Reflections on the INCAP Oriente Longitudinal Study of early life nutrition: An interview with Aryeh Stein

The following two tabs change content below.
David Flood
David Flood, MD, MSc, is a physician with the Guatemalan NGO Wuqu' Kawoq | Maya Health Alliance and resident in Medicine-Pediatrics at the University of Minnesota. He received his medical degree from Harvard Medical School and an MSc in international health policy from the London School of Economics.

aryeh_steinAryeh Stein, MPH, PhD, is a nutritional epidemiologist and Professor in the Hubert Department of Global Health of the Rollins School of Public Health at Emory University. Dr. Stein has conducted groundbreaking research focusing on the critical windows of susceptibility to nutritional deficits and surfeits and how nutrition influences the development of adult chronic disease. Among other research collaborations, he works with the Instituto de Nutrición de Centro América y Panamá (INCAP) and is the principal investigator for the latest follow-up wave of the INCAP Oriente Longitudinal in Guatemala. This study, which began in 1969, is the longest running nutrition trial in the world, and its results have been highly influential in advancing scientific knowledge and policy around child nutrition.

How did you become involved with nutrition research in Guatemala and the INCAP Oriente study?

My undergrad degree is in nutrition, and I had heard about the INCAP study in the mid-1980s when I was an undergraduate. Then I did nutrition fieldwork in Papua New Guinea as a VSO volunteer after college. After my two years in VSO, which is the British analog of Peace Corps, I became a graduate student at Columbia’s school of public health.

Through my work at Columbia, I became very familiar with the concept of early life nutrition. At that time, the idea of longer-term effects of early life nutrition, what’s now called fetal programming or early-life programming, was emerging. Two important studies had been done at Columbia. One was the follow-up of people who had been conceived during the Dutch famine, and that was a project that I became very closely involved with. And the other was a pregnancy nutrition supplementation trial that was conducted in Harlem, as one of a series of studies conducted in the late-1960s trying to test the hypothesis that protein supplementation specifically would improve cognitive development in the offspring. These were a set of studies—there was one done in Taiwan, there was the one done in Harlem in New York City, there were others in Colombia. The INCAP study was done in the same time period with a slightly different design, so I was aware of it.

I continued to work on research related to the Dutch hunger winter, the Dutch Famine studies, with a colleague Bertie Lumey, who’s still at Columbia. I was at Michigan State University as a junior faculty member, and then a position opened up at Emory to work with Reynaldo Martorell, who was the principal investigator of the INCAP study follow-ups. He has been involved with the study since the mid-1970s and did his doctoral work with this project. I joined Emory in 1998 primarily to join the group that was doing this research in Guatemala, and I’ve been working with the group ever since.

The INCAP Oriente study was initiated in the 1960s. At this time, what were some of the biggest unknowns in the field of early nutrition? From what context did this study emerge?

The 1960s was a big period of decolonization throughout the world. All sorts of important social movements were happening. In the nutrition world, similarly, there were a couple of countercurrents. One was “there’s not enough food, there’s famine,” while at the same time others talked about the Malthusian theory that the population growth was outgrowing our food supply.

On another level, the 1960s were the heyday of the Green Revolution, which essentially solved that problem—at least for a couple of generations. One of the theories that was going around as to why most of the intellectual leadership—the scientific successes in the world—were happening in Europe, North America, Japan, and Russia as opposed to what was then known as the “Third World” was that children in poor settings were not having the opportunity for their brains to develop optimally. At this time, an early inference in the concept of early life programming was that protein malnutrition was at the root of this—that the limiting nutrient in the diet of the developing fetus was protein.

In the 1960s, the NIH set up and funded a set of studies explicitly to test this hypothesis. These are the half dozen or so studies I mentioned. They had differences in design. The INCAP Oriente Study was a group-randomized trial. Two villages got atole, two villages got fresco, and everybody within the same village got the same supplement. In other settings, individuals were randomized, which is much more logistically difficult to do, but gives you a cleaner research design if you can implement it.

All of these studies had different designs to test this explicit hypothesis about protein. That was the paradigm at the time. Since then, the paradigm moved on to micronutrients, to quality food, to other factors. Right now, there are a range of studies testing fish oils in pregnancy as a rate-limiting nutrient. As knowledge evolves, as we learn whether or not something really is relevant, then we either adopt it or we move on. That’s how science works.

Photo from Katie Moore

Photo from Katie Moore

The “first thousand days” concept that emphasizes nutrition during the window from pregnancy to a child’s second birthday has not always been the scientific consensus. Can you speak to how results from the Oriente study contributed to the development of this paradigm?

