I previously mentioned that there was a large collection of measurable variables that when improved lead to a longer healthier life. Most people have not heard of these variables or have read about them only in glossy adds designed to encourage the purchase of supplements, vitamins, miracle foods, or weird electromagnetic devices. More magical thinking, more ‘death or a pill’ thinking tricked out in self referential gloating about not being hornswoggled by ‘Big Pharma.”
Basic medical science is genuinely interested in discovering the causes and cure of disease, suffering, aging and even death. Thousands of people, pure of intent, of keen intellect and purpose work long hours in labs, in offices, late in institutions to unlock the processes of disease and the mechanisms of possible cure. If you don’t believe this what I will say won’t make much sense; won’t motivate you to take seriously the work of these scientists, these benefactors of mankind.
Why do I take you back to whether or not you have a basic belief in scientists and the scientific method? Because as a practicing physician for many, many years I’ve grown accustomed to the fog of beliefs, doubts, skepticism and confusion most people work within in their perception of medical science and physician’s recommendations. I well understand the problem but need your essential conviction that real science well done can and does discover things that matter. This is not as silly as it first seems. If you believe, most doctors do for example, that your cholesterol number, whatever that might be is one of the most important things in the world and that behaving in such a way as to drive this number down is all that matters, trumps everything, is the be all and end all of heart disease, risk for stroke and the like then I cannot be of help. For reasons big and small, all of the basic medical literature has always been clear about this, cholesterol is just one number among many and those then are nested together to evaluate your relative risk of vascular disease. If this notion of multivariate relatedness is too much to swallow then what I will say will sound like little more than questioning conventional wisdom- the Holy Grail is Lower Cholesterol- when what I will be trying to do is put this one number in the context of many others that matter as much or more but for which there is no pill to fix or change them and that for this reason you will not have heard as much about these other numbers. The wonderful thing is you have an enormous amount of control over these other important variables; actually much more control than over your total cholesterol number.
With this background in mind I need to make a few other conceptual points before I get to the actual things you can fix and why.
Let me introduce you to four ideas:
- First is the idea to be skeptical about what matters. For example: something might change a number related to disease but not affect or, in some cases, might adversely affect a more important endpoint like death; lowers blood pressure but happens to increase your chance of death or lowers your blood glucose/sugar but not the diseases or death associated with diabetes. Over the years many drugs have been withdrawn after discovering just such problems. An everyday example: most of the common anti-inflammatory and pain medicines like Advil have been around for decades and work very well for the endpoint of pain; the endpoint of death happens to be increased in those who take these medications for they are associated with an as much as a 40% increased chance of heart attack in regular users. Be careful how you choose your desired endpoint. Another example: high homocysteine is associated with stroke, heart disease and premature mental decline, take an array of vitamin B’s and your homocysteine goes down but not the homocysteine related risk for stroke, heart disease and premature mental decline. Oh and there is a known risk of colorectal cancer, breast cancer and prostate cancer if you take the B vitamins.
- Second is to note or watch for abuse or misuse of statistics: you will often hear that use of Lipitor and related drugs ‘reduces heart disease by 30%.’ What this actually means is that 1,000 people need to take the drug to prevent heart related problems in about 3 people. If 1,000 people do not take the drug and 11-12 of them have heart related problems and 1,000 people take the drug and only 8 or 9 people have heart related trouble then that ‘reduces heart disease by 30%. The perception and the facts don’t add up. A reasonable person thinks that “30% reduction” means if 10 people take the drug then it will save 3 people’s lives. What it actually means is that 1,000 people risk the side effects and as of yet unknown long term effects of a drug so that 3 people might not have a problem. Fine, let’s help those 3; eating peanuts three times per week has been shown to have about the same statistical impact as Lipitor yet no one, rightly, has claimed that ‘eating peanuts three times per week reduces heart disease by 30%.’
- Third is ‘number needed to treat’ (NNT): how many people have to take a drug or change a behavior to have a measurable benefit for one person. For example: on the order of one person has to take an antibiotic to be of benefit to one person; this is stretching the case but is true to the nearest rounded integer. For primary prevention, that is to help a population of otherwise seemingly healthy people, at least 200, that is two hundred, 200 or more people have to take Lipitor to be of benefit to one, I said 1, person. The others just get the cost and side effects like memory loss. As you will have noted NNT and misuse of statistics are related and reveal each other. By comparison 7 people need to raise their exercise capacity by 1 MET, I will get to this, a very reasonable thing to do, to help 1 person avoid the same end point, heart attack and so on, as the 200-300 who take Lipitor to help 1 person.
- Fourth is ‘number needed to harm’ (NNH): fine NNT helps me know my chance of benefit of a proposed therapy like exercise or medication but what is my risk of harm. Not what the harm might be but the number of people who need to take the proposed therapy and then wind up having some kind of harm from that therapy. Now this does not tell you whether the harm is a hangnail or death but just the relative risk of harm of any kind. If the NNH is sufficiently high then what the harm is might become less of an issue. If the NNH was 2 for example and the risk was death then I would probably take a pass but if the NNH was 1,000,000 then death might not look like a bad risk if it prevented you from suffering a stroke at 50 and the NNT was 100. Think this way when you worry about flying versus driving.
Now with these ideas in mind you can both police my claims and be a better consumer of pharmaceutical and other medical advice. Let me summarize those three points.
- NNT: number needed to treat
- NNH: number needed to harm
- Abuse/misuse of statistics
- Meaningful ‘clinical endpoints’: ‘am I dead or not’ and not ‘some number went down’ that may or may not matter when treated.