HPV Vaccines Not Proven To Be Unsafe, Or, In praise of the Wise Sages Goldacre and Harford...

We were passed the following link and asked for thoughts. We have a child who is of age to join the NHS HPV Vaccination programme:

http://articles.mercola.com/sites/articles/archive/2013/07/16/hpv-vaccine-effectiveness.aspx

Summary of Objection

If we take what's asserted in this link, and combine it with the important numbers it doesn't provide to give total population information, the numbers given in this link suggest your child might be 10 times safer after taking the vaccine than she would be if she didn't. 

This evidence is in direct conflict with the conclusion the link itself wishes to portray of a drug of questionable efficacy and safety.

Background to the Objection

Thanks to years of reading popular science writings, and in particular thanks to Ben Goldacre's Bad Science and Tim Harford's work on the Radio 4 show More or Less, I am no longer able to read such links without at least a cursorial attempt to work out what's not being said.

I could tackle the material in a visceral, emotional way, based on what we know of the context of the link and it's material, especially in reaction to the source country's known political and anti-vaccination movements. But in a poor imitation of the two Sages I venerate above, let's try some Science and Statistics instead...

Analysis of the article's assertions

The core arguments made to support the (simplified) conclusion "Gardasil is Bad, M'Kay?" in the article are:

• 30,000 adverse effects reported for takers of the vaccine in about 7 years, of which:
• 1600 were serious or life-threatening, 
• 136 people who took the vaccine subsequently died.

These are scary numbers. BUT we're missing vital context:

• How many people were vaccinated with the drug?
• What is the chance of someone dying during the appropriate period anyway?

Because the source link came from the US, I'm going to assume for the rest of this work that all figures will apply to the US only. I'm not convinced my source materials (I'll embed links) are entirely consistent in this, but I'm fairly happy that my numbers will be in the right order-of-magnitude, and I'll try and prove that this is Good Enough to come to some supportable conclusions.

Let's start by answering the "background radiation" question: Given this Vaccine is aimed at girls between about 12 and 20, what's the chance of a girl of this age-group dying of any cause?

• This link suggests a "background rate" of 11 deaths per 100,000 for under-20s in the US. But that's boys and girls. Let's say the rate is about 5 girls per 100,000, per year.
• The US census estimate an average teenage population of 42 Million (boys and girls) between the ages of 10 to 20 

We can then multiply the two together:

•  (21,000,000 / 100,000) * 5 = about 2,160 female deaths per year.

And, just to be clear, we'll call this the "background rate" - it's deaths from all causes, averaged out across the whole US (interestingly, go look it up, most of these will be from car accidents even though they massively favour testosterone=infused boys. Why aren't we making a fuss about Our Precious Children driving these lethal steel boxes? But I digress).

This doesn't make the 136 deaths linked to Gardasil over 7 years doesn't sound too bad after all, when compared to the expected 15,000 females killed from all causes over that time-span. But let's be rigorous, as we don't yet know the percentage female population given the drug. We can put a lower-boundary on this, though, in order to decide when to start worrying:

• 136/7 = ~20 deaths per year linked to the vaccine
• 20/2160 = 0.9% of annual deaths

...So, even without tackling the central problem with the statement "20 people per year taking Gardasil die" (which is, to be clear,  that no-one's attempting to prove whether those deaths have anything to do with the drug), we can be fairly confident that if we discover that Gardasil is given to more than about 1% of the female target population, the death rate is unaffected by the drug. So we need more numbers.

 Wikipedia says 57 million doses were given in roughly the same timescale as the adverse reports, which gives a reasonable upper bound on the number of girls affected. We need to add to this that the standard drug regime has been 3 administered doses per person over a 12-24 month period, so we can estimate that:

• 57,000,000 / 7 / 3   = ~2,700,000 girls given the drug per year

I should be clear that this number is not geographically-bounded, and it's not absolutely clear whether those doses were all administered, vs just manufactured and/or distributed. But it's what we've got to work on. 

At any given time, the female target population aged 10-20 in the US is about 20,000,000 so we can estimate that about 13% of the female population has been being (fully) vaccinated (note: the Wikipedia link referenced above states that about 33% of girls in the US are currently receiving at least 1 dose, so we've immediately got a conflict of information). 

All of a sudden, this turns the study on it's head. Let's do the sums:

• We know the chance of an average girl dying in the US, per year, is about 1 in 20,000

We can now calculate the chance of a girl taking the vaccine dying as:

• 1 / (20 / 2,700,000) = about 1 in 135,000

Or, in other words, according to the material given in the link itself, my child is around 10 times less likely to die after being given the vaccine than she would otherwise.

But my analysis is flawed, because without evidence I've asserted all doses administered to US Girls.  Let's assume only half the doses went into the arms of US girls. This doesn't change matters much: it only means my beautiful daughter is 5 times less likely to die with the vaccine than not. If only 10% of the doses were administered to US girls, then without anything else changing we can still say the drug's not affected the chance of your child being killed this year.

And that in itself is a very simplistic view, because it's based on an assumption that 90% of the doses of the drug went either unused or into the arms of girls from other parts of the world who we've implicitly assumed have a 0% death rate, which is patently false. And we've already got information that suggests the number of vaccinated kids in the US is much higher than my 13% figure so the error could very well go the other way making my kid even safer than I've calculated...

All of which is to say that my analysis is, I am happy to confirm, wrong. But it's probably not more than an order-of-magnitude wrong, and even if that's the case my central refutation of the source material stands.

