Development and validation of a machine learning-supported strategy of patient selection for osteoarthritis clinical trials: the IMI-APPROACH study

Objectives

To efficiently assess the disease-modifying potential of new osteoarthritis treatments, clinical trials need progression-enriched patient populations. To assess whether the application of machine learning results in patient selection enrichment, we developed a machine learning recruitment strategy targeting progressive patients and validated it in the IMI-APPROACH knee osteoarthritis prospective study.

Design

We designed a two-stage recruitment process supported by machine learning models trained to rank candidates by the likelihood of progression. First stage models used data from pre-existing cohorts to select patients for a screening visit. The second stage model used screening data to inform the final inclusion. The effectiveness of this process was evaluated using the actual 24-month progression.

Results

From 3500 candidate patients, 433 with knee osteoarthritis were screened, 297 were enrolled, and 247 completed the 2-year follow-up visit. We observed progression related to pain (P, 30%), structure (S, 13%), and combined pain and structure (P ​+ ​S, 5%), and a proportion of non-progressors (N, 52%) ∼15% lower vs an unenriched population. Our model predicted these outcomes with AUC of 0.86 [95% CI, 0.81-0.90] for pain-related progression and AUC of 0.61 [95% CI, 0.52-0.70] for structure-related progression. Progressors were ranked higher than non-progressors for P ​+ ​S (median rank 65 vs 143, AUC = 0.75), P (median rank 77 vs 143, AUC = 0.71), and S patients (median rank 107 vs 143, AUC = 0.57).

Conclusions

The machine learning-supported recruitment resulted in enriched selection of progressive patients. Further research is needed to improve structural progression prediction and assess this strategy in an interventional trial.

Proteomics to predict the response to tumour necrosis factor-α inhibitors in rheumatoid arthritis using a supervised cluster-analysis based protein score

OBJECTIVE

In rheumatoid arthritis (RA), it is of major importance to identify non-responders to tumour necrosis factor-α inhibitors (TNFi) before starting treatment, to prevent a delay in effective treatment. We developed a protein score for the response to TNFi treatment in RA and investigated its predictive value.

METHOD

In RA patients eligible for biological treatment included in the BiOCURA registry, 53 inflammatory proteins were measured using xMAP® technology. A supervised cluster analysis method, partial least squares (PLS), was used to select the best combination of proteins. Using logistic regression, a predictive model containing readily available clinical parameters was developed and the potential of this model with and without the protein score to predict European League Against Rheumatism (EULAR) response was assessed using the area under the receiving operating characteristics curve (AUC-ROC) and the net reclassification index (NRI).

RESULTS

For the development step (n = 65 patient), PLS revealed 12 important proteins: CCL3 (macrophage inflammatory protein, MIP1a), CCL17 (thymus and activation-regulated chemokine), CCL19 (MIP3b), CCL22 (macrophage-derived chemokine), interleukin-4 (IL-4), IL-6, IL-7, IL-15, soluble cluster of differentiation 14 (sCD14), sCD74 (macrophage migration inhibitory factor), soluble IL-1 receptor I, and soluble tumour necrosis factor receptor II. The protein score scarcely improved the AUC-ROC (0.72 to 0.77) and the ability to improve classification and reclassification (NRI = 0.05). In validation (n = 185), the model including protein score did not improve the AUC-ROC (0.71 to 0.67) or the reclassification (NRI = -0.11).

CONCLUSION

No proteomic predictors were identified that were more suitable than clinical parameters in distinguishing TNFi non-responders from responders before the start of treatment. As the results of previous studies and this study are disparate, we currently have no proteomic predictors for the response to TNFi.

The canine 'groove' model of osteoarthritis is more than simply the expression of surgically applied damage

OBJECTIVE

Recently a new canine model of osteoarthritis (OA; the 'groove' model) has been described. This model is based on surgically applied mechanical damage of the articular cartilage followed by transient forced loading of the affected joint. Ten weeks after surgery this model shows characteristics of OA, mimicking human OA. To establish whether the observed characteristics of degeneration in this model represent the surgically applied damage, or are the results of progressive features of OA, we evaluated this 'groove' model shortly after surgery.

METHODS

In 20 female Beagle dogs, articular cartilage of the weight-bearing areas of the femoral condyles in the right knee was damaged without affecting the underlying bone. After surgery dogs were let out on a patio 5 days/week for 4 h/day. The dogs were forced to load the experimental joint by fixing the contralateral control limb to the trunk 3 days/week. The severity of OA was evaluated at 3 (n = 10) or 10 weeks (n = 10) after surgery. Synovial inflammation, cartilage damage and cartilage matrix turnover were determined.

RESULTS

Ten weeks after surgery osteoarthritic features were found, as described previously. Proteoglycan (PG) synthesis, percentage release of newly formed PG, and that of total amount of PG were enhanced, whereas PG content was significantly diminished (all P < 0.05). Importantly, 3 weeks after surgery these characteristics of OA were not yet evident.

CONCLUSIONS

The present results clearly show that the characteristics observed 10 weeks after induction of joint degeneration in the groove model are not just the expression of the surgically applied damage but are the result of progressive features of (experimental) OA.

Inhibition of COX-2 by celecoxib in the canine groove model of osteoarthritis

OBJECTIVE

In vitro studies showed a beneficial effect of celecoxib on proteoglycan turnover and content of osteoarthritic cartilage. In the present study we evaluated whether these favourable effects of celecoxib could also be demonstrated in vivo.

METHODS

In 24 Beagle dogs, osteoarthritis (OA) was induced in one knee according to the groove model. The animals were divided into three groups and received oral placebo or 100 or 200 mg celecoxib daily, starting directly after surgery. After 15 weeks joint tissue from all dogs was analysed.

