22 July 2011

Science, Statistics, and Truth

Science vs. Truth.

What?!?? vs.Are you implying that they are not the same thing?

Not necessarily - especially when statistics are involved.

A weekly news-and-commentary summary magazine called (of course) The Week, even has a regular column titled “Health Scare of the Week.”

Vitamin D cures cancer.  Vitamin D statistically minimizes recurrence of certain types of cancer.  Those are very different statements, and the “statistically” part depends heavily on sample-size and sample population (what population was checked? Healthy college students? Nursing home residents? Australian life-guards?). Also, measures of statistical significance (T-tests, etc) must be considered.  If you can understand them... (that's why I married a biologist and statistician).

Even then, how does one make sense of a statement like “...found that people with the most Mediterranean diet have up to a 40% lower risk of developing Alzheimer’s disease...” (Nature 475 [14July2011], p. S17)?

How about conflicting scientific papers?  One paper (Nemeroff et al., 2001, Am. J. Psychiatry 158, 906-912), cited in the scientific literature more than 250 times since 2001, said that Paxil was a wonder antidepressant with minimal side-effects. Now there are accusations that the study’s academic authors were “hand-picked” by the drug company, and engaged in gross scientific misconduct.  Among other things, they allowed their names to be attached to the original manuscript, which was actually written by an unacknowledged contractor hired by GlaxoSmithKlein.  The contentious issue of drug-industry influence over medical research is bubbling just under the spill-all-over-the-stove point.

Perhaps the best way to stay sane when faced with conflicting reports like this is first, never take something in popular “news” media too seriously - it’s always been written some distance from the original data, and nearly always has an incentive for a sensationalist slant. Second, never take a scientific study very seriously until there are repeats and duplicate verifications by other groups.

Well, can I gather information on the Internet? Don’t be in too much of a hurry here, and don’t bet your life on it - or the life of someone you love. When our sweet little Sun Conure started showing motor-skill degradation, we took her to a vet. They took blood samples and severely traumatized her for $240. They even warned us that “...she may have a broken leg from the handling...” when she was returned to us. Internet research on the symptoms drew an enormous breadth of opinions by a rather large number of self-anointed experts. “Avian Gout” was a common theme... but this Conure was fed the same food as another Conure in a cage less then 10 cm apart. They were nearly the same age - less than a third their normal life-span. The problem was transparently not food, then, nor was it infection. What’s left? All we could conclude was cancer.  Sunshine had lived with us for 5 years, perfectly healthy until last Summer.

We took such loving care of little Sunshine until she died in December. I tried feeding her her most favorite treats - and she tried valiantly to eat them, apparently just to please me. After she died, I recalled the final letter from My Dad, dying of lung cancer, where he said virtually the same thing of my step-mother -- trying to cook special treats to keep him alive.

Ultimately, Good Science is a highly variable thing, with as many variants as there are people doing (what they at least hope to be) good science. Many good people worry about Science vs. Good Science a lot, and there is even a journal focused on this. Though scientists often contribute to the Journal of the Philosophy of Science, some prominent scientists have felt that the practical effect on their work is limited: “Philosophy of science is about as useful to scientists as ornithology is to birds,” to quote Nobel Laureate physicist Richard Feynman.

So what’s left? By general consensus, the medical profession transitioned from doing more harm than good to its patients to doing more good than harm sometime around 1870. Something started working then, and history suggests it was the introduction of scientific principles, tests, and unbiased trials. An early start was in 1854, when John Snow assembled a map of London, marking cases of children dying of cholera, and saw them centered around a single community water pump on Broad Street. He arranged for the handle of this pump to be removed, and viola - end of epidemic. Stopping child deaths always seems to gather peoples’ attention.

And that raises another issue: when you have limited resources, do you spend them on something like solving tinnitus (the screeching that about 30% of adults like me are hearing every day and night of their lives), or AIDS, which affects roughly 0.3% of the US population, and rarely kills anymore?

Medical science keeps getting better and better - I’m enthralled by weekly stories from my son, a 3rd year Med student. He describes wonderful doctors (and at least one hideous, totally amoral doctor) on his university teaching staff. With the exception of that one doctor-researcher (who has her own agenda), doctors depend on the medical science that itself must depend on statistics - to evaluate the efficacy of one drug or treatment vs. another. You can get better, more reliable statistics with the largest possible sample-group. 238,000 people (The Harvard Nurses Health Study) will give more reliable results (you can have more confidence in them), than a study with 45 men taking a drug and 52 men taking a placebo for a control to evaluate a possible tinnitus solution. Every additional person in a clinical trial, however, costs money. NSF grants being limited, sample populations will almost always be limited as a result.

I said “almost”... read on.

Keep in mind that there are no black-and-white medical fixes or solutions - not even cigarettes vs. no cigarettes. My uncle John died at 96 after chain-smoking for 82 years, while a practicing LDS friend died of lung-cancer almost two years ago at age 45. He had never smoked.

The real problem is that there are too many variables to sort through.

Was one somehow getting more vitamin D in his diet? Less sunlight? More polyphenols? Less exercise? Did one have longer-lived parents? Smoking is as close to a statistical slam-dunk for a known problem that can be avoided, but the example above shows that even with tobacco, there are always glaring exceptions.

If you want the largest statistical sample possible, to get the most reliable numbers, you have to go to the state level: millions of people. Life expectancy is highest, prostate-cancer, lung-cancer, depression, and other malign levels are lowest... in Utah.  It doesn’t take a rocket scientist to figure out a correlation.

Joseph Smith gave it to us in 1835.


No comments:

Post a Comment