While it’s not my usual fare, I saw something on Twitter that reminded me of a thought from long ago and I chased it down. When I was an Internist, it was the early days of treatment guidelines. They were beginning to show up frequently. They began to move into areas of preventive medicine – like when to treat high blood pressure, risk factors for heart disease, diabetes, cholesterol levels, etc. I didn’t like some of them, actually more than some. Many seemed like the suggestions from people based on small differences by people with a monocular focus. Some didn’t bother me. I didn’t mind commenting on smoking, obesity, etc. – the usual suspects. But treating minor blood pressure elevations was anything but benign. The medications of the day made people sick, or at least feel bad. I had no real confidence that the recommendations were true representations of the patient’s real future. It wasn’t the reason I changed specialties, but I didn’t miss the increasingly stringent guidelines when I left. I guess I saw my job as treating sick people. As much as I respected the wisdom and miracles of preventive medicine, I thought many of the then newer recommendations were pushing the preventive metaphor too far.
Here so many years later, that feeling is still with me. I don’t feel comfortable at all with the Statins for minimal lipid elevations, and don’t take them myself. Nowadays, there’s the added concern about industry interference. Today, on the way home from the clinic, I passed a pharmacy with multiple signs stuck by the curb advertising vaccines of many kinds, only one of which made sense to me. Last week at a routine physical, my doctor mentioned several that I question. I didn’t say anything [but the look on his face suggested that he was on to my skepticism]. It’s not a big deal to me, but I do think that the many recommendations and reports that are all over the evening’s news are often examples of very small differences – differences that don’t necessarily make a difference.
So on Twitter, I saw this graph from a recent Japanese study…
Low-Dose Aspirin for Primary Prevention of Cardiovascular Events in Japanese Patients 60 Years or Older With Atherosclerotic Risk Factors
[
click for full text]
… published with this
editorial. It was intended to be a 6.5 year study but was stopped at five years for
futility – a term for
going nowhere, not changing over time.
In the editorial, they review the earlier studies in detail in what seems to be a balanced way and I leave that to you to read for yourself. In my reading all of this, I thought about two things. First, the risk/benefit ration in this case is clear – a baby aspirin a day is near no-risk for the overwhelming majority of people. It is unlikely that this issue gets into Conflicts of Interest or much in the way of industry interference [even if "baby aspirin" costs more than adult doses of aspirin these days].
My point isn’t about baby aspirin. I don’t take them myself, but if I had exertional chest pain, I might. If I ever have an oppressive substernal chest pain radiating down my left arm, I bet I’ll take Aspirin first, then call 911. My point is about the use of statistics in medicine in general. I’m moderately savvy as doctors go about statistics – a numbers guy by nature who likes quantification in almost any flavor. But statistical predictions feel out of hand to me. My specialty, psychiatry, has gone through a very long period where small differences have often been magnified to an outrageous degree and mere statistical significance has been presented as a surrogate for true clinical relevance. It’s not.
That Japanese study was on the news as expected. Should anyone change what they’re doing because of the news reports? I don’t think so. No one study is worth making changes unless it represents a real danger. But considering the general trends in reported research and the rapid dissemination of information in modern times, medicine as a whole would be well placed to spend some real time on the question of clinical relevance. We owe our patients help sifting through the vast amount of information that bombards us from all directions in both our prescribing and in education. My personal preference would be for there to be a specialty or an agency that evaluates all of this minute to minute reporting of preventive medicine advice and tries to separates the wheat from the chaff with the specific charge of turning statistical significance into the more important parameter – clinical relevance which is often a yes-no rather than statistical question.
This is a rather long post and will touch on some ideas in probability.
Aside from the numerous problems with the clinical guidelines being influenced by market forces, there is another problem that exists independently of those. Namely, that the guidelines are applying probabilities from a set of study populations to the individual patient in the office. The probability that a patient in a study, no matter how well designed, will benefit from an intervention is not going to be the same as the probability that a patient in my office will benefit from that intervention.
The reason for this is that the relevant probabilities are conditional. They describe the probability of benefit for a person given that the person meets certain conditions. Now in a well-designed study, attempts are made to keep those conditions as consistent as possible across the participants. That necessarily limits the applicability of the results. The patient in my office almost certainly does not meet those conditions, so the study results almost certainly do not apply.
What I need to know as a clinician is: If the probability that a person in a study who meets conditions A,B,C, D, …, M will benefit is 0.6, what is the probability that my patient who meets only condition A and B but also has condition R and V will benefit. The answer to this is not just unknown, but is mathematically unknowable. I am forced to make a guess, and I can’t even say how likely my guess is to be accurate. The best I can do is use my knowledge of how the various conditions influence the effect of the intervention and hope I am not too far off base.
The use of clinical guidelines to mandate treatment can lead to very bad medicine. They may become a procrustean bed upon which individual patients are either stretched to fit or cut down to size. It is the right and the responsibility of the treating physician to weigh the various conditions for each patient and convey their reasoning to the patient so they can make as informed a decision as possible.
—–
For those interested in how subtle changes in conditions can effect probabilities I offer the following thought experiment:
John and Mary have two children. I want to know the probability that both children are girls.
Condition 1: No information. Answer: Probability that both are girls is 25%.
