Ten Times the Prevalence of Severe Autism over Time, But No Increase vs. 2008
By Katie Weisman, for the SafeMinds’ Research Committee
Thanks to Mark Blaxill and Cynthia Nevison, PhD for the graphs.
On March 28th, thousands of media outlets released the new US autism prevalence numbers of 1 in 68 in 8-year-olds born in 2002 and counted in 2010. These children are 12 years old now. What was glaringly lacking in the media coverage was any critical thought about that actual data, any sign that reporters had actually read the new report or any sign of urgency on behalf of our children.
The CDC’s take away messages were as follows:
- We’re reporting a 30% increase in autism in 2 years, but you don’t need to worry because these kids have always been here. We are just better at counting than we were before.
Comment: There has been a 37-fold increase in reported autism spectrum disorders in the past 30 years – which would be about 1984, hardly the dark ages. Do you really believe that there were this many kids with autism around when you were young? The prevalence was 1 in 2500 in the early 80s or .4/1000; as of this month, it is 14.7/1000. The CDC has yet to conduct a population-based count of people with autism of all ages and severities which would lay this issue to rest. We should all be asking why they haven’t. They continue to say that they still can’t be sure if the increase is real – over and over again – for the past decade.
- It looks like the kids are getting milder – they have less intellectual disability (ID) – and it’s due to broader diagnosis.
Comment: Up until the 2006 data, the ADDM reports only stated the percentage with IQ’s below 70, which were stable (on average) in the early reports (see chart 1) but the averages masked a wide spread in the actual percentages by state. Breakdowns into three categories of ID started in the 2008 data. The percentage of children diagnosed with autism in the newer reports (see chart 1) is consistent with the percentage with ID in the older reports, but the CDC did not supply data on diagnostic categories in the past. What is unclear is what is driving the change in the ID of the children. Are we seeing children who still have autism but are less language impaired?
- There is no possibility that autism and vaccines are connected because the numbers are still going up.
Comment: Assuming that there is, in fact, a smaller percentage of ASD children with Intellectual Disability, those shifts do correspond to the beginning of the phase out of thimerosal in vaccines, which is a plausible explanation as well – but one CDC doesn’t mention. Thimerosal reduction in the recommended childhood vaccine schedule (HepB, Hib and DTaP) started in 1999 and it was phased out over several years. However, shortly thereafter, in the 2002-2003 season, the CDC started encouraging flu shots (most of which contained thimerosal) for infants 6-23 months and in the 2004-2005 flu season flu shots were formally recommended for all infants starting at 6 months of age. Meanwhile, the CDC and ACOG also added influenza vaccines (most of which still contained thimerosal) to the recommendations for pregnant women in all trimesters in 2004. This FDA letter makes clear that thimerosal-containing infant vaccines would still have been administered throughout 2002 – the birth year of the current ADDM report but at amounts, on average, probably less than in the 2000 birth cohort. The exposure to any particular child is an unknown without checking their history.
- If you are a young parent, check out our website to learn the signs of autism and talk to your pediatrician if you have concerns. Early intervention is the answer.
Comment: Where is the interest in prevention? This approach is letting a child fall off a cliff and then trying to catch him. Despite the CDC’s “Learn the Signs” program, the average age of diagnosis in this report is the same as it was a decade ago in the 2000 data – about 53 months or age 4.5. There has been no measurable progress in this area.
Preface – Limitations of the ADDM numbers:
As outlined in the SafeMinds commentary on March, 28th, there are significant limitations to the quality of the data in the ADDM network overall, which I will not repeat here. My purpose here is to point out important data that was missed, further weaknesses in the reports and to show that there are broad questions regarding what is included and excluded that should arise from looking at the ADDM reports. The “spin”, deletion and addition of sections and overall characterization of the reports should be questioned by all those who care about someone with autism.
The chart below is a summary of some key information from the ADDM network over the years. It includes the number of sites included in the ADDM, the average age of first diagnosis with an autism spectrum disorder, the percentages of IQ levels as reported in ASD cases in select states (typically the states in which they have IQ data for more than 70% of the cases), the percentage breakdown of the three primary ASD diagnoses and documentation of regression data.
