Press Briefing Transcript

CDC Weight of the Nation Press Briefing

Monday, May 7, 2012 at 11:45 a.m.

• Audio recording (MP3, 7.43MB)

Jeff McKenna: Welcome to the telebriefing on the CDC study.  The forecast is through 2030.  My name is Jeff.  I'm from the National Center of Chronic Disease and Health Promotion.  We'll hear this morning from dr.  Bill Dietz who directed the division of physical nutrition, activity and obesity.  He'll make brief opening remarks.  Then Eric Finkelstein at Duke's Health Institute and the lead author of the paper who will present remarks.  Then we'll open it up for questions.  We have with us at the front, Dr. Justin Trogdon, a research economist.  If you are able to hear the presentation across the hall a few minutes ago, Justin presented the findings to the whole conference attendees.  We'll start with Dr. Dietz. 

Bill Dietz: Thank you, Jeff. Thank you, Justin for the presentation this morning.  It is a great pleasure to be here and discuss these issues, again.  As you saw this morning, there's some evidence that we, that at the very least, the curve of the increase may have changed at best or a plateau.  In contrast to the findings of this study, which start from a different perspective.  We are happy to address the differences in these two studies.  Regardless of which is correct, we have a serious problem.  If we are at a plateau, children are still obese and still generating substantial costs as the paper that Eric and I were co–authors on, Eric was the lead author in 2009.  That paper demonstrated the costs were $250 billion a year or 10% of the annual health care budget.  Interestingly enough, since that time, there have been other estimates that suggest what we published were underestimates.  We need to correct it for a variety of other factors.  Cost may be as much as two–fold greater.  The real platform belongs to Eric and Justin, the chief authors of the study.  Let me turn to them. 

Eric Finkelstein: Welcome everyone.  Thanks for the opportunity to talk more about our study.  Let me give you a little bit of a background on the study.  Obesity has been skyrocketing.  Everybody is familiar with that.  Prior publications suggest by 2030 or 2020, obesity rates are going to be up to 70%.  Basically, the prior studies plotted obesity prevalence from various studies and straight lined a trajectory.  As you know, as we talked about this morning, obesity prevalence is steadying off.  Right?  The extent to it or the trend is something we'll talk more about.  Certainly, the rise is increasing at a decreasing rate.  So, Justin and I talked about this and we talked to CDC and said instead of modeling future obesity rated on past obesity rates, let's identify the variables that are predictive, create a model that correlates the variables and predict what's going to happen with the variables and use that to predict future increases in obesity rates.  The variables we are talking about are changes in the demographic mix of the population.  We looked at factors such as unemployment rate.  We looked at a variety of pricing including the price of alcohol, the price of fuel.  We looked at the price of fast food.  We looked at healthy food relative to unhealthy food as well as grocery to non–grocery items.  We looked at the number of fast food and full service restaurants for individuals as well as internet access, which is we think a proxy for new technologies.  You heard this morning, there's a difference between them.  It has an advantage in that it's measured at the state level.  We were able to take advantage of state level variations at the variables mentioned.  We corrected in our study for reporting issues related to self–reported height and weight.  We did it just for that.  We used the data because that's the data that has the level to take advantage of.  We have a forecast and estimated a model that allowed for the leveling trend.  So, based on that model, we were able to use the predictions in all of these explanatory variables we found and used that to predict future obesity rates. 

The big headlines you see in this paper are that we predicted by 2030, obesity rates will increase 42%, which is significantly less than the 50 plus percent.  It's greater than a zero percentage increase.  For severe obesity, people roughly 80 pounds or more overweight, we are predicting an increase of double from roughly 5% today to 11% in the future.  Basically, one of the things we say in the story is because of the rising prevalence estimates and growing population, there's going to be roughly 32 million obese adults in the U.S. by 2030.  That would mean an increase of $550 billion in total between now and 2030 due to rising cost of obesity.  Another way to think about that number is if we could keep obesity rates flat or if they were flattening, we would save $550 billion.  So, those are the main take–home points in the paper.  Now, what I would like to do is stop and entertain as many questions as you have and try to clarify questions or differences between these or other implications of the study. 

Jeff McKenna: A couple quick comments, first, if you identify yourself and secondly, if you could use the handheld mic so people on the lines can hear you.  We'll go between the room and the phone lines.  Any questions from the room? 

Operator: On the phone lines, if you would like to ask a question, press star one at this time. 

