In a recent study, economics professor Dr. Gaye DeLong suggests that her findings reveal a positive association between childhood vaccinations and the prevalence of autism in the United States (U.S.). As you are well aware, vaccines have remained a hot topic in the field of autism. With firm believers on both sides of the fence, studies like these often stir up quite a bit of controversy. So before we get too invested in DeLong’s findings, let us take an in-depth look at the research methods behind this study, which are faulty to say the least.
In her study, DeLong set out to investigate the relationship between the rate of children receiving childhood vaccinations and the prevalence of autism within each U.S. state over time. Childhood vaccination rates were obtained from public estimates produced by the Center for Disease Control and Prevention (CDC), which can be accessed here. In the study, DeLong used annual estimates of the percentage of 2-year-olds within each state to receive the entire 4:3:1:3:3 vaccination schedule, as recommended by the CDC. It should be noted that the CDC’s estimates are based on annually conducted surveys that reach roughly 30,000 households with children in the relevant age range. Therefore, the childhood vaccination rates used in this study are approximations based on limited data, not the entire population.
To estimate the number of children with autism in each state, DeLong used Individuals with Disabilities Education Act (IDEA) Data, which can be accessed here. It should be noted that this data is likely not reliable for the purpose of estimating the prevalence of a disability as the data are not based on clinical diagnoses, but rather classifications instituted by school districts. To measure “autism prevalence” in the study, DeLong used data on students classified with autism as well as students classified with speech or language impairment (SLI), her rationale being that the two groups are closely related. Rather than analyze the data separately, DeLong lumped these distinct groups together in her analysis.
The decision to combine these groups is bizarre. As described in the IDEA, Part B Data Dictionary, SLI is “a communication disorder, such as stuttering, impaired articulation, a language impairment, or a voice impairment, that adversely affects a child’s educational performance.” While it may be possible that some students with autism have been classified as having SLI, this certainly does not justify the use of the entire SLI population to determine “autism prevalence.” Furthermore, as demonstrated in the IDEA data, SLI is far more prevalent than autism. For instance, in 2007 approximately 24,515 students, age 8-years-old, were classified with autism in the U.S. as compared to 192,570 with SLI. As you can see, the SLI data completely overwhelms the autism data. So why did DeLong decide to analyze these groups together? Would autism alone have demonstrated a positive association with childhood vaccinations? We can only speculate why this analysis was not included in the study.
To analyze the relationship between childhood vaccinations and “autism prevalence,” DeLong compared annual vaccination estimates for 2-year-olds in each state to statewide rates of autism and SLI in the year those children turned 8. For instance, vaccination estimates from 1995 were compared to rates of autism and SLI in 8-year-olds in 2001. To evaluate trends over time, vaccination estimates from 1995–2001 were measured against rates of autism and SLI from 2001-2007.
The results of the study found childhood vaccinations to be positively associated with autism and SLI. That is, greater percentages of children receiving the entire 4:3:1:3:3 vaccination schedule were found to be associated with increased rates of autism and SLI. When a positive association is found between two given factors, it is imperative to consider other possible explanations for the association. Remember, showing that two things are statistically related does not demonstrate that one caused the other. Researchers typically report such considerations as limitations to their study; however, DeLong fails to consider one possible alternative explanation for her findings.
DeLong used vaccination data starting in 1995 because that is the first year the CDC began publishing annual vaccination estimates. However, DeLong fails to mention that 1995 is also the year the CDC first introduced the 4:3:1:3:3 vaccination series, which added three doses of the hepatitis B vaccine to the recommended vaccination schedule (click here for a history of recommended vaccination schedules). An increase in the use of a newly introduced vaccination schedule is expected over the years following its release. And as previously mentioned, DeLong only included vaccination data for those who received the entire 4:3:1:3:3 series; children missing even one shot were excluded. Pairing this natural increase with the consistent rise in autism diagnoses that we have witnessed over the years may reveal a positive association; however, that does not necessarily mean that one had anything to do with the other.
Childhood vaccination is a controversial topic with high stakes in the field of autism. The danger in declaring a positive association between autism and vaccinations that is based on unsound research methods, like those described above, is that other findings that are grounded on more solid research methods may be discarded as a result. No matter what your beliefs are in regards to vaccinations, I urge you to take caution when interpreting the results of this study.
DeLong, G. (2011). A positive association found between autism prevalence and childhood vaccination uptake across the U.S. population. Journal of Toxicology and Environmental Health, 74, 903-916.