Condition 2: The oldest child is a girl. Answer: Probability that both are girls is 50%.
Condition 3: At least one child is a girl. Answer 33.33%. Note that changing the condition subtly, i.e. knowing the oldest child is a girl vs only knowing that at least one child is a girl, changes the probability significantly.
It can get even more bizarre. Here is a condition which does not seem like it should influence the probability of both children being girls at all. But it does.
Condition 4: At least one child is a girl with curly hair. Answer: It depends. If we ask question in Taipei the probability that both children will be girls given that at least one is a girl with curly hair is close to 50%. If we ask the question in Nairobi then the probability of the same outcome is close to 33%. (It depends on the frequency of curly haired girls in the population.)
So changes in conditions which we would never expect to have an influence on the probability of an outcome clearly do.
Clinically this could be the change from a study where the medication is paid for to the real world where the cost of the medication is a significant stress on the patient, and on and on.
If anyone wants to have the details for the example I used Mickey can give them my email.
Joseph,
Thanks for that long comment. It was, indeed, a thought experiment that was a real addition. As was:
I know just enough statistics to be dangerous. Mickey has helped to dredge up some old concepts and put some new names to new ideas.
I have what I call the three places to the right of the decimal point bias. This means any answer to any problem that results in a number requiring thousands of patients or other action to garner one positive result is due to study error. This is why NNT is an important number in my decision making process.
Not being a doctor I then have to look at the financial incentives for prescribing and the pharma driven advertising and mandates. Many drugs do not look good under this light.
Statins are a favorite in that the NNT becomes very high with minimal lipid elevation. Pharma has used a surrogate end point, lower LDL, as a measurable marker. We then see statins being pushed for everything from dementia to preventive care with no indication.
What is not studied is such links as long term use and the possible promotion of dementia, as mentioned recently in an article. People with very low LDL being pushed to even lower levels and the resulting physical problems such as pain and confusion.
One doctor posted some time ago that if these things are so good they should be put in the water.
My reality is statins produce a measurable result, lower LDL. The have a fairly benign profile in lower doses, but as with all drugs not entirely safe, especially when pushed to extremes. We also need to consider drug interaction and safety. Doses given are not a measure of safety.
This is the process by which I measure medical interventions, simply change the drug and add a very high weight to the pharma advertising and promotion. I have found so many drug and other medical interventions lacking as to generate the big question: Who does the doctor work for, me or the drug company/corporate interest they represent.
The real sadness is doctors are unwilling to give up on their ideas of good medical practice in favor of this corporate model. My 83 year old father is fighting his doctor about having another colonoscopy and driving his blood pressure lower when he has had no colon problems and was admitted to a local hospital with no pulse due to over medication.
Doctors need to not look at the big numbers, but the singular patient in front of them, and then decide on the minimal medications and test necessary for them to live the best life they can live given their condition.
Steve Lucas
I’m not sure I can see any statistical validation in a previously asymptomatic 83-year old having a routine colonoscopy. Even if it were positive, the treatment or something else would likely be fatal before colon cancer was.
Lo and behold the US Preventative Task Force Agrees with me:
http://www.uspreventiveservicestaskforce.org/uspstf/uspscolo.htm
Send your Dad that link and have him show it to the doctor. If he won’t listen to reason, slip Joan Rivers into the conversation.
Mickey,
Thanks for your comment.
Steve,
I am sorry about your father. I agree with your position about the number needed to treat being so high for too many recommended interventions. I remind myself that truly useful developments did not need RTC’s to show their effectiveness, e.g. fire, the wheel, penicillin, clean water, …
Thanks to all.
Steve Lucas
Excellent idea, of course, Dr. Mickey: “a specialty or an agency that evaluates all of this minute to minute reporting of preventive medicine advice and tries to separates the wheat from the chaff…”
I had been taking one low-dose aspirin a day for years with no problems. Then, I took Xarelto for 6 months following an ablation for lone afib. After that, my cardiologist directed me to take one 325mg aspirin a day for 6 months.
A month after I started 325mg/day aspirin, I started having digestive problems. These increased in frequency and severity over the 6 months. When I went off 325mg/day aspirin, I had frequent gut pain.
Over the 6 months, I also developed extensive food intolerance which requires a very, very restricted diet, possibly for years. I can’t eat in restaurants. While I’m looking forward to Thanksgiving dinner today, I will have to be very careful not to eat any trigger foods — turkey, potatoes, and squash are okay (no gravy, nothing spiced, nothing aged or fermented, etc.).
Now I have a gastroenterologist on my payroll, too.
Cardiology frowns on prescribing 325mg aspirin for clot and stroke prevention, it’s been found to be unreliable for that purpose, and the risk of gut damage is considered to be too high for the benefit. But the decision is left to the individual clinician.
Gastroenterologists hate the idea of regular aspirin. Even 81mg daily has been found to frequently cause gut erosion.
This is one situation where I put my trust in my doctor — he’s an excellent electrophysiologist — but I shouldn’t have. I should have done my homework to stay ahead of the game, and I’ll be paying the price for many months.
I would be interested in comments on this study:
http://news.yahoo.com/elderly-cholesterol-lowering-statins-us-study-143822771.html
Haven’t found the NTT for this study but would be curious if it existed and what it was since basically people are being told to take statins strictly based on their age.