MMR Vaccination: Individual Risk/Benefit Analysis



The central question examined here is whether certain vaccines are a good or bad idea for one of us to give our kids. “Us” in this case means individuals in Western or developed nations, especially the United States. Specifically, I have chosen to examine the measles-mumps-rubella (MMR) triple vaccine, which is extremely widespread and has been criticized on the basis of safety.

This is a risk/benefit analysis. A vaccine will have certain negatives associated with it (cost, pain, inconvenience, and side effects ranging from the negligible to the grave). It will have certain positives associated with it (primarily a reduction in the likelihood of contracting an illness). Therefore the question is whether the risks outweigh the benefits, or vice versa.

The question is not about vaccines other than the MMR vaccine, since different vaccines will have differing risk/benefit analyses.

The question is not whether vaccines should be mandated by the government, or any other questions of general policy; nor about whether factors other than vaccines (such as sanitation) influence disease rates; nor about the motivations or political/economic forces at work behind this issue. Any of these may be interesting topics, but do not bear on the original question of whether we should personally vaccinate.

Therefore, what we want to see is evidence on the benefits of the MMR vaccine, and evidence on its harm. The former includes: probability of catching measles, mumps, or rubella in an unvaccinated state; the symptoms and risks of these diseases; and the probability of catching them in a vaccinated state. The latter includes: side-effects of receiving the vaccine; probability of experiencing these side-effects; and how much greater this probability is than the probability of experiencing them in an unvaccinated state.

Standards of Inclusion

The evidence I have chosen to include focuses on primary sources. Specifically, I mean empirically gathered data published in a peer-reviewed medium, and presenting its methods and procedures for perusal. The latter (direct access to the methods) may seem to preclude the necessity of the former (prior peer review), however both seemed necessary for several reasons. For one, only peer review could catch actual factual (i.e. data-based) falsifications or errors. Additionally, peer review would help filter for aspects of study quality (especially the statistical) that I lacked the expertise to notice myself.

Certain primary sources were elided for reasons including the following: weak statistical strength (e.g. only a handful of subjects were studied); irrelevance to the specific issue; and questions raised by other critics about specific issues of methodology or rigor that I lacked the expertise to judge the merit of. (When these last were especially compelling, both the original work and the criticisms have been included, to allow more knowledgable readers to make their own analysis; less compelling studies were simply discarded.) No sources were elided on the basis of potential bias of the researchers, on the grounds that such potential biases were (1) hopefully already cleared by the editorial and peer-review process; (2) unquantifiable; (3) irrelevant as long as data and analysis were valid and untampered with; and (4) unavoidably present, presumably benignly, in an overwhelming number of the bench scientists in this field.

Studies for which the full text were unavailable were used only when the abstracts presented no especially contentious information. For the vast majority of these citations, the full texts and full data were available.

Secondary sources such as commentaries, opinion pieces, and the like were largely disregarded. Those which presented arguments with accompanying cited sources were mined for their primary sources. Those which presented nothing external were perused to suggest paths of research, or duly ignored. Secondary sources such as reviews and meta-analyses were used sparingly, since their conclusions were somewhat opaque.

The only information presented without sources (indicating that sources were used, but I did not bother citing them), or with a source that is not primary and does not itself cite its sources (such as informational websites) are minor facts that are not to my knowledge in contention (such as common symptoms of the mumps). Obviously, my idea of these may differ from the reader’s, so if you contest any of the uncited information here, please say so and I will present the sources. All efforts were made to err on the side of over-citing evidence, particularly because some of the issues here hinge upon otherwise insignificant facts (such as uncommon symptoms of the mumps).

An effort has been made to define medical terms which may be unfamiliar.

Methods and Bias

The majority of these data were culled from PubMed-indexed sources, including third-party databases which it can access. ScienceDirect indices were used less heavily. Google and Google Scholar searches comprised the final major research tool.

Because this amounts to an amateur meta-analysis, some remarks should be made on my personal outlook and bias. While the facts I present are accompanied by their sources, it would potentially be possible for me to influence the results of this analysis by omitting key studies or unfairly wording the presentation of those I do include.

I entered this project with a minor bias in favor of vaccination. Prior to my research, I had no knowledge whatsoever in this field beyond common knowledge, so my bias was based primarily on the general principle of “everything we know may be a lie, but it’s probably not.”

However, I made every effort to examine each piece of evidence with an eye only towards its structural merits. In some cases, where merit seemed ambiguous, I erred on the side of the anti-vaccine position (generally this meant disqualifying a pro-vaccine study).

