The SCOPE Summit for Clinical Ops Executives is an annual conference to drive collaboration and innovation in the clinical research community. I attended the 14th annual SCOPE Summit on February 6-9 of 2023 to connect with my colleagues and other industry experts about what technology advancements are driving growth in the life sciences industry. Here are the three takeaways, I derived from the conference:

1. Digital Measures and the Use of AI/ML 

One theme of discussion at this conference was around using artificial intelligence and machine learning to:

• Accelerate study setup
• Generate content
• Drive automation

Experts discussed approaches to start digitizing protocol and creating CSRs from the digitized protocol using AI/ML programs.

Another approach demonstrated how to use modern tools to collect visit data. For example, it explored how to record a video to send to your doctor to review progress of neurological diseases and how to design trials that can support both decentralized visits and site visits. This would give life sciences organizations the ability to significantly enhance trial progress and identify safety and efficacy issues in an expedited manner. 

2. Pharma Companies are Taking a Leadership Role in Developing New Technologies

Previously, it was standard for technology vendors and IT companies to lead the development of new technologies in the life sciences industry. These organizations would be responsible for identifying potential advancements and capabilities. Based on the conversations that took place at this conference, it is clear that pharma companies are starting to take a more active role.

Pharma companies are leading and delivering solutions, and they want findings and results to be open source. They are telling the technology vendors what they need and how they want to develop it.

They are realizing the importance of including business users from the beginning of the development process, and they are seeing the value of change management.

3. Acceleration and Collaboration in Life Sciences

The pandemic has completely changed the life sciences industry. The industry has dramatically accelerated. Life sciences is no longer an industry that is slow to change.

Post-pandemic life sciences organizations are getting ahead of the curve. They are open to new ideas. They are even open to collaborating with other pharma companies. Life sciences leaders at SCOPE were discussing how they can collaborate on training their AI/ML models and data models so multiple companies can use similar machine learning models. Each organization wouldn’t have to spend time training their own AI/ML models.

Collaboration in the life sciences industry is increasing exponentially.

Decentralized trials and digital measures are transforming how we collect information and data. With such rapid change, it’s an exciting time to be in the life sciences industry, and I expect to see 10 times as much technological advancement at the next conference.

Life Sciences Leaders Turn to Us 

Life sciences organizations’ ability to accelerate transformation is crucial to succeeding in a highly competitive, highly regulated, and quickly-evolving landscape. Life Sciences leaders rely on us for strategic, industry, and technical expertise to achieve their missions in a technologically advancing industry. Our thought leaders are here to support you in achieving business goals and solving your most complex challenges. 

Our Life Sciences practice has a global footprint, with locations across the globe to meet your skill and budget needs at the right time and in the right place. We work seamlessly with our domain experts across Perficient to best execute for our clients’ outcomes. Our comprehensive end-to-end support keeps your organization efficiently moving forward. Extend your capacity with our expert solutions and add day-to-day value where you need it most. 

Have questions? Contact us to discuss your organization’s specific needs. 

–Ruby Lin and Nicolas Frantzen contributed to this blog.

Overview

In life sciences, like many other industries, knowledge is power. Historically, the main challenge has been around sourcing, organizing, and meaningfully surfacing this knowledge. To solve this challenge, one needs to look at innovative and scalable ways, such as AI, to find and organize a wide variety of information from both internal and external sources.

An intelligent knowledge repository that provides relevant and accurate information for consumers, specifically for drug-related information, will require different layers of AI models that can produce predictable and accurate results while quickly adapting to the complex nature of the information.

The dynamic nature of today’s content means we have to work with both structured or unstructured data. Unstructured data (documents, articles, websites, newsletters, blogs, etc.) presents an additional challenge in that it is not readily consumable by systems (other than free-form text, keyword search, etc.). We have to extract the information and organize it first. To do this, we need to curate and structure that data before it can be stored in the knowledge repository, more specifically, a knowledge graph.

Knowledge graphs present significant benefits over traditional databases and tabular schemas in that they allow you to store and maintain relationships between data elements. Once the information is curated, you can use specialized AI models against your knowledge graph to find interesting patterns and predict relationships between pieces of information stored.

