Patient characteristics as predictors of clinical outcome of distraction in treatment of severe ankle osteoarthritis

Authors


  • M.C.L. Hoekstra and B.C. du Pré contributed equally to this work.

Abstract

Osteoarthritis (OA) is a slowly progressive joint disease. Joint distraction can be a treatment of choice in case of severe OA. Prediction of failure will facilitate implementation of joint distraction in clinical practice. Patients with severe ankle OA, who underwent joint distraction were included. Survival analysis was performed over 12 years (n = 25 after 12 years). Regression analyses were used to predict failures and clinical benefit at 2 years after joint distraction (n = 111). Survival analysis showed that 44% of the patients failed, 17% within 2 years and 37% within 5 years after joint distraction (n = 48 after 5 years). Survival analysis in subgroups showed that the percentage failure was only different in women (30% after 2 years) versus men (after 11 years still no 30% failure). In the multivariate analyses female gender was predictive for failure 2 years after joint distraction. Gender and functional disability at baseline predicted more pain. Functional disability and pain at baseline were associated with more functional disability. Joint distraction shows a long-term clinical beneficial outcome. However, failure rate is considerable over the years. Female patients have a higher chance of failure during follow-up. Unfortunately, not all potential predictors could be investigated and other clinically significant predictors were not found. © 2013 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 32:96–101, 2014.

Osteoarthritis (OA) is a slowly progressive joint disease characterized by progressive cartilage damage and loss, changes in peri-articular bone, and often secondary joint inflammation. These changes in tissue structure are to a certain extent associated with pain, stiffness, and functional disabilities. Multiple risk factors, such as age, gender, genetics, obesity, and joint trauma contribute to the initiation of the disease.[1, 2]

In the end stage of ankle OA, joint fusion or joint replacement are final options for treatment, but have their limitations.[3] Joint fusion goes at the expense of joint mobility and joint replacement at the expense of the original joint. Moreover, not in favor of both options, the aging population prefers an active life as long as possible.

To prevent or postpone joint fusion or joint replacement, joint distraction can be a treatment of choice. During joint distraction, the ankle joint is mechanically unloaded for 2–3 months by an external fixation frame. Mechanical contact between the degenerating articular surfaces is eliminated, while intermittent synovial fluid pressure, which is presumed to be essential for nutrition and stimulation of chondrocytes,[4] is maintained, allowing initiation of cartilage repair. Additionally, joint distraction results in significant peri-articular bone turnover during and after treatment. This initial loss in bone density regained after distraction might add biochemically and biomechanically to cartilage repair mechanisms.[5-7]

Joint distraction in the treatment of severe ankle OA shows significant and prolonged improvement of pain and functional disability in open prospective studies as well as in a randomized controlled trial.[8, 9] Despite this on average long-term beneficial clinical effect of joint distraction, part of the patients can be designated as failures, requiring in the years after distraction still the conventional treatment, an arthrodesis or joint replacement.

For example, knee arthroplasty, patient characteristics such as age and gender were found to be predictive of clinical outcome.[10] Similarly, prediction of failure on joint distraction treatment might be of value as it will increase the clinical benefit of joint distraction and with that facilitate implementation of joint distraction in clinical practice.

The aim of the present study was to evaluate the number of failures in the long term after joint distraction in the treatment of severe ankle OA and predict failure and clinical outcome within the first 2 years as well as over time by use of demographics and disease characteristics at baseline before start of treatment.

PATIENTS AND METHODS

Patients

Patients, who underwent joint distraction as a treatment for severe ankle OA, participating in one of the two following prospective studies, were included.

  1. An open prospective multi-center study (n = 75) at the University Medical Center Utrecht (UMCU), The Netherlands, the Onze Lieve Vrouwe Middelares Hospital (OLVM), Antwerp Belgium, and at the University Medical Center Groningen (UMCG), The Netherlands. Patients underwent fixed joint distraction between 1993 and 2001, according to the method as described by Marijnissen et al.[8] All distraction procedures were performed by a single surgeon in each institute. The mean follow up was 8.1 ± 5.1 years. The maximum follow up was 17 years.
  2. A randomized controlled trial (n = 36) at the University of Iowa Hospitals and Clinics, Iowa, USA. Patients were treated with either fixed (n = 18) or joint motion permitted (n = 18) ankle distraction between 2003 and 2007. The mean follow-up was 1.9 ± 0.5 years. The maximum follow-up was 2 years.[9]

