Detrimental effects of overestimating the occurrence of autism.

03Original abstract

Estimates of autism spectrum disorders have been increasing since Kanner (1943, 1944) pointed out an unusual cluster of affective symptoms in children. Specifically, epidemiological estimates have risen from 0.7 in 10,000 (Treffert, 1970) to as high as 1 in 150, as recently proffered by the U.S. Centers for Disease Control and Prevention (CDC; Department of Health & Human Services, 2007). Although the CDC cautions that their estimate is not a true national estimate, it is a remarkable finding, and if the estimates keep rising, they may eclipse prevalence estimates of intellectual disability in the United States, which are typically found to be slightly greater than 1% of the population (Braddock, Emerson, Felce, & Stancliff, 2001; Lee et al., 2001). If Treffert's estimate and the CDC estimate are both correct, one might conclude that autism has increased about 100-fold since 1970. An example of an actual reported increase in the prevalence of autism during a smaller time period was provided by Yazbak (2002, as cited in Rehfeldt & Cuvo, 2004) in which the number of schoolchildren in Rhode Island increased by 1,115% between 1994 and 2002.Is autism really on the rise? Detractors of the autism-is-rising perspective point to the broadening of the autism diagnostic criteria as responsible for the reputed increase (e.g., Fombonne, 2003). Proponents of the autism-is-rising perspective tend to ponder environmental explanations for the increasing incidence estimates, because a genetic influence alone could not account for such increases. It is easy to locate Internet sites positing causes such as toxic metals, chemicals, medications, infections, vaccinations, and certain foods. Others suggest that the recent estimates are correct, but because prior incidence was underestimated, no increase in autism has occurred. As noted by CDC Director Dr. Julie Gerberding (U.S. Department of Health & Human Services, 2007), “Our estimates are becoming better and more consistent, though we can't yet tell if there is a true increase in ASDs [autism spectrum disorders] or if the changes are the result of our better studies” (p. 1).Of course, knowing with certainty if autism is on the rise requires that we ascertain valid autism incidence rates (autism births in a given time period), but determining such estimates has been a challenge. There is no autism gold standard to reference, such as the chromosome aberrations responsible for Down syndrome, or more recently, Fragile X syndrome (Brown, 2002). Researchers are looking for a biological basis for autism, and they have been bolstered by advances in genetic research (see Gupta & State, 2007, for a review) and emerging anatomical and functional imaging studies (Geschwind & Levitt, 2007). An example of a possible biological difference in autism is circulating proteins found from a recent examination of serum proteins in the blood (Corbett et al., 2007). An explanation for the unequal sex ratio in autism (boys outnumber girls by a 3 or 4:1 ratio) might lie in the X chromosome gene skewness of autistic females (Talebizadeh Bittel, Veatch, Kibiryeva, & Butler, 2005). However, to date, no biological or genetic markers for autism have been identified.What if we are grossly overestimating the prevalence of autism? What if it is really 1 in 500 or 1 in 1,000? (Fombonne's [1999] estimate was 1 in 535, derived from a review of 23 epidemiological surveys from 1966 to 1998.) What would be the consequences of this mistake?For one, the search for causes of autism and a possible cure could be impeded if the data are clouded by misdiagnosis. For example, if current autism diagnoses are consistent with the CDC estimate of 1 in 150, and if Fombonne's (1999) conclusion is correct, it is reasonable to assume that in a current investigation of biological indicators of autism, less than one third of the data would reflect information pertaining to autism. Such a gross subject-selection mistake would greatly increase the likelihood of missing the cause.There could be other undesirable consequences of overdiagnosing autism in addition to it impeding etiological research. Although it is clear that early identification is key to early intervention, it is not difficult to imagine the unnecessary distress of parents and social stigma resulting from misdiagnosis. As noted by Maurice (1996, p. 4), “All parents who receive word that there is something wrong with their child experience fear and grief, but a diagnosis of autism seems commonly to produce an overwhelming degree of devastation and confusion.”There could also be undesirable ramifications regarding intervention research and treatment applications. For example, in a hypothetical intervention study, individuals wrongly diagnosed as having autism might be given treatments designed for people who do have autism. Accordingly, the implications of such treatment would not be clearly related to autism, likely yielding misleading conclusions about what treatments are effective for autism. A group treatment study with an autism treatment group composed of participants with autism and (unknowingly) participants without autism increases the chances that (a) an effective treatment for autism will go undetected and (b) an ineffective treatment for autism will appear effective.Overestimation of autism might have treatment implications for other disability groups. In December 2006, President George W. Bush signed a bill that increased federal autism funding by 50%, authorizing $945 million over 5 years (CBS News, 2007). Could this mean that the autism “epidemic” is shifting resources away from people with other disabilities? Suspiciously, from 1987 to 1994, the prevalence of intellectual disability in California decreased in proportion to the increase in autism (Croen, Grether, Hoogstrate, & Selvin, 2002). Given the changed prevalence estimates in