Efforts to develop innovative, lifesaving medicines are always fraught with peril, and the clinical and commercial success of any investigational therapy is never guaranteed. Today, all stakeholders in the healthcare spectrum — pharmaceutical developers, payers, regulators, physicians and patients — are putting their money on the collection and analysis of many different types of real-world evidence (RWE) as a key enabling strategy, to close critical gaps in knowledge, give physicians and patients broader access to therapies, and help payers realize the actual value of those therapies in improving health and reducing costs.
Traditional clinical trials — while still the gold standard for establishing a drug’s risk/benefit profile and appropriate dosing strategy — are not able to fully elucidate how the drug, once launched, will perform under messy, real-world conditions, where patient populations are decidedly more heterogeneous than those studied in the trial. To better assess the full clinical profile and economic impact of high-cost specialty medicines over time, efforts are underway to explore and exploit various forms of real-world data that can be collected during patient care. Then, analytic, modeling and simulation tools, including artificial intelligence (AI) and machine-learning methodologies, can derive meaningful insights from those data.
But the process of using RWE is far from straightforward. Since all data are not created equal, a variety of industry stakeholders are developing standardization frameworks in an effort to establish some form of accountability and reliability, and thus foster greater confidence as RWE is being used to advance a variety of pharma and healthcare objectives.
Today, many data sources — some traditional, others emerging — are being used to improve insight during drug development. These include:
- Insurance claims and billing activities
- Pharmacy and specialty pharmacy dispensing data
- Electronic health records (EHR) data
- Aggregated clinical lab-testing data
- Genomic data
- Patient registries related to specific diseases and products
- Patient-generated data (for instance, from in-home, mobile and wearable devices and sensors)
- Patient-reported outcomes data
- Behavioral data and insights collected from social media
- Purpose-built databases
Third-party aggregators of data records provide another good source of RWE. For example, IQVIA owns a comprehensive, global data portfolio representing more than 530 million de-identified patient records across 10+ markets, with more than 3,000 ready-to-use sources from EHR, hospital, pharmacy and claims data, as well as data related to genomics, mobile-health and patient-reported outcomes.
Similarly, the healthcare AI company Prognos claims to be the largest aggregator of laboratory data related to diagnostic testing in healthcare. Its database, called The Prognos Registry, contains more than 13 billion de-identified, HIPAA-compliant clinical records, representing more than 180 million patients, from more than 140 commercial diagnostic clinical-testing laboratories, according to the company. This enormous cache — which continues to grow over time — spans many different pathologies, in 50 disease states.
Today, a variety of advanced modeling, simulation and artificial intelligence capabilities are being used to analyze data from actual clinical practice to overcome the limitations of narrow or insufficient clinical trial data. Such efforts, according to Camie Britton, Senior Director, Real-world Data Services, Parexel International, include:
- Mining unstructured data (for instance, from EHR clinical notes and sensors) for research purposes
- Creating retrospective cohorts of patients to better understand how a drug works in a defined indication
- Developing and validating algorithms based on EHR clinical notes or claims
- Using such algorithms to distinguish and characterize targeted subpopulations as a means of supporting future research
“Two general approaches are emerging in terms of how drug developers are using RWE — greater use of classical, hypothesis-driven research, and the use of newer data-analytic tools including AI and machine-learning techniques that provide insight based on pattern recognition within vast data sets,” says John Doyle, SVP & GM, Real-world & Analytic Services, IQVIA, and faculty member, Department of Epidemiology, Mailman School of Public Health, Columbia University. “We should be looking at a hybrid approach, but we also need to challenge and ‘stress test’ the pattern-recognition results and inferences that are being made about causation, to ensure that they are based on sound scientific principles.”
The second annual RWE Benchmarking Survey,  from Deloitte’s ConvergeHealth unit, finds that 90% of global pharma companies either have or are building RWE analytical capabilities, and that applications range across the entire product lifecycle for these undertakings.
Leveraging data from off-label prescribing
Once a drug has been approved and enters the marketplace, it is often prescribed to additional patient subgroups (beyond just the narrow patient cohorts enrolled in the clinical trial, or are specified in the approved label indication). The ability to study the drug’s performance in patients using the drug off-label is important, as it will reflect how the drug actually performs among a decidedly more heterogeneous patient population, and thus can help to paint a broader picture of the drug’s full clinical potential.
In some cases, RWE findings from such patients can help the drug maker to seek and gain regulatory approval for additional label indications for that therapy, and can help to inform clinical guidelines and other decision-support tools for use in routine clinical practice.