Lauran Neergaard: Dr. Dietz if you could address the prevalence we are talking about for the average person who doesn't know about the differences between data sets and this technical level issue.  It sounds like what you are saying is we are leveling off, but we are seeing the smaller rises that are going to get us to combine with the population rise.  A total prevalence of 42%.  Is that what you are saying? 

Bill Dietz: The current prevalence is 34%in adults and 17% in children and adolescents.  We don't have comparable pediatric data but the youth/risk behavior survey.  There's a two–fold underestimate self–reported height/weight.  If you look at measure versus self–reported, it's twice as great.  Less of a disparity on the adult side about 25% to 30% report measured.  In the paper, we corrected the data for that disparity.  That, in some respects corrected the under estimates that come from BRFSS to us.  The whole issue of a plateau or not depends on where you start to look and the data you are looking in.  If you start with the data alone and look forward, there are people in the field who argue that the current plateau is actually within the limits of that continuing acceleration.  We chose to use behavior risk factors starting somewhat later than 1970 and capitalizing on the trends occurring within state and that's really the two major differences are state specific data and self–reported data.  We tried to correct for one of those two variables by adjusting the BMI data to the levels.  Does that answer your question?  No?  Okay. 

Lauran Neergaard: For the average person who is going to have a really hard time figuring out you are saying it's leveling off, answer that. 

Bill Dietz: It depends on what you look at. 

Lauran Neergaard: Which do we believe? 

Bill Dietz: My belief is the slope is changed.  Cynthia pointed out, it depends on which group you are looking at.  Because of the boys are still continuing to increase, particularly African American boys.  You know, you can't –– you know, if you look at a trend analysis, it doesn't suggest that women are flat.  There's a lot more variation on –– among men.  As she pointed out, if you look from 1999 and 2000, to 2008, there's a 6% increase in men.  Recently, it's plateaued.  If you look at the most recent data, 1997–'98 forward, men are flat.  If you look at 1999 and go forward, they are increasing.  It's really a matter of choice.  It reflects the uncertainty we have with respect to how things are changing.  I don't know how I can answer it any better than that.  Maybe Eric or Justin can.  Do you want to add anything to that, Eric? 

Eric Finkelstein: I'll give it a go.  What I would say is if you look at the overall estimates even over the last couple years, the latest year is a tiny bit higher than the prior year.  It's not statistically different.  So, statisticians will say since its not different we can't reject the hypothesis its flat.  It doesn't mean they are flat.  The data is larger.  So, that difference would be statistically significant in a larger data set.  I guess to put it simply; I would say the NHANES data shows a slight uptick.  Our estimates are consistent with that.  We go from 34% to 42% in 18 years.  It doesn't take a huge increase in the curve to get there.  To some extent, I think they assess different stories.  If the data set were larger, you would say it's a small but significant difference between the two. 

Lauran Neergaard: Much better. 

David Brown: Yeah, David Brown from the Washington Post.   Your data is projected data.  As you have said a couple times, it comes from a deeper time point in the past than what dr. Ogden was showing.  She showed what was quite convincingly there has been variation in this growth over the last, you know, this trend in American health over the last30 years.  It was flat in the '60s, went up steeply in the '80s and '90s.now it's flattened off.  Do you think they are all within confidence intervals of each other?  To get to Lauran's question, I would think the ordinary reader would say let's take the most recent actual data, which is the last eight years, I guess, that dr. Ogden showed convincingly across almost all ethnic groups, races, sexes, everything that it's basically leveling off or it's going up so marginally that it's effectively level.  Why should we not believe that is the trend that is going to continue on?  Its actual data rather than this modified linear extension from a point in which we know the trend that you're –– that is the trajectory set we know has changed.  You know, the slope has changed from the one that you are using.  It seems to me that there's no way to really square what she showed and what you are predicting. 

Eric Finkelstein: Very thoughtful question, for sure, it's a challenge.  Now, our model, too, takes into account the latest data and allows for that possibility of flattening trend.  We do model that.  Now, we, you know, as I said, we use the state level data and incorporate the additional variables to do the best possible job we can in making predictions.  We don't predict future obesity rates by taking into account past obesity rates.  We look at the trend in other variables.  You might argue if these variables were to change like the relative price of fast food or fast food density or price of grocery versus non–grocery items or any recession changes, the trend may go straight up or down.  Our model accounts for the underlying variables.  With that said, if you did our exact model with exactly as you described it the last couple years of the BRFSS data, because it's showing a slightly greater slope than NHANES, I think you guessed what we have, regardless.  Now, it doesn't take much of a difference between the slopes of the NHANES and the BRFSS to go from 34% to 42% in 18 years.  I should point out, I feel comfortable in these estimates.  The difference we are seeing for severe obesity where we are showing more than doubling is actually consistent as you look among the obese population, if you look at the average weight of that group, it continues to go up.  Our data supports that, especially among minority populations.  Are we going to be right?  Who knows?  The reality is, you know, the data is always based on the past.  It's like trying to predict the stock change.  As things change, new programs come out, new technologies come out, and they matter.  We can't model that.  We don't have access to future information.  You know, I think these are very reasonable estimates of what the future will hold, given the way the past is held.  They are estimates.  Like any estimate, we think they are useful.  Are they going to be accurate?  Time will tell.  If they are not accurate, it's likely because the world changed in such a way that makes the estimates no longer appropriate.  I hope that's a reasonable answer to your question. 

