Liars, Damned Liars, and Statisticians

Home (Main Menu)

Neutrino Nonsense

Vitamin E meta-analysis misleads consumers


"Behind the Good News on Cancer, There's the Same Old Bad News


The good news about America's war on cancer keeps coming. Prostate cancer: Five-year survival of patients is now 94%, up from 43% in 1950. Lung cancer: Patients detected early using spiral CAT scanning expected to have five-year survival of 60%, up from 14%. Colon cancer: American patients now have a five-year survival of 60%, compared with 35% for British patients.

Does this mean that fewer men are dying from prostate cancer, that we should embrace lung cancer screening with spiral computerized tomography (CAT) scans and that cancer care in the United States is far superior to care in the United Kingdom? Maybe not.

The problem is that each report uses what has become the statistic of choice--five-year survival--to deliver good news about the war against cancer. And although five-year survival rates are increasing for all cancers, it may not mean what you think.

Most of us assume that rising five-year survival means that cancer death rates are falling. But in a study for the Journal of the American Medical Assn., my colleagues and I found no relationship between changes in five-year survival and changes in how many Americans die from cancer, as measured by the number of cancer deaths per 100,000 population.

Consider the two cancers with the most dramatic increases in five-year survival since 1950, prostate cancer (from 43% to 94%) and melanoma, or skin cancer, (from 49% to 88%). Despite all the press about recent declines, the current death rate from prostate cancer is actually a little higher than in 1950, and the death rate from melanoma is more than twice as high.

And then consider the case for one of the cancers with the worst five-year survival rates: stomach cancer. There has been only a trivial change in stomach cancer five-year survival since 1950 (from 12% to 19%), but it is America's biggest cancer success story because people die at one-fifth the rate they did in 1950.

To understand how five-year survival could be so misleading, you have to understand the statistic. Five-year survival is the proportion of individuals with cancer who are alive five years after receiving the diagnosis.

The current five-year survival for prostate cancer of 94%, for example, means that if 1,000 men were diagnosed with prostate cancer in 1995, 940 would be alive today. So how can five-year survival go up even when death rates don't change? The best way to understand this paradox is to work through two thought experiments.

First, to keep the math easy, let's consider one group of men--those who died of prostate cancer at age 78 (the median age of death for men with the disease). If everyone in that group had been diagnosed when they were 75, their five-year survival rate would be 0%. But if everyone in that same group had been diagnosed when they were 70, the five-year survival rate would be 100%.

What has changed? Reported five-year survival has gone up, people are aware of a cancer diagnosis for a longer period of time, but everyone still died at age 78. What have we learned? Earlier diagnosis always increases the five-year survival ]statistic. Whether death is postponed is quite a different question.

The second thought experiment requires some familiarity with a very foreign concept: nonprogressive cancer. Most of us think of cancer as a relentlessly progressive disease that spreads throughout the body and ultimately kills. There are some forms of cancer, however, which do not progress and so will never cause symptoms or death ...

Now consider the effect of nonprogressive cancer on five-year survival. Imagine a group of 1,000 men diagnosed with prostate cancer in 1950. Back then, long before our present aggressive screening programs, all would have come to see their doctor because of symptoms. In other words, all would have had progressive cancer.

If 400 were alive five years later, the five-year survival rate would be 400 of the 1,000 or 40%. Today, however, we find cancer not only in patients with symptoms, but also in people who have no symptoms (some of which is progressive, some of which is nonprogressive).

To keep the math easy, let's assume that today a group of 1,000 men are diagnosed because of symptoms and have progressive prostate cancer. In addition, another 1,000 men are diagnosed with nonprogressive cancer. The 1,000 men with nonprogressive disease all will survive five years. So the new five-year survival rate will be 1,400 of 2,000 or 70% ...

Five-year survival, in other words, depends on what constitutes a cancer diagnosis and may not reflect any change in how many people die ...

Five-year survival rates will increase over time, even if death rates don't change, simply because we are telling people they have cancer earlier in their lives and because we are finding more nonprogressive cancer.

Another thing we know is that new diagnostic tests and aggressive screening programs will always look good using five-year survival, even if early diagnosis doesn't help people live one day longer. And because we test so much more frequently than most other countries, we know that international comparisons using five-year survival rates overstate America's ability to provide cancer care.

Dr. H. Gilbert Welch, a Professor at Dartmouth Medical School, Works for the Department of Veterans Affairs in White River Junction, Vt. This Commentary Is Excerpted From an Article in the June 14, 2000, Issue of the Journal of the American Medical Association."

