Thursday, April 29, 2010

Bizarre Things Purported to Cause Autism: Cells from Aborted Fetuses

EXECUTIVE SUMMARY: A pro-life advocacy group called Sound Choice Pharmaceutical Institute has claimed that the increase in autism-spectrum diagnoses in the U.S. over the past two decades is due to the presence of cells from aborted fetuses in vaccines introduced during that period. The evidence they have for this consists of chronological "change points": years after which the rate of autism incidence begins to rise more sharply than it has before. They connect these change points with developments in U.S. immunization practices that might mean more exposure to vaccines made from fetal cells.

I think the idea that those two developments are related is a ridiculous one, first because it is physically impossible that the trace amounts of human cell components still found in the finished vaccines could trigger such a massive immune response even in very sensitive children, and second, because the timing of the change points they've chosen to highlight is an artifact of the autism-incidence statistics they've chosen to use, and third, because many other factors track with those change points that have a more obvious, immediate effect on reported autism incidence, like revisions of the Diagnostic and Statistical Manual (DSM), changes in special-education policy, etc.

There's a blog I used to read occasionally called Autistic Conjecture of the Day --- sadly, it's now open only to invited readers --- that dealt with every idea the blogger had ever encountered for what causes autism. Some of these ideas would be mainstream, some would be cutting-edge, some would be controversial, and some would be just plain off-the-wall.

Since discovering that blog, though, I've loved that idea: simply collecting all the different ideas in one place.

So I've decided I'm going to do my own (intermittent, open-ended) series on Bizarre Things that have been Purported to Cause Autism! Because this is a perseveration of mine, as well, and you can never have too much debunking of wacky ideas on the Internet.

The subject of this inaugural post in the series comes courtesy of fellow feminist skepticblogger Amanda Marcotte: according to pro-life columnist Jill Stanek, and this story on the LifeNews website, the "worldwide jump in the incidence of autism beginning in 1988" can be traced to the use of human fetal cells in the measles-mumps-rubella (MMR) vaccine.

Quoting the pro-life advocacy group Sound Choice Pharmaceutical Institute, Stanek argues that the three "change points" --- points at which the slope of the line on a graph of autism incidence changes --- occurring in 1981, 1988 and 1995 coincide with three changes in which vaccines are recommended for very young children. In 1981, the CDC's Advisory Committee on Immunization Practices (ACIP) approved a rubella vaccine called Meruvax, which is made from live attenuated rubella viruses that have been incubated in human fetal cells from the cell line WI-38; in 1988, a second dose of the MMR vaccine was added to the recommendations; and in 1995, a chicken-pox vaccine (Varivax) was approved that, like the rubella vaccine, contains live attenuated viruses that are grown in human lung cells. The chicken-pox vaccine used cells from two different cell lines, WI-38 and MRC-5, both of which were cultured from lung cells taken from human fetuses* that had been aborted.

I'm going to address the autism-incidence statistics Stanek's sources relied on later in this post; for now, I'd like to concentrate on what is, for me, the biggest problem with her claim that human fetal cells used to grow the viruses used in common childhood vaccines are causing autism: where is the mechanism? The mercury and MMR hypotheses, flawed as they are, at least propose some physical process by which the brain's development can be disrupted as a result of vaccine exposure.

Though neither Stanek nor the newsletter from which she quotes goes into how residual fetal cells in vaccines might have such an effect on an infant's nervous system, this July, 2009 editorial by Sound Choice Pharmaceutical Institute president Theresa A. Deisher makes a bold suggestion:
How could the contaminating aborted fetal DNA create problems? It creates the potential for autoimmune responses and/or inappropriate insertion into our own genomes through a process called recombination. There are groups researching the potential link between this DNA and autoimmune diseases such as juvenile (type I) diabetes, multiple sclerosis and lupus. Our organization, Sound Choice Pharmaceutical Institute (SCPI), is focused on studying the quantity, characteristics and genomic recombination of the aborted fetal DNA found in many of our vaccines.

Preliminary bioinformatics research conducted as SCPI indicates that "hot spots" for DNA recombination are found in nine autism-associated genes on the X chromosome. These nine genes are involved in nerve-cell synapse formation, central nervous system development and mitochondrial function.
So, the hypothesis here is that fetal DNA left in the sample of live virus that makes up the MMR and chickenpox vaccines is merging with people's own genomic DNA and scrambling their genes.

Like all good science fiction, this fantastical scenario springs from fairly plausible conceptual roots: it's true that one type of virus --- the retroviruses (like HIV) --- can embed its tiny genome within its (human) host's much-larger genome, and that some other viruses, like herpesviruses, can keep their genetic material dormant within a host cell, and thereby persist in that host for years, or even decades, with periodic reactivation of its infectious cycle when the host is under physiological stress; it's also true that a few people do suffer severe allergic reactions to some of the ingredients in vaccines, including --- what's most relevant to this post --- cells used to incubate the vaccine's active component, the virus.