All patients had severe ankle OA, mostly post-traumatic, and were considered for arthrodesis. For detailed inclusion criteria see Refs.[8, 9]

Collection of Data

Data collection was performed in a prospective way, before and yearly after surgery[8] gender, age, origin (EU vs. USA), clinical condition (swelling, pain at palpation, etc.), and BMI were collected at baseline. Pain and functional disability were evaluated at baseline and every year after the distraction (from the patients who were not considered failures) using questionnaires. Both parameters are expressed as percentage of the maximum score. In the open prospective study[8] the van Valburg functional disability questionnaire and the box scale for pain were used, while in the randomized controlled trial,[9] the Ankle Osteoarthritis Scale (AOS) was used. Significant correlations were found between the box scale for pain and the AOS pain questionnaire (r = 0.568; p < 0.03) and between the van Valburg functional disability questionnaire and the AOS disability questionnaire (r = 0.714; p < 0.004).[11] In the open prospective study, clinical condition was evaluated by physical examination,[8] while in the randomized controlled trial no data on clinical condition were available. BMI was only evaluated in the randomized controlled trial.

Outcome Parameters

  1. Failure. Failures were defined as patients who underwent an arthrodesis, developed Sudeck's atrophy, underwent osteotomy or underwent a second distraction. Failure at 2 years as well as failure over time during the entire follow-up period was used as endpoint.
  2. Clinical effect. Pain and functional disability as endpoints were defined at 2 years since a 2-year follow-up measurement was available for almost all patients in both cohorts.

Statistical Analysis

Baseline characteristics of the patients from the open prospective study[8] and the randomized controlled trial[9] were compared and tested using the independent t-test and pearson's Chi-square test for continuous and categorical variables, respectively.

Evaluation of Failures of Ankle Distraction

Survival curves with failure as end point were constructed using the Kaplan–Meier method. Survival times were censored at the time of loss to follow-up or maximum follow-up. Due to the study design, the maximum follow-up period of patients from the randomized clinical trial was 2 years.

To assess the association between a specific patient characteristic and failure rate three equally sized subgroups were defined (based on the distribution of the variable) for respectively age, clinical condition, pain and functional disability at baseline. For gender, the two subgroups were defined. Survival curves were compared between subgroups and tested for statistical differences using the log rank test.

Prediction of Failure and Clinical Outcome at 2 Years

Data from all patients with a minimum follow-up of 2 years were used for the analysis. Binary logistic regression analyses were performed with failure at 2 years as dependent variable and linear regression analyses were performed with pain and functional disability at 2 years as dependent variables in separate analyses. Missing data were replaced by last observation carried forward. In case of failure, the last pain and functional disability scores before failure were used for imputation of the 2-year outcome. Gender, age, origin (EU vs. USA), pain, functional disability, clinical condition, and BMI all at baseline were used as independent variables in the regression analyses. Additionally “motion permitted distraction” was included, since joint motion during distraction was allowed in 50% of the patients of the randomized controlled trial.

Univariate and multivariate analyses were performed. In the multivariate analyses all variables were initially included and variables were removed manually based on their estimates (β's, statistical significance, and changes in explained variance (R2) for the linear regression analyses) using a backward stepwise selection procedure. Since this was a first study regarding the prediction of outcome of joint distraction, a liberal p value (i.e., <0.20) was used to select variables in the final model. The variable “origin” is included in every model to account for the different cohorts and set-ups. Dependent on the clinical outcome, pain or function at baseline was added as independent variable to correct for baseline value.

BMI and clinical condition at baseline were not included because these characteristics were not measured in both cohorts. In two separate analyses these two variables were included as extra variables in the regression analyses (either using the data from the open prospective cohort study or the data from the randomized controlled trial).

Prognostic ability of the models was summarized using the area under the curve (AUC) of the receiver operating characteristic curve (ROC) for the analyses regarding failure.

For the analysis regarding pain and disability the explained variance (R2) was used as measure of predictive ability.

Prediction of Time to Failure After Joint Distraction

A univariate and multivariate cox regression analysis was performed with time to failure as dependent variable. Patients were censored at their last visit. The baseline characteristics gender, age, origin, pain, functional disability, and permission of joint motion during distraction, were used as independent variables. A comparable model building approach was used as in the prediction for 2-year outcome.

