The politicization of science

I blogged Wednesday that I thought I smelled politics in the JAMA article on fetal pain and suggested that the authors should have disclosed their political leanings as possible conflicts of interest. Yesterday the Chicago Tribune addressed the issue nicely (thanks to Kevin for directing us to the article).

The author information at the end of the paper says “Financial Disclosures: None reported.” Although no political leanings were disclosed to the JAMA readers one of the authors let her views slip in these comments to the New York Times: Dr. Eleanor A. Drey, one of Dr. Rosen's co-authors, said that as an obstetrician who sometimes performs abortions, she would find it troubling to be compelled to bring up the subject of fetal pain with her patients. "I would be forced to drag them through potentially a lot of misinformation," she said.

We’ve heard howls of indignation about conflict of interest in research until we’re numb, but the outrage has been selective, directed primarily at the financial influence of the pharmaceutical companies. Let’s not forget that political influence and advocacy are also forces that can have corrupting effects on scientific inquiry. Although the JAMA paper may not inform us about fetal pain it will have the beneficial unintended consequence of sparking discussion on an under appreciated form of conflict of interest.

It would be unreasonable to summarily reject a scientific paper based on source or affiliation, be it political or financial. However, disclosure of real or potential conflicts of interest is essential as it gives readers information they need for critical appraisal.

Why study homeopathy----

when it has no physiologic or pharmacologic plausibility in the first place? This is a distortion of evidence based medicine which takes empiricism to an extreme. Straight from the Doc, Medpundit and Medicine and Man have already pointed us to the Lancet article on the subject.

This overview from Homeowatch outlines the field’s history and states that homeopathy was harmless compared to prevailing nineteenth century conventional medical practices, a fact which drove its popularity. It points out the utter irrationality of the theory, which is that a “spirit like essence” remains behind in the water even after all molecules of active substance have been diluted out.

Total lack of biologic plausibility is one reason not to bother to study homeopathy. Another reason, as has already been pointed out by the other bloggers, is that the adherents won’t believe the studies. At least some don’t believe a randomized placebo controlled trial is the proper methodology.

If you want to debunk something, rather than spend all that money on studies, why not just call Randi the magician?

Fetal Pain

In the latest issue of JAMA is an article about fetal pain which makes the case that abortion providers should not have to inform women about the possibility of procedural pain in the fetus, and purports to back up the assertion with science. Go read what Medpundit has to say about this. She has summed it up better than I can although I would offer a few observations and questions.

The authors’ major premise is that pain is a psychological construct. Indeed, the premise sets the evidentiary standard for the rest of the paper, because, in the authors’ own words, it “presupposes” the neuroanatomic connections they believe must be proven to make the case for fetal pain. Conveniently enough, the presupposition makes tissue injury irrelevant. (What if physicians start telling patients their pain is psychogenic)?

JCAHCO had some things to say about this in its pain management initiative of a few years ago. 1) Patients have a right to assessment and management of pain (implicitly surrogates have the right to be informed about pain). 2) Observation of vital signs, crying and reflex withdrawal (which have been observed in premature infants of less than 29 weeks gestation, and which the JAMA authors seem to dismiss as non-evidence of fetal pain) are legitimate pain assessment tools in preverbal individuals.

The authors admit that the evidence does not establish at what point in fetal development the thalamo-cortical pain pathways become functional.

The first sentence of the abstract establishes the politics driving this “science.” Perhaps the authors’ conflict of interest disclosure should have included their political affiliations. The issue of interface between scientific publication and politics has surfaced previously in JAMA and caused trouble for an editor.

Medication safety across the continuum of care

Hospitals have long been concerned about medication errors during the flow of in patient care. Until recently, however, relatively little attention has been paid to the problem of errors at the points of transition between in patient and out patient care. Many hospitalized patients are on a large number of chronic medications. The transition points at admission and discharge are especially vulnerable for such patients because of the need to communicate and document complex medication history. At admission the patient may provide faulty or fragmentary information. At discharge the patient may not understand the doctor’s instructions. Hospitalization often results in one or more changes in the patient’s chronic medication. When medication orders are written in the hospital and when the patient is instructed at discharge these orders and instructions must take into account the patient’s pre-hospital medications and dosages. This is what the Joint Commission on Accreditation of Healthcare Organizations in its 2005 patient safety initiatives terms reconciliation of medications across the continuum of care. This issue was reviewed in an article and an accompanying editorial in the current issue of the American Journal of Health-system Pharmacy. (Open access until September 1 05).