While the list of potential applications for this type of solution is vast, here are several use cases that are specific to the life sciences industry:

• Getting country-specific regulatory requirements for submitting a product for marketing approval in a specific geographic region
• Searching and retrieving literature required for authoring a document
• Understanding the state of current research and clinical trials
• Surfacing real-world evidence (RWE) data
• Addressing a product inquiry from a health care provider

And the list goes on.

While these use cases describe some real and complex challenges, there are readily available solution patterns that can be used to streamline the extraction of information.

Let’s dive further into the process of curating content and storing information into a knowledge repository.

Content Sources

The first consideration is the location of the content sources. The content could be internal or external to the company. The sources could be internal product documents, data from internal systems, medical literature documents, regulatory documents from health authority websites, intelligence from trade associations, etc.

The low-hanging fruit for gaining information would be the marketing or literature references for the drugs and medical products. These are pre-curated contents, and the company can digitalize and organize the information in an appropriate knowledge repository.

Secondly, if there is any existing content, this may need further curation before incorporating it into the repository. The common regulatory agencies, trade associations, and medical literature would provide valuable sources of information.

Due to the volume of external documents, web crawlers and other automated solutions must be leveraged (see how Perficient solved this challenge with Handshake) to automatically source documents and run them through an NLP pipeline that will enrich and organize the information extracted.

Knowledge Repository Example: Ontology

Below is an example of how structured and unstructured content (i.e., regulatory requirements documents) can be stored in the form of a graph repository.

As these documents can be from various health authorities, the ontology includes the contact details of the source (regulatory authority in this example). A hierarchical structure is defined to store documents, and their sections and a recursive hierarchy keeps sections within sections.

Each section is categorized by topic, definition, associated with a specific requirement. As the documents are curated, the ontology can also be automatically enhanced to ensure that the documents can be curated and stored using the ontology. Once the ontology is created, extracted documents can be stored following the ontology.

An AI model can then be trained to auto-curate the documents, based on the defined ontology, to look for and understand those important concepts and information elements necessary to organize and store the documents within the knowledge repository.

Saudi Arabia IB as a Knowledge Graph

The diagram below shows a sample document – Saudi Arabia’s Investigators Brochure (IB) – stored as a knowledge graph using the ontology defined in the previous section.

Conclusion

Though the creation of the ontology is key to the content’s successful curation, AI/ML models can be trained to read any document and store it as a knowledge graph once the ontology is defined. Amazon Neptune, neo4j, Apache Cassandra are some examples of graph databases that are currently available. By combining those graph databases with an AI pipeline (composed of multiple ML and NLP models), you can build a scalable and efficient solution to source and organize information.

With knowledge properly organized, it is possible to tackle the second half of the equation: surface insights and retrieve that information efficiently to end users. Look for another blog post around that topic very soon!

Previously, I dove into how artificial intelligence helps review, and provide statistical analysis to data. The final blog of this series outlines how to be successful with an artificial intelligence implementation.

Setting initial expectations and not promising a magic bullet is a key factor in determining the success of an initiative that focuses on deploying AI to streamline the clinical data review and cleaning process.

Training the machine learning model will determine how accurate the results are. After every phase, evaluating the released functionality, reassigning priorities to backlogged initiatives, and releasing based on prioritized functionalities should be done and closely monitored. The adoption by business users will determine how successful these initiatives are. Other factors that will help in evaluating how success include:

• Accuracy and speed of data review
• Effort needed by humans to reach a milestone
• Improved user experience
• Adoption of the solution by the end-user community

Extending the Roadmap

AI technologies can be used for more initiatives within CDRP. Each of the questions that need to be answered in the initiatives could be addressed by ML models based on how well the models are trained. In addition to CDRP, there are other areas in life sciences – such as safety systems, clinical operations systems, risk-based monitoring systems, and data capture systems – where AI technologies can be used to enrich the data management process.

To learn more about how artificial intelligence – including machine learning – can be used by pharmaceutical and medical device companies to improve the clinical data review and cleansing process, you can download the guide here or submit the form below.