RESULTS

Survival Analysis in the Long Term

Data of 111 patients with severe ankle OA treated with joint distraction and participating in either the open prospective study[8] or randomized controlled trial[9] were available. Baseline characteristics of these patients are depicted in Table 1. No significant differences in age, gender and functional disability score at baseline were found between the populations of the two studies. However, pain at baseline was statistically significantly higher in the open prospective study compared to the randomized controlled study (p < 0.0001).

Table 1. Baseline Characteristics of Patients (n = 111) Treated With Joint Distraction in the Open Prospective Study or Randomized Controlled Trial
 All PatientsOpen ProspectiveRandomized Controlledp-Value
  1. Mean value ± standard deviation is given for age and BMI. Pain, functional ability, and clinical condition, evaluated using questionnaires, are presented as percentage of the maximum score. Clinical condition was only measured in the open prospective study, while BMI is only known from the patients of the randomized controlled trial.
Gender (male %)6057670.347
Age (years)42.7 ± 9.843.3 ± 10.141.4 ± 9.10.333
Pain (%)68.6 ± 15.172.8 ± 13.460.0 ± 14.9<0.0001
Functional disability (%)68.0 ± 14.868.9 ± 16.066.0 ± 12.20.325
Motion permitted distraction (%)16050<0.0001
Clinical condition (%) 72.6 ± 22.4NA
BMI (kg/m2) NA29.8 ± 4.8

The survival curve (Fig. 1A) shows that over a follow up of 12 years, 44% of the patients had failed. Seventeen percent of the patients failed within 2 years and 37% failed within 5 years after joint distraction. Figure 1B shows the amount of patients who are still in follow-up and who had failed per year of follow-up.

Figure 1.

(A) Survival curve of 111 patients with severe ankle osteoarthritis treated with joint distraction. (B) Number of failures per follow-up year in 111 patients. (C) Survival curve of 111 patients with severe ankle osteoarthritis treated with joint distraction, men (n = 67) versus women (n = 44).

No significant difference in the percentage of failures occurred in the first 2 years after joint distraction was found between the open prospective study and the randomized controlled study (respectively 18% vs. 15%, p = 0.647) and no significant difference was found in time to failure between the two study groups (p = 0.637).

The percentage failure was different in women versus men (Fig. 1C); in women there was a 30% failure after 2 years, and in men 30% failure was still not reached after 11 years. Log-rank test confirmed the statistically significant difference in failure during follow-up between men and women (p = 0.001). Evaluation of other subgroups showed no statistically significant differences in failure over time between groups for age (p = 0.095), functional disability (p = 0.850), pain (p = 0.314), and clinical condition (p = 0.618) at baseline.

Percentage Failure and Clinical Outcome 2 Years After Joint Distraction

Six out of the 111 patients were lost to follow-up within 2 years after joint distraction. Of these six patients, one patient died 15 months after treatment, not related to the treatment. The reason for loss to follow-up of the other five patients was unknown. Ultimately, data of 105 patients was available for 2-year follow-up analyses; all these patients had a minimum follow-up of 2 years or were considered failure within 2 years after distraction.

At 2 years, 18 patients (17%) were considered failures, whereas 87 patients (83%) were not. Most patients who were considered failures underwent an arthrodesis (n = 15), two patients developed Sudeck's atrophy, and one patient underwent an osteotomy.

The average scores for pain and functional disability at baseline of the patients who were still in follow-up were respectively 67.1 ± 15.2% and 68.3 ± 15.1% and decreased to 38.2 ± 23.8% and 35.9 ± 23.0% 2 years after joint distraction. For the patients whose treatment failed, the scores at baseline were respectively 75.9 ± 15.0% and 67.4 ± 13.3% and the last observed scores before failure were 70.8 ± 21.3% and 71.1 ± 17.9% respectively.

Prediction of Failure and Clinical Outcome at 2 Years After Joint Distraction

Failure

Table 2 depicts the results of logistic regression analyses with failure at 2 years as dependent variable. In the univariate regression analyses, female gender and pain at baseline were the only two factors that were statistically significantly associated with failure at 2 years after joint distraction treatment (Table 2). Results of the multivariate analysis revealed that female gender was a statistically significant predictor of failure at 2 years, and motion permitted distraction was protective for failure at 2 years (Table 2). Pain at baseline did not contribute statistically significantly in the multivariate model. The AUC–ROC of the model was 0.754 (p = 0.001; CI: 0.64–0.87). In additional regression analyses in the open prospective study cohort and the randomized controlled trial population respectively, clinical condition and BMI were not associated with outcome (data not shown).

