Harnessing Real-World Evidence for the Development of Novel Cancer Therapies

Resourcing real-world evidence (RWE) is becoming an increasingly important asset in developing novel therapies for cancer. In this article, an overview of the benefits and challenges of using these data is provided. Through several case examples we highlight future applications and potential. Resourcing real-world evidence (RWE) is becoming an increasingly important asset in developing novel therapies for cancer. In this article, an overview of the benefits and challenges of using these data is provided. Through several case examples we highlight future applications and potential. The use of population-level data aggregated from electronic health records (EHRs), broadly referred to as RWE, has recently received much attention as a tool to enhance the treatment of cancer; however, use of RWE in the study of cancer is not new. Historic evidence has provided valuable insights into patient subgroups with poor prognosis, as well as cancer incidence as it relates to specific environmental, behavioral, geographic, and socioeconomic factors. What is different now? Emerging RWE data sources are more accessible and have improved resolution with respect to treatment outcomes when compared with traditional claims databases, which greatly expands their utility. RWE can now provide improved insights into the patient experience, such as tracking access to care and the adoption of new treatment options. Emerging data can link the use of specific treatment regimens with efficacy and safety risks, define factors associated with clinical decision making, as well as identify underserved patient populations. Interrogating RWE across large healthcare networks, may help with prescreening of specific, rare patient cohorts for potential participation in clinical trials. Furthermore, advances in data science can develop more complete data sets and provide novel insights. In this article, applications of RWE and advanced analytic techniques for informing and advancing oncology drug development, with insights into future potential, are discussed. Echoing larger patterns of RWE use in improving cancer care, learnings based on RWE are becoming an increasingly important component in developing novel therapies for cancer. A wide range of information can be leveraged, from selection of sites who are likely to care for the study population of interest and perform well on a given clinical trial, to optimizing the likelihood of achieving therapeutic benefits. Tracking the drivers and dynamics of routine clinical practice can be critical for strategic clinical trial planning and execution. For example, RWE-based studies can be used to evaluate available therapies and associated outcomes, thereby informing the statistical design (e.g., predicting hazard ratios) of multi-arm randomized clinical trials (RCTs). In addition to overall survival metrics, well-curated RWE sources and advances in data capture may enable longitudinal follow-up of patients, including ascertainment of time-to-event outcomes such as time to progression and time to treatment discontinuation. These surrogates can identify unmet medical need, and improve the speed of evaluating novel therapies. Furthermore, other information, such as institutional treatment guidelines and management of patient symptoms and therapeutic toxicities may also be valuable for clinical trial planning and design. Identifying highly responsive patients is a cornerstone of developing novel therapeutics for cancer. The advent of expanded genomic testing of patients can provide new opportunities for retrospective associations of tumor genomics and outcomes. It can also contribute to predictions of disease progression, relapse, and risk stratification. A growing number of cases exemplify this potential. For example, a comprehensive review of factors associated with programmed death (ligand) 1 [PD-(L)1] treatments in non-small-cell lung carcinoma (NSCLC) demonstrate that tumor cell staining of PD-L1 is done routinely in clinical practice, and strong staining (≥50%) associates with improved outcomes [1.Khozin S. et al.Real-world progression, treatment, and survival outcomes during rapid adoption of immunotherapy for advanced non-small cell lung cancer.Cancer. 2019; 125: 4019-4032Crossref PubMed Scopus (62) Google Scholar]. Real-world matched clinical and genomic datasets of patients with advanced urothelial cancer, suggest that FGFR alterations trend with poorer overall survival following anti-PD-(L)1 therapy, compared with patients whose tumors are wild type [2.Santiago-Walker A. et al.Predictive value of fibroblast growth factor receptor mutations and gene fusions on anti-PD-(L)1 treatment outcomes in patients with advanced urothelial cancer.J. Clin. Oncol. 2019; 7: 419Crossref Google Scholar]. The latter example indicates therapeutic need, and provides the basis for interpretations of outcome with novel targeted therapies in this population. Various challenges complicate the practical use of RWE in drug development (Table 1). First, a major limiting factor is access to suitable data. The improved aggregation of oncology RWE has been well described in prior reviews [3.Parkin D.M. The evolution of the population-based cancer registry.Nat. Rev. Cancer. 