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 March, 2015

Primary Care Corner with Geoffrey Modest MD: HIV outbreak in Indiana

31 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

HIV outbreak reported in the NY Times​.

Basically, a pretty remarkable HIV outbreak among injection drug users in Indiana:

–71 cases of HIV since mid-December, related to intravenous use of prescription painkiller Opana (long-acting oxymorphone)

–All in Scott County, population 24,000 and 19% live below poverty line. Usually “only” five new cases of HIV/year

–And, the outbreak was so striking that even the Republican governor (Pence), who long opposed needle-sharing programs (which, of course, might have prevented the outbreak), justified a 30-day needle exchange program since “this is all hands on deck. this is a very serious situation” … it’s [needle exchange] a commitment to compassion”. So, I guess this means he thinks that program would work and would be compassionate, just that it shouldn’t normally be available to injection drug users???

–It might even decrease spread of hepatitis Cc???​

Primary Care Corner with Geoffrey Modest MD: Thyroid screening recommendations

31 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

The US Preventive Services Taskforce (USPSTF) just published an update of their recommendations for TSH screening of nonpregnant, asymptomatic adults (see doi:10.7326/M15-0483). For formal USPSTF recommendation, see here.



–hypothyroidism: subclinical = asymptomatic, serum TSH between 4.5-10 (varies a bit by lab), and normal T3 and T4; overt = can be asymptomatic, but high TSH and low T4 levels

–hyperthyroidism: subclinical = asymptomatic, TSH < 0.4 and normal T3 and T4. subdivided into “low but detectable”, with TSH 0.1-0.4, and “undetectable” with TSH <0.1. overt= can be asymptomatic, but has elevated T3 or T4

–Detection: TSH is good for detection of thyroid abnormalities in the general population, but there is measurement variability with TSH (eg, one study did find that TSH levels can vary by as much as 50% on a day-to-day basis. Also, TSH is sensitive to non-thyroidal factors such as being suppressed by acute illness or meds ). Best to confirm asymptomatic abnormalities with repeat confirmatory tests in 3-6 months (except if TSH>10 or <0.01 per other recommendations, though I do tend to repeat prior to further workup or therapy, just to make sure…)

–Benefits of early detection: inadequate evidence that there is clinical benefit in nonpregnant, asymptomatic adult. ie, no clear RCT evidence that detecting abnormalities/doing something improves quality of life, or improves blood pressure, BMI, cognition, bone mineral density or lipid levels (more later)

–Adverse effects of screening include labeling, false-positive results, and overdiagnosis and overtreatment

–Overall assessment: not enough information to recommend screening

— They also note:

–subclinical hypothyroidism is common: 5% of women and 3% of men in US, increasing incidence with aging

–37% of people with subclinical hypothyroidism spontaneously revert without intervention after several years; 2-5% go on to develop overt hypothyroidism

–subclinical hyperthyroidism in about 0.7% of the US population, more in women than men (again, increases with aging)

–1/3 of those with subclinical hyperthyroidism revert to normal over time. 1-2% with TSH<0.1 go on to overt hyperthyroidism. less likely if TSH 0.1-0.45

​–39% of people with TSH of  5.1-10 in a 1996 study received treatment, other data shows increasing % on meds as they age. (This is despite recommendations to the contrary – ie, from their perspective, many people are being treated without clear data)

–risk factors for hypothyroidism in the US (most are from Hashimoto’s thyroiditis) are age, female sex, white race, type 1 diabetes, Down syndrome, family history, goiter, neck irradiation

–risk factors for hyperthyroidism include age, female sex, black race, low iodine intake, ingestion of iodine-containing drugs (eg amiodarone)

This brings up a few issues:

–The USPSTF is pretty stringent in making recommendations, in general requiring high quality positive RCTs to make a strong recommendation. The plus here is that the recommendation is more evidence-based. The minus is that there are lots of things we do (or should do) which just have not had the studies done to back them up.  So, the fact that they find insufficient evidence in the case of thyroid screening does not necessarily mean that it should not be done.  For TSH screening, I basically agree — there really is not enough evidence and studies should be done, especially given how common the issue is.

–One big issue is what is the normal TSH???. A large observational study based on the Natl Health and Nutrition Exam Survey III (see J Clin Endocrinol Metab 92: 4575, 2007) looked specifically at people of different ages who were “disease-free” (no report of thyroid disease, goiter, or taking thyroid meds) and negative antibodies (no thyroid peroxidase or thyroglobulin antibodies, which are typically present with Hashimoto’s thyroiditis), and were not overtly hyper- or hypothyroid (ie, TSH not >10 or <.01).  They looked at the distribution of TSH values per different age groups. It turns out that the plot by age shows a shift in the median and 97.5th percentile noted with aging (a shift to a higher TSH), which suggests that TSH does normally increase with age, and in a step-wise manner by age group. The issue here is that there is more actual thyroid disease with aging, but that if it were just a larger number of people with high TSH, the median should still be in the “normal” place but with a longer tail for the increased incidence of actual thyroid disease. By their analysis if one uses a TSH >4.5 mIU/L as the cutpoint, “70% of the raised values for the 80-yr and older group fall within the 97.5 centile of their age-specific range” (by the way, some smaller studies do not find this. One confounding issue is that the elderly also can have more subclinical hyperthyroidism, which would also affect the median TSH). Other issues which suggest this increasing “normal” TSH is that only a small percentage (about 2.5%/year) actually  do develop overt hypothyroidism (TSH>10), even over a 20-year study. The reason for increasing TSH with aging???? could be medical issues or meds which interfere with the efficiency/sensitivity of the hypothalamic-pituitary feedback system. Or may be a part of healthy aging (slowing down the metabolism to preserve bodily function — though hard to make a strong evolutionary argument here, since not many 70-year olds are strong participants in furthering the species, and the % of the population with TSH>4.5 really begins to climb at age 70).

