Tamiflu side effects have been much in the news and we have concurrently been posting our mega-series on modeling antiviral resistance in influenza control. The two subjects are related in two ways, one obvious (Tamiflu is the main antiviral being stockpiled for influenza control) and one not so obvious: both topics are related to the fact that million, tens of millions or hundreds of millions of doses are contemplated. For antiviral resistance this means even very rare mutations producing a fully transmission-competent resistant virus can spread widely through the population (you will see that shortly in the modeling series). For drug side effects it means that even very rare side effects will occur almost certainly. The same will be true for vaccine use, and for drugs or vaccines used prophylactically (to prevent disease) it is made more difficult in that the victims will be healthy people.
What do we know about the Tamiflu problem? The data come from the country where the drug is used much more than elsewhere, Japan. The widespread use of Tamiflu there — 9 million prescriptions compared to 3 million in all other countries combined — provides the millions of possible exposures that could also produce detectable numbers of rare side effects. Last month two teenagers, both 14 years old, jumped to their deaths while on the drug. This month two more teenagers were injured after jumping from buildings. Fifteen children 10 to 19 have died jumping from buildings while taking the drug.
A news article in Nature by Ichiko Fuyuno (subscription firewall, alas) summarizes some of the other pertinent information:
Tokyo-based Chugai Pharmaceuticals, which distributes the drug in Japan, says that it has reported 289 cases of psychoneurotic effects, including three suspicious deaths, to the Japanese health ministry since the drug was launched there in 2001. Chugai now lists the possibility of severe neurological side effects on the drug’s labelling and has distributed a warning note to hospitals. The Japanese health ministry insists that there is no clear evidence of a link, however, and points out that flu itself can cause symptoms such as abnormal behaviour. Roche agrees, adding that the number of suspicious deaths is tiny compared to the number of people who have been prescribed the drug.
Tamiflu was cleared initially when seven paediatricians and a statistician looked at the side effects of it and other influenza drugs, as well as at the symptoms of flu itself. The study, which ended in February 2006, followed up 2,846 children aged mostly ten or younger who had been diagnosed with flu. The frequency of abnormal behaviour in children who took Tamiflu was 11.9%, compared with 10.6% in those who didn’t take it, which the researchers concluded was not a significant difference.
Shunpei Yokota, head of the study group and a paediatrician at Yokohama City University’s Graduate School of Medicine, admits that the study had shortcomings, including a poor definition for the term ‘abnormal behaviour’. So in February, at the government’s request, Yokota’s team launched a larger study, which will trace 10,000 people aged 0-18 years. The team aims to release the results by this autumn. (Nature, subscription)
Unfortunately there are more complications to the study than just “shortcomings.” There is a whiff of conflict of interest, as well. Yokota and two other members of the study team received fees from the distributor, although not in relation to the study. Meanwhile, the Japanese are advising against use in anyone ages of 10 and 19 unless they are “high risk.” The US and Canada have also asked that warnings be placed on the drug label.
This is part of a generic problem of how to detect and cope with rare adverse reactions to drugs and biologics like vaccines. If there is a one in a million risk of death and you give the drug or vaccine to 100 million people, well, you do the arithmetic. If it happens in a short space of time, say 2 months because of a mass vaccination or antiviral prophylaxis and it happens to otherwise health people, well, you predict the public reaction. The paradoxical result could be that more people would die because of a failure to convince the public the risk benefit equation of dying from a vaccine reaction versus dying of the flu heavily favors the vaccine.
So it’s a problem even if you know it’s going to happen. It’s even bigger if you are caught unawares, as with Vioxx and possibly Tamiflu. The FDA and most other national systems for detecting adverse drug reactions are passive, depending on the voluntary filing of reports by physicians. Most doctors who see adverse reactions don’t bother to report them and aren’t required to. So now there are suggestions that existing repositories of health information be actively “data mined” for adverse drug events.
Last week, Mark McClellan, a former FDA commissioner, told a Senate committee considering new drug-safety legislation that the system “needs to do better than just seeing the tip of the iceberg of a safety problem after it has already hit us”. Health-information technology for drug safety is “an idea whose time has come”, he added.
McClellan and others are pushing the idea of data mining of existing health-record databases as an active surveillance system to pick up early warnings of adverse side effects. By pooling existing databases run by private insurance plans, government agencies and industry, information on well over 100 million people could be studied in something much closer to real time, he says. The FDA could then identify priority questions and the mechanisms for answering them. If such a system had been in place when Vioxx came on the market in 1999, that drug’s dangers could have been detected in months rather than years, McClellan contends.
The same thinking is gaining currency in Europe. “There is great potential in the use of these healthcare databases,” says Panos Tsintis, who oversees the safety of marketed drugs for the European Medicines Agency (EMEA) in London. (Meredith Wadman, Nature [subscription])
Maybe this will work, although my knowledge of data mining technology tells me there are substantial obstacles. Meanwhile, we will have to resign ourselves to the rare adverse event when hundreds of millions are being dosed. It’s a bit like buying a lottery ticket with each capsule, where the prize is death. Even if the risk is small, you at least want to be sure that the pay-off is likely and worth it. I’m not sure we can say that for many of the drugs now taken in high volume that have much more frequent adverse reactions than Tamiflu or a vaccine.
Just one of the many paradoxes of public health practice.