This is an off-topic post on bad science and failures of the academic system. It concerns the depressing news item I read today about the University of Wollongong (not one of Australia’s premier research institutions, but what I had thought to be a legitimate university none-the-less) awarding a PhD to rabid anti-vaxxer, Judy Wilyman, for her conspiracy theory thesis (available here) entitled, “A critical analysis of the Australian government’s rationale for its vaccination policy”.
Since I don’t have time at present to digest a 390 page thesis I thought I would instead take a look, as illustrative example, at Judy’s sole first-author, peer-reviewed publication listed as ‘In support of this thesis’; namely, Wilyman J. 2013. HPV Vaccination programs have not been shown to be cost-effective in countries with comprehensive Pap screening and surgery. Infectious Agents and Cancer [IA&C hereafter] 8:21 (June): pp 1-8. With an impact factor of just 2.36 one might suggest that the Infectious Agents and Cancer journal would struggle to attract the best quality papers, and likewise the services of the most competent referees—but that would be bitchy and we’re not bitchy here at Another Astrostatistics Blog :-). All prejudices about obscure journals aside one might still imagine the papers in IA&C at least parse on face value as sensibly reasoned; well, not so for Wilyman’s paper. Indeed without going further than page 2 I was stuck by two egregious errors that require no scientific training to see them for what they are.
The first error, for which a undergraduate would be severely marked down even in a first-year essay, lies in the following:
“By 2002 scientists were proposing that human papillomavirus (HPV) Type 16 and 18 was the ‘first ever identified necessary cause of human cancer’ . This suggests that cervical cancer does not and will not develop in the absence of HPV DNA . The claim hasn’t been sustained because some investigators observe that HPV infection cannot be found in every patient with cervical cancer .”
where the references Wilyman cites are:
 Haverkos H: Multifactorial etiology of cervical cancer: a hypothesis. MedGenMed 2005, 7(4):57.
 Bosch FX, Lorincz A, Munoz N, Meijer CJL, Shah KV: The causal relation between human papillomavirus and cervical cancer. J Clin Pathol 2002, 55(4):244–265.
If you’re familiar with the use of citations as a means to give authority to your arguments, then you might presume that Haverkos (2005) is an authoritative study presenting some new data, or a new analysis of existing data, that would speak at least in part to the question of whether HPV is indeed a necessary cause of human cervical cancer. Well, you’d be wrong, because Haverkos (2005) is basically a think-piece by a medical doctor working in disease epidemiology who hypothesises that “cervical cancers arise from an interaction between oncogenic viruses and cervical tar exposures” and gives a more-or-less anecdotal historical overview of research regarding cervical tar exposure. His only reference to the question at hand is to observe that “HPV can be found in 90%-95% of patients with cervical cancer worldwide, most frequently HPV-16 (50%), HPV-18 (12%), HPV-45 (8%), and HPV-31 (5%) ” and later muse (with no answer), “How should medical epistemologists interpret findings demonstrating that HPV can be found in roughly 95% of cervical cancer patients but not in the remaining 5%?“. Reference  turns out to be one of Bosch’s earlier papers from 1995 to which his 2002 research represents a significant update (thanks in part to technological advances yielding improvements in the sensitivity of diagnostic tests for HPV genotypes)!
The second error on page 2 of Wilyman’s paper, which is even more transparently false, lies in the following:
“The International Agency for Research on Cancer (IARC) working group also acknowledged in 2005 that there are cofactors that are associated with HPV infection and cancer development . In other words, an HPV infection does not progress to cancer without the co-factors being present.”
with reference to a single page summary report  Cogliano V, Baan R, Straif K, Grosse Y, Secretan B: Carcinogenicity of human papilloma viruses. Lancet Oncol 2005, 6(4):204 rather than the 690 page monograph that contains the detailed findings and interpretation of results by the expert panel. What’s wrong here is that she’s entirely mischaracterised the meaning of the term ‘co-factor’: a co-factor is a variable (e.g. smoker status) other than the one of primary interest (e.g. presence of HPV status) which a case control or retrospective cohort study finds to have a secondary influence on the likelihood of the outcome under study (e.g. cervical cancer; or pre-cancerous lesions). That is, co-factors affect the likelihood of the outcome but are not necessary conditions for it to occur. Take the report in  in which the only reference to co-factors is: “Important cofactors in cancer associated with human papillomavirus infection include coinfection with chlamydia or HIV, smoking, and parity (eg, >3 children); the role of hormonal contraceptives and nutrients is unclear.” Does that sound like it implies “an HPV infection does not progress to cancer without the co-factors being present”? How would that be plausible: do only mothers of three or more children get cervical cancer, or do only smokers get it? Obviously not. For a cogent analysis of the role of, and evidence base for, cofactors in cervical cancer one could do no better than turn to page 278 of the 690 page monograph linked to above.
Anyway, this is as far as I got through Wilyman 2013 before deciding it would be a phenomenal waste of my time to continue debunking what is obviously a deeply flawed piece of ‘scholarship’. I believe that time will show the award of this PhD by the University of Wollongong to be a black mark on their record. The hitherto refusal of UW to release the names of Judy’s thesis examiners nor to enter into any meaningful debate as to the merits of this thesis is to their shame and should be a matter of concern for all UW staff, students, and alumni. Moreover, I think there are serious questions as to the competence of the PhD supervisor who oversaw this ‘research’, Professor Brian Martin; remember, the opportunity cost of employing this guy is that a legitimate researcher is not being employed in his place.