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

Although knee malalignment is assumed to correlate with knee osteoarthritis (OA), it is still unknown whether malalignment precedes the development of OA or whether it is a result of OA. The aim of this study was to assess the relationship between malalignment and the development of knee OA as well as progression of knee OA.

Methods

A total of 1,501 participants in the Rotterdam study were randomly selected. Knee OA at baseline and at followup (mean followup 6.6 years) was scored according to the Kellgren/Lawrence (K/L) grading system. Alignment was measured by the femorotibial angle on radiographs at baseline. Multivariable logistic regression for repeated measurements was used to analyze the association of malalignment with the development and progression of OA.

Results

Of 2,664 knees, 1,012 (38%) were considered to have normal alignment, 693 (26%) had varus alignment, and 959 (36%) had valgus alignment. A comparison of valgus alignment and normal alignment showed that valgus alignment was associated with a borderline significant increase in development of knee OA (odds ratio [OR] 1.54, 95% confidence interval [95% CI] 0.97–2.44), and varus alignment was associated with a 2-fold increased risk (OR 2.06, 95% CI 1.28–3.32). Stratification for body mass index showed that this increased risk was especially seen in overweight and obese individuals but not in non-overweight persons. The risk of OA progression was also significantly increased in the group with varus alignment compared with the group with normal alignment (OR 2.90, 95% CI 1.07–7.88).

Conclusion

An increasing degree of varus alignment is associated not only with progression of knee OA but also with development of knee OA. However, this association seems particularly applicable to overweight and obese persons.

Knee osteoarthritis (OA) is the most common joint disorder and is characterized by abnormal articular cartilage and subchondral bone of the tibiofemoral joint. In The Netherlands, >335,000 of the 16 million inhabitants have knee OA (1). The risk of disability attributable to knee OA alone is as great as that attributable to cardiac disease and greater than that attributable to any other medical condition in elderly persons (2). Knee OA also substantially increases the risk of disability due to other medical conditions (3). Because of aging of the population, the prevalence of OA is expected to increase substantially in the coming decades (4).

Malalignment (valgus or varus) of the knee is assumed to correlate with unicompartmental OA of the knee. However, it is still unknown whether malalignment precedes the development of radiographic OA, whether malalignment is a result of OA, or (even more likely) whether the relationship between malalignment and OA is bidirectional. A reason for the relatively small number of epidemiologic studies dealing with malalignment might be that malalignment is mainly measured by means of the hip–knee–ankle (HKA) angle on full-limb radiographs, assessing the mechanical axis in the knee. Full-limb radiographs are used specifically to determine the HKA angle and, in most cases, to surgically adjust this angle. However, this method is cumbersome, requires specialized equipment and expertise, and is costly (particularly for large epidemiologic studies) (5). Whereas some studies have shown a positive relationship between malalignment and progression of knee OA (6–9), others found that the presence of a varus or valgus deformity, although it is unclear how this was assessed, did not differ between patients in whom OA progressed and those in whom OA did not progress (10). No study has investigated the relationship between malalignment and development of OA using HKA angle measurements.

In clinical practice, obtaining anteroposterior (AP) knee radiographs is the most common way to evaluate knee OA radiographically; these radiographs are also used in most large epidemiologic studies. On AP radiographs, the femorotibial (FT) angle can be measured, which defines the anatomic axis in the knee. It was recently shown that this method correlates moderately (r = 0.75) with measurement of the HKA angle on full-limb radiographs (5). Even more recently, Hinman et al (11) confirmed in another study that the anatomic axis is a valid alternative to the HKA angle for determining frontal plane knee alignment (r = 0.88). This method, however, has not been used to determine the influence of alignment on the development of knee OA. We identified one study in which this method was used to determine the influence of the FT angle on annual cartilage loss measured on magnetic resonance imaging in patients with moderate knee OA; that study showed a significant relationship between medial femoral cartilage loss and the baseline FT angle (12).

Therefore, in the present study we investigated the association between malalignment, based on the FT angle on AP radiographs, and the development of OA in individuals who did not have knee OA at baseline. We also investigated whether malalignment assessed with the FT angle is a prognostic factor for progression of OA in individuals with knee OA at baseline.

PATIENTS AND METHODS

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

Participants.

The current study population comprised participants in the Rotterdam Study, which is an open-population, prospective cohort study on the incidence of and risk factors for chronic disabling diseases. In the Rotterdam Study, all 10,275 inhabitants of one district of Rotterdam (Ommoord) ages 55 years and older were invited to participate. The response rate was 78%, which means that 7,983 men and women participated. All participants gave written informed consent and were visited by a trained research assistant for a home interview. When these participants were willing and able to visit the research center, radiographic and other medical examinations were performed 2 weeks later. Of the 7,983 participants, 6,450 visited the research center for a baseline medical examination, and 3,585 of these revisited the center after 6.6 years of followup (Figure 1). Compared with the total Rotterdam Study population, the population available for followup was significantly younger (70.6 years versus 66.4 years) (13). For the present study, a random subset of 1,501 of the participants available for followup was used, and radiographs of the knees from this group were read.

thumbnail image

Figure 1. Flow chart showing subjects from eligibility to inclusion in the present study.

Download figure to PowerPoint

Baseline measurements were obtained between April 1990 and July 1993, and the followup measurements were obtained between 1996 and 1999 (mean ± SD followup 6.6 ± 0.5 years). The Medical Ethics Committee of the Erasmus Medical Center approved the Rotterdam Study.

Radiographic measurements.

To assess radiographic OA, AP radiographs of the lower extremity with weight-bearing were obtained at 70 KV, a focus of 1.8 mm2, and a focus-to-film distance of 120 cm, using High Resolution G 35 × 43–cm radiographic film (Fujifilm Medical Systems, Stamford, CT). Radiographs of the extended knee were obtained with the patella in central position.

Radiographic OA was graded using the Kellgren/Lawrence (K/L) scale for both tibiofemoral compartments together (14). Two trained readers who were blinded to the clinical status of patients independently evaluated the radiographs of the knee. After each set of 150 radiographs was evaluated, the scores of the 2 readers were assessed. Whenever the K/L grade differed, the 2 readers met to read the radiographs together, and a consensus score was determined. Radiographic OA was defined as being present when the K/L grade was ≥2.

Alignment was measured as the medial angle formed by the femur and tibia (FT angle) (Figure 2), using a method based on that described by Moreland et al (15). Lines were drawn through the middle of the femoral shaft and through the middle of the tibial shaft. The medial angle subtended at the point at which these 2 lines met in the center of the tibial spines was characterized as the anatomic angle (16).

thumbnail image

Figure 2. Method used to measure the femorotibial angle.

Download figure to PowerPoint

The knees were divided into 3 groups. A distinction was made between normal alignment, valgus alignment, and varus alignment. Normal alignment was considered to be an FT angle between 182° and 184°. A knee was defined as valgus when alignment was >184° and as varus when alignment was <182°. These cutoffs were based on values for normal, varus, and valgus alignment for the mechanical axis angle from a full-limb radiograph, as described by Moreland et al (15), with adjustment (+4°) for the offset in valgus direction when measured as FT angle on AP extended-knee radiographs, as reported by Kraus et al (5). We did not use decimals as in the original cutoff values or offset values, because our measurements had only 1° increments.

Two independent observers who were blinded to followup data each measured the FT angle of 1,501 knees. To assess interobserver reproducibility, 98 radiographs were measured by 2 observers blinded to each other's results.

Statistical analysis.

Reproducibility of the FT angle measurements was assessed by calculating intraclass correlation coefficients (ICCs) using a two-way mixed-effect model. Odds ratios (ORs) were calculated using generalized estimating equations (GEEs). ORs represented the likelihood that OA would develop or progress in knees with malalignment at baseline (compared with knees without malalignment at baseline). This is a logistic regression for repeated measurements, to take into account the correlation between the left and right knees. These analyses were executed for valgus and varus alignment, with normal alignment as a reference group. Analyses were adjusted for age (continuous variable), sex, and body mass index (BMI; continuous variable).