–Subclinical hypothyroidism: the potential for detection and treatment is more theoretical, since there are no significant longterm studies with many of the important clinical endpoints to inform practice. There have been a couple of short-term studies on treating subclinical hypothyroidism, showing no effect on blood pressure. Studies on lipids are mixed, but show small effect if any (hypothyroidism is assoc with high LDL). in terms of cardiac effects, one large UK observational study (Arch Intern Med 2012; 172:811) looked at 4735 people with subclinical hypothyroidism (based on a single TSH value of 5-10), followed for 7.6 years, 1/2 on thyroid meds, and found that in those 40-70 yo, 4.2% of those on thyroid meds had ischemic heart disease events vs 6.6% not on meds (a significant HR=0.61). No significant difference in those  >70yo. But they did not account for other meds (statins, aspirin…) which are likely to be different in the >70yo crowd, and this was not an RCT. Of interest, they also found less cancer deaths with thyroxine therapy in the younger group (??suggesting confounding in the study). Some, but not all studies have also found some benefit in the surrogate markers of endothelial function, LV diastolic function. There are also a few studies finding increased heart failure, with a trend in those in the TSH 7-10 range but significant in those in the 10-20 range. But again there are no intervention studies showing benefit to treatment.

–Subclinical hyperthyroidism: autonomous adenomas and multinodular goiters (MNG) are more common causes (one study had 57% of patients with MNG), though overtreatment of hypothyroidism is still a very common cause. 40-60% spontaneously revert over weeks to years (?if they had thyroiditis). In terms of risk for fractures, the effect of T4 is predominantly in cortical bone (eg wrist), least in trabecular (spine) and in-between in mixed (hip). The observational data in a recent study on hip or nonspine fracture was not significantly elevated with subclinical hyperthyroidism (see Ann Intern Med 2014; 161: 189). Other studies have shown bone effect of hyperthyroidism, which tracks with degree of TSH suppression. Atrial fibrillation is a major concern, with increased risk of about 70%, also tracking with the degree of TSH suppression, but clearly present with subclinical hyperthyroidism. Also, heart failure is associated with subclinical hyperthyroidism. Observational data are mixed for mortality, dementia, quality of life.  So, clearest concern is from afib, though we still do not have adequate data on preemptive treatment leading to benefit.

–In terms of adverse effects of labeling/treating: one study found that of 34K women with known hypothyroidism were less likely to report good self-rated health (not clear that this would apply to subclinical patients, but probably…). Also,  a study from 2004 found that 1/4 of patients on levothyroxine were inadvertently on too high a dose (TSH undetectable). Raises unproven potential for fractures or afib.

–One concern I have about asymptomatic vs symptomatic is that it is often hard to tell the difference. Most of the symptoms of both hypo and hyperthyroidism, esp in milder cases, are very nonspecific. And the patient may not even see these as symptoms, since they come on very gradually and the patient may just adapt to them.

–So, how does one piece this all together??  It is clear that we really need long-term RCTs to figure this out. Observational trials can be misleading. And subclinical thyroid dysfunction is such a common condition with such potentially bad outcomes that it is pretty striking it has not been studied (though I suppose that there’s not much $$ for drug companies in it, given generic availability….). The big issue for me is the very real question of what a normal TSH is. One could argue that it might be reasonable to check TSH some time after a patient reaches the ripe old age of 50 just to make sure they don’t have overt thyroid disease and treat that. But if the TSH is in the subclinical region, repeat it in a year. And if okay, maybe repeat in 5 years. This approach acknowledges that even overt thyroid disease, which most believe should be treated, is pretty common and often will not be detected without looking for it. Above and beyond this population approach, it is certainly reasonable based on the above to use targeted testing — eg, if patient has some potentially associated finding (hyponatremia, increased CK, macrocytic anemia, autoimmune disorder, pericardial/pleural effusion, etc, and perhaps even high LDL)​

Primary Care Corner with Geoffrey Modest MD: Treatment of skin infections in era of MRSA

30 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

A study done in 4 centers (Univ of Chicago, San Francisco General Hosp, Harbor-UCLA , and Vanderbilt Univ Med Ctr) looked at the efficacy of clindamycin vs trimethoprim-sulfamethoxazole (TMP-SMX) for uncomplicated skin infections (see N Engl J Med 2015;372:1093-1103​). The question was: which medication is preferred in the current era of community-acquired methicillin-resistant Staph aureus (MRSA), which seems to be extremely common around the country?



–524 patients, including 155 children. 30.5% with abscess (>5cm, in adults), 53.4% with cellulitis, 15.6% with both; the patients were 52.3% male, 53.2% Black, 40.3% White, 28.6% Hispanic. Mean age 27.1. 29.6% children.

