The Loom

i-1c196e24122aab0ea1ad5dc92a919100-papillomavirus.jpgOur culture wars make for strange ironies. The fight over the cervical cancer vaccine is a case in point.

Yesterday news broke that a vaccine for cervical cancer might not be all it’s cracked up to be. Cervical cancer is caused by a virus known as human papillomavirus. It infects epithelial cells in the skin and other surface layers of the body, including the vagina and throat. On rare occasion it causes its host cells to start replicating madly, creating growths that sometimes progress into full-blown tumors. It’s a major menace: the American Cancer Society estimates that it causes 17 percent of all cancer cases–more than 1.8 million a year. Merck has designed a vaccine that turns out to be 98% effective against the types of virus that cause most cases of cancer. And yet, paradoxically enough, doctors found that when women were given the vaccine, it only had a 17% efficacy rate in preventing the lesion that can lead to cancer.

The coverage of the study was massive, but it seemed to me to keep away from the heart of the matter. Many reports zeroed in on the study’s finding that human papillomavirus may cause cancer in the throat through oral sex. Others took note that the paper came out just a couple days after Texas governor Rick Perry announced he would not veto a bill passed by the state legislature blocking his order that all girls in Texas get the vaccine. The move to block the order has been driven by several political groups, such as the Free Market Foundation and Texas Physicians Resource Council. They objected to the state taking over the decisions of parents, and also argued that there were too many questions about the new vaccines. It was possible that other vaccines to other strains of the virus would be more effective. A report in the Baptist Press claimed that the new study supported this claim.

Not having heard of these groups before, I checked out their web sites. When I visited the Free Market Foundation site, I had to smile. They not only fight against the Merck vaccine; they also fight against evolution. Their voters guide (pdf) includes on its list of make-or-break issues, “Present scientific evidence in our public schools supporting intelligent design, and not just evolution, and treat both theories as viable ones on the origin of life. You can catch their president, Kelly Shackelford, on a PBS show coming up about “Darwin versus Design.”

I smiled because I had just finishing reading some new papers with titles like “Multiple Evolutionary Mechanisms Drive Papillomavirus Diversification.” To understand why vaccinating against human papillomavirus is so tough, scientists study how it evolved over millions of years, just as it is evolving today. [More below the fold…]

Viruses sometimes mutate as they replicate inside host cells. Those mutations may be the result of sloppy copying of virus DNA, or of fragments of different strains of viruses recombining into new hybrids. Natural selection favors some of these mutants if they out compete other viruses, or if they can colonize some new niche. In recent years, scientists have been sequencing the genomes of about 150 different papillomaviruses and comparing them to figure out how they have evolved–and how some of them have evolved into cancer triggers. It’s a tough puzzle, because of the way viruses swap genes. All of the genes in a given papillomavirus may have different evolutionary history. In the study I referred to above, researchers in Europe searched for the stretches of DNA in the viruses that provide the clearest picture of their history.

The big debate going on in papillomavirus circles is how all the strains of the virus that infect humans ended up infecting humans. In 2006, for example, researchers at the University of California at Irvine found distinct differences between the strains, leading them to conclude that “all HPV types existed already when humans became a species. Consequently, humans had always suffered from lesions like anogenital cancer, genital warts and common warts.”

But the picture gets more complex when a team of European scientists looked at the papillomaviruses that infect other warm-blooded animals. (The viruses can sometimes cause cancer in them as well, and they are considered a serious threat to some endangered mammals.) The scientists compared how closely related virus strains were to how closely related their hosts were. If the virus’s evolution simply tracked that of its host, you’d expect their evolutionary trees to be mirror images. All human papillomaviruses should be more closely related to one another than any is to viruses in other animals, for example, and of those animal strains, the ones infecting chimpanzees should be closest. There are parts of the virus and host trees that mirror each other, indicating millions of years of coevolution. Some closely related branches of papillomavirus only infect whales and dolphins, for example, and no other species.

But there are also parts where the symmetry is shattered. Some human papillomaviruses all form a single tuft on the tree, but sprouting out of the middle of it is a strain that infects macaque monkeys. The virus leapt between hosts separated by over thirty million years of evolution. Many of these jumps occurred millions of years ago, after which the viruses evolved to adapt to their new hosts and then tracked their hosts’ evolution closely. But in other cases, the jumps were recent–perhaps even in the past few thousand years as humans and animals have spent more and more time in close contact.

As these viruses jump, mix, and evolve, they create a complex ecosystem inside every host. Even the skin of a perfectly healthy human or animal may be rife with different strains of papillomaviruses. Some viruses may do a better job of infecting cells than others, and come to dominate the virus ecosystem. Vaccines that wipe out one or two top cancer-causing viruses may just be making room for others. And unfortunately, the supply of potential sources of cancer–both in humans and in other animals–is practically bottomless. Not only is there a vast pool of these viruses already, but they can evolve rapidly, swapping genes that make them more of a risk to causing cancer.

It may seem strange that someone could simultaneously undermine teaching evolution and use the evolution of viruses to fight against a vaccine program. But this sort of disconnect is nothing new (paging Jeb Bush). Let’s just hope it doesn’t get in the way of more research of the sort I described here.

