I failed my first statistics course. I am reasonably bright but had a phobia of numbers and formulae. More importantly, I was not motivated to learn all of those strange concepts.
After all, I was studying to be a clinical psychologist and help people. I didn’t want to become a statistician or an accountant. It also didn’t help that the statistics professor felt it was his job to intimidate students and make them feel stupid. From his point of view this helped students study hard and learn.
Thank goodness when I took my first statistics course in my PhD program, I totally lucked out and had a brilliant teacher in Dr. Alastair MacLean. His goal was to help his students understand numbers and statistics so that they could accomplish what they wanted.
Since that time, I have learned much more about statistics and numbers. Let me tell you why I am glad I overcame my fear of numbers.
Whenever, I go to the doctor, I ask “How likely is it this treatment will help me?” I usually get something like “It is pretty good’” or “There are randomized trials showing it works.” Nothing too precise.
There is, however, a great statistic to answer that question with. It is called Number Needed to Treat or NNT. NNT refers to the number of patients that are needed to have one person benefit above what would happen naturally. Of course NNTs have to be used in the context of the patients’ values.
So if a disease or disorder is inevitably going to kill me in a year and 1 in 5 patients survive with the treatment, the NNT would be 5 and I would certainly try it unless the side effects were so horrendous that I would rather die. But if the NNT were 100 and it would make my remaining time uncomfortable or if it would cost so much as to make my wife penniless if I did die, I probably would not take it.
Number Needed to Harm or NNH is very similar but the number refers to a specific side effect.
Most diseases have some natural remission and so NNT is a bit more complicated to figure out.
Technically, NNT is the inverse of the absolute risk reduction or 1/absolute risk reduction. It is calculated for two treatments, usually an intervention and a placebo intervention for a defined endpoint. So if a drug for face rash eliminates 80% of the rash within 1 month in 70% of patients and 50% get 80% better on a placebo drug, the NNT would be 1/(0.7-0.5)=5. An NNT of 5 is very good and if the treatment was of modest cost, I would probably do it.
There is a great web site that has carefully collated data and calculated NNTs. It is called The NNT. Here are a few examples from that site.
-Mediterranean diet for those with a previous heart attack, to prevent a second heart attack NNT=18; to prevent death NNT=30; to prevent cancer NNT=30. There were no harms seen.
-Training for falls prevention in the elderly NNT=11 with no harm seen.
-Low dose aspirin each day for a year to prevent a heart attack NNT=2,000 or to prevent stroke NNT=3,000 NNH for a major bleeding event was 3333.
-Statins taken for 5 years for those with known heart disease
- NNT=83 for life saved; NNT=39 to prevent non-fatal heart attack; NNT=125 to prevent stroke
Many treatments do not have NNTs worked out but we will see this statistic increasingly being made available. There are not many NNTs worked out for children’s illnesses because very often drugs are not tested on children.
Another source of information about whether a treatment or a test is likely to do harm or good is Choosing Wisely which I mentioned in my June 12, 2014 blog. Choosing Wisely has developed lists of tests and treatments that you should be talking to your doctor about. But they may do more harm than good. For example bone density tests are often used to see if older people (like me) have weak bones. The problem is that for most the test is not needed and the finding of weak bones often leads to being put on drugs that not only don’t do much good for most people but also have significant risks. There are other things we can all do to keep our bones strong such as stopping smoking, limiting alcohol, exercising and eating well.
As you can see there are things you and I can do to answer the question, “Will this work for me?”
~Dr. Patrick McGrath