–patients were randomized to clindamycin (adults at 300mg tid) or TMP-SMX (on double-strength tablet bid), all for 10 days. Pediatric doses adjusted by body weight

— of those with positive cultures (n=277), 217 had s. aureus [167 with MRSA, 77.0% of the staph infections, with 21 (12.4%) resistant to clinda and 1 resistant to TMP-SMX; 52 were methicillin sensitive]; 32 were strep of different varieties; 10 were proteus, 38 coagulase-negative staph, 15 diphteroids, and a smattering of others. Note: all of the cultures were from those with abscesses, not cellulitis (cellulitis being 53.4% of the skin infections)

–no difference in outcomes between these medications: 80.3% were cured with clindamycin, 77.7% with TMP-SMX in intention-to-treat analysis, but in the 466 evaluable patients, it was 89.5% with clindamycin and 88.2% with TMP-SMX.

–no difference in children vs adults, in those with abscesses vs cellulitis. also no difference in subgroups infected with s. aureus, MRSA, or MSSA. of the 15 patients with clindamycin-resistant staph, 11 were cured (73.3%) vs 91.7% of those sensitive to clinda

–adverse events: similar: 18.9% with TMP-SMX​  and 18.6% with clindamycin, with both antibiotics associated with diarrhea in about 10%, nausea 2.5%, pruritus in 1.5%, rash in 1%. no cases of clindamycin-associated c. diff infections.

So, a few points:

–It is pretty remarkable how well TMP-SMX did, since a reasonable percentage of the patients had strep which is felt not to be sensitive to TMP-SMX. And, it is generally held that cellulitis is much more likely than abscesses to be caused by strep (though we cannot really culture cellulitis), yet the cure rate with TMP-SMX was the same for cellulitis and abscesses. there is some literature suggesting that the way sensitivity testing is done may underestimate TMP-SMX sensitivity. [As a general point, agar-plate antibiotic resistance is not always an accurate reflection of what happens in the body. For example, many organisms causing urinary tract infections seemingly resistant to antibiotics may in fact respond, perhaps related to the high concentration of antibiotics in the urine]

–Several studies have found that antibiotics do not add much to the primary treatment of abscesses (which is: incision and drainage, I&D), so the high cure rate with any antibiotic may not be surprising. So, for example, 73.3% of those with clindamycin-resistant staph in the above study “responded” to antibiotics, but these bacterial isolates were from abscesses and may not have needed antibiotics. But I should add that the literature on this is really all over the place. Some studies suggest that abscesses> 5cm (as in the above study) do better with antibiotics in addition to I&D. Some find that MRSA infections in particular respond better when antibiotics are added. but there are studies to the contrary.

–There have been studies finding that cephalosporins or other b-lactam antibiotics with anti-strep activity work somewhat better than TMP-SMX for skin infections, though given the very high prevalence of MRSA that we find at our health center, we have been mostly using TMP-SMX as the primary agent and with good results​ (again, we only culture those with abscesses). I had a patient with morbid obesity and diabetes a few years ago who had a very large inner thigh abscess and surrounding cellulitis from documented MRSA, sent home from the hospital on linezolid, but I was unable to get the required prior approval and switched him to TMP-SMX​, and he had a great result… In general, we do usually add antibiotics for skin infections because of the articles suggesting benefit in MRSA, and there was even an article suggesting less likely recurrences if use TMP-SMX​.

Note: I have several other articles in the BMJ online blogs on MRSA, including the use of bleach baths, the new Infectious Disease Society guidelines on treatment of skin infections, and skin abscess treatment. See here. ​

Primary Care Corner with Geoffrey Modest MD: Cabg vs newer generation stents

25 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

Two drug-company sponsored papers were published in New England Journal of Medicine comparing percutaneous coronary intervention (PCI) with coronary-artery bypass grafting (CABG), looking specifically at how the newer drug-eluting stents compare with surgery.

  1. a large East Asian non-inferiority study with 27 centers randomly assigned patients with multi-vessel coronary artery disease (CAD) to PCI with everolimus-eluting stent vs CABG (see DOI: 10.1056/NEJMoa1415447).



 –880 patients were enrolled (only 1/2 of the anticipated). Mean age 64, 70% male, BMI 25, 41% diabetic, 67% hypertensive, 20% current smoker, 53% hyperlipidemia. 47% with stable angina, 43% unstable angina, 9% acute MI in past 90 days; ejection fraction 59%; 77% with 3 vessel disease and 23% with 2 vessel disease (they excluded those with clinically significant left main disease)

–primary endpoint: composite of death, MI, or target-vessel revascularization at 2 years

–at 2 years: primary endpoint in 11% in PCI group and 7.9% in CABG (absolute risk diff of 3.1%, non-significant p=0.32 for noninferiority)

–at 4.6 years: primary endpoint in 15.3% in the PCI group and 10.6% in CABG (abs diff of 4.7%, with HR 1.47 and p=0.04). No significant difference in composite of death, MI, stroke; but there was increase in PCI group for repeat revascularization in 11.0% vs 5.4%, p=0.003, and spontaneous MI (ie non-periprocedural MI) in 4.3% vs 1.6%, p=0.02.

–on secondary subgroup analysis, comparing diabetic vs non-diabetic: in diabetics, the rate of primary endpoint was much higher in PCI group (19.2% vs 9.1%, p=0.007) and in non-diabetics was a nonsignificant 12.6% vs 11.7%. But on further scrutiny:

–if look at all primary endpoints, diabetics having PCI vs CABG had HR=2.29 (1.35-3.89) and non-diabetics was nonsignificant (HR=1.16)

–if look at deaths: non-significant difference in PCI vs CABG independent of diabetes status. Similarly if look at “death, MI, or stroke”

–if look only at revascularizations, diabetics had HR of 4.31 (1.76-10.6), and nonsignificant for non-diabetics

  1. an observational New York State registry study looked at long-term mortality after CABG vs PCI with everolimus-eluting stents​ in patients with multi-vessel CAD in the time period 2008-2011 (see DOI: 10.1056/NEJMoa1412168).



 –9223 patients had PCI and 9223 had CABG, mean age 65, 72% male, 80% white/9% black/10% Hispanic, 70% with EF>50%, 35% with prior MI, 45% with 2-vessel dz and the rest with 3. They excluded those with >50% left main lesions.They performed propensity scoring, since more of those with 3-vessel disease got CABG.

–after 2.9 years, those with PCI vs CABG had:

–similar risk of death [3.1% vs 2.9% per year, HR 1.04 (0.93-1.17)]

–higher risk of MI [1.9% vs 1.1% per year, HR 1.51 (1.29-1.77)], but was only significant  in the group who had incomplete revascularization by PCI [ie, raising the untested assumption that one could try PCI and consider CABG in the group who cannot be completely revascularized]

–higher risk of repeat revascularization [7.2% vs 3.1%/year, HR 2.35 (2.14-2.58)]

lower risk of stroke [0.7% vs 1.0%/year, HR 0.62 (0.50-0.76)], especially evident in the first 30 days post-procedure. Though review of the stroke curve shows that even after 4 years, the curves are splaying apart (ie, seems to be increasing strokes over time in the CABG group)

–no difference in outcome in those who had diabetes

So, there are several reasons I am posting about this:

–These drug-eluting stent studies are much more impressive than the earlier stent studies in finding that, for many important endpoints, the stents are pretty much as good as CABG (which is clearly more invasive). The first study found no real difference in nondiabetics, and for diabetics really only for revascularizations. It was notable that all of the data on diabetics above was buried in the supplemental appendix, which requires going onto the New England Journal webpage to access and scrolling down to the middle of the 27 pages…..the second study found higher risk of MI and repeat vascularizations in those with PCI, but no difference in death and a significantly lower risk of stroke. So, this PCI intervention does provide a viable option to the much more aggressive CABG

–I think the first study highlights an increasingly frequent issue in study design: the use of composite endpoints. I believe the rationale is that by grouping several outcomes as the primary outcome, the chance to achieve statistical significance increases. But there may be very different inherent values to these different endpoints, and those are personal values and may be different for the provider and the particular patient. So, for example, I would consider the need for revascularization to be a much less significant outcome, and would be happier with 2 revascularizations than to have a CABG. On the other hand, a stroke would be a much worse outcome for me. And a heart attack in the middle, but closer to the stroke. Another issue obscured by the composite endpoint in this case is the timing of the event.  In both studies there is greater immediate mortality with CABG, which, within the first 30 days, in the large second study was 1.1% vs 0.6% with PCI, and an individual patient may well consider it much more important to survive the first month than having long-term benefit years hence. So just looking at the “primary outcome” obscures the fact that embedded in that outcome are really bad outcomes and some not-so-bad, and therefore the primary outcome of the study may not be very useful in making clinical decisions. It then behooves us guys in primary care to evaluate the risks/benefits of the specific clinical outcomes in the context of the values of the individual patient we are treating.

Primary Care Corner with Geoffrey Modest MD: Unemployment and suicide rates

24 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

A rather disturbing but not unexpected article was just published in Lancet Psychiatry correlating suicide rates with unemployment (see Lancet Psychiatry 2015; 2: 239–45​). About 1 million people per year die by suicide around the world. Prior data on the relationship between unemployment and suicide have been a bit mixed, though several studies have found a relationship and some with a time-lag.  This study looked at the 2008 world-wide economic crisis from a larger perspective than others: using databases from the WHO and the International Monetary Fund, the researchers looked at 63 countries (based largely on sample size and completeness of data) and assessed the long time period of 2000-2011, evaluating  suicide differences by 4 age groups and gender. As background, the 2008 economic crisis led to a 2% decline in world per capita GDP in 2009, and 31.8 million more people were unemployed in 2013 than 2007.



–the 63 countries were divided into 4 regions: Americas (Argentina, Brazil, Canada, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Mexico, Nicaragua, Panama, Paraguay, Suriname, Uruguay, USA); northern and western Europe (the usual suspects, with n=15); southern and eastern Europe (Belarus, Bulgaria, Croatia, Czech Republic, Greece, Hungary, Italy, Macedonia, Moldova, Poland, Portugal, Romania, Russia, Serbia, Slovakia, Slovenia, Spain, Ukraine); and non-Americas, non-Europe (Australia, Egypt, Georgia, Hong Kong, Israel, Japan, Kazakhstan, Kuwait, Kyrgyzstan, Mauritius, New Zealand, Singapore, South Africa, South Korea).

–the overall annual relative risk of suicide decreased  by 1.1% (0.8-1.4%) between 2000 and 2011

–but there was a nonlinear effect in the relationship between unemployment and suicide:

–a higher suicide rate preceded a rise in unemployment by 6 months

–higher suicide rates were in areas with lower baseline unemployment rates

–not much difference in terms of male/female ratio over time (about 4/1) or between the different age groups. There were quite different trends in the different regions, some with increasing and some decreasing suicide rates over the time period, but there was a consistent bump associated with the economic crisis

–overall, the relative risk of suicide associated with unemployment was elevated by 20-30% during the crisis period i​n all world regions (somewhat less in northern/western Europe and somewhat more in southern/eastern Europe)

–the overall excess suicide rate associated with unemployment decreased from 3.53/100K in 2000, to 2.75/100K in 2007, increased to 3.03/100K in 2009, and decreased thereafter to 2.82/100K

–therefore 41,148 excess suicides were associated with unemployment in 2007 and 46,131 in 2009

So, a few striking features.

–there was not a strong age component (one might have thought that the working population would have more significant psych trauma from not working), or much of a gender difference percentage-wise

–despite differences in different regions, some with a trend to increasing and some to decreasing unemployment over the 11 year period, there was a consistent bump in suicides with the economic crisis

–the effect was non-linear: countries used to lower unemployment rates had higher % increases in unemployment-attributable suicides

–the reverse time-lag (increase in suicides predating the upward swing in unemployment) could be from several factors, including the economic contraction beginning before the unemployment rate increases (suggesting that people felt more vulnerable in the period before the countries actually posted their increase in unemployment​), lags in accumulating data on unemployment rates, etc,

–and, perhaps most importantly, this study highlights the huge human and social costs of this economic crisis, one which occurred because of gross malfeasance by banks (though remarkably few bankers have been punished for this world-wide disaster, from which the world is still slowly recovering).

–and, it reinforces our general conception that there is a strong, integral relationship between one’s social environment, in the broadest sense, and one’s health…

Primary Care Corner with Geoffrey Modest MD: Discordance in interpreting breast biopsies

23 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

One factor often not considered in the breast cancer screening algorithm is the accuracy of the pathology report after biopsies. The widespread use of mammograms has led to lots of biopsies (1.6 million women in the US each year), with a reported approx 25% being positive for cancer. From 2011 to 2014 highly experienced pathologists from 8 states reviewed slides and the concordance of their readings was assessed (see JAMA. 2015;313(11):1122-1132​).




–65% of the invited pathologists consented to participate (n=115) and each given 60 slides to review from a total of 240 cases (1 slide/case), including 23 cases of invasive cancer, 73 of ductal ca in situ, 72 with atypical hyperplasia (atypia), and 72 benign cases without atypia

–49% of cases were in women 40-49yo, 50.8% had either heterogeneously dense or extremely dense breast tissue on mammogram, 57.5% were from core biopsies/42.5% excisional

–Pathologist impressions of the slides (all rated 1-5): confident of assessment (1 being very confident) 81% were 2-3; challenge of interpreting (1 being very easy) 78% were 3-4

–The overall concordance rate of diagnostic interpretations between the pathologists was 75.3% (CI 73.4%-77.0%), or 5194 of 6900 interpretations

–Higher discordance was found in women with higher breast density on mammograms (73%) vs lower (77%) (p<0.001) and in pathologists who interpreted lower weekly case volumes (p<0.001), or worked in smaller practices (p=0.034), or nonacademic settings (p=0.007)

–There was also variation depending on the diagnosis:

–benign without atypia (number of interpretations = 2070): concordance rate was 87%, overinterpretation rate 13% (most of the incorrect ones were read as atypia, some DCIS, but several read as invasive carcinoma)

–atypia (number of interpretations = 2070): concordance rate was 48%, overinterpretation rate 17%, underinterpretation rate 35% (most of the incorrect ones were read as DCIS or benign, a few as invasive carcinoma)

–DCIS (number of interpretations = 2097): concordance rate was 84%, overinterpretation rate 3%, underinterpretation rate 13% (most of the incorrect ones were read as benign or atypia, but quite a few as invasive carcinoma)

–invasive ca (number of interpretations = 663): concordance rate was 96%, underinterpretation rate 4% (most of the incorrect ones read as​ DCIS, but there were a few read as benign)

So, there are obvious issues here. It is quite disturbing the degree of variability here, with even totally benign tissue occasionally read as invasive carcinoma and vice versa. Though there are a few caveats. Only one slide per case was available, and in real practice if a question arose, the pathologist would likely check other cuttings/slides. Also, in real life the pathologist might have asked a compatriot to give their opinion. But this study shows that there is significant potential here for errors, creating unnecessary extreme anxiety, incorrect and potentially devastating therapy (chemo,radiation and surgery), or, on the other hand, missing a potentially curable lesion. I think the lessons here are:

–It is reasonable and appropriate for us in primary care to question the pathologist interpretations (ie, they are not always accurate, and i think at least many of us do incorrectly consider them to be a gold standard)

–And perhaps there should be structural changes to how slides are read:

–it seems reasonable to ask pathologists to somehow rate the certainty of their interpretation when they submit a diagnosis (ie, my guess is that the accuracy was higher when the pathologist looked at a slide and felt certain of their interpretation, which happened rarely in the above study)

–perhaps pathologists should regularly look at more than one slide per biopsy, to potentially give them a better ability to self-monitor and self-correct their first interpretation

–perhaps there should be a routine system of getting second opinions on the slide reading​

Primary Care Corner with Geoffrey Modest MD: Coronary angiography or exercise testing for chronic angina??

22 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

The American College of Cardiology meeting this week had several papers of potential clinical importance. One was on the relative utility of using coronary CT angiography (CTA) vs stress testing (functional testing) in patients with likely symptomatic coronary artery disease (CAD) — see DOI: 10.1056/NEJMoa1415516​ — the PROMISE trial, an NIH supported study.


–10,003 symptomatic patients who were felt to need non-urgent cardiac testing were randomly assigned CTA vs exercise testing (regular ETT in 10.2%, nuclear stress testing in 67.5%, or stress echo in 22.4%; 29.4% of these tests were pharmacologic), with median 25 months of follow-up.

–Mean age 60.8, 52.7% women, 22.6% minority, 87.7% had chest pain (72.7%) or dyspnea on exertion (14.9%). Mean pretest prob of CAD was 53.3%. Also 21.4% had diabetes, 65.0% hypertension, 51.1% current or past tobacco use, 32.1% had family history (and patients had a mean of 2.4 of these 5 risk factors). 25% were on b-blockers, 44% on ACE-I/ARB, 46% on statins, 45% on aspirin (which seems a bit low, to me — given the high risk, I would likely have had almost all of them on almost all of these…..)

–Assessed composite primary endpoint of death, MI, hospitalization for unstable angina, or major procedural complication; secondary end points were invasive cardiac cath that did not show obstructive CAD and cumulative radiation exposure


–primary endpoint in 3.3% of the CTA group and 3.0% of the functional-testing: adjusted HR of 1.04 (0.83-1.29), p=0.75. During followup of up to 42 months, there was pretty consistently no difference in the Kaplan-Meier estimates.

–CTA was associated with fewer caths showing no obstructive CAD (3.4% vs 4.3%), p=0.02, though more in the CTA group underwent cath within the first 90 days (12.2% vs 8.1%)

–median cumulative radiation exposure was lower in CTA group (10.0 mSv vs 11.3 mSv) but 32.6% of the functional testing group had no exposure, so the overall mean exposure was higher in the CTA group (12.0 mSv vs 10.1 mSv), p<0.001. Of note, the cumulative radiation exposure also depends on the exposure from the functional test chosen (ie, nuclear test is lots more than echo or regular ETT…), so the exposure in the group with non-nuclear functional testing was much lower than the CTA group.

–also, of those who had cath within the first 90 days, 27.9% of the CTA group and 52.5% of the functional testing group had no obstructive CAD (which raises the possible alternative and untested strategy of functional testing, which could lead to CTA prior to going to invasive cath).

–and revascularization rates (not an endpoint of the study) were higher in those in the CTA group (6.2% vs 3.2%)  (and as noted there was no difference in primary endpoints, raising also the untested assumption in this study of whether all of those who had caths really needed them…)

So, I must admit that the conclusion that CTA did not improve clinical outcomes makes sense on a number of levels.

–As in many of my prior blogs, I am very concerned about radiation exposure, and almost anything that might reduce it is a positive (and this is of huge public health significance, since about 4 million stress tests are done annually in the US). And I would extend my concern to the very frequent use of nuclear stress testing

–On a physiologic level, it makes a lot more sense to do a functional test, such as an exercise test, as opposed to just looking at how the coronary arteries appear on xray. For a few reasons:

–symptoms do not necessarily correlate with degrees of stenosis (there are many asymptomatic patients who have very severe or total coronary occlusion)

–future cardiac events do not correlate with degree of stenosis (old studies have shown that 78-97% of lesions in patients with acute coronary syndromes were <75% stenotic and half were <50% stenotic, reflecting the fact that acute events tend to occur after plaque disruption, and the most disruptable plaques are those with large lipid cores, which also have more inflammatory cells inside, but also may be small in overall size)

–and, it just makes sense to look at how a potential problem affects the individual person, since our relatively primitive models do not reflect the complexity of actual diseases and how the body responds (ie, there may be large-scale changes, such as angiogenesis/developing collaterals, or small-scale ones, such as changes in microvasculature and vascular auto-regulation, and, no doubt, some we know nothing about). It’s a bit like thyroid disease. some patients with the same free T4 levels may have functional hypo- or hyperthyroidism. The best test in general is the TSH, which really reflects how the individual’s body perceives the adequacy of the circulating thyroid hormones.

–and, as another sideline, over the past several years we have been moving towards treating non-urgent angina as a medical problem and not an invasive/surgical one (a conclusion supported by but not proved by this study). I would also suggest that in patients like these, where the pretest probability of CAD was >50%, it probably would have made sense from the beginning to have been on a more aggressive medical regimen.

Primary Care Corner with Geoffrey Modest MD: Exercise and the development of diabetes

21 Mar, 15 | by EBM

By: Dr. Geoffrey Modest.

There are many studies on the relationship between exercise and the development of diabetes. The studies for diabetes prevention have been somewhat conflicting about the intensity and amount of time doing exercise, with some suggesting that doing low-intensity exercise for longer was as good as high-intensity exercise for a shorter period of time. This study in the Annals of Internal Medicine addressed this issue, looking at men and women with abdominal obesity and therefore higher likelihood of developing diabetes, though none with diabetes (see Ann Intern Med. 2015;162:325-334​). In brief:

–300 abdominally obese adults were assigned to one of three 24-week interventions from 2009-2013, of which 217 completed the study. mean age 51.4, BMI 33.4 and waist circumference 110.6 cm (43.5 inches). ​fasting glucose mean was 5.4 mmol/L (97 mg/dL), 2 hr pp glucose mean of 7.3 mmol/L (132 mg/dL).

–all were given 5 weekly sessions. control group had no intervention. groups:

–low-amount, low-intensity exercise (LALI): 180 and 300 kcal/session for women and men, respectively, at 50% of maximum oxygen consumption (eg, routine walking). mean time per session was 31 minutes


–high-amount, low-intensity exercise (HALI): 360 and 600 kcal/session at 50% of max oxygen consumption (eg, routine walking). mean time per session was 58 minutes

–high-amount, high-intensity exercise (HAHI): 360 and 600 kcal/session, at 75% of max oxygen consumption (brisk walking/light jogging). mean time per session was 40 minutes

–no difference between groups in unsupervised additional daily activity or sedentary time

Results, controlling for age and sex:

–waist circumference changes:

–LALI: -3.9 cm [-5.6 to -2.3], p<0.001

–HALI: -4.6 cm [-6.2 to 3.0], p<0.001

–HAHI: -4.6 cm [-6.3 to -2.9], p<0.001

–no statistical difference between the exercise groups

–reduction in 2-hour glucose levels

–adjusting for covariates, only the HAHI group reached significant difference from controls: -0.7 mmol/L (-12.5 gm/dL), with CI of -1.3 to -0.1 mmol/L (-23.5 to -1.5 mg/dL), and p=0.027.

Conclusions: not much difference between different exercise interventions in abdominal obesity, but HAHI had lower sugar levels.

So, I have seen at least 6-8 studies looking at lifestyle interventions in delaying the development of diabetes in high risk people, typically either in those with genetic high risk — eg, those who have 2 parents with diabetes or those from specific high risk groups (eg Pima Indians), or in those with prediabetes (eg the Diabetes Prevention Project, DPP). In general, these studies show that diet and exercise put off the development of diabetes by about 10 years. ​The largest trial was DPP a 2.8 year trial comparing standard of care with an intensive lifestyle intervention vs metformin in nondiabetics with high fasting glucose, impaired 2-hr postload glucose and BMI>24, with the perhaps the surprising results that lifestyle change was dramatically better than metformin (diabetes incidence rates were 4.8 cases/100 with lifestyle intervention, 7.8/100 with metformin, and 11.0 in placebo group). 10-year followup (5.7 years after the trial ended) had 2766 participants, finding those in the original lifestyle group regained some weight (though the modest weight loss with metformin continued) but that the diabetes incidence was reduced by 34% with lifestyle and 18% with metformin as compared to placebo, indicating pretty long-term persistence of the effect of the interventions (there was some continued intervention:  lifestyle sessions were offered to all, those in the intensive lifestyle group were also offered 2 group classes of 4 session/year, those on metformin were offered to continue at 850mg bid if tolerated) — see Lancet 2009; 374: 1677–86​. So, bottom line, lifestyle changes are really helpful. This study unexpectedly had 2 findings: that waist circumference seemed to decrease equivalently with exercise but independent of the intensity and even at the lower level of 150 min/week. But blood sugar decreased only in the high intensity group at 200 minutes/week (this does differ from the American Diabetes Assn guidelines, but I should add that it is unclear what the clinical significance of this is in a non-diabetic population).


Primary Care Corner with Geoffrey Modest MD: Smoking and mortality

20 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

A recent combination of many large cohort observational studies assessed the role of smoking on mortality, looking beyond the accepted 21 smoking-related diseases (see N Engl J Med 2015;372:631-40.)


–current smokers have 2-3x the mortality of never-smokers

–the 2014 Surgeon General’s report estimated that >480K deaths/year in the US are smoking related, but considered only those deaths from those 21 acknowledged diseases

​–the 21 diseases are the usual culprits: the 5 most common are lung cancer, COPD, ischemic heart disease, “other heart disease”, and total stroke (ischemic and hemorrhagic).  Others include less common cancers (esophageal, stomach, lip/oral cavity, liver, colorectal, bladder, kidney and acute myelogenous leukemia), other atherosclerotic issues (aortic aneurysm and “other arterial diseases”), and infections (pneumonia, influenza, TB).

​–the researchers looked at people >55 yo from 5 large US cohorts: Cancer Prevention Study II Nutrition cohort, Nurses’ Health Study I cohort, Health Professionals Follow-up Study cohort, Women’s Health Initiative cohort and National Institutes of Health — AARP Diet and Health Study Cohort. Of these, the NIH-AARP provided about 75% of the male data and the NIH-AARP and WHI each provided >30% of the female data.

