Progress against Economicitis?

Jason Shafrin, the Healthcare Economist, has a nice post explaining how a statistical illusion can make early screening for disease appear much more effective than it really is.

Here is an example using the dreaded disease economicitis.  Let us
divide people into 3 groups.

  • Healthy: You live forever.
  • 1st stage economicitis is asymptomatic. Life
    expectancy when 1st stage economicitis begins is 10 years.  One half of
    economicisits cases are 1st stage.
  • 2nd stage economicitis appears when individuals
    mysteriously grow a third or possibly fourth hand.  Life expectancy with second
    stage economicitis is 2 years.  One half of economicitis cases
    are 2nd stage.

Before any screening was developed, individuals would learn they had
economicitis  when they started growing extra hands.  Thus, documented
life expectancy for those with  economicitis was 2 years, since all
individuals who were recorded as having  economicitis were in the 2nd

Let us assume that a screening technique is now available.  If the screening
device is able to detect 100% of stage 1 and stage 2 economicitis
cases, then we will see that life expectancy will increased to 6 years
(10/2+2/2=6). Statisticians looking at the data may claim the following: “The
economicitis screening test has increased life expectancy after
diagnosis from 2 to 6 years!”

This claim, however, is false since there is no effective treatment for 
economicitis.  The increase in average life expectancy is not due to
any improvement in health care, but only because the relatively healthier
individuals with 1st stage economicitis are now being detected by the

Many years ago, David Plotkin had a article in The Atlantic dealing with this issue and others with respect to breast cancer. The statistics are somewhat out of date but the article remains of real value.


individuals mysteriously grow a third or possibly fourth hand

And those third and fourth hands are invisible.

If you assume everyone passes through both stages, the numbers don't seem to add up in the steady state. However, suppose that in any given year, there is a small risk of getting first-stage economicitis and a larger risk of getting second-stage economicitis? Most people don't benefit from early screening since they skip the first stage, but those who do can drive up the average life expectancy once it becomes known how early they get it and how long they live with it.

the analysis misses a further "statistical illusion" which would also appear to increase the life expectancy. since those with the advanced disease die sooner, there are fewer of them around, so in fact more than half of those diagnosed with the disease will have the less advanced version.

If you want to make the example work with "stages", then change the life expectancy of stage 2 to 10 years. Then pre-screening the life expectancy of someone diagnosed with the disease was 10 years. But now post-screening it is up to 20 years (call it 15 on average, perhaps).

From the way it is described it seems these should be called "type 1" and "type 2". But then if type 1 is asymptomatic why is it even classed with the other one?

This analysis ignores the fact that wide scale screening is only done if we have 1) a cure or 2) effective treatments to prolong life or 3) treatments that only work in earlier stages.

if you could cure economicitis in stage I, then a screening program would make sense and would truly increase life expectancy. however, in the real world, if no cure/treatment existed then there would be no screening program.

This is RIDICULOUS! Everybody knows that you can live for up to 12 years after being diagnosed with Economicitis by a simple treatment of reading Paul Krugman columns faithfully ever month. Once you lose your ability to do economics (as has Paul Krugman), you can live a relatively normal life like most people. You'll still grow the extra hands eventually, but you won't need them once you've adopted to Krugman Cure.


your analysis is based on a faulty assumption. you are assuming that mammograms are 100% accurate. they are not. mammograms will occasionally (IIRC, 3-5% of the time) find cancer where none exists. this does lead to harm to the women as she had to undergo invasive tests (e.g. biopsy, lumpectomy, mastectomy) that she would otherwise have not undergone. these tests carry non-negligible complication rates.

the question then turns into, do mammograms help more women than it hurts? this only holds true for women between 40 or 50 to about 70 or so. in the earlier ages the rate of false positives is too high and in the later ages the likelihood of the woman dying from newly found breast cancer is so much smaller than dying from other causes that the test does not lead to increased survival.

RCH you are still neglecting that the treatment itself has quality of life costs and carries significant risks.

Sometimes the treatment will kill or seriously injure you and it has some unpleasant side effects, so some patients will die or suffer significant reduction in quality of life from treatment for cancer which they did not need. What you have to do is balance the injury suffered by patients who received unnecessary treatment against the benefit to those patients who received needed treatment they would otherwise have missed.


your comments don't make much sense in the context. we are talking about a screening program. by definition that includes a large number of people. whether or not a screening program is beneficial or not depends on whether or not it helps more people than it hurts.

no one is going to stop an individual from getting a test she desires. the doctor may tell you it's not worth it, that more harm will come from it than good (on average) etc, etc. yet if you are willing to pay for the test yourself and assume the risks (i.e. absolve the doctor of all responsibility when the contrast dye in the CAT scan you demanded destroys both of your kidneys) then there will always be some doctor out there who will do the test.

