Humanist Perspectives: issue 153: The Screening and Prevention of Disease

The Screening & Prevention of Disease
by Giles Stevenson

woodcuts in this article by Shawn Sheperd

woodcuts in this article by Shawn Shepherd

Over the centuries of recorded history, teachers, social workers (under various names), nurses, physicians and yes, even politicians, have worked with varying degrees of idealism to promote a healthy and civilized population. Many physicians, dealing week in and week out with devastating and preventable disease (lung cancer, bowel cancer, aids, obesity, diabetes and renal failure) become interested, and sometimes obsessional, on the subje CT of prevention.

Prevention of disease may be attempted by one of two methods. The root cause of the disease may be removed, or one may examine large numbers of people to look for the earliest harbingers of the disease or its precursors so as to prevent progression into full blown lethal clinical symptoms.

The most attractive, but also most difficult approach is to remove the root cause: primary prevention. Cancer is caused by pollutants both inhaled and ingested so the incidence of cancer would be hugely reduced by removal of tobacco, heavy metal pollution of rivers and even medical irradiation. Infectious disease is caused by factors such as close conta CT with farm animals, overcrowding, poverty, malnutrition and contaminated drinking water. The current pandemic of diabetes, hypertension and renal failure is caused by overeating. How simple it should be to deal with these, and yet in practice how difficult!

Since primary prevention is so difficult, secondary prevention of some diseases by population screening becomes attractive. There are many problems for which radiology might have something to offer as a method of screening and they include tuberculosis, lung cancer, foetal abnormality screening in pregnancy, osteoporosis, colon cancer, prostate cancer and whole body CT screening. Examination of a few of these will illustrate the issues around implementation of population screening.

Screening for tuberculosis by chest Xray at a time when lung tuberculosis was common was a successful public health endeavour that saved thousands of lives at little cost. This success led to the use of chest Xrays to screen for lung cancer in the 1950s and 1960s at the time when the importance of smoking was starting to be realized. In 1913 when a patient died from lung cancer, all the medical students were called down to see the autopsy as the disease was so rare “that you will never see another case.” Now it is a pandemic with the death rate among black males in the USA, for example, having gone from almost zero in 1913 to 16 per 100,000 in 1950, and to 80 per 100,000 deaths in 1996. Thousands of chest surgeons around the world spend most of their working life seeing patients with lung cancer caused by smoking, most of whom will die within a year. Not surprisingly, many of these surgeons are keen on screening to dete CT lung cancer earlier before it has time to spread. Is such screening helpful? There were four trials of chest Xrays to screen for early lung cancer in the 1970s and they showed no benefit in reduction of mortality. Can CT scanning do any better? It is clear that CT detects tiny lumps in the lung, far too small to be seen on a chest Xray. Do any trials show benefit? This is where problems begin.

We find that screening trials are complicated by issues that bias the results, and these biases are general to most types of screening for disease. They are of four types and have been called ‘selection bias’, ‘lead time bias’, ‘length time bias’ and ‘overdiagnosis bias’. They are critical in understanding why we should not rush into screening without absolute proof of benefit, however tempted we may be because of the terrible nature of the disease we want to prevent.

woodcut by Shawn Sheperd

Selection bias occurs when selection of people into groups to be screened or not screened is not randomized. It may be that some people with symptoms of the disease choose to be screened, or to avoid screening, or there may be some other subtle selection bias. For example in the Edinburgh breast screening trial which showed a benefit for mammography, excessively large numbers of women in social class 1 managed to get themselves selected into the screening arm of the ‘randomized’ trial. Selection bias distorts the results of a screening trial in unpredictable ways.

Lead time bias reflects earlier detection of a fatal tumour in the screened group. Imagine two identical groups, each with one person with an identical fatal tumour. The tumour in the non screened person might cause death two years after symptoms begin. In the person screened the tumour will be detected before symptoms can occur, perhaps five years before. Even if there is no benefit, and both tumours cause death at the same time, the screened person will have lived seven years after diagnosis, and the non screened person only two years. Thus, although the screening in this instance has not extended life or reduced mortality, the earlier detection of the cancer in the screening group will have given the illusion of benefit because survival is more than five years after diagnosis, whereas in the control group survival was only two years.