[For evolutionary tree-huggers, here’s the key figure from the new paper. Click on it for a bigger view:]

i-9655654fe2aa0db3fa6e40e9ddfb9975-small HPV tree.jpg


  1. #1 Noumenon
    May 11, 2007

    Those “bootstrap support value” numbers sitting on each branch of the tree… those are the probability the computer thinks it found the right spot for the branch?`

  2. #2 Chris Harrison
    May 11, 2007

    Damn my state!

    Great post as usual Carl.

  3. #3 factician
    May 11, 2007

    I need to re-read the paper to be sure, but I think you’re misunderstanding the findings of the vaccine efficacy study (either that or I am). They report 100% efficacy for pre-lesion screening in their per protocol population. They report lower efficacies (depending on the lesion) in their intention to treat population. The difference between these two populations is the the fact that the intention to treat population may already have been infected with the virus as the study began (and before the vaccine could kick in). It’s very likely that the efficacy of this vaccine is fantastic, if you can give it to people before they are exposed to the virus. For folks who get the virus during concurrent with vaccine administration, there is a smaller efficacy. And for folks who already had the virus, they’re pretty much SOL. Of course, I could be totally baked on this. Perhaps I should reread the paper…

  4. #4 Carl Zimmer
    May 11, 2007

    Factician: The paper is definitely a slog, but I do think I’m correct. You might want to look at the editorial that accompanied the papers: “…Another factor explaining the modest efficacy of the vaccine is the role of oncogenic HPV types not included in the vaccine. At least 15 oncogenic HPV types have been identified,4 so targeting only 2 types may not have had a great effect on overall rates of preinvasive lesions.”

  5. #5 factician
    May 11, 2007


    Thanks for the link to the editorial. I think this quote should at least partly clear things up:

    “Why is vaccine efficacy modest in the entire cohort? One factor is the apparent lack of efficacy among subjects with evidence of previous exposure to HPV types included in the vaccine. … Another factor explaining the modest efficacy of the vaccine is the role of oncogenic HPV types not included in the vaccine. …. What can be inferred from these data about the potential effect of vaccination among girls 11 and 12 years of age? The FUTURE trials did not enroll subjects in this age group. Within both trials, subgroups of subjects with no evidence of previous exposure to relevant vaccine HPV types were evaluated separately for vaccine efficacy. In these subgroups, efficacy of nearly 100% against all grades of cervical intraepithelial neoplasia and adenocarcinoma in situ related to vaccine HPV types was reported in both trials. However, it would be important to know the overall rates of grade 2 or 3 cervical intraepithelial neoplasia or adenocarcinoma in situ regardless of HPV types. Without these data, it is difficult to infer both the effectiveness of vaccination and the role of nonvaccine HPV types in overall rates of preinvasive lesions.”

    Sorry for the long quote, but I think the long and the short of it is that in the areas where they gave us data, patients who showed no sign of prior exposure to HPV showed near 100% efficacy – which is crazy-good for a vaccine (but they don’t show us several of the cancer precursors that they should). Most of the data includes patients who were previously exposed. It’s a pity they don’t show us the data for those other precursors for patients who show no sign of pre-exposure. I wonder what that would tell us? I wonder if subdividing into those groups was an afterthought? And I wonder what we’ll see from studies of 10 year olds receiving the vaccine?

    And of course the true pity is that all of these are surrogate outcomes, so it could be 20 years before we have any definitive answer on whether Gardasil actually protects against cancer.

  6. #6 Nick Anthis
    May 11, 2007

    Even the prevention of 17% of cases of such a common type of cancer is pretty damn good. It’s just a shame that the mandatory vaccination program in Texas was DoA.

  7. #7 JIm Hu
    May 12, 2007

    Are trackbacks working? I get a not found error when I ping. Manual trackback:

    HPV and Texas

    … Since I’ve been in NY on sabbatical, I haven’t seen the local coverage of this, but it looks like the story is somewhat more complicated than a simple “wacko Texas Christians run amok” narrative. It appears that Perry ran into a combination of opposition from social conservatives who think the vaccine promotes sex, people who thought he was bought off by Merck, and legislators who thought he was overstepping his authority…

    If the trackbacks are off, I wish Scienceblogs would remove the %$#* link.

  8. #8 Diego
    May 12, 2007


    Basically, that’s correct. There is always a certain degree of uncertainty in any phylogenetic reconstruction. A bootstrap analysis is one way to explore the sensitivity of the data. Multiple replicates are performed with subsamples and the program records how many times certain clades appear in the replicates. The important thing to remember with bootstrap scores is that there’s a strong demarcation between strong support and weak support– anything over 50% is really good and anything below 50% is pretty weak.

  9. #9 Jawon Lee
    May 13, 2007

    It’s a major menace: the American Cancer Society estimates that 17 percent of all cancer cases–more than 1.8 million a year.


  10. #10 Carl Zimmer
    May 14, 2007

    Jawon–Thanks. Text fixed.

  11. #11 luca
    May 15, 2007

    Once again, excellent, excellent piece, Carl. And the comment are very informative too. This blog really makes my day!

    Manual Trackback

  12. #12 Keith Robison
    May 15, 2007

    Your prevalence statistic seems to still be off by a two orders of magnitude; according to the ACS it is (in the U.S.) 11,150 new cases of cervical cancer per year in the U.S. Even if you meant worldwide, the incidence given by Wikipedia wouldn’t get you over a million cases per year.

    It is a menace though, particularly in developing countries. WHO puts it at fifth in rank order of cancer mortality in woment.

New comments have been disabled.