You don't need to be signed in to read BMJ Blogs, but you can register here to receive updates about other BMJ products and services via our site.

Archive for July, 2015

Primary Care Corner with Geoffrey Modest MD: SSRI use in pregnancy and birth defects

29 Jul, 15 | by EBM

By: Dr. Geoffrey Modest

A recent case-control analysis from the National Center on Birth Defects compared women who had children with birth defects to those without, to assess the association of the birth defects with specific SSRIs (selective serotonin reuptake inhibitors) taken from the month before pregnancy and through the third month of pregnancy ​(see BMJ 2015;350:h3190​).



 –17952 mothers of infants with birth defects from US centers in 10 states were compared with 9857 who did not have infants with birth defects. they looked specifically at the birth defects previously reported to be associated with SSRIs, with data from the US National Birth Defects Prevention Study database. Excluded were mothers at higher risk for having children with birth defects for other reasons (diabetes, teratogenic drugs), or those not on SSRIs for psych conditions including depression, bipolar disorder, obsessive compulsive disorder.

–researchers looked at 5 SSRIs (citalopram, escitalopram, fluoxetine, paroxetine, and sertraline), controlling for maternal race/ethnicity, education, smoking and prepregnancy obesity.


–3.0% of mothers were on an SSRI (298 women)

–sertraline was the most commonly used SSRI (38.6% of controls), but there was no association with birth defects (assessed for the previously reported problems of anencephaly, septal defects, anal atresia, limb reduction, omphalocele)

–paroxetine (14.1% of controls) was associated with anencephaly [OR 3.2 (1.6-6.2)]; atrial septal defects [OR 1.8(1.1-3.0)]; right ventricular outflow tract obstruction defects [OR 2.4 (1.4-3.9)]; gastroschisis [OR 2.5 (1.2-4.8)]; and omphalocele [OR 3.5 (1.3-8.0)] (also assessed for previously reported hypospadias, cleft palate)

–fluoxetine (24.8% of controls) was associated with right ventricular outflow tract obstruction defects [OR 2.0 (1.4-3.1)] and craniosynostosis [OR 1.9 (1.1-3.0)] (also assessed for ventricular septal defects, esophageal atresia)

–citalopram (9.7% of controls) and escitalopram (8.7% of controls), no association with birth defects, except for marginal association between citalopram and neural tube defects [OR 1.8 (1.0-3.0)]. (citalopram also assessed for ventricular septal defects, cleft lip/palate, hypospadias; escitalopram assessed for septal defects)

So, this was a pretty large database, with presumably pretty thorough data collection, having a few conclusions (though should be confirmed in other large databases)

–Sertraline, which has evolved into being the most prescribed SSRI used in early pregnancy, seems to be pretty safe, not confirming any of the five previously reported birth defects

–Paroxetine should be avoided if possible. that being said, it is important to bear in mind that depression itself may have a significant morbidity an mortality (maternal and fetal), treating depression is very important and often needs meds, and one SSRI may work a lot better than others. so, the perspective here is that the likely very real  higher risk of fetal abnormalities that they found translates to pretty small absolute risks of anencephaly of 2-7/10,000; right ventricular outflow tract obstruction from 10-24/10,000 (assuming the associations are causal).

–The variability of the birth defect risks suggest that it risk is not intrinsic to the serotonin reuptake inhibition itself but to other structure or pharmacokinetic properties of the meds

–In terms of fetal effects of untreated maternal depression, there was a review and set of recommendations from the  American Psychiatric Association and the American College of Obstetricians and Gynecologists (see doi:10.1016/j.genhosppsych.2009.04.003​), with a few more recent studies. Overall,

–miscarriage rate: some suggestion in the literature of an association, but studies were small and methodologically limited. So, no clear-cut association.

–growth effects: some studies found increased risk of low-birth-weight and small-for-gestational-age infants, but again poor quality data. 2 more recent studies have found some evidence of decreased fetal weight gain and fetal growth restriction.  So, no clear-cut association.

–preterm delivery: again, overall not great studies. more recent ones have found modest effect in decreased breast-feeding and more preterm delivery.  So, no clear-cut association.

–neonatal effects: no clear association with congenital abnormalities. but more neonatal irritability, less activity and attentiveness, and fewer facial expressions; and, small studies have found increased blood cortisol levels and decreased dopamine and serotonin levels (as found in the depressed mothers), and greater neonatal right frontal EEG activation and lower vagal tone.

–longer-term effects: no data on mothers with Major Depressive Disorder (formally assessed), but instead in mothers with prenatal depressive symptoms, finding that kids at 18 months had greater developmental delay. other studies were less impressive.

Bottom line to me: good studies have not been done (that I know of) which assess pregnant women with Major Depressive Disorder (which is the group in the general population most responsive to SSRIs) and look at important maternal and fetal/childhood outcomes, including not just the “hard outcomes” of congenital abnormalities, in utero growth, preterm delivery, but also the short-term effects on the pregnant woman (coping skills, function, general sense of well-being) and the longer-term effects on the child (bonding, general assessment of function, happiness, developmental milestones, psychological effects, etc). In the meantime, I think it is important to treat depression in pregnancy, which may well include using medications (as well as support, counseling), and that the BMJ study provides reassurance that there are several antidepressants which are not associated with evident fetal congenital abnormalities, despite some reports to the contrary.


