Background: Ovarian cancer encompasses a diverse range of neoplasms originating in the ovaries, fallopian tubes, and peritoneum. Despite being one of the commonest gynaecological malignancies, there are no validated screening strategies for early detection. A diagnosis typically relies on imaging, biomarkers, and multidisciplinary team discussions. The accurate interpretation of CTs and MRIs may be challenging, especially in borderline cases. This study proposes a methodological pipeline to develop and evaluate deep learning (DL) models that can assist in classifying ovarian masses from CT and MRI data, potentially improving diagnostic confidence and patient outcomes. Methods: A multi-institutional retrospective dataset was compiled, supplemented by external data from the Cancer Genome Atlas. Two classification workflows were examined: (1) whole-volume input and (2) lesion-focused region of interest. Multiple DL architectures, including ResNet, DenseNet, transformer-based UNeST, and Attention Multiple-Instance Learning (MIL), were implemented within the PyTorch-based MONAI framework. The class imbalance was mitigated using focal loss, oversampling, and dynamic class weighting. The hyperparameters were optimised with Optuna, and balanced accuracy was the primary metric. Results: For a preliminary dataset, the proposed framework demonstrated feasibility for the multi-class classification of ovarian masses. The initial experiments highlighted the potential of transformers and MIL for identifying the relevant imaging features. Conclusions: A reproducible methodological pipeline for DL-based ovarian mass classification using CT and MRI scans has been established. Future work will leverage a multi-institutional dataset to refine these models, aiming to enhance clinical workflows and improve patient outcomes.

The open-label, single-arm, multicentre ORZORA trial (NCT02476968) evaluated maintenance olaparib in patients with platinum-sensitive relapsed ovarian cancer (PSR OC) with a germline (g) or somatic (s) BRCA1 and/or BRCA2 mutation (BRCAm) or a non-BRCA homologous recombination repair mutation (non-BRCA HRRm). Patients were in response to platinum-based chemotherapy after ≥2 prior lines of treatment and underwent prospective central screening for tumour BRCA status, then central gBRCAm testing to determine sBRCAm or gBRCAm status. An exploratory cohort evaluated non-BRCA HRRm in 13 predefined genes. Patients received olaparib 400 mg (capsules) twice daily until investigator-assessed disease progression. Secondary endpoints included overall survival (OS) and safety. 177 patients received olaparib. At the final data cutoff (25 June 2021), median OS from study enrolment was 46.8 (95% confidence interval [CI] 37.9–54.4), 43.2 (31.7–NC [not calculated]), 47.4 (37.9–NC) and 44.9 (28.9–NC) months in the BRCAm, sBRCAm, gBRCAm and non-BRCA HRRm cohorts, respectively. No new safety signals were identified. Maintenance olaparib showed consistent clinical activity in the BRCAm and sBRCAm cohorts; exploratory analysis suggested similar activity in the non-BRCA HRRm cohort. These findings highlight that patients with PSR OC, beyond those with gBRCAm, may benefit from maintenance olaparib.

DigiONE is a pilot European learning health system in precision oncology that aims to identify optimal cancer treatments by learning from every patient, not just those in trials, through privacy-preserving interrogation of their standardized routine electronic health records.

Assessing the long-term impact of cancer on people’s lives is challenging due to confounding issues such as aging and comorbidities. We aimed to investigate this impact by comparing the outcomes of cancer survivors with a matched control cohort. This was a cross-sectional survey of breast, colorectal and ovarian cancer survivors approximately 5 years post-diagnosis and a cohort of age, sex and social deprivation-matched controls who had never had a cancer diagnosis. Eligible participants were invited by post to complete a survey assessing quality of life (QoL), health, identity, healthcare usage and finances. A total of 2075 out of 5734 (36.2

