a must-read article!…

Posted on Wednesday 28 January 2015

An Internal Medicine residency in a charity hospital in Memphis Tennessee in the 1960s was an encounter with Hypertension [High Blood Pressure] of the first kind. The patient population was weighted towards African Americans and we served not only the urban poor of Memphis, but also the rural areas of West Tennessee, Eastern Arkansas, and the Northern Mississippi Delta. If you wanted to learn about Hypertensive Disease, it was the place to be. Besides the ethnic demographic, there was another factor. Not to far from the hospital, there was a huge supermarket that served the same region that was open 24 hours of every day. The dried bean section was stacked floor to ceiling with 25# sacks of White Beans, Pintos, and Black-Eyed Peas. The produce department was mostly greens – Collards, Mustards, Turnip Greens. In the Old South, social status was determined by what part of the pig you ate ["High on the Hog"], and that tradition was apparent at the meat counter: a few hams, butts, and shoulders in a small section at the end, but mostly every other part of a pig, all salt cured – Tails, Feet, Ears, Knuckles, Hocks, Fatback, all thickly encrusted with salt [that you could taste just standing next to the cases]. It was the flavoring for those beans ‘n greens. So it wasn’t just ethnicity working on the Blood Pressure…

  • Malignant Hypertension: The patients would arrive in the ER delirious or in coma with outrageous blood pressure readings. They would have retinal hemorrhages and other vascular eye signs. Kidney function would be compromised. A "stroke" was eminent if it hadn’t happened already. It was a medical emergency, and at the time, the drug of choice was parenteral Reserpine, which did the job. The easiest transition was to oral Reserpine, but the problem was that some fraction of the patients developed a profound melancholic depression [one of the phenomena that lead to the "catecholamine hypothesis" of depression]. So we used the other available drugs of the time – all with plenty of side effects as part of the package.
  • Hypertensive Cardiovascular Disease: These patients were more common. They showed up in the clinics and ER with congestive heart failure of varying intensity: swollen legs, enlarged livers, shortness of breath, sleeping sitting up – with enlarged hearts and Left Ventricular Hypertrophy on EKG [big muscular hearts]. The symptoms cleared with lowering the blood pressure, but they usually got digitalis and diuretics as well.
  • Hypertension: Back then, Hypertension [asymptomatic] was defined as a Diastolic Blood Pressure consistently over 100 mmHg. And there was plenty enough of that around to treat where I was.
My next way station was an Air Force Hospital in the UK – a couple of bases populated with a very different, predomiantly Caucasian group [also younger][also no supermarket like the one in Memphis]. I never saw a case of Malignant Hypertension and not many with Hypertensive Cardiovascular Disease. It was around that time that Treatment Guidelines were beginning to come from the various specialty organizations. The American Heart Association came out with new guidelines for the treatment of High Blood Pressure, and lowered the definition to a Diastolic of 90 mmHg. I didn’t like it. The drugs made people impotent, feel bad. I gave them BP Cuffs to take home. That cured a lot of > 90 mmHg cases ["white coat" syndrome]. I preached weight loss and against salt before tasting. That cured even more. I got a wide BP cuff for fat arms. That was a winner too. I just didn’t feel right treating only a number – making people sick [but I felt guilty when I didn’t]. After that, I was back in Atlanta at a Charity Hospital similar to the one in Memphis training in Psychiatry. When a delirious patient got triaged to the psych floor, the first thing I did was a BP and rolled the Malignant Hypertension cases back downstairs myself to make sure they were seen quickly. Later, when I ran the psych ER, I convinced the medical ER to do a BP before triag on obtunded patients.

