H/T Jredheadgirl for noting reports of a new study showing human papilloma virus (HPV) present in cases of oropharyngeal squamous cell carcinoma (OPSCC) over the past 25 years.
October 5, 2011 — The sharp rise in oropharyngeal cancer associated with human papillomavirus (HPV) and its association with oral sex have been highlighted again, and this time has gained widespread media attention.
This new epidemic of HPV-positive oropharyngeal cancer has been brewing for a while now, as previously reported by Medscape Medical News…
The researchers, headed by Maura Gillison MD, PhD, professor of medicine and Jeg Coughlin Chair of Cancer Research at The Ohio State University Comprehensive Cancer Center in Columbus, report that there has been a dramatic increase in the incidence of oropharyngeal cancer since 1984.
A growing majority of new cases (70%) are associated with HPV, and the vast majority of patients are male.
My interest in HPV has risen over the past couple of years since learning that HPV-16 has been found present not only in cases of cervical cancer (which now has a vaccination jab available), but also in cases of genital and anal cancer, and even in mouth and throat and lung cancer. It’s had me wondering whether the cancer epidemic of the past century may have actually been caused by a virus, rather than by smoking or drinking.
But the new study (which is freely available online), which is based on tissue samples held since 1984, suggests that this is a new form of cancer, with only about 16% of the earliest samples testing positive for HPV as against 71% of the later samples testing positive. Which kind of suggests that HPV wasn’t very prevalent 25 years ago, but is far more prevalent now, and is therefore a mounting new epidemic:
If the current trends continue, HPV-related oropharyngeal cancer will become the major form of head and neck cancer by 2020, and will become the leading HPV-associated cancer in the United States, surpassing cervical cancer, Dr. Gillison said in a statement.
Well, on the face of it, it would certainly look like that. But, on closer inspection, a number of questions arise.
For a start, there are surprisingly few samples in this study, as the accompanying editorial points out.
During the two decades examined, there were approximately 200,000 patients with OPSCC nationwide. … this reconstruction of the history of OPSCC in the United States is based on only 271 patients.
Which works out at about 13 patients for each of the 20 or so years the study covered. That seems to be rather too few to get a definitive fix on whether this is a growing epidemic or not.
Furthermore, they say in the study that it’s a bit difficult to test old samples for HPV.
Because DNA and RNA from formalin-ﬁxed, parafﬁn-embedded tumor specimens are known to degrade over time, laboratory assays were designed to account for this. Specimens were classiﬁed as evaluable or unevaluable for DNA ampliﬁcation by polymerase chain reaction (PCR) by using a real-time TaqMan assay (Applied Biosystems) that ampliﬁed a 58 base-pair (bp) region of a control gene, human endogenous retrovirus-3 (ERV-3)…
We anticipated loss in assay sensitivity because of the age of specimens beyond the exclusion of unevaluable samples. Therefore, HPV prevalence in 69 cervical cancers was used to estimate sensitivity of each laboratory assay across calendar periods, assuming all cervical cancers were HPV-positive.
As I understand this, they tested the samples for common-or-garden ERV-3, and if they could produce a result for them, they included them in the study. And then they adjusted the results they got using similar results from cervical cancer samples which were assumed to all be HPV positive.
But ultimately, what this really means is that they couldn’t actually measure how much HPV-16 there was in the oldest samples, but they had to make an intelligent guess based on how much HPV-16 they found in similar aged cervical cancer samples. So if, say, only 75% of 20 year old, 100% HPV positive cervical samples actually tested positive, you’d get a sampling fraction of 75/100 or 0.75, and if 22% of your 20 year old oropharyngeal samples tested positive, you’d multiply this by a weighting 1/0.75 or 1.33 to give a ‘corrected’ value of 29%. And assuming that the cervical samples (all assumed to have tested positive for HPV) had similarly degraded with age, then you’d find that the weighting would increase with age, rising from 1.0 to higher values with increasing age.
Given that, it’s a bit strange to find that only 0.4% of the oldest oropharyngeal samples tested positive in the results they obtained. Yet the ‘corrected’ values (see right) appear in some cases to be exactly the same as the observed values (e.g. 1984). i.e. the weighting is 1.0.
I may be misreading this graph, of course. It says ‘Positive %’ on the y-axis, but it probably actually means ‘positive fraction‘, and 0.4 isn’t 0.4% but is actually 40%. But why such small corrections? And why are the youngest sample values corrected upwards more than the oldest? A closer examination of the paper revealed the explanation.
…we estimated sampling fractions for inclusion into our study (speciﬁc to age group, sex, race, registry, and calendar year [2-year groups]). The inverse of these sampling probabilities were used as weights to reweight the observed HPV prevalence in tested OPSCCs to all patients within Hawaii, Iowa, and LA (1988 to 2004; periods preceding 1988 were excluded because of sparse data).
So it would appear that the 1984 figure was not reweighted, because of sparse data. And that is why the ‘corrected’ value is the same as the observed value.
So what might the graph have looked like if a correctly reweighted value was used? The weighting is hardly likely to be 1.0. So it would seem reasonable to use some value similar to the other 3 weightings. I estimate these weightings to have been 2.34 for 1990, 1.23 for 1995, and 1.33 for 2000. If it’s assumed that the 1984 weighting would not be smaller than the 1990 weighting, then the reweighted 1984 result would be about 0.38 or 38% at minimum (1). But if the mean increase in weighting between periods (1.11 – 0.1)/2, or 0.5, was also added, then the 1985 weighting would be 2.84, and the reweighted percentage would be 43% (2). And if the overall trend of reweightings (the rate of change of reweighting) was estimated, the reweighted percentage might be as high as 54% (3). And the replotted graph would look like the graph at right.
And instead of the graph showing a growing HPV epidemic with 15% incidence in 1984 rising to a 75% incidence in 2004, it would show a much higher incidence – in the range 38% to 54% – in 1984. Which would suggest that HPV infection was already endemic in 1984.
And this dramatically different conclusion would be the result of simply reweighting a single unweighted number.
However, I’ve only looked at one of the graphs. Of the other three, two of them seem to have had their 1984 figures already mysteriously reweighted to produce values of 40-50%. Nevertheless the authors report a rise in the incidence of HPV-positive oropharyngeal cancer from 16% to 71% over 20 years.
And they may well be right, and this is an entirely new cancer epidemic to complement the prior booze-and-cigs-driven epidemic, which must necessarily be now fading out as fewer and fewer people smoke. This is disease as relay race, and as one epidemic fades out, it passes the baton onto another epidemic! It lends further credence to this earlier, well-established dogma doctrine, but finds a new and rapidly growing epidemic – research into which will need funding, needless to say. And indeed they don’t say any such thing.
But on the other hand, with a slight correction of one of the figures along the lines I have suggested, it would seem to emerge that, far from this being a new epidemic, it shows every sign of being a well-established epidemic, and one which has been under way for at least 25 years. If so, it may well be that the current orthodoxy of booze-and-cigs causation has been horrendously mistaken all along, and a viral cause of cancer has been overlooked for at least 25 years, and quite possibly even 50 or 75 years. Which would be unthinkable. And so nobody is thinking it. Except me.
Perhaps the most percipient observation is to be found in the editorial:
We do not know where we are in the course of the epidemic of oral HPV leading to OPSCC…
Yup. Could be the start of a new epidemic, or the continuation of a very old one. We have no idea.
But what do I know? I’ve only been looking critically at this paper for a few hours today.