Primary Care Corner with Geoffrey Modest MD: Marijuana passing through the generations…

29 Jul, 15 | by EBM

By: Dr. Geoffrey Modest

A pretty striking basic science article was recently published looking at the epigenetic effects of marijuana on rats, which seems to be inherited in the next generation (see doi:10.1038/npp.2015.155​).  There have been a slew of recent articles on marijuana suggesting that its adverse effects are more profound than many of us (myself included) thought (eg, structural changes to the brain, psychosis/scizophrenia, longer-term cognitive problems, risk of addiction to other drugs, lung diseases, etc — see here for some of these articles and critiques), and the positive effects are perhaps overstated by poor-quality studies (see blog, which reviews the data on the beneficial effects of medical marijuana). And, as noted, the marijuana issue is increasingly of concern as marijuana is being promoted as a “safe drug” and is now more commonly smoked than cigarettes by teenagers and young adults. In this context, the rats show us that there may be profound epigenetic effects of marijuana in the offspring of those exposed.  I do not presume to be very knowledgeable about the biochemistry here, so will present the extent of my primitive understanding: basically epigenetics differs from genetics in that there are heritable phenotypic changes that do not reflect genotypic changes (as a brief summary, see here). What I mean is that there can be profound changes in gene expression based on chemical changes to the DNA which typically either turn off or on some section of DNA, resulting in fundamental changes in gene expression. These epigenetic changes, for example, can result from aging, environment, lifestyle, and diseases themselves, and often are mediated by changes in areas of DNA which become methylated (there are other mechanisms as well, such as histone modification). The first epigenetic phenomenon studied in humans was in colon cancer, where DNA hypomethylation can lead to oncogene activation and hypermethylation to silencing tumor suppression genes (in general methylation silences gene expression). There are data from human lung cancers that those with cancer subsequently develop changes in oncogenes and tumor suppressor genes in the lung tissue, which can lead to increasing the potency of the cancer and decreasing the efficacy of chemotherapy (see here for example). There are human data that find that epigenetic modifications are important in the clinical expression of even identified genetic diseases leading to cognitive impairment (eg Prader-Willi syndrome), perhaps in autoimmune diseases, an array of neuropsych problems (schizophrenia, bipolar disease, mood disorders, dementia), etc.  So, as a primary care provider, why is this important???? I think epigenetics really opens the door to the link between the environment/lifestyle and the development of potentially many human effects, including an array of diseases. So, back to the rats…

The impetus of this study was the finding that there were cross-generational effects of parental cannabis exposure in adolescent rats leading to offspring who were more prone to self-administer heroin, have enhanced acute heroin withdrawal, altered striatal synaptic plasticity, and changes in the expression of cannabinoid, dopamine, and glutamatergic receptors (see Neuropsychopharmacology (2014) 39, 1315–1323). The current study used the same rat model, finding that of 16 adolescent rats exposed to tetrahydrocannabinol (THC) before mating, vs 16 non-exposed rats (all rat offspring were raised by non-THC exposed mothers) there were dramatic alterations in differentially methylated regions of the offspring’s brain (specifically, a depletion in gene promotors). They focused on glutmatergic synaptic regulation in the nucleus accumbens (a key area of the brain involved in reward processing and the area most linked to addiction vulnerability), finding profound methylation differences and associated altered mRNA expression in these offspring.

So, basic observations from above:

–the epigenetic effects of various stimuli, including those resulting from environmental exposures, can have profound effects on turning on or off different sections of the DNA, with resulting effects on protein production, etc, and may be associated with a variety of diseases

–although there is “reprogramming” of DNA methylation in offspring to preserve reasonably clean genetic transmission (see 10.1126/science.1229277), there are clearly times when these epigenetic changes do pass through to subsequent generations (well-documented in these rats)

–although I  am not a biological determinist, I think these types of studies do raise important issues about the potential for parental actions to affect offspring directly through epigenetic changes (which is not to say that they determine the behavior of kids. That is really complex and, undoubtedly, the social environment of the kids will play a large role in their own development, including subsequent epigenetic changes)

–it is pretty clear to me in inner-city Boston, that a huge number of people are exposed to the very potent marijuana on the streets now and using that regularly, especially but not only in adolescents. And that usage might increase with the acceptance/legalization of medical marijuana, with perhaps overstated benefits, but providing a more “socially-acceptable” aura, as with cigarettes and alcohol over the years.


–but, bottom line, we should be aware of these epigenetic changes, aware that they can be passed on to offspring even in the context of not using THC during pregnancy itself, and that the effects on the offspring might at least put them at a significant disadvantage in the longrun.