In this 2003 article in Pediatrics on vaccine safety, Drs. Paul Offit and Rita K. Jew discuss the relative dangers posed by all the different vaccine additives: preservatives (like thimerosal, phenol, and 2-phenoxyethanol), adjuvants (aluminum salts), stabilizers (sugars, amino acids like glycine or monosodium glutamate, proteins like gelatin or human serum albumin), and manufacturing residuals (into which category the fetal cells Dr. Deisher is so worried about would fall, along with egg proteins, yeast proteins, formaldehyde and various antibiotic drugs).
Residual quantities of reagents that are used to make vaccines are clearly defined and well regulated by the FDA. Inactivating agents (eg, formaldehyde), antibiotics, and cellular residuals (eg, egg and yeast proteins) may be contained in the final product.
Cellular Residuals

Egg Proteins
Egg allergies occur in approximately 0.5% of the population and in approximately 5% of atopic [i.e., allergically hypersensitive] children. Because influenza and yellow fever vaccines both are propagated in the allantoic sacs of chick embryos (eggs), egg proteins (primarily ovalbumin) are present in the final product. Residual quantities of egg proteins found in the influenza vaccine (approximately 0.02-1.0 μg/dose) are sufficient to induce severe and rarely fatal hypersensitivity reactions in children with egg allergies. Unfortunately, children with egg allergies often have other diseases (eg, asthma) that are associated with a high risk of severe and occasionally fatal influenza infection. For this reason, children who have egg allergies and are at high risk of severe influenza infection should be given influenza vaccine through a strict protocol.

In contrast to influenza vaccine, measles and mumps vaccines are propagated in chick embryo fibroblast cells in culture. The quantity of residual egg proteins found in measles- and mumps-containing vaccines is approximately 40 pg --- a quantity at least 500-fold less than those found for influenza vaccines. The quantity of egg proteins found in measles- and mumps-containing vaccines is not sufficient to induce immediate-type hypersensitivity reactions, and children with severe egg allergies can receive these vaccines safely.
This article does not mention the rubella vaccine, Meruvax, which is propagated in human fibroblast cells in culture, but because the vaccine is prepared in the same way as the measles and mumps vaccines (just using cultures of a different cell type), it's probably safe to assume that residual human cell components are present in the same minute quantities. Chances of severe immune response to residual cell components should be even lower for the rubella vaccine than for the measles or mumps vaccines since the attenuated rubella viruses are grown in human cells as opposed to chicken cells, and human cells are less likely to trigger an immune response in humans than nonhuman cells, since our immune systems work by distinguishing "self" from "non-self" antigens.

As for the DNA from those residual cells somehow integrating itself into people's genomes, I don't think that's even possible. DNA rearrangement like that doesn't happen on its own, whenever two different molecules of DNA encounter each other; retroviruses that inject their genomes into a host cell's genome are able to do so because their protein shells (capsids) also include all the requisite enzymes for doing this --- reverse transcriptase, for converting the virus's RNA genome into a length of double-stranded DNA suitable for patching into a host cell's chromosomal DNA, and, crucially, an enzyme called integrase, which cuts the ends off of both viral and host DNA and then pastes them together. This enzyme exists only in retroviruses and some bacteriophages --- humans have nothing like it, and neither do any of the viruses present in the MMR or chicken-pox vaccines.

Expecting one piece of DNA spontaneously to insert itself into another because they're in the same cell and might happen to bump into each other is as misguided as placing, say, two whole eggs, a bag of flour and a bag of sugar out on your kitchen counter and leaving them there overnight, with the expectation that they will have turned into a cake by the time you get up the next morning.

Finally, let's look at the graph Jill Stanek reproduces in her column, that purports to show an increase in autism incidence coinciding with those three additions to the list of recommended vaccines.

The smaller, inset graph making up this image comes from this 2008 article in the Archives of General Psychiatry, tracking the number of children in California who are a) diagnosed with autism or a related condition like Asperger's syndrome or PDD-NOS, and b) receiving services from the California Department of Developmental Supports.
You can see that this graph has three pairs of lines representing three different age cohorts of children (with the incidence of autism measured two ways for each group; hence the double lines) where the modified version in the SCPI graphic has only one --- the (merged) middle pair of lines, representing the rate of autism diagnoses occurring among four-year-old children each year. The top two sets of lines --- showing the incidence data for four- and five-year-olds --- do trend sharply upward starting around 1995, but for the third set --- three-year-olds --- the rise is much less pronounced, and starts later.

