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.

Cancer- GI

Primary Care Corner by Geoffrey Modest MD: Risks and benefits of longterm PPIs

26 Apr, 17 | by gmodest

​by Dr Geoffrey Modest

The American Gastroenterological Association (AGA) just published a clinical practice update on the risks and benefits of long-term use of proton pump inhibitors (see  ).



RISKS: (these are the authors’ assessment of the quality of the evidence and the effect sizes)

kidney disease: 2 retrospective observational studies found a modest effect size (10-20%) of CKD in those on PPIs, with very low quality of evidence. Mechanism, unclear: ? if those on PPIs had more comorbidities which predispose them to kidney disease?

dementia: retrospective observational studies finding a modest effect size (4-80%), with very low quality of evidence. Presumed mechanism: microglial cells use certain ATPases to degrade beta-amyloid, and PPIs may block these ATPases (which does increase beta-amyloid in mice)

bone fracture: many observational studies, data inconsistent, modest effect size (39% to 4-fold increase), with low to very low quality of evidence. Presumed mechanism: hypochlorhydria-related malabsorption of calcium or vitamin B12, gastrin-induced parathyroid hyperplasia, and/or osteoclast vacuolar proton pump inhibition.

myocardial infarction: though a very small effect was found in an observational study, none found in RCTs. Presumed mechanism: omeprazole decreasing clopidogrel levels and its anti-platelet effect, but a randomized controlled trial comparing those on clopidogrel versus those on clopidogrel plus omeprazole had no difference in cardiovascular event rates.

small intestinal bacterial overgrowth: small studies have found that PPIs lead to bacterial overgrowth in the duodenum/small intestine, only some of which were symptomatic, modest effect size (2-fold to 8-fold increase), low quality of evidence. Presumed mechanism is loss of the bactericidal effects of gastric acid by taking PPIs

non-typhoidal salmonella and Campylobacter infections: increase found in 1 study, not confirmed. modest effect size (2-fold to 6-fold increase). Presumed mechanism: achlorhydria (and studies show that those with pernicious anemia or gastric surgery-induced achlorhydria do seem to have increases in these infections)

spontaneous bacterial peritonitis: observational studies suggest a 2-fold increased risk of SBP (50% to 3-fold increase), very low quality of evidence. Proposed mechanism: achlorhydria leading to gut bacteria changes, leading to changes in intestinal permeability and translocation of bacteria across the intestinal wall

C. diff infections: observational studies suggest 50% increased risk of C diff infection; and changes in bacterial taxa associated with C diff were increased in healthy volunteers after 4-8 weeks of high-dose PPIs. (the risk still pales compared to the rate of C diff with antibiotics). Risk may be higher in children, modest effect size (no increase to 3-fold increase), quality of evidence: low. Proposed mechanism: downstream effects of PPIs on colonic microbiota (see comment below)

pneumonia: seems to be more frequent soon after starting PPIs than after longer-term treatment.   Raises question of perhaps the PPIs were erroneously started for early misdiagnosed pneumonia. pneumonia is not a consistent finding in other studies, modest effect size (though no association in RCTs), very low quality of evidence. Proposed mechanism: upstream effects of PPIs on oropharyngeal microbiome

micronutrient deficiencies (overall 60-70% increase), low or very low quality of evidence:

–Calcium: may be decreased absorption, but not of water-soluble calcium salts or calcium from milk or cheese.

–Iron:  inconsistent data. No association in some Zollinger-Ellison patients on 6 years of PPIs, some association in other studies

–Magnesium: rare cases of profound hypomagnesemia. Observational data on modest positive association

–vitamin B12: most studies finding around 2.4-fold increased risk.

gastrointestinal malignancies: data also mixed. Suggestive data of increased risk in those with untreated H pylori infections, and concern about the profound hypergastrinemia (which has trophic effects on colonic epithelial cells in mice and on human colorectal cancers in vitro),  but population-based retrospective studies have failed to confirm a relationship. (No association in RCTs), modest effect size, very low quality of evidence.



In terms of benefits of PPIs, there are basically moderate to high quality studies supporting their use in:

— GERD with esophagitis or structure (though may not be necessary with non-severe esophagitis, and no long-term data)

— GERD without esophagitis or stricture (though may not be necessary with relatively mild symptoms, and no long-term data)

— Barrett’s esophagus with GERD (no long-term data)

— NSAID bleeding prophylaxis (no long-term data)

— Barrett’s esophagus without GERD (this has low quality of evidence from observational studies only: no RCT, mostly mechanistic thinking that chronic inflammation may lead to esophageal adenocarcinoma and some observational data. But I would also be concerned that these data are based an unusual subset of patients who are asymptomatic yet have had endoscopy that documents Barrett’s, and even observational studies are therefore a tad suspect).



–It is not surprising that the quality of these studies on benefit is higher than the above studies of adverse effects, since these were designed explicitly as intervention trials to look for benefit, probably all supported by drug companies, and controlling for co-morbidities, etc.

–I am also a little concerned that the AGA may be biased towards PPIs, perhaps because gastroenterologists tend to see patients with more severe conditions requiring PPIs, or perhaps financial conflicts-of-interest (as with all specialty societies, since the top academic specialists who often write the guidelines tend to be involved in drug-company-sponsored research).  My real concern with PPIs is that many many outpatients are put on PPIs for marginal reasons, and that very few patients are stepped-down to less aggressive therapy. As mentioned in prior blogs, given the limitations of time a primary care clinician has with patients, when their stomach problem is better with PPIs, it is time to deal with the myriad of other problems, keeping up with standard health maintenance issues, etc etc. The issue of the above potential complications of PPIs are very probably less important clinically than the need for PPIs for those with very clear indications (though I am a bit concerned that these studies are all short-term and it is a bit tenuous to extrapolate to long-term harms). But, the preponderance of studies finding some association of potentially serious adverse effects from PPIs, whether the studies are great or not, reinforces the imperative to avoid using PPIs unless clearly indicated, and, when appropriate, to step-down therapy as soon as possible. My experience is that patients who have endoscopy for dyspepsia are essentially invariably put on PPIs by the gastroenterologists independent of endoscopic findings. And, I have had pretty good success in getting some patients off of them, sometimes just onto prn calcium tablets or H2 blockers. But this may be a time-consuming issue to deal with. And I certainly have many patients for whom either I do not have the time to pursue or who are resistant to stepping down on therapy.

–To me, there is also the perhaps significant general omission in the above article of the effects of PPIs on the microbiome (see here). My guess is that these effects do not necessarily translate clinically into disease, which is not so surprising given the complexity of this process, the multiple variables involved, and the length of time necessary to develop detectable disease (and the studies are too short). But, PPIs are associated with changes in the colonic microbiome to a less healthy one: with significant increases in Enterococcus, Streptococcus, Staphylococcus, and potentially pathogenic E coli species, as well as oral bacteria of the genus Rothia. And decreased Clostridiales.  These changes have been thought to lead to the association with C diff infections, but perhaps with other even unsuspected long-term harms. Though not mentioned specifically in the above article, these microbiome changes do add further credence to the imperative (I think) to minimize PPI usage.

So, my bottom line: PPIs are way overused for marginal indications (it is easy to jump to PPIs for dyspepsia, since they work so well…), but we should really discourage the use of PPIs unless they meet a clear criterion as above, or try to use the step-up approach: start with calcium or H2 blockers, then increase to PPIs when needed, and still try to step-down later; and try to get patients off of PPIs when they have been on them for awhile, unless there is a clear indication to continue.  Though a complicating factor here is that they are available OTC….

for another recent blog on PPI risks and benefits and some additional concerns, see here.


Primary Care Corner with Geoffrey Modest MD: Antibiotics, microbiome changes and colorectal adenoma

21 Apr, 17 | by gmodest

by Dr Geoffrey Modest

There been a few studies over the past suggesting a relationship between the gut microbiome and colorectal cancer, as well as between antibiotic exposure and colorectal cancer. An evaluation of the Nurses’ Health Study recently confirmed prospectively that there was a dose-response curve between women’s prior use of antibiotics and colorectal adenomas (see gutjnl-2016-313413).


— 16,642 women aged at least 60 who had at least one colonoscopy between 2004 and 2010 and had reported their antibiotic use in a 2004 questionnaire, comparing antibiotic users versus nonusers

— mean age 70, family history of cancer in 20%, diabetes in 9%, BMI 25, hormone therapy 20%, regular use of aspirin in 40%, multivitamins in 78%, 20 pack-years of smoking in those who were ever-smokers, 2.3 g of alcohol per day, 6 servings of red meat per week.


— 1195 cases of adenomas were detected

— women who used antibiotics for more than 2 months between the ages of 20 and 39 had a 36% increased risk of adenomas by multivariate analysis, OR =1.36 (1.03-1.79)

— women who used antibiotics for more than 2 months between the ages of 40 and 59 had a 69% increased risk by multivariate analysis, OR = 1.69 (1.24 – 2.31)

— there was a trend between increasing antibiotic use at age 20-39 (p=0.002) and also at 40-59 (p=0.001), in each case with progressively more adenomas when increasing antibiotic use, from no use to 1-14 days, to 15 days-2 months, to >2 months.

— this association was similar for low risk versus high risk adenomas (high-risk being defined as size > 1 cm, with tubulovillous/villous histology, or > 2 detected lesions), though was slightly stronger for proximal lesions.

— there was no association between antibiotic use in the prior 4 years and risk of adenoma [ie, the microbiota were not influenced by recent antibiotic usage]

— women who used antibiotics for a longer duration were overall similar to those who did not in terms of family history, personal disease/screening history, and lifestyle factors, but were more likely to regularly use menopausal hormonal therapy, aspirin, and undergo colonoscopy for symptoms rather than routine screening.


–the Nurses’ Health Study is an ongoing prospective cohort study of 121,700 US female nurses aged 30 to 55 at enrollment in 1976. The advantage of looking at this cohort is the high quality of data collected (which had accurate data both on an array of lifestyle issues as well as medical problems/medications etc, as well as specifically on prior intermittent antibiotic use many years beforehand), and the long-term follow-up

— the presumed mechanism for a relationship between antibiotics and colorectal adenomas is through the effect of antibiotics on the microbiota. For unclear reasons antibiotics may induce either temporary, quasi-stable states, or alternative stable states. The specific microbiota changes associated with colon cancer include depletion of Bacteroides, Firmicutes (Clostridia), and Proteobacteria (Enterobacteriaceae) and enrichment of Fusobacteria.

— of course, though this was a really good prospective study following lots of items (a rather long questionnaire….), there could well be unaccounted-for differences between the antibiotic users and nonusers which could explain the microbiome differences as well as the increase in adenomas. The noted differences between these groups (eg, using postmenopausal hormones, aspirin, having nonscreening colonoscopies) were accounted for, but were there other issues? were there differences in psychosocial issues between the groups? were those on these meds and getting antibiotics more anxious or stressed out (and there is some evidence that increased cortisol levels, often found with stress, can effect changes in the microbiome)? Were these women on the above meds also taking other unassessed meds that could affect the microbiome and adenoma rate (and perhaps leading to the long-term changes in the microbiome)? As with all observational studies, one cannot attribute causality to an association.

–so, I bring this up mostly because this study has a great database, and long-term follow-up, and reinforces many of the articles brought up before regarding the effects of microbiota changes and human disease. And, it provides us with an even stronger imperative to try to decrease antibiotic use, except when clearly indicated. 

See here for an array of articles on the microbiome, including mechanism by which microbiota changes might lead to a variety of diseases including NAFLD, cancer, diabetes, metabolic syndrome, heart disease….  ​

See here for another array of articles, but dealing with the consequences of overuse of antibiotics in humans and livestock and microbial resistance

Primary Care Corner with Geoffrey Modest MD: One-time flex sig?

5 Apr, 17 | by gmodest

by Dr Geoffrey Modest
An observational study in the UK initially reported that a single flexible sigmoidoscopy screening significantly reduced subsequent colorectal cancer incidence as well as colorectal cancer mortality after 11 years of follow-up. They now found similar results after 17 years of follow-up (see S0140-6736(17)30396-3).
— multicenter randomized trial done from 1994-1999, randomized 170,432 men and women aged 55 to 64 to sigmoidoscopy (n= 57,098, of whom 40,621 were screened) vs control (n=112,936).
— mean age 60, 51% women.
— At screening in the intervention group: 18 were referred to surgery directly, 2131 (5%) were referred for a colonoscopy for high risk polyps, and 38,825 (95%) were discharged with low risk polyps or no polyps at all
— primary outcomes were: incidence and mortality of colorectal cancer
—  after 17.1 years of follow-up
   — colorectal cancer was diagnosed in 1230 individuals in the intervention group and 3253 in the control group: distal colorectal cancer was diagnosed in 529 in the interventions group (126 at the time of screening) and 1987 in the control group. Proximal colon cancers were diagnosed in 612 in the intervention group and 1255 in the control group.
   — 353 individuals in the intervention group and 996 in the control group died from colorectal cancer
— intention to treat analyses:
    — colorectal cancer incidence was reduced by 26%, HR 0.74 (0.70-0.80), p<0.0001
    — colorectal cancer mortality was reduced by 30%, HR 0.70 (0.62-0.79), p<0.0001
        — mortality related to distal colorectal cancer deaths was reduced by 46%, HR 0.54 (0.45-0.65), p<0.0001
  — per protocol analyses (ie, those who actually had a sigmoidoscopy in the intervention group):
    — colorectal cancer incidence was reduced by 35%, HR 0.65 (0.59-0.71)
    — colorectal cancer mortality was reduced by 41%, HR 0.59 (0.49 0.70)
        –mortality related to distal colorectal cancer deaths was reduced by 66%, HR 0.34 (0.26-0.46)
— sigmoidoscopy screening had no effect on all-cause mortality or on mortality from proximal colonic lesions (above the reach of the sigmoidoscope)
— the estimated number needed to screen (NNS) to prevent a single colorectal cancer diagnosis over 17 years was 98; it was 191 after the 11 year assessment
— the estimated NNS to prevent a certified death from colorectal cancer over 17 years was 220; it was 489 after 11 years
— no difference in the efficacy of sigmoidoscopic screening between men and women in terms of distal cancer, though the overall effect was less in women [women have been found to have more proximal colon cancers in several studies, including this one, one in the US and one in Norway]. The NNS for women was 165, much higher than men likely because of lower incidence of colorectal cancer in the control group in women than men and the fact that women had more proximal cancers. Also, there was no outcome difference between those 55-59yo versus 60-64yo at the time of the screening
— the 11 year data from the study showed that colorectal cancer incidence was reduced by 33%, distal colorectal cancer incidence by 50%, and colorectal cancer mortality by 43% (similar to above)
–A review of the graphs in the 17-year follow-up article shows that at about 4-5 years post-screening, the curves for colon cancer incidence cross (the intervention group picked up more cancers prior to the 5 year mark, the control group having much more afterwards, with curves then diverging). The incidence of proximal cancer was the same throughout the study. Colorectal cancer mortality diverged after about 5 years, with essentially parallel curves for the last 4 years. This latter finding suggests that follow-up after the 17 year mark would likely continue to show benefit  from a single screen (the annual incidence rate ratio reductions were 74% between years 6 and 10 and 69% between years 11 and 16 post initial screen).
— for reasons that escape me, several studies have found that excision of colonoscopy-discovered right colonic lesions (i.e. beyond the reach of the sigmoidoscopy) do not affect mortality (eg see Baxter NN. Ann Intern Med  2009; 150:1). This finding, in and of itself, adds further support to sigmoidoscopy screening, which is significantly less invasive and has significantly fewer adverse consequences than colonoscopy.
— a pooled analysis of 3 other trials also found extended benefit for flexible sigmoidoscopy screening, though these trials did not find benefit in women >60 years old. The current study authors argue that the screening should be done in people younger than 60 years old anyway, where women seem to have equal benefit.
so, pretty powerful article. The current recommendations have not changed (though are more aggressive in the US than Canada or the UK: (eg see here ). But this article suggests a few things:
–though we may offer patients the recommended screening intervals (as per the recommendations), it is reassuring that the likelihood of cancer or mortality is significantly less even after a single negative sigmoidoscopy (and I have seen a few articles in the past suggesting that colonoscopy screening may also be effective for more than 10 years. eg Nishihara R. N Engl J Med 2013;369:1095, which found protection up to 15 years after a negative colonoscopy)
–it also raises the issue of the natural history of colon cancer, with most of us learning it was on average about 10 years to develop cancer in apparently normal cells (supporting the screening intervals of the recommendations). It seems that people with visually normal exams at screening are at much lower risk of development of cancer for many more years than those with adenomas or early cancers (ie, those with lesions of some sort have a different colonic substrate and are therefore more genetically or environmentally or both predisposed to cancer even after resection than those with apparently normal colons, so it is erroneous to apply the same  finding of metamorphosis to cancer from them to normal individuals.)
–and it even raises tangentially a bigger issue about cancer: smokers with disturbed lungs, such as with COPD, seem to be at higher risk of lung cancer than those with radiographically normal lungs even with the same smoking history, or women with nonmalignant breast changes on biopsy are more predisposed to breast cancer,
–and perhaps even calculations of deleterious effects of radiation exposure is much worse in those with underlying abnormal tissues (which raises yet another issue: doing more mammograms in women with abnormal noncancerous breast tissue, or in smokers for lung cancer screening, might be much more risky than in women or smokers with normal tissue. So, the recommendations to increase radiologic screening in both cases may actually cause more harm in those with abnormal baseline tissues than the current models suggest. though i digress……
See here  , which reviews a population-based study in Norway, similarly finding that a one time flexible sigmoidoscopy decreased deaths from colorectal cancer by 27%, with 10.9 years of follow-up.

Primary Care Corner with Geoffrey Modest MD: USPSTF Colorectal Cancer Screening Guidelines

27 Jun, 16 | by EBM

By Dr. Geoffrey Modest

The USPSTF just published their 2016 guidelines for colorectal cancer screening (see doi:10.1001/jama.2016.5989).


  • Colorectal cancer is the second leading cause of cancer death in the US
  • In 2016, estimates are that 134K people will be diagnosed with the disease, and 49K will die from it.
  • Most frequent age range for diagnosis is 65-74, median age of death is 68
  • About 1/3 of adults never get screened

Recommendations, for asymptomatic adults at average colorectal cancer risk:

  • Screen from age 50-77 (grade A recommendation)
  • Screen from age 76-85 (grade C recommendation), as an individual decision, taking into account the patient’s overall health and prior screening history. specifically focusing on:
    • Adults in this age group who have never been screened (higher likelihood of abnormality)
    • Those healthy enough to undergo treatment if cancer is detected and do not have comorbid conditions that would significantly limit their life expectancy
    • They conclude “with moderate certainty that the net benefit of screening in adults aged 76-85 who have been previously screened in small”
  • The benefit of screening >85yo is likely to be very small, since the time between detection/treatment and the actual mortality benefit can be substantial
  • 7 screening options:
    • Guaiac fecal occult blood test (gFOBT) every year
    • Fecal immunochemical test (FIT) every year
    • Multitargeted stool DNA test (FIT-DNA) every 1 or 3 years (this test combines a FIT with testing for altered DNA biomarkers, has higher sensitivity than FIT but more false positives and leads to more colonoscopies, with their attendant risks). The manufacturer recommends every 3 years
    • Colonoscopy every 10 years
    • CT colonography every 5 years
    • Flex sigmoidoscopy every 5 years
    • Flex sig every 10 years, plus FIT every year
  • They do not prioritize the order, noting that all of these tests have benefit, and that the main goal is screening (with the hope that multiple different options will increase the screening rate)
  • The harms of screening increase with age, going from small in those 50-74, to small-to-moderate in those >75, especially for colonoscopy
  • They provide a bar graph of benefits and harms, noting
    • Life-years gained per 1000 screened: in the 260-270 range for colonoscopy and FIT-DNAevery year; 240-250 for FIT, gFOBT, CT colonoscopy, and flex sig/ FIT; in the 220 range for flex sig and for FIT-DNA every 3 years
    • For colorectal cancer deaths averted per 1000 people screened: 23-24 for colonoscopy, flex sig/FIT and FIT-DNA every year; 22 for FIT, gFOBT, CT colonoscopy; 20 for flex sig or FIT-DNA every 3 years.
    • For harms per 1000 screened: lowest (9-11) for flex sig/FIT, FIT-DNA every 3 years, FIT, gFOBT, CT colonography, and flex sig/FIT; 12 for FIT-DNA every year and highest (15) for colonoscopy
    • For lifetime burden of colonoscopies per 1000 screened: under 1820 for FIT-DNA every 3 years , FIT, CT colonography and flex sig; 2200-2300 for gFOBT and flex sig/FIT; 2662 for FIT-DNA every year; 4069 for colonoscopy.


  • I agree with them that the most important thing is doing the screening, since the benefits really are not that different by which test is done, and that offering noninvasive testing is clearly attractive to many patients (and may increase the screening rate). I would certainly add that any person getting a non-colonoscopic screen should understand and agree to pursuing colonoscopy if positive
  • Note that the above numbers are from different studies in different populations and are not based on head-to-head comparisons of different screening strategies, so the actual numbers above (risks/benefits) are not really strictly comparable. But they do probably provide a reasonable ballpark estimate
  • For the recently released Canadian guidelines, see , which is pretty dismissive of colonoscopy screening, are not in favor of screening those >74 yo, and actually mostly support using gFOBT or FIT every 2 years or flex sig every 10 years. This blog includes my additional concerns about colonoscopy, where the risk does increase significantly with age, and argues that the lesions that benefit from screening are basically the ones the sigmoidoscope reaches.
  • I am concerned about using CT colonoscopy (radiation exposure, competence/accuracy in different settings esp if not done regularly, pick up of incidental non-colon findings which lead to unnecessary testing) and the FIT-DNA testing. For the latter, what does one do if the FIT-DNA is positive but the colonoscopy is negative? Is it a false positive, or is there a small cancer lurking in the crypts? Should there be a follow up colonoscopy in 1 year. What if that is negative — keep repeating it every 1-2 years??? Seems to me to be a black hole, and we should avoid that event horizon.
  • They do not suggest a different strategy for Black or Alaska Native individuals (who have higher incidence and mortality than the general population) because there are no empirical data on the effectiveness of different screening strategies for these populations. And there are some studies showing that equal treatment seems to produce equal outcomes, suggesting that the issue may be more of access than actual risk.
  • So, my own approach is evolving, though I am much more strongly leaning towards FIT testing alone or in combination with flex sig, since these options seem to maximize the benefit and cut in half the number of colonoscopies with their associated intensive prep, conscious sedation, and increased risk of perforation (all worse in older people), and with small apparent benefit.

Primary Care Corner with Geoffrey Modest MD: Coffee and Decreased Colon Cancer

25 Apr, 16 | by EBM

By Dr. Geoffrey Modest

An interesting case-control study found that coffee consumption is associated with a decreased risk of colorectal cancer CRC (see Cancer Epidemiol Biomarkers Prev. 2016; 25(4): 634).


  • 5145 colorectal cancer (CRC) cases were compared with 4097 controls from the Molecular Epidemiology of Colorectal Cancer (MECC) study, a population-based study in northern Israel, beginning in 1998.
  • Mean age 70, 52% male, 60% Ashkenazi/20% Sephardi/13% Arab, 7% first-degree relative with CRC, 58% consuming >= 5 vege servings/d, 35% “sports activity”, 10% current/29% former smokers, 25% daily aspirin, 74% colon/23% rectal cancer). Mean coffee intake in controls was 2.0 servings/d. There were significant differences in both quantity and types of coffee drunk by the different ethnic groups
  • Results, comparing coffee drinkers to nondrinkers:
    • Coffee consumption was associated with 26% lower odds of developing CRC [OR 0.74; (0.64–0.86), p<0.001], controlling for known risk factors (age, sex, ethnicity, vegetable consumption, sports participation, statin use, daily low-dose aspirin, smoking status and family history). Additional controlling for total daily liquid consumption or total calorie consumption did not significantly affect these results.
    • Decaffeinated coffee had 18% lower odds [OR 0.82 (0.68–0.99), p=0.04]
    • Boiled coffeehad 18% lower odds [OR, 0.82 (0.71–0.94), p=0.004].
  • There was a dose-response curve, with p<0.001 for the trend. Compared to <1 serving/day:
    • Intake of 1 to <2 servings/dayhad 22% lower odds for developing CRC [OR 0.78 (0.68–0.90), p< 0.001]
    • 2 to 2.5 servings/dayhad 41% lower odds [OR 0.59 (0.51–0.68), p<0.001]
    • ​>2.5 servings/dayhad 54% lower odds [OR, 0.46 (0.39–0.54) p< 0.001]
  • ​Also, there was an overall inverse relationship between CRC and vegetable consumption (p<0.001), daily low-dose aspirin (p=0.03)​, sports participation (p<0.001), and direct relationship with smoking status (p<0.001), and sex (p<0.001).

So, a bright spot for me given my coffee consumption. But does this all make sense?

  • Coffee has many bioactive components, including chlorogenic acids (powerful anti-oxidants, which modify gene expression and inhibit DNA methyltransferase), polyphenols (which have anti-oxidant and antiproliferative effects, and induce cell-cycle arrest in colorectal cell lines), melanoidins (which may increase colon motility), diterpenes (which may be anticarcinogenic by enhancing defense systems against oxidative stress), and caffeine (may also be anti-oxidant, and limit growth of human colon cancer cells). The actual effects of these constituents is not so clear in humans, since much of the above is through in vitro analyses.
  • Coffee also reduces bile acid secretion
  • Coffee also modifies the microbiome: there are some data from small human experiments that coffee does change the fecal microbiome favorably, increasing Bifidobacterium spp. Another study, presented only as an abstract, looked at the microbiome of coffee consumers in an effort to see how coffee might lower the risk of diabetes, finding in rats that those on high fat diet had less weight gain and less insulin resistance, and more lactobacillus in their microbiome (FASEB journal; 2013; 27:951)​
  • And, coffee does change some bowel functions, such as increasing intestinal motility and stool output
  • Though, it should be added that there are some differences in the above components, depending on the actual coffee beans used, the degree of roasting, and the brewing technique
  • And, although the epidemiologic data are not entirely consistent, most suggest the above association (e.g., see Public Health Nutrition 2013; 16: 346, a meta-analysis finding an overall 15% reduction in colon but not in rectal cancer, especially in Europe and in women), and several show a decreased incidence of liver cancer. Of note in the Israeli study, there was a lower incidence of both colon and rectal cancer, though less impressive for rectal cancers.
  • This study did look at aspirin consumption, which is associated with lower colon cancer risk (see: ), but did not assess NSAID use, also associated in many studies with lower risk of colon cancer
  • Though a case-control study can never be definitive, the constellation of biological plausibility, the consistency with other observational studies, the magnitude of the effect, and the dose-response curve (the more coffee, the better) tends to support the conclusion. There could, however, be reverse causation: did those with GI symptoms from an early cancer, for example, stop drinking coffee?  It would really be hard to have a real randomized controlled study, where thousands of people were randomized to drinking varying amounts of coffee and followed for 5-10 years. So, these case-control studies are about as good as we can get….

For relevant blogs, see: reporting on the Nurses’ Health Studies, finding a decrease in total mortality in coffee-drinkers , a Korean observational study, finding a pretty strong association between coffee-drinkers and decreased coronary artery calcifications

There are also several blogs in the BMJ website on chocolate (another personal addiction), also rich in polyphenols. One showed clinical improvement in those with peripheral artery disease (see )

Primary Care Corner with Geoffrey Modest MD: Colorectal Cancer in Younger People

3 Feb, 16 | by EBM

By Dr. Geoffrey Modest

Colorectal cancer is increasingly being diagnosed in adults <50 yo (i.e., prior to the time of recommended screening by USPSTF), with projections that by the year 2030, 1 in 10 colon cancers and 1 in 4 rectal cancers will be diagnosed in this younger group. A recent population-based cohort study, looked at the disease stage at presentation, treatment patterns by stage, and cancer-specific survival (see DOI: 10.1002/cncr.29716)​. The study was supported by the Agency for Healthcare Research and Quality.


  • The researchers accessed the large SEER database (Surveillance, Epidemiology, and End Results), a source for cancer incidence, survival and prevalence, which captures 28% of the US population and is geographically, racially and ethnically diverse. They looked at all patients 20-79yo diagnosed with histologically confirmed colon or rectal cancer between 1998-2012


  • 258,024 individuals with documented colorectal cancer were in the database, of which 37,847 (14.7%) were too young for routine screening (mean age 42.5 +/- 6 years, 53.5% male)
  • In comparison to the group >50 yo (where mean age was 65.3 +/- 8.5, 54% male)
    • 8 vs 12% were African American (p<0.001), and also more likely to be American Indian/Alaska Native or Asian/Pacific Islander (10.6% vs 8.5%, p<0.001)
    • Right-sided cancer in 20% vs 31.1% (p<0.001)
    • Rectal cancer in 31.2 vs 22.4% (p<0.001)
    • Stage of disease (all with p<0.001):
      • Localized in 35% vs 42.6%
      • Regional in 39.3% vs 36.4%
      • ​Distant in 25.7% vs 21.1%
    • Younger patients were 37%  more likely to have regional vs localized disease (RR 1.37, p<0.001)
    • Younger patients were 58%  more likely to have distant vs localized disease (RR 1.58, p<0.001)
    • 5-year cancer-specific survival, however, was better for younger patients (localized: 95.1 vs 91.9%; regional: 76 vs 70.3%; distant 21.3 vs 14.1%)


  • Pretty striking that 1 in 7 patients have colorectal cancer at <50 yo, and there seems to be an increasing shift to younger people
  • The younger ones tend to present with more advanced disease (perhaps a tribute to our screening the older ones).
  • Their survival curves are somewhat better than the older folks’, perhaps related to a few issues: they tended to receive more aggressive cancer therapy (not shown above), and perhaps in part related to that, tend to have younger bodies which handle the assault of the disease and the therapy better than older patients.
  • Although it may be reasonable to consider screening at a younger age (in the assessment above, the mean age was 42.5, so this would mean huge numbers of more screenings), we really would need to study the real risks and benefits (and, it may be that the colorectal cancers that afflict younger patients are actually different: behave differently, perhaps are less aggressive,  respond differently to therapy, etc. — so the risk/benefit analysis may be very different from those with average age 65.3 as above)
  • But, bottom line, we should really consider colorectal cancer in the differential for younger patients with rectal bleeding, iron-deficiency anemia, or change in bowel habits.

Primary Care Corner with Geoffrey Modest MD: One-time flexible sigmoidoscopy decrease mortality

21 Nov, 14 | by EBM

JAMA had a population-based clinical trial in Norway of flexible sigmoidoscopy (sig) screening (see doi:10.1001/jama.2014.8266). 100K men and women aged 50-64 were randomized to once-only flex sig (10.2K people), combo of flex sig and fecal occult blood testing (FOBT, also 10.2K people) or control (78.2K), with screening in those 55-64 beginning in 1999-2000 and those 50-54 yo in 2001 and followed until 2012.  Those with positive screen (cancer, adenoma, polyp>1cm, pos. FOBT) were offered colonoscopy.


–Adherence to screening was 63%

–After 10.9 yrs, 71 died from colorectal cancer in screened groups (31.4 deaths/100K person-yrs) and 330 in the control group (43.1 deaths/100K person-yrs), abs diff of 11.7 and HR of 0.73 (0.56-0.94). i.e., there was a difference of about 1 death/1000 individuals by sig screening

–Incidence of colon cancer in 253 in the screened group (112.6 cases/100K person-yrs) and 1086 in the controls (141.0 cases/100K person-yrs), abs diff of 28.4 and HR of 0.80 (0.70-0.92)

–Flex sig was beneficial in both age groups: cancer incidence in 50-54yo had HR of 0.68 (0.49-0.94) and in those 55-64 HR was 0.83 (0.71-0.96).

–Adding the FOBT did not change the results

–No complications of flex sig (done in 20.5K men and women), though of the 2816 colonoscopies done, (19.5% of those screened), there were 6 perforations and 4 were admitted for postpolypectomy bleeding

–Of note, there was no difference between screening and controls until 9 years later (for both mortality and cancer detection), which might suggest that the results would be more impressive with longer follow-up.

So, these hazard ratios are pretty comparable to the existing studies (although this was a truly population-based study, unlike the other trials which consisted of volunteers). Also first study to show benefit in those 50-54 years old. The big issue for us is that we tend to do colonoscopies as our first line, even though the complication rate is much higher (as in this study), and the only good data for cancer-specific mortality rate reductions are based on sigmoidoscopy studies. The issue, of course, is that sigs miss proximal tumors, and around 20% of patients with serious proximal lesions have no distal lesions (and therefore be missed by just doing sigs). The rationale for the FOBT is that proximal lesions bleed more than distal ones and FOBT might help uncover them, though the results above showed no benefit from adding FOBT. Also the clinical benefit of finding/resecting proximal tumors is significantly less than with distal tumors, in several studies. This study did find a 10% reduction in proximal colorectal cancer vs 24% reduction for distal cancers (proximal cancers presumably found on colonoscopy). The advent of stool DNA testing might shift the curve more away from routine colonoscopy screening. So, bottom line: we have largely adopted colonoscopy screening based on what makes sense (and to the delight of the gastroenterologists/hospitals — “scoping for dollars” as one of my GI friends put it), though sigs are reasonable alternatives. We should probably discuss this with patients, though there is an uphill battle given the dominant culture on this. Maybe will change with new tests in the future.


Primary Care Corner with Geoffrey Modest MD: Gastric Cancer Screening/Prevention

15 Jul, 14 | by EBM

I have had 2 Cape Verdean patients over the past few years who have developed gastric cancer.  Several months ago I met with a Cape Verdean doctor who confirmed that gastric cancer was relatively common in Cape Verde.  Gastric cancer screening in general does not make sense in the United States given the low prevalence of gastric cancer.  However, many of our patients come from countries with much higher prevalence, prompting this review.  I am posting about this  generally because many of us see patients coming from high prevalence countries.

Most of the data is not great.  There has been mass population screening in Japan since 1983  for individuals over 40, where gastric cancer is the leading cause of cancer death.  A systematic review was done by the Japanese Health Ministry (see doi:10.1093/jjco/hyn017), which only found 10 studies directly related to screening, none of which were randomized controlled trials — only case-control or cohort studies.  They noted in Japan there has been an overall decrease of gastric cancer mortality from 1980 to 2003, from 69.9 to 34.5 per 100,000 in males, and 34.1 down to 13.2 per 100,000 in females.  In their systematic review they found that the best evidence was for barium studies (the most widely used intervention), finding a 40-60% decrease in gastric cancer mortality and a 5 year survival rate of 74-80% for those screened versus 46-56% for the non-screened group.  They found that the data were more mixed and less compelling for endoscopy screening, or blood tests for serum pepsinogen or Helicobacter pylori antibody.  A cohort study in Korea, looking retrospectively at 2485 patients with gastric adenocarcinoma, found that those screened at 4-5 year intervals had a higher risk for gastric cancer than those screened at 2-3 year intervals. Those at the highest risk, people with a family history of gastric cancer and those in their 60s, were found to have a higher stage of gastric cancer when the intervention was performed every 3 years as opposed to annually.  This all led to the recommendation for screening every 2 years by upper GI series or endoscopy for individuals over 40 years old.  Since none of these studies were RCTs, there may be significant biases (lead time bias, length bias, etc.)

There was an RCT in 2004 on H. pylori eradication as a means to prevent gastric cancer ( see JAMA 2004; 291: 187-194).  In this Chinese trial (from Fujian Province, where mortality rate from gastric cancer is 153/100K, and where they have found a 2-4 fold increase in gastric cancer in those H pylori positive) 1630 healthy carriers of H. pylori were enrolled, of whom 988 did not have any precancerous lesions on endoscopy (gastric atrophy, intestinal metaplasia, or gastric dysplasia) at study entry.  The H. pylori status was documented by the endoscopic exam.  Patients were randomly assigned H pylori therapy (a two-week course of omeprazole 20 mg, amoxicillin/clavulanate 750 mg, and metronidazole 400 mg, all twice a day) versus placebo, and followed 7.5  years.  Results:

 –76.4% of patients given triple therapy for H pylori were  successfully treated, per urea breath test.  Those who failed treatment were given quad therapy –the ultimate eradication rate was 83.7%
–Primary outcome (incidence of gastric cancer during followup): No difference, with 7 cases in the H. pylori treatment group and 11 cases in the placebo.
–Secondary outcome (incidence of gastric cancer, comparing those with or without precancerous lesions): In those without precancerous lesions none developed gastric cancer in those treated for H. pylori, 6 did the placebo group, statistically significant.  Of note, the cumulative incidence of cancer in the placebo group was increasing dramatically after about 6 years, whereas those were H. pylori negative remained without cancer (ie very impressive splaying of the curves).
–Smoking and older age were independent risk factors for the development of gastric cancer, with smoking, having a hazards, ratio of 6.2.

so, what is one to do in the United States?  At this point, given the lack of large RCTs, it seems to me to be hard to recommend an aggressive screening program with either upper GI radiography or endoscopy.  However, given the very high prevalence of H pylori infection in many of these patients (including our Cape Verdean patients), and given the known association of H. pylori infection and gastric cancer at least in some high prevalence countries, and given the RCT from Fujian Province,  I personally think that it would be appropriate to screen and treat patients for H pylori infection, using the H. pylori antibody as a reasonable marker of infection.


Primary Care Corner with Geoffrey Modest MD: Colon cancer screening, the latest

23 Oct, 13 | by EBM

2 new articles on colorectal cancer screening in NEJM last month.

1. retrospective review of the nurses’ health study and the health professionals follow-up study (not a randomized controlled trial), of 90K people followed 22 years (see  DOI: 10.1056/NEJMoa1301969).  results:

— total of 1815 incident cancers and 474 colorectal cancer deaths

–with any lower endoscopy done, 43% fewer colorectal cancers in those who had a polypectomy for adenoma, 40% fewer in those with negative sigmoidoscopy, and 56% fewer in those with negative colonoscopy. data consistent for men and women.

–for proximal lesions (prior data suggested less efficacy of screening on proximal lesions), 27% fewer cancers

–for death from colorectal cancers: 41% fewer with screening sigmoidoscopy and 68% fewer with colonosc

–reduced mortality from proximal cancers by 53% with colonoscopy and no difference with sigmoidoscopy

–(not surprisingly) those with cancer diagnosed within 5 years of colonoscopy had more aggressive cancers (CpG island methylator phenotype — CIMP — and microsatellite instability)

–benefit of colonoscopy extended 15 years for both proximal and distal tumors.  the numbers of individuals are significantly smaller in the group getting screening in the 10-15 year range (and it appears that they do even better), though the group with screening >15 years was worse (but still 31% lower than the controls) — so these authors suggest sticking with 10 years as the appropriate screening interval in those with normal initial colonoscopy.

–also, aspirin had no benefit beyond colonoscopy alone

so, this study did show significant benefit of colonoscopy for proximal tumors (there were articles a couple of years ago suggesting that we only do sigmoidoscopy because it is safer/cheaper/easier on the patient and the data on efficacy for proximal tumors was pretty questionable) and a clear/persistent benefit to screening, even if not repeated for more than 15 years

2. 30 year followup of the minnesota colon cancer control study (see DOI: 10.1056/NEJMoa1300720). 47K people studied, randomized to usual care vs annual or biennial fecal occult blood testing (this was a pretty profound study at the time, though 82.5% of the FOBT slides were rehydrated, which creates 4 times as many false positive results — 9.8% per screening in those with rehydrated slides vs 2.4% in those without rehydration — and then leading to 38% of those in the annual and 28% in those in the biennial groups getting colonoscopy (12,246 colonoscopies!!!). in some ways, this perhaps ended up being more of a colonoscopy study, since so many were done…..). the initial study results reported in 1993 (see Mandel J, NEJM 1993; 328: 1365-71) found a 33% decrease in colorectal cancer mortality, leading to the generalized recommendations to screen for colon ca. results of the longterm followup:

–732 deaths from colorectal cancer (by death certificates)

–screening led to 32% decreased colorectal cancer mortality in the annual screening group and 22% in the biennial group through the 30 years of followup. no diff in all-cause mortality. men benefited more than women in the biennial screening group

–no diff in total mortality

so, pretty clear that colon cancer screening is important, has durable benefit over the longterm, and (i think) both studies ultimately reinforce colonoscopy as the most effective tool.



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