We are the Association for Child and Adolescent Mental Health, or ACAMH for short. And this is ACAMH Learn. Welcome to Mind the Kids podcast. This week, I'm going down the rabbit hole with Adriene Beltz of the University of Michigan, psychology professor and hormone expert. We're discussing her JCPP paper, which she claims she never intended to write, but we're certainly glad that she did. I'm Mark Tebbs, and I'm your host today. So my whole career in mental health, from frontline service delivery to director of mental health commissioning. I'm chief executive of mental health charity at the moment and trustee of another mental health organisation. I'm a career coach, and I'm delighted to be hosting these podcasts as we get to speak to leading clinicians and academics about their research. In this episode called "Lessons from the ABCD data revolution," we'll look at the large ABCD Study, a US nationwide powerhouse tracking nearly 12,000 kids over a whole decade. From Adriene Beltz's expertise on hormonal changes, you can see why getting this right is so important. When puberty hits earlier or later than peers, it can spark brain mood and mental changes that last into adulthood. Let's get started. Adriene, welcome. So please to be speaking to you. Yeah, my pleasure. Thanks for taking some time to hear about our work. Oh, really looking forward to it. So, look, let's start with some introductions. So if you could say who you are, where you work, maybe some of your research interests, and any collaborators or colleagues you worked with on the paper, opportunities to give them a bit of a name-check. Yeah, that sounds great. I'm Adriene Beltz. I'm a professor of psychology and also a research professor at the Institute for Social Research at the University of Michigan. That's in Ann Arbour. And broadly, I'm interested in hormone transitions across development. Think prenatal development, puberty, other hormone transitions during life, including things like the menstrual cycle and pregnancy, as well as menopause. OK. And the colleagues you worked with on the paper? Yeah, so this paper was a paper on pubertal development. And I have a team of folks here at the University of Michigan and also some colleagues at Penn State University in Pennsylvania who worked with us. So the colleagues at Penn State were Holly Pham, a graduate student at the time, and in fact, my graduate advisor, Shari Berenbaum. And then colleagues here at the University of Michigan were my close collaborator, Chelsea Kaplan, as well as Mary Heitzig, and our team of students and postdocs and research staff. That includes Tristin Smith, Esmeralda Hidalgo-Lopez, Hannah Becker, Christel Portengen. And that's all of us. A brilliant collaborative effort. So let's dive into-- yeah. So your paper, as I understand it, suggests that the research using the ABCD data set may be oversimplifying our understanding of puberty. So could you just begin by describing what the ABCD data set is and what the core issue that you've been exploring in the paper? Yeah, the ABCD Study is the Adolescent Brain Cognitive Development Study. And it's a massive effort across 21 different data collection sites in the United States. And the goal is to recruit adolescents age 9 and 10 and then follow them up every year for 10 years, really to get a big picture of longitudinal kind of trajectory picture of what adolescent development looks like. One of the key features of ABCD that makes it very unique is every other year-- so the first year and then what we call the two-year follow-up, four-year follow-up, and so on. Every other year, the participant study visits includes an fMRI scan. So this actually makes it the largest study of adolescent neurodevelopment certainly in the country. And this means a larger study. I didn't tell you how many folks that is. The goal is to recruit-- it was initially around 10,000, but actually recruitment has surpassed that. And it's just a smidge under 12,000 kids participating in this longitudinal study. And that means it really has untapped potential. There's so much potential here, especially for understanding certain outcomes. So I'm a developmentalist by training. I talked about how I'm interested in hormone transitions. But one of the challenges we have in developmental research is you see someone at some kind of cross-sectional point in time, some snapshot in time, and say they have a mental health diagnosis. And you want to know. How did we get here? What kind of path was taken to end up in this diagnosis? But then you have to ask, OK, well, what happened last year or two years ago? Then you're looking back. And so it's a really rare opportunity in ABCD that we have so many kids at the start that we can actually look forward. And we can see where they've been. We've assessed where they've been. We've heard from them. We've heard from their parents. And so if there's some sort of outcome that we want to understand down the road, we don't have to look back by recall. We can look back by data. And so that's one of the amazing features of ABCD Study and one of the reasons my group was particularly interested in leveraging these data for understanding what puberty is and isn't related to in the long term. So at the end of adolescence or even early adulthood, let's look back and see what influenced puberty might have had on individuals' trajectories of development. And so that's how we inserted into this study. And then your question, what about pubertal development in this space? We were really curious the core issue we ended up addressing. It's not where we started, but I think you and I will get there. The core issue we ended up addressing is OK, if we're trying to understand these outcomes, and we're looking back to see those trajectories of pubertal