I do not have children, do not intend to in the near future, and received my MMR vaccine long ago.

I have no commercial conflicts of interest.

I am a third-year philosophy undergraduate with no specific training in epidemiology, medicine, or statistics. I was therefore unable to treat some of this material with an advanced critical eye, and my analysis should be interpreted as an informed layman’s.


This document will read like this: first, I will explain the frequency, symptoms, and status of the measles, mumps, and rubella; second, I will explain the known side effects of each vaccine, and of the combined MMR vaccine, including the contentious and unclear effects, such as autism; finally, I will offer some additional remarks pertinent to a final analysis, and sum up the material presented and the situation as it stands.

I will not offer a conclusion on the question of MMR vaccine use or disuse. That decision is left to the reader.

Diseases and Symptomology

Common Features

Natural infection always confers a better and longer-lasting immunity than vaccination. This is true, though in varying degrees, for all of the following diseases.

All of these diseases are highly contagious (airborne). They are all relatively minor in most cases, but with at least one rare but very serious complication.

All were endemic in the US before vaccination.


The measles (or rubeola) is a viral disease characterized by a full-body rash and several flu-like symptoms including coughing, runny nose, and red eyes. It is transmitted via airborne aerosol and considered extremely contagious. It was almost universal in the past, and in many regions of the world is still widespread, but its incidence in the US dropped precipitously in the early 1960s when its vaccine was introduced (see below for sources).

The measles itself is not serious in most cases, except for immune-compromised patients and some other exceptions. The common symptoms are minor. However, complications are common, including diarrhea, and several rare sequelae [diseases caused by it] exist; pneumonia is the most prominent (5%-10% of cases), otitis media (middle ear infection) occurs in 5%-15% of cases, and encephalitis occurs in about 1 per 1,000 cases (1 for all numbers—uncited).

The most serious sequela by far is subacute sclerosing panencephalitis, a rare form of encephalitis (inflammation of the brain) that can only be caused by measles infection. SSPE is degenerative with a high level of mortality, including neurological symptoms like seizure and coma, and has no cure. This one is no joke—you don’t want it. Incidence of SSPE is about 22 per 100,000 measles attacks. (2)

Though complications and hospitalization for the measles are common, deaths are not; the mortality today is about .01-.02 per 100,000 in healthy patients in developed countries (3—uncited but matches my rough calculations.) Mortality in less-developed countries is higher.

The measles is, on the grand scale of human illness, a minor disease. The worst complications are rare and the common ones are bearable. However, it is so radically infectious that almost everybody can expect to contract it in an unimmunized population; therefore by sheer force of mass, with no preventative measures the number of major complications (such as SSPE) becomes large.

The likelihood of contracting measles in an unimmunized population is very high; however, the United States (and most Western countries) are immunized at a level of about 92% today (4). Here we should discuss herd immunity.

Herd immunity is the concept that unimmunized individuals in a population (whether opting out of vaccines, or unable to receive them due to being newborne or immunosuppressed, or unvaccinated due to recent immigration) will receive implicit immunity, due to the fact that there is no chance for the virus to reach them. Person A is infected and interacts with Person B, but Person B is immunized and does not contract the virus. Therefore when Person B interacts with Person C, who received no vaccination, Person C remains uninfected—not only sparing him the illness, but preventing its further spread to Persons D, E, etc. Herd immunity therefore prevents outbreaks of disease from spreading and becoming endemic.

Herd immunity is ranked around a threshold, below which the level of immunity is insufficient to prevent the spread of disease and above which it is sufficient. The herd immunity threshold for measles is judged to be 83-94% (5). With the 92% level of immunity in the US today, we are effectively over this threshold and the measles has been declared no longer endemic (6).

Therefore, there are two consequences of avoiding measles vaccination. On the personal level (which this paper focuses on), your chances of avoiding the measles are very good. The current disease rate (as of 2004) is .01 per 100,000, or 37 individuals throughout the US (7). The disease is not gone—it still crops up occasionally, due to incomplete vaccination, vaccination failure, or immigration of unvaccinated disease-carrying individuals—so nothing is certain, and with any plans to travel to a less-immunized population the odds would change dramatically. However, it would be a reasonable choice, especially if you accept your chances with the symptoms if you do contract the illness.