It’s good to see this issue getting the attention it deserves. Accurate transfer of information at the transition of care seems to be an elusive quality goal. Deficiencies in the process are what we hospitalists often refer to as the “information voltage drop.”

More about evidence based patient decisions

I blogged about this recently. Among other things I pointed out that an essential component of evidence based medicine (EBM) was involvement of the patient in the decision process by providing quality information. Although this important step is often ignored in discussions of EBM it was emphasized as a defining characteristic in the classic paper of David Sackett et.al. entitled “Evidence based medicine: what it is and what it isn’t.” The authors state that the process, among other things, involves “------the more thoughtful identification and compassionate use of individual patients' predicaments, rights, and preferences in making clinical decisions about their care.”

I discussed the concept of number needed to treat (NNT) as an understandable way to discuss the magnitude of treatment effects with patients. Then I ran across this recent study. It seems patients have difficulty with the concept. Decisions were influenced only marginally, if at all, by the NNT. In addition, 24% of patients who made a decision after being informed of the NNT changed their minds following additional explanation of the meaning of NNT. In almost all cases better understanding of the concept caused patients to reject the treatment in question. Worse, this paper from the August 1 issue of American Family Physician paints a dim picture of patients’ health literacy.

This is sobering. In my previous post I illustrated that the process was time consuming and challenging. These papers suggest that it’s even more challenging than we had thought.

Diagnosis of pulmonary embolism: Are we there yet?

A systematic review in the April 27 issue of JAMA was blogged about by Inner Visions on May 8, and has more recently been given the status of POEM as abstracted in the current issue of Cleveland Clinic Journal of Medicine. The take home message was that a normal CT done with the appropriate technique equals pulmonary angiography (PAG) for the exclusion of pulmonary embolism (PE). This sounds simple enough, but questions remain about how this evidence changes our diagnostic approach. What is the initial test of choice? Does CT replace nuclear ventilation and perfusion (V/Q) scanning?

This paper doesn’t answer these questions. Such questions require a broader perspective which takes into account the results of PIOPED I and PIOPED II. The JAMA review evaluated the performance of CT by the standard of freedom from events over three month follow up and compared those results with similar data previously reported for PAG. PIOPED I evaluated V/Q scanning against the reference standard of PAG. PIOPED II evaluated CT against a multi-dimensional diagnostic strategy using non invasive tests and occasional use of PAG. The multi-dimensional strategy was validated as a reference standard by clinical follow up for freedom from events. PIOPED II has not been published, but its results were presented at the Radiological Society of North America 90th Scientific Assembly and Annual Meeting late last year, abstracted here at Medscape (free registration required).

While consideration of all this evidence together with cost and patient tolerability provides no simple answers, I believe the following statements are supported.

1) Invasive PAG is seldom indicated.
2) A normal V/Q scan has the best negative predictive value (NPV) of all tests (note I said normal, not “low probability”). (In PIOPED I no patients with normal V/Q had PE!).
3) When CT results are discordant with clinical probability the PPV and NPV are poor. This was a significant finding in PIOPED II, similar to the findings for discordant V/Q results in PIOPED I. This means that, just as with V/Q scanning, results of CT must be integrated with other clinical data.
4) A CT scan costs more than a V/Q scan. [1].
5) There is no clear modality of choice for initial testing. Careful clinical assessment determines the choice of the initial test(s) which might include D-dimer, leg compression ultrasonography, V/Q scanning or CT.

This POEM adds to our knowledge of testing for PE but it’s just one piece of the puzzle. Be careful of simplistic interpretations.

Who has time for evidence based medicine?

I recently blogged an ideal description of evidence based medicine. I gave examples to point out that it’s onerous, and I implicitly questioned whether it’s realistic. Truth be told I wonder if many doctors even practice it. A commenter did a better job of making the point. “My goodness. If my physician went to all that trouble for me, I'd require resuscitation from the shock. But to tell you the truth, such extreme conscientiousness -- while I applaud it -- is not only impractical in terms of time (yours!) but also overestimates by a long shot what most patients would expect in terms of empowerment.” (Go read the comment in its entirety).