And then I forgot about it. Psychotherapist types aren’t in the BP business. And then I got old, and all my friends [and my wife] were on something for BP. Hers was easy. A home BP cuff cured her [70s and 80s]. One day, I thought seriously about it. I was tired of feeling guilty for insisting on hard evidence before following the guidelines [now 140/90, with a Diastolic BP 80-89 as borderline or pre-hypertension]. But what I realized is that there were several things about this that were important. First, I just didn’t believe it. I read all the long term studies, but I wasn’t impressed. Second, I was a treat-the-sick doctor, not a healthy-lifestyle doctor. Had I stayed in Internal Medicine, I guess I would’ve hired me a healthy-lifestyle nurse practitioner to specialize in seeing to that side of things. It just wasn’t in me to spend a lot of time keeping up with the gajillion guidelines that now pour out of our t.v. sets and journals, and I’m not sure I believed what I read a lot of the time anyway. My point is not about your BP meds, or my rebellious streak, it’s about the limits of population studies, statistics, and the relief I felt reading this article:
New York Times
4 heart attacks are not prevented. When 2,000 People Take a Daily Aspirin for Two Years: 1 Heart Attack is Prevented. People at risk for a first heart attack are often recommended to take aspirin daily to prevent it. Only a very few will actually see this benefit and there’s no way to know in advance who.
By Austin Frakt and Aaron E. Carroll
JAN. 26, 2015

In his State of the Union address last week, President Obama encouraged the development of “precision medicine,” which would tailor treatments based on individuals’ genetics or physiology. This is an effort to improve medical care’s effectiveness, which might cause some to wonder: Don’t we already have effective drugs and treatments? In truth, medical care is often far less effective than most believe. Just because you took some medicine for an illness and became well again, it doesn’t necessarily mean that the treatment provided the cure.

This fundamental lesson is conveyed by a metric known as the number needed to treat, or N.N.T. Developed in the 1980s, the N.N.T. tells us how many people must be treated for one person to derive benefit. An N.N.T. of one would mean every person treated improves and every person not treated fails to, which is how we tend to think most therapies work. What may surprise you is that N.N.T.s are often much higher than one. Double- and even triple-digit N.N.T.s are common.

Consider aspirin for heart attack prevention. Based upon both modifiable risk factors like cholesterol level and smoking, and factors that are beyond one’s control, like family history and age, it is possible to calculate the chance that a person will have a first heart attack in the next 10 years. The American Heart Association recommends that people who have more than a 10 percent chance take a daily aspirin to avoid that heart attack.

How effective is aspirin for that aim? According to clinical trials, if about 2,000 people follow these guidelines over a two-year period, one additional first heart attack will be prevented. That doesn’t mean the 1,999 other people have heart attacks. The fact is, on average about 3.6 of them would have a first heart attack regardless of whether they took the aspirin. Even more important, 1,995.4 people would never have a heart attack whether or not they took aspirin. Only one person is actually affected by aspirin. If he takes it, the number of people who remain heart attack-free rises to 1996.4. If he doesn’t, the number remains 1995.4. But for 1,999 of the 2,000 people, aspirin doesn’t make any difference at all.

Of course, nobody knows if they’re the lucky one for whom aspirin is helpful. So, if aspirin is cheap and doesn’t cause much harm, it might be worth taking, even if the chances of benefit are small. But this already reflects a trade-off we rarely consider rationally. [And many treatments do cause harm. There is a complementary metric known as the number needed to harm, or N.N.H., which says that if that number of people are treated, one additional person will have a specific negative outcome. For some treatments, N.N.T. can be higher than the number needed to harm, indicating more people are harmed than successfully treated.]

Not all N.N.T.s are as high as aspirin’s for heart attacks, but many are higher than you might think. A website developed by David Newman, a director of clinical research at Icahn School of Medicine at Mount Sinai hospital, and Dr. Graham Walker, an assistant clinical professor at the University of California, San Francisco, has become a clearinghouse of N.N.T. data, amassed from clinical trials…
I, of course, immediately raced to Dr. Newman’s NNT website, and there it was:
It wasn’t pleasure in confirmation I felt. It was relief. I guess I felt like I was supposed to believe the AHA guidelines and was conflicted that I didn’t. Was I just adverse to giving those make-you-sick meds? Wanting to be a nice guy? So I felt genuine relief that my persona wasn’t in front of my doctoring in this case.