development, how were those aspects of pubertal development assessed? How were they measured? Who reported on them? And how do we actually make sense of that data that was collected? And so that's really the perspective of this paper or where our perspective came from initially. Right. Brilliant. OK. And is it like live, it's ongoing, it's like here and now? Is it across the United States or a particular geography? Yeah, good questions. Yeah, the 21 different sites are across the United States. If you go on to the ABCD website, they have this lovely little map with pins and all the different 21 sites. Each of the sites can tend to have somewhat of a different focus or emphasis. For instance, University of Michigan is a data collection site for the ABCD Study, and our emphasis here is on understanding and assessing substance use in adolescents. But other sites have different-- so it is across the US. And yes, it's ongoing. The data in our paper, we focused on through the first two years of the study. And the reason for that is there's a time delay in science. You have to have the data available, look at the data, analyse the data, write the data up, go through some sort of publication process. So by the time my group could actually look at how other studies were using the data from ABCD, it's already several years past when those data were released. So our study focused on those first two years. But most recently, I think up through the year five follow-up data have available for researchers to apply to use. There's a data use agreement involved. So a bit over halfway in that 10-year study right now. OK, righty-ho. So you looked at the first couple of years, and your focus was around how puberty was measured and interpreted within the ABCD Study. So what motivated you to look at that specifically? Yeah, with some of my colleagues here at the University of Michigan, we were doing other research on actually pain development and multisite pain development. This is a question that the ABCD Study really has a lot of potential for, because there's not a high percentage of kiddos that end up with multisite pain. This is the kind of pain that doesn't seem to have an external cause. Can't quite figure out why the pain is there. And it seems to be in multiple regions of the body. And so sometimes this is an indicator of what we call centralised pain, which means the body is interpreting the pain, or the brain is interpreting the pain and the different body regions in this way. But there's not a disease process or an injury that could explain the different pain sites. And this is the type of pain that characterises conditions like fibromyalgia, for instance, if that helps make a connection. And there's a large sex difference in this type of pain, where females tend to report higher levels or more sites than males, but this is often understood more in adulthood. And figuring out how these pain trajectories come to be is a pretty open question. And given the large sex difference and some converging elements of research, we thought that hormones might play a role in those sex-differentiated trajectories. And so we had done some studies on multisite pain, what that looks like in the brain and the ABCD study. And we're making some progress. And we finally got to the point where we said, OK, we understand what some of the sex differences in the brain are, related to multisite pain. Now let's see if we can figure out how they get there in relation to puberty. And so we started digging into the puberty data to help link it to those questions we were trying to answer. And as we were digging into it-- and I've done a lot of work on puberty, different types of measurement and assessment. One of our analysts on the project, he said, we're just trying to understand what the initial levels of development look like in these kids. And so our analysts on the project had created some slides. And I looked at the numbers, and I'm like, it's interesting. Based on the measure that was used in the study-- it's called the Pubertal Development Scale-- the participants are asked about a particular aspect of secondary sex characteristic development, for instance, changes in body hair or changes in breast development. And they're asked to say on a 4-point scale, if this aspect of development hasn't begun for you at all, I'm not seeing any changes to-- I saw a bunch of changes in. My development is pretty much complete. That's a 4 on the scale. So that's what I was expecting to see. But when our analysts presented the slides, everything was actually out of 5. And so not averages out of 4, but averages out of 5. And even the averages had different labels from pre-puberty to, I think, all the way to post-puberty. And those weren't the type of indices I'm used to seeing as a puberty researcher. And so we dove in, or as my group likes to say, we crawled into the rabbit hole because we were just trying to make sense of what had happened there. And it took us a little bit of time, but I think we did. So the study focus was about centralised pain. And then as you've got into the study, actually, you've found some interesting facets about how puberty is measured and the use of these scales. So that was almost like a-- was it like a by finding of your initial work? Yes, absolutely. When I present about this paper, this particular paper, in meetings and especially with meetings and friends, I'm like, here's the paper that I never intended to write. And I think many of us in the group felt similarly. It's something we happened upon, and we thought it was important. And so we crawled into that rabbit hole to make sense of it for us and hopefully for others as well, because there's so much incredible potential in the ABCD Study. So if there's different ways we can help other