Is MMR Linked To Autism? Epidemiological Perspectives
Testimony to the Congress of the United States of America
House Committee on Government Reform
April 25, 2001

In a briefing paper for the Institute of Medicine of the National Academy of Sciences of the USA, Soto and colleagues concluded: based on the epidemiological evidence to date and the opinion of the authors , the evidence is inadequate to accept or reject a causal relationship between MMR and autism. [Reference: Soto MA, Cleary SD, and Foster VB, Commissioned Backround Paper, Institute of Medicine Immunization Safety Review Committee, 2001] As summarized in this document, submitted to the House Committee on Government Reform of Congress, on April 25, 2001, I reached the same conclusion independently. My views were peer reviewed shortly before the release of the I.O.M. s briefing document and will be published by Oxford University Press this month. [Reference: Adverse Drug Reactions and Toxicological Reviews, in press].

At the turn of this millennium (December 00 and January 01), the public debate about the safety of trivalent measles-mumps-rubella (MMR) vaccine has been intense. A handful of papers with strong views have been published in the medical press. Extensive coverage in the lay press and electronic media has now brought the controversy to the center of the public s attention. I refer to only one example of lay press coverage, a headline in The Time of London, Biggest study clears MMR [12-XII-2000, p.1].

A cluster of articles in the most recent issue of Adverse Drug Reactions and Toxicological Reviews (2000, 19(3) 1-19) has been welcome to those of us who are agnostic about MMR as a possible determinant of autistic syndromes (AuS) with attendant complications. The lead paper of the cluster is a review of the state of science as reflected in part by pre-licensing studies of the safety of MMR vaccine. In it, Wakefield and Montgomery distilled and underscored viewpoints that have characterized their published and ongoing research. To publish anything about MMR/AuS that does not necessarily support official mainstream policies or governmental interpretation of evidence is a difficult decision for editors of any journal. I commend the editorial integrity of Adverse Drug Reactions and Toxicological Reviews in publishing such controversial articles. It was very wise to publish the evaluations of the four distinguished peer reviewers of the Wakefield-Montgomery article in the same issue. Disquiet among those four British opinion leaders about the evidence on safety of MMR is betrayed by their comments. Nevertheless, the absence of stridor in their written opinions, their objectivity and their willingness to consider all sides of the continuing debate is a refreshing stance and should be a role model. One arresting statement was made by Fletcher and merits prominence: The granting of a Product License was premature. [Emphasis mine.]

There were three important points made by Wakefield and Montgomery that have a bearing on pharmacoepidemiological considerations. Paraphrased, these are the key assertions:

• Studies that assure safety of MMR in reference to autism and AuS cannot be found.
• The length of follow up of immunized children in reported evaluations of the safety of MMR is inadequately short for complications of delayed onset, typically 3 to 6 weeks.
• Regulators and public health authorities have proscribed products other than trivalent MMR in some countries. [Many of us are concerned that parents who believe in immunization but mistrust trivalent MMR must take their children outside the UK for protection, notably against measles.].

I have not found scientifically sound safety studies. I shall present new data supporting the view that British evaluations on safety of MMR in respect to autism invoked inappropriately short lengths of follow up. The intrusion of the authorities in the legitimate freedom of choice of responsible parents by proscribing monovalent products is self evident.

Although the data about biological plausibility of an MMR /autism link has gradually become more persuasive, my experience and expertise is neither in the laboratory nor the clinic.

I will comment specifically on other papers that appeared in recent debate pursuing a better understanding of the epidemiological evedence and the related public health issues.

Fombonne.

Fombonne has written a letter seeking to debunk the very existence of an epidemic of autism. [Fombonne E. Is there an epidemic of autism? Pediatrics 2001; 107(2):411-12]. He is of the view that the prevalence of autism had clearly been underestimated in the past . He writes, further, that the debate on the hypothesis of a secular increase in rates of autism would benefit from a clear recognition of the methodological limitations of existing data . I agree with him about methods. Until now, most of the studies measuring the frequency of occurrence of autism in a defined population estimate prevalence. That includes Fombonne s earlier work. The existence of an epidemic of AuS can only be confirmed with age-specific incidence. Declaring a non-epidemic flies in the face of official statistics in governmental files and several published papers.

