Clinical Oncology Data Research Articles

Towards a Bavarian Oncology Real World Data Research Platform.

In the last decade numerous real-world data networks have been established in order to leverage the value of data from electronic health records for medical research. In Germany, a nation-wide network based on electronic health record data from all German university hospitals has been established within the Medical Informatics Initiative (MII) and recently opened for researcherst' access through the German Portal for Medical Research Data (FDPG). In Bavaria, the six university hospitals have joined forces within the Bavarian Cancer Research Center (BZKF). The oncology departments aim at establishing a federated observational research network based on the framework of the MII/FDPG and extending it with a clear focus on oncological clinical data, imaging data and molecular high throughput analysis data. We describe necessary adaptions and extensions of existing MII components with the goal of establishing a Bavarian oncology real world data research platform with its first use case of performing federated feasibility queries on clinical oncology data. We share insights from developing a feasibility platform prototype and presenting it to end users. Our main discovery was that oncological data is characterized by a higher degree of interdependence and complexity compared to the MII core dataset that is already integrated into the FDPG. The significance of our work lies in the requirements we formulated for extending already existing MII components to match oncology specific data and to meet oncology researchers needs while simultaneously transferring back our results and experiences into further developments within the MII.

Exchanging genomics reports between pathology labs and medical centers using the Minimal Clinical Oncology Data Element (mCODE) FHIR implementation guide.

e13575 Background: The automated transfer of discrete molecular data from pathology labs into the electronic health record (EHR) is rapidly becoming commonplace and is designed to facilitate clinical decision making. Simple genomic changes are represented in existing EHR data transfer standards, but current data standards lack clinical fields that correspond to treatment insights such as associated medications and potential clinical trials. To address this, the mCODE effort defines a consensus oncology data standard that facilitates lightweight transmission of data from patients with cancer. mCODE’s support of genomic data leverages efforts by the HL7 Clinical Genomics Working Group (HL7 CGWG) focusing on oncology. Here, we describe pilot efforts involving both a pathology partner and academic medical centers to extend mCODE FHIR (Fast Healthcare Interoperability Resources) implementation guide (IG) support of genomic data for data interchange and clinical decision support use cases. Methods: From 2021-2022, validation of the mCODE FHIR IG genomic data elements was performed in CodeX, a FHIR accelerator program. These elements encompassed: lab facility, test order information, pathologist, patient demographics, specimen, molecular studies and therapeutic implications. Genomic FHIR data elements were selected for their conciseness and clinical relevance for inclusion in the mCODE FHIR IG under the supervision of two oncology informaticians. As part of the mCODE FHIR IG pilot, synthetic solid tumor molecular patient reports were generated by Tempus Labs. These synthetic molecular reports were transformed into mCODE-conformant FHIR files for transfer. The mCODE IG can be found at http://hl7.org/fhir/us/mcode/ . Results: Over 50 Tempus data elements were analyzed and found to be readily mapped to mCODE genomics elements. Two categories of data elements required additional transformation: 1. Data elements requiring additional genomic database annotations; 2. FHIR syntax constraints like the concatenation of medications in the same drug class. The mCODE file was transmitted to Vanderbilt and the Children's Hospital LA from Tempus for decoding and downstream uses. This effort further informed the HL7 CGWG of needed IG clarifications for compatibility with other standardization efforts and genomic use cases. Conclusions: mCODE’s support of genomic data elements provides a lightweight genomics interchange standard compatible with an existing pathology lab report. Mapping from existing molecular pathology reports to mCODE required nominal alterations. Furthermore, mCODE’s alignment with HL7 genomics efforts ensure FHIR-based compatibility in other clinical domains and with developing international standards. Continued pathology lab and academic medical center collaboration is essential for mCODE refinement and adoption.

Research Goal-Driven Data Model and Harmonization for De-Identifying Patient Data in Radiomics.