The article is about the simplest of all statistics. If there’s a Clinical Trial where 40% respond to placebo and 50% respond to the drug, then the NNT = 1 ÷ (0.50 – 0.40) = 10. That literally means "you have to treat 10 cases to get 1 responder." Actually, 5 would respond, but only 1 would be because of the treatment. The other 4 were going to respond anyway. What could be simpler? And what could be more telling? The NNT is one of a family of measures of the Strength of the Effect of a treatment [see an anatomy of a deceit 3…].

I’m a Stats-Savvy type, but I never heard of these Strength of Effect measurements until I started to look at RCTs a few years ago and a mentor pointed me to them. In a modern world, an RCT that doesn’t report a Strength of Effect index right there next to the p value is suspect of hiding something. This is a must-read article!

COI Statement: My healthy-lifestyle spouse has been out of town for a week and a half. For supper tonight, I had a NY Strip Steak, and there’s a Crock-Pot on the counter with navy beans, onions, and liberal rashers of salty bacon for tomorrow. Culture is hard to transcend!
    January 29, 2015 | 6:12 AM

    Dr. Newman’s NNT website is a great initiative I discovered by chance a few months ago. But there is no Psychiatry section. I wrote to them to ask why (even offered my modest help to create one, as a psychiatrist with some experience in research), but I have had no answer yet. If they received more questions in the same direction, perhaps they would engage… What do you think?

    January 29, 2015 | 8:13 AM


    As one who feels the over prescription of meds transcends psychiatry, this is one of my favorite blog entries. I might have missed this but is there an ideal NNT/NNH ratio that should be an ideal goal?



    James O'Brien, M.D.
    January 29, 2015 | 8:51 AM

    It depends on what the T and the H are. If the T is a life saved and the H is a rash, then you can’t do a straight comparison. If the T is a life saved and the H is a fatal side effect, then it’s apples to apples. That’s why he still gets into the issue of nonnumerical clinical judgment after posting the stats.

    Steve Lucas
    January 29, 2015 | 10:43 AM

    This may be of interest:


    I always find it interesting how doctor’s opinions change after they retire or are dealing with family health issues.

    Steve Lucas

    Bernard Carroll
    January 29, 2015 | 11:00 AM

    Dr. O’Brien’s point about comparing apples with apples is right on target. Moreover, with NNT and NNH the devil is in the details, as always. Being summary clinimetric measures derived from clinical trials, they can be no better than the quality of the primary data as reported. Dr. Mickey has made it clear on this blog that “as reported” sometimes bears only a passing relationship to what is.

    The apples and apples analogy applies also to the clinical context when NNTs are being compared. In the context of acute treatment one wants to see low NNT (high benefit). The NICE regulator in UK has operationally defined that as a NNT of 10 or less. Around 5 or less is nicer. But in the context of preventive trials, NNT may be an order of magnitude greater, while in screening applications NNT may be in the thousands.

    James O'Brien, M.D.
    January 29, 2015 | 2:32 PM

    I have a NNT/NNH question for Dr. Carroll. We hear a lot about Lithium as treatment to prevent suicide in mood disorders. The thing that scares me about that though is the narrow therapeutic index of the drug. Are there any NNT/NNH analyses of that issue?


    January 29, 2015 | 4:42 PM

    I developed gut erosion from daily 325mg enteric-coated aspirin prescribed by my cardiologist after a successful cardiac ablation, even though the evidence for aspirin preventing stroke is very weak.

    After 6 months of the aspirin, I had abdominal pain and extensive food sensitivities. I was quite ill. It took discontinuing the aspirin, a gut healing protocol, and 6 months of a very restricted, boring diet for recovery.

    Later, I found that even 81mg daily aspirin causes gut erosion within 90 days in about 20% of those who take it.

    Bernard Carroll
    January 29, 2015 | 7:36 PM

    To Dr. O’Brien’s questions: from the article you cited one can compute that the NNT for lithium versus placebo in preventing suicide is 40-50 (Figure 2). This effect also held up in studies of just unipolar depression (NNT = 27; Figure 5). These are respectable numbers in the context of prevention, especially considering the low base rate of completed suicides: 0/244 for lithium versus 6/241 for placebo in the head-to-head studies. For a comparison, NNT estimates range from 23-81 for preventing death with 10 years treatment of hypertension. As Dr. Mickey’s post emphasized, one of the strengths of the NNT metric is that it adds nuance to the statement “the treatment works.”