The second consequence is long-term and societal, but significant enough that it bears mentioning. A single individual avoiding vaccination would have no impact, but the population is composed of individuals, and if a growing number of them chose not to vaccinate, then the immunity level would drop. With the loss of only 10% immunity or so, the herd immunity effect would be heavily damaged. The result of this is that the previous calculation—that one’s chances of catching the measles would be minor—would change. For comparison’s sake, in 1960 (prior to the beginning of mass vaccination in 1962), the US saw 441,703 measles cases a year, or about 245 per 100,000 (8) and essentially certain infection at some point in one’s life (9). The immunity level would likely not drop this low, so the odds for a non-vaccinating individual would fall somewhere between the 1960 level and the current one of .01/100,000. This would apply both to voluntary non-vaccinators and involuntary ones.

[An aside: there is some misinformation circulating about the effectiveness of the measles vaccine. This graph, for example (10) is very misleading in terms of its commentary. The data presented are on measles DEATHS, which as discussed are quite low relative to the population; however the spread of the disease itself was relatively unabated prior to mass vaccination. To compare, there were 441,703 cases in 1960 (see above), and 22,231 in 1968 (11). Death rate in 1960 is accurately listed as .2 per 100,000 (12).]

Probability in the US of contracting measles without vaccination is therefore low but variable.

The failure rate [vaccinated individuals who later became infected] of the measles vaccine has varied with different strains and different vaccination schedules. With the current system, it is considered very low, possibly under .2% (13) and generally few outbreaks are attributed to vaccine failure (14).


The mumps, or epidemic parotitis, is a viral disease characterized by swelling of the parotid salivary gland. Its other features resemble flu symptoms: fever in 97%, vomiting in 94%, headache in 89%. Lethargy, abdominal pain, convulsion, difficulty walking, and changes in attitude all occur in under 25% (15), in addition to orchitis [painful testicular swelling] in about 25% of cases and mastitis [swelling of the breasts] at similar rates. Mumps orchitis exhibits testicular atrophy [shrinking] in about 1/3 of cases but sterility rarely. Oophoritis [ovarian swelling] occurs in 5%. (These symptoms only manifest in post-pubescent patients.) (16). It is widespread in unvaccinated populations.

The mumps is a minor illness, however, a number of complications are noted. Meningoencephalitis [resembling both meningitis and encephalitis] is one of the most prominent; aseptic meningitis, generally benign, occurs in up to 10% of cases, and encephalitis, generally non-fatal but with some serious sequelae (paralysis, seizures, etc.) occurs in .02%-.3%. Deafness (varying degrees) occurs in about 4%-15%. Pancreatitis is seen in 4%. Death occurs in only .2 per 100,000 cases or so. Various renal [kidney] and EKG [heart] abnormalities can also manifest. (ibid.).

(Encephalitis occurs around .0074% and aseptic meningitis around .0109% in the normal population without mumps involvement. 17)

Perhaps the most striking complication is spontaneous abortion [miscarriage] in over 25% of mumps cases contracted by pregnant women within their first trimester (ibid.).

Prior to mass vaccination in 1967 (18), mumps was endemic to the US and outbreaks were a regular feature. Estimates on infection rates vary by subgroup from 100 to 6,000 per 100,000 (19). Currently (2004), infection rates are down to 30 per 100,000 (20). 2006 saw 5,783 confirmed + reported cases in 45 states (21). Mumps is still common in less-developed countries. The odds calculation for remaining unvaccinated is therefore the same—including the herd immunity factor—as previously discussed for measles, only with different numbers. Herd immunity threshold for mumps is 75%-86% (22) and current immunity is 92%, identical to measles due to their combined vaccine.

One additional consideration is that, like rubella, widespread but partial vaccination of children for mumps tends to shift the cases that do arise from younger to older patients. Adult patients tend to see more serious complications, especially the spontaneous abortions mentioned above. It is therefore important, if vaccination is performed at all, for population coverage to be near-total wherever possible. (23)

Failure rate for mumps vaccination has varied significantly over time, with both scheduling (second doses are found to be critical for good coverage) and strain (the Rubini strain was found to be essentially worthless, conferring no better coverage than no vaccination—24). The currently-used strain is Jeryl-Lynn, which has shown 86% success (14% failure) with single-dose scheduling (ibid.), though also significantly less (64%, 83%, 82%, 94%) in other studies, highlighting the effect of waning immunity and importance of the second dose (25, 26, 27, 28). There is also evidence suggesting that even natural infection may not confer lifelong immunity (29). Data from one study suggest that patients with the double vaccination (though under a different schedule) saw 90% success; four-fifths of vaccination failures involved patients who had not received a second dose (the strain was not Jeryl-Lynn, but derived from it). Another study saw 91% success with double doses (30). The success rate for the Jeryll-Lynn therefore appears to be at least 90% IF proper scheduling is utilized, though the effects of waning immunity are not entirely clear.


