Last week I commented on the retraction by "Lancet" of an article published by Andrew Wakefield and a number of colleagues in 1998. That article claimed that research evidence supported the hypothesis that vaccination for some childhood diseases played a causal role in the development of autism. Wakefield's paper has been a major focus of the anti-vaccine movement and has encouraged parents to refuse vaccination of their children-- with some predictable results in terms of disease and even death of children.
"Lancet's" retraction of the paper has received attention from bloggers, journalists, scientists, and concerned parents, as we might expect. Criticisms of Wakefield's work have been repeated in a variety of venues. A typical article of this kind was in "Time" magazine, Feb.15 (Claudia Wallis, "Debunked. A pivotal paper linking vaccines and autism is retracted....", p. 18).
The "Time" article has in common with the majority of other critiques its focus on the ethical errors in Wakefield's work. Indeed, there were ethical errors by the carload. Wakefield failed to disclose his financial interests in vaccine protocols or in a relevant lawsuit where he was acting as a consultant; he procured blood samples from children whom he knew personally, thus making it awkward for the children or their parents to hesitate to consent to the procedure; he ordered unnecessary and intrusive tests for some children. All these ethical errors would be reasons for reprimands, for the withdrawal of job opportunities, for ordering Wakefield to attend an ethical training program. His withholding of financial disclosure would be good reason for a journal to refuse publication, to print a correction if the article was already published, or in some cases to retract the published material.
However, repellant as Wakefield's ethical conduct may be, it is not in itself sufficient to support the decision that an article's conclusion is wrong. A researcher might fail to document informed consent, might intimidate parents and children to get them to participate, and might secretly be on the salaried staff of a corporation with strong related business interests, but if his or her conduct of research, collection of data, and analysis of data were beyond criticism, the research itself could still be of value. Another researcher might follow all rules about informed consent and financial disclosure, but if he or she failed to do a good job in designing and carrying out a study, the conclusions drawn from that study would be worthless. Ethical errors raise questions about the researcher's integrity in collecting and handling data, but are not necessarily evidence that the basic work has been done incorrectly. Compliance with ethical rules suggests that the researcher approaches his or her tasks with integrity, but adds no credibility to data or conclusions when it's clear that these do not follow design guidelines. Journals refuse or retract papers when ethical errors are evident, but this is not because their action assures the publication of good material only; it's because success or failure in publication are the major carrots and sticks available to enforce compliance with ethical rules.
I have seen very few comments about the design flaws of Wakefield's retracted paper, so I will briefly repeat those I mentioned last week. A major problem with the paper is the small number of children studied. Errors due to chance are much more likely to occur when a study involves twelve children than when it includes 120 children. Such errors might give the impression that an effect is much larger than it really is, but they could also make it appear that it is much smaller than it would appear to be with more data to consider. This is an especially important point in the light of other studies looking at many thousands of children and failing to reach the conclusions reported by Wakefield. A second major problem has to do with Wakefield's method of ascertaining what symptoms children showed, and at what times in their lives. Rather than using objective records like notes or videotapes dating from the children's early years, Wakefield asked parents to recall events that had occurred eight years or more earlier. In taking this approach, he ignored the fact that human memory reconstructs past events rather than plucking them intact from some sort of bookcase of internal recordings. Previous and subsequent events help to determine how, or even if, reconstruction takes place. Memory research dating back almost a hundred years has shown that memories are influenced by expectation and motivation, and that repeated recall, rehearsal, discussion, and consideration of events change the way memories are experienced. These problems are the real reasons why Wakefield's study cannot be considered to provide evidence genuinely supportive of the conclusion drawn and much promulgated.
How could Wakefield have obtained any objective information to test his hypothesis about vaccines and autism? It's not easy, but there is a method that was used by one autism researcher (asking a different question, not related to vaccines). Philip Teitelbaum, writing in the Proceedings of the National Academy of Sciences in 1998, described requesting home videos of infants and toddlers from the public, and looking for correlations between unusual early movement patterns and later diagnoses of autism. He was thus able to collect a large number of video records. Most of the children developed typically, but Teitelbaum was able to find a small proportion of children with later diagnoses of autism and to compare their unusual movements with the same kinds of movements as they occasionally occurred in the larger group of typical children. A similar analysis of home videos could provide useful information about vaccine effects, if indeed there are any.
[I regret that a series of personal attacks on colleagues and contributors, as well as on me, make it impossible to allow comments on this post].