5519 Background: In the SOLO2 (NCT01874353) trial, maintenance olaparib provided clinically meaningful improvement in OS for PSROC pts with a gBRCA mutation (m) compared with placebo (median 51.7 vs 38.8 months [mo], respectively). The ORZORA trial (NCT02476968) assessed efficacy and safety of maintenance olaparib in PSROC pts with a BRCAm (s or g) or a non-BRCA HRRm. Median progression-free survival (18.0 mo, BRCAm; 16.4, non-BRCA HRRm) was reported at primary data cutoff (DCO). We report final OS analyses. Methods: We conducted an open-label, single-arm, multicenter study of PSROC pts in response to platinum-based chemotherapy (PBC) after ≥2 prior lines of PBC. Pts underwent prospective central screening for tumor BRCAm status (myChoice CDx, Myriad Genetic Laboratories, Inc.), then central gBRCAm testing (BRACAnalysis CDx, Myriad Genetic Laboratories, Inc.) to determine s or g status. An exploratory cohort comprised of pts with predefined non-BRCA HRRm (FoundationOne CDx, Foundation Medicine, Inc.). Pts received maintenance olaparib (400 mg bid; capsules) until progression. OS and time to second progression (PFS2) were secondary endpoints. Results: 181 pts were enrolled (BRCAm n = 145 [s, n = 55; g, n = 87; s/g status unknown, n = 3]; non-BRCA HRRm, n = 33; unassigned, n = 3). At DCO (June 25, 2021), median OS follow-up in censored pts was 42.6 mo in BRCAm and 39.3 mo in non-BRCA HRRm pts. OS and PFS2 are reported in the Table. PBC was received as a subsequent therapy by 33.1% BRCAm, 32.7% sBRCAm, 33.3% gBRCAm, and 45.5% non-BRCA HRRm pts. 177 pts received ≥1 dose of olaparib and were included in safety analyses; 6.2% of pts discontinued because of adverse events (AEs). 37.9% of pts reported grade ≥3 AEs, the most common being anemia (16.4%). Since primary DCO, one new primary malignancy and four myelodysplastic syndrome events occurred. Conclusions: In final OS analyses, maintenance olaparib capsules showed consistent clinical activity in BRCAm and sBRCAm PSROC pts. Exploratory analyses suggest similar activity in non-BRCA HRRm pts. No new safety signals were observed. Findings highlight that PSROC pts, beyond those with a gBRCAm, can benefit from maintenance olaparib. Clinical trial information: NCT02476968. [Table: see text]

Supplementary Figure S1 and supplementary tables

Weakly-supervised classification of histopathology slides is a computationally intensive task, with a typical whole slide image (WSI) containing billions of pixels to process. We propose Discriminative Region Active Sampling for Multiple Instance Learning (DRAS-MIL), a computationally efficient slide classification method using attention scores to focus sampling on highly discriminative regions. We apply this to the diagnosis of ovarian cancer histological subtypes, which is an essential part of the patient care pathway as different subtypes have different genetic and molecular profiles, treatment options, and patient outcomes. We use a dataset of 714 WSIs acquired from 147 epithelial ovarian cancer patients at Leeds Teaching Hospitals NHS Trust to distinguish the most common subtype, high-grade serous carcinoma, from the other four subtypes (low-grade serous, endometrioid, clear cell, and mucinous carcinomas) combined. We demonstrate that DRAS-MIL can achieve similar classification performance to exhaustive slide analysis, with a 3-fold cross-validated AUC of 0.8679 compared to 0.8781 with standard attention-based MIL classification. Our approach uses at most 18% as much memory as the standard approach, while taking 33% of the time when evaluating on a GPU and only 14% on a CPU alone. Reducing prediction time and memory requirements may benefit clinical deployment and the democratisation of AI, reducing the extent to which computational hardware limits end-user adoption.