–and, I wouldn’t be surprised to find these types of epigenetic play a significant role in our common chronic diseases (obesity, hypertension, hyperlipidemia, diabetes…..) in individuals, and perhaps in their offspring as well (ie, it may not just be the genes themselves that predispose people to these diseases, but environmentally-induced changes in methylation etc of the genes which play a big role)



Primary Care Corner with Geoffrey Modest MD: NSAID warning by FDA

29 Jul, 15 | by EBM

By: Dr. Geoffrey Modest

The FDA recently reinforced their existing warning label on the use of non-aspirin NSAIDs and the increased chance of heart attack or stroke (this warning is already on the over-the-counter ones). Their comments:

–the risk of heart attack or stroke can occur within the first weeks of taking NSAIDs

–the risk is greater with higher doses

–although they acknowledge that newer information suggests some differences in NSAID risk, they do not feel it is sufficient to attribute either higher or lower risk to any individual agent

–the increased heart attack/stroke risk is not limited to those with known heart disease or increased risk factors form heart disease, though since their absolute risk is higher in general, they have greater risk of NSAID-induced heart attack or stroke

–those with known heart disease or risk factors are even more likely to die of a heart attack or stroke in the first year after their heart attack

–there is increased risk of heart failure with NSAIDs]

Those of you who have been reading these blogs for awhile know that I have real concerns about the overuse of a couple of drug classes in particular: PPIs and NSAIDs.  This link goes through some of the details from a Danish study — see here and includes some of the data suggesting that there are significant differences between the different NSAIDs. Though I use them quite sparingly, I think the data supporting naproxen over the others is pretty good. However, a large impetus to my doing lots of steroid injections, especially in the elderly, is to spare using systemic NSAIDs (this includes not just joint injections for OA, gout, pauciarticular RA, tendinitis, but I have also had great outcomes at time from injecting discrete back muscle spasms). I also use more topical drugs (topical capsaicin, diclofenac, lidocaine).

See here for the latest blog on PPIs.


Primary Care Corner with Geoffrey Modest MD: Extended anticoag for PE

28 Jul, 15 | by EBM

By: Dr. Geoffrey Modest

JAMA published the PADIS-PE trial of patients having a first unprovoked pulmonary embolism (PE), randomized to stopping anticoagulation after 6 months vs continuing an additional 18 months (see JAMA. 2015;314(1):31-40​).


 –371 patients (mean age 58, 40% >65yo, 50% women, mean BMI 27, 45% with high bleeding risk per the Am College of Chest Physicians rating) with a first episode of unprovoked, symptomatic PE, initially put on 6 months of a vitamin-K antagonist, who were then randomized from 2007-2014 in 14 French centers to continued warfarin vs placebo for 18 months. Target INR = 2.0-3.0. Patients with known major thrombophilia were excluded. Blood was sent for thrombophilia workup on day 0.

–minor thrombophilia (heterozygous factor V Leiden or heterozygous G20210A prothrombin mutation, or factor VIII >90th %) was found in 24% and major thrombophilia (antithrombin deficiency, anticardiolipin antibodies >99th %, or homozygous factor V Leiden or combined thrombophilia) in 4%.

–primary outcome: composite of recurrent venous thromboembolism (VTE) or major bleeding at 18 months; secondary outcome: this composite at 42 months, each component of the composite, and death unrelated to PE or bleeding at 18 and at 42 months (24 months after the end of the study).


–Mean % of time in the target INR range was 69.1%.

–During the treatment period

–primary outcome in 6/184 patients in the warfarin group (3.3%, or 2.3 events per 100 person-years) vs 25/187 in placebo (13.5%, or 10.6 events per 100 person-years), with HR 0.22 (0.09-0.55, p=0.001).  

–in the warfarin group, 3 patients (1.7%) had a symptomatic recurrent VTE, all after discontinuation of warfarin. in the placebo group, 25 patients (13.5%) had symptomatic recurrence, only one of which was after stopping the placebo (p<0.001).

— major bleeding occurred in 4 patients in warfarin group and 1 in the placebo group (statistically nonsignificant)

–After treatment discontinuation

–composite outcome in 27 warfarin-treated (17.7%; 10.0 events per 100 person-years) and 17 placebo-treated (10.3%, 5.7 events per 100 person-years)

–in the warfarin group, 25 patients had symptomatic VTE (9.3 events per 100 person-years, all in the absence of anticoagulation, 4 of which were fatal). in the placebo group, symptomatic recurrence was in 14 patients (4.7 events per 100 person-years, all in the absence of anticoagulation)

–major bleeding occurred in 2 patients on warfarin group (1 while taking warfarin and 1 fatal event happened after stopping it), and 4 in the placebo group (all nonfatal; 2 while on warfarin)

–Review of the graphs showed a dramatic difference during the 18 months after randomization, and a gradual reduction in warfarin protection over the next 24 months of observation when all were off warfarin (see below for the primary endpoint)


​–After 42 months, composite outcome in 33 patients on warfarin (20.8%) and 42 on placebo (24.0%)  [HR 0.75 (0.47-1.18)] —  the benefit of extended warfarin Rx was not maintained

