Tag: ebm

Posted on

Epo, anemia and the lack of placebo controlled trials

By a mistake of communication I was scheduled to give the anemia talk at a recent chronic kidney disease symposium. I would have never selected this topic on my own. I work with Robert Provenzano, one of the Gods of Anemia. Fortunately, this was a happy accident. I loved researching and writing this presentation. The whole experience was an important lesson on the value of working out of your comfort areas.

The anemia saga is well known to all nephrologists and is covered in depth in my presentation but let me recap my version of the story.

Life before Epo was pretty bad. the average dialysis patient received a transfusion more years than not. On the Eve of Epo the transfusion was rate was 16% per quarter!

After the release of Epo, the transfusion rate plummets. It falls by two thirds in a year and continues to fall so that the current rate of 0.3% per quarter is a 98% reduction in transfusions. Revolutionary. And this doesn’t even begin to address the quality of life brought to dialysis patients by higher hemoglobins.
I was in college when this happened and it’s a little hard to imagine how exciting that must have been.  The introduction of Epo launched a million observational trials that all pointed in the same direction: 
Where you found higher hemoglobins you found better patient outcomes. 
It almost didn’t matter what outcome you were interested in: hospitalizations, mortality, regression of left ventricular hypertrophy, quality of life, fatigue score. It didn’t matter, everything was better with higher hemoglobins. I suspect the community was seduced by the observational results but for what ever reason the amount of randomized controlled trial data that emerged was laughably small. By 2007, Epo was eating up 1.8 billion Medicare dollars. It should have been the best studied drug and instead it was among the least studied.
I tried to find Henry Paulson’s signature but had to settle for George Bush’s.
Here is the table from K/DOQI anemia recommendations where they summarized all of the RCT data from the birth of Epo through May 2006 (apparently it omits a single 2005 study).
8 Epo vs Epo studies and 3 placebo controlled trials, 1 placebo controlled trials in pediatrics
The paucity of placebo controlled trials is shocking. No one demonstrated that higher hemoglobin targets had a mortality benefit or regressed LVH compared to placebo. This would not be so problematic if the Epo vs Epo studies had been positive, but those too were negative trials. So we are in the awkward position, a quarter century after the introduction of EPO we cannot conclusively state that the drug does any more than reduce transfusions and improve quality of life. All the mortality reductions, cardiovascular disease protection amount to observational-backed hype (one small rct (N=38) study did show a reduction in LVH).
Favorite slide from the deck
Part of me is outraged but another part understands how difficult it would be randomize dialysis patients to placebo. I couldn’t, in good faith, expose trial subjects to the transfusions, fatigue and weakness that being randomized to the placebo-arm would entail. It fails my personal “Grandma test” (i.e. Would you feel comfortable enrolling your grandma in a placebo controlled trial of Epo? No). I understand that and forgive the dialysis researchers; however I am a little disturbed to witness how in the pre-dialysis CKD population the same pattern of relying on observational data. In CKD these is no transfusion epidemic that needed to be derailed, there is no profound fatigue turning patients to zombies. We had an opportunity to do the right studies to figure out if this expensive Nectar of Thousand Oaks really helped. The paucity of placebo controlled and randomized controlled trials in pre-dialysis patients is embarrassing.

Reading this editorial by Marc Pfeffer crystalized this critical mistake. I spent a fair proportion of the presentation laying out how statins evolved from being used in only the sickest patients to larger and larger populations and how at each step placebo controlled trials were used to prove efficacy. Not every step was a win, statins for heart failure failed, but the key is that cardiologists know that statins don’t improve heart failure because they tested it with two (1, 2) placebo-controlled trials.

Why is nephrology lacking the randomized controlled trials that have defined the huge advancement in cardiovascular disease over the last 25 years?

  • It isn’t due to the severity of illness, the first placebo controlled trial of ACEi in heart failure used a cohort with 40% mortality at 6 months. 
  • It isn’t timing, ACEi for heart failure was developed and proven at the same time as the introduction of Epo.
I don’t know the why but I have a couple of theories. One is the richness of the retrospective data in nephrology, brought to us by the USRDS, blinds us to the importance of prospective data. This could explain why are repeatedly burned by observational studies, see: kT/V, vitamin D, and binders.
The other theory is that, unlike statins and ACEi where there are many vendors producing drugs in the class of interest, there is a total monopoly in the field of ESA. You want to increase the hemoglobin you need to buy Epogen. No generic, no competing ESA. In the ACEi market, having enalepril as the drug studied in the CONSENSUS and SOLVD trials paid huge dividends to Merck. Having in-class competition lead to commercial support of critical research. Amgen had no need for this because the US Patent Office gave them a monopoly. A monopoly that seems to last forever. Why is it that every other drug from the late 80’s is generic: omeprazole, captopril, enalepril, benazepril, metformin, simvastatin, etc. Even drugs from the 90’s are now generic: losartan, lansoprazole. 
Epogen stands alone without generic competition. And unfortunately, largely without placebo controlled trials to back-up mountains of hype.
The deck can be found here: Keynote | PDF
Posted on

One of my friends had a heart attack

Pretty stunning. He’s young and, despite some tobacco and a generous BMI, he seems healthy. He eats whole foods and is physically active. He does not have diabetes or hypertension.

