More on cancer and causes

[lnr]

Ok, so it's probably just bad reporting, but this BBC article, states (for example):

Bowel cancer - Of around 42,000 new cases, being overweight or obese causes 4,800 - I make that 11%. The figures for other cancers are: Kidney (22.4% caused by excess weight), Liver (22%), Ovarian (6.5%!)

But from their own figures there are 14.9 million obese adults in the UK, which is 22.6% of the UK population. Once you add all the overweight people too - you could be talking 63% of the population or more.

I really *really* want to know how they are determining that it was someone's weight which caused their cancer, because surely they don't just mean all the other 37200 people were not overweight? There is definitely *something* missing from this equation.

As others have pointed out there are all sorts of reasons why this campaign is misguided. If it's about population level health it shouldn't be shaming people on billboards. Smoking is something you do, whereas being overweight is something you *are*. Weight loss is not a solved problem. It's horrible for those whose loved ones have cancer - even if they are thin. And it shames those who are overweight and have cancer *even if* their cancer is not linked to obesity (eg leukaemia).

They've even had advice on how to target obesity campaigns - and this is *not* it. I'm utterly furious with them - and with the press's reporting of it.

Comments

[rmc28]

I suspect that a critical mass of the people involved just don't see a problem with fat-shaming people, no matter how carefully one explains it.

(And I have sympathy that stopping smoking is hard - I've seen relatives really struggle with it. But at least they *could* go cold-turkey or switch to e-cigs. I cannot address my weight anything like as straightforwardly.)

[chickenfeet]

I don't think an individual cancer case can ever have a specific cause attached to it. There are risk factors of course but even if someone smoked three packs of cigarettes a day and then got lung cancer one couldn't be sure that that was the cause.

[lnr]

I'm reading through papers and still trying to work out how they come up with their PAFs (population-attributable fractions). I *think* it must be something like this:

New cases of a particular cancer
25% in thin people
75% in fat people

but fat people only make up 63% of the population, so 12% of the cases are linked to being fat? But you don't know which ones of those 75% of fat people are the ones who would have got the cancer anyway. (Figures made up for ease of maths). It would really help if they would spell it out! And you still don't know if the fat caused the cancer, vice versa, or both were caused by something else.

[lnr]

Aha, no, it looks like I've very much got the wrong stick there (not just the wrong end).

This paper is looking pretty interesting, and appears to have the same figures being used in the press.

https://www.nature.com/articles/s41416-018-0029-6

One section I found fascinating:

"Operationalising overweight and obesity prevalence, alcohol consumption, and breastfeeding prevalence as categorical rather than continuous variables is likely to have overestimated the PAFs for these risk factors. However, available exposure prevalence and relative risk data were overwhelmingly categorical, so converting to continuous data would have introduced further uncertainty. RRs comparing categories of people will overestimate the effect for those very near to the category boundary and underestimate the effect for those furthest away from it. If the exposure prevalence distribution is left-skewed (more people near the boundary with optimum exposure), as is the case for overweight and obesity, then the PAF is likely to be an overestimate. This is less of a concern if the within-category distribution is similar in the RR source and the exposure prevalence. However, this information is rarely reported and so the risk of PAF overestimation on this basis cannot be quantified. More accurate PAFs could be calculated if relative risks and exposure prevalence were reported continuously rather than categorically."

(RR is relative risk FWIW). It explains why CR_UK would concentrate on obesity rather than (um) overweightness, given those who are only overweight are likely to be caught by this over-estimation.

It looks like you need to look at relative risk (those in total population getting cancer, versus those not exposed to the factor getting cancer) over a long period of time, in order to calculate the PAF, and it can only ever be an estimate. And I note they have said they only included relative risks if there is considered to be "sufficient" or "convincing" evidence that the risk factor is actually causal. It looks like the risk factors they've used date from various studies/data sources, between 2012 and 2017 (Supplementary material A).

[hilarita]

Also, the big things that drove down smoking were population-level things, e.g. raising taxes sky high, banning smoking almost everywhere except the street, no cigarette advertising, cigarettes hidden away, plain packaging.
Almost none of that, except the packaging reforms, relied on individual smokers changing their lifestyle. Their lifestyle was changed for them. They couldn't light up at work, at the station, at the pub, on the bus or train. At that point, they could well decide to jack it in.
So they should bloody well be attacking the population-level issues, not the people. It's not really a great trade if you reduce your risk of cancer, but instead get an eating disorder.

