Given the low prevalence of sarcoma, international cooperation is necessary to recruit sufficient numbers of patients for surgical trials. The PARITY (Prophylactic Antibiotic Regimens in Tumour Surgery) trial is the first international multicentre trial in orthopedic oncology and successfully achieved target enrollment of 600 patients across 12 countries. It is important to reflect upon the challenges encountered and experiences gained to inform future trials. The objective of this study is to describe recruitment patterns and examine the differences in enrollment across different PARITY sites and identify variables associated with varying levels of recruitment.

Data from this study was obtained from the PARITY trial Methods Centre and correspondence data. We performed descriptive statistics to demonstrate the recruitment patterns over time. We compared recruitment, time to set up, and time to enroll the first patient between North American and international sites, and sites that had dedicated research personnel. Two-tailed non-paired t-tests were performed to compare average monthly recruitment rates between groups with significance being set at alpha=0.05.

A total of 600 patients from 48 clinical sites and 12 countries were recruited from January 2013 through to October 2019. Average monthly enrollment increased every year of the study. There were 36 North American and 12 international sites. North American sites were able to set up significantly faster than international sites (19.3 vs. 28.3 months p=0.037). However, international sites had a significantly higher recruitment rate per month once active (0.2/month vs. 0.62/month, p=0.018). Of active sites, 40 (83%) had research support personnel and 8 (17%) sites did not. Sites with research personnel were able to reach ‘enrolment ready’ status significantly faster than sites without research support (19.6 vs. 30.7 months, p=0.032). However, there was no significant difference in recruitment rate per month once the sites began enrolling (0.28/month vs. 0.2/month, p=0.63). Trial sites that took longer than 1 year to recruit their first patient had 3x lower average recruitment rate compared to sites that were able to recruit their first patient within a year of being enrolment ready.

The PARITY trial is the first multicentre RCT in orthopaedic oncology. The PARITY investigators were able to increase the recruitment levels throughout the trial and generally avoid trial fatigue. This was a North American based trial which may explain the longer start up times internationally given the different regulatory bodies associated with drug-related trials. However, international sites should be considered critical as they were able to recruit significantly more patients per month once active. The absence of research support personnel should not preclude a site from inclusion. These sites took longer to setup but had no difference in monthly recruitment once active. This study will create a framework for identifying and targeting high yield sites for future randomized control trials within orthopaedic oncology to maximum recruitment and resource allocation. Data quality is another consideration that will be addressed in future analyses of the PARITY trial.

Lung cancer is the most common cancer diagnosed, the leading cause of cancer-related deaths, and bone metastases occurs in 20–40% of lung cancer patients. They often present symptomatically with pain or skeletal related events (SREs), which are independently associated with decreased survival. Bone modifying agents (BMAs) such as Denosumab or bisphosphonates are routinely used, however no specific guidelines exist from the National Comprehensive Cancer Center or the European Society of Medical Oncologists. Perhaps preventing the formation of guidelines is the lack of a high-quality quantitative synthesis of randomized controlled trial (RCT) data to determine the optimal treatment for the patient important outcomes of 1) Overall survival (OS), 2) Time to SRE, 3) SRE incidence, and 4) Pain Resolution. The objective of this study was to perform the first systematic review and network meta-analysis (NMA) to assess the best BMA for treatment of metastatic lung cancer to bone.

We conducted our study in accordance to the PRISMA protocol. We performed a librarian assisted search of MEDLINE, PubMed, EMBASE, and Cochrane Library and Chinese databases including CNKI and Wanfang Data. We included studies that are RCTs reporting outcomes specifically for lung cancer patients treated with a bisphosphonate or Denosumab. Screening, data extraction, risk of bias and GRADE were performed in duplicate. The NMA was performed using a Bayesian probability model with R. Results are reported as relative risks, odds ratios or mean differences, and the I2 value is reported for heterogeneity. We assessed all included articles for risk of bias and applied the novel GRADE framework for NMAs to rate the quality of evidence supporting each outcome.