Last night he was asking about whether a stress test could have made a difference. Two years prior, he experienced symptoms consistent with atypical angina. He had two episodes of chest pain, both associated with activity, but not regularly reproducible. My friend discussed these symptoms with a doctor who did an initial work-up to get a Framingham risk score but my friend never followed up. So the question my friend had was simply, “Should an exercise-stress test have been done on my initial presentation?”

So let’s look at the data. During the hospitalization a lipid panel was done which allows us to calculate his Framingham Risk:

Fifteen percent, intermediate risk. Last year the NEJM published a great article (at least by my pee-stained eyes) that looked at the diagnostic yield of cardiac catheterization and non-invasive testing. The data is not very encouraging. Here are the key figures, at least for my question:
The top panel looks at the diagnostic yield of noninvasive heart tests, and stratifies patients based Framingham risk scores. What should be obvious is that the there is much greater differentiation of outcome within series (same stress test result but varying Framingham scores) than compared to fixed risk scores and varying non-invasive test results (i.e. the red arrow is more important than the bue one):
That means that, had my friend gone for a stress test, the likelihood of having a positive cath would have gone from 35% with a negative test to 42% with a positive stress test. It’s hard to imagine that a rational medical plan would change with that additional data. Stated another way, would any doctor recommend a cardiac cath with a pretest probability of 42% but not order that same test with a pre-test probability of 35%? Doubtful.
The other thing which is notable, in panel B is that atypical angina is protective compared to no symptoms at all. My friends symptoms were definitely atypical. Here are the definitions according to UpToDate:
  • Definite or classic angina — Substernal chest discomfort characterized by all of the following characteristics: a typical quality and duration, provocation by exertion or emotional stress, and relief by rest or nitroglycerin
  • Probable or atypical angina — Chest pain with two of the three above characteristics
  • Nonanginal or nonischemic chest pain — Chest pain with one or none of the above characteristics
So it looks like my friend’s primary care doctor was using evidence based medicine. He did not rush to a stress test or other noninvasive test because the Framingham Risk Score was a better predictor of cardiac disease and my friend’s description of the chest pain, oddly, put him at a lower risk of coronary disease than no symptoms at all.
Posted on

Must read article on evidence based medicine

This is worst Photoshopped pill bottle I have ever seen.

You don’t see evidence based medicine called out by name often in the lay-press, but Sunday’s Boston Globe dove into the problems with guidelines and quality of evidence in an in-depth analysis. This is a must read.

But evidence-based medicine is only as strong as the evidence used to support it. The stark reality is that evidence can be weak, biased, or even fraudulent. More guidelines stem from expert opinion than from solid clinical trial evidence. And many physicians who write these recommendations have financial ties to drug companies — with no widely accepted policy to correct for conflicts of interest.

More coverage at Cardiobrief.

Posted on

I was taught that to live a long healthy life avoid alcohol and take your vitamins

Turns out that seemingly sound advice will kill’ya.

Headlines from this week:

Any drinking lowered the risk of heart disease but the more patients drank the greater the protection. And the amount they drank was pretty staggering:

  • light drinking, about a glass of wine or 1.5-beers per day, reduced risk by 35%
  • moderate drinking, 2 glasses of wine or 2-3 beers reduced the risk 51%
  • Moderate and heavy drinking, 5+ glasses of wine or 7+ beers, resulted in 54 and 50% risk reduction.
The health benefits of alcohol were touched upon in ths prior post.

Not only did the patients randomized to folic acid + B12 have a higher rate of developing cancer, they had a higher cancer mortality. Most of the cancer mortality came from excess lung cancer in the folic acid + B12 group. This was despite the fact that there were more smokers in the placebo group (38% vs 40% p=0.01). (JAMA) 

The fall of folic acid was touched on in this prior post.