For instance, in California, there was a 15.6% increase of AuS in 2000 compared to 1999. For 1999 compared to 1998 there had been a 19.0 % rise. This is in a period when there was no change to the diagnostic criteria being used. [Kleffman S. Contra Costa Times Feb 9, 2001] From 1984 to 1998, the increase of persons accrued to the state s AuS system was 273%. [Reference: California Department of Developmental Services. Changes in the population of persons with autism and pervasive developmental disorders in California s Developmental Services System: 1987-1998. A Report to the Legislature; Dept. of Dev.: 1987-1998. A Report to the Legislature; Dept. of Dev. Services, Sacramento, March 1999.]

A report released this month by the California Department of Developmental Services documents a continuing remarkable rising rate of occurrence of full-blown level one DSM IV for autism in the state developmental services system. From January 4, 2001 to April 3, 2001, 700 new children were professionally diagnosed with DSM IV autism, which means a rate of eight new children per day seven days a week. The 700 new children represent the largest quarterly increase in the history of California s system. It also represents an increase of 107 more children than reported in the previous record highest quarter. Autism now accounts for 37 percent of all new cases coming into California s developmental services system. In the period of January 1994 to April 1994, 173 new children had been added to the system. The increase seven years later, i.e. 700 cases for the same quarter of the year, is a 404 percent increase. The figures do not include any component of the autism spectrum of disorders such as pervasive disintegrative disorders (PDD) and disorders not otherwise specified (NOS), Asperger s syndrome, Rett s syndrome or any other spectrum disorder. . [Reference: California Department of Developmental Services. Report to Department of Developmental Services, Sacramento, April 2001.] I cannot categorize the California phenomenon or the British phenomenon described by Kaye as anything other than a true epidemic

Such changes and other major increases in several countries over a period of less than two decades, even when reflected indirectly by prevalence rates, are strongly suggestive evidence. Fombonne s arguments do not explain away such steep rises in occurrence of AuS anywhere. He does not change my opinion mainly because his letter gives inadequate attention to the rate changes of subsets of AuS, such as regressive autism. A worldwide epidemic of autism is in progress. That demands serious scientifically admissible inquiries about possible determinants. Epidemiology has been challenged and epidemiology should respond with the highest attainable rigour.

Kaye, Melero-Montes and Jick.

The Boston Collaborative Drug Surveillance programme may have done the first correctly designed study to measure changes in population frequency of autism. [Kaye J, Melero-Montes M, Jick H. Mumps, measles and rubella vaccine and the incidence of autism recorded by general practitioners: a time trend analysis. BMJ 2001;322:460-463]. The key estimates are annual, age-adjusted incidences. From 1988 to 1993 the adjusted annual incidence (based on birth cohorts) rose from 0.3/10,000 to 2.1/10,000 per year in the General Research Practice Database of the U.K., a seven-fold increase. A lack of correlation of those incidence figures with the prevalence of coverage with MMR in the same cohorts of children is sobering to anyone already fully persuaded that MMR causes AuS.

Nevertheless, observing a key tenet of epidemiology, such an ecological correlation cannot and should not be invoked to test any hypothesis. Further in-depth assessments of the Kaye findings must focus on the diagnostic subcategories of the children affected, the validity of the clinical data and the short span between vaccination date and the date of diagnosis. The Kaye-Jick study is the best published descriptive epidemiological study to date demonstrating that an epidemic of autism exists.

Patja, Peltola and others.

A front page story in The Times of London alleging that Patja cleared MMR as a determinant of AuS was prototypical of misinterpretations of epidemiological studies and particularly of the Finnish Project to the public. There have been many reports in all the lay media suggesting similar conclusions. Manufacturers and the government have taken the same line. Curiously the government did not issue any formal statement in relation to the Finnish study, but they did issue a statement on 12 January, 2001 (from the Medicines Control Agency). Its conclusion is: MMR remains the safest way to protect our children.

A Finnish report evaluates serious adverse events after MMR vaccination during a prospective follow-up of 14 years (The Ped Inf Dis J , 2000, 19; 12:1127-34). 1.8 million individuals received nearly 3 million doses of MMR between 1982 and 1996. 173 potential serious adverse effects were evaluated. These adverse effects had been reported by a passive surveillance system Another original paper from Finland, by Kielinen and others, appearing shortly before (Eur Child & Adolesc Psych, 2000, 9: 162-67) is relevant. The Patja-Peltola Finnish project may have been big but it cannot rule in or rule out risks of autism nor any other major disease. It is important to stress that the Finnish project was passive surveillance in which underreporting is in general serious, typically 80% to 90%, seldom lower than 50%. One misses 50% to 80& of actual cases. Passive surveillance, pioneered by the British Yellow Card system and emulated world-wide, was designed to raise warning flags on safety. The system was never intended to be used the other way round, to confirm safety. Warnings from passive surveys are generally followed by hypothesis-testing and controlled epidemiological studies such as case control or cohort analytic designs. In the Finnish project I find no evidence that the study was set up to be sensitive to AuS, nor that the surveyors or the reporters of events looked for autism events at any time.

With the Kielinen findings it is easy to demonstrate pronounced underreporting in Patja-Peltola. The cumulative incidence (analogous to prevalence) of autistic syndromes in Northern Finland is 13.9 (CI95 12.0-15.7) per 10,000. Conservatively taking the lower boundary of the confidence interval (12.0 per 105), even if there had been no vaccination among the 1.8 million individuals and even if there were no association between MMR and autism, there should have been at least 216 cases of autism in the 1.8 million individuals surveyed. If one had focused on classic autism alone the number would be 79. If the study were capable of detecting autism, it is inconceivable that at least a handful of events of autism should not emerge in the Finnish survey. Curiously the word autism never appears in the abstract of the paper, or in the methods section and is only mentioned once, briefly, in the discussion of the paper. Conversely the mnemonic MMR never appears in the Kielinen article.

A large-scale study as was done in Finland is not automatically well designed or adequately reported because of its size.

Firstly, there were no controls, and this was accepted as a limitation by the authors. Yet there was no discussion about the shortcomings of such uncontrolled surveys.

Secondly, there is no indication in the report about the length of follow-up per individual. Fourteen years was the length of the project yet; the range and the median of follow-up per child are crucial in interpreting the findings.

Thirdly, there is no information given in the paper about the nature or content of the briefings to health care personnel before the study started, in relation to the types of reactions and the inclusion of autism as a reportable side effect. Such briefings affect subsequent reporting of adverse events especially since those reporting an adverse event usually make subjective decisions. It was noted that review of clinical reactions reported after MMR were made with reference to the medical notes and also to the known incubation times of the diseases measles, mumps and rubella, to decide if they could be causally related to the administration of the MMR vaccine. This may explain why AuS events were not reported here at all. In such a large study, we would expect many more diseases and symptoms to be observed within one month of MMR vaccination, by chance alone. Passive reporting, where the event reporter must decide if an event is linked with vaccination, is unlikely to detect rare, delayed or novel adverse events.

Fourthly, it is even less likely that delayed events would be linked with vaccination. Autistic syndromes (AuS) have an insidious onset with significant diagnostic delay. The early symptoms and signs with a closer temporal relationship to vaccination, may be poorly defined and similar to those exhibited by those who were excluded from follow-up in this paper. In AuS, mood and behavioral change may be the first sign, i.e. the zero time for the natural history of an AuS. Clear-cut deterioration or failure to develop usually may not manifest until many months later. Would vague early observations be reported in the Finnish project? Lastly, the ages at vaccination varied from 14 months to adulthood, with members of the Defense Forces, health care personnel and older children also receiving the vaccination. However, there is no breakdown of numbers vaccinated and side effects reported into categories of age. Given that autism among subgroups appears to be the result of damage occurring during a specific, limited window of vulnerability , it would have been helpful to have a precise figure for the numbers of children who actually could have been at risk. Any assertion that the Patja-Peltola paper clears MMR is unfounded. The Finnish Project cannot rule out and cannot rule in MMR as a risk factor for AuS.

Taylor and others.

Commissioned by the Medicines Control Agency in the U.K., this study has probably been the most cited article to substantiate safety of MMR until now. [Taylor B, Miller E, Farrington CP, Petropoulos MC, Favot-Mayou I, Li J, & Waight PA, 1999, MMR vaccine and autism: no epidemiological evidence for a causal association. Lancet; 353: 2026-2029]. Its publication is an earlier one, but a comment on the current situation would not be complete without a critical note about the North Thames project. Taylor studied medical records of 498 prevalent cases of autistic syndrome disorders (AuS) in eight health districts. The study reported the number of patients by year of birth (1982-1996) seeking to identify a temporal trend. The study and its report are seriously, if not fatally flawed, for these reasons, among others:

• Complete ascertainment of all cases of AuS in the eight districts uncertain.
• The authors make statements about an increasing incidence of AuS over time from a prevalent group.
• Imprecise categorization of zero time for the cases.
• Inadequate classification of the various diagnoses within the AuS spectrum, notably unreliable categorization of regressive manifestations.
• Failure to correct for catchup components of the immunization campaign.
• An incorrect analytic method was used. The case series method used by Taylor (similar to the case crossover method) applies primarily to acute events and transient exposures for diseases with a well defined natural history. (Regressive autism, for instance, has anything but such a natural history.) One does not expect AuS to develop acutely or immediately after an exposure to MMR; the onset of autism is typically insidious, highly variable and may occur even years after immunization.
• Failure to discriminate between types of MMR vaccine (Pluserix, Immravax and MMR-II) may have masked a step-up in cases in 1988 due to diagnostic overshadowing. Pluserix and Immravax were both withdrawn due to problems with the Urabe mumps component leading to higher rate of aseptic meningitis.

Paradoxically, the Taylor paper, which is incorrectly interpreted as demonstrating safety, provides much better evidence in the opposite direction consistent with MMR associated with some AuS categories. Moreover, an uncontrolled study is uninterpretable as the basis to demonstrate a link between MMR and AuS or to dismiss it aside unless the findings were dramatic and very clear.

To summarize: views on an epidemic of AuS, on prevalence, which appears to be rising, on incidence and what it all means: Fombonne and Kaye come to differing conclusions. Kaye was the only one to publish age-adjusted incidence. No one has tracked subsets of the spectrum of autistic disorders allowing the lumped data to mask potentially important profiles in relevant subsets of autistic children. Although all we can say with certainty is that autism (combined groups) appear to have become more common since the introduction of MMR, any reasoning concluding that the increases in frequency are probably real is more persuasive to me at this point than arguments that seek to explain away reported steep rises in cumulative incidence. Four-fold, seven-fold and ten-fold rises in prevalence are remarkable.

A war of words should not continue. It would be a major disservice to society and especially to children and families affected by AuS if the current struggles of scientific and regulatory warlords became the status quo. There will never be any winners if such a war were limited to clashes of opinions grounded on questionable assumptions, on statistically underpowered projects, and on incorrectly designed studies with accompanying misinterpretation of findings to the public. I have to be persuaded that anyone is acting in bad faith. We are overdue for an alliance that will confront unanswered questions about the role of MMR and other risk factors in the aetiology of AuS. Answers will emerge quickly. I propose the following research priorities for the next sixty months:

• Investigation in the laboratory and the clinic such as the work conducted by Wakefield in London, O Leary in Dublin and others should be adequately supported. Replication should be done in the USA and Canada. Scientists exploring hypotheses unpopular to the establishment should not be forced to spend a great deal of their time as fundraisers. Laboratory research and clinical investigation should have priority in the MMR-AuS controversy; the silver bullets that will resolve unknowns will most likely emerge from there.
• Three to five carefully designed incidence studies of AuS should be conducted in different countries or regions. The geographic areas must have had different entry dates of MMR to their markets. Precise ascertainment of incidence trends in areas where prevalence studies have already been done is desirable
• Two active surveillance projects patterned after Phase IV simplified cohort studies. The studies should be compatible in definitions and methods and, combined, should have high statistical power to avoid Type II errors when seeking to establish safety. If at all possible they should be controlled. They should be done urgently. Cost should not offer impediments considering the hundreds of millions of MMR doses sold worldwide If one based sample sizes on a model of two cohorts under comparison, given the baseline population rate of AuS of 5 per 10,000 in the appropriate age brackets, specifying a relative risk of 1.5 that should not be missed, an alpha of 0.05 and power of 90% would require 450,000 persons per comparison group. For the combined safety studies, the same order of magnitude would be required, not less than 500,000 new subjects. Again, given the sales volume of MMR worldwide, the profits and the relative ease in categorizing both exposure and outcome, the numbers are not logistically or financially insurmountable. If more than 500,000,000 doses have been sold to date, active surveillance would involve less than 0.001% of exposed children. The cost would be approximately that of the lifetime care of 200 autistic children eligible for state support.
• Two major case control studies, one based on a computerized database and one with concurrent accrual of new cases.