[Click image to open in new window]
Discussion of Intellectual Disability Reporting:
[Click image to open in new window]
The CDC made a big point of promoting the idea that autism prevalence is still going up while the percentage of kids without intellectual disability is rising and the raw numbers on the chart support that trend. However, here is a quote from the 2008 report that provides a different perspective:
In the four sites with IQ test data available on at least 70% of children with ASDs in both the 2006 and 2008 surveillance years, the estimated prevalence of ASD with intellectual disability increased 12% on average (4.2–4.7 per 1,000), while the prevalence of ASD with borderline intellectual ability increased 22% (2.3–2.8 per 1,000), and the prevalence of ASD with average or above-average intellectual ability increased 13% (3.9–4.4 per 1,000).
In other words, if you actually compare apples to apples in the same 4 states, ASD with and without ID increased about the same percentage – certainly not a compelling shift between the two ends of the range and showing the importance of comparing the same states.
No Prevalence Increase in the With Intellectual Disability Subgroup
Also, notice that the “with intellectual disability” prevalence number just reported for the 2002 birth cohort is 4.7 per 1000 children, which is exactly the same as the prevalence reported for the 2000 birth cohort in the quote above, 4.7 per 1000. It is only 12% (5.8% annualized) more than the 1998 birth cohort prevalence of 4.2 per 1000. Unfortunately, the 1992 and 1994 birth cohort reports do not give prevalence statistic for just those with ID so we cannot compare farther back. However, the data that we do have suggests that, after a long period of annualized 8.5% increases, to get from .4/1000 to 4.7/1000, the prevalence of the most severe autism may be flattening – though at a level too high for anyone’s comfort. It is critically important that we look at actual prevalence comparisons in subpopulations and not just at the percentages of the whole population with ASD.
[Click image to open in new window]
Removal of South Carolina and Alabama
This year, after being included in all previous ADDM reports, South Carolina data was not reported due to “not providing suitable data in a timely manner”. No further explanation is given. It seems a little unusual that an experienced team would suddenly be unable to produce the work that they had been doing for a decade. Given that SC had the highest percentage of ID in 2008 at 54% that would have decreased the percentage disparity in the ID categories. Also, Alabama, which despite only 10% access to education records had recorded a high percentage of ASD children with ID in the past, is not part of the current reporting. See Chart 1B below.
[Click image to open in new window]
Change in States Reported Over Time for Intellectual Disability Analysis
To further complicate comparisons of intellectual disability over time, consider the above chart of which states were reported in each birth cohort. The percentages are for those with ID. For each birth cohort, I have indicated the percentage of intellectual disability (IQ<70) in bold if it is mentioned in the report specifically. For some reports, the only source of a percentage is a bar graph so I have put those percentages in regular font as they are taken from the graphs with an approximate adjustment for the ratio of boys to girls in the typical cohort. N/A in the chart means that the state was included but due to the limited nature of the 2004 report (actually an addendum to the 2006 report and issued at the same time), there was no bar graphs to pull approximate percentages from.
Note the following:
The percentage of states included in the ID analysis has varied from 50-64% of the states included for that year.
There has never been a year when the same states were included so one needs to look at “apples to apples” comparisons.
There is huge variation state to state (13-63%) in the percentage of children with ID, which has yet to be explained or investigated. Since we are looking only at cases with ID, there should be less effect of “broader diagnosis” and these variations may reflect actual differences in severe autism due to environmental exposures. This area is ripe for more research.
Here is the direct quote from the 2010 report:
Over the last decade, the most notable change in characteristics of children identified with ASD through the ADDM Network is the growing number who have average or above average intellectual ability. This proportion has increased consistently over time from 32% in 2002, to 38% in 2006, to 46% in 2010, or almost half of children identified with ASD. Concurrently, the proportion of children with ASD and co-occurring intellectual disability has steadily decreased from 47% in 2002, to 41% in 2006, to 31% in 2010. This shift in distribution of intellectual ability among children identified with ASD during 2002–2010 indicates that a large proportion of the observed ASD prevalence increase can be attributed to children with average or above average intellectual ability (IQ >85).