So parents, remind your kids to drink their beer and forget their vitamins.
Posted on

Articles that changed the way I practice: Sodium intake, hypertension and mortality

I have long been skeptical towards the party line that salt intake is a driver of high blood pressure, as I wrote here and here. Though hypertension is nearly unheard of in primitive cultures with sodium intake below 50 mmol/day (1.6 g day), increasing sodium intake has modest effects on blood pressure. Three mm of systolic per 100 mmol of sodium (2.3 grams) according to the Intersalt Study (PDF). This 3 mm of systolic agrees with the change in blood pressure in the DASH-Sodium trial. Similar effect size has been documented in meta-analysis:
  • A 2002 meta-analysis by Lee Hooper of 11 trials of at least 6 months duration found a 1.1/0.6 mmHg reduction from a 35 mmol (810 mg) reduction in sodium intake.
  • A broader meta-analysis published in JAMA in 1998 looked at 114 trials and found a reduction of 3.9/1.9 in hypertensive patients and 1.2/0.3 in normotensive participants.
Despite these seemingly modest results all of the clinical practice guidelines on hypertension have adopted sodium restriction as a key part of blood pressure control:
My position when talking with patients about dietary modifications for high blood pressure had been to mention sodium restriction and weight loss but focus on the DASH diet (PDF) and exercise. But this strategy has recently evolved as I became aware of a pair of studies, one by Cook et al which strengthened the sodium argument and one by Larry Appel which weakened his own DASH research.
The Rise of Sodium
The article by Nancy Cook is a follow-up on the Trial of Hypertension Prevention I and II. These were randomized controlled trials of patients with high normal blood pressure which tried to determine which lifestyle modifications were effective. Patients randomized to sodium reduction were given individual and group counseling sessions on how to reduce sodium in the diet. After 18 months the patients in the TOHP I reduced sodium intake by 44 mmol/day (1 g sodium) and blood pressure fell 1.7/0.8 mmHg. In TOHP II, after 36 months, sodium intake was reduced by 33 mmol/day (750 mg of sodium) and blood pressure fell 1.2/0.7. The decreases in blood pressure in both studies are unimpressive.
Cook went back to these studies, 10 years after TOHP I and 5 years after the completion of TOHP II, and looked at the rate of cardiovascular events (primary outcome: MI , CVA, CABG, PTCA, CV Death). They found a 25% reduction in events in patients in the low sodium group (p=0.04) that increased to 30% reduction when the study was adjusted for baseline sodium excretion and weight. These results are incredible to me, modest reductions in sodium intake that were achieved through patient education had negligible effects on blood pressure but dramatic benefits on morbidity.

The strengths of this evidence comes from two lines of reasoning:
  1. It is a randomised trial. Even though the current data comes from an observational extension of the original RCT, this does not change the fact that we are looking at two groups that were orignially randomized.
  2. This is a study which looks cardiovascular events rather than blood pressure or other intermediate outcomes.
The fall of the DASH
The DASH Trial (Appel 1997) used a diet rich in fruits and vegetables to provide increased fiber and potassium along with other trace minerals. Low-fat dairy products provide increased calcium while keeping the diet low in saturated and total fat. Participants randomized to the DASH diet were served meals with 4-5 servings of fruit, 4-5 servings of vegetables, 2-3 servings of of low fat dairy and <25%>

The results were dramatic:
  • Decreased blood pressure of 5.5/3.0 mmHg
  • Decreased in hypertensives 11.4/5.5 mmHg
  • Maximal blood pressure response occurred after only 2 weeks

The primary weakness in the DASH trials is that I’m not going to provide my patients with all of their food. It is not a clincally relevent intervention. As physicians, all we can do is educate and cousel on diet. Appel did a follow-up study where he did just that and the DASH was no longer so impressive.
The PREMIER Trial randomized patients to three groups:
  1. Control group with no interventions
  2. Standard advice: 18 face-to-face meetings to go over weight loss, and strategies to reduce sodium and alcohol consumption
  3. Standard + DASH: 18 face-to-face meetings with the same contant as the standard group with additional counseling on adopting the DASH diet
Counseling resulted in significant weight loss of 5 kg in both experimental groups versus loss of 1 kg in the control group. There was no difference in physical activity, but physical fitness did improve from baseline all three groups. They didn’t find a reduction in alcohol or sodium intake however there was good separation in the potassium intake with the greatest increase in potassium in the DASH group as would be expected. Both of the experimental groups had greater reductions in blood pressure than the control group. 40% of the patients randomized to the Standard advice and 48% of the patients in the Standard + DASH were able to lower their blood pressure below 120/80. This difference was not statistically significant.

There was no improvement in blood pressure control with the addition of the DASH diet over counseling patients on established risk factors.
Posted on

So how much do we spend on routine daily labs?