Length time bias describes the variation in tumour malignancy. With an aggressive tumour there is a short time between the tumour being detectable by a test and the onset of symptoms. Intermittent screening is unlikely to dete CT such a tumour which will present as an interval tumour, i.e. in the interval between screening visits. Low grade tumours that have a good prognosis are present for a long time before symptoms develop, and so they are over represented in screening populations. Imagine two identical groups each with one rapidly fatal and two very slow growing low grade cancers. In the screened group all three cancers are diagnosed, and one causes death. In the control group only the rapid fatal tumour is diagnosed. Thus the mortality from this cancer in the control group appears to be 100% and in the screened group only 33%, even though there has been no actual benefit.

Overdiagnosis bias is an extreme example of length time bias in which there are early stage low grade tumours which never would have caused death, but by finding them on screening we record them as prevented cancer deaths. 90% of men over 80 have prostate cancer at autopsy though it was never a clinical problem in life. If we screen old men, and operate on small prostate cancers we will cause some deaths, and produce flawed data on the number of lives we have saved, by treating something that would never have caused a problem.

These four biases together mean that all screening trials, even those with no benefit at all, will show

Not only do these three findings not mean necessarily that there is a decreased mortality from the disease for which we are screening, but there may even be an increased mortality. If we want government funding for a screening programme we must show with a randomized controlled trial that there is

  1. a sustained reduced mortality from the disease,
  2. that the harm caused by the trial does not exceed, or even come close to, the good that is done and
  3. hopefully that the overall mortality of the screened group is not higher than the control group, as this might suggest some hidden harm from the screening.

To return now to lung cancer, the early results of CT studies are not very encouraging. One study in 1,500 smokers or ex smokers showed that 70% had small nodules in the lungs needing further investigation, of which 1.4% were cancers. Half of all the nodules operated on were benign and the mortality of the surgery was 3–4%. Three small CT studies showed increased numbers of low grade cancers in the screened group, but similar numbers of advanced ones, a better 5 year survival but no decrease in mortality. In other words CT screening was picking up the low grade less important lung cancers, but not helping people with the aggressive tumours that spread early.

Not only may the money be wasted. The screening may actually be doing harm. The CT scan, one of the greatest imaging advances of the past thirty years, has not come without a cost. A recent review in the medical journal The Lancet suggests that medical irradiation, of which CT scanning now forms the largest part, is responsible for 0.6% of all cancer deaths in the UK, 0.9% in Canada, 1.1% in the USA and 3% in Japan. These are huge numbers, and motivate radiologists not to use CT when other tests such as ultrasound or MRI will do as well, and are available. Apart from the dire CT harm of Xray irradiation there are other potential harms. Hayward has introduced the acronym Vomit – for ‘Victims Of Modern Imaging Technology,’ and described how the wonderful images obtained with the latest CT scanners can show small lesions that are actually of no importance but whose discovery leads to panic and fear when they are reported. Fear and panic as a result of an unimportant Xray finding is not a trivial problem. Leaping into widespread CT screening for lung cancer is thus premature as the benefit is unproved, and detection of small shadows in lungs and liver may lead to unnecessary surgery, and even death.

The difficulties in assessing benefit can only be addressed by a large controlled trial of screening in which subjects are allocated randomly to screening or no screening. This requires very large numbers, and in some areas where public expectations have been raised, this is no longer possible or affordable. For example, in North America, it would no longer be possible to do another randomized controlled trial on mammography. This is a pity as the data on breast screening is very complicated.

woodcut by Shawn Sheperd

Since 1975 most doctors have accepted that mammography for breast screening probably saves lives. Then in 2000 Goetsche and Olsen published a paper in The Lancet claiming that of the six published major trials, the four that showed benefit had major methodological problems, particularly of randomization, and that the two best trials, scientifically, showed no benefit, or even an increase in mortality. This paper unleashed a storm in both the medical and lay press until it began to seem politically incorre CT even to raise the question. The initial Lancet paper was followed by a Cochrane collaboration report that concluded that screening was unjustified, and experts then lined up on both sides. Two further Lancet papers from Sweden and Holland in 2003 claimed reduction in mortality, but were followed by letters from other experts saying that these results were mostly explainable by lead and length time bias, recent increase in additional chemotherapy and by inclusion of ductal carcinoma in situ (DCIS) as cancer. An outstanding USA radiologist and teacher published a partial rebuttal of Goetsche and Olsen’s points, and concluded with the ringing declaration that “informed medical opinion accepts that screening mammography saves lives.”

Should this assertion be sufficient for us? Unfortunately it can not be. Informed medical opinion, sadly, is far from infallible. In 1953 the Dean of Harvard medical school said that “half of what is taught in medical school is wrong. The problem is that we don’t know which half.” Some things don’t change. Today we know that for 20 years informed medical opinion accepted that long term hormone replacement therapy for women saved lives, but this belief was spectacularly wrong. A large recent trial was stopped when the evidence became clear that the participating women’s risk of cardiovascular disease went up, not down, on active hormone replacement therapy. This randomized controlled trial was pivotal in showing that accepted medical belief was wrong. As a result short term therapy for not more than five years maximum is now recommended for control of severe menopausal symptoms, and tens of thousands of women have stopped long term therapy that was previously believed to be helpful.

It takes a brave person to speak out against the combined might of medical opinion, industrial support and the women’s movement, but even now truth about mammography is elusive. The Royal Society of Medicine debated the motion “this house believes that the benefits of breast screening outweigh its risks and costs”, and speaking against the motion, which was carried, Thornton said “The only responsible course is to abstain from voting”. The four biases described above make the data almost impenetrable, and DCIS in the breast is a major problem. ‘Ductal carcinoma in situ’ is a cellular abnormality that is not invasive cancer, and may never become so: indeed it is present in up to 40% of all older women’s breasts at routine autopsy. In 2003 Fletcher concluded, in the New England Journal of Medicine, that the diagnosis of DCIS as cancer is likely to be for the most part, overdiagnosis. 15% of all breast ‘cancer’ found on mammography is DCIS, and inclusion of this in statistics of deaths prevented seriously biases the results of breast screening in favour of screening.

If it is so hard to be certain, with experts in epidemiology and mathematics lined up on each side, the benefits must be small indeed. There is certainly harm from mammography though it has not been well quantified. Mammography converts well women into worried well women. False negatives (19–34% of breast cancers are not shown on mammograms, though double reading by two radiologists reduces this to 15–24%) cause false reassurance and later law suits. False positive mammograms (which is 93% of all positives who are recalled for more testing) cause investigation, morbidity and cost. Approximately seven cancers are found for every 1,000 women screened, which includes on average one DCIS of uncertain significance, three advanced incurable cancers and three cancers where treatment may be decisive. Others have worried about the lack of honesty in promotional literature. One can present results in many ways. A relative reduction in mortality of 34%, an absolute reduction of 0.06%, a survival improvement from 99.02% to 99.08 % and 1 life saved from every 1592 mammograms can all be calculated from the same data, and mean the same thing. The first is much more powerful for promoting mammography. Both Thornton in the UK and Hillman in the USA have complained of an improper lack of truly informed consent in screening programs.

For the time being, mammography is all we have for breast screening, so we do it as well as we possibly can. This does not mean that we should be ‘believers’ that we are necessarily doing more good than harm, as the evidence is so marginal. The public would like reassurance and certainty, but certainty is not to be had. The main benefit from the breast screening programs is that it has led to a stunning improvement in the quality of diagnostic mammography for patients with symptomatic breast disease. Thirty years ago any radiologist was happy to read the occasional mammogram. The screening programmes have shown that better results are obtained by those who read large numbers, and screening programmes now require radiologists to attend regular and frequent educational events and to have their results audited. The money associated with breast screening has stimulated major improvements in equipment, and reduction in radiation dosage. These factors associated with large numbers of mammography law suits in the United States have led to fewer radiologists doing this work and doing it much better than even ten years ago. Unfortunately the numbers of radiologists willing to take on this litigiously dangerous work are diminishing towards a crisis point, especially among new USA graduates.

Colon cancer is a common disease, and not as devastating as lung cancer, as the mortality is only around 50% rather than 90%. From the viewpoint of screening, it is best regarded as two diseases.

woodcut by Shawn Sheperd

One, responsible for 5% of cases, is due to an inherited genetic mutation which leads to rapidly progressive colon cancer usually between the ages of 35 and 60. Families at risk can be identified by this characteristic young age, the presence of the responsible gene can be confirmed, and frequent screening by colonoscopy can be offered. It is a disgrace that provincial registries of these families have not been established as the risk of cancer in these families is between 50 and 75%, and the evidence that cancer death can be prevented by screening is good.

Most colon cancer, at least 75%, is quite different. It arises from a mutation in the wall of the colon that leads to a small polyp that takes on average 25 years to progress into a clinical cancer. This provides a wonderful opportunity for screening to dete CT such polyps. Unfortunately by age 70, some 35% of us have tiny polyps, but only 6% have polyps around 1 cm or more in size which are the ones of significant risk. There is disagreement on the best way to screen, as the cheapest way (testing the stool for blood which has been proved to reduce mortality) is the least efficient, detecting only some 50% of cancers and 5% of polyps. The best test, which is colonoscopy (passing an flexible tube around the bowel to find and remove polyps) is expensive and more dangerous, causing death in from 1 in 5,000 to 1 in 10,000 examinations. CT scan is being investigated as a safer alternative for polyp detection and is looking promising. Several large multi-centre studies are under way or have recently been completed using shorter endoscopes, CT scanners, stool testing for blood with more sensitive agents and colonoscopes. Another ten years or so should produce clearer information on the best way to prevent colon cancer, which unlike breast or lung cancer is uniformly curable when found in its early stages. Even with colon cancer however, screening is not without risk. In a small well defined study of 7,800 people screened with a short flexible endoscope in Britain in the 1990s there was one death of a screened asymptomatic subject. A benign polyp had been found and the man died from the operation to remove it. The study did considerable good as well, because a number of asymptomatic curable cancers were found and removed, as well as many polyps some of which would have become cancers. The result emphasizes however that even an apparently simple colon screening program is not risk free and will cause some deaths in well asymptomatic individuals.

Whole body CT screening is the source of much disagreement among physicians. There is no evidence of any benefit. It is being offered by for-profit centres, and is the cause of much further investigation, with serious complications and even deaths from unimportant disease. Baker has written that CT whole body screening is based on the exploitation of fear and hope, not on an accumulation of data. Another radiologist has described his personal experience of whole body CT screen. Lumps were found in lung and liver, which ultimately proved to be of no importance, but a year and two operations later he was $60,000 poorer and still in pain from the surgery. Hillman has pondered the question of whether an individual has the right to buy screening health care with discretionary income. He suggests the answer may be a qualified yes, provided the individual assumes the financial and personal risk of harm, and also does not impair the ability of others to access health care for sickness. He says that neither of these two conditions is met for whole body screening. The follow-on costs dwarf the initial CT costs, and the follow-on costs are not borne by the individual. As a result there is an impa CT on insurers, or the state, causing increased costs for both public and private health insurance. No randomized studies have been carried out to show any benefit from whole body CT screening and there is already much anecdotal evidence of harm. It is hard to come to any other conclusion than that the promotion of whole body CT screening for profit is unethical.

Muir Gray, the NHS director of screening in the UK has coined a number of salutary screening aphorisms. They include:

  1. All screening programs do harm. Some may also do good.
  2. The harm from a screening program starts immediately, while the good takes longer to appear. Therefore the first effect of any screening program, even an effective one, is to impair the health of a population.
  3. A screening program without false positives will miss too many cases to be effective.
  4. A screening program without false negatives will cause unnecessary harm to the population.
  5. Though insignificant to the population, a single false positive can be of devastating significance to the individual.
woodcut by Shawn Sheperd

In clinical medicine a patient goes to a physician with a problem and expects the doctor to do his or her best. Perfection is neither expected nor possible. With preventive medicine on the other hand, and especially with screening, a higher standard is required. Here the physician is approaching a group of well people and offering them a prescription of treatment or investigation, knowing that it will do some harm, and to a few very great harm. Screening in some areas, such as screening for carcinoma of the cervix, has proved to be of enormous benefit, and it is so tempting to try to offer screening for other devastating diseases which can be detected by modern radiology. This paper is intended to illustrate just why those of us who would like to promote screening must take care to prove a substantial excess of benefit over harm before governments are asked to spend precious health tax dollars on it, and before individuals are invited to accept the risks inherent in being screened. Detailed informed consent that honestly discloses all risks is even more important in screening programs than in routine medical care.

Giles Stevenson is Professor emeritus in Radiology at McMaster University where he worked for 25 years. He now lives on Vancouver Island, and works at the Cowichan distri CT hospital in Duncan. Shawn Shepherd is a painter, sculptor and print-maker, whose work investigates aspects of popular culture such as human identity and information trends. He lives in Victoria, BC.

Shawn Shepherd is a painter, sculptor and print-maker, whose work investigates aspects of popular culture such as human identity and information trends. He lives in Victoria, BC.