We included 132 RCTs comprising 11,161 patients with skeletal metastases from lung cancer. For OS, denosumab was ranked above zoledronic acid (ZA) and estimated to confer an average of 3.7 months (95%CI: −0.5 – 7.6) increased survival compared to untreated patients. For time to SRE, denosumab was ranked first with an average of 9.1 additional SRE-free months (95%CI: 4.0 – 14.0) compared to untreated patients, while ZA conferred an additional 4.8 SRE-free months (2.4 – 7.0). Patients treated with the combination of Ibandronate and systemic therapy were 2.3 times (95%CI: 1.7 – 3.2) more likely to obtain successful pain resolution, compared to untreated. Meta-regression showed no effect of heterogeneity length of follow-up or pain scales on the observed treatment effects. Heterogeneity in the network was considered moderate for overall survival and time to SRE, mild for SRE incidence, and low for pain resolution. While a generally high risk of bias was observed across studies, whether they were from Western or Chinese databases. The overall GRADE for the evidence underlying our results is High for Pain control and SRE incidence, and Moderate for OS and time to SRE.

This study represents the most comprehensive synthesis of the best available evidence guiding pharmacological treatment of bone metastases from lung cancer. Denosumab is ranked above ZA for both overall survival and time to SRE, but both treatments are superior to no treatment. ZA was first among all bisphosphonates assessed for odds of reducing SRE incidence, while the combination of Ibandronate and radionuclide therapy was most effective at significantly reducing pain from metastases. Clinicians and policy makers may use this synthesis of all available RCT data as support for the use of a BMA in MBD for lung cancer.

Population-based studies from the United States have reported that sarcoma patients living in rural areas or belonging to lower socioeconomic classes experience worse overall survival; however, the evidence is not clear for universal healthcare systems where financial resources should theoretically not affect access to standard of care. The purpose of this study was to determine the survival outcomes of soft-tissue sarcoma (STS) patients treated in Ontario, Canada over 23 years and determine if the patient's geographic location or income quintile are associated with survival.

We performed a population-based cohort study using linked administrative databases of patients diagnosed with STS between 1993 – 2015. The Kaplan-Meier method was used to estimate 2, 5, 10, 15 and 20-year survival stratified by age, stage and location of tumor. We estimated survival outcomes based on the patient's geographic location and income quintile. The Log-Rank test was used to detect significant differences between groups. If groups were significantly different, a Cox proportional hazards model was used to test for interaction effects with other patient variables.

We identified 8,896 patients with biopsy-confirmed STS during the 23-year study period. Overall survival following STS diagnosis was 70% at 2 years, 59% at 5 years, 50% at 10 years, 43% at 15 years, and 38% at 20 years. Living in a rural location (p=0.0028) and belonging to the lowest income quintile (p<0.0001) were independently associated with lower overall survival following STS diagnosis. These findings were robust to tests of interaction with each other, age, gender, location of tumor and stage of disease.

This population-based cohort study of 8,896 STS patients treated in Ontario, Canada over 23 years reveals that patients living in a rural area and belonging to the lowest income quintile are at risk for decreased survival following STS diagnosis. We extend previous STS survival reporting by providing 15 and 20-year survival outcomes stratified by age, stage, and tumor location.

Advances in cancer therapy have prolonged patient survival even in the presence of disseminated disease and an increasing number of cancer patients are living with metastatic bone disease (MBD). The proximal femur is the most common long bone involved in MBD and pathologic fractures of the femur are associated with significant morbidity, mortality and loss of quality of life (QoL).

Successful prophylactic surgery for an impending fracture of the proximal femur has been shown in multiple cohort studies to result in longer survival, preserved mobility, lower transfusion rates and shorter post-operative hospital stays. However, there is currently no optimal method to predict a pathologic fracture. The most well-known tool is Mirel's criteria, established in 1989 and is limited from guiding clinical practice due to poor specificity and sensitivity. The ideal clinical decision support tool will be of the highest sensitivity and specificity, non-invasive, generalizable to all patients, and not a burden on hospital resources or the patient's time. Our research uses novel machine learning techniques to develop a model to fill this considerable gap in the treatment pathway of MBD of the femur. The goal of our study is to train a convolutional neural network (CNN) to predict fracture risk when metastatic bone disease is present in the proximal femur.

