John W. Jacobson, PhD, BCBA
One of the questions heard often in the field of developmental disabilities today is, “What causes autism spectrum disorders (ASD)?”
Questions of cause are very difficult questions to address with full candor and confidence, not only in the case of autism spectrum disorders, but also in other conditions for which diagnosis is made based primarily on behavior. For example, some individuals speculate or assert that biological and/or toxic agents in the MMR vaccination cause autism. This article will focus upon that controversy to explore the complexities of cause and effect.
The scientific method is a process by which researchers seek to answer a variety of questions. Some of these questions do not involve causes, while others do. For example, neurological or neuroscientific studies may attempt to determine differences typically present in the structure or neurochemical features of the brains of children with ASD, compared to the brains of same-aged children without diagnosed disabilities. In other words, the question posed is not what causes autism, but rather what differences exist in autistic versus non-autistic brains.
On the other hand, some studies do attempt to identify causes such as causes of a disability. To do so, certain tasks must already have been accomplished. For example, the disability must be well-defined. The decision that the disability is present and that the disability is not present must be accurate. The group that is defined as having the disability must be as homogeneous as possible. That means, for example, researchers should consider the question of what causes autism separately from the question of what causes PDD or Asperger syndrome. This is not to suggest that autism, PDD and Asperger’s are not a spectrum as typically thought, but rather that the question “what causes autism?” is a much more specific question than “what causes autism spectrum disorders.” There is also the consideration that, because the diagnosis of autism and other ASDs is based on breadth and severity of effects, logically, differences in brain structure or other features should be more apparent for these children.
Assuming that we can accurately decide whether children have autism, and we are able to identify biological differences between these children accurately as well, we could then approach the question of cause in two ways. Taking into account possible multiple medical or neurological factors, we can conduct a detailed investigation into the backgrounds and developmental history of the children, identifying events that differ between diagnosed and non-diagnosed children. In this approach, we start with a group of children who are identified, and then look back on their histories. This is termed a retrospective study. Retroactive studies are important to conduct, but there are also challenges associated with them, including documenting that reported events occurred, reliance upon incomplete or differing clinical records, and the fact that universal screening for disabilities like autism usually does not exist, so only the children who happened to be identified and referred are included in the study. Children who are referred are likely to differ in several ways from those who are not referred, and some of these ways may be related to risk factors for autism.
The preferred method for conducting a study of causes is to use a prospective approach. Prospectively, one begins by using outreach to screen a population of children for a disability, to detect instances of the disability that might not otherwise be identified, as well as those who would have been identified. Then, background and history data can be collected on a group that is more likely to actually represent children with autism or any other condition. Historical information could include vaccinations and other medical events (e.g., exposure to general anesthesia, recurrent health conditions), as well as information about child development. More ideally, a prospective study would follow children from birth, and all of the information needed to consider causes could be collected as events occur. This would assure that the information is more complete, and in a standardized form. In the case of autism, even though it is now being identified more frequently, this approach is very difficult to carry out; for each child who may develop autism, there may be from 250 to 1,000 other children who need to be screened and followed. For this reason, some researchers may attempt to use prospective approaches to study causes of autism within larger studies that look at child development in large population groups, and consider a variety of disabilities. This often means that information that is specific to risks for ASD may not be fully collected.
Of course, there are other research designs that can shed light on causes of disabilities. Certainly genetic studies can indicate genetic factors that increase risk for a childhood-onset disability. Research with animals that involve brain surgery during early development that result in behavioral changes akin to those typical of a disability may also be suggestive. Basic research at the level of neurons and the effects of toxic substances and side-effects of medications may also be suggestive. But . . . there is no substitute for actually studying the occurrence of a condition among children prospectively.
Why do we need scientific studies to indicate what the causes of a condition like autism might be? Why isn’t it enough that some research might identify some differences between children with autism and their peers? First, some differences that are identified initially do not necessarily differentiate children with autism from those who are accurately diagnosed as not having autism. For example, research findings have suggested the unexpected presence of measles virus in the gastrointestinal tracks of children with autism, but subsequently at least one report has found this for children without autism as well. This does not mean, in and of itself, that the initial gastrointestinal findings are not possibly suggestive, but does point out the need for careful assessment of the likelihood that particular factors are plausible risk factors. In this case, scientific research needs to address why, if gastrointestinal measles is a risk factor or cause, or reflects a risk factor, some children are affected, and others are not.
But, if many people develop a consensus that a given event–vaccination, for example–is regularly observed to occur shortly prior to detection of autism, is this not sufficient to warrant research on this issue? The short answer is yes—whether observed by parents, clinicians, educators, or researchers, events that may be plausible causes or risk factors for a disability should reasonably be studied. Parents or others in the lives of children with disabilities may certainly detect events that are not apparent or considered by clinicians or researchers. But the fact that a belief is widely held is not, in itself, evidence that the belief is valid or accurate.