Jeff McKenna: Let's pause for a minute from the room and go to the phone.  Any questions on the phone? 

Operator: We have a question on the phone line.  If you would like to ask a question, press star one.  The first question is from Adrian Carrasquino from NBC Latino.  You may ask your question. 

Adrian Carrasquino: Hello.  What did you find concerning Hispanic Americans?  How did their data compare to the overall numbers? 

Eric Finkelstein: I can tell you that in our model, minority populations, African American and Hispanics were predicted of greater rates of obesity as well as severe obesity.  Consistent with the past data, they are likely to see larger increases than the non–Hispanic white population.  Other questions from the phone? 

Operator: We have a question from Melissa Healy from Los Angeles Times.  You may ask your question. 

Melissa Healy: Hi.  Thanks so much for doing this.  I wonder if you can talk a little bit about how rates of overweight to how they trend and how that trend compares with the rates of obesity.  In other words, our greater number of people leaving the overweight category and moving into the obese category or what?  I wonder whether given these kind of projections, I take it you cannot do –– you cannot make estimates of children but young adults.  Can you break down any age brackets that would be meaningful? 

Eric Finkelstein: I think there were a couple questions in there.  Let me answer the best that I can.  First off, we did not specifically model rates of overweight or the BMI in 25 to 30 range but certainly, it's reasonable that if rates of obesity are going up, it needs to come from somewhere else.  We expect to see, it would be at the expense of the overweight as well as the normal weight.  Two other points, in terms of age, consistent with the current data, those aged 45 to 64 were most likely to be obese or severely obese.  That's the age group as people move into that group, obesity rates are higher.  The changing demographic is partly responsible for rising rates of obesity.  Now your final point concerns youth.  In this paper, we didn't actually use data from youth but I have published in the past as well as others there's a very unfortunate correlation that if you are an overweight kid, you are going to be destines for sure to be an obese adult.  As rates of childhood obesity increase, so, too, would we expect that increases obesity in the future.  It's a major challenge we have and speaks to the importance for addressing childhood obesity today as well as the potential benefits of adult obesity down the road.  You may have additional. 

Bill Dietz: Yes.  To add to that, 15% of severe obesity in adults is a consequence of persistent obesity in childhood.  What we are seeing is the past increases and the prevalence of childhood obesity. 

Jeff McKenna: Other questions from the phones? 

Operator: No further questions.  Again, if you would like to ask a question, press star one. 

Jeff McKenna: Other questions from the room? 

David Brown: Yes, David Brown, again.  Can you describe the direction that these different variables have on obesity rates?  Alcohol price, gas prices, fast food, number of restaurants per 10,000.  Some of them are obvious, but some of them aren't.  Also, can you talk about how you use the variables?  Presumably, there's a slope of increase from all of those in the past.  Did you lineally adjust it or adjust it for a better economy later on?  A lot of moving parts. 

