Tuesday, December 21, 2010


One day you read in the paper or hear on TV that being overweight will cause you to die earlier. The following week you hear just the opposite. So what should you believe? Well, it is not easy unless you read the original epidemiological literature and analyze the results yourself.

But, you don’t have to do that because I have already done it.

Recently I posted a blog concerning body weight and death in which I summarized two recent studies which showed that bodyweight and death were related by a U-shaped curve (shown below). That is, if you are too skinny or too fat (obese) you are at a greater risk of death and that just being overweight did not increase the risk.

The data shown in the figure below was taken form McGee DL (2005) Ann Epidemiol 15:87 and Adams KF (2006) N Engl J Med 355:763


As you can see, if you weight too much or too little your risk of death is increased by approximately 1 to 2 percent as compared with the normal weight people. The group numbers indicate bodyweight category: 1. Underweight, 2-4. Normal weight, 5-7 Overweight and 8-10 Obese. These data are for men but the data for women are very similar.

It is clear from these data that people in the overweight groups of 5-7 are not at any greater risk of death than the normal weight groups 3-4. But there is a greater risk for being underweight (groups 1 and 2) or obese (groups 9 and 10).


Now a new study says that both overweight and obese people are at greater risk of death (de Gonzalez et al 2010 N Engl J Med 363 2211). So what should you believe?

As noted above, in the field of epidemiology, which is where all of these kinds of studies come from, disagreement almost always exists. In the case of bodyweight and death the disagreement level is greater than usual and downright ridiculous. For example:





So if the data and conclusions are so contradictory why go on? Why not just accept that the fact that the truth about fat and death will never be clear? Because very recently a paper was published which claims it has uncovered the truth and all other studies should be ignored (de Gonzalez et al 2010 N Engl J Med 363 2211). Right off the bat you realize, it can’t be true, so what is the real story.

This group has combined a large study which made up of 57 other studies (Prospective Studies Collaboration 2009 Lancet 373 1083) and added four more in order to come up with their own set of data from which they conclude that being overweight or obese will cause you to kick off sooner.

However, The Prospective Study showed the same U-shaped curve as we have already discussed. The only difference is that more groups were added in the Obese Category which made the curve slightly J-shaped. The values for Normal Weight average 9.6/yearly deaths /1000 and 9.8/yearly deaths/1000 for the Overweight group. Clearly not different. Hence being overweight poses no addition risk of death. The real skinny and obese groups were at a greater risk of early death but the overweight were not.

The deGonzalez group has taken the above data and combined it with four other studies and the combination was analyzed. Of the four studies listed two had no useable data and the other two are discussed below.

One of these by Hu et al (2004 N Engl J Med 351 2694) shows a linear trend of increase risk of death as a function of body mass index.

This is an example of an upward linear relationship and does not agree with most other studies. However the inclusion of these data in the de Gonzalez study decreases the upward curve of the underweight people and makes it appear that if there is a risk in this group it is very low. Notice that the difference between normal weight (2-3) and overweight (4-6) is 0.51-0.34= 0.17% which is not significant.

The study by Baik et al (2000 Am J Epidemiol 152 264) indicates that underweight people are at somewhat greater risk of death and that normal weight and overweight people are have no significant increase in risk. Whereas obese people are at somewhat greater risk.

Curiously, the study by Klenk et al (2009 Eur J Epidemiol 24 83) was not added to the list. Could it be because the data in Klenk et al 2009 show a U-shaped curve with conclusions about overweight people similar to those discussed above and obtained from McGee 2005 and Adams 2006.

In their final analysis de Gonzalez et al conclude that overweight and obese people are at greater risk of death. However, the small differences between overweight and normal groups are very small and of no biological significance. The authors are able to claim statistical significance because the number of subjects is so large. Thus many studies (some good, some poor, and some bad) many of which are not statistically significant can be brought together and suddenly statistical significance pops out. It is my opinion and the opinion of many others that such lumping of diverse studies into one analysis is a crock of statistical crap.

Obviously I believe that the U-shaped curve holds and overweight people are not at any greater risk of death than normal weight people.

Saturday, December 18, 2010


By James H. Clark

The following is a short summary of my paper which explains why postmenopausal hormones do not cause breast cancer, but instead, increase the level of detection which is a good thing.

Almost all of the studies published concerning postmenopausal hormone replacement therapy (HRT) use the phrase: The risk of breast cancer was increased as a result of HRT. The implication is that during the study HRT caused the cancer to form and grow to a size which could be seen by mammographic screening. This is very unlikely because breast cancers require from 10 to 20 years or more to reach a size that is detectable. Since studies of HRT are usually 5 years or shorter it would be impossible to observe a hormone initiated cancer.

So what are these scientists observing? They are detecting hidden tumors. In some women breast tumors exist in a hidden state (occult) and are not big enough to be detected. However, since some of these tumors are hormone sensitive they grow from the occult size to a size which can be detected by screening and are mistakenly labeled hormone induced cancer.

So every time you read in the paper that some HRT increases the risk of breast cancer, just substitute detection for risk. HRT is not causing breast cancer it is just detecting hidden cancers which can be a good thing.

The following paper explains in more detail the summary points made above.


By James H. Clark

Most papers published on the relationship between estrogen and progestin replacement therapy (EPHRT) of postmenopausal women indicate that such treatment increases slightly the risk of breast cancer. The implication being that these hormones cause breast cancer. Although these results have been discounted by several investigators, their conclusions continued to be quoted as fact in the scientific, news and television media. The purpose of this paper is to show that such studies cannot claim that EPHRT increases the risk of breast cancer. Instead what they have been studying is the small increases in the growth of preexisting cancers which were too small to be detected at the beginning of the studies. Therefore, the only claim they can make is the EPHRT may increase the detection of occult cancers and this may be a good thing.


The length of most EPHRT studies is too short for an increased incidence of BC to be detected if EPHRT were the cause of this increase. This is true because the time between initiation of breast cancer and the time that cancer can be detected varies between 10 to 20 years (Dietel et al. 2005; von Fournier et al. 1980; Koscielny et al. 1985). Since most EPHRT studies are between 5 and 10 yrs any tumor that was initiated by hormone treatment would not be apparent for at least 10 yrs and probably more.

As pointed out by Dietel et al (2005) these estimates of time between inception and detection are based on calculations of tumor doubling time (TDT) which has been reported to be between 23 and 209 days (Spratt et al., 1977; Shackney et al., 1978; Spratt and Spratt, 1985; Haskell, 1985). All things considered the average TDT is considered to be 50-100 days (Spratt et al., 1995). Since a mammary tumor cannot be detected until it reaches approximately 1 cm and would contain approximately 10 billion cells, the time to produce this number of cells would be between 5 and 10 years. This time period is an underestimate because the loss of cells due to apoptosis and the length of the carcinoma in situ, which is very variable and can be several years, are not taken into consideration.


Most studies of EHRT have shown no increased detection of breast canceer and some have shown a decrease. In contrast, EPHRT studies consistently show a small or marginal increase in detection of breast cancer. This probably occurs as a result of estrogen plus progesterone stimulating the more highly differentiated occult cancers which contain receptors for these two hormones. Progesterone appears to have angiogenic effects which may be the cause of an additional effect on growth which is not seen with estrogen alone (Liang 2007 Cancer Res 67 9929; 32: Hyder 1998 Cancer Res 58 392). Whereas, estrogen only exposure may not produce a significant growth response because the tumor is less well differentiated and is incapable of a growth response.


Increased detection as a result of EPHRT means that tumors could be treated sooner and this potentially could reduce morality. This may be the case since these tumors have favorable characteristics which are associated with decreased mortality. Many investigators have found a such characteristics associated with decreased mortality (Kerlinkowske 2003; Newcomb et al 2008)

The increased detection could be due to more frequent screening of women who use HRT than nonusers. However even in studies which adjust for screening bias the tumors HRT tend to be smaller ( Bonnier 1995; Magnusson 1996) of lower grade ( Harding 1996), less advanced stage (Holli 1998; Christante 2008) have fewer positive axillary lymph nodes (Bonnier 1995; Magnusson 1996, Hardin 1996; Squitieeri 1994), lower tumor cell proliferation rate (Oestreicher 2004; Holli 1998, and have other clinically more favorable features (Schnitt 2001; Holli 1998; Chen 2004; Rosenberg 2008; Newcomb 2008).