EBM is indeed rigorous. Formulating a question, translating the question into a search strategy, performing the search and critically appraising what you retrieve is a substantial chore even with the help of the short cuts provided by some medical Internet resources. Of course, your job isn’t finished there. You then must evaluate how well the patient populations represented in the evidence you’ve retrieved match the unique characteristics of your patient and, finally, integrate this evidence with your patient’s preferences, circumstances and values.

Most of us would say we practice EBM but the reality is we often practice only the trappings of EBM. That is, we can quote from randomized controlled trials and talk about meta-analyses, odds ratios and confidence intervals. We can name off a number of things doctors who practice EBM are supposed to do, and to a variable degree, do them. We use Internet resources but often not in a systematic way. With good intentions we put on these appearances and often still don’t really practice EBM as we’re supposed to.

What we have then is a degree of disconnect between the theory and the practice of a revolutionary idea in medicine. This is not unprecedented. It is analogous to another revolution in medical thinking that occurred in the 1960s: the problem oriented medical record, a.k.a. the Weed system. It was the talk of the wards when I was a medical student. Since those days SOAP formatted progress notes have become commonplace, and most medical records have had a problem list, after a fashion. However, this is a far cry from the true practice of the Weed system as it was originally defined and taught. Again, we practiced only the trappings. (In researching for this post I was unable to find any of the original descriptions of the Weed system to provide as links. The articles are old and only the citations are on line. I was able to find a print description in an old copy of Hurst’s The Heart, sixth ed., which contains a description on p. 105. J. Willis Hurst, professor and chairman of the Department of Medicine at Emory, was a champion for the Weed system [1] [2]. Older editions of his text contain elegant descriptions of the problem oriented medical record as applied to clinical practice, which time and space constraints do not permit me to abstract here).

We miss the mark with EBM, in part because we lack time. Reimbursement incentives put doctors under pressure to see large numbers of patients in limited time, and consequently don’t reward the practice of EBM. How then should we regard it? We should embrace it. If the ideal is unattainable it should challenge, not discourage us. It is a challenge I find fascinating, and one that I approach with the hope that as we ascend the learning curve and information resources improve, things will get better and better.

EBM has many dimensions which fascinate me, and I’ll be blogging about it a great deal. Stay tuned.

How and why do doctors read?

It occurs to me that the ways we read medical literature fall into at least two categories. First there’s the type of reading we do to practice evidence based medicine (EBM). That type of reading focuses on the clinical question at hand as it applies to a particular patient. We often use the PICO acronym to help us formulate the question. (In a Population with my patient’s attributes, what is the effect of the Intervention in question versus a Comparison group with respect to Outcomes?). Such a question forms the basis for a search that will retrieve relevant articles. This highly focused and specific type of reading has been variously termed question driven reading, problem based reading and foreground question reading, and often relies on POEMs. It is often done “at the point of care”, e.g. the physician’s office, the exam room or the hospital ward. The reading material might consist of abstracts of randomized controlled trials, a systematic review, or topic summaries from a filtered source such as Bandolier, DARE, TRIP, or Up to Date. This is a relatively new type of reading, not possible before the availability of the Internet and the vast repository of searchable medical databases.

Another type of reading is the more traditional type which, although often patient based, is not as focused. It is sometimes referred to as background question reading. I have recently heard the somewhat derogatory term “reading for the heck of it” to describe this type of reading. Such reading often draws on DOE (disease oriented evidence). It is more likely to take place at home, and the topic might be related to a patient you saw that day, or that week. The source could be a textbook or a narrative review. It would have a broader scope than the former type of reading, and might encompass the clinical features, epidemiology and pathophysiology of a disease.

So why do I make a fuss about this? Well, it seems there’s a ground swell of opinion these days against background reading. This article encourages students to do problem based searching of filtered resources, but not to bother with standard medical journals! This tutorial on EBM warns the reader that DOE is worthless if not dangerous. I’m reading more and more opinions of this sort, many taking aim against narrative reviews, textbooks, experts and the practice of reading about pathophysiology. The purveyors are purportedly doing this to advance the cause of EBM. I would argue that this reflects a narrow view of EBM. The classic paper by Sackett, et. al., which many recognize as a defining article on EBM, clearly points out that expertise and judgment (which I believe derive from background reading) are important components of the process.