Rubella, or German measles (no relation to rubeola, or measles), is a viral disease characterized by a red rash and general flu symptoms. 30%-70% of post-pubescent women will experience acute [short-term] arthritis, up to a month or so; a few experience long-term arthritic symptoms (31, 32—uncited). It is transmitted via airborne aerosol and is moderately contagious. It was widespread in the US prior to mass vaccination and remains common in many other countries.

Rubella is a very minor illness, and often completely asymptomatic [no symptoms noticeable]. A few rare complications are seen, including encephalitis in 1/6,000 cases (with high mortality) and thrombocytopenia [low platelet count] with hemmorhage in 1/3,000 (33). No effort would have been made to immunize against it were it not for one feature—pregnant women who contract rubella during the first portion of their pregnancy (first trimester especially, but up to 20 weeks and ongoing) run an extremely high risk of either spontaneous abortion, delivering a stillborn baby, or delivering a baby with Congenital Rubella Syndrome. CRS babies exhibit defects such as deafness, retardation and various neurological problems, small head size, and poor prognosis of survival. Odds of CRS vary widely with the estimate, the symptoms, and the point in gestation rubella is contracted (earlier is worse), but range from 50%-90% (34). CRS is really bad news; fortunately it’s become almost nonexistent in the US after mass rubella vaccination.

Odds calculations for non-vaccination parallels those for measles and rubella. The rubella vaccine was introduced in 1969 (35). In 1970, 28 cases per 100,000 were seen (56,500 total); today (2004), 0.00 cases per 100,000 (10 total) cases are seen (36). Deaths are negligible; CRS is considered a separate disorder, but with the loss of rubella has been similarly eliminated (37). Rubella vaccination levels in the US are once again 92%; herd immunity threshold is 83%-85% (38).

Like mumps, there is an element of age-shifting that occurs with partial childhood vaccination. Rubella tends to occur more at high school and college years when children are incompletely vaccinated, which is a much higher-risk period for pregnancy than childhood, and consequently for CRS (39). Again, if mass vaccination is performed, it is therefore important for it to be very complete.

Women who foresee becoming pregnant are obviously the most important recipients of rubella vaccines, but it has been seen that reductions in CRS are not noted in targeted vaccination of the female population; widespread immunization of the entire population is needed (40). This seems to be because herd immunity with rubella is less effective at protecting the unimmunized population than it is with other infectious diseases (41, 42).

As with measles and mumps, vaccine failure rates have varied significantly with strain and schedule (43). The current strain is RA 27/3, and has an extremely low failure rate (44, 45).

Adverse Effects of Vaccination

Several side-effects are known to be caused by the individual measles, mumps, and rubella vaccines; others lead from the combined MMR vaccine. All four categories will be presented here.

A brief description should be made of “association,” “causality,” and “biological plausibility,” terms which will see frequently. Briefly, to say that two variables (such as receiving a vaccine and experiencing a symptom) are associated is to say that they tend to occur together; this is usually seen by temporal proximity [close together in time; one usually occurs soon after the other]. Causality is also association, but with the extra element that one variable actually caused the other, which is more difficult to observe in most cases. (The sun rising is associated with people making toast, but causally linked with the sky getting brighter.) Association suggests but does not demand causality. Both association and causality can range from very weak (the relationship seems to exist, but just barely more than pure chance would have created) to very strong (the relationship almost always exists and is very clear). Biological plausibility essentially means that a causal relationship would make sense—it fits with what we know of the human body. If an association is biologically plausible, it means we can articulate how the mechanism of A causing B might work. Biological plausibility makes causality more likely, but lack of it does not make it impossible; if strong evidence of a relationship exists, it falls upon our biological models to change, not the data. Biological plausibility does not denote an actual relationship; it merely makes one more, well, plausible.

There has been an ongoing movement questioning the safety of thimerosal, an ethylmercury compound used as a preservative in several vaccines, and accused by many of contributing to either autism or certain auto-immune disorders. The MMR vaccine does not contain thimerosal and never has, so this issue will not be treated with here.

VAERS is the Vaccine Adverse Event Reporting System, a passive network with which doctors and patients can report adverse reactions they associate with their vaccinations. It was designed to catch rare reactions which can be observed in populations of several million, but not in the pre-release test groups of several thousand. VAERS should be considered a powerful, but blunt tool. Its strength comes from its very large numbers (essentially every patient in the US is a member). Its weaknesses are due to lack of scientific controls on the data, and it probably exhibits both under-reporting (patients/doctors not associating complications with the vaccine) and over-reporting (patients/doctors associating unrelated syndromes with the vaccine).