In contrast to the above studies Chlebowski 2010 found that mortality was increased by EPHRT in the WHI studies. The WHI authors explain that other studies are not as valid as those of the WHI because they were not randomized placebo controlled trials. However it is difficult to disregard the many studies which do not agree with WHI especially since some of them are just as valid, if not more so, than those of the WHI. The WHI studies were far from perfect and have been criticized at length by many scientists.

For instance: Many investigators agree that the WHI study does not even qualify as a randomized placebo-controlled study which is supposed to be superior to other types of studies. The reasons for this statement are: 1. Following randomization the women were free to decide whether to continue their assigned treatment or whether to undergo diagnostic procedures. 2. Almost half of the women were aware of their treatment so there was no valid placebo group. 3. Several warnings were sent to the participants about the detection of increased risks of myocardial infarction, stroke and pulmonary embolism during the study. These problems make the WHI study no better than any observational study with all of their limitations (Clark JH 2006).


Following the release of the WHI results in 2002 there was a decline in the number of detected breast cancers which was associated with a decrease in the number of women taking EPRT. Many investigators have taken this as evidence that EPRT causes breast cancer. However, it is much more likely that the decreased number of breast cancer was due to a decreased detection of occult tumors which would have been observed if EPRT had continued.

As Berry and Ravdin (2007) explained: If EPRT caused breast cancer and millions of women stopped taking EPRT there would be a long slow decline in the number of breast cancers, not a rapid decline between 2002 and 2003 as has been reported. Breast cancer has a long preclinical period that varies in duration from one tumor to another. So a drop in breast cancer incidence would be gradual and not discernable for several years after EPRT was stopped.

A more likely possibility is that a cessation of EPRT removes the hormonal stimulation of receptor positive cancers and their growth rate decreases. If this were the case a discernable and rapid drop in breast cancer numbers would be observed. To use a modified example from Berry and Ravidin (2007): Consider 30 to 40/100,000 women with receptor positive cancers that were growing and destined to be detected in 2003. They had negative mammograms in 2002 and then when the WHI results were announced they stopped taking EPRT. Without the hormonal stimulation their cancers stopped growing and were not detected by the mammograms in 2003. So the overall decrease on breast cancer incidence rates would be immediate and noticeable. This seems to be the case since the rate of decline of breast cancers was rapid between 2002 and 2003 and was parallel to receptor positive cancers (Ravdin 2007).


The preponderance of evidence indicates that EPHRT stimulates some occult breast cancers to grow to a size which is detectable by screening. Such tumors were not caused by the hormonal exposure because the length of time of most studies is too short for tumors to grow to a detectable size. Therefore these hormone responsive tumors are being detected before they would be in nonusers of hormone therapy. This appears to be a good thing since these tumors have favorable characteristics and are associated with a decreased rate of mortality.


The Endocrine Society just published a 65 page review of HRT which agrees and substantiates my views (Santen 2010 J Clin Endocrinol Metabol 95 S1-S66).


Berry DA, Ravdin PM (2007) Breast cancer trials: A marriage between clinical trial evidence and epidemiology. JNCI 99: 1139-1141

Bonnier et al.(1995) Clinical and biologic prognostic factors in breast cancer diagnosed during postmenopausal hormone replacement therapy. Obstet Gynecol 85:11–7

Chen et al (2004) Association of hormone replacement therapy to estrogen and progesterone receptor status in invaseive breast carcinoma. Cancer 101 1490

Chlebowski et al 2010 Estrogen Plus Progestin and Breast Cancer Incidence and Mortality in Postmenopausal Women. JAMA 304 1684

Clark (2006) A critique of the Women’s Health Initiative Studies: estrogen plus progestin (Prempro). Nuclear Receptor Signaling 4, e023

Dietel et al. (2005) Hormone replacement therapy: pathobiological aspects of hormone-sensitive cancers in women relevant to epidemiological studies on HRT: a mini-review. Human Reproduction 20:2052-2060.

Harding et al.(1996) Hormone replacement therapy and tumour grade in breast cancer: prospective study in screening unit. BMJ 312:1646–7.

Haskell CM (1985) Thorax, Unknown Primary—Breast Cancer in Cancer Treatment, 2nd edn. WB Saunders Company

Holli (1998) Low biologic aggressiveness in breast cancer in women using hormone replacement therapy. Clin Oncol 16(9):3115-20.

Hyder et al.(1998) Progestin regulation of vascular endothelial growth factor in human breast cancer cells. Cancer Res 58 392

Kerlikowske et al. (2003) Prognostic characteristics of breast cancer among postmenopausal hormone users in a screened population. J Clin Oncol 21:4314–21.

Koscielny S (1985) A simulation model of the natural history of human breast cancer. Br J Cancer 52:515-524.

Liang et al. (2007) Progestin dependent progression of human breast tumor xenograpfts : a novel model for evaluating antitumor therapeutics. Cancer Res 67 9929

Oestreiche (2004) Hormonal factors and breast tumor proliferation: do factors that affect cancer risk also affect tumor growth? Breast Cancer Res Treat 85:133–42

O'Meara (2001) Hormone replacement therapy after a diagnosis of breast cancer in relation to recurrence and mortality. J Natl Cancer Inst 93:754–61.

Ravdin (2007) et al. The decrease in breast-cancer incidence in 2003 in the United States . NEngl J Med 356 : 1670 – 4 .

Schuetz (2007) Reduced incidence of distant metastases and lower mortality in 1072 patients with breast cancer with a history of hormone replacement therapy. Am J Obstet Gynecol. 196(4):342.e1-9.

Sener (2009) The effects of hormone replacement therapy on postmenopausal breast cancer biology and survival. Am J Surg. 197(3):403-7.

Schnitt (2001) Traditional and newer pathologic factors. J Natl Cancer Inst Monogr 30:22–6.

Shackney et al.(1978) Growth rate patterns of solid tumors and their relation to responsiveness to therapy: an analytical review. Ann Intern Med 89, 107–121.

Spratt JS and Spratt JA (1985) What is breast cancer doing before we can detect it? J Surg Oncol 30, 156–160.

Spratt (1977) Cytokinetic definition of acute and chronic breast cancer. Cancer Res 37, 226–230.

Spratt et al (1995) Rates of growth of human solid neoplasms: part I. J Surg Oncol 60, 137–146.

Squitieri et al. (1994) Carcinoma of the breast in postmenopausal hormone user and nonuser control groups. J Am Coll Surg 178:167–70.

Rosenberg et al. (2008) Menopausal hormone therapy in relation to breast cancer characteristics and prognosis: a cohort study. Breast Cancer Research 10;R78 (doi:10.1186/bcr2145)

Newcomb et al (2008) Prediagonistic use of hormone therapy and mortality after breast cancer. Canceer Epidemiol Biomaarkers Prev 17:864-871.

von Fournier et al. (1980)Growth rate of 147 mammary carcinomas. Cancer 45:2198-2207.

Sunday, October 24, 2010


Just in time for Halloween the Women’s Health Initiative (WHI) investigators are trying to scare women again (Chlebowski et al 2010 JAMA 304 1684). They did this in 2002 with the help of unknowing press and TV journalists who spread the word that hormone replacement therapy (HRT, estrogen plus progestin, Prempro) was really bad for you and you should stop taking it. These authors were wrong then and they are wrong now. The new article says that HRT not only increases the risk of breast cancer but it also causes greater mortality.

For one thing the data in this paper are what polite scientist call marginal which translates to: through it out and forget it. All of the risk data are just barely statistically significant and the absolute risks are very small which means they are of no importance to any individual woman who has taken HRT.

These very weak findings of increased risk of death are in contrast to the many studies which say just the opposite. The WHI authors explain that other studies are not as good as the WHI studies because they are not randomized placebo controlled trials, so just forget them. Well it is hard to forget 15 to 20 studies which disagree with the WHI especially since some of them are just as good as those of the WHI. The WHI people would do well to remember that their studies are far from perfect and have been criticized at length by many scientists.

For instance: Many investigators agree that the WHI study does not even qualify as a randomized placebo-controlled study which is supposed to be superior to other types of studies. The reasons for this statement are: 1. Following randomization the women were free to decide whether to continue their assigned treatment or whether to undergo diagnostic procedures. 2. Almost half of the women were aware of their treatment so there was no valid placebo group. 3. Several warnings were sent to the participants about the detection of increased risks of myocardial infarction, stroke and pulmonary embolism during the study. These problems make the WHI study no better than any observational study with all of their limitations.

In addition to all of these problems, the women in these WHI studies were 12-15 years past the onset of menopause. Thus they were without their pre-menopausal levels of estrogen and progesterone long enough to bring about changes in various bodily functions which are the precursors of disease or of undiagnosed disease. For instance, ovarian hormones are important for maintaining normal structure and function of the blood vascular system. Once vascular disease has begun hormone treatment is not likely to reverse the effects. Proper bone strength is maintained by estrogen and when estrogen is no longer secreted at menopause bones begin to lose calcium and the first stages of osteoporosis begin.

So if anyone you know is or has taken HRT tell them not to be scared. It’s only the big bad WHI monster who makes a loud noise but has no claws.

For more information see my critique of the WHI studies from 2002-2006: Clark JH (2006) Nuclear Receptor Signaling 4, e023

Tuesday, September 28, 2010


In 2002 some scientists published a paper (see ref at end of this article) which said taking hormones after menopause increased risks of heart problems, breast cancer, stroke and blood clots. The news media were quick to exclaim to the world that: BIG BAD HORMONES WERE NOT GOOD FOR YOU.

This caused quite a stir and many physicians advised patients to stop taking their hormone drugs which contained estrogen and progesterone. This is too bad because the conclusions of the Woman’s Health Initiative (WHI) scientist were wrong for the following reasons:


In their first paper, the one in 2002 that caused most of the problems, the WHI authors used a measurement of variation called the nominal 95% confidence interval and they should have used an adjusted 95% confidence interval. Confidence intervals (CIs) are used in tests for statistical significance.

Nominal CIs can be used when a single outcome or result is being tested but in the WHI studies there were multiple outcomes being tested, so they should have used CIs which are adjusted to take into consideration the possible error introduced by multiple testing. Nominal CIs are smaller than adjusted CIs, so when they are used incorrectly they make numbers look significant when they are not.

For example: The relative risk (RR) and nominal CI for risk of breast cancer was 1.26 (1.0-1.59) which is barely significant ; however, when the correct adjusted CI is used (0.83-1.92) we see that it includes the no effect level of 1.0 and would be judged as not significant. This incorrect use of nominal CIs was used throughout the paper.

Such incorrect use of statistical techniques should have disqualified this paper from publication.


One of the biggest contributions to inaccuracies and exaggerations in these studies is the failure to state and emphasize that the actual or absolute risks are small. How can this be? It is really very simple as I have explained in my two earlier Blogs entitled: What is the truth about health risks? and How to understand blog graphs.

The bottom line is you can have very small actual risks accompanied by very large relative risks. Since the relative risk numbers are large they are the ones that show up in the news media. If the small actual risk numbers were released no one would pay any attention to them.

For more comparisons between relative and actual risk see the table below. Remember these claimed increased risks were not significant in the first place so they don’t count.


A good example of this lack of attention was their failure to see that the apparent increased risk of heart disease and breast cancer during a critical time (yr 5) was not caused by an increased number of women with these diseases but a chance decrease in the number of women in the placebo control group. There was no real or significant increase in risk in heart disease or breast cancer but these mistakes were the main reason for stopping the study.

Increases and decreases in risks may occur by chance or there may be some underlying cause. The proper statistical analysis tells us whether these differences are real or just due to chance. All too often in these kinds of studies the authors will make statements like: The risk of breast cancer was increased by hormone treatment. Then they add: although the increase was not significant. In other words the differences could have occurred by chance and there is no real increase. The only part of the above which is published in the press is: Breast Cancer Risks Increased by Hormone Treatment. The headline should have been: Breast Cancer Risks Not Increased by Hormone Treatment.

In a subsequent more detailed paper on the effects of hormone treatment these authors indicate that no significant increased risks of breast cancer were observed (Chlebowski et al 2003). This revised conclusion was not announced by the news media and no retraction of the original incorrect conclusions was made. Most doctors and researchers either do not know of this correction or choose to ignore it.


The WHI studies used data from women who were 12-15 years past menopause before they started taking hormones. This means these women were without their normal levels of estrogen and progesterone long enough to bring about changes in various bodily functions which are the precursors of disease. For instance, ovarian hormones are important for maintaining the blood vascular system in good shape. Once vascular disease has begun as a result of menopause hormone treatment is not likely to reverse the effects. Proper bone strength is maintained by estrogen and when it is no longer secreted at menopause bones begin to lose calcium and the first stages of osteoporosis begin. Most reproductive scientists believe that post menopausal hormones should be used as preventives not corrective therapy.


The best way to do a clinical study of this kind is to do it so that neither the participants nor the doctors know who is getting the drug and who is getting the placebo. This was supposed to be one of the strengths of the WHI studies; unfortunately, the women did not remain blind and as many as 45% of them were not only told what group they were in but also given warnings about the possibility of increased risks for heart disease, breast cancer etc. This destroys the credibility of such a study and when this happen the entire study should have been stopped because of these flaws.


The WHI investigators also evaluated disease outcomes in women who took estrogen alone as a post-menopausal treatment. The authors concluded that estrogen alone did not increase the risk of breast cancer or heart disease and decreased the risks of hip fracture. They also concluded that estrogen alone did increase the risk of stroke. These four papers, most of which were published in 2006, suffer from the same problems and bad sciences as outlined above. They are loaded with extreme variability in the treated and control groups which were not taken into consideration. Once again, they used the wrong measure of variability to test for significance, and called the increase risk of stroke to be significant when it was not. In addition, these estrogen only studies have all of the other problems discussed for the estrogen plus progestin studies.


Most people, including physicians, do not realize that the conclusions in the original 2002 paper were incorrect. Although the WHI investigators indicate in their later papers that no increased risks were found, they fail to point out clearly that their conclusions were incorrect in 2002 and that the study should not have been stopped. Little has been said in the popular media. No retraction, no admission that the WHI study should have been continued, no attempt to set the record straight. So most women and many physicians keep plodding along believing hormones increase risks for these diseases, when they don’t.


Chlebowski RT et al (2003) Influence of estrogen plus progestin on breast cancer and mammography in healthy postmenopausal women. JAMA 289: 3243

Writing Group for the Women’s Health Initiative Investigators. (2002) Risks and Benefits of Estrogen Plus Progestin in Healthy Postmenopausal Women: Principal Results From the Women’s Health Initiative Randomized Controlled Trial. JAMA:288:321-333.

Sunday, September 26, 2010


Surely high fiber diets are good for you?  I always thought so, but I began to wonder when I looked into the evidence.  This concept has suggested by many people. 

One of these beliefs is that high fiber diets will reduce the risk of heart attacks (myocardial infarcts, MI) and other heart diseases(see Mozaffarian et al for references). This belief comes from a paper by Rimm et al (1996) who claimed to demonstrate such an inverse relationship. This seemed to be a reasonable assumption and one I would like to believe, but after analyzing the data I am not convinced.

In the study by Rimm et al (1996) a total of 43,757 male health professionals 40 to 75 years of age were studied for 6 years. During this time 734 cases of myocardial infarction were documented and these were analyzed with respect to dietary intake of fiber. The results of this study are shown in Figure 1. You can see that no significant effect on MI is seen in any group but group 5 which was the highest intake category (28.9 grams/day).

So if you are willing to spend most of you time eating fiber containing foods all day you might be less likely to have a heart attack. However, anytime you have data points which suddenly go from insignificant, as in category 4, to significant, as in category 5 you should worry that the value in category 5 may be a fluke.

Figure 1. Relative risk of myocardial infarct (MI) and dietary intake of fiber.

Each intake category represents increasing intake of total dietary fiber. The data points are relative risks and 95% confidence intervals taken form the data in table 2 of Rimm et al 1996. The quantity of fiber in the diet increases from 1 to 5.

The next study from this same group involved 39,876 women health professionals who were followed for 6 years (Liu et al 2002). The incidence of cardiovascular disease (CVD) and MI were examined in this study and their results for total dietary fiber intake are shown in Fig 2. No significant associations were observed between total fiber intake and CVD (Fig 2 A), and the women in category 4 showed marginal significant decrease in MI (Fig 2 B). Since all other RR are insignificant these data do not support the contention that dietary fiber intake is associated with a decreased risk of CVD or MI.

Figure 2. A. Relative risk of cardiovascular disease (CVD) and B. myocardial infarct (MI) as a result of dietary intake of total fiber. Data taken from tables 3 and 4 in Liu et al 2002.

These authors also examined the association between the various sources of dietary fiber. If any type of fiber would show a protective effect one would think it would be vegetable fiber. Think again. The data in Figure 3 demonstrated that no decrease or increase in relative risks of CVD or MI result from the consumption of various amounts of vegetable fiber.

Figure 3. Relative risk of cardiovascular disease (CVD) and myocardial infarct (MI) as a result of dietary intake of vegetable fiber. Data taken from table 3 and 4 in Liu et al 2002.

If no associations could be shown with total fiber or vegetable fiber, then surely these would be found with soluble fiber. After all, soluble fiber was thought to be primarily responsible for the cholesterol lowering effect of dietary fiber, hence one would expect great things. Several studies claimed this to be true and were summarized in a meta-analysis by Brown et al 1999. However, if you read this paper carefully you would not expect much effect from soluble fiber. Brown et al 1999 found that the concentrations of total cholesterol and LDL cholesterol in the blood were decreased marginally and no effects on HDL cholesterol and triglycerides were found. The reduction of total cholesterol and LDL cholesterol were minus 0.045 mM ( 0.9%)and 0.057 mM ( 2.7%) respectively which are very small and probably physiologically insignificant effects. The authors point out that increasing soluble fiber intake can make only a small contribution to lowering plasma cholesterol. Little did they know, the effect was not just small but it was without consequence as a factor in decreasing the risk of CVD or MI (Figure 4).

 Figure 4 . Relative risk of cardiovascular disease (CVD) and myocardial infarct (MI) as a result of dietary intake of soluble fiber. Data taken from tables 3 and 4 in Liu et al 2002.

Other types of fiber which were included in this study were cereal fiber, fruit fiber and insoluble fiber. These produced no significant reduction in the risk of CVD or MI.


Try as they might, all of the kings men could not put the story of fiber and heart problems together again. No evidence of a protective effect on CVD or MI from eating even very large quantities of any kind of fiber was found in these studies.


Brown L, Rosner B, Willett WW, Sacks FM 1999 Cholesterol lowering effects of dietary fiber: a meta-analysis. Am J Clin Nutr 69:30-42.

Liu S et al 2002 A prospective study of dietary fiber intake and risk of cardiovascular disease among women. J Am Coll Cardiol 39:49-56.

Mozaffarian et al 2006 Trans fatty acids and cardiovascular disease. NEJM 354:1601-1613.

Rimm EB et al 1996 Vegetable, fruit and cereal fiber intake and risk of coronary heart disease among men. JAMA 275: 447-51/

Sunday, August 15, 2010


Overweight and obese people are generally thought to be unhealthy and have more diseases and higher death rate than people of normal weight. However, such information is usually found in articles written by authors who have not examined the actual medical evidence which support these statements. The purpose of this paper is to take a close look at the evidence for some of these claims.

There is no agreement among epidemiologists concerning the increased risk of death in overweight and obese individuals. Some studies indicate that no relationship exist and some that the relationship is the reverse. Some studies agree that obese people are more likely to die than people of normal weight, but there is disagreement concerning overweight people and increased risk of death. Some say yes and others say no. These studies have been summarized by McGee (2005) who concluded that overweight people do not have a significantly higher rate of death than normal weight people.

A recent study claims that even overweight women or men are at a greater risk of dieing than people of normal weight (Adams et al 2006). In their study the relative and actual risks were elevated for under weight, obese men and surprisingly men and women in the normal weight categories 2 and 3 (Table 1). These conclusions are taken from relative risk ratio statistics in column 2. In this table I have added the relative risk as percent (column 3) and the actual risk as percent (column 4) for comparison. It is apparent that the actual risk of death is not great even for obese men in which the increase is only 1.9% greater than the normal reference body weight of 1.3%. Underweight men have an elevated actual risk (3.5%) which is similar to those who are overweight (3.2%). Similar risks were found for all women studied (results not shown).


*Similar results were found for women. **BMI or Body Mass Index is
equal to the body weight in kg divided by height in meters squared.

A BMI calculator can be found at http://www.nhlbisupport.com/bmi/bmicalc.htm

The conclusions that can be drawn from Table 1 are:

1. Underweight and some low normal weight individuals are at an increased risk of death. This important point rarely gets a mention in the news media or fashion magazines.

2. Overweight people are not at an elevated risk of death.

3. Obese people have an elevated risk of death similar to underweight individuals.


In order to examine the effect of bodyweight on risk of death the influence of smoking was considered by Adams et al (2006). To do this data were grouped into the following groups: Never Smoked, Current Smoker and Former Smoker. For the sake of simplicity I have made a table which shows data from never smokers and current smokers (Table 2).


* Actual risk was calculated by dividing the number of individual in each category by 100,000. Data were obtained from table 2 in Adams et al (2006).

If you look only at the results of relative risks and compare never smokers with current smokers you could draw the incorrect conclusions that it is better to be an obese smoker than an obese non-smoker. For example, in group 10 the relative risk for the never smoker is 159%, whereas for the current smoker in this group the risk is 42%. This is a good example of how relative risks can be misleading and why actual risks should be used in these kinds of studies. Comparing the actual risks in these two groups provides the correct answers. The actual risk of death in category 10 for the never smoker group is 2.6% compared to 5.0% for the current smoker group. So just as we suspected being obese and smoking is not good for you.

The reason for the reversal seen with relative risk analysis is that the reference category 4 is a lower number (0.78) in the never smoker group than that of the current smoker group (3.08). Since all weight categories are compared to these numbers the relative risk increases in the smoker groups are lower. For example: in the never smoker group the percent in category 10 is 2.6 compared to 0.78 in the reference group. Therefore the relative risk is 3.25 times as great as the reference value of 0,78 for never smokers and only 1.61 times as great for current smokers (5.0/3.08 = 1.61).

The influence of smoking on the actual risk of death is shown in figure 1. It is clear from these data that current smoking increases the actual risk of death in all BMI categories. The data for women are similar to men (data not shown). In this figure and in table 2 we see that men in the never smoked group show similar actual risks in all categories except the obese groups 9 and 10 and these are not great but they were judged to be significant by Adams et al (2006). The former smokers have a small actual increase in risk in categories 3 – 8 and elevated risks in the underweight, low normal and obese categories. There is no evidence from these data that being overweight increased your risk of death.

Figure 1 Actual risk of death as function of BMI groups in smokers and non-smokers.


My conclusion that being overweight does not increase the risk of death agrees with that of McGee (2005) who analyzed 26 large population studies which included 388,622 individuals.


Data taken form McGee 2005. The data are expressed as relative risks and the 95% confidence interval in parentheses. All groups were compared to normal weight individual defined as having a BMI of 18.5 to 25. Over weight defined as BMI between 25 and 30 and obese as BMI of 30 or greater. Abbreviations: CHD, coronary heart disease; CVD, cardiovascular disease.

From table 3 it is clear that even the relative risks of death for overweight individuals is not different from those of normal weight. Actual risks could not be calculated because insufficient data. These data suggest that obese men and women are at a greater risk of death from all causes, CHD and CVD but not cancer. The lack of any increased risk of cancer in overweight women supports my contention that heavier women do not have higher risk of breast cancer.


1. Underweight and some low normal weight individuals are at an increased risk of death.

2. Overweight people are not at an elevated risk of death.

3. Obese people have an elevated risk of death similar to underweight individuals.

4. Current smoking in obese and under weight individuals increases their risk of death. This is also true for those in the former and never smokers but the risks are not as great.

5. Obesity increases the risk death from heart and vascular diseases but not cancer.

The finding that overweight men and women show no significant increased risk for heart and vascular diseases or death casts doubt on the continual bombardment from the news media, nutrition experts, many physicians, drug companies and food manufactures that being over weight will increase the incidence of many diseases.


McGee DL (2005) Ann Epidemiol 15:87

Adams KF (2006) N Engl J Med 355:763

Saturday, August 14, 2010



For the past few years nutritionist and epidemiologist have been having trouble demonstrating that dietary saturated fats and cholesterol increase the risk of coronary heart disease (CHD). However they kept their noses to the grind stone and have settled on trans fats as the culprit. Trans fats are semi-solid fats which result from the conversion of liquid oils into their partially hydrogenated forms. They are present in most of the good things you eat, everything from donuts to milk. The Food and Drug Administration has gone along with this most recent fat fad and decided that food labels must state the amount of trans fats in all products. The Department of Agriculture has made a key recommendation that trans fat intake should be limited in their new food pyramid guidelines. Several other actions have been taken by various governmental agencies at home and abroad to limit or eliminate trans fats from our diets.

My question is: are trans fats all that bad and is worth the time, money and trouble to take them out of foods. I seem to remember when these same agencies were trying to limit saturated fat and cholesterol in our diets. Actually they still are because they are not able to face the facts that dietary saturated fat and cholesterol do not increase risk of heart disease.


The more important papers in the past 10 years concerned the supposed dangers of trans fats can be found in a review article written by Mozaffarian et al (1). It was important because it was picked up by the news media and given the appropriate level of hype. But, as I have asked before, did anyone in the news media or anywhere else actually read this paper? Well I did and here is my summary of it and the studies it uses to suggest that trans fats are bad. I will also review a couple of others along the way.

For each of the studies I will provide a brief summary for the readers who doesn’t want to cloud their mind with all of the statistical based arguments.


Study 1. Mozaffarian et al (1).

The data in this review article clearly shows no association between dietary saturated fat or cholesterol and the incidence of heart disease. They also provide data which indicates a slight increase in risk of heart disease ( relative risk: 40%, actual risk: 0.12%). This very low increase becomes insignificant when amount of fiber in the diet is taken into account. It seems that people who eat the most trans fats also have the lowest quantity of fiber in their diets. This is obviously related to the attitude: that French fries are better than broccoli any day. So it’s the lack of fiber that is dangerous, not the intake of trans fats.

Study 2. Nurse’s Health Study by Oh et et al (4).

These authors demonstrate in agreement with other studies that dietary total fat, saturated fat and unsaturated fat do not increase the risk of heart disease; however, they do try to implicate trans fat. Their story is poor because the data on estimated risks are only slightly different from no effect. There is no indication that they corrected the data for fiber intake which would reduce any elevated risk even more, so my judgment is the data are too weak to show any actual differences.

Study 3. Zutphen Elderly Study by Oomen et al (5).

Another study which shows a marginal increased risk of heart disease associated with dietary trans fat intake. Too little data and too close to call.

4. The Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study by Pietinen et al (6).

One more study which shows no increased risk of heart disease associated with dietary saturated fat, cholesterol, monounsaturated fats, total triglycerides and saturated fatty acids. Dietary trans fat was not associated with an increase in heart disease but it was associated with marginally with increased risk of death due to heart disease. The numbers involved in this portion of the study are very low and the actual increased risk is also very small (0.12%) I don’t believe there is anything to worry about.


Because this subject is so important and it has insinuated itself into brains of so many scientists, physicians, news media reporters and the public in general, I believe a detailed account of the problems with these studies should be provided. Hence, the following.

In their review Mozaffarian et al (1) present a summary table which shows the relative risk (RR) and the 95% confidence intervals (CI) found in four prospective cohort studies. As usual I will present the data from each of the studies in the review and evaluate the authors risk estimates by making graphs using their data.

Study 1. Ascherio et al (2)

One of these studies which is often quoted as demonstrating that trans fat intake is linked to elevated risk of heart disease was published by Ascherio et al.(2). This is an important paper in many ways because it also presents data that shows no increased risk in the incidence of heart disease due to intake of saturated fat or cholesterol. Hasn’t been too long ago that every one believed they were dangerous dietary bad actors. Many people still do and this is partly due to the failure of the governmental agencies and news media to inform the general public of this change in belief.

In Ascherio et al (2) people were categorized according to the intake of the various kinds of fat (grams/day) and the incidences of heart attack (myocardial infarct, MI) were recorded (Figure 1).

Figure 1. The relative risk (RR) of myocardial infarct (MI) as a function of the intake of saturated fat or cholesterol. The data points show the RR plus and minus the 95% confidence intervals. The intake categories are shown on the X axis and reflect the amount of each fat in grams/day. The lowest amount is 1 and the highest is 5. The dashed line represents the no effect level of 1.0.

It is clear that intake of saturated fat or cholesterol is not related to any increase risk of MI. This conclusion has been confirmed in several more studies some of which are discussed in the following sections.

In this same study the authors examined for any relationship that might exist between trans fat and the risk of heart disease. Their data are shown in Figure 2. There is no significant increase in myocardial infarction (MI) in the 2 - 4 categories which represent from 2.2 to 3.3 grams/day of trans fat. The intake in category 1 was 1.5 gm/day. The only statistically significant group was the people who consumed 4.3 gm/day (category 5). So my conclusion from these data would be that a very weak suggestion of an increase risk of MI was found in group 5. By “weak suggestion” I mean barely statically significant and may not be real.

When the authors adjusted their data for fiber intake there was no hint of and increase in risk due to trans fat intake. It has been suggested that increased dietary fiber intake significantly reduces heart disease (3). The correction for fiber intake is based on the data that show people who consume a fatty diet also consume less fiber. Since fiber is presumed to be protective against MI the incidence increases when dietary fiber intake is low (3). I am not sure we can believe this any more than anything else in these papers. One conclusion you could draw from these data is trans fat is not the problem and it may be the lack of fiber.

Figure 2 The relative risk (RR) of myocardial infarct (MI) without (A) and with (B) correction for dietary fiber intake. The RR plus and minus the 95% confidence intervals are shown as a function of amount of dietary trans fat The intake categories are shown on the X axis and reflect the amount of each trans fat in grams/day. The lowest amount is 1 and the highest is 5.

Adding to my lack of enthusiasm concerning any elevated risk of MI due to dietary intake of trans fats is the following incidence data shown in Figure 3. As you can see the incidence of MI in the three fat groups appear to be almost identical. This makes it very difficult to accept that any significance can be granted to trans fat group and not to the other groups. In most of these kinds of papers the incidence data are not shown. I believe you can guess the reason.

From these data in Figure 3 the actual risks can be calculated. I will do this for trans fat only, the others are similar. The authors calculate a RR of 1.40 with a 95% CI (1.1 -1.79) and would say there is a 40% increase in the risk of MI due to trans fat intake. Note that the lower value for CI is 0.1 away from non-significant. It gets worse, when you calculate the actual increase which is 0.12%. So even if the actual risk were significant the odds of adverse effects from consuming trans fats are small indeed.

Figure 3. Incidence of myocardial infarct as function of dietary fat. Percent incidence calculated from the raw data in Ascherio et al (1). Fat intake categories are the same as in Figure 2 above.


Ascherio et al (1) also examined the risks of fatal heart disease and dietary fat. The number of people who died in these studies was very small which makes any conclusion concerning these data very suspect. The RR’s for increased death rate in the saturated fat group were not significant except in the highest intake category (RR = 2.21 95% CI, 1.38 - 3.54, see Figure 4 A). The actual increased risks is 0.08%. The amount of cholesterol in the diet did not have any significant effect on the risk of death due to CHD (Fig 4 B).

Figure 4. The relative risk of death due to CHD in relation to saturated fat (A) and cholesterol (B)

Trans fat appeared to elevate the risks of death (Fig 5 A); however, when these data were corrected for fiber intake the risks became non-significant (Fig. 5 B). The actual difference in risks between the lowest intake category (0.07%) and the highest (0.12%) was 0.05%. These very small numbers add to the doubt that an statistical significance can be ascribed to these data.

As indicated previously, if any significant increase in the risk of CHD in people who consume a lot of trans fat it is not due to the fat, but the lack of fiber. So don’t blame the fat, blame the lack of fiber

Figure 5 The relative risk of death due to dietary intake of trans fats. The multivariate adjusted relative risks are shown in A and these risks corrected for fiber intake are shown in B.

Study 2 Nurse’s Health Study (NHS) by Oh et al (4).

In this paper women were studied for 20 years and the incidence of heart disease and dietary fat intake were compared. The NHS data are important because they show no increase in risk for the following forms of fat:

· Total fat,

· Saturated fat,

· Mono-unsaturated fat

· Poly-unsaturated fat.

These studies are exemplified by Figure 6 A which shows that dietary saturated fat intake does not increase the risk of coronary heart disease (CHD). Dietary intake of poly unsaturated fat appears to result in a reduced risk of CHD (Figure 6 B). Once again the data are weak and with only one RR (category 5) showing any significant decrease. It is difficult to say with any confidence that the risks of CHD are reduced by eating poly unsaturated fats.

Figure 6 Relative risk of coronary heart disease and dietary intake of saturated and poly unsaturated fat. The energy intake categories represent the percent of energy derived from the specific fat in question. Category 1 being the lowest and 5 the highest.

The results for trans fat study are weak in the opposite direction (Figure 7). That is, they show and upward trend and thus hint at an increased risk of CHD due to dietary trans fat but once again the data in support of this are weak. In category 2 the RR is clearly not significantly different from the no effect line of 1.0. However the level for category 3 is elevated above the no effect line. But, then the RR for category 4 decreases and is not elevated significantly above 1.0. In the final category (5) the RR of 1.33 is elevated but the 95% CI (1.07 -1.66) just barely misses crossing 1.0. No data are provided on the incidence of CHD in each category so I was not able to draw a graph of these data as I did in Figure 3 above. The RR of 1.33 is the one published in the review by Mozaffarian et al (1), and no mention is made of the tenuous nature of its information or that decrease intake of fiber was taken into account.

Figure 7. Relative risk of coronary heart disease and dietary intake of trans fat. The energy intake categories are explained in caption of figure 4.

Study 3. Zutphen Elderly Study by Oomen et al (5). Written for the scientist.

In this study older men (age 64-84) were studied for 10 years and their trans fat intake was correlated with risk of CHD. The authors concluded that there was an increase in CHD with a 2% increase in dietary trans fat. This was based on a RR of 1.28 and (95% CI of 1.01 – 1.61). This looks like another barely significant RR and I advise waiting until the data form next years’ study is available before you make up your mind. No data on incidence or relative risk as a function of fat intake were given in this paper; therefore , it is difficult to judge it’s value.

Study 4. The Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study by Pietinen et al (6).

This is an important study because the authors conclude that risks of CHD are not associated with dietary intake of the following nutrients:

· Saturated fat

· Cholesterol

· Cis-monounsaturated fats

· Total triglycerides,

· Saturated fatty acids

· C12 – C16 Saturated fatty acids.

These authors also report that linoleic and linlenic acids are not associated with increase or decrease risk of CHD. These findings concerning linoleic and linolenic acid are in contradiction to some of the more recent studies which have claimed these two fatty acids decrease the risks of CHD (4).

In their analysis of trans fat they did not show any significant increase in the risk of non-fatal CHD; however they claim that fatal CHD is increased by trans fat intake. Figure 8 shows that the only RR of significance is for the highest intake category ( RR 1.39 95% CI 1.09-1.76). In light of the up and down nature of the preceding values, none of which come close to being statistically significant, it is likely that 1.39 is also not significant. If a correction for fiber intake had been used it probably would not be significant. The actual increase risk in this case is only 0.16% compared to the 39% relative risk.

Figure 8. Relative risk of fatal CHD and dietary intake of trans fats. Data taken from Pietinen et al (6).

Retrospective case-control studies

The remaining studies included in the review by Mozaffarian et al (1) were retrospective case-control studies which have smaller numbers of individuals involved and are less likely to be valid.

In the first study the content of fatty acids in adipose tissue was examined in 671 men in 8 European countries and Israel with acute myocardial infarction (7). The odds ratio was 0.97 (95% CI 0.56-1.67) and obviously not statistically significant. So, no evidence for an increase in the risks of CHD due to fatty acids was found.

The second study also examined the content of fatty acids in adipose tissue but this time in Costa Rica (8). They examined their data by adjusting the values according to the following scheme. First they adjusted the data for income, history of diabetes , history of hypertension, physical activity , smoking, years living in the house and alcohol intake (Figure 9 A). As you can see, this adjustment made the odds ratios increase but not significantly (Fig 9 B). So they tried further adjustments for adipose tissue, linolenic acid and intake of vitamin E, saturated fat and total energy. These adjustments finally gave the authors a statistically significant increase in the odds ratio in group 5 of OR 2.94 (95% CI 1.36-6.37); Figure 9 C. This is the value in table 2 of the review by Mozaffarian et al (1). No mention of the non-significant odds ratios in the first two adjustments. The possibilities for errors in a study containing so many adjustments, some valid and some may be not, make the final odds ratio very suspect and not credible.

Figure 9. Odds ratios and risk of myocardial infarction as a function of the amount of trans fatty acid in adipose tissue. All data taken form table 3 in Baylin et al (8).

The last study listed in the review is an Australian study which involved 209 cases of myocardial infarction (9). The relative risk of MI in the highest category of dietary intake was 2.25 (95% CI 1.16-4.32), and this value was published in Figure 4 in the review by Mozaffarian et al (1). I have plotted the RRs for each intake category in Figure 10 A below and it is obvious the only the last RR is statistically significant. However, this increase in risk was due to decline in the number of people in the control group (see Figure 10 B) which was accompanied by a similar increase in the number of cases of MI. The true RR would have been considerablely lower if the number in the control group had not taken a sudden an unexplained drop.

Figure 10. Dietary intake of trans fat and first myocardial infarct (MI). These data taken from table in Clifton et al (9).


Evidence that trans fatty acids in the diet elevate the risk of CHD is very weak and if the lack of fiber in the diet is taken into consideration there is no significant increased risk of CHD due to dietary trans fat.

These studies demonstrate that the following fats do not increase the risk of CHD:

· Total fat,

· Saturated fat,

· Cholesterol

· Mono-unsaturated fat

· Poly-unsaturated fat

· Cis-monounsaturated fats

· Total triglycerides,

· C12 – C16 Saturated fatty acids.


Some important studies which were not included in the Mozaffarian review (1) will be discussed in this section.

For historical and prophetic reasons the study by Micheal DeBakey’s group published in 1964 should be mentioned (10). They analyzed serum cholesterol values in 1700 patients and found no correlation between cholesterol and atherosclerotic disease. Apparently no one listened and cholesterol was declared the enemy. The battle went on for years, but as you can see from the review of the preceding paper and by ref 11 below, cholesterol has be exonerated by many people.

A large prospective study involving 80,082 women 34 to 59 years of age seems important not only because of its size but for the many different fats studied (11). They found no increase risk of CHD as a result of the following dietary fats:

· Total fat

· Animal fat

· Vegetable fat

· Saturated fat

· Mono-unsaturated fat

· Poly-unsaturated fat

· Cholesterol

The only item on their study menu which showed even the slightest increased risk was trans fat and then only in the highest intake category (Figure 11). It is hard to believe that the RR for this group 1.27 (95% CI 1.03-1.56) is of any significance since the amount of trans fat intake in category 5 differs from category 4 by only 0.05%. This is very weak data point and lacks credibility. No numbers of individuals were given in this paper so the actual risks could not be calculated.

Figure 11. Relative risk of CHD as function of trans fat intake. Data from table 3 in Hu et al (11

In a follow-up study these authors say that diet high in fat and low in carbohydrates is not associated with increased risk of heart disease (12). They seemed surprised by their findings. Maybe they should read their own paper.


Although dietary cholesterol was declared the enemy years ago, the signs that this was not true were provided by none other than Micheal Debakey back in 1964 (10). He has since been proven to be correct. The other highly touted villains in our diet like total saturated fat and animal fat have also bitten the dust. The latest epidemiologist’s horse by the name of trans fats appears to be stumbling and may fall in the future. I believe it has already fallen judging by the marginal data which is said to support its dangers.


1. Mozaffarian et al 2006 Trans fatty acids and cardiovascular disease. NEJM 354:1601-1613.

2. Ascherio A, Rimm EB, Giovannucci EL, et al. Dietary fat and risk of coronary heart disease in men: cohort follow up study in the United States. BMJ 1996;313:84–90.

3. Rimm EB, Ascherio A, Giovannucci E, Spiegelman D, Stampfer MJ, Willett WC 1996 Vegetable, fruit and cereal fiber intake and risk of coronary heart disease among men. JAMA 275:447-51.

4. Oh K, Hu fB, mason JE, Stampfer MJ Willett WC 2005 Dietary fat intake and risk of coronary heart disease in women: 20 years of follow-up of the Nurses’ Health Study. Am J Epidemiol 161:672-679.

5. Oomen CM, Ocke MC, Feskens EJ, et al. Association between trans fatty acid intake and 10-year risk of coronary heart disease in the Zutphen Elderly Study: a prospective population-based study. Lancet 2001;357:746–51.

6. Pietinen P, Ascherio A, Korhonen P, et al. Intake of fatty acids and risk of coronary heart disease in a cohort of Finnish men: The Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study. Am J Epidemiol 1997;145:876–87.

7. Aro et al 1995 Adipose tissue isomeric trans fatty acids and risk of myocardial infarction in nine countries: the EURAMIC study. Lancet 345:273-278.

8. Baylin A et al 2003 High 18:2 trans-fatty acids in adipose tissue are associated with increased risk of nonfatal acute myocardial infarction in Costa Rican Adults, J Nutr 133:1186-1191.

9. Clifton PM, Keogh JB , Noakes M 2004 Trans fatty acids in adipose tissue and the food supply are associated with myocardidal infarction. J Nutr. 134:874-879.

10. Garrett HE, Horning EC, Creech BG, Debakey M 1964 Serum cholesterol values in patients treated surgically for atherosclerosis. JAMA 189:655-659.

11. Hu FB et al 1997 Dietary fat intakes and the risk of coronary heart disease in women. N Engl J Med 337:1491-1499

12. Halton TL et al 2006 Low-carbohydrate-diet score and the risk of coronary heart disease in women. N Engl J Med 355:1991-2002.

Wednesday, August 11, 2010


Even epidemiologists agree that evidence on meat intake and breast cancer is inconsistent and marginal at best. This is an understatement since if you look carefully at their data there is no convincing evidence that any significant relationship exists. But, as usual they are going to try again and in Ferrucci et al.(2009) British J. Cancer 101, 178 they state that they have shown an elevated risk of breast cancer associated with eating red meat.

But, as usual, I have actually examined their data and it just isn’t so. The following figure shows the hazard ratio (similar to a relative risk) and confidence intervals for five groups of women (age 55-74) based on the amount of red meat they consumed.

Group 1 consumed the smallest amount of meat and 5 the most. So you would expect to see an increase in risk as the amount of red meat consumed goes up,- if eating red meat is related to risk. Which as you can see it is not. The response line is flat and the CIs include the no effect level of 1 with the exception of group 2. Since the women in group 2 are consuming much less meat than groups 3-5 this value is a statistical fluke.

This is one of those papers which contains the actual number of cases of breast cancerper number of person years in each food intake group. Thus, an actual percent can be calculated and these are presented in the following table.

So the actual percent difference between group 1 and group 5 is 0.05% or 5/10,000 which is a very small number and would not be considered to be of any statistical or biological significance.

Well if red meat was not going to show an increase the risk of breast cancer maybe other categories will. So they examined:

1. Steak: No increased risk

2. Hamburger: No increased risk

3. Sausage: No increased risk

4. Bacon: No increased risk

5. Pork chops: No increased risk

6. Processed meat: No increased risk


Try as they might there is no evidence that red meat and other kinds of meat increase the risk of breast cancer.

Sunday, August 8, 2010



Some nutritionist and epidemiologists believe that eating red meat and processed meats may increase the risk of cancer. They are not sure because many of the studies of this subject have been contradictory. However, the belief “that eating anything that tastes that good has to be bad” is always bouncing around in their heads. So they keep looking.

This topic is important because the news media love to tell their audience what they should and should not eat. Especially when the scientists say it could cause cancer. Important note: Often the word could gets left out and the headline reads, “EATING STEAK CAUSES BREAST CANCER”.

Cho and co-workers in 2006 set out to see if the occurrence of breast cancer in pre-menopausal women ages 26 to 46 was increased by eating red meat. This was done even though this same group had concluded from their previous analysis of 20 different studies that eating red meat was not associated with increased risk of breast cancer in either pre-menopausal or post-menopausal women (Missmer et al 2002).

But, they might have overlooked something. So they examined the data from 90,659 women over a period of 12 year and documented 1021 cases of invasive breast cancer (Cho et al 2006). Once again they found no association between red meat intake and the relative risk of breast cancer when all cancer were grouped together in agreement with their earlier analysis (Missmer et al 2002; Fig 1-A). Not giving up that easily they decided to divide the women with breast cancer into two groups: Those whose cancers did not contain estrogen and progesterone receptors (Fig 1-B) and those whose cancers did contain these hormone receptors (Fig 1-C). No association between intake of red meat and breast cancer in the women with receptor negative cancers was observed.

The authors then examined the possibility of any association in receptor positive group. They claim their data show the possibility of an increased risk of breast cancer in the women with the greatest intake of red meat (category 5 in Figure 1-C). However the lack of any upward trend in all of the groups with lower intake (2,3 & 4) makes it likely that the value for category 5 is a statistical fluke and is not significant. Certainly the over all evidence for any increased risk is marginal and as Shapiro (1997) has pointed out, when relative risks (RR) are 2 or lower they do not have the resolving power to distinguish between confounding factors and causation. In other words the RR value of 1.97 with a 95% confidence interval of 1.35 to 2.88 is of no real significance and certainly does not say anything about effects of eating red meat.

It seems obvious to me that when the RRs for groups 2 – 4 show no upward trend and the value for group 4 is lower than group 3 and sudden increase as in group 5 is likely to be due to chance.

Figure 1 The relative risks of breast cancer as a function of red meat intake. These data are taken from Table 2 in Cho et al 2006. Category 1 is the lowest intake and 5 is the highest. Error bars represent 95% confidence intervals and dashed line is the no effect level.

Intake categories are the following:

Servings of red meat in each category

1. 3/wk or less

2. > 3/wk to 5/week or less

3. > 5/week to 1/day or less

4. > 1/day to 1.5/day or less

5. > 1.5/day

You might be wondering why anyone would even think that red meat causes breast cancer other than the belief cited above that “anything that good, must be bad”. One is the possibility that red meat might contain sufficient estrogenic hormones to stimulate breast cancer growth. Most scientist believe the exceedingly low amounts of estrogen in red meat would not make any significant contribution to the amount of estrogen present in body of a pre-menopausal women. The ovaries of these women produce amounts of estrogen which far exceed any small amount of some estrogenic substance found in the diet.

Another potential culprit in the epidemiologist’s arsenal are heterocyclic amines (HCA). These are formed when meat is cooked and over cooked and they have been shown to be carcinogenic in some animal models. However when put to the test in human studies they fall short. For example, one of the studies quoted in their paper which reportedly shows the HCAs in red meat are associated with breast cancer was done by Shina et al 2000. The results of this study do not show any association between HCAs and breast cancer for two of the three amines studied (Figure 2A and 2B) and only a marginal, and I believe insignificant, association between 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) and cancer (Figure 2C).

Figure 2. Relative risk of breast cancer as a function of heterocyclic amines (HCA) in the diet. The intake categories represent the estimated amount of HCA in the diet. Category 1 is the lowest and 5 is the highest. Error bars represent 95% confidence intervals and dashed line is the no effect level.

Another possible dietary substance which might be blamed is iron. Iron is a very important constituent of hemoglobin and myoglobin and plays a role in other necessary physiological functions. However, iron has been shown in animal studies to increase mammary cancers when the animals are fed large amounts of iron in the diet and when an iron compound is injected under the skin in rats (Liehr & Jones 2001). Subcutaneous injections and dietary levels of iron which are 10 or more times the normal levels are non-physiological and would be very likely to ever occur in humans. Certainly it is difficult to imagine that 1.5 serving/day of red meat could contain a sufficient quantity of iron to make any difference in the amount of iron present in a woman’s body.


I believe the evidence does not support the idea that eating red meat is associated with an increased risk of breast cancer.


Cho E et al 2006 Red meat intake and risk of breast cancer among premenopausal women. Arch Intern Med 166 2253-2259.

Liehr JG, Jones JS 2001 Role of iron in estrogen-induced cancer. Curr Med Chem 7:839-49.

Shapiro S 1997 Do trans fatty acids increase the risk of coronary artery disease? A critique of the epidemiologic evidence Am J Clin Nutr 66(suppl):1011S-7S

Sinha R et al 2000 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine, a carcinogen in high-temperature-cooked meat, and breast cancer risk. J Natl Cancer Inst. 2000 92(16):1352-4.


Since the kinds of data and graphs in my blogs will be unfamiliar to many of you, I am going to try to explain in this section.

Relative Risk (RR) and Confidence Interval (CI).

Relative risks and confidence intervals are usually expressed numerically or in graphic form. I have made a graph which displays four different RRs and their CIs and will try to explain how to interpret them.

Group 1 has a RR of 1.5 and the CI is 1.1 to 2.5 which are generally shown in text form as RR 1.5 ( CI 1.1-2.5). This RR would be considered marginally significant since the lower range of the CI is just above the line for RR of 1 which is called the no effect level or line.

Remember if the range of the CI includes the no effect line of 1 the data are not statistically significant.

Group 2 also has a RR 1.5 (CI 0.8-2.5) but the lower range of its CI is includes the no effect value of 1.. Therefore it would not be considered statistically significant.

Group 3 has a RR 0.5 (CI 0.2-0.9) which is an example of a RR which is showing
a decreased risk. Since the upper range of the CI does not include the no effect level of 1 this RR would be judged to be statistically significant.

Group 4 has a RR 0.5 (CI 0.2-1.1) which is another example of an RR showing a decreased risk. However, in this case the upper range of the CI includes the no effect level of 1 and would not be considered statistically significant.


When two groups are judged to be statistically significant it just means that the two groups are probably not the same. Something is different,- but what?

Such differences could be due to chance as in studies with large confidence intervals which almost include the no effect level.

These types of studies attempt to show associations or correlations between some event and risk of some disorder or problem. Showing that the data are significant does not prove a cause and effect only the possibility of one.

Tuesday, August 3, 2010


If you believe what governmental and nutritional agents are telling you then the answer is yes. But is it really true.

Sometimes these people don’t know what they are talking about. I have looked into the actual evidence that salt intake is bad for you and found that the question is not settled and the answer is probably, No. I also found that the whole subject has been overblown and misrepresented.

There are many clinical and epidemiological studies on the topic of the relationship of salt intake and increased heart and vascular disease (CVD), stroke and death rate and I have reviewed many of them. These studies fall into three categories:

1. Yes, salt intake may be associated with increase risks of cardiovascular disease (CVD), stroke and mortality (References 1-5).

2. No, salt intake does not increase risks of CVD, stroke and mortality (6,7).

3. Surprise, higher salt intake decreased the risks of CVD (1,8-10).

4. Another Surprise, lower salt intake increased risks of CVD and death (9,10).


How could any government agency have the audacity to state that salt intake is dangerous for your health when the scientific literature is full of papers which display such divergent patterns. Obviously something is wrong. A large part of “Something is Wrong” is people in government agencies and nutritional advisors don’t read the scientific papers which are published on this topic. The scientists who write the papers are guided by a lack of objectivity which is present because of the “Band Wagon Effect”. If they get on the wagon they must remain because their funding, reputation and status depend on it.


If you have had all you can take stop reading; however, if you want a more detailed and somewhat more technical explanation then read on.

Some of this bad information comes from scientists who do not take an honest and careful look at their own data. Their presumption that higher intake of salt will be associated with increased CVD, stroke and mortality over shadows their objective analysis of their own data. As an example of such a study I will use the following paper which incorrectly concluded that high salt intake increased risk of stroke and CVD (11).

In this paper by Strazzullo et al.(11) the results obtained from 19 cohort samples from13 studies have been combined and analyzed by a method called meta-analysis. This method brings together the good, the bad and the ugly and lumps them altogether with the hope that the result will fit the bias of the investigators. In this particular meta-analysis the risk of stroke and salt intake were reported for 14 studies, 10 of which showed no significant effect and 4 showed marginal significance, that is, just barely significant. The results are identical for effect of salt intake on cardiovascular disease: total, 14 studies, 10 showed no significant effect and 4 which showed marginal effects. Most scientists would not believe the conclusions of any study in which 71% of the observations showed no effect and 29% showed a marginal effect.

Yet, Strazzullo et al. conclude that high salt intake is associated with a statistically significant increased risk of stroke and cardiovascular disease. They based their conclusions on relative risks plus or minus confidence intervals (CI). In this kind of analysis if the CI does not include 1.0 then the data are considered to be statistically significant. As the number of subjects get higher the amount of error gets lower. Thus, the confidence interval (CI) gets smaller and is less likely to include the control group number of one.

As an example, a single study with a small number of subjects might have a relative risk (RR) of 1.2 with a CI of 0.6-2.4. Because the CI interval includes 1.0 this means the RR of 1.2 cannot be considered statistically significant. Now if you group together more studies as in a meta-analysis the number of subjects gets much larger and reduces the variability which reduces the interval of the CI. As a result one could obtain a RR of 1.2 with a CI of 1.01-2.0 and since this CI does not include 1.0 the RR is claimed to be statistically significant.

This is exactly what happen in the Strazzullo et al. combined studies which indicated a risk for stroke of RR 1.23 ( CI, 1.06-1.43) and for CVD of 1.17 (CI, 1.02-1.32). The RR of CVD became statistically significant when one of the studies (Alderman 1995) which did not agree with their bias was omitted.  Note that the lowest numbers for CI 1.06 for stroke and 1.02 for CVD are so close to 1.0 you have to hold your breath to claim these RRs are of any real significance.

It should also be noted that demonstrating statistical significance does not necessarily mean that the finding is of any biological significance.

NOTE: In another later blog I will discuss this method in more detail.


1. Tunstall-Pedoe H, Woodward M, Tavendale R, A’Brook R, McCluskey MK. Comparison of the prediction by 27 different factors of coronary heart disease and death in men and women of the Scottish heart health study: cohort study. BMJ. 1997;315:722–9.

2. He J, Ogden LG, Vupputuri S, Bassano LA, Loria C, Whelton PK. Dietary sodium intake and subsequent risk of cardiovascular disease in overweight adults. JAMA. 1999;282:2027–34.

3. Tuomilehto J, Jousilahti P, Rastenyte D, et al. Urinary sodium excretion and cardiovascular mortality in Finland: a prospective study. Lancet. 2001;347:848–51.

4. Nagata C, Takatsuka N, Shimizu N, Shimizu H. Sodium intake and risk of death from stroke in Japanese men and women. Stroke. 2004;15:1543–7.

5. Cook NR, Cutler JA, Obarzanek E, Buring JE, Rexrode KM, Kumanyika SK, Appel LJ, Whelton PK. Long term effects of dietary sodium reduction on cardiovascular disease outcomes: observational follow-up of trials of hypertension prevention (TOHP). BMJ 2007 DOI 10.1136/bmj.39147.604896.55.

6. Kagen A, Popper JS, Rhoads GG, Yano K. Dietary and other risk factors for stroke in Hawaiian Japanese men. Stroke. 1985;16:390–6.

7. Cohen JD, Grandis G, Cutler JA, Neaton JD, Juller LH, Stamler J. Dietary sodium intake and mortality: MRFIT Follow-up Study Results (abstract). Circulation. 1999;100(suppl 1):1–524.

8. Alderman MH, Madhavan S, Cohen H, Sealey JE, Laragh JH. Low urinary sodium is associated with greater risk of myocardial infarction among treated hypertensive men. Hypertension. 1995;25:1144–52.

9. Alderman MH, Cohen H, Madhavan S. Dietary sodium intake and
mortality: the National Health and Nutrition Survey (NHANES I)>. Lancet. 1998;351:781–5.

10. Cohen HW, Hailpern SM, Fang J, Alderman MH. Sodium intake and mortality in the NHANES II Follow-up Study. Am J Med. 2006;119:274e7–5e14.

11. Strazzullo P. et al. Salt intake, stroke and cardiovascular disease: meta-analysis of prospective studies. BMJ 2009:339;b4567.