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.


Primary Care Corner with Geoffrey Modest MD: Prescribing Alcohol to Diabetics?

13 Nov, 15 | by EBM

By Dr. Geoffrey Modest

An Israeli study was just published in which adults with type 2 diabetes were randomized to drinking wine vs water, and an array of cardiometabolic  parameters were assessed (see Ann Intern Med. 2015;163:569-579​) — the CASCADE trial: CArdiovaSCulAr Diabetes & Ethanol trial, which just goes to show you that you can develop an acronym pretty easily for just about anything.



  • 224 patients (baseline: mean age 60, 69% men, BMI 30.0, HDL 1.12 mmol/L or 43.5 mg/dL, LDL 2.41 mmol/L or 93.0 mg/dL, cholesterol/HDL ratio 4.1, fasting plasma glucose 150.4 mg/dL or 8.3 mmol/L, HgbA1C=6.9%, BP 137/78, waist circumference 105 cm, and mean positive metabolic syndrome criteria was 3.1 of 5) were randomly assigned to 150ml of mineral water, white wine or red wine with dinner daily for 2 years. Previously, these subjects had drunk no more than 1 drink of alcohol/week (mean 2.3 g/d)
  • All followed a Mediterranean diet without calorie restriction
  • They also looked at alcohol metabolism/ ADH1B polymorphism, where 36% were CC, 38% were CT and 26% were TT (the TT polymorphism of ADH1B is also called ADH1B*2 rs1229984, and is associated with much faster hepatic clearance of alcohol); and assessed the homeostatic model assessment of insulin resistance score (HOMA-IR, a measure of insulin resistance), which was 5.0 at baseline


  • Those on red wine vs water had: increased HDL [0.05 mmol/L (0.04=0.06), or 2.0 mg/dl (1.6-2.2), p<0.001] and apolipoprotein A1 [0.03 g/L (0.01-0.06),p=0.05], and decreased total cholesterol/HDL ratio by 0.27 [(-0.52 to -0.01), p=0.039]. Those drinking white wine were not significantly different from water in terms of lipid changes (except that, interestingly, both the red and white wine drinkers did have decreases in triglycerides by .09 mmol/L or 7.9 mg/dL with white wine and -0.1 mmol/L and 12.0 mg/dL with red wine). White wine, but not red wine, was associated with significant decreases in fasting plasma glucose levels (decreased 1.0 mmol/L, or 17.2 mg/dL) and HOMA-IR scores.
  • Only when looking at the slow alcohol metabolizers (ADH1B*1 carriers), both red and white wine had better fasting plasma glucose, HOMA-IR​, and hemoglobin A1C levels similarly
  • ​And, sleep quality improved in those drinking wine (p=0.040)
  • But no difference in blood pressure, adiposity, liver function, drug therapy, symptoms, or quality of life
  • Overall, red wine led to decrease in number of components of the metabolic syndrome by 0.34 [(-0.68 to -0.001), p=0.049], not with white wine

So, there are a few issues here

  • There have been many studies and meta-analyses from observational trials suggesting that moderate amounts of alcohol ingestion are cardioprotective, largely attributed to the alcohol content itself. For more discussion of this, see for blogs on alcohol, and​ which argues that the attributable benefits from even small amounts of alcohol consumption are likely related to inherent biases in the observational data collection.
  • This study is the first I’ve seen where there is an actual intervention done. Of course, the intervention is not really through a randomized controlled trial since I suspect that those drinking wine were in fact aware that it was not mineral water. And, one wonders if there might have been other changes in the diet related to drinking wine with the meal, though all were instructed in Mediterranean diet guidelines – i.e., they did not monitor the actual dietary composition, and the glass of wine could have altered the choices or quantities of foods consumed.
  • As noted in prior blogs, not all HDL is the same, and there are nonfunctional and even pro-inflammatory variants (see , and will append an older blog on pro-inflammatory HDL at the bottom). So, looking at HDL numbers may not necessarily translate to cardioprotection.
  • If the findings of the study are indeed valid, they suggest that the effect of wine on diabetic markers is basically through the alcohol itself (since both red and white wine improve the diabetes markers especially in the slow metabolizers who have more sustained blood alcohol levels), while the lipid effects were more evident with red wine, suggesting that its particular components may be protective (perhaps the phenolic compounds: resveratrol and quercetin), perhaps through their antioxidant, endothelial or antiplatelet actions (though in general it is felt that the quantity of these in red wine is insufficient to achieve therapeutic effects)
  • So, interesting study, though I don’t think it provides definitive answers to the question. I.e., I’m not ready to suggest alcohol to patients yet… (and, per the 5/14/12 blog, the real answer will come from looking at real clinical events and not the surrogate markers of changes in cholesterol, since alcohol also increases the proinflammatory variant).