The brains of human beings are structured and function in ways that are the joint product of evolution and experience. One of the well-known biases associated with human perception and thinking is the tendency to conclude that there is a cause and effect relationship between two events, when it can be shown through precise research that this is not the case. Carl Sagan, in his 1997 book, The Demon-Haunted World: Science as a Candle in the Dark, referred to such tendencies as “irreducible human error.” To err in this manner is human, but to insist that reliance on mere consensus is sufficient to accurately identify causes of events, such as the occurrence of autism, is folly. Errors of this type may be even more likely when the identification of a chain of cause and effect is especially important to the person making a judgment about cause and effect; many parents of children with autism believe that identifying the causes of autism, for their child and other children, is important. This may increase the chances that some or many may conclude that certain, unproven events are causes of the disorder, without solid evidence. However, this is a very human thing to do, and clinicians and researchers are prone to do this as well.
The critical distinction that needs to be made is between correlation and causation. Correlation means that two events tend to occur together. When one does not occur, the other tends not to occur as well (called a positive correlation); or that when one occurs, the other tends not to occur (called a negative correlation). Sometimes correlations, like cause and effect, are perceived accurately, and sometimes they are not. But while necessary for showing cause and effect, correlation does not prove cause and effect. Sometimes correlation might be presumed, because of cultural factors; for example, autism is often diagnosed, by definition, at ages when children are subject to frequent vaccinations. Thus vaccinations and autism could be hypothetically correlated, despite the fact that there is no present scientific evidence that this is the case. Correlation does not in itself show causation, because the fact that two events occur together may be influenced, or caused, by a third factor that has been ignored, or that was not studied.
Causation, on the other hand, requires a higher standard of proof than the fact that two events occur together (that is, have a positive correlation). Proving causation, or that an event is a risk factor for a disability, requires that several conditions be met: (1) the purported cause has to consistently or always occur before the purported effect; (2) when the purported cause occurs, the effect regularly occurs; and (3) when the purported cause does not occur, the effect tends not to occur, is less likely to occur than it does generally, or does not occur at all. Other criteria associated with the strengths of prospective studies also need to be met; for example, that the group of people studied is representative of the larger group of people with the condition (in this case, all children with autism or all children with ASD). This can be done by including all children in a general population with the condition, or by randomly sampling the children with the condition. But, if sampling is used, there also must be a sufficient number of children to generalize to the larger group of children, and the required number to do so increases as the complexity and range of issues under study increases.
Where do we stand today in understanding the causes of autism? It is fair to say that researchers are developing a more complete understanding of the neurological factors associated with autism, but some degree of modesty is also appropriate with respect to the predictions that can be made or confidence with which particular neurological findings can be said to characterize autism. Many neuroscientific studies focus on specific aspects of the brain. Therefore, different aspects of the brain have been studied in different samples; there is seldom concrete evidence that these samples are very much alike, or that they represent a larger group of children with autism. This points out the need for independent researchers to conduct studies with other samples, to verify that the findings with one sample also apply to others.
In addition, many neuroscientific studies include small numbers of subjects. As a result, such studies are not able to detect relatively subtle but consistent differences that may exist between individuals who have autism and those who do not, and the studies may not be representative of children with autism more generally. Advances in research design, including identification of subjects with better measures, are addressing these limitations. Neuroscientific knowledge about autism is steadily advancing, but there are, nonetheless, considerations that affect the strength of the conclusions that can be drawn today.
One must also consider that the group of children diagnosed with autism is heterogeneous: some also have diagnoses of mental retardation, while others don’t; some have seizure disorders, while others don’t; some manifested regression or loss of attained skills, while others did not. Although there is a strong (and warranted) presumption that genetic factors play a strong role in the occurrence of autism, the heterogeneity of children with the condition and current research findings suggest that the relevant genetic factors are complex and multiple in nature. At this point one may reasonably argue that the behavioral condition of autism and ASD are final common pathways, or results, of differing genetic factors—that there is no single genetic factor that accounts for occurrence of the condition. Events prior to birth have also been implicated by neuroscientific studies. It may also be that in some cases, environmental events, such as reactions to toxins, may play a role. It may be that all of these factors, and others, are involved as risks or causes.
The Scientific Method
- Observe and describe a phenomenon or group of phenomena.
- Formulate a hypothesis to explain the phenomena.
- Use the hypothesis to predict the existence of other phenomena, or to quantitatively predict the results of new observations.
- Perform experimental tests of the predictions.
- Modify the hypothesis based upon the test results.
- Repeat steps 4 and 5.
- Replicate the tests by several independent experimenters and properly performed experiments.
Note: This article originally appeared in an issue of Science in Autism Treatment, the newsletter of the Association for Science in Autism Treatment (ASAT). It may not be republished or reprinted without advance permission from ASAT. Email us for reprint permission.