–No difference in rate of recurrent VTE, major bleeding, and unrelated deaths

–The risk of recurrent VTE was greatest in the first 6 months after stopping anticoagulation, then increased linearly by 4-5% per year.  the risk of major bleeding with continued warfarin was low, increasing by <2% per year

So, a few issues: 

–this study confirms that when anticoagulation is discontinued, the patient at is at risk for recurrent VTE, whether anticoagulation is for 3-6 month or longer. Also that the risk of major bleeding with anticoagulation is pretty low (confirmed by other studies)

–in the decision analysis of how long to anticoagulate, it is important to remember that the natural history of a venous thrombosis is different from a PE: the likelihood of a recurrent PE is much higher in the latter group, and the case-fatality rate is 4-fold higher than after a proximal DVT. In the above study, 80% of the recurrent VTE events in both groups were from  symptomatic PEs.

–if you and patient decide to stop anticoagulants, there are some potentially useful approaches: for other articles, including one on different and safer ways to stop anticoagulation by checking d-dimer levels, see here and especially here)

–if you decide to stop the anticoagulation, 2 studies have found some efficacy of low dose aspirin (about 30% decreased risk of aspirin vs placebo — better than placebo but not as good as continuing the anticoagulation).

–an (unfortunately) untested hypothesis might be: check for underlying thrombophilia; 6 months of anticoagulation; check d-dimer; in those who are d-dimer negative, offer aspirin and discontinue anticoagulation (after discussing risks/benefits with patient); also look to see if the underlying thrombophilia actually matters (and, if so, then preferentially continue life-long anticoagulation). Short of that, this study adds to other observational studies finding a high risk of recurrent PEs in those with initial symptomatic unprovoked PE, so my inclination (and what I have been doing) is to offer life-long anticoagulation (especially to low risk-of-bleeding patients), with a fallback to checking the d-dimer and changing to aspirin if all is okay.


Primary Care Corner with Geoffrey Modest MD: Lipid lowering in elderly

28 Jul, 15 | by EBM

By: Dr. Geoffrey Modest

There was a recent article looking at statins and fibrates in the primary prevention of vascular events in elderly patients. I will also add to this review a meta-analysis of statin use as primary prevention in the elderly.

  1. A French cohort study reviewed a random sample of patients comparing those on lipid-lowering drugs(LLDs) vs not to assess the incidence of vascular disease (see This 3-city study was designed as a prospective ongoing population based cohort study designed to look at the association between vascular diseases and the risk of dementia. details:

–7484 community-living men and women (63% women, mean age 73.9 and no known vascular events) followed 9.1 years.

–Of the 27.4% on LLDs, 13.5% were on statins (38% of them were on simvastatin and 23% on pravastatin) and 13.8% were on fibrates (73% on fenofibrate). There were some demographic differences: those on LLDs were younger, more likely to be women, and had higher vascular risk (blood pressure, BMI, diabetes, cholesterol). Comparing those on fibrates vs statins in those taking  LLDs: fibrate users were older, had lower diastolic BP, and were less likely to be on antithrombotic agents.


–those on LLDs had a decreased risk of stroke [HR 0.66 (0.49-0.90)]. There was no difference between –those on LLDs did  NOT have a decreased risk of coronary heart disease [HR 1.12 (0.90-1.40)]

  1. A meta-analysis of 8 trials looked at statins in the elderly without known cardiovascular disease (see J Am Coll Cardiol 2013;62:2090–9). details:

–8 very high-quality trials with 24,674 patients (42.7% female, mean age 73), had a mean follow-up of 3.5 years


–statins were associated with 20% decrease in total cholesterol (232 mg/dl to 185 mg/dl), LDL decreased 31% (from 145 to 100 mg/dl), HDL was 50 mg/dl without change, triglycerides decreased 14%  from 147 mg/dl to 126 mg/dl

–statins were associated with reduced risk of MI by 39% [RR 0.606 (0.434-0.847), p=0.003], from 3.9% placebo to 2.7% with statins over 3.5 years; NNT 24 for one year to prevent one MI

–statins were associated with reduced risk of stroke by 24% [RR 0.762 (0.626-0.926), p=0.006], from 2.8%  to 2.1% over 3.5 years; NNT 42 for one year to prevent one stroke

— these numbers translate to: each 1 mmol/l (39 mg/dl) decrease in LDL is associated with a 57.1% MI risk reduction and a 34.5% stroke reduction

–statins were not associated with reduced all-cause mortality [RR 0.941 (0.856-1.035)], or cardiovascular deaths [RR 0.907 (0.686-1.119)]

–no difference in new cancer onset with statins vs placebo

So, these studies bring up a few issues:

–the first study was observational, so it is hard to assess accurately the role of statins and fibrates. The lack of association with cardiovascular outcomes is quite surprising, as is the equivalent efficacy of statins and fibrate (especially since the data on fenofibrate from the FIELD study and several others suggest overall that fenofibrate has much less efficacy than gemfibrozil, which also pales in comparison to statins). Not sure what to make of their results. It turns out that elderly French people do succumb to cardiovascular disease (in 2010, those >85 yo accounted for 43% of deaths from CAD and 49% of deaths from stroke), despite the “French paradox”.

