Sunday, August 29, 2010

My dad says so...

It was 6AM. The sunlight had barely grazed the top of the tower. The morning glory was in full swing. The bucolic English countryside was still mostly asleep. The man stood with his hands and feet tied by leather straps. Eight individuals stood as if in solemn prayer, heads bowed, each in his own recursive thought. The drop distance, calculated based on the prisoner’s height and weight, was figured at six feet and one inches by the Prison Medical Officer who also was in attendance. One of them looked at the time clock overhead and nodded to another, who moved and placed the black hood over the man’s head. The one inch braided rope was tight around his neck.

 The priest prayed for the prisoner’s soul and having done he looked down. The nod that would bring justice to the world against crime without a moment of hesitation occurred. Just before the firmness of the trapdoor dematerialized and the hooded body fell through, a scream emerged, “I am innoc…!” Caught in mid consonant the echoes completed the sentence. The declarative from beyond the grave had reached the living.

His crime was labeled as “murder.” The clues that bore testimony for the evidence included a crow bar that he held in his hand. The wound on the victim was that of a pickaxe, the button near the victim matched the one missing from his overcoat; the bloodstain on his shirt and the treads of his shoes matched the footprints next to the victim. Then there was the death-knell of the numbers of people fitting the description that had been at the victim’s site. He was the one who reported the murder to the authorities and fit the height and general appearances from other observers at the scene. Nothing more needed to be said. It was a slam-dunk – though circumstantial at best.

Twenty-two years later amid the conflicting poisonous thoughts and self-cleansing rituals of a seventy-year-old man, history was purged onto a piece of paper. The dying old man riddled with a cachectic disease expunged himself of his egregious and dehumanizing behavior. He confessed to that murder and other crimes. The folded paper was found in his left hand.

Such are the trials of the human spirit. The judgment dispensed with pomp and vigor with the wrong set of tools to the innocent man.

The same governance applies to statistics and present day interpretation. The H1N1 epidemic that never was more than a seasonal flu, caused a stir worldwide when the UN stepped-in and pronounced that it was a pandemic – a fearful crises. Sometimes in an effort to claim credit for a discovery humans push hard for their personal objectives. The long arm of Tantalus pulls at their collective thoughts makes them see what others cannot and in so doing in their fervor convince, cajole and circumvent and stifle contrary opinion. Once the fever pitch has reduced, a small blurb denounces the myth and the populace is happy that they all live in a clean and sterile world again.

                                 Painting of Tantalus (greed beckons)

To look at the manipulation of small numbers gives us a window into the larger numbers. For instance, if you have a salary of $100 a week and management reduces you by 50%, now your earnings are $50 a week. After a lot of brouhaha and media sponsored criticism, the management agrees to raise your salary back by 50%. Are you back to where you were? No. You are now making $75 a week, since 50% OF $50 is (50+25=75) $75 and not a $100. This is simple but accurate. If some one wants to hide facts they throw in percentages. Listen to the true numbers.

An example worth entertaining: Statins are the new vogue in town, professed as the cardiovascular disease “tamer.” All collective advertisements and the drug labels say that Statins may reduce the relative risk of a cardiac event by 38-46%. That is an impressive statement by any standard. Wow, I can eat the Twinkie and the Crème Brulee and the skin fat of the duck and take a pill without remorse. In fact many a cardiologists, I know go out for a big meal at a restaurant and on their way home pop one of those pills, :It keeps my cholesterol down!” they say. Really. So looking at the real patient numbers of five major studies of a Statin and subdividing them into a basket of 1000 patients for the Statin side and the placebo side gives us the following: 32/1000 patients had a cardiac event while 41/1000 in the placebo category. Add 32 to 41 = 73 now divide by 32 gives you a 43% Relative Risk. Interestingly the cardiologists having been so enamored of this pill have begun to believe in it handsomely, to the point that now they proffer that the “pill” reduces the risk of heart attack. Maybe it is the life-style change, diet change or whole host of advocacies inspired collectively in society that is coming to roost, or maybe it is the “pill” in some small way helping. Further investigating the issues of the side effects of the “pill” are listed as being a paltry 2%. The risks being, muscle fatigue, muscle pain, confusion and cognitive brain function disturbances and liver damage. But comparing 43% to 2% would seem very one-sided and in the realm of the advocates for the “pill.” Yet the 2% is an Absolute Risk and not a Relative Risk. If you were to give the Absolute Benefit from the pill then the answer is 41-32=9. And 9 divided by 1000 = .009% Oops! Here the Side Effects from the “pill” outweigh the benefits don’t they?

  Median means the status of the 50% of those tested.

 “Physiologists must never make average descriptions of experiments because the true relations of the phenomena disappear in the average.” - Claude Bernard.