researchers maximise the time that they spend with the data set, then it's better for all of us. And it's better for the kiddos that we're studying. Brilliant. OK. So what happened when you crawled into the rabbit hole? Yeah So when we crawled into the rabbit hole, we realised that the PDS data-- the 1 to 4 data that we thought was there-- was there. It's there at the item level, but that just a composite score, like an average score had been created. But the average score only included some of the items from the original scale. So there's five items for females, five items for males on the original scale. This composite includes only three items, and then it makes some decisions. It sums the responses from those three items, and then make some decisions so that folks fall into categories of pre-develop-- or pre-puberty, early puberty, mid-puberty, and so on. So it uses a select number of the items and, like I said, make some kind of categorising decisions. That particular metric, that particular composite score has been used a little bit in the scientific literature, but not a lot. And the few times it has been used, some folks have said, fine. Other folks have said, oh, there's some concerns with it. And so we said, well, let's test. We have this amazingly large data set. Let's test two things. First, let's test to see how that what we'll call maybe like the categorical composite score, the one that groups into five scores, how that one looks data-wise compared to the average of the five items on the 1 to 4 scale. So let's just compare those two scores and see what's going on. But then we thought it doesn't really matter how the two scores compare if folks aren't using one or the other. If when researchers access the ABCD data set, if they're not using, say, that categorical score, it doesn't matter how it compares to the more commonly used continuous score. So we also did a systematic review to see of the studies that have used the ABCD data and published on it the first two years of those ABCD data. How were they using the puberty data in general? Were they using either of these scores? So is the score being used, and does it matter? That's basically what we set out to figure out. OK. Right. So I just want to be really clear on this then. So within the study, there are two different potential ways of scoring puberty, a continuous score and categorical score. And the data set gives you the opportunity to look at either or both. Yep-- And then-- --that's-- Yeah, and your study was looking at the research literature to see which studies are using which approach. Yep, yep. And then comparing the two options, if you will, for how to score. And just for any of the puberty researchers who might be out there, there's more options than just the two. We only focused on the two, though, so that's a good point for future studies as well. OK, righty-ho. And so what did you find? We basically found that folks were using the categorical score. It's hard to infer exactly why, but it was the one provided in the data releases. And so it's provided in the data releases. And in some ways, it has this nice heuristic of pre-puberty, early puberty. And putting those labels on points of development helps anyone think about things, or most folks think about things when they label them. So it had some of those attractive qualities. It did seem to be widely used. In fact, the majority of the studies that used use ABCD data had used that kind of categorical score. Some had used the average score, but certainly not the majority. But then when we compared the two scores, we think there was better evidence psychometrically, so in terms of how the numbers appear and what we would expect for adolescent pubertal development. We think that the continuous score actually was looking a little bit more promising. And some of the reasons for that are the score had higher reliability, meaning that folks responded to the five items in similar ways. If they said they were more advanced in aspects of breast growth, like we were saying earlier, they also said they were more advanced in aspects of body hair. So the indices of development grew together more for the continuous score than for the categorical score, for instance. We were also seeing the categorical score in making those cutoffs, whether somebody is early or mid-puberty, for instance. Especially for females, the categorical score really emphasised menarche. So that's the first menstruation. And if individuals had had menstruation-- that's one of the five items on the original scale-- they were automatically put into category 4. And you see this kind of big jump in the data. So instead of a gradual increase across pubertal development, once menarche has occurred, you see this big jump in the data. And that plays out in multiple features when you examine the data, not just plotting it, but in some other cases as well, for instance, related to missing data and whatnot. And so this kind of emphasis on menarche, I think, raises some questions. It could be a very appropriate emphasis, or it might be an overemphasis. But it's one of the reasons that the scores seem to differ. Think that really needs follow-up and future waves of that study. OK, so the continuous scoring seem to have a little bit more internal consistency. Is that right? Whereas the categorization had these jumps around an emphasis around menstruation. And also, was there inconsistencies within the categorization? Inconsistencies within the categorization-- what do you mean? So as in some of the indices suggested say pre-puberty and others would be mid-puberty. That's a really good question. Based on the internal consistency ratings-- it's a statistic called Cronbach's alpha-- that suggests yes, there were more inconsistencies in the