3).

Finally, mode 3 was also of great interest, because the score was significantly lower at baseline in subjects who required a THR during the course of the study, even though it was not significantly different between the OA (non-THR) and control groups at the end of the study, nor did it change in the OA (non-THR) group during the course of the study. This mode may be able to highlight individuals at an increased risk of THR in the near future, despite not being related to an increase in the K/L score. Similar to mode 6, subjects with low mode 3 scores have a sharp transition from the femoral head to the neck, but this is observed in the upper neck rather than the lower neck (Figure 2B).

In conclusion, the ASM can be used to measure and quantify the shape changes that occur in OA of the hip. Of particular interest, it can also identify individuals with a greater risk of requiring a THR in the near future and could detect differences between OA and control subjects at an earlier stage than is possible using current methods (the K/L scale). However, we would need to carry out more detailed investigations to determine whether these findings are representative of early OA or whether they indicate a predisposition to the disease. Mode 6 scores were significantly lower in the OA group than in the control group, both at baseline and at followup, and were lowest in individuals who went on to require a THR during the course of the study. This suggests that this mode of variation may be a marker not only of the development of OA, but also of its severity. The changes in shape observed with the development of OA (in modes 1 and 6) had high statistical power (>0.8) as well as significance, adding confidence to the likelihood of these results being replicated in a larger study. Although this model included effects of age and sex, we did not include other important parameters, such as body mass index or existing OA of the contralateral hip. Extension of the model to a larger segment of the femur or a combination with other measures would be likely to further improve diagnostic power.

The majority of research into OA still focuses on the cartilage. However, our data show that bone-related parameters deserve more attention, even in the early and preclinical stages of OA. The ASM provides a new way to measure the shape of the bone in OA, and provides markers that relate to the progression and prognosis of the disease. In combination with other measures of cartilage (5, 6, 8, 9), bone density and volume (27, 31, 43–46), and trabecular structure (11, 12, 14), the ASM may represent a powerful tool for improved measuring, monitoring, and understanding of OA.

Acknowledgements

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

We are grateful to Anne Weinans for digitizing all of the radiographs.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

Dr. Gregory 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 design. Gregory, Day, Pols, Weinans, Aspden.

Acquisition of data. Gregory, Waarsing, Day, Pols, Reijman, Weinans.

Analysis and interpretation of data. Gregory, Day, Weinans, Aspden.

Manuscript preparation. Gregory, Waarsing, Day, Pols, Reijman, Weinans, Aspden.