I believe both types of reading are essential. Without foreground reading we can’t bring the best and most current evidence to bear on the individual patient’s problem. Without background reading we lack depth of understanding, and can never cultivate the expertise and judgment so essential to the rational and prudent application of evidence.

Now excuse me while I go read a chapter from Harrison’s.

Do drug company promotions impact clinical outcomes? Is there a way to know? Does anybody want to know?

My recent post on the effects of drug industry marketing drew a difficult and insightful question: “RW, how do you propose to measure patient outcomes?”
Can outcome based research be done, or must surrogate evidence suffice? Perhaps patient outcomes in a health care setting (a clinic, an academic medical service, an HMO, or a hospital) could be measured before and after implementation of a “no sample” or a “no drug rep” policy. Is there precedent for this type of study? Yes. There are numerous examples of health system, health care environment, and cultural changes that have been studied for their effects on outcomes including the adoption of heart failure multidisciplinary management programs [2]

Detailing, samples, prescribing, and outcomes

Kevin MD and Blogborygmi cited this article about the influence of drug sampling on residents’ prescribing behavior. There have been numerous articles of this sort, showing various effects of detailing and sampling by pharmaceutical companies, and considerable controversy has been fueled. Many of the citations have been compiled at No Free Lunch and Healthy Skepticism, organizations which are critical of pharmaceutical industry marketing. Some would like to see marketing disappear as we now know it, and are calling on physicians to refuse to accept literature or even small gifts from drug sales representatives. Others would reduce or curtail the pharmaceutical industry’s support of physician CME.

Most all the critics cite the studies as evidence to back up their recommendation to drastically change the landscape of drug company promotion. But is this evidence strong enough to warrant such a recommendation? Check the above web sites, link to the articles and decide for yourself. At the risk of oversimplification the gist of the studies is that physicians’ prescribing behavior is influenced by pharmaceutical industry promotion. That is, marketing works. Some of the studies suggest that the promotions even influence doctors to prescribe outside established guidelines.

But as I browse this extensive collection of links I discover a weakness. I can not find one study about patient outcomes. Only the surrogate endpoints are examined. So, we can learn a lot from these studies about the effect of marketing on doctors’ attitudes and prescribing behaviors, and it’s all very interesting---indeed fascinating---but what about the clinical bottom line? Unfortunately, research to date provides no answers.

Our adventure in evidence-based medicine has taught us about the hazard of reliance on surrogate endpoints. In the 1970s we knew that surrogate hemodynamic endpoints in heart failure were adversely affected by beta blockade, and were taught that beta blockers were bad for heart failure. Only after years of outcome based studies did we realize we were wrong. We also knew that ventricular arrhythmias were harmful, and that since the surrogate endpoint of frequency of ventricular arrhythmias could be decreased by treatment with Class I-C antiarrhythmic drugs, such treatment must be a good thing. Not until completion of the first adequately powered outcome based study, CAST, did we know how tragic such thinking was.

So while the science concerning the effect of drug company promotion is soft, many observers still think the impact on patient outcomes is negative. Although promotions are certainly biased and the material can be deceptive, the information doctors receive is a mixture of good and bad. Must we assume that the bad influence outweighs the good, or is the reverse equally plausible? We must resist simplistic assumptions. Out in the real world varied and sometimes opposing factors are at play. For example, although Vioxx was heavily promoted and associated with harm there are other examples in which drug companies promote numerous beneficial therapies that are evidence-based and known to be under-utilized such as beta blockers and ACE inhibitors for heart failure and low molecular weight heparins for deep vein thrombosis prophylaxis [1] [2] [3]. [4] [5] [6] [7] Who knows how it all balances out?

I don’t rely on drug reps or promotional material to keep current, nor should anyone. There’s no need. Present day information technology makes it unnecessary. We can easily get good quality unbiased information at the point of care. However, a sweeping prohibition against drug reps, samples and industry supported CME is not supported by high level evidence. The jury is still out.