My last installment explained how artificial intelligence (AI) assists search, confidence scores, and data reviews. This blog dives into how artificial intelligence helps review plans, provide statistical analysis to data, and robotic process automation.

Review Plans

Prior to starting data review, each team has its own review plan. The data review team has a data review plan, the safety team has a safety review plan, the analysis team has a statistical analysis plan, etc. These review plans are standard across studies with few changes. An automated program can create review plans based on the metadata information in a study. The review plan will result in a list of tasks, which can be assigned to different user groups.

Based on the previously assigned tasks, the program can automatically assign the tasks and prioritize the tasks for each individual user. This, combined with the prioritized data review, will enable users to prioritize their tasks and complete the data review and the tasks from the review plans.

Statistical Analysis of Clinical Data

Statisticians analyze clinical trial data to understand the efficacy and safety of a drug. Machine learning can be used to assist statisticians in analyzing the data and looking for anomalies, such as:

• Why is it that the subjects in certain age ranges at site 1 are reporting fever in visit 3?
• Are there other sites that are reporting fever for visit 3?
• Is this related to the clinical trial, or is there an outbreak in site 1 region that is causing this anomaly?
• Can I compare this data with publicly available data to check for outbreaks?

There is enough data to train the ML algorithms to help in reviewing the data. Understanding how to use this data and how to perform more-effective exploratory analysis will help statisticians better understand the efficacy and safety of a clinical trial.

Robotic Process Automation

Validation is a manually intensive and time-consuming process. Some companies are considering using robotic process automation for validating their review platforms when new functionalities are added to it or when new versions of the software are released. This will ensure that the available functionalities in the current system are not impacted by new changes and no regression issues are introduced in new versions.

To learn more about how AI – including machine learning – can be used by pharmaceutical and medical device companies to improve the clinical data review and cleansing process, you can download the guide here or submit the form below.

Previously, I analyzed clinical data review platforms. This blog explains how artificial intelligence assists search, confidence scores, and data reviews.

Search

Data managers and reviewers log in to clinical data review platforms (CDRP), and slice and dice the data they want in order to review for missing, wrong, or inaccurate data. If they can search using natural language, they can spend more time on reviewing the data rather than creating complex search criteria. For example, users can write:

• “Show me all demography data where subjects are males, but pregnancy is yes.”
• “Show me all data from adverse events and concomitant medications, highlighting concomitant medications without corresponding adverse events.”
• “Show me all the severe adverse events reported for the third visit.”

Confidence Scores

NLP/NLQ (natural language querying) converts these texts to search criteria, which is then converted to an SQL query. The query is executed, and the results are returned in figure 1 (below). When ML algorithms understand clinical data, they can execute the search criterion and deliver the results. Information extraction with language translation (e.g., English to ML) can be used.

Extending this would include allowing audio input from users, converting audio to text, and then using NLQ to convert the text to corresponding queries. For non-English speaking users, NLP language translators can be used to achieve the same results.

Data Review Prioritization

When clinical data reviewers, medical monitors, statisticians, and safety reviewers are reviewing data in the CDRP, an ML model can analyze the data and apply statistical analysis to determine the probability of whether the data is clean or not. In addition, this model can be used to detect anomalies in the data. Users should be presented with data based on the probability that the data is clean or the data points that require their attention.

The data reviewers can prioritize their review activity based on the input from the ML models. It will be valuable for the users and will significantly reduce the time taken for data review. Users can set the threshold for the data points they want to review first, based on the statistical analysis done by the ML model.

To learn more about how artificial intelligence – including machine learning – can be used by pharmaceutical and medical device companies to improve the clinical data review and cleansing process, you can download the guide here or submit the form below.

My last blog discussed how AI assists to create human-computer systems. This next blog in the series analyzes clinical data review platforms.

While machine learning (ML) and natural language processing (NLP) functionalities are not currently available in today’s clinical data review platforms (CDRP’s), we know they offer tremendous benefits to the clinical data review and cleaning process.

When conducting clinical trials, data is collected from electronic data capture systems and from central and local labs. This data is then combined and stored in a data warehouse. The data is transformed into a format in which the data managers are familiar. Data managers then review and clean the data.