There have been several cases (about 13) of Guillain-Barre syndrome reported temporally associated with the measles vaccine. Causality is not demonstrable, but the association is significant and biologically plausible. It would be extremely rare. (46)

There is a causal link in immunocompromised individuals (such as HIV patients) between the measles vaccine and death. It is extremely rare. (47) Vaccines are contraindicated [not recommended] for immunocompromised individuals.


Aseptic meningitis (similar to that seen in natural mumps infection) is strongly associated and probably causally linked to administration of the Urabe strain mumps vaccine in around 9.1/100,000 cases. The Jeryl-Lynn strain used currently in the US does not seem to suffer from this (48). (Urabe is cheaper and still used in some less-developed countries, however.)


There is a probable causal relationship between the RA 27/3 rubella vaccine and acute arthritis (similar to that seen in natural mumps infection). It occurs in around 13%-15% of post-pubescent women who receive the vaccine (49). (This seems to be around 1/3 to 1/5 of the occurrence in natural infection, but the numbers are too vague for definite comparison.)

There is a possible causal relationship between the vaccine and chronic [long-term] arthritis, but the data are not available (ibid.) Incidence cannot be compared to incidence in natural infections, for which the data are also unavailable, but is very rare (50).


A significant amount of attention has been focused on studying the safety of the MMR vaccine, for the obvious reason that it is administered, often several times, to over 95% of the population in several countries. Renewed attention was brought to bear in the late 1990s, when additional questions were raised about the role of MMR in autism (this is treated with separately at the end of this section).

Most of the side-effects known to occur with MMR vaccination are either minor and general (swelling around the injection site) or directly analogous with the symptoms of the viruses they contain. All three of the vaccines contain live, but attenuated [weakened] viruses; it is therefore possible for similar, but weakened symptoms to exhibit. Many of the adverse reactions discussed here will therefore be familiar.

Fever and rashes are not uncommon in MMR administration (52).

Varying degrees of hearing loss are seen in about 1 per 6-8 million vaccinations (53).

Febrile seizures [heat-related pediatric seizures] are seen in 1 per 3,000 vaccinations. Febrile seizures are common and usually harmless (54).

1 in 40,000 patients experience thrombocytopenia (immune thrombocytopenia purpura), according to a VAERS survey; boys were at a higher risk (55). This is 13 times less than in natural rubella infection. It is usually benign but can vary in severity, extremely rarely causing death (56).

Anaphylaxis [systemic allergic reaction] occurs extremely rarely following vaccination, due to allergic response to one of several components in the vaccine. (57) Rates are given at 1 per million or so. Vaccines are not recommended for patients with allergic responses to any ingredients, and anaphylaxis will generally manifest in the presence of the health professional administering the shot.

Several other spurious and assorted ailments have been associated with MMR, but studies have generally disproven them. Some examples are here (58)

There is ongoing question of whether MMR is causally linked in rare cases with encephalitis and/or aseptic meningitis. It is biologically plausible and temporally associated in several cases. One study showed the incidence of those complictions to be no more frequent than in the normal population (59). Other studies suggest rates of 1.1 per million or up to 3 per million (60), or up to 11 per million, mainly when the Urabe strain is included (61).

One other study, using the VAERS system (see above for the limitations), noted a variety of neurological symptoms following MMR vaccination as well as monovalent [one vaccine only] measles (mumps and rubella showed no such symptoms). Symptoms related to encephalitis and included death, mental regression (see the Autism section below), chronic seizures, motor and sensory deficits, and movement deficiencies. There was no direct link of the symptoms to the vaccination, but there was biological plausibility of causation (as these symptoms are similar to encephalitic sequelae seen with natural measles infection), and there was a temporal association: the onset of symptoms “clustered” in a statistically significant way around a period 8-9 days after vaccination, rather than being evenly distributed across the 15-day monitoring period as would be expected in a random relationship (62). Strong causation cannot be demonstrated due to the lack of data on encephalitis incidence rates in the non-vaccinated population—there aren’t enough people who haven’t been vaccinated—but the work is strong due to the very large sample size. Rough calculation (48 incidents out of a pool of 75,000,000) would place the risk level at .000064%, or .000000064 per 100,000. Underreporting to VAERS was possible, though the potentially lucrative nature of claims (through the compensation program) would seem to limit this. Another confounder is the different vaccine strains that were used through this sample period (1970-1993).