2006; 6: 603-612Crossref PubMed Scopus (283) Google Scholar]. Though academic institutions and commercial data sources can play an important role in developing well curated patient cohorts, operational challenges, such as inconsistent data entry, fragmentation of care, and complex, diverse EHR platforms still hamper use. Research collaborations across multi-institutional EHRs, through either an aggregated or federated approach, require harmonized data. Valuable information, such as medical history, baseline disease characteristics, and investigator-assessed outcomes, are often not captured in a structured format, but more routinely documented in notes by the treating physician or associated staff. Advances in natural language processing can improve knowledge extraction from unstructured notes. For example, improvements have been demonstrated in querying medical reports from myeloma patients across a diversity of sources [4.Loda S. et al.Exploration of artificial intelligence use with ARIES in multiple myeloma research.J. Clin. Med. 2019; 8: 999Crossref Scopus (3) Google Scholar]; however, delivering high quality, reliable data sets still requires augmented review by oncology-experienced personnel.Table 1Overview of Approaches for Addressing Challenges Common to the Practical Use of RWEChallengeApproachRefsLack of suitable data sets for rare patient cohorts•Aggregate data across commercial and academic data sources•Coordinated government efforts such as CanREValue will centralize data across large populations[13.Chan K. et al.Developing a framework to incorporate real-world evidence in cancer drug funding decisions: The Canadian Real-world Evidence for Value of Cancer Drugs (CanREValue) collaboration.BMJ Open. 2020; 10e032884Crossref PubMed Scopus (19) Google Scholar]Inconsistencies that exist between RCT data capture and RWE hamper ability to properly match patients•Impute variables, such as ECOGaAbbreviation: ECOG, Eastern Cooperative Oncology Group. status, using machine learning techniques–High value data fields are not captured in structured medical records and unavailable for use in RWE studies•Natural language processing can extract data from medical notes and enhance manual review–Inconsistencies between institutional EHR systems reduce the fidelity of aggregated data•Multi-institutional ‘prospective’ RWE studies can harmonize different data sources•Future development of consensus-based set of core RWE data elements for cancer[10.Conley R.B. et al.Core clinical data elements for cancer genomic repositories: a multi-stakeholder consensus.Cell. 2017; 171: 982-986Abstract Full Text Full Text PDF PubMed Scopus (10) Google Scholar]a Abbreviation: ECOG, Eastern Cooperative Oncology Group. Open table in a new tab In the context of drug development, there are numerous challenges to be aware of when applying RWE. Relative to RCTs, a core strength of RWE is that it allows the study of contemporaneous larger cohorts with longer follow-up duration. Conversely, RWE data capture is not as standardized or homogeneous as that in RCTs, and therefore may be subject to potential bias, which limits its comparability. Prior efforts to align RWE to that from clinical trials, has yielded mixed results when patient outcomes from RCTs are compared with outcomes from the RWE. Aggregating real-world based short-term (time to treatment discontinuation) and longer-term (overall survival) endpoints for NSCLC patients treated with PD-(L)1 therapies, has been shown to produce outcomes consistent with clinical trials [5.Stewart M. et al.An exploratory analysis of real-world end points for assessing outcomes among immunotherapy-treated patients with advanced non–small-cell lung cancer.JCO Clin. Cancer Inform. 2019; 3: 1-15Crossref PubMed Scopus (52) Google Scholar]; however, such comparisons need to be done with a deep understanding of potential confounders, which limit extrapolation from RWE to RCTs. A study of real-world outcomes in metastatic renal cell carcinoma patients, demonstrates that approximately 40% of patients treated with targeted therapies in the real world would have been ineligible for associated Phase III trials [6.Mitchell A.P. et al.Clinical trial participants with metastatic renal cell carcinoma differ from patients treated in real-world practice.J. Oncol. Pract. 2015; 11: 491-497Crossref PubMed Scopus (57) Google Scholar]. Despite the risks of unrepresentative measurements of clinical benefit, RWE cohorts have even been successfully used in regulatory filings in the absence of a traditional RCT. For example, following approval of the cyclin-dependent kinase (CDK)4/6 inhibitor, palbociclib in combination with an aromatase inhibitor, or fulvestrant, in women with estrogen receptor+/HER2 (human epidermal growth factor receptor 2)- breast cancer, the FDA granted a label expansion to include male breast cancers based solely on retrospective RWE from commercial sources. Due to the use of prescription duration as an outcome surrogate, and the dependence on historic safety risk, the true clinical benefit in these patients was speculative [7.Raphael M.J. et al.Real-world evidence and regulatory drug approval.Nat. Rev. Clin. Oncol. 2020; 17: 271-272Crossref PubMed Scopus (13) Google Scholar]. Finally, by nature, the data used in retrospective RWE cohorts is inflexible, and does not capture important factors in patient care. For example, while retrospective data demonstrate no association between autoimmune disorders (ADs) and real-world outcomes to immune checkpoint blockade (ICB) in patients with advanced NSCLC, these data do not capture the essence of a physician’s decision to treat patients with ADs, hence making safety of ICB in these patients difficult to interpret [8.Khozin S. et al.Real-world outcomes of patients with advanced non-small cell lung cancer (aNSCLC) and autoimmune disease (AD) receiving immune checkpoint inhibitors (ICIs).J. Clin. Oncol. 2019; 37: 110Crossref Google Scholar]. In the future, multidisciplinary efforts will play a major role leveraging RWE in developing new cancer therapies. Building clinical trial-matched RWE data sets for applications such as synthetic control arms (SCAs) often requires data not readily available through structured RWE sources, such as baseline disease characteristics and treatment history. Emerging approaches, such as the successful imputation of patient performance status from RWE [9.Agrawal S. et al.Machine learning imputation of Eastern Cooperative Oncology Group performance status (ECOG PS) scores from data in CancerLinQ discovery.J. Clin. Oncol. 2020; 38e19318Crossref Google Scholar], will help harmonize real world data with clinical trial data capture. Data sparseness and incompatibility across health networks are also being addressed through a consensus-based set of core RWE data elements for cancer [10.Conley R.B. et al.Core clinical data elements for cancer genomic repositories: a multi-stakeholder consensus.Cell. 2017; 171: 982-986Abstract Full Text Full Text PDF PubMed Scopus (10) Google Scholar]. Such techniques will improve the analytic rigor needed to effectively apply SCAs in support of single arm studies. Novel analytic techniques can bring together diverse data types to expand novel insights gained from RWE. Enrichments of high-quality data coupled with machine learning can help find areas of therapeutic need, enhance clinical trial interpretation, and even identify novel targets. For example, artificial intelligence (AI) applied to >50 000 lung cancer cases from the American Society of Clinical Oncology (ASCO) CancerLinQ database was used to predict survival through 360 days of follow up [11.Agrawal S. et al.Development of an artificial intelligence model to predict survival at specific time intervals for lung cancer patients.J. Clin. Oncol. 2019; 37: 6556Crossref Google Scholar]. Molecular medicine is a critical component in advancing cancer patient care. Proposed frameworks, such as the ‘Master Observational Trial’, which is an amalgamation of interventional and prospective observational trials in biomarker-tested patients, will enroll patients regardless of their molecular signatures and track longitudinal outcomes [12.Dickson D. et al.The master observational trial: a new class of master protocol to advance precision medicine.Cell. 2020; 180: 9-14Abstract Full Text Full Text PDF PubMed Scopus (20) Google Scholar]. Machine learning approaches can then be used to develop actionable hypotheses for signature-based therapeutic benefit. Harnessing the potential of RWE will require close cooperation between key stakeholders within the pharmaceutical industry, commercial RWE providers, academic institutions, and government. Fortunately, there are emerging examples where regulators, patient associations, industry, and healthcare providers are partnering to leverage the benefits of collaborative RWE-based data science initiatives. For example, The Canadian Real-World Evidence Value of Cancer Drugs (CanREValue) project framework aims to enhance clinical benefit and value of cancer therapies by making real-time data-driven recommendations and use of novel funding mechanisms for decision-makers/ payers across the country [13.Chan K. et al.Developing a framework to incorporate real-world evidence in cancer drug funding decisions: The Canadian Real-world Evidence for Value of Cancer Drugs (CanREValue) collaboration.BMJ Open. 2020; 10e032884Crossref PubMed Scopus (19) Google Scholar]. Similarly, in Europe, efforts to optimize treatment benefits and leverage RWE for personalized medicine are supported by public/private projects. Here, major advancements are being made in privacy-preserving distributed data access and analytics. If effective, efforts such as these will advance interoperability and data harmonization through automated data capture and aggregation, while maintaining data privacy and enabling robust, clinically informative analyses. Future opportunities point to expanded use of RWE in the support of regulatory drug approvals, exemplified by increasing levels of published guidance from the FDA [14.Gottlieb S. Statement from FDA Commissioner Scott Gottlieb, M.D., on FDA’s new strategic framework to advance use of real-world evidence to support development of drugs and biologics.2018Google Scholar], as well as directed efforts to include real world SCA's in health authority submissions to facilitate regulatory decision making [15.Tallent A. CancerLinQ Partners with FDA to Study Real-World Use of Newly Approved Cancer Treatments. ASCO in Action, 2017Google Scholar] and post-marketing surveillance. Dedicated multidisciplinary efforts and advancing technology can promote a more critical role for RWE by expanding and harmonizing data sets. For patients, this will ultimately enable data-driven treatment decisions through real time data capture and self-improving algorithms. These are tangible goals, and represent a major opportunity to enhance the development of new therapies and ultimately improve cancer care. We gratefully acknowledge the review of this article by Tammy Guld, Senior Director, Clinical Innovation, Janssen Research and Development, and Bart Vannieuwenhuyse, Senior Director, Data Science Lead, Janssen Research and Development.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.