There are various efforts in de-identifying patient's radiation oncology data for their uses in the advancement of research in medicine. Though the task of de-identification needs to be defined in the context of research goals and objectives, existing systems lack the flexibility of modeling data and normalization of names of attributes for accomplishing them. In this work, we describe a de-identification process of radiation and clinical oncology data, which is guided by a data model and a schema of dynamically capturing domain ontology and normalization of terminologies, defined in tune with the research goals in this area. The radiological images are obtained in DICOM format. It consists of diagnostic, radiation therapy (RT) treatment planning, RT verification, and RT response images. During the DICOM de-identification, a few crucial pieces of information are taken about the dataset. The proposed model is generic in organizing information modeling in sync with the de-identification of a patient's clinical information. The treatment and clinical data are provided in the comma-separated values (CSV) format, which follows a predefined data structure. The de-identified data is harmonized throughout the entire process. We have presented four specific case studies on four different types of cancers, namely glioblastoma multiforme, head-neck, breast, and lung. We also present experimental validation on a few patients' data in these four areas. A few aspects are taken care of during de-identification, such as preservation of longitudinal date changes (LDC), incremental de-identification, referential data integrity between the clinical and image data, de-identified data harmonization, and transformation of the data to an underlined database schema.

Effect of first-line pembrolizumab treatment in individuals with advanced non-small cell lung cancer and poor performance status.

e18796 Background: There is limited evidence supporting the optimal use of immune checkpoint inhibitors (ICIs) in NSCLC patients with poor performance status (PS), as clinical trials exclude these patients. In this study, we use real-world oncology data to determine the impact of first line pembrolizumab vs. no treatment in high PD(L)-1 expressing cancers in individuals with advanced NSCLC and ECOG PS ≥2. Methods: We performed a retrospective cohort study of patients with advanced NSCLC with ECOG PS ≥2 between 09/01/2014 and 02/18/2020, using the nationwide Flatiron Health electronic health record (EHR)-derived de-identified database. Patients were included if they were PD(L)-1 high (≥50%) and had clinical and treatment information recorded within 90 days of diagnosis. Real-world overall survival (rwOS) was defined as time from diagnosis to death (censored at last EHR activity). Median rwOS was estimated using weighted Kaplan-Meier methods. A marginal Cox structural model with inverse probability of treatment weighting was used to adjust for selection bias and estimate the effectiveness of pembrolizumab. The inverse probability weights were estimated using an ensemble machine learning technique, Super Learner, based on age, gender, race, practice type and smoking history. Adjusted hazard ratios (aHR) were estimated using weighted Cox proportional hazards models. Stratified analysis was conducted by ECOG PS (2 vs >2). Results: 217 (16%) individuals with advanced NSCLC and high PD(L)-1 expression received no treatment, compared to 546 (39%) individuals who received 1L pembrolizumab. The no-treatment group had a lower proportion of ECOG 2 compared to the pembrolizumab group (Table). Median rwOS in the no-treatment group was 2.4 months, compared to 7.1 months in the pembrolizumab group (p<0.001). In unadjusted survival analyses in the entire cohort and in cohorts stratified by ECOG status, treatment with pembrolizumab was associated with a significantly lower risk of death (hazard ratio [HR]: 0.38, 95% Confidence Interval [CI]: 0.31-0.45). In adjusted analyses, individuals treated with pembrolizumab had improved survival (HR: 0.40, 95% CI: 0.35-0.45). Conclusions: Our analysis of real-world clinical oncology data demonstrated that 1L treatment with pembrolizumab was associated with significantly improved rwOS among individuals with ECOG ≥ 2. [Table: see text]

Structured Data Capture for Oncology.

Lack of interoperability is one of the greatest challenges facing healthcare informatics. Recent interoperability efforts have focused primarily on data transmission and generally ignore data capture standardization. Structured Data Capture (SDC) is an open-source technical framework that enables the capture and exchange of standardized and structured data in interoperable data entry forms (DEFs) at the point of care. Some of SDC’s primary use cases concern complex oncology data such as anatomic pathology, biomarkers, and clinical oncology data collection and reporting. Its interoperability goals are the preservation of semantic, contextual, and structural integrity of the captured data throughout the data’s lifespan. SDC documents are written in eXtensible Markup Language (XML) and are therefore computer readable, yet technology agnostic—SDC can be implemented by any EHR vendor or registry. Any SDC-capable system can render an SDC XML file into a DEF, receive and parse an SDC transmission, and regenerate the original SDC form as a DEF or synoptic report with the response data intact. SDC is therefore able to facilitate interoperable data capture and exchange for patient care, clinical trials, cancer surveillance and public health needs, clinical research, and computable care guidelines. The usability of SDC-captured oncology data is enhanced when the SDC data elements are mapped to standard terminologies. For example, an SDC map to Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) enables aggregation of SDC data with other related data sets and permits advanced queries and groupings on the basis of SNOMED CT concept attributes and description logic. SDC supports terminology maps using separate map files or as terminology codes embedded in an SDC document.

