Apple launches ResearchKit aimed at improving clinical trial data collection – March 10, 2015

In yesterday’s live-streamed event, Apple announced the launch of ResearchKit, an open source software framework aimed at improving clinical trial data collection. Apple’s goal is to make it easy for healthcare researchers and developers to create apps that take advantage of the iPhone’s features and sensors – the lofty (if a bit vague) goal is “to gather new types of data on a scale never available before.” ResearchKit will be available to developers next month – more details here.

Apple’s three-minute video lays out the concept very well, suggesting that “the methods for conducting medical research haven’t changed in decades.” There’s certainly at least a bit of exaggeration in there, but no question things need to become more efficient. Notably, Mount Sinai’s Dr. Eric Schadt mentions diabetes and obesity in the video’s first sentence, two examples of diseases where it’s hard to understand what’s going on. ResearchKit brings potential to address many limitations of clinical trial data collection:

• infrequent data (“going from data once every three months to data once every second”);
• gaining a more real-world understanding of disease (e.g., physical activity passively tracked through the iPhone’s sensors);
• limited scale (“hundreds of millions of people around the world have an iPhone”);
• patient education (some apps allow patients to track their own data and correlate symptoms with actions);
• rapid iteration (e.g., researchers could send a new survey every day); and  
• signing up for studies and providing consent.

Notably, partner Massachusetts General Hospital already has a diabetes app and clinical trial running on the platform: GlucoSuccess. The app asks people with type 2 diabetes and prediabetes to record and track activity, diet, blood glucose measurements, weight, and waist size. The aspirational idea is to “create an unprecedented crowd-sourced database of health behaviors and glucose values.” ResearchKit does read data from Apple’s HealthKit, which itself seamlessly pulls health/fitness information from over 900 apps (e.g., Jawbone activity tracker, Withings scale, Epic’s MyChart EHR). However, since most patients don’t have a glucose meter that sends data right to HealthKit, it’s going to be manual logging in these early days – to our knowledge, only the iHealth meters send glucose values to HealthKit. The hassles of manual data collection are not to be understated, though we do believe this will change over time. Notably, the MGH’s study’s eligibility questions and informed consent are built right into the GlucoSuccess startup screen – what a novel way to enroll and conduct a clinical trial!

It’s early days for these Apple platforms, though if anyone knows how to change an industry by building a digital marketplace, it may be Apple – iTunes and the App Store were revolutionary, and it’s hard to now imagine a world without them. With that in mind, we believe ResearchKit and HealthKit have meaningful potential to dramatically change the way data is collected for research, and beyond that, shared with other apps in the digital ecosystem. What is less clear is the legitimacy of such democratized data collection: will traditional sponsors and benefactors of studies (NIH, JDRF, Helmsley Charitable Trust, device manufacturers, HCPs, payers) embrace these trials and the insights they produce? What will the FDA say about such data collection?

• The scale and democratization that ResearchKit could enable is bound to bring up arguments: data quantity vs. data quality; explanation vs. prediction; RCTs vs. multi-armed, contextual, adaptive clinical trials; sampling bias; privacy; and others. Like any new technology, we imagine it will have pros and cons, and researchers will learn to use it over time.
• Notably, ResearchKit brings a framework for recruiting and obtaining informed consent, something researchers struggle with. It is an exciting prospect in that research could be done more quickly and a scale that previously wasn't possible.
• Having real world, continuous data streams could give researchers never-before-seen insight into diabetes self-care patterns, adherence practices, and life routines. For instance, even in very motivated adult type 1 pumpers, some clinicians notice that ~10-15% of patients will occasionally miss boluses. Those on MDI likely miss more, though it’s hard to get reliable data on that front. Combining Bluetooth-enabled insulin pens (if they ever come out!) and mobile phones for data collection would give a clearer view of how patients are doing in the real world.
• As a reminder, Dexcom’s Gen 5 platform is expected to send data to HealthKit, which will open more possibilities for seamless glucose data collection and an ecosystem of third party apps. Gen 5 was submitted to the FDA in February and Dexcom is cautiously optimistic for FDA approval by the end of this year. We could imagine patients in clinical trials wearing Dexcom CGM, whereby glucose data would seamlessly go to HealthKit and ResearchKit. We’re not sure if Medtronic’s Bluetooth-enabled Guardian Mobile system will be compatible with HealthKit, but assume it is possible.
  