Apparently a crap load:

Several studies have identified the overuse of daily lab testing and how certain interventions can effectively reduce tests ordered. A study by Miyakis et al. examined the effects of disclosing lab test costs on the frequency at which healthcare providers ordered these tests. 24,482 laboratory tests were ordered before the intervention (mean 2.96 tests/patient/day). Among those, roughly 70% were not considered to have contributed towards management of patients (mean avoidable 2.01 tests/patient/day). After costs of tests were disclosed, the avoidable tests/patient/day were significantly decreased (mean 1.58, p = 0.002), but containment of unnecessary ordering of tests gradually waned during the semester after the intervention. (1) A study by Kumwilaisak et al. examined how the implementation of formal guidelines effected how laboratory tests were ordered. 1,117 patients were enrolled. After the institution of the guidelines, the number of laboratory tests decreased by 37% (from 64,305 to 40,877). Furthermore, this result was still present at 1 year. (3)

Posted on

Alcohol and blood pressure: drink up?

Last week I turned in a chapter (sub-chapter?) on life-style modification for the control of blood pressure. This is part of a educational initiative for the National Kidney Foundation of Michigan.
I looked at all of the life-style recommendations in JNC-7:
I hope I will get permission to share my section on this blog because I think my conclusions differ in some important ways from the life-style dogma we hear over and over.
For now I want to share one interesting aspect: moderation of alcohol intake to reduce hypertension.
There is compelling data supporting the position that reducing or moderating alcohol intake reduces blood pressure.
One line of data comes from a recent study of two separate prospective observational trials by Sesso HD, Gaziano MJ, Et al: The Women’s Health Study (28,848 women) and the Physicians Health Study (13,455 men). They looked at initially normotensive men and women and followed them to determine the risk of developing hypertension (10.9 years of follow-up for women and 21.8 years for men) based on their exposure to alcohol. In men the risk of hypertension rose linearly with increased alcohol intake. In women, there was a J-curve, with increased risk of hypertension with teetotallers and decreased risk of hypertension until the risk that bottomed out at 5-6 drinks per week. Drinking at least one drink a day was associated with increased risk of hypertension.
Sesso’s data looks at the risk of developing hypertension, so while it provides evidence for the ill-effects of alcohol it does not answer the question of whether decreasing alcohol will ameliorate hypertension.
Xin X, et al. published a meta-analysis of 14 randomized trials of alcohol reduction in which a reduction of blood pressure was one of the end-points. Trial duration had to be at least 1 week:
Alcohol reduction reduced blood pressure to the degree promised in the JNC7 slides 3.3/2 mmHg:
What intrigued me was not the data on the reduction of blood pressure but the data on overall survival. Because the reason we care about blood pressure is because it is an intermediate outcome which is associated with cardiovascular events. If there is data on the actual outcome rather than the intermediate outcome it should by all means trump the intermediate outcome data.
Two large, high profile studies have looked at alcohol intake and survival. The biggest is a study by Thun et al and published in the New England Journal of Medicine in 1997. The data was taken from the 1.2 million American adults enrolled in the Cancer Prevention Study II and looked at 490,000 who provided complete information on smoking and drinking habits. The study showed that drinkers had a lower rate of cardiovascular disease but a higher rate of alcohol related illness. The statistics of this are interesting, a 40% reduction in the common cause of death (cardiovascular disease) ends up having much more influence than the 200-600% increase in the relatively rare deaths from conditions associated with alcohol intake:
Here is the authors description on the cardiovascular benefits of alcohol:

In contrast, the rates of death from all cardiovas-cular diseases combined were 30 to 40 percent lower among men (relative risk, 0.7; 95 percent confidence interval, 0.7 to 0.8) and women (relative risk, 0.6; 95 percent confidence interval, 0.6 to 0.7) reporting at least one drink daily than among nondrinkers. The largest reduction, in both absolute and relative terms, occurred in mortality from coronary heart disease among drinkers who, at enrollment, had reported heart disease, stroke, or some other indication of preexisting risk of cardiovascular disease.

They added a nice figure which illustrated the relative effects of smoking versus alcohol. The protective effects of alcohol are insignificant when compared to the dangers of smoking:
A second study on the mortality benefit of alcohol used the The Physicians Health Study. This is the same data that Sesso used to show the association of alcohol with the risk of hypertension. Malinski et al. looked at a cohort of subjects with pre-existing hypertension and demonstrated a 40% reduction in cardiovascular mortality with daily drinking as compared to rare or non-existent drinking. So even in the cohort that we are specifically advising to reduce alcohol intake, there is a survival benefit from drinking.
Most of my patients do not drink once a day. My feeling is that when they look at recommendations to reduce drinking they interpret that as they should stop drinking and they probably are actually increasing rather than decreasing their risk of death.
Salut