Here is blog from 5/14/12

Although the vast majority of epidemiologic studies have found HDL to be cardioprotective, there have always been some concerns. HDL is comprised of a diverse group of lipoproteins with significant metabolic heterogeneity. There were a few older studies finding a “pro-inflammatory HDL”, which predisposed people to heart disease. The clinical trial of Torcetrapib, a cholesterol ester transfer protein inhibitor, dramatically increased HDL but was not cardioprotective. The researchers suggested that the HDL was somehow deformed. (This large torcetrapib trial overwhelmed a meta-analysis last year in BMJ, suggesting no benefit to raising HDL). In any event, there is a likely very illuminating article from Harvard school public health (see doi:10.1161/JAHA.111.000232). They had found before that there was occasionally a small apolipoprotein (apo C-III) on some lipoproteins causing a pro-inflammatory and atherogenic response. On LDL particles, this apo C-III caused increased coronary atherosclerosis independent of the LDL itself. They looked at the data from 2 large epidemiologic studies — Nurses health study (NHS, 121K female nurses) and the health professionals follow-up study (HPFS, 52K males), looked at stored serum and assessed the HDL C-III relation to cardiac events.


  • 14% of women in NHS had HDL with apo C-III; 11% of men in HPFS had apo C-III
  • Overall, each standard deviation increase in HDL was assoc with a 21% dec in cardiac events; but for patients without apo C-III, there was a 34% decrease in events and for those with apo C-III there was a statistically significant 18%
  • Looking at the effect of apo C-III in multivariate analysis of other risk factors (all of below statistic signif):
    • Compared to pts with normal wt, overweight and obesity were associated with 7% and 12% lower levels of HDL without apo C-III – i.e., overwt/obesity with lower of the good HDL.
    • Alcohol was assoc with 3% higher levels of both HDL types
    • Smokers had 1% higher levels of HDL with apo C-III (the bad one)
    • Premenopausal women had 9% higher levels of HDL without apo C-III, as did postmenop women on estrogen replacement therapy, vs other postmenop women
    • Per SD increase in triglycerides, 8% lower HDL without apo C-III and 15% increase in HDL with apo C-III
    • Per SD increase in A1C, 4% increase in HDL with apo C-III

(i.e., several of these risk factors which change HDL also lead to more HDL with apo C-III)


Primary Care Corner with Geoffrey Modest MD: Raise the alcohol tax!

15 Apr, 15 | by EBM

By: Dr. Geoffrey Modest

An article was just published in the Washington Post arguing for an increase in the alcohol tax (see here​).


They make the following points:

–There has been a remarkable cheapening of alcohol: a 10-drink/day drinker in 1950 had to spend 45% of their disposable income on the alcohol. today it is 3%.

–Taxes have fallen dramatically: in today’s dollars, the federal tax on a shot of 80-proof whiskey was 90 cents in 1951, now it’s 13 cents (7-fold decrease). Beer tax has decreased 5-fold.

–Cigarettes, in contrast, have less political support and have had record high taxes (and significantly decreasing consumption)

There is a voluminous literature finding that increased alcohol prices through state/national taxes decrease alcohol consumption, alcohol-related morbidity and mortality, and alcohol-related harmful social interactions (violent assaults, drunk-driving, sexually-transmitted infections). A few studies have been published looking at the effects of a small alcohol tax increase in Illinois in 2009, where the tax on beer was increased by 4.6 cents/gallon, wine by 66 cents/gallon, and distilled spirits by $4.05/gallon.  In one study researchers looked at the relationship between this tax increase and sexually transmitted infection (STI) morbidity, finding that state-wide rates of gonorrhea decreased 21% (3506 fewer infections) and chlamydia decreased 11% (5844 fewer infections); these decreases were (not surprisingly) highest among 25-29 year olds (see Addiction 2014; 109: 904–912). In a study just published on-line in the Am J Public Health this group also looked at fatal car crashes, finding that the tax increase was associated with a 26% decrease in fatal alcohol-related car crashes (decreased 9.9/month), with a striking 37% decrease in those <30 yo. Similar reductions were found in those with alcohol levels <0.15 grams/dl (-22%) vs higher (-25%), and there was no significant gend​er or race differences. They controlled for weather, traffic law enforcement, safety policies, and compared their results to those of neighboring Wisconsin where there was no tax increase. Of note, alcohol-related car crashes account for 10K deaths and 500K injuries per year in the US.

So, there have been lots of studies in the public health journals showing that taxes lead to lower levels of harmful-substance use and fewer medical and social consequences. Even with the seemingly paltry tax increase in Illinois, there were pretty dramatic decreases in alcohol-related fatal car crashes and STIs. None of this is surprising. There have been very old observations (eg Prohibition in the US, German seizure of alcohol in France in World War II, Gorbachev’s anti-alcohol campaign in Russia) leading to major decreases in liver cirrhosis mortality, typically within 6-12 months. This all highlights the major health improvement attributable to decreasing alcohol consumption. Seems pretty clear that the current trend in effectively decreasing alcohol taxes/cost is wrong-headed, though unfortunately there are powerful political lobbies in the US which will make it difficult to change this…. (oh, wouldn’t it be great if public health concerns had a bit more of a role in determining policy??).​


Primary Care Corner with Geoffrey Modest MD: Substance use and mental health in youth and adults 2014

2 Apr, 15 | by EBM

By: Dr. Geoffrey Modest

SAMHSA (Substance Abuse and Mental Health Services Administration) just released a 2014 report on where we are at in the US regarding substance abuse and mental health issues (see here). A pretty brief summary:

Youth substance use/mental health (note: there are different surveys used for these data):

–Past month marijuana use, 2002-2013

–slight decrease in adolescents 12-17 to 7.1% (peak in 2011 at 7.9%)

–9th-12th graders without much change, at 23%. 8th-10th graders at 12.5% (sl increase)

–Past-year nonmedical pain reliever use, 2002-2013

–12th grade males: significant decrease, was up to 11.6% in 2002, decreased to mid-10 range til 2009, now down to 8.4%

–12th grade females: pretty flat at around 8% til 2011, then decreased to 5.6%

–age 12-17 males and females was in the 7.2% range, now decreased to 4.5-4.8%

–Past-month illicit drug use, 2009-2013

–age 12-17 with overall decrease in total, White, Latino to 8.8%, but no change in black, at 10.5%

–comparing the different illicits: marijuana most common at 7.1%; nonmedical use of psychotherapeutics next at 2.2%; then hallucinogens, inhalants, cocaine and heroin, all 0.6% or less.

–Past-month cigarette use, 2009-2013

–age 12-17: decreasing in all ethnicities, but highest in white (7.2%), then latino (3.7%), then black (3.2%)

–Past month binge alcohol use, 2008-2013

​–age 12-17: similar in males and females, decreasing to 6.2%. highest in White (7.3%), then Latino (6.3%), then Native American/Alaskan (5.6%), then Black (3.9%), then Asian (2.8%)

–Past year initiated substance use, in age 12-17, from 2009-2013

–alcohol: dec from 10.8% to 9.7%

–marijuana: dec from 5.5% to 4.8%

–cigarettes: dec from 5.2% to 3.7%

–nonmedical use of psychotherapeutics: dec from 3.5% to 2.4%

–and pretty similar %’s of above for white, black, latino

–Major depressive disorders, in age 12-17,in 2013

–overall: 10.7% (2.6 million…)

​–males: highest in Asian (6.5%), then Latino (5.7%) and White (5.6%), then Black (3.0%)

–females: highest in Latino (17.4%), then White (16.3%), then Black (14.4%), and Asian (13.9%)

–Past year depression treatment for major depressive disorder, among age 12-17, in 2103

–around 60-70% did not receive treatment (males 70%, females 59%), and White 58%, Black 71%, Latino 63%

Adult substance use/mental health

–Suicidal thoughts, age 18 and older, in 2013

–highest in 18-25 (7.4%), then decreasing to 4.0% age 26-44, 3.7% age 45-64, and 1.5% if older than 65

–also higher in those not insured (5.7% vs 3.6%) and in those <100% of federal poverty level (6.6% vs 3.5%)

–Serious mental illness, age 18 and older, in 2103

–more female (4.9% vs 3.5%), more in uninsured (5.9% vs 3.9%), and more in those <100% of poverty level (7.7% vs 3.6%)

–did not receive treatment: more men (36.4% vs 28.4%), and more younger (46% in 18-25 yo vs 24% in 45-64), and (not a shocker) more without insurance (49.4% vs 28.6%)

–Alcohol dependence and abuse, age 12 and older, in 2103

–more male (8.7% vs 4.6%), most in the 18-25 age group (13%) then decreasing stepwise through the >65 yo group (2.1%)

–more in uninsured (9.7% vs 6.0%)

–trend decreasing from 2009-2013 (eg, in those 18-25, decreased from 16.1% to 13.0%)

–Illicit drug use, aged 12 and older, in 2013

–highest in 18-25 yo (7.4%),  then decreasing stepwise through the >65 yo group (0.4%)

–Enrollment in substance abuse treatment (single-day counts 2009-2013)

–increasing from 2010 to 2013, from 1.175 million to 1.250 million, pretty evenly divided between “drug use only” at 39.6% and “both drug and alcohol use” at 43.0%

–Past-year treatment for alcohol use, age 12 and older, in 2013

–Highest in 12-17yo (9.8%), then rest in 5.3%-7.3% range

–Of note, 90.6% did not receive treatment and did not perceive a need for treatment!

–People in opioid treatment programs 2009-2013

–methadone: increase from 2009 from 283K to 330K in 2013

–buprenorphine: increase from 2009 from 24K to 48K in 2013

–in 2013, only 13.4% were in treatment, 5.7% perceived need for treatment but did not receive it, and 80.9% did not perceive need for treatment.  those in treatment tended to be older (26-44 yo group, at 18.2%)

So, a few issues from the mass of numbers above:

–These surveys are subject to many biases in terms of reporting, both by sample selection and the general cultural issues (in the largest sense, including those related to ethnicity, as well as those relating to gender, age, local social environment, peer pressure, etc which affect one’s level of introspection/ability to verbalize it as well as interpreting what is “normal”)…all limiting the generalizability and interpretation of the results

–The trends for most of the above regarding substances are in fact getting better!! as well as the use of methadone/buprenorphine, etc

​–One sore point is the psych data: large numbers of people with serious mental illnesses. with lots who are poor and without insurance (there is certainly an aspect of “social drift”, in that those with serious mental illnesses tend to “drift” into poverty, but there is undoubtedly a component of poverty leading to lack of insurance, leading to lack of treatment, leading to lack of ability to function in society, leading to lack of ability to improve the situation…..


Primary Care Corner with Geoffrey Modest MD: Moderate alcohol and cardioprotection????

20 Feb, 15 | by EBM

By: Dr. Geoffrey Modest

alcoholI have been one of those occasionally advocating small amounts of alcohol to some of my patients (ie, those with no personal or family history of addictions  and who were resistant to taking statins despite their increased cardiovascular risk) for several reasons. There have been large meta-analyses (for example, see Ronksley P E et al. BMJ 2011;342:bmj.d671  — meta-analysis of 84 studies with 11 years followup) finding 25% decrease in CAD incidence and mortality and 13% decrease in total mortality in moderate alcohol consumers. And, there was reasonable biological plausibility (reviewed by Brien, et al BMJ 2011) finding increased HDL, slightly decreased LDL, increased apolipoprotein A1, decreased fibrinogen, and increased adiponectin (which is decreased in the setting of insulin resistance/diabetes and negatively correlates with insulin resistance) — all of which should decrease CAD.  There are many clinicians strongly against our promoting a glass of wine with meals, partly because of the overall devastating effects of excessive alcohol intake (and that we as providers should not suggest a toxic chemical) and partly because of very likely biases in the retrospective and uncontrolled studies on moderate alcohol consumption. Some of these biases, for example, include the fact that “non-drinkers” in many of the studies included former drinkers (who from several other studies have more depression and increased mortality than “never drinkers”); in addition, those who drink small amounts of wine with dinner might well take care of themselves better (healthier diet, more exercise, other factors??) than those who either drink much more or do not drink at all. In this light, a new BMJ article found minimal benefit of alcohol (see BMJ 2015;350:h384).


–analyses were done of 10 waves of the Health Survey for England from 1998-2008, in adults >50 yo​.

–the Health Survey is an annual cross sectional survey of the non-institutionalized general English population, which included questions about self-reported average weekly alcohol consumption and consumption on the heaviest day in the past week. the survey assessed “former drinkers” as well as “never drinkers”

–all-cause mortality was based on recorded deaths, with prior consent by the patient to link to the Health Survey database (n=18,368)


–controlling for age (their Model 1), weekly use and use on heaviest day were protective at all levels of consumption and in all age groups except women 50-64 yo (where consumption of >20 units/week was insignificant)

–controlling for age, BMI, economic activity, education, ethnicity, country region, marital status, smoking status, and social class (their Model 2), and not including “former drinkers”, the only groups with protection from alcohol were women >65 yo consuming <=10 units/week or up to 3 units on their heaviest consumption day as well as those who consumed at a frequency of <=2 drinking occasions/month, and men 50-64 yo consuming 15-20 units/week or drinking up to 1.5 units on their heaviest consumption day (though the magnitude of protection in men was “minimal”).

So, this study adds to others which have found that when “former drinkers” are removed from the non-drinker category, the protective effect of alcohol is either attenuated or nullified.


Tim Naimi published an editorial in the journal Addiction, which argues strongly that moderate alcohol consumption should not be considered cardioprotective at this point and that we should not be recommending that patients drink (see doi:10.1111/add.12828). He (and coauthors) cite impressive data from several Mandelian randomization studies (which decreases the likelihood of confounding) which seem to contradict a direct role of alcohol in decreasing both CAD biomarkers as well as several of the attributed “health-promoting” effects of moderate alcohol (cardiac, but also cognitive function, blood pressure, and prenatal exposure with both balance and academic achievement in kids).


Primary Care Corner with Geoffrey Modest MD: Alcohol Consumption and Atrial Fibrillation

14 Oct, 14 | by EBM


A prospective Swedish observational study assessed the level of alcohol consumption and the risk of atrial fibrillation (see DOI: org/10.1016/j.jacc.2014.03.048). In this study, 79,019 men and women from 2 large population-based cohorts (the Cohort of Swedish Men and the Swedish Mammography Cohort), initially free of atrial fibrillation (AKA afib), completed alcohol ingestion questionnaires (as part of 350-item questionnaire), followed for >12 years (with 860K person-years of followup).


–7245 incident cases of afib

–Compared with current drinkers of <1 drink/week (with 12 grams alcohol/drink), multivariate relative risks were:

–1-6 drinks/wk: RR=1.01 (0.94-1.09)

–7-14 drinks/wk: RR=1.07 (0.98-1.17)

–15-21 drinks/wk: RR=1.14 (1.01-1.28)

–>21 drinks/wk: RR=1.39 (1.22-1.58)

–No difference in association with alcohol by gender. Above RRs were similar if excluded binge drinkers (>5 drinks on single occasion)

–In their meta-analysis of 7 prior studies with 12554 incident cases of afib, they found relative risks:

–for 1 drink/day: RR=1.08 (1.06-1.10)

–for 2 drinks/day: RR=1.17 (1.13-1.21)

–for 3 drinks/day: RR=1.26 (1.19-1.33)

–for 4 drinks/day: RR=1.36 (1.27-1.46)

–for 5 drinks/day: RR=​1.47 (1.34-1.61)

So, pretty impressive association in both their study and the meta-analysis that moderate alcohol consumption is associated with the development of atrial fibrillation, and that this association is not just with binge drinking (“holiday heart”). However, on the other side, is the lower risk of coronary artery disease with moderate alcohol consumption (though there is argument that these studies showing benefit of alcohol are pretty biased). Seems reasonable to me (though untested) that people at high risk of developing afib (e.g., some combination of advanced age, systolic hypertension, heart failure, increased BMI, mitral valve disease, frequent PACs or increased PR interval) be advised to avoid alcohol on a regular basis.


Primary Care Corner with Geoffrey Modest MD: Meds for Alcohol Relapse

11 Aug, 14 | by EBM

a comprehensive systematic review and meta-analysis of studies looking at pharmacotherapy for alcohol use disorders (which are associated with 3x increase in early mortality, only 30% of patients receive treatment, and <10% are prescribed meds to decrease alcohol consumption) – see Reviewed 123 studies, almost all RCTs, with 22.8K participants. Mean age in 40’s. studies from 12-52 weeks long. Mostly enrolled patients after detox or at least 3 days off alcohol. Results (significant findings only):

–acamprosate tid (a glutamine antagonist and γ-aminobutyric acid agonist): NNT (# needed to treat) to prevent return to any drinking=12;

–naltrexone 50mg/d: NNT to prevent any drinking =20; NNT to prevent return to heavy drinking=12.

–In trials comparing these 2 drugs – no signif difference

–injectable naltrexone: no effect on return to drinking or heavy drinking, but 4.6% reduction in heavy drinking days

–nalmefene (one tab daily, off-label):  decrease of 2 heavy drinking days/month

–topiramate (off-label): 9% fewer heavy drinking days

–adverse effects: naltrexone and nalmefene – withdrawal from trials higher, esp for dizziness and gi effects  (# needed to harm=48). No signif adverse effects for acamprosate or topiramate

–they did look at studies of other off-label meds (aripiprazole, atomoxetine, desipramine, fluvoxamine, gabapentin, imipramine, olanzapine, ondansetron, and paroxetine – all with only one trial. Multiple trials for baclofen, buspirone, citalopram, fluoxetine, quetiapine, sertraline, valproic acid and varenicline), insufficient evidence to support their use.

–No significant benefit of disulfuram in the 4 studies found (data inconsistent. Some found that fewer drinking days in those returning to drinking. Also small numbers of patients).

–Nalmefene and topiramate considered  “somewhat effective”, but disulfiram is not.

–in many of these meds, improved results when combined with psychosocial intervention, esp studies with acamprosate and oral naltrexone.

So, pretty striking how little we are using meds, some of which have pretty good results for a very difficult, chronic, debilitating condition for many patients. Acamprosate and oral naltrexone are the best, with acomprosate being a t.i.d. med (less convenient) and contraindicated with severe renal impairment. Naltrexone being more convenient (once a day), but contraindicated if acute hepatitis, liver failure, concurrent opioid use, and with more adverse effects. The potential benefit of increasing treatment coverage to 40%, as per the authors, would reduce alcohol-attributable mortality by 13% in the European Union. Cost of a month supply of acamprosate or naltrexone is $140. ?cost of nalmefene (prescribed as one pill daily, but i am unable to get a price for it). best if meds supplemented with psychosocial intervention.


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