Patient participation in medical decisions: Is it evidence-based?

Yesterday DB of DB’s Medical Rants blogged about patients participating in treatment decisions, citing this article in the Journal of General Internal Medicine. Interestingly, the study found that about half of patients surveyed would rather the physician make final decisions. He makes the point, and I agree, that we should allow patients to be involved to the extent they wish. But I would caution that unless we apply rigorous principles of evidence-based medicine to the discussion with the patient, the decision making process becomes flawed. So if patients’ decisions need to be evidence-based just as doctors’ decisions do how do we bring that about? Is there enough time in our busy practices to do it?

Consider these case examples. You are counselling your patient about upcoming knee replacement. You plan to give enoxeparin for DVT prophylaxis. How do you inform the patient? Sure, you know that it’s the right thing to do and all the experts recommend it, but the patient deserves evidence. So, you do a Pub Med search (or, somewhat more easily, consult a filtered resource such as Up To Date) and cull out the studies you critically appraise as valid and applicable to your patient. In order to advise your patient on the magnitude of benefit of the proposed treatment you look for, or calculate the absolute risk reduction (ARR) for the proposed treatment as well as the absolute risk increase (ARI) for bleeding. In order to translate this into language the patient can understand you then, from the absolute risk reduction for DVT and the absolute risk increase for bleeding, calculate the number needed to treat (NNT) and number needed to harm (NNH) respectively. You then have the discussion with the patient, modifying it as necessary to take into account any unique attributes of your patient which might increase bleeding risk. If the patient is to be discharged early after the surgery you must also provide information concerning the cost of continued enoxeparin at home, taking into account the patient’s financial condition and payer sources. Then you ask the patient and any concerned family members if they have questions, answer them as they arise, and document your discussion in the medical record. Time consuming, eh?

Of course if this is not your first patient to receive enoxeparin for orthopedic prophylaxis maybe you’ve done the drill before, but the discussion still takes a lot of time. And your next patient may present a clinical problem you haven’t researched in the last couple of months. Let’s say you’re discharging a patient after an acute coronary syndrome (ACS). You want to put the patient on a statin drug because you know it’s a good thing to do and you are going to explain it to the patient. To merely say “this is good for you” or “here, take this pill because it will decrease your risk of another heart attack” might sound paternalistic in this day of medical consumerism. So, again, you must present the patient with evidence. You do the search and find a raft of studies, from which you choose the ones of the highest level of evidence and applicability to your patient. As you critically appraise the studies you ask how the study in question applies to your patient. How do the lipid entry criteria and average lipid levels of the study patients match up with your patient’s lipid results? What is the number needed to treat and the number needed to harm (your patient tells you a friend of hers heard that cholesterol lowering medications could destroy the kidneys)?

Your next patient arrives from the nursing home with severe sepsis. As multiple concerned family members attend the patient and ask numerous questions you are considering whether to give activated protein C. You are aware of the positive results of the PROWESS study, but are concerned about the risks of serious bleeding. You are aware of the strict and very detailed inclusion criteria for patient selection, and you’ve heard of some recent controversy that has arisen concerning the use of activated protein C, and you need to explain it all to the family. You are already behind schedule after the long discussion with your previous patient about statin drugs.

What’s a doctor to do? I have no pat answer. Our patients deserve to be included in decision making but if they participate they must have the high quality and specificity of information they need to make valid choices.

An under-appreciated cause of dilated cardiomyopathy

The current issue of Annals of Internal Medicine contains a paper highlighting the importance of familial dilated cardiomyopathy. In a series of “idiopathic” dilated cardiomyopathy (DCM) the investigators found evidence of DCM or subclinical abnormalities in left ventricular function in relatives of about one third of patients. Previous estimates have cited similar or even higher prevalence of familial etiology of DCM. This points out that familial DCM is a very important etiology of heart failure and may be the most common cause of idiopathic DCM.

In my experience, when we encounter a new case of DCM we do the standard work up to screen for valve disease, coronary disease, thyroid disease, hemochromatosis as well as toxins (particularly ethanol) and if that screen is negative we say “it must have been a viral myocarditis.” I always wonder if that diagnosis is correct. I have no doubt it exists but I think it tends to be overdiagnosed.