–the second study was quite impressive and confirms/extends the primary prevention data on statins into the elderly.  We know that elderly, even without known vascular disease, are still more likely to die from a vascular cause than any other problem. We also know that although the relative risk reduction is pretty similar in primary and secondary prevention trials (around 30% for all of the studies), the actual event rate (absolute risk) is so much higher in those with known atherosclerotic vascular disease as well in the elderly (>2/3 of those >65yo have cardiovascular events) that the similar relative risk reduction with statins translates to a much higher absolute risk reduction/protection. The Framingham Study found that of those free of coronary artery disease at age 70, 35% of men and 24% of women still had a lifetime risk of CAD. By the way, this brings up another reason why I disagreed with the recent AHA guidelines. They reinforce treading softly in the elderly and only using moderate intensity statins, commenting that in general only moderate intensity statins be used over age 75, with more intense statins only on an individualized basis.  My continuing approach to treat to a target LDL of <100 for primary prevention in those at high risk, which often includes the elderly, which frequently leads me to use the higher intensity therapy. And, high intensity statins are really well tolerated even in the elderly (I have been using high intensity statins in the elderly for many, many years and have had no, as in zero, problems with them). Another reason to treat hyperlipidemia in the elderly is that studies have mostly found significant decreases in cardiovascular events within 6-12 months of starting a statin, usually within a reasonable time-frame for elderly patients.

–the lack of clear association between statins and all-cause or cardiovascular deaths is not so surprising. There was a favorable trend to decreased events, though not statistically significant, but this is likely because of the small numbers of cardiovascular deaths (about 1% of the patients) that occurred over a pretty short follow-up period.​

Primary Care Corner with Geoffrey Modest MD: Implantable defibrillators post-MI

28 Jul, 15 | by EBM

By: Dr. Geoffrey Modest


A lead article in JAMA was a study of Medicare patients who had an MI with a low ejection fraction (EF) to see what % actually had a cardioverter-defibrillator (ICD) implanted and how the patients did (see JAMA. 2015;313(24):2433-2440). This was a retrospective observational study.


–10,381 patients had an MI with EF<35% from 441 US hospitals, between the years of 2007-2010

–mean age 76, 40% female, 10% nonwhite. 65% had NSTEMIs and 75% had in-hospital revascularization


–cumulative ICD implantation rate by 1 year was 8.1%

–ICDs more often done in patients with prior CABG (31% had ICD vs 20% without); those who had cardiogenic shock (13% had one vs 8% not), and in male patients (38% less likely for females to get ICD).

–the 2-year mortality rate was much lower in those who had ICD placement: 15.3 events per 100 patient-years with ICD (128 deaths in 838 patient-years) vs 26.4 events per 100 patient-years without (3033 deaths in 11479 patient-years), with adjusted HR of 0.64 (0.53-0.78)

There are some important results of the study:

–one likely useful lesson is that those patients with profound LV dysfunction post-MI, should usually receive another echocardiogram 40 days later, and if EF is depressed, should be considered for an ICD. The MADIT-2 trial (N Engl J Med 1996;335:1933-40) did find an impressive 35% decrease in all-cause mortality after only 20 months, at the expense of an increased hospitalization for heart failure, in patients with LVEF<30 and at least 30 days post-MI. It is undoubtedly true that waiting the 40 days, when the patient is an outpatient, will lead to fewer ICDs being implanted (as with statins post-MI: if not started in the hospital, they may never be given later). The recommendations for an ICD are those with an EF<35% from an MI and NYHA class 2-3 after 40 days (and optimal medical therapy), or those with EF<30% and NYHA class 1. The basis for the recommendation to wait 40 days after an MI is a few studies.

–one older article (see Ann Intern Med. 2001;134(6):451) tracked echocardiograms after MIs, finding that 22% of patients with low EF totally normalized their EF, and 36% more had partial functional recovery, with 53% having > 5% improvement of EF over their post-MI baseline. And in the JAMA study above, 75% of those with MIs had revascularization – mostly PCI, which is more likely to lead to high levels of myocardial sparing and recovery over the short-term (after the stunned myocardial wakes up to its new reality). One concern in this Annals study is that there was not optimal medical therapy. Not clear that patients received aspirin, statin or b-blocker (this study was actually one testing the efficacy of ramipril post-MI, beginning on day 14 and with varying doses. So, the myocardial recovery rate they found might have been much higher with our current optimal medical therapy. And the rate of arrhythmias lower.)