Lets stay with medicine for a bit longer and look at how statistical manipulations can harm the psyche of the unsuspecting. The responses to therapy in oncology (cancer medicine) used to be categorized as Compete Response (All discernable evidence of disease obliterated or 100% reduction), Partial Response (Where 50% of the discernable disease by diagnostic tests such as scans and X-Rays was reduced) and Minor Response (Less than 50% reduction) Then suddenly one day entering into the 21st Century, the ivory tower decided that Partial Response would be changed to represent disease that had been reduced by only 30%. Why the change one asks? Well, you would show more responders in a study touting benefits of the particular potion and therefore the benefits would sound better and people would rave about it. Really? So instead of monkeying around with these response rates why not just look at Overall Survival Rate or OS. Now that hurts the manipulators! OS cannot be manipulated. Since, there is very limited progress in the absolute survival rates in three-decades does not mean that therapy has not helped. It has in mollifying the disease of cancer, caring for the sick and giving them a sense of wellbeing and chance to enjoy time. Unfortunately progress is not in the form of a giant leap but more in the realm of mini-steps. There is no need to make more of it then you should. Uncertainty is a fact of life. Manipulating to present it a certain way is the function of selfishness or self-righteousness, either way it is abandoning the truth. Unfortunately it is widely used by the media, you know, TV, Newspapers, Magazines etc. Mostly because either they don’t understand or don't really care because of compelling time constraint to put it on the editor’s desk. Unfortunately, it is also as mentioned, sometimes inadvertently used by the white coats for self-aggrandizement or by the government-types- you don’t say- for political messaging. Note I don’t use the word reasoning because to reason requires concept deliberation and understanding!

Another wrinkle in the pages of concepts is the correlation and causality issue. To muddy the waters in the pristine white sands of our thinking, I am reminded of a question posed by a ten-year-old: “Does listening to loud music cause pimple?” I mulled over that one for a while.

“Where did you hear that?”
“My dad told me.”
“The truth is that loud music through the ear buds and headphones cause a hearing loss but do not cause pimples.
“But he is right. My older brother and his friends have pimples and they listen to loud music all the time.” He whimpered.
“You see young kids love Rock bands and Heavy Metal which they listen to in high volume. But the act of listening does not cause the pimples. It only hurts your hearing.”
“But my dad says so…!”
“And he is right for you not to listen to loud music.”

You see the problem. This little boy was scared of the gigantic yellow-tipped angry-red pimples his sibling had and wanted no part of it, but he was conflicted with the causality to ask the question. This brings us to a problem of a Confirmatory Bias, where, if I had answered yes to his question to keep his thinking on the same track, he would go on through puberty in the dark about pimples and loud music. As one grows older one substitutes a different "governing power" for the "dad" and lives happily in fear.

My best friend told me in her youth she followed her father around in household repairs. While he was nailing down a board, she asked him, “Dad why does the nail bend every time you hit it with a hammer?” He answered, “It bends because you are talking too much.” After a while surveying the results quietly she asked, “Why does the nail bend even when I am not talking?”
“Darn, he said this finger keeps getting in the way.” Just like in truth, data keeps getting in the way of a fabled story with hastily built foundations.

The armchair analysis of any scientific study hastens to mind a smattering of causes, a mixture of bias, chance, inference, manipulation and motivation. In medicine since we deal with humans there are multiplicity of competing reasons to objectify a cause with aplomb. The science of medicine is not the same as the science of physics dealing with forces, speeds and vectors defined in units. Medicine does not lend itself to absolutes, it conspires to make stealthy attributes where none are expected and remove those attributes from somewhere else where they do not belong. 

The rudiments of statistical thinking however dates back to 1828 when Pierre Charles Alexandre Louis did some simple mathematical inquiry into blood-letting for pneumonia and infected patients and found that the widely practiced method had no beneficial effect but possibly a deleterious effect. The procedure was ended slowly as the word caught on. There was a minimum of correlational analysis contrary to our present day, where correlations are used to imply causation all the time. The more the analysis is reviewed by the experts in the Media and amongst the laity the more it earns itself a seat amongst the full-fledged irrefutable causes for the malady under review. With that everyone is happy that science has marched on and left in its wake another great victory. Not so fast, because the next turn in the road shows the fallacy of this incomplete link through correlation. Reality is vindicated eventually.