That raises the question of just how common viral myocarditis really is as a cause of DCM. Controversy surrounds the issue of how many patients actually have myocarditis, and how many cases of myocarditis are actually viral. This small series found a viral etiology in around 20% to 30% of DCM, all Coxsackie B.

So, what can we make of this? Here are some key points.
1) Familial cardiomyopathy is a common cause of DCM.
2) Familial DCM is important to recognize because family members may benefit from screening via electrocardiography and echocardiography.
3) Gene testing in DCM should be limited to specific (and infrequent) situations. This review of the genetics of cardiomyopathy discusses genetic testing and makes some useful practical points.
4) Viral myocarditis is a known cause of cardiomyopathy but its relative importance is controversial. The American College of Cardiology Guidelines state that myocardial biopsy is indicated only in unusual specific circumstances, and the guidelines do not recommend viral serologic testing.

Doctors get bad grades in antithrombotics 101

In the July 11 issue of Archives of Internal Medicine is a survey from the National Anticoagulation Benchmark and Outcomes Report Steering Committee showing poor adherence to guidelines for antithrombotic therapies in a variety of clinical settings. Aspirin for myocardial infarction, warfarin for atrial fibrillation, treatment of venous thromboembolism (VTE) and prophylaxis for orthopedic procedures were evaluated.

Although practice patterns were found lacking in all areas one finding concerning treatment of VTE was particularly interesting. Almost half of patients with acute VTE had unfractionated or low molecular weight heparin discontinued too soon (before the INR was at least 2.0 for two consecutive days). As pointed out in the Archives article, the guidelines call for continuation of some form of heparin (unfractionated or low molecular weight) for at least four days AND until the INR is at least 2.0 for two consecutive days.

I suspect this failure of adherence is driven by pressure to discharge patients quickly and by a popular misconception that patients are adequately anticoagulated with warfarin as soon as the INR reaches 2.0. Such an approach is not only outside the guidelines but also pharmacologically irrational. The vitamin K dependent factors decay at different rates following the initiation of warfarin. Although factor VII, with the shortest half-life, is the major determinant of the INR the patient’s state of anticoagulation is determined by the other factors which have longer half lives. Put another way, when warfarin is initiated the initial rise in INR does not reflect anticoagulation of the patient. Worse yet, protein C, a vitamin K dependent anticoagulant has a short half-life similar to that of factor VII, thus rendering the patient hypercoagulable when the INR first rises. Here is a nice little review of the acute effects of warfarin from the Cleveland Clinic Journal of Medicine.

Atorvastatin in end stage renal disease

This week's New England Journal of Medicine reported a study of atorvastatin treatment for patients with type 2 diabetes undergoing hemodialysis. There was no difference from placebo in the primary composite endpoint, a relative risk of .82 for combined cardiac events, and a relative risk of 2.03 for fatal stroke. There was no difference in non-fatal stroke or all cause mortality.

The media are already hyping this study, citing the “surprising” findings and highlighting a “two fold increase in deadly stroke” due to the statin. So, let’s take a moment for a sober assessment.

The results are disappointing, but how surprising, really? The attributes of the study population suggest that the principal dyslipidemia was the metabolic syndrome, characterized by high triglycerides, low HDL, and presumably small dense LDL particles. (The mean pre-treatment LDL cholesterol was 126 and the mean triglyceride level was around 265). It has long been known that statins are of limited effectiveness in treating the metabolic syndrome. Multiple other atherogenic factors are known to be associated with end stage renal disease including hypercoagulability, elevated levels of lipoprotein (a), elevated homocysteine, elevated angiotensin II, and disturbances in calcium and phosphate metabolism. Such factors are not mitigated by statin treatment. We would not expect LDLC to be the principal player in the vast atherogenic milieu of ESRD.

The finding regarding fatal stroke is unexplained. The authors point out in the discussion section of the paper that it could be due to chance. True to form, the popular media have cited only the more sensational increase in relative risk. The actual absolute excess in stroke mortality, though not trivial, was somewhat less impressive (2%, number needed to harm 50).

An essential tenet of evidence-based medicine is critical appraisal the literature, something we don't usually find in the lay press.