The relationship between valgus or varus alignment and the development of knee OA (K/L grade ≥2) was assessed for knees with K/L grades 0 and 1 together and for knees with K/L grade 0 and K/L grade 1 separately.

Progression of OA of the knee was defined by subtracting the K/L grade at baseline from the K/L grade at followup (outcome ≥1) and was assessed in knees with K/L grade 2. Knees with K/L grades 3 and 4 were excluded due to very few cases, and knees that were not able to progress any further (K/L grade 4 at baseline) were also excluded. Again, alignment was assessed as normal versus valgus and varus.

Furthermore, given the role of being overweight in the development of knee OA (17, 18), additional analyses were executed stratified for participants who were obese, those who were overweight, and those who were not overweight. The BMI was used for this stratification, with the following 3 groups: <25 kg/m2, ≥25 kg/m2 and <30 kg/m2, and ≥30 kg/m2. Overweight was defined as a BMI of 25–30 kg/m2, and obesity was defined as a BMI of ≥30 kg/m2. All statistical tests were 2-sided and were conducted using an alpha error of 0.05. SPSS version 11.0 software (SPSS, Chicago, IL) and SAS version 8.2 software (SAS Institute, Cary, NC) were used for all analyses.

RESULTS

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

Baseline characteristics of the cohort.

The study population comprised 1,501 participants. The mean age was 66.4 years, and almost 60% of the participants were women. Table 1 presents the baseline characteristics of the total study population. The interreader assessment of the FT angle measurement showed high reproducibility (ICC 0.90).

Table 1. Baseline characteristics of the study population (n = 1,501)*
Characteristic 
  • *

    Except where indicated otherwise, values are the percent. Information on alignment was missing for 13% of left knees and 9.5% of right knees. K/L = Kellgren/Lawrence.

Female sex59.4
Age, mean ± SD years66.4 ± 6.7
Body mass index, mean ± SD kg/m226.3 ± 3.6
Femorotibial angle, mean ± SD 
 Left knee183.9 ± 3.6
 Right knee183.3 ± 3.3
Varus alignment 
 Left knee20.8
 Right knee25.4
Valgus alignment 
 Left knee34.5
 Right knee28.8
Normal alignment 
 Left knee31.7
 Right knee36.2
K/L grade 0 
 Left knee73.0
 Right knee69.8
K/L grade 1 
 Left knee13.9
 Right knee14.2
K/L grade 2 
 Left knee12.1
 Right knee14.9
K/L grade 3 
 Left knee1.1
 Right knee1.1
K/L grade 40.0
 Left knee0.0
 Right knee0.1

Among the 3,002 knees, information on 338 knees (11%) was missing due to the absence (or illegibility) of the radiographs. The remaining 2,664 knees were divided into 3 categories as described above. Of these, 1,012 knees (38%) were considered to have normal alignment, 693 (26%) had varus alignment, and 959 (36%) had valgus alignment. Figure 3 shows the distribution of the FT angle over all knees.

thumbnail image

Figure 3. Histogram showing the distribution of alignment of the knees.

Download figure to PowerPoint

Development of knee OA.

Among knees without OA (K/L grade 0 or 1) at baseline, OA had developed at followup in 45 (5.5%) of the 819 knees with valgus alignment, in 43 (7.4%) of the 579 knees with varus alignment, and in 35 (3.9%) of the 892 knees with normal alignment. In GEE logistic regression analyses, valgus versus normal alignment at baseline was associated with a borderline significantly increased risk of developing OA for those with K/L grade 0 or 1 at baseline (OR 1.54, 95% confidence interval [95% CI] 0.97–2.44). The knees with varus alignment showed a significant increase in the risk of developing OA for those with K/L grade 0 or 1 at baseline (OR 2.06, 95% CI 1.28–3.32) (Table 2). Similar estimates were obtained when knees with K/L grade 0 and knees with K/L grade 1 at baseline were analyzed separately, although valgus alignment was no longer associated with a significantly increased risk of developing OA (Table 2).