XPA: DNA Repair Protein of Significant Clinical Importance.

The nucleotide excision repair (NER) pathway is activated in response to a broad spectrum of DNA lesions, including bulky lesions induced by platinum-based chemotherapeutic agents. Expression levels of NER factors and resistance to chemotherapy has been examined with some suggestion that NER plays a role in tumour resistance; however, there is a great degree of variability in these studies. Nevertheless, recent clinical studies have suggested Xeroderma Pigmentosum group A (XPA) protein, a key regulator of the NER pathway that is essential for the repair of DNA damage induced by platinum-based chemotherapeutics, as a potential prognostic and predictive biomarker for response to treatment. XPA functions in damage verification step in NER, as well as a molecular scaffold to assemble other NER core factors around the DNA damage site, mediated by protein–protein interactions. In this review, we focus on the interacting partners and mechanisms of regulation of the XPA protein. We summarize clinical oncology data related to this DNA repair factor, particularly its relationship with treatment outcome, and examine the potential of XPA as a target for small molecule inhibitors.

Screening MRI in Women With a Personal History of Breast Cancer.

Screening MRI is recommended for individuals at high risk for breast cancer, based on genetic risk or family history (GFH); however, there is insufficient evidence to support screening MRI for women with a personal history (PH) of breast cancer. We compared screening MRI performance in women with PH vs GFH of breast cancer. We analyzed case-series registry data, collected at time of MRI and at 12-month follow-up, from our regional Clinical Oncology Data Integration project. MRI performance was compared in women with PH with those with GFH. Chi-square testing was used to identify associations between age, prior history of MRI, and clinical indication with MRI performance; logistic regression was used to determine the combined contribution of these variables in predicting risk of a false-positive exam. All statistical tests were two-sided. Of 1521 women who underwent screening MRI from July 2004 to November 2011, 915 had PH and 606 had GFH of breast cancer. Overall, MRI sensitivity was 79.4% for all cancers and 88.5% for invasive cancers. False-positive exams were lower in the PH vs GFH groups (12.3% vs 21.6%, P < .001), specificity was higher (94.0% vs 86.0%, P < .001), and sensitivity and cancer detection rate were not statistically different (P > .99). Age (P < .001), prior MRI (P < .001), and clinical indication (P < .001) were individually associated with initial false-positive rate; age and prior MRI remained statistically significant in multivariable modeling (P = .001 and P < .001, respectively). MRI performance is superior in women with PH compared with women with GFH. Screening MRI warrants consideration as an adjunct to mammography in women with a PH of breast cancer.

COST OF FEBRILE NEUTROPENIA TREATMENT IN BOSNIA AND HERZEGOVINA.

Febrile neutropenia is common chemotherapy complication significantly impacting patient’s outcomes, quality of life and costs, too. Febrile neutropenia (FN) often leads to hospitalization, the need for intravenous antibiotics and use of granulocyte colony-stimulating factors (GCSF) in order to avoid its complications. Cost of febrile neutropenia is well described in literature, but no study has been performed in Bosnia and Herzegovina. We have conducted observational cohort study with aim to describe and present costs of FN treatment from payers’ (provider’s) perspective. Only direct medical costs from one middle-sized oncology Clinic in Bosnia and Herzegovina database have been included and presented. We found that overall cost in five months period (January-May 2015) were almost 30.000 euros, or 1.0035 euro per episode/patient in average. The highest cost are allocated to hospitalization (40%), followed by GCSF (36%), while rest of costs are generated by laboratory tests performed and drug application. Proportion of costs is in line with other published studies even with huge differences in absolute values, mainly to low prices of services in Bosnia and Herzegovina. This was the first study on costs of chemotherapy induced FN in Bosnia and Herzegovina suggesting that significant costs are inquired by this conditions and that further research should be performed including larger patient population and other clinical oncology data, including pharmacoeconomic analysis.

Implementation of EMR and effects on health care delivery.

250 Background: Electronic health records have potential to enhance health care safety and quality by enabling effective communication, Clinical Decision Support technology, and generation of structured data for analysis. Despite the fact that EHRs have been shown to improve health care processes, the rapid growth of technology used to support medical care has resulted in the creation of enormous volumes of data used to assess and manage healthcare delivery. Valuable data is often siloed in multiple IT systems or captured in free-text clinical notes and non-searchable documents (unstructured data) making it impossible to quickly develop and act on real-time longitudinal data. Methods: Flatiron (FI) operates a "big data" software platform that aggregates, cleans and structures deep clinical oncology data thus facilitating understanding of operational, clinical and financial analytics related to cancer care. The platform proposes to solve these fundamental data challenges, enabling a range of real-time quality measures. It is best understood as three related components: the first is the technical integration between FI and providers' IT and data systems. That integration serves as the plumbing through which data flows securely from the provider to FI. The second is FI’s structured database, which transforms raw clinical, operational, and billing data into clean, structured, longitudinal, real-time patient data. The third is the FI user interface portal, which is a web-based analytics tool that facilitates access to the continually updated structured database. Results: Over 10 institutions have partnered with FI to deploy the platform. The full integration of an institution can be achieved in 12 to 16 weeks with daily updates to the analytics after the initial phase. Over 15,000 patients are expected to be fully integrated in the next 3-6 months. Initial QOPI and other quality measures have been included and customizable dashboards linking charge databases with clinical outcomes have been created. Dashboards with quality metrics will be presented in the conference. Conclusions: Implementation of EMRs alone does not improve healthcare delivery. Data on colorectal cancer patients will be presented at the meeting.

Physician inter-annotator agreement in the Quality Oncology Practice Initiative (QOPI) manual abstraction task.

226 Background: The Quality Oncology Practice Initiative (QOPI) relies upon the accuracy of manual abstraction of clinical data from paper-based and electronic medical records (EMR’s). While there is no “gold standard” to measure manual abstraction accuracy, measurement of inter-annotator agreement (IAA) is a commonly agreed-upon surrogate. We quantified the IAA of QOPI abstractions on a cohort of cancer patients treated at Beth Israel Deaconess Medical Center. Methods: The EMR charts of 49 patients (20 colorectal cancer; 18 breast cancer; 11 non-Hodgkin lymphoma) were abstracted by separate physician abstractors in the Fall 2010 and Fall 2011 QOPI abstraction rounds. The Fall 2011 abstractors were unaware that the charts had been previously abstracted. We analyzed data elements that were common to both rounds, had data entry for both rounds, and did not contain information expected to change after the Fall 2010 round. Cohen’s kappa (κ) was calculated for encoded data; raw levels of agreement and magnitude of discrepancies were calculated for numeric and dated data. IRB approval was obtained and all investigators completed appropriate human subjects research training. Results: 109 data elements with 2,035 paired entries met the above criteria; four narrative elements were not analyzed. Overall IAA for 1,496 coded entries was κ=0.75; median IAA for n=85 individual coded elements was κ=0.84 (interquartile range 0.30-1.00). Overall IAA for 421 dated entries was 73%; median IAA for n=17 individual dated elements was 67% (interquartile range 61-86%). The median discrepancy for the 113 discrepant dated entries was +6 days (range -217 to +391 days). Conclusions: This study establishes a baseline level of accuracy for a complex medical abstraction task with clear relevance for the oncology community. Despite the fact that all information was available in the EMR, there was disagreement for objective elements (e.g. cancer stage) as well as subjective elements (e.g. patient wellbeing). Given that the observed κ is considered only fair IAA, and that the rate of date discrepancy is high, caution is necessary in interpreting the results of QOPI and other manual abstractions of clinical oncology data.

SU‐E‐I‐88: Realistic Pathological Simulations of the NCAT and Zubal Anthropomorphic Models, Based on Clinical PET/CT Data

In the present study a patient-specific dataset of realistic PET simulations was created, taking into account the variability of clinical oncology data. Tumor variability was tested in the simulated results. A comparison of the produced simulated data was performed to clinical PET/CT data, for the validation and the evaluation of the procedure. Clinical PET/CT data of oncology patients were used as the basis of the simulated variability inserting patient-specific characteristics in the NCAT and the Zubal anthropomorphic phantoms. GATE Monte Carlo toolkit was used for simulating a commercial PET scanner. The standard computational anthropomorphic phantoms were adapted to the CT data (organ shapes), using a fitting algorithm. The activity map was derived from PET images. Patient tumors were segmented and inserted in the phantom, using different activity distributions. The produced simulated data were reconstructed using the STIR opensource software and compared to the original clinical ones. The accuracy of the procedure was tested in four different oncology cases. Each pathological situation was illustrated simulating a) a healthy body, b) insertion of the clinical tumor with homogenous activity, and c) insertion of the clinical tumor with variable activity (voxel-by-voxel) based on the clinical PET data. The accuracy of the presented dataset was compared to the original PET/CT data. Partial Volume Correction (PVC) was also applied in the simulated data. In this study patient-specific characteristics were used in computational anthropomorphic models for simulating realistic pathological patients. Voxel-by-voxel activity distribution with PVC within the tumor gives the most accurate results. Radiotherapy applications can utilize the benefits of the accurate realistic imaging simulations, using the anatomicaland biological information of each patient. Further work will incorporate the development of analytical anthropomorphic models with motion and cardiac correction, combined with pathological patients to achieve high accuracy in tumor imaging. This research was supported by the Joint Research and Technology Program between Greece and France; 2009-2011 (protocol ID: 09FR103).

Automated uro‐oncology data collection: the Cancer Research Uro‐Oncology Database

To develop a relational database (Cancer Research Uro-Oncology Database, CRUD) to enable automatic data collection on all urological malignancies within our region, as there is increasing emphasis on good data collection for surgical patients with cancer, and numerous overlapping systems that are amassing data on the same patients. Links have been established between pathological databases, multidisciplinary team data-collection systems and patient-survival monitoring facilities, providing accurate pathology, treatment and survival data on all of uro-oncology patients. We are also developing individual modules for the oncological surgeons within our unit that are compatible with the British Association of Urological Surgeons (Section of Oncology), and have plans to connect to the Medical and Clinical Oncology data systems in the future. Pre-existing protocols for fresh tissue, plasma and urine collection have been incorporated within CRUD via a tissue-tracking system, to comply with the Human Tissue Act 2004, and link these samples to accurate clinical, pathological and survival data. Many research and audit projects have already used these data, including the construction of a 274-case tissue microarray for renal cell carcinoma, microRNA hybridization arrays and analysis of 900 nephrectomy cases from the past three decades. Our work over the past 3 years in Oxford has established numerous links with organizations collecting data on our uro-oncological patients, and we are now able to collect this excellent combined data on all of these patients, in an automated manner.

URETHRAL TUMOR RECURRENCE FOLLOWING CYSTECTOMY AND URINARY DIVERSION: CLINICAL AND PATHOLOGICAL CHARACTERISTICS IN 768 MALE PATIENTS

We evaluated the incidence and risks of urethral recurrence following radical cystectomy and urinary diversion in men with transitional cell carcinoma of the bladder. Clinical and pathological results were evaluated in 768 consecutive male patients undergoing radical cystectomy with intent to cure for bladder cancer with a median followup 13 years, including 397 (51%) who underwent orthotopic urinary diversion with a median followup of 10 years and 371 (49%) who underwent cutaneous urinary diversion with a median followup of 19 years. Demographically and clinically these 2 groups were well matched with the only exception being longer median followup in the cutaneous group (p <0.001). Urethral recurrence was analyzed by univariate and multivariable analysis according to carcinoma in situ, tumor multifocality, pathological characteristics (tumor grade, stage and subgroup), the presence and extent of prostate tumor involvement (superficial vs stromal invasion) and the form of urinary diversion (cutaneous vs orthotopic). A total of 45 patients (6%) had urethral recurrence at a median of 2 years (range 0.2 to 13.6), including 16 (4%) with an orthotopic and 29 (8%) with a cutaneous form of urinary diversion. Carcinoma in situ and tumor multifocality were not significantly associated with an increased risk of urethral recurrence (p = 0.07 and 0.06, respectively). The presence of any (superficial and/or stromal invasion) prostatic tumor involvement was identified in 129 patients (17%). Prostate tumor involvement was associated with a significantly increased risk of urethral recurrence (p = 0.01). The estimated 5-year chance of urethral recurrence was 5% without any prostate involvement, increasing to 12% and 18% with superficial and invasive prostate involvement, respectively. Patients undergoing orthotopic diversion demonstrated a significantly lower risk of urethral recurrence compared with those undergoing cutaneous urinary diversion (p = 0.02). Patients without any prostate tumor involvement and orthotopic diversion (lowest risk group) demonstrated an estimated 4% year chance of urethral recurrence compared with a 24% chance in those with invasive prostate involvement undergoing cutaneous diversion (highest risk group). On multivariate analysis any prostate involvement (superficial and/or invasive) and urinary diversion form remained independent and significant predictors of urethral recurrence (p = 0.035 and 0.01, respectively). At long-term followup urethral tumor recurrence occurs in approximately 7% of men following cystectomy for bladder transitional cell carcinoma. Involvement of the prostate with tumor and the form of urinary diversion were significant and independent risk factors for urethral tumor recurrence. Patients undergoing orthotopic diversion have a lower incidence of urethral recurrence compared with those undergoing cutaneous diversion. Although prostate tumor involvement is a risk factor for urethral recurrence, it should not preclude orthotopic diversion, provided that intraoperative frozen section analysis of the urethral margin is without evidence of tumor.

On the Use of Cause-Specific Failure and Conditional Failure Probabilities: Examples From Clinical Oncology Data

Abstract Nonparametric maximum likelihood estimation of the probability of failing from a particular cause by time t in the presence of other acting causes (i.e., the cause-specific failure probability) is discussed. A commonly used incorrect approach is to take 1 minus the Kaplan-Meier (KM) estimator (1 – KM), whereby patients who fail of extraneous causes are treated as censored observations. Examples showing the extent of bias in using the 1-KM approach are presented using clinical oncology data. This bias can be quite large if the data are uncensored or if a large percentage of patients fail from extraneous causes prior to the occurrence of failures from the cause of interest. Each cause-specific failure probability is mathematically defined as a function of all of the cause-specific hazards. Therefore, nonparametric estimates of the cause-specific failure probabilities may not be able to identify categorized covariate effects on the cause-specific hazards. These effects would be correctly identified by cause-specific cumulative hazard or KM plots in which the extraneous causes of failure are treated as censored observations. Examples are provided. Finally, nonparametric graphical representation of the two distinct cause-specific failure components of the mixture model (i.e., the probability of ever failing from a particular cause and the time-to-failure distribution given that a patient will fail of that cause) are presented. The difficulty in extrapolating the nonparametric estimates beyond the range of observed failure times is highlighted. In addition, the mathematical relationship of a single covariate z and the two cause-specific failure components is shown for the case where z acts multiplicatively on the cause-specific hazards. Examples are considered using the important prognostic factors in adult soft tissue sarcoma (STS) of the extremity and adult acute lymphoblastic leukemia (ALL). In each of the mathematical and actual examples considered, the characteristic that was associated with a higher hazard rate of failure from the disease was also associated with a higher probability of ever failing from the disease as well as a shorter time-to-failure distribution given that failure due to the disease will occur. Thus it may be quite common in chronic disease situations to find prognosticators that are associated with both cause-specific failure components.

On the Use of Cause-Specific Failure and Conditional Failure Probabilities: Examples from Clinical Oncology Data

Abstract Nonparametric maximum likelihood estimation of the probability of failing from a particular cause by time t in the presence of other acting causes (i.e., the cause-specific failure probability) is discussed. A commonly used incorrect approach is to take 1 minus the Kaplan-Meier (KM) estimator (1 – KM), whereby patients who fail of extraneous causes are treated as censored observations. Examples showing the extent of bias in using the 1-KM approach are presented using clinical oncology data. This bias can be quite large if the data are uncensored or if a large percentage of patients fail from extraneous causes prior to the occurrence of failures from the cause of interest. Each cause-specific failure probability is mathematically defined as a function of all of the cause-specific hazards. Therefore, nonparametric estimates of the cause-specific failure probabilities may not be able to identify categorized covariate effects on the cause-specific hazards. These effects would be correctly identified ...