Thread by @ProfDFrancis

Happy Memorial Day Weekend to all cardiology fellows!

I thank m'learned friend Prof Simon Thom for showing me the funniest paper of the year.

I will discuss it tomorrow.

What approach should I take?

#cardiotwitter

Thank you to everyone for paying attention to the buses I sent round that said there would by £350,000,000 per week for your country's health system, if you voted the way I wanted you to.

Now I can tell @rallamee that I am only listening to the Will of the People.

The problem with starting an article with a clear, true and useful statement, like this...

... is that the only way is down.

The problem with expressing it like that, is that everyone nods in agreement, without understanding what is being said.

Here is why.

What would you like to do about your government wasting money?

Well that was easy.

What about falsely condemning people to the death penalty?

Everyone nods, and for the same reason when you ask:

"Should your research be more biased, same, or less?"

and

"Should research be more polluted by the placebo effect or less?"

Everyone agrees they desire less rubbishness.

I avoid this approach, because while the Frank Harrells and the Janet Wittes of this world all understand exactly what is being said, many people don't, and skipping the understanding of WHAT one is opposing, is a disaster.

The best description I have ever heard of the reason we need blinding was from this man, a genius of presentation skills: Paul Ridker, at the @CVCTForum meeting in 2018.

Off the cuff he said the following on the podium:

You want both arms of the trial to have interpretations and treatments on a level footing, before they enter the trial, AND AFTER.

Randomization only balances things that happened BEFORE the trial.

Blinding balances things that happen AFTER the trial begins.

"But why should I care what happens *after* a trial begins, with successful randomization?", I can hear you all thinking.

What is the single largest cause of Resistant Hypertension?

Clue 1
ncbi.nlm.nih.gov/pubmed/23337469

Clue 2

And why does this happen?

What is the best way of reminding yourself that there is something wrong with you that can kill you or permanently disable you, and that you should take the meds to prevent it.

Hmmmm.... Me thinks I am being trolled.

Rest assured, people who go into hospital to have a hole drilled into them preceded by a long panic-inducing consent mentioning death, bleeding and whatnot, will experience a mental slap in the face that this is a Serious Condition, which helps them remember their meds.

That is good for them health-wise, but it would be wrong to ATTRIBUTE THIS to the hole-drilling etc, lest we descend to the level of witch-doctors, homeopaths and other purveyors of gibberish.

By all means use every tool we have to improve people's health, but let's not be unscientific about attributing its mechanism.

In the case of renal denervation, when patients DID NOT know whether they had received the real or placebo procedure, their BP dropped just 2 mmHg more with denervation than placebo, with all data documented automatically by machine.

nejm.org/doi/full/10.10…

That wonderfully designed trial, led by Deepak Bhatt who never smiles and George Bakris who never doesn't, also reported what the HUMAN staff documented.

Staff, who again DID NOT know who had which treatment, found 2mm more BP reduction with denervation than placebo.

The brilliant insight from Symplicity 3 is that BLINDED STAFF and BLINDED MACHINES, both agree on the difference between the effects of denervation and placebo: 2 mmHg.

Neither the machine nor the humans knew which patient was which, so there was no chance of inadvertant tilting

Fortunately for curious people like you and me, people spent billions more dollars running various other designs of trial.

One popular design was to let everyone know which patients were the "Chosen Ones".

When you do this, TWO things change.

First, the behaviour of the patient.

Second, the behaviour of the staff.

Patients randomized to denervation, attended a major cardiac procedure with elaborate precautions etc, and adhered better to their meds, their automatic, machine-documented pressures falling not 2 but 10 mmHg more than their control arm counterparts.

thelancet.com/journals/lance…

Second, the staff change behaviour in documenting numbers. This is NOT INTENTIONAL CHEATING. This is staff trying to do the right thing.

When the patient has undergone a BP-lowering procedure and the BP is (for example) markedly higher, what do you think they do?

I have asked "what would you do?" of audiences of doctors on 4 continents. They overwhelmingly vote for "measure again". The reason is NOT to cheat, but because they are scared they haven't properly relaxed the patient, and they want to AVOID writing an unrepresentative value.

The other way to document BP is to have a portable machine record automatically, it a few dozen times, across day and night.

How do you think they answered this question:

"If an ABPM average reading was unexpectedly high, what would you do?"

Of course they would! They are not frauds.

"If *I* as an individual doctor have failed to relax the patient in my office, that's my fault, and I want to do a good job, so I correct the situation...

"On the other hand, I didn't follow the patient round all day and all night, stressing him out! So there is no way you can pin any of those measurements on me! It's not my fault and I am NOT going to meddle with it. The result is the result. Get lost."

As soon as you tell the doctor the clinical background of a patient, i.e. they have had a groundbreaking magical breakthrough treatment from heaven, or not, they change their behaviour during documentation of things that CAN BE BLAMED ON THEM. We doctors are conformists.

This powerful influence is easily seen in the unblinded trial data.

The MACHINES document a ~10 mmHg benefit of denervation,

but the HUMANS document a ~25 mmHg benefit of denervation

We explain this in more detail in this article ncbi.nlm.nih.gov/pubmed/25467244

In short, denervation seems to have a small beneficial physical effect, 2-5 mmHg, when you take into account later (blinded) trials.

This is valuable, because it doesn't require anyone to take tablets, and there is a big cohort of people who simply don't want to do so.

To this physical blood pressure lowering effect (let us generously call it 5) is a further 10 mmHg of remembering-value.

A hole in the leg instead a knot in the handkerchief.

And, for HUMAN-documented rather than MACHINE-documented, another 15.

James Howard, being a fan of euphemistic periphrasis, calls this element "Asymmetric Handling of Unexpected Values".

I call it cheating.

Anyway, there you go.

Up to 5 mmHg of blood pressure lowering
10 mmhg of reminder value.
And 15 mmHg of (ahem) ditching unexpected office values.

That's your 30 mmHg magic denervation effect.

There is one shocking truth I have learned from the denervation escapade:

Without blinding, randomization can be worse than useless.

You and the patient put in all the effort, but are not rewarded with the truth.

Or, to put it another way:

If you are about to do an unblinded trial of a symptom-relieving treatment:

Don't bother. Just make up the results. We know it will be positive.

Embezzle the money. Spend 1/3 on cars, 1/3 on women, and waste the rest.

You will save patients the effort and benefit the economy.

Back to the paper that has the unenviable award of being the funniest of the year.

No it isn't. Only by stupid people who want to fail to understand what is happening.

If I have at considerable cost worked out that 1 egg + 1 egg = 2 eggs, then when I move on to cows, I don't want to say that 1 cow + 1 cow = ஜ cows.

Science is about finding generalisable principles that work, not creating artificial silos.

The only reason I would flip-flop terminology is if I wanted to sell you ஜ cows for the price of 3, and I needed you to fail to understand what I was saying, until you handed over the $.

I like this sentence because despite some years of education, I can't understand what it means. While some people wake up and do a DropAndGiveMe50 or whatever, I just open up my laptop and read this to myself a few times, until my head hurts.

You know how in the interventional world, we have "discovered" the need for blinding?

Boy, we are so smart!

Like Columbus, discovering America.

Only thing is, there were some dudes who discovered it already, through the accident of living there.

Just as the rest of medicine managed to understand very quickly in the 1950's:

Balance the arms BEFORE the trial with randomization.

Balance the arms AFTER the trial begins, with BLINDING.

To blind patients and staff in a trial of tablets, we need a placebo tablet.

For a trial of procedures, we need a placebo procedure.

It really is no more difficult than that. Unless you want to sell ஜ bananas, in which case you will quite rightly make it more confusing.

So this is what puzzles me.

When we in intervention, in Columbus-like dignity, discover the need for blinding, and the fact that everyone other than us had considered it normal scientific procedure for the past >60 years, how do we describe the situation?

I think this means, it is obvious to everyone in the world, and has been for 60 years, BUT NOT TO US.

So we don't understand why we have to do it

We have our own ways of doing things, and we reserve the right to continue thinking the way we do. We don't care how silly we sound

The next bit is cataclysmically baffling.

It's like Columbus saying, "My grandfather went to America, and found it to be really surprising. But I haven't been there yet, so it is unnecessary and irrelevant.

Hmmmph."

The article reports that blinding (placebo control) was discovered by interventionists in the 1950's, and permitted the discovery that internal mammary LIGATION, a treatment for angina at the time, was worthless.

OK, and what was this new era that was "ushered in"?

If a trial published in 1955, "ushered in" a new era that took over half a century to arrive, what do you think of the quality of ushering?

How about this?

Did ORBITA test for the MECHANISM of improvement of exercise time from clinical stenting?
Or did it test whether there WAS an improvement in exercise time arising from placing the stent versus placebo procedure?

In case you are in doubt, in clinical stenting we take great care to impress upon the patient how wonderfully wide open their artery now is, and how it can't cause ischaemia now.

In ORBITA we didn't tell them whether they had a stent, and certainly not how wonderful it was.

This is why ORBITA was not a test of the MECHANISM of the angina relief we seem to see in clinical practice. The patients did NOT get the full experience, they got a stent or not, with as little superadded shenanigans as possible.

How about this?

For over a decade, unblinded studies had reported ~30 mmHg effect sizes for denervation. We fell over ourselves to come up with the most ludicrous claims of effects sizes: 40 or even 50 mmHg were claimed during peaks of exuberance.

(Names withheld to protect the guilty)

All of that effort was WORSE THAN USELESS.

Not only was it a waste of money, AND made holes in people for no gain in truthful knowledge, BUT ALSO it actively created FALSE INFORMATION.

Symplicity-3 was a much-needed shaft of light in this gloomy morass, the first trial to balance management not only BEFORE but also AFTER randomization, by applying BLINDING.

Yes it was research, like the other trials, but only in the way that these are all carbon.

Symplicity was the first trial to provide genuine information. The rest of the information available at that time was, unfortunately, all a load of rubbish.

The best use for the unblinded trial data is like that of the coal, to keep yourself warm in winter

8-(

So I consider this sentence to be 180 degrees backwards.

Symplicity 3 was a giant, and at that time the only, setFRONT in the investigation of the technology. Everything else was the setback.

I found this next bit particularly amusing.

Does technical failure of a denervation catheter ONLY occur in blinded trials?

Regression to the mean is the process by which when you select people who are particularly happy and ebullient today, the chances are, they will be less so tomorrow, because some of them rose to the top of the joyfulness ladder on that day due to some chance event.

Likewise, if you make a list of all the people in your ward with the highest potassiums, or in your clinic with the highest BPs, and recheck them (without doing anything) on another day, ON AVERAGE they will almost certainly have gone down statistically significantly.

wtf!!

for avoidance of doubt, "blinding" means that AFTER the procedure, all the people who were in the lab, shut up about it, and do not speak to the patient, nor tell the other caregivers whether the real procedure was done or not.

They do NOT wear blindfolds!

OK, back to regression to the mean.

One beautiful feature of a randomized controlled trial is that the data can regress to the mean all it likes.

We... don't... give a damn!

BOTH arms will regress. We only care about the difference between arms.

So, happy days!

Errr... no!

That's backwards.

The benefits of any *proposed new therapy*, which we are going to persuade people to have, which will cost money and will make holes in them, must be justified ... blah blah.

Now I am getting a bit angry. I can feel my BP rising. I think I need denervation. About 10 or 20 denervation procedures should be enough, I think...

The reason I am angry is
(1) "trainable outcome of exercise time" has been exactly what we have used to identify effective drugs. Every single anti anginal has had to prove itself on this same endpoint, AGAINST PLACEBO, otherwise it heads for the trash can.

Yes people often do a little better on the second than the first exercise test, BUT THAT HAPPENS IN BOTH ARMS, hence the term "randomized controlled trial".

Doh.

And (2) WTF is "the subjective symptom of angina"?

Patients have angina.
It is a symptom.
It is experienced by the patient, not the doctor!
That means it is subjective!

It is frankly ridiculous to criticise symptoms for being subjective - that is EXACTLY what they are supposed to be.

Go to the back of the class.

No, on second thoughts, go out the door and into the next-but-one class down the hall.

This next bit is multi-level ignorance

First, let's take a poll:

Of the patients randomized, what proportion at the time of randomization had ONE OR MORE positive ischaemia tests:

Clue:

eurointervention.pcronline.com/article/swimmi…

Second, if you are measuring at what level of FFR *symptoms* begin, how should you do it?

Well, here is some data from one of those designs.

ahajournals.org/doi/pdf/10.116…

Specifically:

Some people were hoping for a cliff at 0.80.
We were expecting a slope.

But there wasn't a cliff.
There wasn't even a slope.

There was a great big nothing.

(Think there's a bit of a slope?
Think longer.
If >0.80 is less ischaemia, should there be less benefit or more?
Ah, yes, now you get the point.
Let's not talk up that tiny tilt of a slope as a "signal" now that we realise its the wrong way round.)

OK too much laughter for one day, my friends, so let's wind down.

Remember how denervation delivers its 30 mmHg?

5 real mmHg, like this screw that keeps falling out of my laptop. Small, but important.

10 of knot-in-handkerchief/hole-in-leg reminder to take meds

15 of cheating

Suppose you observe that with unbiased automatic BP measurement,

(a) UNBLINDED trials, the active arm patients underwent the denervation experience and the controls didn't, and there was a ~10mmHg difference in BP change between arms

(b) but in the BLINDED trials, no patient knows what they had, and there was virtually no difference in BP change between arms,

Would you conclude:

Well, this was their conclusion:

Just as a reminder:

Blinding means the PATIENT behaves the same between arms (in terms of tablet-taking, for example)

and the DOCTOR behaves the same between arms (in terms of which number to write down for the BP).

If you take out the doctor component of bias, by automatic blood pressure measurements, that's great, but you still need to take out the patient component.

If you do blinding, it is all straightforward. All done in one go.

I think this waffle means "We don't like it, but we don't know what we don't like about it, so we are just gonna say some impressive-sounding, but vague stuff."

Another ludicrous implication is that placebo controlled trials are dangerous and should be skipped:

... Skipped, and instead replaced by non-placebo controlled trials, whose results will be positive, so we can go and do these procedures to millions of people across the world.

Their complications are no concern to the writers, apparently.

And now, the coup de crass.

In an article peppered with nonsense, this is truly moronic, looming like Everest itself over the lofty Himalayas.

In a randomized controlled trial we do not say "You are in Arm A" to patients who are in Arm B (or even in Arm A!)

We say, truthfully,

"We do not know what arm you are going to be in, until the randomization."

Later, the patient has sedation, randomization and their procedure (done by staff who are not blindfolded!)

When they wake up, there are new staff, who know it is their job to have received handover from the cath lab without knowing whether it was real or placebo. They know it is their job to treat the patient for both eventualities.

When the patient asks "Did I have the real thing", these second set of staff (i.e. everyone in the world other than the few people in the lab for the procedure) says, truthfully,

"We don't know. As arranged, the people doing the procedure have not told us or you."

So let's take a poll, as I know my boss will haul me into his office sometime next week when he learns I have used the word "moronic".

Each of you has the benefit of only one brain, unlike the article which has had several.

In your opinion, who is being "actively deceived"?

Oops. I think @rallamee has been alerted that I have sneaked out of the Twitter closet this weekend while she was out of internet contact on holiday.

I'd better go quiet.

One piece of good news though. My next tweetorial has been sponsored by a well known car maker.

Bye-bye!

One final thought though:

This is why we purveyors of a procedure must never be the ones to lead the discussion on the value of the procedure.

We are only human, and vulnerable to a massive drop in IQ when thinking about activities we love.

If the article-writers knew they were making ridiculous points, they would not do so, and certainly not in the Lancet.

We are all cursed with rose-tinted spectacles when discussing our beloved speciality.

Love is blind.

Important erratum raised by Kyle Sheldrick @K_Sheldrick

I am naughty to call the 15 "cheating".

There are three people involved here:

1. The staff making the measurements.

They are not cheating. They are trying to AVOID failing to have relaxed the patient properly. [The impact is the same as that of cheating, but this is not the motivation for their actions.]

2. The paper authors.

They are not cheating. They simply forgot, when designing the trial, about the need to balance the management of the trial arms AFTER the trial begins, with blinding.

They forgot at design-time, and aren't suddenly going to remember during writing-time.

How would you define "cheating"?

Group 3 are the people who discuss the results of the trial once published. (They may or may not be the same as the trial staff, group 1, and paper authors, group 2).

However they are now looking back on data, in the full knowledge of the Howardogram ...

How would you categorise the activity of people who, with all this knowledge,

(a) still insinuate that Symplicity-3 is a failure (when in fact it was a major scientific success), or

(b) oppose balancing management post-randomization (i.e. blinding & placebo), or

(c) Try to rebrand "placebo" as "sham", so that they can pull it away from your mental category of "obviously necessary feature" and into the category of "active deception"