However, unlike the tightly screened patient populations enrolled in a formal clinical trial, patients seeking treatment for any condition under real-world conditions often have one or more comorbidities, and are already taking other medications that could lead to drug-drug interactions (DDIs). Those DDIs, coupled with unpleasant side effects or adverse events, affordability issues and other patient behaviors, often conspire to reduce the patient’s adherence to therapy, and poor adherence to therapy is one of the leading factors to undermine the clinical effectiveness and health outcomes that were demonstrated in the trial.
For many specialty drugs that are aimed at diseases with critical unmet medical need — particularly in oncology and rare or orphan diseases — fast-track or conditional regulatory approval provides an accelerated pathway for regulatory approval. “Giving patients faster access to these drugs is fantastic, but fast-track approval leaves an unfinished story still to be told,” says Doyle of IQVIA. “RWE becomes the only option to round out that conditional approval, to demonstrate safety, effectiveness and tolerability, across more diverse populations in real-world clinical settings over longer time horizons.”
Patient or product registries can yield data from a sufficient number of patients to enable statistically relevant subgroup analyses, and these findings can be used to seek expanded label indications and help demonstrate the therapy’s full value proposition to payers and Health Technology Assessment (HTA) organizations. “Such examples are exciting and they are really happening today, but frankly these RWE techniques are still being grossly underutilized to validate, extend and complement the findings of these clinical trials,” he adds.
Meanwhile, Zhen Su, MD, Chief Medical Officer for North America at EMD Serono cautions: “Monitoring how the drug responds in specific sub-populations in actual clinical settings can certainly help to elucidate who responds better to treatment, enabling drug developers to negotiate for label expansion for the drug. But drug developers must also be prepared for the fact that such efforts may also reveal patients who do not end up responding well over time, and this can lead FDA to shrink the label indication for that therapy.” This happened in June, when FDA announced it was restricting the use of Merck & Co.’s Keytruda (pembrolizumab) and Roche’s Tecentriq (atezolizumab) in first-line bladder cancer to patients who express high levels of PD-L1, revising the labels for both drugs to include a requirement for testing for PD-L1 expression. Treatment continuation can then be considered for patients responding to the drugs who are not eligible for cisplatin-containing chemotherapy, according to FDA. 
While RWE has historically been used to monitor post-launch safety and adverse events, drug developers are increasingly incorporating RWE-driven insights further upstream to improve clinical trial design and execution. For instance, RWE can be used to assess and validate specific biomarker hypotheses, or to create a “synthetic” control arm based on RWE derived from historic or contemporaneous patients being treated in actual practice and then use that synthetic arm in lieu of an actual control arm during the clinical trial.
“The need for clinical trials to run a control arm often creates redundancy and added time and cost for the drug developer, and for some diseases and drugs, it’s unethical to put patients into a control arm of a trial,” says Su of EMD Serono. “By comparison, the ability to create a synthetic control arm based on RWE can reduce drug-development costs and enable better treatment paradigms. This approach is especially helpful to improve and streamline the regulatory approval process for investigational therapies that are intended to address unmet medical need in rare diseases.” Su cites the example of Bavencio (avelumab), an immunotherapy developed by EMD Serono and parent company Merck KGaA for the treatment of metastatic Merkel cell carcinoma (MCC), saying: “Bavencio successfully used real-world data as a reference for efficacy, because it’s a rare disease and thus trial enrollment was challenging.”
Similarly, RWE is also making its mark further downstream, to support regulatory submissions and/or later label expansions (for instance, to pursue additional indications, additional patients cohorts or additional clinical endpoints), and to inform negotiations with payers related to formulary placement, drug pricing and the negotiation of value-based contracts or outcomes-based agreements that tie reimbursement rates more explicitly to the performance of the therapy under messy real-world conditions.
“More than ever, drug manufacturers know the trial won’t provide the final evidence package — it’s just the first part of the entire package — and they know that the strategic use of RWE can enable a deeper understanding of value,” says Sarah Alwardt, PhD, VP of HI and HEOR Operations at McKesson, adding: “Since everyone understands that clinical trial findings are limited, there’s a huge appetite for these types of retrospective studies using RWE. It’s a change in mindset that’s very exciting.”