RESULTS

Induction of OA resulted in macroscopic and histological damage of cartilage, changes in cartilage proteoglycan turnover, loss of cartilage matrix proteoglycans and slight synovial inflammation, all characteristic of early OA. Surprisingly, none of the parameters was significantly changed upon celecoxib treatment. Synovial fluid prostaglandin E(2) levels were dose-dependently diminished by celecoxib, demonstrating that the celecoxib had reached the joint in sufficient amounts. Using an in vitro setup, canine cartilage under degenerative conditions was favourably influenced by celecoxib, demonstrating that canine cartilage is sensitive to celecoxib.

CONCLUSION

The present study showed a chondroneutral effect of celecoxib on the characteristics of experimentally induced OA in vivo, in contrast to the observed beneficial effect in vitro. It could be that celecoxib had been beneficial to degenerated cartilage in vivo but that these effects were counteracted by increased loading of the affected joint and the associated progression of OA, occurring because of the well-known analgesic effects of celecoxib.

Ankle images digital analysis (AIDA): digital measurement of joint space width and subchondral sclerosis on standard radiographs

OBJECTIVE

Reliable evaluation of joint space width and subchondral sclerosis of osteoarthritic joints is difficult. The present study describes a new digital method to analyse standard radiographs of the ankle.

DESIGN

Standardized radiographs were taken of the ankle of 12 patients with severe osteoarthritis (OA) under full weight-bearing conditions, before treatment and 1 year after initiation of treatment. Treatment consisted of 3 months distraction of the tibio-talar joint, for which clinical benefit has been shown previously. The width of the joint space was measured on digitized images of the radiographs by means of the newly developed semi-automatic digital technique called AIDA (Ankle Images Digital Analysis) and by means of the most widely used conventional analogue measurements. In addition, AIDA was used to assess subchondral sclerosis by measuring the intensity of the radiograph at fixed positions at the bone-cartilage interface.

RESULTS

AIDA appeared to be a reliable method for measuring small changes in joint space width and subchondral sclerosis because the intra- and interobserver variation was small. Mean JSW for two observers was 1.96 and 2.00 mm, with mean differences between two observations of 0.05 and -0.01, respectively. Mean subchondral sclerosis in the tibia was 1.52 and 1.61 with mean differences between two observations of, respectively, 0.00 and 0.03. In addition to conventional measurements, AIDA could demonstrate a decrease in subchondral sclerosis as a result of joint distraction of 71% and 69% after 1 year for talus and tibia, respectively.

CONCLUSION

The use of AIDA is preferable to the conventional analogue method for evaluating the severity of ankle OA, because the method provides quantitative data not only for the joint space width but also for subchondral sclerosis.

Joint distraction in treatment of osteoarthritis (II): effects on cartilage in a canine model

OBJECTIVE

From a clinical point of view, joint distraction as a treatment for osteoarthritis (OA) of hip and ankle has been demonstrated to be very promising. Pain, joint mobility and functional ability, the most important factors for a patient with severe OA, all improved. Although radiographic joint space enlargement in a significant number of patients suggested cartilage repair, actual cartilage repair remains difficult to evaluate. Therefore the present study was initiated to evaluate the actual effects of joint distraction on cartilage.

METHODS

For this purpose a canine model for OA, anterior cruciate ligament transection (ACLT) was used. Sixteen weeks after ACLT articulating Ilizarov joint distraction of the knee was carried out. Absence of mechanical contact between articular surfaces and presence of intra-articular intermittent fluid pressure, characteristics of Ilizarov joint distraction, were confirmed. Twenty-five weeks after ACLT joint tissue of the dogs was analyzed.

RESULTS

Biochemical analysis showed that after joint distraction the abnormal cartilage proteoglycan (PG) metabolism, characteristic for OA, had changed to a level found in control joints. Moreover, a mild degree of inflammation, present after ACLT, was reduced upon joint distraction. PG-content and histological cartilage degeneration had not (yet) improved within the time of treatment.

DISCUSSION

Results suggest that the promising clinical results of Ilizarov joint distraction in patients with OA are accompanied by changes in cartilage metabolism. A change in proteoglycan turnover, indicating normalization of overall chondrocyte function, might in the long term, with normal joint use, lead to actual repair of cartilage.

Joint distraction in treatment of osteoarthritis: a two-year follow-up of the ankle

OBJECTIVE

Osteoarthritis (OA) is a degenerative joint disease with an incidence exceeding 10% of the adult population. In end stages, OA can result in severe restriction of activity and consequent disability. For these severe cases of OA, no effective remedy is available yet. Joint distraction is a new experimental approach in treatment of OA. Studied in retrospect it appeared that this treatment results in prolonged clinical improvement in the case of ankle OA. Presently the results of a prospective two-year follow-up are given.

METHODS

Patients with severe ankle OA, who were considered for joint fusion (arthrodesis), were treated with distraction, by use of an Ilizarov external ring fixator. Distraction was carried out for 3 months during which full weight bearing (walking) was allowed. Standardized clinical examination was conducted before and at yearly intervals after treatment, including physical examination, functional ability questionnaire, pain scale, joint mobility, and radiographic evaluation.

RESULTS

More than two thirds of the patients improved significantly as shown by physical examination, functional ability questionnaires and pain scale; effects were progressive in the second year of follow-up. On average, joint mobility and radiographic joint space were preserved, whilst improvement was observed in a significant number of patients.

DISCUSSION

Results of the present prospective study confirm the findings of the previous retrospective study and suggest that Ilizarov joint distraction is a promising treatment for severe ankle OA, at least delaying the need for a joint fusion. Considering the high prevalence of OA and the lack of a remedy for this disorder, Ilizarov joint distraction as a treatment for OA may have great medical, social and economic impact.