Table 2. Association between alignment at baseline and development of knee OA after 6.6 years of followup in knees without OA at baseline*
AlignmentNo. of knees without OA at baselineNo. of knees with OA at followupOR95% CIP
  • *

    No osteoarthritis (OA) at baseline was defined as a Kellgren/Lawrence (K/L) grade of 0 or 1. OR = odds ratio; 95% CI = 95% confidence interval.

  • Adjusted for age, sex, and body mass index.

K/L grades 1 and 2     
 Normal892351 (reference)
 Valgus819451.540.97–2.440.065
 Varus579432.061.28–3.320.003
K/L grade 0     
 Normal752191 (reference)
 Valgus694271.630.89–3.000.112
 Varus484231.951.02–3.730.043
K/L grade 1     
 Normal140161 (reference)
 Valgus125181.520.77–3.010.232
 Varus95202.121.00–4.460.049

Progression of OA.

We also examined the relationship between baseline alignment and progression of knee OA (increase in the K/L score of ≥1 at followup in knees with a K/L score at baseline of ≥2). Progression occurred in 8 (6.3%) of the 128 knees with valgus alignment, 11 (11.6%) of the 95 knees with varus alignment, and 6 (4.8%) of the 125 knees with normal alignment. In GEE logistic regression analyses, valgus versus normal alignment was not significantly associated with progression of knee OA (OR 1.39, 95% CI 0.48–4.05). In knees with varus alignment, analyses showed significantly increased odds for progression (OR 2.90, 95% CI 1.07–7.88) compared with knees with normal alignment (Table 3).

Table 3. Association between alignment at baseline and progression of knee OA after 6.6 years of followup in knees with OA at baseline*
AlignmentNo. of knees with OA at baselineNo. of knees with progression at followupOR95% CIP
  • *

    Osteoarthritis (OA) at baseline was defined as a Kellgren/Lawrence (K/L) grade of ≥2. OR = odds ratio; 95% CI = 95% confidence interval.

  • Adjusted for age, sex, and body mass index.

Normal12561 (reference)
Valgus12881.390.48–4.050.550
Varus95112.901.07–7.880.037

Analyses stratified by BMI.

For stratification according to being overweight, fewer knees were included in the analyses due to the absence of 7 BMI measurements. Among individuals (123 knees) with K/L grade 0 or 1 at baseline and in whom OA developed, 30 persons had a BMI of <25 kg/m2, 62 persons had a BMI of 25–29.99 kg/m2, and 31 persons had a BMI of ≥30 kg/m2. Comparing the valgus group with the normal alignment group, a nonsignificant OR of 1.42 (95% CI 0.78–2.60) for the development of OA was observed for individuals with a BMI of ≥25 kg/m2 but <30 kg/m2, and a significant OR of 3.25 (95% CI 1.14–9.27) was associated with a BMI of ≥30 kg/m2. In the non-overweight group, the analyses showed a nonsignificant OR of 1.08 (95% CI 0.41–2.80) for the group with valgus alignment compared with the group with normal alignment (Table 4).

Table 4. Association between alignment at baseline and development of knee OA after 6.6 years of followup in knees without OA at baseline, stratified for BMI (kg/m2)*
AlignmentNo. of knees without OA at baselineNo. of knees with OA at followupOR95% CIP
  • *

    OA = osteoarthritis; BMI = body mass index; OR = odds ratio; 95% CI = 95% confidence interval.

  • Adjusted for age and sex.

