Using RWE to inform relative effectiveness estimates and trial design: a case study in metastatic melanoma

Context

A key objective of Work Package 1 of GetReal was to develop a framework for incorporating real-world data (RWD) into decision-making. Case studies were constructed to explore the different ways that RWD may be used to help demonstrate the relative effectiveness of new medicines.

Melanoma is a form of skin cancer that first manifests as an asymmetric primary tumour. It may spread to other parts of the body, which is referred to as metastatic melanoma. Treatments depend on a number of factors, including the health and preferences of the person, and the location, number and rate of spread of the metastases. Since 2011, a number of new medicines for metastatic melanoma have become available, which have been subject to health technology assessments in different European countries.

What was examined in this case study?

The case study on metastatic melanoma aimed to:

  • explore options for the combination of real-world evidence (RWE) and randomised controlled trial (RCT) data to improve extrapolation of long-term survival
  • explore the use of social media to collect patient perspectives on quality of life (QoL). (see here for more about social media)

Extrapolation of long-term survival
Overall survival (OS) is a key outcome measure for assessment of the long-term benefits of treatments for metastatic melanoma. However, traditional RCTs are carried out over a shorter time period, and so extrapolation of the results is needed to estimate long-term outcomes (such as OS) beyond the trial period. In this case study, long-term extrapolation for OS was examined using the following combinations of real-world data (RWD) and RCTs:

  • Base case: Individual participant data (IPD) from RCT data to perform extrapolation of OS using parametric survival models.
  • Approach 1: IPD from RCT data + IPD simulated from RWD (published Kaplan-Meier curves of registry publications).
  • Approach 2: IPD from RCT data + IPD re-created from SEER-Medicare registry (based on summary data).
  • Approach 3: IPD from RCT data + IPD simulated from RWD + IPD re-created from SEER-Medicare registry.

Social media to collect patient perspectives
As OS trends increase with the introduction of new therapies, the QoL associated with prolonged survival becomes increasingly important to decision makers. However, in both RCTs and registries, it is often difficult logistically to capture QoL data. Therefore, the case study evaluated the extent to which patient perspectives on QoL could be assessed via social media. Melanoma patient perspectives on QoL were collected in two ways:

  • survey of patient perspectives on QoL
  • use of forum posts to identify patient perspectives on QoL.

What ‘effectiveness challenge(s)’ was addressed in this case study?

Extrapolation of long-term survival:

  • Insufficient length of trial outcomes.

Social media to collect patient perspectives:

  • Absence of understanding of how reported adverse events are managed in practice.
  • Gaps in QoL measurements.

What were the findings and conclusions?

Difficulties obtaining registry data
The metastatic melanoma case study initially aimed to collect IPD from three European registries (in the Netherlands, France and Italy). Applying for data access involved complying with the general procedure of each registry. But despite aiming to comply, all three registries declined the requests for data access or were unable to provide data in time for inclusion in this case study. Access to RWD to inform regulatory, HTA and development decisions is important, and this difficulty accessing registry data is a key finding in itself. This was supported by input from stakeholders, who faced difficulties accessing RWD in the form of IPD from registries across their different fields of work. (See a summary of the GetReal WP1 experience of accessing registry data here)

Extrapolation of long-term survival
Results from each approach:

  • Base case: high level of uncertainty in the results in the long-term.
  • Approach 1: extrapolation beyond RCT IPD was difficult and there was high a level of uncertainty in the results beyond 4 years; addition of simulated IPD from RWD may further reduce uncertainty but there were some technical concerns.
  • Approach 2: blending different proportions of the RCT data with the re-created IPD registry data resulted in a reduction of uncertainty but the similarity between the populations in the RCT and registry data was questionable.
  • Approach 3: uncertainty was further reduced by using a combination of both sources of RWD and RCT.

More experience using combinations of RWD and RCT data with access to better quality IPD registry data is needed to establish the accuracy and validity of the approaches explored in this case study.

Social media to collect patient perspectives
Social media provides a valuable tool in assessing patient perspectives on QoL. However, results are difficult to generalise to the broader melanoma patient population. Differences also emerge between what patients and carers consider important for QoL. Available melanoma-specific QoL questionnaires do not seem to correlate fully with what patients view as important.

What do stakeholders say?

Extrapolation of long-term survival
Stakeholder feedback on this case study was obtained through a workshop. The different approaches were considered to be acceptable in theory, but there were concerns about the level of acceptance of the assumptions made (which needs to be discussed between relevant stakeholders) and, even more important, the low quality of registry data available for this case study, including the lack of IPD.

There were serious concerns about the use of aggregate registry data, as well as the external validity of the registries that were available. However, the group generally felt that these methods of incorporating RWD could be considered acceptable in some situations. The group also felt that more work on further case studies is needed, and that the assessment should include whether IPD from registries would improve the analyses.

The interpretation of the results from any of these analyses using RWD is dependent on the transparency and communication of the biases and limitations associated with the data.

See more feedback from the workshop here.

Social media to collect patient perspectives
Stakeholders expressed that, despite some issues with validation of results based on social media, and their generalisability, this data source may complement more traditional research methods. However, opinions remained divided as to whether this new source of data can be used in practice in the near future.

See more feedback from the workshop here and more stakeholder feedback on using of social media as a source of data here.

Key contributors

Amr Makady, Zorginstituut Nederlands (ZIN)
Rachel Kalf, Zorginstituut Nederlands (ZIN)
Michael Lees, Bristol-Myers Squibb
Prof. Keith Abrams, University of Leicester