IntroductionMuch drug development and published analysis for epithelial ovarian cancer (EOC) focuses on early-line treatment. Full sequences of treatment from diagnosis to death and the impact of later lines of therapy are rarely studied. We describe the establishment of an international network of cancer centers configured to compare real-world treatment pathways in UK, Portugal, Germany, South Korea, France and Romania (the Ovarian Real-World International Consortium; ORWIC).Methods3344 patients diagnosed with EOC (2012-2018) were analysed using a common data model and hub and spoke programming approach applied to existing electronic medical records. Consistent definition of line of therapy between sites and an efficient approach to analysis within the limitations of local information governance was achieved.ResultsMedian age of participants was 53-67 years old and 5-29% were ECOG >1. Between 62% and 84% of patients were diagnosed with late-stage disease (FIGO III-IV). Sites treating younger and fitter patients had higher rates of debulking surgery for those diagnosed at late stage than sites with older, more frail patients. At least 21% of patients treated with systemic anti-cancer therapy (SACT) had recurrent disease following second-line therapy (2L); up to 11 lines of SACT treatment were recorded for some patients. Platinum-based SACT was consistently used across sites at 1L, but choices at 2L varied, with hormone therapies commonly used in the UK and Portugal. The use (and type) of maintenance therapy following 1L also varied. Beyond 2L, there was little consensus between sites on treatment choice: trial compounds and unspecified combinations of other agents were common.DiscussionSpecific treatment sequences are reported up to 4L and the establishment of this network facilitates future analysis of comparative outcomes per line of treatment with the aim of optimizing available options for patients with recurrent EOC. In particular, this real-world network can be used to assess the growing use of PARP inhibitors. The real-world optimization of advanced line treatment will be especially important for patients not usually eligible for involvement with clinical trials. The resources to enable this analysis to be implemented elsewhere are supplied and the network will seek to grow in coverage of further sites.

Introduction/Background Artificial intelligence (AI) approaches applied to digital pathology have shown promise in supporting morphological differentiation of ovarian carcinoma subtypes from resection specimen whole slide images (WSIs). However, no existing studies have compared the use of WSIs from primary versus interval debulking surgery (IDS), a clinically relevant parameter given that subtyping is not routinely performed for post-neoadjuvant chemotherapy cases, although their inclusion would help meet the demand for data-intensive modern AI approaches. This study applies an AI-based analysis to determine the appropriateness of including both of these specimen types. Methodology We used a standard supervised classification technique (attention-based multiple instance learning) to classify the five commonest ovarian carcinoma subtypes. This was applied to compare performance on an independent test set of primary resections (100 WSIs, 30 patients), following training with a dataset comprising primary resections alone and a second dataset with the addition of IDS resections (1415 WSIs; 963 primary resections, 452 IDS from 338 patients; 201 primary resections, 137 IDS). Training and test data were from 368 patients with ovarian malignancies managed at Leeds Teaching Hospitals NHS Trust. Results The held-out test set of primary resection specimen WSIs showed a better classification performance by training with additional IDS specimens compared to primary resections alone. Accuracy was improved from 63.00% to 74.98% by using the combined training dataset, with an area under the curve increasing from 0.8370 to 0.9311. Conclusion The addition of post-neoadjuvant therapy IDS specimens to training datasets for classifying ovarian carcinoma subtypes is both appropriate and unlikely to reduce the accuracy of model performance, whilst increasing the amount of image training data available. The present model was trained with a single set of hyperparameters, and the extent of the benefit seen by including IDS specimens may vary in different scenarios, which will be the focus of our future work. Disclosures The authors declare no conflict of interest.

This study evaluates the quality of published research using artificial intelligence (AI) for ovarian cancer diagnosis or prognosis using histopathology data. A systematic search of PubMed, Scopus, Web of Science, Cochrane CENTRAL, and WHO-ICTRP was conducted up to May 19, 2023. Inclusion criteria required that AI was used for prognostic or diagnostic inferences in human ovarian cancer histopathology images. Risk of bias was assessed using PROBAST. Information about each model was tabulated and summary statistics were reported. The study was registered on PROSPERO (CRD42022334730) and PRISMA 2020 reporting guidelines were followed. Searches identified 1573 records, of which 45 were eligible for inclusion. These studies contained 80 models of interest, including 37 diagnostic models, 22 prognostic models, and 21 other diagnostically relevant models. Common tasks included treatment response prediction (11/80), malignancy status classification (10/80), stain quantification (9/80), and histological subtyping (7/80). Models were developed using 1-1375 histopathology slides from 1-776 ovarian cancer patients. A high or unclear risk of bias was found in all studies, most frequently due to limited analysis and incomplete reporting regarding participant recruitment. Limited research has been conducted on the application of AI to histopathology images for diagnostic or prognostic purposes in ovarian cancer, and none of the models have been demonstrated to be ready for real-world implementation. Key aspects to accelerate clinical translation include transparent and comprehensive reporting of data provenance and modelling approaches, and improved quantitative evaluation using cross-validation and external validations. This work was funded by the Engineering and Physical Sciences Research Council.

