Thursday, July 15, 2010

Polyp Prevention Trial and homocysteine

I don't know where to start with this. Let's begin with one of my all time favourite studies, the Polyp Prevention Trial, PPT. For those of us who have forgotten, this quote from the abstract sums up PPT:

"This study failed to show any effect of a low-fat, high-fiber, high-fruit and -vegetable eating pattern on adenoma recurrence even with 8 years of follow-up"

Now, would anyone like to guess how much the PPT cost? I've no idea, but I guess it cost enough that you are not going to accept a result like that after spending all of those tax dollars (thank goodness it wasn't pounds sterling, it's bad enough funding the FSA on this side of the Pond). So there is now an on-going avalanche of subgroup analysis papers flooding out from PPT, many of which are utterly hysterical. Some actually show, by accident, the injury done to many of the people in the intervention arm of the study. From these new papers you would NEVER guess the main trial primary end point was an utter destruction of the hypothesis that flatulence is good for you.

Let's start with that lovely sub group who ate all the dry beans. Upping your beans by 370% appeared to drop your risk for adenoma recurrence to an OR of 0.35 in the intervention arm. This is music to the fibre-ophile ear. The effect is massive. It really should show in observational studies if it is real. It doesn't.

Sub group analysis of the control group, observational only, suggests that contributing to global warming is NOT justified for colorectal cancer prevention. There is no excuse for flatulence. I dunno. Maybe the intervention trial is correct. But don't hold your breath, just your nose.

That last sub group analysis is particularly interesting once you get hold of the full text (many thanks Anna). It only looks at the control arm, so it's observational in nature. They blood sampled folks at the start of the study then counted polyps after four years on whatever shade of the SAD happened to be chosen by people who have had a polyp removed but no diet advice given (ie the lucky ones).

What is topical to me at the moment is that homocysteine (HCy) correlation with polyp recurrence. HCy at enrolment is a very good predictor of polyp recurrence in this group of untutored polyp formers over the next 4 years.

The suggestion, by the authors, is that reducing HCy will reduce adenoma recurrence. Really, someone should test this hypothesis. Maybe a trial increasing fruit, vegetables and fibre should be done to see if these miracle foods will reduce adenoma formation over a four or even eight year period?

Oh, it's been done has it? Ah yes, that's where this data trawl has come from! I remember, yes, that's it, eight years of fruit and fibre does bugger all to prevent polyp recurrence. Silly of me to forget. There was an intervention arm to PPT!

Remember that and re read the very carefully worded abstract... Would you have guessed?

Before I get on to a dietary intervention which might lower HCy (if you care, all this is observational), let's have a look at another subgroup analysis of the PPT data.

I now want to look at what happened in the intervention arm. We can tease out a sub group, the folks who really complied with the intervention, and compare them with the people in the intervention arm who complied poorly.

That's this subgroup analysis.

A total of 210 "super compliers" were noted and compared to controls. I can't really tell if the controls were on the SAD or were those less compliant members of the intervention arm, probably the latter. The odds ratio for adenoma recurrence of supercompliance was 0.65 and statistically significant.

So 210 people out of the 1,905 in the intervention arm benefited. That means that 1,695 people didn't benefit from four years of advice.

But hold on a minute. When the initial study compared intervention with non intervention there was no benefit from intervention compared to the SAD. So if 210 people did better than average, then 1,695 people MUST HAVE DONE WORSE than average. That is, they did worse than those on the SAD. You would have expected a 50:50 split, obviously half should do better and half should do worse than average!

EDIT: Of course if they compared 210 super compliers to the 1,695 injured people they possibly got p < 0.05 because these 1,695 did particularly badly. I wonder if p would have been < 0.05 if they had compared the 210 to the whole SAD group? Probably not would be my guess. Or they would have done it! Me, cynical? Surely not... END EDIT

We can summarise: 1,695 people were injured by being in the intervention arm of the PPT.

I would also point out that "super compliers" are super compliers. These are the sort of people who meticulously take their placebo tablets in clinical trials and consistently do better than those who are lackadaisical with consuming their placebo tablets. Mike Eades has blogged about it nicely here.