–a few trials actually looked at immediate placement of ICDs in patients with MIs and reduced EFs (<35% or <40%): neither the DINAMIT trial (NEJM 2004; 351: 2481) nor the IRIS trial (NEJM 2009; 361: 1427)  found any difference in all-cause mortality by implanting ICDs right away

–subgroup analysis in the JAMA study showed that the mortality benefit was the same in patients < and >80 years old, and this suggests pretty strongly that we should still consider ICDs in older patients who were underrepresented in the clinical trials.​

But, I do not really trust this study (which basically promotes more patients receiving very expensive devices, for which there are clinically-important adverse effects), and for a variety of reasons:

–this was a quick-and-dirty study, data-mining a large database, but missing important elements: it was not randomized, so those who received ICDs may be very different from those who did not (eg, were some very sick patients excluded from getting ICDs because they were too sick, then only to die without an ICD?). Also, perhaps most importantly, they did not have any data on the EF 40 days after the patients had their MIs

–although the current JAMA study was sponsored by the Agency for Healthcare Research and Quality (AHRQ), the authors had a litany of drug and medical device company affiliations, including to a maker of ICDs (Boston Scientific). And needless to say, if there were a dramatic increase in ICD use, the magnitude of profit increase to these medical device companies would be mind-boggling to us mere mortals and add significantly to our health care costs.

–one last point. Especially in those areas with very rapid diagnostic and therapeutic advances (eg cardiology, or HIV treatments a few years ago), we often confront the problem that by the time a study is published, common practice has already moved beyond the starting point of the study, and the baseline medical management used in the study is no longer relevant.  In the current case, the ICD guidelines were based on older studies that did not really answer the questions that we want to know, and typically with suboptimal medical management compared to what we do now. The MADIT-2 study was done in 1996 and 75% of medical controls were on amiodarone, only 2/3 on ACE-I, and a minority on b-blockers. No comment on aspirin or statins. This was not even designed as a post-MI low EF study comparing patients with currently-conceived optimal medical therapy vs the same with ICD. Another big study, the SCD-HeFT trial (N Engl J Med 2005;352:225-37) included patients with both ischemic and nonischemic cardiomyopathy, and there was no clear breakdown of the meds the ischemic group was taking (so hard to know if they had optimal med therapy, though I doubt it given the study was published in 2005), and did not include patients who were NYHA 1 class. So, for example, are ICDs superior to medical management if all of the control patients were on b-blockers (which might prevent arrhythmias — though, the SCD-HeFT study showed no efficacy of amiodarone over placebo), plus aspirin (potentially preventing thrombosis and associated arrhythmias), plus ACE-I (potentially decreasing the vascular remodeling and the direct vasotoxic effects of aldosterone), plus a statin (perhaps preventing thromboses as well as improving endothelial function and being antioxidant, and thereby perhaps decreasing arrhythmias)?  So, it is actually hard to know with certainty if, with current aggressive medical management, an ICD would provide significant protection to all  patients with persistent low EF post MI, even after 40 days….

So, this article brings up several important issues. One is the role of drug and medical supply companies as drivers for expensive meds/devices, with major input into articles published in our best medical journals, guideline writing, etc. Another issue is that medical management if forever changing, leading to the pretty frequent problem of a moving target: soon after an article is published, medical management may have changed substantially enough that an intervention might not be useful. With all of this lack of certainty, it seems to me appropriate to repeat the echo after 40 days, and if persistently <35%, still to suggest an ICD to the patient. Also, a real concern in the sometimes hectic practice of primary care is to figure out a mechanism to make sure that important issues, including repeating the echo a month or so post-MI, are not lost in the shuffle, amidst the multitude of other issues we deal with pretty much on a daily basis for all patients we see…

Primary Care Corner with Geoffrey Modest MD: Medical use of marijuana

27 Jul, 15 | by EBM

By: Dr. Geoffrey Modest

JAMA published a systematic review and meta-analysis of the medical use of cannabinoids (see JAMA. 2015;313(24):2456-2473.). This is rather timely as several states are moving forward implementing new procedures for use of medical marijuana (as of may 2015, 23 states and Washington DC). The researchers evaluated 79 studies with 6462 participants.



–Nausea/vomiting from chemotherapy: 28 studies, most with high risk of bias.

-all studies showed cannabinoid benefit (some not reaching statistical significance), whether vs placebo or active comparator (procholorperizine, chlorpromazine, domperidone, and with single studies on alizapride, hydroxyzine, metoclopramide and ondansetron)

-cannabinoids vs placebo: also higher average number of patients showing complete response with cannabinoids [OR 3.82 (1.55-9.42); 3 trials]

–Appetite stimulation in HIV infection. 4 studies, all at high risk of bias

-some evidence that cannabinoids led to increased weight vs placebo, though the one trial comparing it to megastrol found the latter led to more weight gain [by the way, I have had pretty good success with cyproheptadine, with many fewer adverse effects vs megastrol]

-there is also limited evidence that cannabinoids are associated with increased appetite, greater percentage of body fat, reduced nausea and improved functional status.

–Chronic Pain, 28 studies, 2 of which were at low risk of bias

-cannabinoids better than placebo, assessing reduction of pain of at least 30%  [OR 1.41 (0.99-2.00); 8 trials]. Greatest benefit in the 1 trial using smoked THC, with OR 3.4 (1.03-11.48). This was true using a variety of pain scales and measurements.

-no difference in benefit if neuropathic pain [OR 1.38 (0.93-2.03); 6 trials] vs cancer pain [OR 1.41 (0.99-2.00); 2 trials].