The expanded use of RWE is already helping different stakeholders at different points throughout the biopharmaceutical product lifecycle to expand and calibrate their understanding of the actual clinical effectiveness, long-term health outcomes, cost impact and risk/benefit ratios for various therapies, says Britton of Parexel International, who notes that the growing list of RWE-informed efforts includes:
- Identifying new drug targets and better biomarkers
- Finding new uses for existing and late-stage products
- Identifying undiagnosed, underdiagnosed and misdiagnosed illness
- Identifying sites with higher enrollment probability and capability to aid clinical trial recruitment
- Predicting patients who have higher risk for adverse events
- Predicting medication compliance (or lack thereof) in defined patient cohorts (which can inform mitigation efforts)
- Using AI and machine-learning techniques to recognize trends and patterns in social media and healthcare sensor data (to inform clinical study development and define study objectives, patient-relevant endpoints and study procedures)
“Postmarketing studies using RWE are more common now than ever,” notes Michael Fronstin, GM, Real World Evidence, Kantar Health. “Early signal detection through observational studies of newly approved medications is critical to minimizing the potentially broad impact of an unforeseen risk, should a new drug prove to be unsafe within certain patient cohorts.”
Eli Lilly’s new compound, Engality (galcanezumab-gnlm), is expected to receive FDA approval by the end of 2018 for the treatment of migraine headaches. To further understand the burden of migraine and the barriers that preventive and acute treatment options face, Eli Lilly recently announced a study dubbed the ObserVational Survey of the Epidemiology, tReatment and Care Of MigrainE (OVERCOME). According to the company, this two-year study will be the largest of its kind in migraine, gathering real-world data from 40,000 patients. 
Overcoming clinical trial limitations
By design, the patient population enrolled in any given clinical trial is limited, with targeted patient cohorts handpicked using strict inclusion and exclusion criteria, and enrollment tending to favor younger, healthier patients compared to the entire population in that disease category. Similarly, trial investigators are typically given formal training and required to follow consistent protocols for drug administration and data collection. Collectively, these factors help to greatly reduce variability as the clinical trials are carried out, and thus allow the investigational therapy to show its best self, in a trial population for whom co-morbidities and drug adherence issues have been eliminated or minimized.
By contrast, “when you have someone who is not being studied and poked and prodded in a highly controlled clinical trial, their outcomes will be different,” says Alwardt of McKesson. “By studying data from realworld prescribing, you can dig into which factors are helping to drive improved outcomes in specific patient sub-populations. In other instances, the outcomes will be worse and you can try to identify and reconcile the issues (such as co-morbidities, drug-drug interactions, dosing, complex-administration or adherence issues) that may be to blame.”
“Historically, clinical trials have been primarily focused on efficacy and safety, but in recent times clinical effectiveness has risen as a key differentiator to be studied in the trial,” adds Kantar Health’s Fronstin. “Effectiveness comes in many forms, but is often associated with patient outcomes, which may be clinical, humanistic or economic in nature.”
“Real-world data, such as patient-reported outcomes (PROs), quality of life, disease progression and epidemiology, exposure patterns and factors driving reffectiveness, can inform which parameters need to be captured in the trial design,” says Billy Amzal, MSc, MPA, PhD, SVP, Real World Decision & Data Analytics, Analytica Laser, a Certara company. “Patient registries and social media through opt-in surveys offer a good source of such patient preference data.”
“The patient’s perspective and preferences, captured through PROs, play a particularly important part of demonstrating value in rare diseases and childhood disorders, and focus groups involving patients and other stakeholders such as nurses and physicians, patient association members and caretakers) can help to elicit what types of attributes the drug maker should be looking for when designing the trial, adds Sumeet Bakshi, MBBS, MBA, VP, Real World Data Solutions at Analytica Laser. However, he adds “We find a lot of drug manufacturers coming to us very late in the design phase of a clinical trial, asking for us to identify the most relevant PROs.”
“To incorporate PROs into trials, drug developers are analyzing real-world data to identify specific patient populations that may be experiencing incremental burden or negative health outcomes as a result of their disease,” adds Fronstin of Kantar Health. “Using this insight, patient-reported outcomes can help to shape the trial design.”
Similarly, while the drug’s ability to demonstrate a favorable risk/benefit profile within this limited patient cohort is sufficient to earn regulatory approval, this traditional approach often leaves more questions than it answers, in terms of whether the approved drug can be used in other patient cohorts that were not involved in the trials (such as pregnant women, elderly or pediatric patients, those with renal or hepatic impairment, those with unstudied co-morbidities and so on). In the absence of trial-derived data, physicians are then left to prescribe drugs off-label. This raises potential clinical and ethical concerns for physicians (due to a lack of comprehensive data), and often creates a situation where insurance coverage is restricted or denied, leaving patients to manage the entire cost burden of treatment.