Given the above, do you feel that the CDC has accurately reported the situation? Or is this “spin” to say the numbers are still going up, but you don’t need to worry?
Elimination of Regression Reporting
In the early years of the ADDM network, regression was reported consistently in about 13-30% of the children counted. They also reported a plateau in development in an additional 3-10% of the cases. The relative consistency is noteworthy (compared with much of the data reported) in the 2000, 2002 and 2006 collection years (see Chart 1). Then in the past two reports for the 2008 and 2010 data, the entire section of the study reporting regression breakdown was eliminated. These numbers were based on actual documentation in the case files and a quote in the 2002 report states, “Therefore, these results should be considered a minimal estimate of plateau and regression among ASD cases”. In 2009, the ADDM team did a separate study looking at regression in the network and reported the following:
This study evaluated the phenomenon of autistic regression using population-based data. The sample comprised 285 children who met the autism spectrum disorder (ASD) case definition within an ongoing surveillance program. Results indicated that children with a previously documented ASD diagnosis had higher rates of autistic regression than children who met the ASD surveillance definition but did not have a clearly documented ASD diagnosis in their records (17-26 percent of surveillance cases). Most children regressed around 24 months of age and boys were more likely to have documented regression than girls. Half of the children with regression had developmental concerns noted prior to the loss of skills. Moreover, children with autistic regression were more likely to show certain associated features, including cognitive impairment. These data indicate that some children with ASD experience a loss of skills in the first few years of life and may have a unique symptom profile.
http://www.ncbi.nlm.nih.gov/pubmed/19535466?log$=activity Why was the reporting of regression eliminated in the most recent reports, given the finding that these children might have a “unique symptom profile”?
Timeliness of Reporting the Surveillance
In a 2007 study (Braun et al.) of the 2002 data collection, CDC made the following statement:
Although the ADDM sites participating in the 2002 surveillance year represent multiple grant cycles, the estimated time required for this surveillance year, from start of funding to reporting of results, was approximately 3–4 years. Once the surveillance system has been instituted at a site, these limitations to timeliness are greatly reduced for future surveillance years. As ADDM Network surveillance methods have evolved, the time required to make data available has decreased. Multiple surveillance years can now be conducted concurrently, and clinician review has been restructured to increase efficiency.
So, if the time to actually make the data available has decreased, why are the 2010 data only being reported in the spring of 2014? There has been no improvement in a decade of the timeliness of the reporting of the autism surveillance. Our children deserve better.
Where is the Study of the 4 Year-Olds? The 2010 surveillance funding included collection of cases in 4 year olds with ASD in 6 states. The states included are Arizona, Missouri, New Jersey, Missouri, Utah and Wisconsin. Missouri and Wisconsin are states that don’t access education records, but this should be less of an issue in a 4 year old population since they are not school age yet. It will be interesting to see if the South Caroline data for the 4 year-olds is included or not. The primary question is when that data will be published. With fewer records to collect, shouldn’t it have been published ahead of the 8 year-old numbers?
Complete Elimination of the Within State Comparisons of Prevalence
In the 2008 ADDM report, there are 3 pages looking at within state comparisons of prevalence of ASDs by sex, race and IQ. The comparisons are done for various combinations of the 2002, 2006 and 2008 reporting years. This entire analysis is missing from the 2010 report with no explanation given.
I have created the chart below to try to get a bird’s eye view of the ADDM prevalences over time. At first glance, the following thoughts come to mind:
Arizona’s growth curve spiked between 2004 and 2006 and has been slowing ever since.
Arkansas’s growth curve spiked between the 2000 and 2002 birth cohorts. What happened?
Florida’s growth curve spiked between the 1998 and 2000 birth cohorts. More below.
Georgia’s growth curve doubled between the 2000 and 2002 birth cohorts. This is the longest running surveillance site and has had very steady growth over time. What changed?
Missouri’s growth curve spiked between the 1996 and 1998 birth cohorts and has come down to almost flat since then.
