In Focus

Turning big data into personalised diabetes care Researchers are beginning to harness the power of big data to improve research and care for patients with diabetes. Dara Mohammadi reports. ”Big is beautiful” isn’t an expression often uttered by diabetes and obesity researchers. Unless, that is, they are talking about data. New ways of capturing and analysing patient data mean researchers can scour bigger and bigger datasets, with the power to tease out the often subtle effects genetic and lifestyle interactions can have on a person’s risk of diabetes. The question is, with so much data and the need for collaboration between all the people involved in gathering it, what’s the best way to harness its full potential? And how do we ensure that the outcomes improve patients’ lives? In September, Nick Wareham, codirector of the Institute of Metabolic Science at the University of Cambridge, Cambridge, UK, chaired a meeting on global data for diabetes and obesity research at the 51st annual meeting of the European Association for the Study of Diabetes in Stockholm, Sweden. It was a meeting to discuss the European Union-funded InterConnect project, which aims to use a big-data approach to improve interventions and treatments for both type 1 and 2 diabetes by studying the interplay between genetic risk factors and lifestyle. “Although there’s heterogeneity within diet, physical activity, and disease outcome within Europe, the variation between countries and continents around the world is much greater”, says Wareham. “The challenge is how do we move from effectively studying variation within populations in Europe to studying variation between populations, bringing in data from researchers in all different continents around the world?” Wareham outlines the three existing potential strategies, each not entirely suitable for a project of this scale. One option is to do a big, international

study from scratch, but this would be too expensive and time-consuming. The second is to share the results from separate analyses of cohorts, an approach viable for studies of, say, genome-wide associations, but one that would be too administratively burdensome for research teams in a project as complex and nuanced as InterConnect. The third would be to have a central deposition of data for analysis, but data-transfer laws across borders would preclude this. “So the InterConnect project proposes a different approach”, says Wareham, “one around the notion of federated meta-analysis of data—the data stay where they are but you take the analysis to them using a particular type of computer infrastructure.” In this instance it’s the European Union-funded BioSHaRE tool, a strategy used by other projects, such as the Enhancing Neuro Imaging and Genetics through Meta-Analysis (ENIGMA) Network, which aims to crack the genotype–phenotype relation in complex neurological and psychiatric disorders. To take advantage of the data available worldwide, says Wareham, researchers must get over the much vaunted call for sharing of data—it’s an administrative minefield when dealing with personal medical data. “What we actually want is the optimisation of information that exists”, he says. “Data sharing is just a means to an end.” He also feels that a culture change has to happen if we are to take advantage of these new opportunities of globalscale research. “Real-world researchers have to change. Funders have to change how they fund research and how they think about optimising the use of existing data. And stakeholders need to say this is what they want their data to be used for.”

www.thelancet.com/diabetes-endocrinology Vol 3 December 2015

The first question that InterConnect will ask will be how a mother’s physical activity affects birthweight and adiposity in her newborn infant. Wareham’s own work is in trying to understand how innate susceptibility, through genetics or early developmental exposures, affects disease risk in later life. He is a coordinator of the InterAct Study, which pulls together data from EPIC-Europe (European Prospective Investigation into Cancer and Nutrition) in ten European countries. InterAct is designed to investigate the interaction between genetic and lifestyle behavioural factors and risk of type 2 diabetes and has assembled more than 4 million person-years of follow-up in more 300 000 people. Since baseline in the 1990s, there have been more than 12 000 incident cases of diabetes. For each patient, Wareham and his colleagues have clinical data and blood samples, allowing assessment of standard biochemical predictors and nutritional biomarkers. With genetic information overlaid, he says, InterAct is a powerful source for both the development of public health recommendations and generation of future, smaller, hypothesis-driven studies, a necessary add-on to any large-scale venture. Osama Hamdy is an endocrinologist at Joslin Diabetes Center and Harvard Medical School, MA, USA, who is aware of the benefit of amassing big databases, not only to shape new questions about the disease, but to also improve clinical practice and patient care. Joslin Diabetes Centre started collating electronic medical records more than a decade ago: their database now numbers more than 8000 patients. This system allows clinicians to quickly identify high-risk

Published Online November 9, 2015 http://dx.doi.org/10.1016/ S2213-8587(15)00429-5 See more on mobile applications for diabetes management see Comment page 921 For the InterConnect project see http://www.interconnectdiabetes.eu/ For the BioSHaRE tool see https://www.bioshare.eu/ For the ENIGMA Network see http://enigma.ini.usc.edu/ protocols/imaging-protocols/