“There are conversations happening today that have never happened before, and in clinical trial design, there is increased appetite among both drug developers and regulators to better understand that RWE can help improve trial design,” says Alwardt of McKesson. For instance, she notes that “In oncology in particular, one of the clinical trial endpoints is almost always progression-free survival (PFS), but PFS is really hard to measure in the real world due to variation from doctor to doctor in terms of how they evaluate and capture such data.” Instead, getting more relevant, real-world endpoints into the clinical trial design will help to create a more powerful trial baseline and more realistic expectations for how the drug will perform once it enters the market.
Innovative value-based contracting or risk-sharing agreements — formally called risk-based contracts or outcomes-based agreements — have been explored for a growing number of high-cost specialty medications. Such innovative contracting arrangements provide a mechanism for biopharma companies and health insurers to both put more skin in the game when it comes to managing the high cost or risk of today’s specialty therapies. Such arrangements aim to tie contracted drug prices and reimbursement rates more closely to actual clinical outcomes that are demonstrated by collecting and analyzing RWE from clinical practice. “The whole idea is to manage uncertainty at the time the manufacturers are committing to a given clinical performance for their drug and payers deciding for a price and reimbursement,” says Amzal of Analytica Laser.
According to Deloitte, 14 of the 20 companies polled in its 2nd annual RWE Benchmarking Survey are currently engaged in value-based contracting, although many industry observers note that drug companies and payers have tended to remain tight-lipped about publicizing such agreements, hoping to preserve their own competitive advantage. Such secrecy thwarts the sharing of lessons learned, and undermine efforts to create a consistent framework or best practices to improve the data collection and expectations and to guide the complex negotiations that are needed to create a win-win for both sides through these innovative contracting schemes.
All data are not created equal
While interest in and use of RWE throughout pharma and healthcare continues to grow, many industry observers have expressed concern and frustration over the inherent variability and inconsistency of the RWE data sources that are increasingly being pressed into service. Such challenges, according to Britton of Parexel include:
- Heterogeneous data and lack of interoperability
- Data incompleteness, inconsistency and inaccuracy
- Data ownership and data-use agreement status
- Data privacy regulation / patient-level authorization / ongoing patient consent management for RWD usage
- Lack of data standards
- Lack of standard unique patient identifiers to enable linking across data sources
- Multiple data sources and the need for ‘big data’ aggregation and ‘data lake infrastructure’
- Shortage of biomedical and informatics expertise to successfully leverage modeling, simulation and AI opportunities
Similarly, Bakshi of Analytica Laser notes investigators must be aware of the limitations of the available data, and design any studies accordingly. “Claims data was set up to manage insurance claims — it was never intended to be used for conducting comparative effectiveness studies, but many are doing that today, so you have to identify the gaps and limitations.”
Many are calling for some type of industry standards or guidelines, to establish some criteria for data consistency.
“The lack of comprehensiveness, consistent quality and standardization threatens to undermine trust and limit the full potential of using RWE to advance many objectives in pharma,” says Doyle of IQVIA. “Today with so many parallel efforts, the use of RWE is being democratized, which should prompt research design innovation, but how do you validate and pressure-test the various approaches for validity and reliability? I am eager to see what type of guidance FDA ultimately issues with regard to RWE guidance.” He notes that FDA’s 2017 guidance issued on the use of RWE in the field of medical devices  provides a lot of hints about FDA’s thinking on this topic.
Additional industry-wide initiatives are already underway to address some of these challenges. For instance, in 2017, the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) and the International Society for Pharmacoepidemiology (ISPE) created a joint task force which is working to establish consistent methodologies and procedural practices to improve the consistency and reliability of using RWE 
Similarly, the GetReal Initiative (www.imi-getreal.eu), established in 2013 by an EU public-private consortium consisting of pharmaceutical companies, academia, HTA agencies, regulators and patient organizations, is a working to develop a variety of tools and resources for RWE.
- Second Annual Real-World Evidence (RWE) Benchmarking Study, 2018, www.deloitte.com/insights
- FDA, Use of Real-World Evidence to Support Regulatory Decision-Making for Medical Devices, issued August 31, 2017; https://www.fda.gov/downloads/medicaldevices/deviceregulationandguidance/guidancedocuments/ucm513027.pdf
- Berger, Marc., et al., Good Practices for Real-World Data Studies of Treatment and/or Comparative Effectiveness: Recommendations from the Joint ISPOR-ISPE Special Task Force on Real-World Evidence in Healthcare Decisionmaking,