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: Drink Coffee and Live Longer

7 Dec, 15 | by EBM

By Dr. Geoffrey Modest

An analysis of the Nurses’ Health Studies (NHS, which began in 1976 with RNs aged 30-55 and  NHS2 which began in 1989 with RNs aged 25-42) and the Health Professionals Follow-up Study (HPFS, which began in 1986, with male health professionals aged 40-75) assessed total and cause-specific mortality in drinkers of caffeinated and decaffeinated coffee (see  or DOI: 10.1161/CIRCULATIONAHA.115.017341).


  • 74,890 women in NHS, 93,054 women in NHS2, and 40,557 men in HPFS, with extensive semi-quantitative food frequency questionnaires (131 items, administered every 4 years), and 4,690,072 person-years of follow-up. 95% white.
  • Up to 30 years of follow-up, with extensive data on known or suspected confounders (biennial questionnaires asking about age, weight, smoking, physical activity, medication use, fam history of diabetes, and self-reported diagnoses including hypertension, hypercholesterolemia, cardiovasc disease, and cancer), and 31,956 deaths


  • Never coffee drinkers, included 12% of those in NHS, 30% in NHS2, and 17% in HPFS
  • Drinkers of >5 cups/d, included 24% of those in NHS, 35% in NHS2, and 25% in HPFS
  • Compared to non-drinkers, coffee consumption of 1-5 cups/d was associated with decreased total mortality (HR for death was 0.95 for <=1cup/d, 0.91 for 1.1-3 cups/d, 0.93 for 3.1-5 cups/d, all of which were statistically significant, and a nonsignificant HR of  1.02 for >5 cups/d). No diff between caffeinated and decaffeinated coffee consumption
  • When analysis was restricted to never smokers (there were 10,505 deaths in 2,451,970 person-years in this group), compared to nondrinkers:
    • <=1cup/d, HR for mortality 0.94 (0.89-0.99)
    • 1-3 cups/d, HR for mortality 0.92 (0.87-0.97)
    • 1-5 cups/d, HR for mortality 0.85 (0.79-0.92)
    • >5 cups/d, HR for mortality 0.88 (0.78-0.99)
    • Overall p<0.001 for the trend
    • The association with total mortality was especially strong for those <70yo; no differences if stratify by diet quality (they calculated the Alternate Healthy Eating Index), BMI, physical activity, sex, and the cohort studied (NHS vs HPFS)
    • For caffeinated coffee, the trend had p<0.001. For decaffeinated coffee, the trend had p=0.022 (still significant)
    • Association with 1-cup increments of coffee/d found significant inverse associations for cardiovascular disease, neurological diseases, and suicide. No significant association with total cancer mortality

So, really great news…

  • Why does this make sense, especially since the association was not really with caffeine?
    • For cardiovascular disease: chlorogenic acid, lignans, quinides, trigonelline, and magnesium in coffee reduce insulin resistance and systemic inflammation
    • For neurological diseases (other studies have also found coffee consumption was associated with lower risk of Parkinson’s): for MPTP-induced Parkinson’s, there is loss of striatal dopamine transporter binding sites, and dopaminergic neurons, which is attenuated by caffeine
    • For suicides (also found in other studies): ???coffee has antidepressant effects, also it turns out that mice in stressful situations given caffeine have less stress-induced changes in their brains and behavior, but this doesn’t really answer the question about coffee or esp about decaffeinated coffee in humans
  • And, unfortunately, this study just reveals an association, so hard to say that this is causal. Though this was based on really long-term studies of people with extensive food diaries and controlling for many of the underlying possible confounders (and, perhaps the major confounder of smoking). But, then again, it is all self-reported/not verified, and is in a largely white population of professionals, limiting the generalizability of its conclusions.
  • But, at least coffee is not being demonized as being bad for your health….. In fact, it probably should be added to the food pyramid, and at the bottom (right next to the dark chocolate).

Primary Care Corner with Geoffrey Modest MD: Increasing Mortality of US White Population

18 Nov, 15 | by EBM

By Dr. Geoffrey Modest

​The NY Times had a striking article from 2 Princeton economists (including Angus Deaton, who just won the Nobel Prize in Economics) finding that white Americans aged 45-54 with no more than a high school education had an increase in death rates, largely attributed to​ suicides and substance abuse from alcohol and prescription opiates. In terms of racial comparisons, white Americans have a mortality rate of 415/100K, African-Americans are at 581/100K (though the gap is closing), and Hispanics have a much lower rate of 262/100K. For the NY Times article, see . The full article is also freely available (see doi: 10.1073/pnas.1518393112​).

Will summarize the article in more detail:

  • If the white mortality rate in those 45-54 had continued to decline at the rate it was from 1979-1998, 488,500 deaths would have been avoided in the period 1999-2013. This loss of life is comparable to lives lost during the AIDS epidemic though mid-2015.
  • Between 1970 and 2013, there was a 44% overall decrease in mortality for those 45-54 yo, attributed to behavioral change, prevention and treatment. Similar changes were seen in other wealthy countries. In the US, this was a decrease of 2%/year. But after 1998, in contrast to other rich countries’ mortality rates which continued to decline at 2%/yr, the mortality rates for US non-Hispanic whites rose by 0.5%/yr. See the first graph below: all groups (including US Hispanic) are trending down, with the solitary exception of US whites​ (in graph, USW=US whites, USH=US Hispanics).
  • The increases in mortality were largely in US whites with high school degree or less (37% increased mortality over 1999-2013, with an increase of 134/100K) and less so for those with some college but no BA (31%) and those with BA or more (32%).
  • In those 45-54 group, mortality was dramatically increased for drug and alcohol poisonings, but also for suicides and chronic liver diseases, with not much change for diabetes and decreased mortality from lung cancer (see 2nd graph below). In 2013, drug overdoses were highest in white non-Hispanic, and death from cirrhosis and chronic liver disease fell for blacks and rose for whites.
  • The increased mortality in white non-Hispanics increased for each 5 year grouping from age 30 to 64, but the increases were most dramatic in the 45-54 groupings (see last graph on below). Not shown was those 65-74, who continued to have decreased mortality at 2%/yr.
  • Morbidity increased in parallel to mortality: overall, comparing data from 1997-99 to 2011-13, there were significant changes in self-assessed health status, pain, psychological distress, difficulties with activities of daily living, and alcohol use. For example, one in three white non-Hispanics aged 45-54 reported chronic joint pain in 2011-13. The Kessler six questionnaire, which assessed serious mental illness, increased from 3.9% to 4.8% of respondants. There were significant decreases (2-3% range) in walking 1/4-mile, climbing 10 steps, standing or sitting for 2h, shopping, and socializing with friends; and the activities overall were limited by physical and mental health issues. Of note, though obesity played a role, increases in midlife morbidity increased both in those with BMI >30 and those <30.
  • “Concurrent declines in self-reported health, mental health, and ability to work, increased reports of pain, and deteriorating measures of liver function all point to increasing midlife distress.”​

So, pretty impressive/depressing. The CDC estimates that for each prescription pain killer death in 2008, there were 10 admissions for abuse, 32 ER visits for misuse/abuse, 130 people who were abusers or dependent, and 825 nonmedical opiate prescription users (so, mortality is really just the tip of the iceberg in terms of the full societal effects). One aspect in the current discourse not really addressed is the effects of economic insecurity and lack of adequate social supports in the US (it is notable that the economic downturn was worldwide over the past 7-8 years, but the increases in mortality were confined to the US). And the greatest change was in non-Hispanic whites with little education (unclear why US Hispanics did relatively better, but there may have been less of a dramatic change in their economic and social well-being during the downturn, as well as differences in social/community supports. International surveys indicate that overall Latin Americans have higher subjective well-being, despite lower incomes). Just raises yet again the 1879 Virchow quote: “Don’t crowd diseases (epidemics) point everywhere to deficiencies of society?”

Primary Care Corner with Geoffrey Modest MD: Sleep Deprivation and Colds

11 Sep, 15 | by EBM

By Dr. Geoffrey Modest

There have been several articles finding an association between short sleep duration and various infectious diseases. The current study was a better-documented clinical trial supporting this (see SLEEP 2015;38(9):1353–1359). Details:

  • 164 healthy adults (94 men and 70 women, mean age 30) volunteered for the study
  • These volunteers were monitored for 7 consecutive days for their sleep duration and continuity, both by subjective questionnaires and by wrist actigraphy (which correlates well with polysomnography, the gold standard)
  • Then they were given nasal drops containing rhinovirus 39 and subsequently monitored for 5 days in a quarantined hotel. A “clinical cold” was defined if they were both infected and met the illness criteria (objective measures of mucous production and nasal congestion)


  • 124 of 164 participants (75.6%) were infected with rhinovirus (antibody measurements before and 28d after viral exposure, and daily nas
    al rhinovirus cultures)
  • 48 (29.3% of them) developed a clinical cold
  • Objective sleep duration (by actigraphy) was associated with an increased likelihood of developing a clinical cold, with odds ratios of developing a cold, as compared to those sleeping >7hrs per night:
    • <5 hrs/night had OR=4.50 (1.08-18.69)
    • <5-6 hrs/night had OR=4.24 (1.08-16.71)
    • 6-7 hrs/night had a nonsignificant  OR=1.66 (0.40-6.95)
  • These results were independent of prechallenge antibody levels, demographics (e.g. SES, education), season of the year, BMI, various psychological variables (e.g. perceived stress), and health practices (e.g. smoking, physical activity, alcohol)
  • Sleep continuity was not associated with developing a cold
  • Sleep duration was not related to getting the infection, just in getting sick


I bring up this study for a few reasons:

  • This study fits in with the very large literature on stress and diseaseover the past many decades, showing that both acute and chronic stressors, both physical and mental, can create huge changes in both hormones (essentially all hormones are affected), and in immunologic function (esp T cell and natural killer NK cell function), perhaps largely through the effects of the cortisol increases from the stressors. Of note, at least in several studies which directly looked at this: the issue is not so much the intensity of the stressor, but how the stressor is perceived by the person; and the physiologic effect of the stressor (including cortisol levels) is moderated by the degree of social support that the person has (not surprising that there are mediators to the effect of stress on individuals, since people exposed to the same physical or mental stressor often have different physiological reactions to it).
  • It brings up the limitations of the “germ theory”. Perhaps the main conclusion (to me) of this study is that infectious diseases (at least the vast majority of them) reflect a complex interplay between the characteristics of the bug and the host response to that bug. It is not just the presence of an external stimulus, but its interaction with the human organism.  Not everyone exposed to an organism gets infected (in this study 75.6% actually got infected, though this was not related to decreased sleep) and not everyone who gets infected develops disease (in this study 29.3% got a cold, which was related). In addition, not assessed in this study, there is great variability in the disease severity in those who get the disease, perhaps related to the individual’s immunologic response that is partly determined by the social environment. One summary article noted that even relatively acute sleep deprivation (less than one week) is associated with increases in cortisol, decreases in TSH, increases in prolactin, increases in growth hormone, and  increases in ghrelin/decreases in leptin, which might cause the munchies found with sleep deprivation (see, as well as the whole issue of the journal Brain, Behavior, and Immunity, volume 18, 2004 devoted to the hormonal/immunological effects of sleep deprivation​). Other studies have found sleep deprivation related to increases in inflammatory markers (e.g. C-reactive protein), down regulation of T cell production of interleukin-2, overall decreased proliferative capacity of T cells in vitro, and decreases in NK cells. Of interest, there seems to be a bidirectional association between sleep deficiency and inflammation, with studies finding that cognitive behavioral therapy for older adults with insomnia leads to decreased levels of systemic inflammation (Sleep. 2014; 37: 1543-52).
  • And, as the graph below showed, overall we are getting less sleep over the past 50 years.
  • The presumed association between disease and social conditions, by the way, has been articulated for a long time. For example, Virchow in 1879 wrote: “Don’t crowd diseases (epidemics) point everywhere to deficiencies of society?”​

Primary Care Corner with Geoffrey Modest MD: Stress and peptic ulcers

1 Feb, 15 | by EBM

By: Dr. Geoffrey Modest

Older studies have shown an association between stress and peptic ulcer disease. At least the ones I’ve seen have not controlled for the use of NSAIDs or the presence of H pylori. However, from newer data, 16-31% of ulcers are not associated with either of these precipitating factors. The current prospective population-based study was done in Denmark, in which the researchers collected blood samples as well as an inventory of psychological, social, behavioral and medical data in 1982-3, and reinterviewed these patients in 1987-8 and 1993-4, finding that psychological stress did indeed increase the risk of ulcers (see​).

–3379 adults without prior history of ulcer disease were enrolled, with subsequent data on 2809 of them in 1987-8 and 2410 in 1993-4. Pretty evenly distributed in the 30-60 year age range.
–socioeconomic status (SES) was calculated from education, occupation, employment status. Stressors included working more than 40 hours/week and an assessment if the person had economic, work, family, housing , or personal problems. Subjects also answered 22 items from the Mental Vulnerability Scale (a questionnaire used by the military to screen potential recruits), a validated scale which assesses “somatization, neuroticism, depression, and anxiety.”
–a stress index was calculated which combined the baseline Mental Vulnerability; tranquilizer use; economic, work, family, housing or personal problems; unemployment; and working >40 hours/week — with a score of 0-10.​


–43% were H pylori positive, 16% were taking NSAIDs, 56% were current smokers, 39% in the lower two SES categories
–76 people were diagnosed with an ulcer over the course of the study (documented on endoscopy/radiology exam): 39 duodenal ulcers and 30 gastric ulcers
–ulcers were significantly more common in those in the highest tertile of stress scores (3.5%) vs the lowest (1.6%), with adjusted odds ratio of 2.2 (CI 1.2-3.9, p<.01). This did not change after adjusting for IgG antibodies to H pylori, alcohol consumption, or sleep duration, with a per-point odds ratio for the stress index being 1.19 (CI 1.09-1.31, p<0.001), and a clear dose-response (the higher the stress score, the more likely to have an ulcer)
–the adjusted stress relationship was lower after controlling additionally for SES, with per-point odds ratio for the stress index being 1.17 (CI 1.07-1.29, p<0.001), and further by controlling for smoking, use of NSAIDs and lack of exercise, with per-point odds ratio for the stress index being 1.11 (CI 1.01-1.23, p=0.04).
–there was a similar risk of ulcer related to stress in those who were H pylori positive or negative, or in those both H pylori negative and not on NSAIDs.
–in multivariate analysis, stress, SES, smoking, H pylori infection, and use of NSAIDs were independent predictors of ulcers.

One big plus for this study was its timing: it predated H pylori testing (and treatment), either H2-blockers and subsequently proton-pump inhibitors were available only as prescriptions and not over-the-counter, and the study was done largely before widespread use of low-dose aspirin; all of these factors significantly decrease the confounding that would exist if the study were done now. Purported mechanisms by which stress could cause ulcers include by increases in stomach acid secretion, effects on the hypothalamic-pituitary-adrenal axis (essentially all hormones are affected by stress, many of them mediated by the stress-related cortisol elevations) which can affect healing, hormone-mediated changes in blood flow to the stomach, or cytokine-mediated impairment of mucosal defenses. Some of the stress effect is likely mediated by smoking (which in one study accounted for one-third of the ulcerogenic effect of stress), alcohol or poor sleep (though these latter 2 were not confirmed in the Danish study above). Of note, there was no synergy in the Danish study between stress and H pylori in ulcer development. To me, this is a pretty impressive study despite its being observational, given the quality of the data they collected prospectively, the fact that there is a dose-response curve with each increase in their stress scale associated with increased likelihood of ulcer disease, and their controlling for many known or likely associations for ulcers.

So, not so shocking a finding (I think most of my patients are aware of the link between stress and ulcers….). However, the National Institute of Diabetes and Kidney Diseases of the NIH in 2012 notes “peptic ulcers are not caused by stress”. So, I guess my patients were right all along…

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