For many patients, current ovarian cancer treatments offer limited clinical benefit. For some therapies, it is not possible to predict patients' responses, potentially exposing them to the adverse effects of treatment without any therapeutic benefit. As part of the automated prediction of treatment effectiveness in ovarian cancer using histopathological images (ATEC23) challenge, we evaluated the effectiveness of deep learning to predict whether a course of treatment including the antiangiogenic drug bevacizumab could contribute to remission or prevent disease progression for at least 6 months in a set of 282 histopathology whole slide images (WSIs) from 78 ovarian cancer patients. Our approach used a pretrained Hierarchical Image Pyramid Transformer (HIPT) to extract region-level features and an attention-based multiple instance learning (ABMIL) model to aggregate features and classify whole slides. The optimal HIPT-ABMIL model had an internal balanced accuracy of 60.2% +- 2.9% and an AUC of 0.646 +- 0.033. Histopathology-specific model pretraining was found to be beneficial to classification performance, though hierarchical transformers were not, with a ResNet feature extractor achieving similar performance. Due to the dataset being small and highly heterogeneous, performance was variable across 5-fold cross-validation folds, and there were some extreme differences between validation and test set performance within folds. The model did not generalise well to tissue microarrays, with accuracy worse than random chance. It is not yet clear whether ovarian cancer WSIs contain information that can be used to accurately predict treatment response, with further validation using larger, higher-quality datasets required.

