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 October, 2015

Primary Care Corner with Geoffrey Modest MD: Atrial Fibrillation and Dementia

30 Oct, 15 | by EBM

By Dr. Geoffrey Modest

The studies on the relationship between atrial fibrillation (AF) and dementia are inconsistent. This long-term, prospective population-based observational Rotterdam Study provided insight into this relationship (see doi:10.1001/jamaneurol.2015.2161).


  • 6514 participants (mean age 68, 59% female, BMI 26, 10% diabetic, BP 139/74, chol/HDL ratio of 290/54, 42% former/23% current smokers, 8% with coronary artery disease, and 3% with heart failure)
  • 318 (5%) had prevalent AF at start of study; 994 (15%) developed incident dementia over 20 years of follow-up


  • Prevalent AF was associated with increased risk of dementia, with HR 1.33 (1.02-1.73), meaning a 33% increase vs those without prevalence AF
  • Of 6196 without prevalent AF (79K person-years of follow-up), 723 (12%) developed incident AF and 932 (15%) developed incident dementia
  • Incident AF was associated with an increased risk of dementia in those <67yo, with HR 1.81 (1.11-2.94), but was not significant in those >67yo
  • The risk of dementia was strongly related to the duration of exposure to AF:  those younger patients with the longest exposure had HR 3.30 (1.16-9.38, with p=0.003 for trend), though not in those >67yo
  • The above relationships were independent of clinical stroke, and didn’t vary by whether the dementia was classified as “Alzheimer” or “non-Alzheimer”

The above study found an impressive relationship between AF and dementia in those <67yo, with an apparent dose-response curve (the longer the AF exposure, the more dementia). There are some important caveats:

  • The definition of dementia was basically a clinical one, relying on the MIni-Mental State Exam and Geriatric Mental State Schedule, with further clinical differentiation of Alzheimer vs non-Alzheimer. These tests are insensitive to picking up early dementia, especially in high-functioning adults
  • There was no systematic neuroimaging, so unclear if there were either subclinical strokes, or even asymptomatic emboli (lacunar strokes are associated with cognitive decline)
  • The above study did not have any information about whether the patients with AF were anticoagulated (nor how good the level of anticoagulation was)
  • ​It is difficult to attribute causality in an observational study: was the issue that the AF caused cerebral hypoperfusion (AF is associated with decreased cardiac output) or small cerebral emboli leading to cognitive decline? Or that both AF and cognitive decline share similar causes (those who had AF were more likely to be smokers, have lower HDL levels, be on treatment for hypertension, be diabetics)?
  • ​AF is often undetected in asymptomatic individuals, since it is often intermittent. In a study as the above one, with clinic visits/EKGs done every 3-4 years, we may well miss asymptomatic intermittent AF. Perhaps we should have a more aggressive approach to AF diagnosis?? 24 or more hour monitoring every year??

So, the real issues in primary care are:

  • AF is really common and increasing as the population ages
  • The data on rhythm control (e.g., reverting to normal sinus rhythm) vs rate control (just controlling the ventricular rate) did not assess long-term cognitive issues (e.g. the AFFIRM and RACE trials, which somewhat favored rate control, basically assessed mortality, strokes/hemorrhages, serious arrhythmias, and used somewhat toxic antiarrhythmics and did not maintain full anticoagulation in those with rhythm control). It is certainly possible that restoration of normal sinus rhythm is in fact better in the prevention of cognitive decline.
  • But, it is hard to maintain normal sinus rhythm with these pretty toxic meds. Even with radiofrequency ablation, there are not-so-infrequent AF recurrences (meta-analysis of ablation found only about 50% were in NSR after 3 years with a single ablation, though increases to about 80% with multiple ablations, though ablations themselves are also associated with emboli –e.g., see JACC 2013; 62:531 or Circ 2010; 122:1667.) and who knows how sustained normal rhythm will actually be in longer-term studies? Or even if some who have “NSR” still have occasional episodes of AF which could be associated with small, inapparent emboli? Maybe we should aim for sinus rhythm but continue with anticoagulation??
  • And, it would be very interesting to know if either rhythm control (as best as we can measure it) or adequate anticoagulation diminishes cognitive decline (one possible study might be in those with low CHA2DS2-VASc scores, who are not generally given anticoagulation, to be randomized to anticoagulation vs not, to see if there is a long-term cognitive difference. There is a study lending some support to this — those in AF on anticoagulants and more often in therapeutic range had less dementia –see  Heart Rhythm 2014: 11:2206 — though this was not a randomized control study, so there may have been important differences in those achieving better control than the others.)
  • But, the finding above that the duration of AF being associated with more cognitive decline, is somewhat reassuring since it is likely that those with less evident AF seem to be at lower risk, supporting the concept that decreasing the numbers of episodes of AF may be helpful.
  • In some ways the real value of a quick-and-dirty study as the Rotterdam one above serves to raise a much bigger issue: we perhaps should not be so complacent with just achieving rate control, the current general approach, but should pursue studies to see if another very important endpoint (cognitive decline) might be better achieved by more aggressive attempts to reinstate sinus rhythm.​

Primary Care Corner with Geoffrey Modest MD: Hep C Treatment and Hep B Reactivation

29 Oct, 15 | by EBM

By Dr. Geoffrey Modest

Thanks to Dr. Jon Pincus for his suggestions on this blog.

A brief report was just published with 2 cases of hepatitis B virus (HBV) reactivation during successful treatment of hepatitis C  virus (HCV) –see Clin Infect Dis 2015;61:1304​.

Details of the Cases:

  • 55yo male with chronic HBV and genotype 1a HCV infections. The patient had failed 2 prior courses of interferon-based therapy and had underlying compensated cirrhosis. His HBV viral load prior to the new treatment for HCV was 2300 IU/ml and had been consistently <2000 before; ALT 62 and AST 59, bili 0.7, normal INR. Hepatitis B e antigen was negative and antibody positive. Given sofosbuvir and simeprevir for the HCV. The HCV VL (viral load) was undetectable by week 4 of treatment. By week 7 he developed symptoms of hepatitis and by week 8 had tender hepatomegaly and jaundice, with ALT 1495, AST 1792, bili 12.2 and INR 1.96. HCV VL was still undetectable, no evidence of drugs or acetaminophen to cause the hepatitis, and his HBV VL increased to 22 million. They stopped the HCV treatment, and began tenofovir/emtricitabine. By week 14, symptoms had resolved, and LFTs were back to baseline. At week 28 (20 weeks after stopping HCV drugs), his HCV and HBV VLs were both undetectable
  • 57 yo male with chronic HCV genotype 1a, ALT of 54, and hepatitis B serologies of: positive hep b core antibody (HBcAb), negative hep b surface antigen (HBsAg) and antibody (HBsAb), and undetectable HBV VL. He had received interferon-based therapy 5 years before. Begun on sofosbuvir and simeprevir for HCV with undetectable HCV VL at 2 weeks, though HBV VL increased to 353. At 4 weeks, HCV VL undetectable, but HBV VL increased to 11,255. Patient remained asymptomatic and LFTs remained normal. Tenofovir was added and both HCV and HBV VLs were undetectable after 12 weeks of therapy.

So, raises a few issues:

  • Treatment of HCV infection can cause reactivation of HBV, so it is important to test for HBV infection in those with HCV.
  • There can be significant HBV flares which are asymptomatic, as in the second patient. Also in this case, there can be HBV flares with no detectable serum HBV VL beforehand.
  • In general I usually consider that anyone with an isolated HBcAb (which I do see pretty commonly, and as in the second case above) and an undetectable HBV VL are likely to either have a false-positive HBcAb or an old infection with waning and undetectable HBsAb (or, rarely, be in the window of an acute infection with a clearing HBsAg and prior to the emergence of a detectable HBsAb, and actually have the IgM HBcAb, if you check it). This study raises the issue that even with hepatitis B serologies that appear to indicate no active infection, there is still the possibility of a dormant HBV infection which may energize with HCV treatment (as sometimes happens in those getting transplants or biologicals with the attendant immunosuppression). This phenomenon is referred to as an “occult HBV infection”, found in people who do not have detectable HBsAg in the blood, where there is HBV in the liver but sometimes with no VL found in the blood. The prevalence of this varies in studies, perhaps due to differing definitions/methodology, but seems to be more frequent in those with HCV infection (also in injection drug users and those with HIV or hepatocellular carcinoma) — see  doi:10.1128/CMR.00018-11.​ Interestingly, an older study from 1999 found 20 people without any of the serologic markers of HBV infection​ had liver biopsies finding HBV sequences, more commonly in those with concommitant HCV infection (see​ ), raising the importance to check for acute HBV in anyone with acute hepatitis, even if recently checked serologies are totally normal. And those with HCV and occult HBV together had more cirrhosis (and, those without HCV but only occult HBV had more “cryptogentic liver disease”)
  • The editorialists (see Clin Infect Dis 2015;61:1307) feel that the likely pathophysiology, with some experimental evidence in support, is that HCV creates a host immune state in the liver which suppresses HBV replication, and that the effective treatment of HCV disrupts this control.  for clinical management, they suggest the following:
    • Always check the full array of hepatitis B serologies (HBsAg, HBsAb, HBcAb), and if negative, immunize.
    • Check HBV VL prior to starting HCV therapy in those who are either HBsAg or HBcAb positive.
    • If the HBV VL is negative, repeat “at least every 4 weeks”, and initiate treatment if they reach the usual criteria to start HBV therapy.
    • ​Those with evidence of immune recovery from prior HBV infection (with both HBsAb and HBcAb positive) should get one HBV VL at 4 weeks, even though their chances of reactivation are low. If the 4 week VL is negative, then no further testing is needed.
    • The editorialists are less aggressive in treatment of those with asymptomatic HBV reactivation, noting that the optimal treatment is unknown (the second patient above had dramatically increasing HBV levels but was asymptomatic). I personally would argue that, in the absence of clear studies, treatment to prevent acute clinical HBV infection seems reasonable, given the real morbidity and mortality of HBV and the relatively benign nature of current HBV treatment (i.e., so I would have treated the second patient above, though the editorialists were equivocal).


Primary Care Corner with Geoffrey Modest MD: Academic Conflicts of Interest

26 Oct, 15 | by EBM

By Dr. Geoffrey Modest

conflict of interest

Two issues on conflict of interest in medicine just came out in 1 online blog HMS-in the News on 9/30/15 (a regular email highlighting whenever Harvard Medical School people make it into the news, even apparently when the news isn’t so great).

  1. The BMJ just printed an article on the prevalence and compensation of academic leaders, professors and trustees on US healthcare company boards.  As we know (and as has been mentioned several times in prior blogs), we are getting more and more clinical guidelines, and these guidelines increasingly come from specialty societies (vs the good ol’ days, when the NIH did them). And, with the increasing role of the private sector (drug companies, etc.) in academia and in funding academic research, more of the specialty society leaders are financially tied into the private sector. There has been great general concern about conflicts-of-interest (COIs), and many of the larger specialty societies (e.g. Am Acad of Cardiology) are increasingly explicit about exploring and revealing COIs. Coincidentally, the Annals of Internal Medicine just published the Guidelines International Network position paper on disclosures of conflicts of interest and their management, citing 9 principles (see Ann Intern Med. 2015;163:548-553). Although many of us are deeply concerned that COIs in clinical guidelines could seriously distort the quality and integrity of these guidelines (as we should be), it turns out that there is a much more profound COI in academia, assessing data from 2013 (see BMJ 2015;351:h4826).


  • 442 healthcare companies had publicly accessible disclosures on their boards of directors
  • 180 (41%) had one or more academically affiliated directors (166 drug companies, 107 biotech companies, 94 medical equipment and supply companies, and 75 healthcare provider companies)
  • These directors were affiliated with 85 geographically diverse non-profit academic institutions, including 19 of the 20 NIH funded medical schools and all of the 17 US News and World Report “honor roll hospitals”
  • These 279 academically affiliated directors include: 73 leaders (17 CEOs, 11 vice presidents, 15 university presidents/provosts/chancellors, and 8 medical school deans/presidents), 121 professors including 16 department chairs, and 85 trustees
  • The total compensation to them was $54,995,786, with median individual compensation of $193,000
  • In addition, these directors owned 59,831,477 shares of company stocks, with median individual share being 50,699 shares (stock shares, of course, are tied to company performance)

So, this brings up some pretty big issues:

  • Although guideline writers and higher-ups in specialty societies are under more scrutiny about COI’s, it is clear that the involvement of these highly-respected individuals could still distort the integrity of the recommendations
  • However, the stakes are much higher/the issue is more profound for leaders in academic institutions being on company boards: as noted in the BMJ article: “unlike consultants who are compensated to provide expertise on a specific issue, directors are subject to fiduciary responsibility to company shareholders to advance the general interests of the company and increase profits”: i.e., by being on the board of a for-profit company, their explicit role is to increase the profits of the company. This is a whole level of COI above the scientist who may feel forced to accept drug company money in order to do their research. The issue with the researchers is that in the climate of public-sector cuts and the ideological and financial shift to increasing private sector support, they actually may not have so much choice (that is not to say this is a good thing, just that this is the current perverse reality for the ongoing array of political changes since 1980/Reagan presidency, with defunding public research support and shifting research to the private sector). Of course several of the leading researchers have done quite well financially as a result of this relationship (project funding, speaking engagements, ownership of small companies, etc.). But, it is really a much more significant infiltration of some of the most highly regarded academic hospitals and medical schools to have their leaders contractually (by being on the for-profit board) have an obligation to make the company profitable (of course, this COI really manifests itself when the “for profit mission of industry competes with the non-profit taxpayer funded clinical and research missions of academic medical and research institutions”, per the BMJ article. All of this is compounded by the fact that the company-based compensation is so high for these directors that it might well dwarf their academic compensation.
  • The Washington Post interviewed Wallid Gellad, one of the study leaders, who noted “Part of the impetus for [doing this study] is it’s remarkable that my $15 lunch with a pharmaceutical company is on a public Web site, but this {board of directors] information is not. It is publicly available, but you have to do a lot of searching and connecting”. As a local Boston example, “Elizabeth Nabel, the president of Brigham and Women’s Hospital, and a professor of medicine at Harvard Medical School, for example, accepted a role on the board of the medical device company Medtronic last year. According to the company’s proxy statement, she received $71,800, a prorated amount since she started midway through the fiscal year.” Despite Harvard’s own policy on faculty members serving on a company’s board of directors, which comments that as an individual’s authority within Harvard increases, “the scrutiny applied by HMS (and the affiliated institution) will similarly increase in view of the individual’s scope of authority”.  She does not exactly serve as a great role model, or seem to be in line with the HMS stated policy …. Especially since >50% of Medtronics business is in cardiac devices, $8.8 billion per year per their website, in a hospital which is remarkably cardiology-focused.

(For full article, see​ )

  1. As a somewhat related issue, the Boston Globe had an article on Harvard Medical School noting that 5 years ago they strengthened their COI rules in response to a US Senate investigation into MD payments from drug and medical-device companies. There is currently a backlash at Harvard, with some researchers claiming that the stricter regulations have “hurt innovation”. The rules that they are looking at are the “research support rule, which bars faculty who own equity in a company from receiving research grants or contributions from that company” and the clinical research rule, which prohibits “faculty from conducting clinical trials on a company’s product if they have equity in or earn at least $10,000 in income from that company”. As above, there have been shifts in the research climate and public funding which have pushed this, and one researcher commented that “professors like him didn’t set out to be entrepreneurs but have felt compelled to launch companies to test their discoveries because the pharmaceutical industry has moved away from basic research, shifting the risk to academic scientists and their institutions. Federal grant money has also become harder to get, prompting academics to launch startups that can collect funds from investors.” The reality is that the cleanest way to avoid COIs would be to re-establish a robust public sector system of funding and monitoring the research, with clearcut COI boundaries. As commented by Marcia Angell, the proposed posed changes in COI at Harvard will “be going along with the times — and the times are that everything is for sale.”

(For full article, see​ )​

EBM Author Blog: Diet Quality Inversely Associated with the Risk of COPD

26 Oct, 15 | by Kelly Horwood, BMJ

Evidence-Based Medicine Author Blog


Dr Rania A Mekary, Department of Social and Administrative Sciences, MCPHS University 

A higher overall diet quality is inversely associated with the risk of chronic obstructive pulmonary disease (COPD) in men and women

The obvious association between some modifiable risk factors such as tobacco smoke, chemicals, and dust and chronic obstructive pulmonary disease (COPD) has long been established.  Other well-established non-modifiable risk factors include age and genetics.  The question is whether there are other non-modifiable risk factors that we, and in particular physicians, need to be aware of.

According to recent prospective cohort studies, the answer is yes.  It is overall diet quality.

What is meant in here is not a specific food, a food group, or a few food groups per se; it is rather the overall diet quality as reflected by the Alternate Healthy Eating Index 2010 (AHEI-2010).  The AHEI-2010 is based on food and nutrients consistently associated with a lower risk of cardiovascular diseases, diabetes, or cancer.  More specifically, a higher AHEI-2010 diet score reflects high intakes of of whole grains, polyunsaturated fatty acids, nuts, and long chain omega-3 fats and low intakes of red/processed meats, refined grains, and sugar sweetened drinks.

A recent prospective cohort study assessed the relationship between AHEI-2010 diet score and COPD risk in men and women.


  • Data was collected from the Nurses’ Health Study cohort with 73,228 females and 1137,106 person-years and from the Health Professionals Follow-Up Study with 47,026 males and 521,764 person-years
  • The AHEI-2010 was identified from the different Food Frequency Questionnaires. Self-reported newly diagnosed COPD was defined by the affirmative response to a physician’s diagnosis of chronic bronchitis or emphysema and by the report of a diagnostic test at diagnosis.


  • Among women: 723 cases of newly diagnosed COPD; incidence rate= 64/100,000 person-years;
  • Among men: 167 cases of newly diagnosed COPD; incidence rate= 32/100,000 person-years;
  • Pooled analysis of men and women, AHEI-2010 diet score was inversely associated with newly diagnosed COPD [Multivariate-adjusted HR=0.67; 95%CI= 0.53 to 0.85 comparing top to bottom quintiles; P-trend<0.001)
  • Similar results were seen in women (RR=0.69; 95%CI= 0.53 to 0.90) and men (RR=0.60; 95%CI= 0.34 to 1.03), although the results were not significant in men; this could be due to lack of power.
  • Similar results were seen among ex-smokers and current smokers.

So what are we to take from this?

It is important to realize that the overall diet quality is an important risk factor for COPD; physicians need to stress the importance of an overall healthy diet when consulting their patients.  Whether these findings (AHEI-2010 & COPD) could differ between men and women remains unclear and merits further studies.

In fact, the root of the question is whether we have less incidence of COPD among men than women.

Read the full commentary here.



Primary Care Corner with Geoffrey Modest MD: Carotid Occlusion and Just Medical Therapy

22 Oct, 15 | by EBM

By Dr. Geoffrey Modest

The USPSTF recently restated their prior recommendation to NOT check for asymptomatic carotid stenosis (see Ann Intern Med. 2014;161:356-362​). This is based both on the potential risks of both screening and treatment, and some studies showing decreased benefit: current medical therapy may be associated with fewer strokes than when the initial studies on carotid endarterectomy were done. These older studies had found that in patients with asymptomatic carotid stenosis, carotid endarterectomy when done in a referral facility with consistently low peri-procedural morbidity/mortality, the procedure decreased the overall risk of strokes by about 50%. As a consequence of these earlier studies, a study in​2011 found that more than 90% of carotid interventions were for asymptomatic carotid stenoses in the US Medicare population (vs 0% in Denmark!!). One of the concerns about medical management of internal carotid stenosis was that progression to occlusion was likely devastating. A new study in JAMA Neurology confirmed that the current medical therapies are in fact associated with much lower progression to carotid occlusion, and that occlusion itself was rarely associated with clinical stroke, further supporting the USPSTF recommendation (See doi:10.1001/jamaneurol.2015.1843).


  • Retrospective review of data from 2 stoke prevention clinics in Canada from 1990 (when they began annual carotid ultrasound surveillance) until 2013, looking at ipsilateral stroke or TIA, death from ipsilateral stroke, or death from unknown cause
  • 3681 patients in the database (mean age 66, 71% men, 78% hypertensive with mean BP 148/80 presumably on meds, 68% hyperlipidemic), with carotid stenosis measured by ultrasound/Doppler


  • 316 (8.6%) were asymptomatic before an index occlusion that occurred during the observation period
  • Of these 316 patients, 13% had prior MI, 21% diabetes, 22% smoked/ 53% had quit smoking/12% never smoked
  • 80% of these occlusions occurred prior to 2002, when medical therapy was less intensive, with decreasing frequency over time (254 cases before 2002, 39 in 2002-7, and 7 after 2010)
  • ​Only 1 patient had an ipsilateral stroke at the time ofcarotid occlusion, and 3 had ipsilateral stroke during follow-up (all before 2005)
  • Of the 316 who had a carotid occlusion, 71 died over mean of 7.2 years after the occlusion, with the major known causes being MI (16%), cancer (13%), sudden death (11%), sepsis (11%), though 23% were of unknown cause
  • The total carotid plaque area was significantly related inversely to event-free survival, as was age and male sex (but not the % of stenosis)
  • ​Similar results were found for contralateral occlusion