Our fracture risk prediction tool was developed by analysis of prospectively collected data of consecutive MBD patients presenting from 2009–2016. Patients with primary bone tumors, pathologic fractures at initial presentation, and hematologic malignancies were excluded. A total of 546 patients comprising 114 pathologic fractures were included. Every patient had at least one Anterior-Posterior X-ray and clinical data including patient demographics, Mirel's criteria, tumor biology, all previous radiation and chemotherapy received, multiple pain and function scores, medications and time to fracture or time to death.

We have trained a convolutional neural network (CNN) with AP X-ray images of 546 patients with metastatic bone disease of the proximal femur. The digital X-ray data is converted into a matrix representing the color information at each pixel. Our CNN contains five convolutional layers, a fully connected layers of 512 units and a final output layer. As the information passes through successive levels of the network, higher level features are abstracted from the data. The model converges on two fully connected deep neural network layers that output the risk of fracture. This prediction is compared to the true outcome, and any errors are back-propagated through the network to accordingly adjust the weights between connections, until overall prediction accuracy is optimized. Methods to improve learning included using stochastic gradient descent with a learning rate of 0.01 and a momentum rate of 0.9.

We used average classification accuracy and the average F1 score across five test sets to measure model performance. We compute F1 = 2 x (precision x recall)/(precision + recall). F1 is a measure of a model's accuracy in binary classification, in our case, whether a lesion would result in pathologic fracture or not. Our model achieved 88.2% accuracy in predicting fracture risk across five-fold cross validation testing. The F1 statistic is 0.87.

This is the first reported application of convolutional neural networks, a machine learning algorithm, to this important Orthopaedic problem. Our neural network model was able to achieve reasonable accuracy in classifying fracture risk of metastatic proximal femur lesions from analysis of X-rays and clinical information. Our future work will aim to externally validate this algorithm on an international cohort.

Advances in cancer therapy have prolonged cancer patient survival even in the presence of disseminated disease and an increasing number of cancer patients are living with metastatic bone disease (MBD). The proximal femur is the most common long bone involved in MBD and pathologic fractures of the femur are associated with significant morbidity, mortality and loss of quality of life (QoL).

Successful prophylactic surgery for an impending fracture of the proximal femur has been shown in multiple cohort studies to result in patients more likely to walk after surgery, longer survival, lower transfusion rates and shorter post-operative hospital stays. However, there is currently no optimal method to predict a pathologic fracture. The most well-known tool is Mirel's criteria, established in 1989 and is limited from guiding clinical practice due to poor specificity and sensitivity. The goal of our study is to train a convolutional neural network (CNN) to predict fracture risk when metastatic bone disease is present in the proximal femur.

Our fracture risk prediction tool was developed by analysis of prospectively collected data for MBD patients (2009–2016) in order to determine which features are most commonly associated with fracture. Patients with primary bone tumors, pathologic fractures at initial presentation, and hematologic malignancies were excluded. A total of 1146 patients comprising 224 pathologic fractures were included. Every patient had at least one Anterior-Posterior X-ray. The clinical data includes patient demographics, tumor biology, all previous radiation and chemotherapy received, multiple pain and function scores, medications and time to fracture or time to death. Each of Mirel's criteria has been further subdivided and recorded for each lesion.

We have trained a convolutional neural network (CNN) with X-ray images of 1146 patients with metastatic bone disease of the proximal femur. The digital X-ray data is converted into a matrix representing the color information at each pixel. Our CNN contains five convolutional layers, a fully connected layers of 512 units and a final output layer. As the information passes through successive levels of the network, higher level features are abstracted from the data. This model converges on two fully connected deep neural network layers that output the fracture risk. This prediction is compared to the true outcome, and any errors are back-propagated through the network to accordingly adjust the weights between connections. Methods to improve learning included using stochastic gradient descent with a learning rate of 0.01 and a momentum rate of 0.9.