    The other half of your question concerned the low therapeutic index for lithium. Low therapeutic index means toxicity appears at doses/blood levels not far above the therapeutic range. One of the ways we earn our keep is to assume the professional responsibility on behalf of our patients for managing and preventing lithium toxicity. Educating the patient and forming a good therapeutic alliance is essential in this process. Lithium does have irksome side effects – tremor, metallic taste, frequent urination – but the risk of frank lithium toxicity has been exaggerated since the introduction of on-patent competing mood stabilizer drugs (anticonvulsants and atypical antipsychotic drugs, which have their own complement of side effects). Another scare with lithium is chronic renal disease. The section on this topic in the textbook by Goodwin and Jamison is worth reading for perspective here: the scare is greatly overstated.

    Full disclosure: one of my teachers in Melbourne was John Cade, who pioneered the modern clinical use of lithium. Another teacher in Melbourne was Samuel Gershon who later did a great deal to get lithium accepted in the US finally in the early 1970s. We used lithium routinely in Australia through the 1950s and 1960s. I established one of the earliest lithium clinics in the US at The University of Michigan in 1973. We treated hundreds of patients and we had no major toxicity incidents. It is important, however, that we physicians call the shots on how often patients are seen rather than be constrained by some management directive.

    James O'Brien, M.D.
    January 29, 2015 | 8:09 PM

    Wow, great answer and more than I bargained for. I knew the Australian would come through on this topic.

    The biggest side effect I see with patients is weight gain. I don’t worry about renal failure as much as metabolic syndrome.

    Do you have the reference link to those stats? Thanks again.

    James O'Brien, M.D.
    January 29, 2015 | 8:10 PM

    Sorry, you were referencing my link, I see the numbers.

    January 29, 2015 | 8:29 PM

    Hmm, one could infer from these above comments here that my earlier point of view, that maybe we have maxed out psychopharmacology years ago and that we need to relook at the standards like Lithium, MAOIs, and the first gen antipsychotics. I know the people who’s income streams depend on the likes of Lilly, Astra Zeneca, and these days Abilify’s kingpin, er, developer, don’t want conversations like this, but, hell, I don’t care who I offend when it comes to advocacy for patient care!

    Placebo and poisons, sad so many miss that latter half to the pharmacology equation in place for decades and decades…

    James O'Brien, M.D.
    January 29, 2015 | 8:53 PM

    I’m firmly convinced MAOIs are better than SSRIs especially for resistant depression. I think the caveat that there is difference in efficacy between the old and new antidepressants was the influence of DSM3 and not real science.

    The one thing about the lithium data above is there is not one death from lithium overdose which may not be representative. I’d like to run the number on some other studies.

    Bernard Carroll
    January 29, 2015 | 9:19 PM

    Well, in the overall data there were 4 suicides in 893 patients taking lithium. Those 4 occurred in the trials where lithium was compared to another drug rather than to placebo (Figure 2). The comparison is 6 suicides in 241 patients taking placebo. From these numbers the upper boundary of the NNT is 50, as I mentioned earlier. I think in your previous comment you may have been conflating lithium overdose with suicide while taking lithium.

    James O'Brien, M.D.
    January 29, 2015 | 10:27 PM

    Good point. Suicide vs accidental overdose is important distinction. Suicide by any means is apples to apples, accident OD is another H. Yes, I was aiming to ascertain accident ODs as I assumed suicide by the drug was subsumed under the original number. I agree NNT 50 is pretty decent for prevention. I suppose another hypothetical source of fatal H would be weight gain leading to a fatal medical outcome but that would be hard to quantify in any study. Lots of nuances.

