Large Cohorts: Toward Routine Databases for Public Health Science

In 1985, Per Magnus was hired by the Norwegian medical and social science research councils to figure out how epidemiological research could best be developed in Norway. Magnus was a young physician with a PhD, a specialty in medical genetics, and an extraordinary talent and interest in science. To complete his mission, he chose to travel to six countries to interview epidemiologists. At Harvard University, he spoke with Walter Willett.¹ In his report from 1986, Magnus wrote that Willett was responsible for implementation of a large prospective study on the effect of nutrition on disease later in life. This study, according to Magnus, would recruit 60 000 male health professionals. They would receive a comprehensive questionnaire on nutrition and be followed up every other year onwards. The study was modeled on the Nurses' Health Study (NHS), a similar study of 120 000 nurses who had been followed successfully biannually since 1976. According to the report, Willett was convinced that if one wanted progress in epidemiological research on chronic diseases, it was vital to develop large, prospective cohort studies. Such cohorts would provide much stronger evidence than smaller inconclusive studies. Therefore one should not scatter and waste funding on small studies, said Willett.

THE NORWEGIAN MOTHER AND CHILD COHORT STUDY

Magnus went on to do just as Willett had recommended. He was one of the pioneers of the Norwegian Mother and Child Cohort Study (MoBa), and the only one who continues to have a leading role in the study after nearly 25 years.

After a long and demanding period of planning and fundraising, MoBa initiated recruitment in 1999. When the last child was born into the study in 2009, there were altogether about 284 000 participants including children, mothers and fathers.² Whereas NHS has recruited health professionals, MoBa is population-based, and pregnant women and their partners were recruited from all over the country. Follow-up continued throughout pregnancy and later at intervals following birth, mainly by questionnaires, additional data collection in selective substudies, and linkage to routine registries such as the Medical Birth Registry of Norway with data on all pregnancies and births in Norway, the Norwegian Patient Registry with data on all hospital and out-patient clinic diagnoses, and the Norwegian Prescription Database with data on all prescriptions. Eventually, 41% of invited women consented to participate. This included more than 114 000 children, 95 000 mothers, and 75 000 fathers. The questionnaires query health, dietary intake, socioeconomic factors, child development and behavior, and other factors relevant to child and parental health. Data from ultrasound examinations were collected from midwives during pregnancy. Blood samples (plasma, full blood, and DNA) were obtained from both parents during pregnancy and from mothers at birth. A urine sample was collected from the mother during pregnancy. From the child, a blood sample (plasma, full blood, DNA, and RNA) was taken from the umbilical cord right after birth. By mid 2016, about 60 000 participants have been genotyped; of which most are mother, father, and child trios.

NHS inspired MoBa not only as a large, prospective cohort, but also with its focus on nutrition. MoBa researchers drew from knowledge, approaches and questionnaires in NHS when the MoBa food frequency questionnaires were designed.³

SUCCESS AND FAILURE

To establish and run large cohorts is a daunting enterprise. In her recent book on birth cohorts, Helen Pearson describes how it requires risk taking, immense endurance, tolerance for uncertainty, and considerable personal stamina, stubbornness, and visionary capacities to lead such projects.⁴ And even with all that in place, funding may be difficult to obtain, recruitment may be slow and low, and politics may be too complicated. A cohort may fail altogether, such as the US National Children's Study and the UK Life Study have done. And if it does not fail, survival on a shoe string budget may be necessary for unpredictable periods of time when the cohort is out of fashion or funding and leadership is at a low. But when they thrive, large prospective cohorts do deliver knowledge, better and more comprehensive than most other sources of empirical data. Just the way NHS does. We believe that MoBa also is well on its way to prove Willett’s points from 1986.

TOWARD INFRASTRUCTURES FOR SCIENCE

In the wake of the publication of the human genome project in 2001, the scientific opportunities of large, prospective, population-based cohorts has been promoted repeatedly by prominent leaders in science.⁵ The concept of cohorts has reached beyond epidemiology and public health into genetics and molecular biology, and even into the domains of randomized controlled trials where many now acknowledge the scientific value of registry-based epidemiology as a complement to randomized controlled trials, often as the only feasible alternative, and sometimes as a superior approach to observe reality and support causal inference. A recent grand cohort effort is the Precision Medicine Initiative launched by President Obama in 2015 and followed up closely by Francis Collins.⁶

Given their popularity in science, why are cohorts so hard to establish and manage? There are several plausible explanations. One of them is that although we speak highly of scientific evidence as the basis for health systems, what it takes to achieve more, better, and faster research results is not sufficiently understood. The demand for trustworthy knowledge in real time is increasing rapidly. At the same time, there are no efficient systems in place to meet this demand. Large prospective population based cohorts, whether “thin” as in routine registries or “thick” as in cohorts with extensive collection of data on exposures (such as nutrition), genotyping, phenotyping, and biological samples are the optimal infrastructures to support science for health care and public health. Cohorts provide the best basis we have for scientific knowledge for human health. However, there is a need to develop infrastructures through which cohort and registry data are routinely collected, securely and professionally managed, and easily accessed. Subsequently, such infrastructures can provide unprecedented opportunities for science; for new discoveries and trustworthy knowledge for patients, citizens, health services, and public health; as well as for human curiosity and creativity.