News blog

Q&A: US children’s study revisits strategy

Leaders of the US National Children's Study have struggled to define a sampling strategy for the 100,000 children they aim to recruit.

Image source/Corbis

The US National Children’s Study (NCS) came into being in 2000 as an ambitious effort to track influences on the health of 100,000 US children from before birth to age 21. Now the nature of that effort has been significantly redrawn, in a revised recruiting plan issued on 24 July by study leaders at the US National Institutes of Health (NIH) in response to complaints from US lawmakers.

Among the politicians’ main concerns: the study has spent US$1 billion since 2000 to enrol, as of mid-June, 3,650 subjects.

Part of its problem was its expensive recruiting approach, much of it relying on going door to door in 105 far-flung locations that were scientifically selected to represent the entire US population, to obtain a ‘national probability sample’. But in February, study leaders proposed abandoning that approach in favour of a ‘convenience sample’ drawn mainly from the client lists of various health-care providers.

That change drew criticism from many corners, because it meant that the study would lose much of its power as a tool for teasing out disease risk factors and prevalences for all US children. Senate lawmakers who oversee the study’s funding told its leaders to “make every possible effort” to enrol a statistically representative sample of children. This month their counterparts in the US House of Representatives ordered them not to make any changes to the current design until the Institute of Medicine (IOM) has reviewed the proposed changes.

According to the revised plan, the study would still recruit 100,000 subjects: some 40,000 from a “birth cohort” recruited from a geographically dispersed, nationally representative selection of hospitals and birthing centres; another 50,000 or so in a “prenatal cohort” of pregnant women recruited from clinics and physicians’ offices that refer patients for delivery to that same set of hospitals; and 10,000 recruited from the same clinics and offices, but before they become pregnant — a ‘preconception cohort’. That group of 10,000 would also be used for deep dives to assess particularly under-represented populations, or those with specific exposures. (A group of pregnant women exposed to fracking is an example the document puts forward).

Last week Nature spoke with Steve Hirschfeld, the paediatric oncologist who is director of the NCS, the day after the revised plan was presented to the study’s advisory committee:


How does this new plan get at Congress’s desire for you to address the nation’s diversity and health disparities?

Our definition of health-care providers is broad: we include family practitioners, obstetricians, paediatricians and hospitals and birthing centres. The hospitals and birthing centres in the United States are the places where close to 100% of women go for their deliveries. We can have assurance of diverse populations and reaching people with health disparities, because even people with no prenatal care go to a hospital for delivery.

The other mechanism is we have a supplemental recruitment. This [cohort] will be used to target populations that are underrepresented on the basis of demographics, exposures or other factors. We’ll go wherever the science and the data lead us. That’s a separate effort.

The House committee that funds your study has required an IOM review before you can make any changes. Such a review usually takes months to conduct. Are you hamstrung in the meantime?

Our goal all along — and we have been in conversation with members of the IOM — our plan was when we reach a point where we have a sufficiently robust document that could describe [our plan] to an adequate level of detail… we would have an external review. For us, that’s no change. Our timing was to have this done in the fall.

The 2013 budget statement you sent to Capitol Hill in February said that you were dropping national probability sampling in favour of a less costly, convenience sample. Are you now saying: “We were wrong?

The statement said that this was an option, and that we were discussing other options… . It said very affirmatively that we were going to discuss [other] proposals [as well as convenience sampling.]

So a convenience sample is now off the table?

The [national] probability sample will be the birth cohort, that’s clear. We will enrol the prenatal cohort using probability techniques, but because of the nature of how you identify pregnant women, and who has access to health care, we are not going to package that as a probability sample.

And the preconception cohort, if we do one — again because of resources — is going to be a non-probability sample.

How do you mesh the 105 far-flung study locations that were scientifically chosen in 2004 with your newly proposed sample based on hospitals and birthing centres? Will they have to be in those 105 locations, or are you dispensing with those?

We don’t know yet. We have to work out the plan. And once we have the plan we will have an independent evaluation as to whether some or all of the current locations mesh with the plan. All of them, none of them, that’s something we can’t make a statement [about now] because we lack the details.

But our intent is that we will develop this plan and have some independent entity review the plan and make some determination about what the locations would be.

Given the nature of your project, would the sampling hospitals under the revised plan need to be huge institutions, or could smaller hospitals in smaller places qualify?

If we’re following a national probability sample, then all locations within a selected area would have an opportunity. Now, statistically, what often happens, is the more births a place has, the more likely it is to be selected based on proportions. But the short answer is yes: a tiny birthing centre could well be in the sample frame, that is, the list of facilities from which you would be recruiting participants in the study.

Some principal investigators are doubtful that you can retain subjects who are recruited at or about the time of delivery. They think that the odds of keeping a subject in the study are much better if the mother is recruited during pregnancy and builds a relationship with study staff before the stress of labour and delivery. How do you respond?

There are birth-cohort studies where women are recruited at birth and the [willingness to enrol in the study] is really very high, as much as 90%. We are addressing that now at three pilot centres… with hundreds of women anticipated in each category, we can get some sense on several key questions: what is the efficiency… what is the retention rate; what is the quality of information you can gather?

Your plan would require participating hospitals to keep cord blood and all placental samples from all deliveries, disposing later of samples from women who didn’t agree to enrol. Is this realistic?

We have spoken to other people in other birth-cohort studies and the answer is that if you are clear about your communications and systematic… it’s a plausible system. So yes that’s an issue to bring up; it’s a logistical problem to be solved but we have information that it’s quite a solvable problem.

Given that you can’t move forward with the main study until the IOM weighs in on the proposed new study design, how can you effectively spend the $165 million or more that Congress looks likely to give you in 2013?

We hope we’re not stymied. We are running 40 Vanguard [pilot] Centers and have budgets to go forward for each of the 40 for the coming year. We also are running formative research programmes. And no matter where we’re going to be doing the main study recruitment, we have a centralized logistical setup: our informatics systems; expanding our repository capacity; there are logistical preparations that we need to make centrally. No matter where the study occurs, we have to have these things in place.

When is the earliest that the main study could launch?

We intend to be able to enrol participants in 2014.


Comments are closed.