Statistical analysis. Gregory, Waarsing, Day, Aspden.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  • 1
    Sack KE. Osteoarthritis: a continuing challenge. West J Med 1995; 163: 57986.
  • 2
    Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis. Ann Rheum Dis 1957; 16: 494502.
  • 3
    Wluka AE, Wang Y, Davis SR, Cicuttini FM. Tibial plateau size is related to grade of joint space narrowing and osteophytes in healthy women and in women with osteoarthritis. Ann Rheum Dis 2005; 64: 10337.
  • 4
    Maheu E, Cadet C, Marty M, Dougados M, Ghabri S, Kerloch I, et al. Reproducibility and sensitivity to change of various methods to measure joint space width in osteoarthritis of the hip: a double reading of three different radiographic views taken with a three-year interval. Arthritis Res Ther 2005; 7: R137585.
  • 5
    Duryea J, Zaim S, Genant HK. New radiographic-based surrogate outcome measures for osteoarthritis of the knee. Osteoarthritis Cartilage 2003; 11: 10210.
  • 6
    Schmidt JE, Amrami KK, Manduca A, Kaufman KR. Semi-automated digital image analysis of joint space width in knee radiographs. Skeletal Radiol 2005; 34: 63943.
  • 7
    Cohen ZA, McCarthy DM, Kwak SD, Legrand P, Fogarasi F, Ciaccio EJ, et al. Knee cartilage topography, thickness, and contact areas from MRI: in-vitro calibration and in-vivo measurements. Osteoarthritis Cartilage 1999; 7: 95109.
  • 8
    Peterfy CG, van Dijke CF, Janzen DL, Gluer CC, Namba R, Majumdar S, et al. Quantification of articular cartilage in the knee with pulsed saturation transfer subtraction and fat-suppressed MR imaging: optimization and validation. Radiology 1994; 192: 48591.
  • 9
    Gandy SJ, Brett AD, Dieppe PA, Keen MC, Maciewicz RA, Taylor CJ, et al. Measurement of cartilage volumes in rheumatoid arthritis using MRI. Br J Radiol 2005; 78: 3945.
  • 10
    Raynauld JP, Martel-Pelletier J, Berthiaume MJ, Beaudoin G, Choquette D, Haraoui B, et al. Long term evaluation of disease progression through the quantitative magnetic resonance imaging of symptomatic knee osteoarthritis patients: correlation with clinical symptoms and radiographic changes. Arthritis Res Ther 2005; 8: R21.
  • 11
    Lynch JA, Hawkes DJ, Buckland-Wright JC. Analysis of texture in macroradiographs of osteoarthritic knees using the fractal signature. Phys Med Biol 1991; 36: 70922.
  • 12
    Messent EA, Buckland-Wright JC, Blake GM. Fractal analysis of trabecular bone in knee osteoarthritis (OA) is a more sensitive marker of disease status than bone mineral density (BMD). Calcif Tissue Int 2005; 76: 41925.
  • 13
    Papaloucas CD, Ward RJ, Tonkin CJ, Buckland-Wright JC. Cancellous bone changes in hip osteoarthritis: a short term longitudinal study using fractal signature analysis. Osteoarthritis Cartilage 2005; 13: 9981003.
  • 14
    Link TM, Vieth V, Langenberg R, Meier N, Lotter A, Newitt D, et al. Structure analysis of high resolution magnetic resonance imaging of the proximal femur: in vitro correlation with biomechanical strength and BMD. Calcif Tissue Int 2003; 72: 15665.
  • 15
    Blumentkrantz G, Lindsey CT, Dunn TC, Jin H, Ries MD, Steinbach LS, et al. A pilot, two-year longitudinal study of the interrelationship between trabecular bone and articular cartilage in the osteoarthritic knee. Osteoarthritis Cartilage 2004; 12: 9971005.
  • 16
    Lindsey CT, Narasimhan A, Adolfo JM, Jin H, Steinbach LS, Link T, et al. Magnetic resonance evaluation of the interrelationship between articular cartilage and trabecular bone of the osteoarthritic knee. Osteoarthritis Cartilage 2004; 12: 8696.
  • 17
    Sandini L, Arokoski JP, Jurvelin JS, Kroger H. Increased bone mineral content but not bone mineral density in surgically treated knee and hip osteoarthritis. J Rheumatol 2005; 32: 19517.
  • 18
    Li B, Aspden RM. Composition and mechanical properties of cancellous bone from the femoral head of patients with osteoporosis or osteoarthritis. J Bone Miner Res 1997; 12: 64151.
  • 19
    Chappard C, Peyrin F, Bonnassie A, Lemineur G, Brunet-Imbault B, Lespessailles E, et al. Subchondral bone micro-architectural alterations in osteoarthritis: a synchrotron micro-computed tomography study. Osteoarthritis Cartilage 2006; 14: 21523.
  • 20
    Coats AM, Zioupos P, Aspden RM. Material properties of subchondral bone from patients with osteoporosis or osteoarthritis by microindentation testing and electron probe microanalysis. Calcif Tissue Int 2003; 73: 6671.
  • 21
    Day JS, Ding M, van der Linden JC, Hvid I, Sumner DR, Weinans H. A decreased subchondral trabecular bone tissue elastic modulus is associated with pre-arthritis cartilage damage. J Orthop Res 2001; 19: 9148.
  • 22