Background: Cancer is treated using multiple modalities (e.g. surgery, radiotherapy and systemic therapies) and is frequently associated with adverse events that affect treatment delivery and quality of life. Regular adverse event reporting could improve care and safety through timely detection and management. Information technology provides a feasible monitoring model, but applied research is needed. This research programme developed and evaluated an electronic system, called eRAPID, for cancer patients to remotely self-report adverse events. Objectives: The objectives were to address the following research questions: is it feasible to collect adverse event data from patients’ homes and in clinics during cancer treatment? Can eRAPID be implemented in different hospitals and treatment settings? Will oncology health-care professionals review eRAPID reports for decision-making? When added to usual care, will the eRAPID intervention (i.e. self-reporting with tailored advice) lead to clinical benefits (e.g. better adverse event control, improved patient safety and experiences)? Will eRAPID be cost-effective? Design: Five mixed-methods work packages were conducted, incorporating co-design with patients and health-care professionals: work package 1 – development and implementation of the electronic platform across hospital centres; work package 2 – development of patient-reported adverse event items and advice (systematic and scoping reviews, patient interviews, Delphi exercise); work package 3 – mapping health-care professionals and care pathways; work package 4 – feasibility pilot studies to assess patient and clinician acceptability; and work package 5 – a single-centre randomised controlled trial of systemic treatment with a full health economic assessment. Setting: The setting was three UK cancer centres (in Leeds, Manchester and Bristol). Participants: The intervention was developed and evaluated with patients and clinicians. The systemic randomised controlled trial included 508 participants who were starting treatment for breast, colorectal or gynaecological cancer and 55 health-care professionals. The radiotherapy feasibility pilot recruited 167 patients undergoing treatment for pelvic cancers. The surgical feasibility pilot included 40 gastrointestinal cancer patients. Intervention: eRAPID is an online system that allows patients to complete adverse event/symptom reports from home or hospital. The system provides immediate severity-graded advice based on clinical algorithms to guide self-management or hospital contact. Adverse event data are transferred to electronic patient records for review by clinical teams. Patients complete an online symptom report every week and whenever they experience symptoms. Main outcome measures: In systemic treatment, the primary outcome was Functional Assessment of Cancer Therapy – General, Physical Well-Being score assessed at 6, 12 and 18 weeks (primary end point). Secondary outcomes included cost-effectiveness assessed through the comparison of health-care costs and quality-adjusted life-years. Patient self-efficacy was measured (using the Self-Efficacy for Managing Chronic Diseases 6-item Scale). The radiotherapy pilot studied feasibility (recruitment and attrition rates) and selection of outcome measures. The surgical pilot examined symptom report completeness, system actions, barriers to using eRAPID and technical performance. Results: eRAPID was successfully developed and introduced across the treatments and centres. The systemic randomised controlled trial found no statistically significant effect of eRAPID on the primary end point at 18 weeks. There was a significant effect at 6 weeks (adjusted difference least square means 1.08, 95% confidence interval 0.12 to 2.05; p = 0.028) and 12 weeks (adjusted difference least square means 1.01, 95% confidence interval 0.05 to 1.98; p = 0.0395). No between-arm differences were found for admissions or calls/visits to acute oncology or chemotherapy delivery. Health economic analyses over 18 weeks indicated no statistically significant difference between the cost of the eRAPID information technology system and the cost of usual care (£12.28, 95% confidence interval –£1240.91 to £1167.69; p > 0.05). Mean differences were small, with eRAPID having a 55% probability of being cost-effective at the National Institute for Health and Care Excellence-recommended cost-effectiveness threshold of £20,000 per quality-adjusted life-year gained. Patient self-efficacy was greater in the intervention arm (0.48, 95% confidence interval 0.13 to 0.83; p = 0.0073). Qualitative interviews indicated that many participants found eRAPID useful for support and guidance. Patient adherence to adverse-event symptom reporting was good (median compliance 72.2%). In the radiotherapy pilot, high levels of consent (73.2%) and low attrition rates (10%) were observed. Patient quality-of-life outcomes indicated a potential intervention benefit in chemoradiotherapy arms. In the surgical pilot, 40 out of 91 approached patients (44%) consented. Symptom report completion rates were high. Across the studies, clinician intervention engagement was varied. Both patient and staff feedback on the value of eRAPID was positive. Limitations: The randomised controlled trial methodology led to small numbers of patients simultaneously using the intervention, thus reducing overall clinician exposure to and engagement with eRAPID. Furthermore, staff saw patients across both arms, introducing a contamination bias and potentially reducing the intervention effect. The health economic results were limited by numbers of missing data (e.g. for use of resources and EuroQol-5 Dimensions). Conclusions: This research provides evidence that online symptom monitoring with inbuilt patient advice is acceptable to patients and clinical teams. Evidence of patient benefit was found, particularly during the early phases of treatment and in relation to self-efficacy. The findings will help improve the intervention and guide future trial designs. Future work: Definitive trials in radiotherapy and surgical settings are suggested. Future research during systemic treatments could study self-report online interventions to replace elements of traditional follow-up care in the curative setting. Further research during modern targeted treatments (e.g. immunotherapy and small-molecule oral therapy) and in metastatic disease is recommended. Trial registration: The systemic randomised controlled trial is registered as ISRCTN88520246. The radiotherapy trial is registered as ClinicalTrials.gov NCT02747264. Funding: This project was funded by the National Institute for Health Research (NIHR) Programme Grants for Applied Research programme and will be published in full in Programme Grants for Applied Research; Vol. 10, No. 1. See the NIHR Journals Library website for further project information.

Since the emergence of COVID-19, deep learning models have been developed to identify COVID-19 from chest X-rays. With little to no direct access to hospital data, the AI community relies heavily on public data comprising numerous data sources. Model performance results have been exceptional when training and testing on open-source data, surpassing the reported capabilities of AI in pneumonia-detection prior to the COVID-19 outbreak. In this study impactful models are trained on a widely used open-source data and tested on an external test set and a hospital dataset, for the task of classifying chest X-rays into one of three classes: COVID-19, non-COVID pneumonia and no-pneumonia. Classification performance of the models investigated is evaluated through ROC curves, confusion matrices and standard classification metrics. Explainability modules are implemented to explore the image features most important to classification. Data analysis and model evalutions show that the popular open-source dataset COVIDx is not representative of the real clinical problem and that results from testing on this are inflated. Dependence on open-source data can leave models vulnerable to bias and confounding variables, requiring careful analysis to develop clinically useful/viable AI tools for COVID-19 detection in chest X-rays.