New Jersey’s growth curve may be flattening. However, the fact sheet on New Jersey reports a staggering 1 in 45 children with ASD; 1 in 28 boys and 1 in 133 girls who are now 12 years old. This was lost in the news reports.
Pennsylvania’s growth curve doubled between the 1998 and 2000 birth cohorts.
South Carolina’s growth curve had a huge spike between the 1996 and 1998 birth cohorts.
Utah reported a decrease in autism of almost 13% over the 2000 to 2002 birth cohorts. This is the biggest drop reported in ADDM in a state with good access to records. Yet there was no mention of it in the CDC’s press coverage. Interestingly, all the press out of Utah reported rates as “holding steady”; none mentioned the actual drop in prevalence.
West Virginia’s growth curve was dramatic from the 1992-1994 birth cohorts but there has been no follow up for 10 years.
All of these changes suggest the possibility of environmental factors, particularly the large spikes in certain 2 year periods. Again, more research is needed.
Chart 2 – Summary of State Prevalence Reporting in the ADDM Network – 1992-2002 Birth Cohorts
In the chart below, the shaded states do not have access to most educational records and have consistently been shown to underreport autism prevalence compared to states with access to education records.
Bolded percentages are annual percentage changes in prevalence rates. CDC typically reports two year changes at a time so these percentages are roughly half of what the media reports. It was necessary to do this to compare 2 year changes to 4 or 6 year changes, though these numbers will not reflect variations within the periods of time. We would have to have annual instead of biannual collection of data to know for sure what happened in a given state. The percentage annualized growth rates in prevalence are useful to see how the growth curves are changing in various states over time.
[Click image to open in new window]
What is going On in Florida?
In the 2008 report, there was a glaring signal that the CDC neither reported, nor followed up. They found 211 children with ASD of which 52.9% were Hispanic (Miami Dade County). The Hispanic prevalence rate was 8.2 per 1000 compared to a white prevalence of 4.2 per 1000. This is the only place in the 10 years of ADDM reporting that a Hispanic rate significantly (in this case almost doubled) exceeded the white rate. Typically, minorities are diagnosed at lower rates than white children in the same areas. The median earliest age of ASD diagnosis was 3 years, 6 months which suggests a more severe population and is lower than the median age of first diagnosis network-wide. The chances that this represents an autism “cluster” are strong and it should be investigated. Florida also reported an enormous prevalence change from 2006-2008 – a 71% increase in 2 years or 31% annually – the largest 2 year change ever in the ADDM reporting. Yet Florida was not funded for the 2010 cycle so we don’t know what happened next.
Press Involvement at the CDC and the Impact of PDD-NOS Non-Inclusion
Lastly, there is the question I found on the case status of children in the ADDM network. A January, 2014 study (Maenner et al.) investigating the potential impact of the DSM-5, based on ADDM data, included several interesting paragraphs:
To calculate the potential impact on prevalence, we applied DSM-5ASD criteria to 2 groups of 8-year-old children under surveillance for the years 2006 and 2008: (1) the 6577 children who met both ADDM Network ASD criteria based on the DSM-IV-TR and our operationalized DSM-5ASDcriteria and (2) the 1020 children who did not meet ADDM Network ASD criteria but could plausibly meetDSM-5 criteria. These 1020 children all technically met DSM-IV-TR criteria for PDD-NOS, but the clinician reviewers did not classify them as ASD cases for surveillance purposes; for most of these children, the clinician reviewers concluded that the behaviors were better accounted for by another disorder.
Children with a history of developmental regression were more likely to meetDSM-5ASDcriteria than those without a history (89.4% vs. 79.0%, P < .001), and children with intellectual disability were more likely to meet DSM-5ASDcriteria than children with an IQ greater than 70 (86.6%vs 82.5%, P < .001). Children with a history of regression remained more likely to meet DSM-5ASD criteria than children without a history of regression after controlling for intellectual disability.