Top -- Home


Commentator One wrote:

A plot of Super-K "neutrino" events vs. cosine theta (where theta is the angle on the celestial sphere with the Sun at 0 degrees) gives a spike near zero degrees, possibly indicating the presence of neutrinos above background noise. However, if events are plotted against theta, the spike goes away; and all that can be seen is background noise.

Commentator Two wrote:

The plot vs angle is meaningless. The correct plot is the one vs cosine theta, because the path length of the neutrino through the Earth is directly proportional to the cosine of the zenith angle. The probability of oscillation due to MSW effect is directly proportional to the path length. So the cosine plot directly reflects oscillation probability, and as you already point out for us, shows a definite effect, proving that the Kamiokande results are valid and that MSW oscillation happens to neutrinos passing through the Earth.

[Pontecorvo in 1967 proposed that neutrinos might oscillate, from one type to another, if a mass difference exists among three varieties of neutrinos. A theory of how such oscillations might change a neutrino of one type into another as it passes through matter along its path has been worked out by Mikheyev, Smirnov and Wolfenstein (1985) and is now referred to as the "MSW effect".]

Commentator One wrote:

distance through the Earth = (cosine theta) x D,

where D is the diameter of the Earth. With the Sun at angle theta = 0 degrees on the celestial sphere, cosine theta = 1; and d = D. So "neutrinos" from the Sun would have the longest path through the Earth to reach Super-K.

This distance is important because as Commentator Two has noted. "The probability of oscillation due to MSW effect is directly proportional to the path length. ..." 'd' is directly proportional to cos(theta), and according to theory, the probability of matter-induced neutrino oscillations ('p', the MSW effect) is directly proportional to 'd'. Therefore, 'p' is also directly proportional to cos(theta).

Commentator Two wrote:

It is not a matter of assuming what is to be proved, it is a matter of showing that the experimental results are consistent with the predictions of the theory.

Commentator One writes:

I guess my problem is that I'm bothered by the close connection between cosine theta, d, and "the probability of matter-induced neutrino oscillations." Perhaps "assuming what is to be proved" is too strong; but I would like a situation where "events" and "angle theta" are not, a priori, tied to a particular theory. There should be equal opportunity for a null hypothesis.

So I'm going to take a stab at another approach.

Let's picture the celestial sphere as a golf ball with dimples of the same size and the Sun located at dimple [theta=0]. While there is only one dimple at theta = 0, there are many dimples at theta = 90 degrees. Thus, if there were no Sun at dimple [theta=0], we would expect to receive more neutrinos or other bacgroud noise from theta = 90 than from theta = 0.

We could ask: "How many more events must we receive from dimple [theta=0] when the Sun is present than when the Sun is absent to conclude that, when the Sun is present, we are seeing something more than backrgound noise?"

To get a fix on the number of events which would constitue bacground noise, we could look at theta = 180 degrees, on the opposite side of the celestial sphere (golf ball) from the Sun. However, we should expect more background noise when looking at theta = 0 than at theta = 180 because, at theta = 0, we are looking at a closeby star while, at theta = 180, we are looking at nothing in particular.

If we plot dimples against cosine theta, there are as many dimples with a cosine theta between 0 and .05 as there are between .95 and 1; and thus the number of background events against cosine theta should give a flat graph with a blip at theta = 0 because of the BACKGROUND NOISE from the Sun. So we can rephrase the question as follows: "How big must the blip be at cosine theta=0 relative to the flat background graph to conclude that we are seeing something more that backround noise from the Sun?" This "background noise" from the Sun may or may not be partly due to neutrinos passing through the Sun.

If we plot events against theta, there is a blip near dimple 0:

                Figure 1

    |               _____
  E |            __/     \__
  v |         __/           \__
  e |       _/                 \_
  n |     _/                     \_
  t |    /                         \
  s |  _/                           \_
    |_/                               \ 
    |                                  \
    0                 90               180
    Angle theta on the celestial sphere (golf ball)

If we plot events against cosine theta, the blip becomes a spike.

Figure 2

    | |
    | |
    | |
    | |__________________________________
     Cosine theta

How much of this spike is due to solar neutrinos and how much is due to background noise? Would the plot look like Figure 3 if the background noise due to the Sun were removed?

Figure 3

     Cosine theta

As long as we work with cosine theta, there is no way to tell how much of the spike is due to background noise and how much is due to solar neutrinos.

As an aside, it would be nice to have data from theta = 0 during the daytime versus during the nighttime, to compare solar neutrino flux from the Sun when it comes through the earth versus from the direction of the sky.

If the nighttime "neutrino" flux is equal to the flux during daytime, that might be evidence for solar neutrinos (if nothing else could pass through the earth) since only a negligible part of the spike would be due to background noise from the Sun. I could not find any such data.

A null hypothesis could be "The differnce between events at theta = 0 and theta = 180 is due to background noise, not to solar neutrinos." Is there some way to look at the data to test the proposed null hypothesis?

One approach is as follows: if chunks of latitude with equal number of "dimples" were paired, then a line of constant slope should be obtained except where the effect of the Sun is seen as a spike. In order to obtain data, I eyeballed Figure 3 in SuperK.html (SAA site) to produce the followimg table:

Angle "Event"

175    .09
165    .18
155    .29
145    .33
135    .40
125    .57
115    .57
105    .62
95     .63
85     .62
75     .63
65     .55
55     .55
45     .45
35     .45
25     .46
15     .30
5      .28

I then plotted the data as follows:

 E                                                     - 
 v                                                 -      
 e .6                                (115,65)   - XX (95,85)
 n                                        XX-  (105,75)
 t                                     - (125,55)
 s .5                    (145,35)  -    
                      X    X   -    X
 a               (155,25)  -     (135,45)
 t .4                  -   
 A             -  
 n .3      -  X (165,15)
 g      X 
 l  -    (175,5)
 e .2
85 .1
     0      .1      .2      .3      .4      .5      .6      .7

         Events at Angles 175  to  95

A line of constant slope is obtained, but there is no spike. The inevitable conclusion seems to be that the null hypothesis is correct.

For more information: Autodynamics

Top -- Home

Alliance for Natural Health PRESS RELEASE, 11-Nov-04

``Vitamin E meta-analysis misleads consumers

The following release is made by the ANH in the wake of a raft of negative publicity on Vitamin E following the meta-analysis released by Johns Hopkins Medical Institutions.

The meta-analysis to be published next January in the Annals of Internal Medicine (Volume 142, Issue 1), suggests that mortality in people taking over 400 IU of Vitamin E a day is increased. The study, published electronically on the journal's website,re-analyses data from 19 clinical trials involving Vitamin E published between 1993 and 2004 and was undertaken by researchers at Johns Hopkins Medical Institutions.

In the view of the Alliance for Natural Health, which represents doctors, practitioners, consumers and leading-edge companies with interests in sustainable healthcare and natural therapies, it is of paramount importance that the study results are interpreted with caution.

Scaremongering headlines, based largely on misinterpretations of the study, which have begun to appear today, do nothing to help the development of responsible self-care patterns in consumers which are urgently required in the face of escalating heart disease and cancer rates.

Limitations of the study:

In a widely circulated press release publicising the study, Dr Edgar Miller, lead author of the meta-analysis at Johns Hopkins University School of Medicine, claims that "If people are taking a multivitamin, they should make sure it contains no more than a low dose of vitamin E; our study shows that use of high-dose vitamin E supplements certainly did not prolong life, but was associated with a higher risk of death."

However, the study has a number of very important limitations which mean that sweeping generalisations such as those made by Dr Miller should not be made. Such statements could be regarded as irresponsible particularly as they are likely to cloud consumer views over the importance of food supplementation as a means of compensating for the now well demonstrated inadequacies of the typical, western diet.

Some of the most important limitations of the study are:

All trials in the study were performed using only one of the seven forms of Vitamin E, namely alpha-tocopherol, which is often used in its synthetic form. There is now strong scientific evidence which demonstrates that other forms of Vitamin E, as found in natural food sources, are much more effective as antioxidants ...

The studies were undertaken in the main on elderly persons suffering chronic diseases, so it is not possible to conclude that higher supplemental doses of Vitamin E promote disease or increase death among healthy people. This limitation is acknowledged in the Johns Hopkins study, the authors indicating, "we could not evaluate the generalizability of our findings to healthy adult populations."

Although the study includes 19 trials and is referred to as a meta-analysis, there is huge variation between the scale of the various trials and in the case of the 11 high dose trials, the bulk of the data come from just two large studies, namely the Cambridge Heart Antioxidant Study [CHAOS] and the Medical Research Council/British Heart Foundation Heart Protection Study [MRC/BHF HPS].

This latter study shows nearly double the number of deaths compared with all the other high dosage studies combined, and four times as many as in the other high dose studies with increased mortality. The study used a combination of synthetic beta-carotene and synthetic Vitamin E. [Combining Vitamin E and beta-carotene has been shown to increase deaths among smokers.]

The follow-up periods for the 19 studies were often short, averaging from 1.4 to 8.2 years. In many cases the chronic diseases suffered by study participants would not be able to be significantly impacted over shorter time periods ...

Misinterpretations of the study ...

The biggest issues affecting consumer choice and governmental regulation are related to study interpretation, both by the authors and others subsequently ...

The authors of the study make no mention of the good body of evidence that shows that synthetic forms of Vitamin E, or purified natural forms that are limited to alpha-tocopherol, have the ability to reduce the body's absorption of other forms of Vitamin E (e.g. Handelman GJ et al, J Nutr 115; 807-13,1985; Burton GH et al, Am J Clin Nutr 67; 669-84, 1998; Huang HY, Appel LJ J Nutr 133; 3137-40, 2003) which are much more powerful antioxidants ,,,

Negative implications of the study: ...

Recommended Daily Allowances, although largely discredited by the scientific community and bearing no relationship to the levels required for optimum health, are still used by many governments as the yardstick by which to determine acceptable intakes of vitamins and minerals,

It is a very rich irony that the European Commission has seen fit to allow only the alpha-tocopherol form of Vitamin E in its new Food Supplement Directive, set to come in to force across Europe in August 2005 unless challenged successfully. The Directive bans the natural, food-forms of Vitamin E and some 300 forms of vitamin and mineral presently on the European market. In many cases these food forms have been shown to be considerably more beneficial to health.

The Alliance for Natural Health is presently challenging the Directive in the European Court of Justice. The case is expected to be heard in early 2005.

Last week in Bonn, Germany, United Nations' Codex Committee on Nutrition and Foods for Special Dietary Uses finalised the text of an international guideline for vitamins and minerals. The next stage will be the development of guidelines for maximum dosages, which will use a so-called evidence-based risk assessment approach.

In September [2004], the Food & Agricultural Organisation and the World Health Organisation released a consultation document* for which comments must be received by 10 December. The timing of the release of the Johns Hopkins meta-analysis seems set to give the regulators justification to be excessively restrictive, particularly if the media can be relied on to exaggerate and distort the findings of the study.

Whatever levels are developed for the international Codex guidelines are then likely to filter back into national or regional legislative systems, including the Food Supplements Directive in Europe. There is no doubt that even the United States' Dietary Supplement Health & Education Act (DSHEA), which has been viewed as the bastion of liberal natural healthcare legislation for some 10 years, is now deeply under threat.

The Alliance for Natural Health, and associated campaigns around the world, are working to protect the interests of millions of consumers worldwide who have adopted natural therapies as their primary healthcare approach, as opposed to use of drugs.

Please join the Alliance for Natural Health free of charge, or make a donation now, to support our work. For information, please visit our website ''


Carotech Communication Regarding Johns Hopkins vitamin E Meta-analysis

11-12-04, Carotech, Inc.

``TO "E" OR NOT TO "E" ?

This Technical Communication is in response to an article published on Nov 10th 2004 in the Annals of Internal Medicine. Researchers at Johns Hopkins examined 19 different vitamin E studies between 1966 and 2004 to a meta-analysis. The total number of subjects (age 47 - 84 years old) in these 19 studies was 135,967. The dosages of vitamin E ranged from 16.5 to 2000 IU per day. The meta-analysis suggests that too much of vitamin E (400IU or more per day) increases the risk of all-cause mortality.

Meta-analyses are often highly speculative because of the different variables in each of the studies such as source of vitamin E (natural or synthetic), study duration, health/disease condition of subjects, etc. Hence, they by no means offer definitive proof of anything, due to the lack of uniform protocols and patient groups

Perhaps - on the other hand, it goes to show that a single nutrient vitamin E (ie : alpha-tocopherol - synthetic or natural) is not the panacea. It is against conventional wisdom to take mega-doses of one nutrient without considering the potential side effects.

As a matter of fact, we have seen this before - in 1996 with the beta-carotene debacle (The ATBC and CARET studies). These two studies provide evidence that taking beta-carotene alone rather than a multi-carotenoids (beta-carotene, alpha-carotene, gamma-carotene, lycopene, lutein - as produced in nature), may increase the cancer risks among smokers. This may be because all these carotenoids work synergistically as a team - recharging and supporting each other to confer the health benefits.

Similarly, a high dosage of alpha-tocopherol alone has been shown to deplete the body's gamma-tocopherol. Despite alpha tocopherol's action as an antioxidant, gamma tocopherol is required to effectively remove the harmful peroxynitrite-derived nitrating species. Because large doses of dietary alpha tocopherol displace gamma tocopherol in plasma and other tissues, the current wisdom of vitamin E supplementation with primarily alpha tocopherol should be reconsidered. Other forms of vitamin E - gamma-tocopherol, delta-tocopherol and certainly tocotrienols have been proven to have unique health properties.

Taking a single form of vitamin E (ie : alpha-tocopherol alone) denies the very fact that nature put seven (7) different forms of tocopherols and tocotrienols there for a reason ...''

Best Times -- Top -- Home -- What's New