Most data collected by hospitals as a consequence of the delivery of routine care is not utilised for analytics or organisational intelligence. This project aims to develop tools to enhance the utilisation of routinely collected cancer data within hospitals across England. This was achieved by developing a web application using open source tools to provide health care professionals and hospital managers with easy to use, interactive analytics for cancer data. The application uses data items hospitals in England are mandated to collect as part of the Cancer Outcomes and Services Dataset (COSD), to provide clinical insight into survival outcomes, population distributions, service demands, waiting times, geographical case distributions and treatment information in real-time or near real-time. Development was guided by end user needs through the use of panels of clinical and non-clinical end users.

ObjectivesTo develop and validate tests to assess the risk of any cancer for patients referred to the NHS Urgent Suspected Cancer (2-week wait, 2WW) clinical pathways.SettingPrimary and secondary care, one participating regional centre.ParticipantsRetrospective analysis of data from 371 799 consecutive 2WW referrals in the Leeds region from 2011 to 2019. The development cohort was composed of 224 669 consecutive patients with an urgent suspected cancer referral in Leeds between January 2011 and December 2016. The diagnostic algorithms developed were then externally validated on a similar consecutive sample of 147 130 patients (between January 2017 and December 2019). All such patients over the age of 18 with a minimum set of blood counts and biochemistry measurements available were included in the cohort.Primary and secondary outcome measuressensitivity, specificity, negative predictive value, positive predictive value, Receiver Operating Characteristic (ROC) curve Area Under Curve (AUC), calibration curvesResultsWe present results for two clinical use-cases. In use-case 1, the algorithms identify 20% of patients who do not have cancer and may not need an urgent 2WW referral. In use-case 2, they identify 90% of cancer cases with a high probability of cancer that could be prioritised for review.ConclusionsCombining a panel of widely available blood markers produces effective blood tests for cancer for NHS 2WW patients. The tests are affordable, and can be deployed rapidly to any NHS pathology laboratory with no additional hardware requirements.

Introduction More people are living with and beyond a cancer diagnosis. There is limited understanding of the long-term effects of cancer and cancer treatment on quality of life and personal and household finances when compared to people without cancer. In a separate protocol we have proposed to link de-identified data from electronic primary care and hospital records for a large population of cancer survivors and matched controls. In this current protocol, we propose the linkage of Patient Reported Outcomes Measures data to the above data for a subset of this population. The aim of this study is to investigate the full impact of living with and beyond a cancer diagnosis compared to age and gender matched controls. A secondary aim is to test the feasibility of the collection of Patient Reported Outcomes Measures (PROMS) data and the linkage procedures of the PROMs data to electronic health records data. Materials and methods This is a cross-sectional study, aiming to recruit participants treated at the Leeds Teaching Hospitals National Health Service Trust. Eligible patients will be cancer survivors at around 5 years post-diagnosis (breast, colorectal and ovarian cancer) and non-cancer patient matched controls attending dermatology out-patient clinics. They will be identified by running a query on the Leeds Teaching Hospitals Trust patient records system. Approximately 6000 patients (2000 cases and 4000 controls) will be invited to participate via post. Participants will be invited to complete PROMs assessing factors such as quality of life and finances, which can be completed on paper or online (surveys includes established instruments, and bespoke instruments (demographics, financial costs). This PROMs data will then be linked to routinely collected de-identified data from patient’s electronic primary care and hospital records. Discussion This innovative work aims to create a truly ‘comprehensive patient record’ to provide a broad picture of what happens to cancer patients across their cancer pathway, and the long-term impact of cancer treatment. Comparisons can be made between the cases and controls, to identify the aspects of life that has had the greatest impact following a cancer diagnosis. The feasibility of linking PROMs data to electronic health records can also be assessed. This work can inform future support offered to people living with and beyond a cancer diagnosis, clinical practice, and future research methodologies.

