11 Feb 2010


Recently, our government officials were asked about incidents of Guillain-Barre syndrome and intrauterine deaths in people who received the new swine flu vaccine. Every time, their response was monotonous. “There is no evidence that the risk of Guillain-Barre syndrome (or intrauterine deaths or any complications, depending on the original question) is increased with the use of the vaccine.”

According to the tone of the media, their claim was hardly reassuring. Many still fear that the government is tricking them to use an unsafe product. (The effect was better illustrated at http://ccszeto.blogspot.com/2010/01/armor.html.) Although I do not really think the vaccine is dangerous, I have the uneasy feeling that the government may intentionally or unintentionally confuse the public.

In evidence-based medicine, the lack of evidence can never be used as evidence against a concept or hypothesis. For example, if no one has ever tested a drug in a certain disease, you can just say you do not know if it works but cannot state it does not work. The same also applies to harm. If you never look for the harm, how would there be evidence?

In the last two months, the efficacy data of the swine flu vaccine became available. [Greenberg et al. N Engl J Med 2009;361:2405-13, Zhu et al. N Engl J Med 2009;361:2414-23, Clark et al. N Engl J Med 2009;361:2424-35, Arguedas et al. N Engl J Med 2010;362:370-2, Nolan et al. JAMA 2010;303:37-46] Notably, all these studies excluded pregnant women. Similar to our government, WHO and CDC also encourage pregnant women to receive the new vaccine and reassure the public that studies on pregnant women are ongoing. If we search ClinicalTrial.gov, we can see that a handful of studies are recruiting pregnant women to try different brands of the swine flu vaccine. Each study recruits around 150 to 300 women. Thus, in the end we probably will have safety data of 1000 to 2000 women. It is reassuring, isn’t it?

This leads to the question of statistical power. In 2003, data from the US National Center for Health Statistics showed an average fetal mortality rate of 6.9 deaths per 1000 births. Using online calculators, we can estimate that a sample size of 2000 subjects can achieve a confidence level of approximately 0.5% at a 95% confidence interval. Although this confidence level is already respectable, it also means that we can conclude the new vaccine causes harm only if the rate of fetal death has doubled. Therefore, the importance of post-marketing surveillance cannot be overemphasized.

In the end, the vaccine probably will result in more benefit than harm at the community level by reducing the number of people infected with the virus. However, the risk-to-benefit ratio is less certain at the individual level. People who stress the mortality rate of swine flu is higher than the harm of the vaccine should consider the fact that people not receiving the vaccine may not contract the virus at all. While the motives of various health authorities are understandable, mixing up the above concepts can be misleading.

1 comment:

  1. Absence of evidence is not evidence of absence