Data Detectives Track Emerging Drug Resistance: WWARN

The issue: Antimalarial drug resistance

Over the past fifty years, malaria researchers have watched in horror as resistance to the standard antimalarial drugs chloroquine and sulfadoxine-pyrimethamine has spread across the globe. In 2007, resistance to artemisinin, the most effective drug at that time against the deadliest form of malaria, was first reported in Cambodia, and it has since spread to Thailand, Vietnam, Myanmar, Laos, and South China. Today, there are early reports that artemisinin resistance might have reached India.

In 2009, scientists, public health agencies and policymakers established the WorldWide Antimalarial Resistance Network (WWARN) to create a global database to track and understand this worrying trend. WWARN’s goal is to generate the tools and evidence necessary to track antimalarial drug resistance, understand the factors affecting the efficacy of these drugs, and develop new ones, all by re-using individual patient data gathered by researchers, NGOs, and pharmaceutical companies in more than 70 malaria-endemic countries.

A new approach: Repurposing data

In the field of poverty-related infectious diseases, data is scarce and scattered across institutions around the world. To solve this problem, WWARN isn’t simply extracting information from existing publications and aggregating it into traditional meta-analyses. (Although these studies are useful, they can sometimes be misleading because of publication bias and variation in methodologies between the different studies.) Instead, WWARN has been collating and standardizing anonymized individual patient data (IPD) from many unrelated trials and studies conducted across the endemic regions. The aim is to integrate all this information into one single dataset, thereby increasing the statistical power needed to answer key questions in malaria research.

“When you have access to individual patient data, you get a level of granularity about what happened to those particular patients that you can’t extract from a typical, aggregated meta-analysis,” says WWARN director Philippe Guerin, Professor of Epidemiology and Global Health at the University of Oxford. “You can identify patients who might be considered outliers in a single study, but who may be representative of a subgroup that responds differently to a disease or treatment compared to the general population. When we have enough of them in these individual patient meta-analyses, we can figure out if they really do have a different profile and then start to understand the implications of this.” For instance, this data makes it possible to study the efficacy of particular drugs for particular subgroups or key target populations—malnourished children, for example—who may be underrepresented in individual trials.

WWARN’s repository now holds more than 180,000 individual malaria-patient data files from around 600 different databases. These files, including clinical, in vitro, molecular and pharmacological data as well as data on the quality of medicines, are freely available for researchers to re-use.

Barriers to data-sharing

Historically, global health researchers, policymakers and those running disease control programs have had a relatively poor track record of collaboration with one another, and data-sharing was not common practice within the malaria community. Although many scientists recognized data-sharing’s potential to facilitate global surveillance of antimalarial resistance, it was difficult to convince the wider community of its value. “We approached investigators and said: ‘Would you like to share your data and trust us to do something good with it?’, and this approach did not work,” recalls Professor Guerin. “Just sharing data for the sake of sharing data was not sufficiently convincing for them.”

Solutions: Finding the right approach

Progress accelerated when WWARN began to focus on concrete research questions. “It was only by articulating the kind of research questions we wanted to answer—and making the investigator part of the reflection and refinement of that question—that we provided an incentive for them to engage and share data with the platform,” Professor Guerin explains. “Data sharing is not a goal but a means to generate new scientific evidence, and it has to be done in an equitable manner to be successful.”

WWARN has now developed a model in which dataset contributors are invited to help define a question to be addressed, join a study group, contribute data for secondary analyses, and co-author the resulting publication. Professor Guerin advises identifying a meaty, stimulating question which cannot be answered by a single study or small research consortium, and which piques the interest of the research community more generally.

Even with this model in place, it took time for the WWARN investigators to assemble the data researchers needed to do their work, and this created additional stress. “We were under pressure from our funders and stakeholders to answer the questions more rapidly, while still trying to convince investigators to share their data, and we knew that without a critical mass of information we would not be able to answer that question most effectively,” says Guerin. “One lesson that we’ve learned is that it takes time both to convince people and to do this kind of retrospective work, but when you know you’re on the right track then it’s easier to keep going,” he says.

Impact: Improved antimalarial therapy for children

WWARN and its dataset have already made a difference. For instance, they have helped researchers optimize the dosage of artemisinin-based combination therapies (ACTs) for young children. Even though children under five are the most vulnerable group affected by malaria, dosing studies are usually conducted in adults and the results extrapolated to children—but by pooling data on 7,072 individual patients collected from 26 different studies, WWARN showed that young children of a certain weight range with uncomplicated malaria were much more likely to fail treatment with one of the newly registered ACTs, dihydroartemisinin-piperaquine (DP) because they were receiving a sub-optimal dose. “For one particular weight band of kids who were receiving a lower dose, we found that they had a four-times higher risk of treatment failure,” says Professor Guerin. The study was published in PLoS Medicine in 2013. Subsequent modeling by the group informed a new proposed dosage regimen for children, a recommendation the World Health Organization adopted in 2015.

WWARN has also been investigating the efficacy of ACTs in malnourished children. Although many studies have been published on this issue, their conclusions have been mixed. By pooling data from 14,327 individuals, WWARN found that malnourished children have a higher risk of treatment failure with the common ACT artemether-lumefantrine. “We believe that this is extremely important, and it suggests that treatment regimens should be tailored to a child’s condition,” says Guerin.

At the request of health communities working on specific infectious diseases, the WWARN model has now expanded into other disease areas as the Infectious Diseases Data Observatory (IDDO), also led by Professor Guerin and based at the Centre for Tropical Medicine & Global Health at University of Oxford. IDDO applies the methods and programs WWARN developed for malaria to other poverty-related infections, including visceral leishmaniasis, Chagas disease, schistosomiasis, soil-transmitted diseases, and Ebola.

For many neglected tropical diseases that affect the world’s poorest people, the scarcity and far-flung nature of the available data has made it difficult for researchers to make progress. In this context, Guerin and his colleagues at WWARN believe that data-sharing has the potential to make a huge difference. Indeed, Professor Guerin says, “the rarer the data, the greater the value of this kind of approach.”