In healthcare, process mining has been used in many case studies to discover and analyse process models of patient treatments. Process mining is generally applied to analyse the event log of patient treatments as extracted from the Electronic Health Record (EHR). In this study, we proposed an approach to combine the event log of patient treatments with the clinical user access log of the hospital information system to analyse system usage during patient treatments. Our case study combined an event log of breast cancer patients receiving chemotherapy treatments in the Leeds Cancer Centre with the user access log in the hospital information system. The event log of patient records during chemotherapy was extracted from the EHR system. The clinical user access log was extracted from the Splunk web-based log management system in the hospital. Combining records from those two logs has been useful to provide information on system usage during patient treatment. Our experiment focused on the GPTab, a functionality that allows clinicians during consultations to check on patient records on their GP visits. We applied both statistical and clinical evaluations to ensure that the findings are statistically correct and clinically meaningful. We captured the phenomena of the decreasing number of patients on the subsequent cycles of chemotherapy and when GPTab has been used during the course of chemotherapy. This approach is potentially useful for general cases to analyse system usage during process execution and can be applied to investigate the effects of system changes to process executions.

The main challenge in the pathway analysis of cancer treatments is the complexity of the process. Process mining is one of the approaches that can be used to visualize and analyze these complex pathways. In this study, our purpose was to use process mining to explore variations in the treatment pathways of endometrial cancer. We extracted patient data from a hospital information system, created the process model, and analyzed the variations of the 62-day pathway from a General Practitioner referral to the first treatment in the hospital. We also analyzed the variations based on three different criteria: the type of the first treatment, the age at diagnosis, and the year of diagnosis. This approach should be of interest to others dealing with complex medical and healthcare processes.

e13554 Background: The value of real-world evidence derived from the care of patients managed outside the context of clinical trials is well recognised. However, the ability to link data from multiple centres, especially those from different countries, is complicated by complex legal and information governance differences. The Oncology Evidence Network is a collaboration of large hospital centres, with strong clinical informatics capabilities in six countries in Europe and Asia working with the support of an industrial partner to provide high quality, real world data reflecting routine clinical care. We have developed an efficient workflow based on a study-specific common data model (CDM) clinically validated at each site and analysed with a single analysis script, which embeds a set of data quality rules. Local implementation allows each centre to generate analytical outputs aligned across the different sites without the need for any patient level data to leave the participating site. This approach has been designed and tested in Epithelial Ovarian Cancer (EOC) patients. Methods: A CDM was agreed using expert advisors from each centre. Clinical alignment was achieved through iterative assessment of clinical vignettes, to ensure common definitions of clinical assessment, prognosis, and treatment algorithms in EOC patients. A data guide detailing variable level derivations and validation rules, general data coding principles, and conversions/codes from international coding systems was developed. The analysis scripts were implemented as a bespoke package (OpenOvary) in R. The package includes functions to validate the data against the CDM, and generate a standard output including tables, numerical summaries and Kaplan-Meier analysis of progression and overall survival. Results: 2,925 patient records from 6 centres across 6 countries were included in the study with 27 key data items curated by each centre. Treatment data is available detailing relevant surgical procedures and their outcomes, and regimens of SACT throughout patients’ care from diagnosis to death. Data completeness was generally high for key data items, with missing data ranging from 0-16% for FIGO stage at diagnosis and 0-14% for tumour morphology. The CDM and R script will be made publicly available for other centres to adopt and facilitate analysis of their local data. Conclusions: This collaboration has brought together a substantial body of data describing the care and outcomes for EOC patients. A CDM and flexible shared analysis approach enabled unified analysis and reporting whilst avoiding the transfer of patient level data and its pooling into a common database. The process of clinical and data alignment has generated a replicable model for rapid extension to other study centres to join the EOC study, or application to other disease areas.