Table 2. Results of Logistic Regression Analyses With Failure at 2 Years After Joint Distraction as Dependent Variable
 UnivariateMultivariate
 OR (95% CI)pOR (95% CI)p
  1. OR, odds ratio; 95% CI, 95% confidence interval; p, significance level. The AUC–ROC (area under the receiver operating characteristic curve) of the multivariate model was 0.754 (p = 0.001; CI: 0.64–0.87). The model was corrected for “origin.”
Gender4.94 (1.6–15.2)0.0055.42 (1.70–17.29)0.004
Age0.10 (0.9–1.0)0.810  
Pain1.05 (1.0–1.1)0.033  
Functional disability1.00 (1.0–1.0)0.808  
Clinical condition1.03 (1.0–1.1)0.159  
Motion permitted distraction0.24 (0.0–1.9)0.1830.132 (0.012–1.46)0.098
BMI0.91 (0.7–1.1)0.407  

Pain

Table 3 depicts results of linear regression analyses with pain at 2 years after joint distraction as dependent variable. In the univariate linear regression analysis pain and functional disability at baseline predicted pain at 2 years after joint distraction while motion permitted distraction showed to be protective for pain at 2 years after joint distraction. In the multivariate linear regression model, functional disability at baseline and male gender predicted more pain, while only motion permitted distraction was a statistically significant protector for pain at 2 years after joint distraction. The explained variance of this model was marginal, 14% (R2 = 0.144).

Table 3. Results of Linear Regression Analyses With Pain at 2 Years After Joint Distraction as Dependent Variable
 UnivariateMultivariate (R2 = 0.144)
 β (95% CI)pβ (95% CI)p
  1. β, regression coefficient; 95% CI, 95% confidence interval; p, significance level. R2 of the multivariate model was 0.144. The model was corrected for “origin.”
Gender9.66 (−0.6 to 19.9)0.0647.51 (−2.61 to 17.62)0.144
Age0.02 (−0.5 to 0.6)0.939  
Pain0.40 (0.08–0.73)0.0150.14 (−0.26 to 0.53)0.492
Functional disability0.35 (0.003–0.7)0.0480.25 (−0.12 to 0.62)0.187
Clinical condition0.22 (−0.1 to 0.5)0.116  
Motion permitted distraction−19.0 (−32.1 to −6.0)0.005−20.3 (−37.02 to −3.04)0.021
BMI−1.11 (−2.9 to 0.7)0.215  

Additional multivariate regression analyses including clinical condition or BMI as extra independent variables, using data from the open prospective study cohort and the randomized controlled trial population respectively, showed clinical condition to be predictive (β: 12.62, CI: −0.203 to 25.45, p = 0.054) in a model (R2 = 0.100) with pain (β: 0.005, CI: −0.468 to 0.478, p = 0.982) and gender (β: 12.62, CI: −0.203 to 25.45, p = 0.054). BMI at baseline was not found to predict pain at 2 years.

Functional Disability

Table 4 depicts the results of univariate and multivariate linear regression analyses with functional disability at 2 years after joint distraction as dependent variable.

Table 4. Results of Linear Regression Analyses With Functional Disability at 2 Years After Joint Distraction as Dependent Variable
 UnivariateMultivariate (R2 = 0.128)
 β (95% CI)pβ (95% CI)p

Note

  1. Only 18 patients underwent a joint motion permitted distraction. R2 of the multivariate model was 0.128. The model was corrected for “origin.”
  2. β, regression coefficient; 95% CI, 95% confidence interval; p, significance level.
Gender7.55 (−2.5 to 17.7)0.141  
Age0.21 (−0.3 to 0.7)0.429  
Pain0.38 (0.06–0.7)0.0200.29 (−0.094 to 0.673)0.138
Functional disability0.38 (0.05–0.7)0.0270.28 (−0.09 to 0.65)0.136
Clinical condition0.10 (−0.2 to 0.4)0.438  
Motion permitted distraction−12.57 (−25.68 to 0.54)0.060−21.06 (−37.89 to −4.23)0.015
BMI−0.48 (−2.6 to 1.7)0.651  

In the univariate linear regression analysis pain and functional disability at baseline predicted functional disability at 2 years. In the multivariate linear regression analysis motion permitted distraction was statistically significantly associated with impaired functional disability at 2 years after joint distraction. Pain and functional disability at baseline were associated with this outcome measurement as well. In additional regression analyses in respectively the open prospective study cohort and the randomized controlled trial population, clinical condition and BMI were not associated with outcome (data not shown).