The larger, background graph comes from this article (full text here) in the March 15, 2010 issue of Environmental Science & Technology, in which two EPA researchers look at autism-incidence trends worldwide over the past fifty years or so by looking at three large autism-incidence studies from the U.S. (this report from the California state Department of Health and Human Services' Department of Developmental Supports), Denmark (Lauritsen et al., 2004), and Japan (Honda et al., 2005; full text here) and estimating each subject's year of birth, and then plotting the number of autistic people born between 1960 and 2000.

The EPA study found just one "changepoint year": 1988. After that point, the slope of each line representing each of their four data sets (each of the three already cited, plus a "worldwide" data set that includes all of those three put together, plus a bunch of other studies too small, or too variable in timing or methodology, to be analyzed on their own) jumps to a value several times that of the slope of the line prior to that point.

That discrepancy is also reflected in the raw data pre- and post-1988: in California, the average incidence of DSM-IV Autistic Disorder was 5.7 cases per 10,000 children in the years before 1988; after 1988, it was 20.8 cases. In Denmark, 1988 marked an increase from 0.6 to 6.6 cases of ICD-10 Autistic Disorder per 10,000 children; worldwide, the change was from 6.0 to 24.2 autistic children per 10,000 live births.

I did not forget to mention Japan; the Japanese data starts in 1988.

No changepoint data could be calculated for the Kohoku Ward, Japan, data set (Table 1), as AD cumulative incidence increased continuously over the entire period of study (1988-1996)(Figure 1a).

As Mark Chu-Carroll of Good Math, Bad Math remarks about another crucial data discrepancy*** hinging on that unlucky year:

[T]he earliest data in their analysis comes from a different source than the latest data. They've got some data from the US Department of Education (1970->1987) and some data from the California Department of Developmental Services (1973->1997). And those two are measuring different things; the US DOE statistic is based on a count of the number of 19 year olds who have a diagnosis of autism (so it was data collected in 1989 through 2006); the California DDS statistic is based on the autism diagnosis rate for children living in California.

So - guess where one of their slope changes occurs? Go on, guess.


The slope changed in the year when they switched from mixed data to California DDS data exclusively. Gosh, you don't think that might be a confounding factor, do you?

While the absence of any Japanese data from before 1988 wouldn't affect the change-point calculations for California or Denmark, it does make me a bit more skeptical of the worldwide changepoint, which was around 1989. It does help a bit to know that there are other bits and pieces of data from around the world helping to fill in the patchwork --- while it still might be a bit lopsided, depending on how the various studies are distributed in time, it's at least not as much of a foregone conclusion that the slope will change in 1988.

Besides the statistical uncertainties arising from such an uneven, patched-together data set, there are also some confounding factors arising from the fact that all this data came from administrative databases maintained by government human-services agencies. This introduces changes in policy as other potential influences on how many people show up labeled "autistic" (see this old post by Michelle Dawson for more on how this works).

The authors of the EPA study, Michael E. McDonald and John F. Paul, understand this and are accordingly cautious about whether their observed increase in autism incidence is "real" or not:

The three studies that were selected for inclusion in our analysis had sufficient record length for time trend analysis because they collected administrative data for programmatic services to autistic children. Each of these studies, which showed cumulative incidence of autism increasing (Figure 1a), had the advantage of using the data collections from a single administrative database that covered a well-defined geographic region. This was not true of our worldwide data set, where differences in method of data collection, record length, and geographic area may all have contributed substantially to an increase in AD cumulative incidence (Figure 1b).

Administrative databases also have the advantage of a relatively consistent methodology over periods of time. However, all three of our selected databases have some methodological changes associated with their long-term data collections. A changing of the diagnostic criteria and the broadening of the definition of autism to PDD occurred during the data collection within the Danish and California databases, but a single diagnostic criterion was applied consistently within the Kohoku Ward study. As we were aware of the issues with broadening and changing diagnostic criteria, in our study selection criteria we chose explicitly to focus on AD, which has had relatively consistent diagnostic criteria since about 1978. However, this does not necessarily mean that the diagnostic criteria have been consistently applied in practice over this time frame. It does appear that AD criteria have been applied fairly consistently since about 1994 in the Danish database and across the United States with the use of ICD-10 and DSM-IV, respectively. A recent analysis of the California database from the early 1990s through about 2006 suggests that changing diagnostic criteria may account for a 2.2-fold higher cumulative incidence of autism, relative to the 7-fold increase observed over 11 birth cohorts. It is unknown how consistently previous AD criteria were applied or how the application of the current criteria compare with past criteria.

