Pioneering work that transformed medicine and public health by better distinguishing cause from correlation has been recognised in a ceremony at Leuven University (KU Leuven) in the presence of HM King Philippe; the inaugural $1 million Rousseeuw Prize for Statistics was awarded by the King Baudouin Foundation to James Robins of Harvard University and to four colleagues who developed his work.
Established by Peter Rousseeuw, Professor of Statistics at KU Leuven, the biennial prize awarded on Wednesday rewards excellence in statistical research which has a significant impact on everyday life.
Insights from James Robins and his collaborators into “causal inference” permitted statisticians to move beyond observations that phenomena move in tandem – correlation – to indicating whether one is responsible for another - causality. Robins received half the prize, with the rest shared by Miguel Hernán (Harvard University), Thomas Richardson (University of Washington), Andrea Rotnitzky (Universidad Torcuato di Tella, Argentina), and Eric Tchetgen Tchetgen (University of Pennsylvania). All four were trained or deeply influenced by Robins and still work with him.
Does ice cream give you a rash?
Determining cause and effect from data can be a lot harder than it looks, certainly without empirical experiments, because several factors can occur together. As a simplified example, take jellyfish stings and ice cream sales. Both tend to rise and fall together. But jellyfish stings are also correlated with how many people swim in the sea on a given day. And how much power is used for air conditioning. Causal inference makes it possible to determine from data that bathing, and not air conditioning or ice cream, raises the risk of jellyfish stings. This is but an illustration, in fact causal inference is applied to important real questions in medicine.
Revolution in statistics
The laureates’ work has provided new insights and statistical methods for addressing central questions in epidemiology. For example, what is the effect of a long-term medical treatment? And, if beneficial, what treatment strategies are optimal? One result has been guidelines on when to best initiate antiretroviral therapy in people with HIV.
Robins showed that it was especially difficult to interpret data from studies that measured exposures or treatments repeatedly over time. This is due to feedback mechanisms. HIV patients who start antiretroviral treatment are on average sicker than patients who do not start treatment. Not factoring that in could lead to interpreting mortality data to mean the treatment was doing more harm than good. Imagine, then, that antiretrovirals do boost immunity in some. But those patients will then receive a different treatment afterward which causes a feedback issue in the data, making it harder to draw correct conclusions. Robins solved this methodological problem in a series of ingenious papers in the 1980s, laying the foundations for a long line of innovations by all five laureates, which has helped launch a causal revolution in statistics.
Impact ranging from epidemiology to economics
Their work has also influenced economics, psychology and other fields. In a number of important cases, it has demonstrated that disparities between conclusions drawn from empirical and non-empirical studies are due to the use of outdated statistical methods. Among areas where the impact has been significant are in understanding the effects of post-menopausal hormone therapy on coronary heart disease, of statin therapy on cancer and on the benefits of anti-inflammatory therapy for Covid-19 patients
The work honoured by the Rousseeuw Prize has completely transformed the way in which statisticians, epidemiologists, and others infer the effects of interventions, treatments, and exposures to potentially harmful substances. It has greatly improved the overall reliability of causal analysis in medicine and public health, with great benefit to society.
Peter Rousseeuw, who founded the prize to help ensure statistics attracts the talent and investment which the Nobel and other prizes offer to other branches of science, said the inaugural laureates had debunked long-held assumptions that it was impossible to determine causation without controlled experiments:
“They have proved that it is possible, and they applied their innovative techniques to important questions, such as the effect of post-menopausal hormone therapy on heart disease, and the effect of statin medication on cancer. Afterward, clinical trials have confirmed their findings,” said Rousseeuw. “Their work can also be used in the fight against epidemics, and for public health, for instance to decide whether certain environmental substances should be avoided or banned.”