We used average classification accuracy and the average F1 score across test sets to measure model performance. We compute F1 = 2 x (precision x recall)/(precision + recall). F1 is a measure of a test's accuracy in binary classification, in our case, whether a lesion would result in pathologic fracture or not. Five-fold cross validation testing of our fully trained model revealed accurate classification for 88.2% of patients with metastatic bone disease of the proximal femur. The F1 statistic is 0.87. This represents a 24% error reduction from using Mirel's criteria alone to classify the risk of fracture in this cohort.

This is the first reported application of convolutional neural networks, a machine learning algorithm, to an important Orthopaedic problem. Our neural network model was able to achieve impressive accuracy in classifying fracture risk of metastatic proximal femur lesions from analysis of X-rays and clinical information. Our future work will aim to validate this algorithm on an external cohort.

Total hip arthroplasty (THA) is one of the most successful and effective treatments for advanced hip osteoarthritis (OA). Over the last 5 years, Canada has seen a 17.8% increase in the number of hip replacements performed annually, and that number is expected to grow along with the aging Canadian population. However, the rise in THA surgery is associated with an increased number of patients at risk for the development of an infection involving the joint prosthesis and adjacent deep tissue – periprosthetic joint infections (PJI). Despite improved hygiene protocols and novel surgical strategies, PJI remains a serious complication. No previous population-based studies has investigated PJI risk factors using a time-to-event approach and none have focused exclusively on patients undergoing THA for primary hip OA. The purpose of this study is to determine risk factors for PJI after primary THA for OA using a large population-based database collected over 15 years. Our secondary objective is to determine the incidence of PJI, the time to PJI following primary THA, and if PJI rates have changed in the past 15 years.

We performed a population-based cohort study using linked administrative databases in Ontario, Canada in accordance with RECORD and STROBE guidelines. All primary total hip replacements performed for osteoarthritis in patients aged 55 or older between January 1st 2002 – December 31st 2016 in Ontario, Canada were identified. Periprosthetic joint infection as the cause for revision surgery was identified with the International Classification of Diseases, 10th Edition (ICD-10), Clinical Modification diagnosis code T84.53 in any component of the healthcare data set.

Data were obtained from the Institute for Clinical Evaluative Sciences (ICES).

Demographic data and outcomes are summarized using descriptive statistics. We used a Cox proportional hazards model to analyze the effect of surgical factors and patient factors on the risk of developing PJI. Surgical factors include the approach, use of bone graft, use of cement, and the year of surgery. Patient factors include sex, age at surgery, income quintile and rurality (community vs. urban). We compared the 1,2,5 and 10 year PJI rates for patients undergoing THA each year of our cohort with the Cochran-Armitage test. Less than 0.1% of data were missing from all fields except for rurality which was lacking 0.3% of data.

A total of 100,674 patients aged 55 or older received a primary total hip arthroplasty for osteoarthritis from 2002–2016. We identified 1034 cases of revision surgery for prosthetic joint infection for an overall PJI rate of 1.03%. When accounting for patients censored at final follow-up, the cumulative incidence for PJI is 1.44%. Our Cox proportional hazards model revealed that male sex, Type II diabetes mellitus, discharge to convalescent care, and having both hips replaced during one's lifetime were associated with increased risk of developing PJI following primary THA. Importantly, the time adjusted risk for PJI was equal for patients operated within the past 5 years, 6–10 years ago, or 11–15 years ago. The surgical approach, use of bone grafting or cement were not associated with increased risk of infection. PJI rates have not changed significantly over the past 15 years. One, two, five and ten-year PJI rates were similar for patients undergoing THA in all qualifying years.

Analysis of a population-based cohort of 100,674 patients has shown that the risk of developing PJI following primary THA has not changed over 15 years. The surgical approach, use of bone grafting or cement were not associated with increased risk of infection. Male sex, Type II diabetes Mellitus and discharge to a rehab facility are associated with increased risk of PJI. As the risk of PJI has not changed in 15 years, an appropriately powered trial is warranted to determine interventions that can improve infection rate after THA.