    Mark Hochhauser, Ph.D.
    January 30, 2015 | 7:05 AM

    When I went to Newman’s NNT website, I was struck by the caveats for the study he presented. The data came from some studies that perhaps shouldn’t have been published, yet it’s all uncritically “acceptable.” Plus, NNTs don’t come with a margin of error, or a standard deviation, or an acknowledgement that NNT is an estimate, as in ENNT. I’m troubled by how researchers can be so completely confident in a number based on some seriously flawed studies.

    January 30, 2015 | 8:44 AM

    I’ve been able to start a conversation in twitter with two of the developers of Dr. Newman’s NNT. They seem quite interested in developing a psychiatry section and ask for topics to be tackled. Any suggestions?

    James O'Brien, M.D.
    January 30, 2015 | 12:05 PM

    What we talked about in another thread…SSRI/SNRIs.

    Doctor Z
    January 30, 2015 | 12:21 PM

    Agree that a daily aspirin is unlikely to prevent a heart attack, but I wonder how often it may prevent a stroke following a heart attack?

    James O'Brien, M.D.
    January 30, 2015 | 2:30 PM
    Bernard Carroll
    January 30, 2015 | 3:09 PM

    To Dr. Hochhauser concerning confidence intervals for the NNT: there are standard methods of calculating the NNT confidence intervals, though I agree they are not routinely presented. It is becoming more common to see them, however. Here is a useful place to start – and check out some of the references in this article, too.

    January 30, 2015 | 9:08 PM

    Funny, I was prescribed lithium for three years, by three doctors, and none of them, said “boo” about blood tests. My V.A. psychiatrist too me off it, evidently the doses were very high. My kidneys thank him.

    I got my first computer in 1998 and the best use of it for me is researching pharmaceutical drugs and learning what I’ve learned here. I’m just saying “no” to a lot of offers these days, and keep finding that my condition improves without drugs that were prescribed with little to no information. Though I’m quite grateful for the drugs I do need and do benefit from, I’m glad to be on the tail end of ending my last prescribing cascade and am ready to have a very frank talk with my neurologist on limiting our visits to once a year and not doing routine MRIs. They can’t tell me anything I don’t know already know and cannot be used to identify any particular relationship between a lesion and reduced ability save a lesion on the optic nerve which can be spotted by an ophthalmologist.

    Over-medicalizing, over-diagnosing, and over-prescribing are very bad habits. I saw an article recently challenging yearly exams, and for most people, I think that’s a swell idea. Though I do like the idea of the Stanford 21, my best friend out his liver was enlarged from a doctor who did a hands on exam.

    Our healthcare system is so busy with busy work that it’s difficult to get a timely appointment when a pressing health problem does present itself.

    I’m rooting for a saner and more healthy health care system, like everyone here, and seeing numbers put into their place so that individuals can receive the treatment they need.

    Gad Mayer
    January 31, 2015 | 4:12 AM

    @NabuccoDream: Interesting; I had presumed that they avoided psychiatry because of the complexity and non-dichotomous nature of many of its outcome variables. A preliminary, off-hand list for them could include:
    -remission from acute depressive episode with SSRI/TCA/MAOi
    -relapse prevention with the above
    -remission from manic and depressive episodes in bipolar disorder with different medications
    -relapse prevention with the above
    -remission from acute psychosis in schizophrenia with APs
    – relapse prevention with the above
    Notice I would choose to calculate NNTs for remission and not for response, as the the definition of response is very inconsistent in psychiatric research.

    January 31, 2015 | 4:28 PM

    Thanks a lot Gad! I also thought that they were avoiding the section due to the intrinsic “problems” of Psychiatry (and it’s probable that, when facing them, they just quit. I have tried to warn them, haha!).

    In my opinion, it is very important to put a psychiatry section in there, even if it is only to say what we already know: “current evidence is rubbish and we cannot generate any useful information. More studies, with precisely defined outcomes and unflawed methodology/reports, are needed”. If you take a look at Cochrane’s systematic reviews for psychiatric drugs, that’s the kind of conclusion you will find almost without exception. But it has to be said, loud and clear, again and again.

    The remission / response point is a good one, and that is the kind of thing that needs to be considered. I will approach them with that. If you would like to contact them directly, I am sure they will appreciate.

Sorry, the comment form is closed at this time.