–Spasticity from multiple sclerosis or paraplegia, 14 studies, 2 of which had low risk of bias

-cannabinoids demonstrated greater improvement of spasticity by Ashworth scale, vs placebo, though not reach statistical significance.

-greater average improvement in spasticity

–Depression, no good studies.

-secondary analysis of studies for other indications did not find improvement in depression

–Anxiety disorder, 1 trial at high risk of bias

-cannabinoids associated with statistically significant decrease in anxiety on a visual analog scale (p=0.01)

–Sleep disorder, 2 studies

        -improvement in sleep apnea/hypopnea index (mean difference -19.64, p=0.02)

-compared to amitriptyline in patients with fibromyalgia, cannabinoids were associated with improvements in insomnia but amitriptyline had greater sleep restfulness.

-in cannabinoid trials for other indications (MS, pain), secondary analysis noted improvement in sleep quality and sleep disturbance

–Psychosis, 2 studies both at high risk of bias

-no benefit

–Glaucoma, 1 remarkably small trial (n=6)

-no difference in intraocular pressures in patients with glaucoma [rather surprising to me, but a really small and therefore suspect study]

–Movement disorder from Tourette syndrome, 2 studies, 36 participants

-significant improvement in tic severity

–Adverse events (AEs)

-in 29 studies, risk of any AE had OR 3.03 (2.42-3.80)

-in 34 studies, risk of serious AE had OR 1.41 (1.04-1.92)

-in 29 studies, risk withdrawal due to AE had OR 2.94 (2.18-3.96)

-major AEs were dizziness, dry mouth, nausea, fatigue, somnolence, euphoria, disorientation, drowsiness, confusion, balance, hallucination, paranoia [newer studies which I have blogged on over the past couple of years have raised significant questions about longer-term cognitive problems, some finding structural brain changes, and psychosis – though it is hard to know for sure which came first: the marijuana leading to anatomic and physiologic brain changes or vice-versa]

So, overall not great data (pretty poor-quality data, per the Cochrane risk of bias tool). The best quality data supporting cannabinoid use were for chronic neuropathic or cancer pain, and spasticity from MS. Less good data supported use for nausea/vomiting from chemotherapy, weight gain in HIV, sleep disorders and tics from Tourette syndrome. Very low quality data for anxiety disorders. Other outstanding issues include: whether there are significant differences either between different cannabinoids or their mode of administration (orally, sublingually, topically; or if smoked, inhaled, mixed with food or made into tea). It should be emphasized that any medication with so many AEs make it very difficult to have a real placebo control (ie, the patients are pretty likely to know if they are on the active med vs placebo….)​

Primary Care Corner with Geoffrey Modest MD: Metformin, CKD, and death

23 Jul, 15 | by EBM

By: Dr. Geoffrey Modest 

A rather unusual quick-and-dirty study was just published looking at the relationship between metformin use and mortality, as well as lactic acidosis, in those with severe chronic kidney disease, CKD. This study was based on a window of time until 2009, when metformin was used in Taiwan without restriction in those with renal disease (See The researchers looked at a large database of diabetics with severe CKD — (those with diabetes plus a primary diagnosis of chronic kidney disease and those receiving erythropoiesis-stimulating agents, which was prescribed for 85% of non-dialysis patients with stage 5 chronic kidney disease –ie,  GFR<15 ml/min), and compared outcomes of those who were prescribed metformin to  a propensity-match scored group of patients not on metformin (a mathematical tool to attempt to adjust for patients with similar levels of comorbidities), and looked at outcomes.


–813 metformin users were compared to 2439 non-users, with no difference in 30 baseline clinical and socioeconomic variables. Patients had a serum creatinine of at least 6 mg/dL (the researchers did not have more specific/individual information)

–median follow-up of 2.1 years (range 0.3-9.8). Mean age 67.2, mean eGFR in men was 10 ml/min and 7 ml/min in women. All-cause mortality was reported in 434 (53%) of metformin users and 1012 (41%) of non-users.


–after multivariate adjustment, metformin use was associated with increased all-cause mortality [HR 1.35 (1.20-1.51, p<0.001)] and was consistent among all subgroups

–the increased mortality was dose-dependent: no increase in the groups on <= 500mg metformin/d (nonsignificant 14% increase), or 501-1000 mg/d (nonsignificant 30% increase), but was significant for those on >1000 mg/d (57% increased risk)

–for those with “metabolic acidosis”, which includes those with lactic acidosis, there was no significant increase in those on metformin, independent of metformin dose and without any trend of increase in those on the highest dose.

–the incidence of metabolic acidosis was pretty small: 1.6 vs 1.3 events per 100 patient-years ([HR 1.30 (0.88-1.93, p=0.19)]

So, pretty striking study in that those with very advanced renal disease DID NOT have any increase in lactic acidosis — the feared complication, since this is a highly lethal condition, metformin is excreted unchanged in the urine and accumulates with renal dysfunction, and the kissing-cousin of metformin (phenformin, which was used in the US) was associated with lots of lactic acidosis-related deaths.