Administrative data may also be prone to diagnostic substitution, where children with multiple diagnoses may be identified differently over time depending on which diagnosis allows the individual to receive administrative services. In British Columbia, Canada, changes in the assignment of special education codes may account for at least one-third of the increase in autism prevalence from 1996 to 2004. However, in the California data set, diagnostic substitution from the category of mental retardation to autism could not account for increased autism from 1987 to 1994.


Studies with AD assessment in children occurring before age 10 may show an apparent increase in autism because of earlier ages of diagnosis in recent years, relative to historic underidentification at the same assessment age. In fact, Parner and coinvestigators examined recent cohorts (1996 and 1997; 1998 and 1999) in the Danish database and found that at least some of the AD increase was attributable to earlier diagnosis. In the California database, a shift toward a younger age at diagnosis also was found and contributed to about 12% of the observed increase in autism from 1990 to 1996, based on assessment at age 10. Thus, earlier diagnosis contributed to increase in AD cumulative incidence in at least two of our selected studies and, likely, to studies in our worldwide data set (Table S1, Supporting Information).

In addition to finding that changes in diagnostic criteria and earlier age at diagnosis do contribute to some of the observed increase in cumulative AD incidence in the California database for 1990-2006, Hertz-Picciotto and Delwiche also found that the inclusion of milder cases of autism contributed to the increase. This contribution was not as much as that resulting from changing diagnostic criteria but was more than that contributed by earlier age at diagnosis. Differential migration of autistic children into the state was also found to play a minor role in the increase. The investigators suggest that wider awareness of autism, greater motivation to seek services, and increased funding for services also may contribute to increasing cumulative AD incidence, but these factors could not be documented or quantified.

I've already spent a ridiculously long time**** writing this post, so I'm not going to try to unearth every single policy change that might affect autistic children's enrollment in state disability programs. I can tell you, though, that each of the change points SCPI identifies corresponds with historical events much likelier to account for changes in how many people receive a given DSM diagnosis than the addition of a few new vaccines to the recommended vaccine regimen: in 1980, 1987 and 1994 the DSM underwent major revisions. While I'm not sure about the DSM-III, I know for sure that the DSM-IV greatly relaxed the diagnostic requirements for Autistic Disorder from what they were in the DSM-III-R, and added the category of Asperger's syndrome. (The 1995 changepoint, identified by SCPI and taken from a study not used by the EPA analysts, reflects all the different autism-spectrum diagnoses, not just Autistic Disorder). Changing diagnostic criteria, as well as a sea change in people's ideas about what sort of person might be called "autistic," seem to me much more likely to be behind the increasing rate of autism diagnosis than exposure to a few femtograms of human fetal DNA fragments via vaccines.

*(It is hard to find sources on the Internet that can verify that claim --- while most laboratories that use those cell lines, or sell them for other researchers to use, will classify the cells as human fetal cells, and describe their cell type, growth characteristics and other things it would be useful to know if you were going to try to culture those cells, they don't go into huge detail about the circumstances of those cell lines' establishment. Why would they? But I was able to find original articles describing the cell lines --- MRC-5 was established in 1970 by British researchers J. P. Jacobs, C. M. Jones and J. P. Baille, and described in this 1970 letter to Nature; WI-38 was established sometime in the 1960s at the Wistar Institute in Philadelphia, Pennsylvania, by cell biologists Leonard Hayflick** and Paul S. Moorhead, according to a method described in this 1961 article (full text here) in Experimental Cell Research. That paper only describes the creation of cell lines WI-1 through WI-25, though; WI-38 came later. Papers using WI-38 cells usually cite either Hayflick and Moorhead (1961), or this later article by Hayflick (1965), also published in Experimental Cell Research. Hayflick and Moorhead (1961) do not go into detail about the source of their human tissue cells; they only mention that the source of the lung tissue used to establish some of the cell lines, which would include WI-38, is a three-month-old fetus. So, weird as it might sound, it's actually true that some vaccines are produced using cells derived from aborted fetuses.)

**AnneC, if you're reading, you might have heard of this guy! He discovered the "Hayflick limit," or the maximum number of times a given cell can divide before it deteriorates past functioning. He's done lots of research on the biology of aging and life extension, the latter of which I know is one of your perseverations.

***He seems to be using the SCPI newsletter as his reference, and not trying to tease out which errors are theirs and which were already in the EPA's somewhat more conservative analysis. The problems with the Department of Education data he mentions would apply only to the SCPI's findings, since the EPA researchers did not use any of that data. The problems he has with "hockey-stick analysis" in general, though, would seem to apply to both parties. SCPI just compounds the EPA's error by doing multiple such analyses on their already-too-small data set.

****A little over two weeks.

McDonald, M., & Paul, J. (2010). Timing of Increased Autistic Disorder Cumulative Incidence Environmental Science & Technology, 44 (6), 2112-2118 DOI: 10.1021/es902057k

No comments: