Autism is a pervasive developmental disorder that is behaviorally defined and is characterized by impairments in three areas: social interactions, reciprocal verbal and nonverbal communication and the range of interests and activities (1). The current definition of autism has been refined and broadened as compared to the original description of Kanner's from 1943 (2). Nowadays, persons with autism are considered to present with neurodevelopmental abnormalities that have such wide range of behavioral symptoms and severity that they are collectively referred to as pervasive developmental disorders (PDDs) in the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders, Text Revised (1). Within the group of five pervasive developmental disorders, a narrower term of autism spectrum disorders (ASDs) is used to refer to: autistic disorder, Asperger's disorder and pervasive developmental disorder – not otherwise specified. In this article we will use the term "autism” and "autistic” broadly to refer to the whole range of ASDs.

Since Kanner's 1943 classic paper (2) describing a pattern of behavior that he referred to as "early infantile autism” and through the 1990s, autism had been thought of as a rare condition (3). It is not the case nowadays. Scientists and journalists refer to "an epidemic of autism” (4). Autism is said to be more prevalent than spina bifida, cancer, or Down syndrome (5). Fombonne (6) writes that available data places "PDDs, particularly autistic disorder, among the most prevalent medical conditions of childhood” (p. 7). Research on the incidence/prevalence of this disorder is important for a better understanding of its etiology, assessment and services available for individuals with autism. Knowledge of the incidence (i.e., the number of new cases occurring over time) would be more informational about the alleged "epidemic”, but incidence studies would require identification of a definite time of onset in a defined population at risk. This is a difficult task because the wide range of behaviors necessary for a diagnosis of autism may not be apparent in all children at a specific age. Thus, most of the studies determine prevalence (i.e., the extent of a problem across a population at a particular point in time) rates.

The first three epidemiological studies report prevalence of autism in the range of 0.7 to 4.5 per 10 000 children (7, 8, 9). However, over time there has been a marked increase in those rates (10, 11). The US Department of Education reported an astonishing 556% rise in prevalence of autism between 1991 and 1997 (12). Several review studies indicate that the current rate for PDD is between 30 and 60 cases per 10 000 children, with a quarter of those meeting the full criteria for autism (13). According to Fombonne (11), the best estimate for the prevalence rate of autistic disorder is 10 per 10 000. The Centers for Disease Control reported in 2007 an estimated prevalence rate for ASDs 1 case in 150 children, and in 2009 – 1 case in 110 children (14, 15). Despite the fact that most of the information comes from the United Kingdom and the United States, epidemiological data have been gathered from 14 countries (6) and increases are reported worldwide (16).

Wing and Potter (3) list several potential reasons for this marked rise in prevalence: 1) changes in the diagnostic criteria, 2) varied methodology of epidemiological studies, 3) increased professional and public awareness of autism, 4) expansion in therapeutic and educational services, and 5) genuine rise in incidence of ASDs. For the purposes of the present paper, we will elaborate on the last assumption. If there were a true increase in incidence of ASDs it could be attributable to some environmental hazard (13). Many suggestions have been made considering the influence of environmental factors on autism: prenatal exposure to chemical agents such as thalidomide and valproic acid, as well as to infectious agents such as the rubella and influenza viruses, postnatal influences of diets, environmental pollutants, antibiotics, vaccinations, and neurotoxins such as mercury present in preservatives used for some vaccines (3, 17, 18, 19). None of those, however, has been yet scientifically validated. The purported link between measles-mumps-rubella (MMR) vaccine and ASDs has received much public and political attention (20, 21, 22) and will be the focus of the remainder of the article. Information presented below is a summary of several review articles and reports (23, 24, 25, 26, 27, 28, 29, 30, 31).