An association between MMR (specifically the measles component) and encephalitis seems reasonable, and the Urabe strain is probably causally linked. Whether there is a very rare causal link with the other strains or not seems somewhat academic considering the frequency in question.

[An aside about scheduling: the current US vaccination schedule—which places the second dose around 4-6 years of age—seems to produce fewer clinical events than placing it at 10-12 years, which some alternate schedules suggest. This is not rock-solid evidence, however. 63]

Women thinking of becoming pregnant are generally tested for rubella immunity, and (re-)vaccinated if necessary. It is recommended as a precautionary measure that these women wait a month before becoming pregnant, due to some suggestions that the vaccine could endanger the fetus. This is not substantiated and it has been shown that vaccination during pregnancy produced no CRS (64 , 65), but it is still generally recommended on the precautionary principle [there is no harm in the practice, and potential benefit].

MMR and Autism

The elephant in the room with regard to MMR’s adverse effects is autism.

The story began in 1998, when Andrew Wakefield published a clinical study in The Lancet (66—this is on a private site, but appears to be identical to the published version and has the full text available). The study was focused on twelve children with forms of developmental regression, such as autism, and also certain gastrointestinal problems. Most of the 12 children had recently received MMR or similar vaccines which their parents or physician associated with the onset of their symptoms. The primary aim of the study was to establish a link between the neurological and gastrointestinal conditions, a state which Wakefield later called “autistic enterocolitis”; his hypothesis was that the GI state led to the mental defects. The vaccine issue added another element, essentially suggesting that environmental factors could have precipitated the disorder.

The autistic enterocolitis issue has undergone research on other fronts, but it is not of interest to us, since this report is not about novel forms of autism. We are interested in MMR’s link to autism, and the extent of it presented in the Wakefield study is this: 12 children experienced varying degrees of mental degeneration temporally associated with MMR or measles vaccines. The time period ranged from 24 hours to 2 months. The authors note two other case studies that observed temporal associations between unspecified vaccinations and autistic symptoms in similar numbers of patients (these studies were not available) and suggest several biological models. They note that they have not proven an association between MMR and the syndrome described and advise further research.

The fallout from this report was significant, leading to many families in the UK avoiding vaccination and a drop in the MMR immunity level. Ten of the study’s authors, noting that the work had not demonstrated that MMR caused autism and regretting the confusion, formally retracted the interpretation in the original study (67). Wakefield did not sign this retraction, but would continue to make it clear that the study had not demonstrated the aforementioned causal relationship (68).

Numerous other studies followed, mainly epidemiological studies interested in determining whether the MMR vaccine could or could not be causally linked to the onset of autism.

This type of epidemiological study functions by drawing up statistics on large groups of individuals, either in an isolated homogenous [all the same] or general randomized [different but randomly so, creating no skew in the average] population. With statistics on—for instance—which patients were vaccinated and developed autism, compared to how many patients were not vaccinated and developed autism, an association and possibly causation can be demonstrated or refuted.

The most useful studies would be those directly comparing vaccinated versus unvaccinated populations in large numbers. Unfortunately, this is difficult to accomplish given to the high rates of vaccination in developed countries; it’s difficult to find unvaccinated examples, and given that—when present at all—autism seems to develop only very rarely, sample sizes need to be very large to be useful. (See the VAERS study mentioned above.)

There are two studies of this type I was able to find. The first is a very large Danish study which directly compares autism (and autistic-spectrum disorders) incidence rates in MMR vaccinated and unvaccinated babies born over a 7-year period. It uses reliable data and a sample pool of half a million people; it found no difference in autism occurrence between the two groups (69). Rough odds calculations are: .00167 chance of any autistic disorder, .000717 chance of autism, .000958 chance of any autistic spectrum disorder.

Several statistical and methodological criticisms have been made of this study, such as the fact that choosing a period and studying everyone born within it will include some babies too young to receive their vaccine before the end, and therefore possibly being vaccinated and developing symptoms after the end of the study, being incorrectly categorized as having no adverse response. Most of these criticisms seem to be valid, but not influencing the overall drive of the results (which is that no relation exists, or if it does it is too rare to show up at these numbers). However, several statistical criticisms have been made of the study, suggesting that different workings of the data actually produce the opposite result (a significant relationship between vaccination and autism). One such piece, of which Wakefield is a contributor, is here (70), and the statistical details in question are completely over my head; I cannot judge on these points. I encourage any readers with a better grounding in this field to judge for themselves; however, I also strongly encourage them to examine only the methods described in the critique piece, ignoring the conclusions and if possible calculating the results on their own. The reason is because the publication of the latter is the Journal of American Physicians and Surgeons, which has an overt political stance. I distrust any publication with a clear bias and therefore distrust its peer-review process. However, judge the material on its own merit.