you are still underestimating the false positives. breast cancer may be one of the most common female cancers, yet overall it is still fairly rare. that means the number of people who undergo invasive tests as a result of a false positives dwarf the number of people whose lives are saved. how would you feel if you underwent a modified radical mastectomy as result of a falsely positive mammogram reading? (i literally saw this happen a few months ago. a woman underwent a MRM for what turned out to be a benign enlarged lymph node that looked just like a cancerous lesion on mammogram). what do you plan on telling all those women? how many women's bodies are you willing to disfigure to lengthen one life?

if someone has cancer or will have eventually they won't care about false positives or risks of dying from other things.
this is just plain false and leads to be question if you have any medical experience at all. a 65 year old male newly diagnosed with prostate cancer isn't going to stop worrying about his congestive heart failure and COPD. it'll take at least 20 years for the prostate cancer to metastasize and kill him. by that time he'll have died twice from his CHF. and yes, they will care about false positives. do you think the person who had a false positive CAT scan indicating rectal cancer is going to be happy that he now has a permanent colostomy as a result of removal of half his colon?

but the statistics don't measure the pros and cons for an individual.
yes, they do. look up the work up of a deep venous thrombosis (DVT) and pulmonary embolism (PE). it may not be cancer, but it is an easily digestible example. the amount of testing that is done is determined based on what risk level you fall into. breast cancer is done the same way...and you can find algorithms online.

A small percentage of women would receive a priceless gift of extra life from mammograms and a lot of other women would be inconvenienced, and perhaps a few even die early from radiation. Somehow that needs to be weighed up in the choice, whether saving a small statistically marginal group is worth it or not.

isn't this exactly what you disagreed with earlier? when i explained that in young women the rate of false positives is too high mainly because the breast tissue itself is dense which makes differentiating cancer difficult and due to the relatively low rates of cancer in young women. and when i explained how in much older women even if you discovered cancer it would affect their life expectancy because they would die of something else beforehand. why should these women undergo invasive tests and potentially dangerous chemo and radiation for something that isn't even going to kill them?

"isn't this exactly what you disagreed with earlier? when i explained that in young women the rate of false positives is too high mainly because the breast tissue itself is dense which makes differentiating cancer difficult and due to the relatively low rates of cancer in young women. and when i explained how in much older women even if you discovered cancer it would affect their life expectancy because they would die of something else beforehand. why should these women undergo invasive tests and potentially dangerous chemo and radiation for something that isn't even going to kill them?"

No, the difference is here the doctors and statisticians are telling patients what is good for them. Why should doctors advise people that they should prefer to die of one disease instead of another?

And the idea that the patient has the ultimate decision is a copout. The doctor with his authority tells a patient who might get cancer not to have a mammogram that might save them. One what basis do they play God with that patient's life by giving this advice? This is not informed consent, and people who got cancer might well sue doctors with a class action suit over this. People generally believe their doctors and some people will die of cancer who might have had no complications from a mammogram.

I know you can respond that someone has to make this decision or no one could advise anyone. But I am saying experts don't really know how this decision affects people beyond a statistical equivalence in dying from one thing as opposed to another.

As another example look how much money was spent to protect people against terrorism, and if you spent that money instead in research on cancer, designing car safety better, etc. Logically you could argue to people it is better to have terrorists flying planes into buildings occasionally and have less cancer, but no one would vote for you. Why is that if the algorithms are so self evident? People value a freedom from terror more than a lower mortality rate. The real factors people care about are just not represented in these studies.

Say for example a 1000 women in their 40's walk into a doctor's office over a year. Also say that X of them know they will have breast cancer and will die without detection from a mammogram. Y of them will die if they have a mammogram. If you advise the X women not to have a mammogram they will consider the advise to be crazy because it is sentencing them to death, as would the advise to the Y women on having a mammogram.

Once women know their fate then it becomes equivalent to another medical situation that happens all the time. Say the X women know they have breast cancer and will die without an operation, and the Y women will die if they have the operation. In this case the decision to operate will be based on how the patient feels, their family and friends, the chance they might live without the operation, have more time with their family instead of a sudden death in surgery, etc. This is how real medical decisions are made. You don't advise people that it is a toss up whether they prefer to die of an operation or cancer, or that this is somehow presumed to be equivalent in an algorithm. I suspect the doctors might find they had a higher mortality rate than the patients if they told them that.

The lack of knowledge in the case of breast cancer makes it become something else which it never was, a statistical argument better suited to insurance premium calculations.

My brother in law, a pulmonologist, called this "lead time bias". Wikipedia has a short article,

Well put Albatross. I am also making the point that statistics are based on a normal curve and there is a tendency to bias the results on who or what is perceirved as normal. The people on the edges end up being seen as deviates or abnormal.

For example a shoe manufacturer might decide just to make shoes for average sized feet and regard big and small sizes as abnormal. Statistically this looks correct but it ignores the long tail of customers with unusual needs. A small book store might only stock the most popular books on the theory these are what average or normal people want to read. Amazon and Netflix though stock a larger range so people with more abnormal tastes find what they want there, and so they become bigger businesses.

In the same way medicine tends to think of normal disease and normal patients, and makes decisions on such as mammograms like this. However the abnormal patient might have other factors which need mammograms or they are more dangerous. This is seen more and more for example where people with certain genes (predisposing them to breast cancer or being susceptible to radiation) would have had the wrong advice by being treated as average.

Here is an example of this:

" Japanese study in 2001 looked at the usefulness of combining mammograms with ultrasound during breast cancer screening.16

The researchers evaluated 15,139 women during a five-year period and found that the combination of mammograms and ultrasounds increased a doctor’s ability to detect breast cancer by an impressive 29%. They also found that cancers detected with the addition of ultrasound screening were more likely to be discovered earlier, and therefore were more susceptible to treatment.

An even more recent study published this year examined the ability of sonograms on their own to detect breast cancers in women with dense breasts, as mammograms done on women with dense breasts are less sensitive at detecting cancer. This study, done between January 2000 and January 2002, examined 1,517 women with dense breasts and normal mammograms.17 Sonograms done on these women detected seven cancers, leading the researchers to conclude that “screening breast sonography in a population of women with dense breast tissue is useful in detecting small breast cancers that are not detected on mammography or clinical breast examination. The use of sonography as an adjunct to screening mammography in women with increased risk of breast cancer and dense breasts may be especially useful.†

There's a tendency to see mammograms as the normal screening and sonograms and MRI as unusual, and they end up in the long tail of small numbers of doctors using them.

As another example it depends on the level of radiationa dn radioactive materials used with younger women"

OBJECTIVE: A mammography unit with both a molybdenum anode and a rhodium anode, filtered with molybdenum and rhodium, respectively, was evaluated to determine which types of women would benefit from the dose savings of the rhodium combination despite some loss of contrast. SUBJECTS AND MATERIALS: In 100 women, the molybdenum anode and molybdenum filtration (Mo/Mo) were used to obtain mammograms of the right breast, and the rhodium anode and rhodium filtration (Rh/Rh) were used for mammograms of the left breast. All mammograms were obtained at 26 kVp. All milliampere-second values used to radiograph the breasts of these women were recorded. Mammograms of 54 women (30 with previous mammograms available), representing the four types of breasts as defined by the American College of Radiology, were interpreted by three radiologists. Each mammogram was assigned a grade for breast type, preference (Rh/Rh, Mo/Mo, or previous mammograms), contrast, and sharpness. RESULTS: Overall, mammograms obtained by using the Mo/Mo combination were preferred. However, for images of types 3 and 4 breasts, Rh/Rh was preferred twice as often as it had been for mammograms of types 1 and 2 breasts. The mean glandular dose for all breast types when the Rh/Rh combination was used was 42% of the dose used for the Mo/Mo combination. For a 6-cm-thick dense breast, the Rh/Rh combination required 40% of the dose required for the Mo/Mo combination. CONCLUSION: Mammograms obtained with the Rh/Rh combination carried an overall decrease in contrast and mean glandular dose. However, for young women and some women with large dense breasts, the Rh/Rh mammograms were equivalent to or better than the mammograms obtained with the Mo/Mo combination. Effective use of Rh/Rh units requires careful selection of women based on age or the amount of glandular tissue seen on previous mammograms.

AJR Am J Roentgenol 1994 Jun;162(6):1313-1317

Women who take estrogen are more likely to get breast cancer but less likely to die from it. That's because they are motivated to have more mammograms. So even though a study might encourage women not to take estrogen they live longer by taking it.

"Another study, The Breast Cancer Detection Demonstration Project, analyzed over 2000 women out of a database of 46,000+ participants before concluding that the risk of breast cancer is increased every year a woman takes hormone drugs. Another study tracked over 10,000 women at risk for breast cancer plus over 8,000 women at risk for endometrial cancer for 5 years. It concluded that women who take estrogen drugs without progestins for at least 6 years have a four-times increased risk of invasive endometrial cancer, with no increase in breast cancer. But women who take estrogen drugs with progestin drugs have about a 50% increased risk of breast cancer over those who don't. The size and consistency of these studies is hard to argue with.

One of the things that has recently emerged from breast cancer/drug studies is that the combination of estrogen and progestins dramatically increases breast density. This may confound the results of mammograms. Yet women who do take drug hormones may have reduced mortality because they are more likely to get a mammogram and have early detection since they are seeing a physician on a regular basis. The answer, of course, is for women to see a doctor regularly whether or not they’re taking prescription drugs."

Dietary factors are probably more important than a genetic predisposition to breast cancer. So whether to have mammograms is also affected by what the patient is eating, but this is ignored in many studies:

"According to the Breast Cancer Fund, a woman’s risk of contracting breast cancer was 1 in 22 in the 1940s. Today, it is 1 in 7. There is no end to the theories as to why this risk has increased. “Endocrine disrupters† (chemicals that mimic hormones) are a likely suspect. They are wreaking havoc on wildlife and clearly affect brain cells in the developing embryo.67 So far, however, studies have failed to show a link between breast cancer and blood levels of these chemicals. Still, they remain suspect—especially in combination with other factors.

Mainstream dogma is that exposure to estrogen causes breast cancer. By “estrogen,† the mainstream means the body’s own estrogens. This line of thinking always links variables (such as having/not having children or the age at which menopause occurs) to estrogen exposure and, hence, breast cancer risk. While this viewpoint appears to have some validity, a few things are wrong with it, including the thorny question of why, all of a sudden, exposure to something that has been a part of the human body for eons would cause cancer. It also skirts the question of why long-term use of birth control pills containing estrogens does not increase the risk of breast cancer.68

Genes are another possible explanation for breast cancer. This depressing theory implies that whether or not people get breast cancer is beyond their control and that nothing can be done about it, except having the breasts removed as a preventive measure.69 New research may put an end to the notion that there is nothing a person can do about “bad genes.†

“Bad genes† do not necessarily come from parents. Sometimes they come from the environment. Eighty-five percent of the “family risk† for breast cancer may come from something besides an inherited gene.70 Moreover, it has now been discovered that there are genes that can modify “bad genes.†71,72 In other words, you may not have to live with “bad genes.†

In addition, a new study shows that even if a person has a genetic predisposition toward breast cancer, the cancer does not necessarily activate unless the person encounters something in the environment that activates it.73 For some women, that “something† could be meat. For the first time, eating meat has been linked to genes and breast cancer.73 Families tend to share not only genes but recipes as well, and it is becoming clear that what you eat may be more important than what you were born with.

In studies that search for the cause of breast cancer, certain things consistently emerge. One is that diets rich in vegetables, soy, and green tea reduce cancer risk, and diets rich in animal fats (especially from red meat) increase risk.73-79 In a study from the Barbara Ann Karmanos Cancer Institute at Wayne State University in Detroit, beef, pork and vegetables accounted for 85% of the alterations to DNA in women, with meat causing damage and vegetables preventing it.80 Damaged DNA lays the groundwork for cancer.

The case of red meat is interesting not only because cooking it creates carcinogens, but also because the use of hormone implants in cows (which dates back about 50 years) coincides with the beginning of a major increase in breast cancer in North America.81 Countries with the highest rates of breast (and prostate) cancer also are the countries that allow such implants. North America’s breast cancer rate is the world’s highest—higher than all of South America and northern and southern Europe combined.82 Australia and New Zealand, which allow hormones to be implanted in cattle, have similarly high rates of breast cancer. In Europe, such implants are banned.

It is not hard to figure out why. Cattle implants contain 17 beta-estradiol and other strong steroids, including synthetic estrogens. Cows are repeatedly implanted, and the implants are in the cows when they are slaughtered. Guidelines published by the US Department of Agriculture and the University of Nebraska advise implanting the strongest drug last, 70 days before slaughter.83 The strongest implants last 90-120 days. Besides being in the cows at the time of slaughter, over time the hormones build up in fat.84 Fifty percent of the hormones contained in a steak may be in the fat.84 Neither the FDA nor the USDA monitors the use of hormone implants, or tests for residues in beef. Testing for the metabolites of estradiol alone would be a major undertaking, as there are more than a dozen such metabolites, and this is just one estrogen. Cows are given other hormones as well, including “male† hormones. Heifers are fed melengesterol acetate, a synthetic progesterone used for birth control and promoting rapid weight gain.

It has been demonstrated that a diet high in beef fat activates hormone-related genes.85 Zeranol, a synthetic estrogen cow implant, causes breast cancer cells to grow in the test tube. The amount of Zeranol needed to cause this growth is 30 times less than the amount that the FDA deems to be safe.86 A follow-up study being conducted at Ohio State University hopes to ascertain how much Zeranol ends up on the dinner plate and in the tissue of women with breast cancer.87 The study, which began in 2002, is still in progress. Data from approximately 200 women have been collected and are being analyzed. This important study may shed some light on at least one hormone implant. Studies on the total amount of all hormones added to American beef have yet to be conducted."

Comments for this post are closed