935

Andrzej Wojcicki/Science Photo Library

In Focus

For more on iPhone versus Android users see http://www. forbes.com/sites/ toddhixon/2014/04/10/whatkind-of-person-prefers-aniphone/

936

patients and provide them with the care they need. They can also assess their own performance by checking what proportion of patients have met, say, their BMI or cholesterol targets. “We’ve been trying to encourage other centres and practices across the country to do the same”, says Hamdy, “but it’s baby steps right now while everybody transfers over to electronic records and starts using systems that can communicate with each other.” Once that happens, he says, high-risk patients in smaller centres can be easily referred for specialist care. It could also speed up recruitment for trials. Though, he points out, such plans are pie in the sky at the moment, in view of the stringent laws around data sharing. “Sharing information about patients must be extremely safe and shouldn’t compromise patient privacy”, says Hamdy. “We need a way to have secure transaction, like with secure bank transfers of money online. We to have the same level of secure transaction between hospitals and centres for this to work.” Across town from the Joslin Centre at another Harvard-affiliated hospital, Massachusetts General Hospital, MA, USA, Stanley Shaw, cofounder and codirector of the hospital’s Center for Assessment Technology and Continuous Health (CATCH) is experimenting with a different type of big data project with a potentially even wider reach. In March, 2015, Shaw launched GlucoSuccess, an iPhone app for

patients with type 2 diabetes, on Apple’s then-brand-new research platform, ResearchKit. As with the other four apps launched with the platform that were developed by other research teams—for heart disease, Parkinson’s disease, breast cancer, and asthma—Shaw’s GlucoSuccess lets users input data about their disease, either actively though answering questions on the touchscreen or passively via the phone’s on-board accelerometer and gyroscope, which can measure physical activity. The researchers have recruited more than 5000 participants and plan to discover more about how specific behaviours relate to blood glucose in individuals and populations. “It’s a step up from the way these types of data are typically collected in large observational studies”, says Shaw, giving the National Health and Nutrition Examination Survey (NHANES) or the Nurses’ Health Study in the USA, which use written questionnaires, as an example. “They are limited because the questionnaires are taken periodically and are prone to recall bias, which can really lose resolution in your data. Here [with Apple’s ResearchKit] we have the capacity to put a data gathering instrument literally in the palm of people’s hands in form of a smart phone, so that they can contribute data every day and in real-life situations.” This type of research is not without bias either. For the time being at least, trial participants are limited to only those with an iPhone, a population that polls have shown to be likely to be richer and better educated than Android users. Also the veracity of data gathered is under question— how, for instance, do researchers know that the person entering the data is who they say they are? Shaw and his colleagues are taking these problems in their stride. Availability of such research apps on all smartphones is surely inevitable, and confidence in the reliability of data can be improved by making sure that the participant has a vested interest in

their data being correct. “If we’re going to move into this way of doing science, we have to keep the study population motivated and engaged by making the process mutually beneficial”, says Shaw. “We can offer them analysis of their own data, which can help them better manage their disease. Based on their glucose readings, we can send them updates showing how what they did or what they ate affected their glucose control. The idea is to make research a more two-way street.” Shaw thinks that such approaches will help researchers refine some commonly held assumptions about how people’s lifestyles can affect their disease. A preliminary look at GlucoSuccess data, he says, already shows subsets of patients that “pop out” from the data. “Some people show very strong inverse correlation between the number of steps they take and their glucose level, and in others the correlation is less obvious.” Such heterogeneity, Shaw says, comes from collecting richer datasets than were available before, with data collected while people are leading their daily lives, not just when they are coming in for their annual checkup or research visit. He thinks such approaches can do the same for epidemiology as high-throughput sequencers did for genetics. “The genetics has really speeded ahead”, he says, “but resolution of phenotypic measurements and behaviours and environmental exposure have really lagged behind. If our ability to make these broader measurements could catch up to the genetic measurements, and we can do cross comparisons between datasets, then the genetic measurements would become much more powerful.” It’s the effect that big data can have on patient care that excites him most. “I think we can go beyond that one-sizefits-all epidemiological approach to understanding how a disease affects the lives of our patients as individuals.”

Dara Mohammadi

www.thelancet.com/diabetes-endocrinology Vol 3 December 2015

Turning big data into personalised diabetes care.

Turning big data into personalised diabetes care. - PDF Download Free
565B Sizes 1 Downloads 10 Views