The second study is an examination of cases from the National Childhood Encephalopathy Study, which is a data-mine somewhat similar to VAERS. The sample size is not as large as the Denmark study, and in several ways it is less powerful, but it still has some merit. The study works, in a way, “backwards”—it takes a sample of patients who have shown neurological problems (770 total), matches them against a control group (1500 healthy individuals), and determined how many from each group had been vaccinated within a set time period before their symptoms manifested, or in the control group, merely an equivalent reference date (7-14 days was used as the window). No significant difference was found between the two groups (71).

Several criticisms can be made against this piece. For one, it is relatively “small”—only 6 cases and 10 controls were present in each vaccinated group. Of course, this is because autism is rare no matter how you stack it, but it still decreases the strength of the study; a larger pool, yielding larger case/control groups, would have been better, though unfortunately not available. The other criticism is that the study only allowed acute symptoms to manifest within a two-week period after vaccination; though many of Wakefield’s patients fell in this window, it is quite easy to argue that autism might take longer to develop in some cases.

One more study uses a similar comparison method, with better numbers (64 unvaccinated, 105 vaccinated) and one major advantage—it attempts to address the criticisms against short monitoring windows in this and other studies by monitoring its patients for a lengthy period (up to 15 years). It then compares the age of autism diagnosis in the unvaccinated group with the age of diagnosis in the vaccinated groups, and compares age of diagnosis in the vaccinated groups against the age of vaccination. All groups show similar ages of diagnosis, and no groups show a temporal clustering between vaccination and diagnosis age; no increase in autism risk is seen in the periods after vaccination, either. (72). This study addresses the time concern appealingly, though once again its numbers could be larger.

The remainder of studies on this issue tend to use time-based methods which compare MMR vaccination rates against autism incidence rates in a large population and look for a correlation. (Indeed, the original Wakefield piece itself noted than an obvious way to test their theory would be to examine autism rates in the UK after the MMR vaccine was introduced.) The strength of these is in their potentially large numbers. The weakness is in the variety of confounding factors that are introduced with such large numbers and limited control. Most of these studies either attempt to determine if autism rates “stepped up” in the period immediately following MMR introduction, or they chart autism rates against MMR uptake rates and see whether the slopes match (or have any correlation).

This UK study finds around 300 autistic individuals in a population of tens of thousands, and matches their incidence against MMR vaccination over 11 years. Autism rates climbed at a steady rate; MMR rates held almost steady (over 95%) for the entire period, showing no correlation, unless invisible confounding factors were increasing the autism rates the entire time. (73—see figure 2 especially).

[Incidentally, the issue of rising autism rates comes up again and again. It is not clear whether actual autism rates are rising or other factors, such as changes in diagnostic definitions, are increasing the rate of diagnosis; see 74 for one take. It is an interesting topic, but not relevant for us.]

A Canadian study uses the time-correlate method with a large base (27,000), with 180 pervasive development disorder patients. (Autism is a form of PDD.) Interestingly, 82% were male; males have been shown repeatedly as much more susceptible to these disorders than females. Thimerosal was also used as a variable, which is beyond our scope here. Again, autism rates were found to rise continually and MMR rates held fairly constant, showing no association. The same criticisms of the above relevant studies above can be applied here, but this is otherwise a fairly strong piece (75).

Numerous other similar studies exist, but most of them use sample sizes small enough that they have little significance.

There has been no statistically significant studies that suggest a positive association between MMR and autism.

We now come to a more varied collection of evidence, much of which is compelling but none of which has the same weight of actual causal evidence as what we’ve seen above.

Dan Olmsted, a reporter, undertook some journalistic research and found that—by all appearances—the Amish of Lancaster County, who did not vaccinate, and several other largely non-vaccinating groups, had very low rates of autism (far lower than the regular population). His hypothesis was that the independent variable was the thimerosal they had avoided, or some other aspect of the vaccines (76 for the gist of the work).

Though interesting, there is no scientific weight to these observations. There are inummerable ways in which a group like the Amish differ from the larger population, and there is absolutely no way to isolate vaccination as a variable without some further degree of testing.