Medicine is indebted and also held in hostage by the brilliance of the bespectacled nearly blind, with full bearded pipe smoking man named Ronald Aylmer Fisher. Fisher could not read due to his eyesight and held most of his opinions through lectures and social communications. He learned to visualize with an imagination unlike any other through multidimensional thinking and came up with the concept of Randomization. This he acquired from W.J. Gossett of the Student t-test fame and merged the two processes of randomization and the t-test to come up with a scientific terminology called analysis of variance or ANOVA utilized widely in science. Unfortunately for all the mathematically developed scientific study method the foundation of proving the null hypothesis to be true is based on a 5% error. All medical studies are based on a CI of 95%. CI stands for Confidence Interval, which means that there is a 5% chance that, the observation of a difference between the studied drug and placebo or standard care or any other comparator is by chance and there is a 95% confidence that it did not. Putting that in perspective one would be hard-pressed to say it’s okay to take the bungee cord jump with a 5% chance of ripping the bungee cord and plummeting to death. I don’t think any one would take that chance, if it were true. Would you? Yet that is what all experiments in science are based on, the decisions of the FDA for drug approval and the studies conducted under the auspices of scientific rigor are based upon the Confidence Interval. In essence you are trying to eliminate 2.5% of the data from the left and right tails of the Normal Distribution Curve (Bell Curve)

Another pesky problem in medicine is that the burgeoning list of medical journals consist large number of published studies made up of small number of patients and therefore subject to question, as authors hurry to publish a result and be recognized. The problem with the small number of patient data reveals a world-class error in interpreting reality. Case in point was an ISIS-4 study that negated a Meta analysis of 7 small trials of 1301 patients proffering that intravenous Magnesium supplementation reduced the risk of death after a heart attack. ISIS-4 study with 58,050 patients randomized to with and without Magnesium, found NO benefit at a cost of millions of dollars. The small numbers with even a single death in one arm makes the other side “statistically significant.” And while we are at it, let me address the famous p-value. The statistically significant p-value of equal or less than 0.05 was Ronald Fisher’s hunch and not a mathematically determined end point. It just made sense to him. But he was dealing with mathematics and agriculture and not human lives. His hunch lives on and gathers more steam as time passes in the stream of humanity.

Imagine that a study showing a difference between 99.82% and 99.88% yields a benefit of 34% (p-value 0.003). The absolute benefit is 0.06%! It would take 1592 patients to show a single patient benefit! This would make for a splash in the medical journals touting the relative risk reduction numbers of 34% in the abstract and the headlines. The reality remains buried in small italics in the belly of the misleading beast if one has the time to review it.

Where is the Life we have lost in living?
Where is the wisdom we have lost in knowledge?
Where is the knowledge we have lost in information? – T.S. Eliot

You can see that manipulation of data to prove something is within the purview of statisticians and mathematicians nowadays to prove the expected bias that the scientist wishes to prove. Ouch! The tactics used nowadays is reductio-ad-absurdum. Reduce the information to its infinite parts and rebuild it to make a mountain out of a molehill (my definition).

In economics as in medicine along with other disciplines, there is a prejudiced view of thought labeled, ”observer bias.” Here one sees what wants to see and stacks the odds in his or her favor to the exclusion of the truth and reality. Therefore the very act of observation is prejudiced from the beginning to give the result that one has contemplated. Upon reaching the desired result, the cry of “Eureka!” is convulsively expelled for others to rejoice in the “discovery.” Thus the art and science of experimentation is corrupted to suit the thinking rather than the other way around.

                     Archimedes: the fulcrum and weight displacement

In reality and with real experimentation most brilliant departures from the norm are made through slow methodical and non-judgmental beginnings. The Archimedes “Eureka” was a boiling and brewing thought process that congealed when he stepped into the bathtub to displace water. 

Newton and the apple, 

the Wright Brothers and the geometry of the wing following many mishaps before success was achieved, confirm the fact that innovation and progress is never made through manipulation or wanton misrepresentation of data. It is the slow contemplative advance till the thought has reached the perfect temperature to yield its virtues like the color of browned sugar.

Large scale Economics suffer from the same quakes that want to absolve the readers from all possible damaging thoughts. When things are going well and the economy is humming along at a fine pace, the demand for products goes up hence the price of the product goes up. Initially the company has to hire more people and the cost of labor has to be translated into the price and then the price maintained to yield a profit for the risk the company has taken. This therefore leads to a higher CPI (Consumer Price Index) and a PPI (Produce Price Index). Ultimately this translates to the rising cost of products, which leads to inflation. The government looks at the inflation rate and reports it monthly. But wait, when the prices are sky-rocketing and the fear that the dollar is not going far enough to buy the necessary amounts of the products, the government steps in to appease the masses by saying, we have to look at the “core” inflation which is ex food and energy. Oh okay. I understand. Maybe.

And then there is the quarterly ritual of the GDP (Gross Domestic Product) Why is always revised downwards after the initial figure? Because once the dancing and partying has ended the revised numbers are far away in the future to affect that past. The partying has already occurred. The moment enjoyed and so what is wrong with a 0.6% downward revision. Move on old chap, you say, don’t be a spoiled sport. Okay?