–total of 421,378 men and 532,651 women were followed from 2000-2011, mean age 65, 92% white, 50% drinkers, 9% current smokers and 50% former smokers.


–total deaths 181,377, including 16,475 current smokers

​–17% of the excess mortality was not in the 21 diseases per the US Surgeon General, with the most significant increases in current smokers for:

–renal failure, RR=2.0

–intestinal ischemia, RR=6.0

–hypertensive heart diseease, RR=2.4

–infections, RR=2.3

–various respiratory diseases, RR=2.0

–breast cancer, RR=1.3

–prostate cancer, RR=1.4

–others, though smaller numbers of deaths overall, included liver cirrhosis (RR=2.6), cancers of unknown site (RR=3.0), essential hypertension and hypertensive renal disease (RR=2.5)

–the specific death rates in general also increased with number of cigarettes smoked daily

–in general, the rates were increased in former smokers, with the RR declining as the number of years since quitting increased

–all of above associations were adjusted for current alcohol consumption. but as a side analysis looking at non-drinking smokers, the RR of liver cirrhosis still had RR=2.0 (somewhat lower), but for breast cancer was unchanged at 1.4

Conclusion: if one were to include these excess smoking-related deaths, the attributable annual mortality would be >60K more than the Surgeon General would have estimated by the 21 identified diseases

So, even more reason to not smoke, or to stop (and the sooner, the better). One can speculate about the mechanisms (eg, is death from renal failure from atherosclerotic disease, or the known dose-dependent increase in albuminuria?, is increased infections related to immune dysfunction, locally and/or systemically?). But, bottom line, smoking is not good for your health….. (and the above give us even more issues to discuss with patients)

Primary Care Corner with Geoffrey Modest MD: Pertussis vaccine — not quite up to snuff

19 Mar, 15 | by EBM

By: Dr. Geoffrey Modest

I’ve commented in a few blogs about the resurgence of pertussis, despite immunizing adults with the Tdap. I thought it might be useful to elaborate — there have been a few insights into what seems to be happening.

–In the bad old days, pertussis infection was pretty rampant in the US

–With the development of the whole inactivated  pertussis vaccine in the 1940s, the number of cases plummeted from 157/100K to 1/100K in 1973.

–There have been gradual increases in pertussis since 1982, attributed to the fact that immunity from pertussis is not life-long, as for example, it is with measles. In 2012 there were 42,000 cases in the US. Some of this is undoubtedly ascertainment bias: for example, a study of adolescents/adults with prolonged cough actually looked at it and found that 13-20% had b. Pertussis infection (see Clin Microbiol Rev 2005; 18: 326). Also, newer studies are using PCR and finding much more b. pertussis overall than previously.

–Because of adverse effects (esp high fevers, with about 1 in 330 kids getting temperatures > 105 deg, and about 1/2 getting temps >100.4 —  by the way, these were treatable with acetaminophen, and we used to suggest giving the kid acetaminophen prior to the vaccine, which used to prevent many of the complications), an acellular vaccine was developed in the 1990s based on pieces of the b. pertussis bacterium, renaming the old DPT as DTaP, and was used exclusively in kids (the “a” was for acellular).

–However, it became clear that the immunogenicity of the acellular pertussis component was not as strong as the older inactivated pertussis vaccine (see, for example, NEJM 2012; 367: 785). A California study looked at kids who received their fifth/final dose of DTaP from 2006-2011, a time-span which included the large pertussis outbreak in 2010. They looked at 277 kids who developed pertussis (PCR positive) vs 3318 PCR-negative controls and determined when they received their last DTaP vaccine, finding that those who developed PCR-positive pertussis were more likely to receive their DTaP earlier, with an odds of acquiring pertussis increasing 42% per year (see NEJM 2012; 367: 1012).

–As pertussis infections became more common, the CDC  recommended adults be vaccinated with the acellular vaccine (Tdap) once, though we are now experiencing pertussis outbreaks in adults within a few years after this vaccine was administered (making the suggestion of one vaccine in one’s adult life a tad suspect).

–Infant baboons were infected with b. pertussis bacteria at age 7 months, with 2 groups having been vaccinated before (with either acellular or whole-cell pertussis vaccines at ages 2, 4,and 6 months), a group naive to infection and unvaccinated, and another group who had prior infection. Results: colonization of the nasopharynx was no different in those who were unvaccinated and naive to infection and those immunized with the acellular vaccine (there was a gradual decrease in colonization after 14 days, cleared by 30 days). There was dramatically less colonization in those given whole-cell vaccine (almost all cleared by 14 days). And there was no colonization in those previously infected. Those with nasopharyngeal colonization after acellular vaccine were able to transmit pertussis infection to naive animals.

Also, they found that the actual immunologic T-cell response of those given the whole-cell vaccine was the same as seen in baboons who had natural infection, whereas the acellular vaccine produced a significantly different immunologic response (see doi/10.1073/pnas.1314688110).

–So, let’s see: we have a serious infection which does not create life-long immunity, an old vaccine that really worked well but had too many adverse effects, a new vaccine which not only does not elicit the same immunologic response as the natural infection but seems to work only transiently and probably does not do much to affect colonization or potential transmission….. About time to develop a more effective pertussis vaccine????

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