BMI <25     
 Normal347111 (reference)
 Valgus28981.080.41–2.800.881
 Varus287111.240.50–3.100.640
BMI ≥25 and <30     
 Normal442191 (reference)
 Valgus432231.420.78–2.600.252
 Varus240202.021.07–3.840.031
BMI ≥30     
 Normal10351 (reference)
 Valgus98143.251.14–9.270.028
 Varus52125.061.71–14.940.003

When we compared the group with varus alignment with the group with normal alignment, a significant OR of 2.02 (95% CI 1.07–3.84) was observed for persons with a BMI ≥25 kg/m2 but <30 kg/m2, and an even stronger OR (5.06 [95% CI 1.71–14.94]) was associated with a BMI of ≥30 kg/m2. The analyses in the non-overweight group showed a nonsignificant OR (1.24 [95% CI 0.50–3.10]) (Table 4). Stratification for the analyses of progression in knees with OA (K/L grade ≥2 at baseline) yielded very few cases in each stratum, especially in the non-overweight group, and was deemed clinically and statistically meaningless.

DISCUSSION

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

Until now, only animal model data have supported a link between preexisting varus or valgus alignment and development of knee OA (19). Using data from a large population-based cohort study, our study is the first to show that malalignment is a risk factor for development of knee OA. It also further strengthened the concept that malalignment is positively related to progression of knee OA.

A clear limitation of the present study is that we did not obtain full-limb radiographs for accurate measurement of mechanical alignment in the population studies. Although the anatomic axis of the tibia is assumed to be straight (20), the bowing curvature of the tibia might lead to differences between anatomic alignment (measured by knee angle) and mechanical alignment using the entire tibia. Kraus et al (5) reported a mean offset of ∼4° in the valgus direction for anatomic alignment compared with mechanical alignment. For this reason, we used this reported offset in our definitions of normal, varus, and valgus alignment. However, Kraus et al and Hinman et al (11) also reported that measurements of mechanical alignment and measurement of anatomic alignment were strongly but not optimally correlated (r = 0.75 and r = 0.88, respectively). Therefore, the strength of the relationship between alignment and the increase in the K/L score might have been stronger if a full-limb assessment of alignment had been used and mechanical alignment had been measured directly.

This difference in measurement technique may explain the stronger relationship (OR 4.1) between knee angle and radiographic progression seen in the study by Sharma et al (7). Another explanation for the difference in ORs between our study and that of Sharma et al might be the difference in methods used to assess progression of knee OA. Sharma and colleagues used an increase of >1 grade in severity of joint space narrowing, whereas we used the K/L scale. Both methods are used to assess OA, but the sensitivity to measure change is higher for the methods measuring joint space width than for methods using the K/L score (21). Moreover, the sensitivity to change also depends on the manner in which radiographs are obtained. Sharma et al used a semiflexed, fluoroscopically confirmed knee radiograph, which is a better radiographic method than the standing, weight-bearing, fully extended AP radiographs used in our study (7). In fully extended AP radiographs, it is not always possible to acquire the fully extended position; pain and stiffness of the joint make it harder to completely extend the knee joint, which provides a smaller joint space width, which might overestimate disease progression (22). Therefore, it might be noteworthy that the reported relationships, although less strong, were detected despite the radiographic methods used. Finally, simple differences in the cohorts, for example a higher mean BMI in the population described by Sharma et al, might explain the stronger relationship in that cohort.

Another point of discussion is that the adduction moment, rather than alignment, reflects the dynamic load on the medial compartment of the knee (8). The adduction moment of the knee is a major determinant of medial-to-lateral load distribution; thus, it is responsible for the biomechanical abnormality of medial compartment knee OA. Sharma et al reported that dynamic load during gait correlated with disease severity in tibiofemoral knee OA. They suggested that the magnitude of the adduction moment possibly influences the structural outcome in medial compartment knee OA. Although the FT angle measured in our study might be a proxy for unfavorable load as a risk factor, this might also be the case for the HKA angle. Using univariate analyses, Miyazaki et al (8) showed that the HKA angle, as well as the adduction moment, were related to radiographic progression; however, in multivariable analyses only the adduction moment was significantly related, and the HKA angle lost its relationship and significance.