  • Particularly exciting is the potential for an ecosystem of third party apps that leverage freely flowing diabetes device data. For example, an exercise app that takes glucose data and makes recommendations or analyzes what’s happening. Or a food app like Tidepool’s Nutshell that can track bolus and glucose history and help patients improve mealtime dosing. The possibilities are endless, as noted by Dr. Joe Cafazzo and Howard Look at November’s FDA meeting on Device Interoperability and Bolus Calculators. The challenge is getting all the devices to talk to each other and drop data in the same place. One key difference between HealthKit and ResearchKit is that the latter is open source and could potentially be ported to other platforms.
• ResearchKit and HealthKit do bring questions about the data model used for diabetes. Collecting a single time-stamped glucose value or insulin dose is fairly easy (though still possible to get wrong, as we've seen with Apple’s first version of the Health app, which excluded mmol/l). However, getting a data model that accurately collects and represents the myriad of ways that pumps do boluses and basals is where it gets really complicated. To the extent that Apple adopts a robust diabetes data model that can represent all of these ways of delivering insulin, ResearchKit + HealthKit could be very significant innovations.
• Today’s move from Apple continues a theme of big tech companies moving into healthcare (see our 2014+2015 reflections piece).
  
  • Apple: Aside from yesterday’s news, we been following the news of the Apple Watch since it was first unveiled last September. As a reminder, Dexcom plans to offer apps for the upcoming Apple Watch, which is slated to launch on April 26.
  • Google: The tech giant recently changed the way it displays health information in Search, making information much more accessible and convenient. In addition, work continues on the smart contact lens with Novartis, though it’s unclear how much focus the glucose monitoring application is getting. Novartis’ 4Q14 call seemed to shift focus away from the diabetes application towards broader ophthalmologic indications. Last July, Novartis CEO Joe Jimenez said that a glucose sensing prototype would ideally be available for R&D reviews by early 2015, and the product could be on the market in about five years. Google has also rolled out Helpouts, a novel platform over which patients can discuss medical issues (a number of people are already offering services related to diabetes) with providers via live video.
  • Facebook: Last October, Facebook expressed interest in exploring patient communities and preventative care apps.      
  • Amazon: Last September, it was reported that Amazon executives had met with multiple top-level members of the FDA. Speculation continues that Amazon may enter healthcare – everything from offering services to clinics to selling implants has been hypothesized.
  • Microsoft: In January, the Seattle-based giant teamed up with a local JDRF to host a Nightscout installation workshop. The company was also linked to rumors that its smartwatch would include a glucose monitor, though these were ultimately dispelled.
  • Samsung: Samsung announced a partnership in February 2014 with UCSF to accelerate validation and commercialization of promising new sensors, algorithms, and digital health technologies for preventive health solutions. Meanwhile, WellDoc announced a multi-stage collaboration with Samsung last November to improve the lives of people with type 2 diabetes and explore next generation diabetes devices and product offerings. The partnership brings a natural synergy between WellDoc’s FDA-approved Mobile Prescription Therapy (BlueStar) and Samsung devices and new FDA cleared healthcare platform.

Close Concerns Questions

Q: To what extent will ResearchKit improve the efficiency and quality of data collection in clinical trials?

Q: To what extent will the traditional research decision makers and benefactors take ResearchKit and HealthKit seriously?

Q: Will FDA take the data gathered from ResearchKit and HealthKit seriously? Where is the right balance between data quality and data quantity?

Q: Will more diabetes devices become compatible with Apple’s HealthKit? Will Apple’s HealthKit emerge as a standard platform, similar to iTunes?

Q: How could ResearchKit be used in pharmaceutical trials?

Q: In what ways could ResearchKit expand our knowledge of the psychosocial realms of diabetes and obesity?

Q: What are Apple’s plans for diabetes? What about Google, Facebook, Amazon, Microsoft, and Samsung?

Q: How will the FDA’s downclassification of secondary display of CGM data affect the pace of innovation? Will an ecosystem of apps actually emerge?

-- by Adam Brown and Kelly Close