A few comments:

–metformin was likely used at such a high rate in Taiwan, even after the US FDA proscribed its use in men with creatinine>1.5 and women with creatinine>1.4 in 1994, because metformin has the best data in preventing both cardiovascular and all-cause mortality in diabetics (metformin was the preferred biguanide used in Europe and has been used on many millions of people since 1957)

–there were data from other studies suggesting cardiovascular benefit in those with moderate CKD (eGFR of 30-60 ml/min)

–the cause of the 35% increased all-cause mortality in the high-dose metformin users is unclear.

–on the one hand, metformin users seemed less sick (shorter duration of diabetes than non-users, had less retinopathy, and were less likely to go on to dialysis).  The increased mortality was even in the subgroup of those on metformin monotherapy (suggesting that hypoglycemia was not the cause of death), and metformin use was associated with an 86% increased risk of admission for cardiovascular disease before death [HR 1.86 (1.31-2.23), p<0.0001]. And, there was no increased death risk in those on non-metformin oral anti-diabetic agents. So, all of this suggests that metformin was the bad actor.

–on the other hand,  this was a quick-and-dirty study with much missing information: were there unknown and uncontrolled reasons why the Taiwanese doctors chose metformin for some patients over others? did those on metformin actually have worse renal function than those not? (and there may well be increasing atherosclerotic risk with decreasing renal function). Were there unaccounted for variables in their propensity-matching– eg they did not have data on the intensity of the diabetes control or even the smoking history. Perhaps physicians treating those at the highest risk of heart disease (smokers, with diabetes and ESRD) might have been disproportionately prescribed metformin to lower their cardiac risk? Or perhaps more patients with poorly-controlled diabetes and perhaps higher risk of heart disease got metformin???

–the lack of association with lactic acidosis is pretty striking in this cohort with severe CKD. Part of the lactic acidosis fear is that some studies have found increased lactic acid levels in those on metformin who have CKD, but the relationship between high lactate levels and clinical lactic acidosis is not clear, leading to recommendations NOT to follow lactic acid levels.

So, I would still not use metformin in those with eGFR<30 (see prior blog for more information on this) but the dreaded fear of the highly lethal lactic acidosis is unlikely to be an issue at any level of renal dysfunction…


Primary Care Corner with Geoffrey Modest MD: Tenofovir nephrotoxicity

23 Jul, 15 | by EBM

By: Dr. Geoffrey Modest

 A study just published looked at tenofovir disoproxil fumarate (TDF) renal toxicity, finding that urine dipstick evaluation is inadequate (see  AIDS 2015, 29:941–946). Details:

 –43 patients with TDF nephrotoxicity (mean age 54.7, 53% men, 37% white/32% african-american/19% latino), and having biopsies showing proximal tubular injury and typical mitochondrial abnormalities for TDF nephropathy. Median duration of TDF therapy was 9.5 months, with the earliest case after 2 weeks. Baseline serum creatinine 1.2 mg/dl, increasing to 4.9 after the TDF.


–37 cases reported proteinuria by dipstick, 60% of these cases had at least 2+ proteinuria ​(8% 0-trace, 32% 1+) — note: the dipstick proteinuria primarily measures albuminuria.

–27 patients had urine protein quantified (either 24-hour collection or spot urine protein-to-creatinine ratio), median proteinuria was 1742 mg/day — note: this is total protein measured

–10 of these 27 patients had concomitant urinary albumin measured, with median 236 mg/g creatinine

–so, the mean urine albumin-to-protein ratio (uAPR) was 0.17

–therefore, TDF nephrotoxicity is predominantly associated with very high urinary protein excretion but not much albuminuria

–there was also a comparison arm of HIV patients undergoing kidney biopsy for other reasons (dx of focal segmental glomerulosclerosis, diabetes, hepatitis C proliferative glomerulonephritis, lupus) and never exposed to TDF had elevations of both proteinuria (2570 mg/gCr) and albuminuria (2029 mg/gCr)

I raise this issue because I think many of us use dipsticks or urinary albumin excretion to assess nephropathy (and appropriately with diabetes, or HIV-associated nephropathy). This study strongly suggests that urinary protein excretion may be a much better marker of nephrotoxicity in those on TDF. So, even though this study looked at patients with pretty advanced nephropathy (median creatinine of 4.9 mg/dl), it does suggest that using a urinary dipstick to detect early TDF nephropathy is probably the wrong thing to do: the uAPR is likely more sensitive (ie, this would support the algorithm: check urinary protein-to-creatinine ratio, and if elevated, then do a uAPR). I would assume that the issues would be the same for other TDF users, such as those on treatment of chronic hepatitis B infection. One hopefully positive change is the likely approval of tenofovir alafenamide, which seems to be as effective clinically as TDF for HIV infection, but has much less nephrotoxicity (eg, see recent findings brought out in the CROI/HIV meeting)


EBM blog homepage

Evidence-Based Medicine blog

Analysis and discussion of developments in Evidence-Based Medicine Visit site

Creative Comms logo

Latest from Evidence-Based Medicine

Latest from EBM