Because of the arbitrariness of the cutoff point for the absence of radiographic OA, we also assessed the risk of development of OA in knees with K/L grades 0 and 1 at baseline separately. Knees with K/L grade 2 were used to analyze progression. However, data for individuals who had K/L grade 1 at baseline with progression to K/L grade ≥2 that were used for the analysis of development of OA can also be used to assess progression. For progression, however, the association tended to be stronger in persons with K/L grade 2 than in those with K/L grade 1. The risk of development of OA was similar for individuals with a baseline K/L grade of 0 or 1, indicating that perhaps the usual definitions for non-OA (K/L grades 0 or 1) and OA (K/L grade ≥2) might be acceptable after all.

Of the 6,450 participants who visited a research center for a baseline examination, 3,585 revisited the center after 6.6 years of followup. A possible limitation of our study is potential health-based selection bias. The subjects in the present study had to be mobile enough to visit the research center at baseline and had to survive the followup period or be healthy enough to visit the research center at followup. In other words, individuals with the most severe symptoms were most likely not included. Therefore, it seems probable that in this younger and healthier population with less frequent lower-limb disability and pain, the prevalence of radiographic knee OA or progression of radiographic OA at followup may have been underestimated. Based on the report by Sharma et al that malalignment is related to a faster functional decline in knee OA (7), we emphasize that the relationship reported in our study may be an underestimation.

Based on our analysis, we now know that malalignment predates knee OA, and this might be attributable to genetic, posttraumatic, or developmental factors. Unfortunately, our study does not provide any data on former knee injury and possible subsequent meniscectomy; further studies on the background of malalignment are necessary.

Our results show a stronger relationship between malalignment and the risk of development of OA in the overweight group than in the non-overweight group. The relationship might even be absent in the non-overweight group, suggesting that an unfavorable load is much less harmful in non-overweight persons. However, our study showed a low number of knees with disease progression, a common problem in population-based studies, and therefore the power was too low (too few cases in each stratum, especially in the non-overweight group) to compare the relationships between varus or valgus alignment and risk of progression of OA in the non-overweight, overweight, and obese groups. Therefore, we do not yet know whether being overweight or obese increases the relationship between malalignment and progression of knee OA in a similar manner. Interestingly, Felson et al showed that high body weight increases the risk of structural progression of knee OA, but its effect on progression appears to be limited to knees with moderate malalignment (17). Together with the findings from our study, these results suggest that these factors mutually influence each other.

The present results indicate a need to investigate the effectiveness of preventive interventions to reduce the risk of varus load, especially in overweight persons. For secondary prevention, more active interventions (besides weight reduction) such as wedged insoles should be evaluated.

In conclusion, we used FT angle measurements to confirm relationships between malalignment and progression of knee OA. This study is the first to show an association between malalignment at baseline and the risk of development of knee OA; this association seems particularly applicable to overweight and obese persons.

Acknowledgements

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

We are very grateful to Dr. Odding for scoring the radiographs of the knee, and F. van Rooij, E. van der Heijden, R. Vermeeren, and L. Verwey for collecting followup data. Moreover, we thank the participating general practitioners, the pharmacists, the many field workers at the research center in Ommoord, and, of course, all of the participants.

AUTHOR CONTRIBUTIONS

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

Dr. Bierma-Zeinstra 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. Brouwer, van Tol, Pols, Bierma-Zeinstra.

Acquisition of data. Brouwer, van Tol, Bergink.

Analysis and interpretation of data. Brouwer, van Tol, Bergink, Bernsen, Reijman, Bierma-Zeinstra.

Manuscript preparation. Brouwer, van Tol, Bergink, Belo, Bernsen, Reijman, Pols, Bierma-Zeinstra.

Statistical analysis. Brouwer, van Tol, Bernsen, Bierma-Zeinstra.

REFERENCES

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