California, one might speculate even further about this possibility of shifting resources away from people with other disabilities. Indeed, in a recent review of U.S. special education data, Shattuck (2006) found that declines in the administrative prevalence estimates of intellectual and learning disabilities were significantly associated with corresponding increases in autism prevalence. Both Croen et al. (2002) and Shattuck (2006) suggested that the increased autism prevalence was due to symptom substitution (reclassification of children from one diagnostic category to another).Clearly, prevalence estimates for autism and funding for the treatment and cure of autism have been on the rise. If there is more funding available for autism services, clinicians may be predisposed toward the diagnosis if it brings the child in contact with more services. In pointing to the methodological challenges in behavioral diagnosis of autism, Grether (2006) noted the likelihood of diagnostic bias associated with eligibility requirements for autism services. It is easy to see how estimates and funding could cycle upward together. Indeed, a recent study by Skellern, Schluter, and McDowell (2005) revealed that over one half of psychiatrists and developmental pediatricians surveyed in Australia upgraded children with ambiguous symptoms of autism to an official diagnosis of autism so they could receive enhanced services. However, if researchers discover that autism estimates are substantially inflated, and if those estimates are related to service availability, this is yet another potential detrimental treatment effect of overestimating the occurrence of autism: Would there be a retraction of funding such that people who really do have autism would lose services? Perhaps a “cry-wolf” effect would cause legislators to be slow to fund a true epidemic in the future. Given an average special education cost of approximately $40,000 per child per year for intensive autism intervention (Ganz, 2006), legislators might be disinclined to maintain such intervention funding, despite a cost–benefit analysis estimate by Jacobson, Mulick, and Green (1998) pointing to immense savings from early intensive behavioral intervention.To be fair, there are a number of counterarguments to the position in this commentary. With respect to the assertion that symptom substitution accounted for the lowering of intellectual disability prevalence while autism estimates increased, one can argue that improved diagnostic procedures permitted proper identification of individuals previously misdiagnosed as having intellectual disability (Cohen, personal communication, June 2007). Given this perspective, reclassification would be due to accurate symptom identification rather than expedient symptom substitution. This is not an unreasonable viewpoint, given that approximately 70% of individuals who have autism also satisfy the diagnostic criteria for intellectual disability (Sigman, Ungerer, Mundy, & Sherman, 1987).A more intriguing counterargument points to the possibility that researchers may have found the true prevalence of autism. Recent epidemiological studies on autism have become more scientifically rigorous, and the most recent of these pointed to prevalence estimates that are similar to the CDC estimate of 1 in 150 (see reviews by Frombonne, 2005; Grether, 2006). The most recent findings are derived from the same or similar diagnostic instruments and corroborated from several countries, which suggests that the estimates might have stabilized.Another consideration is the inherent difficulty in biological identification of a disorder. For autism, it is similar to a “chicken-and-egg” endeavor: Although the disorder may be determined by some biological mishap, researchers must first identify those who have the disorder (in this case, using a behavioral definition) before we identify biological marker(s). This task may be all the more difficult given the contention that there are multiple forms of autism (Gillberg & Coleman, 2000). If several subtypes of autism exist, the disorder might be similar to intellectual disability in the sense that numerous biological causes of intellectual disability have been identified. This multiple-subtype hypothesis would seem to be consistent with higher prevalence estimates.In summary, autism estimates have risen dramatically in recent years, but researchers do not know if these increases are due to a true upsurge, changes in diagnostic criteria, better epidemiological studies, or a combination of such factors. Moreover, no one knows if the true prevalence of autism is rising or stable, or if the expanded behavioral definition has impeded or facilitated biological research into the disorder. Despite the recent research suggesting that prevalence estimates may have reached asymptote, it is possible that sociocultural influences will push the estimates higher, and this practice could backfire. It is clear that overdiagnosing autism would produce unnecessary parental distress, but it also might impede the search for a cause, obscure treatment research, and misguide policy decisions. If the trend of increasingly higher estimates of autism continues, the probability of detrimental effects due to overdiagnosing the disorder may also increase. In a recent review of diagnostic and treatment issues related to autism, Shattuck and Grosse (2007, p. 134) concluded that, “The growing number of children diagnosed with ASD and the efforts of well-organized advocacy groups have increased pressure on policy makers and service systems to improve and expand diagnostic and treatment services.” Ironically, the rapidly increasing estimates of autism have fueled greater interest and resource allocation for the cause, treatment, and possible cure for the disorder, but overidentification of autism might also be hampering these aspirations.

Intellectual and developmental disabilities, 2008 · doi:10.1352/2008.46:243-246