Eric Finkelstein: It's a great question.  Of course, we need to make some assumption about how the variables would move in the future.  I can walk you through what we hypothesize going in and then talk about how we predicted the future.  Certainly, things like internet access, which we think is a proxy for sedentary technology.  As internet access increases, it will positively influence rates of obesity.  Fast food restaurants or the number of restaurants, if it's easier to access the restaurants, people eat out more.  It would increase rates of obesity.  The relative price of healthier versus non–healthy foods, as healthier foods become more expensive, we thought people would switch out and consume more less healthy food.  If food, over all, is more expensive we worry that people would switch to the cheaper, more calorie dense foods.  That, too, would be increase obesity rates.  Same story with fast food prevalence.  Fuel prices, our hypothesis is tricky, but low fuel prices, people will drive more and walk less.  Alcohol, we figured alcohol certainly high calorie, so the prices of alcohol go down, people will consume more alcohol and they might end up gaining weight.  Those were generally the primary hypothesis and why we chose those. There′s a theory in why they matter.  Other things influence obesity rates such as programs.  We didn't have access to those data.  This is what we had.  So, essentially, what we did was looked at the trends and variables over the period of analysis and we modeled the trends.  We fit a best fit curve.  Sometimes it looks linear, other times curvy linear.  For example, internet access, you know, in 1990 when we started our study was 2% or 3%.  In 2008, it was well above 50%, closer to 60%.  Of course that could never go above 100%.  We think it would peak. We fit it within a reasonable logistic looking line.  We took the variables and took our predictions and ran them through the model.  That's how we came up with our estimates.  I want to make clear, predicting obesity is tricky.  No one variable showed up as being the cause of obesity.  If we knew what it was, we would change it.  These things, in and of themselves combine what is predicted.  The entire model explained 5% of the variation in rates of obesity in individuals.  It's not that surprising.  It's really hard to predict who is going to be obese and who is not.  There's genetics, environment and lots of factors.  So, you know, we think the model had some explanatory power. Predicting obesity is tough business. There are so many factors in play that this is a challenging test to undertake.  Is that helpful? 

Jeff McKenna: One more time to the phones.  Any questions on the phone? 

Operator: We have several.  The first is from Elise Vie beck from The Hill newspaper.  Ask your question. 

Elise Viebeck: Thank you very much.  I'm interested in the fuel price variable in terms of what you found on that question and also what your findings could mean for other policy makings.  Things outside the anti–obesity fight per se but it seems relevant talking about the power of the study.  Thank you very much. 

Eric Finkelstein: Fuel prices actually did not show up statistically significant in the models.  I can certainly speculate as to why it could be.  Maybe people didn't change their behavior.  We didn't see any real change there.  Now, that's not to say that maybe long term sustained fuel prices wouldn't generate changes.  Over a period of analysis and fuel prices we saw, we didn't see anything.  Maybe for policy implications, I don't know if you have comments, Dr. Dietz? 

Bill Dietz: No, I don't.  I think it's hard to speculate on policy implications of a model with projections. 

Jeff McKenna: Other questions from the phone. 

Operator: yes, a question from Katie Leslie.  You may ask your question.  

Katie Leslie: Good afternoon.  Thanks for taking my call.  I was hoping you could speak about your predictions for various parts of the country, specifically the south.  I believe we have higher rates of obesity.  Is there an expectation we'll have higher rates in obesity moving forward? 

Eric Finkelstein: It's a good question.  Now, I should say I don't have an answer for you.  Our model uses the state level variation to generate it.  It makes it harder to make predictions for a state.  Our model is a national model that uses data.  You could probably speculate, you know, if states have very different prevalence of fast food or very different changes in any of these variables we describe, demographics or relative prices how it might influence the results.  In this analysis, we were unable to take specific state level or regional level variations.  We needed it to come up with the national estimates.  Other questions on the phone? 

Operator: This comes from Nellie Bristol with congressional quarterly.  You may ask your question. 

Nellie Bristol: Hi, I was wondering, I know you say you can't take into account policies into any degree.  Could you look at things like deserts or menu labeling?  If you can take into account fast food, could you take into account more access to fresh fruits and vegetables and what that might mean for obesity? 

Bill Dietz: As I said, in this analysis, we used the variables we had access to at the state level.  If we didn't have access to it, that's why it's not in the model.  Now, I can speak specifically to menu labeling because I did a study on it.  In our study, we were unable to show in a particular fast food labeling that it changed anybody's behavior.  Other studies found modest effects.  I think the extent to which menu labeling is going to have an effect is yet to be seen.  We may find longer term studies to see the implications of that.  Same with other policies.  If we take them into account, I think we might have a better answer for you.  Within this paper, we weren't able to do it.  I don't know if you have other comments? 

Eric Finkelstein: No.  I think it's an important question.  There is modeling under way with a group that's called comvent. It's looking at how policy initiatives affect the prevalence of childhood obesity.  That work is just getting under way. 

Jeff McKenna: Other questions on the phone? 

Operator: No further questions. 

Jeff McKenna: Final call, questions from the room?  David and Lauran. 

David Brown: Yes, David Brown, again.  Can you describe a little, you mentioned the different variables explain 5% of, I guess, what predicts obesity.  You know, again, to a layperson, it doesn't seem like a lot.  Yet, you also said the variables were really kind of where the action was or in terms of, you know, setting the slope and where you end up.  So, if you could sort of explain that and what sort of things, obviously genetics is one of them.  What sort of explains the other 95%? 