The other big problem is that you can live perfectly well without smoking. You can't live perfectly well without eating. This makes CRUK's approach deeply infuriating on so many levels.

[nou]

You and commenters on this and your other post have pointed out the problems with using population statistics to claim that individuals need to change. Another interesting thing to ask is why they aren’t using evidence that could actually support their claim. Why not put up statistics showing that X% of fat people who didn’t lose weight got cancer, but only Z% of fat people who did lose weight got the same cancer? (Z<X obv.)

These studies already exist for things like alcohol abuse — you can show that people who’ve stopped abusing alcohol have a lower risk of certain cancers than otherwise-equivalent people who’ve continued to abuse alcohol. (An example chosen solely because I remember it from copyediting it for the authors.) But there doesn’t seem to be any such evidence for fat people who’ve stopped being fat. If they want to say losing weight reduces your risk of cancer, why aren’t they trying to prove it?

[lnr]

*applause* - I'm pretty sure it's now reasonably well known that metabolically people who have lost weight are not the same as people who have never been overweight, it's one of the reasons it's so hard to keep weight off if you do lose weight. I've not seen anything to show what that does in terms of cancer risk indeed.

[naath]

The population interventions also put people taking up smoking; it's a whole lot easier to never start than to stop.

[naath]

Because they do not have a sufficiently large population of formerly fat people to use for this experiment /cynic. (N needs to be huge, cancer is not that common fortunately)

[lnr]

Yeah I notice the BBC is linking calorie counting apps and eating disorders today too!

[lnr]

That may be true! I've asked them if they have any such research - smoking/non-smoking/ex-smoking is one that comes to mind that has been studied a lot. And contraceptive pill users/former users/never too.

[kaberett]

(HPV? It's not the *sole* cause of cervical cancer -- e.g. risk factors are higher for people who also smoke -- but 99+% of cervical cancer is associated with HPV.)

[chickenfeet]

That's a good point. HPV is weird. It also has an impact on head and neck tumours.

[chickenfeet]

It is important to look at data over long periods of time as there are often lags. For example changes in the incidence of lung cancer tend to lag changes in the incidence of smoking by about 20 years (at least that's what it looks like from fifty years or so of data in Ontarioo).

[lnr]

The trouble is of course that weight changes over time. I've got about 18 years of data on my weight I think, and it's varied from a starting point of 95kg or so between 68kg (min) and 112kg (max) in that time - currently just under 105. Or in BMI terms between about 24 and 38.

Suddenly I'm actually really impressed that I'm only 10kg heavier than I was nearly 20 years ago.

[bens_dad]

Last time I heard, no one had yet managed to show that overweight/obese people reduce their rate of dying by losing weight, even allowing for ill-health being a significant reason for losing weight.
If this still so, there may not be much interest in researching whether weight-loss does reduce cancer risk.

I am probably being cynical but their adverts (eg on buses*) make me think that CRUK is not interested in reducing the incidence of cancer, but aim to find a cure.
* Adverts like "We *will* cure cancer, (small print) but we will do it faster if you give us more money".

[zenithed]

Yeah, this kind of screams residual confounding (i.e. associated risk factors that you can't control for) to me, especially given it involves a meta-analysis so they might not even have access to the original data - though tbf I've only had time to skim the paper.

I think the main take-away is the one you identify though - what are people meant to do with this? Even if the results were valid, we already know that just giving health warnings doesn't massively change behaviour for smoking or alcohol, let alone for something as complex as weight loss. Maaaybe as a way of driving structural change this could be useful, but as a public-facing campaign it just seems to risk accentuating stigma.

[lnr]

I think you're probably right, I tried to challenge someone's fat shaming on twitter yesterday, and as a result I'm now seeing lots of people "liking" the tweets that replied to me justifying it because obesity is more dangerous than riding a bike without a helmet.

[lnr]

The supplementary data has links to where they got all their figures from, but I haven't gone and looked at the sources either.

Rather good open letter/petition here:

https://twitter.com/laurathomasphd/status/1147023926721175553