How about Stocks? Everyday the TV and most every Internet has a stock symbol look up price depicted with a green colored numbers if the stock went up that day or in the dreaded red color if it ventured down due to selling pressure. However if you are not a day-trader then the more important issue is to look for the fundamental and technical value of the stock price action. For instance if it is on a steady decline, but had an up tick on a certain day or week why get trapped there? If for instance a stock not paying any dividends languishes at a constant price, it is like parking your money in the mattress. 

Microsoft comes to mind; if you put money in MSFT in 1999 you would be down on your principle investment today even with the one time dividend! (Just the facts, I am not busting the great company that spawned work for thousands)

“Stars, hide your fires, let not light see my black and deep desires!” cries Macbeth in the face of dark reason. Yeah, there is plenty of light here if you are willing to flip the switch. The generators are humming with energy. So then let us begin with the issue of “Pay for Performance,” euphemistically written as P4P. What in the name of Sam does that mean? How does one contemplate performance in medicine? That patients with terminal disease if they are not cured then the treating physician has a poor performance record and therefore not get paid for services? 

Or for that reason a diabetic with four or five vessel disease and small coronary arteries cannot sustain the post operative recovery in 5 days and be sent home means that the cardiac surgeon has maybe rendered poor care? Or a General Practitioner has failed to prescribe the right medicine for a raging influenza and the patient succumbs to the disease? I mean some one is not thinking, the dark suits in the government maybe have never ever thought of the real issues in medicine.  P4P has hidden dangers written in bold all over it:  Rosenthal JAMA 2005 (294) and BMJ May 2010 (340) articles view these implementations and their dangers.

An example in the United Kingdom where there was a performance related issue with “Ambulance Response Time” for emergencies. The day it was enacted the graph for showing compliance skyrocketed to the peak. Interesting and rapid compliance one would say upon viewing the graph and the newspaper would shower accolades and the National Health System government auditors would bask in their glory of ridding the pesky delays that cost human lives. What happened was manipulation of numbers by the ambulance drivers and the real emergencies got delayed as the less emergent patients nearby were recast as Category A. As a result the unintended consequences cloaked devil came out to play the pipe. What the policymakers do not contemplate are the unintended consequences. Exactly what they wanted to avoid – delays in emergency – happened and cost lives. And these issues of policy reliance on poor information are evidence at every turn of the screw. Those that reason, anticipate and those that anticipate also hesitate in proclamations. A fall from 10 feet above hurts the same whether someone says it is only three feet.

Oh and I can't let this one go. The latest buzzwords in medicine are “Evidence Based Medicine.” Sounds good. You are therefore making decisions on some collected robust data. Now for the three ring binder full of questions, what does one consider evidence? Does eating a high carbohydrate diet as proposed in the past by cardiology societies and the American Medical Association recommendations still apply? No!

Does heavy marathon running constitute good exercise for a healthy heart in face of contrary data showing excess calcium buildup due to trauma? (recent data)

And of course we would all laugh heartily if we were to recommend cigarette smoking to cure chronic cough – proposed in 1899 edition of Merck Manual. 

Then there is the amputation of the limbs to cure Pernicious anemia (that is why it is pernicious because the cause remained unknown and death rate was high) changed to eating raw liver and then regurgitated food products from a doctor to finally a Vitamin B12 sublingual pills and injections. And lets not forget the previously mentioned “blood-letting.” Medicine flows like the river forever changing its shores and its depth. To rely on the past evidence does not do a patient good when better treatment is available and as of yet has not met muster of the ivory tower crowd, but since the “evidence” has not been collected for writing articles it is withheld from patient care. Oh me Oh my! (Read Dr. Jerome Groopman and Pamela Hartzband's article on Evidence Based Medicine, "Why quality care is dangerous" in the Wall Street Journal 2005 @

The consummate skeptic is welcome into the statistical paradise as he considers this jungle of pitfalls with the questioning eye and makes the learned decisions. So there are a few lessons here for all of us: Reading the information with an eye towards the motives, Considering the methodological hiccups in the experiment, Counting the real numbers where percentages are used, Considering the differences between Relative and Absolute Risks and above all Questioning with the same word you dogged your father with, Why!

For Further Reading:

Innumeracy, John Allen Paulos, (Penguin 2000)
The Numbers Game, Michael Blastland and Andrew Dilnot (Gotham Books 2009)
How to Lie with Statistics, Darrell Huff (W.W. Norton 1993)
Biostatistics, The bare Essentials Second Edition; Geoffery Norman and David Steiner (BC Decker 2000)
Big Fat Liars, Morris E. Chafetz, MD (Nelson Current 2005)
Reckoning with Risk, Gerd Gigerenzer (Penguin 2003)

No comments:

Post a Comment