Prediction of Failure After Joint Distraction Over Time

In the univariate cox regression analysis only gender was associated with failure over time (Table 5). In multivariate cox-regression analyses gender (p: 0.002, hazard ratio: 0.350, CI: 0.181–0.678) was found to be an independent factor that predicts increased chance of failure over time and motion permitted distraction (p: 0.150, hazard ratio: 0.200, CI: 0.022–1.786) was found to be an independent factor that predicts decreased chance of failure over time.

Table 5. Cox Regression Analyses With Failure After Joint Distraction as Dependent Variable
 UnivariateMultivariate
 HR (95% CI)pHR (95% CI)p

Note

  1. Only 18 patients underwent a joint motion permitted distraction. The model was corrected for “origin.”
  2. HR, hazard ratio; 95% CI, 95% confidence interval; p, significance level.
Gender2.83 (1.46–5.48)0.0022.86 (1.48–5.53)0.002
Age1.01 (0.98–1.04)0.688  
Pain1.03 (1.00–1.05)0.062  
Functional disability1.01 (0.99–1.03)0.355  
Clinical condition1.00 (0.99–1.02)0.879  
Motion permitted distraction0.28 (0.04–2.06)0.2090.20 (0.022–1.786)0.150
BMI0.92 (0.75–1.14)0.434  

DISCUSSION

The present study demonstrates that on average slightly more than two third of the patients with severe ankle OA considered for a joint fusion or prosthesis but treated with joint distraction benefit from this treatment for over 5 years. From the eight patients who reached a 15-year follow-up until now, slightly more than half of the patients have still benefit. Failure to treatment was highest in the first 5 years of follow-up. The present study, merged results of an open prospective EU study and a randomized controlled US trial. This results in a unique group of patients, but also has its drawbacks. Long-term follow-up data were not availabe for all included patients, due to the shorter follow-up period of the US trial and due to the fact that patients did not reach the total follow-up time yet. In the patients evaluated, pain and functional disability scores of patients, whose treatment did not fail, decreased significantly in follow up and remained stable over the years thereafter.

Since the set-up of both studies did not focus on the baseline characteristics that could predict outcome, we were not able to evaluate all potential predictive parameters, such as radiological severity of the disease, activity level of patients, or preceding treatments. Nor could we include details on the original trauma with potential joint instability as a result as these data were not available in sufficient detail. This is a limitation of our study as potential predictors of response might have been missed. From the parameters evaluated, only female gender could be identified as a predictor of failure and distraction with motion as a predictor for the degree of clinical benefit. Data show that female gender independently predicts a high chance of failure. These results are in accordance with the results of articulated hip distraction in treatment of femoral head avascular necrosis in adolescence described by Gomez et al.[12] that also found a better outcome in male patients. It is unknown what causes the difference between men and women. Although BMI was measured in the USA cohort (n = 31), potential associations of gender and BMI were not found. Surprisingly, in contrast to cell transplantation techniques in case of treatment of local defects[13] of which successful outcome strongly depends on age, efficacy of distraction appeared despite a rather broad age distribution 42.7 ± 9.8 (mean ± SD) years, not to be predicted by age.

Prospective data of other possible variables such as hormonal influence, or differences in exercise or occupational behavior before and after distraction were not available. Therefore, further research is needed to determine the cause of the gender difference in outcome of ankle distraction.

Characteristics of the patients, such as age, BMI, clinical condition, pain, and functional disability at baseline did not predict the clinical outcome of ankle distraction consistently and predictive ability was low. Patient characteristics that predict the chance of failure differ from those that predict the clinical outcome 2 years after joint distraction. Gender appears the best predictor of failure, whereas, joint motion permitted distraction is the most important predictor of improved functional ability score and decrease in pain score at 2 years after joint distraction. The randomized controlled trial from Saltzman et al.[9] already showed that joint motion permitted distraction results in a larger improvement of functional ability and decrease in pain score than fixed distraction. The present study shows that also when distraction with motion was compared to the whole group including patients from the open prospective study, this finding was consolidated. This argues for the use of hinged distraction instead of fixed distraction.

In conclusion, joint distraction in severe OA shows a prolonged clinical beneficial outcome. However, over the years failure rate becomes considerable, with the highest chance of failure in women. For those with sustained clinical benefit, hinged distraction favors this clinical benefit.

Ancillary