There are two aspects to this. In the short term, the Oriente study was seen to be a null study. In other words, if you think back, the original study design was to test the hypothesis that improved nutrition will lead to better cognitive development in children. The original study was designed to test children through the age of 7 and then leave them alone.

When the study was concluded, the papers that came out of it in the late-1970s and early 1980s suggested that there was very little benefit to the early life nutrition. And that’s often been the case in other studies as well. It takes a while for the benefits to become noticeable. There’s a period of latency. If children continue in their normal environment and the brain isn’t challenged by better schools, for example, then the improved nutrition doesn’t have a chance to play out.

One of Reynaldo Martorell’s key contributions to this project was doing a survey in 1988 of the children who had been in the original study, at which time the children were between 11-25 years old. Some of them had finished their schooling, some of them were still in school, some of them had started to have jobs. At that point in time, he was able to clearly show that early-life exposure, while having no impact in the short term, had made a significant difference in the longer term.

The subsequent field work that I was involved with in the early 2000s clearly showed that being exposed to the nutrition program in the first couple of years of life was equivalent cognitively to a year of school in terms of contribution to literacy, ability to do math puzzles, and so on. In fact, kids who had been in the intervention group stayed in school longer, particularly the girls.

And then a key paper that John Hoddinott led showed that, at least in men as few women were in the paid labor force, nutrition supplementation in early life was associated with a 30-40% increase in wages. So that really suggests that although you may not be improving things in the short term, in the long term, improved nutrition makes a big difference.

Coming back to the specifics of what the Oriente study has done for this, it’s the longest running nutrition trial, and it’s in a low-income country where the lessons can be generalizable to many other settings.

Was the evidence from the Oriente study the most important data in developing the “Thousand Days” paradigm?

The Oriente study is one of a set of pieces of evidence. The Dutch Hunger Winter was another critical area of research because it allows us to identify timing within pregnancy on outcomes, and it really highlights early gestation as a key period of fetal development. It has led to questions about epigenetic remodeling and so on that the Oriente study cannot do because we don’t have the ability to separate out periods of exposure well enough.

The other issue has been in regards to sample size. The Oriente study is 1500 people. For some questions, 1500 people is just too small to be able to look at things; you need larger samples. But the Oriente study has certainly been one of the very important pieces of evidence because of the length of follow-up and the randomized trial design.

Photo from Daniel Agee

Photo from Daniel Agee

Nutrition epidemiologists and researchers design and assess trials differently now than in the 1960s when the Oriente study was initiated. Hindsight is always 20/20, of course. But as you continue to push forward this study more than 50 years after it began, what are some of the things had been different about the study design or data you wish had been collected from the beginning?

First, I would have tried to have more communities. In the Oriente study, there are two pairs of communities, and for statistical modeling it’s a very strong limitation that we’ve had to develop ways to work around. Having more communities would lead to a larger sample size as well, which gives you more ability to look at what is it about particular villages and particular contexts that makes things different. And it provides better control for village-level differences and ensuring that randomization works.

Another would be more biological sampling. There was no biological sampling in the children in the 1970s. There’s anthropometry, cognitive testing, dietary intake data, and morbidity data. But there’s no biological data—no blood samples, no stool samples. We can’t look at the microbiome and all sorts of other things that would be interesting. There was also no environmental data at an exposure level—no air monitors, no pollutant monitors, no measures of the quality of the water. These are things that one would love to be able to throw into a big data model of child growth and development.

Additionally, there are no good measures of maternal-child interaction. We know much more now about what it is that helps children grow physically and cognitively. But those things weren’t measured.

The scientific output of the Oriente study has been incredible—hundreds of peer-reviewed papers. For people who wish to learn more about the study, where is a good place to start?

I would go to two places. To get a basic idea about the study, there’s the cohort profile, a paper published in the International Journal of Epidemiology in 2008. I was the first author on that one, but all of our work is teamwork. That’s a good place to get a brief overview of the core study design.

There was a supplement to the Journal of Nutrition in 2010 because there was a symposium at the American Society of Nutrition focusing on the Oriente study. There are multiple papers about the study in this supplement.

There’s also an detailed monograph after every big wave of field work, but to get an overall view of the impact of the study, I would look at Journal of Nutrition symposium.

You are leading the latest round of follow-up data collection for the Oriente study, which is taking place right now. What are you most excited about in this wave?