Eric Finkelstein: That's a tough question.  I mean, explaining 5% of variation in hard to understand things is actually about –– it may sound bad, but it's not that far out of the ordinary.  I think you were trying to predict who is going to be an alcohol or get cancer, you wouldn't do much better.  It's hard to do.  I don't really have a good answer to you.  I mean, I don't know if you can think of a variable that is obvious that we are missing.  I'm not sure what it is.  If we put it in our model, I suspect it wouldn't do much to our results.  We know genetics.  People say genetics explains up to 70% of variation in an individual's weight.  If you believe that, it means 30% is left to be explained.  Now, Justin, for example, published on peer effects.  Maybe what's happening in school or your community is explaining the extra couple percentage points.  I wish we could have done a better job on built environment.  I think it has a significant effect.  The ability to walk and be active in your community would matter.  I think, you know, yeah, I was thinking for kids, certainly the school environment is the single largest factor.  There's certainly data to show it matters for kids.  For adults, a work sight wellness program may have an effect.  I have been involved in work studies.  We continue to study weight.  By and large, what we see with all weight loss programs is you see a short term effect.  Long term sustained weight loss is still largely elusive.  One thing that is fair to mention is we don't talk about medicine or devices, so, certainly, for some people, drugs that they take that have weight gain for side effects or drugs for weight reduction, those things actually have been proven effective.  It matters to the extent as they change in the future, new technologies and programs come online, if we can include them in the model or real life, we would see those things matter. 

Lauran Neergaard: Could you give more discussion of the severe obesity?  That was a scary number to be doubling.  What may be behind that? 

Eric Finkelstein: I think that's a great question.  I think it's something we haven't talked so much about.  So, if you look at the rise in obesity prevalence going back to the 1970s to today, people have certainly seen the scary slopes and slides from CDC.  I think lost in that discussion is among the group of individuals with a BMI over 30, their average weight continues to increase.  In fact, the numbers show in a year or two.  In the last ten years, it's gone up tremendously.  Those individuals have significantly greater risk of diabetes and heart disease and in fact, a large reduction in mortality compared to those of a BMI in the 30 to 35 range.  They are more expensive.  So, for that group –– actually, the reality is other than surgery, there's not a lot that's proven helpful for them.  This is a group that's at risk of health implications, premature mortality and yet they are increasing at a greater rate than the rates of obesity.  You know, the explanation of that, I guess I would argue is that the world has changed in such a way that allows people to be that overweight.  I think 50, 60, 70 years ago, it's impossible.  You couldn't sustain the caloric intake to be that heavy.  Now you can.  So, given people's predisposition and gene interaction and inactivity, workplace technology and access to unhealthy foods, people can be that heavy and are continuing to do so at increasing rates despite of health implications.  It's a major challenge.  We need to concern ourselves with the cost of that as a population. 

Bill Dietz: Just to reiterate what I said earlier, there are few studies that is look at what happened to children that were overweight as children and remain overweight as adults.  The one exception is the study in Louisiana which has 30 years of follow up data on children who were overweight or obese.  As I said, about 50% of adults with severe obesity, meaning a BMI greater than or equal to 40 were children or adolescence with persistent obesity.  One of the things to add to what Eric said, if you look at the cost, they are proportionately increases or disproportionately among the obese. 

Jeff McKenna: We have another person on the phone.  Please ask your question. 

Operator: We have a question from China Millman with the Pittsburgh Post–Gazette.  You may ask your question. 

China Millman: Hi.  Thanks for taking my question.  I was wondering whether you considering income as a variable or whether there is data available so you could consider that? 

Eric Finkelstein: That's a good question.  We actually use unemployment rate in our model.  We used race and ethnicity and household income.  Average household income within the state.  We didn't look specifically at relative inequality.  Now, I suspect if we could get at fine neighborhood variable levels, this might have explanatory power.  At the state level, I don't think it would have helped us much.  There's so much within neighborhood variation that wouldn't be captured in a state level model.  To your point, there is research that shows being, sort of relative disparity within your community puts pressure on individuals and may have effects.  Our model wasn't able to take that into account. 

Jeff McKenna: One last call for questions in the room.  Thank you very much for your participation.  To remind folks, if you have additional questions or would like more resources.  Names are in your press packets.  They are available to line up the appropriate responses.  Thank you for attending today.  Thanks to everyone on the phone as well. 

Operator: This concludes today's conference.  At this time, you may disconnect your line. 

###
U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES