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: Coffee and decreased mortality

8 Aug, 17 | by

by Dr Geoffrey Modest

2 articles just came out finding that coffee drinkers had decreased all-cause mortality.


One study was from 10 European countries (see doi:10.7326/M16-2945).



— 521,333 people enrolled in EPIC (European Prospective Investigation into Cancer and Nutrition) from 10 European countries, followed 16.4 years

— 41,693 deaths occurred (18,003 were from cancer, 9106 were from circulatory diseases, 2380 were from cerebrovascular diseases, 3536 from ischemic heart disease, 1213 from digestive disease, 1589 respiratory, 1571 external causes, and 418 from suicide)

— Dietary intake was assessed through different validated instruments, baseline biomarkers were assessed (albumin, alkaline phosphatase, ALT, AST, GGT, CRP, A1c, HDL, lipoprotein(a))

— analysis was based on the number of cups of coffee drunk daily: 0, <1, 1 to 2, 2 to 3, and >3 cups per day, where a cup was defined as 237 mL

— Mean coffee consumption was 300 mL per day in quartile 2 and 855 per day in quartile 4 for men; 253 for quartile 2 and 684 for quartile for in women

— mean age 50, BMI 25, 23% university educated

— there were significant increases in coffee consumption as: smoking increased, less physical activity by men but increasing activity by women, increased consumption of red and processed meats and fewer fruits and vegetables, and alcohol consumption



— compared with coffee non-consumers, those in the highest quartile of consumption had significantly lower all-cause mortality by multivariate analysis:

— men, 12% decrease, HR 0.88 (0.82-0.95), p<0.001 for trend, and lowest incidence was in quartile 3, 16% decrease

— women, 7% decrease, HR 0.93 (0.87-0.98), p=0.009 for trend, and lowest incidence in quartiles 2 and 3, 10% decrease

— there was no evidence of heterogeneity by countries, and overall similar trends were found for caffeinated vs decaffeinated coffee, though the association was somewhat more pronounced for decaffeinated

— for digestive disease mortality (one-third of which was from liver disease, and the results were largely from the mortality benefit from liver disease, including from cirrhosis), (multivariate analysis):

— men, 59% decrease, HR 0.41 (0.32-0.54), p<0.001 for trend

— women, 40% decrease HR 0.60 (0.46-0.78), p<0.001 for trend

— for cardiovascular disease mortality (multivariate analysis):

— for women, 22% decrease, HR 0.78 (0.68-0.90), p<0.001 for trend

— for cancer mortality (multivariate analysis):

— No  statistically significant difference in overall cancer incidence in men, though in general lower coffee consumptions seem to be somewhat better

— Statistically significant inverse association (ie benefit) between medium to low coffee consumption and lung cancer mortality in men, and liver cancer mortality and men and women. No difference in respiratory disease mortality

— for women, 31% increase In ovarian cancer, HR 1.31 (1.07-1.61), p=0.015 for trend

— higher coffee consumption also associated with lower serum alkaline phosphatase, ALT, AST, GGT. And in women, lower CRP, lipoprotein (a), and A1c levels



–2.25 billion cups of coffee are consumed worldwide daily. Several US studies (see prior blogs listed at the end) all suggested decreases in all-cause mortality with coffee consumption, though large studies have not been done in the European population.

–one advantage of this study is that it covers many different countries where coffee is made in different ways (some older studies have found differences depending on how the coffee was brewed)

–Overall, reviewing the breakdown of their data, moderate coffee consumption seemed to have the most significant benefit

–Probably the most important confounder in this study was the association between coffee consumption and smoking. Controlling for smoking showed some strengthening of the association between coffee drinking and the reduced risk for death. Coffee consumption was also associated with reduced risk of all-cause mortality among ever-smokers as well as other subgroups at higher risk. And limiting the analysis to never-smokers also demonstrated an inverse relationship with cancer, circulatory, digestive, and respiratory diseases.​

–This was still an observational study, so it is difficult to assess causality given the possibility of unexpected/unmeasured confounders. The results were similar when the analysis was limited to those reporting “excellent” or “good” health, probably making reverse causality less likely, where sicker people more likely to die might drink less coffee.



And, a US study found a lower risk of death in non-white coffee consumers (see doi:10.7326/M16-2472​)



–185,855 participants in the MEC (Multiethnic Cohort) prospective population-based cohort study between 1993 and 1996, with evaluation of mortality until 2012, average follow-up of 16.2 years

–this co​hort included African-Americans, Japanese, Native Hawaiians, Latinos and whites

–coffee intake was assessed at baseline

–58,397 people died during 3,195,484 person-years of follow-up



–adjusting for smoking, number of cigarettes smoked, age, sex, BMI, education, physical activity, alcohol consumption, total energy intake, energy  from fat, and pre-existing illnesses, compared to non-drinkers:

–1 cup coffee/day was associated with 12% decreased total mortality, HR 0.88 (0.85-0.91)

​–2-3 cups coffee/day was associated with 18% decreased total mortality, HR 0.82 (0.79-0.86)
​–>3 cups coffee/day was associated with 18% decreased total mortality, HR 0.82 (0.78-0.87)

​–the trend for increasing coffee consumption and decreasing mortality was highly statistically significant, with p<0.001

–similar trends for caffeinated and decaffeinated coffee, though benefit more pronounced in those drinking caffeinated coffee, and were similar for men and women.

–benefit was found for all of the different race/ethnic groups,  though did not reach significance for Native Hawaiians (there were many fewer Native Hawaiians in the study group)

–inverse relationships (ie benefits) were found in never-smokers, those <55yo, and those without reporting a chronic disease

–and inverse associations were found for deaths from heart disease, cancer, respiratory disease, stroke, diabetes, and kidney disease



–these results were similar to the findings in the PLCO (Prostate, Lung, Colorectal and Ovarian Cancer Screening trial) finding an 18% reduction in coffee drinkers, and the National Institutes of Health-AARP Diet and Health Study, finding a 10% reduction in men and 13% in women, though these studies were largely in white populations

–this study was limited by the assessment of coffee consumption, since it was self-reported and only assessed at baseline

–Why does this relationship make sense? it is not just the caffeine, since there were pretty similar benefits to caffeinated and decaffeinated coffee

–For cardiovascular disease: chlorogenic acid, lignans, quinides, trigonelline, and magnesium in coffee reduce insulin resistance and systemic inflammation. And, the improvement in inflammatory markers​ in the above study supports the finding of cardiovascular benefit (and also lower risk of diabetes in some studies)

–For hepatic disease (one of the strongest findings above): there is experimental evidence that caffeine has antifibrotic effects on hepatocytes, and also reduces fat accumulation, oxidative stress, and liver inflammation in mice

–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): not sure what to make of this. Some studies have found that coffee has antidepressant effects, also it turns out that mice in stressful situations given caffeine have fewer stress-induced changes in their brains and behavior, but this doesn’t really answer the question about coffee or especially about decaffeinated coffee in humans (of note, the number of suicides in the first study was pretty small and therefore more likely related to statistical aberration)


so, overall these studies add to many others finding health benefits to coffee consumption. which is really good news for many of us.


Here are links to some relevant prior blogs:

–see a Korean study finding that moderate coffee consumption is associated with lower coronary artery calcium scores

— see an analysis of the Nurse’s Health Study and the Health Professionals Follow-up Study, finding decreased all-cause mortality, both in smokers and non-smokers

–see the Molecular Epidemiology of Colorectal Cancer study, finding that there was a dose-response curve with increasing intakes of coffee being associated with lower odds of colorectal cancer

— see the International Agency for Research on Cancer of the World Health Organization study, finding that consuming hot drinks in general (greater than 65°C) was “probably carcinogenic to humans”, but for caffeine, there seems to be decreased incidence of bladder cancer, liver cancer, endometrial cancer and possibly breast cancer, with a comment “overall coffee drinking was evaluated as unclassifiable as to its carcinogenicity to  humans “. So, hot liquids are bad, but caffeine is probably okay-to-good

Primary Care Corner with Geoffrey Modest MD: leisure time activity and lower cancer risk

27 Feb, 17 | by EBM

By Dr. Geoffrey Modest

There have been a plethora of articles in the past year on the beneficial effects of exercise. I will use the next several blogs to sample some of these.

One article looked at the beneficial effects of leisure-time physical activity on 26 cancer types (see doi:10.1001/jamainternmed.2016).


  • 44 million participants from 12 prospective US and European cohorts had self-reported leisure-time physical activity at baseline (1987 to 2004). Leisure-time physical activity levels were assessed as cohort-specific percentiles on a continuous basis. Hazard ratios are based on high vs low activity levels (comparing the 90th versus 10th percentiles of activity)
  • Median age 59 years, 57% females, BMI 26.
  • 186,932 participants with cancer were included in the analysis
  • Moderate activity in general was defined as intensity of >=3 more METS; vigorous activity as >=6 METS (see below for a definition for METS, or Metabolic Equivalents


  • There was a lower risk for 13 cancers with higher levels of leisure-time physical activity (the statistical models controlled for age, sex, smoking, alcohol, race/ethnicity, education; as well as specific risk factors for some cancers, such as hormone therapy, age at menarche, age at menopause, and parity for several of the female-only cancers, etc.):
    • Esophageal adenocarcinoma, decreased 42%, HR 0.58 (0.37-0.89)
    • Liver, decreased 27%, HR 0.73 (0.55-0.98)
    • Lung, decreased 26%, HR 0.74 (0.71-0.77)
    • Kidney, decreased 23%, HR 0.77 (0.70-0.85)
    • Gastric cardia, decreased 22%, HR 0.78 (0.64-0.95)
    • Endometrial, decreased 21%, HR 0.79 (0.68-0.92)
    • Myeloid leukemia, decreased 20%, HR 0.80 (0.70-0.92)
    • Myeloma, decreased 17%, HR 0.83 (0.72-0.95)
    • Colon, decreased 16%, HR 0.84 (0.77-0.91)
    • Head and neck, decreased 15%, HR 0.85 (0.78-0.93)
    • Rectal, decreased 13%, HR 0.87 (0.80-0.95)
    • Bladder, decreased 13%, HR 0.87 (0.82-0.92)
    • Breast, decreased 10%, HR 0.90 (0.87-0.93)
  • But there were higher risks of:
    • Malignant melanoma, increased 27%, HR 1.27 (1.16-1.40)
    • Prostate cancer, increased 5%, HR 1.05 (1.03-1.08): but specifically for non-advanced prostate cancer (HR 1.08), with no association for advanced prostate cancer (HR 0.99)
    • Cancers that did not reach statistical significance included non-Hodgkin’s lymphoma (though this was borderline significant at the P=0.05 level, with an 8% decrease with increased leisure-time activity), thyroid, gastric non-cardia, soft-tissue, pancreas, lymphocytic leukemia, ovary, and brain
    • Controlling for BMI decreased the statistical significance for esophageal carcinoma, and rendered the associations with endometrial cancer, liver and gastric cardia to be nonsignificant
    • Associations were generally similar with overweight/obese versus normal weight individuals
    • Smoking status modified  the association for lung cancer but not the other smoking-related cancers


  • This observational study found the quite impressive result that exercise was associated with major decreases in 13 cancers (10, after adjusting for BMI), and the decrease was 20+ % (i.e., really large) for 7 of them.
  • There are obvious concerns with such a study, including the fact that they compared only the top 10th percentile to the lowest 10th percentile of leisure-time activity, which also likely includes more unaccounted-for biases (e.g., those in the highest percentile group being much more likely to have generally healthy lifestyles, which may not be fully reflected in the multivariate analysis). Also, as with any meta-analyses, there are bound to be significant differences in the methodology of each individual study, making the strict combination of them less rigorous. The measures of physical activity were self-reported, and the cutpoints of high vs low varied between the individual studies. Also, some of the measurements (e.g. BMI) were considered as dichotomous variables (either above or below 25) which could conceal their true contribution (i.e. a BMI of 25.1 may confer a very different risk from a BMI of 35.1; on the other hand a BMI of 24.9 may not be so different from a BMI of 25.1)
  • Another issue is that leisure-time activity reflects only part of the picture. Many of the old studies only looked at leisure-time activity because they could not figure out how to incorporate work-related activity into the metric. Work-related activity requires very detailed analyses of individual workplaces, given that people doing the same job in different workplaces may have amounts of physical exertion, depending on such factors as degree of automation, how large the workplace is and what the division-of-labor is, and, in general, how the specific job was structured, including the role of labor unions requiring employers to decrease the intensity or potential risks of many jobs. Also, since leisure-time activity typically reflects voluntary participation that would reinforce the above-stated potential bias that these people lead generally healthier lifestyles.
  • Another recent systematic/meta-analysis (see 174 world-wide studies, looked at the levels of total physical activity (leisure-time, occupational, domestic, transportation) and the risk of breast cancer, colon cancer, diabetes, ischemic heart disease, and ischemic strokes, finding:
    • Overall, major gains for all outcomes occurred at lower levels of physical activity (3000-4000 MET minutes/week)
    • Even 600 MET/week (the lowest level in the studies), had a 2% lower risk of diabetes (vs no reported physical activity)
    • But going from 600 to 3600 MET minutes/week reduced the risk additionally by 19%. further increases did not add much (e.g., 0.6% if increase from 9000 to 12000 MET minutes/week)
    • At higher levels of physical activity (>8000 MET minutes/week), they found:
      • 14% reduction in breast cancer
      • 21% reduction in colon cancer
      • 28% reduction in diabetes
      • 25% reduction in ischemic heart disease
      • 26% reduction in ischemic stroke
    • But, looking at the curves: for all endpoints but breast cancer, there was the most dramatic improvement going from about 1500 to 4000 MET minutes/week, with leveling off thereafter. For breast cancer, the curve showed a relatively linear decline with more activity. [Remember: this study included all physical activity: i.e. it is hard to translate the current US recommendation of 75-150 minutes/week of exercise into the above 1500-4000 MET minutes/week.]​
  • Potential mechanisms connecting exercise with decreased cancer include: decreased body fat (body fat could confer various risks, including increased estradiol levels; they did note that BMI did decrease the association with several cancers, however I would add that BMI is not the most specific measurement of body fat, and does not differentiate from the much more metabolically active and less healthy visceral fat from subcutaneous fat); also many/most hormonal systems are changed with exercise, including cortisol levels (which in themselves affect most other hormone levels), male and female sex steroids, insulin and insulin-like growth factors, and adipokines (and many of these hormone systems could be related to carcinogenesis, e.g. by altering immune function); as well as changes in inflammation, oxidative stress (which are especially related to  visceral fat), and the reduced colonic transit time which could affect colon cancer incidence.

So, a pretty quick and dirty study, but it does really reinforce the potential role of exercise in cancer prevention. This becomes even more of an issue given the predictions that the global cancer burden will increase dramatically (one model suggesting a doubling by the year 2030), especially as unhealthy lifestyles such as smoking and poor diet increase in resource-poor countries: there are increasing obesity trends and less physical activity as more people move to crowded and often quite polluted cities — these changes are associated with a pretty dramatic shift from mortality associated with infectious diseases to that associated with chronic, western-type diseases).

From blog of 1/3/17 on cardiovascular fitness as a vital sign:

  • As a quick guide to METs:
    • Light activity (<3 METs): includes walking 2.5 mph (2.9 METs)
    • Moderate activity (3-6 METS): includes walking 3.0 mph (3.3 METs), walking 3.4 mph (3.6 METs), stationary biking (light effort) 5.5 METs
    • Vigorous activity (>6 METs):  jogging (7.0 METs), calisthenics/pushups/situps (8.0 METs), rope jumping (10.0 METs)​

Primary Care Corner with Geoffrey Modest MD: understated cervical cancer mortality and hpv in men

16 Feb, 17 | by EBM

By Dr. Geoffrey Modest

2 recent articles looked at US cervical cancer mortality and hpv infections in men.

  1. The New York Times reported a huge racial gap in cervical cancer deaths in the United States (see ). They referred to an article which calculated a much higher death rate from cervical cancer overall in the US than previously found, with an increased disparity between black and white women (see DOI: 10.1002/cncr.30507).


  • The age-standardized rate for cervical cancer death reported by the National Center for Health Statistics from 2000-2012 was 3.2/100K in white women and 5.7/100K in black women
  • However, these results were not corrected for the prevalence of hysterectomies, and given that hysterectomies are significantly more common in black women, the above statistics understated the cervical cancer death rates (since these cancers are essentially eliminated in people who’ve had hysterectomies for benign reasons, especially if the cervix is removed).


  • The overall prevalence of hysterectomies was 20% for women >20 years old, higher in black women for all ages between 45-69, peaking for both white and black women at ages 65-69, but this peak hysterectomy rate was 58% of black women vs 43% of white women [remarkably high numbers overall and shockingly so for older black women!!]
  • Correcting for the prevalence of hysterectomies, the mortality rate was 10.1/100K in black woman and 4.7/100K in white women.
  • Based on this, the disparity in mortality rates was underestimated by 44% over the published NCHS numbers.
  • The highest corrected rate was in black women > 85 years, with a death rate of 37/100K vs 11/100K for white women!!!
  • Using this corrected analysis, the rate of cervical cancer deaths in white women decreased at 0.8% per year, whereas for black women the annual decrease was 3.6%.


  • Each year more than 12,000 women in the US are diagnosed with cervical cancer and more than 4000 women die from it.
  • For the women in this study, the likelihood of a supracervical hysterectomy (i.e., leaving the cervix) was <2% (data from before 2004), though now is closer to 4-7.5%. So more individual data, even if available, would not have altered the results much
  • Cervical cancer is largely preventable through screening, and screening rates may be lower in poor and minority areas, although published results on this are equivocal. The data are clearer that black women tend to present with more advanced disease and may receive different treatments than white women, e.g. less surgery and more radiation for the same stage of cancer. For example, a recent large study of more than 15,000 patients with advanced cervical cancer found that more than half did not receive treatment considered to be standard of care, mostly those who were black and poor.
  • The corrected cervical cancer mortality rates in black American women is similar to those of many resource-poor countries in Latin America, Asia (excluding Japan), the Caribbean, and Africa (including sub-Saharan Africa). For white women, their corrected cervical mortality rates are similar to those of Europe, Australia, and Japan
  • The corrected mortality is significantly higher in black vs white women in all age groups, except those aged 20-29 and 35-39 (though increases pretty dramatically in women older than this)
  • Some of the methodologic weaknesses of the study include the potential for biases related to incomplete data and the merging of 2 unrelated databases with very different methodologies. The data on age-standardized death rates came from the National Center for Health Statistics (they also looked at the SEER database, which did not include every state and notably did not include Louisiana), whereas the data on hysterectomy prevalence came from the Behavioral Risk Factor Surveillance System survey, which is based on interviews.
  • It was quite striking to me when I was working in Chicago many many years ago that a very large number of my middle-aged to older black female patients had had hysterectomies when living in the Southern US, but were unaware they even had the surgery (many reported having had some surgery, but were never told that the doctors had done a hysterectomy). We were told that this was a not uncommon method of enforced birth control for black women…. So, since some were unaware they even had the surgery (and may be part of the older women now), the racial disparity may be even greater.

So, it is quite striking that black women in the United States have such a high death rate from cervical cancer. And, perhaps a real concern is that with the repeal of the Affordable Care Act, which does cover such screenings, there may be less access for many people for appropriate screenings. In addition, I am very concerned that the upcoming, likely attacks on Planned Parenthood and other clinics providing cervical cancer screening etc., will decrease access especially for poor women and women of color.

Relevant prior blogs: which reviews the American Society of Clinical Oncology guidelines presents data from a study in the Netherlands, suggesting that negative cervical HPV screening in women over age 40 supports a strategy of screening every 10 years which was a review of urinary screening for HPV, with my concern that clinicians will be doing far fewer pelvic exams (which certainly has its pluses, since these can be invasive and uncomfortable procedures for women), but with the caveat that I have seen several younger clinicians feeling less comfortable doing pelvic exams even when clinically indicated presents a study suggesting that we should do Pap smears in HIV patients, even post-hysterectomy


  1. Another article came out looking at the prevalence of genital HPV infections as well as vaccination rates in US adult men, from the National Health and Nutrition Examination Survey (NHANES) of 2013- 2014 (see doi:10.1001/jamaoncol.2016.6192).


  • NHANES collects information of representative cross-section samples of the US population.
  • 1868 men aged 18 to 59 were examined and DNA was extracted from self-collected penile swab specimens for HPV genotype.
  • Demographic and vaccination information was gathered by self-report.


  • The overall general HPV infection prevalence in males aged 18-59 was 45.2% (i.e., 34.8 million men). bimodal pattern, with peaks age 28-32 and another 58-59
  • The infection prevalence with at least one high risk HPV subtype by DNA testing was 25.1%
  • The overall prevalence of infection for subtypes covered by the HPV-9 valent vaccine was 15.1% (the 9-valent vaccine covers 90% of subtypes responsible for cervical cancer in women)
  • The specific very high-risk subtype prevalences: 4.3% for HPV-16 (3.3 million men), 1.7% for HPV-18 (1.3 million men)
  • In vaccine-eligible men, the prevalence of infection with at least one HPV strain targeted by the HPV-4 valent vaccine was 7.1% and by the HPV 9-valent vaccine was 15.4%
  • Among vaccine-eligible men, HPV vaccination coverage was 10.7% (i.e. more than 25 million vaccine-eligible men did not receive the vaccination)


  • HPV is the most commonly known sexually transmitted infection in the US. An estimated 79 million people in the US are infected with HPV, half of new infections occurring before age 24. There was a study about 10 years ago finding that 50% of women in their first year at college acquired HPV infection. In men, an estimated 160,000 are infected annually with low-risk HPV infections
  • In men, an estimated 9000 HPV-related cancers occur annually, responsible for 63% of penile cancers, 91% of anal cancers and 72% of oropharyngeal cancers (the oral HPV infection rates are around 10% for men and 4% for women). HPV can also cause recurrent respiratory papillomatosis
  • Men seem to clear HPV infection pretty quickly, with a study of 290 men finding that the 12-month risk of acquiring a new infection was 29%, with the median time to clearance being 5.9 months (Giuliano AR. J Infect Dis 2008; 198: 827). So, it seems likely that the point prevalence in the above study significantly understates the life-time acquisition rate for HPV, which is similar to that of women.
  • BUT, the major public health issue for men is that they can transmit this infection to women, potentially leading to cervical cancer, with a significant morbidity and mortality (as in first article above)
  • The CDC therefore published their recommendations for HPV vaccination: females aged 11 to 26; males aged 11 to 21, but from 21-26 being “recommended for persons with a risk factor (medical, occupational, lifestyle, or other indication”). Probably makes sense to support male vaccination till age 26, similar to the female recommendations

So, bottom line: these studies are very concerning, since on the one hand HPV infections are wide-spread and the number of unvaccinated men who are vaccine-eligible is staggering; on the other hand, cervical cancer death rates are quite high in the US and with a pretty dramatic black-white differential.

Primary Care Corner with Geoffrey Modest MD: Normal BMI/Exercise Lower Cancer Risk

23 Sep, 16 | by EBM

By Dr. Geoffrey Modest

The International Agency for Research on Cancer (IARC) working group just assessed the relationship between overweight/obesity and cancers, finding 8 more cancers associated with obesity (see Lauby-Secretan B. N Engl J Med 201; 375: 794). They relied on over 1000 epidemiological/observational studies to assess this association, since there really are no large randomized clinical intervention trials with long-term follow-up assessing the effects of weight-loss vs maintaining weight to see if there is a difference in cancer incidence.

  • Background, worldwide estimates:
    • In 2014: 640 million adults in 2014 (an increase by a factor of 6 since 1975) were obese
    • In 2013: 110 million children and adolescents (an increase by a factor of 2 since 1980) were obese
    • In 2014: prevalence of obesity was 10.8% among men, 14.9% among women, and 5.0% among children; and globally more people are overweight or obese than are underweight.
    • In 2013: 4.5 million deaths worldwide were caused by overweight and obesity; the obesity-related cancer burden represents up to 9% of the cancer burden among women in North America, Europe, and the Middle East.
    • In 2012: 1 million new cancer cases and 8.2 million cancer-related deaths
  • The 8 new cancer associations:
    • Colon or rectum, RR = 1.3, with positive dose response relationships (e., the more overweight, the higher the risk)
    • Gastric cardia, RR = 1.8, with positive dose response relationships
    • Liver, RR = 1.8, with positive dose response relationships
    • Gallbladder, RR = 1.3, with positive dose response relationships (though in their analysis, comparing the top vs bottom decile of activity, this achieved a P=0.06 only)
    • Pancreas, RR = 1.5, with positive dose response relationships
    • Kidney, RR = 1.8, with positive dose response relationships
    • Esophageal adenocarcinoma, RR=8, with positive dose response relationships
  • In general the relative risks increased from 1.2 to 1.5 for overweight and from 1.5 to 1.8 for obesity for cancers of the colon, gastric cardia, liver, gallbladder, pancreas and kidney
  • These results were consistent in different geographic regions, and were similar for men and women
  • The previously known cancers with associations:
    • Breast cancer in postmenopausal women, RR of 1.1 per 5 BMI units, esp in estrogen-receptor positive tumors
    • Endometrial cancer: RR=1.5 for overweight,5 for BMI 30-35, 4.5 for BMI 35-40, and 7.1 for BMI>40
    • Ovarian cancer (epithelial): RR=1.1
    • Multiple myeloma, RR=1.2 for overweight, 1.2 for BMI 30-35, 1.5 for BMI 35-40, and 1.5 for BMI>40
    • Meningioma, RR = 1.5
    • Thyroid, RR=1.1
  • And there is some limited evidence of an obesity association with male breast cancer, fatal prostate cancer, and diffuse large B-cell lymphoma
  • For breast cancer, there was an association between increased BMI at the time of diagnosis and reduced survival
  • In terms of weight loss: the quality of the data are not great, but there are some suggestions that weight loss (including by bariatric surgery) may reduce the breast and endometrial cancer risks.
  • As supporting evidence:
    • Animal data (different animals) confirm an association between obesity and cancer at many different sites
    • Animal data also supports the effect of limiting weight gain vs food ad libitum for some cancers (mammary gland, colon, liver, pancreas, skin, pituitary) but inverse relationship with others (prostate, lymphoma, leukemia)


  • As with all of these observational studies, association does not imply causality. For example, is it the obesity itself which is associated with cancer? Or, are there specific things that obese people do differently than normal weight ones (e.g., eating certain oncogenic foods? not exercising enough? living in more toxic environments?)
  • The above results were similar for BMI and waist circumference when that data was available (waist circumference has a higher correlation with visceral obesity, which is the metabolically more active obesity associated with metabolic syndrome, increased inflammatory markers, )
  • In many of the above associations, the associations persisted in studies using mendelian randomization (see , which describes mendelian randomization and some of its limitations, but overall it is a process that assesses known genetic markers for a disease to help assess causality (to differentiate in this case whether the causality is if those genetically predisposed to obesity are more likely to get the cancer, not vice-versa or as independent phenomena)
  • Possible mechanisms: increased body fat is associated with multiple metabolic and endocrine changes (sex hormones, insulin and insulin-like growth factor, inflammation), which could promote tumor initiation and/or growth
  • It is important to keep in mind the strength of the associations above. Typically, in observational studies, a relative risk of under 1.5-2 often does not pan out as being really significant, despite the fact that it can be really significant in randomized controlled trials. So, a bit of a caution in over interpreting the above results for many of the cancers. The dose-response relationship does add some support the associations, however.


Another recent article came out on the relationship between physical activity and cancer (see doi:10.1001/jamainternmed.2016), finding that leisure-time physical activity was associated with lower risk of many cancers. Details:

  • 12 prospective US and European cohorts with self-reported physical activity from 1987-2004, including 1.44 million participants, looking at 26 different cancers
  • Mean age 59 (19-98), 57% female, mean follow-up 11 years (7-21), mean BMI 26, 54% ever-smokers
  • 186,932 cancers diagnosed
  • Leisure-time activity, defined as high if 6 or more METs. Median activity was 8 MET-h/week (equivalent to 150 minutes of moderate-intensity exercise, e.g. walking)
  • Results:
    • High vs low leisure-time activity was associated with lower risk of:
      • Esophageal adenocarcinoma (HR 0.58, i.e., 42% decreased risk)
      • Liver cancer (HR 0.73)
      • Lung cancer (HR 0.74)
      • Gastric cardia (HR 0.78)
      • Endometrial (HR 0.79)
      • Myeloid leukemia (HR 0.80)
      • Myeloma (HR 0.83)
      • Colon (HR 0.84)
      • Head and neck (HR 0.85)
      • Rectal (HR 0.87)
      • Bladder (HR 0.87)
      • Breast (HR 0.90)
    • In aggregate, there was a 7% lower risk of total cancer in those performing higher levels of physical activity [HR 0.93 (0.90-0.95)]
    • Adjusting for BMI (nullied the relationship above for liver, gastric cardia and endometrium) but otherwise only a small attenuation of the risk, on the order of 5-11% of the HR’s. Smoking status affected lung cancer but not the others
    • Some cancers were associated with more activity
      • Melanoma (HR 1.27)
      • Prostate cancer (HR 1.05)


  • One striking finding is the overlap of cancers which seem to be affected by both BMI and exercise, reinforcing that these lifestyle/environmental issues seem to be particularly important.
  • But, one needs to be particularly careful in meta-analyses in general and huge ones in particular: it is very hard to get granular data over time (what is “ever-smokers”? a few cigarettes at the beginning of the study? stopping smoking 2 packs/day near the end of the study?); how often did they track information, such as changes in BMI or physical activity over time? Was it just a one-shot assessment at the beginning of the study? And how did they then quantitate these typically changing variables over such a long follow-up?  This data acquisition is done differently in different studies, so how is this all put together mathematically? It is pretty striking the range of ages (19-98) and years of follow-up (7-21) in the individual studies, suggesting they were pretty heterogeneous. And, in general, the people in this large meta-analysis were reasonably lean (BMI=26), so it may be difficult to really control for BMI in their data (they divided the patients into BMI <25 vs >25, but did not have the BMI spread of the IARC study). This limits the interpretation of their finding in this exercise study that 3 of the highest risk cancers in the AIRC study for BMI had no relationship to exercise when controlling for BMI.
  • They only looked at leisure-time physical activity. It seems pretty intuitive that people with very physical jobs do have more exercise at work than those with office jobs (i.e., many of my patients are on their feet all day, walking around cleaning office buildings, etc. And it seems they should get some “exercise” credit for that.) There are not great studies which have looked at occupationally-related exercise, probably because it is hard to measure on an individual basis: even those with the same job category may have very different amounts of exercise if they clean a small office vs a large automated office building)
  • One concern is that the burden of obesity and lack of exercise is increasing, especially with migration to larger cities and with increasing Westernization around the world
  • But one potentially positive finding is that exercise is associated with lower cancer risk independent of BMI for many cancers (with above caveat): it is much easier to help people do exercise than to achieve sustained weight loss (see ). And there are reasonable postulated mechanisms by which exercise could decrease cancer: hormonal changes (sex steroids, insulin and insulin-like growth factos, adipokines; similar to the BMI mechanisms postulated above) as well as nonhormonal (decrease inflammation, improve immune function/surveillance, decrease oxidative stress, and increase GI transit time, the latter of which could decrease colon cancer incidence)
  • There are still many questions, even if one accepts the conclusions of these studies
    • Does instituting a more aggressive exercise program lead to decreased cancer (i.e., an intervention study would provide stronger conclusions than an observational study)
    • And how much exercise works? Is there a threshold? Is it different for different cancers? (this might be important in different parts of the world where different cancers predominate)
  • But, the real bottom line is that there have been many studies over the years showing that lifestyle/environment are associated with pretty much all of the chronic diseases in the world. The above studies simply reinforce the association with cancer. And it offers us as clinicians yet another way to talk with patients about the importance of a healthy lifestyle. The association with cancer may be a particularly useful tool in motivating patients to avoid progressing to a less healthy lifestyle over time or instituting changes to improve their lifestyle (for better or worse, patients given equal mortality scenarios from cancer or heart disease, for example, are more afraid of the cancer one…it just sounds scarier)

Primary Care Corner with Geoffrey Modest MD: Coffee, Hot Drinks, and Cancer

4 Jul, 16 | by EBM

By Dr. Geoffrey Modest

A working group of the International Agency for Research on Cancer (IARC), the specialized cancer agency of the World Health Organization, reviewed data on the relationship between caffeine as well as hot beverages with cancer, finding that caffeine is good but hot beverages bad (See The full Monograph is in press, but Lancet Oncology published a summary.


  • Caffeine had been listed in 1991 as “possibly carcinogenic to humans” based on limited data of increase in bladder cancer. But:
    • Bladder cancer: newer, better data (>1000 observational and experimental studies) suggest that there is no consistent relationship, and raised the concern in older studies of potential confounding from cigarettes or occupational exposures.
    • Liver cancer: there seems to be an inverse correlation, based on case-control studies, with decreasing risk by 15% for each cup of coffee/day. Beneficial effects on liver fibrosis and cirrhosis
    • Breast cancer: meta-analysis of 1 million women suggests no association or ?modest inverse one
    • Pancreas and prostate: no increase
    • Endometrium: likely inverse associationwith coffee
    • Other cancers (lung, colorectal, stomach, esophagus, oral cavity, ovary, brain, childhood leukemia): data inconsistent or not evaluable because of biases, etc.
    • Conclusion: “overall coffee drinking was evaluated as unclassifiable as to it carcinogenicity to humans”
  • Hot drinks, IARC classified Mate, served traditionally at a very hot temperature (>65°C), as “probably carcinogenic to humans”. Mate is made from dried leaves of Ilex paraguariensis (Mate Yerba) with varying caffeine content and mainly consumed in South America, and less so in Middle East, Europe, North America.
  • Esophageal cancer: significantly increased risk, but only when drunk very hot or hot. and similar results have been found for other very hot drinks
  • Animal studies: very hot water (65-70°C) increased the incidence of nitrosamine-induced esophageal tumors. And hot water (>65°C) can act as a tumor promoter, perhaps through chronic cell injury
  • Conclusion: “drinking very hot beverages at above 65°C was classified as ‘probably carcinogenic to humans’, but “drinking mate that is not very hot was ‘not classifiable as to its carcinogenicity to humans'”



Primary Care Corner with Geoffrey Modest MD: Profits From Cancer Drugs – Further Drug Company Shenanigans

11 Mar, 16 | by EBM

By Dr. Geoffrey Modest

BMJ just had another of their many articles on drug company malfeasance, this one on mega-profits from their packaging of cancer drugs​​ (see BMJ2016;352:i788). The authors are from Memorial Sloan Kettering in New York and the University of Chicago.


  • Background:
    • The issue is that cancer drugs dosed by body size are often in single-dose vials and are typically packaged in quantities larger than needed.
    • The leftover drug, which could have been used for another patient, needs to be used within 6 hours and only in specialized pharmacies.
    • Hospitals and doctors also profit by buying these meds through the “buy and bill” system (they buy the drugs, then bill the insurers but with a mark-up which can vary widely — the current Medicare mark-up is 4% but can be as high as 58% if the hospital has a 340B pharmacy (see below), for commercial insurers it is 22% for doctors and 142% for hospitals and even higher through a 340B pharmacy). And they benefit more from buying more drug than they use.
  • Extent of the problem:
    • The researchers looked at the top 20 cancer drugs that are dosed by body size and packaged as single-dose vials (which they say accounts for 93% of all sales of such drugs)
    • Their methodology: estimate how often vial-sharing occurred (using Medicare claims, from which they can ascertain the amount of drug used that did not total the full vial contents); then calculate what they thought would be the most efficient packaging method based on the general US population height/weight measurements from the National Health and Nutrition Examination Survey, which they adjusted “to mirror a cancer patient population”
    • Their estimate: 10% of the drugs are discarded. But this seemingly small % translates to $1.8 billion from the $18 billion in sales. the amount discarded varies by drug (from 1-33%), for example:
      • Rituximab: 7% of $3.9 billion was discarded, for a loss of $254 million (and drug company profit of the same)
      • Carfilzomib: 33% of $697 million was discarded, for a pretty similar loss of $231 million
    • Sensitivity analyses did not provide much variation:
      • If the drugs were prescribed at the highest dose suggested by the FDA, the overall loss would be $1.4 billion
      • If every cancer patient weighed 10% lower than the survey population, the loss would be $2 billion
    • Why are there such variations in waste?
      • It really varies by drug packaging, e.g.:
        • Bendamustine (used for leukemia) is packaged in varying sized vials (25, 45, 100, and 180mg), so there can be pretty precise dosing/minimal waste
        • Bortezomib (used to treat multiple myeloma) only comes in one size (3.5 mg vial), which is much larger than the average dose of 2.5 mg, with $309 million in waste (in the UK the drug is sold in 1mg vials)
        • ​Pembrolizumab was initially marketed as 50-mg vials, then was changed so that 100-mg vials were the only ones made in the US (vs continuing the 50-mg vials in Europe). The average 70 kg patient needs 140mg of drug (ie three 50-mg vials, with minimal waste, or two 100-mg vials with lots of waste). So, the net effect of the repackaging is an extra $1.2 billion for the company over the next 5 years.
      • Hospital and doctor billing, and the effects on patients
        • As above, there are considerable markups by hospitals and doctors, which is projected to translate to > $1 billion in 2016
        • In terms of patient costs, about 1/2 of cancer patients have Medicare B, which has 20% copay and no upper limit, and 14% of them have no additional coverage for their coinsurance.
      • Other diseases:
        • Infliximab (one of largest selling drugs in the US at $4.3 billion in 2015, used for a variety of inflammatory conditions) is only available as 100mg single dose vials, which by above type of calculations leads to $500 million in extra revenues for the drug company.
        • ​Omalizumab (for asthma), used in 75 mg dosing increments, is available only as 150 mg vials
      • Conflicting regulations
        • One issue is conflicting regulations: e.g.
          • FDA: encourage drug companies to package drugs responsibly to minimize waste
          • Medicare: okay to share open vials with other patients, if precautions taken
          • But, CDC: only use single dose vials for one patient.

So, a few points:

  • Yet again we are confronted with drug companies making huge profits by manipulating the system. The Feds deserve some credit/blame for not having a consistent directive on using leftovers of these extremely expensive drugs, as I do all the time with leftover lidocaine or immunizations from multi-use bottles (there may be issues of shelf-life that vary with different drugs, of course). But the bigger regulatory let-down is that, unlike almost all other countries, we have no regulatory restraints on the drug industry. They can charge whatever they want. As we see again and again. For example, when the new drug for hepatitis C infection Harvoni came out, there was not even the pretense that the price reflected drug company reported costs (however inflated they might be), but instead were considered “reasonable” given the cost of treating the complications of otherwise untreated hep C.
  • The 340B program began in 1992 to “stretch scarce federal resources as far as possible, reaching more eligible patients and providing more comprehensive services.” The goal of the 340B pharmacy act was to make meds more available at a lower price in areas of need, not to allow larger profits for hospitals and doctors.
  • The magnitude of these profits is shocking/mind-numbing (and, remember, the $2 billion/yr is only for the top 20 cancer drugs). Just think how many more uninsured/underinsured people in the US could receive accessible, affordable care if the $$ were channeled there instead of drug company coffers…..

Primary Care Corner with Geoffrey Modest MD: Microbiome and Cancer

15 Dec, 15 | by EBM

By Dr. Geoffrey Modest

Well, here is yet another microbiome blog, but this one suggests that the gut microbiome and associated immunologic changes could affect cancer development and therapy. There were 2 articles in an issue of the journal Science which assessed this (002)

  1. One study looked at how changes in the microbiome affect host immunity and the natural response to implanted cancer (see Science 2015; 350 (6264): 1084). Prior studies have found a role for the intestinal microbiome in human systemic immune responses.


  • 2 genetically-similar mice from 2 different suppliers (JAX and TAC) with 2 different intestinal microbiomes (257 different bacterial taxa) had different responses to implanted melanoma, with the JAX mice having higher tumor-specific T-cell responses, intratumor CD8 T-cell accumulation, and improved survival
  • Cohousing the JAX and TAC mouse eliminated the difference, as did transferring fecal specimens from the JAX mice into the TAC mice (but not if gavage saline or TAC fecal material into TAC mice)
  • The clinical effect of having the JAX microbiome was equivalent to using a specific antibody immunotherapy (PD-L1), though the combo of this immunotherapy with the more immunogenic JAX microbiome was even better (human patients with evidence of endogenous T cell response also do better with immunotherapies)
  • In assessing the genus-level taxa of the differences in the microbiomes of JAX and TAC mice, the presence of Bifidobacterium species in particular showed the positive association with antitumor T-cell responses (400-fold). And Bifidobacterium species (including B. breve and B. longum) administered to TAC mice led to improved tumor control, which was abrogated in CD-8 depleted mice, suggesting that the effect of this bacterial species was not direct but was mediated through modifying endogenous immunologic responses
  • The same basic finding was found with bladder cell implants
  • Lactobacillus had no effect on tumor growth
  • Further experiments suggested that the antitumor effect of Bifidobacterium was not related to indirect effects of the Bifidobacterium​ on other bacterial species but was acting on its own. In addition, Bifidobacterium was shown to alter lymphocytic dendritic cell activation and improved tumor-specific CD8 T-cell function.
  1. Another study (which I will not summarize much) found that antibodies targeting a regulator of T cell activation (ipilimumab, a human monoclonal antibody useful in patients with metastatic melanoma) only worked if the gut microbiome had Bacteroides species present, especially B. fragilis (see Science 2015; 350 (6264): 1079)

So, I am continually amazed at both the extent of the rather profound impacts of the microbiome in preserving many health-related outcomes as well as its remarkable fragility. And, I really do not want to be reductionist — the Bifidobacterium seems to be protective in a few studies, along with other bugs such as Bacteroides. But there are undoubtedly complex interactions between the different species, some known and some not, which are essential for a healthy microbiome. And some of these commensals may be important for some functions and others for other functions. I.e., the answer is not simply to take certain probiotics (which may be helpful), but to do everything we can to preserve the microbiome naturally. As noted in several of themicrobiome blogs, the issue is to emphasize a healthy life style (esp. diet and exercise), eat good foods (e.g. avoid many of the artificial stuff around us, such as artificial sweeteners), and avoid toxins as much as we can (antibiotics, both given unnecessarily in the medical setting, and especially in use in agriculture and in animals).  No doubt, the microbiome is often able to regenerate (and we are lucky that many of our natural systems, including the environment are able to repair themselves from natural and human interruptions), but as with all such systems, there may be tipping points beyond which there is no return….

For other blogs on the microbiome: associates PPI use and C diff infections showing short term salivary and long-term gut microbiome changes after a single dose of antibiotics​ noting increased asthma in kids with early antibiotic exposure, as well as some data on celiac disease (also highlights possible protective effect of Bifidobacterium)​ showing microbiome changes with international travel and use of antibiotics looked at the effect of red meat on the microbiome (increasing TMAO levels, which are strongly pro-atherogenic) and showing that metformin induces positive changes in the microbiome which decrease insulin resistance finding that non-caloric artificial sweeteners induce microbiome changes causing increased insulin resistance and, several additional articles in on recommendations to limit antibiotic use or to use the most targeted and specific antibiotic when necessary in order to inflict the least harm on the microbiome

Primary Care Corner with Geoffrey Modest MD: Low Dose Aspirin and Colon Cancer Risk

9 Sep, 15 | by EBM

By Dr. Geoffrey Modest

The case for aspirin use for primary prevention of cardiovascular disease is a bit of a toss-up. However, there may be additional roles for aspirin (and NSAIDs) in cancer prevention (see doi:10.7326/M15-0039). Over the past 20-30 year, there have been many studies in animals showing a potential benefit of aspirin and NSAIDs in preventing colorectal neoplasia. And many observational studies in humans have found a reduced risk of colorectal cancer. More data is needed, though one of the lingering questions is the dose of aspirin associated with decreased colorectal cancer. This current study cross-references the Danish cancer registry with the pharmacy registry, with pretty complete data.


  • Population-based cohort study of people in Northern Denmark, comparing aspirin and NSAID use in 10,280 cases of documented colorectal cancer with 102,800 controls, between the years of 1994-2011
  • In Denmark, there is a pretty complete registry of patients with histologically-verified first diagnosis of colorectal cancer; also, >90% of total sales of low-dose aspirin are prescriptions and tracked in their systems. Of the NSAIDs, only ibuproben 200mg (15% of total Danish sales of NSAIDs) is over-the-counter, so the vast majority are tracked.
  • Median age 69.8, 54.8% male, 78% never used low-dose ASA, 54% never used NSAIDs, 12% on hormone replacement therapy, 15% statins, 7% diabetic, 20% with rheumatologic disease, 19% cardiovascular dz


  • Those continuously taking low-dose aspirin (75,100, or 150mg/d)  for >5 years had a 27% reduction in colorectal cancer risk [OR 0.73 (0.54-0.99)]
  • Those taking NSAIDs consistently for >5 years esp at the highest doses, hada 36% reduction in colorectal cancer risk [OR 0.64 (0.52-0.80)], and especially so in those NSAIDs with significant COX-2 selectivity
  • The data for ever-use (i.e., not taking continuously) was non-significant for aspirin, but significant for NSAIDs (6% risk reduction)

So, how does one piece this all together???

  • The data on cancer is largely from observational studies (and therefore susceptible to lots of biases), though the Danish one seems to be pretty tight. I am adding some of my older blogs at the end here, to reference some of these older cancer studies. It is notable that the aspirin data in the Danish study found potential effectiveness in those on low-dose aspirin. Their data on NSAIDs is also pretty strong and consistent with older data, though I would argue that NSAIDs have lots of other adverse effects (blood pressure, heart failure, renal failure…), and the most effective ones (COX-2 selective) are pretty consistently bad for the heart.
  • There have been several primary prevention studies in the use of aspirin for cardiovascular disease. Overall, the data, in this pretty low risk group (as opposed to secondary prevention studies) is: there is a significant 20% decrease [RR 0.80 (0.67-0.96)] in risk of first nonfatal MI, and non-significant decreases in total mortality and stroke, but a 54% increase in major extracranial bleeds [RR 1.54(1.30-1.82)] and a 12% decrease in cancer incidence [RR 0.88(0.80-0.98)]. For example these numbers translate to — for a 60 yo man over 10 years:
    • At low cardiovasc risk: 5 fewer nonfatal MIs, 4 more significant extracranial bleeds
    • At mod cardiovasc risk (eg 10-yr risk of about 5%): 17 fewer nonfatal MIs, 16 more significant extracranial bleeds, approx 1 more intracranial bleed
    • At high cardiovasc risk (10-year risk of about 10%): 27 fewer nonfatal MIs, 22 more significant extracranial bleeds, approx 1 more intracranial bleed
    • At low cancer risk: 1 fewer cancers
    • At mod cancer risk: 6 fewer cancers
    • At high cancer risk: 12 fewer cancers
  • As with most things patient-wise, it is useful to look at the individual instead of the aggregate data. Those in the primary prevention group who are healthy, active, have no risk factors, are at low risk for cancer, probably should not be on aspirin. Those on the opposite end of the spectrum (no cardiovasc disease, but smoker with lots of other risk factors; or at high risk of cancer) probably should be. All patients should get a sense of the risks and benefits, taking into account where they are in the spectrum of potential disease (heart has the best data, though I do include cancer risk) along with how they value potential outcomes (e.g. long-term potential protection from cancer and heart disease, esp nonfatal MI) vs risk of major bleed, including the much lower risk of intracranial hemorrhage.

Here are the older blogs: raises the issue of cancer prevention as part of the equation of its use in primary prevention has 2 studies: one showing decreased cancer mortality, the other decreased metastatic disease finding decreased ovarian cancer finding decreased prostate cancer metastases and mortality

Primary Care Corner with Geoffrey Modest MD: Chemotherapy at End of Life

1 Sep, 15 | by EBM

By Dr. Geoffrey Modest

A recent multi-institutional longitudinal cohort study assessed the use of chemotherapy in patients with progressive metastatic cancer, assessing survival, quality of life (QOL), and the relationship to performance status (see doi:10.1001/jamaoncol.2015.2378).


  • The American Society of Clinical Oncology (ASCO) in 2012 published a list of their “Choosing-Wisely” 5 opportunities to improve care and decrease costs, noting that chemotherapy use among patients with no evidence of clinical benefit was the most wasteful, unnecessary and widespread practice in oncology, especially in those with poor performance status with an Eastern Cooperative Oncology Group (ECOG) score of 3 or more (“capable of only limited self-care, confined to bed or chair >50% of waking hours”). Older studies from the 1980s have confirmed that those with poor performance had low response rates to chemotherapy, high rates of toxic effects, and shorter survival.
  • But, despite lack of evidence to support the practice, a paper in 2012 found that 40% of patients with non-small cell lung cancer and  ECOG score of 3-4 received palliative chemotherapy, with almost no hope of benefit. A Norwegian study also found that 53% of patients with ECOG performance score of 2 (in bed <50% of time)and 16% with scores of 3-4 received palliative chemotherapy, with 10% having chemotherapy in the last 30 days of their lives.


  • 312 patients were reviewed with a diagnosis of end-stage metastatic cancer,  refractory to at least 1 line of chemotherapy and MD-estimated life expectancy of <6 months. 158 received chemotherapy and 154 did not.
  • 8% male, average age 58.6, 12.4 years of education, 61.5% white/20.5% black/16.7% Hispanic.
  • The baseline performance status was assessed on average 3.8 months before death: 122 patients had ECOG score of 1 (somewhat restricted in strenuous physical activities but ambulatory), 116 with score of 2 , 58 with score of 3


  • On multiple logistic regression, those receiving chemotherapy were much more likely to be in an academic medical center than community clinic (adjusted odds ratio of 17.1, with the vast majority receiving chemotherapy at the academic centers at Yale, Simmons in Dallas, and Dana-Farber in Boston; about 50/50 at the West Haven VA and Parkland; and a significant minority in New Hampshire Oncology-Hematology); had better ECOG performance scores (AOR=0.67); had pancreatic (AOR= 4.07) and breast cancer (AOR=2.45); and were younger (AOR=0.96)
  • Patients’ risk of death was NOT significantly associated with whether they received chemotherapy or not, after adjusting for enrollment site and baseline ECOG score
  • In assessing QOD (QOL near death, asking the caregiver most knowledgeable about the health care of the patient in his/her final week of life by a validated questionnaire), by performance status:
    • ECOG 1: lower QOD !!! [odds ratio 0.35 (0.17-0.75), p=0.01]
    • ECOG 2: QOD [relative risk 1.06 (0.51-2.21), p=0.87, NS]
    • ECOG 3: QOD [relative risk 1.34 (0.46-3.89), p=0.59, NS]

So, a few issues:

  • This was a retrospective analysis of a cohort study and is therefore limited by not having individual data on the people who had chemotherapy vs not. Also there was likely a significant selection bias in those choosing a prestigious academic medical center vs a community-based center for their care.
  • But, it is still quite striking that of those in the academic medical centers 79% had chemotherapy, vs 46% at the VA and 22% in the community-based practice.
  • One of the most useful findings was that despite no difference in survival (again, not the most conclusive data, given limitations of study design/data), those with the best performance scores as assessed by their closest caregivers​​ <4 months prior to death actually had much worse QOL when given chemotherapy.
  • A related issue, I have some concern about palliative care being a separate discipline from oncology. Although I assume that oncologists in general do get significant training in end-of-life management, I have heard examples of oncologists just referring patients to palliative care specialists to discuss these issues. To the extent that there is a separation, this creates the unfortunate dynamic of oncologists pushing therapies, even not terribly beneficial ones, on the one side, and the palliative care specialists pushing for a more rational discussion of the benefits and risks of therapy (which they may not understand fully, given lack of oncology training in many cases). Without an integrated, coherent approach, the patient is caught in the middle, perhaps being emotionally swayed by the thought that there is a potential therapy. In addition, oncologists historically have had a significant financial interest in giving meds. There was a major rollback of the huge profit margin oncologists had received prior to 2005 for administering medications they bought and administered. Now there is a 6% markup of the Average Sales Price, though Congress realized that this would dramatically decrease oncologist income, so they significantly increased the fees paid for chemotherapy administration. Coincident with this reimbursement change, “physicians switched from dispensing the drugs that experienced the largest cuts in profitability… to other high-margin drugs” –see Health Affairs July 2010; 29: 1391-9.  And the office-based administration fees skyrocketed in some cases (I suspect more so in the large cancer centers). So there still is a significant financial incentive for many oncologists to give drugs… and the concern is that this financial incentive, especially if the oncologist is not so involved in the individual’s end-of-life/palliative care issues, could lead to inappropriate/unnecessary, potentially harmful, and expensive care.

Primary Care Corner with Geoffrey Modest MD: Melanoma increasing (a lot)

23 Jun, 15 | by EBM

By: Dr. Geoffrey Modest

MMWR just released an analysis of melanoma incidence, mortality trends, and projections in the US from 1982-2030 (see here). Melanoma is common (5th most common cancer in men and 7th in women) and is associated with the most skin cancer deaths, with deaths occurring most frequently in younger people (average of 20.4 years of potential life lost). Costs of treatment are high ($3.3 billion/yr in the US). And melanoma is largely preventable — 90% are attributable to skin damage from UV light exposure, with sunburns being a significant risk factor and 40% of US people reporting sunburn every year. There is clear efficacy of sun-protective behaviors in decreasing UV skin exposure, such as sunscreen, protective clothing. In addition there are suggestive data that cumulative UV radiation exposure over time may also be a culprit. results from MMWR:

–in 2011: 65,647 invasive melanomas in the US, with age-adjusted incidence rate of 19.7/100K, increasing with age and highest in non-Hispanic whites, in women 15-49yo, and in men >50.

–in 2011: 9,128 melanoma deaths, with age-adjusted death rate of 2.7/100K, higher in non-Hispanic whites, increasing with age, and higher in men (4.0) than women (1.7)

–from 1982-2011: melanoma incidence rates increased (doubling), though mortality remained the same.

–without intervention, there are projected to be 112,000 new cases in 2030 with annual cost for treatment projected to increase 252.4% from 2011-2030 (from $454 million to $1.6 billion)

–a comprehensive skin care prevention program (modeled after the one in Australia, which raised community awareness through mass media campaigns, programs in schools and workplaces, provider education, etc, had projected savings of $2.30 for every $1 spent) is estimated to prevent 20% of US melanoma cases from 2011-2030, averting 21,000 cases/yr, and reducing spending $250 million/yr (saving $2.7 billion from 2020-2030).

So, this seems like a largely preventable problem. There are pretty easy individual strategies available, including not using indoor tanning salons and using appropriate sunscreen protection — correct amount, reapplication rates, etc.  But, as in Australia, the major potential impact would be from concerted, community-oriented and community-based public health initiatives. Some of the above gender differences may be attributable to the increased female use of indoor tanning salons (a 2013 report found that approx 1/3 of non-Hispanic white women aged 16-25 use indoor tanning salons each year) and decreased male use of sunscreen protection. It is also important to remember that there still is a melanoma risk in darker skinned individuals. Overall black incidence in 2011 was 1.0/100K, vs 24.6/100K in non-Hispanic whites; but the mortality in black americans was 0.4/100K vs 3.1/100K reflecting a 3-fold increased mortality-to-incidence ratio vs non-Hispanic whites. This higher mortality is perhaps related to data suggesting that darker-skinned people report more frequent sunburns, are less likely to use sun-protection strategies, have a higher incidence of poorer-survival melanomas perhaps related to later diagnosis, a lower perceived risk by patients and providers, and more melanomas in non-sun-exposed areas.

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