categorical. But there is a caveat there, which is that if you recall, I said that categorical score only uses three of the five available items. And the continuous score uses all of the available items. And so we know when you have more items, things just tend to actually be more consistent. When you have fewer items, the inconsistencies cost you more. And so that's just something to think about in terms of the consistency of the scales. So just having more items in general, all other things held equal. We know having more items is better for consistency. Right. Yeah, cool. Makes sense. Makes perfect sense. OK, so what's the implications of this difference? Tell us why all of this matters. Why all of this matters-- because going back to where we started with why does pubertal development matter. If we think puberty-- and we know from the vast literature now over the last 20, 30 years that puberty does matter, not just for adolescent development, mental health, cognition, as well as physical health outcomes. But there's more and more research coming out showing that adolescent development, particularly when an individual goes through puberty compared to their peers, that those aspects of pubertal timing matter even into adulthood, because it sets folks into a developmental track of related influences. It's about hormones and how they influence the brain and behaviour. But as hormones influence the body, and as other people see bodies changing, it changes how people interact with an individual. And so you can see lots of different ways that pubertal development maybe triggers a sequence of events for some individuals that end up mattering for things like adult mental health. And if we're using something like the ABCD Study to get a prospective look, to understand, OK, what are those influences, who does puberty matter most for, and if there's outcomes down the road that we want to try to prevent how and when might we intervene, it's really important to know what aspects of puberty you're measuring and how and when they happen. And so really measurement fidelity and using these data well matters very broadly, in my opinion, as a puberty researcher, for accurately understanding some of these trajectories. And so that's the big picture how we might-- zooming in a bit, how we might end up using that in research is-- I think the bottom line of our study is be thoughtful about pubertal assessment. And there's even a metacomment there about secondary data broadly. Secondary data, large data sets like ABCD can be amazing, and perhaps only if we use those data well. And so especially for things like pubertal development, make thoughtful choices that make sense for your study. The categorical score might make sense for some questions and some studies. The continuous score might make sense for some questions and some studies. Just understand the aspects of the measure that are going into the composite score that you're using and what it does and doesn't say about pubertal development. Know the strengths and limitations of the measure before you plug it into your model. And I think that's the metapoint about our paper. Don't just use composites that are available because they're available. Make sure they make sense first. Yeah, OK. I thought from the early conversation, there was a sense that the continuous scoring was better. But you're saying that it's more horses for courses. It's matching it to the research questions that you're trying to answer. I do think for the ways we looked at it in this paper, the continuous measure is better. There's more validity for its use broadly in the field. And that means that the existing literature can inform what the numbers mean in this study and how to predict. I think the reliability of the measure matters. There's less missing data on this continuous measure and that matter. So yeah, I do think the continuous measure-- if someone asked me and I didn't know anything about their research question, I would say check out the continuous measure. But I don't want to rule out the possibility that the categorical measure might be a good option in some cases, depending on what the research question might be or the nature of the sample might be. There's so many possibilities in large studies like ABCD, so I don't want to say never use that option. I don't think that would be appropriate because I can't envision all of the possibilities, but I would put most of my eggs in the continuous measure basket, if that makes sense. That's a good description. And am I correct in interpreting? So with the continuous score, it's more holistic. It incorporates lived experience of puberty as well as the biological aspects you talked about earlier. Is that a correct interpretation? Actually, the aspects of lived experience aren't part of the original assessment. So there's five items in the PDS, and they all have to do with secondary sex characteristics. So I already mentioned breast growth and body hair changes and menarche. There's also skin changes and a growth in height. So those are the five options for females. Some are overlapping, and some are different for males. And so they're all about secondary sex characteristics. Just the point I was trying to make earlier is that they are indeed secondary sex characteristics, and they're often visible to others. And individuals know if they're having those experiences before or after their peers. And so just those secondary sex characteristics, because they're influence how people interact with teens, how teens think about themselves. So they just set up that potential cascade of influences. And I think ultimately in studies of puberty, that's part of what we're getting at too. But this particular measure is only assessing the secondary sex characteristics themselves