The hypothesis relating MMR immunization and the onset of symptoms of ASD was advanced by Andrew Wakefield and his colleagues in an 1998 article (32) describing 12 children with inflammatory bowel conditions and developmental disorders, primarily autism. For 8 out of the 12 children either the parent or the physician associated the MMR vaccination with the onset of behavioral problems and regression in the child's functioning. The average latency between the receipt of the injection and the occurrence of the symptoms was 6 days. Additionally, 9 children were diagnosed with lymphoid nodular hyperplasia in the terminal ileum as determined by endoscopy. A hypothesis was put forth that there is a new variant of ASD (regressive autism characterized by gastrointestinal symptoms) that originates from the MMR vaccine. The authors proposed the following sequence of events: 1) MMR produces inflammation in the intestines, 2) inflammation in the gut results in the change in intestinal barrier function that allows for the passage of toxic neuropeptides, 3) peptides disregulate the endogenous opioid system and subsequently cause central nervous system damage, which in turn results in developmental regression. This hypothesis has been called "gut-mediated toxic encephalopathy hypothesis” or "Wakefield hypothesis” (27). Despite the fact that Wakefield's study was heavily criticized on methodological grounds (i.e., small number of cases, no unaffected comparison group, possibility of a coincidental, not causal, temporal relation between the MMR vaccine and autism), the fact that subsequent studies by the same group of researchers did not support the original hypothesis (33) and the fact that in 2004 10 of the 13 authors of the 1998 paper asked to "formally retract the interpretation placed upon these findings” (34), the original publication raised great interest and public attention with regards to safety of the MMR vaccine. The result was a drop in the number of children immunized. In England, for example, MMR immunization rates dropped from greater than 90% prior to 1998 to a low of 75% in 2003-2004 (28) and cases of measles increased from 56 cases in 1998 to 1.370 cases in 2008 (35). Suspicions about the MMR vaccine spread to the USA. Many parents refuse or delay administration of the vaccination and ask the physicians to use three separate vaccines instead of the three-in-one shot. Needless to say, Wakefield's article had a far-reaching impact that had to be mitigated by epidemiological studies investigating in-depth the purported link between MMR and autism.

There are at least 20 epidemiological studies related to MMR and ASDs that have been published since the 1998 Wakefield's article. Sixteen of them were closely scrutinized and evaluated in the Immunization Safety Review. Out of the 16, nine are controlled observational studies (36, 37, 38, 39, 40, 41, 42, 43, 44), four are ecological studies (45, 46, 47, 48), and three are studies based on passive reporting system (49, 50, 51). Five other studies (52, 53, 54, 55, 56) were not included in the ISR review, but are discussed in other reviews. The 20 studies were conducted in six different countries (the United States, the United Kingdom, Finland, Denmark, Sweden, and Japan), used at least 12 distinct data sources, and employed a variety of study designs, including time-series analysis, cross-sectional analysis, ecologic analysis, case control, and retrospective cohort. The majority of studies were designed to address specific hypotheses that stemmed from Wakefield's study. These hypotheses are:1) ASD rates are higher among children who have received the MMR vaccine as opposed to those who have not, 2) increased rates of ASD occur as a consequence of the MMR vaccine, 3) the onset of ASD is temporally associated with receipt of the MMR vaccine, and 4) there is a new variant of ASD related to the MMR vaccine (31).

Only one study examined Hypothesis 1 (40). Danish researchers conducted a retrospective cohort study of all children born in Denmark between January 1991 and December 1998. A total of 537 303 children were included in the cohort, 440 655 (82%) of whom were vaccinated with the MMR vaccination. The researchers analyzed the relative risk of autistic disorder and other ASDs in vaccinated and unvaccinated children. Analysis was adjusted for age, calendar period, sex, birth weight, gestational age, mother's education, and socioeconomic status of the family. The results showed no statistically significant differences in rates of autism and ASDs in those two populations. Additionally, there was no relation between the age at the time of vaccination, the time since vaccination, or the date of vaccination and the development of ASD. The authors concluded that their "study provides three strong arguments against a causal relation between MMR vaccination and autism” (p. 1480): 1) the risk of autism was similar in vaccinated and unvaccinated children, 2) there was no temporal clustering of cases of autism at any time after immunization, 3) neither autism nor other ASDs were associated with the MMR vaccination.

Six studies examined Hypothesis 2 (39, 43, 44, 45, 47, 48).

Dales et al. (45) investigated whether there is correspondence between the trends in MMR coverage and numbers of ASD cases. The authors conducted a retrospective analysis of MMR immunization rates among children born between 1980 and 1994 who were enrolled in California kindergartens and cases of children born in these years who were diagnosed with autism and were enrolled in the California Developmental Services system. The results show essentially no correlation between those two variables. Between the years 1980 and 1994, the increase in the coverage of the MMR vaccination was 14% and the increase was observed for the cohort born in 1988 – before that year and after that year the data were stable. As for the autism caseloads, there was a steeply increasing trend (a relative increase of 572%) beginning in 1985 and continuing to 1994. The authors concluded that their results "do not support the hypothesis that increasingly widespread MMR immunization of young children is associated with the marked secular trend of increasing number of autism cases” (p. 1185).