Once cleaned, the data is transformed into common data models (e.g., CDISC SDTM) and used for generating submission documents to the FDA. Machine learning models can be used to check patterns in data, and if there are irregularities or missing data, bring it to the attention of data managers for further review.

Data from prior studies is available for ML algorithms and models and can be learned from. Every clinical trial has certain milestones to achieve, and for each milestone, there are certain criteria to be met and associated documentation to be generated. ML models can be trained to understand if a trial is ready for a certain milestone. If not ready, it can determine the bottleneck, and it can make predictions for how long it will take to reach the milestone based on historical knowledge.

It can generate the documentation needed for a milestone and send it to humans for approval before submitting to the FDA. For future clinical trials, the algorithms can answer questions such as, “How long will it take for me to enroll ‘N’ number of subjects for an oncology study?” and “How long will it take to reach a milestone based on the trial protocol document?”

To learn more about how artificial intelligence – including machine learning – can be used by pharmaceutical and medical device companies to improve the clinical data review and cleansing process, you can download the guide here or submit the form below.

Previously, I discussed artificial intelligence (AI) enhancing clinical data review processes. This blog discusses how AI assists to create a human-computer system.

Humans and machines each have their own strengths. On the one hand, machines are good at processing and analyzing large volumes of data with high speed and accuracy. On the other hand, humans are good at making decisions based on data, interacting with other humans, and applying general intelligence to the data.

When it comes to reviewing clinical data, a human-computer system will perform better than either standalone method. The AI initiatives that revolve around clinical trials require humans and machines to work together.

Machine learning models can understand clinical data, do an initial analysis of the data, and perform an initial cleaning of the data using statistical analysis. Based on the clean/dirty probability, humans (data reviewers) can prioritize their review activity. They can search data using NLP and obtain the status of their activities using a self-documenting platform that gives them a current summary of the clinical trial state.

Analysis using machine learning models can provide more insight into clinical data and enable humans to determine the safety and efficacy of the trial.

To learn more about how artificial intelligence – including machine learning – can be used by pharmaceutical and medical device companies to improve the clinical data review and cleansing process, you can download the guide here or submit the form below.

Ensuring that clinical trial data is accurate and that clinical trials are safe and effective, is a time-consuming and a manually intensive process. Many life sciences companies have implemented home-grown or off-the-shelf clinical data review platforms and defined the reviewing and cleaning processes to be used with them.

Many of these systems have proven to be challenging to use, inflexible, and created frustration among users.

This guide discusses how artificial intelligence (AI) – including machine learning (ML), including deep learning (DL) and natural language processing (NLP) – can be used by pharmaceutical and medical device companies to improve the clinical data review and cleaning process. These technologies enable drugs and devices to reach the market faster and more safely and effectively.

To learn more about the clinical trial data process, you can download the guide here, or you can submit the form below.

Enhancing the clinical data review and cleaning process using available AI technologies, will enable pharma companies to release new drugs to the market following an effective process and ensuring the safety & efficacy of the drugs released.

During the past several years, many pharmaceutical companies implemented home-grown or off-the-shelf Clinical Data Review Platforms. They defined the processes followed by reviewers/monitors to review and clean data. Both options have their challenges with regards to ease of use, flexibility and complexity of the product. Clinical data review and cleaning is currently a time-consuming, manually-intensive process.

The Current Process

The current process is causing delays in meeting milestones and releasing of drugs with required validations and approvals. This process can be improved using well-trained machine learning (ML) and natural language processing (NLP) models. These models provide a user-friendly, self-documenting solution for various teams reviewing, cleaning and analyzing the clinical data. With the proposed initiatives to improve Clinical Data Review Platforms, pharma companies can analyze and understand clinical data with minimal effort and cost.

There are various functionalities that can help to search and analyze data, streamline tasks based on review plans, prioritize tasks based on milestone and provide current study status. For each, ML, NLP or a combination of ML and NLP models can be used.

I will be adding more posts with details on these functionalities. The use of AI technology will have a major strategic impact on focus and differentiation. Artificial intelligence will significantly reduce the cost and effort needed for reviewing data.