We are, for the first time, collecting detailed biological sample data on the adult members of the cohort. Each individual cohort member comes in, we’re updating their economic and medical histories, we’re doing the clinical examination, looking at fasting and post-load blood draws, studying body composition by deuterium-labeled water, we’re doing a physical fitness assessment, and we’re measuring physical activity using pedometers. We’re collecting a wide range of biological data and updating the socioeconomic data our study has been very rich in. We’re going to bring the samples to the Emory metabolomics and lipid core lab to run detailed assays.

Our primary hypothesis is that the individuals who got better nutrition as children are better positioned to handle the obesogenic environment in Guatemala. They will respond less to a glucose challenge in terms of a raised glucose profile and to a lipid challenge in terms of a raised lipid profile. Importantly, it’s not that these people won’t be overweight and obese; it’s that the short-term response to an obesogenic load will be muted.

If you go back to the concept of early life programming, one of the key components is the concept of mismatch. The environment in which you were conceived should match the environment in which you find yourself as an adult. And if it doesn’t, then you start to get the metabolic consequences. The point is that the early life improvement in nutrition puts people on a path for which nutrition is not the limiting factor.

Photo from Katie Moore

Photo from Katie Moore

INCAP has historically conducted research in Maya communities, but the Oriente study only included non-indigenous Maya (Ladino) villages. Could you talk about the history of this decision? And do you think the study would have turned out differently in any way if Maya had been included?

The study was originally designed to test whether better nutrition improves the cognitive development of children. So you had to be able to measure cognitive development in children. The study team made the decision that while there are Spanish-language tools to assess cognitive development, they could not be implemented without bias in the Maya communities. And it would not be feasible to develop and validate development assessment tools to assess the Mayan kids who don’t speak Spanish as a first language over the course of the study.

The study question could have been answered in any undernourished community. If the study had utilized Mayan children, the investigators would have faced other challenges such as which Mayan language group to work in. There would have been generalizability issues. And you’d spend all this time developing tools that would not have use outside of that language group. I don’t think this issue poses any threat to validity of the findings; I think the study is generalizable anywhere.

It’s important to remember that the children in the Oriente communities were highly undernourished. During the initial trial, the stunting rate at age two was 87%, and there was not a single child at age two that had a height-for-age Z-score greater than zero—above the reference median. Later, between the mid-1960s and the mid-2000s, over two generations, the stunting prevalence in the same villages had dropped to about 17%. Some of this reduction may have been improved nutrition resulting from the intervention program, but these are the enrollees’ grandchildren and there are other factors at play. The people in these villages are living in a very different world than they were in the 1960s. Development has helped the kids in the long term.

I’m not an expert on what’s been happening in the Mayan communities, but I know that really there are two Guatemalas—the Mayan Guatemala and the Ladino Guatemala—and that the disparities are wide. There is a need for intensive, sustained work to help Mayas develop economically and improve nutrition. But it’s a long-term process. It takes generations.

You mentioned that one limitation of the Oriente study is sample size. You and other investigators in the COHORTS group have attempted to address this limitation by bringing together data from multiple birth cohorts, including the Oriente study. What findings from this collaboration have you found most interesting and exciting?

The COHORTS collaboration pools data from five relatively large cohorts that were each established at birth or before birth, and, at the time the consortium was set up in 2007, had enrolled children who had reached at least mid-late adolescence. In addition to the Guatemalan Oriente study, the cohorts also come from Brazil, South Africa, India, and the Philippines.

In the main data set, there are about 8000 individuals with data starting at birth and now going through at least age 20. As you can imagine, each study was conceived, conducted, implemented, and funded separately. Data collection methods, data periodicity, exposure measures, and outcome measures all differed among the cohorts. So one of the really exciting and interesting challenges has been to see what’s similar and dissimilar among the cohorts given all the differences in methodologies.

What really surprises us every time we look at the data is how consistent the patterns are across the cohorts—the commonalities across cohorts are quite striking. The commonalities far outweigh the differences.

We have been able to look at some key factors in child development that were collected in a way that we believe are common enough to be useful, such as information about growth during particular periods of early life. For example, Linda Adair led a paper in 2013 looking at linear growth and adiposity. We have just published a paper that looked at maternal age as a predictor of child outcomes. So we’ve been creative in our use of creating common variables across disparate cohorts to get at the question of limitations of sample size that any one cohort has. As with any pooling of data, we lose some of the richness of the detail within each individual cohort, but we gain sample size and generalizability.

One Trackback