So, a few issues:

  • This observational study is consistent with others that suggest that optimal medical therapy seems to be strongly associated with decreases in carotid artery stenosis progression, and that the effect of complete occlusion is not as dire as feared (likely thanks to the rich collateral circulation in the brain).
  • I think the presence of carotid stenosis, whether symptomatic or not, should trigger an aggressive medical response, since atherosclerotic disease in the carotids is a marker of atherosclerotic disease elsewhere (including the coronaries, i.e. it is a “coronary artery disease equivalent”).
  • The cornerstones of this aggressive medical response should include lifestyle interventions (diet, exercise, not smoking), as well as high-potency statins, optimizing blood pressure and diabetes control, and antiplatelet drug therapy (see
  • There is a recently started 2-center study which will formally answer the question of medical vs surgical management: the CREST-2 trial (Carotid Revascularization and Medical Management for Asymptomatic Carotid Stenosis Study) is a 4-year trial which will compare maximal medical therapy vs that plus carotid endarterectomy in one wing, and that plus carotid stent in the other, with results expected by December 2020.
  • But, for now, it the data support not screening for asymptomatic carotid stenosis and that the primary treatment is aggressive lifestyle and medical management.

Primary Care Corner with Geoffrey Modest MD: Medication in Elderly with Comorbidities

21 Oct, 15 | by EBM

By Dr. Geoffrey Modest

BMJ printed a new population-based cohort study looking at guideline-recommended drugs and deaths in older adults with multiple chronic conditions (see BMJ 2015;351:h498). These guidelines were typically based on randomized control studies with younger people and a single chronic condition.


  • 8578 older adults (mean age 77 with 36% >80yo, 59% women, 87% white), having multiple chronic conditions: hypertension (HTN)  92%, hyperlipidemia (HL) 77% , diabetes (DM) 40% , coronary artery disease (CAD) 39%, depression (DEP) 26%, heart failure (HF) 20%, atrial fibrillation (AF) 19%, chronic kidney disease (CKD) 12%, and thromboembolic disease 6%.
  • Data were from Medicare Current Beneficiary Survey cohort, a nationally representative sample of Americans >65yo, followed through 2011
  • The 9 study drugs: RAS blockers (ACE-I or ARBs) were used in 54%, statins in 53%, thiazides 47%, b-blockers 47%, calcium channel blockers (not differentiated by class) in 33%, SSRIs/SNRIs 21%, warfarin 14% , metformin 14%, clopidogrel 13%. 54% took at least 3 of the 9 study drugs. Mean total number of drugs (including nonstudy drugs) was 10. They also tracked changes in meds over the study period
  • Median follow-up of 24 months. 15% (1287) people died during follow-up.


  • Mortality over the 24 months: 27% in those with AF, 19% with CAD, 17% with DM, 33% with HF, 11% with HL, 15% with HTN
  • For the specific drugs (all of below were statistically significant), the HR for mortality was:
    • ​b-blockers: adjusted HR of 0.59 for AF, 0.70 for CAD,  0.68 for HF, and 0.48 for combo of AF/CAD/HL/HTN, 0.59 for HF/CAD/HL/HTN [note: an HR of 0.59 means a 41% decrease in mortality]
    • Statins:  HR 0.75 for CAD, 0.75 for DM, 0.68 for HL, and 0.65 for combo DM/CAD/HL/HTN, 0.68 for HF/CAD/HL/HTN, 0.70 for DEP/CAD/HL/HTN
    • Calc channel blockers: HR 0.78 for AF, 0.85 for HTN, and 0.69 for DM/CAD/HL/HTN, 0.71 for AF/CAD/HL/HTN and 0.72 for HF/CAD/HL/HTN
    • Thiazides: no significant benefits for HTN or any of the combo comorbidities
    • RAS blockers: 0.72 for HF, 0.80 for HTN, 0.82 for CAD, and 0.73 for AF/CAD/HL/HTN, 0.77 for HF/CAD/HL/HTN
    • Clopidogrel: no significant benefit for AF or CAD, or any of the combos (aspirin use could not be tracked in this database)
    • SSRI/SNRI: no significant benefit for DEP or any of the combos
    • Metformin: no significant benefit for DM or any of the combos
    • Warfarin: 0.69 for AF, 0.44 for thromboembolic dz, but no benefit for any of the combos​

So, this study, I think, is important for several reasons:

  • It largely reinforces what we are already doing, treating older patients with comorbidities based on the usual randomized controlled trials which typically limited the age range to younger patients and those usually limited to a single disease (e.g., excluding those with renal failure, etc.). And, of course, as people get older, they regularly and routinely develop multiple comorbid conditions
  • The study shows that in patients with multiple common comorbidities, the usual medications do improve mortality, even in a pretty short-term 2-year study. And the association between drug use and mortality was pretty similar across patterns of coexisting comorbidities, suggesting that similar benefits were evident despite the presence of comorbidities. This last finding supports the utility of randomized control trials limited to a single disease and then being applied more generally, at least in the above diseases/medications.​​
  • Although there is empirical evidence that observational studies usually have similar results to controlled intervention studies (see Cochrane Database Syst Rev 2014;4:MR000034), one has to remain somewhat skeptical that there could be unexpected biases. This study was quite good in that it incorporated certain social comorbidities (e.g. functional level, amount of time in the hospital, and living in a nursing facility), but still is open to the potential for other potential biases
  • For example, it is pretty clear that many of the cardiac meds do well. The hardest one for me to accept is that metformin does not have clear benefit, though on each analysis, there was a clear non-significant trend to benefit. My guess is that there is a strong selection bias here: those who are sicker do not get metformin. My bet is that they have a little (or lot) of renal dysfunction, or heart failure, etc., which scare the providers away from using metformin, leaving only the healthier elderly on it (and with a lower likelihood of showing mortality benefit in the healthier subgroup over a short 2-year study).  [Though, there are strong arguments that metformin is still a safe drug in many of these cases, esp. at a lower dose: See]
  • There have been increasing studies showing that, for example, anticoagulation for the very common condition of atrial fibrillation seems to be safer in the elderly than we thought in the old days, and I do have several patients into their late 80’s/early 90’s doing very well on them. And we know that the benefits of statins are typically evident within 6 months of starting them (based on studies of mostly younger people but some elderly) – i.e., these drugs seem to work well and are pretty well tolerated in the elderly.
  • So, bottom line, this study provides some pretty strong scientific rationale for continuing to treat elderly patients with their common multiple medical conditions with the same meds we have been using based on studies of younger people with single diseases. ​But, it is also important to bear in mind that this study only looked at mortality, which is not the only important end-point….

Primary Care Corner with Geoffrey Modest MD: International Travel and Gut Microbiome Changes

20 Oct, 15 | by EBM

By Dr. Geoffrey Modest

Last March I sent out a blog reviewing a Finnish study that found changes in intestinal colonization in those traveling to a variety of places, including South Asia and Sub-Saharan Africa, looking at a few specific resistant bacteria (extended-spectrum beta-lactamase-producing Enterobacteriaceae and carbapenemase-producing Enterobacteriaceae), and finding significant colonization overall but especially in those who developed traveler’s diarrhea and even much more so in those who took antibiotics for it (see ).

The current study adds greatly to the concern about GI microbiome changes with travel by looking at the larger microbiome, not just a couple of concerning species as above (see doi:10.1128/AAC.00933-15​).


  • They used “explorative shotgun metagenomic sequencing” (???? you might ask — turns out to be a new technology to do quantitative analysis of pretty much all known antibiotic-resistance genes, >300 genes, in a single assay looking at the total DNA of the gut, whether the bacteria is cultivable or not) in 35 Swedish students aged 23-34, 74% female, comparing their fecal samples before and after exchange programs in the Indian peninsula (n=18) or in Central Africa (n=17), from 2010-2013. The median travel duration was 34 days. None were exposed to antibiotics, though 23 of the 35 did participate in hospital-based patient work. In doing the metagenomic sequencing, they limited their assessment to genes conferring verified resistance phenotypes (i.e., those genes known to be associated with resistant bacterial infections).
  • 69% (12 of 17 going to Central Africa and 12/18 to the Indian peninsula) had travelers’ diarrhea for a median of 4 days.
  • None were exposed to antibiotic treatment in the 6-month period before or during the travel, though those who went to Central Africa did have antimalarial chemoprophylaxis (mostly mefloqine, which may have some broader antibacterial effect).


  • The overall taxonomic diversity and composition of the microbiome was pretty stable from before to after travel, though there was an increase in Proteobacteria in 25 of the 35 students (Proteobacteria have been associated with inflammatory conditions in the large bowel, including ulcerative colitis, difficile enteritis, and chronic HIV).
  • BUT, there was an overall increase in the abundance of antibiotic resistance genes, especially for those encoding resistance to:
    • ​Sulfonamides (2.6-fold increase)
    • Trimethoprim (7.7-fold increase)
    • Beta-lactams (2.6-fold increase)
    • And significant increases, but less profound, for aminoglycosides and tetracycline
    • Of​ those traveling to the Indian peninsula, 12 acquired extended-spectrum beta-lactamase-producing (ESBL) E. coli, though none from Africa. Of some concern, these were not picked up on the above genetic sequencing, likely because the sensitivity of the assay was insufficient for low-abundant genes.
    • The above microbiome changes were not significantly related to those who reported experiencing traveler’s diarrhea (and, as mentioned, none had taken antibiotics)

So, a few observations:

  • This metagenomic technique appears to be pretty powerful in assessing microbiome changes, though not perfect (e.g., it missed some ESBL E. coli)
  • In general, we think of the major changes in microbiome being from use of antibiotics, poor infection prevention/control within health care systems, poor sanitation, antibiotic pollution of the environment (e.g. massive use of antibiotics for animals), international food trade and travel. I would add to this list other medications and dietary/lifestyle changes (as delineated in prior blogs), though I think this study does raise the issue of travel-acquired changes even more. We do not know if there was any transmission of antibiotic-resistant genes back home or even if the microbiome changes continued/for how long (the final stool samples were a median of 3 weeks after return from travel, with max of 4 months). We don’t even know if the quantitative increase in antibiotic resistance genes is associated with clinical disease. In addition, it would be useful to know if others living with the travelers had any changes in their microbiomes. but, overall, the microbiome changes from travel are still concerning and have the very real potential to lead to difficult-to-treat diseases spreading throughout the world, and doing so more often as more antibiotic resistance continues to develop.

Primary Care Corner with Geoffrey Modest MD: Obesity in Kids and Cardiometabolic Risk

19 Oct, 15 | by EBM

By Dr. Geoffrey Modest

So, no great surprise, but it seems that cardiometabolic risk factors track the degree of obesity in children and young adults (see N Engl J Med 2015;373:1307-17​). A cross-sectional analysis of overweight and obese kids/young adults (age 3-19) in the National Health and Nutrition Examination Survey from 1999-2012 looked at measured height and weight, along with an array of cardiometabolic risk factors (lipids, A1c, etc.), assessing the relationship over different obesity levels.

Background (from the CDC):

  • Childhood obesity (defined as BMI >95th %ile) has more than doubled in children and quadrupled in adolescents in past 30 years: % of children aged 6-11 in the US who were obese increased from 7% in 1980 to 18% in 2012; for adolescents 12-19yo, it increased from 5% to 21%.
  • In 2012 >1/3 of children and adolescents were overweight or obese
  • Despite recent declines in obesity prevalence in preschool-aged kids, obesity is still way too prevalent: overall for those 2-19 yo, the prevalence “has remained fairly stable at about 17% and affects about 12.7 million children and adolescents for the past decade”, though the prevalence in those 2-5 yo has decreased significantly from 13.9% in 2003-4 to 8.4% in 2011-2, at which time the prevalence was 17.7% in 6-11 yo and 20.5% in 12-19 yo; the prevalence was highest in Hispanics (22.4%) and non-Hispanic blacks (20.2%) vs non-Hispanic whites (14.1%)

Details of study:

  • 8579 individuals (53.7% white, 16.5% black, 24.0% Hispanic; 52% male) with BMI>85th %ile, of whom 46.9% were overweight (BMI 85-95th%), 36.4% had class I obesity (95-120% of the 95th %ile), 11.9% had class II obesity (120-140% of the 95th %ile or BMI≥35), and 4.8% had class III obesity (≥140% of the 95th %ile, or BMI≥40).

Results, as progress from overweight to class I to class II to class III obesity:

  • LDL: 94.6, 98.4, 98.2, 96.5 (p=0.131, non-significant)
  • HDL: 49.4, 46.7, 43.5, 41.3 (p<0.001)
  • Systolic BP: 108.5, 111.0, 112.6,116.2 (p<0.001)
  • Diastolic BP 57.0, 58.8, 58.7, 64.5 (p<0.001)
  • Fasting triglycerides: 91.0, 113.2, 113.3, 143.2 (p<0.001)
  • Glycohemoglobin 5.15, 5.20, 5.30, 5.37 (p<0.001)
  • Fasting glucose: 93.2, 95.1, 96.7, 96.5 (p=0.001)
  • And, overall, these risk factors did have a sex difference: males did worse. In fact, the only significant ones for females were: HDL, systolic/diastolic BP, glycohemoglobin and glucose; and for each of these, the prevalence in males was much higher.

So, although this study tracked only the surrogate markers of cardiometabolic parameters (not so likely to have cardiac clinical events at this age…..), this study is important because:

  • Obesity in kids tracks with obesity in adults
  • This study, vs older ones, looks at levels of obesity and differences in risk factors, showing a graded response overall: the worse the obesity, the worse the risk factors. And, this study, I think, justifies subdividing obesity in kids into different levels (since there are differences in attributable cardiometabolic risks), as is done with adults.
  • And, though clinical events are the gold standard, autopsy studies have shown that there are fatty streaks in pretty much everyone aged 15-34; and there are advanced atherosclerotic lesions in 2% of men/0% of women​ aged 15-19 and 20% of men/8% of women aged 30-34. So, actual disease does begin early and, per usual, is seems better to deal with risk factors early on, before clinical disease manifests itself….

Also, for prior blog on the relationship between pediatric obesity and increased left ventricular mass from the Bogalusa Heart Study, see


Primary Care Corner with Geoffrey Modest MD: Pulmonary Embolism Evaluation

16 Oct, 15 | by EBM

By Dr. Geoffrey Modest

The Clinical Guidelines Committee of the Am College of Physicians published a paper detailing what they consider to be “best practice advice” for the evaluation of patients with suspected acute pulmonary embolism –PE– (see doi:10.7326/M14-1772), focusing on what they see as the overuse of CT scans and plasma D-dimer tests.