There is a significant body of work that examines some of the Wakefield claims from a clinical approach, the same angle he began from. These perform a spectrum of tests on patients that have been vaccinated, or exhibit autism, and seek to further explore complex body signs related to the biological mechanisms involved. These are completely beyond my ability to interpret, and probably beyond most readers; however, some commentators suggest that these support the Wakefield hypothesis. If true, they would do so in the sense of reinforcing the biological plausibility of the relation, rather than by actually demonstrating its existence. Better trained readers may be able to learn more from these; see Part H here (77).

Finally, there are a number of reviews that have been published that take a large net of the previously-demonstrated data (such as what has been presented here), analyzes it according to certain parameters of rigor, then presents a conclusion based on the body of evidence. Most of these either conclude that no link between MMR and autism have been demonstrated, or conclude that it has been refuted, but because the reviewers’ process is always essentially a black box—studies go in, conclusion comes out—I found them of limited application here unless I wanted to simply trust their opinion. Perhaps the review that would have been the most interesting would be the one published by the Cochrane Library, an evidence-based medicine resource (78). Their review purported to examine the entire body of available evidence with rigorous standards, ending up including 39 different articles and concluding that the evidence was largely inadequate. However, the review is not available without purchasing it directly from Cochrane, and the abstract is inadequate for our purposes, so I have ignored it.


The overall MMR paradigm appears something like this.

For a Westerner today, neither measles, mumps, nor rubella are common diseases, and it is therefore fairly unlikely he will contract them, with or without vaccination. However, they do have some rare complications which can be very serious, so it becomes partly a question of what level of risk he is comfortable with. This is particularly true if the odds of him contracting one of the diseases increases, for instance by decreased immunization rates in the population or by travelling into higher-risk areas.

Certain individuals are at a higher risk than others. For instance, I would say with near certainty that a doctor working in less-developed nations or a non-immune woman who could become pregnant should unquestionably be vaccinated. A mechanic living in the heartland would need it less. Anyone vaccinating should, if possible, be vaccinated with the current formulations and in the correct schedule to maximize their immunity.

In some cases, it may be preferable to contract the wild strains of these viruses and receive natural immunity. For instance, catching asymptomatic rubella (perhaps not even noticing) and becoming immune thereafter would be very appealing to many. Again, this would have to fall under the lens of the risk-based decision taking into account the possible dangers of the illnesses—the complication here is that these vaccinations are best performed on the very young, who cannot give meaningful input on their outlook and future habits.

One of the interesting interpretations of this debate is to note that it is, in some ways, almost meaningless. The diseases are rare and serious adverse effects are (except for some exceptions like CRS) generally extremely rare. Adverse effects of the vaccines are rare and serious effects are extremely rare. In some sense, anything one does with respect to this matter will probably turn out fine.

The trouble is that some of the low-odds possibilities here are very, very bad. Undergoing seizures from your vaccination or permanent deafness from the mumps—or death from anything—is nothing to joke about. Russian roulette, even with one bullet among a million empty chambers, is unappealing to most.

The autism question is a special case, because—although Wakefield’s paper precipitated the movement—it has gathered momentum chiefly due to parents anecdotally reporting sudden regression immediately following after vaccination. It is tempting to call these accounts temporal coincidence (everyone is vaccinated, plenty of people get autism, they tend to happen around the same time, so they’ll sometimes go together), emotionally charged by parents looking for something to blame. However, the truth is that none of this is relevant; all that is particularly relevant is the objective data about actual incidence rates and associations.

In the strictest analysis, the data on the autism issue are inconclusive. The body of epidemiological evidence is relatively broad, but due to the blunt nature of such studies, it is always possible that confounding factors have brushed out a subtle wrinkle in the data. Though there has been no direct evidence demonstrating a link, the evidence against it is also not been bulletproof.

However, it is reasonable to call an association between MMR and autism somewhere between unlikely and uncommon. The body of evidence may be flawed, but it is improbable that it is all so flawed that it is completely wrong—that the sum hundreds of thousands of data involved have all contributed inaccurate input. If, by chance, this did happen, it would still support the latter possibility, which is that autism is an extraordinarily rare effect of vaccination—so rare that it was able to hide behind minor statistical fluctuations. In some sense, these conclusions seem identical. If something is either not real, or so rare that it rounds to zero, the ramifications for most individuals are the same. In this sense, the autism-MMR link is at least as “disproven” as many accepted facts in the scientific oeuvre. It is not unreasonable to ask for additional scientific attention to the matter, however, which will probably focus on the clinical unless further large pools of data can be obtained and analysed by even-more-rigorous methods. Work on peripheral issues, such as Wakefield’s enterocolitis theory and the overall question of autism diagnosis and etiology, certainly merits effort.

Final conclusions are left to the reader.