Fombonne and Chakrabarti (39) conducted a cross-sectional study to examine whether a new variant of autism, characterized by regression and bowel symptoms, is associated with MMR. The authors stated that if there were a new phenotype of autism, at least one of the following six predictions would have to be supported by empirical data: "1) childhood disintegrative disorder (CDD) has become more frequent, 2) the mean age of first parental concern for autistic children who are exposed to MMR is closer to the mean immunization age than in children who are not exposed to MMR, 3) regression in the development of children with autism has become more common in MMR-vaccinated children, 4) the age of onset for autistic children with regression clusters around the MMR immunization date and is different from that of autistic children without regression, 5) children with regressive autism have distinct symptom and severity profiles, and 6) regressive autism is associated with gastrointestinal symptoms and/or inflammatory bowel disorder” (p. 1). Three samples of children were used: the main sample (96 children born between 1992 and 1995 with a pervasive developmental disorder diagnosis, 99% received the MMR vaccine thus called a "post-MMR sample”) and two comparison samples (comparison sample 1 – 68 children born between 1987 and 1996 with a diagnosis of PDD, most of these children were likely to have been exposed to the MMR vaccination, thus called "post-MMR sample”); comparison sample 2 – 99 individuals born between 1954 and 1979 with a diagnosis of autism, none of them received the MMR vaccine, thus called "pre-MMR sample”). The experimenters tested the six hypotheses mentioned above and obtained the following results: 1) there was no evidence that CDD rate was increased among children who were exposed to MMR immunization, 2) there was no difference across the pre-MMR sample and the two post-MMR samples in the mean age at which parents became concerned about autistic symptoms of their child, 3) there was no evidence that regressive autism has increased in frequency, 4) when compared with parents of autistic children without regression, parents of children with regressive autism did not become concerned at an earlier age or at an age closer to the MMR immunization date, 5) no statistically significant difference was found between the group of children with regression and the group without regression, suggesting a great similarity for patterns and levels of autistic symptomatology, and 6) no association was found between gastrointestinal symptoms and regression. The authors concluded that there is "lack of evidence for a new phenotype of MMR-induced autism” (p. 7).

Gillberg and Heijbel (47) compared via a case series analysis proportions of autistic cases in high and low MMR coverage periods. The authors reanalyzed data obtained from a population study of autism performed in the late 1980s. Seventy four children (55 with the diagnosis of autistic disorder and 19 with the diagnosis of atypical autism) were divided into two groups based on era of birth, as a proxy for exposure to the MMR vaccine. The MMR vaccine was introduced for 18-month-old children in Sweden in 1982, so the pre-MMR group consisted of children born between January 1, 1975 and June 30, 1980 and the post-MMR group consisted of children born between July 1, 1980 and December 31, 1984. The authors hypothesized that the post-MMR group would be at higher risk of developing autism if there were a correlation between MMR vaccine and autism. The prediction was not correct, as the analysis showed that 47 children with diagnosed autism or atypical autism were born in the earlier period and 27 children in the later period. Thus, the authors concluded that there is not an association between MMR vaccine and autism.

Kaye at al. (48) investigated in a time series analysis the relation between increasing rates of ASD and changes in rates of the MMR vaccine coverage in the UK. The authors obtained from the general practice research data base information about 305 cases of autism among children aged 12-years or younger, who were diagnosed in the years 1988-1999. The MMR vaccine was introduced in the UK in 1988. The estimated annual incidence of diagnosed autism had increased sevenfold from 0,3 per 10 000 person-years in 1988 to 2,1 per 10 000 person-years in 1999. Further analyses were conducted for 114 boys born between 1988 and 1993 and diagnosed between 1990 and 1999 in an attempt to assess more precisely the possibility of a temporal association between MMR vaccine and autism. The four year risk of diagnosed autism increased nearly fourfold, from 8 per 10 000 for boys born in 1988 to 29 per 10 000 for boys born in 1993. In contrast, the rates of MMR vaccination were stable (about 97%) for each successive annual birth cohort. The authors explain that if the MMR vaccine had played a role in developing autism, the risk of autism in successive birth cohorts would be expected to stop increasing within a few years from when the vaccine began in full use. The results speak to the contrary. Thus, the authors concluded that their study provides "evidence against a causal relation between MMR vaccination and the risk of autism” (p. 462).