Conclusions

At this point, I've done enough Science and Statistics to stop worrying about what the article's trying to tell me, because it's clearly not being backed up by sufficient facts to make it's argument robust. I probably ought to tackle the "incapacity rate" number (1,600 over 7 years) to strengthen my case, but I just can't be bothered: The central thesis of the link is that I should be scared of giving my daughter this vaccine because it might do her harm, but the "facts" it quotes to support that thesis suggest the exact opposite. I see no reason to believe it's assertions over serious harm are just as flawed.

But let me be absolutely clear: I'm not suggesting that the HPV Vaccines are some sort of wonder-drug that magically improve your kid's chances of survival, and I'm not going to make any assertions of the efficacy or safety of the drugs concerned. All I've sought to do is to tackle the fundamental gaps present in the information presented by the source link and see if they can be supported by available evidence, and my conclusion is that they can't, and that the authority of the assertions made and conclusions drawn by that article are therefore, at best, questionable.

All I'm trying to suggest is that you should take the claims made in the source link with a VERY large pinch of salt and back them up with a decent review of other materials when making your decisions. 

First-world problems

Here I sit, a privileged member of the middle classes in one of the most privileged countries on the planet. 99% of the global population have serious life-affecting problems (you know: hunger, poverty, disease, shelter) which I fervently hope will never cross my effete brow.

OK, with all that out of the way:

ADSL. Copper Wiring, in particular, has been vexing me somewhat recently.

I've had, for the last 4 years, a very reliable connection that over time has increased in speed from 2MBit/S up to about 10MBit/S briefly. For most of that time I've been nicely and reliably connected at about 6Mbit/S, or in english numbers about 600KByte/Sec peak.

Now it's clear that there's only so far you can go with old copper wires originally put in to support a fairly basic-quality voice link. ADSL-Max (or whatever it's branded as; that which delivers ADSL2+ at least) is, from my experience, as fast as we're going on this infrastructure.

I'm about 3KM in cable distance from the exchange. I get about 40db attenuation, and have consistently gotten 1/2 to 2/3rds of what I've been promised (all those "up to XXX" ads - very disingenuous). Frankly, as long as it's stable and reliable speed has been a second priority.

My exchange got enabled for ADSL2+ last summer. I saw a re-sync and speed up - temporarily - from about 6Mbit to about 8Mbit down (and occasionally from 100Kbit to 500Kbit up). Then my ISP did their bit of the upgrade in April. Speed (which had dropped back to 6Mbit) went back up to 8Mbit (occasionally 9) and upload went up to 1Mbit then back to 500Kbit. All well and good.

Then the bad weather came at the end of May. Oh, and I got a new router - must declare my own actions in this I guess - in an effort to get faster speeds (a DG834GT apparently has a better chipset for higher-noise environments. Apparently). Worked great for exactly 1 day, then I went through 2 weeks of connection hell. Up, down, re-syncs, speed ups, slow downs and for considerable periods of time a line that went up and down on a 2-minute cycle. Ever tried web browsing when you've got a connection for 30 seconds in every 2 minutes?

I can't say what solved it - calling my ISP and having them "monitor the connection" certainly made a big difference, and it's been up for 8 days now. And what's worse is that I can't say what caused it either - notionally it would be plugging in a new router, but the problems persisted when I reverted to the old router. In fact they got worse.

The biggest correlation was a period of extremely heavy rain. Is it possible, therefore, that my internet connection depends on how damp the bits of copper connecting me to the world get?

Roll on the Fibre upgrade, that's all I can say...

Tokenism

Stories like this one on the BBC website set me off something chronic. Let us not dwell on the particular details in this case, but focus on the wider concerns.

We live in a very, very shallow world. Complex issues must be brought down to black & white, right & wrong, in an instant to capture the fleeting attention of the fickle masses before moving on to the next story.

But we're adults, we're intelligent beings. We can observe the world and what goes on and understand that most things don't have simple yes/no, right/wrong positions. There's a bit of truth in both sides of a position; positive things to be said even for the views of those most unlike us.

And yet we've taken this yes/no 5-minute-debate-and-the-problem's-sorted, boiled-down-to-a-10-second-soundbite presentation of all the world's various troubles and it's become institutionalised. If the problem's only presented as a shallow face-value issue, then we only need to take shallow face-value steps to resolve it, right?

Bankers take risks and get big bonuses? Oh, legislate the bonuses away and the problem's solved. Never mind tackling the risk-taking culture that underlies the industry, or the fact that the industry is able to take such rewards without contributing much back to the society that enables it in the first place, that's hard and requires thought and effort. The news picks out a 1-in-a-million brutality and blurts it at us for days? Well, all adults are suspect and we'll legislate their access to minors at all times. Never mind that most bad things that happen to kids are caused by people they know and trust - that would require a functional society to support families in trouble and provide guidance and assistance instead of blame and hate.

And then we get to the fluff.  Population Sub-Group X are under-represented in Power Structure Y.  So we'll just mandate that there must be "n" members of X promoted / appointed to Y. Never mind whether there are better candidates with more appropriate skills & experience. Never mind whether Group X has been empowered and given access to the resources necessary to make a good fist of being in Y.

Never mind what the longer term consequences are for a society which is learning that you don't need skills, or experience, or intelligence to solve complex problems. You just need a transiently plausible Issue, a loud voice, and access to Media to sort it out.



I really, really should know better than to let this get to me. Aren't I old enough and cynical enough to just let it pass me by and get on with my taudry life?