and in fact in individuals or their parents' perception of them. OK, righty-ho. Yeah, I think we can all remember back to puberty and the sensitivity around your body and having to change in locker rooms and things. It's like it's a tough time. Yeah, and a lot of the literature shows, especially for I mentioned pubertal timing earlier. So not just the development of secondary sex characteristics, but when they happen in comparison to peers seems to be a particularly poignant aspect of psychological development. In general, females develop before males. And so that means if you're among the first females to develop, you're among the first of your whole peer group. So you're undergoing those changes before everybody else, and they're often noticeable. And there's a large literature and a mental health outcomes associated with early pubertal timing and females. What's less studied and where I think ABCD has a lot of potential and why it's so important we measure puberty well is in that same line of thought, it means that late developing males are also the last of their peer group. And so the very last ones of everybody-- everyone else has developed and matured. And then there's things like facial hair growth and voice cracking that happens last for some set of males. And that has mental health consequences as well, knowing that everybody else has passed you in development. And studying pubertal development in males has traditionally been harder than in females, because there's been a lot of emphasis in females on menarche, consistent with that categorical measure, as we were talking about earlier. And the same equivalent indicators in males aren't studied as much or as reliably. And so having data like PDS data across 10 years to help us shine the light on pubertal consequences for everybody and for individuals who might not have had their experiences represented as much in the literature is really important. Yeah, absolutely. And then you mentioned a little bit about parent measures and child measures. Could you tell us a little bit more about that? Yeah, so another benefit of ABCD is that those five puberty items that we've been talking about and that 4-point scale that they get rated on. The parents rate them about their child, or a parent rates them about their child. And the child rates them about themselves. And what we found in our paper-- and this is very consistent with the broader literature-- is that at least in the beginnings of pubertal development, parents actually have more reliable ratings than the youth themselves have. And we think the reason for that is-- and like I said, there's a larger literature on this-- is that parents know what puberty is. They know what they're looking for. They know what those skin changes are. And they notice these signs, perhaps in their kids before the kids really notice what's going on with themselves. Or sometimes you'll see-- and we mention this briefly in the paper too. You see these what we call regressions and reporting, which is that a kid will say, oh, yeah, I'm definitely having skin changes. And so they'll select like a 3 when responding to the skin changes item. And then a year later, they actually select a 1. It's like, well, last year, you said you were definitely having skin changes. And now you say skin changes haven't started at all. What has changed? Perhaps the kid's knowledge about what those skin changes actually are, because they've had a year. They've seen some of their peers. You maybe have experiences with oily skin or acne, and they're like, oh, but that's actually not me. And so early on in the beginning stages of pubertal development, parents actually seem to be perhaps more reliable reporters than the youth themselves. We expect, though, that as puberty progresses, that actually changes, because as adolescents unfolds, as puberty unfolds, kids become understandably and appropriately more private about their bodies and their experiences. And so by the end of puberty, it's reasonable to expect that the kids will actually be more reliable reporters than the parents. So the parents can see when it starts, but the kids then are like, OK, this is my body. I know what they're talking about now. And they become more reliable reporters near the end of pubertal development. And so my joke is, if this was the paper that my group didn't want to write in like three years, we might have to do a second one. Yeah. OK. And I'm interested about almost the implications of your work in that clinical practise area. So what are the-- what does your work suggest in terms of the importance of considering puberty carefully when studying emotional and behavioural challenges? Yeah, it's a really good question. Thinking about things in the clinic, the assessments of puberty that tend to be done in the clinic, if they're done at all, are often the gold standard in the field. So we've been talking about the Pubertal Development Scale in different ways to code it up and get a measure of it. But the gold standard in the field is actually something called Tanner staging, which is actually in five categories, which might also lead to some of the appeal, the initial appeal of that categorical measure. But Tanner staging involves an evaluation of the youth's body. For females, it's about breast growth is assessed and pubic hair development. And for males, it's about testicular volume and pubic hair development. And so those are the clinical measures. At least in the US, they're not done as a course of-- or in the course of a regular annual wellness visit or check-in. But they might be done by, say, a paediatric endocrinologist in a specialist sense. So that gold standard assessment by a medical health professional is something that is not typically done in a research setting. So