Their advice:

  • Use validated clinical prediction rules to estimate the pretest probability of PE (e.g. Wells prediction rules or revised Geneva score, included below)
  • Patients who have low pretest probability of PE and who also meet all of the Pulmonary Embolism Rule-Out Criteria (PERC, included below) should NOT have a D-dimer done
  • Patients with an intermediate pretest probability of PE, or low pretest probability but do not meet all of the PERC, should have a high-sensitivity D-dimer done
  • In patients over 50yo, use age-adjusted D-dimer thresholds (agex10ng/ml, rather than a universal cutoff of 500 ng/ml)
  • In those with D-dimer levels below the above threshold, do NOT do any imaging (studies show way too many are done in ERs with no improvement in patient outcomes, though lots of radiation exposure and expense (see more reviews on excesses of radiation exposure)
  • In patients with high pretest probability of PE, go directly to CTPA (CT pulmonary angiography). Use ventilation-perfusion scans only if contraindication to CTPA or CTPA not available. Do NOT get D-dimer​ in addition

A few supportive points:

  • D-dimer has low specificity, but a normal high-sensitivity D-dimer in recent studies had a 99.5-100% sensitivity for excluding PE on CT scan
  • The PERC (which is not a screening tool for all patients but only those with signs/symptoms potentially suggestive of PE) in a large meta-analysis has found that the overall proportion of missed PEs was 0.3%, with pooled sensitivity of 97%. In those with score of “0”,  they state “the risk for PE is lower than the risks of testing”
  • Of note, pregnancy, heart failure or stroke are not part of the Geneva score because they do NOT add to its predictive performance
  • They do stress that alternative approaches may be okay: e.g., if someone has concerns for PE, are hemodynamically stable, and have lower extremity symptoms, it is reasonable to get leg ultrasound, and anticoagulate if positive (i.e., spare the radiation exposure, since the patient will be treated anyway). Though one issue is length of therapy here. Those who get a PE are at higher risk for a recurrent PE than those who just have a DVT. They comment that for patients with “cardiothoracic symptoms, the need for long-term anticoagulation can be determined after the initial treatment period”. I’m not sure what that means exactly. As I have mentioned in previous blogs, there are some interesting data supporting either checking D-dimer levels before stopping therapy or 3-4 weeks after stopping it (I actually do both), and in those with normal D-dimers one can stop therapy (this is not accepted universally as strategy, in part because those with normal D-dimers can still get PEs, though the studies suggest they are much less likely. I do discuss the risks and benefits of stopping anticoagulation with patients and try to make a joint strategy) — See​ for more detailed articles/critiques.



Wells prediction rules:


Revised Geneva score:




Primary Care Corner with Geoffrey Modest MD: Low LDL and Diabetes Risk

15 Oct, 15 | by EBM

By Dr. Geoffrey Modest

As is evident in several studies, there is a relationship between the use of statins and the development of diabetes (approx 9% increase in diabetes incidence). A recent study looked at the relationship between LDL levels themselves and the development of diabetes (see DOI 10.1007/s00125-015-3762-x).


  • Data from the Framingham Heart Study offspring cohort, with 6011 people and 14120 person-observations (mean age 50, 56% women, mean LDL 125 mg/dl) who were not on any lipid-modifying or antihypertensive medication, followed a mean of 4.5 years
  • Assess the development of diabetes (fasting glucose >125 mg/dl or put on glucose-lowering meds), comparing that to their LDL levels as well as a genetic risk score (GRS) –different genetic changes in single-nucleotide polymorphisms (SNPs) which affect LDL levels


  • 312 people (2.2%) developed diabetes
  • A higher LDL level was associated with a lower risk of diabetes in a graded fashion [OR per SD decrement was 0.81 (0.70-0.93, p=0.004)]
  • The GRS was similarly associated with incident diabetes, both in direction and in magnitude [OR per SD decrement was 0.85 (0.76-0.96, p=0.009)]
  • The increased risk of diabetes was similar across age, sex, BMI, fasting glucose, HDL, or triglyceride levels

A few background issues:

  • It may be more than a coincidence that 2 drugs (niacin and statins) that lower LDL levels are both associated with the development of diabetes
  • Other studies have also found that the GRS for LDL was also associated with diabetes
  • 3 observational studies have found lower LDL levels in people with insulin resistance and diabetes
  • A large Danish database of patients with familial hypercholesterolemia and very high LDL levels (86% of whom had LDL receptor mutations) has found a lower incidence of diabetes (overall OR of 0.45 for those with LDL receptor mutations), with a graded association: the more severe the mutation (assoc with higher LDL levels), the lower the risk of diabetes (see JAMA 2015;313:1029), raising the question that the issue is not statins per se, but the role of the upgraded LDL receptor in predisposing people to diabetes, and raising further the issue of that receptor’s role in glucose homeostasis (see JAMA 2015;313:1016).
  • Patients having variants of the HMG Co-A reductase gene, the enzyme targeted by statins, have increased blood sugar and diabetes risk (i.e., both in those on statins and those with this genetic variant, both of which increase the expression of LDL receptors). the authors postulate that pancreatic β-cells, in the setting of familial hypercholesterolemia, have decreased cholesterol uptake (genetic impairment of LDL receptors) and therefore improved β-cells function and survival (there are a slew of animal studies and some human tissue culture ones supporting the conclusion that those with enhanced LDL receptor activity and the resulting cholesterol-laden pancreatic β-cells have impaired function of the β-cells, see JAMA 2015;313:1029).

So, what does this all mean?? It raises a few issues: perhaps it is not the statins that are causing diabetes, but the lower LDL levels themselves. and, if it turns out that this is mediated through the increased LDL receptors as is likely, and this increase is augmented by statins as well as some mutations, then perhaps developing other drugs which lower LDL levels but do not affect the LDL receptors might be more beneficial (by the way, the PCSK9 inhibitors increase LDL receptor concentrations even more than statins). But at this point, all analyses suggest that lowering of LDL is still beneficial overall, despite the potential development of diabetes. See for a review of statin adverse effects.

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