The best way to remain in the dark about safety is not to look for it. One way not to look is to underpower the studies but pretend to be looking is to study very small samples. Some authors write that the rarity of the outcome makes it difficult to derive statistically definitive findings. Given the rarity of AuS, hundreds of thousands of children must be followed. The reluctance of regulators to require rigorous safety studies and the unwillingness of manufacturers to conduct them is noteworthy. Most importantly, whether one conducts active surveillance safety projects or case control studies, with a controlled design or without, the length of follow up after immunization is critically important.

A follow up of three to six weeks from MMR immunization to look for the occurrence of AuS negates any opportunity to discover an association if present. My colleagues and I at McGill University, in collaboration with British clinical experts, are nearing conclusion of a study of the natural history of 493 British patients with AuS who were also exposed to MMR. The source of data for each child is the complete medical record, that is, the depository of all GP notes, entries from community resources, and specialists reports. (The records, each spanning several years, are generally many inches thick.) Because adequate length of follow up is so important in any active safety study, preliminary findings of our study are given here. [Spitzer WO, et al. The natural history of autistic syndromes in British children exposed to MMR. In preparation.] Among 372 eligible children (all criteria met, abstracting done, data clean, double-coding and double-entry reconciled) the median delay from first exposure to MMR to the earliest manifestations of AuS recorded by a professional clinician (zero time) was 1.1 yearsb and the range was from 33 days to 7.2 years. The median delay between MMR exposure and formal diagnosis was 2.6 years and the range 6 months to 11.8 years.

Only definitive safety studies of active surveillance with a pre-determined sample size and with an adequate length of follow-up can give evidence about the safety of a product. Thus, whilev understanding and supporting Professor Fletcher s opinion that granting a license may have been premature for MMR, I would have preferred a conditional license until valid safety studies were completed in the light of these concerns. Proper drug safety studies cannot be implemented at all with uncommon or unexpected adverse events until a new product is marketed.

Autism is such a serious outcome that at least one controlled epidemiological study that explores aetiology should be done. In the 18 months I have been studying the MMR/AuS controversy I have also met many of the affected children and their families. As a former clinician and as a public health doctor I have been gradually convinced that the aggregate manifestations of autism are an outcome at least as serious as death. Despite the stoic, indeed heroic conviction of most parents, despite their invincible optimism, autistic children act as dead spirits in ostensibly healthy bodies (except for gastrointestinal complications) resulting in a very long terminal condition. The families cannot mourn and come to closure about the tragedy. Families of a child who actually died can. One wonders what the position of the authorities might be if the possible adverse event of a vaccine were postulated to be premature death? A controlled study is clearly justified.

As this comment is released to the written record, a group of epidemiologists and biostatisticians is in the early phases of protocol formulation for The Intercontinental Case Control Study of MMR and Autistic Syndromes. At the moment we anticipate the need of at least 3,000 new cases of autism and 6,000 unaffected controls, in 13 centers from 9 candidate countries across three continents. The full detailed protocol will be published in the peer-reviewed press later in 2001 to allow international constructive critique before going to the field in any country. Final results should be published by the end of 2004.

I assert that I am a strong supporter of vaccination in general; it is the most important pillar of effective primary prevention in the history of public health.

We would not wish to replicate the Japanese experience with MMR. In Japan, loss of public confidence in vaccination per se against mumps, measles and rubella after the sudden withdrawal of their MMR vaccine which had caused aseptic meningitis led to low uptake of vaccination against measles and significant mortality in outbreaks of wild strain infection.

In prudence, I also believe that each new vaccine should be scrutinized for safety in the post-marketing period, one by one. I continue agnostic, but today the general burden of scientific evidence makes it difficult to set aside an MMR/autism link. It is for those completely convinced of MMR s safety, especially those profiting from the hundreds of millions of doses of MMR sold, to finance the research that could set aside legitimate concerns. Sponsorship must be undertaken by an entity without vested interests or publicized prior opinions. Many scientists who are concerned about the unanswered questions are not crusaders, including myself. I have no family members with autism. I am a consultant to affected families in the UK, when asked, pro bono. I am very worried by autism as an adverse event. I would be delighted to be proven wrong and to have my concerns minimized or dispelled by the highest standard of laboratory research, by well designed clinical studies, by valid safety projects and by rigorous aetiologically oriented epidemiology. Until then, it is premature to offer any assurance about the safety of the measles-mumps-rubella vaccine.