these research measures that involve self-report or parent report seems to be a reasonable balance between reliability and intrusiveness when it comes to pubertal assessment. So just a comment there about how puberty is assessed clinically. But thinking about it in terms of perhaps clinical psychology and behavioural outcomes, mental health outcomes, specifically, yeah, I go back to the initial point that if we really want to understand who is at risk and when we might intervene, we have to know what is actually going on with pubertal development as reliably and accurately as we can. So I really think in a lot of ways this is research bench question. And when we get enough of the answers at the bench, then we'll be able to inform, make our way up to the bedside, as the saying goes, where we can then say, OK, now we know that for girls or for females, it's about this aspect of puberty, or for males, it's about that aspect of puberty. And the timing is this way, or the timing is that way. And that information can be passed up into a clinical setting then to say, oh, I have a patient or a client here with this set of characteristics that does or doesn't align with what we're finding about puberty. And that might then inform some recommendations about how to proceed. I think the important thing to remember, though, is that puberty is a normative process. It's reasonable to have some behaviour changes associated with these big body changes and the context in which they occur. So I think a lot of the implications of the work aren't about changing puberty or its timing, but they continue to be about the things we consistently say with respect to support youth, create safe spaces where they can share their problems, education about puberty and what these body changes are. So there's some appreciation that these are normative. And even though the changes truly do feel weird to a teen and the situation they're in, that those are the appropriate feelings to have about those types of changes. And so just continue to support our youth clinically. Yeah, absolutely. I'm also curious around whether the kind of categorization systems we're talking about are culturally sensitive. And I imagine the experience of puberty is not like universal or homogeneous. Does your research touch on any of those aspects? Yeah, those are great questions and things I'm hoping that the large ABCD Study will be able to tell us about, at least within different contexts within the US. We do have 21 different sites, and there's ways to look at those sites separately or just for variances and those sites statistically. And so I think there's a lot of possibilities and what we can learn from the ABCD Study and puberty there. Going beyond that, other cultures and contexts and what puberty means in those places is also really important. And I'm hoping as we can make some inroads about how we even assess puberty, at least in this culture and context, than other-- it will inform and empower others to see, well, whether those assessments are reasonable in other cultures and contexts as well. They may, or they may not be. Yeah, OK. So yeah, it's the same thing, isn't it? We got to get the measurement, right? Right. Measurement, right. I know. Yeah. Yep. A bit of a broken record going on here, but I'm a psychologist at heart. Yeah, for sure. Absolutely. So, look, we're coming to the end of the podcast. So are there any final messages maybe to researchers that are working in the area and thinking about how they measure puberty? Yeah, I think the final-- I guess maybe just a reiteration is my final point, that large open data sets-- ABCD and others are amazing resources. They allow us to prospectively understand what pubertal development maybe does or doesn't mean, or who it means something for, and maybe who it doesn't matter for as much. And there's a lot of potential in those questions. We just have to be mindful about what we're measuring and exactly how we're measuring it, and how we're creating those composites within our data sets, and just make informed choices about them that makes sense for your study, and say how you made those choices. One thing we didn't talk about is that in going through this massive literature of studies that have used the ABCD Study data set and trying to figure out what aspects or what assessments of pubertal development were used, we often had a hard time even knowing what was done, which composite was used. So provide sufficient information about reporting so that folks can, other researchers can reproduce the findings and then hopefully build on them and shine the brightest light on adolescent development that we can. [INAUDIBLE] If people want to find out more about the work, the study, is there a resource or a website that people could check out? Yeah, sure. It's at abdcstudy.org. Brilliant. It's been lovely speaking to you. I've really enjoyed the podcast. Yes, same here, Mark. Thanks again for taking the time and helping get the word out about this important study and pubertal development broadly. Right. Thank you. I really enjoyed that. It yeah, sent some shivers down my back about my own experiences of puberty. And I guess I was a little bit worried about talking about complex data sets. But actually, I think that was unpacked in a really human and understandable way. So as always, please leave a comment or review a suggestion. Share with colleagues. Get yourself a free ACAMH Learn account from www.acamhlearn.org, and you'll be able to get free CPD/CME certificate for listening to any of the podcasts. Next week is the last in the series, and we'll be off to Nashville to speak to Professor Julie Taylor from the Vanderbilt University Medical Centre as we look at the issue of autistic youth transition from child to adult services.