So did the 210 super compliers benefit from the diet or were they just the sort of people who comply with authority and automatically do well despite their doctor's best efforts? There was no scope for an adherer effect in the SAD non-intervention arm because they were left alone with no religious group therapy support. Oops, I mean dietary advice...

So was everyone injured in the intervention group but a few had their injury offset by the warm glow of study compliance?

Don't hold your breath waiting to see THAT get discussed in an NIH funded paper. We're looking to justify the cash spent on a superb study which has made no friends among people with control over the purse strings...

Anyhow, back to HCy. We all know from nauseous observational stuff that PPT intervention should have dropped HCy (I've not seen this analysis anywhere yet but I guess it did) but it still didn't drop adenoma recurrence and it probably worsened it in that vast majority of the participants who weren't super compliers.

What sort of HCy levels are present in the SAD arm of PPT and how do they stratify against adenoma recurrence risk? This is from the sub analysis of the control arm I wrote about earlier:

To summarise the right hand column, folks with no relapse had median HCy around 12.5micromol/l. Those groups with the worst recurrences generally had median HCy over 14micromol/l. Low HCy is "associated", observationally, with lower adenoma relapses.

Of course those who ate the most plants, fibre and general pre-poop had the lowest HCy... Really begs why the trial totally flopped. Something to do with the difference between observational studies and intervention studies I think! There may be no funding for basic science (ask my wife!) but there seems to be plenty for this sort of dogma support BS!

However, within the dogma that HCy is controlled by eating leaves, how bad must it get if you eat almost no plants, with the emphasis on peeled potatoes when you do indulge in vegi-cide, as little fibre as possible and as much animal fat as practical, say >70% of calories?

Well the HCy comes out, when eating to the Optimal Diet, with a median of about 10.7micromol/l. And relatively few plants were murdered (blame Elizabeth) in the process.

No leaves, no fibre, lots of saturated fat. The occasional spud. Please remember that the data from the OD is observational in nature too. You have to guess what the HCy level was pre OD.

Counter intuitive isn't it? But science does that sometimes. Just look at the money spent on supporting the long dead hypothesis that eating leaves is in some way beneficial for humans. There's a powerful need for self justification of the pig's-ear that PPT turned out to be. Depending on your point of view of course!



Dr. B G said...


Naughty. You are making me snort the coconut cream in 100% saturated fat coffee out of my nose.


Peter said...

Hah, success then!



Asclepius said...

"vegi-cide" - just added that to my lexicon.


Jenny said...

Homocysteine appears to be an interesting marker, but usng supplements to change its value has not lowered risk in any study I've seen.

It's a shame they didn't measure fasting insulin rather than C-peptide. C-peptide is a uselss measurement as the same sample tested twice in the same lab gives significantly different results. It is useful ONLY for determining if an insulin user is still making some insulin in the pancreas. Anyone else should have insulin measured.

My guess is that higher fasting insulin would correlate with a slightly higher growth of cancerous polyps as it does with other cancers.

Peter said...

Thanks Jenny re C-peptide. It doesn't correlate well and there are many observational studies correlating fasting insulin to most of the common cancers, colorectal included... I was being niggled about that. I was also wondering why they didn't measure fasting insulin. Silly question really when they are promoting a low fat diet.


blogblog said...

In Britain vegetables were generally considered unfit for human consumption until the mid-1600s. They were mostly grown as animal fodder.

Peter said...

What a great idea!

And the flip side is the scene in a Thomas Hardy novel where a supply of meat (destined for the high and mighty) was rendered inedible through being badly contaminated with mud and road gravel and so discarded. The peasants still ate it with relish but were careful to chew without closing their teeth on the grit...


Aaron Blaisdell said...

Beans, beans, are bad for the heart,
The more you eat, the more you fart,
The more you fart, the worse you'll heal,
So strike those beans from every meal!

David said...

so what is it about being an obsessive, fastidious automaton scrupulously complying with authority that is so healthy?

Peter said...

Hi David, now there's a great PhD project, but you might have to self fund it....


blogblog said...

Dr Peter Clifton, director of the CSIRO's Nutrition Clinic, told The Sunday Telegraph there was ``zero evidence'' that eating fruit and vegetables could protect against cancer.

The CSIRO is Australia's premier research organisation.

Peter said...


OMG. Perhaps I was correct....