    Hunter DJ, Hart D, Sneider H, Bettica P, Swaminathan R, Spector TD. Evidence of altered bone turnover, vitamin D and calcium regulation with knee osteoarthritis in female twins. Rheumatology (Oxford) 2003; 42: 13116.
  • 23
    Plumb MS, Aspden RM. High levels of fat and (n-6) fatty acids in cancellous bone in osteoarthritis. Lipids Health Dis 2004; 3: 12.
  • 24
    Aspden RM, Scheven BA, Hutchison JD. Osteoarthritis as a systemic disorder including stromal cell differentiation and lipid metabolism. Lancet 2001; 357: 111820.
  • 25
    Dieppe P. Osteoarthritis: time to shift the paradigm. This includes distinguishing between severe disease and common minor disability [editorial]. BMJ 1999; 318: 1299300.
  • 26
    Reikeras O, Hoiseth A. Femoral-neck angles in osteoarthritis of the hip. Acta Orthop Scand 1982; 53: 7814.
  • 27
    Moore RJ, Fazzalari NL, Manthey BA, Vernon-Roberts B. The relationship between head-neck-shaft angle, calcar width, articular cartilage thickness and bone volume in arthrosis of the hip. Br J Rheumatol 1994; 33: 4326.
  • 28
    Vossinakis IC, Karnezis IA, Parry K, Learmonth ID. Radiographic associations for “primary” hip osteoarthrosis: a retrospective cohort study of 47 patients. Acta Orthop Scand 2001; 72: 6008.
  • 29
    Beck M, Kalhor M, Leunig M, Ganz R. Hip morphology influences the pattern of damage to the acetabular cartilage: femoroacetabular impingement as a cause of early osteoarthritis of the hip. J Bone Joint Surg Br 2005; 87: 10128.
  • 30
    Jacobsen S, Sonne-Holm S. Hip dysplasia: a significant risk factor for the development of hip osteoarthritis. A cross-sectional survey. Rheumatology (Oxford) 2005; 44: 2118.
  • 31
    Burger H, van Daele PL, Odding E, Valkenburg HA, Hofman A, Grobbee DE, et al. Association of radiographically evident osteoarthritis with higher bone mineral density and increased bone loss with age: the Rotterdam Study. Arthritis Rheum 1996; 39: 816.
  • 32
    Cootes TF, Taylor CJ. Active shape models. In: HoggD, BoyleR, editors. British Machine Vision Conference. Third ed. Berlin, Germany: Springer-Verlag; 1992. p. 26675.
  • 33
    Cootes TF, Taylor CJ, Cooper DH, Graham J. Active shape models: their training and application. Computer Vision and Image Understanding 1995; 61: 3859.
  • 34
    Reijman M, Hazes JM, Pols HA, Bernsen RM, Koes BW, Bierma-Zeinstra SM. Role of radiography in predicting progression of osteoarthritis of the hip: prospective cohort study. BMJ 2005; 330: 11837.
  • 35
    Cootes TF, Hill A, Taylor CJ, Haslam J. Use of active shape models for locating structures in medical images. Image and Vision Computing 1994; 12: 35565.
  • 36
    Cootes TF, Taylor CJ. Statistical models of appearance for medical image analysis and computer vision. Proceedings of SPIE (The International Society for Optical Engineering)2001; 4322: 23648.
  • 37
    Gregory JS, Testi D, Stewart A, Undrill PE, Reid DM, Aspden RM. A method for assessment of the shape of the proximal femur and its relationship to osteoporotic hip fracture. Osteoporos Int 2004; 15: 511.
  • 38
    Long LR, Thoma GR. Use of shape models to search digitized spine X-rays. Proceedings of the 13th IEEE Symposium on Computer-Based Medical Systems; 2001: 13.
  • 39
    Thodberg HH, Rosholm A. Application of the active shape model in a commercial medical device for bone densitometry. Image and Vision Computing 2003; 21: 115561.
  • 40
    Smyth PP, Taylor CJ, Adams JE. Vertebral shape: automatic measurement with active shape models. Radiology 1999; 211: 5718.
  • 41
    Koo S, Gold GE, Andriacchi TP. Considerations in measuring cartilage thickness using MRI: factors influencing reproducibility and accuracy. Osteoarthritis Cartilage 2005; 13: 7829.
  • 42
    Mills HJ, Horne JG, Purdie GL. The relationship between proximal femoral anatomy and osteoarthrosis of the hip. Clin Orthop Relat Res 1993; 288: 2058.
  • 43
    Laforgia R, Specchiulli F, Solarino G, Nitti L. Radiographic variables in normal and osteoarthritic hips. Bull Hosp Jt Dis 1996; 54: 21521.
  • 44
    Marcelli C, Favier F, Kotzki PO, Ferrazzi V, Picot MC, Simon L. The relationship between osteoarthritis of the hands, bone mineral density, and osteoporotic fractures in elderly women. Osteoporos Int 1995; 5: 3828.
  • 45
    Beattie KA, Boulos P, Duryea J, O'Neill J, Pui M, Gordon CL, et al. The relationships between bone mineral density in the spine, hip, distal femur and proximal tibia and medial minimum joint space width in the knees of healthy females. Osteoarthritis Cartilage 2005; 13: 8728.
  • 46
    Stewart A, Black A, Robins SP, Reid DM. Bone density and bone turnover in patients with osteoarthritis and osteoporosis. J Rheumatol 1999; 26: 6226.