Since excluding cases would have a significant impact on the prevalences reported in those two years, I wrote to Dr. Maenner with questions about those children. After a delay, I received an incomplete answer back from Dr. Maenner through the CDC press office. I do not normally put private e-mails online, but since this one has the press office’s blessing, that e-mail exchange is available here: http://www.safeminds.org/?p=6266
This chart shows the impact on what was reported in 2006 and 2008 of including those children who met DSM-IV-TR PDD-NOS criteria but were not included as cases.
[Click image to open in new window]
This graph shows the potential effect of the culling of PDD-NOS cases (in chart above) on the annualized growth percentages for autism. Whether or not that culling was present in the earlier and later cohorts is a key question. The CDC case definition has not changed over time. Per my e-mails with the CDC, this analysis has not been done for the earlier years. I have sent the question regarding the 2002 birth cohort and I am awaiting a response. If the culling was not done for the 2002 birth cohort, this suggests two things – that the prevalence growth curve is dropping in the latest cohorts and that the “without ID” percentages may have been bumped up in the 2002 cohort if the same ratio of PDD-NOS kids were not excluded.
[Click image to open in new window]
The Bottom Line
The CDC ADDM network reports raise far more questions than they answer and the changes over time make it difficult to answer even the most basic questions about what is going on with autism in the United States. The changes in the information reported, the states included and the way results are presented to the press should all raise red flags. If the information is accurate, then there are many opportunities for follow-up research on environmental factors that have been overlooked or ignored.
Overall, there are patterns that suggest CDC is doing its best to maintain that rates are still going up rapidly but can be explained by the inclusion of children with less intellectual disability while at the same time, arguing that vaccines (thimerosal in particular) did not play a role in this change. The real picture has a lot more shades of gray. It will be important to see what is happening with the 4 year olds when that data is published and to watch what is reported in 2016 based on the 2004 birth cohort, keeping in mind that the CDC already had that data in 2013, before releasing the current report.
For those interested: All CDC ADDM reports and community reports are available here:
Addendum on CDC Autism Spending:
The Centers for Disease Control show no urgency in adding more funding for autism despite their over 11 billion dollar total budget. I have gone through the CDC budgets as far back as they are available on their website: http://www.cdc.gov/fmo/topic/budget%20Information/index.html
A line item for autism was first created in the CDC budget in 2007. This chart shows the total autism spending to maintain the ADDM network, some additional research and the Learn the Signs program. Note that these expenditures are part of the annual appropriations provided to the CDC and are not appropriated under the CAA. Note that these numbers vary slightly from the CDC’s report to IACC which reports the following: FY 2010 – $24,710,050, FY 2011 – $23,942,225 and FY 2012 – $23,348,012. The
IACC numbers show the amount spent on autism slowly declining. The budget reports show funding basically flat for the most recent 4 years.
[Click image to open in new window]
Even given the 1 in 68 prevalence just reported, there is no urgency around finding out what is happening in the other 38 states. See page 370 for the CDC’s current plan to maintain only 12 ADDM sites for fiscal 2015.
See page 9 for the flat 2015 budget request for autism.
Considering the Costs:
- The costs associated with a child with autism average $17,000 per year in excess of a typical child. That increases to $21,000 excess for a child with severe autism. Overall, the estimated societal cost of caring for children with autism was over $9 billion dollars in 2011. (Lavalle et al. Pediatrics, 2014)
- The lifetime cost of care for an individual on the spectrum is 3.2 million dollars. Ganz, 2007.
- In 2005, the average annual medical costs of a Medicaid -enrolled child with autism were $10,709 – about 6 times higher than a child without autism. Peacock, et al. J. of Developmental Behavioral Pediatrics, 2012.
- In addition to medical costs, behavioral intervention for a child with autism can cost $40,000-$60,000 per year. Amendah, et al., Autism Spectrum Disorders, Oxford University Press 2011.
To put this whole thing in perspective, the cost of monitoring the other 36 states that have never been monitored should be about quadruple the current CDC budget or $92 million dollars. That would cost about $76 per individual under 21. However, if you use the Ganz number multiplied by 1.2 million children (not counting adults) then the overall lifetime cost to society of those children is $3,840,000,000,000 – 3.84 trillion dollars. It seems like a small price to pay to get a handle on the problem